Professional Documents
Culture Documents
Financial Theory and Corporate Policy - Copeland
Financial Theory and Corporate Policy - Copeland
Theory and
Corporate
Policy/
THOMAS E. COPELAND
Professor of Finance
University of California at Los Angeles
Firm Consultant, Finance
McKinsey & Company, Inc.
J. FRED WESTON
Cordner Professor of Managerial Economics and Finance
University of California at Los Angeles
•
••
ADDISON-WESLEY PUBLISHING COMPANY
Reading, Massachusetts • Menlo Park, California • New York
Don Mills, Ontario • Wokingham, England • Amsterdam
Bonn • Sydney • Singapore • Tokyo • Madrid • San Juan
This îook is dedicated to our wives, Casey and June, who
have provided their loving support; and to the pioneers in
the development of the modern theory of finance:
Hirshleifer, Arrow, Deîreu, Miller, Modigliani, Markowitz,
Sharpe, Lintner, Jensen, Fama, Roll, Black, Scholes,
Merton, Ross, and others cited in the pages that follow.
Without their intellectual leadership this text could not
exist.
Over the past 30 years a branch of applied microeconomics has been developed
and specialized into what is known as modern finance theory. The historical
demarcation point was roughly 1958, when Markowitz and Tobin were working
on the theory of portfolio selection and Modigliani and Miller were working on
capital structure and valuation. Prior to 1958, finance was largely a descriptive
field of endeavor. Since then major theoretical thrusts have transformed the field
into a positive science. As evidence of the changes that have taken place we need
only look at the types of people who teach in the schools of business. Fifty years
ago the faculty were drawn from the ranks of business and government. They
were respected and experienced statesmen within their fields. Today, finance
faculty are predominantly academicians in the traditional sense of the word. The
majority of them have no business experience except for consulting. Their interest
iii
iV PREFACE
practical question of how to apply the concepts to difficult and complex real-
world problems. Toward this end, Chapters 2 and 3 are devoted to capital
budgeting, Chapter 14 shows how to estimate the cost of capital for a large,
publicly held corporation, and Chapter 16 determines the value of the same
company. Chapter 18, another change in this edition, emphasizes the theory and
evidence on topics of interest to chief financial officers: pension fund management,
interest rate swaps, and leveraged buyouts. Throughout the text we attempt,
wherever feasible, to give examples of how to apply the theory. Among other
things we show how the reader can estimate his or her own utility function,
calculate portfolio means and variances, set up a cross-hedge to reduce the
variance of equity returns, value a call option, determine the terms of a merger
or acquisition, use international exchange rate relationships.
In sum, we believe that a sound foundation in finance theory requires not
only a complete presentation of the theoretical concepts, but also a review of the
empirical evidence that either supports or refutes the theory as well as enough
examples to allow the practitioner to apply the validated theory.
We have tried to move all the central paradigms of finance theory into the first
half of the book. In the second edition this motivated our shifting the option
pricing material into Chapter 8. In this third edition we decided to add a com-
pletely new chapter on futures markets—Chapter 9. It covers traditional material
on pricing both commodity and financial futures, as well as newer issues: why
futures markets exist, why there are price limits in some markets but not others,
and empirical evidence on normal backwardation and contango.
In the materials on portfolio theory we have added a section on how to use
T-bond futures contracts for cross-hedging. In Chapter 7 we have updated the
literature review on the Capital Asset Pricing Model and the Arbitrage Pricing
Model. Chapter 8 contains new evidence on option pricing. The materials on
capital structure (Chapters 13 and 14) and on dividend policy (Chapters 15 and
16) have been completely rewritten to summarize the latest thinking in these
rapidly changing areas of research.
Chapter 18 is completely new. Many topics of importance to chief financial
officers are applications of finance theory. Pension fund management, interest
rate swaps, and leveraged buyouts are the examples developed in this chapter.
Chapters 19 and 20 on mergers and acquisitions, restructuring, and corporate
control represent up-to-date coverage of the burgeoning literature. Similarly,
Chapters 21 and 22 reflect the latest thinking in the field of international financial
management.
We made numerous other minor changes. In general, we sought to reflect all
of the new important literature of finance theory—published articles and treatises
as well as working papers. Our aim was to keep the book as close as possible to
the frontiers of the "state-of-the-art" in the literature of finance theory.
Vi PREFACE
At UCLA we use the text as a second course in finance for MBA students and
as the first finance course for doctoral students. We found that requiring all
finance majors to take a theory-of-finance course before proceeding to upper-
level courses eliminated a great deal of redundancy. For example, a portfolio
theory course that uses the theory of finance as a prerequisite does not have to
waste time with the fundamentals. Instead, after a brief review, most of the course
can be devoted to more recent developments and applications.
Because finance theory has developed into a cohesive body of knowledge, it
underlies almost all of what had formerly been thought of as disparate topics.
The theory of finance, as presented in this text, is prerequisite to security analysis,
portfolio theory, money and capital markets, commercial banking, speculative
markets, investment banking, international finance, insurance, case courses in
corporation finance, and quantitative methods of finance. The theory of finance
can be, and is, applied in all of these courses. That is why, at UCLA at least,
we have made it a prerequisite to all the aforementioned course offerings.
The basic building blocks that will lead to the most advantageous use of this
text include algebra and elementary calculus; basic finance skills such as discount-
ing, the use of cash flows, pro-forma income statements and balance sheets;
elementary statistics; and an intermediate-level microeconomics course. Conse-
quently, the book would be applicable as a second semester (or quarter) in
finance. This could occur at the junior or senior undergraduate year, for MBAs
during the end of their first year or beginning of their second year, or as an
introductory course for Ph.D. students.
The end-of-chapter problems and questions ask the students not only to feed back
what they have just learned, but also to take the concepts and extend them beyond
the material covered directly in the body of the text. Consequently, we hope that
the solutions manual will be employed almost as if it were a supplementary text.
It should not be locked up in the faculty member's office, as so many instructor's
manuals are. It is not an instructor's manual in a narrow sense. Rather, it is a
solutions manual, intended for use by the students. Anyone (without restriction)
can order it from the publisher. We order it, through our bookstore, as a
recommended supplemental reading.
Understanding of the theory is increased by efforts to apply it. Consequently,
most of the end-of-chapter problems are oriented toward applications of the
theory. They require analytical thinking as well as a thorough understanding of
the theory. If the solutions manual is used, as we hope it will be, then students
who learn how to apply their understanding of the theory to the end-of-chapter
problems will at the same time be learning how to apply the theory to real-world
tasks.
PREFACE Vii
E. ACKNOWLEDGMENTS
We have received help from many persons on the three editions of the book. We
especially benefited from the insightful corrections, clarifications, and suggestions
of Eugene Fama, Herb Johnson, and Kuldeep Shastri. Nai-fu Chen and Ronald
Bibb wrote Appendixes B and D, respectively. Ron Masulis rewrote Chapter 5.
We also wish to acknowledge the help of the following: Ed Altman, Enrique
Arzac, Dan Asquith, Warren Bailey, Gerry Bierwag, Diran Bodenhorn, Jim
Brandon, Michael Brennan, William Carleton, Don Chance, Nai-fu Chen, Don
Chew, Kwang S. Chung, Halimah Clark, Peter Clark, S. Kerry Cooper, Larry
Dann, Harry and Linda E. DeAngelo, Dirk Davidson, David Eiteman, Chapman
Findlay, Kenneth French, Dan Galai, Robert Geske, Mark Grinblatt, C. W.
Haley, Ronald Hanoian, Iraj Heravi, David Hirshleifer, Tom Ho, Chi-Cheng
Hsia, William C. Hunter, Ashok Korwar, Clement Krouse, Steven Lippman,
Stephen Magee, Dubos Masson, Bill Margrabe, Charles Martin, Ronald Masulis,
David Mayers, Guy Mercier, Edward Miller, Merton Miller, Timothy J. Nantell,
Ron Necoechea, Jorge:Nielson, R. Richardson Pettit, Richard Pettway, Richard
Roll, Shigeki Sakakibara, Eduardo Schwartz, Jim Scott, Jandhyala Sharma, Kil-
man Shin, Ron Shrieves-, Keith Smith, Dennis Soter, Joel Stern, Sheridan Titman,
Brett Trueman, Jim Wansley, Marty Weingartner, Richard West, Randy West-
erfield, Robert Whaley, Stuart Wood, and Bill Ziemba.
For their considerable help in preparation of the text, we thank Susan Hoag
and Marilyn McElroy. We also express appreciation for the cooperation of the
Addison-Wesley staff: Steve Mautner, Herb Merritt, and their associates.
There are undoubtedly errors in the final product, both typographical and
conceptual as well as differences of opinion. We invite readers to send suggestions,
comments, criticisms, and corrections to the authors at the Anderson Graduate
School of Management, University of California, Los Angeles, CA 90024. Any
form of communication will be welcome.
1
2 THE THEORY OF FINANCE
Introduction: Capital
Markets, Consumption,
and Investment
A. INTRODUCTION
The objective of this chapter is to study consumption and investment decisions made
by individuals and firms. Logical development is facilitated if we begin with the sim-
plest of all worlds, a one-person/one-good economy. The decision maker, Robinson
Crusoe, must choose between consumption now and consumption in the future. Of
course, the decision not to consume now is the same as investment. Thus Robinson
Crusoe's decision is simultaneously one of consumption and investment. In order to
decide, he needs two types of information. First, he needs to understand his own sub-
jective trade-offs between consumption now and consumption in the future. This
information is embodied in the utility and indifference curves depicted in Figs. 1.1
through 1.3. Second, he must know the feasible trade-offs between present and future
consumption that are technologically possible. These are given in the investment and
production opportunity sets of Figs. 1.4 and 1.5.
From the analysis of a Robinson Crusoe economy we will find that the optimal
consumption/investment decision establishes a subjective interest rate for Robinson
Crusoe. Shown in Fig. 1.5, it represents his (unique) optimal rate of exchange between
consumption now and in the future. Thus interest rates are an integral part of con-
sumption/investment decisions. One can think of the interest rate as the price of
3
4 INTRODUCTION: CAPITAL MARKETS, CONSUMPTION, AND INVESTMENT
deferred consumption or the rate of return on investment. After the Robinson Crusoe
economy we will introduce opportunities to exchange consumption across time by
borrowing or lending in a multiperson economy (shown in Fig. 1.7). The introduction
of these exchange opportunities results in a single market interest rate that everyone
can use as a signal for making optimal consumption/investment decisions (Fig. 1.8).
Furthermore, no one is worse off in an exchange economy when compared with a
Robinson Crusoe economy and almost everyone is better off (Fig. 1.9). Thus an ex-
change economy that uses market prices (interest rates) to allocate resources across
time will be seen to be superior to an economy without the price mechanism.
The obvious extension to the introductory material in this chapter is the invest-
ment decision made by firms in a multiperiod context. Managers need optimal deci-
sion rules to help in selecting those projects that maximize the wealth of shareholders.
We shall see that market-dete'rmined interest rates play an important role in the cor-
porate investment and production decisions. This material will be discussed in depth
in Chapters 2 and 3.
The answer to the question "Do capital markets benefit society?" requires that we
compare a world without capital markets to one with them and show that no one is
worse off and that at least one individual is better off in a world with capital markets.
To make things as simple as possible, we assume that all outcomes from investment
are known with certainty, that there are no transactions costs or taxes, and that deci-
sions are made in a one-period context. Individuals are endowed with income (manna
from heaven) at the beginning of the period, y o , and at the end of the period, y,.
They must decide how much to actually consume now, Co , and how much to invest
in productive opportunities in order to provide end-of-period consumption, C 1 . Every
individual is assumed to prefer more consumption to less. In other words, the mar-
ginal utility of consumption is always positive. Also, we assume that the marginal
utility of consumption is decreasing. The total utility curve (Fig. 1.1) shows the utility
of consumption at the beginning of the period, assuming that the second-period con-
sumption is held constant. Changes in consumption have been marked off in equal
increments along the horizontal axis. Note that equal increases in consumption cause
total utility to increase (marginal utility is positive), but that the increments in utility
become smaller and smaller (marginal utility is decreasing). We can easily construct
a similar graph to represent the utility of end-of-period consumption, U(C 1 ). When
combined with Fig. 1.1, the result (the three-dimensional graph shown in Fig. 1.2)
provides a description of trade-offs between consumption at the beginning of the
period, Co , and consumption at the end of the period, C 1. The dashed lines represent
contours along the utility surface where various combinations of C o and C1 provide
the same total utility (measured along the vertical axis). Since all points along the
same contour (e.g., points A and B) have equal total utility, the individual will be in-
different with respect to them. Therefore the contours are called indifference curves.
CONSUMPTION AND INVESTMENT WITHOUT CAPITAL MARKETS 5
Consumption, Co
Looking at Fig. 1.2 from above, we can project the indifference curves onto the con-
sumption argument plane (i.e., the plane formed by the Co, C= axes in Fig. 1.3). To
reiterate, all combinations of consumption today and consumption tomorrow that
lie on the same indifference curve have the same total utility. The decision maker
whose indifference curves are depicted in Fig. 1.3 would be indifferent as to point A
with consumption (C oa , C ia) and point B with consumption (C oî, Cy)). Point A has
more consumption at the end of the period but less consumption at the beginning
than point B does. Point D has more consumption in both periods than do either
points A or B. Point D lies on an indifference curve with higher utility than points
A and B; hence curves to the northeast have greater total utility.
U(Co , ) U(C=)
/
/ /
/ /
/ /
/ / /
/ /
/ /
tzl /
K C() )
Figure =.<
Trade-offs îetween îeginning and end-of-period
consumption.
6 INTRODUCTION: CAPITAL MARKETS, CONSUMPTION, AND INVESTMENT
Figure 1.3
Indifference curves representing the time
preference of consumption.
The slope of the straight line just tangent to the indifference curve at point B
measures the rate of trade-off between C o and C, at point B. This trade-off is called
the marginal rate of substitution ( MRS) between consumption today and consumption
tomorrow. It also reveals the decision maker's subjective rate of time preference, r 1,
at point B. We can think of the subjective rate of time preference as an interest rate
because it measures the rate of substitution between consumption bundles over time.
It reveals how many extra units of consumption tomorrow must be received in order
to give up one unit of consumption today and still have the same total utility. Mathe-
matically, it is expressed as'
ac
MRS = acoi = —(1 + r i). (1.1)
U= const.
Note that the subjective rate of time preference is greater at point A than at point
B. The individual has less consumption today at point A and will therefore demand
relatively more future consumption in order to have the same total utility.
Thus far we have described preference functions that tell us how individuals will
make choices among consumption bundles over time. What happens if we introduce
productive opportunities that allow a unit of current savings/investment to be turned
into more than one unit of future consumption? We assume that each individual in
the economy has a schedule of productive investment opportunities that can be
arranged from the highest rate of return down to the lowest (Fig. 1.4). Although we
have chosen to graph the investment opportunities schedule as a straight line, any
decreasing function would do. This implies diminishing marginal returns to invest-
ment because the more an individual invests, the lower the rate of return on the mar-
ginal investment. Also, all investments are assumed independent of one another and
perfectly divisible.
Equation (1.1) can be read as follows: The marginal rate of substitution between consumption today
and end-of-period consumption, MRS2, is equal to the slope of a line tangent to an indifference curve
given constant total utility roC i /aC oil u _ consts • This in turn is equal to the individual's subjective rate of
time preference, —(1 + ri).
CONSUMPTION AND INVESTMENT WITHOUT CAPITAL MARKETS 7
X
Total investment
Figure 1.5
The production opportunity set.
See Problem 1.6 at the end of the chapter for an example of how to make the transition between the
schedule of productive investment opportunities and the consumption argument plane.
8 INTRODUCTION: CAPITAL MARKETS, CONSUMPTION, AND INVESTMENT
Individual 2
Individual I
Yo
Co
Figure 1.6
Individuals with different indifference curves choose
different production/consumption patterns.
CONSUMPTION AND INVESTMENT WITH CAPITAL MARKETS 9
2, who has a lower rate of time preference (Why?), will choose to invest more than
individual 1.
A Robinson Crusoe economy is characterized by the fact that there are no opportu-
nities to exchange intertemporal consumption among individuals. What happens if
instead of one person—many individuals are said to exist in the economy? Inter-
temporal exchange of consumption bundles will be represented by the opportunity
to borrow or lend unlimited amounts at r, a market-determined rate of interest.'
Financial markets facilitate the transfer of funds between lenders and borrowers.
Assuming that interest rates are positive, any amount of funds lent today will return
interest plus principal at the end of the period. Ignoring production for the time
being, we can graph borrowing and lending opportunities along the capital market
line in Fig. 1.7 (line WO ABW1). With an initial endowment of (yo , y i ) that has utility
equal to U1, we can reach any point along the market line by borrowing or lending
at the market interest rate plus repaying the principal amount, X,. If we designate
the future value as X 1, we can write that the future value is equal to the principal
amount plus interest earned,
X, = X 0 + rX 0 , X -= (1 + r)X 0 .
Co
Figure 1.7
The capital market line.
3
The market rate of interest is provided by the solution to a general equilibrium problem. For simplicity,
we assume that the market rate of interest is a given.
10 INTRODUCTION: CAPITAL MARKETS, CONSUMPTION, AND INVESTMENT
Similarly, the present value, Wo , of our initial endowment, (y o , y } , is the sum of cur-
rent income, Yo, and the present value of our end-of-period income, Yi(1 + r - }:
Yi
Wo = Yo + • (1.2)
(} + r
Referring to Fig. 1.7, we see that with endowment (y o , y , we will maximize utility
by moving along the market line to the point where our subjective time preference
equals the market interest rate. Point B represents the consumption bundle (Ct, , Cl)
on the highest attainable indifference curve. At the initial endowment (point A , our
subjective time preference, represented by the slope of a line tangent to the indiffer-
ence curve at point A, is less than the market rate of return. Therefore we will desire
to lend because the capital market offers a rate of return higher than what we subjec-
tively require. Ultimately, we reach a consumption decision (Co, CT) where we maxi-
mize utility. The utility, U 2 , at point B is greater than the utility, U1, at our initial
endowment, point A. The present value of this consumption bundle is also equal to
our wealth, Wo :
Wo = + • (1.3)
+ r
This can be rearranged to give the equation for the capital market line:
= Wo(} r — (} + r Q, (1.4)
and since W0(1 + r = W1, we have
Cr = — (1 + r C]. (1.5)
Thus the capital market line in Fig. 1.7 has an intercept at W1 and a slope of —(1 + r .
Also note that by equating (1.2) and (1.3) we see that the present value of our endow-
ment equals the present value of our consumption, and both are equal to our wealth,
Wo . Moving along the capital market line does not change one's wealth, but it does
offer a pattern of consumption that has higher utility.
What happens if the production/consumption decision takes place in a world
where capital markets facilitate the exchange of funds at the market rate of interest?
Figure 1.8 combines production possibilities with market exchange possibilities. With
the family of indifference curves U }, U 2 , and U3 and endowment (y o , y i at point A,
what actions will we take in order to maximize our utility? Starting at point A, we
can move either along the production opportunity set or along the capital market
line. Both alternatives offer a higher rate of return than our subjective time preference,
but production offers the higher return, i.e., a steeper slope. Therefore we choose to
invest and move along the production opportunity frontier. Without the opportunity
to borrow or lend along the capital market line, we would stop investing at point
D, where the marginal return on productive investment equals our subjective time
preference. This was the result shown for consumption and investment in a Robinson
Crusoe world without capital markets in Fig. 1.5. At this point, our level of utility
CONSUMPTION AND INVESTMENT WITH CAPITAL MARKETS 11
Co
Po Co Yo Wo W0
Figure 1.8
Production and consumption with capital markets.
Fisher separation theorem. Given perfect and complete capital markets, the pro-
duction decision is governed solely by an objective market criterion (represented
by maximizing attained wealth) without regard to individuals' subjective prefer-
ences that enter into their consumption decisions.
An important implication for corporate policy is that the investment decision
can be delegated to managers. Given the same opportunity set, every investor will
make the same production decision (P o , P 1) regardless of the shape of his or her in-
difference curves. This is shown in Fig. 1.9. Both investor 1 and investor 2 will direct
the manager of their firm to choose production combination (P o , P 1). They can then
take the output of the firm and adapt it to their own subjective time preferences by
borrowing or lending in the capital market. Investor 1 will choose to consume more
than his or her share of current production (point A) by borrowing today in the cap-
ital market and repaying out of his or her share of future production. Alternately,
investor 2 will lend because he or she consumes less than his or her share of current
production. Either way, they are both better off with a capital market. The optimal
production decision is separated from individual utility preferences. Without capital
market opportunities to borrow or lend, investor 1 would choose to produce at point
Y, which has lower utility. Similarly, investor 2 would be worse off at point X.
In equilibrium, the marginal rate of substitution for all investors is equal to the
market rate of interest, and this in turn is equal to the marginal rate of transforma-
tion for productive investment. Mathematically, the marginal rates of substitution
for investors i and j are
MRS, = MRS.] = —(1 + r) = MRT.
Thus all individuals use the same time value of money (i.e., the same market-deter-
mined objective interest rate) in making their production/investment decisions.
Figure 1.9
The investment decision is independent
of individual preferences.
MARKETPLACES AND TRANSACTIONS COSTS 13
The importance of capital markets cannot be overstated. They allow the efficient
transfer of funds between borrowers and lenders. Individuals who have insufficient
wealth to take advantage of all their investment opportunities that yield rates of
return higher than the market rate are able to borrow funds and invest more than
they would without capital markets. In this way, funds can be efficiently allocated
from individuals with few productive opportunities and great wealth to individuals
with many opportunities and insufficient wealth. As a result, all (borrowers and
lenders) are better off than they would have been without capital markets.
D. MARKETPLACES AND
TRANSACTIONS COSTS
The foregoing discussion has demonstrated the advantages of capital markets for
funds allocation in a world without transactions costs. In such a world, there is no
need for a central location for exchange; that is, there is no need for a marketplace
per se. But let us assume that we have a primitive economy with N producers, each
making a specialized product and consuming a bundle of all N consumption goods.
Given no marketplace, bilateral exchange is necessary. During a given time period,
each visits the other in order to exchange goods. The cost of each leg of a trip is T
dollars. If there are five individuals and five consumption goods in this economy,
then individual 1 makes four trips, one to each of the other four producers. Individual
2 makes three trips, and so on. Altogether, there are [N(N — 1)]/2 = 10 trips, at a
total cost of 10T dollars. This is shown in Fig. 1.10. If an entrepreneur establishes a
central marketplace and carries an inventory of each of the N products, as shown
in Fig. 1.11, the total number of trips can be reduced to five, with a total cost of 5T
dollars. Therefore if the entrepreneur has a total cost (including the cost of living) of
less than 10T — 5T dollars, he or she can profitably establish a marketplace and
everyone will be better off.'
Figure 1.10
A primitive exchange economy with no
central marketplace.
In general, for N individuals making two-way exchanges, there are (I) = N(N — 1)/2 trips. With a mar-
ketplace the number of trips is reduced to N. Therefore the savings. is [N(N — 1)/2 — N]T .
14 INTRODUCTION: CAPITAL MARKETS, CONSUMPTION, AND INVESTMENT
Figure 1.11
The productivity of a central
marketplace.
Borrowing Rate
Individual 1
Lending rate
Individual 2
Co
Figure 1.12
MarketsWith different borrowing and lending rates.
PROBLEM SET 15
SUMMARY
The rest of the text follows almost exactly the same logic as this chapter, except that
from Chapter 4 onward it focuses on decision making under uncertainty. The first
step is to develop indifference curves to model individual decision making in a world
with uncertainty. Chapter 4 is analogous to Fig. 1.3. It will describe a theory of choice
under uncertainty. Next, the portfolio opportunity set, which represents choices among
combinations of risky assets, is developed. Chapters 5 and 6 are similar to Fig. 1.5.
They describe the objects of choice the portfolio opportunity set. The tangency be-
tween the indifference curves of a risk-averse investor and his or her opportunity set
provides a theory of individual choice in a world without capital markets (this is dis-
cussed in Chapter 6). Finally, in Chapter 7, we introduce the opportunity to borrow
and lend at a riskless rate and develop models of capital market equilibrium. Chapter
7 follows logic similar to Fig. 1.8. In fact, we show that a type of separation principle
(two-fund separation) obtains, given uncertainty and perfect capital markets. Chapters
10 and 11 take a careful look at the meaning of efficient capital markets and at
empirical evidence that relates to the question of how well the perfect capital market
assumption fits reality. The remainder of the book, following Chapter 11, applies
financial theory to corporate policy decisions.
PROBLEM SET
1.1 Graphically demonstrate the Fisher separation theorem for the case where an individual
ends up lending in financial markets. Label the following points on the graph: initial wealth,
Wo ; optimal production/investment (Po , P1 ); optimal consumption (Co, CD; present value of
final wealth, W.
16 INTRODUCTION: CAPITAL MARKETS, CONSUMPTION, AND INVESTMENT
1.2 Graphically analyze the effect of an exogenous decrease in the interest rate on (a) the
utility of borrowers and lenders, (b) the present wealth of borrowers and lenders, and (c) the
investment in real assets.
1.3 The interest rate cannot fall below the net rate from storage. True or false? Why?
1.4 Graphically illustrate the decision-making process faced by an individual in a Robinson
Crusoe economy where (a) storage is the only investment opportunity and (b) there are no
capital markets.
1.5 Suppose that the investment opportunity set has N projects, all of which have the same
rate of return, R*. Graph the investment set.
1.6 Suppose your production opportunity set in a world with perfect certainty consists of the
following possibilities:
A $1,000,000 8%
B 1,000,000 20
C 2,000,000 4
D 3,000,000 30
REFERENCES
Investment Decisions:
The Certainty Case
A. INTRODUCTION
The investment decision is essentially how much not to consume in the present in
order that more can be consumed in the future. The optimal investment decision
maximizes the expected satisfaction (expected utility) gained from consumption over
the planning horizon of the decision maker. We assume that all economic decisions
ultimately reduce to questions about consumption. Even more fundamentally, con-
sumption is related to survival.
The consumption/investment decision is important to all sectors of the economy.
An individual who saves does so because the expected benefit of future consumption
provided by an extra dollar of saving exceeds the benefit of using it for consumption
today. Managers of corporations, who act as agents for the owners (shareholders)
of the firm, must decide between paying out earnings in the form of dividends, which
may be used for present consumption, and retaining the earnings to invest in pro-
ductive opportunities that are expected to yield future consumption. Managers of
not-for-profit organizations try to maximize the expected utility of contributors—
those individuals who provide external funds. And public sector managers attempt
to maximize the expected utility of their constituencies.
The examples of investment decisions in this chapter are taken from the corpo-
rate sector of the economy, but the decision criterion, which is to maximize the
present value of lifetime consumption, can be applied to any sector of the economy.
For the time being, we assume that intertemporal decisions are based on knowledge
of the market-determined time value of money the interest rate. Furthermore, the
17
18 INVESTMENT DECISIONS : THE CERTAINTY CASE
interest rate is assumed to be known with certainty in all time periods. It is non-
stochastic. That is, it may change over time, but each change is known with certainty.
The interest rate is assumed not to be a random variable. In addition, all future
payoffs from current investment decisions are known with certainty. And finally, there
are no imperfections (e.g., transactions costs) in capital markets. These assumptions
are obviously an oversimplification, but they are a good place to start. Most of the
remainder of the text after this chapter is devoted to decision making under uncer-
tainty. But for the time being it is useful to establish the fundamental criterion of
economic decision making the maximization of the net present value of wealth,
assuming perfect certainty.
The most important theme of this chapter is that the objective of the firm is to
maximize the wealth of its shareholders. This will be seen to be the same as maxi-
mizing the present value of shareholders' lifetime consumption and no different than
maximizing the price per share of stock. Alternative issues such as agency costs are
also discussed. Then the maximization of shareholder wealth is more carefully de-
fined as the discounted value of future expected cash flows. Finally, techniques for
project selection are reviewed, and the net present value criterion is shown to be
consistent with shareholder wealth maximization.
To say that the goal of the firm is the maximization of its shareholders' wealth is
one thing, but the problem of how to do it is another. We know that interpersonal
comparison of individuals' utility functions is not possible. For example, if we give
individuals A and B $100 each, they will both be happy. However, no one, not even
the two individuals, will be able to discern which person is happier. How then can
a manager maximize shareholders' utility when individual utility functions cannot
be compared or combined?
The answer to the question is provided if we turn to our understanding of the
role of capital markets. If capital markets are perfect in the sense that they have no
frictions that cause the borrowing rate to be different from the lending rate, then (as
we saw in Chapter 1) Fisher separation obtains. This means that individuals can
delegate investment decisions to the manager of the firm in which they are owners.
Regardless of the shape of the shareholders' individual utility functions, the man-
agers maximize the owners' individual (and collective) wealth positions by choosing
to invest until the rate of return on the least favorable project is exactly equal to
the market-determined rate of return. This result is shown in Fig. 2.1. The optimal
production/investment decision, (P o , P1), is the one that maximizes the present value
of the shareholders' wealth, Wo . The appropriate decision rule is the same, indepen-
dent of the shareholders' time preferences for consumption. The manager will be
directed, by all shareholders, to undertake all projects that earn more than the market
rate of return.
FISHER SEPARATION 19
Individual 1
Individual 2
Co
Figure 2.1
Separation of shareholder preferences from the
production/investment decision.
If the marginal return on investment equals the market-determined opportunity
cost of capital, then the shareholders' wealth, Wo , is maximized. Individual share-
holders can then take the optimal production decision (Po, P 1) and borrow or lend
along the capital market line in order to satisfy their time pattern for consumption.
In other words, they can take the cash payouts from the firm and use them for current
consumption or save them for future consumption, according to their individual
desires.
The separation principle implies that the maximization of the shareholders'
wealth is identical to maximizing the present value of their lifetime consumption.
Mathematically, this was demonstrated in Eq. (1.3):
Ci
Wo = Co + + r•
Even though the two individuals in Fig. 2.1 choose different levels of current and
future consumption, they have the same current wealth, Wo . This follows from the
fact that they receive the same income from productive investments (P o , P 1 ).
Because exchange opportunities permit borrowing and lending at the same rate
of interest, an individual's productive optimum is independent of his or her resources
and tastes. Therefore if asked to vote on their preferred production decisions at a
shareholders' meeting, different shareholders of the same firm will be unanimous in
their preference. This is known as the unanimity principle. It implies that the man-
agers of the firm, in their capacity as agents of the shareholders, need not worry
about making decisions that reconcile differences of opinion among shareholders.
All shareholders will have identical interests. In effect, the price system by which
wealth is measured conveys the shareholders' unanimously preferred productive de-
cisions to the firm.
20 INVESTMENT DECISIONS: THE CERTAINTY CASE
So far, we have shown that in perfect markets all shareholders will agree that man-
agers should follow a simple investment decision rule: Take projects until the mar-
ginal rate of return equals the market-determined discount rate. Therefore the
shareholders' wealth is seen to be the present value of cash flows discounted at the
opportunity cost of capital (the market-determined rate).
Shareholders can agree on the decision rule that they should give to managers.
But they must be able to costlessly monitor management decisions if they are to be
sure that management really does make every decision in a way that maximizes their
wealth. There is obviously a difference between ownership and control, and there is
no reason to believe that the manager, who serves as an agent for the owners, will
always act in the best interest of the shareholders. In most agency relationships the
owner will incur nontrivial monitoring costs in order to keep the agent in line. Con-
sequently, the owner faces a trade-off between monitoring costs and forms of com-
pensation that will cause the agent to always act in the owner's interest. At one extreme,
if the agent's compensation were all in the form of shares in the firm, then monitor-
ing costs would be zero. Unfortunately, this type of scheme is practically impossible
because the agent will always be able to receive some compensation in the form of
nonpecuniary benefits such as larger office space, expensive lunches, an executive
jet, etc. At the opposite extreme, the owner would have to incur inordinate monitoring
costs in order to guarantee that the agent always makes the decision the owner would
prefer. Somewhere between these two extremes lies an optimal solution. The reader
who wishes to explore this classic problem in greater depth is referred to books by
Williamson [1964], Marschak and Radner [1972], and Cyert and March [1963],
and to articles by Jensen and Meckling [1976], Machlup [1967], Coase [1937], and
Alchian and Demsetz [1972] as good references to an immense literature in this area.
The issue is also explored in greater depth in Chapter 14 of this text.
In spite of the above discussion, we shall assume that managers always make
decisions that maximize the wealth of the firm's shareholders. To do so, they must
find and select the best set of investment projects to accomplish their objective.
D. MAXIMIZATION OF
SHAREHOLDERS' WEALTH
after-tax cash flows paid out by the firm.' After-tax cash flows available for consump-
tion can be shown to be the same as the stream of dividends, Dive, paid to shareholders.
The discounted value of the stream of dividends is
coDiv,
So = E
, =0 (1 + ks)
(2.1)
where So is the present value of shareholders' wealth (in Fig. 2.1 it is Wo) and k s is
the market-determined rate of return on equity capital (common stock).
Equation (2.1) is a multiperiod formula that assumes that future cash flows paid
to shareholders are known with certainty and that the market-determined discount
rate is nonstochastic and constant over all time periods. These assumptions are main-
tained throughout this chapter because our main objective is to understand how
the investment decision, shown graphically in Fig. 2.1 in a one-period context, can
be extended to the more practical setting of many time periods in a manner consistent
with the maximization of the shareholders' wealth. For the time being, we shall ig-
nore the effect of personal taxes on dividends, and we shall assume that the discount
rate, k s, is the market-determined opportunity cost of capital for equivalent income
streams. It is determined by the slope of the market line in Fig. 2.1.
One question that often arises is: What about capital gains? Surely shareholders
receive both capital gains and dividends; why then do capital gains not appear in Eq.
(2.1)? The answer to this question is that capital gains do appear in Eq. (2.1). This can
be shown by use of a simple example. Suppose a firm pays a dividend, Div l , of $1.00
at the end of this year and $1.00(1 + g)e at the end of each year thereafter, where the
growth rate of the dividend stream is g. If the growth rate in dividends, g, is 5% and
the opportunity cost of investment, k„ is 10%, how much will an investor pay today
for the stock? Using the formula for the present value of a growing annuity stream,
we get 2
Div, $1.00
So = = $20.00.
ks — g .10 — .05
Next, suppose that an investor bought the stock today for $20 and held it for five
years. What would it be worth at the end of the fifth year?
Div6
S5 = •
ks — g
Since much of the rest of this chapter assumes familiarity with discounting, the reader is referred to
Appendix A for a review.
2 The formula used here, sometimes called the Gordon growth model, is derived in Appendix A. It as-
sumes that the dividend grows forever at a constant rate, g, which is less than the discount rate, g < ks.
22 INVESTMENT DECISIONS: THE CERTAINTY CASE
Therefore the value of the stock at the end of the fifth year would be
$1.2763
S, = = $25.5256.
.10 — .05
The value of the stock at the end of the fifth year is the discounted value of all div-
idends from that time on. Now we can compute the present value of the stream of
income of an investor who holds the stock only five years. He or she gets five divi-
dend payments plus the market price of the stock in the fifth year. The discounted
value of these payments is So .
Except for a one-cent rounding difference, the present value of the stock is the same
whether an investor holds it forever or for only, say, five years. Since the value of the
stock in the fifth year is equal to the future dividends from that time on, the value
of dividends for five years plus a capital gain is exactly the same as the value of an in-
finite stream of dividends. Therefore Eq. (2.1) is the discounted value of the stream of
cash payments to shareholders and is equivalent to the shareholders' wealth. Because
we are ignoring the taxable differences between dividends and capital gains (this will
be discussed in Chapter 15, "Dividend Policy"), we can say that Eq. (2.1) incorporates
all cash payments, both dividends and capital gains.
and payoffs resulting from merger or acquisition. Stock dividends, which involve no
cash flow, are not included in our definition of dividends.
We can use a very simple model to show the difference between the economic
definition of profit and the accounting definition. Assume that we have an all-equity
firm and that there are no taxes.' Then sources of funds are revenues, Rev, and sale
of new equity (on m shares at S dollars per share). Uses of funds are wages, salaries,
materials, and services, W&S; investment, I; and dividends, Div. For each time period,
t, we can write the equality between sources and uses of funds as
Rev, + m,S, = Div, + (W&S), + It . (2.2)
To simplify things even further, assume that the firm issues no new equity, i.e., m,S, = 0.
Now we can write dividends as
Div, = Rev, — (W&S), — It , (2.3)
which is the simple cash flow definition of profit. Dividends are the cash flow left
over after costs of operations and new investment are deducted from revenues. Using
Eq. (2.3) and the definition of shareholders' wealth [Eq. (2.1)], we can rewrite share-
holders' wealth as
c° Rev, — (W&S), —
So = (2.4)
t=o (1 + k s)t
The accounting definition of profit does not deduct gross investment, It , as investment
outlays are made. Instead, the book value of new investment is capitalized on the
balance sheet and written off at some depreciation rate, dep. The accounting definition
of profit is net income,
NI, = Rev, — (W&S), — dep t . (2.5)
Let AA, be the net change in the book value of assets during a year. The net change
will equal gross new investment during the year, I t , less the change in accumulated
depreciation during the year, dep,:
We already know that the accounting definition of profit, NI„ is different from the
economic definition, Div,. However, it can be adjusted by subtracting net investment.
This is done in Eq. (2.7):
Rev, — (W&S), — dept — (It — dept)
So =
-0 (1 +
E NI t — AA, (2.7)
t=o (I + kjt
3
The conclusions to be drawn from the model do not change if we add debt and taxes, but the arithmetic
becomes more complex.
24 INVESTMENT DECISIONS: THE CERTAINTY CASE
The main difference between the accounting definition and the economic defini-
tion of profit is that the former does not focus on cash flows when they occur, whereas
the latter does. The economic definition of profit, for example, correctly deducts the
entire expenditure for investment in plant and equipment at the time the cash outflow
occurs.
Financial managers are frequently misled when they focus on the accounting
definition of profit, or earnings per share. The objective of the firm is not to maximize
earnings per share. The correct objective is to maximize shareholders' wealth, which
is the price per share that in turn is equivalent to the discounted cash flows of the
firm. There are two good examples that point out the difference between maximizing
earnings per share and maximizing discounted cash flow. The first example is the
difference between FIFO (first-in, first-out) and LIFO (last-in, first-out) inventory
accounting during inflation. Earnings per share are higher if the firm adopts FIFO
inventory accounting. The reason is that the cost of manufacturing the oldest items
in inventory is less than the cost of producing the newest items. Consequently, if the
cost of the oldest inventory (the inventory that was first in) is written off as expense
against revenue, earnings per share will be higher than if the cost of the newest items
(the inventory that was in last) is written off. A numerical example is given in Table
2.1. It is easy to see how managers might be tempted to use FIFO accounting tech-
niques. Earnings per share are higher. However, FIFO is the wrong technique to use in
an inflationary period because it minimizes cash flow by maximizing taxes. In our
example, production has taken place during some previous time period, and we are
trying to make the correct choice of inventory accounting in the present. The sale
of an item from inventory in Table 2.1 provides $100 of cash inflow (revenue) regard-
less of which accounting system we are using. Cost of goods sold involves no current
cash flow, but taxes do. Therefore with FIFO, earnings per share are $0.45, but cash
flow per share is ($100 — $30)/100 shares, which equals $0.70 per share. On the other
hand, with LIFO inventory accounting, earnings per share are only $0.06, but cash
flow is ($100 — $4)/100 shares, which equals $0.96 per share. Since shareholders care
only about discounted cash flow, they will assign a higher value to the shares of the
company using LIFO accounting. The reason is that LIFO provides higher cash
flow because it pays lower taxes to the government.' This is a good example of the
In 1979 the Internal Revenue Service estimated that if every firm that could have switched to LIFO had
actually done so, approximately $18 billion less corporate taxes would have been paid.
TECHNIQUES FOR CAPITAL BUDGETING 25
E. TECHNIQUES FOR
CAPITAL BUDGETING
5
See Chapter 11 for a discussion of empirical research on this issue.
6
See Chapter 11 for a discussion of empirical evidence relating to this issue.
26 INVESTMENT DECISIONS: THE CERTAINTY CASE
Three major problems face managers when they make investment decisions.
First, they have to search out new opportunities in the marketplace or new technol-
ogies. These are the basis of growth. Unfortunately, the Theory of Finance cannot
help with this problem. Second, the expected cash flows from the projects have to
be estimated. And finally, the projects have to be evaluated according to sound deci-
sion rules. These latter two problems are central topics of this text. In the remainder
of this chapter we look at project evaluation techniques assuming that cash flows
are known with certainty, and in Chapter 12 we will assume that cash flows are
uncertain.
Investment decision rules are usually referred to as capital budgeting techniques.
The best technique will possess the following essential property: It will maximize
shareholders' wealth. This essential property can be broken down into separate
criteria:
• All cash flows should be considered.
• The cash flows should be discounted at the opportunity cost of funds.
• The technique should select from a set of mutually exclusive projects the one
that maximizes shareholders' wealth.
• Managers should be able to consider one project independently from all others
(this is known as the value-additivity principle).
The last two criteria need some explanation. Mutually exclusive projects are a set
from which only one project can be chosen. In other words, if a manager chooses
to go ahead with one project from the set, he or she cannot choose to take on any
of the others. For example, there may be three or four different types of bridges that
could be constructed to cross a river at a given site. Choosing a wooden bridge ex-
cludes other types, e.g., steel. Projects are also categorized in other ways. Independent
projects are those that permit the manager to choose to undertake any or all, and
contingent projects are those that have to be carried out together or not at all. For
example, if building a tunnel also requires a ventilation system, then the tunnel and
ventilation system should be considered as a single, contingent project.
The fourth criterion, the value-additivity principle, implies that if we know the
value of separate projects accepted by management, then simply adding their values,
Vf , will give us the value of the firm, V. In mathematical terms, if there are N projects,
then the value of the firm is
N
V= (2.8)
j =1
This is a particularly important point because it means that projects can be consid-
ered on their own merit without the necessity of looking at them in an infinite variety
of combinations with other projects.
There are four widely used capital budgeting techniques: (1) the payback method,
(2) the accounting rate of return, (3) the net present value, and (4) the internal rate
of return. Our task is to choose the technique that best satisfies the four desirable
properties discussed above. It will be demonstrated that only one technique—the net
TECHNIQUES FOR CAPITAL BUDGETING 27
present value method—is correct. It is the only technique that is always consistent
with shareholder wealth maximization.
To provide an example for discussion, Table 2.2 lists the estimates of cash flow
for four projects, each of which has a five-year life. Since they are mutually exclusive,
there is only one that will maximize the price of the firm's stock; i.e., there is only
one that will maximize shareholders' wealth. We would normally assume at this point
that all four projects are equally "risky." However, according to the assumption used
throughout this chapter, their cash flows are known with certainty; therefore their
risk is zero. The appropriate discount rate in a world with no risk is the risk-free
rate (e.g., the Treasury Bill rate).
It is not too hard to find real-world examples of projects that have negative future cash flows and can-
not îe aîandoned. A good example is nuclear power plants; at the end of their useful life they must îe
decommissioned at consideraîle expense.
28 INVESTMENT DECISIONS: THE CERTAINTY CASE
in its choice between project A and a second project that paid $900 in the first year
and $100 in the second. Both projects would have the same payback period. We
reject the payback method because it violates (at least) the first two of the four
properties that are desirable in capital budgeting techniques.'
See Problem 2.10 at the end of the chapter. It demonstrates that the payback technique also violates
the value-additivity principle.
TECHNIQUES FOR CAPITAL BUDGETING 29
We have discounted each of the cash flows back to the present and summed them.
Mathematically, this can be written as
E NCFt
N
NPV = (2.10)
t= I. (1 + k)t °'
where NCF, is the net cash flow in time period t, l o is the initial cash outlay, k is
the firm's weighted average cost of capital, and N is the number of years in the
project. The net present values of the four projects are:
Project A, NPV = —407.30;
Project B, NPV = 510.70;
Project C, NPV = 530.85;
Project D, NPV = 519.20.
If these projects were independent instead of mutually exclusive, we would reject A
and accept B, C, and D. (Why?) Since they are mutually exclusive, we select the proj-
ect with greatest NPV, project C. The NPV of a project is exactly the same as the
increase in shareholders' wealth. This fact makes it the correct decision rule for capital
budgeting purposes. More will be said about this when we compare the NPV rule
with the internal rate of return.
9
The reader who wishes to brush up on the algebra of discounting is referred to Appendix A.
30 INVESTMENT DECISIONS: THE CERTAINTY CASE
We can solve for the IRR on project C by trial and error. (Most pocket calculators
have programs that can quickly solve for the IRR by using similar iterative tech-
niques.) This is done in Table 2.3 and graphed in Fig. 2.2.
Figure 2.2 shows that the NPV of the given set of cash flows decreases as the
discount rate is increased. If the discount rate is zero, there is no time value of money,
and the NPV of a project is simply the sum of its cash flows. For project C, the NPV
equals $1250 when the discount rate is zero. At the opposite extreme, if the discount
rate is infinite, then future cash flows are valueless, and the NPV of project C is its
current cash flow, - $1000. Somewhere between these two extremes is a discount rate
that makes the present value equal to zero. Called the IRR on the project, this rate
equates the present value of cash inflows with the present value of cash outflows. The
IRRs for the four projects are:
Project A, IRR = -200%;
Project B, IRR = 20.9%;
Project C, IRR = 22.8%;
Project D, IRR = 25.4%.
If we use the IRR criterion and the projects are independent, we accept any project
that has an IRR greater than the opportunity cost of capital, which is 10%. Therefore
Discount rate
Figure <.<
NPV of project C at different discount rates.
COMPARISON OF NET PRESENT VALUE WITH INTERNAL RATE OF RETURN 31
we would accept projects B, C, and D. However, since these projects are mutually
exclusive, the IRR rule leads us to accept project D as the best.
As the example shows, the net present value and the internal rate of return can favor
conflicting project choices. The net present value favors project C, whereas the IRR
favors project D. Both techniques consider all cash flows and both use the concept
of the time value of money in order to discount cash flows. However, we must choose
from among the four mutually exclusive projects the one project that maximizes
shareholders' wealth. Consequently, only one of the two techniques can be correct.
We shall see that the NPV criterion is the only one that is necessarily consistent with
maximizing shareholders' wealth.
Figure 2.3 compares projects B, C, and D. For very low discount rates, project B
has the highest net present value; for intermediate discount rates, project C is best;
and for high discount rates, project D is best. The NPV rule compares the three proj-
ects at the same discount rate. Remember, 10% was not arbitrarily chosen. It is the
market-determined opportunity cost of capital. We saw earlier in the chapter that
this market-determined discount rate is the one managers should use if they desire
to maximize the wealth of all shareholders. Consequently, no other discount rate is
appropriate. Project C is the best project because it gives the greatest NPV when the
opportunity cost of funds invested is 10%.
The IRR rule does not discount at the opportunity cost of capital. Instead, it
implicitly assumes that the time value of money is the IRR, since all cash flows are
NPV
1500
B
C
D
1000
500
30%
Rate
10% 20% Project D
ft.. Project C
Project B
—
500
Figure 2.3
Comparison of three mutually exclusive projects.
32 INVESTMENT DECISIONS: THE CERTAINTY CASE
discounted at that rate. This implicit assumption has come to be called the reinvest-
ment rate assumption.
—1,600
1 10,000
2 —10,000
10
The number of combinations for five projects is
(5) + ( 5) + r5
)+ ) + ( 5) + (5) = 32
75
\OJ \I \2)
Imagine the number of combinations that would have to be considered if there were 50 projects.
34 INVESTMENT DECISIONS: THE CERTAINTY CASE
the second year, the high-speed pump produces $10,000 less oil because the well has
been depleted. The question is whether or not to accept the rapid pumping technique,
which speeds up cash flows in the near term at the expense of cash flows in the long
term. Figure 2.4 shows the NPV of the project for different discount rates. If the
opportunity cost of capital is 10%, the NPV rule would reject the project because
it has negative NPV at that rate. If we are using the IRR rule, the project has two
IRRs, 25% and 400%. Since both exceed the opportunity cost of capital, the project
would probably be accepted.
Mathematically, the multiple IRRs are a result of Descartes' rule of signs, which
implies that every time the cash flows change signs, there may be a new (positive, real)
root to the problem solution. For the above example, the signs of cash flows change
twice. The IRR is the rate that causes the discounted value of the cash flows to equal
zero. Hence we solve the following equation for IRR:
—1,600 10,000 —10,000
NPV = 0 = + 1 +
(1 + IRR)° (1 + IRR) (1 + IRR)2 '
—1,600(1 + IRR)2 + 10,000(1 + IRR) — 10,000
0=
(1 + IRR)2
0 = 1,600(1 + IRR)2 — 10,000(1 + IRR) + 10,000.
This is clearly a quadratic equation and has two roots. It has the general form
ax 2 + bx + c 0
and can be solved using the quadratic formula
—b + — 4ac
x= •
2a
NPV
Rate
Figure 2.4
Multiple internal rates of return.
COMPARISON OF NET PRESENT VALUE WITH INTERNAL RATE OF RETURN 35
as the IRR. For the example of the oil-well pump, we see that when it is viewed prop-
erly, the IRR gives the same answer as the NPV. We should reject the project be-
cause the internal rate of return is less than the opportunity cost of capital.
Up to this point we have made the implicit assumptions that the firm has no debt
and that there are no corporate taxes. This section adds a note of realism by pro-
viding a definition of cash flows for capital budgeting purposes, given debt and taxes.
In particular, we shall see that some cash flows, such as interest paid on debt and re-
payment of principal on debt, should not be considered cash flows for capital bud-
geting purposes. At the same time, we shall demonstrate, by using an example, that
there is only one definition of cash flows that is consistent with shareholder wealth
maximization.
To understand discounted cash flows it is also necessary to have a rudimentary
understanding of the opportunity cost of capital of the firm. Chapter 13 discusses the
cost of capital in great depth; however, the basics will be given here. The firm receives
its investment funds from two classes of investors: creditors and shareholders. They
provide debt and equity capital, respectively. Both groups expect to receive a rate of
return that compensates them for the level of risk they accept." Debt holders receive
a stream of fixed payments and can force the firm into receivership or bankruptcy
if they do not receive payment. On the other hand, shareholders receive the firm's
11
The assumption that future cash flows are known with certainty must be relaxed at this point, in order
to allow risk-free debt and risky equity. The reader who is interested in the related theoretical problems
is referred to Chapters 13 and 14.
CASH FLOWS FOR CAPITAL BUDGETING PURPOSES 37
residual cash flows that remain after all other payments are made. Consequently, the
interest rate paid to debt holders is less than the required rate of return on equity
because debt is less risky.
It is important to understand that projects undertaken by the firm must earn
enough cash flow to provide the required rate of return to creditors, repayment of
the face amount of debt, and payment of expected dividends to shareholders. Only
when cash flows exceed these amounts will there be any gain in shareholders' wealth.
When we discount cash flows at the weighted average cost of capital, this is exactly
what we are saying. A positive NPV is achieved only after creditors and shareholders
receive their expected risk-adjusted rates of return.
In order to provide an example of this very important concept, consider the fol-
lowing (somewhat artificial) situation. A firm is going to be created from scratch. It
will require an initial investment, I, of $1000 for equipment that will depreciate at
the rate of $200 per year. Owners have decided to borrow $500 at 10% interest. In
other words, the before-tax coupon rate on debt capital, k d , is 10%. The expected
annual cash flows for the project are implicit in the pro forma income statement
given in Table 2.6. We shall assume that shareholders require a rate of return of 30%
in order to compensate them for the riskiness of their position. Thus the cost of
equity, k 5, is 30%.
To provide the simplest possible example, assume that all the cash flows are per-
petual, i.e., the firm has no growth. This assumption has the effect of keeping the
firm's market value debt-to-equity ratio constant through time.' Perpetual cash
flows are obtained, first, by writing a consol bond that never matures and pays a
coupon of $50 each year; and second, by investing $200 annually to replace the
depreciation of the equipment.
Table 2.7 details the exact cash flows assuming that the project is held for five
years. At the end of five years the firm will be sold for its market value. Shareholders
will receive the cash, use some of it ($500) to pay off bondholders, and keep the
remainder.
12 Without a constant debt-to-equity ratio, the weighted average cost of capital would change through
Current cash flows are $500 provided by creditors and $500 from equity holders;
outflows are $1000 paid for the equipment. In years 1 through 5 the project returns
$700 in cash after the cash costs of production ($600) are subtracted from revenues
($1300). Then depreciation, a noncash charge ($200), is deducted, leaving $500 in
earnings before interest and taxes. The deduction of $50 of interest expenses leaves
taxable income of $450. After taxes, there is $225 in net income. To compute free
cash flows available for payment to shareholders, depreciation ($200), a noncash
charge, must be added back, and replacement investment ($200), a cash outflow, must
be subtracted. Thus residual cash flow available to shareholders is $225 per year.
Shareholders' wealth, S, is the present value of their stream of residual cash flows,
discounted at the cost of equity capital, k 3 = 30%. Recalling that their stream of re-
sidual cash flows continues forever, we can compute their wealth as shown below.'
— Residual cash flow $225
S = $750.
ks .3
The present value of bondholders' wealth, B, is the present value of their perpetual
stream of coupon payments discounted at the market cost of debt, kb:
Interest payments = $50
B = = $500.
kî .10
Thus we see that the market value of the firm, V, is expected to be
V = B S = $500 + $750 = $1250.
Note that the present value of debt and equity are not affected by the fact that
they will be sold at the end of year 5. The new bondholders and shareholders simply
take over ownership of their streams of cash, paying $500 and $750, respectively. As
shown in the last row of Table 2.7, the shareholders receive $1250 in year 5 for the
firm but must pay $500 to bondholders. Note also that the present value of share-
holders' wealth is $750, but they had put up $500 of the initial investment. Therefore
their change in wealth, AS, is $750 minus $500, which equals $250. We shall see that
this is exactly the same thing as the NPV of the project.
Instead of working through the complicated procedure given above, it will be
easier to analyze capital budgeting projects by defining cash flows fo] capital budgeting
pu]poses and discounting them at the firm's weighted average cost of capital. First,
what is the weighted average cost of capital (k = WACC)? As shown in Eq. (2.12)
below, it is the after-tax market cost of debt capital, k b(1 'r e), multiplied by the
percentage of the market value of the firm owned by creditors, [B/(B + S)], plus the
cost of equity, ks , multiplied by the percentage of the firm's value owned by share-
holders, [S/(B + S)]. Note that 2, is the firm's marginal tax rate.
14
Depreciation is a noncash charge against revenues. If there are other noncash charges, they should also
be included here.
15
An equivalent definition is
NCF for cap. budgeting = ANI + Adep + (1 — T d)A(k d D) — Al, (2.13a)
where ANT stands for the change in net income, the accounting definition of profit, and Akd D is the change
in the coupon rate, k d , on debt times the change in the face value of debt, D. Although sometimes easier
to use, it obscures the difference between cash flows for capital budgeting purposes and the accounting
definition of profit.
40 INVESTMENT DECISIONS: THE CERTAINTY CASE
Operating NCF =
Year Cash Flow Depreciation Tax* Cash Flow
0 —1000 —1000
1 700 200 250 250
2 700 200 250 250
3 700 200 250 250
4 700 200 250 250
5 700 200 250 250 + 1250
individual projects may require that debt be paid off over the life of the project. With-
out this assumption, the cost of capital would have to change each time period.
Another relevant issue worth pointing out is that the definition of cash flows for
capital budgeting purposes includes all incremental cash flows attributable to a proj-
ect. Too often, analysts forget that the total investment in a project includes working
capital requirements as well as the cash outlays for buildings and equipment. Working
capital includes any changes in short-term balance sheet items such as increases in
inventories, accounts receivable, and accounts payable that are expected to result
from undertaking a project. Net working capital requirements are the difference be-
tween changes in short-term assets and short-term liabilities.
PROBLEM SET
2.1 Basic capital budgeting problem with straight-line depreciation. The Roberts Company has
cash inflows of $140,000 per year on project A and cash outflows of $100,000 per year. The
investment outlay on the project is $100,000; its life is 10 years; the tax rate, 'C c , is 40%. The
opportunity cost of capital is 12%.
42 INVESTMENT DECISIONS: THE CERTAINTY CASE
a) Present two alternative formulations of the net cash flows adjusted for the depreciation
tax shelter.
b) Calculate the net present value for project A, using straight-line depreciation for tax
purposes.
2.2 Basic capital budgeting problem with accelerated depreciation. Assume the same facts as in
Problem 2.1 except that the earnings before depreciation, interest, and taxes are $22,000 per
year.
a) Calculate the net present value, using straight-line depreciation for tax purposes.
b) Calculate the net present value, using the sum-of-the-years digits method of accelerated
depreciation, for tax purposes.
2.3 Basic replacement problem. The Virginia Company is considering replacing a riveting ma-
chine with a new design that will increase the earnings before depreciation from $20,000 per
year to $51,000 per year. The new machine will cost $100,000 and have an estimated life of
eight years, with no salvage value. The applicable corporate tax rate is 40%, and the firm's
cost of capital is 12%. The old machine has been fully depreciated and has no salvage value.
Should it be replaced by the new machine?
2.4 Replacement problem when old machine has a positive book value. Assume the same facts
as in Problem 2.3 except that the new machine will have a salvage value of $12,000. Assume
further that the old machine has a book value of $40,000, with a remaining life of eight years.
If replaced, the old machine can, at present, be sold for $15,000. Should the machine replace-
ment be made?
2.5 Cash flows. The Cary Company is considering a new investment that costs $10,000. It will
last five years and have no salvage value. The project would save $3000 in salaries and wages
each year and would be financed with a loan with interest costs of 15% per year and amortiza-
tion costs (repayment of principal on the loan) of $2000 per year. If the firm's tax rate is 40%
and its after-tax cost of capital is 20%, what is the net present value of the project? [Note:
The annuity factor for five years at 20% is 2.991.]
2.6 Calculate the internal rate of return for the following set of cash flows.
t1: 400
t2: 400
t3: —1000
t o: —1000
t1: 100
t2: 900
t3: 100
t4: —100
t5: —400
2.8 The Ambergast Corporation is considering a project that has a three-year life and costs
$1200. It would save $360 per year in operating costs and increase revenue by $200 per year.
It would be financed with a three-year loan with the following payment schedule (the annual
rate of interest is 5%).
PROBLEM SET 43
Repayment of
Payment Interest Principal Balance
If the company has a 10% after-tax weighted average cost of capital, has a 40% tax rate, and
uses straight-line depreciation, what is the net present value of the project?
2.9 The treasurer of United Southern Capital Co. has submitted a proposal to the board of
directors that, he argues, will increase profits for the all-equity company by a whopping
55%. It costs $900 and saves $290 in labor costs, providing a 3.1-year payback even though
the equipment has an expected 5-year life (with no salvage value). If the firm has a 50% tax
rate, uses straight-line depreciation, and has a 10% weighted average after-tax cost of capital,
should the project be accepted? Income statements before and after the project are given in
Tables Q2.9A and Q2.9B, respectively.
Table Q2.9A
Table Q2.9B
After Year 1 Year 2 Year 3 Year 4 Year 5
2.10 The cash flow for projects A, B, and C are given below. Calculate the payback period
and net present value for each project (assume a 10% discount rate). If A and B are mutually
exclusive and C is independent, which project, or combination of projects, is preferred using (a)
the payback method or (b) the net present value method? What do the results tell you about
the value-additivity properties of the payback method?
Project
Year A
0 —1 —1 —1
1 0 1 0
2 2 0 0
3 —1 1 3
2.11 Calculate the internal rate of return on the following set of cash flows, according to
Teichroew's economic interpretation of internal rate of return. Assume that the opportunity
cost of capital is 10%.
0 — 5,000
1 10,000
2 — 3,000
REFERENCES
Alchian, A., and H. Demsetz, "Production, Information Costs, and Economic Organization,"
American Economic Review, 1972, 777-795.
Bierman, H., Jr., and S. Smidt, The Capital Budgeting Decision, 4th ed. Macmillan, New York,
1975.
Bodenhorn, D., "A Cash-Flow Concept of Profit," Journal of Finance, March 1964, 16-31.
Coase, R. H., "The Nature of the Firm," Economica, 1937, 386-405.
Cyert, R. M., and J. G. March, A Behavioral Theory of the Firm. Prentice-Hall, Englewood
Cliffs, N.J., 1963.
Gagnon, J.-M., "The Purchase-Pooling Choice: Some Empirical Evidence," Journal of Ac-
counting Research, Spring 1971, 52-72.
Hirshleifer, J., Investment, Interest, and Capital. Prentice-Hall, Englewood Cliffs, N.J., 1970.
Hong, J.; R. S. Kaplan; and G. Mandelker, "Pooling vs. Purchase: The Effects of Accounting
Mergers on Stock Prices," Accounting Review, January 1978, 31-47.
Jensen, M., and W. Meckling, "Theory of the Firm: Managerial Behavior, Agency Costs and
Ownership Structure," Journal of Financial Economics, October 1976, 305-360.
Machlup, F., "Theories of the Firm: Marginalist, Behavior, Managerial," American Economic
Review, March 1967, 1-33.
REFERENCES 45
Marschak, J., and R. Radner, Economic Theory of Teams (Cowles Foundation Monograph
22). Yale University Press, New Haven, 1972.
Stern, J., "Earnings per Share Doesn't Count," Financial Analysts Journal, July—August 1974,
39-43.
Sunder, S., "Relationship between Accounting Changes and Stock Prices: Problems of Mea-
surement and Some Empirical Evidence," Empirical Research in Accounting: Selected
Studies, 1973, 1-45.
, "Stock Price and Risk Related to Accounting Changes in Inventory Valuation," Ac-
counting Review, April 1975, 305-315.
Teichroew, D., An Introduction to Management Science: Deterministic Models. Wiley, New
York, 1964, 78-82.
Williamson, 0. E., The Economics of Discretionary Behavior: Managerial Objectives in a The-
ory of the Firm. Prentice-Hall, Englewood Cliffs, N.J., 1964.
3
The basic problem of time valuation which Nature sets us is always
that of translating the future into the present, that is, the problem of
ascertaining the capital value of future income.
Irving Fisher, The Theory of Interest, Macmillan, New York, 1930, 14
A. INTRODUCTION
Although Chapter 2 introduced the net present value (NPV) criterion, there were
several implied assumptions that require further investigation. For example, all the
illustrations in Chapter 2 assumed that mutually exclusive projects had the same life
and scale. What happens when these assumptions are relaxed? This question is dealt
with in the first part of this chapter.
Next, we will turn the usual capital budgeting problem around and attempt to
determine the optimal life for a project with growing cash flows. For example, when
should growing trees be harvested or aging whisky be bottled? Finally, let us suppose
that the firm is operating under a fixed budget. How will this affect the project selec-
tion process?
The above topics are not usually covered in introductory finance texts. One
reason is that they require more than an introductory level of mathematical sophis-
tication. For example, the optimal harvest problem requires calculus optimization
techniques (see Appendix D), and multiperiod constrained capital budgeting requires
linear programming The reader who is not interested in the mathematics need read
only the introduction and conclusion to sections C.2 and D.2 of this chapter.
Section E discusses the problem of capital budgeting in an inflationary environ-
ment. Needless to say, there has been growing interest over the last decade in this
important applied problem. The reader should be cautioned, however, that solution
techniques assume that future rates of inflation are known with certainty. Therefore
46
CAPITAL BUDGETING TECHNIQUES IN PRACTICE 47
Chapter 2 argued that the NPV and the internal rate of return (IRR) techniques of
capital budgeting were the most sophisticated of the four commonly used criteria.
They both consider cash flows (not earnings per share) and discount them in order
to take into account the time value of money. Yet the question often arises: Do cor-
porations actually employ these techniques?
A survey of large corporations conducted by Klammer [1972] and reported in
the Journal of Business has provided an estimate of the actual usage of different capital
budgeting techniques. His results are duplicated in Table 3.1.
Percentage Using in
Technique 1970 1964 1959
Profit contribution analysis required:
For over 75% of projects 53 53 50
For 25%-757 of projects 41 40 34
For less than 25% of projects 6 7 16
Total 100 100 100
Minimum profitability standards required:
For most projects 77 65 58
For some projects 13 23 20
For few projects 10 12 22
Total 100 100 100
Most sophisticated primary evaluation standard:
Discounting (rate of return or present worth) 57 38 19
Accounting rate of return 26 30 34
Payback or payback reciprocal 12 24 34
Urgency 5 8 13
Total 100 100 100
Approximately 180 firms responded in 1970, 150 in 1964, and 145 in 1959. For
our purposes the most interesting statistic is the most sophisticated primary evalua-
tion standard. Note the increased usage of discounted cash flow techniques such as
the IRR or NPV and the simultaneous decrease in payback. With the advent of com-
puter technology and pocket calculators it is very easy to use the more sophisticated,
and more correct, discounted cash flow techniques. A study by Schall, Sundem, and
Geijsbeek [1978] sampled 424 large firms (with a 46.4% response rate) and found
that 86% use discounted cash flow methods, most of them combined with a payback
or ARR analysis.
One of the interesting implications of capital budgeting is that the net present
value of a project is equal to the expected increase in shareholders' wealth. This means
that the moment a firm publicly reveals that it has undertaken a positive NPV pro-
ject, the market price of the firm's stock should increase by the project's NPV, even
though no cash inflows from the project have yet been received. Empirical confirma-
tion of this idea is provided by McConnell and Muscarella [1985]. They studied the
effect of the announcement of capital expenditure plans for a sample of 658 corpora-
tions over the interval 1975 through 1981. Capital expenditure announcements were
separated into four categories: (1) an announcement of an annual capital budget that
is an increase from the previous year's budget, (2) a decrease from the previous year's
budget, (3) an announcement of an increase in the current year's previously an-
nounced budget and (4) a decrease in the current year's previously announced budget.
Theory predicts that if managers accept positive NPV projects, any announcement
of an increase in planned capital expenditures should result in an increase in the firm's
stock price, whereas any announcement of a decrease will imply fewer positive NPV
opportunities and result in a decline in the stock price. Table 3.2 summarizes the
McConnell and Muscarella results. The two-day announcement period returns (i.e.,
the day the news appeared in the Wall Street Journal and the following day) were
significantly positive for budget increases announced by industrial firms and signifi-
cantly negative for budget decreases. These results confirm that the market reacts
immediately to news about capital expenditure plans, at least for industrial firms.
Industrial firms
All budget increases 273 1.30% .18% 5.60
All budget decreases 76 —1.75 .18 — 5.78
Public utility firms
All budget increases 39 .14 .11 .07
All budget decreases 17 — .84 .22 —1.79
Adapted from McConnell, J. and C. Muscarella, "Corporate Capital Expenditure Decisions and
the Market Value of the Firm," Journal of Financial Economics, September 1985, 399-422.
PROJECTS WITH DIFFERENT LIVES 49
Public utilities, however, are a different story. They are regulated and earn only their
weighted average cost of capital. Hence on average they have few positive NPV pro-
jects. The empirical evidence indicates no statistically significant market reaction to
public utility announcements of capital expenditure changes.
All examples used in Chapter 2 compared projects with the same life. Now we turn
our attention toward methods of choosing among mutually exclusive projects with
different lives. We begin by demonstrating the correct technique. It uses the NPV
rule assuming that projects are replicated indefinitely at constant scale. A good ex-
ample of a replicable project is the harvesting of trees. After the harvest the identical
acreage is replanted, and the project is started anew at the same scale. Next we will
borrow from Hirshleifer [1970] to show why the NPV criterion (when correctly for-
mulated) is superior to the IRR criterion, given that projects have different lives and
are replicable.
Then we have
NPV(N, oo) = NPV(N)(1 + U + U 2 + • • • + Un). (3.2)
Multiplying both sides by U, this becomes
U[NPV(N, co)] = NPV(N)(U + U 2 + • • • + + Un + } ). (3.3)
Subtracting Eq. (3.3) from (3.2) gives
NPV(N, co) — UNPV(N, oo) = NPV(N)(1 — Un + '),
NPV(N)(1 — U" ±1)
NPV(N, Go) = •
1— U
And taking the limit as the number of replications, n, approaches infinity gives
1
lim NPV(N, co) = NPV(N) = NPV(N) [,
co U 1 [11(1 + O]Nd '
[ (1 + ON
NPV(N, oo) = NPV(N) (3.4)
(1 + k)N — 1
Equation (3.4) is the NPV of an N-year project replicated at constant scale an in-
finite number of times. We can use it to compare projects with different lives because
when their cash flow streams are replicated forever, it is as if they had the same
(infinite) life. Furthermore, the NPV of an infinitely replicated project is the same
as the value of shareholders' wealth. It is the present value of the entire stream of
projects.
In our example the value of the two-year project, A, replicated at constant scale
forever is
= $2.36.
And for project B, the three-year project, we have
[ (1 + .10) 3
(1.10) 2
NPV(3, 00) = NPV(3)
= ($.50)[13333
]
= $2.02.
Consequently, we would choose to accept project A over project B because when
the cash flows are adjusted for different project lives, A provides the greater wealth.
Another way of comparing the projects is to multiply the NPVs of the infinitely
PROJECTS WITH DIFFERENT LIVES Si
replicated projects by the opportunity cost of capital to obtain what is called the
annual equivalent value, which is given in Eq. (3.5).1
[ k(1 + k)" 1.
kNPV(N, co) = NPV(N) (3.5)
(1 + k)N — 1]
This decision rule is equivalent to that provided by Eq. (3.4) as long as the projects
being compared have equal risk. If they have different risk, the annual equivalent
should not be used. For example, suppose that we were considering two projects
with the same NPV as project A in the previous example but that they have different
risk. The computation of NPV(N, co) will tell us that the project with higher risk
contributes less to the value of the firm because it will have a lower NPV(N, oo). Yet
when the lower NPV(N, co) of the higher risk project is multiplied by a higher risk-
adjusted opportunity cost, k, it is possible to reach the opposite conclusion. For an
example, work Problem 3.3.
A. USING THE SIMPLE NPV RULE TO SOLVE THE DURATION PROBLEM. Assume
that we own a growing stand of trees. Let the revenue, Rev„ that can be obtained
from harvesting them at time t be represented by the expression
Rev, = 10,000/1 -1- t.
Also, let the initial cost, c, be $15,000 and let the opportunity cost of capital be 5%
1
Note that Eq. (3.5) is equivalent to
NPV(N)
kNPV(N, co) =
Annuity factor
where the annuity factor = [1 — (1 +
52 MORE ADVANCED CAPITAL BUDGETING TOPICS
In $
1 1 .0 —
9.0 ► Time
2 5 6 7 8 10
Figure 3.1
Tree-harvesting problem.
2
Appendix A contains a complete reference to the mathematics of continuous compounding.
PROJECTS WITH DIFFERENT LIVES 53
B. USING THE IRR RULE TO SOLVE THE DURATION PROBLEM. Next, we would like
to compare the simple NPV result with the harvest time that maximizes the IRR
of the project. The IRR is the rate that sets the NPV of the project equal to zero.
Mathematically, this is
(IRR)t c.
NPV = 0 = Rev,e-
Adding c to both sides and taking the natural logarithm, we have
ln Rev, — (IRR)t = ln c,
and solving for IRR, we have
1 Revt
IRR = ln •
t c
Substituting in the revenue function, we obtain
[10,000(1 + t)1/ 2 1
IRR = In
t 15,000
If we try different values of t, the project life that maximizes the IRR at a value of
9.98% is four years.
We can find this result graphically in Fig. 3.1 by rotating a line that passes
through an intercept of ln c (at t = 0, if the natural logarithm of the present value
of revenue is equal to ln c, then the NPV of the project is zero) until it is just tangent
to the revenue curve. The point of tangency gives the optimal harvest time, four
years, and the slope of the line is the maximum IRR, 9.98%.
It is frequently argued that the IRR rule gives the best solution to the simple
duration problem. However, this is incorrect because the IRR rule implicitly assumes
that the funds provided by the project can continuously be reinvested in projects
with proportional expansion of scale. In other words, we started out in our example
with an investment of $15,000. After four years, we would reinvest
$15,000e1R") = $22,359,
and so on, in ever increasing amounts. So long as the IRR is greater than the op-
portunity cost of capital (5% in our example), the present value of an infinite rep-
lication of a proportionately growing stream of projects is infinite This is patently
absurd.
54 MORE ADVANCED CAPITAL BUDGETING TOPICS
C. USING THE NPV RULE WITH CONSTANT SCALE REPLICATION. The correct
formulation of the optimal duration problem is to assume that the project can be
replicated indefinitely at constant scale. In the tree-harvesting problem this is equiv-
alent to assuming that once the trees are harvested, the same acreage is replanted so
that the project begins again at constant scale.
If the project is reformulated with constant scale replication, the NPV of an
infinite stream of projects would be
NPV = -c + (Rev, - c)e -kt + (Rev, - c)e - 2kt . .
The second term is the present value of the revenue received at the time of harvest
less the cash outlay for replanting at constant scale. The third term is the present
value of the cash flows at the time of the second harvest, and so on. The NPV of
this stream is
Rev, - c
NPV = c+ ekt
To maximize, we set the derivative of the NPV with respect to project life, t, equal
to zero.'
dNPV dRev, (Rev, - c)k dRev, (Rev, - c)k
0
1 -e
=
dt dt dt 1 -e
Using the numbers from our example, we get
5,000 (10,000(1 0 112 - 15,000).05
(1 + t)'12 1 _ e .05t
Table 3.4 shows the values at the left- and right-hand sides of the solution for various
values of t. The optimal duration is approximately 4.6 years. This answer lies between
the solution for simple duration with the NPV rule (9 years) and replication with
proportionately increasing scale using the IRR rule (4 years).
3
dNPV d
= [ c + (Rev, - c)(ekt — 1) -1] = 0
dt dt
dRev,
= dt (ekt — 1r + [(Rev, - - 1)(e" - 1)'ke" = 0
dRev,
— k
t 1) (Rev, - c)kem = 0
dt (e
dRev, k(Rev, - c)
dt 1 — e-k[
CONSTRAINED CAPITAL BUDGETING PROBLEMS 55
the data from the tree-harvesting proîlem, recall that all three projects have the same
scale îecause each requires an outlay of $=P,000. However, their lives vary îetween
4 and 9 years. An important question is: How much would you pay to purchase the
forestry operation, assuming that you harvest after t years, then replant, thereîy
replicating the project at constant scale every t years forever, and that the time value
of money to you is P%? The answer is $=8,P0P. It is the current value of the forestry
operation îecause it represents the present value of the cash stream provided îy the
operation into the indefinite future. It is the present value of a strategy of harvesting
every 4.6 years.
The aîove example demonstrates that the correct procedure for comparing
projects with different lives is the same as the correct solution to the optimal duration
proîlem. Both require that NPV maximization îe formulated as the maximization
of the NPV of a stream of projects replicated at constant scale.
A capital îudgeting constraint implies that the firm can oîtain only N dollars of
funding at a fixed cost of capital. Implicitly, the cost of capital in excess of N dollars
is infinite. Therefore the firm is limited to a fixed îudget. Most economists would agree
that strict capital constraints simply do not exist in the real world. For example,
consider a small firm with a "îudget" of only a few thousand dollars of capital that
suddenly acquires a new patent for economically converting garîage into gasoline.
56 MORE ADVANCED CAPITAL BUDGETING TOPICS
Certainly the firm would not find it very difficult to raise large amounts of money
even though its initial budget was quite limited. As long as capital markets are reason-
ably efficient, it will always be possible for a firm to raise an indefinite amount of
money so long as the projects are expected to have a positive net present value.
Weingartner [1977] discusses capital rationing in terms of situations imposed
from within the firm and those imposed by the capital market. Self-imposed expendi-
ture limits may arise to preserve corporate control or reflect the view of owners of
closely held firms that the sale of the firm as a whole at a future date will provide a
greater present value of wealth than the piecemeal sale that may permit faster growth.
Externally imposed capital rationing could result from an attitude of the capital
markets that providing funds beyond a specific amount would lead to increased risks
of high bankruptcy costs—so high that feasible interest rates would not be adequate
compensation. An aspect of this is the "Penrose effect," which holds that the organi-
zational problems of obtaining and training additional personnel are large. Hence
growth that involves increasing the organization's size by more than some percentage,
e.g., 50%, in one year is fraught with high risks of organizational inefficiencies, which
increase risks of bankruptcy and give rise to high costs owing to the loss of efficiency
of a previously effectively functioning organization system or firm.
Although it is hard to justify the assumption of limited capital, nevertheless we
shall review various decision-making techniques, assuming that capital constraints do
in fact exist.
NPV, followed by project 3, then 2, and finally 4. Suppose that there is a capital con-
straint that limits our spending to $300,000 or less. Now we would accept only proj-
ects 2 and 3 because they have the greatest NPV among those combinations of
projects that use no more than $300,000. The logic leading to this decision is formal-
ized by what is known as the present value index (PVI).
The PVI is defined as the present value of cash inflows divided by the present
value of cash outflows:
Present value of inflows
PVI = (3.6)
Present value of outflows
When used correctly, it is equivalent to maximizing the NPV of a set of projects
subject to the constraint that project outlays be less than or equal to the firm's budget.
The PVI for each project is given in Table 3.6. The PVI of excess funds, not invested
in any of the projects, is always assumed to be equal to 1.0. (Why?)
The objective is to compare all sets of projects that meet the budget and find
the one that maximizes the weighted average PVI. For example, if projects 2 and 3
are selected, the weighted average PVI is
125 °00 175'000
PVI = ' (1.13) + (1.11) = 1.1183.
300,000 300,000
It is computed by multiplying the PVI of each project by the percentage of the total
budget allocated to it. If project 1 is selected, no additional project can be undertaken;
therefore the PVI for project 1 is
100,000
,000
PVI = 200 (1.15) + (1.00) = 1.1000.
300,000 300,000
Project 1 is not preferred to projects 2 and 3 because $100,000 must be invested in
marketable securities that have a PVI of 1.0 (i.e., their cost is always equal to the
present value of their cash inflows).
The PVI can be used to solve simple problems where there is a one-period cap-
ital constraint. It can also be used to compare mutually exclusive projects of dif-
ferent scale, although a simple comparison of NPVs provides the same result. For
example, refer to projects A and B (at the beginning of this section). The first project
cost $1,000,000 and had an NPV of $1,000, whereas the second cost only $10 and
had an NPV of $500. If these are the only two projects available to the firm and
if they are mutually exclusive, we can evaluate them by comparing their PVIs with
the assumption that the firm has a $1,000,000 budget.'
The PVI of project A is
1,000,000 (1,001,000 )
= 1.001,
1,000,000 1,000,000
We could also assume any budget whatsoever as long as it is greater than $1,000,000 without changing
the results. (Why?)
58 MORE ADVANCED CAPITAL BUDGETING TOPICS
Period-1 Period-2
Project Outlay Outlay NPV
1 12 3 14
2 54 7 17
3 6 6 17
4 6 2 15
5 30 35 40
6 6 6 12
7 48 4 14
8 36 3 10
9 18 3 12
in period 2 it is $20. The problem is to find the set of projects that maximizes NPV
and satisfies the cash constraints. Weingartner [1963] solved the problem by using
linear programming. If we designate b1 as the NPV of each project, é i as the frac-
tion of each project that is accepted, c o as the cash outlay used by the jth project
in the tth time period, and C ] as the cash budget, the linear programming problem
is written as
MAX E b i é j , (P]imal p]oblem) (3.7)
é i < 1.
If S ] and qi are designated as slack variables, the primal problem can be rewritten as
MAX E (3.8)
subject to E + St = Cr ,
é i + q j = 1.
The objective is to choose the set of weights, é j , that maximizes the combined NPV
of all projects. The constraints require (1) that the set of projects undertaken use less
cash than is budgeted and (2) that no more than 100% of any project be undertaken.
Every linear programming problem has a counterpart called the dual p]oblem
where the primal constraints appear in the dual objective function and the primal
decision variables become dual constraints. The dual for this problem can be written
as
MIN E p t C, + 1,1 j • 1, (Dual p]oblem) (3.9)
subject to E p t c ti + kt i ÿ = bi ,
Pt, Ili O.
The dual introduces three new variables, p t , ki p and y j . The last, y; , is a slack variable.
We can define pi by using the notion of complementary slackness. When one of the
projects enters into the primal solution (i.e., é i > 0), then the corresponding constraint
in the dual is binding and therefore yi = 0. Thus the dual constraint can be written
as an equality:
E ptco + = bi,
or, solving for pi, we have
= b ./ ÿ E (3.10)
Therefore p j may be thought of as the difference between the NPV of a project's cash
flows, bi , and the imputed value of the outlays needed to undertake the project,
p i c ti . Conceptually, it is similar to the PVI in the single-period problem because
60 MORE ADVANCED CAPITAL BUDGETING TOPICS
it is a measure of the benefit (NPV) minus the imputed cash cost of a project. Finally,
pt may be thought of as the implicit one-period discount rate caused by the cash
constraints in the linear programming problem. It is the value of relaxing the cash
constraint by $1, the shadow price. This can be demonstrated by using (3.10) and
noting that for fractionally accepted projects, where the constraint is binding, pi = 0.
Therefore (3.10) becomes
E p r cr; = bi.
This says that the NPV of a project, bi, equals the discounted value of its cash flows.
Consequently, p, is the one-period discount rate:
Pt= (3.11)
+ trt+i
Table 3.8 shows the linear programming problem and solution to the Lorie-Savage
problem. Projects 1, 3, 4, and 9 are accepted, and projects 6 and 7 are fractionally
accepted into the optimal solution, which has an NPV of $70.27. Cash constraints
in both time periods are binding since the primal slacks, S1 and S 2 , are zero. The
project with the greatest net "benefit" is project 4 with p 4 = 10.45. Although it has
a smaller NPV than several of the other projects, it also uses less cash than the others.
Finally, a comparison of the dual variables, p, and p 2 , tells us that the value of
relaxing the second-period constraint is greater. In other words, providing extra cash
in the second period would increase the firm's NPV more than providing extra cash
in the first time period. We can also use the values of p, to calculate the one-period
implicit interest rates:
—
1
ir2 = 635%,
Pi
=
- P<
2
r3 = 46.4%.
P2
Linear programming solutions to capital budgeting have great versatility and
have been applied to many types of problems. However, it is difficult to justify the
existence of capital constraints in the first place. Also, once uncertainty is introduced
as a major consideration, linear programming models fail to handle it adequately.
For these and other reasons, linear programming models have become less popular
in recent years.
The United States has experienced persistent inflation since 1966 at levels exceeding
the moderate price level changes of previous peacetime periods. What effects does
this have on the results of capital budgeting analysis? We can analyze the impacts
of inflation by using an illustrative example to clarify the new influences introduced.
Let us begin with the standard capital budgeting case in which inflation is absent.
The expression for calculating the NPV of the investment is shown in Eq. (3.12):
NCF
NPV = E t
t =1 (1 + k)t
o (3.12)
6
For articles on this subject see Van Home [1971] and Cooley, Roenfeldt, and Chew [1975]; also see
their exchange with Findlay and Frankle [1976].
62 MORE ADVANCED CAPITAL BUDGETING TOPICS
where k is the required rate of return in nominal terms, I) is the anticipated annual
inflation rate over the life of the project, and r is the real rate of return. For our
example, Eq. (3.13) would be
(1 + .09)(1 + .06) = (1 + .09 + .06 + .0054).
When the cross-product term, .0054, is included, we have .1554 as the required
rate of return in nominal terms.
It is at this point that some biases in capital budgeting under inflationary con-
ditions may be introduced. The market data utilized in the estimated current capital
costs will include a premium for anticipated inflation. But while the market remembers
to include an adjustment for inflation in the discount factor, the cash flow estimates
used by the firm in the capital budgeting analysis may fail to include an element to
reflect future inflation. Given that the cost of capital (observed using market rates of
return) already includes expected inflation, the decision maker can correct for inflation
either (a) by adding an estimate of inflation to the cash flows in the numerator or
(b) by expressing the numerator without including an adjustment for inflation and
removing an inflationary factor from the market rate in the denominator.
It is more natural to utilize market data and to explicitly incorporate estimates
of the anticipated inflation rate in the cash flows in the numerator. But either way,
the difficult problem is how to estimate the expected future inflation rate. Some ideas
on how to do this will be discussed in the next section of this chapter.
Sound analysis requires that the anticipated inflation rate be taken into account
in the cash flow estimates. Initially, let us assume that an inflation rate of 6% is
applicable to the net cash flows as well as to the discount rate. We take this step in
CAPITAL BUDGETING PROCEDURES UNDER INFLATION 63
were greater than on the cash inflows. Some have suggested that this influence has
îeen sufficiently widespread and that it accounts for the sluggish rate of capital invest-
ment in the United States during the late =970's and early =980's.
The situation we illustrated initially was that failure to take inflation into account
in the expected cash flows resulted in an erroneous capital îudgeting analysis. A
project was accepted which, when measured correctly, produced a return îelow the
required rate of return. The allocation of capital would îe unsound if the îias in the
analysis due to inflation had not îeen taken into account. In our second and more
▪
complex example, inflation caused the cash outflows to grow at a higher rate than
the cash inflows. As a consequence, the expected NPV of the project was negative.
Making the inflation adjustment does not always necessarily result in a negative NPV
for the project, it simply results in a more accurate estimate of the net benefits from
the project positive or negative.
Throughout Chapters 2 and 3 there has been the implicit assumption that interest
rates are nonstochastic and constant in a multiperiod setting. This has been a conve-
nient but misleading assumption. Interest rates are not constant through time. The
yield on a particular financial instrument (a bond, for example) is a function of the
length of time to maturity. The yield also depends on the risk of the security, but we
shall continue to assume, for the time being, that there is no risk.
Figure 3.2 shows the term structure of interest rates for United States Treasury
securities at three points in time, March 1976, August 1981, and May 1987. Each
point on the graph gives the yield to maturity, o R T , on a bond that is bought today
(at time zero) and that matures T years hence. The bond is assumed to be default
free. Default free means that there is no uncertainty about the nominal payments
promised by the bond. There are, however, other kinds of risk. For example, un-
expected changes in future interest rates will induce risk because the market value
of the bond will change when interest rate expectations do.
20
15 • • August 3, 1981
E
0
May 11, 1987
0 10 • •
• ogeow..•■•=i■
•
5 March 1, 1976
3 4 5 6 7 8 9 10 11 12 13 14
Time to maturity in years
Figure 3.2
The yield to maturity on U.S. Treasury securities.
66 MORE ADVANCED CAPITAL BUDGETING TOPICS
example: A bond promises to pay a 15% coupon at the end of each year for three
years, then pay a face value of $1000. We observe the current market price, Bo , of the
bond to be $977.54. The yield to maturity on the bond may be computed by solving
for OR, in the following present value formula:
T
Coupon, Face
+ value
= (3.15)
B0 t=i (1 + oR TY (I + o)RO T
T
150 + 1000
977.54 = •
t=i (1 + 0RTY (1 + °RO T
Solving iteratively, we find that the yield to maturity, o R T , is 16%.
The term structure shows the yield to maturity for all bonds of all maturities.
In March 1976 and May of 1987 the term structure was upward sloping. Long-term
bonds paid higher yields than short-term bonds. In August 1981 the opposite pattern
existed. The term structure was downward sloping. One thing that both term struc-
ture patterns have in common is that the interest rate is not constant. The yield on
securities clearly depends on when they mature.
To keep things simple, we assume that forward rates are one-period rates and that all prices and yields
are observed at the present time. One could also discuss N-period forward rates observed at any time, t.
For an excellent discussion of the term structure, the reader is referred to Financial Rates and Flows by
James C. Van Home [1978].
Note that like the internal rate of return the yield to maturity implicitly assumes that funds are rein-
vested at the implied yield. This contrasts with the assumption of discounting at the market-determined
rate to get a net present value.
•
Figure 3.3
A simple term structure example.
22
21 •
20
19 •
18
•
17
♦ Yield to
16 •• maturity
15
14
13 Observed
12 ao
forward rate
11
1 2 3
Time to maturity (years)
It can be computed by taking the ratio of the (geometric product of the) yields to
maturity on the three- and two-period bonds:
(1 + 0R 3)
3
2 13
1 + ' — (1 + 0R 2) 2
do? The logical action would be to sell short the two-year bonds for $718.18. If we
are right, interest rates in the second year will be higher than the market expects,
and we will make a capital gain. If enough people believe the observed forward rate
is too low, the decreased demand for two-year bonds will lower their prices until
the implied forward rate rises to 17%. Then interest rates will be in line with revised
expectations.
If there were no transactions costs and if there were no uncertainty, then there
is every reason to believe that the unbiased expectations theory of the term structure
would hold. Implied forward rates would be exact forecasts of expected future rates.'
Future interest rates become more uncertain the further into the future one tries
to predict. Attempts to deal with this fact have led to theories of a liquidity premium
in the term structure of interest rates." Hicks argues that a liquidity premium exists
because a given change in interest rates will have a greater effect on the price of long-
term bonds than on short-term bonds. Hence there is greater risk of loss (and, one
should add, a greater probability of gain) with long-term bonds.'' Consequently, risk-
averse investors will require a higher yield in order to hold longer-term bonds. This
extra yield is called a liquidity premium.
To illustrate the sensitivity of bond prices to changes in the interest rate, consider
two bonds. They both pay $120 per year in coupons and have a $1000 face value,
but one has 5 years to maturity and the other has 10. If current interest rates are
10%, the present value of the 5-year bond, B 5 , is
5
$120 $1000
B5
=t (1 + + (1 + .10)5 = $1075.82,
and the present value of the 10-year bond is
to $120 $1000
B1 = (1 ± .10)' + (1 + .tor = $1122.89.
9
Empirical evidence on the validity of the expectations hypothesis is discussed in Chapter 9.
1
° For example, see Hicks [1946], Keynes [1936], Kessel [1965], Hirshleifer [1972], and Woodward
[1979].
11
A more lengthy discussion of bond risk, including default risk and duration, can be found in the
Appendix to Chapter 14.
THE TERM STRUCTURE OF INTEREST RATES 69
70 Liquidity
premium
Time to maturity
Figure 3.4
The liquidity premium.
What happens to the market prices of the bonds if the interest rate increases from 10%
to 15%, a 50% increase? By similar calculations, we find that
B! = $899.44 and B„ = $849.44.
The market value of the 5-year bond has decreased 16.4%, whereas the 10-year bond
has fallen by 24.4%. Clearly, the 10-year bond is more sensitive to the increase in inter-
est rates. This is the risk Hicks had in mind when he described the liquidity premium.'
Figure 3.4 shows how the liquidity premium is posited to change with the time
to maturity on bonds. Note that the liquidity premium increases for bonds of longer
maturity, but that it increases at a decreasing rate. Figure 3.5 shows how the liquidity
premium is added to unbiased expectations in order to arrive at the observed term
structure.
Using monthly returns data on U.S. Treasury bills (1964-1982) and on portfolios
of U.S. Government bonds (1953-1982), Fama [1984b] investigated term premiums
in bond returns. He found statistically reliable evidence that expected returns on
longer-term bills exceed the returns on one-month bills, but that the premium did not
increase monotonically with maturity (as shown in Fig. 3.4, for example); rather, it
tended to peak at around eight or nine months. The high variability of longer-term
bond returns made it impossible to draw any conclusions about liquidity premia in
their returns.
12 See the Appendix to Chapter 14 for an explanation of "duration," which measures the sensitivity (price
elasticity) of bond prices to changes in interest rates.
70 MORE ADVANCED CAPITAL BUDGETING TOPICS
0 Liquidity
premium
Observed term structure
Unbiased
expectations
Liquidity premium
Figure 3.5
The liquidity premium added to a decreasing term structure (a) and to an increasing
term structure (b).
cash flows from the project. Capital-intensive firms that construct long-term plant
and equipment will prefer to issue long-term debt rather than rolling over short-term
debt, and less capital-intensive firms will prefer to borrow short-term. Insurance com-
panies with long-term liabilities (their life insurance policies) prefer to lend long term.
Thus the market segmentation hypothesis argues that suppliers and users of funds
have preferred habitats. Interest rates for a given maturity are explained mainly by
the supply and demand for funds of that specific maturity.
While the market segmentation hypothesis can explain why implied forward and
expected future rates may differ, the direction and magnitudes are not systematic. Re-
call that the Hicksian liquidity premium causes forward and future rates to differ
systematically, depending on the maturity of the bonds.
discount factor is
(1.16) -3 = [(1.21)(1.1507)(1.1210)] = .6407.
The correct discount factors are given in the last column of Table 3.13. When the cash
flows are appropriately discounted, the NPVs of projects A and B are $5.08 and $5.21,
respectively. Now project B is preferred over A.
When the term structure is downward sloping, as in our simple example, a firm
that uses the long-term rate (the three-year rate) to discount all cash flows will tend
to overestimate the NPVs of projects. Of course, when the term structure is upward
sloping, the opposite bias exists. In addition, as the example has shown, it is possible
for the wrong project to be selected if the information given in the term structure is
ignored.
It has been suggested that the term structure provides the best estimate of expected
inflation." If so, a downward-sloping term structure implies that investors expect
near-term inflation to be higher than long-term. An upward-sloping term structure
(removing the liquidity premium) implies the opposite. If the firm's capital budgeting
procedure discounts nominal cash flows (cum inflation) at market rates, the cash flow
estimates should reflect inflation on a year-by-year basis.
Perhaps the single most important decision faced by management is the selection of
investment projects that maximize the present value of shareholders' wealth. Therefore
it is hardly surprising that much of the literature in finance focuses on the capital
budgeting problem. Both this chapter and its predecessor have emphasized capital
budgeting techniques. However, the story is far from complete. Throughout we have
maintained the assumption that future cash flows are known with certainty and can
be estimated without error. In addition, we assumed that the opportunity cost of
capital (the discount rate) was given.
Chapters 4 through 7 introduce the reader to a world where decisions must be
made under the assumption of uncertainty. It is not until Chapter 12 that we return,
for a second time, to the important capital budgeting decision. However, at that time
13
For empirical evidence consistent with this point of view see Fama [1975].
72 MORE ADVANCED CAPITAL BUDGETING TOPICS
we will be able to discuss the problem of project selection under uncertainty. Fortu-
nately, the inclusion of uncertainty does not materially change the subject matter
presented in Chapters 2 and 3. However, some important extensions to project selec-
tion techniques will be introduced. Finally, the logical cycle is completed in Chapter
13, when we discuss the determination of the appropriate opportunity cost of cap-
ital in a world of uncertainty. At that time all the necessary elements will have been
covered under the assumption of uncertainty. They include the correct definition of
cash flows for capital budgeting purposes, determination of the appropriate cost of
capital, and a proof of why the NPV criterion is consistent with shareholder wealth
maximization.
PROBLEM SET
3.1 The Johnson Company is considering the following mutually exclusive projects:
3.3 The cash flows for two mutually exclusive projects with different lives are given below:
0 —10.00 —10.00
1 6.00 6.55
2 6.00 6.55
3 6.55
PROBLEM SET 73
The opportunity cost of capital for project A is 10%, but project B is much riskier and has a
40% cost of capital.
a) What is the simple NPV of each project?
b) What is the NPV(N, co) of each project?
c) What is the annual equivalent value [see Eq. (3.5)] of each project?
d) Which project should be accepted? Why?
3.4 The Hansen Company is considering four mutually exclusive projects as follows:
a) Compute the NPV and IRR of each project and rank the investments from best to worst
under each method. What factors are responsible for the differences in rankings between
the two approaches?
b) Compute the PVI for each project and rank the alternatives. What are the implicit
assumptions of the PVI method with respect to scale and duration of projects? When is
it appropriate to use the PVI method?
c) If the projects are mutually exclusive, which should be accepted if they are independent?
Why?
3.5 The Dandy Candy Company is considering two mutually exclusive projects. They are the
only projects available. The risk-free rate is 5%. The cash flows from the projects are known
with certainty and are given below:
3.7 You are given the following information: The Dorkin Company has made an investment
of $40,000, which is expected to yield benefits over a five-year period. Annual cash inflows of
$90,000 and annual cash outflows of $75,000 are expected, excluding taxes and the deprecia-
tion tax shelter. The tax rate is 40%, and the cost of capital is 8%. Dorkin Company uses
straight-line depreciation.
a) Compute the NPV of the investment.
b) On investigation, you discover that no adjustments have been made for inflation or
price-level changes. The data for the first year are correct, but after that, inflows are
expected to increase at 4% per year, outflows are expected to increase at 6% per year,
and the annual rate of inflation is expected to be about 6%. Reevaluate the NPV of the
project in light of this information.
3.8 The Baldwin Company is considering investing in a machine that produces bowling balls.
The cost of the machine is $100,000. Production by year during the five-year life of the machine
is expected to be as follows: 5,000 units, 8,000 units, 12,000 units, 10,000 units, and 6,000 units.
The interest in bowling is declining, and hence management believes that the price of
bowling balls will increase at only 2% per year, compared with the general rate of inflation
of 5%. The price of bowling balls in the first year will be $20.
On the other hand, plastic used to produce bowling balls is rapidly becoming more
expensive. Because of this, production cash outflows are expected to grow at 10% per year.
First-year production cost will be $10 per unit.
Depreciation of the machine will be straight-line for five years, after which time the salvage
value will be zero. The company's tax rate is 40% and its cost of capital is 15%, based on the
existing rate of inflation. Should the project be undertaken?
3.9 The yields to maturity on five zero coupon bonds are given below:
1 12.0%
2 14.0
3 15.0
4 15.5
5 15.7
a) What is the implied forward rate of interest for the third year?
b) What rate of interest would you receive if you bought a bond at the beginning of the
second year and sold it at the beginning of the fourth year?
REFERENCES
Baumol, W. S., and R. E. Quandt, "Investment and Discount Rates under Capital Rationing,"
Economic Journal, June 1965, 317-329.
Bernhard, R. H., "Mathematical Programming Models for Capital Budgeting A Survey,
Generalization and Critique," Journal of Financial and Quantitative Analysis, June 1969,
111-158.
REFERENCES 7P
Bierman, H., Jr., and S. Smidt, The Capital Budgeting Decision, 4th ed. Macmillan, New York,
=97P.
Carleton, W., "Linear Programming and Capital Budgeting Models: A New Interpretation,"
Jou]nal of Finance, Decemîer =969, 8<P-833.
Cooley, P. L.; R. L. Roenfeldt; and I. K. Chew, "Capital Budgeting Procedures under Inflation,"
Financial Management, Winter =97P, =8-<7.
Cox, J. C.; J. Ingersoll, Jr.; and S. A. Ross, "A Theory of the Term Structure of Interest Rates."
Unpuîlished working paper, Stanford University, =980.
Culîertson, J. M., "The Term Structure of Interest Rates," Qua]te]ly Jou]nal of Economics,
Novemîer =9P7, 489-P04.
Fama, E. F., "Short-term Interest Rates as Predictors of Inflation," Ame]ican Economic Review,
June =97P, <69-<8<.
, "The Information in the Term Structure," Jou]nal of Financial Economics, Decemîer
=984a, P09-P<8.
, "Term Premiums in Bond Returns," Jou]nal of Financial Economics, Decemîer =984î,
P<9-P46.
, and G. W. Schwert, "Asset Returns and Inflation," Jou]nal of Financial Economics,
Novemîer =977, ==3-=46.
Findlay, M. C., and A. W. Frankle, "Capital Budgeting Procedures under Inflation: Cooley,
Roenfeldt and Chew vs. Findlay and Frankle," Financial Management, Autumn =976,
83-90.
Fisher, I., "Appreciation and Interest," Publications of the Ame]ican Economic Association,
August =896, <3-<9, 9=-9<.
Hicks, J. R., Value and Capital, <nd ed. Oxford University Press, London, =946.
Hirshleifer, J., Investment, Inte]est, and Capital. Prentice-Hall, Englewood Cliffs, N.J., =970.
, "Liquidity, Uncertainty, and the Accumulation of Information," in Carter and Ford,
eds., Essays in Hono] of G. L. S. Shackle. Basil Blackwell, Oxford, =97<.
Kessel, R. A., The Cyclical Behavio] of the Te]m St]uctu]e of Inte]est Rates. National Bureau
of Economic Research, New York, =96P.
Keynes, J. M., The Gene]al Theo]y of Employment, Inte]est and Money. Harcourt, Brace and
World Inc., =936.
Klammer, T., "Empirical Evidence of the Adoption of Sophisticated Capital Budgeting Tech-
niques," Jou]nal of Business, July =97<, 387-397.
Lorie, J. H., and L. J. Savage, "Three Proîlems in Capital Rationing," Jou]nal of Business,
Octoîer =9PP, <<9-<39.
Lutz, F. A., "The Structure of Interest Rates," Qua]te]ly Jou]nal of Economics, Novemîer =940,
36-63.
Malkiel, B. G., "Expectations, Bond Prices, and the Term Structure of Interest Rates," Qua]te]ly
Jou]nal of Economics, May =96<, =97-<=8.
McConnell, J., and C. Muscarella, "Corporate Capital Expenditure Decisions and the Market
Value of the Firm," Jou]nal of Financial Economics, Septemîer =98P, 399-4<<.
McCulloch, J. H., "An Estimate of the Liquidity Premium," Jou]nal of Political Economy,
January—Feîruary =97P, 9P-==9.
Meiselman, D., The Te]m St]uctu]e of Inte]est Rates. Princeton University Press, Princeton,
N.J., =966.
76 MORE ADVANCED CAPITAL BUDGETING TOPICS
Modigliani, F., and R. Sutch, "Innovation and Interest Rate Policy," Ame]ican Economic
Review, May =966, =78-=97.
Myers, S. C., "A Note on Linear Programming and Capital Budgeting," Jou]nal of Finance,
March =97<, 89-9<.
Nelson, C., The Te]m St]uctu]e of Inte]est Rates. Basic Books, New York, =97<.
Roll, R., The Behavio] of Inte]est Rates- An Application of the Efficient Ma]ket Model to U.S.
T]easu]y Bills. Basic Books, New York, =970.
Sargent, T., "Rational Expectations and the Term Structure of Interest Rates," Jou]nal of
Money, C]edit and Banking, Feîruary =97<, 74-97.
Schall, L.; G. Sundem; and W. Geijsîeek, Jr., "Survey and Analysis of Capital Budgeting
References," Jou]nal of Finance, March =978, <8=-<87.
Van Home, J. C., "A Note on Biases on Capital Budgeting Introduced îy Inflation," Jou]nal
of Financial and Quantitative Analysis, January =97=, 6P3-6P8.
, Financial Rates and Flows. Prentice Hall, Englewood Cliffs, N.J., =978.
Walker, C. E., "Federal Reserve Policy and the Structure of Interest Rates on Government
Securities," Qua]te]ly Jou]nal of Economics, Feîruary =9P4, <<-<3.
Weingartner, H. M., Mathematical P]og]amming and the Analysis of Capital Budgeting P]oblems.
Prentice-Hall, Englewood Cliffs, N.J., =963.
, "Capital Rationing: N Authors in Search of a Plot," Jou]nal of Finance, Decemîer
=977, =403-=43<.
Wood, J. H., "Expectations, Error and the Term Structure of Interest Rates," Jou]nal of Political
Economy, April =963, =60-=7=.
Woodward, S., "The Liquidity Premium and the Solidity Premium," Ame]ican Economic Review,
June =983, 348-36=.
4 We wish to find the mathematically complete principles which define
"rational behavior" for the participants in a social economy, and
derive from them the general characteristics of that behavior.
J. Von Neumann and 0. Morgenstern, Theory of Games and Economic
Behavior, Princeton University Press, Princeton, 1947, 31
T he T heory of Choice:
Utility Theory Given
Uncertainty
Economics is the study of how people and societies choose to allocate scarce resources
and distribute wealth among one another and over time. Therefore one must under-
stand the objects of choice and the method of choice. The following two chapters
(Chapters 5 and 6) are devoted to the objects of choice faced by an investor. Here,
we focus on the theory of how people make choices when faced with uncertainty.
Later on—once the theory of choice and the objects of choice are understood—we
shall combine the two in order to produce a theory of optimal decision making under
uncertainty. In particular, we shall study the allocation of resources in an economic
society where prices provide a system of signals for optimal allocation. There are,
however, other means of allocation. Instead of using prices, we might allow an in-
dividual or committee to make all the allocation decisions, or we might program
allocational rules into an algorithm run by machine.
We shall begin with a discussion of the axioms of behavior used by economists.
However, before rushing into them, we must recognize that there are other theories
of behavior. Social sciences such as anthropology, psychology, political science, socio-
biology, and sociology also provide great insight into the theory of choice. And very
early in this chapter we shall be forced to recognize that individuals have different
tastes for the time preference of consumption and different degrees of risk aversion.
77
78 THE THEORY OF CHOICE: UTILITY THEORY GIVEN UNCERTAINTY
Figure 4.1
Indifference curves for various types of choices: (a) Choice between consumption goods
under certainty; (b) choice between consumption and investment under certainty;
(c) choice between risk and return.
Economic theory recognizes these differences but has little to say about why they
exist or what causes them.' The other social sciences study these problems. However,
as we shall see, there is much one can say about the theory of choice under uncertainty
without, for example, understanding why a 70-year-old person is more or less risk
averse than the same person at age 20, or why some people prefer meat, whereas
others prefer vegetables.
The theory of investor choice is only one corner of what has come to be known
as utility theory. Most students are already familiar with the microeconomic price
theory treatment of choices among various bundles of perishable commodities such
as apples and oranges at an instant in time. The indifference curves that result are
shown in Fig. 4.1(a). Another type of choice available to individuals is whether to
consume now or to save (invest) and consume more at a later date. This is the utility
theory of choices over time, which is fundamental for understanding interest rates.
This type of one-period consumption/investment decision was discussed in Chapter 1
and is illustrated in Fig. 4.1(b). Our main concern here is the choice between timeless
risky alternatives, which we call the theory of investor choice. The theory begins with
nothing more than five assumptions about the behavior of individuals when con-
fronted with the task of ranking risky alternatives and the assumption of nonsatiation
(i.e., greed). The theory ends by parameterizing the objects of choice as the mean
and variance of return and by mapping trade-offs between them that provide equal
utility to investors. These mappings are indifference curves for timeless (or one-period)
choices under uncertainty. They are shown in Fig. 4.1(c), and are used extensively
in Chapters 6 and 7.
An interesting exception is an article by Rubin and Paul [1979] that suggests a theory of why people
exhibit different attitudes toward risk at different stages in their lives.
FIVE AXIOMS OF CHOICE UNDER UNCERTAINTY 79
2
The notation and much of the conceptual outline follow the development found in Fama and Miller
[1972].
3
The symbol used to indicate preference ( >- ) is not a mathematical inequality. It can rank only preferences.
For example, an individual may prefer one Picasso to two Rembrandts, or vice versa.
The reason for bounding y on only one side or the other is to eliminate the possibility of x y z, in
which case any a would satisfy the indifference condition required by the gamble.
80 THE THEORY OF CHOICE: UTILITY THEORY GIVEN UNCERTAINTY
u and a second gamble, this time between x (with probability a < ) and z, then if
a, is greater than a < , y is preferred to u.
These are known as the axioms of cardinal utility. They boil down to the following
assumptions about behavior. First, all individuals are assumed to always make com-
pletely rational decisions. A statement that "I like Chevrolets more than Fords and
Fords more than Toyotas but Toyotas more than Chevrolets" is not rational. Second,
people are assumed to be able to make these rational choices among thousands of
alternatives—not a very simple task.
The axiom of strong independence is usually the hardest to accept. To illustrate
it, consider the following example. Let outcome x be winning a left shoe, let y be a
right shoe, and let z also be a right shoe. Imagine two gambles. The first is a 50/50
chance of winning x or z, i.e., a left shoe or a right shoe. The second gamble is a
50/50 chance of winning y or z, i.e., a right shoe or a right shoe. If we were originally
indifferent between the choice of a left shoe (by itself) or a right shoe (by itself), then
strong independence implies that we will also be indifferent between the two gambles
we constructed. Of course, left shoes and right shoes are complementary goods, and
we would naturally prefer to have both if possible. The point of strong independence
is that outcome z in the above examples is always mutually exclusive. In the first
gamble, the payoffs are a left shoe or a right shoe but never both. And in the second
gamble, the payoffs are a right shoe or a right shoe but never both. The mutual
exclusiveness of the third alternative z is critical to the axiom of strong independence.
Having established the five axioms, we add to them the assumption that indi-
viduals always prefer more wealth to less. In other words, people are greedy. The
marginal utility of wealth is always positive. This assumption, in conjunction with
the other five, is all that is needed to provide a complete development of utility theory.
Next, we need to answer the question, How do individuals rank various com-
binations of risky alternatives? We can use the axioms of preference to show how
preferences can be mapped into measurable utility. How do we establish a utility
function that allows the assignment of a unit of measure (a number) to various alter-
natives so that we can look at the number and know that if, e.g., the utility of x is
35 and the utility of y is 27, then x is preferred to y? To do this we need to discuss
two properties of utility functions.
The utility function will have two properties. First, it will be order preserving. In
other words, if we measure the utility of x as greater than the utility of y, U(x) > U(y),
it means that x is actually preferred to y, x >- y. Second, expected utility can be used
DEVELOPING UTILITY FUNCTIONS 81
Then we can use Axiom 5 (ranking) so that the probabilities a(x and a(y) can be
interpreted as numerical utilities that uniquely rank x and y. By Axiom 5,
If a(x) > a(y , then x y.
a Figure 4.2
ca(v) Elementary gambles.
82 THE THEORY OF CHOICE: UTILITY THEORY GIVEN UNCERTAINTY
Figure 4.3
Outcome z compared with a gamble
between outcomes x and y.
/é (z)
z—
1
— 0(z)
Now, we have already established, by Axioms 4 and 5, that the utilities of x and y
can be represented by their probabilities, i.e., U(x = a(x) and U(y = a(y . Therefore
the above gamble can be rewritten as
z G[a, b : mé(z U(x + (} — [}(z U(y ].
Finally, by using Axioms 4 and 5 a second time, it must be true that the unique prob-
ability of outcome z can be used as a cardinal measure of its utility relative to the
elemental prospects of a and b. Therefore we have
U(z = mé(z U(x + (1 — mé(z U(y . (4.1)
In this way we have shown that the correct ranking function for risky alternatives is
expe3ted utility. Equation (4.1) says that the utility of z is equal to the probability of
x times its utility plus the probability of y times its utility. This is an expected utility
that represents a linear combination of the utilities of outcomes.
In general, we can write the expected utility of wealth as follows:
E[U(W ] p i U(Wi .
Given the five axioms of rational investor behavior and the additional assumption
that all investors always prefer more wealth to less, we can say that investors will
always seek to maximize their expected utility of wealth. In fact, the above equation
is exactly what we mean by the theory of choice. All investors will use it as their ob-
jective function. In other words, they will seem to calculate the expected utility of
wealth for all possible alternative choices and then choose the outcome that maxi-
mizes their expected utility of wealth.
a Figure 4.4
1---(5(z)(a(x))4-0-130)),:x(y) Outcome z related to elementary
Z --
prospects a and b.
—a(x))+0 —)(3(z))(1—ct,0,)) b
DEVELOPING UTILITY FUNCTIONS 83
Now, we can use the properties of utility functions to demonstrate how our utility
function might îe constructed. Suppose we arîitrarily assign a utility of -=0 utiles
to a loss of $=000 and ask the following question: When we are faced with a gamîle
with proîaîility a of winning $=000 and proîaîility (= - a) of losing $=000, what
proîaîility would make us indifferent îetween the gamîle and $0.0 with certainty?
Mathematically, this proîlem can îe expressed as
0 - G(=000, -=000 : a)
or
U(0) = aU(=000) + (= - oc)U( - =000).
Suppose that the proîaîility of winning $=000 must îe .6 in order for us to îe
indifferent îetween the gamîle and a sure $0.0. By assuming that the utility of $0.0
with certainty is zero and suîstituting U( - =000) = -=0 and at = .6 into the aîove
equation, we can solve for the utility of $=000:
(= - cx)U( - =000)
U(=000) =
(= - .6)( - =0)
- 6.7 utiles.
6
By repeating this procedure for different payoffs it is possiîle to develop a utility
function. Taîle 4.= shows various gamîles, their proîaîilities, and the utility of pay-
offs for a risk-averse investor. The cardinal utility function that oîtains for the set
of preferences indicated in Taîle 4.= is given in Fig. 4.P.5
Utility Index
30
10
(5000) (3000) (1000) Monetary
gains and
1 000 3000 5000 7000 losses
(10)
(30)
(50)
(70)
(90)
(110)
(130)
Figure 4.5
Cardinal utility function. (From Dividend Policy and
Enterprise Valuation, by James E. Walter. Copyright ©
1967 by Wadsworth Publishing Company, Inc., Belmont,
Calif. Reprinted by permission of the publisher.)
An important thing to keep in mind is that utility functions are specific to indi-
viduals. There is no way to compare one individual's utility function to another's.
For example, we could perform an experiment by giving two people $1000. We would
see that they are both happy, having just experienced an increase in utility. But whose
utility increased more? It is impossible to say! Interpersonal comparison of utility
functions is impossible. If it were not, we could establish a social welfare function
that would combine everyone's utility, and we could then use it to solve such prob-
lems as the optimal distribution of wealth. We could maximize society's utility by
taking wealth from individual i and giving it to individual j. However, it is not pos-
sible to know how real-world utility functions for different individuals should be ag-
gregated. It follows that group utility functions, such as the utility function of a firm,
have no meaning.
Another important property of cardinal utility functions is that we can sensibly
talk about increasing or decreasing marginal utility. This can best be illustrated with
an example taken from the centigrade and Fahrenheit temperature scales. Consider
two outcomes: the freezing point of water and its boiling point. Call them x and y,
respectively. Each scale may be likened to a function that maps various degrees of
heat into numbers. Utility functions do the same thing for risky alternatives. The
difference between two outcomes is marginal utility. On the centigrade scale the dif-
ESTABLISHING A DEFINITION OF RISK AVERSION 85
ference between freezing and boiling is 100°C. On the Fahrenheit scale the difference
is 180°F. The ratio of the "changes" is
212° — 32°
100°-0° = 1.8.
If the two scales really do provide the same ranking for all prospects, then the ratio
of changes should be the same for all prospects. Mathematically,
U(x) — U(y)
= constant,
tP(x) — 0(Y)
where U(•) and ■N•˝ are the two utility Eunctions. Compare any two points on the
two temperature scales and you will see that the ratio oE changes between them is a
constant- i.e.- 1.8. Hence changes in utility between any two wealth levels have exactly
the same meaning on the two utility Eunctions- i.e.- one utility Eunction is just a
"transEormation" oE the other.
C. ESTABLISHING A DEFINITION OF
RISK AVERSION
Having established a way of converting the axioms of preference into a utility func-
tion, we can make use of the concept to establish definitions of risk premia and also
precisely what is meant by risk aversion. A useful way to begin is to compare three
simple utility functions (Fig. 4.6) which assume that more wealth is preferred to less—
in other words, the marginal utility of wealth is positive MU(W) > 0. Suppose that
we establish a gamble between two prospects, a and b. Let the probability of receiving
prospect a be a and the probability of b be (1 — a). The gamble can be written as
before: G(a, b: a). Now the question is this: Will we prefer the actuarial value of the
gamble (i.e., its expected or average outcome) with certainty or the gamble itself?
In other words, would we like to receive $10 for sure, or would we prefer to "roll
the dice" in a gamble that pays off $100 with a 10% probability and $0 with a 90%
U(W) U(W)
U(b)
U(b)
U(a)
U(a)
a b a
(b) (c)
Figure 4.6
Three utility functions with positive marginal utility: (a) risk lover; (b) risk neutral; (c) risk
averter.
86 THE THEORY OF CHOICE: UTILITY THEORY GIVEN UNCERTAINTY
U(W)=1nW
3.4=U($30) W U(W)
1 0
3.0
5 1.61
10 2.30
2.30=U[E(W)] — 20 3.00
1.97=E[U(W)] 30 3.40
1.61=U($5)
I
1.0 I I
I I
I I
5 7.17 10 20 30
a E(W) b
Figure 4.7
Logarithmic utility function.
probability? A person who prefers the gamble is a risk lover; one who is indifferent
is risk neutral; and one who prefers the actuarial value with certainty is a risk averter.
In Fig. 4.7, we have graphed a logarithmic utility function: U(W) = ln(W). The gam-
ble is an 80% chance of a $5 outcome and a 20% chance of a $30 outcome. The
actuarial value of the gamble is its expected outcome. In other words, the expected
wealth is
E(W) = .8($5) + .2($30) = $10.
The utility of the expected wealth can be read directly from the utility function:
U[E(W)] = 2.3. That is, if an individual with a logarithmic utility function could
receive $10 with certainty, it would provide him or her with 2.3 utiles. The other
possibility is the utility of the gamble. We know from Eq. (4.1) that it is equal to the
expected utility of wealth provided by the gamble.
E[U(W)] = .8U($5) + .2U($30)
Because we receive more utility from the actuarial value of the gamble obtained with
certainty than from taking the gamble itself, we are risk averse. In general, if the
utility of expected wealth is greater than the expected utility of wealth, the individual
will be risk averse. The three definitions are:6
If U[E(W)] > E[U(W)], then we have risk aversion. (4.2a)
If U[E(W)] = E[U(W)], then we have risk neutrality. (4.2b)
If U[E(W)] < E[U(W)], then we have risk loving. (4.2c)
6
These definitions can be found in Markowitz [1959].
ESTABLISHING A DEFINITION OF RISK AVERSION 87
Note (see Fig. 4.6) that if our utility function is strictly concave, we will be risk averse;
if it is linear, we will be risk neutral; and if it is convex, we will be risk lovers.
It is even possible to compute the maximum amount of wealth an individual
would be willing to give up in order to avoid the gamble. This is called a risk premium.
Suppose that Mr. Smith is faced with the gamble illustrated in Fig. 4.7, has a current
level of wealth of $10, and has a logarithmic utility function. How much will he pay
to avoid the gamble? If he does nothing, he has an 80% chance of ending up with
$5 (a decline of $5) and a 20% chance of ending up with $30 (an increase of $20).
The expected utility of the gamble has already been determined to be 1.97 utiles.
From the logarithmic utility function in Fig. 4.7 we see that the level of wealth that
provides 1.97 utiles is $7.17. On the other hand, Smith receives an expected level of
wealth of $10 (equal to his current wealth) if he accepts the gamble. Therefore, given
a logarithmic utility function, he will be willing to pay up to $2.83 in order to avoid
the gamble. We shall call this the Markowitz risk premium. If Smith is offered insur-
ance against the gamble that costs less than $2.83, he will buy it.
We shall adopt the convention of measuring the risk premium as the difference
between an individual's expected wealth, given the gamble, and the level of wealth
that individual would accept with certainty if the gamble were removed, i.e., his or
her certainty equivalent wealth. There is another convention that might be called the
cost of the gamble. It is defined as the difference between an individual's current
wealth and his or her certainty equivalent wealth. Note that in the first example,
given above, expected wealth and current wealth were identical because the expected
change in wealth was zero. Thus there was no difference between the risk premium
and the cost of the gamble. To illustrate the difference between the two definitions,
consider the following example. A risk-averse individual has the same logarithmic
utility function as in Fig. 4.7 and the same current wealth, i.e., $10, but the gamble
is a 10% chance of winning $10 and a 90% chance of winning $100. We can compute
the following numbers:
Current wealth = $10,
Expected wealth = $101,
Certainty equivalent wealth = $92.76.
Our convention will be to define the risk premium as the difference between expected
wealth and certainty equivalent wealth, i.e.:
Risk premium = Expected wealth — certainty equivalent wealth
= $101 — $92.76 = $8.24.
This measures, in dollars, the risk premium associated with the gamble. Note, how-
ever, that since the gamble is favorable (we can only win if we take it), we would be
willing to pay a positive amount to take the gamble. The cost of the gamble is
Cost of the gamble = Current wealth — certainty equivalent
= $10 — $92.76 = $ — 82.76.
88 THE THEORY OF CHOICE: UTILITY THEORY GIVEN UNCERTAINTY
Decreasing marginal utility is probably genetically coded because without it we would exhibit extreme
impulsive behavior. We would engage in the activity with the highest marginal utility to the exclusion of
all other choices. Addictive behavior would be the norm.
Students not familiar with Taylor's series approximations are referred to Appendix D.
9
We assume that the third absolute central moment of 2 is of smaller order than QZ (normally it is of
the order of 4).
ESTABLISHING A DEFINITION OF RISK AVERSION 89
U"(W
7E =- 1 (4.6b)
U](W •
This is the Pratt-Arrow measure of a local risk premium. Since is always positive,
the sign of the risk premium is always determined by the sign of the term in paren-
theses. We shall define the measure of absolute risk aversion (ARA) as
U"(W
ARA = • (4.7)
U](W
It is called absolute risk aversion because it measures risk aversion for a given level
of wealth. The Pratt-Arrow definition of risk aversion is useful because it provides
much more insight into people's behavior in the face of risk. For example, how does
ARA change with one's wealth level? Casual empiricism tells us that ARA will prob-
ably decrease as our wealth increases. A $1000 gamble may seem trivial to a billion-
aire, but a pauper would probably be very risk averse toward it. On the other hand,
we can multiply the measure of absolute risk aversion by the level of wealth to obtain
what is known as relative risk aversion (RRA):
U"(W
RRA = W • (4.8)
U](W
Constant relative risk aversion implies that an individual will have constant risk
aversion to a proportional loss of wealth even though the absolute loss increases as
wealth does.
We can use these definitions of risk aversion to provide a more detailed examina-
tion of various types of utility functions to see whether or not they have decreasing
ARA and constant RRA. The quadratic utility function has been used widely in
academic literature. It can be written (for W < al2b
The Pratt-Arrow definition of risk aversion provides useful insights into the properties
of ARA and RRA, but it assumes that risks are small and actuarially neutral. The
Markowitz concept, which simply compares E[U(W)] with U[E(W)], is not limited
by these assumptions.
An interesting comparison of the two measures of risk premiums is offered in
the following example. An individual with a logarithmic utility function and a level
of wealth of $20,000 is exposed to two different risks: (1) a 50/50 chance of gaining
or losing $10 and (2) an 80% chance of losing $1,000 and a 20% chance of losing
COMPARISON OF RISK AVERSION IN THE SMALL AND IN THE LARGE 91
$10,000. What is the risk premium required by the individual faced with each of these
risks? Note that the first risk is small and actuarially neutral, so that it approximates
the assumptions that were used to derive the Pratt-Arrow risk premium. The second
risk, however, is large and very asymmetric.
The first risk is a small, actuarially neutral gamble, so the Pratt-Arrow measure
of the risk premium [Eq. (4.6b)] should yield a result almost identical to the Markowitz
measure. The Pratt-Arrow measure is
U"(W
= -101 U]( W
The variance of the first risk is
= E p i( 2 — E(2 2
The certainty equivalent wealth level that would make us indifferent to our current
level of wealth, given the gamble and a lower but certain level of wealth, is the level
of wealth that has a utility of 9.903487428. This is
W = e l n (w) = $19,999.9974998.
Therefore we would pay a risk premium as large as $.0025002. The difference between
the Pratt-Arrow risk premium and that of Markowitz is negligible in this case.
If we repeat similar computations for the second risk in the above example, the
Pratt-Arrow assumptions of a small, actuarially neutral risk are not closely approxi-
mated. Nevertheless, if we apply the Pratt-Arrow definition, the risk premium is cal-
culated to be $324. The Markowitz risk premium for the same risk is the difference
92 THE THEORY OF CHOICE: UTILITY THEORY GIVEN UNCERTAINTY
between expected wealth, $17,200, and the certainty equivalent wealth, $16,711, or
$489. Now the dollar difference between the two risk premia is much larger."
The above example illustrates the difference between risk aversion for small,
actuarially neutral risks, where the Pratt-Arrow assumptions are closely approxi-
mated, and risk aversion in the large, where the magnitude of the gamble is large or
where it is not actuarially neutral. In general, the Markowitz measure of a risk pre-
mium is superior for large or asymmetric risks. This does not mean that the Pratt-
Arrow definition of risk aversion is not useful. As we have seen, the intuition provided
by the definition of risk aversion was useful for distinguishing between various types
of concave utility functions.
E. STOCHASTIC DOMINANCE
So far we have discussed the axioms of investor preference, then used them to develop
cardinal utility functions, and finally employed the utility functions to measure risk
premia and derive measures of risk aversion. Clearly, any investor, whether risk averse
or not, will seek to maximize the expected utility of his or her wealth. The expected
utility rule can be used to introduce the economics of choice under uncertainty. An
asset (or portfolio) is said to be stochastically dominant over another if an individual
receives greater wealth from it in every (ordered) state of nature. This definition is
known as first-order stochastic dominance. Mathematically, asset x, with cumulative
probability distribution F x( W), will be stochastically dominant over asset y, with
cumulative probability distribution Gy( W), for the set of all nondecreasing utility
functions if
" Had we calculated the cost of the gamble instead of the risk premium, we would have subtracted the
certainty equivalent wealth, $16,711, from the individual's current wealth, $20,000, to find that the indi-
vidual would have paid up to $3,289 to avoid the gamble.
STOCHASTIC DOMINANCE 93
Figure 4.8
An example of first-order stochastic
dominance.
F(W)
of expected utility:
where
U(W) = the utility function,
W = the level of wealth,
f(W)= the frequency distribution of wealth.
The utility functions in Fig. 4.9 are linear, but they could just as easily be any of the
set of increasing functions that we are comparing with any set of nonincreasing func-
tions. Expected utility is the sum of the utilities of all possible levels of wealth weighted
by their probability. For a given frequency of wealth, fi(W), in the top half of Fig. 4.9,
the increasing utility function assigns higher utility to the level of wealth offered by
asset x than by asset y. This is true for every frequency. Consequently, the expected
utility of wealth from asset x is greater than that from asset y for the set of increasing
utility functions (i.e., all utility functions that have a positive marginal utility of
wealth). Of course, the opposite would be true for utility functions nonincreasing in
wealth.
Second-order stochastic dominance not only assumes utility functions where mar-
ginal utility of wealth is positive; it also assumes that total utility must increase at
a decreasing rate. In other words, utility functions are nondecreasing and strictly
concave. Thus individuals are assumed to be risk averse. Asset x will be stochastically
dominant over asset y for all risk-averse investors if
f(W)
f(W)
U(W)
increasing
Figure 4.9
First-order stochastic dominance and expected
utility.
greater than the accumulated area for x, îelow any given level of wealth. This implies
that, unlike first-order stochastic dominance, the cumulative density functions can
cross. Figure 4.=0 provides a graphic example, this time assuming normal distriîu-
tions. Oîviously, asset x will dominate asset y if an investor is risk averse îecause
they îoth offer the same expected level of wealth (y„ = t ) and îecause y is riskier. It
f(W) f(W)
f (W)
W)
g' (111)
Awl 4W
I
I
F(W) U(W) Linear
I
AU < = = = = A, - $ Concave
ilx = 13-1 ,
(a) (î)
Figure 4.=0
An example of second-order stochastic dominance.
STOCHASTIC DOMINANCE 95
f[Gy(W)—F,(W)1dW
µx = Py
Figure 4.11
Graphical representation of the sum of the differences
in cumulative probabilities.
has greater variance. The second-order stochastic dominance criterion requires that
the difference in areas under the cumulative density functions be positive below any
level of wealth, W. Up to the mean, G),( W) is strictly greater than Fx( W). Beyond
the mean, the opposite is true. Figure 4.11 shows that the sum of the differences
between the two cumulative density functions is always greater than or equal to zero;
therefore x dominates y.
Figure 4.10(b) ties the concept of second-order stochastic dominance back to the
notion of maximizing expected utility." The concave utility function of a risk averter
has the property that the increase in utility for constant changes in wealth declines
as a function of wealth. Therefore if we select a given frequency of wealth such as
f}( W), it maps out equal changes in wealth A W, and 4W2 . The difference in utility
between x and y below the mean is much greater than the difference in utility for
the same change in wealth above the mean. Consequently, if we take the expected
utility by pairing all such differences with equal probability, the expected utility of x
is seen to be greater than the expected utility of y. If the individual were risk neutral,
with a linear utility function, the differences in utility above and below the mean
would always be equal. Hence a risk-neutral investor would be indifferent relative
to x and y.
Stochastic dominance is an extremely important and powerful result. It is prop-
erly founded on the basis of expected utility maximization, and even more important,
it applies to any probability distribution whatsoever. This is because it takes into
account every point in the probability distribution. Furthermore, we can be sure that
if an asset demonstrates second-order stochastic dominance, it will be preferred by
all risk-averse investors, regardless of the specific shape of their utility functions. We
could use stochastic dominance as the basis of a complete theory of how risk-averse
investors choose among various risky alternatives. All we need to do is find the set
of portfolios that is stochastically dominant and then select a portfolio from among
those in the set.'
11
The graphical presentation given here is intuitive and not meant to be a proof of the fact that second-order
stochastic dominance maximizes expected utility for risk-averse investors. For proof, the reader is referred
to Hanoch and Levy [1969].
12
There is a growing body of literature that uses this concept. The interested reader is referred to Bawa
[1975], Whitmore [1970], Porter, Wart, and Ferguson [1973], Levy and Kroll [1976], Vickson and Altman
[1977], Jean [1975], and Kira and Ziemba [1977].
96 THE THEORY OF CHOICE: UTILITY THEORY GIVEN UNCERTAINTY
If the distribution of returns offered by assets is jointly normal, then we can maximize
expected utility simply by selecting the best combinations of mean and variance."
This is computationally much simpler than stochastic dominance but requires that we
restrict ourselves to normal distributions. Every normal distribution can be com-
pletely described by two parameters: its mean and variance—return and risk. If
we adopt utility functions that maximize expected utility of end-of-period wealth
(assuming a single-period model), it is easy to show the relationship between wealth
and return:
—W
R.= °•
Wo
If the end-of-period wealth from investing in asset j is normally distributed with mean
W and variance then the return on asset j will also be normally distributed with
mean E(Ri = [(E(W mWo — 1] and variance al, = (6i r/ W(2)).
Assuming that the return on an asset is normally distributed with mean E and
variance a 2, we can write our utility function as"
U = U(RJ ; E, a).
Our expected utility is
13
By jointly normal we mean that all assets are individually normally distributed, and in addition, their
interrelationships (covariances) obey the normal probability laws. This concept is developed further in
Chapter 6.
14
This proof can be found in Tobin [1958]. Also note that the proof applies equally well to any continuous,
symmetric two-parameter distribution.
USING MEAN AND VARIANCE AS CHOICE CRITERIA 97
Figure 4.12
Indifference curves for a risk-averse
investor.
aR
and when R = -oo, then Z = — Do, and when R = 3o, then Z = a . Now, by using
15
the change of variables technique from integral calculus, we can rewrite (4.14) as
Next, we take the derivative of the expected utility with respect to a change in the
standard deviation of return:1 6
dE( U)
(dE -
= U](E + aZ + Z f(Z; 0, 1) dZ 0. (4.17)
du - do-
An indifference curve is defined as the locus of points where the change in the ex-
pected utility is equal to zero. Therefore (4.17) has been set equal to zero, and the
solution of the equation represents an indifference curve. Separating terms, we have
dE .f.°°
0= U](E + 62 f(Z; 0, 1) dZ + U](E + 62 Zf(Z; 0, 1) dZ.
do- f
dE U](E + 62 Zf(Z; 0, 1) dZ
> 0. (4.18)
do- U](E + 62 f(Z; 0, 1) dZ
The denominator must be positive because of the assumption that marginal utility,
U](E + o-Z , must always be positive. People always prefer more return to less. The
15
Since f(R; E, a = (}ma f(Z; 0, 1), it follows that
}6
af(Z m06 = 0.
98 THE THEORY OF CHOICE: UTILITY THEORY GIVEN UNCERTAINTY
numerator will be positive only if we have a risk-averse investor with a strictly con-
cave utility function. The marginal utility of every negative value of Z in Fig. 4.13
is greater than the marginal utility of an equally likely positive value of Z. Because
this is true for every pair of outcomes ± Z, the integral in the numerator of (4.18)
is negative, and the (entire) numerator is positive. Consequently, the slope of a risk
averter's indifference curve in Fig. 4.12, i.e., his or her marginal rate of substitution
between mean and variance, is everywhere positive, excepting when o- = 0 where the
slope is also zero."
The indifference curves in Fig. 4.12 will be used throughout the remainder of
the text to represent the theory of choice for risk-averse investors. Any points along
a given indifference curve provide us with equal total utility. For example, we would
not care whether we were at point A in Fig. 4.12, which has no risk, at point B with
higher risk and return, or at point C. They all lie on the same indifference curve.
Moving from right to left across the family of indifference curves provides us with
increasing levels of expected utility. We would prefer point D on indifference curve
3 to point C on indifference curve 2, even though D has a lower return. The reason,
of course, is that it has a much lower risk, which more than makes up for the lower
return. The easiest way to see that expected utility increases from right to left is to
fix the level of risk at and then note that the expected return increases as we
move from curve 1 to curve 4. Although the indifference curves in Fig. 4.12 appear
17
The convexity of the utility function can be shown as follows. Let (E 1 , i) and (F2 , o-2) be two points
on the same indifference curve so that they have the same expected utility. If a third point is constructed
to be a weighted average of the first two, (E1 + E 2)/2, (a, + 02)/2, the indifference curve is convex, if
for every Z,
E2 + al + a2
U (E 1 +
}U(E i + 0 1 Z) + + o- 2Z) < z).
2 2
In the case of declining marginal utilities, this is obviously true because the utility of the second point,
(E2 , 02), will be less than twice the utility of the first. Consequently,
E[U(E1 E2
, 61 72
)1> E[U(E 1, o- 1)] = E[U(E 2, 02)],
2 2
and the third point, which is a weighted average of the first two, lies above the indifference curve. This
is shown graphically in Fig. 4A.
E(R) Figure 4A
Convexity of the risk averter's
(E2, 0 2)
indifference curve.
(Ei+E2 cri + (7 2 )
Ei+E2 2 2
2
a
0 R
0
'= 1+ 0
2 0-
2
2
A MEAN-VARIANCE PARADOX 99
Slope equals
U(Z) MU of (—Z) Slope equals
7-
MU of (+Z)
v.
v
f(Z)
Figure 4.13
Graphic representation for
U'(E o-Z)Zf(Z; 0, 1)dZ < 0.
to be parallel, they need not be. The only requirement is that they never touch or
cross.
G. A MEAN-VARIANCE PARADOX
Although it is convenient to characterize return and risk by the mean and variance
of distributions of return offered by assets, it is not always correct. In fact, it is correct
only when the returns have a normal distribution. Consider the following example.
Two companies with equal total assets and exactly the same distribution of net
operating income differ only with regard to their financial leverage. Table 4.2 shows
their respective income statements in different, equally likely, states of nature.
The mean and standard deviation of earnings per share for firm A are $5 and
$1.41, respectively. For firm B, they are $7 and $2.82. These alternatives are plotted
in Fig. 4.14. According to the mean-variance criterion, individual I would be indif-
ferent between the risk-return combinations offered by A and B. Individual II, who
is less risk averse, would prefer alternative B, which has a greater return. Finally,
individual III would prefer alternative A, which has lower risk. The paradox arises
when we reexamine the earnings per share offered by the two firms. The earnings
per share for firm B are equal to or greater than the earnings per share for firm A
in every state of nature. Obviously, the mean-variance criterion provides misleading
results. No investor with positive marginal utility would prefer firm A.
=00 THE THEORY OF CHOICE: UTILITY THEORY GIVEN UNCERTAINTY
Fi]m A Fi]m B
Assets Liaîilities Assets Liaîilities
Deît 0 Deît =0,000
Equity <0,000 Equity =0,000
$<0,000 $<0,000 $<0,000 $<0,000
The trouîle with trying to apply the mean-variance criterion to the aîove proî-
lem is that the distriîution of outcomes is not normal. Instead, it is a rectangular
distriîution with equal proîaîilities for each state of nature. However, we can use
second-order stochastic dominance regardless of the shape of the proîaîility distri-
E(eps)
IIIî IIIa
I// IIî
•
0000•Illa
,
..,■1 / ■••
•
/A
/ /
a(eps)
= <
Figure 4.=4
A mean-variance paradox.
A MEAN-VARIANCE PARADOX =0=
îution." This is done in Taîle 4.3. Because the accumulated area under the dis-
triîution of earnings per share offered îy firm B is always less than or equal to the
accumulated distriîution for firm A, we can say that B clearly dominates A. The
density functions and cumulative density functions are shown in Fig. 4.=P.
This mean-variance paradox example demonstrates very clearly the shortcomings
of a theory of choice that relies on the (somewhat heroic) assumption that returns
eps
=0
Cumulative proîaîility
=.0
eps
l0
Figure 4.=P
Stochastic dominance applied to the mean-variance
paradox.
First-order stochastic dominance also oîtains in this example. We have used second-order dominance
îecause we assume a risk-averse decision maker.
102 THE THEORY OF CHOICE: UTILITY THEORY GIVEN UNCERTAINTY
are normally distributed. Nevertheless, much of the remainder of this text will as-
sume that returns are in fact normally distributed.
Utility theory is founded on the axioms of Von Neumann and Morgenstern [1947]
and the elegant mathematics which follows logically from them. Furthermore, the
basic results increasing marginal utility, risk aversion, and decreasing absolute risk
aversion—seem to conform to economists' casual empiricism. There has been almost
no empirical testing of the axioms or of their implications, at least not by economists.
Psychologists, however, have been busy testing the validity of the axioms. Do in-
dividuals actually behave as described by the axioms? The answer seems to be a
resounding no they do not.
Kahneman and Tversky [1979, 1986] point out that the way decisions are framed
seems to matter for individual decision making. They give the following example
where people are asked to decide between surgery and radiation therapy for cancer
treatment.
Survival Frame
Surgery: Of 100 people having surgery, 90 live through the postoperative period,
68 are alive at the end of the first year, and 34 are alive at the end of five years.
Radiation Therapy: Of 100 people having radiation therapy, all live through the
treatment, 77 are alive at the end of one year, and 22 are alive at the end of five
years.
Mortality Frame
Surgery: Of 100 people having surgery, 10 die during surgery or the postopera-
tive period, 32 die by the end of the first year, and 66 die by the end of five
years.
Radiation Therapy: Of 100 people having radiation therapy, none die during
treatment, 23 die by the end of one year, and 78 die by the end of five years.
The information in both frames is exactly the same, yet when presented with the
survival frame, 18 percent preferred radiation, and when presented with the mortal-
ity frame, 44 percent preferred radiation a significant difference. The framing effect
was not smaller for experienced physicians or for statistically sophisticated business
students.
If individual decision making is not adequately described by the Von Neumann
and Morgenstern axioms, then it becomes necessary to rethink the descriptive validity
of expected utility theory. No widely accepted answer to this problem has appeared,
but it is safe to say that the foundations of mathematical utility theory have been
shaken by the empirical evidence. Much work remains to be done.
PROBLEM SET 103
SUMMARY
The logic of the theory of investor choice can best be summarized by listing the
series of logical steps and assumptions necessary to derive the indifference curves
of Fig. 4.12.
• First, the five axioms of rational behavior were described.
• The expected utility rule was derived from the axioms.
• Cardinal utility functions were derived from the axioms.
• We assumed positive marginal utility. This and the expected utility rule were used
to argue that individuals will always maximize the expected utility of wealth.
• Risk premia were defined and a Pratt-Arrow measure of local risk aversion was
developed.
• Stochastic dominance was shown to be a general theory of choice that maximizes
expected utility for various classes of utility functions.
• Mean-variance indifference curves (which exhibit second-order stochastic dom-
inance for normally distributed returns) were developed as a parametric theory
of choice.
In Chapter 6 we shall use the mean-variance theory of choice as embodied in the
mean-variance indifference curves to describe the manner in which investors actually
choose optimal portfolios.
PROBLEM SET
4.1 State in your own words the minimum set of necessary conditions needed to obtain
mean-variance indifference curves like those graphed in Fig. Q4.1.
Mean
Variance
Figure Q4.1
Mean-variance indifference curves.
4.2 Figure 4.6 shows the utility curve of a risk lover. What does the indifference curve of a
risk lover look like?
4.3 You have a logarithmic utility function U(W) = In W, and your current level of wealth
is $5000.
104 THE THEORY OF CHOICE: UTILITY THEORY GIVEN UNCERTAINTY
a) Suppose you are exposed to a situation that results in a 50/50 chance of winning or
losing $1000. If you can buy insurance that completely removes the risk for a fee of $125,
will you buy it or take the gamble?
b) Suppose you accept the gamble outlined in (a) and lose, so that your wealth is reduced
to $4000. If you are faced with the same gamble and have the same offer of insurance
as before, will you buy the insurance the second time around?
4.4 Assume that you have a logarithmic utility function for wealth U(W) = In (W) and that
you are faced with a 50/50 chance of winning or losing $1,000. How much will you pay to
avoid this risk if your current level of wealth is $10,000? How much would you pay if your
level of wealth were $1,000,000?
4.5 Given the exponential utility function U(W) =
a) Graph the function, assuming a > 0.
b) Does the function exhibit-positive marginal utility and risk aversion?
c) Does the function have decreasing absolute risk aversion?
d) Does the function have constant relative risk aversion?
4.6 What kind of utility function of wealth might be consistent with an individual gambling
and paying insurance at the same time?
4.7 Suppose that A>B>C>D and that the utilities of these alternatives satisfy U(A) +
U(D) = U(B) + U(C). Is it true that U(-B + is greater than U(4/1 + +D) because the former
has a smaller variance? Why or why not?
4.8 A small businesswoman faces a 10% chance of having a fire that will reduce her net worth
to $1.00, a 10% chance that fire will reduce it to $50,000, and an 80% chance that nothing
detrimental will happen, so that her business will retain its worth of $100,000. What is the
maximum amount she will pay for insurance if she has a logarithmic utility function? In other
words, if U(W) = in W, compute the cost of the gamble. [Note: The insurance pays $99,999 in
the first case; $50,000 in the second; and nothing in the third.]
4.9 If you are exposed to a 50/50 chance of gaining or losing $1000 and insurance that
removes the risk costs $500, at what level of wealth will you be indifferent relative to taking
the gamble or paying the insurance? That is, what is your certainty equivalent wealth? Assume
your utility function is U(W) = —
4.10 Consider a lottery that pays $2 if n consecutive heads turn up in (n + 1) tosses of a fair
coin (i.e., the sequence of coin flips ends with the first tail). If you have a logarithmic utility
function, U(W) = In W, what is the utility of the expected payoff? What is the expected utility
of the payoff?
4.11 (Our thanks to David Pyle, University of California, Berkeley, for providing this problem.)
Mr. Casadesus's current wealth consists of his home, which is worth $50,000, and $20,000 in
savings, which are earning 7% in a savings and loan account. His (one-year) homeowner's
insurance is up for renewal, and he has the following estimates of the potential losses on his
house owing to fire, storm, etc., during the period covered by the renewal:
30,000 30 + AVLT
40,000 27 + AVL 2
50,000 24 + AVL 3
Mr. Casadesus expects neither to save nor to dissave during the coming year, and he does not
expect his home to change appreciably in value over this period. His utility for wealth at the
end of the period covered by the renewal is logarithmic, i.e., U(W) = ln(W).
a) Given that the insurance company agrees with Mr. Casadesus's estimate of his losses,
should he renew his policy (1) for the full value of his house, (2) for $40,000, or (3) for
$30,000, or (4) should he cancel it?
b) Suppose that Mr. Casadesus had $320,000 in a savings account. Would this change his
insurance decision?
c) If Mr. Casadesus has $20,000 in savings, and if his utility function is
U( W) = — 200,000 W -
should he renew his home insurance? And if so, for what amount of coverage?
[Note: Insurance covers the first x dollars of loss. For simplicity, assume that all losses occur
at the end of the year and that the premium is paid at the beginning of the year.]
4.12 Assume that security returns are normally distributed. Compare portfolios A and B,
using both first- and second-order stochastic dominance:
5A > aB !
A — aB !
A<UB
EA= E, EA> EB EA< E,
4.13 Given the following probability distributions for risky assets X and Y:
Probability X, Xi Probability Yi Y,
.1 —10 .2 2
.4 5 .5 3
.3 10 .2 4
.2 12 .1 30
a) If the only available choice is 100% of your wealth in X or 100% in Y and you choose on
the basis of mean and variance, which asset is preferred?
b) According to the second-order stochastic dominance criterion, how would you compare
them?
106 THE THEORY OF CHOICE: UTILITY THEORY GIVEN UNCERTAINTY
4.14 You have estimated the following probabilities for earnings per share of companies A
and B:
Probability A
.1 0 —.50
.2 .50 —.25
.4 1.00 1.50
.2 2.00 3.00
.1 3.00 4.00
a) Calculate the mean and variance of the earnings per share for each company.
b) Explain how some investors might choose A and others might choose B if preferences
are based on mean and variance.
c) Compare A and B, using the second-order stochastic dominance criterion.
4.16 Consider the following two risky scenarios for future cash flows for a firm:
Project I Project 2
.2 4,000 .4 0
.6 5,000 .2 5,000
.2 6,000 .4 10,000
Given that the firm has fixed debt payments of $8000, and limited liability, which scenario
will shareholders choose and why? How would your answer change if there were not limited
liability?
4.17 (Our thanks to Nils Hakansson, University of California, Berkeley, for providing this
problem.) Two widows, each with $10,000 to invest, have been advised by a trusted friend to
put their money into a one-year real estate trust which requires a minimum investment of
$10,000. They have been offered a choice of six trusts with the following estimated yields:
REFERENCES 107
Trust -2 -1 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
A .4 .2 .2 .2
B .1 .1 .1 .1 .1 .1 .1 .1 .1
C .2 .2 .2 .2 .2
D .2 .2 .1 .1 .1 .1 .2
E .4 .6
F .2 .2 .1 .1 .1 .1 .1 .1
Before making up their minds, they have called on you for advice.
a) The first widow leaves you unsure as to whether she is risk averse. What advice can you
give her?
b) The second widow shows definite risk aversion. What is your advice to her?
4.18
a) Reorder the six real estate trusts in Problem 4.17, using the mean-variance criterion.
b) Is the mean-variance ranking the same as that achieved by second-order stochastic
dominance?
REFERENCES
Levy, H., and Y. Kroll, "Stochastic Dominance with Riskless Assets," Journal of Financial
and Quantitative Analysis, December 1976, 743-778.
Markowitz, H., Portfolio Selection. Yale University Press, New Haven, 1959.
Porter, R. B.; J. R. Wart; and D. L. Ferguson, "Efficient Algorithms for Conducting Stochastic
Dominance Tests of Large Numbers of Portfolios," Journal of Financial and Quantitative
Analysis, January 1973, 71-82.
Pratt, J. W., "Risk Aversion in the Small and in the Large," Econometrica, January—April, 1964,
122-136.
Rubin, P. H., and C. W. Paul II, "An Evolutionary Model of Taste for Risk," Economic In-
quiry, October 1979, 585-596.
Tobin, J., "Liquidity Preference as a Behavior toward Risk," Review of Economic Studies,
February 1958, 65-86.
Tversky, A., and D. Kahneman, "Rational Choice and the Framing of Decisions," Journal of
Business, October 1986, S251—S278.
Vickson, R. G., "Stochastic Dominance for Decreasing Absolute Risk Aversion," Journal of
Financial and Quantitative Analysis, December 1975, 799-812.
, and M. Altman, "On the Relative Effectiveness of Stochastic Dominance Rules: Ex-
tension to Decreasingly Risk-Averse Utility Functions," Journal of Financial and Quan-
titative Analysis, March 1977, 73-84.
Von Neumann, J., and 0. Morgenstern, Theory of Games and Economic Behavior. Princeton
University Press, Princeton, N.J., 1947.
Walter, J. E., Dividend Policy and Enterprise Valuation. Wadsworth, Belmont, Calif., 1967.
Whitmore, G. A., "Third Degree Stochastic Dominance," American Economic Review, June
1970, 457-459.
, and M. C. Findlay, Stochastic Dominance. Lexington Books, D.C. Heath and Co.,
Lexington, Mass., 1975.
Ziemba, W. T., and R. G. Vickson, eds., Stochastic Optimization Models in Finance. Academic
Press, New York, 1975.
5
In this formulation the objects of choice are not derivative statistical
measures of the probability distribution of consumption opportunities
but rather the contingent consumption claims themselves set out in
extensive form.
J. Hirshleifer, "Efficient Allocation of Capital in an Uncertain World,"
American Economic Review, May 1964, 80
State-Preference Theory'
Finance deals with investment decisions of individuals and firms linked through the
supply and demand for securities in the capital market. Firms borrow capital for
investment in real assets by selling securities; individuals obtain claims to firms' real
assets by investing in securities. Thus securities present opportunities for intertempo-
ral shifts of consumption through the financing of productive activities. Individual
consumption/investment decisions that determine aggregate security demand, and
firm investment decisions that determine aggregate security supply are both affected
by security prices. By equating security supply and demand, security prices yield a
consistent set of firm and individual investment decisions. In this chapter, we will
analyze how optimal individual investment decisions and optimal firm investment
decisions are determined under uncertainty for a given set of security prices.
In Chapter 4 we found that, under specified conditions, individual decision making
under uncertainty is accomplished by maximizing expected utility of end-of-period
wealth. This decision criterion was shown to be valid when individuals are rational,
prefer more wealth to less, and follow the five axioms of choice under uncertainty.
Implicitly, it was also assumed that individuals can assess a security's probability
distribution of end-of-period payoffs. It was shown that the expected utility criterion
is a very simple way of choosing among mutually exclusive investments having dif-
ferent probability distributions of end-of-period payoffs. By choosing the investment
Ronald W. Masulis was the primary author of the chapter text and has benefited from lecture notes on
this topic by Herbert Johnson.
109
110 STATE-PREFERENCE THEORY
with the highest expected utility, the optimal investment is determined, thus con-
densing a choice across N probability distributions of end-of-period payoffs into a
comparison among N expected utility values.
In this and following chapters, we wish to move beyond the individual's choice
problem of mutually exclusive investments to the more general problem of portfolio
decision making, i.e., the optimal choice of investing in more than one risky security.
This is equivalent to the problem of choosing an individual's probability distribution
of end-of-period wealth that is consistent with the set of available risky securities and
the individual's initial wealth. The individual's choice problem is to find that port-
folio or linear combination of risky security quantities that is optimal, given his or
her initial wealth and tastes. We assume a perfect capital market to ensure that there
are no costs of portfolio construction.
Figure 5.1
Elementary state—contingent claim.
States of Nature 1, 2, 3, .. . , S
Prosperity
Normalcy
Recession
Depression Market securities j, k, . J
Securities are
♦ defined by patterns
of payoffs under
Payoffs different states.
High
Medium
Low
Zero
Figure 5.2
States, payoffs, and securities.
One person might think the security with payoff (0, 1, 0) is worth more than someone else does, but if
the security cannot be formed from a portfolio of existing market securities, then these virtual prices that
different people would assign to this hypothetical security need not/be the same.
5
See Appendix A to this chapter for a general method of determining whether a complete market exists.
DERIVATION OF PURE SECURITY PRICES 113
D. DERIVATION OF PURE
SECURITY PRICES
Given that we know the state-contingent payoff vectors of both the market securities
and the pure securities, we wish to develop the relationship between the prices of the
market securities and pure securities in a perfect and complete capital market.
The following notation will be used throughout this chapter:
Th= prices of pure securities,
p j = prices of market securities,
ns = state probabilities individuals' beliefs about the relative likelihoods
of states occurring,
Q, = number of pure securities.
Let us begin with an analogy. The Mistinback Company sells baskets of fruit,
limiting its sales to only two types of baskets. Basket 1 is composed of 10 bananas
and 20 apples and sells for $8. Basket 2 is composed of 30 bananas and 10 apples
and sells for $9. The situation may be summarized by the payoffs set forth in Table
5.1.
Using the relationships in Table 5.1, we can solve for the prices of apples and
bananas separately. Let us denote apples by A, bananas by B, the baskets of fruit by
1 and 2, and the quantity of apples and bananas in a basket by QJA and Q113 , re-
spectively. Using this notation, we can express the prices of the two baskets as follows:
6 While a complete market may appear to require an unreasonably large number of independent securities,
Ross [1976] showed that in general if option contracts can be written on market securities and market
securities have sufficiently variable payoffs across states, an infinite number of linearly independent security
and option payoffs can be formed from a small number of securities.
114 STATE-PREFERENCE THEORY
Only p, and p, are unknown. Thus there are two equations and two unknowns, and
the system is solvable as follows (substitute the known values in each equation):
(1) $8 = pA 20 + pB10, (2) $9 = pA10 + p,30.
Subtract three times Eq. (1) from Eq. (2) to obtain p A :
$ 9 = PA 10 PB30
—$24 = —pA60 — pB30
—$ 15 = —
PA 50
PA = $. 30.
Then substituting the value of p A into Eq. (1), we have
$8 = ($.30)20 + pB10 = $6 + pB10,
$2 = p,10,
PB = $.20.
Given that we know the prices of the market securities, we may now apply this
same analysis to the problem of determining the implicit prices of the pure securities.
Consider security j, which pays $10 if state 1 occurs and $20 if state 2 occurs; its
price is $8. Security k pays $30 if state 1 occurs and $10 if state 2 occurs; its price is
$9. Note that state 1 might be a gross national product (GNP) growth of 8% in real
terms during the year, whereas state 2 might represent a GNP growth rate of only
1% in real terms. This information is summarized in Table 5.2.
Any individual security is similar to a mixed basket of goods with regard to
alternative future states of nature. Recall that a pure security pays $1 if a specified
state occurs and nothing if any other state occurs. We may proceed to determine the
price of a pure security in a manner analogous to that employed for the fruit baskets.
The equations for determining the price for two pure securities related to the situa-
tion described are
PiQi i + P2Q.i 2 = pi ,
It should be emphasized that the Pi of $.20 and the /32 of $.30 are the prices of the
two pure securities and not the prices of the market securities j and k. Securities j
and k represent portfolios of pure securities. Any actual security provides different
payoffs for different future states. But under appropriately defined conditions, the
prices of market securities permit us to determine the prices of pure securities. Thus
our results indicate that for pure security 1 a $.20 payment is required for a promise
of a payoff of $1 if state 1 occurs and nothing if any other states occur. The concept
of pure security is useful for analytical purposes as well as for providing a simple
description of uncertainty for financial analysis.
Capital market equilibrium requires that market prices be set so that supply equals
demand for each individual security. In the context of the state-preference frame-
work, one condition necessary for market equilibrium requires that any two securities
or portfolios with the same state-contingent payoff vectors must be priced identically.'
Otherwise, everyone would want to buy the security or portfolio with the lower price
and to sell the security or portfolio with the higher price. If both securities or port-
folios are in positive supply, such prices cannot represent an equilibrium. This con-
dition is often called the single-price law of markets.
If short selling is allowed in the capital market, we can obtain a second related
necessary condition for market equilibrium, i.e., the absence of any riskless arbitrage
profit opportunity. To short sell a security, an individual borrows the security from
a current owner and then immediately sells the security in the capital market at the
current price. Then, at a later date, the individual goes back to the capital market
This condition implies the absence of any first-order stochastically dominated market securities. Other-
wise the former payoff per dollar of investment would exceed the latter payoff per dollar of investment
in every state. The latter security would be first-order stochastically dominated by the former security.
116 STATE-PREFERENCE THEORY
and repurchases the security at the then-current market price and immediately returns
the security to the lender. If the security price fell over the period of the short sale,
the individual makes a profit; if the security price rose, he or she takes a loss. In
either case the short seller's gain or loss is always the negative of the owner's gain
or loss over this same period.
When two portfolios, A and B, sell at different prices, where PA > pB, but have
identical state-contingent payoff vectors, we could short sell the more expensive port-
folio and realize a cash flow of p,, then buy the less expensive portfolio, for a negative
cash flow of pB. We would realize a positive net cash flow of (p,— pB), and at the
end of the period, we could at no risk take our payoff from owning portfolio B to
exactly repay our short position in portfolio A. Thus the positive net cash flow at
the beginning of the period represents a riskless arbitrage profit opportunity. Since
all investors are assumed to prefer more wealth to less, this arbitrage opportunity is
inconsistent with market equilibrium.
In a perfect and complete capital market, any market security's payoff vector
can be exactly replicated by a portfolio of pure securities. Thus it follows that when
short selling is allowed, the no—arbitrage profit condition requires that the price of
the market security be equal to the price of any linear combination of pure securities
that replicates the market security's payoff vector.
F. ECONOMIC DETERMINANTS OF
SECURITY PRICES
To understand how (1) affects security prices, we need to recognize that a riskless
security can always be constructed in a complete capital market simply by forming
a portfolio composed of one pure security for each state. Hence the payoff on this
portfolio is riskless since a dollar will be paid regardless of what state is realized. In
the case of three states the price of this riskless portfolio is the sum of the prices of
the three individual pure securities p i + p, + p 3 = .8, for example. The price of a
riskless claim to a dollar at the end of the period is just the present value of a dollar
ECONOMIC DETERMINANTS OF SECURITY PRICES 117
where 0 < ps < 1. Thus the pure security's expected rate of return is
$1 • it $1
E(Rs) = s 1 = — — 1, where 0 < 0, < 1,
Ps Bs
The property that individuals prefer to consume a dollar of resources today, rather than consume the
same dollar of resources tomorrow, is called time preference for consumption. As discussed in Chapter 1,
an individual's marginal rate of time preference for consumption is equal to his or her marginal rate of
substitution of current consumption and certain end-of-period consumption. In a perfect capital market,
it was also shown that the marginal rates of time preference for all individuals are equal to the market
interest rate.
118 STATE-PREFERENCE THEORY
be the same only when the prices of the pure securities vary proportionally with the
state probabilities.
The third determinant of security prices, and a second cause for differences in
these prices, is individuals' attitudes toward risk when there is variability in aggregate
wealth across states. Assuming that individuals are risk averse, they will diversify by
investing in some of each pure security to ensure that they are not penniless regard-
less of what state is realized.' In fact, if the prices, Os's, of expected payoffs of a dollar
contingent on a particular state occurring were the same for all states (and thus the
expected rates of return of pure securities are all equal), then each risk-averse indi-
vidual would want to invest in an equal number of each pure security so as to elimi-
nate all uncertainty about his or her future wealth. Not everyone can do this, however,
since aggregate wealth is not the same in every state; i.e., there is nondiversifiable risk
in the economy and it must be borne by someone. Consider the following example.
End-of-period aggregate wealth can be one, two, or three trillion dollars, depending
on whether the depressed, normal, or prosperous state occurs; then the average in-
vestor must hold a portfolio with a payoff vector of the form (X, 2X, 3X). Because
individuals are risk averse, dollar payoffs are more valuable in states where they have
relatively low wealth, which in this example is state 1. In order for people to be in-
duced to bear the risk associated with a payoff vector of form (X, 2X, 3X), pure se-
curity prices must be adjusted to make the state 1 security relatively expensive and
the state 3 security relatively cheap. In other words, to increase demand for the rela-
tively abundant state 3 securities, prices must adjust to lower the expected rate of
return on state 1 securities and to raise the expected rate of return on state 3 securities.
If aggregate wealth were the same in some states, then risk-averse investors would
want to hold the same number of pure securities for these states and there would be
no reason for prices of expected dollar payoffs to be different in these states. Investors
would not want to hold unequal numbers of claims to the states with the same aggre-
gate wealth because this would mean bearing risk that could be diversified away,
and there is no reason to expect a reward for bearing diversifiable risk. So it is the
prospect of a higher portfolio expected return that induces the risk-averse investors
to bear nondiversifiable risk. Thus risk aversion combined with variability in end-of-
period aggregate wealth causes variation in the prices (Os's) of dollar expected payoffs
across states, negatively related to the aggregate end-of-period wealth or aggregate
payoffs across states. This in turn causes like variations in the pure security prices.
There is a very important condition implicit in the previous discussion. We found
that when investors are risk averse, securities that pay off relatively more in states
with low aggregate wealth have relatively low expected rates of return, whereas secu-
rities that pay off relatively more in states with high aggregate wealth have relatively
high expected rates of return. Since aggregate wealth is equal to the sum of the pay-
offs on all market securities, it is also termed the payoff on the market portfolio. Secu-
rities with state-contingent payoffs positively related to the state-contingent payoffs
on the market portfolio, and which therefore involve significant nondiversifiable risk
bearing, have higher expected rates of return than securities that have payoffs nega-
This also requires the utility function to exhibit infinite marginal utility at a zero wealth level.
OPTIMAL PORTFOLIO DECISIONS 119
tively or less positively related to the payoffs on the market portfolio, and which
therefore involve little nondiversifiable risk bearing. We will return to this important
condition in Chapter 7.
It follows from this analysis that a pure security price can be decomposed into
three factors:
$ 1Es $1 r 1+r 1
PS = 70, = = ns
1 + E(RS) 1+r 1 + E(RS) ]
$1 r E(RS) — r i
= 7E 1 , where E(RS) r.
1+r s 1 + E(RS)
The first factor is an end-of-period dollar payoff discounted to the present at the
riskless rate. It is multiplied by the second factor, which is the probability of payoff.
The third factor, in brackets, is a risk adjustment factor. Note that if investors are
all risk neutral, the expected rate of return on all securities will be equal to the riskless
interest rate, in which case the above risk adjustment factor (i.e., the factor in brackets)
becomes one. In summary, security prices are affected by (1) the time value of money,
(2) the probability beliefs about state-contingent payoffs, and (3) individual prefer-
ences toward risk and the level of variability in aggregate state-contingent payoffs
or wealth (i.e., the level of nondiversifiable risk in the economy).
Now that we have developed the basic structure of state-preference theory, we will
return to the problem of optimal portfolio choice in a perfect and complete capital
market. This will then be followed by an analysis of a firm's optimal investment prob-
lem, also in a perfect and complete capital market. Since any portfolio payoff pattern
can be constructed from the existing market securities or from a full set of pure securi-
ties in a complete capital market, we can obtain the same optimal portfolio position
whether we frame the analysis in terms of market securities or pure securities. Since
pure securities are much simpler to analyze, we will phrase the optimal portfolio
problem in terms of these securities. Thus we can write an individual's expected utility
of end-of-period wealth as E gsU(Q,), where Q,, = number of pure securities paying
a dollar if state s occurs. In this context Q, represents the number of state s pure secu-
rities the individual buys as well as his or her end-of-period wealth if state s occurs.
Now consider the problem we face when we must decide how much of our initial
wealth, W0 , to spend for current consumption, C, and what portfolio of securities to
hold for the future. We wish to solve the problem'
10 This formulation assumes that the utility function is separable into utility of current consumption and
utility of end-of-period consumption. In principle, the utility functions, u(C) and U(Q,), can be different
functions.
120 STATE-PREFERENCE THEORY
subject to
psQ, + $1C Wo . (5.2)
That is, we are maximizing our expected utility of current and future consumption
subject to our wealth constraint. Our portfolio decision consists of the choices we
make for Qs , the number of pure securities we buy for each state s. Note that there
is no explicit discounting of future utility, but any such discounting could be absorbed
in the functional form for U(Qs). In addition, the ps's include an implicit market dis-
count rate. There is no need to take an expectation over u(C), our utility of current
consumption, since there is no uncertainty concerning the present.
There are two ways to maximize expected utility subject to a wealth constraint.
We could solve (5.2) for one of the Q;s, say Q 1, and then eliminate this variable from
(5.1). Sometimes this is the easiest way, but more often it is easier to use the Lagrange
multiplier method (see Appendix D at the end of the book):
where the prime denotes partial differentiation with respect to the argument of the
function. Next, we take partial derivatives with respect to Q 1, Q 2 , and so on. For
each Q t , we will pick up one term from the expected utility and one from the wealth
constraint (all other terms vanish):
are .4 and .6 and the state probabilities are 4 and 4, respectively. The Lagrangian
function is
L = ln C + in Q 1 + in Q 2 — 2(.4Q, + .6Q 2 + C — 10,000),
and the first-order conditions are
OL 1 1
A which implies C = —A ,
(a) = = (),
OL 1
(b) 3,21 .42 = 0, which implies Qi =
a421
1.22 '
01, 2 1
= 0, which implies Q2 =
.92 '
01 34
)° 22 22
aL
OA = 10,000 — C — .4Q 1 — .6Q 2 = 0.
(d) —
4_ .6
(d') 10,000,
1..22 + .92
and multiplying by 2 yields
1
(d") 1 + 3 + 3 = 10,0002, which yields 2=
5,000
Now, substituting this value of 2 back into Eqs. (a), (b), and (c) yields the optimal
consumption and investment choices, C = $5,000, Q 1 = 4166.7, and Q 2 = 5555.5.
Substituting these quantities back into the wealth constraint verifies that this is indeed
a feasible solution. The investor in this problem divides his or her wealth equally
between current and future consumption, which is what we should expect since the
risk-free interest rate is zero [i.e., ps = 1 = 1/(1 + r)] and there is no time preference
for consumption in this logarithmic utility function. However, the investor does buy
more of the state 2 pure security since the expected rate of return on the state 2 pure
security is greater. Because the utility function exhibits risk aversion, the investor also
invests some of his or her wealth in the state 1 pure security.
In this example we assumed that the investor is a price taker. In a general equi-
librium framework, the prices of the pure securities would be determined as part of
the problem; i.e., they would be endogenous. The prices would be determined as a
result of individuals' constrained expected utility maximization (which determines
the aggregate demands for securities), and firms' optimal investment decisions (which
determine the aggregate supplies of securities). The critical condition required for
equilibrium is that the supply of each market security equal its aggregate demand.
In a complete capital market this equilibrium condition can be restated by saying
that the aggregate supply of each pure security is equal to its aggregate demand.
122 STATE-PREFERENCE THEORY
H. PORTFOLIO OPTIMALITY
CONDITIONS AND PORTFOLIO
SEPARATION
and
ztU 1(2t) Pt
= for any two states s and t. (5.8)
nstr(Q5) Ps
In both cases, the optimal allocation of wealth represents choosing C and the Qs's
so that the ratio of expected marginal utilities equals the ratio of market prices for
the C and the Qs's. That is, the optimal consumption and investment choices involve
choosing points on the various indifference curves (curves of constant expected utility)
that are tangent to the associated market lines. This is equivalent to choosing con-
sumption and investment weights so that the slopes of the indifference curves (which
are defined as the negative of the marginal rates of substitution) representing current
consumption and future consumption contingent on state t (as in Fig. 5.3) or rep-
resenting future consumption contingent on state s and state t (as in Fig. 5.4) are
equal to the slopes of the respective market lines (representing the market exchange
rates, e.g., — pdps).
An alternative way of stating the optimality conditions of the above portfolio is
that the expected marginal utilities of wealth in state s, divided by the price of the
state s pure security, should be equal across all states, and this ratio should also
be equal to the marginal utility of current consumption. This is a reasonable result;
if expected marginal utility per pure security price were high in one state and low
Figure 5.3
Optimal consumption/investment
decisions.
Indifference curve
(Qt)
Slope =
u' ( C)
PORTFOLIO OPTIMALITY CONDITIONS AND PORTFOLIO SEPARATION 123
Figure 5.4
Optimal portfolio decisions.
Indifference curve
(1'(Qt)
Irt ,(Q0
Slope = irs u
in another, then we must not have maximized expected utility. We should increase
investment in the high expected marginal utility security at the expense of the security
yielding low expected marginal utility. But as we do that, we lower expected marginal
utility where it is high and raise it where it is low, because a risk-averse investor's
marginal utility decreases with wealth (his or her utility function has a positive but
decreasing slope). Finally, when Eq. (5.8) is satisfied, there is no way left to increase
expected utility.'
When investors' portfolio choices over risky securities are independent of their
individual wealth positions, we have a condition known as portfolio separation. This
condition requires that there are either additional restrictions on investor preferences
or additional restrictions on security return distributions.' Under portfolio separa-
tion, investors choose among only a few basic portfolios of market securities. Thus the
importance of having a complete market is greatly decreased. Recall that when capital
markets are incomplete, individuals are limited in their choices of state-contingent
payoff patterns to those payoff patterns that can be constructed as linear combina-
tions of existing market securities. However, with portfolio separation, investors will
12
This entire procedure will ordinarily not work if the investor is risk neutral instead of risk averse. A
risk-neutral investor will plunge entirely into the security with the highest expected return. He or she would
like to invest even more in this security but, being already fully invested in it, cannot do so. Equation (5.8)
will not hold for risk neutrality.
13
Cass and Stiglitz [1970] proved that for arbitrary security return distributions, utility functions with
the property of linear risk tolerance yield portfolio separation. The risk tolerance of the utility function
is the reciprocal of the. Pratt-Arrow measure of absolute risk aversion discussed in Chapter 4. Thus a
linear risk-tolerance utility function can be expressed as a linear function of wealth:
— U'(W)/U"(W) = a + bW. (5.9)
If investors also have homogeneous expectations about state probabilities and all investors have the
same b, then there is two fund separation where all investors hold combinations of two basic portfolios.
Utility functions exhibiting linear risk tolerance include the quadratic, logarithmic, power, and ex-
ponential functions. Ross [1976] proved that for arbitrary risk-averse utility functions a number of classes
of security return distributions (including the normal distribution, some stable Paretian distributions, and
some distributions that are not stable Paretian, e.g., fat-tailed distributions with relatively more extreme
values) yield portfolio separation.
124 STATE-PREFERENCE THEORY
often find that the infeasible payoff opportunities would not have been chosen even
if they were available. Thus under portfolio separation, investor portfolio decisions will
often be unaffected by whether or not the capital market is complete.
Portfolio separation has been shown to depend on the form of the utility function
of individuals and the form of the security return distributions. In the special case
where investor utility functions of wealth are quadratic, or security returns are joint-
normally distributed, portfolio separation obtains. With the addition of homogeneous
expectations, portfolio separation provides sufficient conditions for a security-pricing
equation. This security-pricing relationship can be expressed in terms of means and
variances and is called the capital asset pricing model." The resulting form of the
security-pricing equation is particularly convenient for formulating testable proposi-
tions and conducting empirical studies. As we shall discuss in the following chapters,
many of the implications of portfolio separation in capital markets appear to be
consistent with observed behavior and have been supported by empirical tests.
14
The capital asset pricing model is discussed in detail in Chapter 7.
15
Hirshleifer [1964, 1965, 1966] and Myers [1968] were among the first papers to apply state-preference
theory to corporate finance problems.
FIRM VALUATION, THE FISHER SEPARATION PRINCIPLE 125
In shifting from firm decision making under certainty to decision making under
uncertainty, it is important to know under what conditions, if any, Fisher separation
continues to hold. It can be shown that firms that are maximizing the price of current
shares are also maximizing current shareholders' expected utility when the capital
market is (1) perfectly competitive and frictionless and (2) complete. The first condition
ensures that firm actions will not be perceived to affect other firms' market security
prices, whereas the second ensures that the state-space "spanned" by the existing set
of linearly independent market securities (i.e., the set of risk opportunities) is un-
affected by the firm's actions. Thus firm actions affect shareholders' expected utility
only by affecting their wealth through changes in the firm's current share price. This
is analogous to the certainty case in Chapter 1.
The two conditions of a perfect and complete capital market ensure that the prices
of a full set of pure securities can be obtained from the prices of the market securities,
and vice versa, given the state-contingent payoff vectors of the market securities. As
a result the firm's objective function of maximizing current share price can be phrased
in terms of a portfolio of pure securities that replicates its shares. The firm's objective
function then becomes E Qis ps , where 12;s is defined as the state s—contingent end-
of-period payoff on firm j's existing securities. In this formulation the price of a firm's
current shares is determined by (1) the firm's state-contingent production function
Q i, = OA, s), which transforms current resources into state-contingent future pay-
offs, and (2) the initial investment which represents the current cost to the firm of
producing its state-contingent payoff. It follows that the price Y, for which the current
owners could sell the firm prior to making the investment is
= E Ty<i, - (5.10)
where the firm's technological constraints are captured in its production function
Qs = Nip s).
For divisible investment projects, the production function is assumed to be a
continuous differentiable function of that exhibits diminishing returns to scale and
has the property that zero investment yields zero output. The firm's optimal invest-
ment scale is determined by the first-order condition
d
Yi = E ps 01.(I s) — 1 = 0. (5.11)
d/ j "
For indivisible investment projects with finite scale the optimal investment rule is to
accept all projects with positive net present value. In this context, (5.10) represents
the net present value of the project's state-contingent net cash flow.
It is important to note that acceptance of positive NPV investments increases the
price of the firm's current stock and therefore the wealth and expected utility of all
current shareholders in a perfect and complete capital market. Since all shareholders
are made better off by these investment decisions, these firm investment decisions are
unanimously supported by all the firm's current shareholders. However, if the capital
market is incomplete or imperfect, this is not necessarily true, because the firm's
investment decisions may affect the price of other firms' shares or the feasible set of
126 STATE-PREFERENCE THEORY
State 1 100 10
State 2 30 12
Firm A's stock price = 62; initial investment cost
of its project = 10.
Proposed
Stock Investment Project
State 1 40 12
State 2 90 6
Firm B's stock price = 56; initial investment cost
of its project = 8.
state-contingent payoffs. As a result, increasing the price of a firm's shares may not
increase the wealth of all current shareholders (since the prices of some of their other
shareholdings may fall) and may not maximize shareholder expected utility (since the
opportunity set of feasible end-of-period payoffs may have changed).16
Let us now consider an example of a firm investment decision problem in a two-
state world of uncertainty. Assume that all investors are expected utility maximizers
and exhibit positive marginal utility of wealth (i.e., more wealth is preferred to less).
Consider the following two firms and their proposed investment projects described
in Tables 5.3 and 5.4.
To determine whether either firm should undertake its proposed project, we need
to first determine whether the capital market is complete. Since the state-contingent
payoffs of the two firms' stocks are linearly independent, the capital market is com-
plete. In a complete market, the Fisher separation principle holds, so that the firm
need only maximize the price of current shares to maximize its current shareholders'
expected utility. This requires that the firm invest only in positive net present value
16
See DeAngelo [1981] for a critical analysis of the unanimity literature and a careful formulation of the
conditions under which it holds in incomplete and complete capital markets.
FIRM VALUATION, THE FISHER SEPARATION PRINCIPLE 127
investment projects, which requires knowing the pure security prices in the two states.
These two prices can be obtained from the market prices of the two firms' stocks
and their state-contingent payoffs by solving the two simultaneous equations
100p 1 + 30p 2 = 62,
40p1 + 90p2 = 56,
to obtain the solution p i = .5 and p 2 = .4. To calculate the net present value of the
two projects, we use the NPV definition in (5.10):
NPVA = 10p1 + 12p2 — Io = 10(.5) + 12(.4) — 10 = — .2
and
NPV„, = 12p, + 6p 2 — /=,1 = 145) + 6(.4) — 8 = .4.
Since firm A's project has a negative NPV, it should be rejected, whereas firm B's
project should be accepted since it has a positive NPV.
In examining this optimal investment rule, it should be clear that the prices of
the pure securities affect the firm's investment decisions. It follows that since these
security prices are affected by (1) time preference for consumption and the productivity
of capital, (2) the probability of state-contingent payoffs, and (3) individual preferences
toward risk and the level of nondiversifiable risk in the economy, firm investment
decisions are also affected by these factors.
We have applied state-preference theory to the firm's optimal investment decision
while assuming that the firm has a simple capital structure represented by shares of
stock. However, it is also possible to allow the firm to have more complicated capital
structures, which include various debt, preferred stock, and warrant contracts. In
doing this, state-preference theory can be used to address the important question of
a firm's optimal financing decision.' For this purpose it has been found useful to
order the payoffs under alternative states. One can think of the payoffs for future
states as arranged in an ordered sequence from the lowest to the highest payoff.
Keeping in mind the ordered payoffs for alternative future states, we can specify the
conditions under which a security such as corporate debt will be risk free or risky."
The state-preference model has also been very useful in developing option pricing
theory. By combining securities with claims on various portions of the ordered pay-
offs and by combining long and short positions, portfolios with an infinite variety of
payoff characteristics can be created. From such portfolios various propositions with
regard to option pricing relationships can be developed.'
17
There are many examples of the usefulness of state-preference theory in the area of optimal capital
structure or financing decisions; see, e.g., Stiglitz [1969], Mossin [1977], and DeAngelo and Masulis [1980a
and 1980b].
18
For applications of this approach see Kraus and Litzenberger [1973] and DeAngelo and Masulis
[1980a].
19
For some further applications of state-preference theory to option pricing theory see Merton [1973]
and Ross [1976], and for application of both theories to investment and financing decision making see
Appendixes B and C or Banz and Miller [1978].
128 STATE-PREFERENCE THEORY
SUMMARY
Wealth is held over periods of time, and the different future states of nature
will change the value of a person's wealth position over time. Securities represent
positions with regard to the relation between present and future wealth. Since
securities involve taking a position over time, they inherently involve risk and
uncertainty.
The states of nature capture a wide variety of factors that influence the future
values of risky assets. Individuals must formulate judgments about payoffs under
alternative future states of nature. From these state-contingent payoffs and market
prices of securities, the prices of the underlying pure securities can be developed in
a complete and perfect capital market. Given these pure security prices, the price of
any other security can be determined from its state-contingent payoff vector. Con-
ceptually, the equilibrium prices of the pure securities reflect the aggregate risk pref-
erences of investors and investment opportunities of firms. Furthermore, the concept
of a pure security facilitates analytical solutions to individuals' consumption/portfolio
investment decisions under uncertainty.
The state-preference approach is a useful way of looking at firm investment deci-
sions under uncertainty. In a perfect and complete capital market the net present
value rule was shown to be an optimal firm investment decision rule. The property
that firm decisions can be made independently of shareholder utility functions is
termed the Fisher separation principle. State-preference theory also provides a con-
ceptual basis for developing models for analyzing firm capital structure decisions and
the pricing of option contracts. Thus the state-preference approach provides a useful
way of thinking about finance problems both for the individual investor and for the
corporate manager.
In summary the state-preference model has been shown to be very useful in a
world of uncertainty. It can be used to develop optimal portfolio decisions for in-
dividuals and optimal investment rules for firms. We have found that in perfect and
complete capital markets a set of equilibrium prices of all outstanding market secu-
rities can be derived. Further, these prices have been shown to be determined by (1)
individual time preferences for consumption and the investment opportunities of
firms, (2) probability beliefs concerning state-contingent payoffs, and (3) individual
preferences toward risk and the level of nondiversifiable risk in the economy.
A number of important concepts developed in this chapter will be found to be
very useful in the analysis to follow. In Chapter 6 we develop the fundamental prop-
erties of the mean-variance model, where the concept of diversifiable and nondiver-
sifiable risk takes on added importance. In Chapter 7 the mean-variance framework
is used to develop market equilibrium relationships that provide an alternative basis
for pricing securities. Known as the capital asset pricing model, this model, like the
state-preference model, has the property that securities having relatively high levels
of nondiversifiable risk have relatively higher expected rates of return. The concept
of arbitrage is fundamental to the development of arbitrage pricing theory (APT) in
Chapter 7 and the option pricing model (OPM) in Chapter 8.
PROBLEM SET 129
PROBLEM SET
5.1 Security A pays $30 if state 1 occurs and $10 if state 2 occurs. Security B pays $20 if state
1 occurs and $40 if state 2 occurs. The price of security A is $5 and the price of security B is
$10.
Payoff
Security k 24 10 pk = 20
5.3 Interplanetary starship captain Jose Ching has been pondering the investment of his recent
pilot's bonus of 1000 stenglers. His choice is restricted to two securities: Galactic Steel, selling
for 20 stenglers per share, and Nova Nutrients, at 10 stenglers per share. The future state of
his solar system is uncertain. If there is a war with a nearby group of asteroids, Captain Ching
expects Galactic Steel to be worth 36 stenglers per share. However, if peace prevails, Galactic
Steel will be worth only 4 stenglers per share. Nova Nutrients should sell at a future price of
6 stenglers per share in either eventuality.
a) Construct the payoff table that summarizes the starship captain's assessment of future
security prices, given the two possible future states of the solar system. What are the prices
of the pure securities implicit in the payoff table?
b) If the captain buys only Nova Nutrients shares, how many can he buy? If he buys only
Galactic Steel, how many shares can he buy? What would be his final wealth in both
cases in peace? At war?
c) Suppose Captain Ching can issue (sell short) securities as well as buy them, but he must
be able to meet all claims in the future. What is the maximum number of Nova Nutrients
shares he could sell short to buy Galactic Steel? How many shares of Galactic Steel could
he sell short to buy Nova Nutrients? What would be his final wealth in both cases and
in each possible future state?
d) Suppose a third security, Astro Ammo, is available and should be worth 28 stenglers per
share if peace continues and 36 stenglers per share if war breaks out. What would be the
current price of Astro Ammo?
e) Summarize the results of (a) through (d) on a graph with axes W, and W2.
130 STATE-PREFERENCE THEORY
f) Suppose the captain's utility function can be written U = WP W. If his investment is re-
stricted to Galactic Steel and/or Nova Nutrients, what is his optimal portfolio, i.e., how
many shares of each security should he buy or sell?
5.4 Ms. Mary Kelley has initial wealth W0 = $1200 and faces an uncertain future that she
partitions into two states, s = 1 and s = 2. She can invest in two securities, j and k, with initial
prices of pi = $10 and pk = $12, and the following payoff table:
Payoff
Security s=1 s=2
j $10 $12
k 20 8
a) If she buys only security j, how many shares can she buy? If she buys only security k,
how many can she buy? What would her final wealth, Ws, be in both cases and each state?
b) Suppose Ms. Kelley can issue as well as buy securities; however, she must be able to meet
all claims under the occurrence of either state. What is the maximum number of shares
of security j she could sell to buy security k? What is the maximum number of shares of
security k she could sell to buy security j? What would her final wealth be in both cases
and in each state?
c) What are the prices of the pure securities implicit in the payoff table?
d) What is the initial price of a third security i for which Q„ = $5 and Qi2 = $12?
e) Summarize the results of (a) through (d) on a graph with axes W1 and W2.
f) Suppose Ms. Kelley has a utility function of the form U Wt. Find the optimal
portfolio, assuming the issuance of securities is possible, if she restricts herself to a portfolio
consisting only of j and k. How do you interpret your results?
5.5 Two securities have the following payoffs in two equally likely states of nature at the end
of one year:
Payoff
Security s=1 s=2
$10 $20
k 30 10
Security j costs $8 today, whereas k costs $9, and your total wealth is currently $720.
a) If you wanted to buy a completely risk-free portfolio (i.e., one that has the same payoff in
both states of nature), how many shares of j and k would you buy? (You may buy fractions
of shares.)
b) What is the one-period risk-free rate of interest?
c) If there were two securities and three states of nature, you would not be able to find a com-
pletely risk-free portfolio. Why not?
REFERENCES =3=
P.6 Suppose there are only two possiîle future states of the world, and the utility function is
logarithmic. Let the proîaîility of state =, n,, equal 3, and the prices of the pure securities, p,
and p 2 , equal $0.60 and $0.40, respectively. An individual has an initial wealth or endowment,
Wo , of $P0,000.
a) What amounts will the risk-averse individual invest in pure securities = and <?
î) How will the individual divide his or her initial endowment îetween current and future
consumption?
<°
[Hint- Use the wealth constraint instead of the Lagrange multiplier technique.]
REFERENCES
<0 Proîlem P.6 was suggested îy Professor Herî Johnson of the University of California, Davis.
=3< STATE-PREFERENCE THEORY
If we have n market securities and n states of the world, but we are not sure the n
market securities are independent, we can find out by taking the determinant of the
payoffs from the securities: a nonzero determinant implies independence. For exam-
ple, the set of pure securities is independent since
1 0 0 1 0 0
1 1
0 1 0 = 1; but 0 1 1 =0
1 1
0 0 1 1 1 1
implies that the security payoffs (1, 0, 0), (0, 1, 1), and (1, 1, 1) are not linearly
independent.
We can use Appendix B, "Matrix Algebra," found at the end of the book, to form
a portfolio of pure securities from an arbitrary complete set of market securities. This
involves computing the inverse of the payoff matrix for the actual securities. For ex-
ample, if (1, 0, 0), (0, 1, 1), and (0, 1, 3) are available, then define
1 0 0\
A= (0 1 1
0 1 3/
1 0 0
1 1
IA1 = 0 1 1 = 2 0 0.
1 3
0 1 3
133
134 STATE-PREFERENCE THEORY
Let X ii be the amount of the jth security one buys in forming the ith pure security,
and let X be the matrix formed from the X. Then we require that
/1 0 0
XA = I where I= 0 1 0
\0 0 1/
is the identity matrix and also the matrix of payoffs from the pure securities. Hence
X = A - 1. In the present example
/2 0 0\ /1 0 0\
A -= =4 0 3 —1 = 0
\0 —1 1/ \0 —1
-=
We then multiply X times A or equivalently A A to obtain a matrix of payoffs
from the pure securities. We have:
/1 0 0\ /1 0 0\ /1 0 0\
0 z —1 0 1 1 = 0 1 0
\0 \ O 1 3/ \0 0 1/
We can now see that the purpose of finding the inverse of A is to obtain directions
for forming a portfolio that will yield a matrix of payoffs from the pure securities
the identity matrix. Recall that the three securities available are: (1, 0, 0), (0, 1, 1),
and (0, 1, 3). To obtain the pure security payoff (1, 0, 0), we buy the security with that
pattern of payoffs under the three states. To obtain (0, 1, 0), we buy z of (0, 1, 1) and
sell short z of (0, 1, 3). To obtain (0, 0, 1), we sell short z of (0, 1, 1) and buy z of (0, 1, 3).
Appendix B to Chapte] 5:
Use of P]ices fo] State-
Contingent Claims in
Capital Budgeting
Banz and Miller [1978] develop estimates of prices for state-contingent claims that
can be applied in capital budgeting problems of the kind discussed in Chapter 12.
They employ option pricing concepts discussed later in this book in Chapter 8. Sim-
ilar methodologies were developed about the same time by Garman [1978] and by
Breeden and Litzenberger [1978].
Banz and Miller note that a fundamental breakthrough was provided in Ross
[1976], who demonstrated that by selling or buying options on a portfolio of existing
securities, investors could obtain any desired pattern of returns "investors could
span the return space to any degree of fineness desired" [Banz and Miller, 1978, 658].
Banz and Miller present their estimates of state prices in a format similar to stan-
dard interest tables. Like other interest tables, the estimates of state prices can in
principle be used by any firm in any industry or activity (subject to some important
cautions and qualifications). Thus the reciprocals (minus one) of the state prices com-
puted are analogous to single-period interest rates. Banz and Miller handle the multi-
period case by assuming stability in the relations between initial states and outcome
states. Thus the two-period matrix is simply the square of the one-period matrix, the
three-period matrix is the product of the one-period matrix and the two-period matrix,
the four-period matrix is the product of the one-period matrix and the three-period
matrix, and so on. In equation form,
,
Vfl = V (B5.1)
The perpetuity matrix is the one-period matrix times the inverse of the identity matrix
minus the one-period matrix, or V(/ —
Their computations for a V matrix of real discount factors for three states of the
world is provided in Table B5.1. In the definition of states in Table B5.1 the state
=3P
1Ç6 STATE-PREFERENCE THEORY
Implied Annual
Real Riskless
State = < 3 Row Sum Rate
8 years ( V 8 )-
= .P=76 .<8<8 .=64< .9646 .004P
< .P<07 .<84P .=6P< .9704 .0038
3 .P<39 .<863 .=66< .9764 .0030
9 years ( V 9 )-
= .P=PP .<8=7 .=636 .9608 .004P
< .P=86 .<834 .=64P .966P .0038
3 .P<=8 .<8P= .=6PP .97<4 .003=
=0 years (V'°):
= .P=34 .<806 .=6<9 .9P69 .0044
< .P=6P .<8<3 .=639 .96<7 .0038
3 .P=97 .<840 .=649 .9686 .003<
Perpetuity
[V(/ - V) -1 ]-
= =3<.P0 7<.4= 4<.0P <46.9,6 .0040
< =33.3= 7<.8P 4<.30 <48.46 .0040
3 =34.=4 73.<9 4<.PP <49.98 .0040
îoundaries are defined over returns on the market. The conditional means are ex-
pected market returns under alternative states. The elements of any matrix V may
îe interpreted îy use of the first group of data. The .P<P= represents the outcome
for state = when the initial state was also state =. The .<93P represents the outcome
for state < when state = was the initial state. The .=73P represents the outcome for
state 3 when state = was the initial state. By analogy the .=6=< represents an outcome
for state 3 when state 3 was the initial state.
For equal proîaîilities the current price of a claim to funds in a state in which
funds are scarce (a depression) will îe higher than in a îoom state when returns are
more favoraîle. Thus a project with most of its payoffs contingent on a îoom will
have a lower value per dollar of expected returns than a project whose payoffs are
relatively more favoraîle during a depression.
The vector of gross present values of the project, Gk will îe
P _
Gk = Vté„(t), (BP.<)
t=i
where X k(t) is a vector whose elements represent the expected real cash flows of
project k in year t, assuming the economy is in state i. The summation is performed
over time periods ending in p, the last period during which the project's cash flows
are nonzero in any state.
1Ç8 STATE-PREFERENCE THEORY
An example of how the "interest factors" in Taîle BP.= can îe applied is îased
on the illustration presented îy Banz and Miller. The Omega Corporation is analyzing
an investment project whose cash flow pattern in constant =980 dollars (ignoring
taxes and tax shields) is presented in Taîle BP.<.
The Banz-Miller example is sophisticated in illustrating that îoth the level and
risk of the cash flows vary with the degree of competition. In our example we modify
their estimates of the cumulative proîaîilities of competitive entry, using 0 in the
year of introduction, .3 one year later, .6 two years later, and = three years later. The
risk-adjusted gross present value vector for the project was set forth in Eq. (BP.<).
For the assumptions of our example, the particular gross present value vector is G k =
Vg„, + V 2(0.7é,, + 0.3X0 + V 3(0.4X,, + 0.6X,) + V 4 [V(/ - V)']é e . We use the
values of V and its powers as presented in Taîle BP.= to oîtain the results shown
in Taîle BP.3.
If the initial investment were $=<36 in every state of the economy, the project
would not have a positive net present value if the economy were depressed or normal.
However, the net present value would îe positive if the economy were strong. If
initial investment costs had cyclical îehavior, particularly if supply îottlenecks devel-
oped during a îoom, investment outlays might vary so strongly with states of the
world that net present values could be positive for a depressed economy and negative
for a booming economy.
The Banz-Miller use of state prices in capital budgeting is a promising applica-
tion of the state-preference model. Further applications will provide additional tests
of the feasibility of their approach. More work in comparing the results under
alternative approaches will provide increased understanding of the advantages and
possible limitations of the use of state prices as discount factors in capital budgeting
analysis.
Appendix C to Chapter 5:
Application of the SPM in
Capital Structure Decisions
The state-preference model (SPM) can also be used to analyze financing decisions
if estimates of state prices as provided in the Banz-Miller paper are available. While
capital structure decisions are discussed in Chapters 13 and 14, the following ex-
planation is self-contained and provides an overview of some key aspects of the sub-
ject. In addition to the formal framework an illustrative numerical example will be
developed.
The symbols that will be employed are listed and explained in Table C5.1. In
the capital structure analysis three broad categories of alternative outcomes may
take place. The outcomes are defined by the amount of net operating income achieved
in relation to the amount of the debt obligations incurred. These three alternative
140
APPENDIX C TO CHAPTER 5 141
outcomes are specified by column (1) in Table C5.2. Under outcome 1 the state-
dependent net operating income, X s, is equal to or greater than the amount of the
debt obligations, B. Debt holders will receive the total amount of promised payments,
B. Equity holders will receive the after-tax income remaining after the debt payments.
The amounts received by debt holders and equity holders are listed in columns (2)
and (3) in Table C5.2.
Under outcome 2, the state-dependent net operating income is less than B but
positive. Debt holders will receive the net operating income less bankruptcy costs
that may be incurred. Equity holders will receive nothing. If the state-dependent in-
come is negative, neither debt holders nor equity holders will receive anything.
In Table C5.3 the applicable state prices are multiplied times what the debt
holders receive under alternative outcomes to determine the value of the debt holders'
receipts in each state. Similarly, what the equity holders receive is multiplied by the
state prices to give the value of equity. The value of the firm in each state is the
sum of the market value of debt and the market value of equity.
This formal framework is next utilized in an illustrative numerical example. The
basic data to be utilized are set forth in Table C5.4. In Table C5.5 the data are used
to calculate the value of the firm under alternative debt levels. On the left-hand column
labeled Debt Levels we begin by specifying the amount of debt and the resulting rela-
tionships between X,, the EBIT under alternative states, and the promised debt pay-
ment. The applicable formulas for calculating the state-contingent value of the firm
are specified for each level of debt. Utilizing the illustrative data from Table C5.4, we
can then obtain the value of the firm's state payoff for alternative debt levels.
When the firm is unlevered (B = 0) its value is equal to the after-tax EBIT times
the state price for each state, summed over all the states. The resulting value is $408.
When debt is $200 the value of the firm under state 1 is zero. For states 2, 3,
and 4 both the debt and equity have value as shown in Table C5.5. When the value
of debt is $500, again the firm has no value under state 1. Under states 2, 3, and 4
both debt and equity have value. The total value of the firm sums to $550.
Similarly, for a debt level of $800 the value of the firm is calculated as $386.
For a value of debt of $2000 the value of the firm is $350.
The debt level that results in the highest indicated value of the firm as shown
by Table C5.5 is $500. For alternative levels of debt the value of the firm is lower.
AllOUHI 30Na<i33H==(=-HIVIS
Taîle CP.3 Formulas for the Value of the Firm under
Alternative Outcomes
Amount of Value of Value of
2s in Debt Debt Holders] Equity Equity Holders]
Relation Holders Re3eipts Holders Re3eipts Value of the
to B Re3eive in State s Re3eive in State s Firm in State s
This result follows from the numerical values chosen for the illustration. The example
illustrates the tax advantage of debt. Also, substantial bankruptcy costs are postu-
lated. In addition, the bankruptcy costs are assumed to have a substantial fixed ele-
ment as well as to rise with the amount of resources that may become available under
each of the alternative states. Further aspects of capital structure decisions will be
developed in Chapters 13 and 14.
6
The results of a portfolio analysis are no more than the logical
consequence of its information concerning securities.
Harry Markowitz, Portfolio Selection, Yale University Press,
New Haven, 1959, 205
Objects of Choice:
Mean-Variance
Uncertainty
145
146 OBJECTS OF CHOICE: MEAN-VARIANCE UNCERTAINTY
can better allocate the funds provided by shareholders and creditors among scarce
capital resources.
This chapter begins with simple measures of risk and return for a single asset
and then complicates the discussion by moving to risk and return for a portfolio
of many risky assets. Decision rules are then developed to show how individuals
choose optimal portfolios that maximize their expected utility of wealth, first in a
world without riskless borrowing and lending, then with such opportunities.
Suppose the task at hand is to describe the relevant features of a common stock to
a friend who is an investor. What are the really crucial facts that you should com-
municate? You could start off by giving the company's name, say Bayside Cigar Co.
Then you would discuss the financial ratios of the company: its earnings per share,
its inventory turnover, its financial leverage, its interest coverage, and so on. All
these data are merely one way of getting at what is crucial—How will your friend's
wealth position be affected if he or she invests in Bayside Cigar? Consequently, it is
wise to talk about measures of the effect on relative wealth at the end of an invest-
ment period. The terminology used is end-of-period wealth.
The link between end-of-period wealth and an initial dollar investment is the
rate of return. For the time being, we will not specify what calendar interval we
are working with except to say that it is a single time period. If the initial invest-
ment is $1 and the final wealth is $W, then the investor's rate of return, R, is
R= W — I
(6.1)
As you see, this is the same expression as that used for the present or future value
formulas for one time period.
If end-of-period wealth is known with certainty, then so is the present value of the
investment and the rate of return. However, this is seldom the case in the real world.
Even short-term default-free bonds such as U.S. Treasury Bills are not completely
risk free (although later on we shall use them as a close approximation to a risk-
free security).
For risky assets often the best that can be done is to assign probabilities to
various possible outcomes. Suppose the current price (P0) of Bayside Cigar is $25
per share and you tell your friend that after a careful analysis the best estimate of
the price per share at the end of the time period is given in Table 6.1.
MEASURING RISK AND RETURN FOR A SINGLE ASSET 147
1. Measures of Location
It is desirable to develop some statistics that can summarize a wide set of pos-
sible outcomes. The most commonly used statistics are measures of location and
dispersion. Measures of location are intended to describe the most likely outcome
in a set of events. The most often used measure of location is the mean or ex-
pectation. It is defined as (the tilde, -, is used to designate randomness)
41) = pi X i , (6.2)
i =1
The expected or mean return is the expected price less the current price divided by
the current price.
- E(P) - Po 26.50 - 25 = .
E(R) = 06 or 6%. (6.3)
Po 25
Implicitly, we have used two probability properties of the expected value operator
to obtain Eq. (6.3).
Property 1. The expected value of a random variable )7 plus a constant a is equal
to the expected value of the random variable plus the constant:
E(X- + a) = E(X) + a (6.4)
Property 1 can be proved by using the definition of expected value. Since the random
variable is (X + a), we take its expectation by substituting (X i + a) for X i in Eq. (6.2):
E(aé) = E p i(aé i ).
i=i
Then by expanding the sum, we have
E(al) = p i aé + p 2 aé 2 + • • • + pn aé„.
E(a.k)= a pi é i .
i=i
And finally, recognizing that E p é, = E(é ), we have
i
-
Figure 6.= is a histogram for the set of numbers. Note that most of the probabil-
ity (in fact 53.3%) lies between —1 and 20. However, the mean, which assigns equal
weight to all observations in this case, gives 28.13 as the best measure of location.
The median is 13. Clearly, in this case, where we have a distribution of outcomes
that is skewed to the right, the median is a better measure of location than the mean
is. Later on, when we actually look at empirical distributions of security returns, the
choice of mean return as the best measure of central tendency will depend a great
deal on whether or not the actual distributions are skewed.
The last measure of location to be considered is the mode. It is defined as the
most frequent outcome. In the above example it is the number 13, which occurs five
times, or the interval between 6 and 13, which contains 23.3% of the probability. The
mode is not often used as a measure of location for empirical distributions of security
returns because security returns are real numbers (i.e., they can take on any decimal
value) and consequently do not repeat themselves frequently.
2. Measures of Dispersion
So far we have looked at statistical measures that can be used to best describe
the most likely outcome when out friend invests in Bayside Cigar. An investment of
$1000 can be expected to bring an end-of-period wealth of $1060. (Why?) But the
question still remains What risk is being taken? There are five measures of dispersion
we could use: the range, the semiinterquartile range, the variance, the semivariance,
and the mean absolute deviation. Each of these has slightly different implications for
risk.
The range is the simplest statistic and is defined as the difference between the
highest and lowest outcomes. For an investment in one share of Bayside Cigar (see
I
I --, Skewed to right
===== = . ii al
—P0 P0 =00 =P0 <00
150 OBJECTS OF CHOICE: MEAN-VARIANCE UNCERTAINTY
Table 6.1) the worst outcome is $20 and the best outcome is $40. Therefore the range
is $20. However, the range is a very poor descriptive statistic because it becomes
larger as sample size increases. Whenever the underlying probability distribution of
investment outcomes is being estimated e.g., by looking at observations of past
performance the estimated range will increase as more observations are included
in the sample.
The semiinterquartile range is the difference between the observation of the 75th
percentile, X.75, and the 25th percentile, X .„, divided by 2:
X.25
Semiinterquartile range = (6.6)
2
Unlike the range, this statistic does not increase with sample size and is therefore
much more reliable.' For the set of 30 numbers that we were using earlier (in Table
6.2) the semiinterquartile range is
27.0 — 4.5
Semiinterquartile range = 2 = 11.25.
1
The interested reader is referred to Cramer [1961, 367-370] for proof that sample quantiles converge
to consistent estimates as sample sizes increase.
MEASURING RISK AND RETURN FOR A SINGLE ASSET 151
VAR(P)
P,;
And of course this is exactly the formula used earlier to compute the variance of
return from our knowledge of the variance of prices.
The next section of this chapter uses the properties of the mean and variance
that we have developed here in order to discuss the mean and variance of a portfolio
of assets. At this point we could summarize the investment opportunity offered by
Bayside Cigar by saying that the expected price is $26.50 with a standard deviation
of $5.39. Or else we could say that the expected return on this investment is 6% with
a standard deviation of 21.54%. However, before moving on, it will be useful to
contrast the variance as a measure of risk with the semivariance and the mean absolute
deviation.
One problem with the variance is that it gives equal weight to possibilities above
as well as below the average. However, suppose that risk-averse investors are more
concerned with downside risk. The semivariance is a statistic that relates to just that
risk. It is defined as the expectation of the mean differences below the mean, squared.
Mathematically, the definition is as follows. Let
x. f X, — E(X) if X, < E(X)
10 if X, > E(1)'
then
SEMIVAR = E[(X,)2 ]. (6.10)
If the semivariance is used as a measure of risk, an increase in the probability of
events above the mean will change risk only slightly because the only effect would
be to increase the mean slightly. For example, the semivariance of return for Bayside
Cigar is
2
SEMIVAR = .1( — .20 — .06)2 + .2( — .10 — .06)2 + .4(0 — .06)
= 1.332%.
MEASURING PORTFOLIO RISK AND RETURN 153
But if the probability of a 60% return (in Table 6.1) were to increase to .2 while the
probability of a 20% return fell to .1, the impact on semivariance would be slight.
The new expected return would be 10% and the semivariance would increase to 2.1% .
Given the same change in probabilities, the variance would increase from 4.64% to
7.2%.
Both the variance and the semivariance are sensitive to observations distant from
the mean because the mean differences are squared. Squaring gives them greater
weight. A statistic that avoids this difficulty is the mean absolute deviation (MAD),
which is defined as the expectation of the absolute value of the differences from the
mean:
MAD = E[1X 1 — E(I)1]. (6.11)
For the Bayside Cigar example, the mean absolute deviation is
MAD = .11( — .2 — .06)1 + .21( — .1 — .06)1 + .41(0 — .06)1
+.21(.2 — .06)1 + .11(.6 — .06)1
= 16.4%.
Although for the most part we shall measure risk and return by using the variance
(or standard deviation) and the mean return, it is useful to keep in mind that there
are other statistics that, in some situations, may be more appropriate. An under-
standing of these statistics helps to put the mean and variance into proper perspective.
From this point we assume that investors measure the expected utility of choices
among risky assets by looking at the mean and variance provided by combinations
of those assets. For a financial manager, the operating risk of the firm may be mea-
sured by estimating the mean and variance of returns provided by the portfolio of
assets which the firm holds: its inventory, cash, accounts receivable, marketable se-
curities, and physical plant. For a portfolio manager, the risk and return are the
mean and variance of the weighted average of the assets in his or her portfolio.
Therefore, in order to understand how to manage risk it becomes necessary to explore
the risk and return provided by combinations of risky assets.
f(R)
1.994
1.760
1.210
Figure 6.2
A normal distribution [E(R) = .1, a = .2].
follow this distribution. An example is given in Fig. 6.2. The frequency of a return
is measured along the vertical axis, and the returns are measured along the horizontal
axis. The normal distribution is perfectly symmetric, and 50% of the probability lies
above the mean, 15.9% above a point one standard deviation above the mean, and
2.3% above a point two standard deviations above the mean. Because of its symmetry
the variance and semivariance are equivalent measures of risk for the normal distribu-
tion. Furthermore, if you know the mean and standard deviation (or semivariance)
of a normal distribution, you know the likelihood of every point in the distribution.
This would not be true if the distribution were not symmetric. If it were skewed to
the right, e.g., one would also need to know a measure of skewness in addition to
the mean and standard deviation, and the variance and semivariance would not be
equivalent.
The equation for the frequency of returns, R, which are normally distributed, is
given below:<
1 - ( =/ < )[(R — E(R))/°]2
f(R) = e . (6.12)
a OR
If we know the mean, E(R), and the standard deviation, a, of the distribution, then
we can plot the frequency of any return. For example, if E(R) = 10% and a = 20%,
then the frequency of a 13% rate of return is
f(.13) =
1
e
- (112)[(.13 -.=0)/.<]<,
.207c
f(.13) = 1.972.
<
Of course IL is "pi," the ratio of the circumference and the diameter of a circle, and e is the base of
natural logarithms.
MEASURING PORTFOLIO RISK AND RETURN 155
Probability Yi
.2 117. — 370
.2 9 15
.2 25 2
.2 7 20
.2 —2 6
To simplify matters we have assumed that each pair of returns [2 i , Yi] has equal
probability (Prob = .2). The expected value of 2 is 10%, and the expected value of
Y is 8%. The variances are computed below.
3
From this point on, the tilde, -, will be used in the text to designate a random variable only when it is
needed to prevent ambiguity.
MEASURING PORTFOLIO RISK AND RETURN =P7
E(R p a(R p
/ 9.0-
,e •
10.0 - / 8.0
4/ 7.0
9.0 — / 6.0
/ 5.0
8.01 I I I I ■a
0 25 50 75 100 0 25 50 75 100
e
(a) (b)
Figure 6.3
The portfolio return mean and standard deviation as a function of the percentage
invested in risky asset 2.
For each 1% decline in "a" there will be a 2% decline in expected return. The relation-
ship between the portfolio standard deviation, o-(R p , and the weight in asset 2 is non-
linear and reaches a minimum. Later on, we will show how to determine the portfolio
weights that will minimize portfolio risk.
Figure 6.4 plots the portfolio mean and standard deviation on a single graph.
Each point represents a different weight in asset 2. The solid portion of the line
represents all combinations where the weights in asset 2 range between 0% and 100%.
9.0
8.5
8.0
2 3 4 5 7 8 **„.,
MEASURING PORTFOLIO RISK AND RETURN 159
If we can sell an asset short without restriction, then the dashed portions of the lines
in Fig. 6.4 are feasible. Selling short means that you sell an asset that you do not
already have. For example, it might be possible to sell short 50% of your wealth in
asset X (even though you do not already own shares of asset X) and buy 150% of
asset Y. If you sell X short, you should receive the proceeds, which you can then
use to buy an extra 50% of Y. This is not possible in the real world because investors
do not receive funds equal to the value of securities which they sell short. Nevertheless,
for expositional purposes, we assume that short sales are not constrained. The mean
and variance of the above short position are calculated below:
E(R p) = — .5E(X) + 1.5E(Y)
= — .5(.10) + 1.5(.08) = 7.0%.
VAR(R p) = ( — . 5) 2 VAR(X) + (1.5)2 VAR(Y) + 2( — .5)(1.5)COV(X, Y)
= .25(0076) + (2.25)(.00708) + 2( — .75)( — .0024) = .02143.
-
o (R p) = OTAR(R p) = 14.64%.
Now that we have developed ways of measuring the risk (variance) and return
(mean) for a portfolio of assets, there are several interesting questions to explore.
For example, what happens if the covariance between X and Y is zero—i.e., what
happens if the two securities are independent? On the other hand, what happens if
they are perfectly correlated? How do we find the combination of X and Y that gives
minimum variance?
Obviously, if returns on the two assets are independent, i.e., if the covariance between
them is zero, then the correlation between them will be zero. Such a situation is
shown in Fig. 6.5, which is a scatter diagram of two independent returns.
The opposite situation occurs when the returns are perfectly correlated, as in
Fig. 6.6, in which the returns all fall on a straight line. Perfect correlation will result
Y Figure 6.5
Independent returns.
160 OBJECTS OF CHOICE: MEAN-VARIANCE UNCERTAINTY
Y Figure 6.6
Perfectly correlated returns.
in a correlation coefficient equal to 1. To see that this is true we can use the fact that
Y is a linear function of 2. In other words, if we are given the value of 2, we know
for sure what the corresponding value of Y will be. This is expressed as a linear
function:
Y = a + b2.
We also use the definition of the correlation coefficient. First, we derive the expected
value and standard deviation of Y by using Properties 1 through 4:
E(Y) = a + bE(2 ,
VAR(Y) = b2VAR(X),
oy = box .
The definition of the correlation coefficient is
COV(X, Y E[(2 — E(2 (Y — E(Y ] •
rxy =
ox oy uxay
By substituting the mean and variance of Y, we obtain
E[(2 — E(2 (a + b2 — a — bE(2 ]
r xy =
o- „b a- x
E[(2 — E(2 b(2 — E(2 ] box 1
= =
box b
box
Therefore the correlation coefficient equals +1 if the returns are perfectly correlated,
and it equals —1 if the returns are perfectly inversely correlated.' It is left as an
exercise for the student to prove that the latter is true. The correlation coefficient
ranges between + 1 and —1:
—1 < rxy < 1. (6.18)
For the example we have been working with, the correlation between 2 and Y is
COV(X, Y) —.0024
rxy =
axcry (.0872)(.0841) = —.33.
By rearranging the definition of the correlation coefficient [Eq. (6.17)] we get
another definition of covariance whereby it is seen to be equal to the correlation
The linear relationship îetween Y and 2 for perfect inverse correlation is Y = a — b2.
MEASURING PORFOLIO RISK AND RETURN 161
da
2 2
a(o- „ + o- - 2rxya- x ) - - -<
y + ]xyoxo ÿ o = 0.
Solving for the optimal percentage to invest in é in order to obtain the minimum
variance portfolio, we get
<
- r xy6 x6 y
a *ÿ (6.21)
+ 63<
2] xyo- xc ] y
Continuing with the example used throughout this section, we see that the minimum
variance portfolio is the one where
.00708 — (—.33)(.0872)(.0841)
a* = = .487.
.0076 + .00708 — 2( —.33)(.0872)(.0841)
The portfolio return and variance for the minimum variance portfolio are
E(R p ) = aE(é) + (1 ÿ a)E(ê)
= .487(.10) + (.513)(.08) = 8.974%.
VAR(R P) = a 2 VAR(X) + (1 — a)2 VAR(Y) + 2(a)(1 — a)] xyo- xo- y
= (.487)2(.0076) + (.513)2(.00708) + 2(.487)(.513)(—.33)(.0872)(.0841)
= .0018025 + .0018632 — .0012092 = .0024565.
= 4.956%.
P
The student who wishes to review the mathematics of maximization is referred to Appendix D.
162 OBJECTS OF CHOICE: MEAN-VARIANCE UNCERTAINTY
The minimum variance portfolio is represented by the intersection of the dashed lines
in Fig. 6.4.
Probability X
.2 —1.408% —3%
.2 17.258 15
.2 3.777 2
.2 22.443 20
.2 7.925 6
or, = 1.0376y = 8.72%,
o-y = 8.41%,
COV(X, Y) = r„,o-xo-y = .007334.
Figure 6.7
Risk-return trade-offs for two assets.
MEASURING PORTFOLIO RISK AND RETURN 163
variance argument plane, we take a look at the definitions of mean and variance
when r xy = }:
E(R p = aE(2 + (} — a E(Y ,
(6.22)
VAR(R p) = a 2 o-!. + (} — a 2o + 2a(} — a o- x o-y .
Note that the variance can be factored:
VAR(R p) = [au x + (} — a a,] 2 ;
therefore the standard deviation is
u(R, = au x + (} — a o-y . (6.23)
The easiest way to prove that the curve between A and B is a straight line is to
show that its slope does not change as a, the proportion of the portfolio invested
in 2, changes. The slope of the line will be the derivative of expected value with re-
spect to the weight in 2 divided by the derivative of standard deviation with respect
to the weight in 2:
dE(R p dE(R p mda
Slope =
do-(R p du(R p mda
The derivative of the expected portfolio return with respect to a change in a is
dE(R p =
E(X) E( Y ),
da
and the derivative of the standard deviation with respect to a is
do-(R p
= a. ÿ ay.
da
By substituting this weight into the equations for mean and standard deviation we
can demonstrate that the portfolio has zero variance:
E(R p = .49095(.10) + (1 — .49095)(.08) = 8.982%,
o-(R p = .49095(.0872) — (1 — .49095)(0841) = 0%.
Because standard deviations cannot be negative, the two roots of Eq. (6.25) need to
be interpreted as follows. So long as the percentage invested in 2 is greater than
or equal to 49.095% (which is aû, the minimum variance portfolio), the standard
deviation of the portfolio is
o-y
u(R p = ao- x ÿ (1 — a)o- y if a (6.25a)
ax + ay
On the other hand, if less than 49.095% of the portfolio is invested in 2, the standard
deviation is
6Y
o-(R p = (} — a o- y — ao- x if a< (6.25b)
+ ay
MEASURING PORTFOLIO RISK AND RETURN 165
We can use these results to show that the line segments AC and CB are linear. The
proof proceeds in precisely the same way that we were able to show that AB is linear
if r xy = 1. For the positively sloped line segment, AC, using Eq. (6.24), we have
dE(R p =
E(2 — E(Y ,
da
set is bounded by the triangle ACB. Intuitively, any set of portfolio combinations
formed by two risky assets that are less than perfectly correlated must lie inside the
triangle ACB and will be convex.
The concepts developed in this section can now be used to discuss the way we,
as investors, are able to select portfolios that maximize our expected utility. The port-
folio mean return and variance are the measures of return and risk. We choose the
percentages of our wealth that we want to invest in each security in order to obtain
the required risk and return. We have shown the choices that are possible if two risky
assets are perfectly correlated, perfectly inversely correlated, and where their corre-
lation lies between —1 and + 1. We have also seen how we can find the minimum vari-
ance portfolio. Later in this chapter these results will be extended from the two-asset
case to portfolios of many assets, and we will discuss an example wherein a corporate
treasurer may use portfolio theory to reduce the risk (variability) of shareholders'
wealth.
The assumption of no risk-free asset is the same as saying that there are no borrowing
or lending opportunities. In other words, this section shows how a single individual
(Robinson Crusoe) will choose his optimal portfolio of risky assets in a world where
there is no opportunity for exchange. As we shall see, the following discussion is
analogous to the Robinson Crusoe economy described in Chapter 1 except that the
objects of choice are risk and return rather than consumption and investment. The
results are also similar. Robinson Crusoe's optimal portfolio will be that where his
subjective marginal rate of substitution between risk and return is exactly equal to
the objectively determined marginal rate of transformation (along his mean-variance
opportunity set) between risk and return. At his optimal portfolio the equality be-
tween MRS and MRT determines his subjective price of risk. Later on, in section
E.5, we shall introduce a marketplace with opportunities to exchange by borrowing
and lending unlimited amounts of money at the risk-free rate. This exchange economy
setting will show the existence of a single market-determined price of risk. All in-
dividuals and their agents (firms, for example) will use the market price of risk for
optimal decisions in the face of uncertainty.
In the chapter on utility theory we saw that indifference curves for the risk-averse
investor were convex in the mean-variance plane. Figure 6.8 shows a family of in-
difference curves as well as the convex set of portfolio choices offered by various
percentages of investment in two risky assets. If we know our risk-return trade-off
and also know the possibilities offered by combinations of risky assets, we will maxi-
mize our expected utility at point C in Fig. 6.8. This is where our indifference curve
is tangent to the opportunity set offered by combinations of X and Y. Each indif-
ference curve maps out all combinations of risk and return that provide us with the
same total utility. Moving from right to left in Fig. 6.8, we know (from Chapter 4)
OPTIMAL PORTFOLIO CHOICE: THE EFFICIENT SET WITH TWO RISKY ASSETS 167
Rip)
•
vy Iv / III 11
..........•A
/., ( I
1 ,..-
,„e- H f
....00 i /
B
G
.I .
/•D
• •/°E4ftft■../
F
► a(;?p)
Figure 6.8
Optimal portfolio choice for a risk-averse investor
and two risky assets.
that indifference curve I has less total utility than indifference curve II, and so on.
We could put all our money in one asset and receive the risk and return at point F,
which is on indifference curve I, but of course we can do better at points B and E,
and best at point C (on indifference curve III). Points G, H, and I have higher total
utility than point C, but they are not feasible because the opportunity set offered by
the risky assets does not extend that far.
An important feature of the optimal portfolio that we choose in order to maximize
our utility is that the marginal rate of substitution between our preference for risk
and return represented by our indifference curves must equal the marginal rate of
transformation offered by the minimum variance opportunity set. The slope of the
dashed line drawn tangent to our indifference curve at point C is our marginal rate
of substitution between risk and return. This line is also tangent to the opportunity
set at point C. Hence its slope also represents the trade-off between risk and return
offered by the opportunity set. Therefore the way we can find a utility-maximizing
portfolio is to try different portfolios along the opportunity set until we find the one
where the marginal rate of transformation between risk and return along the mini-
mum variance opportunity set just equals the marginal rate of substitution along our
indifference curve:
MRSE4TJ = MRTE4;?.
The fact that this point is unique is guaranteed by the convexity of our indifference
curve and the convexity of the upper half of the minimum variance opportunity set.
Let us take a look at Fig. 6.9. Suppose we find ourselves endowed with a portfolio
that has the mean-variance opportunities at point A. By changing the percentage of
our wealth in each of the risky assets, we can reach any point along the minimum
variance opportunity set. At point A the marginal rate of transformation between
return and risk along the minimum variance opportunity set is equal to the slope of
the line DAF. The low slope indicates that we will get rid of a lot of risk in exchange
for giving up only a little return. On the other hand, the slope of our indifference
curve, U 1, the slope of the line CAB at point A, indicates our subjective trade-off
168 OBJECTS OF CHOICE: MEAN-VARIANCE UNCERTAINTY
Figure 6.9
The utility maximizing choice equates
the marginal rates of substitution
and transformation.
between return and risk, i.e., our marginal rate of substitution. At point A, where we
already have a relatively high level of risk, we are willing to give up a lot of return in
order to get rid of a little risk. If we can move along the opportunity set toward
point E without incurring any cost, we will clearly do so because the opportunity
set at point A allows us to trade off return and risk at a more favorable rate than
we require (according to our indifference curve). We will continue to move along the
opportunity set until we reach point E. At this point we attain the highest possible
expected utility on indifference curve U 2 . Furthermore, the marginal rate of trans-
formation between return and risk along the opportunity (the slope of line HEG) set
is exactly equal to the marginal rate of substitution along the indifference curve (also,
the slope of tangent line HEG). Thus we have shown that a necessary condition for
expected utility maximization is that the marginal rate of substitution must equal
the marginal rate of transformation. This also implies that at the optimum portfolio
choice, we have a linear trade-off between return, E(Rp), and risk, o-(R p). 6
Even though different investors may have the same assessment of the return and
risk offered by risky assets, they may hold different portfolios. Later we shall discover
that when a riskless asset is introduced into the opportunity set, investors will hold
identical combinations of risky assets even though they have different attitudes toward
risk. However, in the current framework for analysis, we assume that investors have
homogeneous beliefs about the opportunity set, that no risk-free asset exists, and
that investors have different indifference curves, which reflect their differing attitudes
toward risk.' Figure 6.10 shows three different indifference curves and the investment
opportunity set. Investor III is more risk averse than investor II, who in turn is more
risk averse than investor I. (Why is this true?) Consequently, they each will choose
to invest a different percentage of their portfolio in the risky assets that make up the
opportunity set.
6
For an excellent mathematical development of this fact, see Fama and Miller [1972, Chapter 6].
Homogeneous beliefs mean simply that everyone has exactly the same information so that they all
perceive exactly the same opportunity set.
OPTIMAL 131 E: THE EFFICIENT SET WITH TWO RISKY ASSETS 169
Figure 6.10
Choices by investors with different indifference curves.
Note that rational investors will never choose a portfolio below the minimum
variance point. They can always attain higher expected utility along the positively
sloped portion of the opportunity set represented by the line segment EDCBA. This
concept leads to the definition of the efficient set.
Efficient set. The efficient set is the set of mean-variance choices from the in-
vestment opportunity set where for a given variance (or standard deviation) no
other investment opportunity offers a higher mean return.
The notion of an efficient set considerably narrows the number of portfolios from
which an investor might choose. In Fig. 6.10, e.g., the portfolios at points B and F
offer the same standard deviation, but B is on the efficient set because it offers a
higher return for the same risk. Hence no rational investor would ever choose point
F over point B, and we can ignore point F. Point B is stochastically dominant over
point F. It is interesting to note, however, that investors will hold positions in an
asset or portfolio at point F. No one will hold F by itself; rather it will be held as
part of portfolios that lie along the efficient set.
Interesting special cases of the efficient set for two risky assets occur when their
returns are perfectly correlated. Figure 6.11 shows perfect correlation, and Fig. 6.12
Figure 6.11
e Two perfectly correlated assets.
170 OBJECTS OF CHOICE: MEAN-VARIANCE UNCEL.
Figure 6.12
Two assets with perfect inverse
correlation.
a(R )
shows perfect inverse correlation. In both cases, the efficient set is linear. In Fig. 6.11
it is line X Y and in Fig. 6.12 it is line XZ.
In general the locus of feasible mean-variance opportunities can be found by
solving either of the following two mathematical programming problems. The first
defines the minimum variance opportunity set, and the second defines the efficient
set.
Programming Problem 1:
MIN 62(R p) subject to E(R p) = K. (6.26a)
Programming Problem 2:
MAX E(R p) subject to 6 2(R p) = K. (6.26b)
Note that the minimum variance opportunity set is found by finding all combinations
that give the lowest risk for a given return. The efficient set is the locus of highest
returns for a given risk. If we write out the first problem at greater length,
subject to
If one of the two assets, R f , has zero variance, then the mean and variance of the
portfolio become
E(R p = aE(2 + (} — a R f ,
VAR(R p) = a 2VAR(2 .
We have assumed that the risk-free asset is R f . Its variance and its covariance with
the risky asset are zero; therefore the second and third terms in the general expression
for variance, Eq. (6.20), are equal to zero, and portfolio variance is simply the variance
of the risky asset.
Knowledge of the mean and variance of a portfolio with one risk-free and one
risky asset allows us to plot the opportunity set in Fig. 6.13. It is linear. Proof of
linearity proceeds in the same way as earlier proofs. All we need to do is show that
the slope is independent of a, the percentage of the portfolio invested in the risky asset.
The change in expected return with respect to the percentage invested in 2 is
dE(R p =
E(2 Rf,
da
E(K p V
a>1
2 0 0
E(2 ----------------------- vas
Borrowing
-
0 -(a <}
Lending
Y
R
,,a <0
Lending and
short selling
a(2
Figure 6.13
Opportunity set with one risky and one risk-free asset.
172 OBJECTS OF CHOICE: MEAN-VARIANCE UNCERTAINTY
Consequently, the line VX Y must be linear because its slope does not change with the
percentage invested in X.
It is usually assumed that the rate of return on the risk-free asset is equal to the
borrowing and lending rate in the economy. In the real world, of course, the borrowing
and lending rates are not equal. One possible cause is transactions costs, i.e., frictions
in the marketplace. However, like physicists who assume that friction does not exist
in order to derive the laws of mechanics, economists assume that asset markets are
frictionless in order to develop price theory. A frictionless world to an economist is
one where all assets are infinitely divisible and where there are no transactions costs.
In such a world, the borrowing rate would equal the lending rate for risk-free assets.
We shall use this assumption to develop a theory for the equilibrium price of risk, then
provide empirical evidence that indicates that in spite of several unrealistic assump-
tions, the theory describes reality surprisingly well.
Given the assumption that the borrowing rate equals the lending rate, YXV is
a straight line. To reach portfolios along the line segment XV it is necessary to borrow
in order to invest more than 100% of the portfolio in the risky asset. Note that borrow-
ing is analogous to selling short the risk-free assets. Therefore along the line segment
XV the percentage invested in X is greater than 1; in other words, a > 1. The mean
and standard deviation of the portfolio along this portion of the line are
E(R p) = aE(X) + (1 — a)R f ,
o-(R p) = aa x .
the other hand, when we decide to invest more than 100% of our portfolio in the
risk-free asset, we must sell short the risky asset. Assuming no restrictions on short
sales (another assumption necessary for frictionless markets), the mean and variance
of the portfolio for a < 0 are
E(R p) = (1 — a)R f + aE(X),
6(R p) =
Note that because negative standard deviations are impossible, the absolute value of
a is used to measure the standard deviation of the portfolio when the risky asset is sold
short. The line segment YZ represents portfolio mean and variance in this case.
What about the efficient set for portfolios composed of one risk-free and one
risky asset? Clearly no risk-averse investor would prefer line segment YZ in Fig. 6.13
because he or she can always do better along the positively sloped line segment YXV.
Therefore the efficient set is composed of long positions in the risky asset combined
with borrowing or lending. Why then do we observe short sales in the real world?
The answer, of course, is that not all people hold the same probability beliefs about
the distributions of returns provided by risky assets. Some investors may believe
that the expected return on asset X is negative, in which case they would sell short.
In equilibrium, however, we know that so long as investors are risk averse, the final
price of the risky asset X must be adjusted so that its expected rate of return is greater
than the risk-free rate. In equilibrium, assets of higher risk must have higher expected
return.
OPTIMAL PORTFOLIO CHOICE: MANY ASSETS 173
Until now it has been convenient to discuss portfolios of only two assets. By gen-
eralizing the argument to many assets, we can discuss several important properties
such as portfolio diversification, the separation principle, and the Capital Market
Line. We can also provide a realistic example of how a corporate chief financial officer
who views his or her firm as a portfolio can control the risk exposure of shareholders.
We begin by developing the mean and variance for portfolios of many assets.
E(R p = E[w}R } + w 2 R 2 + w é R é ],
As was the case for a portfolio with two assets, the expected portfolio return is simply
a weighted average of the expected return on individual assets. This can be rewritten
as
The definition of portfolio variance for three assets is the expectation of the sum of the
mean differences squared:
The portfolio variance is a weighted sum of variance and covariance terms. It can be
rewritten as
3 3
VAR(Rp (6.28)
) = i=1 j=1
where w, and w; are the percentages invested in each asset, and is the covariance
of asset i with asset j. You will recall from the discussion of covariance earlier in the
text that the variance is really a special case of covariance. The variance is the co-
variance of an asset with itself. For example, when i = 2 and j = 2, then we have
w2w 2622 , which is the same thing as v2VAR(R 2). Therefore Eq. (6.28) contains three
variance and six covariance terms.
If we replace the three assets with N, Eqs. (6.27) and (6.28) can be used as general
representations of the mean and variance of a portfolio of N assets. We can also write
Eqs. (6.27) and (6.28) in matrix form,' which for two assets looks like this:
The expected portfolio return is the (1XN) row vector of expected returns, [E(R} ),
E(R2 )] = R', postmultiplied by the (N X1) column vector of weights held in each asset,
[w1w 2] = W. The variance is the (NXN) variance-covariance matrix, t, premultiplied
and postmultiplied by the vector of weights, W. To see that the matrix definition of
the variance is identical to Eq. (6.28), first postmultiply the variance-covariance matrix
by the column vector of weights to get
+ w26121
VAR(R p) = [w1w 2 ]
wian + W2 622
Postmultiplying the second vector times the first, we have
VAR(Rp) = wf,o- „ + w1w2a12 + w2w1a21 + W2 622*
ample, we might construct portfolio A by holding 50% of our wealth in asset X and
the remaining 50% in asset Y. Next, let W 2 be the (NX1) column vector of weights
used to construct portfolio B. For example, we might have 25% in X and 75% in
Y. If t is the (NXN) variance-covariance matrix, then the covariance between the
two portfolios is defined as
COV(RA, RB) = W1 W2
[511 5
121
Wlb
= [w la W2a] (6.29)
5 6
21 22 W2b
and postmultiplying the row vector, Wri , by the column vector above, we obtain
To show that this matrix result is indeed the same as the traditional definition, we
begin with the usual covariance equation
RB = W lbRx W 2bRy'
Substituting these expressions as well as their expected values into the covariance
definition, we have
Note that this is exactly the same as the expanded covariance expression obtained
from the matrix definition, Eq. (6.29).
The matrix definitions of portfolio mean, variance, and covariance are particularly
powerful and useful because the size of the vectors and matrices can easily be ex-
panded to handle any number of assets. The matrix form also lends itself naturally
to computer programs.
176 OBJECTS OF CHOICE: MEAN-VARIANCE UNCERTAINTY
WSTL , RSTL = the weight and return on the firm's short-term liabilities,
WLTD, 1
-LTD = the weight and return on the firm's long-term debt.
Suppose that the firm's chief financial officer (CFO) is concerned that a tough
anti-inflationary policy will cause a decline in inflation and in interest rates. The
result would be a decline in the market value of the property held by the company
(its major asset) and an increase in the market value of the firm's long-term debt.
The net effect would be a dramatic rise in the firm's debt-to-assets ratio and a drop
in the market value of equity. To hedge against this risk the CFO has decided to
buy T-bond futures contracts (an investment that we shall assume, for the sake of
convenience, requires no cash outlay).9 A long position in T-bond futures is expected
to be a hedge for two reasons. First, when inflation and interest rates fall the market
value of the T-bond futures will rise to offset an expected decline in the market
value of the firm's assets. Second, the T-bond position will hedge against an increase
in the market value of the firm's debt liabilities.
Given that T-bond futures will provide a hedge, the CFO must determine the
optimal number, N, of T-bond futures contracts to buy. Too few will not provide
Table 6.5 Market Value Balance Sheet for NR Inc.
9
Chapter 9 provides a complete description of futures contracts.
OPTIMAL PORTFOLIO CHOICE: MANY ASSETS 177
an adequate hedge. Too many will overhedge. If P„ is the current price of a $100,000
face-value T-bond contract, V is the market value of the firm, and R TB is the return
on T-bond futures, then the return on equity, given the hedge position, becomes
NPTB
N TB — (6.31)
-14S = WST AT A + WLT AkT A WSTLRSTL WLTDRLTD
TD + RTB.
V
Z= the variance-covariance matrix of all assets and liabilities in the firm's hedge
portfolio.
To find the optimal hedge portfolio, we can take the derivative of Eq. (6.32) with
respect to N and set the result equal to zero:
d VAR(Rs) 2P TB x-, p2
= 1 ' TB ,.,—1 NT =
L win rscrif Yrs 7- v2 " TB"
+ 0.
dN V i '
Note that ri,„is the correlation between the ith portfolio asset (or liability) and
T-bond futures contracts, i is the standard deviation of the ith asset, and aTB is
the standard deviation of the T-bond futures contract return. Solving for N, the
optimal number of futures contracts, we have
Vri TB - i
N = E ' • (6.33)
P TBC TB
Equation (6.33) shows that the hedging asset, T-bond futures, affects shareholders'
risk through the correlation betwen T-bond futures returns and the returns on the
firm's other assets and liabilities, rti , F . 10 The actual values for the parameters in-
volved are
rsTA,TB = 0, rSTL,TB = 0
,
r
LTA,TB = — .6725, rLTD,TB = . 7834,
a LTD = .0908, r LT A = .0482,
TB = .0766, VTB = $70,250.
The correlations confirm the CFO's suspicion that T-bond futures will be a good
hedge against changes in the market value of long-term assets and long-term debt.
10
The expression ri,TaudaTs is equal to the slope in a linear regression of the returns on the ith asset
or liability on T-bond futures.
178 OBJECTS OF CHOICE: MEAN-VARIANCE UNCERTAINTY
T-bond futures returns are negatively correlated with assets and positively correlated
with debt. Substituting the above values into Eq. (6.33) we have
The numbers reveal that the CFO needs to buy 1205 contracts as a hedge against
changes in the value of assets and 1269 contracts to hedge against changes in the
value of long-term debt.
There are, of course, many ways a CFO might choose to reduce the risk of share-
holders. This example shows that a total of 2474 T-bond contracts (i.e., $247.4 million
worth) provides the optimal hedge. This example illustrates one way of conceptuali-
zing the firm as a portfolio of risky assets. Whether or not the CFO should hedge
in the first place and the set of other possible hedge techniques are topics that will
be discussed later in the text.
11
In general, if N securities are analyzed the variance-covariance matrix will have -RN — 1)N different
covariance elements and N variance elements.
12
For proof, see Merton [1972].
OPTIMAL PORTFOLIO CHOICE: MANY ASSETS 179
Figure 6.14
The investment opportunity set with
many risky assets.
a(R)
Figure 6.15
The efficient set with one risk-free and many risky
assets.
180 OBJECTS OF CHOICE: MEAN-VARIANCE UNCERTAINTY
Figure 6.16
Dominance of the linear efficient set.
asset. Investor I, who is the least risk averse, will borrow (at the risk-free rate) to
invest more than 100% of his or her portfolio in the risky portfolio M. However, no
investor will choose to invest in any other risky portfolio except portfolio M. For
example, all three could attain the minimum variance portfolio at point B, but none
will choose this alternative because all can do better with some combination of the
risk-free asset and portfolio M. Next we shall see that portfolio M can be identified
as the market portfolio of all risky assets. All risky assets are held as part of risky
portfolio M.
5. A Description of Equilibrium
In section C we analyzed a Robinson Crusoe economy where there was no
opportunity for exchange. Robinson Crusoe's optimal portfolio resulted from maxi-
mizing his expected utility, given his risk preferences, subject to the feasible set of
mean-variance trade-offs offered by a combination of two risky assets. In section D
we saw how a linear efficient set could be formed from one risky and one risk-free
asset. So far in section E, we have first described the opportunity set with risky
assets, then with one risk-free asset and many risky assets.
The introduction of a risk-free asset may be thought of as creating an exchange
or market economy where there are many individuals. Each of them may borrow
or lend unlimited amounts at the risk-free rate. With the introduction of an ex-
change economy, we shall be able to describe a fundamental principle called two-
fund separation. Analogous to Fisher separation found in Chapter 1 (where everyone
used the market-determined time value of money to determine consumption/invest-
ment decisions), two-fund separation implies that there is a single market-determined
equilibrium price of risk (which is used in portfolio decisions). This concept will
prove extremely useful later on in the text when, e.g., we want to conceptualize the
opportunity cost of capital for projects of different risk.
OPTIMAL PORTFOLIO CHOICE: MANY ASSETS 181
If—in addition to the earlier assumption of equality between the borrowing and
lending rate that follows, given frictionless capital markets—we add the assumption
that all investors have homogeneous (i.e., identical) beliefs about the expected distri-
butions of returns offered by all assets, then all investors will perceive the same efficient
set. Therefore they will all try to hold some combination of the risk-free asset, R f ,
and portfolio M.
For the market to be in equilibrium, we require a set of market-clearing prices. All
assets must be held. In other words the existence of an equilibrium requires that all
prices be adjusted so that the excess demand for any asset will be zero. This market-
clearing condition implies that an equilibrium is not attained until the single-tangency
portfolio, M, which all investors (with homogeneous expectations) try to combine with
risk-free borrowing or lending, is a portfolio in which all assets are held according to
their market value weights. If V is the market value of the ith asset, then the per-
centage of wealth held in each asset is equal to the ratio of the market value of the
asset to the market value of all assets. Mathematically,
vvi = N '
E
where w i is the weight of the ith asset in the market portfolio and E V is the total
market value of all assets. Market equilibrium is not reached until the tangency port-
folio, M, is the market portfolio. Also, the value of the risk-free rate must be such
that aggregate borrowing and lending are equal.
The fact that the portfolios of all risk-averse investors will consist of different
combinations of only two portfolios is an extremely powerful result. It has come to
be known as the two-fund separation principle. Its definition is given below.
Two fund separation. Each investor will have a utility-maximizing portfolio that
is a combination of the risk-free asset and a portfolio (or fund) of risky assets that
is determined by the line drawn from the risk-free rate of return tangent to the
investor's efficient set of risky assets.
The straight line in Fig. 6.16 will be the efficient set for all investors. This line has come
to be known as the capital market line. It represents a linear relationship between
portfolio risk and return.
Capital market line (CML). If investors have homogeneous beliefs, then they all
have the same linear efficient set called the capital market line.
Figure 6.17 is a graph of the capital market line. The intercept is the risk-free rate,
R f , and its slope is [E(Rm) — R f ] / 0-(Rm). Therefore the equation for the capital market
line is
E(Rm) — R f cy(R p).
E(R p) Rf u(Rm) (6.34)
It provides a simple linear relationship between the risk and return for efficient port-
folios of assets. Having established the principle of two-fund separation and defined
182 OBJECTS OF CHOICE: MEAN-VARIANCE UNCERTAINTY
a(:R p )
a(R m )
Figure 6.17
The capital market line.
the capital market line, we find it useful to describe the importance of capital market
equilibrium from an individual's point of view.
We wish to compare expected utility-maximizing choices in a world without capi-
tal markets (as depicted in Fig. 6.9) with those in a world with capital markets (seen
in Fig. 6.18). As in Chapter 1, a capital market is nothing more than the opportunity
to borrow and lend at the risk-free rate. Chapter 1 emphasized that in a world with
certainty everyone was better off, given that capital markets existed and Fisher separa-
tion obtained. Now we have extended this result to a world with mean-variance un-
certainty. Everyone is better off with capital markets where two-fund separation
obtains.
Figure 6.18 shows us endowed with the mean-variance combination at point A.
With a capital market, we always have two choices available. We can move along the
mean-variance opportunity set (by changing our portfolio of risky assets), or we can
Figure 6.18
Individual expected utility maximization
in a world with capital markets.
► a(kp )
OPTIMAL PORTFOLIO CHOICE: MANY ASSETS 183
move along the Capital Market Line by borrowing or lending. Initially, at point A, the
trade-off between return and risk is more favorable along the opportunity set than
along the market line. Therefore we will move along the opportunity set toward point
B where the marginal rate of transformation between return and risk on the oppor-
tunity set is equal to our subjective marginal rate of substitution along our indiffer-
ence curve. In the absence of capital markets, we would have maximized our expected
utility at point B. This would be the Robinson Crusoe solution, and our level of
utility would have increased from U 1 to U2. However, if we have the opportunity
to move along the Capital Market Line, we can be even better off. By moving to
point M, then borrowing to reach point C, we can increase our expected utility from
U 2 to U 3 . Therefore we have three important results. First, nearly everyone is better
off in a world with capital markets (and no one is worse off). Second, two-fund
separation obtains. This means that everyone, regardless of the shape of his or her
indifference curve, will decide to hold various combinations of two funds: the market
portfolio and the risk-free asset. And third, in equilibrium, the marginal rate of sub-
stitution (MRS) between return and risk is the same for all individuals, regardless of
their subjective attitudes toward risk.
If the marginal rate of substitution between risk and return is the same for every
individual in equilibrium, then the slope of the Capital Market Line is the equilibrium
price of risk (EPR):
E(R — R f
EPR = MRSE(R,)
a(Rp) — (6.35)
o-m( R ,i •
The implication is that decision makers, e.g., managers of firms, can use the market-
determined equilibrium price of risk to evaluate investment projects regardless of the
tastes of shareholders. Every shareholder will unanimously agree on the price of risk
even though different shareholders have different degrees of risk aversion. Also the
marginal rates of substitution between risk and return for the ith and jth individuals
in equilibrium will equal the marginal rate of transformation, and both will be equal
to the equilibrium price of risk.
E(R„, Rf
MRS, = MRS; = = MRT.
3r( R m)
Next, and in Chapter 7, we turn our attention to the problem of measuring risk.
We have already established that variance is an adequate measure of risk for port-
folios of assets; however, it is not particularly useful when we wish to evaluate the
risk of individual assets that do not lie on the efficient set. Nor is it possible, given
our current development of the theory, to compare a single risky asset with a well-
diversified portfolio. Therefore it is necessary to distinguish between portfolio risk and
the contribution of a single asset to the riskiness of a well-diversified portfolio (such
as the market portfolio).
To set the framework for the difference between portfolio risk and individual asset
risk, we observe the average return and variance of return calculated for a single asset,
184 OBJECTS OF CHOICE: MEAN-VARIANCE UNCERTAINTY
Bayside Cigar, and for a 100-stock portfolio of randomly selected common stocks.
Return on the assets was defined as total return, i.e., dividends, Dive , plus capital
gains, 1é, — Pt _ 1 . The equation for a monthly return is given below:
P e — 1é 1 _ 1 + Dive
Re =
Pt-1
Data were collected for the 306 months between January 1945 and June 1970.13 The
average monthly return on Bayside Cigar was .45%, which is approximately 5.4% per
year, and the standard deviation was 7.26%. By comparison, the 100-stock portfolio
had an average return of .91% per month or 10.9% per year. Its standard deviation
was 4.45%. Normally, one would expect the standard deviation of a well-diversified
portfolio to be lower than for a single asset, and the empirical results bear this out.
But we also know that riskier assets should have higher returns. Therefore if standard
deviation is the appropriate measure of risk for an individual asset, then Bayside Cigar
should have a higher return. But it does not! We shall see in the next chapter that the
resolution to this apparent paradox is that although the standard deviation is appro-
priate for measuring the risk of an efficient portfolio, it is not the appropriate measure
of risk for individual assets or for comparing the riskiness of portfolios with the
riskiness of assets.
13
" See Modigliani and Pogue [1974].
PORTFOLIO DIVERSIFICATION AND INDIVIDUAL ASSET RISK 185
This expression can be separated into variance and covariance terms as follows:
1 x-,
N 1 N N
VAR(R ) =
2 4 a
(6.36)
P N i- 1 " + N2
j
Suppose that the largest individual asset variance is L. Then the first term, the variance
term, is always less than or equal to
1 N
N „ E L = LN
i 1 N2
L
=N,
and as the number of assets in the portfolio becomes large, this term approaches
zero:
L
lim — = 0.
N-~ co N
On the other hand, the covariance terms do not vanish. Let cii; be the average co-
variance. Then in the right-hand term in Eq. (6.36), there are (N 2 — N) covariance
terms, all equal to "du; therefore the right-hand term can be rewritten as
1 NÆ N
N 2 (N2 N)cJ.1i = N 2 N2 a
and the limit as N approaches infinity is
N2 _ N_
lim (,N ,o-, J•
Ni
N2 J (6.37)
Consequently, as we form portfolios that have large numbers of assets and that are
better diversified, the covariance terms become relatively more important.
Fama [1976] has illustrated this result empirically.' His results are shown in
Fig. 6.19. He randomly selected 50 New York Stock Exchange (NYSE) listed securi-
ties and calculated their standard deviations using monthly data from July 1963 to
June 1968. Then a single security was selected randomly. Its standard deviation of
return was around 11%. Next, this security was combined with another (also ran-
domly selected) to form an equally weighted portfolio of two securities. The standard
deviation fell to around 7.2%. Step by step more securities were randomly added to
the portfolio until all 50 securities were included. Almost all the diversification was
obtained after the first 10 to 15 securities were randomly selected. In addition the
portfolio standard deviation quickly approached a limit that is roughly equal to the
average covariance of all securities. One of the practical implications is that most of
the benefits of diversification (given a random portfolio selection strategy) can be
achieved with fewer than 15 stocks.
Ibbotson and Sinquefield [1986], using monthly data between 1926 and 1985,
have computed the geometric mean and standard deviation of returns for value-
weighted portfolios of broad classes of assets, e.g., common stocks (the Standard and
14
See Fama [1976, 253-254].
186 OBJECTS OF CHOICE: MEAN-VARIANCE UNCERTAINTY
a(Ri,)
.111
.097
.082
.067
.053 •
•• •
.038 • .
I I 1 I I I I I I 1
5 10 15 20 25 30 35 40 45 50N
Figure 6.19
The standard deviation of portfolio return as a function of
the number of securities in the portfolio. (From Fama, E. F.,
Foundations of Finance, 1976, reprinted with permission of
the author.)
Poor's 500), small stocks (the smallest quintile on the New York Stock Exchange),
and bonds. Table 6.6 summarizes their results. Note that the standard deviation of
the portfolio of common stocks was 21.2% per year, whereas the standard devia-
tion of small stocks (not a randomly selected portfolio) was about 70% larger at a
level of 36.0% per year. These data show the limits of diversification for different
classes of securities.
Still another way of looking at the risk of a single asset is to evaluate its contri-
bution to total portfolio risk. This can be done by taking the partial derivative of
the expression for portfolio variance [Eq. (6.28)] with respect to w1, the percentage
invested in the ith risky asset:
a VAR(R )
awi
P
= 2W i0 + 2 wp0 . (6.38)
Geometric Standard
Portfolio Mean Deviation
Value weighted common stocks 9.8% 21.2%
Smallest NYSE quintile stocks 12.6% 36.0%
High-grade long-term corporate bonds 4.8% 8.3%
Long-term U.S. government bonds 4.1% 8.2%
U.S. Treasury Bills 3.4% 3.4%
Consumer Price Index 3.1% 4.9%
Source: Ibbotson and Sinquefield [1986, 25].
PORTFOLIO DIVERSIFICATION AND INDIVIDUAL ASSET RISK 187
E p )
Figure 6.20
The capital market line.
188 OBJECTS OF CHOICE: MEAN-VARIANCE UNCERTAINTY
Since the three securities clearly have different distributions of return, they also have
different prices. Even though we know the mean and variance of return offered by
assets I, J, and K, we cannot be sure what prices they will have in equilibrium. The
reason is that we do not know their joint distribution with all other assets. The
missing information is the variance-covariance matrix of all assets. In Chapter 7 we
shall see that variance is not the appropriate measure of risk for an individual asset.
This was the point of Eqs. (6.37) and (6.38). As the number of assets in a portfolio
increases, the risk that an asset contributes to a portfolio reduces to be exclusively
the covariance risk. Therefore the portion of an asset's risk that is uncorrelated with
the economy can be avoided at no cost. No rational investor will pay a premium to
avoid diversifiable risk. On the other hand, because covariance risk cannot be diver-
sified away, investors will pay a premium to escape it. Therefore covariance is the
relevant measure of risk for an asset because it measures the contribution of an
individual asset to the variance of a well-diversified portfolio.
SUMMARY
This chapter has combined our knowledge of the theory of investor choice (utility
theory) with the objects of investor choice (the portfolio opportunity set) to show
how risk-averse investors wishing to maximize expected utility will choose their opti-
mal portfolios. We began with simple measures of risk and return (and simple proba-
bility theory) and ended with portfolio theory. Finally, we saw that when a risk-free
asset exists, the opportunity set can be reduced to the simple, linear Capital Market
Line. Given frictionless capital markets and homogeneous investor expectations, all
individuals will choose to hold some combination of the risk-free asset and the market
portfolio.
PROBLEM SET
6.1 Historically, the empirical distributions of stock prices on the NYSE have been skewed
right. Why?
6.2 Given the following relationship between x and y,
y = a + bx, b < 0,
prove that x and y are perfectly negatively correlated.
6.3 Given the following hypothetical end-of-period prices for shares of the Drill-On
Corporation,
Probability .01 .05 .07 .02 .10 .30 .20 .15 .05 .05
End-of-period
price per share 0 35.00 38.57 40.00 42.00 50.00 55.00 57.00 60.00 69.00
E(X)7) = ?
6.5 Using the definition of portfolio variance, prove that a perfectly hedged stock portfolio
that is 100 shares long and 100 shares short is perfectly risk free.
6.6 Given the variance-covariance matrix
24 —10 25
—10 75 32
25 32 12_
.2 I 18 0
.2 II 5 —3
.2 III 12 15
.2 IV 4 12
.2 V 6 1
190 OBJECTS OF CHOICE: MEAN-VARIANCE UNCERTAINTY
a) Calculate the mean and variance of each of these variables, and the covariance between
them.
b) Suppose x and y represent the returns from two assets. Calculate the mean and vari-
ance for the following portfolios:
6.10 (Our thanks to Nils Hakansson, University of California, Berkeley, for providing this
problem.) Two securities have the following joint distribution of returns, r = and r2:
P{r 1 = — 1.0 and r < = .15} = .1,
P{r 1 = .5 and r, = .15} = .8,
P{r = = .5 and r < = 1.65} = .1.
a) Compute the means, variances, and covariance of returns for the two securities.
b) Plot the feasible mean-standard deviation [E(R), a] combinations, assuming that the two
securities are the only investment vehicles available.
c) Which portfolios belong to the mean-variance efficient set?
d) Show that security 2 is mean-variance dominated by security 1, yet enters all efficient
portfolios but one. How do you explain this?
PROBLEM SET 191
e) Suppose that the possibility of lending, but not borrowing, at 5% (without risk) is added
to the previous opportunities. Draw the new set of [E(R), o-] combinations. Which port-
folios are now efficient?
6.11 Suppose a risk-averse investor can choose a portfolio from among N assets with inde-
pendently distributed returns, all of which have identical means [E(R i) = E(Ri)] and identical
variances (al = ol). What will be the composition of his optimal portfolio?
6.12 Given decreasing marginal utility, it is possible to prove that in a mean-variance frame-
work no individual will hold 100% of his or her wealth in the risk-free asset. Why? [Hint: The
answer requires an understanding of the shape of investors' indifference curves as well as the
capital market line.]
6.13 Given that assets X and Y are perfectly correlated such that Y = 6 + .2X and the prob-
ability distribution for X is
Probability X
.1 30%
.2 20
.4 15
.2 10
.1 — 50
What is the percentage of your wealth to put into asset X to achieve zero variance? Graph
the opportunity set and the zero variance point.
6.14 A market value balance sheet for the Carr Commercial Bank is given below in millions
of dollars:
Assets Liabilities
1000 1000
The standard deviations and correlations between returns on asset and liability categories
(excepting equity) are as follows:
a(D) -= .03.
a) What is the standard deviation of the equity holder 's position?
192 OBJECTS OF CHOICE: MEAN-VARIANCE UNCERTAINTY
Suppose the bank decides to hedge by taking a position in T-bond futures contracts. You
are given the following information:
V„ = $90,000 for a $100,000 face-value T-bond contract,
r
.9,
0
TB,US = rTB,L = .5, rTB,STL , rTB,D • 3.
Should the bank take a long or short position in T-bond futures? How many futures con-
tracts should they buy/sell? How much is the standard deviation of equity reduced?
REFERENCES
Cramer, H., Mathematical Methods in Statistics. Princeton University Press, Princeton, N.J.,
1961.
Fama, E. F., Foundations of Finance. Basic Books, New York, 1976.
, and M. Miller, The Theory of Finance. Holt, Rinehart and Winston, New York, 1972.
Ibbotson, R. G., and R. Sinquefield, Stocks, Bonds, Bills and Inflation: 1986 Yearbook. Ibbotson
Associates, Inc., Chicago, Illinois, 1986.
Markowitz, H. M., Portfolio Selection: Efficient Diversification of Investment (Cowles Founda-
tion Monograph 16). Yale University Press, New Haven, 1959.
Merton, R., "An Analytic Derivation of the Efficient Set," Journal of Financial and Quanti-
tative Analysis, September 1972, 1851-1872.
Modigliani, F., and G. Pogue, "An Introduction to Risk and Return: Concepts and Evidence,"
Financial Analysts Journal, March—April and May—June, 1974, 68-80, 69-85.
Sharpe, W., "A Simplified Model for Portfolio Analysis," Management Science, January 1963,
277-293.
, Portfolio Theory and Capital Markets. McGraw-Hill, New York, 1970.
Tobin, J., "Liquidity Preference as Behavior towards Risk," Review of Economic Studies, Feb-
ruary 1958, 65-86.
7
Lucy: "I've just come up with the perfect theory. It's my theory that
Beethoven would have written even better music if he had been
married."
Schroeder: "What's so perfect about that theory?"
Lucy: "It can't be proved one way or the other!"
Charles Schulz, Peanuts, 1976
Market Equilibrium:
CAPM and APT
A. INTRODUCTION
The greater portion of this chapter is devoted to extending the concept of market
equilibrium in order to determine the market price for risk and the appropriate mea-
sure of risk for a single asset. One economic model used to solve this problem was
developed almost simultaneously by Sharpe [1963, 1964], and Treynor [1961], while
Mossin [1966], Lintner [1965b, 1969], and Black [1972] developed it further. The first
model we will discuss is usually referred to as the capital asset pricing model (CAPM).
It will show that the equilibrium rates of return on all risky assets are a function of
their covariance with the market portfolio. A second important equilibrium pricing
model, called the arbitrage pricing theory (APT), was developed by Ross [1976]. It is
similar to the CAPM in that it is also an equilibrium asset pricing model. The return
on any risky asset is seen to be a linear combination of various common factors that
affect asset returns. It is more general than the CAPM because it allows numerous
factors to explain the equilibrium return on a risky asset. However, it is in the same
spirit as the CAPM. In fact, the CAPM can be shown to be a special case of the APT.
The organization of the chapter is to first develop the CAPM and its extensions,
then to summarize the empirical evidence relating to its validity. Thereafter the APT
will be developed and empirical evidence on it will be described. We begin with a list
of the assumptions that were first used to derive the CAPM.
193
194 MARKET EQUILIBRIUM: CAPM AND APT
Proof of the CAPM requires that in equilibrium the market portfolio must be an
efficient portfolio. It must lie on the upper half of the minimum variance opportunity
set graphed in Fig. 7.1. One way to establish its efficiency is to argue that so long as
investors have homogeneous expectations, they will all perceive the same minimum
variance opportunity set.' Even without a risk-free asset, they will all select efficient
portfolios regardless of their individual risk tolerances. As shown in Fig. 7.1, individ-
For a more rigorous proof of the efficiency of the market portfolio see Fama [1976, Chapter 8].
DERIVATION OF THE CAPM 195
ual I chooses efficient portfolio B, whereas individual II, who is less risk averse,
chooses efficient portfolio C. Given that all individuals hold positive proportions of
their wealth in efficient portfolios, then the market portfolio must be efficient because
(1) the market is simply the sum of all individual holdings and (2) all individual
holdings are efficient.
Thus, in theory, when all individuals have homogeneous expectations, the market
portfolio must be efficient. Without homogeneous expectations the market portfolio
is not necessarily efficient and the equilibrium model of capital markets that is derived
in the next section does not necessarily hold. Thus the efficiency of the market port-
folio and the capital asset pricing model are inseparable, joint hypotheses. It is not
possible to test the validity of one without the other. We shall return to this important
point when we discuss Roll's critique later in the chapter.
Figure 7.2 shows the expected return and standard deviation of the market portfolio,
M, the risk-free asset, R f , and a risky asset, I. The straight line connecting the risk-
free asset and the market portfolio is the capital market line. For example, see Fig.
6.17, in Chapter 6. We know that if a market equilibrium is to exist, the prices of all
assets must adjust until all are held by investors. There can be no excess demand.
In other words, prices must be established so that the supply of all assets equals the
demand for holding them. Consequently, in equilibrium the market portfolio will con-
sist of all marketable assets held in proportion to their value weights. The equilibrium
proportion of each asset in the market portfolio must be
Market value of individual asset
wi (7.1)
Market value of all assets
A portfolio consisting of a% invested in risky asset I and (1 — a)% in the market
portfolio will have the following mean and standard deviation:
E(ic) = aE(ii i) + (1 — a)E(IZ,„), (7.2)
1/2,
2 2 „
p ) = [a + (1 — a) 6„, + 2a(1 — a)o- (7.3)
196 MARKET EQUILIBRIUM: CAPM AND APT
Figure 7.2
The opportunity set provided by combinations of
risky asset I and the market portfolio, M.
where
We shall see shortly that the market portfolio already contains asset I held according
to its market value weight. In fact the definition of the market portfolio is that it
consists of all assets held according to their market value weights. The opportunity
set provided by various combinations of the risky asset and the market portfolio is
the line /M/' in Fig. 7.2. The change in the mean and standard deviation with respect
to the percentage of the portfolio, a, invested in asset I is determined as follows:
0E(RP)
E(R A) — E(1L), (7.4)
as
8
6(k) _ + (1 - a)2 o-,2„ + 2a(1 — a)o-i„,] -112
2L
ea
X [2ao-i2 — 26,2n. + 2a6n2 + 26im — 4aaim]. (7.5)
Sharpe's and Treynor's insight, which allowed them to use the above facts to deter-
mine a market equilibrium price for risk, was that in equilibrium the market portfolio
already has the value weight, wi percent, invested in the risky asset I. Therefore the
percentage a in the above equations is the excess demand for an individual risky asset.
But we know that in equilibrium the excess demand for any asset must be zero. Prices
will adjust until all assets are held by someone. Therefore if Eqs. (7.4) and (7.5) are
evaluated where excess demand, a, equals zero, then we can determine the equilibrium
price relationships at point M in Fig. 7.2. This will provide the equilibrium price of
DERIVATION OF THE CAPM 197
The final insight is to realize that the slope of the opportunity set /M/' provided by
the relationship between the risky asset and the market portfolio at point M must
also be equal to the slope of the capital market line, R/M.
As established in Chapter 6, the capital market line is also an equilibrium rela-
tionship. Given market efficiency, the tangency portfolio, M, must be the market port-
folio where all assets are held according to their market value weights. Recall that
the slope of the capital market line in Eq. (6.34) is
E(Rm — Rf
am
where am is the standard deviation of the market portfolio. Equating this with the
slope of the opportunity set at point M, we have
E(R m — R f = E(}4, — E(R m
ÿ 2
(au. a. mam
This relationship can be arranged to solve for E(k i as follows:
711n
E(14_,) = Rf + [E(k m — R f ] 2 • (7.9)
m
Equation (7.9) is known as the 3apital asset pri3ing model, CAPM. It is shown graph-
ically in Fig. 7.3 where it is also called the se3urity market line. The required rate of
E(FR
Security market line
E(1 m )
Rf
gm = 1
Figure 7.3
The capital asset pricing model.
198 MARKET EQUILIBRIUM: CAPM AND APT
return on any asset, E(k i) in Eq. (7.9), is equal to the risk-free rate of return plus a
risk premium. The risk premium is the price of risk multiplied by the quantity of
risk. In the terminology of the CAPM, the price of risk is the slope of the line, the
difference between the expected rate of return on the market portfolio and the risk-
free rate of return.' The quantity of risk is often called beta,
COV(R1, R„,)
Ni — 2 =
• (7.10)
VAR(R m)
It is the covariance between returns on the risky asset, I, and market portfolio, M,
divided by the variance of the market portfolio. The risk-free asset has a beta of zero
because its covariance with the market portfolio is zero. The market portfolio has a
beta of one because the covariance of the market portfolio with itself is identical to
the variance of the market portfolio:
ni COV(R,,„ 1?„,) VAR(R m)
fi = — 1.
VAR(R m) VAR(R m)
There are several properties of the CAPM that are important. First, in equilibrium,
every asset must be priced so that its risk-adjusted required rate of return falls exactly
on the straight line in Fig. 7.3, which is called the security market line. This means,
for example, that assets such as I and J in Fig. 6.20, which do not lie on the mean-
variance efficient set, will lie exactly on the security market line in Fig. 7.3. This is true
because not all the variance of an asset's return is of concern to risk-averse investors.
As we saw in the previous chapter, investors can always diversify away all risk except
the covariance of an asset with the market portfolio. In other words, they can diversify
away all risk except the risk of the economy as a whole, which is inescapable (un-
diversifiable). Consequently, the only risk that investors will pay a premium to avoid
is covariance risk. The total risk of any individual asset can be partitioned into two
parts—systematic risk, which is a measure of how the asset covaries with the economy,
and unsystematic risk, which is independent of the economy.
Total risk = systematic risk + unsystematic risk, (7.11)
2 Note that the CAPM terminology is somewhat different from that used in Chapter 6. Earlier, the equi-
librium price of risk was seen to be the marginal rate of substitution between return and risk and was
defined as
E(R„,) — R1
Using this definition for the price of risk, the quantity of risk is
COV(R i , R,,,)
Because a,,, the standard deviation of the market, is assumed to be constant, it does not make much dif-
ference which terminology we adopt. Hereafter, risk will be [I and the price of risk will be [E(R,„) — Rf].
PROPERTIES OF THE CAPM 199
Table 7.1 Ris'' and Return for Bayside Cigar and a 100-Stock Portfolio
Annual Return Standard Deviation Beta
100-stock portfolio 10.9% 4.45% 1.11
Bayside Cigar 5.4 7.25 .71
Mathematical precision can be attached to this concept by noting that empirically the
return on any asset is a linear function of market return plus a random error term
which is independent of the market:
= a; + +
This equation contains three terms: a constant, which has no variance; a constant
times a random variable, Mc; and a second random variable, Z".i, which has zero
covariance with Rm . Using Properties 3 and 4 of random variables (given in Chapter
6), we can immediately write the variance of this relationship as
o-4 = b(7
J m
2
+ cd. (7.12)
The variance is total risk; it can be partitioned into systematic risk, bo-ni2, and un-
systematic risk, (72. It turns out that b1 in the simple linear relationship between
individual asset return and market return is exactly the same as f3 in the CAPM. 3
If systematic risk is the only type of risk that investors will pay to avoid, and
if the required rate of return for every asset in equilibrium must fall on the security
market line, we should be able to go back to the example of Bayside Cigar Company
and resolve the paradox introduced in Chapter 6. Table 7.1 summarizes the empirical
findings. We know that if investors are risk averse, there should be a positive trade-
off between risk and return. When we tried to use the standard deviation as a measure
of risk for an individual asset, Bayside Cigar, in comparison with a well-diversified
portfolio, we were forced to make the inappropriate observation that the asset with
higher risk has a lower return. The difficulty was that we were using the wrong
measure of risk. One cannot compare the variance of return on a single asset with
the variance for a well-diversified portfolio. The variance of the portfolio will almost
always be smaller. The appropriate measure of risk for a single asset is beta, its
covariance with the market divided by the variance of the market. This risk is non-
diversifiable and is linearly related to the rate of return [E(R i) in Eq. (7.9)] required
in equilibrium. When we look at the appropriate measure of risk, we see that Bayside
Cigar is less risky than the 100-stock portfolio, and .we have the sensible result that
lower risk is accompanied by lower return.
Table 7.2 shows the realized rates of return and the betas of many different assets
between January 1945 and June 1970. The calculations are taken from an article by
Modigliani and Pogue [1974] that used monthly observations. In most cases the risk-
return relationships make sense. Consumer product companies such as Swift and Co.,
3
The interested reader is referred to Appendix C on linear regression for proof that the slope coefficient,
b p equals
bi = COV(R J, RJ/VAR(R,„).
200 MARKET EQUILIBRIUM: CAPM AND APT
Table 7.2 Rates of Return and Betas for Selected Companies, 194-19 I U
Average Standard
Annual Return Deviation Beta
From F. Modigliani and G. Pogue, "An Introduction to Risk and Return," reprinted
from Financial Analysts Journal, March-April 1974, 71.
Bayside Cigar, and American Snuff are all less risky than the market portfolio (repre-
sented here by the NYSE index). On the other hand, steel, electronics, and auto-
mobiles are riskier. Figure 7.4 plots the empirical relationship between risk (measured
by beta) and return for the companies listed in Table 7.2. The linearity of the rela-
tionship appears to be reasonable, and the trade-off between risk and return is positive.
A more thorough discussion of empirical tests of the CAPM will be given later in this
chapter.
A second important property of the CAPM is that the measure of risk for in-
dividual assets is linearly additive when the assets are combined into portfolios. For
example, if we put a% of our wealth into asset X, with systematic risk of /3„, and b%
of our wealth into asset Y, with systematic risk of fl y , then the beta of the resulting
portfolio, )6,, is simply the weighted average of the betas of the individual securities:
Figure 7.4
An empirical security market line.
18
16
14
12
10
8
PROPERTIES OF THE CAPM 201
Proof of this follows from the definition of covariance and the properties of the mean
and variance. The definition of the portfolio beta is
The fact that portfolio betas are linearly weighted combinations of individual asset
betas is an extremely useful tool. All that is needed to measure the systematic risk
of portfolios is the betas of the individual assets. It is not necessary to solve a qua-
dratic programming problem [see Eqs. (6.26a) and (6.26b)] to find the efficient set.
It is worth reiterating the relationship between individual asset risk and portfolio
risk. The correct definition of an individual asset 's risk is its contribution to portfolio
risk. Referring to Eq. (6.28), we see that the variance of returns for a portfolio of
assets is
NN
V AR(ic) = 0-2 (k) = W OF i1, (6.28)
i=1 j=1
which can be rewritten as'
To see that E w1(E = E wiCOV(Ri , R,), consider a simple three-asset example. Rewriting the
left-hand side, we have
Wi(E Wj 6i = Wl(Wlal 1 + W2612 + W3613)
6
22
32
6
23
6
33_
W2
W3
as the risk of security i in portfolio p. However, at the margin, the change in the con-
tribution of asset i to portfolio risk is simply
Therefore covariance risk is the appropriate definition of risk since it measures the
change in portfolio risk as we change the weighting of an individual asset in the
portfolio.
Although the use of systematic risk and undiversifiable risk have arisen in the
literature as synonyms for covariance risk, they are somewhat misleading. They rely
on the existence of costless diversification opportunities and on the existence of a
large market portfolio. The definition of covariance risk given above does not. It
continues to be relevant, even when the market portfolio under consideration has
few assets.
because it provides a quantifiable measure of risk for individual assets, the CAPM is
an extremely useful tool for valuing risky assets. For the time being, let us assume
that we are dealing with a single time period. This assumption was built into the deri-
vation of the CAPM. We want to value an asset that has a risky payoff at the end
of the period. Call this Pe . It could represent the capital gain on a common stock
or the capital gain plus a dividend. If the risky asset is a bond, it is the repayment
of the principal plus the interest on the bond. The expected return on an investment
in the risky asset is determined by the price we are willing to pay at the beginning
of the time period for the right to the risky end-of-period payoff. If P, is the price
we pay today, our risky return, R1, is
e
—
Po (7.17)
Po
The CAPM can be used to determine what the current value of the asset, P o , should
be. The CAPM is
COV(Ri, Rn,)
E(Rj) = Rf + [E(R,n) — R f]
VAR(Ri,i)
which can be rewritten as
— Rf
E(Ri) = R f /1. COV (R‘ j, R„,), where = E(Rm) (7.18)
VAR(Rm)
USE OF THE CAPM FOR VALUATION 203
Note that 2 can be described as the market price per unit risk. From Eq. (7.17) and
the properties of the mean, we can equate the expected return from Eq. (7.17) with
the expected return in Eq. (7.18):
E(Pe) — P,
= R f + 2 COV( ,
We can now interpret P o as the equilibrium price of the risky asset. Rearranging the
above expression, we get
E(:15 e)
- (7.19)
PO = 1 + R f + COV( fz j, iz )'
which is often referred to as the risk-adjusted rate of return valuation formula. The
numerator is the expected end-of-period price for the risky asset, and the denominator
can be thought of as a discount rate. If the asset has no risk, then its covariance with
the market will be zero and the appropriate one-period discount rate is (1 + R f ), i.e.,
(1 + the risk-free rate). For assets with positive systematic risk, a risk premium,
COV(J, /Z.), is added to the risk-free rate so that the discount rate is risk adjusted.
An equivalent approach to valuation is to deduct a risk premium from E(Pe) in
the numerator, then discount at (1 + R f ). The covariance between the risky asset
and the market can be rewritten as
E(Pe)
PO =
R f + 2(1/Po)COV(P e, Rni)
we can derive the certainty equivalent valuation formula:
E(Pe) — ),COV(15 e , .R,n)
Po= • (7.20)
1 + Rf
The risk-adjusted rate of return and the certainty equivalent approaches are
equivalent for one-period valuation models. It is important to realize that in both
cases value does not depend on the utility preferences of individuals. All one needs
to know in order to determine value is the expected end-of-period cash payoff, the
quantity of risk provided by the asset, the risk-free rate, and the price of risk (which
are market-determined variables). Consequently, individuals who perceive the same
distribution of payoffs for a risky asset will price it in exactly the same way regardless
of their individual utility functions. The separation of valuation from attitudes toward
risk is a consequence of two-fund separation. This was discussed in Chapter 6, section
E.5.
204 MARKET EQUILIBRIUM: CAPM AND APT
E(R)
Project K: required return
E(Tzk )
Project K: expected return
K
Figure 7.5
The cost of equity using the CAPM.
EXTENSIONS OF THE CAPM 205
the projected rate of return on project K, Rk , is higher than the cost of equity for the
firm, E(RJ). But the project also is riskier than the firm because it has greater sys-
tematic risk. If the managers of the firm were to demand that it earn the same rate as
the firm [lc, = E(RA, the project would be accepted since its projected rate of return,
Rk , is greater than the firm's cost of equity. However, this would be incorrect. The
market requires a rate of return, E(Rk ), for a project with systematic risk of flk , but
the project will earn less. Therefore since Rk < E(Rk) the project is clearly unaccept-
able. (Is project L acceptable? Why?)
Because the CAPM allows decision makers to estimate the required rate of return
for projects of different risk, it is an extremely useful concept. Although we have as-
sumed no debt or taxes in the above simple introduction, Chapter 13 will show how
the model can be extended to properly conceptualize more realistic capital budgeting
and cost of capital decisions.
Virtually every one of the assumptions under which the CAPM is derived is violated
in the real world. If so, then how good is the model? There are two parts to this
question: (1) Is it possible to extend the model to relax the unrealistic assumptions
without drastically changing it? (2) How well does the model stand up to empirical
testing? The first part is the subject of this section of the chapter. Surprisingly, the
model is fairly resilient to various extensions of it.
1. No Riskless Asset
First, how will the model change if investors cannot borrow and lend at the risk-
free rate? In other words, how is the CAPM affected if there is no risk-free asset that
has constant returns in every state of nature? This problem was solved by Black
[1972]. His argument is illustrated in Fig. 7.6. Portfolio M is identified by the
dm
Figure 7.6
The capital market line with no risk-free rate.
206 MARKET EQUILIBRIUM: CAPM AND APT
5
investors as the market portfolio that lies on the efficient set. Now, suppose that we
can identify all portfolios that are uncorrelated with the true market portfolio.' This
means that their returns have zero covariance with the market portfolio, and they
have the same systematic risk (i.e., they have zero beta). Because they have the same
systematic risk, each must have the same expected return. Portfolios A and B in
Fig. 7.6 are both uncorrelated with the market portfolio M and have the same ex-
pected return, E(R z . However, only one of them, portfolio B, lies on the opportunity
set. It is the minimum variance zero-beta portfolio and it is unique. Portfolio A also
has zero beta, but it has a higher variance and therefore does not lie on the minimum
variance opportunity set.
We can derive the slope of the line E(R Z M by forming a portfolio with aM in
the market portfolio and (1 — a M in the minimum variance zero-beta portfolio. The
mean and standard deviation of such a portfolio can be written as follows:
E(R p = aE(R m + (} — a E(R z ,
12
0-( R p = [a r, + (} — a .<} + 2a(} — a r zmo zo 0 .
2 2 -
But since the correlation, r zm , between the zero-beta portfolio and the market port-
folio is zero, the last term drops out. The slope of a line tangent to the efficient set at
point M, where 100% of the investor's wealth is invested in the market portfolio, can
be found by taking the partial derivatives of the above equations and evaluating them
where a = 1. The partial derivative of the mean portfolio return is
aE(R p
E(Rm — E(R z ,
Oa
20_2z +2a0.2]
8
a(Rp) _ 7, [ a2 um
2 + a 2
2 0}]}}2[2 aam z .
as
5 Note, however, that the extension of the CAPM that follows can be applied to any efficient portfolio, not
For portfolios with many assets, we are interested in identifying the portfolio that (a) has zero covariance
with the market portfolio and (b) has the minimum variance. The solution will be the vector of weights
that satisfies the following quadratic programming problem.
MIN a =W,t W1
-
Subject to W', VV,„ = o ,„, = 0
Wle = 1,
where
= the variance of the zero-beta portfolio,
W', = the row vector of weights in the minimum variance zero-beta portfolio
( W1 is a column vector with the same weights),
= the variance/covariance matrix for all N assets in the market,
W,, = the vector of weights in the market portfolio,
11„, = the covariance between the zero-beta portfolio and the market—which must equal zero,
e = a column vector of ones.
EXTENSIONS OF THE CAPM 207
Taking the ratio of these partials and evaluating where a = }, we obtain the slope of
the line E(R Z M in Fig. 7.6:
8E(R p m0a E(R m — E(R z
(7.22)
Oo-( R p mea am
Furthermore, since the line must pass through the point [E(R. , a(R m ], the intercept
of the tangent line must be E(R z . Consequently, the equation of the line must be
[E(R. — E(R z } 0_
E(R p = E(R z + P•
(7.23)
This is exactly the same as the capital market line [Eq. (6.34)] except that the expected
rate of return on the zero-beta portfolio, E(R z , has replaced the risk-free rate.
Given the above result, it is not hard to prove that the expected rate of return on
any risky asset, whether or not it lies on the efficient set, must be a linear combination
of the rate of return on the zero-beta portfolio and the market portfolio. To show
this, recall that in equilibrium the slope of a line tangent to a portfolio composed of
the market portfolio and any other asset at the point represented by the market port-
folio must be equal to Eq. (7.8):
OE(R p m0a E(R} — E(R m
- (7.8)
0o (R p m0a a= o (a im —
If we equate the two definitions of the slope of a line tangent to point M [i.e., if we
equate (7.8) and (7.22)], we have
E(R m — E(R z _[E(R, — E(R ni ]o-,,
aim um2
Solving for the required rate of return on asset i, we have
E(R, = (} — ,6, E(R z + ,6 }E(R m , (7.24)
where
= = COV(R i , R ni mor T2p.
Equation (7.24) shows that the expected rate of return on any asset can be written as
a linear combination of the expected rate of return of two assets the market port-
folio and the unique minimum variance zero-beta portfolio (which is chosen to be
uncorrelated with the market portfolio). Interestingly, the weight to be invested in the
market portfolio is the beta of the ith asset. If we rearrange (7.24), we see that it is
exactly equal to the CAPM [Eqs. (7.9) and (7.21)] except that the expected rate of
return on the zero-beta portfolio has replaced the rate of return on the risk-free asset:
E(R, = E(R z + [E(R m — E(R, }6,. (7.25)
The upshot of this proof is that the major results of the CAPM do not require
the existence of a pure riskless asset. Beta is still the appropriate measure of sys-
tematic risk for an asset, and the linearity of the model still obtains. The version of
the model given by Eq. (7.25) is usually called the two-fa3tor model.
208 MARKET EQUILIBRIUM: CAPM AND APT
One limitation of the two-factor model is that it relies rather heavily on the as-
sumption that there are no short sales constraints. In other words, investors are
assumed to be able to sell shares that they do not already own, then use the proceeds
to purchase other shares. Empirically, almost all asset returns have positive correla-
tions. This makes it virtually impossible to construct a zero-beta portfolio composed
of only long positions in securities. Therefore the unconstrained use of short sales is
a practical necessity to obtain zero-beta portfolios. With short positions the correla-
tion between asset returns is reversed. For example, if you have sold IBM short, you
make positive returns when the asset price falls. In general, zero-beta portfolios would
have to be composed of both long and short positions of risky assets. Ross [1977]
has shown that in a world with short sales restrictions and no riskless asset the linear
CAPM is invalid. Thus to obtain the CAPM in a linear form [Eqs. (7.9) and (7.25)]
we require either (1) a risk-free asset that can be freely short sold or (2) no constraints
on short sales.
f(R)
o "Fat-tailed"
♦ At
Return
Figure 7.7
The empirical distribution of daily stock returns.
There are various theories that explain the empirical distribution of daily returns. The interested reader
is referred to Fama [1965a] for the stable Paretian hypothesis and to Clark [1973] for the subordinated
stochastic process hypothesis.
EXTENSIONS OF THE CAPM 209
Return
tion that arises is, How can investors make choices based on mean and variance if
the actual distribution of security prices is such that a variance does not exist? Fama
[1965b] has shown that as long as the distribution is symmetric (and stable), investors
can use measures of dispersion other than the variance (e.g., the semiinterquartile
range) and the theory of portfolio choice is still valid.'
If security returns are measured over longer periods of time, their distribution is
better approximated by a lognormal distribution that has positive skewness. 9 Figure
7.8 compares a normal with a lognormal distribution. There is no limit to the positive
returns that may be realized on a successful investment, but the maximum negative
return is minus 100%. This explains why the distribution of annual returns, e.g.,
tends to be lognormal. The CAPM makes no provision for investor preference for
skewness; it is therefore an empirical question whether or not the model fits reality
well enough to permit us to ignore the fact that the empirical distribution of returns
is not normal. The empirical evidence is reviewed in the next section of this chapter.
Usually we refer to distributions that are stable under addition. This means that the addition of two
distributions (e.g., two normal distributions) will result in the same type of distribution.
9
The natural logarithm of a lognormal distribution is normally distributed. See Eq. (6.12) for the equation
of a normal distribution.
210 MARKET EQUILIBRIUM: CAPM AND APT
Mayers [1972] shows that when investors are constrained to hold nonmarketable
assets that have risky (dollar) rates of return, RH, the CAPM takes the following form:
E(Ri) = R f + .1.[K,COV(Ri, Rm) + COV(R3, RH )], (7.26)
where
E(R„,) — R f
A= -
Vnio + COV(R„„ RH)
V„, = the current market value of all marketable assets,
RH = the total dollar return on all nonmarketable assets.
In this version of the model, A may be interpreted as the market price per unit risk
where risk contains not only the market variance, ol but also the covariance between
the rate of return on marketable assets and the aggregate dollar return on non-
marketable assets. This result is obtained by first deriving an individual's demand
curves for holding marketable assets, then aggregating them to obtain Eq. (7.26),
which is the return on a marketable asset required by the market equilibrium. There
are three important implications. First, individuals will hold different portfolios of
risky assets because their human capital has differing amounts of risk. Second, the
market equilibrium price of a risky asset may still be determined independently of
the shape of the individual's indifference curves. This implies that the separation
principle still holds. There is still an objectively determined market price of risk that
is independent of individual attitudes toward risk. No variable in Eq. (7.26) is sub-
scripted for the preferences of the ith individual. Both the price of risk and the amount
of risk depend only on properties of the jth asset, the portfolio of all marketable
assets, and the portfolio of aggregated nonmarketable assets. Third, the appropriate
measure of risk is still the covariance, but we must now consider the covariance
between the jth risky asset and two portfolios, one composed of marketable and a
second of nonmarketable assets."
'° See Fama and Schwert [1977] for an empirical test of the model set forth by Mayers.
EXTENSIONS OF THE CAPM 211
If the risk-free rate is stochastic, investors are exposed to another kind of risk,
namely, the risk of unfavorable shifts in the investment opportunity set. Merton
shows that investors will hold portfolios chosen from three funds: the riskless asset,
the market portfolio, and a portfolio chosen so that its returns are perfectly negatively
correlated with the riskless asset. This model exhibits three-fund separation. The third
fund is necessary to hedge against unforeseen changes in the future risk-free rate. The
required rate of return on the jth asset is
E(R;) = r f + y i [E(R„,) — rf] + y 2 [E(R N) — r (7.28)
where
RN = the instantaneous rate of return on a portfolio that has perfect negative
correlation with the riskless asset,
1
Aim 3 Nm fi jN 13
jmfl Nm
Yi — 1
72 =
P Nm
Merton argues that the sign of y 2 will be negative for high beta assets and positive
for low beta assets. As we shall see in the next section, which discusses the empirical
tests of the CAPM, Merton's argument is consistent with the empirical evidence.
We shall leave a complete discussion on the Brennan model to Chapters 15 and 16,
which cover the theory and empirical evidence related to the corporate dividend
policy decision. For now it is sufficient to note that Brennan's model predicts that
higher rates of return will be required on assets with higher dividend yields. In other
words, investors do not like dividends because they must pay ordinary income tax
rates on dividends but only capital gains rates on stock price increases. (The 1986
tax code sets the capital gains rate equal to the ordinary income rate, but still allows
capital gains to be realized at the option of the investor—a tax-timing option.)
The CAPM is a simple linear model that is expressed in terms of expected returns
and expected risk. In its ex ante form, we have
E(RJ) = R f + [E(Rm) — R (7.30)
Although many of the aforementioned extensions of the model support this simple
linear form, others suggest that it may not be linear, that factors other than beta are
needed to explain E(R;), or that R f is not the appropriate riskless rate. Therefore
with so many alternative possibilities a great deal of energy has been devoted to the
empirical question: How well does the model fit the data?
There have been numerous empirical tests of the CAPM, so many in fact that
it would be fruitless to mention all of them. Also, the literature is interwoven with
many serious and difficult econometric problems that must be confronted in order to
provide the best empirical tests of the model." Most of the econometric subtleties
are beyond the scope of this text and are therefore ignored. However, in the opinion
of the authors, the tests of the CAPM summarized below represent the best of the
work that has been done to date.
The first step necessary to empirically test the theoretical CAPM is to transform
it from expectations or ex ante form (expectations cannot be measured) into a form
that uses observed data. This can be done by assuming that the rate of return on
any asset is a fair game." In other words, on average the realized rate of return on
an asset is equal to the expected rate of return. We can write the fair game as follows:
= E(Rit) + fl;Ow + (7.31)
where
6,„ = Rmt — E(R„„),
E(6„,t) = 0,
e ft = a random-error term,
11 For papers that discuss some of the econometric problems involved in testing the CAPM, the reader
is referred to Miller and Scholes [1972], Roll [1977, 1981], Scholes and Williams [1977], Dimson [1979],
and Gibbons [1982].
12
Chapter 10 explains the theory of efficient capital markets that describes a fair game at length. Also,
empirical evidence is presented that suggests that the market is in fact a fair game.
EMPIRICAL TESTS OF THE CAPM 213
E(e jt ) = 0
COV(ei„ (5,„) = 0,
COV(eit , E- jt — 1) = 0,
[31, = COV(R 1„ Rmt)/VAR(Rmt).
Equation (7.31) is seen to be a fair game because if we take the expectation of both
sides, the average realized return is equal to the expected return. In other words, on
average you get the return you expected:
E(R jt = E(R it .
If we use the CAPM assumption that asset returns are jointly normal, then ,6; in the
fair game model is defined in exactly the same way as in the CAPM. By substitut-
ing E(R i from the CAPM into Eq. (7.31), we obtain
Rit = RD + [ E(Rmt) — RDA + fli [ R., — E(Rmt)]
= RD + ( Rtn , — R ft )f3 e ft .
Finally, by subtracting R ft from both sides, we have
R j, — R ft = (Rmt — Rft)I3i ei„ (7.32)
which is the ex post form of the CAPM. We derived it by simply assuming that returns
are normally distributed and that capital markets are efficient in a fair game sense.
Now we have an empirical version of the CAPM that is expressed in terms of ex post
observations of return data instead of ex ante expectations.
One important difference between the ex post empirical model and the ex ante
theoretical model is that the former can have a negative slope, whereas the latter can-
not. After the fact we may have experienced a state of nature where the market rate of
return was negative. When this happens the empirical security market line will slope
downward as in Fig. 7.9(a). On the other hand, the theoretical CAPM always requires
1.0
(a) (b)
Figure 7.9
(a) Ex post CAPM; (î) ex ante CAPM.
214 MARKET EQUILIBRIUM: CAPM AND APT
the ex ante expected return on the market to be higher than the risk-free rate of
return, as shown in Fig. 7.9(b). This is because prices must be established in such a
way that riskier assets have higher expected rates of return. Of course, it may turn
out that after the fact their return was low or negative, but that is what is meant by
risk. If a risky asset has a beta of 2.0, it will lose roughly 20% when the market goes
down by 10%.
When the CAPM is empirically tested it is usually written in the following form:
Kpt —
"Yo +7'ifip+ Ept, (7.33)
where
= R„„— R f „
R'p„ = the excess return on portfolio p, (R pt — R ft ).
This is the same as Eq. (7.32) except that a constant term, 70, has been added. Exactly
what predictions made by the CAPM are tested in Eq. (7.33)? The predictions should
meet the following criteria:
a) The intercept term, y o , should not be significantly different from zero. If it is dif-
ferent from zero, then there may be something "left out" of the CAPM that is
captured in the empirically estimated intercept term.
b) Beta should be the only factor that explains the rate of return on a risky asset.
If other terms such as residual variance, dividend yield, price/earnings ratios, firm
size, or beta squared are included in an attempt to explain return, they should
have no explanatory power.
c) The relationship should be linear in beta.
d) The coefficient of beta, y„ should be equal to (Rmt — R ft ).
e) When the equation is estimated over very long periods of time, the rate of return
on the market portfolio should be greater than the risk-free rate. Because the
market portfolio is riskier, on average it should have a higher rate of return.
The major empirical tests of the CAPM were published by Friend and Blume
[1970], Black, Jensen, and Scholes [1972], Miller and Scholes [1972], Blume and
Friend [1973], Blume and Husick [1973], Fama and Macbeth [1973], Basu [1977],
Reinganum [1981b], Litzenberger and Ramaswamy [1979], Banz [1981], Gibbons
[1982], Stambaugh [1982], and Shanken [1985b]. Most of the studies use monthly
total returns (dividends are reinvested) on listed common stocks as their data base.
A frequently used technique is to estimate the betas of every security during a five-
year holding period, by computing the covariance between return on the security and
a market index that is usually an equally weighted index of all listed common stocks.
The securities are then ranked by beta and placed into N portfolios (where N is
usually 10, 12, or 20). By grouping the individual securities into large portfolios chosen
to provide the maximum dispersion in systematic risk, it is possible to avoid a good
part of the measurement error in estimating betas of individual stocks. Next, the port-
folio betas and returns are calculated over a second five-year period and a regression
similar to Eq. (7.33) is run.
EMPIRICAL TESTS OF THE CAPM 215
With few exceptions, the empirical studies agree on the following conclusions:
a) The intercept term, 7 0 , is significantly different from zero, and the slope, y l , is less
than the difference between the return on the market portfolio minus the risk-free
rate.' The implication is that low beta securities earn more than the CAPM
would predict and high beta securities earn less.
b) Versions of the model that include a squared beta term or unsystematic risk find
that at best these explanatory factors are useful only in a small number of the time
periods sampled. Beta dominates them as a measure of risk.
c) The simple linear empirical model [Eq. (7.33)] fits the data best. It is linear in
beta. Also, over long periods of time the rate of return on the market portfolio
is greater than the risk-free rate (i.e., y, > 0).
d) Factors other than beta are successful in explaining that portion of security re-
turns not captured by beta. Basu [1977] found that low price/earnings portfolios
have rates of return higher than could be explained by the CAPM. Banz [1981]
and Reinganum [1981b] found that the size of a firm is important. Smaller firms
tend to have high abnormal rates of return. Litzenberger and Ramaswamy [1979]
found that the market requires higher rates of return on equities with high divi-
dend yields. Keim [1983, 1985] reports seasonality in stock returns a January
effect.
Figure 7.10 shows the average monthly returns on 10 portfolios vs. their systematic
risk for the 35-year period 1931-1965 (taken from the Black-Jensen-Scholes study
[1972]). The results shown here are typical. The empirical market line is linear with a
positive trade-off between return and risk, but the intercept term is significantly dif-
ferent from zero. In fact, it is 9.79 standard deviations away. This forces us to reject
the CAPM, given the empirical techniques of the previously mentioned studies. In
addition, the ability of other variables such as price/earnings ratios to explain the
portion of returns that are unexplained by the CAPM suggests either (1) that the
CAPM is misspecified and requires the addition of factors other than beta to explain
security returns or (2) that the problems in measuring beta are systematically related
to variables such as firm size. Work that is consistent with this second point of view
has been published by Rosenberg and Marathe [1977], who find that beta can be
predicted much better if variables such as dividend yield, trading volume, and firm
size are added to the predictive model. Roll [1981] suggests that infrequent trading
of shares in small firms may explain much of the measurement error in estimating
betas.
Gibbons [1982], Stambaugh [1982], and Shanken [1985b] test the CAPM by
first assuming that the market model is true i.e., that the return on the ith asset is
a linear function of a market portfolio proxy such as an equally weighted market
portfolio:
13
Empirical studies have used a 90-day Treasury bill as a proxy for the risk-free rate, and they have
also laboriously calculated the return on the zero-beta portfolio. Either approach results in an intercept
term significantly different from zero.
216 MARKET EQUILIBRIUM: CAPM AND APT
.06
Intercept = .00519
Standard error = .00053
Slope = .01081
.04 Standard error = .00050
Average monthly returns
.02
xi oixox'xagx
X/
.00
—02 I I I I
.0 .5 1.0 1.5 2.0
Systematic risk
Figure 7.10
Average monthly returns vs. systematic risk for 10
portfolios, 1931-1965. (From Studies in the Theory
of Capital Markets, edited by Michael C. Jensen.
Copyright © 1972 by Praeger Publishers, Inc.
Reprinted by permission of Holt, Rinehart and
Winston.)
The market model, Eq. (7.34), is merely a statistical statement. It is not the CAPM.
The CAPM e.g., Black's [1972] two-factor version actually requires the intercept
term, E(12,) in Eq. (7.25), to be the same for all assets. The two-factor CAPM is true
across all assets at a point in time,
E(R i) = E(Rz) [E(R,n) — E(Rz)1(3i . (7.25)
Gibbons [1982] points out that the two-factor CAPM implies the following constraint
on the intercept of the market model
a i = E(Rz)(1 — fl i) (7.35)
for all securities during the same time interval. When he tests restriction (7.35), he
finds that it is violated and that the CAPM must be rejected.
The empirical evidence has led scholars to conclude that the pure theoretical form
of the CAPM does not agree well with reality. However, the empirical form of the
model, which has come to be known as the empirical market line,
/ .1
Rat = liot + 1 1tfiit + eft, (7.36)
does provide an adequate model of security returns. The practitioner who wishes to
have unbiased estimates of the empirical market line parameters, you and ji lt, estimated
THE PROBLEM OF MEASURING PEFORMANCE: ROLL'S CRITIQUE 217
each month from January 1935 through June 1968, is referred to Fama [1976]. Ob-
viously, if one can estimate a security's beta for a given period, then by knowing the
empirical market line parameters, he can estimate the security's required rate of return
from Eq. (7.36).
One of the potentially most useful applications of the securities market line in its ex
post form [Eq. (7.32)] or the empirical market line [Eq. (7.36)] is that they might be
used as benchmarks for security performance. The residual term, eft, has been inter-
preted as abnormal because, as shown in Fig. 7.11, it represents return in excess of
what is predicted by the security market line.
Roll [1977] takes exception to this interpretation of cross-section abnormal per-
formance measures and to empirical tests of the CAPM in general. In brief, his ma-
jor conclusions are:
1. The only legitimate test of the CAPM is whether or not the market portfolio (which
includes all assets) is mean-variance efficient.
2. If performance is measured relative to an index that is ex post efficient, then from
the mathematics of the efficient set no security will have abnormal performance
when measured as a departure from the security market line."
3. If performance is measured relative to an ex post inefficient index, then any
ranking of portfolio performance is possible, depending on which inefficient index
has been chosen.
Return
Figure 7.11
Abnormal return.
14
It is important to note that Roll does not take exception to time series measures of abnormal performance
such as those described by the market model in Chapter 11.
218 MARKET EQUILIBRIUM: CAPM AND APT
This is a startling statement. It implies that even if markets are efficient and the CAPM
is valid, then the cross-section security market line cannot be used as a means of
measuring the ex post performance of portfolio selection techniques. Furthermore,
the efficiency of the market portfolio and the validity of the CAPM are joint
hypotheses that are almost impossible to test because of the difficulty of measuring
the true market portfolio.
To understand Roll's critique, we must go back to the derivation of the zero-
beta portfolio. Recall that if there is no risk-free asset, it is still possible to write the
security market line as a combination of the market portfolio and a zero-beta portfolio
that is uncorrelated with the market index. Therefore the expected return on any
asset could be written as a two-factor model:
E(Ri) = E(Rz) + [E(Rm ) — E(RAI31. (7.37)
Roll points out that there is nothing unique about the market portfolio. It is always
possible to choose any efficient portfolio as an index, then find the minimum variance
portfolio that is uncorrelated with the selected efficient index. This is shown in Fig.
7.12. Once this has been done, then Eq. (7.37) can be derived and written as
E(Ri) = E(Rz ,,) + [E(R,) — (7.38)
Note that the market portfolio, Rn„ has been replaced by any efficient index, RI , and
the beta is measured relative to the selected efficient index, flu . Also, the zero-beta
portfolio is measured relative to the index, Rz,,. Because the expected return on any
asset can be written as a linear function of its beta measured relative to any efficient
index, it is not necessary to know the market index. One only need know the com-
position of an efficient index in order to write Eq. (7.38). Furthermore, if the index
turns out to be ex post efficient, then every asset will fall exactly on the security
market line. There will be no abnormal returns. If there are systematic abnormal
returns, it simply means that the index that has been chosen is not ex post efficient.
Figure 7.12
Two index portfolios with their respective orthogonal
portfolios.
THE ARBITRAGE PRICING THEORY 219
The Roll critique does not imply that the CAPM is an invalid theory. However,
it does mean that tests of the CAPM must be interpreted with great caution. The fact
that portfolio residuals exhibited no significant departures from linearity merely
implies that the market index that was selected (usually an equally weighted index
of all listed shares of common stock) was ex post efficient. In fact, the only way to
test the CAPM directly is to see whether or not the true market portfolio is ex post
efficient. Unfortunately, because the market portfolio contains all assets (marketable
and nonmarketable, e.g., human capital, coins, houses, bonds, stocks, options, land,
etc.), it is impossible to observe.
1. The Theory
Formulated by Ross [1976], the arbitrage pricing theory (APT) offers a testable
alternative to the capital asset pricing model. The CAPM predicts that security rates
of return will be linearly related to a single common factor—the rate of return on
the market portfolio. The APT is based on similar intuition but is much more general.
It assumes that the rate of return on any security is a linear function of k factors as
shown below:
= E(k i ) + ha :E l + + b ik f k Ei , (7.39)
where
= the random rate of return on the ith asset,
E(k i) = the expected rate of return on the ith asset,
b ik = the sensitivity of the ith asset's returns to the kth factor,
F k = the mean zero kth factor common to the returns of all assets under
consideration,
= a random zero mean noise term for the ith asset.
As we shall see later on, the CAPM may be viewed as a special case of the APT
when the market rate of return is assumed to be the single relevant factor.
The APT is derived under the usual assumptions of perfectly competitive and
frictionless capital markets. Furthermore, individuals are assumed to have homo-
geneous beliefs that the random returns for the set of assets being considered are
governed by the linear k-factor model given in Eq. (7.39). The theory requires that
the number of assets under consideration, n, be much larger than the number of fac-
tors, k, and that the noise term, E i, be the unsystematic risk component for the ith
asset. It must be independent of all factors and all error terms for other assets.
The most important feature of the APT is reasonable and straightforward. In
equilibrium all portfolios that can be selected from among the set of assets under
consideration and that satisfy the conditions of (a) using no wealth and (b) having
no risk must earn no return on average. These portfolios are called arbitrage port-
folios. To see how they can be constructed, let w i be the change in the dollar amount
220 MARKET EQUILIBRIUM: CAPM AND APT
w i = 0. (7.40)
i =
If there are n assets in the arbitrage portfolio, then the additional portfolio return
gained is
RP =
i=i
= E w,E(iz i) + Ew i b il P l + • • • + w,b,kfk + (7.41)
wi 1/n, (7.42a)
n chosen to be a large number, (7.42b)
wibik = 0 for each factor. (7.42c)
Because the error terms, ""i , are independent the law of large numbers guarantees that
a weighted average of many of them will approach zero in the limit as n becomes large.
In other words, costless diversification eliminates the last term (the unsystematic or
idiosyncratic risk) in Eq. (7.39). Thus we are left with
At first glance the return on our portfolio appears to be a random variable, but we
have chosen the weighted average of the systematic risk components for each factor
to be equal to zero (1 wi b ik = 0). This eliminates all systematic risk. One might say
that we have selected an arbitrage portfolio with zero beta in each factor. Conse-
quently, the return on our arbitrage portfolio becomes a constant. Correct choice of
the weights has eliminated all uncertainty, so that R p is not a random variable.
Therefore Eq. (7.41) becomes
Rp = E wi E(k i). (7.43a)
Recall that the arbitrage portfolio, so constructed, has no risk (of any kind) and
requires no new wealth. If the return on the arbitrage portfolio were not zero, then
it would be possible to achieve an infinite rate of return with no capital requirements
and no risk. Such an opportunity is clearly impossible if the market is to be in equi-
THE ARBITRAGE PRICING THEORY 221
librium. In fact, if the individual arbitrageur is in equilibrium (hence content with his
or her current portfolio), then the return on any and all arbitrage portfolios must be
zero. In other words,
Rp = ,E(k) = 0. (7.44)
Eqs. (7.40), (7.42c), and (7.44) are really statements in linear algebra. Any vector
that is orthogonal to the constant vector, i.e.,15
wi) • e = 0, (7.40)
An algebraic consequence of this statement is that the expected return vector must
be a linear combination of the constant vector and the coefficient vectors. Algebrai-
cally, there must exist a set of k + 1 coefficients, A O , A,, , /1., such that
ECRT — A) + Alba + (7.45)
Recall that the b a are the "sensitivities" of the returns on the ith security to the
kth factor. If there is a riskless asset with a riskless rate of return, R f , then b,„ = 0 and
15 Note that Eq. (7.40) says that the sum of the investment weights equals zero. This is really a no-wealth
constraint. No new wealth is required to take an arbitrage position. Recall that e is a column vector of ones.
222 MARKET EQUILIBRIUM: CAPM AND APT
Figure 7.13
The arbitrage pricing line.
Suppose that asset returns are determined by two underlying factors such as unan-
ticipated changes in real output and unanticipated inflation. The arbitrage pricing
theory can easily account for the effect of changes in both factors on asset returns.
Because the capital asset pricing model relies on a single factor (the market index),
it cannot do as well. Using the CAPM is a little like being lost in the clouds while
piloting a private plane. You call the air controller and ask, "Where am I?" If the
controller is using a unidimensional model like the CAPM, he or she is likely to
respond, "Two hundred miles from New York City." Obviously, this is not a very
helpful answer. A multidimensional model like the APT would be more useful. It
would be nice to know latitude, longitude, and altitude.
Figure 7.14 illustrates the same point. The factor loadings (or factor sensitivities),
bit and b i2 , for our two hypothetical factors—changes in unanticipated real output
and changes in unanticipated inflation—are plotted on the axes. The origin represents
the risk-free rate that is the rate of return received when an asset has zero beta in
both factors. Points along the diagonal dashed lines have equal expected return but
not the same risk. For example, all points along the line OJ have an expected rate
of return equal to the risk-free rate but are not riskless portfolios. If the risk-free rate
is 10%, one can obtain that rate either with a truly riskless portfolio that pays 10%
in every state of nature or with a second portfolio that has positive sensitivity to one
factor and negative sensitivity to the other factor.
Suppose the arbitrage pricing model, Eq. (7.47),
E(Ri) ÿ R f = [6 ÿ R f ]b [a2 ÿ R
is estimated to have the following numerical values: R f = 10%, 6 1 = 20%, and 5 2 =
15%. If b it is plotted on the vertical axis, then the vertical intercept for a given
return E(Ri) is
E(Ri) — R f
= vertical intercept = o —R
i f
î il for unanticipated
changes in real output
•
---,--.
---.. -,.
=.P,- -,-
-ik / (index portfolio)
---,
-,.. "--- -,./..)
=.0 ---.. ,,, .4,
-. -,.. P .,,,,)
-,,-..,. --.-,..
, --.. p]
J .5
AK
2. A Numerical Example
To illustrate how arbitrage pricing might be employed in practice, suppose that
empirical work reveals that the expected returns on assets can be explained by two
factors, F 1 and F<. Table 7.3 shows the subjectively estimated returns on three assets
(X, Y, and Z) and the changes in (orthogonal transformations of) the two factors for
five equally likely states of nature. To make the problem tractable, assume that all
factors and returns are normally distributed. In addition, suppose we know that the
expected return on the risk-free asset, R f , is 10%.
One of the requirements of the factor analysis program usually used to test the
arbitrage pricing model is that we are working with linear transformations of the
underlying factors. The transformations must be orthogonal, i.e., the product of their
row and column vectors must equal zero. This is shown below for the transformed
factors in Table 7.3:
—5.00
38.48
[ —10 — 5 25 40 50] 8.00 = 0.
—1.44
0
How can we tell from the bewildering set of numbers in Table 7.3 whether or not
there are any arbitrage opportunities? And if there are, how can we take advantage
of them by forming an arbitrage portfolio?
If there are only two factors that govern all returns, then the APT becomes
E(Ri) = R f + — R f ]b i , + [82 - R f ]b i2 .
The data from Table 7.3 can be used to compute all the terms on the right-hand side
of this equation. The factor loadings (or sensitivities) are the same as beta, Eq. (7.47a),
given the assumption of normally distributed returns and orthogonal transformations
of the factors. Using asset X as an example, we need to calculate
b COV(X, (5 ) 285.0
1
Note that the equilibrium return, E(Ri), on assets X and Z is the same as the projected
return, Ri , computed from the data. Hence no arbitrage opportunities exist for trading
in these two assets. On the other hand, the projected return on asset Y, R y, is 25%
when computed from the data, and the market equilibrium return, E(R)), is only 17%.
226 MARKET EQUILIBRIUM: CAPM AND APT
é)(51 - P=)
.<( - 66.<3)( -30) = 397.38
.<(P9.70)( -<P) = - <98.P0
<8P.00
.<( - <0.00)(P) = -<0.00 b„1 = = .P
P70.00
.<( - <3.47)(<0) = -93.88
.<(P0.00)(30) = 300.00
COV(X, S i ) = <8P.00
Therefore îy selling the correct proportions of assets é and Z and îuying asset Y,
we can form an arîitrage portfolio that requires no new capital, has no change in
risk, and earns a positive rate of return.
Suppose that we currently have one third of our wealth in each of the three assets.
How should we change our portfolio to form a riskless arîitrage position? It turns
out that so long as there are more assets than factors, there are virtually an infinite
numîer of ways of forming arîitrage portfolios. But let us suppose that we desire to
put the maximum investment into asset Y without actually selling short either é or
Z. Our investment proportion in Y would go from one third to one; thus the change
in holdings of asset Y would îe w y = Ç. We also require that the change portfolio
have zero îeta in each factor and that it need no net wealth. These conditions are
stated in Eqs. (7.40) and (7.4<c), which are repeated îelow:
( 3
E w,) e = 0. (7.40)
i ==
3
Thus we would sell all our holdings in assets X and Z and invest the proceeds in
asset Y. This strategy would require no new wealth and would have no change in
risk. Note that our risk position before making the change was
4(.5) + 4(1.0) + 4(1.5) = 1.0 for factor 1,
4(2) + 4(1.5) + 4(1.0) = 1.5 for factor 2.
After undertaking the new investment strategy our risk will be
0(.5) + 1(1.0) + 0(1.5) = 1.0 for factor 1,
0(2) + 1(1.5) + 0(1.0) = 1.5 for factor 2.
Since our systematic risk has not changed, the extra systematic risk created by the
arbitrage portfolio is zero." Our originally projected portfolio return was
4(11%) + 4(25%) + 4(23%) = 19.67%.
But after investing in the arbitrage portfolio, we project a rate of return of
0(11%) + 1(25%) + 0(23%) = 25%.
Thus the arbitrage portfolio increases our return by 5.33% without changing our
systematic risk.
Figure 7.15 provides a visual description of the example problem. Expected rates
of return are plotted on the vertical axis, and asset betas in each of the two factors
are plotted along the horizontal axes. Note that the expected returns for assets X
and Z plot exactly on the arbitrage pricing plane. They are in equilibrium. But asset
Y plots above the plane. Its return lies considerably above what the market requires
16
Because there is idiosyncratic risk to contend with, total risk will in fact change. However, with well-
diversified arbitrage portfolios this problem vanishes because diversification reduces idiosyncratic risk
until it is negligible.
228 MARKET EQUILIBRIUM: CAPM AND APT
Figure 7.15
Arbitrage pricing plane for two factors.
for its factor loadings, b y, and by2 . Hence an arbitrage opportunity exists. If enough
people take advantage of it the price of asset Y will rise, thereby forcing its rate of
return down and back into equilibrium.
Papers by Gehr [1975], Roll and Ross [1980], Reinganum [1981a], and Chen [1983]
have tested the APT using data on equity daily rates of return for New York and
American Stock Exchange listed stocks. The usual empirical procedure has the fol-
lowing steps:
1. Collect a time series of daily stock returns data for a group of stocks.
2. Compute the empirical variance-covariance matrix from the returns data.
3. Use a (maximum-likelihood) factor analysis procedure to identify the number of
factors and their factor loadings, b,.
4. Use the estimated factor loadings, b„, to explain the cross-sectional variation
of individual estimated expected returns and to measure the size and statistical
significance of the estimated risk premia associated with each factor.
The Roll and Ross [1980] study used daily returns data for NYSE and AMEX
companies listed on the exchanges on both July 3, 1962, and December 31, 1972.
EMPIRICAL TESTS OF THE ARBITRAGE PRICING THEORY 229
There were a maximum of 2619 daily returns for each stock. The 1260 securities
selected were divided alphabetically into groups of 30. For each group of 30 the
procedure described above was carried out. The analysis showed that there are at
least three and probably four "priced" factors. There may be other zero-priced factors,
but this procedure cannot identify them because their regression coefficients in step 4
would be zero.
One of the frustrating things about using factor analysis to test the APT is that
this procedure cannot tell us what the factors are. However, we can reject the APT
if a specified alternative variable such as the total variance of individual returns, firm
size, or the asset's last period return were to be significant in explaining the expected
returns. Roll and Ross [1980], after correcting for the problem that positive skew-
ness in lognormal returns can create dependence between the sample mean and the
sample standard deviation, found that the total variance of security returns does not
add any explanatory power for estimated expected returns. Therefore the APT could
not be rejected on this basis. Although a different procedure was employed by Chen
[1983], he was able to confirm this result. He also finds that the asset's last period
return adds no explanatory power.
Currently there is a question whether or not firm size can be used to refute the
APT because it adds explanatory power to the factor loadings. Reinganum [1981a]
finds that it does. His test consisted of estimating the factor loadings in year (t — 1)
for all securities, then combining securities with similar loadings into control port-
folios. In year t, excess security returns are computed by subtracting the daily control
portfolio returns from the daily security returns. Finally, with excess returns in hand,
the securities are ranked on the basis of the market value of all the firm's outstanding
common stock at period (t — 1). The APT predicts (if factor loadings are stationary
across time) that all deciles of the market value ranked excess returns should have
the same mean. Reinganum finds that there are significant differences between the
mean excess returns and rejects the APT. Chen [1983], on the other hand, finds that
firm size adds no explanatory power. His procedure uses Reinganum's data for the
market value of each firm's equity. He divides the sample of firms into two groups—
those with greater than the median market value and those with less. Then portfolios
are formed from the high- and low-value firms so that the following conditions are
satisfied: (1) each security in the portfolio has nonnegative weight and the weight
should not be too far from 1/n, where n is the number of securities in the portfolio;
and (2) the resultant two portfolios have exactly the same factor loadings (arbitrage
risk factors) in each factor. The factor loadings are determined by using returns data
from odd days during each test period; the even-days returns from the same test period
are used for measuring the average portfolio returns of the high- and low-valued port-
folios. If the APT is correct the returns of the two portfolios should not be statistically
different because they are selected to have the same "risk" as determined by the factor
loadings. In only one of the four periods tested is the difference in returns statistically
different at the 95% confidence level. Therefore Chen argues that firm size effects have
insufficient explanatory power to reject the APT.
There is one parameter in the APT namely, the intercept term .10 = R that
should be identical across groups of securities when the model is estimated separately
230 MARKET EQUILIBRIUM: CAPM AND APT
for each group during a time period. Other factors need not be the same because the
factor loadings are not unique from group to group. For example, factor 1 in group
A might correspond to factor 3 in group B. The intercept term, 20 , however, has the
same meaning in each group because it is the return on an asset that has no sensitivity
to the common factors. Roll and Ross [1980] tested for the equivalence of the 20
terms across 38 groups and found absolutely no evidence that the intercept terms were
different. Again, the APT could not be rejected.
A direct comparison of the APT and the CAPM was performed by Chen [1983].
First, the APT model was fitted to the data as in the following equation:
k i = 10 + a l b l + • • • + „b in + Ei . (APT)
Then the CAPM was fitted to the same data
R i = )s, o + + (CAPM)
Next the CAPM residuals, q i , were regressed on the arbitrage factor loadings, 5,„
and the APT residuals, e i , were regressed on the CAPM coefficients. The results
showed that the APT could explain a statistically significant portion of the CAPM
residual variance, but the CAPM could not explain the APT residuals. This is strong
evidence that the APT is a more reasonable model for explaining the cross-sectional
variation in asset returns.
Although it is mathematically impossible to use factor analysis to unambiguously
identify the underlying factors that drive security returns, Chen, Roll, and Ross
[1983] have correlated various macroeconomic variables with returns on five port-
folios that mimic the underlying factors. Their conclusions provide valuable insight
into what the underlying factors might be. Four macroeconomic variables were
significant:
1. An index of industrial production.
2. Changes in a default risk premium (measured by the differences in promised
yields to maturity on AAA versus Baa corporate bonds).
3. Twists in the yield curve (measured by differences in promised yields to maturity
on long- and short-term government bonds).
4. Unanticipated inflation.
The economic logic underlying these variables seems to make sense. Common
stock prices are the present values of discounted cash flows. The industrial produc-
tion index is obviously related to profitability. The remaining variables are related
to the discount rate.
The intuition behind these factors is useful for portfolio management. For ex-
ample, it has often been stated that common stocks are not a good hedge against
inflation. Although it is true if one holds an equally weighted portfolio of all stocks,
the logic of factor analysis suggests that there is a well-diversified subset of common
stocks that is in fact a good hedge against inflation. Since the factors are mutually
orthogonal, one can (at least in principle) choose a portfolio which is hedged against
inflation risk without changing the portfolio sensitivity to any of the other three
above-mentioned factors.
PROBLEM SET 231
SUMMARY
This chapter has derived two theoretical models, the CAPM and the APT, that
enable us to price risky assets in equilibrium. Within the CAPM framework the
appropriate measure of risk is the covariance of returns between the risky asset in
question and the market portfolio of all assets. The APT model is more general.
Many factors (not just the market portfolio) may explain asset returns. For each
factor the appropriate measure of risk is the sensitivity of asset returns to changes
in the factor. For normally distributed returns the sensitivity is analogous to the
beta (or systematic risk) of the CAPM.
The CAPM was shown to provide a useful conceptual framework for capital
budgeting and the cost of capital. It is also reasonably unchanged by the relaxation
of many of the unrealistic assumptions that made its derivation simpler. Finally,
although the model is not perfectly validated by empirical tests, its main implications
are upheld—namely, that systematic risk (beta) is a valid measure of risk, that the
model is linear, and that the trade-off between return and risk is positive.
The APT can also be applied to cost of capital and capital budgeting problems.
Chapter 12 will show how the APT can be used for capital budgeting in a multi-
period framework. The earliest empirical tests of the APT have shown that asset
returns are explained by three or possibly four factors and have ruled out the vari-
ance of an asset's own returns as one of the factors.
PROBLEM SET
7.1 Let us assume a normal distribution of returns and risk-averse utility functions. Under
what conditions will all investors demand the same portfolio of risky assets?
7.2 The following data have been developed for the Donovan Company, the manufacturer
of an advanced line of adhesives:
7.3 The following data have îeen developed for the Milliken Company:
The yield to maturity on Treasury Bills is .066 and is expected to remain at this point for
the foreseeaîle future. Calculate the following:
a) The expected market return.
î) The variance of the market return.
c) The expected return for the Milliken Company.
d) The covariance of the return for the Milliken Company with the return on the market.
e) Write the equation of the security market line.
f) What is the required return for the Milliken Company?
7.4 For the data in Taîle Q7.4, perform the indicated calculations.
7.P For the data in Taîle Q7.P, calculate the items indicated.
Percentage
GM Change Dividend Percentage Deviation Variance Covariance
Year Price in Price Yield Return of Returns of Returns with Market
7.6 What are the assumptions sufficient to guarantee that the market portfolio is an efficient
portfolio?
7.7 In the CAPM is there any way to identify the investors who are more risk averse? Ex-
plain. How would your answer change if there were not a riskless asset?
7.8 Given risk-free borrowing and lending, efficient portfolios have no unsystematic risk. True
or false? Explain.
7.9 What is the beta of an efficient portfolio with E(Rd= 20% if R f = 5%, E(Rm) = 15%,
and o-„, = 20%? What is its rr? What is its correlation with the market?
7.10 Given the facts of Problem 7.9, and that the common stock of the Rapid Rolling Cor-
poration has E(Rk ) = 25% and ol = 52%, what is the systematic risk of the common stock?
What is its unsystematic risk?
7.11
a) If the expected rate of return on the market portfolio is 14% and the risk-free rate is 6%,
find the beta for a portfolio that has an expected rate of return of 10%. What assump-
tions concerning this portfolio and/or market conditions do you need to make to calculate
the portfolio's beta?
b) What percentage of this portfolio must an individual put into the market portfolio in
order to achieve an expected return of 10%?
7.12 You believe that the Beta Alpha Watch Company will be worth $100 per share one year
from now. How much are you willing to pay for one share today if the risk-free rate is 8%,
the expected rate of return on the market is 18%, and the company's beta is 2.0?
7.13 Given the following variance-covariance matrix and expected returns vector (for assets
X and Y, respectively) for a two-asset world:
0 1,
= RI = [.2 .1],
0 0064]'
234 MARKET EQUILIBRIUM: CAPM AND APT
a) What is the expected return of a zero-beta portfolio, given that 50% of the index port-
folio is invested in asset é and in asset Y?
b) What is the vector of weights in the global minimum variance portfolio?
c) What is the covariance between the global minimum variance portfolio and the zero-
beta portfolio?
d) What is the equation of the market line?
7.14 Given the following variance-covariance matrix, calculate the covariance between
portfolio A, which has 10% in asset 1 and 90% in asset 2, and portfolio B, which has 60% in asset 1
and 40% in asset 2:
.01 — .021
— .02 .04i .
7.15 Suppose that securities are priced as if they are traded in a two-parameter economy. You
have forecast the correlation coefficient between the rate of return on Knowlode Mutual Fund
and the market portfolio at .8. Your forecast of the standard deviations of the rates of return
are .25 for Knowlode, and .20 for the market portfolio. How would you combine the Knowlode
Fund and a riskless security to obtain a portfolio with a volatility (beta) of 1.6?
7.16 You currently have 50% of your wealth in a risk-free asset and 50% in the four assets below:
If you want an expected rate of return of 12%, you can obtain it by selling some of your
holdings of the risk-free asset and using the proceeds to buy the equally weighted market
portfolio. If this is the way you decide to revise your portfolio, what will the set of weights in
your revised portfolio be? If you hold only the risk-free asset and the market portfolio, what
set of weights would give you an expected 12% return?
7.17 The market price of a security is $40, the security's expected rate of return is 13%, the
riskless rate of interest is 7%, and the market risk premium, [E(R„,) ÿ R f ], is 8%. What will
be the security's current price if its expected future payoff remains the same but the covariance
of its rate of return with the market portfolio doubles?
7.18 Suppose you are the manager of an investment fund in a two-parameter economy. Given
the following forecast:
E(R„,) = .16, o-(R,„) = .20, R f = .08,
a) Would you recommend investment in a security with E(R ;) = .12 and COV(R 1, R„,) = .01?
[Note- Assume that this price change has no significant effect on the position of the
security market line.]
b) Suppose that in the next period security R i has earned only 5% over the preceding period.
How would you explain this ex post return?
REFERENCES 235
.0081 0
=[ 0 .0025].
Asset A has an expected return of 30%, and asset B has an expected return of 20%. Answer
the following questions:
a) Suppose investor I chooses his "market portfolio" to consist of 75% in asset A and 25%
in asset B, whereas investor J chooses a different "market portfolio" with 50% in asset A and
50% in asset B.
Weights chosen by I are: [.75 .25]
Weights chosen by J are: [.50 .50]
Given these facts, what /3 will each investor calculate for asset A?
b) Given your answer to part (a) above, which of the following is true and why?
1. Investor I will require a higher rate of return on asset A than will investor J.
2. They will both require the same return on asset A.
3. Investor J will require a higher rate of return on asset A than will investor I.
c) Compute the zero-beta portfolios and the equations for the security market line for each
investor.
7.21 Ms. Bethel, manager of the Humongous Mutual Fund, knows that her fund currently
is well diversified and that it has a CAPM beta of 1.0. The risk-free rate is 8% and the
CAPM risk premium, [E(R,„) — RA, is 6.2%. She has been learning about APT measures
of risk and knows that there are (at least) two factors: changes in the industrial production
index, ,51, and unexpected inflation, 5 2 . The APT equation is
E(R i) — R f = [S, — R + [S, — R f ib
E(R i ) = .08 +[.05]b11 + [.11]b12 .
a) If her portfolio currently has a sensitivity to the first factor of b p, = — .5, what is its
sensitivity to unexpected inflation?
b) If she rebalances her portfolio to keep the same expected return but reduce her exposure
to inflation to zero (i.e., bp2 = 0), what will its sensitivity to the first factor become?
REFERENCES
Banz, R. W., "The Relationship between Return and Market Value of Common Stocks,"
Journal of Financial Economics, March 1981, 3-18.
Basu, S., "Investment Performance of Common Stocks in Relation to Their Price-Earnings
Ratios: A Test of the Efficient Markets Hypothesis," Journal of Finance, June 1977,
663-682.
<36 MARKET EQUILIBRIUM: CAPM AND APT
Black, F., "Capital Market Equiliîrium with Restricted Borrowing," Jou]nal of Business, July
=97<, 444-4PP.
, M. C. Jensen. and M. Scholes, "The Capital Asset Pricing Model: Some Empirical
Tests," in Jensen, ed., Studies in the Theo]y of Capital Ma]kets. Praeger, New York, =97<,
79-=<4.
Blume, M., "Portfolio Theory: A Step toward Its Practical Application," Jou]nal of Business,
April =970, =P<-=73.
, "On the Assessment of Risk," Jou]nal of Finance, March =97=, =-=0.
, and I. Friend, "A New Look at the Capital Asset Pricing Model," Jou]nal of Finance,
March =973, =9-34.
, and F. Husick, "Price, Beta and Exchange Listing," Jou]nal of Finance, May =973,
<83-<99.
, and R. Stamîaugh, "Biases in Computed Returns: An Application to the Size Effect,"
Jou]nal of Financial Economics, Novemîer =983, 387-404.
Bower, D.; R. Bower; and D. Logue, "Arîitrage Pricing Theory and Utility Stock Returns,"
Jou]nal of Finance, Septemîer =984, =04=-=0P4.
Breeden, D. T., "An Intertemporal Asset Pricing Model with Stochastic Consumption and
Investment Opportunities," Jou]nal of Financial Economics, Septemîer =979, <6P-<96.
Brennan, M. J., "Taxes, Market Valuation and Corporation Financial Policy," National Tax
Jou]nal, Decemîer =970, 4=7-4<7.
Chamîerlain, G., and M. Rothschild, "Arîitrage, Factor Structure, and Mean-Variance Anal-
ysis on Large Asset Markets," Economet]ica, Septemîer =983, =<8=-=304.
Chen, N. F., "Some Empirical Tests of the Theory of Arîitrage Pricing," Jou]nal of Finance,
Decemîer =983, =393-=4=4.
, and J. Ingersoll, Jr., "Exact Pricing in Linear Factor Models with Finitely Many
Assets," Jou]nal of Finance, June =983, 98P-988.
; R. Roll; and S. Ross, "Economic Forces and the Stock Market: Testing the APT and
Alternative Asset Pricing Theories," Working Paper #<0-83, UCLA, Decemîer =983.
Clark, P. K., "A Suîordinated Stochastic Process Model with Finite Variance for Speculative
Prices," Economet]ica, January =973, =3P-=PP.
Copeland, T. E., and D. Mayers, "The Value Line Enigma (=96P-=978): A Case Study of
Performance Evaluation Issues," Jou]nal of Financial Economics, Novemîer =98<,
<89-3<<.
Cornell, B., "Asymmetric Information and Portfolio Performance Measurement," Jou]nal of
Financial Economics, Decemîer =979, 38=-390.
Dhrymes, P.; I. Friend; and B. Gultekin, "A Critical Reexamination of the Empirical Evidence
on the Arîitrage Pricing Theory," Jou]nal of Finance, June =984, 3<3-346.
Dimson, E., "Risk Measurement When Shares are Suîject to Infrequent Trading," Jou]nal of
Financial Economics, June =979, =97-<<6.
Dyîvig, P., "An Explicit Bound on Individual Assets' Deviations from APT Pricing in a Finite
Economy," Jou]nal of Financial Economics, Decemîer =983, 483-496.
, and S. Ross, "Yes, the APT Is Testaîle," Jou]nal of Finance, Septemîer =98P,
==73-==88.
Fama, E. F., "The Behavior of Stock Market Prices," Jou]nal of Business, January =96Pa,
34-=0P.
REFERENCES <37
, "The Capital Asset Pricing Model (CAPM), Short Sales Restrictions and Related
Issues," Jou]nal of Finance, March 1977, 177-184.
, "A Simple Approach to the Valuation of Risky Streams," Jou]nal of Business, July
1978, 453-475.
Rubinstein, M. E., "A Mean-Variance Synthesis of Corporate Financial Theory," Jou]nal of
Finance, March 1973, 167-182.
Scholes, M., and J. Williams, "Estimating Betas from Non-synchronous Data," Jou]nal of
Financial Economics, December 1977, 309-327.
Shanken, J., "Multi-Beta CAPM or Equilibrium APT?: A Reply," Jou]nal of Finance, Septem-
ber 1985a, 1189-1196.
, "Multivariate Tests of the Zero-beta CAPM," Jou]nal of Financial Economics, Septem-
ber 1985b, 327-348.
Sharpe, W. F., "A Simplified Model for Portfolio Analysis," Management Science, January
1963, 277-293.
, "Capital Asset Prices: A Theory of Market Equilibrium under Conditions of Risk,"
Jou]nal of Finance, September 1964, 425-442.
Stambaugh, R., "On the Exclusion of Assets from Tests of the Two-Parameter Model: A Sen-
sitivity Analysis," Jou]nal of Financial Economics, November 1982, 237-268.
Treynor, J., "Toward a Theory of the Market Value of Risky Assets," unpublished manuscript,
1961.
Vasicek, 0. A., "Capital Market Equilibrium with No Riskless Borrowing," March 1971.
Mimeograph available from the Wells Fargo Bank.
Verrecchia, R. E., "The Mayers-Rice Conjecture—A Counterexample," Jou]nal of Financial
Economics, March 1980, 87-100.
8. . . option pricing theory is relevant to almost every area of finance.
For example, virtually all corporate securities can be interpreted as
portfolios of puts and calls on the firm.
J. C. Cox, S. A. Ross, and M. Rubinstein, "Option Pricing:
A Simplified Approach," Journal of Financial Economics, September 1979, 230
A. INTRODUCTION
The theory of option pricing has undergone rapid advances in recent years. Simul-
taneously, organized option markets have developed in the United States. On April
26, 1973, the Chicago Board of Options Exchange (CBOE) became the first orga-
nized exchange for trading standardized options contracts. By the end of 1974, in
terms of share equivalents, volume on the CBOE was larger than that on the Amer-
ican Stock Exchange. By the end of 1976 there were 1337 registered CBOE ex-
change members, and the bid-ask price for a seat was $55,000 to $62,000.1 This
phenomenal growth in option trading has been catalyzed, at least in part, by the
By comparison, during December 1976, nine seats on the New York Stock Exchange exchanged hands
at prices ranging from $58,000 to $80,000.
240
A DESCRIPTION OF THE FACTORS THAT AFFECT PRICES 241
standardization of contracts, which has had the effect of lowering the transaction
costs of option trading.
There are many types of options, and at first the terminology is confusing with
calls, puts, straps, strips, straddles, spreads, in-the-money options, out-of-the-money
options, and so forth. This nomenclature can be greatly simplified if we recognize
that all contracts are made up of four basic securities: puts, calls, stocks (the underly-
ing asset), and default-free bonds. A call option is a contract that is contingent on
the value of an underlying asset. For example, a call option on the CBOE allows its
holder to purchase a share of stock in the underlying company at a fixed price, usually
called the exercise price or the striking price, for a fixed length of time. Table 8.1 is
a clipping from the Wall Street Journal of October 4, 1977. We see, for example, that
three Bethlehem Steel call options were being traded based on the value of its com-
mon stock, which closed that day at $19i per share. The first option had an exercise
price of $20 and three maturity dates, the third Friday in October 1977, January
1978, and April 1978.
Should the price of the common stock have climbed above $20 per share, a trader
who held the call option with an exercise price of $20 could have exercised his or
her option and kept the difference between the exercise price and the stock price. A
put is exactly the opposite of a call option. The holder of a put has the right to sell
the stock at the exercise price any time up to and including the maturity date of the
put. For example, the holder of a January put option on Avon with the exercise price
of $45 could have made a profit if the stock, which was then selling for $47 per share,
fell below $45 before the third Friday in January.
To keep the theory simple for the time being, we assume that all options can be
exercised only on their maturity date and that there are no cash payments (such as
dividends) made by the underlying asset. Options of this type are called European
options. They are considerably easier to price than their American option counter-
parts, which can be exercised at any date up to maturity.
A quick look at the option prices in Table 8.1 shows that at least three factors
are important for the market value of an option. Since most of the options on the
CBOE are call options, we shall, for the moment, confine our discussion to the de-
termination of their value. Obviously, the higher the value of the underlying asset,
S, the greater the value of an option written on it, ceteris paribus. Alcoa, American
Express, Burlington Northern, Digital Equipment, and Walt Disney all have call
options with an exercise price of $40 and a maturity date on the third Friday in
January. Figure 8.1 clearly shows that the value of the call increases as a function
of the value of the stock for a given exercise price and maturity date. Note also that
the option still has a positive value even though the stock price is less than the exer-
cise price. As long as investors believe that there is a chance that the stock price will
Table 8.1 CBOE Option Price Listing
— Oct — —Jan—
Quotations
P
Alcoa ...45 66 Y2 27 2 41 234 433/4 3 1
2 143/4 6344
/, 38
Mon 50
50 21 13 /. 77 14 6
Alcoa 5 1-16 101 1
1 4346 60 482 31/2 75 436 459 51/2 6334
Am Exp 35 25 23/4 26 31/4 1 41/2 371/2 176 a 1634
/, 18 11/2 3716
Pitne y .. 15 3 a a a
Am Exp 40 166 1-16 6 3
Pitney ..20 1 1-16 20 1
/. 50 1/2 163/,
Am Tel .60 197 2 70 21/2 76 27e 62 15 1376 5 141/2 b b 8336
Am Tel .65 53 1-16 210 7
6 86 46 62 Tuesday, October 4, 1977 Proc G 70
3
3 /e 88 472 10 614 8346
Proc G 80 74
Atl R 276 21/2 70 41/4 154 51/4 521/2 Closing prices of all options. Sales unit Proc G 90 5 1-16 83 11-16 79 1 1/2 8336
Atl R 32 1-16 62 46 33 11/4 5246
..15 29 31/2 21 334 23 4 184
usually is 100 shares. Security description Rite A
Avon ... 45 226 2'/e 105 3 216 334 47 ..20 41 1
6 63 /. 30 11-16 181/2
Rite A 3
Avon P 45 257 3
/e 234 1 5-16 223 2 1-16 47 214 20 2 11-16 411/2
Avon 444 1-16 247 11-16 283 1 5-16 47
includes exercise price. Stock close is New St Cal . .40 47 11/2 21
St fa
I Cavi l ..45 S 1-16 41 7-16 24 7
/e 411/2
Avon p 44 33/e 48 4l/e 186 5 47 York Stock Exchange final price. p-Put op- . 35 15 7-16 15 1 3-16 16 13/4 343/4
BankAm 25 84 1/4 32 3
/4 30 11/4 243/4 . 40 a a B V, 2 7-16 3434
BankAm 30 25 1-16 a a b b 2434 tion. o-Old shares. W .. 20 a a 8 97/e a a 7914
T
Tandy y
Beth S 191 7-16 108 1 5-16 129 216 197/e - Oct — — Jan — — Apr — N.Y. . .25 41 41/2 35 51/2 2 61/. 29V.
Tandy
Beth S 25 3 1-16 161 1
/. 198 9-16 1976 Option & price Vol. Last Vol. Last Vol. Last Close ..30 201 7-16 160 21/2 52 3 291/.
Tandy
Beth S _30 a a 5 1-16 25 3-16 1976 Xerox .. 45 268 846 282 876 271 946 533/4 Tady .. 35 13 1-16 268 '/2 b b 791/4
Bruns ...10 a a 17 23/e 14 2 51-16 1216 Xerox .. 50 1306 31/2 411 476 203 574 5346 a 25 15 346 9 1
3 /2 7 4 281/4
T
Teex . acoco
Bruns .. 15 4 1-16 93 3-16 69 7-16 121/2 Xerox .. 60 130 1-16 545 3
6 578 11/2 533/. 30 384 1-16 275 3
/e 56 9-16 281/4
Burl N . 40 2 31/2 a a 4 47/e 431/2 — Nov — — Feb — — May — .45 83 451 67 1'h 34 2 6 437/e
3
U Carb
Burl . 45 57 1
4 62 1 36 41 2 431/2 A E P 241/2 67 3-16 15 5-16 b b 241/4 .50 4 1-16 73 5-16 84 VA 437e
U Carb
Burl N . 50 a a 9 5-16 25 11-16 431/2 A E P b b b b 10 1/2 241/. a a 30 1-16 b b 4376
U Carb 60
Burrgh . 50 35 19 194 19 b b 69 Am Hos 20 5 4 a a a a 24 b 73 4 7
/e 17 5 Me
U S St .. .25 b
Burrgh 60 306 9 50 1046 27 111/2 69 Am Hos 25 18 6 20 11/4 a a 24 . 33 456 5-16 804 11/2 200 2 3-16 293/e
3
U S St
Burrgh 70 420 70 384 31/4 219 41/2 69 A M P 1 344 a a a a 21336 ..35 a a 183 5-16 103 6 2946
U S St
3
C (call price)
X=_$ 35
12
Digital equipment #1
11 X = $ 40
10 S=X
9 II
8 •
7
6 01Alcoa
1■
.0 4 ` T = January, 109 days
0
0°
1 Disney 3
American Jr 2
Expressi_ 0001110 45°
IP 1
S (stock price)
36 37 38 39 40 4 42 43 44 45 46 47 48
Figure 8.1
The relationship between the call price and the stock price.
exceed the exercise price before the option matures, the option will be valuable. Three
of these companies, American Express, Digital Equipment, and Disney, have a second
option with a lower exercise price of $35. Note that the relationship between the call
price and the stock price has shifted upward. The lower the exercise price, the greater
the value of a call option. The third obvious factor that affects call prices is the length
of time to maturity. A quick look at any option in Table 8.1 shows that the longer
the time to maturity, the higher the value of the option. The reason is that with more
time to maturity there is a greater chance that the stock price will climb higher above
the exercise price. Hence options with longer maturity have higher prices. In fact, a
call option that has an infinite maturity date will have the same value as the stock,
regardless of its exercise price. This is because the option will never be exercised.
(Why not?)
In addition to the stock price, the exercise price, and the time to maturity, there are
two other important (but less obvious) factors that affect the option's value: (1) the
instantaneous variance of the rate of return on the underlying asset (common stock)
and (2) the risk-free rate of return. The holder of a call option will prefer more vari-
ance in the price of the stock to less. The greater the variance, the greater the prob-
ability that the stock price will exceed the exercise price, and this is of value to the call
holder. A call option is a type of contingent claim. In other words, the option
holder gains only under the condition that the stock price exceeds the exercise price at
the maturity date. Suppose we hold options and are faced with the two hypothetical
stock price distributions shown in Fig. 8.2. Both distributions have identical means, but
one has a larger variance. Which would we prefer if both have an identical exercise
price, X? Recalling that option holders gain only when the stock price is greater than
the exercise price, it becomes clear that we will prefer the option on the security that
244 PRICING CONTINGENT CLAIMS: OPTION PRICING THEORY AND EVIDENCE
f(S)
Figure 8.2
Hypothetical distributions of stock prices.
has the higher variance because the cumulative probability of receiving a gain is
greater for a security of this sort.' This points out an important difference between
the value of options and the value of the underlying asset. If we hold the asset, we
receive payoffs offered by the entire probability distribution of outcomes. If we are
risk averse, we will dislike higher variance, which means that we will require high
returns along with high variance. On the other hand, if we hold an option, we receive
payoffs only from the tail of the distribution. The contingent-claim feature of options
makes higher variance desirable.
The value of higher variance is also illustrated in the following example. Suppose
that a company has borrowed long-term debt with fixed interest payments of $8000
per year and that it finds itself with one of the two investment projects below:
Project 1 Project 2
Both projects have identical expected cash flows of $5000. However, if the share-
holders accept project 1, the firm will surely go bankrupt because all possible cash
flows are less than the debt commitment of $8000. On the other hand, if they accept
project 2, which has higher variance, there is a 40% chance that they will be able to
pay off their debt obligation and have $2000 left over. Obviously they will choose
the riskier project because it offers them a 40% chance of a positive value. This ex-
ample further illustrates the fact that holders of contingent claims, i.e., holders of
options, will prefer more variance to less. It also introduces the notion that the share-
holders of a firm are really holders of call options on the market value of the firm.
2 This example is given merely as an illustration. The example that follows is more accurate and more
general.
COMBINING OPTIONS, A GRAPHIC PRESENTATION 245
If the value of the firm is less than the required debt payoff (the exercise price on
the option), shareholders will allow their option to expire unexercised and turn over
the assets of the firm to bondholders. If the value of the firm exceeds the debt payoff,
they will exercise their option by paying off the debt holders and keeping any excess
for themselves. In many of the chapters that follow we shall utilize option pricing
theory for applications in corporate financial policy.
The final factor in determining the value of an option is the risk-free rate of in-
terest. Of all the factors it is the least intuitive. Black and Scholes [1973] have shown
that it is possible to create a risk-free hedged position consisting of a long position
in the stock and a short position (where the investor writes a call) in the option. This
insight allows them to argue that the rate of return on the equity in the hedged posi-
tion is nonstochastic. Therefore the appropriate rate is the risk-free rate, and as it
increases so does the rate of return on the hedged position. This implies that the value
of the call option will increase as a function of the risk-free rate of return. The me-
chanics of forming the risk-free hedge, as well as a more precise exposition of the
logic, will be given later in the chapter.
The preceding intuitive description shows that five factors are important in deter-
mining the value of a European option: the price of the underlying asset, S; the exercise
price of the option, X; the instantaneous variance of the returns of the underlying
asset, 62; the time to maturity of the option, T; and the risk-free rate, r f . This may be
written in functional form as
c = f(S, X, a', T,r f ), (8.1)
and the partial derivatives of the call price, c, with respect to its various arguments are:
ac ac ac ac ac >
as > o, ax < o, ao .2
> o, 0. (8.2)
OT > O' O f
One of the most fascinating features of options is that they can be combined in many
different ways to provide almost any desired pattern of payoffs. For the sake of sim-
plicity, assume that European put and call options have the same maturity date and
the same underlying asset, and that the exercise price is set equal to the asset price.'
A graphic representation of the value of buying or selling a call option as a function
of changes in the stock price is given in Fig. 8.3. When selling a call, we receive the
call price now. If the stock price stays the same or falls, the option will mature un-
expired, and we keep the future value of the sale price, + erf T C. This is the horizontal
portion of the dashed line in Fig. 8.3(a) with an intercept at + erf T C. If the stock
price rises, we lose a dollar for each dollar it rises. This is the portion of the dashed
line with an intercept at + erf T C and a slope of —1. Buying a call is the opposite of
3
We also assume that capital markets are frictionless and there are no taxes. This implies, among other
things, that the risk-free borrowing rate equals the risk-free lending rate.
246 PRICING CONTINGENT CLAIMS: OPTION PRICING THEORY AND EVIDENCE
per fT
•
• _pe r fT
e•
•
•_ 4.
Sell a call, —C ♦ / Sell a put, —P
(a) (b)
Figure 8.3
Payoffs from put and call options where S = X.
selling a call. If we sell a put, we receive +P dollars now and lose a dollar for every
dollar the stock price falls below the exercise price. This is represented by the dashed
line in Fig. 8.3(b). The solid line, which represents buying a put, is just the opposite.
The payoffs for long and short positions for stocks and risk-free pure discount
bonds are shown in Fig. 8.4. If we hold a long position in a stock, we gain or lose
a dollar for every dollar the stock price changes. If we hold a bond, we receive the
same payoff regardless of changes in the stock price because a risk-free bond is pre-
sumed to have identical payoffs irrespective of which state of the world obtains.
These elemental securities may be combined in various ways according to the
following relationship:
S + P = B + C. (8.3)
Buying a share of stock and buying a put written on that share yield the same payoff
as holding a bond and buying a call. Alternatively, holding a portfolio made up of
AW AW
Short N.., A Long
in stock7% in stock,
• +S
•
• BerfT Buy a bond, +B
•
•
•
•
■ AS AS
e♦
(a) (b)
Figure 8.4
Payoffs from stock or bond.
COMBINING OPTIONS, A GRAPHIC PRESENTATION 247
S+P—C=B
■ AS
\ Buy put, +P
■
■
Sell call, —C
long positions in the stock and the put and a short position in the call is equivalent
to the perfectly risk-free payoff offered by holding the bond. This can be seen by
simply rearranging Eq. (8.3) as follows:
S P — C = B. (8.4)
Figure 8.5 shows how the payoffs can be combined graphically by adding the payoffs.
The bold dashed line is the payoff from holding a put and a share of stock. On the
graph it is the vertical sum of the payoffs of the two securities. Note that there is no
risk if the stock price falls, because the put gains a dollar in value for every dollar
the stock loses. In the terminology of Wall Street we say that the put insures against
downside losses. If we also sell a call, any upside variability is eliminated, too, and
the result is a perfectly risk-free payoff. Thus a portfolio composed of a long position
in a stock, a put on the stock, and a short position in a call on the stock is completely
risk free.
The reader may use the above graphic analysis to investigate the payoff patterns
of many different securities. One often hears of straddles, spreads, straps, and strips.
They are defined as follows:
• Spread. A combination of put and call options in a single contract, with the
exercise price of the put usually less than the exercise price of the call. Because
of the put-call parity relationship (discussed in the next section) the value of a
spread is less than for a straddle.
• Straddle. A combination of put and call options in the same contract where the
exercise price and maturity date are identical for both options. A straddle is
graphed in Fig. 8.6.
• Straps and strips. Combinations of two calls and one put, and two puts and one
call, respectively.
A straddle loses money for small changes in the stock price and gains money for
large changes. This may seem to be "a good deal," but let us keep in mind that the
248 PRICING CONTINGENT CLAIMS: OPTION PRICING THEORY AND EVIDENCE
AW Figure 8.6
Payoff on a straddle.
AS
• • MEM • =MN • ■IN,
Buy a put, + P
Buy a call, + C
Straddle = C + P
market values of the put and call options are determined in a way that already
incorporates the market's best estimate of the variance in the price of the underlying
security. The greater the variance, the more we have to pay for the put and call
options. Therefore greater price fluctuations will be needed to make a profit. In the
final analysis the securities will always be priced to yield a fair return for their
riskiness.
Option pricing theory has many applications in corporate finance. Black and Scholes
[1973] were the first to point out that the equity in a levered firm is really a call
option on the value of the firm. Later articles have shown how option pricing applies
to many different corporate finance topics such as dividend policy, capital structure,
mergers and acquisitions, investment policy, spinoffs, divestitures, convertible debt
and warrants, and abandonment decisions. Needless to say, the more we know about
option pricing theory, the better we shall understand corporate financial management.
To introduce equity as a call option on the assets of a levered firm, assume that
the firm has only two sources of capital: equity and risky debt. The debt is a zero
coupon bond, has a face value D, and matures T years from now. It is secured by
the assets of the firm, but bondholders may not force the firm into bankruptcy until
the maturity date of the bond. The firm pays no dividends.
We saw, in Eq. (8.3), that any risky portfolio can be constructed from the four
basic building blocks
S + P = B + C. (8.3)
For a CBOE call option the underlying asset is a share of stock, S. For our current
discussion, it is the market value of the firm, V. The equity in a levered firm, S, is really
a call option on the value of the firm. If, on the maturity date, the value of the firm,
V, exceeds the face value of the bonds, D, the shareholders will exercise their call
PUT-CALL PARITY 249
Payoffs at Maturity
IfV<D IfV>D
Shareholders' position:
Call option, S 0 ( V — D)
Bondholders' position:
Risk-free bond
Put option, P — (D — V) 0
Total for bondholders V
Sum of stakeholder positions 0+V =V V— D+D=V
option by paying off the bonds and keeping the excess. On the other hand, if the
value of the firm is less than the face value of the bonds, the shareholders will default
on the debt by failing to exercise their option. Therefore at maturity the shareholders'
wealth, S, is
S = MAX[0, V — D]. (8.5)
V = (B — P) + S. (8.6)
Equation (8.6) tells us that the value of a risky asset, the levered firm, can be parti-
tioned into two parts. The equity position, S, is a call option, and the risky debt
position (B — P), is equivalent to the present value of the risk-free debt, minus the
value of a European put option, P. At maturity the bondholders receive
Table 8.2 shows how the payoffs to equity and risky debt add up to equal the value
of the firm at maturity. We are assuming that there are no taxes and no bankruptcy
costs paid to third parties (e.g., lawyers and the courts). At maturity the entire value
of the firm is divided between bondholders and shareholders. If the firm is successful,
i.e., if V > D, the bondholders receive the face value of the riskless bond, D, and their
put option is worthless. If the firm is bankrupt, they still receive the face value of the
riskless bond, but a put option has in effect been exercised against them because they
lose the difference between the face value of the riskless debt, D, and the market value
of the firm, V. They gain D but lose (D — V); therefore their net position is V, the
market value of the firm in bankruptcy.
The fact that the equity in a levered firm is really a call option on the value of
the firm's assets will provide many insights throughout this chapter and in the re-
mainder of the text. Now let us turn our attention to the problem of how to value
a call option.
250 PRICING CONTINGENT CLAIMS: OPTION PRICING THEORY AND EVIDENCE
E. PUT-CALL PARITY
Table 8.1 shows some securities with both put and call options written against them.
For example, Avon has puts and calls with exercise prices of $45 and $50. We show
below that for European options there is a fixed relationship between the price of
put and call options with the same maturity date that are written on a single under-
lying security. This relationship, derived by Stoll [1969], is called put-call parity. It
implies that if we know the price of a European call on an asset, we can easily deter-
mine the price of a European put on the same asset.
Suppose we have a portfolio where we purchase one share of stock, one put
option, and sell (write) one call option. Both options are written on the share of
stock. Also, they have the same maturity date, T, and the same exercise price, X. At
maturity all states of the world can be divided into those where the stock price is
less than the exercise price, S < X, and those where it is greater than or equal to the
exercise price, S > X. The payoffs from the portfolio in either state are listed below.
If S < X:
a) You hold the stock
b) The call option is worthless 0
c) The put option is worth X—S
d) Therefore, your net position is X
If S > X:
a) You hold the stock
b) The call option is worth —(S — X)
c) And the put option is worthless 0
d) Therefore, your net position is X
No matter what state of the world obtains at maturity, the portfolio will be worth
$X. Consequently, the payoff from the portfolio is completely risk free, and we can
discount its value at the risk-free rate, r f . Using discrete discounting, this is
X
S o + Po — c o =
1 + rf
(1 + r f )S, — X
(co — Po) = (8.8)
1 + rf
Note that the interest rate, rf , is a one-period rate but that the time period need not
equal a calendar year. For example, if the option expires in six months and r is an
11
annual rate, then we can replace (1 + r f ) in Eq. (8.8) with (1 + r 1) '. Equation (8.8)
is referred to as the put-call parity relationship for European options. A special case
occurs when the exercise price, X, is set equal to the current stock price, S. When
SOME DOMINANCE THEOREMS 251
S = X, we have
rSo
Co — Po = >0. (8.9)
1 +/r1
This shows that when all the valuation parameters are identical (the same stock price,
instantaneous variance, exercise price, time to expiration, and risk-free rate) and the
exercise price equals the stock price, the call option will have greater present value
than the put option. It explains why the dashed call line in Fig. 8.6 lies below the
put line.
An equivalent continuous compounding formula for put-call parity is
c o — P o = So — Xe - rf T, (8.10)
where r f is the annual risk-free rate and T is the time to maturity (in years) of the
put and call options. The put-call parity relationship is extremely useful for the val-
uation of European options because if we know the value of a European call, the
put-call parity relationship also gives the value of a corresponding put.
The value of a call option has been described as a function of five parameters: the
price of the underlying asset, S; the instantaneous variance of the asset returns, a 2;
the exercise price, X; the time to expiration, T; and the risk-free rate, r f :
c = f(S, a 2, X, T, r f ). (8.1)
Perhaps even more interesting are some factors that do not affect the value of an
option. For example, the option price does not depend on investor attitudes toward
risk, nor does it depend on the expected rate of return of the underlying security.
This section of the chapter provides a logical, rather than descriptive or intuitive,
framework for understanding why these five parameters affect option value and why
investor attitudes toward risk and the rate of return on the underlying security do
not.
All the following discussion is based on the notion of stochastic dominance,
which was introduced in Chapter 4. We shall use first-order stochastic dominance,
which says that one asset will be preferred by all investors (be they risk averse, risk
neutral, or risk loving) if the return that it offers is superior in every state of nature
to the return offered by a second asset. If this is true, we say that the first asset is
stochastically dominant over the second. Clearly, if all the following analysis is based
on this simple notion, the value of a call option will not depend on individual risk
preferences.
Before developing the analysis and some related theorems it is useful to spell out
in detail the assumptions that have been used in developing valuation models for
252 PRICING CONTINGENT CLAIMS: OPTION PRICING THEORY AND EVIDENCE
options:
• Frictionless capital markets with no transactions costs or taxes and with infor-
mation simultaneously and costlessly available to all individuals.
• No restrictions on short sales.
• Continuous asset trading with all asset prices following continuous and stationary
stochastic processes.4
• Nonstochastic risk-free rate (constant over time).5
• No dividends. 6
Most of these assumptions are self-explanatory and are consistent with efficient capital
markets. By continuous stochastic processes we mean that the price of the underlying
asset can vary over time but does not have any discontinuities or jumps. In other
words, we could graph the price movement over time without lifting our pen from
the paper. A stationary stochastic process is one that is determined in the same way
for all time periods of equal length. In particular the instantaneous price variance
does not change over time. If the underlying asset is a common stock, we assume no
dividend payments so that there are no jumps in the stock price. It is well known
that the stock price falls by approximately the amount of the dividend on the ex-
dividend date.
Our objectives are (1) to show the boundary conditions that limit the values a
call option can take and (2) to prove that American calls on nondividend-paying
stocks will optimally not be exercised prior to maturity. The reader who wishes to
carefully study the theorems of option pricing is referred to the seminal work of
Merton [1973b]. We shall adopt the convention that European calls that can be
exercised only at maturity will be written with a lowercase c, whereas American calls
that can be exercised any time will be written with an uppercase C.
The payoff to a call option at maturity is (1) the maximum of zero if the stock
price is less than the exercise price or (2) the difference between the stock price and
the exercise price if the stock price exceeds the exercise price, i.e.,
C > c = MAX[0, S — X] 0. (8.11)
Clearly, the call price must be nonnegative. Also, because the American call, C,
can be exercised prior to maturity, its value must be greater than or equal to the
European call value, c.
Equation (8.11) also tells us that the call price can never fall below (S — X).
Additionally, the option price will never exceed the price of the stock on which it is
written. Even if the exercise price, X, is zero and the option never matures, it can be
worth at most S. Even in this extreme case the option may be worth less than the
stock because shareholders have voting rights, whereas option holders do not.
C(S, T, 2 C ›# C —
Figure 8.7
Boundaries for the value of a call option.
The preceding discussion serves to limit the possible values that option prices
may take relative to the stock price. The results are illustrated in Fig. 8.7. The call
option is a function of the stock price, S; the time to maturity, T; and the exercise
price, C(S, T, 2 . Its value is nonnegative, less than S, and greater than S — 2. Note
that the boundaries C < S and C > S — 2 are 45° lines.
Next, we shall prove that an American call option written on a nondividend-
paying stock will not be exercised before the expiration date of the option. Along
the way we will further bound the value that call options can take and we will see
why the call value will increase when the risk-free rate does. The result we wish to
prove is stated as Theorem 8.1.
Theorem 8.}. An American call on a nondividend-paying stock will not be ex-
ercised before the call expiration date.
To prove this, we first assume that B(T is the current price of a risk-free zero
coupon bond. Given positive interest rates and assuming that the bond pays $1 upon
maturity, we have'
B(T = ($} e - rf T , (8.12)
where r is the one-year risk-free rate and T is the number of years (or fraction of a
year) to maturity. We shall adopt the convention that T, > T2 > • • • > T n ; therefore
0 = B(3o) < B(T 2 < B(T 2 < • • < B(0 = $1.
Now let us consider two portfolios. Portfolio A represents the purchase of one Eu-
ropean call for 3(S, T, 2 dollars and 2 bonds for 2B(T dollars.' Portfolio B is
the purchase of one share of stock for S dollars. Table 8.3 demonstrates the relation-
ship between the terminal values for the two portfolios. If the stock price is less than
This is the continuous discounting version of the more familiar discrete discounting formula
$1
B(T = = ($1)(1 +
(} + rf T
In this proof we have defined T as the time to maturity for the call option.
254 PRICING CONTINGENT CLAIMS: OPTION PRICING THEORY AND EVIDENCE
the exercise price at the expiration date, the option will expire unexercised, with no
value, and portfolio A will be worth X dollars. But since X > S, portfolio A will be
worth more than portfolio B, which is one share of stock. On the other hand, when
the stock price is greater than the exercise price, portfolios A and B have the same
payoff. In any state of nature portfolio A pays an amount greater than or equal to
portfolio B. Therefore, in order to prevent dominance, portfolio A must have a higher
price than portfolio B:
c(S, T, X) + XB(T) > S.
This restriction may be rearranged to obtain
c(S, T, X) > MAX[0, S — XB(T)].
Finally, from (8.12) we have
-
c(S, T, X) > MAX[0, S — e rf T X]. (8.13)
Equation (8.13) applies to a European call, but we have already discussed the fact
that an American call is always at least as valuable as an equivalent European call;
therefore
C(S, T, X) > c(S, T, X) > MAX[0, S — e - rf T X]. (8.14)
Furthermore, if exercised, the value of an American call is MAX[0, S — X], which
is less than MAX[0, S — XB(T)], since B(T) = e_rf T, which is less than one, for pos-
itive r f . Consequently, the holder of an American option can always do better by
selling it in the marketplace rather than exercising it prior to expiration. This is an
important result because European options are much simpler than American options.
Theorem 8.1 further limits the set of feasible prices for call options because the
requirement that
c(S, T, X) > MAX[0, S — XB(T)]
is more restrictive than
c(S, T, X) > MAX[0, S — X].
This is shown in Fig. 8.8. Also, it is now possible to demonstrate, in a plausible
fashion, that the call price will increase when the risk-free rate increases. Suppose
SOME DOMINANCE THEOREMS 255
Figure 8.8
Further limitation of the feasible set of call prices.
the stock price is $50, the exercise price is $30, and the option expires in one year.
If the risk-free rate is 5%, the lower bound on the option price will be $21.46. If the
risk-free rate changes to 10%, the lower bound increases to $22.85. Intuitively, the
call option is worth more because an investor has to pay less today to acquire the
risk-free discount bond that guarantees $1 at the end of the year. This makes port-
folio A in Table 8.3 more valuable relative to portfolio B.
Theorem 8.2 introduces more realism into the analysis by showing what happens
to the value of an American call option when the underlying stock pays a dividend.
Since most firms pay dividends, investors who hold CBOE call options must be
careful; the options are not protected against a drop in value when the underlying
stock falls in price because it goes ex-dividend.
Theorem 8.2. Premature exercise of an American call may occur if the underlying
security (common stock) pays dividends (and if the option is inadequately pro-
tected against the dividend payment).
In December 1976, General Motors stock was selling at around $75 per share.
Call options were outstanding with an exercise price of $60. On the next day, the
company was scheduled to go ex-dividend with a dividend of $3 per share. This
implied that the stock price would fall to approximately $72 per share. CBOE call
options provide no protection against dividend payments, and hence option holders
found themselves with the following dilemma Before the ex-dividend date the option
price could not fall below S — X, or $15. (Why?) On the following day, everyone
knew the stock price would fall to around $72 per share and that the option price
would also fall (it fell to around $12i'). On one day their option was worth $15 and
on the next they knew it would have a lower price. Obviously, the rational thing to
do was to exercise the option just before the stock went ex-dividend.
The rationality of the above example can be demonstrated by assuming that
a security makes a certain dividend payment, Div, on the expiration date of an option.
Consider two portfolios. Portfolio A is one European call and X + Div bonds. Port-
folio B is one share of stock. Table 8.4 shows the terminal values of the two portfolios.
The value of A is greater than that of B when the stock price is less than the exercise
256 PRICING CONTINGENT CLAIMS: OPTION PRICING THEORY AND EVIDENCE
We shall discuss two derivations of the (Black-Scholes) option pricing model (OPM).
The first, a closed-form solution, was provided by Black and Scholes [1973]. They
recognized that given the assumption of frictionless markets and continuous trading
9
The reader who is interested in the valuation of call options written on dividend-paying stocks is referred
to Roll [1977].
DERIVATION OF THE OPTION PRICING FORMULA 257
At the end of one time period the stock price may increase to uS with probability q
or decrease to dS with probability 1 — q. Figure 8.9 provides a simple example where
the current stock price is $20. There is a 50/50 chance (i.e., q= .5) it will increase to
$24 or fall to $13.40 by the end of the period. Note that the downward multiplier for
the stock price, d, must be less than one and greater than zero. This assumption en-
sures that the stock price will not fall below a value of $0, no matter how many time
periods are eventually added. If there are n periods, then
lim d" = 0 iff 0 < d < 1.
n co
Of course, there is no upper bound on the value that the stock price may take.
Next, denote 1 + r f as one plus the riskless rate of interest over the single time
period (1 + r f = 1.1 in our example). The derivation requires that u > 1 + r f > d. If
1 — q
dS = $13.40
258 PRICING CONTINGENT CLAIMS: OPTION PRICING THEORY AND EVIDENCE
cu = MAX[0, uS —X] = $3
1 —q
cd = MAX[0, dS — X] = $0
Figure 8.10
Payoffs for a one-period call option.
these inequalities did not hold, then riskless arbitrage opportunities would exist. Also,
for convenience, we assume that r f > 0.
Now, imagine a call option c, with an exercise price of X = $21 written on
the stock. The payoffs for the call are shown in Fig. 8.10. Given our numerical
example, there is a 50/50 chance of ending up with MAX[0, uS — X] = $3, or
MAX[0, dS — X] = $0. The question is, How much would we pay for the call right
now?
To answer this question, we begin by constructing a risk-free hedge portfolio
composed of one share of stock, S, and m shares of a call option written against the
stock. Figure 8.11 shows the payoffs for the hedge portfolio. If the end-of-period pay-
offs are equal, the portfolio will be risk free. Equating the payoffs, we have
uS — /mu = dS — mc d ,
and solving for m, the number of call options to be written against the share of stock,
we have
S(u — d)
m= (8.16)
Cu — Cd
Substituting the numbers from our example problem, we see that the hedge ratio
is
$20(1.2 — .67)
m= = 3.53.
$3 — $0
Thus the riskless hedge portfolio consists of buying one share of stock and writing
3.53 call options against it. The payoffs in the two states of nature are identical as
shown below.
Before we can determine what rate of return this payoff represents, we must figure
out the call price, c, in order to know exactly what our original investment was.
♦
uS — m3 r, Figure 8.11
q The payoffs for a risk-free hedge
portfolio.
S — m3
}—q
dS — m3 d
We also know that because the hedge portfolio is constructed to be riskless, the
current value of the portfolio multiplied by one plus the risk-free rate must equal
the end-of-period payoff. Mathematically, this is
(1 + r f (S — m3 = uS — m3 u ,
S[(} + r f — u] + m3 u
3 = ♦ •♦ (8.17)
in(1 rf
Substituting the hedge ratio, m, into this equation and rearranging terms, we can solve
for the value of the call option:
u
3 = [3ul r f 3d — (} + : (} + r } . (8.18)
u d u —d
It can be simplified by letting
(} + r f — d u — (} + r f
p = and 1 p=
u—d u—d
Thus we have
3 = [p3 u + (} — p 3 3,] ± (} + r f . (8.19)
We shall call p the hedging probability. It is always greater than zero and less than
one, so it has all the properties of a probability. In fact p is the value q would have
in equilibrium if investors were risk neutral. Referring back to Fig. 8.9, a risk-neutral
investor would require only the risk-free rate on an investment in the common stock;
hence
(1 + r f S = quS + (} — q dS,
and solving for q, we have
(1 + r f — d
=
q
Li - 3r
Thus p = q for a risk-neutral investor, and Eq. (8.18), which gives the value of a call,
can be interpreted as the expectation of its discounted future value in a risk-neutral
world. Of course, this does not imply that in equilibrium the required rate of return
on a call is the risk-free rate. A call option has risk similar to that of buying the
stock on margin."
" Buying on margin means that part of the investment in the stock is borrowed. In fact the exact pay-
offs of the call option can be duplicated by buying (c„ — cd )/[(u — d S] shares of stock and [u3 d — do„]m
(u — d (} + r f units of the risk-free bond. See Cox, Ross, and Rubinstein [1979].
260 PRICING CONTINGENT CLAIMS: OPTION PRICING THEORY AND EVIDENCE
Continuing with our numerical example, we can use Eq. (8.19) to solve for the
value of the call option:
c = [pcu + (1 — p)c d ] (1 + r f )
The preceding derivation of the value of a call option depends critically on the
existence of a hedge portfolio and on the fact that the call option must be priced so
that the risk-free hedge earns exactly the risk-free rate of return. If the call had a
higher (or lower) price the hedge portfolio would earn more (or less) than the riskless
rate, and opportunities to earn risk-free arbitrage profits would exist.
There are three interesting features of the call pricing formula:
• It does not depend on q, the probability of an upward movement in the stock
price. Consequently, even though investors might have heterogeneous expecta-
tions about q, they will still agree on the call value relative to its other param-
eters, namely, u, S, X, and rt (in the one-period model). The stock price itself
aggregates investors' diverse opinions regarding q.
• Individuals' attitudes toward risk are irrelevant in deriving the call option for-
mula. All that is required for our results is that people prefer more wealth to less
so that arbitrage profits are eliminated.
• The only random variable on which the call value depends is the stock itself.
It does not depend, e.g., on the market portfolio of all securities.
The next logical step is to extend the one-period model to many time periods in
order to show how an option's time to maturity affects its value. First consider the
two-period graphs of the stock prices and the call option payoffs as shown in Figs.
2 11
8.12 and 8.13. We assume that the two-period risk-free rate is simply (1 + r f ) . Next,
we can employ the one-period option pricing model, Eq. (8.19), to solve for c„ and
2 u 2 S = $28.80
q
uS = $24.00
q q(1 — q)
1—q dS = $13.40 (1
d 2 S = $8.98
Figure 8.12
Stock prices with a two-period binomial process: S = $20, u = 1.2, d = .67.
q
Cu
q(1 — q)
q(1 — q)
C
ud = Cdu = MAX[0, udS —X] = $0
1—q cd
(1 — q) 2
Figure 8.13
Two-period binomial call payoffs: S = $20, u = 1.2, d = .67, X = $21.
c d , the values of the one-period options that started the end of the first period:
As before, we can construct a riskless hedge during the first period to prevent risk-
less arbitrage opportunities. The result gives the following equation for the present
value of the call:
c = [pcu + (1 — p)cd ] + (1 + r f ).
12
One can easily imagine how this iterative technique lends itself easily to a computer program.
262 PRICING CONTINGENT CLAIMS: OPTION PRICING THEORY AND EVIDENCE
possible values the call can have after two time periods:
Another way of looking at Eq. (8.21) is to say that the call value is equal to the ex-
pected two-period payoffs (where the expectation uses the hedging probabilities, p and
1 — p) discounted at the risk-free rate.
Note that the bond price in the "up" state of nature when interest rates go down,
B,„ is greater than when they go up, Bd. Figure 8.15 illustrates the bond valuation
tree. The bond price is stochastic until its maturity because interest rates are.
Suppose we have a call option, with exercise price X = $1000, written on the
default-free bond. Let c, be the market value of the call at period t. Its payoffs, if
exercised, are illustrated in Fig. 8.16. Note that if the call had the same life as the
default-free bond, the call would have the same payout in all final states of nature
and would therefore be worthless. The call on a default-free bond must have a shorter
life than the bond. Hence the call option illustrated in Fig. 8.16 has a two-period life,
whereas the underlying bond matures after three periods.
To value the call option, we have to create a risk-free hedge of one bond minus
the value of m calls written against it. A perfect hedge will give the same payoff in
13One of the problems with the analysis of options on bonds is that when interest rates change, the
present values of all coupons shift in a correlated fashion not captured by our simple binomial process.
Another problem is that the variance of the interest rate process is not stationary. The simple example
given in the text ignores both these important problems.
DERIVATION OF THE OPTION PRICING FORMULA <63
q 2 = .<P u ] f = =4.4%
u] f = =<%
q( 1 q) = .25
] f = 10% ud] f du] f = 10.2%
1 ÿ q .5 = ÿ q)q = .<P
d] f = 8.P%
Figure 8.=4
A îinomial process for interest rates.
D + coup = $==00
Figure 8.=P
Expected îond prices.
Co = $89.07
Figure 8.=6
Payoffs for a call written on a default-free îond assuming the call is
exercised at period t.
We also know that the current value of the hedge portfolio, multiplied îy I + ]ft,
will equal the end-of-period payoff,
(Bt ÿ mc,)(1 + ] ft ) = ÿBd,t+1 + coup — mc,,, ±=. (8.<4)
264 PRICING CONTINGENT CLAIMS: OPTION PRICING THEORY AND EVIDENCE
Substituting the value of m from Eq. (8.23) into Eq. (8.24) and solving for the call
price, c„ we have
Later on in the text, we shall use options on bonds to evaluate variable rate loans
with CAPS, i.e., variable rate loans with an upper bound on the rate of interest.
Many corporate loans and mortgage loans have this format.
The numbers in the rows of Pascal's triangle are the coefficients for a binomial
expansion, say [p + (1 — p)] 2' . For example, if T is 3 and we want to know the
coefficients in [p + (1 — p)] 3, they are 1, 3, 3, and 1 as in 1 • p 3 + 3 • p 2(1 — p) + 3 •
p(1 — p) 2 + 1 • (1 — p) 3 . In general the probability of observing n heads in T trials is
- n,
Pr}rtlY} = (coef)pn(1 —
DERIVATION OF THE OPTION PRICING FORMULA 265
where (coef) is the correct coefficient taken from Pascal's triangle. Thus the prob-
ability of observing two heads in three flips of a fair (i.e., p = .5) coin is
Pr{n = 21 T = 3} = 3(.5)2(.5) = .375, where coef = 3.
For large numbers Pascal's triangle is cumbersome. Another way of figuring out the
coefficient is to use combinatorial notation as shown below:
T!
coef =
n ) (T — n)!n! .
The term in parentheses can be read "The number of combinations of T things chosen
n at a time." For example, what are all the possible ways of getting two heads out
of three coin flips? There should be three different ways of doing it, and they are:
hht, hth, thh. The right-hand term uses factorial notation to compute the number of
combinations. The term T! (read T factorial) is the product of all the numbers from
T down to 1. Thus if we want to compute the number of combinations of two heads
in three coin flips, we have
T! 3! 3.2.1
= 3.
(T — n)!n! (3 — 2)!2! (1)2 • 1
By the way, 0! is always defined as being equal to 1.
The binomial probability of observing n heads out of T trials, given that the
probability of a head is p, can be written
T 7)
=4
The reader who wants to read more on binomial trials is referred to Feller [1968, Chapter 6].
"---\
66) PRICING CONTINGENT CLAIMS: OPTION PRICING THEORY AND EVIDENCE
ÿ é(1 + ]f)-T
LT T! ,
ÿ p)T -n1 (8.28)
n = a (, n)-11-
The second bracketed expression is the discounted value of the exercise price,
X(1 + ] f ) - T multiplied by a complementary binomial distribution B(n M a1 T, p). The
complementary binomial probability is the cumulative probability of having
in-the-money options (i.e., where n M a) where the probabilities are the hedging
probabilities determined by the risk-free hedge portfolio. The first bracketed expres-
sion is the stock price, S, multiplied by a complementary binomial probability. It
may be interpreted in the same way if we let
[u/(1 + ] f )]p and 1 — p' = [d/(1 + ] f )](1 ÿ p).
We then have
n(1 T
p) -n
und T [ u d
p
(1 ] f ) T (1 + ]1) P
(1 + ] 4 (1 P)
= (1)' )n (1 P])T
n
DERIVATION OF THE OPTION PRICING FORMULA 267
The binomial model for the pricing of a European call option can be summarized
as follows:
c = SB(n T, p') ÿ é(1 + ] f ) - T B(n M a1T, p), (8.29)
where
(1 + ] f ) ÿ d = u
P and p' [
uÿd (1 + ] f)
becomes larger, the calendar interval between binomial trials becomes shorter and
shorter until, in the limit, we have a continuous stochastic process.' Note that in
this section of the chapter we are interpreting T as one time interval e.g., 1.6
years—where previously we used T as the number of periods until maturity. A con-
tinuous stochastic process has the stock price constantly changing, so its path can
be drawn without ever lifting pen from paper.
Of particular concern is the way that the annual risk-free rate, r f , the up and
down movements, u and d, and the binomial process are to be handled as the number
of time intervals, n, becomes infinite. If we define r f as the rate of return for one year,
and let j be the rate that is compounded n times in interval T (where T is a fraction
of a year, e.g., six months), then in the limit we have'
n IT
bin (1 + = = (1 + r f ). (8.30)
n 3o n/T
Equation (8.30) shows how an annual rate of interest can be converted into the rate
of interest for a binomial model with n binomial trials per year. Next, we need to
know how the up and down movements, u and d, in a single binomial trial relate to
the annual standard deviation of a stock's rate of return. Cox, Ross, and Rubinstein
[1979] prove the following relationships:
ult7n
u = e
and (8.31)
-7 m
d = e ' ` Tin.
The relationships given in Eq. (8.31) are extremely useful for translating continuous
time variables such as the annualized standard deviation, a, into discrete variables
such as u and d for use in the binomial option pricing model.
The continuous-time option pricing formula, derived by Black and Scholes [1973],
is given below: t '
c = SN(d i ) — Xe -r fT N(d2), (8.32)
where
ln(S/X) + r f T 1 /—
(8.32a)
I- 2 a V T'
d, =
u N/T
c/ 2 = d, — a (8.32b)
15 The binomial formula can also be used to model a jump stochastic process as a limiting case. See Cox,
Ross, and Rubinstein [1979, 254-255] for the derivation. With a jump process the stock price will usually
move in a smooth deterministic way but will occasionally experience sudden discontinuous jumps.
16
For proof, see Appendix A at the end of the book.
7
The appendix to this chapter gives the stochastic calculus derivation.
VALUATION OF AN AMERICAN CALL WITH NO DIVIDEND PAYMENTS 269
The terms N(d i and N(d 2 are the cumulative probabilities for a unit normal variable
z where, e.g., N(— 3o = 0, N(0) = .!, and N(3o = 1.0. Mathematically, this is
where f(z is distributed normally with mean zero and standard deviation one.
The binomial model is rewritten below so that it can be readily compared with
the Black-Scholes model:
-
3 = SB(n T, p] — 2(} + r f B(n > a}T, p . (8.29)
The two equations look very similar. The variables S and 2 are exactly the same
and Eq. (8.30) shows the relationship between (1 + r f -T and e -rf T. Cox, Ross, and
Rubinstein [1979] have proved that as n, the number of binomial jumps per year,
becomes large, the two formulas converge because
B(n > a}T, p] N(d, and B(n > a}T, p N(d 2 .
Thus the binomial option pricing formula contains the Black-Scholes formula as a
limiting case.
Both formulas will be used throughout the remainder of the text, and the student
should be familiar with how to use them. The next section gives a numerical example
using both artificial data and data for Digital Equipment call options. Because Digital
Equipment paid no dividends during the valuation period, the Black-Scholes and
binomial pricing models can be used directly, without further modification.
Figure 8.17
Illustration of N(di).
To evaluate (8.32) we first calculate the value of d 1. The time to maturity, three
months, must be expressed as a fraction of a year, i.e., one fourth of a year. Setting
T = .25, and substituting in the values of the other parameters, we get
ln(50/45) + .06(.25) 1
d, = + 025
2
.12036
+ .1118 = .65.
.2236
Using (8.32b), we can solve for d 2 :
d 2 = d, — o- T = .65 — .2 V.25 = .4264.
Substituting these values back into (8.32), we have
c = SN(.65) — e -rf TXN(.4264).
Recall that N(-) are cumulative probabilities for a unit normal variable. Therefore
N(d 1) is the cumulative probability from minus infinity to + .65 standard deviations
above the mean (which is defined to be zero for a unit normal distribution). The
probability contained in the shaded area of Fig. 8.17 will give us the value of N(d 1).
Table 8.7 ("Areas under the Normal Curve") (which appears at the end of the chapter)
shows that if d 1 = .65, the cumulative probability from the mean (it = 0) to .65 is
approximately .242. If we add this to the cumulative probability from minus infinity
to zero (which equals .5), we get
°
N(d 1 ) = 5-00 f(z)dz + Sod ' f(z)dz
= .5 + .242 = .742.
Repeating this procedure for N(d 2), we get N(d 2) = .6651. Substituting these prob-
abilities into the call valuation formula, we have
c = 50(.742) — e ' 6(.25)(45)(6651)
= 37.10 — .9851(45)(.6651)
= 37.10 — 29.48 = $7.62.
VALUATION OF AN AMERICAN CALL WITH NO DIVIDEND PAYMENTS 271
Stock
Price d1 N(d 1) d2 N(d 2) Call Price Given
$30 —1.63 .052 —1.85 .032 $ .14
40 — .35 .363 — .57 .284 $ 1.93 T = 3 months
50 .65 .742 .43 .665 $ 7.62 r f = .06
60 1.47 .929 1.24 .893 $16.15 a 2 = 20%
70 2.15 .984 1.93 .973 $25.75 X = $45
Figure 8.18
Call pricing example.
Table 8.5 gives the value of the call option for various stock prices and Fig. 8.18
plots the call price as a function of the stock price. Note that the call has little value
until the stock price rises to the point where it is near the exercise price (X = $45).
When the stock price is well below the exercise price, the option is said to be "out-
of-the-money," and the call will not be worth much. On the other hand, when the
stock price is above the exercise price, the option is "in-the-money," and its value
increases until in the limit it reaches S — Xe_rf T for very high stock prices.
When pricing a real-world call it is important to keep in mind (1) that the Black-
Scholes formula cannot be used if the common stock is expected to pay a dividend
during the life of the option, (2) that a CBOE call option is not really a simple option
but rather an option on an option, and (3) that the instantaneous variance is not
stationary over time. If the common stock of a firm is really an option on the assets
of the firm, then a call option written against the common stock is really an option
on an option. The Black-Scholes formula tends to misprice deep out-of-the-money
options and deep in-the-money options. One possible reason is that the simple OPM
does not accurately price compound options, and that the bias increases as the stock
price moves away from the exercise price.
It is also of interest to contrast the call prices obtained from the binomial model
with those given by the Black-Scholes model. Suppose we assume there are only two
time periods (each six weeks long). How closely does the binomial model with only
272 PRICING CONTINGENT CLAIMS: OPTION PRICING THEORY AND EVIDENCE
two periods approximate the Black-Scholes continuous time call price of $7.62, given
a $45 exercise price and three months to maturity?
First, we need to convert the annual effective risk-free rate, 6%, into a semi-
quarterly rate, j. This is done below for T = three months, i.e., .25 years:"
2/.25
1+ ) =1+ .06,
2/.25
j/8 = .731%.
Next, we need to convert the annualized standard deviation, a = .2 =- .4472, into
the up and down variables of the option pricing formula. Using Eq. (8.31) we have
a,/ T In = e.4472,/.25/2 = 1.1713,
u = e
T = e —.4472./.2512 = . 8538.
crs,rTn
d = e-
c u ç $13.89
cd= $2.40
c dd ç M›X[0, d 2 S — X] ç
Figure 8.19
Binomial call payoffs.
18
See Eq. (8.30) or Appendix A at the back of the book.
VALUATION OF AN AMERICAN CALL WITH NO DIVIDEND PAYMENTS 273
Figure 8.20
The binomial option pricing closely approximates the Black-Scholes result.
Therefore
cu = [.4835(23.60) + .5165(5.00)] + 1.00731 = $13.8916.
A similar calculation reveals that
cd = [pc„ + (1 — p)cdd] (1 + r f ) = $2.4000.
Now the above results for cu and cd can be used to solve for the current call value, c:
c = [pc„ + (1 — p)cd ] (1 + r1)
= [.4835(13.8916) + .5165(2.4000)] + 1.00731
= $7.90.
From the same data the Black-Scholes call value was computed to be $7.62. A
two-period binomial approximation is reasonably accurate in this case. However, as
the number of binomial trials is increased, the accuracy of the binomial approxima-
tion improves considerably. Figure 8.20 shows how the binomial approximation ap-
proaches the Black-Scholes answer as n increases. It is fairly easy to write a computer
program to estimate the binomial pricing formula using either the iterative technique
as illustrated above or Eq. (8.29).
by using the average of the bid and ask quotes on U.S. Treasury Bills of approxi-
mately the same maturity as the option. The only missing piece of information is the
instantaneous variance of the stock rate of return. Several different techniques have
been suggested for estimating it (e.g., see Latane and Rendleman [1976] or Parkinson
[1977]). We shall use the implicit variance estimated from one call price in valuing
the others. The implicit variance is calculated by simply using the actual call price
and the four known exogenous parameters in the Black-Scholes formula, Eq. (8.32),
to solve for an estimate of the instantaneous variance. We did this, using the January
45s on Digital Equipment, which were priced at $41: on October 4. The estimate of
instantaneous variance was approximately 7.84% (this is a standard deviation of 28%).
Substituting our estimates of the five parameters into the Black-Scholes valuation
equation, we can estimate the price of the April 45s as follows:
c= SN(d,) — e'f'XN(d 2),
where
r f = .062, T = 199/365, S = $46.75, X = $45, 6 = .28,
ln(S/X) + r f T 1
= + o- \/T, d2 = d, — afT.
2
The estimated call price turns out to be $5.58, whereas the actual call price is $6.00.
If we repeat the procedure for the October 45s (now r f = .059 and T = 17/365), the
VALUATION OF AN AMERICAN CALL WITH NO DIVIDEND PAYMENTS 275
estimated call price is $2.28, whereas the actual price is $2.94. Since both the estimated
prices are lower than the actual prices, our estimate of the instantaneous variance is
probably too low.
The above examples show how the Black-Scholes valuation model may be used
to price call options on nondividend-paying stocks. Roll [1977] and Geske [1979a]
have solved the problem of valuing American calls when the common stock is as-
sumed to make known dividend payments before the option matures.' However,
the mathematics involved in the solution is beyond the level of this text.
A riskless hedge portfolio will contain Qs shares of stock and Q, call options written
against it. Its dollar return per unit time will be approximately
dS dc
Qs (T
c t — Q, ( d t )-
If we write 1/N(d 1) call options for each share of stock (i.e., Q, = 1), the return on
the hedge is approximately zero, as shown below:
( ) 1 (dc)
1• = 0.
dt dc/dS
If we use the Digital Equipment April 45s to hedge against 100 shares of Digital
Equipment common stock, then we would have to write 100 times 1/N(d 1) options.
Computing N(d 1) we have
ln(S/X) + r f T 1
d1 =
o- ± 2 a fi;
ln(46.75/45) + .062(199/365) 1
+ (.28),\/199/365
.28V199/365 2
= .451409.
19
Also see Whaley [1981].
20
Equation (8.33) is the exact solution even though the derivative is complicated by the fact that N(d 1 )
is a function of S. The curious reader is referred to Galai and Masulis [1976] for the math.
276 PRICING CONTINGENT CLAIMS: OPTION PRICING THEORY AND EVIDENCE
Knowledge of put-call parity, Eq. (8.8), and the call option pricing formula is sufficient
to value a European put option. Unfortunately, American put options can be exercised
before maturity. Therefore put-call parity does not hold for them and they must be
evaluated directly.
All known solutions to the American put valuation problem involve computerized
numerical methods. Solutions have been provided by Parkinson [1977], Brennan
and Schwartz [1977], and Cox, Ross, and Rubinstein [1979]. Because we have already
made use of the binomial approach to option pricing in section G of this chapter,
the easiest thing to do is to show how it may be employed to value American puts
on nondividend-paying common stock.
To provide a concrete example, suppose we know the following facts for a two-
period American put:
u = 1.2 = the multiplicative upward movement in the stock price, u > 1 + r f > 1,
d = .6 = the multiplicative downward movement in the stock price, d < 1 + r f ,
1 + r f = 1.1 = one plus the risk-free rate,
PRICING AMERICAN PUT OPTIONS 277
1—q
MAX[0, X — dS] = $20
1 —q
mdS +Pd
278 PRICING CONTINGENT CLAIMS: OPTION PRICING THEORY AND EVIDENCE
u = MAX[0, X — u 2 S] = $0
Figure 8.23
Payoffs for a two-period American put.
or
1 1 (8.35)
Po = [PP d + ( — P)P.] ± ( + rf),
p_ — (1 + rf ) (1 + r f — d
(} = • (8.36)
u —d u— d
This one-period put valuation equation can easily be extended recursively into
a multiperiod framework to derive a binomial pricing equation for European puts
similar to that for European calls (Eq. 8.29). However, our objective in this section is
to price American puts that may be exercised before maturity. Figure 8.23 shows the
payoffs for a two-period American put. Note that the problem is complicated by the
fact that the put holder may decide to exercise his or her option at the end of the first
period if the value from exercising, say, 2 — dS is greater than the market value of the
put, Pd . Given the numbers chosen for our example, this is exactly what will happen.
For example, use the one-period put formula to evaluate Pu and Pd , the put values
at the end of the first period. First, we know that the hedging probability is
u — (1 + r 1.2 — 1.1
P= f = .167.
u—d 1.2 — .6
Pu = [PPud + (1 — P)Puu] ± (1 r f
Given that the stock price has fallen to dS = $30 at the end of the first period, a ratio-
nal investor can either exercise the option and receive X — dS = $50 — $30 = $20 or
hold it, in which case it is worth $15.45. Obviously it is better to exercise early and
not hold the option to the end of the second period. This fact makes it difficult to
come up with a closed-form solution to the American put valuation problem (see
Geske and Johnson [1984]). However, a computer program that uses an iterative
technique starting with the set of possible final payoffs in n periods and working
backward, as illustrated in Fig. 8.23, can solve for the present value of American puts.
The opportunity to exercise early makes an American put option have a value
equal to or greater than its European counterpart. For the numerical example we
have been using, the put option is worth $4.64 if it is an American put and only $3.95
if it is a European put.
All the models discussed so far have assumed that the stock price is generated by a
continuous stochastic process with a constant variance. Changes in the assumptions
about the distribution of stock prices that change the model are discussed in part 1
of this section. There are also various special circumstances (such as unusual types
of contracts) that cause us to consider alterations in the model parameter definitions.
For example, the exercise price might be a random variable or it may change in a
predetermined fashion over the life of the option. Extensions of this type are discussed
in part 2.
dS = ,u dt + dz, (8.37)
where
= the instantaneous expected rate of return (measuring the drift
on the random walk of the stock price through time),
= the instantaneous standard deviation of the rate of return, assumed constant,
dt = a small increment in time,
dz = a Wiener process.
If the stock price does not follow a continuous path through time, then it follows
either a pure jump process or a mixed diffusion -jump model.
280 PRICING CONTINGENT CLAIMS: OPTION PRICING THEORY AND EVIDENCE
The pure jump model was introduced by Cox and Ross [1975]. The stock price
path through time is described as a deterministic movement upon which are super-
imposed discrete jumps. It can be written as
dS = dt + (k — dm
k— 1
= dt + (8.38)
1 —)dt
Adt 0
where
77 = a continuous-time Poisson process,
= the intensity of the process,
k — 1 = the jump amplitude.
Equation (8.38) says that the percentage jump on the stock price is composed of a
drift term, dt, and a term dir that with probability 2 dt will jump the percentage stock
change to (k — 1) and with probability (1 — 2 dt) will do nothing.
The mixed diffusion-jump model, developed by Merton [1976], is something of
a mixture of the continuous and pure jump models. Its plausibility comes from the
intuition that stock prices seem to have small, almost continuous movements most
of the time but sometimes experience large discrete jumps when important new infor-
mation arrives.
The next three models relax the Black-Scholes assumption that the instantaneous
standard deviation is constant through time.
The constant elasticity of variance model was derived by Cox [1975] and Cox
and Ross [1976]. Its mathematical statement is
dS = µS dt + QSa 2 dz, (8.39)
where
a = the elasticity factor (0 < a < 2).
Note that if a = 2, the constant elasticity of variance model reduces to the Black-
Scholes model. For a < 2 the standard deviation of the return distribution moves
inversely with the level of the stock price. The intuitive appeal of such an argument
is that every firm has fixed costs that have to be met regardless of the level of its in-
come. A decrease in income will decrease the value of the firm and simultaneously
increase its riskiness. Either operating or financial leverage may explain the inverse
relationship between variance and stock price.
The compound option model of Geske [1977] shows how to price options on op-
tions. This may seem to be a complex abstraction until one considers that the equity
in a levered firm is really an option on the assets of the firm. An example of this was
given in section D of this chapter. A CBOE-listed call option is therefore an option
on the firm's equity, which in turn is an option on the firm's assets. Thus a CBOE
call option is really an option on an option. In Geske's model a compound call option
is a function of seven variables instead of the usual five variables found in the Black-
EXTENSIONS OF THE OPTION PRICING MODEL 281
Scholes model:
c = f(S, X, r f , T, o, D/V, t B).
The two new variables are D/V, the ratio of the face value of debt to the market value
of the firm, and t B, the time to maturity of the firm's debt. Under certain conditions
the compound option model can be made equal to the constant elasticity of variance
model.
The displaced diffusion model by Rubinstein [1983] focuses on the assets side of
a firm's balance sheet and divides it into a portfolio of riskless assets and risky assets.
Like Geske's model it is a compound option formula and requires two parameters in
addition to those of the Black-Scholes model. They are (1) the instantaneous volatility
of the rate of return on the firm's risky assets and (2) the proportion of total assets
invested in risky assets.
Section K of this chapter discusses the empirical evidence on option pricing for-
mulas. Most of the research along these lines is very recent, and therefore the con-
clusions are tentative. Yet the evidence seems to provide support for the extensions
of the Black-Scholes model over the original model.
21
If the firms were levered the problem would be much the same except that we would have to deal with
options on options.
282 PRICING CONTINGENT CLAIMS: OPTION PRICING THEORY AND EVIDENCE
Note that V2 is the instantaneous proportional variance of the change in the ratio of
the asset prices, SA /SB, and p Ai, is the instantaneous correlation between them. The
option is less (more) valuable if the assets are positively (negatively) correlated.
Another interesting option is the truncated option discussed by Johnson [1981].
Examples of a truncated option are an option on a firm consisting of two divisions,
an option entitling the owner to a choice between two risky assets, and competing
tender offers. Consider the case of an option on a firm with two risky divisions.
Rubinstein's [1983] displaced diffusion model is a special case because it considers
an option on a firm consisting of one riskless and one risky asset. Here we are exam-
ining the value of an option on a portfolio of two risky assets. If A l and A 2 are the
lognormally distributed returns of the two divisions, the payoff to this type of trun-
cated option at maturity is
C = MAX[0, A, + A 2 — X].
The actual solution is too complex for this text; however, the interested reader is
referred to Johnson [1981] for the mathematics of this and three other truncated
option cases. One of the implications is that an option on a portfolio of risky assets
is less valuable than the corresponding portfolio of options.
Merton [1973b] solved the problem of pricing options when the value of the
underlying asset and the risk-free rate are both stochastic. Option problems that as
yet have no known solution are (1) options with stochastic variances and (2) options
with random maturity dates.
Tests of the option pricing model (OPM), are different from those of the CAPM be-
cause options are contingent claims on an underlying asset, the stock price, that is
directly observable. This fact, however, does not eliminate the problem that empirical
tests of the OPM are joint tests of market efficiency and the validity of the model.
In addition there are two practical problems: option prices must be recorded syn-
chronously with prices of the underlying asset, and data must allow unbiased esti-
mation of the OPM parameters.
There are three broad categories of OPM empirical tests. The most obvious are
tests of the absolute price level of options to determine whether model prices are
biased relative to market prices and to investigate the profitability of trading rules
based on portfolios of mispriced options. One difficulty with these tests is that taxes
and transactions costs must be taken into account in order to determine the net profit
from any trading rule. A second form of test is based on violations of option pricing
boundary conditions such as those implied by Theorems 8.1 and 8.2. Significant and
persistent violations of these boundary conditions would imply either market ineffi-
ciency or that the OPM is incorrect. The third form of testing is based on the perfor-
mance of hedge portfolios, i.e., combinations of options and other assets. Riskless
EMPIRICAL EVIDENCE ON THE OPTION PRICING MODEL 283
hedge portfolios that earn returns above the risk-free rate are indications of a failure
of either the OPM being tested or of market efficiency.
C
(A C A S) — (C C S) r At.
OS OS f
The first expression is the dollar return on the hedge portfolio, A VH , where AC is
the change in the value of a call option and where 3C/3S = N(d i ) is the number of
shares multiplied by AS, the change in the price per share. The second expression,
which is subtracted from the first in order to obtain excess returns, is the dollar return
on a risk-free position. Theoretically, the difference between the two terms should be
equal to zero because the portfolio is chosen to be a risk-free hedge. Therefore the
portfolio should have zero beta and earn the risk-free rate.
The option position was maintained throughout the life of the option. The risk-
free hedge was adjusted daily by buying or selling shares of stock in order to maintain
the proportion 0C/8S = N(d,). At the end of each day, the hedged position was as-
sumed to be liquidated so that the daily dollar return could be calculated. The option
position was then immediately reestablished and a new hedge position constructed.
Black and Scholes computed the systematic risk of the hedge portfolio by re-
gressing its excess returns against a market index. The results verified that it has a
beta not significantly different from zero (even though the hedge was not adjusted
continuously).
22 Of course, there is no such thing as perfect dividend protection. For example, if shareholders were to
issue a liquidating dividend equal to the value of the firm's assets, the value of common stock would fall
to zero, and no amount of dividend protection could keep the value of a call option from falling to zero.
284 PRICING CONTINGENT CLAIMS: OPTION PRICING THEORY AND EVIDENCE
Their results showed that (given ex post estimates of actual variances of the re-
turns on the underlying stock over the holding period), in the absence of transactions
costs, buying undervalued contracts and selling overvalued contracts at model prices
produced insignificant average profits. However, when ex ante variances were esti-
mated from past stock price histories, buying undervalued contracts and selling over-
valued contracts at model prices resulted in significant negative excess portfolio
returns. The same procedure, when repeated using market prices, yielded substantial
positive excess returns. These results indicate that the market uses more than past
price histories to estimate the ex ante instantaneous variance of stock returns. But
when actual variances are used in the model, it matches actual option prices quite
accurately.
When the transaction costs of trading in options were included, the implied prof-
its vanished. Therefore even though the option market does not appear to be efficient
before transaction costs are taken into account, there is no opportunity for traders
to take advantage of this inefficiency.
1. Using ex post hedge returns, trading strategies (in the absence of transaction
costs) that were based on the Black-Scholes model earned significant excess
returns.
EMPIRICAL EVIDENCE ON THE OPTION PRICING MODEL 285
Bhattacharya [1983] used CBOE transaction data from August 1976 to June
1977. He looked at three different boundary conditions. An immediate exercise test
was composed of situations where the trader could earn more from exercising imme-
diately than from keeping his or her option alive. For a sample of 86,137 transactions
there were 1,120 immediate exercise opportunities assuming zero transactions costs.
However, after transactions costs, even a member of the CBOE or NYSE would have
realized negative average trading profits. Similar results were obtained for tests of
dividend-adjusted lower bounds using the lower bound for European call options
and for pseudo-American call options.
Taken together, the two studies mentioned above seem to indicate that the Black-
Scholes OPM predicts option prices very well indeed. So well in fact, that excess
returns can only be earned in the absence of transaction costs. However, once trans-
action costs are introduced into the trading strategies, excess profits vanish. This
confirms the usual result that nonmember traders cannot beat the market. Prices are
efficiently determined down to the level of transaction costs.
These hedges are based on the gross terminal profit from engaging in a long or short
hedge constructed from American options and the stock. The terms are as defined
in the put-call parity equation (Eq. 8.8) where rf is the risk-free rate of return covering
286 PRICING CONTINGENT CLAIMS: OPTION PRICING THEORY AND EVIDENCE
the life of the option. The last term in each equation is the terminal value of the divi-
dends where Div; is assumed to be the known nonstochastic dividend paid during
the life of the option, and 6; is the length of time between the dividend and the expira-
tion date of the options. The strongest assumption, of course, is that dividends were
nonstochastic. However, the virtue of a test based on put-call parity is that it is not
necessary to make any assumptions about which version of the option pricing model
is best. If put-call parity holds, then there are no arbitrage profits and the market
is efficient, regardless of how options are valued.
Klemkosky and Resnick find their results to be consistent with put-call parity
and with efficiency for the registered options markets. If $20 is the minimum trans-
actions cost for a member firm to take a hedge position, then only 27% of the hedges
were profitable. If $60 is the minimum transactions cost for a nonmember investor,
then only 7% of the hedges were profitable.
Bhattacharya [1980] avoids difficulties (1) and (2) by creating hypothetical hedge
portfolios based on simulated B-S option values. If a neutral hedge is adjusted daily
(continuously would be even better), the excess return should be zero if the B-S for-
mula is correct. The only observed data inputs are the stock price, the stock price
variance estimated directly from stock data during the life of the hedge, and the risk-
free rate. Bhattacharya's results show no operationally significant mispricing by the
B-S formula except for at-the-money options very close to maturity where the B-S
model overvalues options.
the Black-Scholes model Eq. (8.32). Then, by assuming that the B-S model correctly
prices at-the-money options with at least 90 days to expiration, they are able to esti-
mate the percent deviation of observed call prices from B-S call prices. They conclude
that:
1. The Black-Scholes model prices are on average less (greater) than market prices
for in-the-money (out-of-the-money) options.
2. The extent to which the Black-Scholes model underprices (overprices) an in-the-
money (out-of-the-money) option increases with the extent to which the option
is in-the-money (out-of-the-money) and decreases as the time to expiration
decreases.
3. The Black-Scholes model prices of out-of-the-money options with less than 90
days to expiration are, on the average, greater than market prices; but there does
not appear to be any consistent relationship between the extent to which these
options are overpriced by the B-S model and the degree to which these options
are out-of-the-money or the time to expiration.
The second MacBeth and Merville paper [1980] compares the Black-Scholes
model against the constant elasticity of variance (CEV) model. The primary difference
between the two models is that the B-S model assumes the variance of returns on
the underlying asset remains constant, whereas the constant elasticity of variance
model assumes the variance changes when the stock price does. Empirical evidence
on the relationship between the level of stock prices and the rate of return variance
is somewhat mixed. Blattberg and Gonedes [1974] suggest that the variance may
change randomly through time. Rosenberg [1973] finds that it follows an autoregres-
sive scheme. And Black [1976] observes that the variance of returns varies inversely
with stock prices. For the six securities that MacBeth and Merville studied the vari-
ance relative to the stock price seems to decrease as the stock price rises. Using their
estimates of the constant elasticity of variance, they find that the Cox model fits the
data better than the Black-Scholes model. Their empirical results are also consistent
with the compound option model of Geske [1979b].
Beckers [1980] also compares the constant elasticity of variance model with the
Black-Scholes model. First, however, he uses 1253 daily observations (September 18,
1972, to September 7, 1977) for 47 different stocks to test the Black-Scholes assump-
tion that the stock variance is not a function of the stock price. The data reject this
hypothesis. The variance was an inverse function of the stock price for 38 of the 47
stocks a result consistent with the constant elasticity of variance model. This is
sometimes called the volatility bias. When testing simulated Black-Scholes call prices
against the constant elasticity of variance prices, Beckers found that the CEV model
gives higher option prices than B-S for in-the-money options. This is consistent with
the empirical work of MacBeth and Merville, who found that the B-S model under-
values in-the-money options.
Geske and Roll [1984a] use transactions data for all options traded at midday
on August 24, 1976 (a sample of 667 different options on 85 stocks). A subsample of
288 PRICING CONTINGENT CLAIMS: OPTION PRICING THEORY AND EVIDENCE
119 options on 28 different stocks with zero scheduled dividends during their remain-
ing life was identified within the main sample. Using regression analysis, Geske and
Roll demonstrate that the original time, in versus out-of-the-money, and volatility
biases are present in the entire sample. Next, they show that in the nondividend
sample the time and "money" biases are significantly reduced but the volatility bias
remains large. However, by correcting the volatility estimates of all stocks by using
a "Stein shrinker" technique, the volatility bias is reduced. Geske and Roll conclude
that the time and money biases may be related to improper model treatment of early
exercise (the dividend problem), whereas the volatility bias problem may be more
related to statistical errors in variance parameter estimates.
6 Rubinstein-Nonparametric Tests of
Alternative Option Pricing Models
The five extensions of the option pricing model, discussed in section J, have been
compared by Rubinstein [1985]. He used the MDR (market data report) data base
of the Chicago Board of Options Exchange, which has been consolidated into the
Berkeley Options Database. The data is a time-stamped record, to the nearest second,
including option trade prices, quotes, and volume, coupled with the stock price at
the corresponding time during the day.
Rubinstein's experimental design was to select matched pairs of options, e.g., all
options belonging to the same stock, observed on the same day during the same con-
stant stock price interval, having the same exercise price, and falling within a pre-
determined range of out-of-the-money values (e.g., S/X = .75 to .85). Pairing in this
case was on the basis of different times to maturity. For example, one option might
fall within the 71 to 120 day range and another on the 171 to 220 day range. There
were actually 373 of these particular matched pairs (overall there were 19,094 pairs
based on differences on time to expiration and 12,239 pairs based on equal time to
maturity but different exercise prices). If the Black-Scholes formula is unbiased there
should be a 50/50 chance that the implied option variance for the shorter maturity
option is higher than for the longer maturity option. In fact, 94.1% of the shorter
maturity options had higher variance. Thus in this case the Black-Scholes formula
could be rejected.
Nonparametric tests of options paired either by differences in time to maturity
or in exercise prices were performed for two time periods: (1) August 21, 1976, to
October 21, 1977, and (2) October 24, 1977, to August 31, 1978. Two interesting con-
clusions were found. First, if the time to maturity is held constant, then the Black-
Scholes model is biased, but the direction of the bias is different in the two time
periods that were investigated. During the 1976-1977 interval in-the-money options
were undervalued by the Black-Scholes formula. This confirms the work of MacBeth
and Merville [1979]. The direction of the bias is reversed during the 1977-1978
period. No one knows why the bias should reverse. The second conclusion was that
although some of the alternative option pricing models (e.g., the displaced diffusion
model) were more compatible with the empirical results, none of them were superior
in both time periods.
SUMMARY 289
Studies that have used different versions of the option pricing model to try to
find economically exploitable biases have been unsuccessful in doing so when trans-
actions costs were deducted from trading rule profits. From this one can conclude
that the option pricing models fit observed prices well in an economic sense. Also,
the results are consistent with market efficiency.
On the other hand, there are statistically significant biases in the Black-Scholes
model. This suggests that some alternative (perhaps composite) model can do better.
Much work remains to be done in testing the OPM. Rubinstein's [1985] study sug-
gests that none of the alternative models (the jump model, the mixed diffusion-jump
model, the constant elasticity of variance model, the compound option model, or the
displaced diffusion model) can explain all the biases all the time. Also, it remains
unclear which of the observed biases are caused by model misspecification and which
are due to statistical problems in estimating the model parameters.
SUMMARY
Closed-form solutions to the option pricing problem have been developed relatively
recently. Yet their potential for application to problems in finance is tremendous.
Almost all financial assets are really contingent claims. For example, common stock
is really a call option on the value of the assets of a levered firm Similarly, risky
debt, insurance, warrants, and convertible debt may all be thought of as options.
Also, option pricing theory has implications for the capital structure of the firm, for
investment policy, for mergers and acquisitions, for spin-offs, and for dividend policy.
Much of the rest of this book is devoted to exploring applications of the theories dis-
cussed so far: state-preference theory, the capital asset pricing model, arbitrage pricing
theory, and option pricing. As we shall see, option pricing plays a major role in
shaping our thinking
We have established that option prices are functions of five parameters: the price
of the underlying security, its instantaneous variance, the exercise price on the option,
the time to maturity, and the risk-free rate. Only one of these variables, the instan-
taneous variance, is not directly observable. Even more interesting is the fact that
the option price does not depend (1) on individual risk preferences or (2) on the
expected rate of return on the underlying asset. Both results follow from the fact that
option prices are determined from pure arbitrage conditions available to the investor
who establishes perfectly hedged portfolios.
Much remains to be done to empirically test the validity of the option pricing
model in general and of various versions of it such as the Black-Scholes model, the
jump diffusion model, the constant elasticity of variance model, etc. The empirical
results thus far tend to accept the option pricing model in the sense that differences
between the prices it predicts and observed market prices are not economically sig-
nificant. On the other hand, statistically significant departures from the Black-Scholes
model have been discovered, but as yet no single superior model has been found.
290 PRICING CONTINGENT CLAIMS: OPTION PRICING THEORY AND EVIDENCE
z .00 .01 .02 .03 .04 .05 .06 .07 .08 .09
0.0 .0000 .0040 .0080 .0120 .0160 .0199 .0239 .0279 .0319 .0359
0.1 .0398 .0438 .0478 .0517 .0557 .0596 .0636 .0675 .0714 .0753
0.2 .0793 .0832 .0871 .0910 .0948 .0987 .1026 .1064 .1103 .1141
0.3 .1179 .1217 .1255 .1293 .1331 .1368 .1406 .1443 .1480 .1517
0.4 .1554 .1591 .1628 .1664 .1700 .1736 .1772 .1808 .1844 .1879
0.5 .1915 .1950 .1985 .2019 .2054 .2088 .2123 .2157 .2190 .2224
0.6 .2257 .2291 .2324 .2357 .2389 .2422 .2454 .2486 .2517 .2549
0.7 .2580 .2611 .2642 .2673 .2704 .2734 .2764 .2794 .2823 .2852
0.8 .2881 .2910 .2939 .2967 .2995 .3023 .3051 .3078 .3106 .3133
0.9 .3159 .3186 .3212 .3238 .3264 .3289 .3315 .3340 .3365 .3389
1.0 .3413 .3438 .3461 .3485 .3508 .3531 .3554 .3577 .3599 .3621
1.1 .3643 .3665 .3686 .3708 .3729 .3749 .3770 .3790 .3810 .3830
1.2 .3849 .3869 .3888 .3907 .3925 .3944 .3962 .3980 .3997 .4015
1.3 .4032 .4049 .4066 .4082 .4099 .4115 .4131 .4147 .4162 .4177
1.4 .4192 .4207 .4222 .4236 .4251 .4265 .4279 .4292 .4306 .4319
1.5 .4332 .4345 .4357 .4370 .4382 .4394 .4406 .4418 .4429 .4441
1.6 .4452 .4463 .4474 .4484 .4495 .4505 .4515 .4525 .4535 .4545
1.7 .4554 .4564 .4573 .4582 .4591 .4599 .4608 .4616 .4625 .4633
1.8 .4641 .4649 .4656 .4664 .4671 .4678 .4686 .4693 .4699 .4706
1.9 .4713 .4719 .4726 .4732 .4738 .4744 .4750 .4756 .4761 .4767
2.0 .4772 .4778 .4783 .4788 .4793 .4798 .4803 .4808 .4812 .4817
2.1 .4821 .4826 .4830 .4834 .4838 .4842 .4846 .4850 .4854 .4857
2.2 .4861 .4864 .4868 .4871 .4875 .4878 .4881 .4884 .4887 .4890
2.3 .4893 .4896 .4898 .4901 .4904 .4906 .4909 .4911 .4913 .4916
2.4 .4918 .4920 .4922 .4925 .4927 .4929 .4931 .4932 .4934 .4936
2.5 .4938 .4940 .4941 .4943 .4945 .4946 .4948 .4949 .4951 .4952
2.6 .4953 .4955 .4956 .4957 .4959 .4960 .4961 .4962 .4963 .4964
2.7 .4965 .4966 .4967 .4968 .4969 .4970 .4971 .4972 .4973 .4974
2.8 .4974 .4975 .4976 .4977 .4977 .4978 .4979 .4979 .4980 ,4981
2.9 .4981 .4982 .4982 .4982 .4984 .4984 .4985 .4985 .4986 .4986
3.0 .4987 .4987 .4987 .4988 .4988 .4989 .4989 .4989 .4990 .4990
PROBLEM SET
8.1 What is the value of a European call option with an exercise price of $40 and a maturity
date six months from now if the stock price is $28, the instantaneous variance of the stock
price is .5, and the risk-free rate is 6%?
PROBLEM SET 291
8.2 What is the price of a European put if the price of the underlying common stock is $20,
the exercise price is $20, the risk-free rate is 8%, the variance of the price of the underlying
stock is .36 (that is, a = .6), and the option expires six months from now?
8.3
a) Graph changes in wealth, A W, vs. changes in the price of the underlying security, AS, for a
portfolio where you sell one call option and sell one put option (both with the same X, T, a,
andr1). Would this be a good strategy if you have inside information that leads you to expect
the instantaneous variance of the underlying security will increase?
b) Graph A W against AS for a portfolio where you buy a call and sell a put. Would this be a
good strategy if you expect an increase in the instantaneous variance?
8.4 Assume you are a senior financial analyst at Morgan Stanley. You are asked by a client
to determine the maximum price he or she should be willing to pay to purchase Honeywell
call options having an exercise price of $45 and expiring in 156 days. The current price of
Honeywell stock is 448, the riskless interest rate is 7%, and the estimated rate of return variance
of the stock is a 2 = .0961. No dividends are expected to be declared over the next six months.
8.5 Given two European put options that are identical except that the exercise price of the
first put, X 1 , is greater than the exercise price of the second put, X2, use first-order stochastic
dominance and equilibrium in a perfect capital market to prove that one of the puts must
have a higher price than the other. Which put option has the higher price? [Hint: Determine
the relevant states of the world.]
8.6 Consider a firm with current value of $5,000,000 and outstanding debt of $4,000,000 that
matures in 10 years. The firm's asset rate-of-return variance is .5. The interest on the debt is
paid at maturity, and the firm has a policy of not paying cash dividends. Use the OPM to
determine the change in the prices of the firm's debt and equity if there is an unanticipated
rise in the rate of inflation of 5%, which raises the riskless nominal interest rate from 5% to
10%. Which class of security holders benefits from the rise in r f ?
8.7 Figure 8.3 graphs the value of a call option as a function of the value of the underlying
stock. Graph the value of a call option (vertical axis) against
a) a, the instantaneous standard deviation of the returns on the underlying asset;
b) r f , the risk-free rate;
c) T, the time to maturity.
8.8 What are the conditions under which an American put would be exercised early on a stock
that pays no dividends?
8.9 Consider the case of a firm with secured debt, subordinated debentures, and common
stock, where the secured debt and subordinated debentures mature at the same time. Find the
equations for the values of the three classes of securities using the OPM framework. Assume
no dividends or interest payments prior to the debt's maturity and a lognormal distribution
of the future value of the firm's assets, 17"„ as shown in Fig. Q8.9, where V = market value of
the firm, S = market value of the stock, Bs = market value of the senior debt, Bi = market
value of the junior debt, Ds = face value of the senior debt, Di = face value of the junior debt.
8.10 Why will the value of an American put always be greater than or equal to the value of
a corresponding European put?
8.11 Options listed for Digital Equipment were used in the text as an example of option price
estimation using implicit variance. The implicit variance from the January 45 option resulted in
estimated call prices lower than actual call prices for the April 45 and October 45 options.
292 PRICING CONTINGENT CLAIMS: OPTION PRICING THEORY AND EVIDENCE
Prob V t
V = market value of the firm
S = market value of the stock
Bs = market value of the senior debt
Bf = market value of the junior debt
D, = face value of the senior debt
D j = face value of the junior debt
D, D, +D1
Figure Q8.9
Assuming the Black-Scholes OPM is correct, and that all assumptions of the model are met in
the marketplace: What hedge (i.e., riskless) portfolio can be formed to make arbitrage profits
with Digital Equipment April 45 options?
8.12 The share price of Honeybear Inc. is $44.75. Call options written on Honeybear have an
exercise price of $40 and mature in 71 days. The risk-free rate is 62%, and the instantaneous
price variance of Honeybear is 9.61% (i.e., o- = .31). What actions must you take in order to
create a perfect hedge in the following situations:
a) If you own 100 shares of Honeybear stock, how many call options must you buy (sell)?
b) If you own five put option contracts, how many shares of stock do you need?
c) If you own one call contract, how many put contracts do you need?
8.13 After a call contract is created, the outcome must be a zero-sum game; i.e., the call writer
may win or lose $N, but the call buyer will experience an opposite return of exactly $N and
consequently their aggregate payoff is zero. Given this fact, can you explain how they could
both enter into the contract anticipating a positive return?
8.14 Suppose that the government passes a usury law that prohibits lending at more than 5%
interest, but normal market rates are much higher due to inflation. You have a customer, a
Ms. Olsen, who wants to borrow at 20% and can put up her $100,000 store as collateral.
Rather than refusing her request you decide to create a five-year contract with the following
terms: You hold title to the store and receive the right to sell her the store for $X at the end
of five years. If you decide to sell she must buy. In return you give her $80,000 in cash (the
amount she wants to borrow) and the right to buy the store from you for $X at the end of
five years. How can this contract provide you with a 20% annual rate of return on the $80,000?
REFERENCES
Ball, C. A., and W. N. Torous, "Bond Price Dynamics and Options," Journal of Financial and
Quantitative Analysis, 1983, 517-531.
, "On Jumps in Common Stock Prices and Their Impact on Call Option Pricing,"
Journal of Finance, March 1985, 155-173.
Beckers, S., "The Constant Elasticity of Variance Model and Its Implications for Option
Pricing," Journal of Finance, June 1980, 661-673.
REFERENCES 293
, "A Note on Estimating the Parameters of the Diffusion-Jump Model of Stock Returns,"
Journal of Financial and Quantitative Analysis, March 1981, 127-139.
Bhattacharya, M., "Empirical Properties of the Black -Scholes Formula under Ideal Condi-
tions," Journal of Financial and Quantitative Analysis, December 1980, 1081-1106.
, "Transaction Data Tests on the Efficiency of the Chicago Board of Options Exchange,"
Journal of Financial Economics, August 1983, 161-185.
Black, F., "Studies of Stock Price Volatility Changes." Proceedings of the Meetings of the
American Statistical Association, Business and Economics Statistics Section, Chicago,
1976.
, and M. Scholes, "The Valuation of Option Contracts and a Test of Market Efficiency,"
Journal of Finance, May 1972, 399 -418.
, "The Pricing of Options and Corporate Liabilities," Journal of Political Economy,
May—June 1973, 637-659.
Blattberg, R., and N. Gonedes, "A Comparison of the Stable and Student Distributions as
Stochastic Models for Stock Prices," Journal of Business, 1974, 244-280.
Boyle, P., and D. Emmanuel, "Discretely Adjusted Option Hedges," Journal of Financial
Economics, September 1980, 259-282.
Brennan, M., and E. Schwartz, "The Valuation of American Put Options," Journal of Finance,
May 1977, 449-462.
Butler, J., and B. Schachter, "Unbiased Estimation of the Black-Scholes Formula," Journal of
Financial Economics, March 1986, 341-357.
Chiras, D., and S. Manaster, "The Informational Content of Option Prices and a Test of
Market Efficiency," Journal of Financial Economics, June—September 1978, 213-234.
Courtdon, G., "The Pricing of Options on Default-Free Bonds," Journal of Financial and
Quantitative Analysis, March 1982, 75 -100.
Cox, J., "Notes on Option Pricing I: Constant Elasticity of Diffusions," unpublished draft,
Stanford University, September 1975.
; J. Ingersoll; and S. Ross, "A Theory of the Term Structure of Interest Rates," Econo-
metrica, March 1985, 385-407.
Cox, J., and S. Ross, "The Pricing of Options for Jump Processes," Working Paper No. 2-75,
University of Pennsylvania, April 1975.
, "The Valuation of Options for Alternative Stochastic Processes," Journal of Financial
Economics, January—March 1976, 145-166.
, and M. Rubinstein, "Option Pricing: A Simplified Approach," Journal of Financial
Economics, September 1979, 229-263.
Cox, J., and M. Rubinstein, Options Markets. Prentice-Hall, Englewood Cliffs, N.J., 1985.
Feller, W., An Introduction to Probability Theory and Its Applications, Vol. I, 3rd ed. John
Wiley and Sons, New York, 1968.
Finnerty, J., "The Chicago Board of Options Exchange and Market Efficiency," Journal of
Financial and Quantitative Analysis, March 1978, 29 -38.
Fisher, S., "Call Option Pricing When the Exercise Price Is Uncertain and the Valuation of
Index Bonds," Journal of Finance, March 1978, 169-186.
Galai, D., "Tests of Market Efficiency of the Chicago Board of Options Exchange," Journal
of Business, April 1977, 167-197.
29< PRICING CONTINGENT CLAIMS: OPTION PRICING THEORY AND EVIDENCE
, "On the Boness and Black-Scholes Models for the Valuation of Call Options," Jou]nal
of Financial and Quantitative Analysis, March =978, =P-<7.
, and R. Masulis, "The Option Pricing Model and the Risk Factor of Stock," Jou]nal
of Financial Economics, January—March =976, P3-8<.
Garman, M., "An Algeîra for Evaluating Hedge Portfolios," Jou]nal of Financial Economics,
Octoîer =976, 403-4<8.
Geske, R., "The Valuation of Corporate Liaîilities as Compound Options," Jou]nal of Finan-
cial and Quantitative Analysis, Novemîer =977, P4=-PP<.
, "The Pricing of Options with Stochastic Dividend Yield," Jou]nal of Finance, May
=978, 6=7-6<P.
, "A Note on an Analytical Valuation Formula for Unprotected American Call Options
on Stocks with Known Dividends," Jou]nal of Financial Economics, Decemîer =979a,
37P-380.
, "The Valuation of Compound Options," Jou]nal of Financial Economics, March =979î,
63-8=.
, and H. Johnson, "The American Put Option Valued Analytically," Jou]nal of Finance,
Decemîer =984, =P==-=P<4.
Geske, R., and R. Roll, "Isolating the Oîscured Biases in American Call Pricing: An Alternative
Variance Estimator," UCLA Working Paper No. 4-84, Feîruary =984a.
, "On Valuing American Call Options with the Black-Scholes European Formula,"
Jou]nal of Finance, =984î, 443-4PP.
Gould, J., and D. Galai, "Transactions Costs and the Relationship îetween Put and Call
Prices," Jou]nal of Financial Economics, July =974, =0P-=<9.
Hsia, C., "Relationships among the Theories of MM, CAPM, and OPM," mimeograph, UCLA,
=978.
Johnson, H., "Three Topics in Option Pricing," Ph.D. dissertation, UCLA Graduate School
of Management, =98=.
Klemkosky, R., and B. Resnick, "Put-Call Parity and Market Efficiency," Jou]nal of Finance,
Decemîer =979, ==4=-==PP.
Kruizenga, R. J., "Introduction to the Option Contract," reprinted in Cootner, ed., The Random
Cha]acte] of Stock Ma]ket P]ices. MIT Press, Camîridge, Mass., =964, 377-39=.
Latane, H., and R. J. Rendleman, Jr., "Standard Deviations of Stock Price Ratios Implied in
Option Prices," Jou]nal of Finance, May =976, 369-38<.
MacBeth, J., and L. Merville, "An Empirical Examination of the Black-Scholes Call Option
Pricing Model," Jou]nal of Finance, Decemîer =979, ==73-==86.
, "Tests of the Black-Scholes and Cox Call Option Valuation Models," Jou]nal of
Finance, May =980, <8P-300.
Margraîe, W., "The Value of an Option to Exchange One Asset for Another," Jou]nal of
Finance, March =978, =77-=86.
Merton, R., "An Intertemporal Capital Asset Pricing Model," Economet]ica, Septemîer =973a,
867-887.
, "The Theory of Rational Option Pricing," Bell Jou]nal of Economics and Management
Science, Spring =973î, =4=-=83.
, "On the Pricing of Corporate Deît: The Risk Structure of Interest Rates," Jou]nal
of Finance, May =974, 449-470.
REFERENCES 295
Black and Scholes [1973] were the first to provide a closed-form solution for the
valuation of European calls. They recognized that given the assumption of friction-
less markets and continuous trading opportunities, it is possible to form a risk-free
hedge portfolio consisting of a long position in the stock and a short position in the
European call written on that stock. If the stock price changes over time, the risk-
free hedge can be maintained by continuously readjusting the proportions of stock
and calls. The value of the hedge portfolio, VH, can be expressed as the number of
shares of stock, Qs, times the price per share, S, plus the quantity of calls, Q„ times
their price:
The change in the value of the hedge portfolio is the total derivative of (A8.1):
Of course, the stock price moves randomly over time. We assume that it follows a
geometric Brownian motion process. Its rate of return can be described as
dS
dt + 6 dz, (A8.3)
where
296
APPENDIX TO CHAPTER 8: DERIVATION OF THE BLACK-SCHOLES OPTION 297
Since the option's price is a function of the stock's price, its movement over time
must be related to the stock's movement over time. Black and Scholes [1973] show
that if the stock price follows a geometric Brownian motion process such as (A8.3),
then using stochastic calculus (which is far beyond the mathematical capabilities as-
sumed for this text) and employing a technique known as Ito's lemma, one can ex-
press the change in the option price by the following stochastic differential equation:
ac ac a2c
dc = 0,2s2 dt.
dS + dt + (A8.4)
OS at 2 aS2
Note that the only stochastic term in the expression for de is dS. The others are
deterministic.
Substituting (A8.4) into (A8.2), we obtain
ac 1 0c -2 2
d17, = Qs dS + Qc [ dS + dt + 0 5 dd. (A8.5)
OS at 2 02S2
As mentioned earlier, the insight that Black and Scholes provided was to notice
that it is possible to continuously adjust the hedge portfolio, VH, so that it becomes
risk free. How this may be done is illustrated in Fig. A8.1. The curved line, labeled
c(S, T, X), represents the theoretical relationship between the call price and the stock
price. If we buy one share of stock and sell a number of call options equal to the
inverse of the slope of a line tangent to the curve c(S, T, X), we can create a riskless
hedge. For example, let us suppose that the current stock price is $15, and the option
price is $5. We form a hedge portfolio by selling two calls (for which we receive $10)
and by buying one share of stock for $15. Our net equity position would be $5. This
is illustrated in Table A8.1. Next, for the sake of argument, let us assume that the
stock price rises to $20 and the option price to $7i. We gain $5 from the stock and
lose $54 on the two call options. Our net equity has decreased by $.75. A similar
small loss may be expected if the stock price falls by $5. Of course, for smaller changes
in the stock price, our equity loss is even smaller. And if we continuously adjust our
hedge portfolio to maintain a ratio of stock to call options of 1/(0/3S), the hedge will
C=
Figure A8.1
Forming a continually adjusted riskless hedge portfolio.
298 PRICING CONTINGENT CLAIMS: OPTION PRICING THEORY AND EVIDENCE
be perfectly risk free. Therefore the risk-free hedge portfolio will earn the risk-free
rate in equilibrium if capital markets are efficient. This is really the main insight. It
is possible, given continuous trading, to form risk-free hedges with only two securities,
the underlying asset and a call option. This equilibrium relationship is expressed as
dVH
= rf dt. (A8.6)
H
This fact and the fact that the riskless hedge is maintained by purchasing one share
of stock and selling 1/(Oc/3S) calls,
1
(A8.7)
Qs = 1, Q,— (ac/as) ,
allows us to substitute (A8.6) and (A8.7) into (A8.5) in order to simplify it. This is
cone below:
2
dVH = r f dtVH = (1) dS — —0-5- -[ass, dS + atat dt + 21 as2 0 s dtl,
OS 0 c 2- 2
ac ac t a 2c 2
S2 .
at rf VH
OS ) 2 OS 2
Substituting (A8.1) for V, and using (A8.7) again, we have
( al 1 a2c
OC 2 2
-
-- =rf(SQs
Ot as - 2 s 2a S
2
S ac 1 0 C -2 2
= rfc — r
0 S . (A8.8)
f as 2 as 2
Equation (A8.8) is a nonstochastic differential equation for the value of the option.
By using the notion of a riskless hedge, Black and Scholes eliminate the only sto-
chastic term, dS. Now the nonstochastic differential equation may be solved subject
to the boundary conditions that at the terminal date the option price must be
c = MAX[0, S — X]
and that, for any date,
c(S = 0, T, X) = 0.
APPENDIX TO CHAPTER 8: DERIVATION OF THE BLACK-SCHOLES OPTION 299
Black and Scholes [1973] transform the equation into the heat exchange equation
from physics to find the following solution: )
- 2
{1n(S/ X) + [r f + (0- 2m2 ]T} }n(S /X) + [r f — (a /2)] T
3 = S • N e -rf T 2 • N
ci N/T 3r }7"
(A8.9)
where all variables are as defined previously except that
N(• = the cumulative normal probability of a unit normal variable; where, e.g.,
N(— = 0, N(0 = .!, and N(oo) = 1.0;
N(• = ,f z f(z dz, where f(z is distributed normally with mean zero and
standard deviation one.
=
Note that once we are given the call pricing solution [Eq. (A8.9)], we see that the hedge ratio in terms
of the number of calls per share is
Q, = — }}(83l8S = — }}N(d i ,
where N(d i is the first term in braces in Eq. (A8.9). This fact is needed in order to continuously con-
struct risk-free hedges.
9
Futures trading would seem to be one of those marvels that ought to
be invented if it did not already exist. Yet the number of futures
markets is surprisingly small: in the whole world there are probably
not more than 60 or 70 . . . and not more than 40 or 50 commodities
traded on them.
H. S. Houthakker, "The Scope and Limits of Futures Trading," in
Abramovitz, M., ed., The Allocation of Economic Resources. Stanford
University Press, Stanford, 1959, 159
A. INTRODUCTION
In January 1987 the Wall Street Journal listed futures contracts on commodities
(e.g., grains, livestock, food, fiber, metals, petroleum products, and wood), on foreign
currencies (the British pound, Canadian dollar, Japanese yen, Swiss franc, German
mark, and Eurodollar), on financial instruments (Treasury bonds, notes, and bills),
and on stock indices (the Standard and Poor's 500, the NYSE Composite index, the
Value Line index, and the Major Market index). Figure 9.1 shows a sampling of
futures quotations. We can broadly divide futures contracts into three categories:
commodities futures, financial futures, and futures on indices.
One chapter alone cannot cover all the details of every futures contract. Our
objective is to describe the general features of futures contracts (such as standard-
ization, delivery, clearing, open interest, and price limits) in section B, the theory of
the pricing of futures contracts in section C, empirical evidence in section D, and
synthetic futures and options on futures in section E.
B. GENERAL CHARACTERISTICS OF
FUTURES CONTRACTS
1. Definition
When you buy a forward contract, you write a contract today at a stated price—
but pay no cash—for promised future delivery of the underlying asset at a specified
300
GENERAL CHARACTERISTICS OF FUTURES CONTRACTS 301
FUTURES PRICES
Lifetime Open Lifetime Open
Thursday. January 29, 1987. Open High Low Settle Change High Low Interest
Open High Low Settle Change High Low Interest
Open Interest Reflects Previous Trading Day. -FINANCIAL-
-GRAINS AND OILSEEDS-
CORN (COT) 5,000 bu.; cents per bu. Lifetime Open BRITISH POUND (IMM)-25,000 pounds; $ per pound
Mar 1553/4 157 1551/4 156 + 1/4 2421/2 1523/4 49,974 Mar 1.5315 1.5320 7.5275 1.5290 - .0010 1.5390 1.3725 29,282
Open High Low Settle Change High Low Interest
May 1613/4 763 1611/2 1621/4 + 3/) 242 1601/4 28,038 June 1.5140 1.5170 1.5090 1.5120 - .0015 1.5225 1.3600 1,537
July 167 1671/4 166 167 + 1/2 227 1641/2 27,817 -METALS 8 PETROLEUM- Sept 1.4950 1.5010 1.4915 1.4950 - .0015 1.5080 1.3420 198
Sept 1693/4 1703/4 7697/, 7701/2 + V) 2011/2 1681/4 6,951 Est vol 3,682; vol Wed 9,286; open int 31,111, +334.
Dec 3
1741/2 1751/4 174 /4 1751/4 + 4 197 1711/4 18,769 COPPER (CMX)-25,000 lbs.; cents per lb. CANADIAN DOLLAR UMW-100,000 dirs.; S per Can
Mr88 7 7
1813/4 1823/4 1813/4 182 /4 + /4 2033/4 1791/2 2,801 Feb . 60.90 + .45 .._ 0 Mar .7414 .7460 .7412 .7441 -.0016 2474 .6770 15,371
May 1841/2 1851/4 1841/2 1851/4 + 4 195i1 183 642 Mar 60.75 61.25 60.45 61.10 + .45 70.00 58.20 38,217 June .74100 .7437 .7400 .7419 -.0016 .7457 .6995 3,794
Est vol 20,000; vol Wed 20,647; open int 134,991, +372. May 61.25 61.75 60.95 61.70 + .45 70.00 58.20 23,425 Sept .7380 .7412 .7380 .7397 -.0016 .7436 .6950 700
OATS (CBT) 5,000 bu.; cents per bu. July 61.80 62.20 61.50 62.20 + .45 70.00 58.20 7,155 Dec .7368 .7400 .7368 .7375 -.0016 .7431 .6980 419
Mar 1581/2 1591/4 1571/4 1571/2 - 1 1653/4 108 2,685 Sept 62.55 61.55 62.55 62.60 + .45 70.35 59.40 3,142 Est vol 3,742; vol Wed 8,029; open int 20,321, -329.
May 153 153 7/4 1523/4 1523/4 - 1 15774 110 2,398 Dec 62.80 62.95 62.65 63.25 + .45 69.50 60.05 1,949 JAPANESE YEN (IMMI 12.5 million yen; 5 per yen (.00)
July 1413/4 143 1411/2 1421/4 - 11/4 1461/2 1133/4 998 Mom 63.50 63.85 63.45 63.95 + .45 65.65 60.70 297 Mar .6611 .6616 .6574 .6577 -.0045 .6670 .5850 33,497
Sept 1361/4 136 7/4 136 136 - V) 142 1241/2 540 May .... - 64.25 + .45 64.85 60.90 184 June .6650 .6650 .6612 .6613 -.0045 .6707 .6111 2,139
Dec 139 139 1381/4 1381/4 - 1 141 138 116 Est vol 4,800; vol Wed 5,698; open int 74,415, -467. Est vol 12,901; vol Wed 21,188; open int 35,718, -619.
Est vol 500; vol Wed 370; open int 6,732, +30. GOLD (CM)(1-100 troy oz.; S per troy oz. SWISS FRANC (IMM)-125,000 francs-0 per franc
Feb 410.30 413.00 408.60 412.60 + 2.00 451.00 337.30 24,300 Mar .6699 6709 .6659 .6679 -.0031 .6709 .5125 34,352
SOYBEANS (CBT) 5,000 be.; cents per bu. .6736 .6747 .6700 .6717 -.0332 .6747 .5870 1,875
Apr 414.50 417.20 472.60 416.60 + 1.90 456.00 350.00 47,164 June
Mar 4991/4 5011/2 4983/2 5013/43 + 274 576 4831/2 27,706
.6785 .6785 .6745 .6757 -.0032 .6785 .5960 333
489 13,425 June 418.50 421.00 416.50 420.50 + 1.90 456.00 350.50 15,484 Sept
May 4971/2 5001/4 497V) 4993/4 + 21/4 574 Est vol 26,034; vol Wed 35,919; open int 36,566, +331.
19,981 Aug 421.80 424.00 421.20 424.30 + 1.90 460.00 356.00 11,325
July 4971/2 499% 497 4991/4 + 21/4 5771/2 491 W. GERMAN MARK (IMM)-125,000 marks; S per mark
3
21/4 5401/2 4901/2 5,106 Out 425.00 426.00 425.00 428.00 + 1.90 465.00 361.00 8,354
Aug 4953/4 498 495 /4 4973/4 Mar .5626 .5630 .5590 .5604 -.0030 .5661 .4370 60,659
Dec 430.00 430.00 427.10 431.70 4- 1.90 470.00 365.00 72,755
Sept 488 4881/4 4867/4 487./4 + 3 512 4831/2 2,289 .5658 .5658 .5618 .5633 -.0030 .5692 .4850 2,869
7
4791/2 7,179 Fb88 432.50 434.00 432.50 435.60 + 1.90 473.00 371.50 9,541 June
Nov 4821/4 484 /4 4813/4 4843/4 + 23/4 514 Apr Sept .5687 .5687 .5655 .5660 -.0030 .5725 .4868 469
Ja88 488 4901/2 488 4901/2 + 21/2 519 4871/2 341 438.00 438.00 438.00 439.50 + 1.90 479.00 378.00 4,474
3
113 June 441.40 441.40 441.40 443.40 + 1.90 484.00 399.00 4,105 Est vol 26,784; vol Wed 43,910; open int 64,025, -88.
Mar 495 4977/4 495 497 4 + 2 506 495
EURODOLLAR (LIFFE)-51 million; pis of 100%
Est vol 16,000; vol Wed 18,218; open int 76,246, -156. Aug 445.60 445.60 445.60 447.60 + 1.70 483.00 426.00 1,815
Oct .... 452.00 + 1.60 461.50 429.00 979 Mar 93.86 93.89 93.86 93.87 + .03 94.33 90.80 14,122
WHEAT (CBT) 5,000 be.; cents per bu. Est Vol 47,000; vol Wed 68,202; open Int 140,323, -181. June 93.95 93.99 93.95 93.96 + .03 94.15 90.85 7,090
Mar 287 2901/2 2861/2 2901/2 + 21/2 308 243V) 13,301 Sept 93.91 93.94 93.90 93.91 + .04 94.03 91.65 733
May 2681/4 2711/2 2681/4 271 + 2 300 2331/4 3,908 SILVER (CMX)-5,000 troy oz.; cents per troy oz. Dec 93.77 93.80 93.77 93.77 + .05 93.88 91.96 1,803
July 2501/2 2521/4 250 2521/4 + 11/4 255 2231/4 8,272 Mar 563.0 567.0 561.0 565.5 + 2.0 770.0 511.0 37,908 93.57 93,57 93.55 93.56 + .06 93.67 92.08 760
Mr88
Sept 250 251 2493/4 251 .... 2541/2 233 1,512 May 568.5 572.5 567.0 571.3 + 1.0 752.0 520.0 16,326 93.94 93.14 93.31 93.32 + .06 93.39 91.99 502
June
Dec 2561/2 257% 256 2571/2 + 1/2 260 2473/4 402 July 574.0 578.0 572.5 577.0+ 1.0 746.0 524.0 11,116 93.08 93.W 93.07 93.07 + .06 93.13 92.37 147
Sept
Est vol 6,093; vol Wed 3,440; open Int 27,416, +373. Sept 580.0 582.0 580.0 583.0 + 2.0 729.2 532.0 6,656 Ent Vol 5.514: vol Wed 6,186; open int 28,197, +678.
Dec 588.5 593.0 5885 591.9 + 2.0 694.0 543.5 12,745
-LIVESTOCK 8 MEAT- TREASURY BONDS (C87)-0100,000; pis. 32nds of 100%
Mr88 598.5 598.5 598.5 600.8 + 2.0 657.0 552.6 6,608
May 607.2 + 2.2 666.0 567.0 2,895 Mar 100-15 100-16 100-03 100-04 - 2 7.987 + .006 215,791
CATTLE-FEEDER (CME) 44,000 lbs.; cents per lb. July .... .. 613.9 + 2.4 653.0 580.0 1,545 June 99-10 99-17 99-05 99-06 - 3 8.083 + .010 38,227
Jan 66.20 6620 66.05 65.10 - .20 66.40 56.75 577 Sept 621.0 621.0 621.0 620.5 + 2.6 646.0 588.0 Sept 98-19 98-20 98-09 98-09 - 3 8.176 + .010 8,776
412
Mar 67.00 67.45 66.70 66.92 - .15 67.45 56.75 4,728 Est vol 10,000; vol Wed 17,252; open int 96,295, -7,161. Dec 97-22 97-24 97-72 97-12 - 4 8.271 + .014 3,466
Apr 66.05 66.20 65.80 65.90 - .25 66.22 57.00 2,871 Mr88 96-26 96-28 96-16 96-16 - 4 8.363 + .013 2,966
May 64,75 65.00 64.55 64.60 - .25 65.15 58.50 2,851 SILVER (CBTI-1,000 troy or.; cents per troy az. June 95-31 96-01 95-21 95-21 - 4 8.454 + .014 1,772
Aug 63.80 64.05 63.75 63.80 - 25 64.15 58.85 719 Feb 557.0 561.0 556.0 560.0 + 1.0 725.0 510.0 1,663 Sept 95-06 95-00 94-27 94-27 - 4 8.542 + .014 2,234
Sep 63.20 63.25 63.05 63.05 - .15 63.30 57.70 127 Apr 565.0 569.0 563.0 567.0 + 2.0 660.0 519.0 5,289 Dec 94-13 94-14 94-02 94-02 - 4 8.628 + .014 436
Oct 63.05 63.15 62.90 62.90 - .15 63.55 59.45 164 June 571.5 576.0 571.0 575.0 + 3.5 694.0 525.0 3,034 Est vol 145,000; vol Wed 233,552; opn int 273,703, -1,171
Est vol 1,120; vol Wed 2563; open int 12,077, +706. Aug 578.8 580.0 578.0 580.0 + 2.0 638.0 535.0 312 TREASURY NOTES (CBT)-$100,000; pis. 32005 of 100%
HOGS (CME) 30,000 lbs.; cents per ib. Oct 5853 588,0 5855 588.0 + 3.5 650.0 542.0 109 Mar 104-79 104-19 104-14 104-15 - 1 7.361 + .005 53,914
Feb 49.50 50.35 49.30 50.20 + .37 56.00 39.95 7,630 Dec 593.0 595.0 590.0 594.5 + 3.5 650.0 542.0 2,571 June 103-24 103-27 103-22 103-22 - 1 7.470 + .004 6,881
Apr 44.80 45.65 44.65 45.30 + .17 90.75 37.20 10,542 F1388 - 601.0 + 3.5 625.0 569.5 225 Est vol 15,000; vol Wed 17,008; open int 60,861, +589.
.
June 46.75 47.50 46.70 47.30 + .30 51.85 39.00 4,741 Est vol 3,000; vol Wed 1,661; open int 13,248, +5. TREASURY BILLS (1MM)-$1 mil.; pts. of 100%
July 46.80 47.27 46.60 47.05 + .15 50.90 40.20 2,655 CRUDE OIL, Light Sweet (NYM) 42,000 gal.; S per bbl. Discount Open
Aug 44.30 44.80 44.20 44.40 - .17 49.75 42.75 1,892 Mar 18.55 18.71 18.55 18.66 + .08 24.90 10.55 47,257 Open High Low Settle Chg Settle Chg Interest
Out 40.70 40.90 40.45 40.57 - .30 45.70 40.10 471 Apr 18.48 10.58 18.43 78.55 + .09 18.97 10.50 44,223 Mar 94.68 94.69 94.64 94.65 .... 5.35 .... 28,157
Dec 41.55 41.75 40.85 40.92 - .47 45.75 40.55 349 May 18.34 18.41 18.28 78.37 + .07 18.75 10.70 18,813 June 94.77 94.78 94.74 94.74 - 01 5.26 + .01 10,960
Est vol 7,529; vol Wed 10,252; open Int 28,424, +923. AM. 18.16 18.18 18.05 18.76 + .06 18.47 10.70 9,714 Sept 94.76 94.77 94.74 94.73 - .01 527 + .01 1,631
PORK BELLIES (CME) 40,000 lbs.; cents per lb. July 18.01 18.06 77.98 78.03 + .04 18.27 10.85 4,663 Dec 94.67 94.67 94.65 94.62 - .03 5.38 + ,03 1,072
Feb 64.00 6437 63.20 63.25 - .95 81.85 56.40 3,498 Aug 17.90 17.99 17.88 17.96 + .03 18.20 13.95 1,885 Mr88 94.51 94.51 94.44 94,46 - .02 5.54 + .02 613
Mar 62.80 63.45 62.20 62.55 - .57 80.67 56.30 3,679 Sept 17.90 17.95 17.90 17.91 + .03 17.95 15.10 697 June 94.25 - .03 5.75 + .03 236
May 63.50 63.90 62.90 63.12 - .60 79.87 58.00 2,198 Oct 17.90 18.00 17.90 17.91 + .03 18.05 15.40 456 Est Vol 3,121; vol.! Wed 7,768; open int 42731, +327.
July 62.10 62.90 62.10 62.20 - ,37 77.60 59.60 1,570 J088 18.06 -( .03 77.97 17.90 102
Aug 61.05 61.05 60.30 60.60 - .65 74.60 58.90 895 Est vol .15,701; vol Wed 31,653; open int 127,929, +2,556. MUNI BOND INDEX(CBT)41,000; times Bond Buyer MBt
Est vol 5,641; vol Wed 6,907; open int 11,933, +209. HEATING OIL NO. 2 (NYM) 42,000 gal.; $ per gal. Open
Feb .5090 3165 5080 3149 + .0042 .6100 .3450 9,701 Open High Low Settle Chg High Low Interest
Mar .5112 .5270 .5112 .5798 + .0086 .5530 .3225 30,124 Mar 101-05 101-09 101-00 101-04 + 8 102-03 93-05 5,674
Apr .5015 .5090 .5015 .5084 + .0069 .5300 .3250 15,424 June 99-24 180-24 99-17 99-30 + 7 100-28 94-10 285
-FOOD 8 FIBER-
May .4920 .4960 .4910 .4951 + .0053 .5130 .3175 7,574 Est Vol 2,000; vol Wed 3,073; open int 5,959, -245.
COFFEE (CSCEI-37,500 lbs.; cents per lb. June .4825 .4860 .4800 .4851 + '.0053 .5020 3075 2,148 The index: Close 102-14; Yield 7.05.
Mar 126.35 126.40 124.00 124.11 - 3.79 290.33 122.50 8,115 July .4820 .4840 .4820 .4845 + .0055 .5050 .3200 654 S&P 500 INDEX (CME) 500 times index
May 129.45 128.50 126.20 126.55 - 3.45 297.83 124.90 4,366 Aug .4880 .4880 .4860 .4880 + .0040 .4980 .3850 194 Mar 278.80 278.80 272.60 275.15 - 2.40 282.60 227.35 707,279
July 131.15 131.15 120.65 118 70 - 3.95 266.25 127.25 2,244 Sept _. _ - .4975 + .0035 .4980 .4035 141 June 280.00 280.30 274.00 276.75 - 2.55 284.35 228.90 5,516
Est vol 20,108; Vol Wed 23,611; open int 66,042, +359. Sept 281.45 281.75 27630 278.15 - 2.60 285.40 229.90 241
ORANGE JUICE (CTN)-15,010 ms.; cents per lb. Est vol 83,263; vol Wed 73,783; open int 113,053, +2,313.
Mar 118.50 120.50 118.40 119.60 + 1.45 733.80 83.90 5,846 The index: High 276.85; Low 272.54; Close 274.24-1.16
May 119.75 120.65 119.30 120.35 + 1.35 134.80 84.50 2,743 NYSE COMPOSITE INDEX (NYFE) 500 times index
ACC-Amex commodities Corp., CBT-Chicago
July 119.80 121.25 119.80 120.95 + 1.30 135.30 84.75 Mar 15830 15830 155.00 15655 - 7.15 15850 128.93 10,192
645 Board of Trade; CME -Chicago Mercantile Ex-
Sept 721.00 121.50 121.00 121.95 + 1.35 134.70 107.50 June 159.35 15945 155.95 157.50 - 1.20 159.45 131.05 1,701
630 change; CW3- Commodity Exchange, New York; Sept 159.80 160.15 158.00 158.35 - 1.25 160.15 133.55
Nov 121.00 121.50 121.00 122.95 + 1.50 135,50 108.00 402 CRCE -Chicago Rice 8 Cotton Exchange; CSCE - 455
Jan 121.50 121.75 121.50 123.00 + 1.55 135.90 115.00 Est vol 11,077; vol Wed 9,913; open int 12,393, -347.
168 Coffee, Sugar 8 Cocoa Exchange, New York; CTN - The index: High 75744; Low 155.38; Close 156.19 -.53
Est vol 1,200; vol Wed 1,501; open int 10,635, -240. New York Cotton Exchange; IPEL - International Pe.
SUGAR-WORLD (CSCE)-112,000 lbs.; cents per lb. KC VALUE LINE INDEX (KC) SOO times index
troleum Exchange of London; 1MM-- International
Mar 7.89 7.93 7.75 7.85 - .06 9.64 5.75 49,127 Mar 247.90 248.00 243.40 243.70 - 2.90 251.40 210.00 9,478
Monetary Market at CME, Chicago; KC- Kansas City
May 7.95 8.03 7.87 7.93 - .06 9.82 6.00 27,533 Board of Trade; LIEF E - London International Finan- June 245.50 245.50 241.80 241.95 - 3.05 250.20 219.50 1,185
July 7.95 8.05 7.92 7.94 - .08 9.91 6.17 11,202 Est vol 4,000; vol Wed 2,974; open int 10,719, -613.
cial Futures Exchange; MCE MidAmerica Commod-
Sept 8.00 8.00 8.00 7.89 - .14 8.30 6.24 100 The index: High 149.92; Low 248.49; Close 248.97,-.18
ity Exchange; MPLS-Minneapolis Grain Exchange; MAJOR MKT INDEX ( CBT) 8250 times index
Oct 8.05 8.13 8.00 8.00 - .73 9.60 6.41 14,940 NYFE -New York Futures Exchange, unit of New Feb 420.00 422.70 412.70 417.70 - 1.35 430.00 364.10 4,683
Mr88 8.43 8.48 8.37 8.37 - 8.89 7.04 4,640 York Stock Exchange. NYM- New York Mercantile
Est vol 12,586; vol Wed 29,308; open int 107,543, +1,561. Mar 421.00 423.50 413.00 418.35 - 1.65 430.90 341.00 1,859
Exchange; PBOT- Philadelphia Board of Trade; Est vol 12,009; vol Wed 9,226; open int 6,605, -968.
WPG Winnipeg Commodity Exchange. The index: High 421.93; Low 413.67; Close 417.58 -1.08
Figure 9.1
Selected futures contract prices, January 29, 1987. (Reprinted with permission from the
Wall Street Journal 0
Dow Jones and Co. Inc., January 29, 1987. All rights reserved.)
ti me and place in the future. At the time of delivery you receive the asset you purchased
and pay the contract price. Your profit (or loss) on the delivery date is the difference
between the market value of the asset and the contract price. A futures contract is
si milar to a forward contract except that changes in the contract price are settled
daily. There is nothing unusual about forward and futures contracts. In fact they are
302 FUTURES CONTRACTS AND MARKETS
commonplace. For example, when you contract for purchase of a car that will be
delivered six weeks from now (at the dealer's lot), you are buying a forward contract.
If the future value of the car is known with certainty, the current value of the forward
contract is easy to determine. For example, suppose you know the car will be worth
$10,000 six weeks from now; then the current price of the contract, o F T , is
O F T = E(S T ), (9.1)
where T is the time until delivery of the asset, and E(ST) is the expected price of the
asset (the car) on the delivery date.' For our simple example, since there is no un-
certainty, the contract price is $10,000. If there were a secondary market for futures
contracts on cars, you could resell your contract for future delivery of the car, and
the futures contract price would be $10,000.
Alternately, if you were able to purchase the car immediately, you would pay the
spot price, which is the current market price of the commodity. The spot price today
would be the expected spot price six weeks from now, discounted back to the present
at the continuous annual risk-free rate, r f = 10%, since there is no uncertainty:
So = E(S T)e —r fT
= 10,000e -10(6 / 52)
= 10,000(988528)
= $9,885.28.
One pays the spot price for immediate delivery and the futures price at the time of
future delivery.
When the forward/futures contract matures, six weeks from now, you receive a
car worth $10,000 from the dealer and pay $10,000 in cash. Your net profit at delivery
is zero because there was no uncertainty in our simple example. Had the future value
of the car been uncertain, you could have gained or lost the difference between the
market value of the car at the delivery date, T, and the $10,000 contract price that
you must pay at T. In section C we shall examine some of the complications that
arise in pricing futures contracts when the future commodity price is uncertain, when
storage of the commodity is costly, and when taxes must be taken into consideration.
2. Standardization
Of course, if a secondary market existed, other buyers would like to know exactly
what kind of car they would be buying (e.g., a blue four-cylinder Chevrolet with
bucket seats and a four-speed manual transmission). The same is true of actual futures
markets. The asset to be delivered must be standardized as much as is practical.
1
Notation shall be standardized as follows: The T period futures (or forward) contract priced at time t
will be written as ,F T , and the spot price at time t as S,. Whenever it is useful to subscript interest rates,
the interest rate between t and T will be written as ITT.
GENERAL CHARACTERISTIC OF FUTURES CONTRACTS 303
Standardization helps to make the market large enough to attract active trading and
to provide liquidity.
As an example of standardization the New York Mercantile Exchange (NYMEX)
defines a contract in "light, sweet" crude oil as follows:2
1. The seller agrees to deliver 1,000 U.S. barrels (at 60°F. with a tolerance of ±2%)
of "light, sweet" crude oil meeting the following specifications:
a) Sulfur—.50% or less by weight,
b) Gravity—not less than 34° nor more than 45° API.
Price adjustments can be made for oils departing from the aforementioned
specifications.
2. Delivery shall be made FOB (free on board) at any pipeline or storage facility
in Cushing, Oklahoma.
3. Delivery shall take place no earlier than the first calendar day and no later than
the last calendar day of the delivery month. Delivery months are the six consec-
utive calender months following the current calendar month (in Fig. 9.1 these
are the March through August 1987 contracts) as well as the following quarterly
contracts (September, October, and January 1988 in Fig. 9.1) up to 18 months out.
4. Speculative position limits are imposed. No person shall own or control a net
long or short position in all months combined of more than 5000 contracts and in
the month preceding delivery no more than 750 contracts for the delivery month.
Although the definition of the commodity to be delivered seems very precise, to the
extent that variations exist, the seller has an implied delivery option and will, if possible,
deliver the lowest quality product at the latest possible date. The value of the implied
delivery option varies from contract to contract and is implicit in the futures price.'
2
Actual contract terms have been simplified for expositional purposes. Contact the NYMEX for full details.
3
For research on the value of the implied delivery option see Hemmler [1987] or Gay and Manaster
[1984].
-
no further transactions and the market closes at $17.96 per barrel. The Wall Street
Journal reports (see Fig. 9.1) that the market for August contracts opened at $17.90,
the high was $17.99, the low was $17.88, and the closing price was $17.96. The price
change from the previous day's close was $0.03.
At the end of the trading day, each trader's position is marked to market by the
clearinghouse in its daily settlement operation. As shown in Fig. 9.2, Mr. A's net posi-
tion for the day is a gain of $40. This amount of money is credited to his interest-
bearing account at the clearinghouse. On the other hand, Mr. D lost $20 and he must
pay the clearinghouse. In actuality, only brokers belong to the clearinghouse and it is
their accounts that are settled daily. Each brokerage firm then acts as a clearinghouse
for its clients.
GENERAL CHARACTERISTIC OF FUTURES CONTRACTS 305
Open
interest
Time
Contract Contract
begins trading expires
Figure 9.3
Typical pattern for open interest over the life of a futures contract.
Note that four contracts were traded during the day in our example. Actual
trading volume in August contracts is not reported in the Wall Street Journal, but
total trading volume for all oil contracts was estimated at 25,701. In addition to
volume per day, the Wall Street Journal reports open interest, which is the total num-
ber of contracts outstanding as of the previous trading day. In most futures markets
the open interest, as illustrated in Fig. 9.3, is relatively low during the early months
of a contract, when it still has a long time before expiration. Then it rises as hedgers
and speculators become more active in the market, and finally it falls rapidly as the
contract expiration approaches. For example, in crude oil we see that the open in-
terest in January 1988 contracts, which have almost two years to maturity, is only
102, whereas for the March 1987 contracts, open interest is 47,257. In some futures
markets, e.g., stock index futures, there cannot be any open interest at the contract
expiration date because it is not possible to actually deliver the underlying asset the
stock index. Even when the commodity can be delivered, relatively few of the futures
positions (less than 3% on average) end in actual delivery of the commodity involved.
Later on, in the theory section of the chapter, we shall discuss some reasons why
open interest is so high relative to the number of contracts that result in delivery.
4. Margins
The example in Fig. 9.2 implicitly assumed that all traders invested an amount
equal to the full value of the underlying commodity. This is rarely the case. Usually,
the futures trader is required to put up only enough money to insure that the prob-
ability of reaching a negative equity position in one day is quite small. Each futures
market has its own regulations, but the initial margin when a position is first opened
is usually only 5% to 10% of the total value of the contract. The maintenance margin
is the minimum amount of equity that the account may have and is usually set at 75%
to 80% of the initial margin. If losses drive down the value of equity in the account
306 FUTURES CONTRACTS AND MARKETS
below the maintenance margin, then the investor receives a margin call requiring that
additional cash (or interest-bearing certificates) be placed into the account to bring
the equity in the account above the maintenance margin. If the investor fails to meet
the requirements of the margin call, then the broker may close out the investor's
futures position.
There is, of course, nothing that requires an investor to use a margin account.
The effect of trading on margin is to leverage any position so that the systematic and
unsystematic risks are both greater per dollar of investment. High margin has given
commodity futures markets the reputation of being very risky when in fact, as we
shall see later on (in section D), 100% margin positions have about the same variance
as common stock portfolios, although they have very different covariances with the
market portfolio.
There are two commonly cited reasons for having margin requirements. First, a
margin on a futures contract represents a performance bond, which serves to protect
the integrity and reputation of the futures exchange and to protect the middleman
(known as the futures commission merchant) from customer default. Second, it is
often argued (by regulators) that higher margin requirements serve to reduce price
volatility caused by speculative activity.
Hartzmark [1986] analyzes the effect of changes in margin requirements on fu-
tures markets and concludes (1) that when margin levels are raised (lowered) the
number of open contracts falls (rises), (2) that there is weak evidence to support the
conclusion that there is an inverse relationship between margin changes and trading
volume, (3) that there are significant but unpredictable changes in the composition
of traders in the market, and (4) that there is no systematic or significant relationship
between margin changes and price volatility.
5. Price Limits
Another interesting feature of commodities markets is price limits. The U.S. Com-
modity Futures Trading Commission, which regulates trading on U.S. commodity
exchanges, places limits on the extent to which futures prices are allowed to vary
from day to day. For example, a simplified description of the price limits on frozen
concentrated orange juice futures (after 1979) is that: (1) prices may move no more
than $0.05 per pound ($750 per contract); (2) when three or more contract months
have closed at the limit in the same direction for three successive business days, the
limit is raised to $0.08 per pound; and (3) on the last three days before the near
contract's expiration, its limit is $0.10 per pound. It is not unusual for the price to
move up (or down) the limit (i.e., up or down by $0.10 per pound) for several days
without any trading taking place.
There are arguments for and against price limits For example, Roll [1984] notes
that the orange juice futures price is rendered informationally inefficient by the im-
position of price limits on price movements because prices respond to weather changes
(especially freezes, which damage the crop) slower than they otherwise might in the
absence of price limits. Brennan [1986b], however, provides an economic rationale
GENERAL CHARACTERISTIC OF FUTURES CONTRACTS 307
for price limits. Against the clear costs on market participants imposed by prohibiting
mutually beneficial trades at prices outside the price limits, he suggests that a benefit
of price limits is that their imposition allows lower margin requirements than would
otherwise prevail. Margin requirements and price limits are substitutes in ensuring
contract performance without costly litigation. If margin requirements are costly,
then having lower margin requirements is a benefit that results from price limits.
That margin requirements are costly is itself a debated proposition. Black [1976],
e.g., argues that the opportunity cost of margins is zero with daily settlement because
the value of the futures position goes to zero. This argument, however, fails to account
for the costs associated with the initial margin. Others (e.g., Anderson [1981]) believe
that margin positions have no opportunity cost because they can be satisfied with
interest-bearing securities. However, Telser [1981a] provides a sensible argument for
costly margin requirements, namely, that interest-bearing securities such as Treasury
Bills are part of the holder's precautionary balance and if they are committed for
use as margin, they are unavailable for other uses. Brennan [1986b] also points out
that the bonding feature of margin requirements helps to avoid costly litigation that
would otherwise be needed to enforce daily settlement on futures contracts.
Brennan [1986b] is also able to explain why some futures markets have price
limits while others do not. A futures market with daily price limits helps to prevent
default because the losing trader cannot be absolutely certain that defaulting is the
best thing to do. For example, if the price moves down the limit, the trader may be
subject to a maintenance margin call but will not necessarily be wiped out. Further-
more, the trader does not know what the futures price will be when the daily price
limits no longer apply. Therefore he will tend to meet the maintenance margin rather
than defaulting. An analogy (provided by Phil Dybvig of Yale) is that you take your
aging car to the mechanic and ask him to fix it. The car is only worth $4000 and the
mechanic knows it will take $4500 to complete all repairs. Rather than telling you the
total cost of repair, and having you respond by junking the car instead of repairing
it, the mechanic gives you a price limit. He says the first repair will cost $450. Once
this is paid for, he announces that the second repair will cost $450. And so it goes.
Of course, your ability to estimate the total cost of repairs is crucial. The reason the
mechanic can fool you is that on average the total cost of the repair is, let's say, only
$550. It pays you to pay $450 now because you are "saving" $100 on average. But
occasionally you have bad luck and can be persuaded to pay up even when the even-
tual repair is much more. Similar, argues Brennan, is the investor's ability to estimate
what the equilibrium futures price will be when the price limits are lifted. If there is
an active spot market where spot prices are good predictors of futures prices (e.g.,
in interest rate, currency, stock index, and possibly metals futures) price limits will
serve little use because investors can use the spot prices to learn the bad news. Also,
price limits on short-term contracts should be larger or nonexistent because spot and
futures prices are equal at maturity. Note that for orange juice futures the price limit
is doubled for near-maturity contracts. For agricultural commodities futures with
some time before delivery is due, there is usually no spot market (it's hard to trade
orange juice when the crop is still in blossom) and price limits serve a useful role.
308 FUTURES CONTRACTS AND MARKETS
First, it is interesting to take a look at the fundamental question of why futures mar-
kets exist at all. What purpose do they serve, and why have some futures markets
prospered while others have withered and failed? Next, how are futures contracts
valued? Since there are so many different kinds of contracts, we will try to simplify
matters by proceeding in three stages. Initially, we will discuss a generalized model
of futures contracts (provided by Samuelson [1965]) to see how expected futures
prices should be expected to vary randomly through time even though the variance
of futures prices may increase, decrease, or remain constant as the contract life de-
clines. Next we will turn to the pricing of financial futures contracts where arbitrage
with the spot financial contracts helps to determine the market price. Finally, we will
THE THEORY OF FUTURES MARKETS AND FUTURES CONTRACT PRICING 309
discuss commodities futures, where arbitrage in the spot contract is not always so
easy.
The changing variance is important for those who wish to price options on futures contracts. Recall
that the Black-Scholes option pricing model assumes a constant variance; hence it may not do well for
pricing options on futures contracts. See section E of this chapter.
310 FUTURES CONTRACTS AND MARKETS
Spot price, St
•
$80.—
T E(St+T) VAR(Sl+T)
$40 \ , 0 $80 0
1 40 QE
$20 — • 2 20 6E (a 2 + 1)
3 10 01(a 4 a2 1)
$10
•
t+1 t+2 t+3 Time co 0 olj(1 — a2 )
Figure 9.4
A hypothetical spot price that obeys
the autoregressive scheme in Eq. (9.2).
To replicate Samuelson's proof, assume that storage costs and interest rates are
zero, and that the spot price, St+ „, obeys the following stationary autoregressive
scheme:
St+ } = aS, + e„ a < 1. (9.2)
For the purposes of a numerical example, let the constant a equal 1; the error term,
e„ be distributed normally with mean zero and standard deviation, cr,; let the initial
spot price be $80; and let the covariance between the error term and the spot price
be zero, i.e., COV(S„ et ) = 0. Given that a = 2, the expected spot price one period
ahead is one half of the previous period's spot price, as illustrated in Fig. 9.4. Needless
to say, this is an aberrant example because the spot price does not behave randomly.
However, as we shall soon see, the futures price will in fact be random even though
the spot price is not.
Begin by deriving the mean and variance of the spot price. Today's spot price,
S„ is a constant with no variance. Next period's spot price, St+ ,, according to Eq.
(9.2) is
St+i = aS, + e„ (9.3)
and using the properties of random variables (from Chapter 6) its mean and variance
are
E(St+ 1) = aE(S ,) since E(et) = 0 (9.4)
and
V AR(St +1) = E[aSt + et — aE(S t)] 2 (9.5)
= E(02 = 0.
THE THEORY OF FUTURES MARKETS AND FUTURES CONTRACT PRICING 311
The mean and variance of the spot price two periods hence can be derived by sub-
stituting Eq. (9.3) into the definition of St+2 , namely,
Having described the behavior of spot prices, we turn to the problem of pricing
futures contracts. The price, F T , of the futures contract at time t (today) on the com-
modity to be delivered at time T (e.g., three periods hence) in the absence of storage
costs, interest rates, and a risk premium will be the expected spot price in period
three. This may be written as
F 3 = Et(S, ± 3 )
t
= Et(a 3 St a 2 s, +1 + aet+ 2 + + 3)
= a 3 St since Et(a 2 at+ i ) = Et(aet+ 2) = Et(et+3) = 0. (9.8)
The futures price one period hence is still the expected spot price in period three,
but the expectation must be made at t 1; therefore
t+ 1F3 = Et+ i(St+3)
= Et+ ((a 3 St + a 2 rt+1 aEt+ 2 ± et + 3)
3 2
= a St a et+ 1 since Et+ i (aet+ 2) = E(et+ 3 ) = 0. (9.9)
Notice that the expectation, taken at t + 1, of the error term at t + 1 does not vanish
because the error already exists at t + 1.
Next, Samuelson proves that the expected futures price does not change through
time. The change in the futures price from t to t + 1 is
3 2 3
tF3 t+ iF3 = a St + a 8, + — a St , (9.10)
and the expected futures price change, evaluated at time t, is zero since Et(a 2 et 1) = 0.
Thus even though the spot price changes in a known fashion, the futures price is not
expected to change. Given no storage costs or interest rates, it is a random walk with
no drift. The intuition is quite simple. Futures contracts are written for delivery at
a single point in time. Hence the pattern of spot prices is irrelevant. All that counts
is today's estimate of the expected spot price at maturity. For example, in Fig. 9.4
the futures price today for delivery in period t + 2 is $20 even though today's spot
price is $80. Since expected information about the spot price at maturity is random
and unbiased, E(et+T ) = 0, the futures price is a random walk with zero drift.
The variance of the futures price from t to t + 1 is taken from Eq. (9.10) as follows:
VAR[t ,i F 3 — t F3] = Et+1[(a 2 et+ 1)1
= a 4cr„
2 (9.11)
and in general the futures price variance is
YA,
K ,F 3 — ,F 3] 001,
VAR[t+2F 3 — t+1 F 3 ] = (9.12)
VAR[t +3 F 3 — t+2 F 3] =
Thus if a < 1, the variance of the futures price increases as we get closer to the matu-
rity of the contract, but if a = 1, so that the spot price is a random walk, then the
C. THE THEORY OF FUTURES MARKETS AND FUTURES CONTRACT PRICING 313
Prices
Spot price
=
E(S, T
) --i------- September
futures price
= =
= =
I = =
= =
Time
January April September
Figure 9.5
Hypothetical spot and futures prices.
standard deviation of the futures price is (y T, is constant across time, and is equal
to the standard deviation of the spot price.
The basic intuition of Samuelson's model is that if we assume a stationary auto-
regressive process in spot prices, then the variance of futures prices will increase as
the contract nears maturity. Far distant contracts will exhibit relatively lower vari-
ances because autoregressive prices will have a long interval to correct themselves.
Near-maturity contracts will be more variable because prices have little time to cor-
rect. Of course, it may be too much to expect stationarity in the price-generating
process. Many commodities, especially grains, have critical points in their growing
seasons when weather dramatically affects the potential harvest. One might expect
greater variance during these seasons than during other points in time.
Using daily data for 9 commodities between 1966 and 1980, Anderson [1985]
finds evidence supporting both hypotheses the Samuelson hypothesis that the fu-
tures price variance increases nearer maturity and the seasonal production hypothesis
that variance is higher at critical information points. Milonas [1986] uses a slightly
larger data base and, after removing seasonalities, finds a strong maturity effect on
variance for 10 of 11 commodities tested including financial futures (T-bills, T-bonds,
and the GNMA, or Government National Mortgage Association) and metals (copper,
gold, and silver).
The preceding analysis helps to make the point that there is no necessary relation-
ship between today's spot price and today's futures price, which is the expectation
of the spot price at the delivery date.6 For example, the futures price for September
contracts in Fig. 9.5 is the expected September spot price, E(S, „). The January spot
price is below the expected September spot price (i.e., the futures price), and the April
spot price is above it.
6
In section E, we shall prove that, for financial futures at least, the current futures price must be a risk-
adjusted expected spot price.
314 FUTURES CONTRACTS AND MARKETS
Life Yield
Futures contract which requires delivery on March 22, 198X 90 days 8.0%
of a 90-day T-bill maturing on June 20, 198X
T-bill maturing on June 20, 198X t67 12.0
T-bill maturing on March 22, 198X 77 14.0
What position should you take in order to earn an arbitrage profit? First, remember
that the futures contract that matures on March 22 results in the delivery of a 90-day
T-bill (maturing on June 20 with an expected yield of 8.0%). Therefore if you hold
a portfolio consisting of the futures contract and the T-bill maturing on March 22,
' Later on, in section E, we shall return to the problem of pricing financial futures on risky assets.
THE THEORY OF FUTURES MARKETS AND FUTURES CONTRACT PRICING 315
Figure 9.6
Time line for arbitrage example.
your position should be exactly equal to holding the June 20 T-bill, which matures
in 167 days. This equivalence is illustrated in Fig. 9.6. In order for there to be no
arbitrage, the product of the yields to maturity on the two shorter instruments should
equal the yield to maturity on the longer instrument.' If there were no arbitrage,
then
(1.12)167/360,
(1.14)77/360(1.08)90/
360
=
(1.02842)(1.01943) = (1.05398),
1.04840 = 1.05398.
The 167-day yield is high relative to the product of the 77- and 90-day yields. There-
fore the price of the 167-day T-bill is too low and it should be bought. If the face
value on the 167-day T-bill is $1,000,000, today it is worth 9
PV = 1,000,000(1.12) - 167/360
= 1,000,000(94879)
= 948,786.08.
The cash necessary to purchase the T-bill can be raised by borrowing this amount
for 77 days at 14%, which is the 77-day interest rate currently prevailing in the mar-
ket.' Therefore in 77 days you will pay back a future value of
360
FV = 948,786.08(1.14) 77 /
= 948,786.08(1.02842)
= 975,752.39.
In addition to borrowing, you simultaneously sell short a T-bond futures contract
with a $1,000,000 face value. On March 22 when the futures contract matures you
will deliver the 167-day T-bill, which will have 90 days to maturity on March 22, in
order to cover your short position in the futures contract, and you receive the following
The convention in bond markets is to calculate the yield to maturity on the basis of a 360-day year.
9
T-bills pay no interest and are sold on a discount basis.
1°
Capozza and Cornell [1979] have shown that the actual cost of borrowing in this case is approximately
50 basis points above the market yield. In our example this raises the cost of borrowing to 14.5% and
the interest paid by $913.78. Profits will be reduced accordingly.
316 FUTURES CONTRACTS AND MARKETS
amount in cash:
90/360
PV = 1,000,000(1.08)-
= 1,000,000(.98094)
= 980,943.65.
This is more than enough to repay your loan that comes due on March 22. In fact,
your net arbitrage profit on March 22 is
980,943.65 proceeds from the short position
— 975,752.39 amount due on the loan
5,191.26 profit
Thus you earn a $5,191.26 arbitrage profit without taking any risk whatsoever and
without investing any capital of your own. This is called a self-financing riskless
arbitrage. There is no risk involved because you locked in the arbitrage by owning
the 167-day T-bill and simultaneously borrowing and shorting the futures contract
against it. No matter how prices changed in the interim before March 22, your profit
was assured. When the loan came due, you received cash for delivering a T-bill that
you owned. The deal was self-financing because you put up none of your own money.
Thus given the existence of arbitrage between spot and futures markets, financial
futures contracts should be priced according to the expectations theory of Eq. (9.13).
Empirical evidence by Rendleman and Carabini [1979] indicates that when brokerage
costs, bid-ask spreads, and borrowing costs are taken into account, no pure arbitrage
opportunities could be found. They conclude that "the inefficiencies in the Treasury
bill futures market do not appear to be significant enough to offer attractive invest-
ment alternatives to the short-term portfolio manager" Capozza and Cornell [1979]
concluded that near-term contracts were priced efficiently but that longer-term con-
tracts tended to be underpriced; however, none of these discrepancies could have been
arbitraged owing to the cost of shorting the spot bill necessary to establish the appro-
priate position.
Pricing stock index futures is more difficult than pricing interest rate futures for
several reasons." First, the market value of the stock index portfolio is affected by
the fact that the portfolio of stocks pays dividends, but the index portfolio is only a
weighted average of the stock prices. The index does not receive dividends. Therefore
the stock index futures price must subtract the present value of expected dividends
paid by the index stocks before the futures contract matures. Second, futures are taxed
differently than the underlying stock index portfolio. All gains and losses are marked
to market at year's end with 40% taxed at the short-term capital gains rate and 60%
at the long-term capital gains rate. Hence capital gains taxes cannot be deferred on
the futures contract, whereas they may on the underlying securities. Cornell and
French [1983b] show how to price futures on stock indices and test their model on
data taken from the first seven months of stock index futures trading.
11
Stock index futures are traded on the Standard and Poor's 500 index, the New York Stock Exchange
Composite index, and the Value Line index.
THE THEORY OF FUTURES MARKETS AND FUTURES CONTRACT PRICING 317
A. FUTURES PRICES AND STORAGE. The traditional view explains the current
futures price as the expected spot price,minus the cost of storage (interest foregone,
warehousing, and shrinkage), and minus a convenience yield. Costs of storage are
obvious, but convenience yield is much like a liquidity premium, usually being described
as the convenience of holding inventories because many commodities (e.g., wheat) are
inputs in the production process (e.g., bread making) or as the convenience of having
inventory to meet unexpected demand. The theory of storage predicts low convenience
yields when inventories are plentiful and high convenience yields when stockout is
more likely. Telser [1958] and Brennan [1986a] have provided empirical estimates of
the convenience yield that are consistent with the theory. Fama and French [1987]
have provided evidence that marginal convenience yields vary seasonally for most
agricultural and animal commodities but not for metals.
According to the storage theory, the futures price of a T-period contract observed
at time t is given by
tF T = Stet"T? +-tWT — t CT, (9.15)
where Stet"T is the current spot price compounded by the interest rate between the
current time, t, and the delivery date, T; where ,WT is the storage cost between now
and delivery; and where t C T is the convenience yield (in dollars) between now and
delivery.
If storage costs and convenience yields are very low, then we would predict that
prior to delivery the futures price is below the expected spot price,
Prices
•
■■ c
4
4
%6
/26
0 6
•
[•[••[[
E(S T )
so •"•211111...9
0"a1.0
.0 "1.
, a,;01,-
I •I#.-\ bac‘(*
-0
■ Time
T
Figure 9.7
Normal backwardation and contango.
will rise (on average) through time until, at delivery, the futures price equals the spot
price.
Also illustrated in Fig. 9.7 is a situation where the futures price is above the ex-
pected spot price:
t F T > E(S T) = St errT .
Called contango, this is just the opposite of normal backwardation. If hedgers need
to go long, or if the convenience yield is negative owing to oversupply, then they
must pay a premium for futures contracts in order to induce speculators to go short.
Next, write down the definition of a one-period rate of return for an investor who
holds the risky commodity. If Si, is the current spot price of the ith commodity and
E(SiT) is the expected spot price at the time of delivery, T, we have
E(SiT ) — Si,
E(R i) = (9.18)
Si,
Combining Eq. (9.18) with Eq. (9.17), we have a certainty equivalent model for the
spot price of the commodity:
— [E(Rm) — R i016
Sio — E(S (9.19)
1 + Rf
where
COV(R i, R m)
Qi = 2(R.) = the systematic risk of the ith commodity.
0-
Finally, a futures contract allows an investor to purchase an asset now but to defer
payment for one period; therefore the current price of the futures contract, ,F iT, must
be the current spot price multiplied by a future value factor:12
„F iT = + R f ). (9.20)
Multiplying both sides of the certainty equivalent model, Eq. (9.19), by (1 + R1), and
noting that the result is equivalent to Eq. (9.20), we have
The futures price, „F a, equals the expected spot price minus a risk premium based
on the systematic risk of the commodity.
The CAPM approach, Eq. (9.21), argues that systematic risk should be important
in the pricing of futures contracts but leaves out storage costs and convenience yields.
On the other hand, the traditional approach, Eq. (9.15), ignores the possibility that
systematic risk may affect the equilibrium prices of commodity futures contracts. We
now turn to the empirical evidence, which is still in its infancy, to provide some clues
to which theory is correct or whether some combination of them best describes com-
modities futures prices.
D. EMPIRICAL EVIDENCE
There are two interrelated lines of empirical research on futures contracts. One focuses
on comparing models of futures prices to see which best explains the data. The other
asks how well information is reflected in futures prices Are they good forecasts of
future spot prices? These ideas are interrelated because we need a good model to
12
This is the same argument used in Eq. (9.14) for financial futures.
Ç20 FUTURES CONTRACTS AND MARKETS
forecast future spot prices, and îecause the futures markets must îe informationally
efficient if the models are to have any hope of working.
Bodie and Rosansky [=980] provide a comprehensive analysis of rates of return
on commodities futures prices îetween =9P0 and =976. Taîle 9.= summarizes their
results. Perhaps the most interesting fact is that an equally weighted portfolio of <3
commodities futures had roughly the same return and standard deviation as the
equally weighted common stock portfolio." Also, the correlation matrix (Panel B)
indicates that stocks and futures are negatively correlated with each other. Further-
more, common stock returns are negatively correlated with inflation, whereas com-
modity futures are positively correlated. One interpretation of these facts is that
different factors (in the arîitrage pricing model) affect commodities and stocks. A
randomly chosen portfolio of common stock is a îad hedge against unexpected
inflation, îut a well-diversified commodity portfolio is a good hedge. Apparently,
stocks have very different factor sensitivities to the factors in the arîitrage pricing
model (APM) (especially unexpected inflation) than do commodities. If this interpre-
tation is true, then it is not surprising that Bodie and Rosansky [=980] and Dusak
[=973] îoth found that the CAPM does a poor joî of explaining commodities returns.
After all, a market portfolio of common stock is usually employed as the proxy for
=3
None of the futures positions were îought on margin.
EMPIRICAL EVIDENCE 321
the market portfolio (a single-factor model), and commodity returns are negatively
correlated with common stock. Bodie and Rosansky found that 15 out of 23 com-
modities had negative betas and that the security market line, estimated using com-
modities data, had a negative slope. From these results we must conclude that the
CAPM fails to explain (even approximately) the returns on commodities. We are
waiting for someone to apply the APM to this problem.
On a more positive note, Bodie and Rosansky do find evidence of normal back-
wardation, where Dusak had not, because they used a longer time period. Mean
returns in excess of the risk-free rate averaged 9.77% for commodity futures. Chang
[1985] also finds evidence of normal backwardation for wheat, corn, and soybeans
over the interval 1951 to 1980. Fama and French [1987] find marginal evidence of
normal backwardation when commodities are combined into portfolios but conclude
that the evidence is not strong enough to resolve the long-standing controversy about
the existence of nonzero-risk premia. If there is a risk premium for commodity futures
contracts, it is probably not related to the CAPM beta, and it may be time varying.
The second interesting question is whether or not futures prices are good fore-
casts of expected spot prices. The variability of spot prices will depend on seasonal
supply and demand shocks, and on the availability of inventory to cushion them.
One might expect that commodities with high inventory relative to production-
e.g., precious metals (gold, silver, and platinum) allow price shocks to be trans-
mitted freely from one period to another. This implies that a demand shock today
will affect both today's spot price and expected spot prices. If spot prices are linked
in this fashion there is little left for futures prices to explain. French [1986] and Fama
and French [1987] argue that futures prices cannot provide forecasts that are reliably
better than the current spot price unless the variance in the expected spot price
changes is a large fraction of the variance of the actual spot price changes. For metals
this is not true, and they find that futures prices predict expected spot prices no better
than do current spot prices. However, they find reliable evidence that futures prices
are good forecasts for animal and agricultural commodity expected spot prices. This
is consistent with the role of inventories because commodities are affected by rela-
tively large storage costs and production seasonals.
Roll [1984] studied the orange juice futures market. Although the commodity
is frozen and therefore not perishable, only a small amount (about 10%) is carried
over in inventory from one year to the next. Almost all (98%) of U.S. production
takes place in central Florida around Orlando. Short-term variations in supply due
to planting decisions is low (because oranges grow on trees that require 5 to 15 years
to mature) and short-term variations in demand are also low. All these facts imply
that orange juice futures prices should be heavily influenced by weather, particularly
by cold temperatures. Roll reports that the orange juice futures price is a statistically
significant predictor of the forecast error of the U.S. National Weather Service. That
is, futures prices predict the weather better than the weather service. Also, Roll finds
(1) that price limits decrease the informational efficiency of the market and (2) that
there is much more variability in futures prices than can be explained by the weather
or any other measurable short-term supply or demand shock.
322 FUTURES CONTRACTS AND MARKETS
In sum, the existing empirical evidence indicates that inventories are important
in explaining the ability of futures prices to predict expected spot prices. Also, there
is weak evidence to support normal backwardation, but the risk premium may be
time varying and is not related to a CAPM beta. In addition, it was reported earlier
in the chapter that Brennan [1986a] and Fama and French [1987] have found evi-
dence consistent with the existence of convenience yields that vary through time with
inventory levels.
I. Synthetic Futures
We can create a synthetic forward contract by buying a European call, C, on the
underlying risky asset or commodity with time to maturity T and exercise price X,
equal to the forward price, O F T , and simultaneously writing a European put, P, with
the same time to maturity and exercise price. Table 9.2 shows that the end-of-period
payoffs for the synthetic forward contract are equal to the payoffs on an actual
forward contract. At maturity, if the price of the risky asset, S, is less than the exercise
price, X = O F T , then the call is worthless and the put is exercised against us, creating
a loss of O F T — S T dollars. Had we held the forward contract, we would have lost
exactly the same dollar amount. Alternately, if the risky asset is worth more than
the exercise price at maturity, the call pays the difference between the asset price and
the delivery price, S T — O F T , which is the same as the forward contract payout. Either
way, the synthetic forward contract and the real forward contract have identical
payouts.
The initial investments in the synthetic and real forward contracts must also be
identical they must both have zero initial outlay. This fact allows us to establish
an interesting result, namely, the expectations hypothesis for pricing financial futures.
We begin by writing down the formula for put-call parity from Chapter 8:
oC — P = S — Xe-rfT .
o o
Next, our synthetic futures position has an exercise price, X, equal to the current
forward price, O F T . Therefore
Co — Po = So — OF Te-rfT (9.22)
VA = CO PO 0 — (O F ST) ST — OFT
VB = forward contract ST — OFT ST — O F T
VA = VB VA = VB
SYNTHETIC FUTURES AND OPTIONS ON FUTURES 323
OW
Figure 9.8
Constructing a synthetic future.
We also know that the synthetic forward contract requires zero cash outlay; therefore
we can set Eq. (9.22) equal to zero:"
fT 0,
Co Po = So — OF Te r
14
This is the same as Eq. (9.14).
324 FUTURES CONTRACTS AND MARKETS
have emphasized throughout, the synthetic future is really a forward contract since
no daily marking to market is required.
2. Options on Futures
Options are traded on futures on stock market indices, on Treasury instruments,
on foreign exchange rates, and on some metals. Since these options are merely con-
tingent claims on the underlying risky assets, one would think that they could be
priced using the Black-Scholes formula. Unfortunately, this is not quite true. Rama-
swamy and Sundaresan [1985] have shown (1) that even with constant interest rates,
premature exercise may be optimal and (2) that the fact that interest rates are sto-
chastic is crucial for pricing options on futures.
To show why early exercise may be optimal, even with constant interest rates,
it is useful to recognize that an option on a futures contract has the same payouts
as an option on a portfolio with the same random price realizations as the futures
contract but which pays a continuous dividend at the riskless rate of interest. Consider
a generalization of Eq. (9.14), where the futures price was shown to be the spot price
multiplied by a compounding factor:
t F T = SterT. (9.14)
The generalization is to define the rate of compounding as the risk-free rate plus (or
minus) a risk premium, (5:
6
trT = rf(Tt) T,t•
15
Their model assumes the variance of the futures contract price is constant over the life of the contract
but empirical evidence indicates that it increases.
PROBLEM SET 325
SUMMARY
Futures contracts give one the right to receive delivery of a risky asset or commodity
at a predetermined future date at a price agreed upon today. No cash changes hands
at the time of purchase of the contract. Unlike forward contracts, which are usually
bilateral contracts between two parties, futures contracts are marked to market each
day via a clearinghouse. This procedure provides liquidity to the market and allows
open interest to far exceed the quantity of the underlying asset to be delivered.
Most futures contracts are traded on margin, which has an opportunity cost to
investors. Price limits on futures contracts are (imperfect) substitutes for margin levels,
and therefore we can predict that contracts on assets with active spot markets will
have low margins and no price limits, but when spot markets are thin (as with orange
juice futures) price limits will play an important role.
Futures contract prices are determined by storage costs, by convenience yields,
and probably by a risk premium, although the risk premium is not a function of the
CAPM beta. The variance of futures prices appears to increase as the contract comes
closer to maturity. Futures prices provide better forecasts of future spot prices than
do current spot prices for those commodities where inventory levels are relatively
low, e.g., for agricultural and animal commodities.
PROBLEM SET
9.1 Most futures contracts have fairly short lives, usually less than 18 months. Why are there
not futures contracts with longer lives?
9.2 Suppose you observe the following yields on T-bills and T-bill futures contracts on January
5, 1991:
Yield
risk-free rate is 7% per year. You believe the stock price will rise and wish to create a synthetic
forward contract position for delivery of 100 shares six months hence.
a) How do you construct the synthetic futures position; how much must you borrow or lend?
b) What is your expected profit if you believe the share price will be $42 six months from now?
9.5 On January 29, 1987, you could buy a March 1987 silver contract for $5.610 per ounce
and at the same time sell a March 1988 contract for $6.008 an ounce.
a) Exactly what would you have done had you taken these positions?
b) If the annual riskless rate of interest were 8%, would the position be profitable? Why
or why not?
9.6 Suppose you believe your portfolio, which has a beta of 1.0, has been selected to outperform
other portfolios of similar risk but you know you cannot predict which way the market will
move. If it goes down, you will outperform the market but will still have a negative rate of
return. What can you do to alleviate your timing risk?
9.7 Suppose you are convinced that the spread between long- and short-term rates will widen,
whereas everyone else thinks it will remain constant. Unfortunately, you do not know whether
the general level of interest rates will go up or down. What can you do?
9.8 Your bank is exploring the possibility of using T-bond futures to minimize the exposure
of shareholders to changes in the interest rate. The market value of major assets and liabilities
is given in the balance sheet below:
The economics staff has used the rates of return on the asset and liability positions to compute
the following long-run standard deviations and correlations:
If the current market value is $80,000 for a T-bond futures contract with a $100,000 face value,
how many T-bond contracts will be needed to minimize shareholders' risk exposure to interest
rate fluctuations?
REFERENCES
Anderson, R., "Comments on 'Margins and Futures Contracts, — Journal of Futures Markets,
Summer 1981, 259-264.
, "Some Determinants of the Volatility of Futures Prices," Journal of Futures Markets,
Fall 1985, 331-348.
REFERENCES 3<7
, and J. Danthine, "The Time Pattern of Hedging and the Volatility of Futures Prices,"
Review of Economic Studies, April =983, <49-<66.
Ball, C., and W. Torous, "Futures Options and the Volatility of Futures Prices," Jou]nal of
Finance, Septemîer =986, 8P7-870.
Black, F., "The Pricing of Commodity Options," Jou]nal of Financial Economics, March =976,
=67-=79.
Bodie, Z., and V. Rosansky, "Risk and Return in Commodities Futures," Financial Analysts
Jou]nal, May-June =980, <7-39.
Breeden, D., "Consumption Risk in Futures Markets," Jou]nal of Finance, May =980, P03-P<0.
Brennan, M. J., "The Supply of Storage," Ame]ican Economic Review, March =9P8, P0-7<.
Brennan, M. "The Cost of Convenience and the Pricing of Commodity Contingent Claims,"
Working Paper, UCLA, =986a.
, "A Theory of Price Limits in Futures Markets," Jou]nal of Financial Economics, June
=986î, <=3-<34.
Brenner, M.; G. Courtadon; and M. Suîrahmanyam, "Options on the Spot and Options on
Futures," Jou]nal of Finance, Decemîer =98P, =303-=3=7.
Capozza, D., and B. Cornell, "Treasury Bill Pricing in the Spot and Futures Markets," Review
of Economics and Statistics, Novemîer =979, P=3-P<0.
Carlton, D., "Futures Markets: Their Purpose, Their History, Their Growth, Their Successes
and Failures," Jou]nal of Futu]es Ma]kets, Fall =984, <37-<7=.
Carter, C. A.; G. C. Rausser; and A. Schmitz, "Efficient Asset Portfolios and the Theory of
Normal Backwardation," Jou]nal of Political Economy, April =983, 3=9-33=.
Chang, E., "Returns to Speculators and the Theory of Normal Backwardation," Jou]nal of
Finance, March =98P, =93-<08.
Constantinides, G. M., "Capital Market Equiliîrium with Personal Tax," Economet]ica, May
=983, 6==-636.
Cootner, P., "Returns to Speculators: Telser vs. Keynes," Jou]nal of Political Economy, August
=960, 396-4=4.
Cornell, B., and K. French, "The Pricing of Stock Index Futures," Jou]nal of Futu]es Ma]kets,
Spring =983a, =-=4.
, "Taxes and the Pricing of Stock Index Futures," Jou]nal of Finance, June =983î,
67P-694.
Cornell, B., and M. Reinganum, "Forward and Futures Prices: Evidence from the Foreign
Exchange Markets," Jou]nal of Finance, Decemîer =98=, =03P-=04P.
Cox, J. C.; J. E. Ingersoll; and S. A. Ross, "The Relation îetween Forward Prices and Futures
Prices," Jou]nal of Financial Economics, Decemîer =98=, 3<=-346.
Dusak, K., "Futures Trading and Investor Returns: An Investigation of Commodity Market
Risk Premiums," Jou]nal of Political Economy, Novemîer—Decemîer =973, =387-=406.
Fama, E., and K. French, "Business Cycles and the Behavior of Metals Prices," Working Paper
# 3=-86, UCLA, =986.
, "Commodity Futures Prices: Some Evidence on the Forecast Power, Premiums, and
the Theory of Storage," Jou]nal of Business, January =987, PP-73.
Figlewski, S., "Margins and Market Integrity: Margin Setting for Stock Index Futures and
Options," Jou]nal of Futu]es Ma]kets, Fall =984, 38P-4=6.
French, K., "Detecting Spot Price Forecasts in Futures Prices," Jou]nal of Business, April =986,
P39-PP4.
3<8 FUTURES CONTRACTS AND MARKETS
Gay, G., and S. Manaster, "The Quality Option Implicit in Futures Contracts," Jou]nal of
Financial Economics, Septemîer =984, 3P3-370.
Grauer, F., and R. Litzenîerger, "The Pricing of Commodity Futures Contracts, Nominal
Bonds and Other Assets under Commodity Price Uncertainty," Jou]nal of Finance, March
=979, 69-83.
Hansen, L. P., and R. J. Hodrick, "Forward Exchange Rates as Optimal Predictors of Future
Spot Rates: An Econometric Analysis," Jou]nal of Political Economy, Octoîer =980,
8<9-8P3.
Hartzmark, M. L., "The Effect of Changing Margin Levels on Futures Market Activity, the
Composition of Traders in the Market, and Price Performance," Jou]nal of Business,
April =986, S=47—S=80.
Hazuka, T., "Consumption Betas and Backwardat on in Commodity Markets," Jou]nal of
Finance, July =984, 647-6PP.
Hemmler, M., "The Quality Delivery Option in Treasury Bond Futures Contracts," Ph.D. dis-
sertation, University of Chicago, =987.
Hicks, J. R., Value and Capital, <nd ed. Oxford University Press, Oxford, =946.
Hirshleifer, D., "Risk, Equiliîrium, and Futures Markets," Ph.D. dissertation, University of
Chicago, =984.
, "Residual Risk, Trading Costs and Commodity Risk Premia," Working Paper, UCLA,
=986.
Houthakker, H. S., "The Scope and Limits of Futures Trading," in Aîramovitz, ed., The
Allocation of Economic Resou]ces. Stanford University Press, Stanford, =9P9.
Jagannathan, Ravi, "An Investigation of Commodity Futures Prices Using the Consumption-
Based Intertemporal Capital Asset Pricing Model," Jou]nal of Finance, March =98P,
=7P-=9=.
Kaldor, N., "Speculation and Economic Staîility," Review of Economic Studies, Octoîer =939,
=-<7.
Keynes, J. M., "Some Aspects of Commodity Markets," Mancheste] Gua]dian Comme]cial,
Eu]opean Reconst]uction Se]ies, Section =3, March <9, =9<3, 784-786.
Kolî, R., Unde]standing Futu]es Ma]kets, Scott, Foresman and Company, Glenview, Ill. =98P.
, and R. Chiang, "Improving Hedging Performance Using Interest Rate Futures,"
Financial Management, Autumn =98=, 7<-79.
, "Immunization and Hedging with Interest Rate Futures," Jou]nal of Financial Re-
sea]ch, Summer =98<, =6=-=70.
Kolî, R.; G. Gay; and J. Jordan, "Are There Arîitrage Opportunities in the Treasury-Bond
Futures Market?" Jou]nal of Futu]es Ma]kets, Fall =98<, <=7-<30.
Milonas, N., "Price Variaîility and the Maturity Effect in Futures Markets," Jou]nal of Futu]es
Ma]kets, Fall =986, 443-460.
Modest, D., and M. Sundaresan, "The Relationship îetween Spot and Futures Prices in Stock
Index Futures Markets: Some Preliminary Evidence," Jou]nal of Futu]es Ma]kets, Spring
=983, =P-4=.
Oldfield, G., and C. Rovira, "Futures Contract Options," Jou]nal of Futu]es Ma]kets, Winter
=984, 479-490.
Pashigian, B. P., "The Political Economy of Futures Market Regulation," Jou]nal of Business,
April =986, SPP—S84.
REFERENCES 3<9
The purpose of capital markets is to transfer funds between lenders (savers) and
borrowers (producers) efficiently. Individuals or firms may have an excess of produc-
tive investment opportunities with anticipated rates of return that exceed the mar-
ket-determined borrowing rate but not enough funds to take advantage of all these
opportunities. However, if capital markets exist, they can borrow the needed funds.
Lenders, who have excess funds after exhausting all their productive opportunities
with expected returns greater than the borrowing rate, will be willing to lend their
excess funds because the borrowing/lending rate is higher than what they might other-
wise earn. Therefore both borrowers and lenders are better off if efficient capital
markets are used to facilitate fund transfers. The borrowing/lending rate is used as
an important piece of information by each producer, who will accept projects until
the rate of return on the least profitable project just equals the opportunity cost of
external funds (the borrowing/lending rate). Thus a market is said to be allocationally
efficient when prices are determined in a way that equates the marginal rates of return
(adjusted for risk) for all producers and savers. In an allocationally efficient market,
scarce savings are optimally allocated to productive investments in a way that bene-
fits everyone.
To describe efficient capital markets it is useful, first of all, to contrast them with
perfect capital markets. The following conditions are necessary for perfect capital
330
DEFINING CAPITAL MARKET EFFICIENCY 331
markets:
• Markets are frictionless; i.e., there are no transactions costs or taxes, all assets
are perfectly divisible and marketable, and there are no constraining regulations.
• There is perfect competition in product and securities markets. In product mar-
kets this means that all producers supply goods and services at minimum average
cost, and in securities markets it means that all participants are price takers.
• Markets are informationally efficient; i.e., information is costless, and it is received
simultaneously by all individuals.
• All individuals are rational expected utility maximizers.
Given these conditions both product and securities markets will be both allocation-
ally and operationally efficient. Allocational efficiency has already been defined, but
what about operational efficiency? Operational efficiency deals with the cost of trans-
ferring funds. In the idealized world of perfect capital markets, transactions costs are
assumed to be zero; therefore we have perfect operational efficiency.' However, we
shall see later, when we focus on empirical studies of real-world phenomena, that
operational efficiency is indeed an important consideration.
Capital market efficiency is much less restrictive than the notion of perfect capital
markets outlined above. In an efficient capital market, prices fully and instantaneously
reflect all available relevant information. This means that when assets are traded,
prices are accurate signals for capital allocation.
To show the difference between perfect markets and efficient capital markets we
can relax some of the perfect market assumptions. For example, we can still have
efficient capital markets if markets are not frictionless. Prices will still fully reflect all
available information if, e.g., securities traders have to pay brokerage fees or if an
individual's human capital (which, after all, is an asset) cannot be divided into a
thousand parts and auctioned off. More important, there can be imperfect competi-
tion in product markets and we still have efficient capital markets. Hence if a firm
can reap monopoly profits in the product market, the efficient capital market will
determine a security price that fully reflects the present value of the anticipated stream
of monopoly profits. Hence we can have allocative inefficiencies in product markets
but still have efficient capital markets. Finally, it is not necessary to have costless
information in efficient capital markets. This point is discussed in greater detail in
section E of this chapter.
Still, in a somewhat limited sense, efficient capital markets imply operational
efficiency as well as asset prices that are allocationally efficient. Asset prices are correct
signals in the sense that they fully and instantaneously reflect all available relevant
information and are useful for directing the flow of funds from savers to investment
projects that yield the highest return (even though the return may reflect monopolistic
practices in product markets). Capital markets are operationally efficient if interme-
diaries, who provide the service of channeling funds from savers to investors, do so
at the minimum cost that provides them a fair return for their services.
I
Note that even in perfect markets the minimum cost of transferring funds may not be zero if the transfer
of funds also involves risk bearing.
332 EFFICIENT CAPITAL MARKETS: THEORY
Fama [1970, 1976] has done a great deal to operationalize the notion of capital
market efficiency. He defines three types of efficiency, each of which is based on a
different notion of exactly what type of information is understood to be relevant in
the phrase "all prices fully reflect all relevant information."
1. Weak form efficiency. No investor can earn excess returns by developing trading
rules based on historical price or return information. In other words, the infor-
mation in past prices or returns is not useful or relevant in achieving excess
returns.
2. Semistrong-form efficiency. No investor can earn excess returns from trading rules
based on any publicly available information. Examples of publicly available in-
formation are annual reports of companies, investment advisory data such as
"Heard on the Street" in the Wall Street Journal, or ticker tape information.
3. Strong form efficiency. No investor can earn excess returns using any information,
whether publicly available or not.
Obviously, the last type of market efficiency is very strong indeed. If markets were
efficient in their strong form, prices would fully reflect all information even though
it might be held exclusively by a corporate insider. Suppose, e.g., we know that our
company has just discovered how to control nuclear fusion. Even before we have a
chance to trade based on the news, the strong form of market efficiency predicts that
prices will have adjusted so that we cannot profit.
Rubinstein [1975] and Latham [1985] have extended the definition of market
efficiency. The market is said to be efficient with regard to an information event if
the information causes no portfolio changes. It is possible that people might disagree
about the implications of a piece of information so that some buy an asset and others
sell in such a way that the market price is unaffected. If the information does not
change prices, then the market is said to be efficient with regard to the information
in the Fama [1976] sense but not in the Rubinstein [1975] or Latham [1985] sense.
The Rubinstein-Latham definition requires not only that there be no price change
but also that there be no transactions. Hence it is a stronger form of market efficiency
than even the Fama strong-form efficiency mentioned above.
The notion of efficient capital markets depends on the precise definition of information
and the value of information.' An information structure may be defined as a message
about various events which may happen. For example, the message "There are no
clouds in the sky" provides a probability distribution for the likelihood of rain within
the next 24 hours. This message may have various values to different people depend-
ing on (1) whether or not they can take any actions based on the message and (2)
For an excellent review of the economics of information, see Hirshleifer and Riley [1979].
A FORMAL DEFINITION OF THE VALUE OF INFORMATION 333
what net benefits (gain in utility) will result from their actions. For example, a mes-
sage related to rainfall can be of value to farmers, who can act on the information
to increase their wealth. If there is to be no rain, the farmers might decide that it
would be a good time to harvest hay. On the other hand, messages about rainfall
have no value to deep-pit coal miners because such information probably will not
alter the miners' actions at all.
A formal expression of the above concept defines the value of an information
structure, V(q), as
V(ii) q(m) MAX E p(e m)U(a, — V( 110), (10.1)
a e
where
q(m) = the marginal probability of receiving a message m;
p(e m) = the conditional probability of an event e, given a message m;
U(a, e) = the utility resulting from an action a if an event e occurs;
we shall call this a benefit function;
V(17 0 ) = the expected utility of the decision maker without the information.
According to Eq. (10.1), a decision maker will evaluate an information structure
(which, for the sake of generality, is defined as a set of messages) by choosing an ac-
tion that will maximize his or her expected utility, given the arrival of a message.
For example, if we receive a message (one of many that we could have received)
that there is a 20% chance of rain, we may carry an umbrella because of the high
"disutility" of getting drenched and the low cost of carrying it. For each possible
message we can determine our optimal action. Mathematically, this is the solution
to the problem:
MAX p(elm)U(a, e).
a
Finally, by weighting the expected utility of each optimal action (in response to all
possible messages) by the probability, q(m), of receiving the message that gives rise
to the action, the decision maker knows the expected utility of the entire set of mes-
sages, which we call the expected utility (or utility value) of an information set, V(q).
The following example applies the value-of-information concept to the theory of
portfolio choice. We will choose our optimal portfolio as a combination of two funds:
either the risk-free asset that yields 6%, or the market portfolio that may yield 16%
or 10% or —5%. 3 Assume that we know the standard deviation of the market port-
folio, a m , with certainty. Figure 10.1 shows the linear efficient set (the capital market
line) for two of the three possible states of the world. As risk-averse investors, in
this case, we will maximize our expected utility by choosing the portfolio where our
indifference curve is tangent to the efficient set.
To calculate the value of an information set we need to know the payoff function
U(a, e), which tells us the utility of having taken a course of action, a, when an event
3
A more general, îut also more complicated, example would assign a continuous proîaîility distriîution
to the possiîle returns offered îy the market portfolio.
334 EFFICIENT CAPITAL MARKETS: THEORY
E(ii p ) U7 = 40
U4= 26
U1= 10
e3
E(i?,,)= 16%
o- (RP )
E(R,p ) = —5%
Portfolio 1 Portfolio 3 Qm
Figure 10.1
Optimal choices for two states of the world.
or state of the world, e, occurs. For the sake of convenience, we will label the three
states of the world by their market returns: e 3 = 16%, e 2 = 10%, and e 1 = —5%,
If the market return were 16%, we would choose portfolio 3 in Fig. 10.1, which
is where indifference curve U 7 is tangent to the capital market line e 3 . On the other
hand, we would put all our portfolio in the risk-free asset (portfolio 1) if the market
return were known to be —5%. This occurs where indifference curve U 4 passes
through R f . If we choose portfolio 3, and if the market rate of return really is 16%,
then our payoff is U 7 = 40. But if we make a mistake and choose portfolio 3 when
the market rate of return turns out to be —5%, our portfolio is suboptimal. After
the fact, we would have been much better off with portfolio 1, the risk-free asset.
Our utility for holding portfolio 3 when state 1 obtains (point A in Fig. 10.1) is U 1 =
10 less our regret, which is the difference between where we actually are and where
we would like to be.' In this case our regret is the difference between U 4 and U 1.
Therefore our net utility is U 1 —(U 4 —U 1) = 2U1 — U 4 = —6. When we receive a
message that provides estimates of the likelihood of future states of the world, we
will choose an action (in our example this amounts to choosing a portfolio) that
will maximize our expected utility, given that message. The utility provided by each
portfolio choice (i.e., each action) in each state of the world can be taken from Fig.
10.2, which is similar to Fig. 10.1 except that it gives all possible portfolios and
states of the world. The corresponding benefit function, U(a, e), is given in Table 10.1.
In addition to a benefit matrix it is also necessary to have an information struc-
ture (a Markov matrix) that gives the probability that an event will actually occur,
given that a particular message has been received. The two obvious polar cases are
Needless to say, the computation of "regret" as suggested here is not necessarily accurate and is only
for purposes of illustration. However, it is consistent with the fact that utilities may be state contingent.
(See Chapter 5, "State-Preference Theory.")
A FORMAL DEFINITION OF THE VALUE OF INFORMATION 335
Cardinal utilities
U1= 10 % in Rf % in R„,
U2= 20 Portfolio 1 100.0 0
U3= 25 Portfolio 2 77.5 22.5
U4= 26 Portfolio 3 27.5 72.5
U5 = 30
U6= 32
U7 = 40
Portfolio 2 (a2)
Portfolio 3 (a3)
Portfolio 1 is 100% in Rf (at )
Figure 10.2
Optimal actions, given various states of the world.
perfect information and no information. Their matrices are given in Table 10.2. If
the information structure is perfect, receipt of a message implies that a given state
of the world will occur with certainty. This makes it easy to select a course of action
that results in the highest utility. If you receive m 1, then the state of the world e l
will obtain with certainty; therefore the best action is a 1, with expected utility U4 = 26.
Thus for each message it is possible to find the optimal action. Implicitly this pro-
cedure amounts to solving Eq. (10.1), which tells us the utility value of information.
ml m2 M3 17'1 3 m2 m3 M1 M2 M
3
e1 1.0 0 0 e1 3 3 e1 .6 .3 .1
e2 0 1.0 0 e2 3 e2 .2 .5 .3
e3 0 0 1.0 e3 3 e3 .2 .2 .6
336 EFFICIENT CAPITAL MARKETS: THEORY
Given perfect information, the optimal action (and its utility) for each message is seen
to be:
mi a l (invest in portfolio 1) 26
M2 a 2 (invest in portfolio 2) 30
m3 a 3 (invest in portfolio 3) 40
These actions represent the solution to one part of the value of information, i.e., Eq.
(10.1):
MAX p(e m)U(a, e).
a e
Finally, the utility of each action that results from a message is weighted by the prob-
ability of the message. In our example each of the three messages is assumed to be
equally likely; therefore, using the notation of Eq. (10.1), we have q(m i ) = 4, q(m 2) = 3,
and q(m 3 ) = 4. The utility value of perfect information is
V(r1 2) = 4(26) + 4(30) + 4(40) = 32.
It is also useful to compute the dollar value of this information set to the ith in-
dividual. Of course, different individuals will have different utility functions and there-
fore different demand curves for a given information set. But for now we will focus
only on the dollar value of information to one individual.' Suppose his utility func-
tion is
U(W) = 10 ln(W — $100).
Then if the expected utility of perfect information is 32 utiles, the corresponding
dollar value is
32 = 10 ln(W(g 2 ) — $100),
3.2 = ln( W(17 2) — $100),
$24.5325 = W(q 2) — $100,
$124.53 = W(q 2).
This is the increment to wealth that the ith individual would receive were he or she
to acquire the information (at no cost).
Next, consider the value of no information. In Table 10.2 we see that no infor-
mation means that all messages are identical. Each says the same thing: "All events
are equally likely." For example, suppose a friend asks whether or not a movie was
worth seeing. If we always reply yes, whether or not the movie is good, then the mes-
sage contains no information. If we always say no, the result is the same no infor-
mation. In the example at hand, when asked about the probability of a state of the
5
For more on this topic see Huang, Vertinsky, and Ziemba [1977].
A FORMAL DEFINITION OF THE VALUE OF INFORMATION 337
And if we receive message 3 the expected utilities of our actions are the same (obviously,
since the three messages are the same). Regardless of the message, our optimal action
is always the same—invest in portfolio 2; and our expected utility is 22.67. As with
perfect information the selection of optimal actions, given various messages, is the
first part of the problem. The second part is to weight the expected utility of an
action, given a message, by the probability of the message. The result is the utility
value of information. The utility value of no information is
V(r 0 ) = 4(22.67) + 4(22.67) + 4(22.67) = 22.67.
Using the same utility function as above, the dollar value of information in this case
is W(g o ) = $109.65. The difference between the dollar value of perfect information
and no information is the maximum gain from information for the ith individual:
Similarly,
Finally, if we weight the utility of the optimal actions, given the three messages, by
the probability of the messages, we have the utility value of the noisy information
structure:
Its corresponding dollar value, for the ith individual, is W(171) = $112.52.
Equation (10.1) can be used to evaluate any information structure. It also points out
some ideas that are only implicit in the definition of efficient markets. Fama [1976]
defines efficient capital markets as those where the joint distribution of security prices,
we do not already know. If the distribution of prices in time period t (which was
predicted in the previous time period t— 1 and based on the information structure
the market uses) is not different from the prices predicted by using all relevant
information from the previous time period, then there must be no difference between
the information the market uses and the set of all relevant information. This is the
essence of an efficient capital market it instantaneously and fully reflects all relevant
information. Using information theory, this also means that net of costs, the utility
value of the gain from information to the ith individual, must be zero:
V. (10.3)
(i1i) — V i
( 1o) O.
For example, consider capital markets that are efficient in their weak form. The
relevant information structure, rh, is defined to be the set of historical prices on all
assets. If capital markets are efficient, then Eq. (10.2) says that the distribution of
security prices today has already incorporated past price histories. In other words,
RATIONAL EXPECTATION AND MARKET EFFICIENCY 339
it is not possible to develop trading rules (courses of action) based on past prices
that will allow anyone to beat the market. Equation (10.3) says that no one would
pay anything for the information set of historical prices. The value of the information
is zero. Empirical evidence on trading rules that use past price data is discussed in
section F of this chapter.
It is important to emphasize that the value of information is determined net of
costs. These include the cost of undertaking courses of action and the costs of trans-
mitting and evaluating messages. Some of these costs in securities markets are trans-
actions costs: e.g., brokerage fees, bid-ask spreads, costs involved in searching for the
best price (if more than one price is quoted), and taxes, as well as data costs and
analysts' fees. The capital market is efficient relative to a given information set only
after consideration of the costs of acquiring messages and taking actions pursuant
to a particular information structure.
The utility value of information has three parts: (1) the utilities of the payoffs, given
an action; (2) the optimal actions, given receipt of a message; and (3) the probabilities
of states of nature provided by the messages. We are interested in understanding how
the individual's decision-making process, given the receipt of information, is reflected
in the market prices of assets. This is not easy because it is impossible to observe
the quantity and quality of information or the timing of its receipt in the real world.
There is even disagreement among theorists about what information will be used by
investors. For example, Forsythe, Palfrey, and Plott [1982] identify four different
hypotheses. Each hypothesis assumes that investors know with certainty what their
own payoffs will be across time, but they also know that different individuals may
pay different prices because of differing preferences.
The first hypothesis is particularly nonsensical (call it the naive hypothesis) in
that it asserts that asset prices are completely arbitrary and unrelated either to how
much they will pay out in the future or to the probabilities of various payouts. The
second hypothesis, call it the speculative equilibrium hypothesis, is captured in a quote
taken from Keynes's General Theory [1936, 156]:
Professional investment may be likened to those newspaper competitions in which the
competitors have to pick out the six prettiest faces from a hundred photographs, the prize
being awarded to the competitor whose choice most nearly corresponds to the average
preferences of the competitors as a whole; so that each competitor has to pick, not those
faces which he himself finds the prettiest, but those which he thinks likeliest to catch the
fancy of the other competitors, all of whom are looking at the problem from the same
point of view. It is not a case of choosing those which, to the best of one's judgement, are
really the prettiest, nor even those which average opinion genuinely thinks the prettiest.
We have reached the third degree where we devote our intelligences to anticipating what
average opinion expects the average opinion to be. And there are some, I believe, who
practice the fourth, fifth and higher degrees.
340 EFFICIENT CAPITAL MARKETS: THEORY
We might debate about what Keynes really meant, but one interpretation is that
all investors base their investment decisions entirely on their anticipation of other
individuals' behavior without any necessary relationship to the actual payoffs that
the assets are expected to provide. The third hypothesis is that asset prices are sys-
tematically related to their future payouts. Called the intrinsic value hypothesis, it
says that prices will be determined by each individual's estimate of the payoffs of an
asset without consideration of its resale value to other individuals. The fourth hy-
pothesis may be called the rational expectations hypothesis. It predicts that prices are
formed on the basis of the expected future payouts of the assets, including their
resale value to third parties. Thus a rational expectations market is an efficient market
because prices will reflect all information.
To make these hypotheses more concrete it is useful to review an experiment by
Forsythe, Palfrey, and Plott [1982]. They set up an oral double auction market that
had two time periods, one asset, and three "types" of individuals. An oral double
auction market is one where individuals can call out both buying and selling prices
for an asset. The largest market of this type is the New York Stock Exchange. All
participants knew exactly how much the asset would pay them in each time period.
They also knew that the asset would pay different amounts to the other market
participants, but not how much. Thus the asset had different values to different
individuals. Table 10.3 shows the experimental parameters. If you held an asset at
the end of the first time period, you received 300, 50, or 150 "francs," depending on
whether you were individual type I, II or III.' The differences in the franc payouts
across individuals were designed to reflect differences among individual preferences
and information sets at an instant in time. Each individual was endowed with 10,000
francs in working capital and two shares.' The 10,000 franc endowment was paid
back at the end of the second time period (this is the fixed cost column in Table
10.3), but any dividends received and any trading profits were kept by the market
participants.
The interesting question is, What will the market equilibrium prices be at the
end of each of the two time periods? If either of the intrinsic value or rational
expectations hypotheses are true, the period B price should be 300 francs (or very
6
Each "franc" was worth $0.002.
No short sales were allowed; thus the supply of assets was fixed.
RATIONAL EXPECTATIONS AND MARKET EFFICIENCY 34=
Price — — —— —. - --
Year 4 — _
Year _ _ R_
Year
A=B A B A B A B A vv B AIB
700 A B A B
=
I i
600 --= ---= --= —= ---= —I —
I
=00
0
38= <70 398 <78 4<P <88 4P4 <87 478 <9< 496 <90 P4= <96 P76 <9P
Average price
Figure =0.3
Spot price equiliîria in an experimental market. (From R. Forsythe, T. Palfrey, and
C. R. Plott "Asset Valuation in an Experimental Market,". reprinted from Economet]ica,
May =98<, PP0.)
close to it) îecause there are three people competing for that payoff. While the hy-
potheses agree aîout the second-period price, they make different predictions con-
cerning the first-period price. The intrinsic value hypothesis predicts that people
will îid their own values. For example, if individual type III holds the asset for two
periods, he or she can collect 400 francs in dividends (=P0 in period A and <P0 in
period B). This will also îe the predicted first-period equiliîrium price îecause it
represents the high îid. The rational expectations hypothesis predicts that the first-
period equiliîrium price will îe 600 francs îecause type I individuals can collect a
300 franc dividend in the first time period, then sell the asset for 300 francs to type
II individuals in the second time period.
Figure =0.3 shows the results of eight replications of the experiment for two
periods each. Note that the period B price quickly converges on the anticipated
equiliîrium value of 300 francs. This result alone repudiates the naive value and the
speculative equiliîrium hypotheses îecause the asset value is clearly îased on its
actual second-period payout. The first-period price starts out at 400 francs, seemingly
verifying the intrinsic value hypothesis; îut it then rises gradually until, îy the eighth
replication of the experiment ("year" 8), it closes in on the rational expectations value.
The experiment seems to confirm the rational expectations hypothesis, îut why
did it take so long to do so? The reason is that prices are determined in the first
period îefore second-period prices are known. In order for type I individuals to îid
the full 600 franc value, they need to have the information that the second-period
342 EFFICIENT CAPITAL MARKETS: THEORY
value is really 300 francs. Obviously they do not know this during the first trial of
the experiment, but they learn quickly.
If instead bidding had taken place in both period A and B markets simulta-
neously, perhaps the speed of adjustment to a rational expectations equilibrium would
have been faster. In another experiment, Forsythe, Palfrey, and Plott [1982] opened
a futures market. Everything remained the same as before except bidding for period
B holdings was held concurrently with the period A spot market, and the payoffs
were as shown in Table 10.4.
The rational expectations hypothesis predicts that the period A price will be
845 francs, whereas the intrinsic value hypothesis predicts 403 francs. They both
predict a period B price of 442 francs. The results are shown in Fig. 10.4. Even in
the first trial, the period A spot price closed at 742 francs, much closer to the rational
expectations prediction. In subsequent trials ("years") the closing prices were even
closer to the results predicted by the rational expectations hypothesis. Perhaps the
Price ____ 9_
Year Year S
A 1 B A i B A 1 B A B A I B A B
950
850
_ ...._ ..- _ .. .. ..•. . .„....- .. . .. . ---
...:••
750 .....
650 ---I _I
550
250
742 806 800 825 831 31 period spot
371 425 429 435 435 437 period B futures
Average price
Figure 10.4
Rational expectations with a futures market. (From R. Forsythe, T. Palfrey, and C. R. Plott,
"Asset Valuation in an Experimental Market," reprinted from Econometrica, May 1982, 554.)
MARKET EFFICIENCY WITH COSTLY INFORMATION 343
most valuable implication of the experiment is that it clearly demonstrates the useful-
ness of futures markets. By allowing simultaneous trading in both markets, the speed
with which information is made public is increased through price transactions. In-
formation about the future value of assets is revealed today.
In the rational expectations equilibria described above, all traders knew with
certainty what their own payoffs would be in each time period but did not know the
asset clearing price because other individuals had different payoffs in the same state
of nature. These different payoffs represent a form of heterogeneous expectations.
In the above experiments all market participants were equally well informed. There
is a different way of looking at heterogeneous expectations, however. Suppose that
some traders are better informed about which states of nature will actually occur.
Furthermore, suppose that different individuals have different information about
which states will occur. For example, suppose investor I knows for sure that a Re-
publican will be elected president but knows nothing else. Investor J, on the other
hand, knows that both houses of Congress will be Democratic but knows nothing
else. The question is this: Will market prices reflect the full impact of both pieces
of information as though the market were fully informed, or will prices reflect only
some average of the impact of both pieces of information? If prices reflect all infor-
mation, the market is said to be fully aggregating; otherwise it is only averaging
prices.
Very little is known about whether real-world capital markets fully aggregate
information or merely average it. A fully aggregating market, however, would be
consistent with Fama's [1970] definition of strong-form market efficiency. In a fully
aggregating market even insiders who possess private information would not be able
to profit by it. One mechanism for aggregation has been suggested by Grossman
and Stiglitz [1976] and Grossman [1976]. 8 In a market with two types of traders,
"informed" and "uninformed," informed traders will acquire better estimates of future
states of nature and take trading positions based on this information. When all
informed traders do this, current prices are affected. Uninformed traders invest no
resources in collecting information, but they can infer the information of informed
traders by observing what happens to prices. Thus the market prices may aggregate
information so that all traders (both informed and uninformed) become informed.
In Chapter 11 we will suggest that capital markets do not instantaneously and fully
aggregate information because empirical evidence on insider trading reveals that
insiders can and do make abnormal returns.
If capital markets are efficient, then no one can earn abnormal returns, but without
abnormal returns there is no strong incentive to acquire information. Random selec-
tion of securities is just as effective. How, then, can prices reflect information if there
Actually, this idea can be traced back to Hayek's classic article, "The Use of Knowledge in Society"
[1945].
344 EFFICIENT CAPITAL MARKETS: THEORY
r = normal return = 6%
d = competitive advantage = 2 x
c2 = cost of analysis = 8%
c 1 = cost with no analysis = 4%
is no incentive to search it out and use it for arbitrage? How can a securities analysis
industry exist?
The above argument may have some merit in a world with costless information
because all investors would have zero abnormal returns.' However, it is probably
premature to predict the demise of the security analysis industry or to argue that
prices are uninformative. Grossman and Stiglitz [1976, 1980] and Cornell and Roll
[1981] have shown that a sensible asset market equilibrium must leave some room
for analysis. Their articles make the more reasonable assumption that information
acquisition is a costly activity. Because of its simplicity, the Cornell and Roll model
is discussed below.
We want to analyze the rational behavior of individuals when information is
useful in the sense that having it will improve one's decisions but also where in-
formation is costly. Imagine two simple strategies. The first is to pay a fee, say c 2 =
8%, for acquiring valuable information. This is called the analyst strategy. The
opposing strategy is to pay a minimal fee, say c 1 = 4%, for the right to trade. Call
this the random selector's strategy. Table 10.5 shows the net payoffs for various two-
way trades involving all combinations of analysts and random selectors. Note that
the "normal" rate of return, r, is 6% (c 1 < r < c 2).
The example in Table 10.5 assumes that the costly information doubles the
competitive advantage, d, of an analyst whenever he or she trades with a random
selector. The analyst, being better informed, grosses 12% and nets 4% after paying
8% for the information. Conversely, the random selector finds his or her gross return
halved when trading with an analyst. He or she grosses only 3% and nets —1% after
paying 4% on transactions costs. When an analyst trades with another analyst there
is no competitive advantage because they possess the same information. Thus their
gross return is only 6% and they net —2%.
A stable equilibrium can exist (1) if all trading is anonymous and (2) if the ex-
pected payoff to the
10
analysis strategy equals the expected payoff to the random
selection strategy. If p is the probability of utilizing an analyst's strategy and
9
Abnormal returns are returns in excess of what can be earned for a given level of risk.
19
Anonymous trading is necessary so that uninformed random selectors will actually consummate a trade
with an analyst with superior information. One service provided by brokers is to ensure the anonymity
of their clients.
MARKET EFFICIENCY WITH COSTLY INFORMATION 345
The expected payoff to analysis [the left-hand side of Eq. (10.4)] is the probability that
an analyst will confront another analyst multiplied by the net payoff given this event,
plus the probability of confronting a random selector multiplied by the net payoff
in that event. Similar logic produces the expected payoff for a random selector [the
right-hand side of Eq. (10.4)]. Solving Eq. (10.4) for p, the probability of using analysis,
we have
r(} — d + 3 2 — 3 }
P= (10.5)
2r — rd — rmd
A mixed stable strategy is one where 0 < p < 1, that is, where analysts and random
selectors will coexist in equilibrium. The necessary conditions for a mixed stable
strategy are
r(d — } > 3 2 — 3 } and r(1 — 1/d) < 3, — 3 }. (10.6)
These conditions can be derived from the definition of the equilibrium probability,
p, Eq. (10.5). We know that the "normal" rate of return, r, is greater than zero and
the competitive advantage, d, is greater than one. Therefore the denominator of Eq.
(10.5), must be negative:
2r — rd — rmd < 0,
2d — d 2 — 1 < 0,
2
(d — } > 0, since d > 1. QED
It follows that the numerator of Eq. (10.5) must also be negative if the probability,
p, is to be positive. Therefore
r(1 — d + 3 2 — 3 } < 0,
r(d — } > 3 2 — 3 },
and we have derived the first necessary condition in Eq. (10.6). Also, in order for
p < 1 the numerator of Eq. (10.5) must be greater than the denominator (since both
are negative numbers). This fact gives us the second necessary condition:
If there is no net economic profit when the mixed stable strategy evolves, then
there will be no incentive for new entrants to disturb the equilibrium. This zero profit
condition is equivalent to setting both sides of Eq. (10.4), the expected payoff equa-
tion, equal to zero. This results in two equations, which when equated and simplified
346 EFFICIENT CAPITAL MARKETS: THEORY
Historically it was possible to test certain predictions of the efficient markets hy-
pothesis even before a theory of risk-bearing allowed comparison of risk-adjusted
returns. For example, if the riskiness of an asset does not change over time or if its
risk changes randomly over time, then there should be no pattern in the time series
of security returns. If there were a recurring pattern of any type, investors who rec-
ognize it could use it to predict future returns and make excess profits. However, in
their very efforts to use the patterns, they would eliminate them.
Three theories of the time series behavior of prices can be found in the literature:
(1) the fai]-game model, (2) the ma]tingale or subma]tingale, and (3) the ]andom walk.
The fair game model is based on the behavior of average returns (not on the entire
probability distribution). Its mathematical expression is
Pi,t+i Pit E(Pi,t+) TO Pit
ei,t+ =
Pit
13
P j,t +1 ÿ E( J,t +
(10.7)
Pit
where
= the actual price of security j next period,
E(PJ,t+1 th)= the predicted end-of-period price of security j given the current
information structure, u„
ei ,t+ , = the difference between actual and predicted returns.
STATISTICAL TESTS UNADJUSTED FOR RISK 347
Note that (10.7) is really written in returns form. If we let the one-period return be
defined as
P j,t + 1 — P jt
r+1=
Pit
then (10.7) may be rewritten as
= rj,t+ 1 E(ri,t+ 1111t)
and
E(e1,t+1) = E[ri,t+1 — E(rj,t +1 nt)] = 0. (10.8)
A fair game means that, on average, across a large number of samples the expected
return on an asset equals its actual return. An example of a fair game would be games
of chance in Las Vegas. Because of the house percentage, you should expect to lose,
let us say, 10%; and sure enough, on the average that is what people actually lose. A
fair game does not imply that you will earn a positive return; only that expectations
are not biased.
Given the definition of a fair game in Eq. (10.7), a submartingale is a fair game
where tomorrow's price is expected to be greater than today's price. Mathematically,
a submartingale is
> Pit.
In returns form this implies that expected returns are positive. This may be written
as follows:
E(P + 1 t)
= E(ri,t+11111) > 0. (10.9a)
Pit
Random walks are much stronger conditions than fair games or martingales because
they require all the parameters of a distribution (e.g., mean, variance, skewness, and
kurtosis) to be the same with or without an information structure. Furthermore,
successive drawings over time must (1) be independent and (2) be taken from the
same distribution. If returns follow a random walk, then the mean of the underlying
distribution does not change over time, and a fair game will result.
Most empirical evidence indicates that security returns do not follow a process
that has all the properties of a random walk. This makes sense because the condi-
tion that the entire underlying probability distribution of returns remain stationary
through time is simply too strong. It is reasonable to believe that because of changes
in the risk of a firm, the variance of stock returns will change over time. This, in fact,
appears to be the case. The fair-game model makes no statement about the variance
of the distribution of security returns, and consequently, the nonstationarity of return
variances is irrelevant to its validity."
A statistical difference between fair games and random walks is that the latter
hypothesis requires that all drawings be independently taken from the same distribu-
tion, whereas the former does not. This means that the random walk requires that
serial covariances between returns for any lag must be zero. However, significant
serial covariances of one-period returns are not inconsistent with a fair game. To see
this, suppose that the relevant information structure consists of past returns. In other
words, assume weak-form market efficiency. When Eq. (10.7) is written in returns
form, we have
Ej,t+ = Ej,t+ 1 — E(rj,,± I'm + 1, • • • (10.11)
and
gei,t+i = O.
Note that the fair game variable, Ei ,t+ ,, is the deviation of the return in period t + }
from its conditional expectation, i.e., the residual. If the residual is a fair game, then
it must have zero serial covariance for all lags. Yet even though the residual is a fair
game variable, the conditional expectation of returns for t + 1 can depend on the
return observed for t. Therefore the serial covariances of returns need not be zero.
The serial covariance for one-period returns is12
E[(r j ,t+ , — E(r j ,t+ , (r it — E(r jt ] = COV(r j ,, ± r jt
11
For example, consider a situation where random drawings are taken randomly from two normal distri-
butions that have a mean return of zero but different return variances. The expected value of a large
sample of alternative drawings would be zero; therefore we have a fair game. However, the experiment
violates the random walk requirement that all drawings be taken from the same distribution.
12
The reader who is unfamiliar with covariances is referred to Chapter 6. In general the covariance
between two random variables, x and y, is
COV(x, y) = E[(x — E(x (y — E(y ].
STATISTICAL TESTS UNADJUSTED FOR RISK 349
But the fair game in residuals, Eq. (10.11), does not imply that E(ri,, ± 1 vit) =
We have the result that the deviation of return for t 1 from its conditional expecta-
tion is a fair game, but the conditional expectation of return itself can depend on the
return observed for t. Therefore serial covariances of one-period returns are not in-
consistent with a fair game model. However, they are inconsistent with a random
walk because the latter requires that successive drawings be independent (a serial
covariance of zero for all lags).
Fama [1965] has presented evidence to show that the serial correlations of one-
day changes in the natural logarithm of price are significantly different from zero for
11 out of 30 of the Dow Jones Industrials.' Furthermore, 22 of the 30 estimated
serial correlations are positive. This, as well as evidence collected by other authors,
shows that security returns are not, strictly speaking, random walks. However, the
evidence is not inconsistent with fair-game models or, in particular, the submartingale.
Direct tests of the fair-game model were provided by Alexander [1961] and Fama
and Blume [1966]. They used a technical trading filter rule, which states: Using price
history, buy a stock if the price rises x%, hold it until the security falls x%, then
sell and go short. Maintain the short position until the price rises x%, then cover
the short position and establish a long position. This process is repeated for a fixed
time interval, and the performance according to the filter rule is then compared with
a buy-and-hold strategy in the same security. Because each security is compared with
itself, there is no need to adjust for risk.
Filter rules are designed to make the investor a profit if there are any systematic
patterns in the movement of prices over time. It is only a matter of trying enough
different filters so that one of them picks up any serial dependencies in prices and
makes a profit that exceeds the simple buy-and-hold strategy.
The filter rule tests have three important results. First, they show that even before
subtracting transactions costs, filters greater than 1.5% cannot beat a simple buy-
and-hold strategy. Second, filters below 1.5%, on the average, make very small profits
that because of frequent trading can beat the market. This is evidence of a very short-
term serial dependence in price changes. However, it is not necessarily evidence of
capital market inefficiency. First one must subtract from gross profits the cost of
taking action based on the filter rule. Fama and Blume [1966] show that even a
floor trader (the owner of a seat on the NYSE) must pay at least .1% per transaction.
Once these costs are deducted from the profits of filters that are less than 1.5%, the
profits vanish. Therefore the capital market is allocationally efficient down to the
13
To show that the logarithm of successive price changes is a good approximation of returns, assume
one-period continuous compounding:
Pt+ I = Pt e", where t = 1,
P,
P — —P
In Pt+ — In P, = t+ , where r=P
P, P,
350 EFFICIENT CAPITAL MARKETS: THEORY
level of transactions costs. The smaller the transactions costs are, the more opera-
tionally efficient the market is, and smaller price dependencies are eliminated by
arbitrage trading. Capital markets are efficient in their weak form because the return
on a portfolio managed with price-history information is the same as a buy-and-hold
strategy that uses no information. Therefore the value of messages provided by filter
rules is zero. Technical trading does not work."
The third inference that can be drawn from filter tests is that the market appears
to follow a submartingale. All the securities tested had average positive returns. This
makes sense because risky assets are expected to yield positive returns to compensate
investors for the risk they undertake.
Statistical tests and filter rules are interesting and present evidence of weak-form
efficiency but are limited by the fact that they cannot compare assets of different risk.
The CAPM provides a theory that allows the expected return of a fair-game model
to be conditional on a relevant costless measure of risk." If the CAPM is written
as a fair game, we have
ejt = Rjt E
(Rjt 3
1 jt),
14
See Ball [1978] for a discussion of filter rules and how to improve them as tests of market efficiency.
15
Note that the discussion that follows also applies the arbitrage pricing theory if one allows the expected
return to depend on multiple factor loadings (i.e., multiple betas).
" For a detailed explanation of the CAPM and empirical tests of it, see Chapter 7.
THE JOINT HYPOTHESIS OF MARKET EFFICIENCY AND THE CAPM 351
E(Rit )
Figure 10.5
The CAPM as a fair game.
expected return is interpreted as an abnormal return, eft , and can be taken as evidence
of market inefficiency if the CAPM is correct.
The CAPM is derived from a set of assumptions that are very similar to those
of market efficiency. For example, the Sharpe-Lintner-Mossin derivation of the
CAPM assumes:
• All investors are single-period expected utility of wealth maximizers whose utility
functions are based on the mean and variance of return.
• All investors can borrow or lend an indefinite amount at the risk-free rate, and
there are no restrictions on short sales.
• All investors have homogeneous expectations of the end-of-period joint distribu-
tions of returns.
• Securities markets are frictionless and perfectly competitive.
In Chapter 11 we shall report the results of several empirical studies that use the
CAPM as a tool for analyzing capital market efficiency. However, one should always
keep in mind the fact that the CAPM and capital market efficiency are joint and
inseparable hypotheses. If capital markets are inefficient, then the assumptions of the
CAPM are invalid and a different model is required. And if the CAPM is inappropriate,
even though capital markets are efficient, then the CAPM is the wrong tool to use
in order to test for efficiency.
Various sophisticated empirical tests of the CAPM by Black, Jensen, and Scholes
[1972], Black and Scholes [1974], and Fama and MacBeth [1973] show that the
CAPM fits reality surprisingly well. However, because the theoretical CAPM assumes
market efficiency, any empirical results that show that on the average there are no
significant deviations from the model are merely consistent with market efficiency.
They do not necessarily prove market efficiency because the model might be wrong.
Therefore any test of market efficiency that uses the CAPM to adjust for risk is, as
mentioned before, a joint test of the CAPM that assumes market efficiency for its
derivation and of market efficiency itself.
352 EFFICIENT CAPITAL MARKETS: THEORY
One may also ask the question: "If I can accurately predict systematic risk,
13;t, I can also predict the expected rate of return on an asset; doesn't this mean that
I can beat the market?" The answer, of course, is: "Probably not." If the information
necessary to estimate is publicly available and if markets are efficient in their semi-
strong form, then prices will instantaneously and fully reflect all the information
relevant for estimating flit, the expected return of the security will fall exactly on the
security line, and no abnormal returns will be observed.
Perhaps the most interesting use of the CAPM is to examine historical situations
to see whether or not the market was efficient for a particular set of information.
If the CAPM is valid (we shall assume it is, but keep in mind that it is a joint hy-
pothesis with market efficiency), then any evidence of persistent deviations from the
security market line can be interpreted as evidence of capital market inefficiency with
regard to a particular information set. Chapter 11 is devoted to tests of market effi-
ciency with regard to various information sets.
SUMMARY
The hypothesis of capital market efficiency has attracted a great deal of interest and
critical comment. This is somewhat surprising because capital market efficiency is a
fairly limited concept. It says that the prices of securities instantaneously and fully
reflect all available relevant information. It does not imply that product markets are
perfectly competitive or that information is costless.
Capital market efficiency relies on the ability of arbitrageurs to recognize that
prices are out of line and to make a profit by driving them back to an equilibrium
value consistent with available information. Given this type of behavioral paradigm,
one often hears the following questions: If capital market efficiency implies that no
one can beat the market (i.e., make an abnormal profit), then how can analysts be ex-
pected to exist since they, too, cannot beat the market? If capital markets are efficient,
how can we explain the existence of a multibillion dollar security analysis industry?
The answer, of course, is that neither of these questions is inconsistent with efficient
capital markets. First, analysts can and do make profits. However, they compete with
each other to do so. If the profit to analysis becomes abnormally large, then new
individuals will enter the analysis business until, on average, the return from analysis
equals the cost (which, by the way, includes a fair return to the resources that are
employed). As shown by Cornell and Roll [1981], it is reasonable to have efficient
markets where people earn different gross rates of return because they pay differing
costs for information. However, net of costs their abnormal rates of return will be
equal (to zero).
As we shall see in the next chapter, the concept of capital market efficiency is
important in a wide range of applied topics, such as accounting information, new
issues of securities, and portfolio performance measurement. By and large the evi-
dence seems to indicate that capital markets are efficient in the weak and semistrong
forms but not in the strong form.
PROBLEM SET 353
PROBLEM SET
10.1 Suppose you know with certainty that the Clark Capital Corporation will pay a dividend
of $10 per share on every January 1 forever. The continuously compounded risk-free rate is
5% (also forever).
a) Graph the price path of the Clark Capital common stock over time.
b) Is this (highly artificial) example a random walk? A martingale? A submartingale? (Why?)
10.2 Given the following situations, determine in each case whether or not the hypothesis of
an efficient capital market (semistrong form) is contradicted.
a) Through the introduction of a complex computer program into the analysis of past stock
price changes, a brokerage firm is able to predict price movements well enough to earn a
consistent 3% profit, adjusted for risk, above normal market returns.
b) On the average, investors in the stock market this year are expected to earn a positive
return (profit) on their investment. Some investors will earn considerably more than others.
c) You have discovered that the square root of any given stock price multiplied by the day
of the month provides an indication of the direction in price movement of that particular
stock with a probability of .7.
d) A Securities and Exchange Commission (SEC) suit was filed against Texas Gulf Sulphur
Company in 1965 because its corporate employees had made unusually high profits on
company stock that they had purchased after exploratory drilling had started in Ontario
(in 1959) and before stock prices rose dramatically (in 1964) with the announcement of
the discovery of large mineral deposits in Ontario.
10.3 The First National Bank has been losing money on automobile consumer loans and is
considering the implementation of a new loan procedure that requires a credit check on loan
applicants. Experience indicates that 82% of the loans were paid off, whereas the remainder
defaulted. However, if the credit check is run, the probabilities can be revised as follows:
Favorable Unfavorable
Credit Check Credit Check
Loan is paid .9 .5
Loan is defaulted .1 .5
An estimated 80% of the loan applicants receive a favorable credit check. Assume that the
bank earns 18% on successful loans, loses 100% on defaulted loans, suffers an opportunity
cost of 18% when the loan is not granted but would have been successful, and an opportunity
cost of 0% when the loan is not granted and would have defaulted. If the cost of a credit check
is 5% of the value of the loan and the bank is risk neutral, should the bank go ahead with
the new policy?
10.4 Hearty Western Foods, one of the nation's largest consumer products firms, is trying
to decide whether it should spend $5 million to test market a new ready-to-eat product (called
Kidwich), to proceed directly to a nationwide marketing effort, or to cancel the product. The
expected payoffs (in millions of dollars) from cancellation vs. nationwide marketing are given
354 EFFICIENT CAPITAL MARKETS: THEORY
below:
Action
No acceptance 0 —10
Marginal 0 10
Success 0 80
No acceptance .6
Marginal .3
Success
If the firm decides to test market the product, the following information will become available:
Probability
No Acceptance .9 .1 0
Outcome Predicted
Marginal .1 .7 .2
by the Test Market
Success .1 .3 .6
For example, if the test market results predict a success, there is a 60% chance that the nation-
wide marketing effort really will be a success but a 30% chance it will be marginal and a
10% chance it will have no acceptance.
a) If the firm is risk neutral, should it test market the product or not?
b) If the firm is risk averse with a utility function
U(W)= In(W + 11),
should it test market the product or not?
10.5 The efficient market hypothesis implies that abnormal returns are expected to be zero.
Yet in order for markets to be efficient, arbitrageurs must be able to force prices back into
equilibrium. If they earn profits in doing so, is this fact inconsistent with market efficiency?
10.6
a) In a poker game with six players, you can expect to lose 83% of the time. How can
this still be a martingale?
b) In the options market, call options expire unexercised over 80% of the time." Thus the
option holders frequently lose all their investment. Does this imply that the options market
is not a fair game? Not a martingale? Not a submartingale?
17
See Chapter 8 for a description of call options.
REFERENCES 355
10.7 If securities markets are efficient, what is the NPV of any security, regardless of its risk?
10.8 From time to time the federal government considers passing into law an excess profits
tax on U.S. corporations. Given what you know about efficient markets and the CAPM, how
would you define excess profits? What would be the effect of an excess profits tax on the
investor?
10.9 State the assumptions inherent in this statement: A condition for market efficiency is that
there be no second-order stochastic dominance.
REFERENCES
Alexander, S. S., "Price Movements in Speculative Markets: Trends or Random Walks," In-
dustrial Management Review, May 1961, 7-26.
Ball, R., "Filter Rules: Interpretation of Market Efficiency, Experimental Problems and Aus-
tralian Evidence," Accounting Education, November 1978, 1-17.
Black, F.; M. Jensen; and M. Scholes, "The Capital Asset Pricing Model: Some Empirical
Tests," in Jensen, ed., Studies in the Theory of Capital Markets. Praeger, New York, 1972,
79-124.
Black, F., and M. Scholes, "The Effects of Dividend Yield and Dividend Policy on Common
Stock Prices and Returns," Journal of Financial Economics, May 1974, 1-22.
Copeland, T. E., and D. Friedman, "The Market for Information: Some Experimental Results,"
Working Paper 5-87, UCLA Graduate School of Management, 1986.
Cornell, B., and R. Roll, "Strategies for Pairwise Competitions in Markets and Organizations,"
Bell Journal of Economics, Spring 198.1, 201-213.
Fama, E. F., "The Behavior of Stock Market Prices," Journal of Business, January 1965, 34-105.
, "Efficient Capital Markets: A Review of Theory and Empirical Work," Journal of
Finance, May 1970, 383-417.
, Foundations of Finance. Basic Books, New York, 1976.
, and M. Blume, "Filter Rules and Stock Market Trading Profits," Journal of Business,
January (spec. supp.) 1966, 226-241.
Fama, E. F., and J. MacBeth, "Risk, Return and Equilibrium: Empirical Test," Journal of
Political Economy, May—June 1973, 607-636.
Finnerty, J. E., "Insiders and Market Efficiency," Journal of Finance, September 1976, 1141-
1148.
Forsythe, R.; T. Palfrey; and C. R. Plott, "Asset Valuation in an Experimental Market," Econo-
metrica, May 1982, 537-567.
Green, J. R., "Information, Efficiency and Equilibrium," Discussion Paper No. 284, Harvard
Institute of Economic Research, March 1974.
Grossman, S. J., "On the Efficiency of Competitive Stock Markets Where Trades Have Diverse
Information," Journal of Finance, May 1976, 573-586.
, and J. Stiglitz, "Information and Competitive Price Systems," American Economic
Review, May 1976, 246-253.
, "The Impossibility of Informationally Efficient Markets," American Economic Review,
June 1980, 393-408.
3P6 EFFICIENT CAPITAL MARKETS: THEORY
Harrison, J. M., and D. M. Kreps, "Speculative Investor Behavior in a Stock Market with
Heterogeneous Expectations," Qua]te]ly Jou]nal of Economics, May =978, 3<3-336.
Hayek, F. H., "The Use of Knowledge in Society," Ame]ican Economic Review, Septemîer
=94P.
Hirshleifer, J., Investment, Inte]est, and Capital. Prentice-Hall, Englewood Cliffs, N.J., =970.
, and J. Riley, "The Analytics of Uncertainty and Information—An Expository Survey,"
Jou]nal of Economic Lite]atu]e, Decemîer =979, =37P-=4<=.
Huang, C. C.; I. Vertinsky; and W. T. Ziemîa, "Sharp Bounds on the Value of Perfect Infor-
mation," Ope]ations Resea]ch, January–Feîruary =977, =<8-=39.
Jaffe, J., "The Effect of Regulation Changes on Insider Trading," Bell Jou]nal of Economics
and Management Science, Spring =974, 93-=<=.
Keynes, J. M., The Gene]al Theo]y of Employment, Inte]est and Money. Harcourt Brace, New
York, =936.
Latham, M., "Defining Capital Market Efficiency," Finance Working Paper =P0, Institute for
Business and Economic Research, University of California, Berkeley, April =98P.
Lucas, R. E., "Expectations and the Neutrality of Money," Jou]nal of Economic Theo]y, April
=97<, =03-=<4.
Marschak, J., Economic Info]mation, Decisions, and P]edictions, Selected Essays, Vol. <. Reidel,
Boston, =974.
Miller, R. M.; C. R. Plott; and V. L. Smith, "Intertemporal Competitive Equiliîrium: An
Empirical Study of Speculation," Ame]ican Economic Review, June =98=, 448-4P9.
Plott, C. R., and S. Sunder, "Efficiency of Experimental Security Markets with Insider Infor-
mation: An Application of Rational Expectations Models," Jou]nal of Political Economy,
August =98<, 663-698.
Ruîinstein, M., "Securities Market Efficiency in an Arrow-Deîreu Economy," Ame]ican Eco-
nomic Review, Decemîer =97P, 8=<-8<4.
Samuelson, P. A., "Proof that Properly Anticipated Prices Fluctuate Randomly," Indust]ial
Management Review, Spring =96P, 4=-49.
Smith, V. L., "Experimental Economics: Induced Value Theory," Ame]ican Economic Review,
May =976, <74-<79.
Sunder, S., "Market for Information: Experimental Evidence," Working Paper, University of
Minnesota, =984.
PART II
Corporate
Policy: Theory,
Evidence, and
Applications
T
HE FIRST PART OF THE text covers most of what
has come to be recognized as a unified theory of decision
making under uncertainty as applied to the field of finance.
The theory of finance, as presented in the first half of the text, is applicable to a wide
range of finance topics. The theoretical foundations are prerequisite to almost any
of the traditional subject areas in finance curricula; e.g., portfolio management,
corporation finance, commercial banking, money and capital markets, financial in-
stitutions, security analysis, international finance, investment banking, speculative
markets, insurance, and case studies in finance. Since all these topics require a thorough
understanding of decision making under uncertainty, all use the theory of finance.
The second half of this text focuses, for the most part, on applications of the
theory of finance to a corporate setting. The fundamental issues are: Does financing
matter? Does the type of financing (debt or equity) have any real effect on the value
of the firm? Does the form of financial payment (dividends or capital gains) have any
effect on the value of claims held by various classes of security holders?
Because these issues are usually discussed in the context of corporate finance
they may seem to be narrow. This is not the case. First of all, the definition of a
corporation is very broad. The class of corporations includes not only manufacturing
firms but also commercial banks, savings and loan associations, many brokerage
houses, some investment banks, and even the major security exchanges. Second, the
debt equity decision applies to all individuals as well as all corporations. Therefore
although the language is narrow, the issues are very broad indeed. They affect almost
every economic entity in the private sector of the economy.
357
358 CORPORATE POLICY: THEORY, EVIDENCE, AND APPLICATIONS
As we shall see, the theoretical answer to the question "Does financing matter?"
is often a loud and resounding "Maybe." Often the answer depends on the assumptions
of the model employed to study the problem. Under different sets of assumptions,
different and even opposite answers are possible. This is extremely disquieting to the
student of finance. Therefore we have presented empirical evidence related to each
of the theoretical hypotheses. Frequently, but not always, the preponderance of evi-
dence supports a single conclusion.
It is important to keep in mind that hypotheses cannot be tested by the realism
of the assumptions used to derive them. What counts for a positive science is the
development of theories that yield valid and meaningful predictions about observed
phenomena. On the first pass, what counts is whether or not the hypothesis is con-
sistent with the evidence at hand. Further testing involves deducing new facts capable
of being observed but not previously known, then checking those deduced facts
against additional empirical evidence. As students of finance, which seeks to be a
positive science, we must not only understand the theory, but also study the empirical
evidence in order to determine which hypothesis is validated.
Chapter 11 is devoted to various empirical studies related to the efficient market
hypothesis. Most of the evidence is consistent with the weak and semistrong forms
of market efficiency but inconsistent with the strong form. In certain situations,
individuals with inside information appear to be able to earn abnormal returns. In
particular, corporate insiders can beat the market when trading in the securities of
their firm. Also, block traders can earn abnormal returns when they trade at the
block price, as can purchasers of new equity issues. The last two situations will surely
lead to further research because current theory cannot explain why, in the absence
of barriers to entry, there appear to be inexplicable abnormal rates of return. Chapter
12 returns to the theoretical problem of how to evaluate multiperiod investments in
a world with uncertainty. It shows the set of assumptions necessary in order to extend
the simple one-period CAPM rules into a multiperiod world. It also discusses two
interesting applied issues: the abandonment problem, and the technique for dis-
counting uncertain costs.
Chapter 13 explores the theory of capital structure and the cost of capital. This
is the first of the corporate policy questions that relate to whether or not the value
of the firm is affected by the type of financing it chooses. Also, we define a cost of
capital that is consistent with the objective of maximizing the wealth of the current
shareholders of the firm. This helps to complete, in a consistent fashion, the theory
of project selection. Capital budgeting decisions that are consistent with shareholder
wealth maximization require use of the correct technique (the NPV criterion), the
correct definition of cash flows (operating cash flows after taxes), and the correct cost
of capital definition.
Chapter 14 discusses empirical evidence on whether or not the debt-to-equity
ratio (i.e., the type of financing) affects the value of the firm. This is one of the most
difficult empirical issues in finance. Although not conclusive, the evidence is consistent
with increases in the value of the firm resulting from increasing debt (up to some
range) in the capital structure. However, much work remains to be done in this area.
Chapter 14 also provides a short example of how to actually compute the cost of
capital.
CORPORATE POLICY: THEORY, EVIDENCE, AND APPLICATIONS 359
Chapter 15 looks at the relationship between dividend policy and the value of
the firm. There are several competing theories. However, the dominant argument
seems to be that the value of an all-equity firm depends on the expected returns from
current and future investment and not on the form in which the returns are paid out.
If investment is held constant, it makes no difference whether the firm pays out high
or low dividends. On the other hand, a firm's announcement of increase in dividend
payout may be interpreted as a signal by shareholders that the firm anticipates per-
manently higher levels of return from investment, and of course, higher returns on
investment will result in higher share prices.
Chapter 16 presents empirical evidence on the relationship between dividend
policy and the value of the firm that, for the most part, seems to be consistent with
the theory—namely, that dividend policy does not affect shareholders' wealth. The
chapter also applies the valuation models (presented in Chapter 15) to an example.
Chapter 17 uses the subject of leasing to bring together a number of further
applications of capital structure and cost of capital issues. We also illustrate how
option pricing can help clarify the nature of an operating lease under which the lessee
may exercise a contractual right to cancel (with some notice and with moderate
penalties). Chapter 18 discusses several applied topics of interest to chief financial
officers pension-fund management, executive compensation, leveraged buyouts,
ESOP's and interest rate swaps.
Chapters 19 and 20 consider the widespread phenomenon of mergers. They begin
with the proposition that without synergy, value additivity holds in mergers as it
does in other types of capital budgeting analysis. Mergers do not affect value unless
the underlying determinants of value—the patterns of future cash flows or the ap-
plicable capitalization factors are changed by combining firms. Empirical tests of
mergers indicate that the shareholders of acquired firms benefit, on the average, but
the shareholders of acquiring firms experience neither significant benefit nor harm.
Chapters 21 and 22 conclude the book by placing finance in its increasingly
important international setting. A framework for analyzing the international financial
decisions of business firms is developed by summarizing the applicable fundamen-
tal propositions. The Fisher effect, which states that nominal interest rates reflect
anticipated rates of inflation, is carried over to its international implications. This
leads to the Interest Rate Parity Theorem, which states that the current forward
exchange rate for a country's currency in relation to the currency of another country
will reflect the present interest rate differentials between the two countries. The
Purchasing Power Parity Theorem states that the difference between the current spot
exchange rate and the future spot exchange rate of a country's currency in relation
to the currency of another country will reflect the ratio of the rates of price changes
of their internationally traded goods. We point out that exchange risk is a "myth"
in the sense that departures from fundamental parity theorems reflect changes in
underlying demand and supply conditions that would cause business risks even if
international markets were not involved. The fundamental relations provide the
principles to guide firms in adjusting their policies to the fluctuations in the exchange
rate values of the currencies in which their business is conducted.
The only valid statement is that the current price embodies all
knowledge, all expectations and all discounts that infringe upon the
market.
C. W. J. Granger and 0. Morgenstern, Predictability of Stock Market
Prices, Heath Lexington Books, Lexington, Mass., 1970, 20
Empirical evidence for or against the hypothesis that capital markets are efficient
takes many forms. This chapter is arranged in topical order rather than chronological
order, degree of sophistication, or type of market efficiency being tested. Not all the
articles mentioned completely support the efficient market hypothesis. However, most
agree that capital markets are efficient in the weak and semistrong forms but not in
the strong form. The majority of the studies are very recent, dating from the late 1960s
and continuing up to the most recently published papers. Usually capital market
efficiency has been tested in the large and sophisticated capital markets of developed
countries. Therefore one must be careful to limit any conclusions to the appropriate
arena from which they are drawn. Research into the efficiency of capital markets is
an ongoing process, and the work is being extended to include assets other than com-
mon stock as well as smaller and less sophisticated marketplaces.
Before discussing the empirical tests of market efficiency it is useful to review the three
basic types of empirical models that are frequently employed. The differences between
them are important. The simplest model, called the market model, simply argues that
361
362 EFFICIENT CAPITAL MARKETS: EVIDENCE
B. ACCOUNTING INFORMATION
Market efficiency requires that security prices instantaneously and fully reflect all
available relevant information. But what information is relevant? And how fast do
security prices really react to new information? The answers to these questions are
of particular interest to corporate officers who report the performance of their firm
to the public; to the accounting profession, which audits these reports; and to the
Securities and Exchange Commission, which regulates securities information.
The market value of assets is the present value of their cash flows discounted at
the appropriate risk-adjusted rate. Investors should care only about the cash flow
ACCOUNTING INFORMATION 363
those occurring plus or minus 12 months around the announcement of the inventory-
accounting change. He then used the estimated ,6";„ the actual risk-free rate, and the
actual market return during the 24-month period around the announcement date to
predict the expected return.' Differences between estimated and actual returns were
then averaged across all companies for each month. The average abnormal return in
a given month is
1 N
AR E = — e., where N = the number of companies.
N 1=1
The cumulative average return (CAR) is the sum of average abnormal returns over
all months from the start of the data up to and including the current month, T:
CAR = AR E ,
E=t
where
T = the number of months being summed (T = 1, 2, . . , M),
M = the total number of months in the sample.
If there were no abnormal change in the value of the firm associated with the switch
from FIFO to LIFO, we should observe no pattern in the residuals. They would
fluctuate around zero and on the average would equal zero. In other words, we would
have a fair game. Figure 11.1 shows Sunder's results. Assuming that risk does not
change during the 24-month period, the cumulative average residuals for the firms
switching to LIFO rise by 5.3% during the 12 months prior to the announcement of
the accounting change. This is consistent with the fact that shareholders actually value
cash flow, not eps. However, it does not necessarily mean that a switch to LIFO
causes higher value. Almost all studies of this type, which focus on a particular phe-
nomenon, suffer from what has come to be known as postselection bias. In this case,
firms may decide to switch to LIFO because they are already doing well and their
value may have risen for that reason, not because of the switch in accounting method.
Either way, Sunder's results are inconsistent with the fact that shareholders look only
at changes in eps in order to value common stock. He finds no evidence that the
switch to LIFO lowered value even though it did lower eps.
More recently Ricks [1982] studied a set of 354 NYSE- and AMEX-listed firms
that switched to LIFO in 1974. He computed their earnings "as if" they never
switched and found that the firms that switched to LIFO had an average 47% increase
in their as-if earnings, whereas a matched sample of no-change firms had an average
2% decrease. Ricks also found that the abnormal returns of the switching firms were
significantly lower than the matched sample of no-change firms. These results are
inconsistent with those reported by Sunder.
The studies above indicate that investors in efficient markets attempt to evaluate
news about the effect of managerial decisions on cash flows not on eps. This fact has
Sunder used a moving-average beta technique in his second study [1975]. However, it did not substan-
tially change his results.
ACCOUNTING INFORMATION 365
.125 .125
110 firms switching 22 firms' switching
to LIFO from LIFO
• •
.075 .075
•
..•• •••••••••
•• •••• ••
0
72
0 .025 - vir 4D.
— ...
•••• • 0.• •
—.125 L I —.0125 ù ù
—12 —7 —2 3 8 13 —12 —7 —2 3 8 13
Months from the date of change Months from the date of change
Figure 11.1
Cumulative average residuals for 24 months around the
accounting change. (From S. Sunder, "Relationship between
Accounting Changes and Stock Prices: Problems of
Measurement and Some Empirical Evidence," reprinted from
Empirical Research in Accounting: Selected Studies, 1973, 18.)
direct implications for the accounting treatment of mergers and acquisitions. Two
types of accounting treatment are possible: pooling or purchase. In a pooling arrange-
ment the income statements and balance sheets of the merging firms are simply added
together. On the other hand, when one company purchases another, the assets of
the acquired company are added to the acquiring company's balance sheet along with
an item called goodwill. Goodwill is the difference between the purchase price and
the book value of the acquired company's assets. Regulations require that goodwill
be written off as a charge against earnings after taxes in a period not to exceed 40
years. Because the writeoff is after taxes, there is no effect on cash flows, but reported
eps decline. The fact that there is no difference in cash flows between pooling and
purchase and the fact that cash flows, not eps, are the relevant information used by
investors to value the firm should convey to management the message that the ac-
counting treatment of mergers and acquisitions is a matter of indifference.' Yet many
managements prefer pooling, presumably because they do not like to see eps decline
owing to the writeoff of goodwill. No economically rational basis for this type of
behavior can be cited.
In a recent empirical study tiong, Kaplan, and Mandelker [1978] tested the effect
of pooling and purchase techniques on stock prices of acquiring firms. Using monthly
Prior to the 1986 Tax Reform Act, the Internal Revenue Service (IRS) allowed the book value of the
assets of the acquired firm to be written up upon purchase. This reduced the amount of goodwill created,
but even more important, it created a depreciation tax shield that did not exist in a pooling arrangement.
Therefore cash flows for purchase were often higher than pooling. In these cases purchase was actually
preferable to pooling, at least from the point of view of the acquiring firm.
366 EFFICIENT CAPITAL MARKETS: EVIDENCE
data between 1954 and 1964, they compared a sample of 122 firms that used pooling
and 37 that used purchase. The acquired firm had to be at least 3% of the net asset
value of the acquiring firm. Mergers were excluded from the sample if another merger
took place within 18 months, if the acquiring firm was not NYSE listed, or if the
merger terms were not based on an exchange of shares. (This last criterion rules out
taxable mergers.)
Using the simple time-series market model given below, they calculated cumula-
tive abnormal residuals:
ln = oci + ln + ui„
where
= return on the jth security in time period t,
o ; = an intercept term assumed to be constant over the entire time period,
=6'; = systematic risk assumed to be constant over the entire time period,
Rmt = market return in time period t,
nit = abnormal return for the jth security in time period t.
When the cumulative average residuals were centered around the month of the actual
merger, the patterns revealed no evidence of abnormal performance for the sample
of 122 poolings. This is shown in Fig. 11.2. Therefore there is no evidence that "dirty
pooling" raises the stock prices of acquiring firms. Investors are not fooled by the
accounting convention.
These results are just as important for acquiring firms that had to write off good-
will against their after-tax earnings because they used the purchase technique. As
shown in Fig. 11.3, there is no evidence of negative abnormal returns, which is what
we would expect if investors looked at eps. Instead, there is weak evidence that share-
holders of acquiring firms experienced positive abnormal returns when the purchase
technique was used. This is consistent with the hypothesis that investors value cash
flows and that they disregard reported eps.
The empirical studies of Sunder [1973, 1975], and Hong, Kaplan, and Man-
delker [1978] provide evidence on what is meant by relevant accounting information.
Cumulative
abnormal
residual
.04—
.02 — •
• • • • • • •
0•• • • • • °•• ••• •:•••
—.02 v.:.
4
—.0 -- —50 —40 —30 —20 —10 10 20 30 40 50 60
Figure 11.2
Cumulative abnormal residuals for 122 poolings with market value greater than book
value in the month relative to merger. (From H. Hong, R. S. Kaplan, and G. Mandelker,
"Pooling vs. Purchase: The Effects of Accounting for Mergers on Stock Prices," reprinted
with permission of Accounting Review, January 1978, 42.)
ACCOUNTING INFORMATION 367
Cumulative
abnormal •
residual •
• •••• .
.04— • • e• •
••
• •
•
.
•
• • ..
•-• -0 41.
0.
• • ••
• • . % . .
.02
• • •
••• . • %
• •• o„
0
•
• •1
II. 1.• I I I •r• i• i I I
••• ••• • •• .0
••••
—.02 • •
•
By relevant we mean any information about the expected distribution of future cash
flows. Next, a study by Ball and Brown [1968] provides some evidence about the
speed of adjustment of efficient markets to new information.
Earnings data and cash flows are usually highly correlated. The examples dis-
cussed above merely serve to point out some situations where they are not related
and therefore allow empiricists to distinguish between the two. Ball and Brown used
monthly data for a sample of 261 firms between 1946 and 1965 to evaluate the useful-
ness of information in annual reports. First, they separated the sample into companies
that had earnings that were either higher or lower than those predicted by a naive
time series model. Their model for the change in earnings was
AN/i, = a + 1), Arn, + e ft , (11.2)
where
ANIii = the change in earnings per share for the jth firm,
Am, = the change in the average eps for all firms (other than firm j) in the market.
Next, this regression was used to predict next year's change in earnings, AN/ j,,,,:
ANI;,, +1 = a + b1 Amt +1, (11.3)
where
a, b = coefficients estimated from time series fits of Eq. (11.2) to the data,
Am t +i = the actual change in market average eps during the (t + 1)th time period.
Finally, estimated earnings changes were compared with actual earnings changes. If
the actual change was greater than estimated, the company was put into a portfolio
where returns were expected to be positive, and vice versa.
Figure 11.4 plots an abnormal performance index (API) that represents the value
of $1 invested in a portfolio 12 months before an annual report and held for T
368 EFFICIENT CAPITAL MARKETS: EVIDENCE
Abnormal
performance
index
1.12—
1.10
Variable 2
1.08
1.06
1.04
1.02
1.00
Total sample
0.98
0.96
0.94
0.92
0.90
0.88
I I
12 10 8 6 4 2 0 2 4 6
Month relative to Annual Report announcement date
Figure 11.4
Abnormal performance index of portfolios chosen
on the basis of differences between actual and
predicted accounting income. (From R. Ball and
P. Brown, "An Empirical Evaluation of Accounting
Income Numbers," reprinted with permission of
Journal of Accounting Research, Autumn 1968, 169.)
where
N= the number of companies in a portfolio,
T= 1, 2, . . . , 12,
sit = abnormal performance measured by deviations from the market model.
A quick look at Fig. 11.4 shows that when earnings are higher than predicted, returns
are abnormally high. Furthermore, returns appear to adjust gradually until, by the
time of the annual report, almost all the adjustment has occurred. Most of the infor-
ACCOUNTING INFORMATION 369
mation contained in the annual report is anticipated by the market before the annual
report is released. In fact, anticipation is so accurate that the actual income number
does not appear to cause any unusual jumps in the API in the announcement month.
Most of the content of the annual report (about 85% to 90%) is captured by more
timely sources of information. Apparently market prices adjust continuously to new
information as it becomes publicly available throughout the year. The annual report
has little new information to add.
These results suggest that prices in the marketplace continuously adjust in an
unbiased manner to new information. Two implications for the corporate treasurers
are: (1) significant new information, which will affect the future cash flows of the firm,
should be announced as soon as it becomes available so that shareholders can use
it without the (presumably greater) expense of discovering it from alternative sources;
and (2) it probably does not make any difference whether cash flow effects are reported
in the balance sheet, the income statement, or footnotes—the market can evaluate
the news as long as it is publicly available, whatever form it may take.
The Ball and Brown study raised the question of whether or not annual reports
contain any new information. More recent studies by Aharony and Swary [1980],
Joy, Litzenberger and McEnally [1977], and Watts [1978] have focused on quarterly
earnings reports where information revealed to the market is (perhaps) more timely
than annual reports.' They typically use a time series model to predict quarterly
earnings, then form two portfolios of equal risk, one consisting of firms with earnings
higher than predicted and the other of firms with lower than expected earnings. The
combined portfolio, which is long in the higher than expected earnings firms and
short in the lower than expected earnings firms, is a zero-beta portfolio that (in per-
fect markets) requires no investment. It is an arbitrage portfolio and should have
zero expected return. Watts [1978] finds a statistically significant return in the quarter
of the announcement a clear indication that quarterly earnings reports contain new
information. However, he also finds a statistically significant return in the following
quarter and concludes that "the existence of those abnormal returns is evidence that
the market is inefficient."
Quarterly earnings reports are sometimes followed by announcements of divi-
dend changes, which also affect the stock price. To study this problem, Aharony and
Swary [1980] examine all dividend and earnings announcements within the same
quarter that are at least 11 trading days apart. They conclude that both quarterly
earnings announcements and dividend change announcements have statistically sig-
nificant effects on the stock price. But more important they find no evidence of market
inefficiency when the two types of announcement effects are separated. They used
daily data and Watts [1978] used quarterly data, so we cannot be sure that the con-
clusions of the two studies regarding market inefficiency are inconsistent. All we can
say is that unexpected changes in quarterly dividends and in quarterly earnings both
have significant effects on the value of the firm and that more research needs to be
done on possible market inefficiencies following the announcement of unexpected
earnings changes.
See also articles by Brown [1978], Griffin [1977], and Foster [1977].
370 EFFICIENT CAPITAL MARKETS: EVIDENCE
Using intraday records of all transactions for the common stock returns of 96
(large) firms, Paten and Wolfson [1984] were able to estimate the speed of market
reaction to disclosures of dividend and earnings information. In a simple trading
rule, they bought (sold short) stocks whose dividend or earnings announcements ex-
ceeded (fell below) what had been forecast by Value Line Investor Service. The initial
price reactions to earnings and dividend change announcements begin with the first
pair of price changes following the appearance of the news release on the Broad Tape
monitors. Although there was a hint of some activity in the hour or two preceding
the Broad Tape news release, by far the largest portion of the price response occurs
in the first 5 to 15 minutes after the disclosure. Thus, according to Patell and Wolfson,
the market reacts to unexpected changes in earnings and dividends, and it reacts very
quickly.
C. BLOCK TRADES
During a typical day for an actively traded security on a major stock exchange,
thousands of shares will be traded, usually in round lots ranging between one hundred
and several hundred shares. However, occasionally a large block, say 10,000 shares
or more, is brought to the floor for trading. The behavior of the marketplace during
the time interval around the trading of a large block provides a "laboratory" where
the following questions can be investigated: (1) Does the block trade disrupt the
market? (2) If the stock price falls when the block is sold, is the fall a liquidity effect,
an information effect, or both? (3) Can anyone earn abnormal returns from the fall in
price? (4) How fast does the market adjust to the effects of a block trade?
In perfect (rather than efficient) capital markets all securities of equal risk are
perfect substitutes for each other. Because all individuals are assumed to possess the
same information and because markets are assumed to be frictionless, the number of
shares traded in a given security should have no effect on its price. If markets are less
than perfect, the sale of a large block may have two effects (see Fig. 11.5). First, if it
is believed to carry with it some new information about the security, the price will
change (permanently) to reflect the new information. As illustrated in parts (c) and (d)
of Fig. 11.5, the closing price is lower than the opening price and it remains low
permanently. 4 Second, if buyers must incur extra costs when they accept the block,
there may be a (temporary) decline in price to reflect what has been in various articles
described as a price pressure, or distribution effect, or liquidity premium, as shown in
parts (a) and (c). Figure 11.5 depicts how hypothesized information or price pressure
effects can be expected to show up in continuous transactions data. For example, if
the sale of a large block has both effects [Fig. 11.5(c)], we may expect the price to
fall from the price before the trade ( — T) to the block price (BP) at the time of the
block trade (BT), then to recover quickly from any price pressure effect by the time of
4
The permanent decline in price is also tested by looking at the pattern of day-to-day closing prices. Ev
dence on this is reported in Fig. 11.6.
BLOCK TRADES 371
Price Price
BP
Time Time
Open —T BT +T Close Open —T BT +T Close
(a) (b)
Price Price
New information
BP ---fi I
I I I
Time Time
Open —T BT +T Close Open —T BT +T Close
(c) (d)
Figure 11.5
Competing hypotheses of price behavior around the sale of a large block.
the next trade ( + T) but to remain at a permanently lower level, which reflects the
impact of new information on the value of the security.
Scholes [1972] and Kraus and Stoll [1972] provided the first empirical evidence
about the price effects of block trading. Scholes used daily returns data to analyze
345 secondary distributions between July 1961 and December 1965. Secondary dis-
tributions, unlike primary distributions, are not initiated by the company but by
shareholders who will receive the proceeds of the sale. The distributions are usually
underwritten by an investment banking group that buys the entire block from the
seller. The shares are then sold on a subscription basis after normal trading hours.
The subscriber pays only the subscription price and not stock exchange or brokerage
commissions. Figure 11.6 shows an abnormal performance index based on the market
model and calculated for 40 trading days around the date of a secondary distribution.
The abnormal performance index falls from an initial level of 1.0 to a final value of
.977 just 14 days after the sale, a decline of 2.2%. On the day of the secondary dis-
tribution, the average abnormal performance was —.57.. Because this study uses only
close-to-close daily returns data, it focuses only on permanent price changes. We have
characterized these as information effects [Fig. 11.5(c) and (d)]. Further evidence that
the permanent decline in price is an information effect is revealed when the API is
partitioned by vendor classification. These results appear in Table 11.2.
On the day of the offering the vendor is not usually known, but we may presume
that the news becomes available soon thereafter. One may expect that an estate liqui-
dation or portfolio rebalancing by a bank or insurance company would not be moti-
vated by information about the performance of the firm. On the other hand, corporate
372 EFFICIENT CAPITAL MARKETS: EVIDENCE
Performance index
1.05 —
1.04 —
Market adjustments
1.03 —
1.02 -
1.01—
1.00 — .........................
.99 — •
.98— . ••
• ..........
.97 -
.96 -
.95 I I
—25 —20 —15 —10 —5 0 5 10 15
Day relative to distribution day
Figure 11.6
Abnormal performance index on days around a
secondary distribution. (From M. Scholes, "The Market
for Securities: Substitution vs. Price Pressure and the
Effects of Information on Share Prices," reprinted with
permission of Journal of Business, April 1972, 193.)
Copyright © 1972, The University of Chicago Press.
API
No. of
Observations —10 to 0 to
in Sample Category +10 Days + 10 Days
192 Investment companies and mutual funds —2.57. —1.47.
31 Banks and insurance companies —.3 —0.0
36 Individuals —1.1 —.7
23 Corporations and officers —2.9 — 2.1
50 Estates and trusts — .7 —.5
From M. Scholes, "The Market for Securities: Substitution vs. Price Pressure and the Effects of Infor-
mation on Share Prices," reprinted with permission of Journal of Business, April 1972, 202. Copyright
© 1972, The University of Chicago Press.
insiders as well as investment companies and mutual funds (with large research staffs)
may be selling on the basis of adverse information. The data seem to support these
suppositions. Greater price changes after the distribution are observed when the seller
is presumed to have a knowledgeable reason for trading.'
Mikkelson and Partch [1985] studied a sample of 146 registered and 321 non-
registered secondary offerings between 1972 and 1981. Using daily data, they find an
5
A second test performed by Scholes showed that there was no relationship between the size of the dis-
tribution (as a percentage of the firm) and changes in the API on the distribution date. This would lead
us to reject the hypothesis that investment companies and mutual funds may have had an impact because
they sold larger blocks.
BLOCK TRADES 373
Price (logarithmic)
•
Closing price (P_ }
\
1.8610% (Et)
•Price prior to block (PPB)
T Closing price (Po)
• --t--
1.1380% (E3)
.7130% (E2)
✓ i • 1
Block Price (BP)
• Time
End of day —1 End of day 0
Figure 11.7
Intraday price impacts of block trading. (From A. Kraus
and H. R. Stoll, "Price Impacts of Block Trading on the
New York Stock Exchange," reprinted with permission
of Journal of Finance, June 1972, 575.)
374 EFFICIENT CAPITAL MARKETS: EVIDENCE
have a block price of $49.43, but by the end of the day the price would have recovered
to $49.79.
The Scholes and Kraus-Stoll studies find evidence of a permanent price decline
that is measured by price drops from the closing price the day before the block trade
to the closing price the day of the block transaction. These negative returns seem to
persist for at least a month after the block trade. In addition, Kraus and Stoll found
evidence of temporary intraday price pressure effects. The implications of these find-
ings are discussed by Dann, Mayers, and Raab [1977], who collected continuous
transactions data during the day of a block trade for a sample of 298 blocks between
July 1968 and December 1969. The open-to-block price decline was at least 4.56% for
each block in the sample. The reason for restricting the sample to blocks with large
price declines was to provide the strongest test of market efficiency. If an individual
or a group of investors can establish a trading rule that allows them to buy a block
whose open-to-block price change is at least 4.56%, then sell at the end of the day,
they may be able to earn abnormal profits. This would be evidence of capital market
inefficiency.
Testing a trading rule of this type takes great care. Normally, a block trade is not
made publicly available until the trade has already been consummated and the trans-
action is recorded on the ticker. The semistrong form of market efficiency is based
on the set of publicly available information. Therefore a critical issue is: Exactly how
fast must we react after we observe that our —4.56% trading rule has been activated
by the first publicly available announcement that occurs on the ticker tape? Figure
11.8 shows annualized rates of return using the —4.56% rule with the purchase made
x minutes after the block and the stock then sold at the close. Returns are net of
actual commissions and New York State transfer taxes. For both time periods that
are reported, we would have to react in less than five minutes in order to earn a
positive return. Such a rapid reaction is, for all practical purposes, impossible. It
seems that no abnormal returns are available to individuals who trade on publicly
available information about block trades because prices react so quickly. Fifteen
minutes after the block trade, transaction prices have completely adjusted to unbiased
estimates of closing prices. This gives some idea of how fast the market adjusts to new,
unexpected information like a block trade.
What about people who can transact at the block price? Who are they and do
they not earn an abnormal return? Usually, the specialist, the floor trader (a member
of the NYSE), brokerage houses, and favored customers of the brokerage houses can
participate at the block price. Dann, Mayers, and Raab show that with a —4.56%
trading rule, an individual participating in every block with purchases of $100,000 or
more could have earned a net annualized rate of return of 203% for the 173 blocks
that activated the filter rule. Of course, this represents the maximum realizable rate
of return. Nevertheless, it is clear that even after adjusting for risk, transactions costs.
and taxes, it is possible to earn rates of return in excess of what any existing theory
would call "normal." This may be interpreted as evidence that capital markets are
inefficient in their strong form. Individuals who are notified of the pending block
trade and who can participate at the block price before the information becomes
publicly available do in fact appear to earn excess profits.
However, Dann, Mayers, and Raab caution us that we may not properly under-
BLOCK TRADES 37P
480 B =00 K -
=<0-= A 470= P0 K — —
=0 K ........
=00
3<0
80 3=0
60-
% Rate of return
=0/=/68 — 3/3=/69 7/=/69 — =</3=/69
2 40 40 —
Figure ==.8
Annualized rates of return on the —4.P6% rule. (From L.
Dann, D. Mayers, and R. Raaî, "Trading Rules, Large
Blocks, and the Speed of Adjustment," reprinted from Jou]nal
of Financial Economics, January =977, =8.)
stand all the costs that a îuyer faces in a îlock trade. One possiîility is that the
specialist (or anyone else) normally holds an optimal utility-maximizing portfolio. In
order to accept part of a îlock trade, which forces the specialist away from that port-
folio, he or she will charge a premium rate of return. In this way, what appear to îe
aînormal returns may actually îe fair, competitively determined fees for a service
rendered—the service of providing liquidity to a seller.
To date, the empirical research into the phenomenon of price changes around a
îlock trade shows that îlock trades do not disrupt markets, that markets are efficient
in the sense that they very quickly (less than =P minutes) fully reflect all puîlicly
availaîle information. There is evidence of îoth a permanent effect and a (very)
temporary liquidity or price pressure effect as illustrated in Fig. ==.P(c). The market
is efficient in its semistrong form, îut the fact that aînormal returns are earned îy
individuals who participate at the îlock price may indicate strong-form inefficiency.
376 EFFICIENT CAPITAL MARKETS: EVIDENCE
D. INSIDER TRADING
A direct test of strong-form efficiency is whether or not insiders with access to in-
formation that is not publicly available can outperform the market.6 Jaffe [1974]
collected data on insider trading from the Official Summary of Security Transactions
and Holdings published by the Securities and Exchange Commission. He then defined
an intensive trading month as one during which there were at least three more
insiders selling than buying, or vice versa. If a stock was intensively traded during
a given month, it was included in an intensive-trading portfolio. Using the empirical
market line, Jaffe then calculated cumulative average residuals. If the stock had
intensive selling, its residual (which would presumably be negative) was multiplied
by —1 and added to the portfolio returns, and conversely for intensive buying. For
861 observations during the 1960s, the residuals rose approximately 5% in eight
months following the intensive-trading event, with 3% of the rise occurring in the
last six months. These returns are statistically significant and are greater than trans-
actions costs. A sample of insider trading during the 1950s produces similar results.
These findings suggest that insiders do earn abnormal returns and that the strong-
form hypothesis of market efficiency does not hold.
Jaffe also investigated the effect of regulation changes on insider trading. Two of
the most significant changes in security regulation resulted from (1) the Cady-Roberts
decision in November 1961, when the SEC first exercised its power to punish insider
trading and thus established the precedent that corporate officials trading on insider
information were liable for civil prosecution; and (2) the Texas Gulf Sulphur case in
August 1966, when the courts upheld the earlier (April 1965) SEC indictment of
company officials who had suppressed and traded on news about a vast mineral
strike. After examining abnormal returns from intensive insider-trading samples
around dates of these historic decisions, Jaffe was forced to the following conclusion:
The data could not reject the null hypothesis that the enforcement of SEC regulations
in these two cases had no effect on insider trading in general. At best the regulations
prohibit only the most flagrant examples of speculation based on inside information.
A study by Finnerty [1976] corroborates Jaffe's conclusions. The major differ-
ence is that the Finnerty data sample was not restricted to an intensive trading
group. By testing the entire population of insiders, the empirical findings allow an
evaluation of the "average" insider returns. The data include over 30,000 individual
insider transactions between January 1969 and December 1972. Abnormal returns
computed from the market model indicate that insiders are able to "beat the market"
on a risk-adjusted basis, both when selling and when buying.
A study by Givoly and Palmon [1985] correlates insider trading with subse-
quent news announcements to see if insiders trade in anticipation of news releases.
The surprising result is that there is no relationship between insider trading and
news events. Although insiders' transactions are associated with a strong price move-
The Securities and Exchange Commission defines insiders as members of the board of directors, corpo-
rate officers, and any beneficial owner of more than 10% of any class of stock. They must disclose, on a
monthly basis, any changes in their stock holdings.
NEW ISSUES 377
ment in the direction of the trade during the month following the trade, these price
movements occur independent of subsequent publication of news. This leads to the
conjecture that outside investors accept (blindly) the superior knowledge and follow
in the footsteps of insiders.
One of the interesting implications of the empirical work on insider trading is
that it is consistent with the point of view that markets do not aggregate information.
In Chapter 10, fully aggregating markets were described as those that reflect all
available information even though it is not known to all market participants. In a
fully aggregating market an insider should not be able to make abnormal returns
because his trading activity would reveal his information to the market. The evidence
on profitable insider trading shows that this is clearly not the case.
E. NEW ISSUES
There has been a long history of articles that have studied the pricing of the common
stock of companies that is issued to the public for the first time. To mention a few, the
list includes papers by the Securities and Exchange Commission [1963], Reilly and
Hatfield [1969], Stickney [1970], McDonald and Fisher [1972], Logue [1973],
Stigler [1964], and Shaw [1971]. They all faced a seemingly insoluble problem: How
could returns on unseasoned issues be adjusted for risk if time series data on preissue
prices were nonexistent? Any estimate of systematic risk, e.g., requires the computa-
tion of the covariance between time series returns for a given security and returns
on a market portfolio. But new issues are not priced until they become public. An
ingenious way around this problem was employed by Ibbotson [1975]. Portfolios
of new issues with identical seasoning (defined as the number of months since issue)
were formed. The monthly return on the XYZ Company in March 1964, say two
months after its issue, was matched with the market return that month, resulting
in one pair of returns for a portfolio of two months seasoning. By collecting a large
number of return pairs for new issues that went public in different calendar months
but that all had two months seasoning, it was possible to form a vector of returns
of issues of two months seasoning for which Ibbotson could compute a covariance
with the market. In this manner, he estimated the systematic risk of issues with
various seasoning. Using the empirical market line, he was able to estimate abnormal
performance indices in the month of initial issue (initial performance from the offering
date price to the end of the first month) and in the aftermarket (months following
the initial issue). From 2650 new issues between 1960 and 1969, Ibbotson randomly
selected one new issue for each of the 120 calendar months.
The estimated systematic risk (beta) in the month of issue was 2.26, and the
abnormal return was estimated to be 11.4%. Even after transactions costs, this
represents a statistically significant positive abnormal return. Therefore either the
offering price is set too low or investors systematically overvalue new issues at the
end of the first month of seasoning. Later evidence shows that the aftermarket is
efficient; therefore Ibbotson focused his attention on the possibility that offering
prices determined by the investment banking firm are systematically set below the
378 EFFICIENT CAPITAL MARKETS: EVIDENCE
I Maximum offering price > market price > offering price Gain Parity
II Maximum offering price > offering price > market price Parity Loss
III Maximum offering price = offering price > market price Parity Loss
IV Market price > maximum offering price = offering price Gain Parity
fair market value of the security. Regulations of the SEC require a maximum offering
price for a new issue, which is usually filed two weeks in advance of the actual offering,
although it can be adjusted in some cases.' The actual offering price is set immedi-
ately before the offering. The existence of a regulation that requires the actual
offering price to be fixed creates the possibility of a "Heads I lose, tails you win"
situation for the underwriter. Table 11.3 shows the four possibilities that can occur
in a firm commitment offering (the underwriting syndicate buys the issue from the
firm for the offering price less an underwriting spread, then sells the issue to the
public at the fixed offering price). The best the underwriter can do is achieve a parity
situation with no gain or loss. This happens whenever the market price turns out to
be above the offering price (situations I and IV). Obviously, the investment banker
does not want the market price to equal or exceed the maximum offering price
(situations III and IV). This would infuriate the issuing firm and lead to a loss of
future underwriting business. Therefore we usually observe situations I and II. But
if the investment banking firm receives adequate compensation from its underwriting
spread for the risk it undertakes, and if it cannot gain by setting the offer price lower
than the market price, then why do we not observe offer prices (which, after all, are
established only moments before the issues are sold to the public) set equal to the
market value? Why can investors systematically earn an abnormal return of 11.4%
during the first month of issue? This conundrum, like the difference between the
block price and the closing price on the day of the block, cannot easily be explained
by existing finance theory.
What about new issue performance in the aftermarket, i.e., for prices from the
first market price onward? Figure 11.9 shows abnormal returns (based on the empiri-
cal market line) in the aftermarket for six-month holding periods and the significance
tests (t-tests). The 9 periods other than the initial offering period include only 2
periods with results that are statistically different from zero (and returns in these 2
periods are negative). Ibbotson concludes that the evidence cannot allow us to reject
the null hypothesis that aftermarkets are efficient, although it is interesting to note
that returns in 7 out of 10 periods show negative returns.
Figure 11.10 shows plots of changes in systematic risk in the aftermarket; note
the decline. The results show that the systematic risk of new issues is greater than
In most cases the maximum offering price is set high enough to cause little concern that it may ac-
tually constrain the actual offering price.
NEW ISSUES 379
.=P0 4.000 —
.=00 <.000
.0P0 0.000
çMùNùMç
NOMMEN
—.0P0 I I I I I —4.000 I I I I I
0 =< <4 36 48 60 0 =< <4 36 48 60
aMonths of seasonin g (n), n = =, 7, =3, =9, <P, 3=, 37, 43, 49, PP ; r = 6 ; reg ressions =-6, 7-=<, ..., PP-60.
Figure ==.9
Aînormal returns for issues of different seasoning. (From R. Iîîotson, "Price Performance
of Common Stock New Issues," reprinted from Jou]nal of Financial Economics, Septemîer
=97P, <P4.)
3.000 6.000
Beta t - test
<.000 4.000
111•111■
MIII=111
=.000 <.000
.000 .000
—=.000 —<.000
0 =< <4 48
36 60 0 =< <4 36 48 60
aMonths of seasonin g (n), n = =, 7, =3, =9, <P, 3=, 37, 43, 49, PP ; ] = 6 ; re gressions =-6,
7-=<, ..., PP-60. These t-statistics are designed to test if p„, ], 0 > =.
Figure ==.=0
Systematic risk of issues with different seasoning. (From R. Iîîotson, "Price Performance
of Common Stock New Issues," reprinted from Jou]nal of Financial Economics, Septemîer
=97P, <60.)
380 EFFICIENT CAPITAL MARKETS: EVIDENCE
the systematic risk of the market (which always has a beta equal to one) and that
their systematic risk is not stable in that it drops as the new issues become seasoned.
Weinstein [1978] studied the price behavior of newly issued corporate bonds
by measuring their excess holding period returns. Excess returns were defined as the
difference between the return on the ith newly issued bond and a portfolio of seasoned
bonds with the same (Moody's) bond rating. Data were collected for 179 new issues
between June 1962 and July 1974. Weinstein's conclusions for newly issued bonds are
similar to those of Ibbotson [1975] for newly issued stock, namely, that the offering
price is below the market equilibrium price but that the aftermarket is efficient.
Weinstein found a .383% rate of return during the first month and only a .06% rate
of return over the next six months.
F. STOCK SPLITS
Why do stocks split, and what effect, if any, do splits have on shareholder wealth?
The best known study of stock splits was conducted by Fama, Fisher, Jensen, and
Roll [1969]. Cumulative average residuals were calculated from the simple market
model, using monthly data for an interval of 60 months around the split ex date for
940 splits between January 1927 and December 1959. Figure 11.11 shows the results.
It plots the cumulative average return for the stock split sample. Positive abnormal
returns are observed before the split but not afterward. This would seem to indicate
that splits are the cause of the abnormal returns. But such a conclusion has no
economic logic to it. The run-up in the cumulative average returns prior to the stock
split in Fig. 11.11 can be explained by selection bias. Stocks split because their price
has increased prior to the split date. Consequently, it should hardly be surprising
that when we select a sample of split-up stocks, we observe that they have positive
0.11
.• •
••
.
0 .° 1 1 1 1 1 1 1 I I I 1
—29 —20 —10 0 10 20 30
Month relative to split, m
STOCK SPLITS 381
••••.....
..•
0.33 0.33
0.22 0.22
.....••••
0.11— 0.11
r•
..*
0 1.-'1 • 1 1 1 1 1 1 1 1 1 1 0 1•:1 I 1 1 1 1_ 1 I I 1 1 1
-29 -20 -10 0 10 20 30 -29 -20 -10 0 10 20 30
Month relative to split, m Month relative to split, m
(a) (b)
Figure 11.12
Cumulative average residuals for splits with (a) dividend increases and (b) decreases. (From
E. F. Fama, L. Fisher, M. Jensen, and R. Roll, "The Adjustment of Stock Prices to New
Information," reprinted with permission of International Economic Review, February 1969,
15. Copyright © International Economic Review.)
abnormal performance prior to the split date. Selection bias occurs because we are
studying a selected data set of stocks that have been observed to split.
Farna et al. [1969] speculated that stock splits might be interpreted by investors
as a message about future changes in the firm's expected cash flows. They hypoth-
esized that stock splits might be interpreted as a message about dividend increases,
which in turn imply that the managers of the firm feel confident that it can maintain
a permanently higher level of cash flows. To test this hypothesis the sample was di-
vided into those firms that increased their dividends beyond the average for the
market in the interval following the split and those that paid out lower dividends.
The results, shown in Fig. 11.12, reveal that stocks in the dividend "increased" class
have slightly positive returns following the split. This is consistent with the hypothesis
that splits are interpreted as messages about dividend increases.' Of course, a div-
idend increase does not always follow a split. Hence the slightly positive abnormal
return for the dividend-increase group reflects small price adjustments that occur
when the market is absolutely sure of the increase. On the other hand, the cumulative
average residuals of split-up stocks with poor dividend performance decline until
about a year after the split, by which time it must be very clear that the anticipated
8 This does not imply that higher dividend payout per se causes an increase in the value of the firm. In
Chapter 15 "Dividend Policy" we shall see that higher dividends are interpreted as signals that the future
cash flows from the firm will increase.
382 EFFICIENT CAPITAL MARKETS: EVIDENCE
dividend increase is not forthcoming. When we combine the results for the dividend
increases and decreases, these results are consistent with the hypothesis that on the
average the market makes unbiased dividend forecasts for split-up securities and
these forecasts are fully reflected in the price of the security by the end of the split
month.
A more recent study by Grinblatt, Masulis, and Titman [1984] used daily data
and looked at shareholder returns on the split announcement date as well as the split
ex date. They examined a special subsample of splits where no other announcements
were made in the three-day period around the split announcement and where no
cash dividends had been declared in the previous three years.' For this sample of
125 "pure" stock splits they found a statistically significant announcement return of
3.44%. They interpret stock split announcements as favorable signals about the firm's
future cash flows. Surprisingly, they also find statistically significant returns (for their
entire sample of 1360 stock splits) on the ex date. There is no good explanation for
this result, and it is inconsistent with the earlier Fama et al. study that used monthly
returns data.
In the same study, Grinblatt, Masulis, and Titman [1984] confirm earlier work
on stock dividends by Foster and Vickrey [1978] and Woolridge [1983a, 1983b].
The announcement effects for stock dividends are large, 4.90% for a sample of 382
stock dividends and 5.89% for a smaller sample of 84 stock dividends with no other
announcements in a three-day period around the stock dividend announcement. One
possible reason for the larger announcement effect of a stock dividend is that retained
earnings must be reduced by the dollar amount of the stock dividend. Only those
companies that are confident they will not run afoul of debt restrictions that require
minimum levels of retained earnings will willingly announce a stock dividend. Another
reason is that convertible debt and warrant holders are not protected against dilution
caused by stock dividends. As with stock splits, there was a significant positive re-
turn on the stock dividend ex date (and the day before). No explanation is offered
for why the ex date effect is observed.
The results of Fama et al. [1969] are consistent with the semistrong form of
market efficiency. Prices appear to fully reflect information about expected cash flows.
The split per se has no effect on shareholder wealth. Rather, it merely serves as a
message about the future prospects of the firm. Thus splits have benefits as signaling
devices. There seems to be no way to use a split to increase one's expected returns,
unless, of course, inside information concerning the split or subsequent dividend
behavior is available.
One often hears that stocks split because there is an "optimal" price range for
common stocks. Moving the security price into this range makes the market for
trading in the security "wider" or "deeper"; hence there is more trading liquidity.
Copeland [1979] reports that contrary to the above argument, market liquidity is
actually lower following a stock split. Trading volume is proportionately lower than
its presplit level, brokerage revenues (a major portion of transactions costs) are
9
However, 11% of the pure samples declared a dividend within one year of the stock split.
PERFORMANCE OF MANAGED PORTFOLIOS 383
proportionately higher, and bid-ask spreads are higher as a percentage of the bid
price.' Taken together, these empirical results point to lower postsplit liquidity.
Hence we can say that the market for split-up securities has lower operational ef-
ficiency relative to its presplit level. Ohlson and Penman [1985] report that the
postsplit return standard deviation for split-up stocks exceeds the presplit return
standard deviation by an average of 30%. Lower liquidity and higher return variance
are both costs of splitting.
Brennan and Copeland [1987] provide a signaling theory explanation for stock
splits and show that it is consistent with the data. The intuition can be explained
as follows. Suppose that managers know the future prospects of their firm better than
the market does. Furthermore, assume that there are two firms with a price of $60
per share which are alike in every way except that the managers of firm A know it
has a bright future while the managers of firm B expect only average performance.
Managers of both firms know that if they decide to announce a split, their share-
holders will suffer from the higher transactions costs documented by Copeland [1979].
However, the successful firm A will bear these costs only temporarily, while firm B
will bear them indefinitely. Hence firm A will signal its bright future with a stock
split and the signal will not be mimicked by firm B. As a result, A's price will rise
at the time of the announcement so as to reflect the present value of its future pros-
pects. Furthermore, the lower the target price to which the firm splits, the greater
confidence management has, and the larger will be the announcement residual.
Empirical results by Brennan and Copeland [1987] confirm this prediction.
G. PERFORMANCE OF MANAGED
PORTFOLIOS
1. Mutual Funds
Mutual funds allege that they can provide two types of service to their clients.
First, they may minimize the amount of unsystematic risk an investor must face. This
is done through efficient diversification in the face of transactions costs. Second, they
may be able to use their professional expertise to earn abnormal returns through
successful prediction of future security prices. This second claim is contradictory to
the semistrong form of capital market efficiency unless, for some reason, mutual fund
managers can consistently obtain information that is not publicly available.
A number of studies have focused their attention on the performance of mutual
funds. A partial list includes Friend and Vickers [1965], Sharpe [1966], Treynor
[1965], Farrar [1962], Friend, Blume, and Crockett [1970], Jensen [1968], Mains
[1977], Henricksson [1984], and Grinblatt and Titman [1986]. Various performance
10
The bid price is the price that a potential buyer offers, say $20, and the ask price is what the seller
requires, suppose it is $20-i. The bid-ask spread is the difference, specifically Si:.
384 EFFICIENT CAPITAL MARKETS: EVIDENCE
R • — R ft
Reward to variability ratio = (11.4)
u•
R
Treynor index = (11.5)
P.;
Abnormal performance = an = (Rit — R ft ) — [Si(Rmt — R ft )], (11.6)
where
Of these, the abnormal performance measure [Eq. (11.6)] makes use of the CAPM.
It was developed by Jensen [1968], who used it to test the abnormal performance
of 115 mutual funds, using annual data between 1955 and 1964. If the performance
index, a, is positive, then after adjusting for risk and for movements in the market
index, the abnormal performance of a portfolio is also positive. The average a for
returns measured net of costs (such as research costs, management fees, and brokerage
commissions) was — 1.1% per year over the 10-year period. This suggests that on the
average the funds were not able to forecast future security prices well enough to cover
their expenses. When returns were measured gross of expenses (excepting brokerage
commissions), the average a was — .4% per year. Apparently the gross returns were
not sufficient to recoup even brokerage commissions.
In sum, Jensen's study of mutual funds provides evidence that the 115 mutual
funds, on the average, were not able to predict security prices well enough to out-
perform a buy-and-hold strategy. In addition, there was very little evidence that any
individual fund was able to do better than what might be expected from mere random
chance. These conclusions held even when fund returns were measured gross of
management expenses and brokerage costs. Results obtained are consistent with the
hypothesis of capital market efficiency in its semistrong form, because we may as-
sume that, at the very least, mutual fund managers have access to publicly available
information. However, they do not necessarily imply that mutual funds will not be
held by rational investors. On the average the funds do an excellent job of diver-
sification. This may by itself be a socially desirable service to investors.
More recently, Mains [1977] has reexamined the issue of mutual fund perfor-
mance. He criticizes Jensen's work on two accounts. First, the rates of return were
underestimated because dividends were assumed to be reinvested at year's end rather
than during the quarter they were received and because when expenses were added
back to obtain gross returns, they were added back at year's end instead of continu-
ously throughout the year. By using monthly data instead of annual data, Mains
is able to better estimate both net and gross returns. Second, Jensen assumed that
PERFORMANCE OF MANAGED PORTFOLIOS 385
mutual fund betas were stationary over long periods of time [note that 13' has no
time subscript in Eq. (11.6)]. Using monthly data, Mains obtains lower estimates of
Si and argues that Jensen's estimates of risk were too high.
The abnormal performance results calculated for a sample of 70 mutual funds
indicate that as a group the mutual funds had neutral risk-adjusted performance on
a net return basis. On a gross return basis (i.e., before operating expenses and trans-
actions costs), 80% of the funds sampled performed positively. This suggests that
mutual funds are able to outperform the market well enough to earn back their
operating expenses. It is also consistent with the theory of efficient markets given
costly information. Recall from Chapter 10 that the theoretical work of Cornell and
Roll [1981] and Grossman [1980] predicts a market equilibrium where investors
who utilize costly information will have higher gross rates of return than their un-
informed competitors. But because information is costly, the equilibrium net rates
of return for informed and uninformed investors will be the same. This is just what
Main's work shows. Mutual funds' gross rates of return are greater than the rate on
a randomly selected portfolio of equivalent risk, but when costs (transactions costs
and management fees) are subtracted, the net performance of mutual funds is the
same as that for a naive investment strategy.
11A partial list of Value Line—related studies is: Shelton [1967], Hausman [1969], Black [1971], Kaplan
and Weil [1973], Brown and Rozeff [1978], Holloway [1981], and Copeland and Mayers [1982].
386 EFFICIENT CAPITAL MARKETS: EVIDENCE
Black [1971] performed the first systematic study utilizing Jensen's abnormal
performance measure as given in Eq. (11.6). Black's results indicate statistically sig-
nificant abnormal performance for equally weighted portfolios formed from stocks
ranked 1, 2, 4, and 5 by Value Line and rebalanced monthly. Before transactions
costs, portfolios 1 and 5 had risk-adjusted rates of return of + 10% and —10%,
respectively. Even with round-trip transactions costs of 2%, the net rate of return for
a long position in portfolio 1 would still have been positive, thereby indicating eco-
nomically significant performance. One problem with these results is the Jensen
methodology for measuring portfolio performance. It has been criticized by Roll
[1977, 1978], who argues that any methodology based on the capital asset pricing
model will measure either (1) no abnormal performance if the market index portfolio
is ex post efficient or (2) a meaningless abnormal performance if the index portfolio
is ex post inefficient.'
Copeland and Mayers [1982] and Chen, Copeland, and Mayers [1986] mea-
sured Value Line portfolio performance by using a future benchmark technique that
avoids selection bias problems associated with using historic benchmarks as well as
the known difficulties of using capital asset pricing model benchmarks.' The future
benchmark technique uses the market model (described in section A of this chapter)
fit using data after the test period where portfolio performance is being measured.
The steps in the procedure are:
1. Using a sample from after the test period, calculate the market model equation
for the portfolio being evaluated.
2. Use the parameters of the model as a benchmark for computing the portfolio's
unexpected return during a test period.
3. Repeat the procedure and test to see whether the mean unexpected return is
significantly different from zero.
In other words, rather than using a particular (perhaps suspect) model (such as the
CAPM) of asset pricing as a benchmark, estimate the expected returns directly from
the data. The future benchmark technique is not without its problems, however. It
assumes that the portfolio characteristics (e.g., risk and dividend yield) remain essen-
tially the same throughout the test and benchmark periods.
Copeland and Mayers find considerably less abnormal performance than Black,
who used the Jensen methodology. Where Black reported (roughly) 20% per year for
an investor who was long on portfolio 1 and short on portfolio 5, Copeland and
Mayers find an annual rate of return of only 6.8%. Moreover, only portfolio 5 had
statistically significant returns. Nevertheless, any significant performance is a potential
violation of semistrong market efficiency. Thus Value Line remains an enigma.
12
For a more complete discussion of Roll's critique, see Chapter 7.
13
Using historic benchmarks creates a selection bias problem because Value Line uses a variant of the
"relative strength" criterion to choose rankings. Portfolio I stocks tend to have abnormally high historic
rates of return; thus subtracting these rates from test period returns would tend to bias the results against
Value Line.
PERFORMANCE OF MANAGED PORTFOLIOS 387
Stickel [1985] uses the future benchmark methodology to measure the abnormal
performance resulting from changes in Value Line rankings. He finds statistically
significant returns for reclassifications from rank 2 to rank 1 that are three times as
large as the returns from reclassifications from 1 to 2. Upgradings from 5 to 4 were
not associated with significant abnormal returns. He concludes that the market reacts
to Value Line reclassifications as news events, that the price adjustment takes place
over a multiple-day period, and that the size of the adjustment is larger for smaller
firms.
1 4 The net asset value received at maturity is the market value of the securities in the fund at that date,
There are several explanations for Thompson's results. First, the market may be
inefficient, at least for tax-exempt institutions that could seemingly be able to profit
from the above-mentioned trading rule. Second, so long as taxable investors persist
in holding closed-end shares the gross rates of return before taxes may have to exceed
the market equilibrium rate of return in order to compensate for unrealized tax liabil-
ities. Third, abnormal return measures based on the capital asset pricing model may
be inappropriate for measuring the performance of closed-end fund capital shares
that are call options.
An interesting paper by Brauer [1984] reports on the effects of open-ending 14
closed-end funds between 1960 and 1981. Funds that were open-ended had larger
discounts from net asset value (23.6% versus 16.2%) and lower management fees (.78%
versus 1.00%) than funds that were not open-ended. Large discounts provide share-
holders with greater incentive to open-end their funds and lower management fees
imply less management resistance. These two variables were actually able to predict
which funds would be open-ended. In addition, Brauer reports that most of the (large)
abnormal returns that resulted from the announcement of open-ending were realized
by the end of the announcement month a result consistent with semistrong-form
market efficiency.
The problem of analyzing dual funds is not yet completely resolved. The observed
discounts (premia) on capital shares may be attributable to (1) unrealized capital
gains tax liabilities, (2) fund holdings of letter stock, (3) management and brokerage
costs, or (4) the option nature of capital shares. The relative importance of these
factors has not yet been completely resolved. There is no good explanation for why
all funds selling at a discount have not been open-ended. In addition, there remains
some question about whether or not abnormal returns can be earned by utilizing
trading rules based on observed discounts (premia). Thompson's [1978] work suggests
that abnormal returns are possible, whereas Ingersoll [1976] finds no evidence of
abnormal returns.
Any predictable pattern in asset returns may be exploitable and therefore judged as
evidence against semistrong market efficiency. Even if the pattern cannot be employed
directly in a trading rule because of prohibitive transactions costs, it may enable
people who were going to trade anyway to increase their portfolio returns over what
they otherwise may have received without knowledge of the pattern. Two statistically
significant patterns in stock market returns are the weekend effect and the turn-of-
the-year effect.
French [1980] studied daily returns on the Standard and Poor's composite port-
folio of the 500 largest firms on the New York Stock Exchange over the period 1953
to 1977. Table 11.4 shows the summary statistics for returns by day of the week. The
negative returns on Monday were highly significant. They were also significantly
negative in each of the five-year subperiods that were studied.
WEEKEND AND YEAR-END EFFECTS 39=
indices of NYSE and AMEX stocks was 9.31% for calendar years 1963-1980 inclu-
sive. During those same years, the average return for the five days of the turn-of-the-
year (last day of December and first five days of January) was 3.45%. Thus, about
37% of the annual differential is due to just five trading days, 67% of the annual
differential is due to the first twenty trading days of January plus the last day of
December."
The most likely cause of the year-end effect is tax selling. At least there is a sig-
nificant correlation between the realized rates of return during the year and the size
of the turn-of-the-year price recovery. Whether or not this phenomenon is exploitable
with a trading rule remains to be seen. However, an individual who was going to
transact anyway can benefit by altering his or her timing to buy in late December
or sell in early January.
SUMMARY
Most evidence suggests that capital markets are efficient in their weak and semistrong
forms, that security prices conform to a fair-game model but not precisely to a ran-
dom walk because of small first-order dependencies in prices and nonstationarities
in the underlying price distribution over time, and that the strong-form hypothesis
does not hold. However, any conclusions about the strong form of market efficiency
need to be qualified by the fact that capital market efficiency must be considered
jointly with competition and efficiency in markets for information. If insiders have
monopolistic access to information, this fact may be considered an inefficiency in the
market for information rather than in capital markets. Filter rules (described in Chap-
ter 10) have shown that security prices exhibit no dependencies over time, at least
down to the level of transactions costs. Thus capital markets are allocationally effi-
cient up to the point of operational efficiency. If transactions costs amounted to a
greater percentage of value traded, price dependencies for filter rules greater than
1.5% might have been found.
At least in two instances, special types of "abnormal" returns could not be ex-
plained. Block traders who can buy at the block price and sell at the market close
could earn annual abnormal returns of over 200% per year even after transactions
costs. Individuals who could buy new issues at the subscription price and sell at the
end of the month could earn annual abnormal returns of 11.4% per month (this is
over 350% per year). Although both these results may be interpreted as strong-form
inefficiencies, the authors were quick to point out that they may simply represent
fair returns for services by the block positioner or the investment banker. It is best
to say at this point that we do not know.
Most of the studies reviewed in this chapter have used data from the stock mar-
ket. However, there is evidence that other markets are also efficient. Roll [1970]
showed that prices in the Treasury bill market obey a fair game model. Schwert
[1977] concluded that the prices of New York Stock Exchange seats follow a multi-
plicative random walk. Stein [1977] examined the auction market for art and found
it efficient. Larson [1964] looked at corn futures, and Mandelbrot [1964] investi-
PROBLEM SET 393
gated spot prices in cotton. In addition to these studies, we should mention in passing
that there are many other topics related to the question of market efficiency that
have not been discussed here.
PROBLEM SET
11.1 Roll's critique of tests of the CAPM shows that if the index portfolio is ex post efficient,
it is mathematically impossible for abnormal returns, as measured by the empirical market
line, to be statistically different from zero. Yet the Ibbotson study on new issues uses the cross-
section empirical market line and finds significant abnormal returns in the month of issue and
none in the following months. Given Roll's critique, this should have been impossible. How
can the empirical results be reconciled with the theory?
11.2 In a study on corporate disclosure by a special committee of the Securities and Exchange
Commission, we find the following statement (1977, D6):
The "efficient market hypothesis"—which asserts that the current price of a security
reflects all publicly available information even if valid, does not negate the necessity of a
mandatory disclosure system. This theory is concerned with how the market reacts to
disclosed information and is silent as to the optimum amount of information required or
whether that optimum should be achieved on a mandatory or voluntary basis; market
forces alone are insufficient to cause all material information to be disclosed.
Two questions that arise are:
a) What is the difference between efficient markets for securities and efficient markets for
information?
b) What criteria define "material information"?
11.3 In your own words, what does the empirical evidence on block trading tell us about
market efficiency?
11.4 Which of the following types of information provides a likely opportunity to earn abnor-
mal returns on the market?
11.5 Mr. A has received, over the last three months, a solicitation to purchase a service that
claims to be able to forecast movements in the Dow Jones Industrial index. Normally, he
does not believe in such things, but the service provides evidence of amazing accuracy. In each
of the last three months, it was always right in predicting whether or not the index would
move up more than 10 points, stay within a 10-point range, or go down by more than 10
points. Would you advise him to purchase the service? Why or why not?
394 EFFICIENT CAPITAL MARKETS: EVIDENCE
+1600% 600%
+1400 400
+1200 200
+1000 - +1000
+800 +800
Record of Value Line rankings for timeliness
+600 April =6, =96P-Decemîer <8, =983 +600
+500
(Without allowance for changes in rank) +500
+400 - +400
+300 +300
+200 - +200
+100 - +100
.. ..... •
N
0 .■ _ 0
- 20 - -20
Group = +=<9P%
- 40 - -40
Group < + 68=%
Average + 339% /
-60 -
Group 3 + 3<8% / - -60
/
Group 4 + =37%
--- Group P + 3P%
-80 -80
1965 19661967 1968 1969 1970 1971 972 1973 1974 19751976 1977 1978 1979 980119811198211983
Group =96P =966 =967 =968 =969 =970 =97= =97< =973 =974
= +33.6% - 3.=% +39.<% +3=.<% -=7.7% - 8.9% +<6.P% +=0.=% -=7,=% -<3.=%
< +=8.9 - 6.0 +3=.9 +<6.3 -=6.3 - 4.0 +=7.4 + 7.P -<6.< -<7.8
3 + 8.9 - 9.7 +30.= +<=.4 -<0.7 - P.P +=<.< + 6.< -<7.0 -<8.P
4 + 0.8 - 7.< +<P.= +<P.= -<6.8 -==.7 +=4.< + 3.< -<9.= -33.6
-
P - =.< - =<.4 +<8.4 +<P.9 -3P.7 -=3.= +=0.P + <.9 -43.= 36.8
Avg +=0.= - 7.9 +<9.9 +<4.6 -<<.= - 7.P +=4.9 + P.P -<7.7 -<9.6
1965
through
Group =97P =976 =977 =978 =979 =980 =98= =98< =983 1983
= +P=.6% +3P.3% +=P.8% +=9.8% +<P.6% +P0.<% -=.9% +33.7% +<P.<% +=<9P%
< +P3.0 +36.3 +=<.7 +=6.= +30.8 +37.4 +0.7 +<9.0 -k <.< + 68=
3 +P<.9 +33.8 + P.< + 9.< +<7.6 +<0.8 +<.7 +<P.P +<6.7 + 3<8
4 +48.4 +36.= - 0.< + <.4 +<3.= +=3.< -0.9 +=8.P +3P.< + =37
P +4<.= +38.< - <.8 + 4.0 +39.9 + 8.4 -4.< +=9.9 +30.0 + 3P
Avg +P=.< +3P.= + P.8 + 9.6 +<8.0 +<3.4 +0.9 +<P.0 +<7.P + 339
Figure Q==.8
Eighteen-year record of actual forecast assumes all rank changes have îeen followed. (From
A. Bernhard, "The Value Line Investment Survey," Investing in Common Stock, Arnold
Bernhard and Company, Inc. © Value Line, Inc. Reprinted with permission.)
REFERENCES 395
11.6 The Ponzi Mutual Fund (which is not registered with the SEC) guarantees a 2% per
month (24% per year) return on your money. You have looked into the matter and found that
they have indeed been able to pay their shareholders the promised return for each of the 18
months they have been in operation. What implications does this have for capital markets?
Should you invest?
11.7 Empirical evidence indicates that mutual funds that have abnormal returns in a given
year are successful in attracting abnormally large numbers of new investors the following year.
Is this inconsistent with capital market efficiency?
11.8 The Value Line Investment Survey publishes weekly stock performance forecasts. Stocks
are grouped into five portfolios according to expected price performance, with Group 1 com-
prising the most highly recommended stocks. The chart of each portfolio's actual performance
over an 18-year period (Fig. Q11.8) assumes that each of the five portfolios was adjusted on
a weekly basis in accordance with Value Line's stock ratings. The chart shows that the port-
folios' actual performances are consistent with Value Line's forecasts. Is this evidence against
an efficient securities market?
11.9 In each of the following situations, explain the extent to which the empirical results offer
reliable evidence for (or against) market efficiency.
a) A research study using data for firms continuously listed on the Compustat computer
tapes from 1953 to 1973 finds no evidence of impending bankruptcy cost reflected in stock
prices as a firm's debt/equity ratio increases.
b) One thousand stockbrokers are surveyed via questionnaire, and their stated investment
preferences are classified according to industry groupings. The results can be used to
explain rate of return differences across industries.
c) A study of the relationships between size of type in the New York Times headline and
size of price change (in either direction) in the subsequent day's stock index reveals a sig-
nificant positive correlation. Further, when independent subjects are asked to qualify the
headline news as good, neutral, or bad, the direction of the following day's price change
(up or down) is discovered to vary with the quality of news (good or bad).
d) Using 25 years of data in exhaustive regression analysis, a Barron's writer develops a sta-
tistical model that explains the 25-year period of stock returns (using 31 variables) with
minuscule error.
REFERENCES
Aharony, J., and I. Swary, "Quarterly Dividend and Earnings Announcements and Stock-
holders' Returns: An Empirical Analysis," Journal of Finance, March 1980, 1-12.
Alexander, S. S., "Price Movements in Speculative Markets: Trends or Random Walks," Indus-
trial Management Review, May 1961, 7-26.
Ball, R., and P. Brown, "An Empirical Evaluation of Accounting Income Numbers," Journal of
Accounting Research, Autumn 1968, 159-178.
Barker, A., "Evaluation of Stock Dividends," Harvard Business Review, July—August 1958,
99-114.
Bernhard, A., Investing in Common Stocks. Arnold Bernhard & Co., Inc., New York, 1975.
, Value Line Methods of Evaluating Common Stocks. Arnold Bernhard & Co., Inc., New
York, 96 pp., 1975.
396 EFFICIENT CAPITAL MARKETS: EVIDENCE
Black, F., "Yes, Virginia, There Is Hope: Tests of the Value Line Ranking System," Graduate
School of Business, University of Chicago, May =97=.
; M. Jensen; and M. Scholes, "The Capital Asset Pricing Model: Some Empirical
Tests," in Jensen, ed., Studies in the Theo]y of Capital Ma]kets. Praeger, New York, =97<,
79-=<=.
Black, F., and M. Scholes, "The Effects of Dividend Yield and Dividend Policy on Common
Stock Prices and Returns," Jou]nal of Financial Economics, May =974, =-<<.
Boudreaux, K., "Discounts and Premiums on Closed-End Funds: A Study on Valuation,"
Jou]nal of Finance, May =973, P=P-P<<.
Branch, B., "A Tax Loss Trading Rule," Jou]nal of Business, April =977, =98-<07.
Brauer, G., — Open-Ending' Closed-End Funds," Jou]nal of Financial Economics, Decemîer
=984, 49=-P08.
Brennan, M., and T. Copeland, "A Signalling Model of Stock Split Behavior: Theory and
Evidence," UCLA working paper, March =987.
Brimmer, A., "Credit Conditions and Price Determination in the Corporate Bond Market,"
Jou]nal of Finance, Septemîer =960, 3P3-370.
Brown, L., and M. Rozeff, "The Superiority of Analyst Forecasts as Measures of Expectations:
Evidence from Earnings," Jou]nal of Finance, March =978, =-=6.
Brown, S., "Earnings Changes, Stock Prices and Market Efficiency," Jou]nal of Finance, March
=978, =7-<8.
Chen, N. F.; T. Copeland; and D. Mayers, "A Comparison of Single and Multiple Factor
Portfolio Performance Methodologies," Working Paper <-86, UCLA, January =986.
Chottiner, S., and A. Young, "A Test of the AICPA Differentiation îetween Stock Dividends
and Stock Splits," Jou]nal of Accounting Resea]ch, Autumn =97=, 367-374.
Copeland, T. E., "Liquidity Changes Following Stock Splits," Jou]nal of Finance, March =979,
==P-=4=.
, and D. Mayers, "The Value Line Enigma (=96P-=978): A Case Study of Performance
Evaluation Issues," Jou]nal of Financial Economics, Novemîer =98<, <89-3<=.
Cornell, B., and R. Roll, "Strategies for Pairwise Competitions in Markets and Organizations,"
Bell Jou]nal of Economics, Spring =98=, <0=-<=3.
Dann, L.; D. Mayers; and R. Raaî, "Trading Rules, Large Blocks and the Speed of Adjust-
ment," Jou]nal of Financial Economics, January =977, 3-<<.
Dyl, E., "The Effect of Capital Gains Taxation on the Stock Market," Ph.D. dissertation,
Stanford University Graduate School of Business, August =973.
Ederington, L., "The Yield Spread on New Issues of Corporate Bonds," Jou]nal of Finance,
Decemîer =974, =P3=-=P43.
Fama, E. F., "The Behavior of Stock Market Prices," Jou]nal of Business, January =96P, 34-=0P.
, "Efficient Capital Markets: A Review of Theory and Empirical Work," Jou]nal of
Finance, May =970, 383-4=7.
, Foundations of Finance. Basic Books, New York, =976.
Fama, E. F., and M. Blume, "Filter Rules and Stock Market Trading Profits," Jou]nal of
Finance, May =970, <<6-<4=.
Fama, E. F.; L. Fisher; M. Jensen; and R. Roll, "The Adjustment of Stock Prices to New
Information," Inte]national Economic Review, Feîruary =969, =-<=.
REFERENCES 397
Fama, E. F., and J. MacBeth, "Risk, Return and Equilibrium: Empirical Test," Jou]nal of
Political Economy, May—June 1973, 607-635.
Farrar, D. E., The Investment Decision unde] Unce]tainty. Prentice—Hall, Englewood Cliffs,
N.J., 1962.
Finnerty, J. E., "Insiders and Market Efficiency," Jou]nal of Finance, September 1976, 1141-
1148.
Foster, G., "Quarterly Accounting Data: Time Series Properties and Predictive Ability Re-
sults," Accounting Review, January 1977, 1-21.
Foster, T., III, and D. Vickrey, "The Information Content of Stock Dividend Announcements,"
Accounting Review, April 1978, 360-370.
French, D., "The Weekend Effect on the Distribution of Stock Prices," Jou]nal of Financial
Economics, December 1984, 547-560.
French, K., "Stock Returns and the Weekend Effect," Jou]nal of Financial Economics, March
1980, 55-69.
Friend, I.; M. Blume; and J. Crockett, Mutual Funds and Othe] Institutional Investo]s. McGraw-
Hill, New York, 1970.
Friend, I.; F. E. Brown; E. S. Herman; and D. Vickers, A Study of Mutual Funds. U.S. Govern-
ment Printing Office, Washington, D.C., 1962.
Friend, I., and D. Vickers, "Portfolio Selection and Investment Performance," Jou]nal of
Finance, September 1965, 391-415.
Gibbons, M., and P. Hess, "Day of the Week Effects and Asset Returns," Jou]nal of Business,
October 1981, 579-596.
Givoly, D., and D. Palmon, "Insider Trading and the Exploitation of Inside Information: Some
Empirical Evidence," Jou]nal of Business, January 1985, 69-87.
Granger, C. W. J., and 0. Morgenstern, P]edictability of Stock Ma]ket P]ices. Heath Lexington
Books, Lexington, Mass., 1970.
Griffin, P., "The Time-Series Behavior of Quarterly Earnings: Preliminary Evidence," Jou]nal
of Accounting Resea]ch, Spring 1977, 71-83.
Grinblatt, M.; R. Masulis; and S. Titman, "The Valuation Effects of Stock Splits and Stock
Dividends," Jou]nal of Financial Economics, December 1984, 461-490.
Grinblatt, M., and S. Titman, "A Comparison of Measures of Abnormal Performance on a
Sample of Monthly Mutual Fund Returns," Working Paper 13-86, UCLA, 1986.
Grossman, S., "The Impossibility of Informationally Efficient Markets," Ame]ican Economic
Review, June 1980, 393-408.
Gultekin, M., and N. B. Gultekin, "Stock Market Seasonality: International Evidence," Jou]nal
of Financial Economics, December 1983, 469-481.
Hausman, W., "A Note on the Value Line Contest: A Test of the Predictability of Stock Price
Changes," Jou]nal of Business, July 1969, 317-320.
Henricksson, R., "Market Timing and Mutual Fund Performance: An Empirical Investigation,"
Jou]nal of Business, January 1984, 73-96.
Holloway, C., "A Note on Testing in Aggressive Investment Strategy Using Value Line Ranks,"
Jou]nal of Finance, June 1981, 711-719.
Hong, H.; R. S. Kaplan; and G. Mandelker, "Pooling vs. Purchase: The Effects of Accounting
for Mergers on Stock Prices," Accounting Review, January 1978, 31-47.
Ç98 EFFICIENT CAPITAL MARKETS: EVIDENCE
Iîîotson, R., "Price Performance of Common Stock New Issues," Jou]nal of Financial Eco-
nomics, Septemîer =97P, <3P-<7<.
Ingersoll, J., Jr., "A Theoretical and Empirical Investigation of the Dual Purpose Funds: An
Application of Contingent Claims Analysis," Jou]nal of Financial Economics, January—
March =976, 83-=<4.
Jaffe, J., "The Effect of Regulation Changes on Insider Trading," Bell Jou]nal of Economics
and Management Science, Spring =974, 93-=<=.
Jensen, M., "The Performance of Mutual Funds in the Period =94P-64," Jou]nal of Finance,
May =968, 389-4=6.
, "Risk, the Pricing of Capital Assets, and the Evaluation of Investment Portfolios,"
Jou]nal of Business, April =969, =67-<47.
, "Capital Markets: Theory and Evidence," Bell Jou]nal of Economics and Management
Science, Autumn =97<, 3P7-398.
Joy, M.; R. Litzenîerger; and R. McEnally, "The Adjustment of Stock Prices to Announce-
ments of Unanticipated Changes in Quarterly Earnings," Jou]nal of Accounting Resea]ch,
Autumn =977, <07-<<P.
Kaplan, R. S., and R. Roll, "Investor Evaluation of Accounting Information: Some Empirical
Evidence," Jou]nal of Business, April =97<, <<P-<P7.
, and R. Weil, "Risk and the Value Line Contest," Financial Analysts Jou]nal, July—
August =973, P6-60.
Keim, D., "Size-Related Anomalies and Stock Return Seasonality: Further Empirical Evi-
dence," Jou]nal of Financial Economics, June =983, =3-3<.
Kraus, A., and H. R. Stoll, "Price Impacts of Block Trading on the New York Stock Exchange,"
Jou]nal of Finance, June =97<, P69-P88.
Larson, A. B., "Measurement of a Random Process in Futures Prices," in Cootner, ed.,
The Random Cha]acte] of Stock Ma]ket P]ices. MIT Press, Camîridge, Mass., =964,
<=9-<30.
Lindvall, J., "New Issue Corporate Bonds, Seasoned Market Efficiency, and Yield Spreads,"
Jou]nal of Finance, Septemîer =977, =0P7-=067.
Lintner, J., "The Valuation of Risky Assets and the Selection of Risky Investments in Stock
Portfolios and Capital Budgets," Review of Economics and Statistics, Feîruary =96P, =3-37.
Litzenîerger, R., and H. Sosin, "The Structure and Management of Dual Purpose Funds,"
Jou]nal of Financial Economics, March =977, <03-<30.
Logue, D. E., "On the Pricing of Unseasoned Equity Offerings: =96P-=969," Jou]nal of Finan-
cial and Quantitative Analysis, January =973, 9=-=04.
Mains, N. E., "Risk, the Pricing of Capital Assets, and the Evaluation of Investment Portfolios:
Comment," Jou]nal of Business, July =977, 37=-384.
Malkiel, B., "The Valuation of Closed-End Investment Company Shares," Jou]nal of Finance,
June =977, 847-8P9.
Mandelîrot, B., "The Variation of Certain Speculative Prices," in Cootner, ed., The Random
Cha]acte] of Stock Ma]ket P]ices. MIT Press, Camîridge, Mass., =964, 307-33<.
McDonald, J. G., and A. K. Fisher, "New Issue Stock Price Behavior," Jou]nal of Finance,
March =97<, 97-=0<.
Mikkelson, W., and M. Partch, "Stock Price Effects and the Costs of Secondary Distriîution,"
Jou]nal of Financial Economics, June =98P, =6P-=94.
REFERENCES 399
Mossin, J., "Security Pricing and Investment Criteria in Competitive Markets," Ame]ican Eco-
nomic Review, December 1969, 749-756.
Ohlson, J., and S. Penman, "Volatility Increases Subsequent to Stock Splits: An Empirical
Aberration," Jou]nal of Financial Economics, June 1985, 251-266.
Patell, J., and M. Wolfson, "Anticipated Information Releases Reflected in Call Option Prices,"
Jou]nal of Accounting and Economics, August 1979, 117-140.
, "The Intraday Speed of Adjustment of Stock Prices to Earnings and Dividend An-
nouncements," Jou]nal of Financial Economics, June 1984, 223-252.
Pratt, E., "Myths Associated with Closed-End Investment Company Discounts," Financial
Analysts Jou]nal, July—August 1966, 79-82.
Reilly, F. K., and K. Hatfield, "Investor Experience with New Stock Issues," Financial Analysts
Jou]nal, September—October 1969, 73-80.
Reinganum, M., "The Anomalous Stock Market Behavior of Small Firms in January: Empirical
Tests for Tax-Loss Selling Effects," Jou]nal of Financial Economics, June 1983, 89-104.
Repo]t of the Adviso]y Committee on Co]po]ate Disclosu]e to the Secu]ities and Exchange Com-
mission. U.S. Government Printing Office, Washington, D.C., November 1977.
Ricks, W., "The Market's Responses to the 1974 LIFO Adoptions," Jou]nal of Accounting
Resea]ch, Autumn 1982, 367-387.
Roendfelt, R., and D. Tuttle, "An Examination of Discounts and Premiums of Closed-End
Investment Companies," Jou]nal of Business Resea]ch, Fall 1973, 129-140.
Roll, R. The Behavio] of Inte]est Rates. Basic Books, New York, 1970.
, "A Critique of the Asset Pricing Theory's Tests," Jou]nal of Financial Economics,
March 1977, 129-176.
, "Ambiguity When Performance Is Measured by the Securities Market Line," Jou]nal
of Finance, September 1978, 1051-1069.
, "The Turn-of-the-Year Effect and the Return Premia of Small Firms," Jou]nal of Po]t-
folio Management, Winter 1983, 18-28.
Scholes, M., "The Market for Securities: Substitution vs. Price Pressure and the Effects of
Information on Share Prices," Jou]nal of Business, April 1972, 179-211.
Schwert, W., "Stock Exchange Seats as Capital Assets," Jou]nal of Financial Economics, Jan-
uary 1977, 51-78.
Securities and Exchange Commission, Repo]t of the Special Study on Secu]ities Ma]kets. U.S.
Government Printing Office, Washington, D.C., 1963.
Sharpe, W. F., "Mutual Fund Performance," Jou]nal of Business, January 1966, 119-138.
Shaw, D., "The Performance of Primary Stock Offerings: A Canadian Comparison," Jou]nal
of Finance, December 1971, 1103-1113.
Shelton, J., "The Value Line Contest: A Test of the Predictability of Stock Price Changes,"
Jou]nal of Business, July 1967, 251-269.
Stein, J. P., "The Monetary Appreciation of Paintings," Jou]nal of Political Economy, October
1977, 1021-1036.
Stickel, S., "The Effect of Value Line Investment Survey Rank Changes on Common Stock
Prices," Jou]nal of Financial Economics, March 1985, 121-144.
Stickney, C. P., Jr., "A Study of the Relationships of Accounting Principles and Common Stock
Prices of Firms Going Public," Ph.D. dissertation, Florida State University, Tallahassee,
1970.
<00 EFFICIENT CAPITAL MARKETS: EVIDENCE
Stigler, G., "Puîlic Regulation of Security Markets," Jou]nal of Business, April =964, ==7-=4<.
Sunder, S., "Relationship îetween Accounting Changes and Stock Prices: Proîlems of Mea-
surement and Some Empirical Evidence," Empi]ical Resea]ch in Accounting- Selected
Studies, =973, =-4P.
, "Stock Price and Risk Related Accounting Changes in Inventory Valuation," Ac-
counting Review, April =97P, 30P-3=P.
Thompson, R., "The Information Content of Discounts and Premiums on Closed-End Fund
Shares," Jou]nal of Financial Economics, June—Septemîer =978, =P=-=86.
Treynor, J. L., "How to Rate Mutual Fund Performance," Ha]va]d Business Review, January—
Feîruary =96P, 63-7P.
Watts, R., "Systematic 'Aînormal' Returns after Quarterly Earnings Announcements," Jou]nal
of Financial Economics, June—Septemîer =978, =<7-=P0.
Weinstein, M., "The Seasoning Process of New Corporate Bond Issues," Jou]nal of Finance,
Decemîer =978, =343-=3P4.
Woolridge, J. R., "Ex-date Stock Price Adjustment to Stock Dividends: A Note," Jou]nal of
Finance, March =983a, <47-<PP.
, "Stock Dividends as Signals," Jou]nal of Financial Resea]ch, Spring =983î, =-=<.
1û
Former Student: Professor, this is the same examination you gave to
my class when I was a student twenty years ago. Don't you ever change
the questions?
Professor: The questions don't change—just the answers.
A. INTRODUCTION
Chapters 2 and 3 discussed capital budgeting given the assumption that all future
cash flows were known with certainty. The appropriate discount rate was assumed to
be the risk-free rate, and the chapters focused on selection of discounting techniques
consistent with the goal of maximizing the net present value of shareholders' wealth.
Subsequent chapters introduced uncertainty in the context of a one-period equilib-
rium pricing model, the capital asset pricing model (CAPM).
In this chapter we introduce some of the difficulties implicit in the use of the
CAPM to determine the appropriate multiperiod risk-adjusted discount rate for
capital budgeting purposes. Given multiperiod uncertainty, under what conditions
can one use the following formula to determine the NPV of risky projects?
NC Fit
NPV 1 = (12.1)
0 [1 + E(R j)]t'
401
402 CAPITAL BUDGETING UNDER UNCERTAINTY: THE MULTIPERIOD CASE
where
NPV; = the net present value of project j,
NCF J, = the net cash flow of project j in time t,
E(Ri) = the risk-adjusted required rate of return for project j.
In particular, we are interested in the conditions under which the required rate of
return on the project, frequently called the weighted average cost of capital, can be
determined by the CAPM as written below:
E(R;) = r f [E(Rm) — r (12.2)
Presumably, one would use current estimates of the risk-free rate, r f , the expected
rate of return on the market, E(R m), and the systematic risk of the project, /31, in
order to determine the multiperiod discount rate, E(Ri).
In the first half of this chapter, which deals mainly with theoretical issues, we
review the results of three articles. First, Bogue and Roll [1974] show that the prob-
lem may not be as simple as suggested in Eq. (12.1). They show that a much more
complex procedure becomes necessary if we consider a world where the risk-free rate
in future time periods is not known with certainty. Later Fama [1977] shows the set
of assumptions necessary to use the risk-adjusted discounting procedure of Eq. (12.1),
and Constantinides [1980] shows the minimum set of assumptions necessary for the
multiperiod CAPM to be valid.
Section D of the chapter shows how to use the arbitrage pricing theory for capital
budgeting purposes. While the technique is not easier to use in practice than the
risk-adjusted discount rate, it has the advantage that it requires less stringent as-
sumptions to be empirically valid.
Sections E and F deal with two applied issues. Both sections assume that the
risk-adjusted discount rate may be used for multiperiod capital budgeting under un-
certainty. The simpler of the two problems shows how to adjust the risk-adjusted
discount rate when comparing cost (rather than total income) data for mutually
exclusive projects. The second issue has come to be known as the abandonment prob-
lem. How should one evaluate the residual value of investment assets? We shall see
how option pricing can be applied to this problem, then extend the analysis to dis-
cretionary temporary shutdown of operations, rather than permanent abandonment.
Bogue and Roll [1974] analyze capital budgeting of risky projects in a multiperiod
framework and conclude that it may be incorrect to discount cash flows by using
the single-period risk-adjusted discount rate. However, under specified conditions,
the investment decision for a multiperiod project can be made with a one-period
MULTIPERIOD CAPITAL BUDGETING WITH "IMPERFECT" MARKETS 403
forecast. If appropriate secondary markets exist for the project, only a one-period
analysis is required for a decision. The firm makes a comparison between the current
investment outlay and the value of the forecast of cash flows during the first period
plus the forecast end-of-period secondary market price.
In addition, even with imperfect secondary markets for physical capital, some
investment decisions can still be made with one-period forecasts. If the machine is
acceptable on the basis of its one-period cash flow plus its net salvage value after the
first period, the possibility of values in subsequent periods will only add to the
acceptability of the project. Because the single-period analysis is important in its own
right, and also because it will be used in a dynamic programming framework to solve
the multiperiod problem, Bogue and Roll start the analysis with a single-period
valuation model. They begin with the CAPM in value form. We derive their basic
valuation expression by starting with the CAPM in return form as shown in Eq.
(12.3):
COV(R' 1, Rmi )
E(k 1) = [E(km1) r10] 2(k ) (12.3)
6 m1
By definition,
0-2(R m i) = 62 rmi 1) =
Vnio ( Vm0)
1 2
62(Vml)'
f7m1 1
COV(k1, km1) = COV (V1 1, 1) = COV(17i,
Vo Vrn0 VoVmo
where
V0 = the certain current value of the firm,
Vl = the uncertain end-of-period value of the firm (including any dividends paid
over the period),
-
Vm1 = the uncertain end-of-period value of the market portfolio,
rf 0 = the risk-free rate of interest over the period.
Then the security market line (SML) is
E(171) [E(Vtni) I Vmo )COV(Vi, Pm1)
V70
= rf -F rfol ( vini) (12.4)
17„,, VO
Substituting and rearranging, we can obtain a certainty equivalent value for the firm.
First, we multiply both sides by V0 :
E(17i ) — Vo = Vo r fo [E(P„,} ) (1 + rfo)Vmd [COY( 17
mi)/(72( Vmi)].
Next, we subtract Vorfo and E(171) from both sides and change signs:
(1 + r f 0 )V, = E(V1) — [E(17,,} ) — (1 + r fo )Vmo ][COV(17,, i7m1)162(Vini)].
404 CAPITAL BUDGETING UNDER UNCERTAINTY: THE MULTIPERIOD CASE
Note that there is the implicit assumption that the pro1ect returns cash only in the nth time period.
MULTIPERIOD CAPITAL BUDGETING WITH "IMPERFECT" MARKETS 405
where
En _ 1 = state of the world at time n - 1,
= market price of risk at n - 1,
Ff (n — 1) = risk-free rate at n - 1 (assumed to be stochastic),
E(Xnlgn -1) = conditional expectation at n - 1 of cash flows at n.
This enables us to obtain the incremental value at n - 1:
E(ie ,i 1En _ 1 ) — 1 C0VP?n , 1711mn ÿ 1)
A Vn _ = X n _ + (12.9)
1 -I- F
f(n 1)
The result in Eq. (12.9) can be generalized to the recursive relationship for the
incremental value at any time
time k:
1 14) — ;ik COV[AT7„,,, 17m(k + 1)1 4]
k (12.10)
1 + rfk
From Eq. (12.10) we see that for an n-period project, an n-period infinite-state
dynamic programming problem must be solved. Each step involves an application
of the one-period valuation model with the parameters depending on the state of the
world at the beginning of that particular period. Next, the nature of this solution is
illustrated for the special case of a project with a single cash inflow two periods in
the future. Using Eq. (12.10) with the risk-free interest rate expression placed on the
left-hand side, we can obtain the incremental value for period 1 and period 0 as
shown below:
i7/ (1 + = Al2 g1) ;116*12 5 i7M2 g1), (12.11)
AV0(1 + r fo ) = E(4171 ) - 20 00V(41-71, V.1). (12.12)
Note that the second-period risk-free rate, is currently uncertain and will not
be revealed until the end of the first period. Next, we take expectations of Eq. (12.11),
and making use of the covariance identity,
COV(5C, = E(X)E(Y),
2
we have
E(AV=)E(1 + Ffi ) + COV(A1-7,,Ffi) = E(X <) — EP1COV(k2, Vm2 1 E1)1
2
The covariance identity follows directly from the definition of covariance:
COV(I, = E[(1 - E(1))(k - E(k))]
= E[(XY - E(X)Y - E(17)X + E(X)E(Y)]
= - E(X)E(Y) - E(Y)E(X) + E(X)E(Y)
= E(XY) - E(X)E(Y).
This result is applied to the left-hand side of Eq. (12.11) because it is the expectation of the product of
two random variables.
406 CAPITAL BUDGETING UNDER UNCERTAINTY: THE MULTIPERIOD CASE
We then solve Eq. (12.12) for E(4 V1) and substitute in the preceding expression and
arrange terms. The result is
A Vo(1 + rfo)E(1 + f f 1 ) = E(I2) — ECI 1COVOZ 2 ,
— .10 00V(4171, T7,,1)E(1 + Fp) — COV(A Fp). (12.13)
Equation (12.13) can also be written in the following form: A Vo , the present value of
the single uncertain cash flow two periods in the future, is equal to
E(X 2 ) EL):1C011()?2 , T7„,2 1"J1)] .10 00V(A, V„, i )
—
(1 + rfo)E(1 + F.f1) (1 + r f 0 )E(1 + F f ,) (1 + r f 0 )
COV (4171, F11)
(12.14)
(1 + rfo)E(1 + 1711)
The first of the four terms on the right-hand side of (12.14) is the two-period dis-
counted current expectation of the uncertain cash flow two periods in the future.
Subtracted from it are three risk premiums: (1) covariation risk within the second
period—i.e., beta risk; (2) covariation risk of the intermediate value of the project,
which may be thought of as a reinvestment opportunity cost related to the sale of
rights to the cash flow after one period has elapsed; and (3) the risk premium for
interest rate fluctuations over the two time periods, which could cause changes in
the project's value at intermediate periods.
In its rearranged form, Eq. (12.13) is in the same form and spirit as the single-
period pricing model (Eq. 12.7). However, it contains two additional risk premia that
have to be deducted from the two-period discounted current expectation of the single
uncertain cash flow two periods in the future. The first is the covariation risk of the
intermediate value of the project, and the second is the risk premium charged for
the risk of interest fluctuations over the two time periods.
Bogue and Roll conclude by observing that if the errors in probability assess-
ments of the cash flows over the multiple time periods are not systematically biased,
stockholders can diversify away most of the error as the number of projects becomes
large. Thus unbiased misassessments can be diversified in the personal portfolios of
stockholders, whereas the use of a wrong capital budgeting criterion will result in
aggregate errors that stockholders will not be able to reduce by diversification. In
concept the multiperiod capital budgeting problem must utilize a valuation expression
that includes two additional risk measures over and above the discounted current
expectation of cash flows and the usual covariation of those flows with total market
values.
C. AN EXAMINATION OF ADMISSIBLE
UNCERTAINTY IN A MULTIPERIOD
CAPITAL ASSET PRICING WORLD
In the previous section we presented Bogue and Roll's suggestion that in a two-period
context the present value of the firm (Eq. 12.14) cannot be calculated by simply dis-
counting the certainty equivalent cash flows at the end of the second time period
AN EXAMINATION OF ADMISSIBLE UNCERTAINTY 407
back to the present. In addition, it is necessary to subtract two additional risk premia:
(1) a term for the covariation risk of the intermediate value of the project and (2) a
term for the risk of fluctuations in the risk-free rate over the two time periods.
Fama [1977] reexamines the multiperiod capital budgeting problem under un-
certainty and clarifies the Bogue and Roll analysis by showing that within a CAPM
world, certain types of uncertainty that are allowed by Bogue and Roll are inad-
missible. He then shows that given the CAPM assumptions the last two terms of Eq.
(12.14) vanish, and it is possible to use the risk-adjusted rate of return approach to
capital budgeting as suggested in Eq. (12.1).
If we assume that the firm has net cash earnings, IZ-t, at time t and no cash flows
at any other time, the recursive relationship for the value of the firm at t — 1 can be
written in a form similar to Eq. (12.5):
t — 0,COV(k „ kit
} (12.15)
4-} =
1 + r ft
where 0, = [E(R mt — r Amo-2( k nit . This is a certainty-equivalent expression for the
value of the firm at t — 1. The firm's value at t — 1 can also be expressed using the
risk-adjusted discount rate to compute the present value of the expected end-of-period
cash flows:
,E()7t)
Vt-} = (12.16)
} + E(k ]
Vt — 2 = -+-
r r f,t— }
3
The Merton [1973] study is discussed briefly in Chapter 7.
408 CAPITAL BUDGETING UNDER UNCERTAINTY: THE MULTIPERIOD CASE
and at t = 0 it is
t -1 ( )
Vo =
k=1 1 ± r fk
Finally, using (12.16), we obtain
170 = ( E( (12.17)
k=1 + rfki 1 + E(kt)/
In this case the appropriate discount rates prior to period t are the risk-free rates
because there is no uncertainty until period t. For period t the risk-adjusted rate is
given by the CAPM relationships.
Of course the previous assumption is unreasonably strong. However, if Vt _ 1 is
to be uncertain prior to t — 1, the uncertainty must be introduced in a fashion con-
sistent with the CAPM. Suppose that the cash flow in period t is estimated in an
unbiased fashion in period t — 1 conditional on all information available at that time.
This process can be expressed as
Xt = Et Alt)( 1 + Et) = Et- gt) + Et-1(U, (12.18)
.
where E,_ }(,k,) is the expected value of X'„ and Et is a random variable with expected
value equal to zero. Prior to t — 1 the expected value itself is a random variable.
This process evolves in the following fashion:
:Olt) = Et-1(1t)(1 + Et) = Er- i(it) + Er - ()Zt)Et. (12.19)
Again, the expected value of Et, conditional on the availability of information at i - =,
is equal to zero. Therefore, given rational expectations, the value of cash flow at time
t, X„ evolves as a martingale.' Note that k is the change in the expected value of
X, per unit of Er _ }(X,):
fr(k-t Er-}( Zt Er()ZT)
= 1. (12.20)
E,_ 1(it) E,_ 1(X„)
Substituting (12.18) into (12.15), we have the value of the firm as of t — 1:
1
17, _ = [E,_ 2()?t) + Et 2(U- 1][ ]* (12.23)
1 + E(Rt)
The implication of (12.23) is that the value of the firm at t — 1 is perfectly correlated
with Et _ 1(X t), which is the assessment of the expected value of earnings turning up
at t — 1.
Taking the expectation of (12.23), we see that the expected value of 17t _ 1 as of
t — 2 is
COV(Vt- =, - 1) (12.26)
Et 7// t- 1)
2(1
and from (12.20),
cov[E, _ t ), -1 ]
= cov(k_ ,, t- (12.27)
Et- 2(k t)
Therefore the covariance between the value of the firm at t — 1 and the market port-
folio per unit of Et _ 2(Vt _ ,) is identical to the covariance between the expected value
of earnings and the market portfolio per unit of Et _ 2( X t). Now the value of the firm
as of t — 2 may be written as
Ot-=COV(f2t-=,
Vt- 2 -
1
+ r 1 ,t - 1
= Et _ 2(17t — 1) (12.28)
[ 1- + E(k(-1)1
l
We can rewrite Et _ 2(I7;_ i ) by using (12.24) to obtain
1
V- 2 = Et— 2(kt)[ 1
The market value of the firm at r is the expected value at r of the earnings to be
realized at time t, discounted at the risk-adjusted discount rates for each of the periods
between i and t.
Fama points out that if the CAPM is assumed to hold, only uncertainty about
E,(X,) is admissible. Uncertainty about the risk-adjusted discount rates, E(Ri ,1), ,
E(R,), is not admissible. The risk adjustments in the discount rates arise because of
the uncertain evolution through time of the expected value of cash flow.
If we are to obtain the usual solution to the multiperiod capital budgeting prob-
lem, we must also assume that the risk-free rate, r f „ the covariance, COV(k,
and the risk-adjusted rate, E(R,), are constant through time. If so, we obtain
Eo(X,)
=
[1 + E(R)]t
which, of course, is equal to Eq. (12.1) for an example with cash flow only in the last
time period.
Bogue and Roll [1974] show that if the expected risk-free rate and therefore the
expected portfolio opportunity set are stochastic, then the multiperiod capital bud-
geting problem is not easily solved. The investor must not only consider systematic
risk in the usual CAPM sense but must also take into account two additional factors:
(1) the risk of fluctuations in the risk-free rate and (2) the covariation risk of the in-
termediate value of the project. These results are consistent with various multiperiod
versions of the CAPM (e.g., Merton [1973] and Long [1974]) that assume a stochastic
risk-free rate.
Fama [1977] carefully examines the types of variability admissible under a
stationary CAPM that assumes that the portfolio opportunity set is nonstochastic.
In general the only admissible form of uncertainty is in the expected cash earnings
in time t, assessed as of time r < t. The risk-adjusted discount rates in each future
time period are known with certainty at time 2. Given the somewhat unpalatable
assumptions of the stationary multiperiod CAPM, we have the result that the usual
textbook treatment of multiperiod capital budgeting under uncertainty is reasonable.
USING THE ARBITRAGE PRICING THEORY 411
Recall, from Chapter 7, that the assumptions of the arbitrage pricing theory (APT)
are less restrictive than those of the capital asset pricing model (CAPM), especially
in its multiperiod form. The APT does not require that the market portfolio be
observable, nor does it require intertemporal stationarity in the investment oppor-
tunity set. Rather, it requires only that at any instant in time there be no available
arbitrage opportunities.
Ross [1979] shows how the APT can be used to value risky income streams, and
Gehr [1981] shows how the idea can be applied specifically to the multiperiod capital
budgeting problem. Gehr's example uses the APT to avoid the problems implicit in
using a multiperiod risk-adjusted discount rate, but it is no panacea because it re-
quires that we estimate a relationship between the future price of (some) publicly
traded asset and future cash flows for a project.
412 CAPITAL BUDGETING UNDER UNCERTAINTY: THE MULTIPERIOD CASE
rf
Economic Subjective Risk-free Comparison
Period State Probability Project CF Rate Stock
0 current 1.0 —8,000 NA $ 9.00
1 bad .3 8,000 10% $ 8.00
1 good .7 15,000 10% $25.50
5
The price per share of the stock is $9, and implicitly the price per unit of the risk-free asset is $1.
USING THE ARBITRAGE PRICING THEORY 4=3
CF = $180, S = $40
State C
S = $30, rf = 10%,
CF = $150
State A State D
S= $10,rf = 12%, CF = $160, S =$30
10 = $-200
CF = $130, S = $12
State E
S = $8, r f = 8%,
CF = $120
State B
State F
CF = $120,S = $8
Figure 12.1
Two-period APT example, first stage.
about the risk tolerances of the firm (or investors) that evaluated the project. All this
information is implicitly included in the relative prices (across time and across states
of nature) of the risk-free asset and the risky comparison stock. As long as their prices
are true equilibrium prices, then the market information tells us all that we need to
know.
The technique can readily be extended to a multiperiod economy, even one where
the risk-free rate changes through time. Figure 12.1 gives data for a two-period
example. The probabilities of states of nature are not given because they are not
explicitly needed to solve the problem. Note that the risk-free rate is state con-
tingent. If a favorable state of the economy prevails in year 1, then r f = 10%; other-
wise it is 8%. Also, the correlation between the project and the comparison stock
need not be stationary through time.
The project can be evaluated by finding the cash equivalent portfolios between
years 1 and 2, then using this information to find the year 0 cash equivalent port-
folio.' The two sets of simultaneous equations given below are used to determine the
period 1 cash equivalent portfolios:
State C 40Q, + 1.10Q, = 1801 Q, = 2, Q,. = $90.91
State D 30Q, + 1.10Q, = 160f VA = $150.91
State E 12Q, + 1.08Q, = 130' Q, = 2.5, Qr = $92.59
State F 8Q, + 1.08Q, = 120, V, = $112.59.
6
If this procedure sounds similar to the Cox, Ross, Rubinstein binomial model, derived in Chapter 8, it
is. The arbitrage pricing theory, the option pricing model, and state-preference theory (even the CAPM)
are all consistent with each other.
414 CAPITAL BUDGETING UNDER UNCERTAINTY: THE MULTIPERIOD CASE
S = $8, CF = $120,
VB = $112.59
State B
Year 0 Year 1
As shown in Fig. 12.2, the values of the cash equivalent portfolios in States A and
B are $150.91 and $112.59, respectively. The value of the project, in State A, e.g., is
the cash flow that it provides in that state of nature, plus the cash equivalent value,
VA. This information can be used to write the appropriate set of payoffs for the period
1 cash equivalent portfolio as shown below:
State A 30Q, + 1.12Q,, = 150 + 150.91
State B 8Q# + 1.12Q,. = 120 + 112.59.
Solving, we find that Q, = 3.11 shares and Qr = $185.49. Thus the value of the cash
equivalent portfolio for the two-year project is
PV = (3.11 shares)($10.00/share) + $185.49
= $216.59,
and the net present value of the project is
NPV = PV — / = $216.59 — $200 = $16.59.
Therefore the project should be accepted.
Despite the obvious theoretical advantages of using the arbitrage pricing theory
for capital budgeting, there remains the practical problem of estimating the state-
contingent prices of the comparison stock and the risk-free asset. Obviously this is
no easy task. Whether or not decision makers will find this new technique of practi-
cal value remains to be seen. Who knows, maybe a decade from now, questionnaires
on capital budgeting techniques will ask sophisticated firms whether or not they
employ the APT capital budgeting technique. Only time can tell.
Applied capital budgeting problems are almost always multiperiod, and frequently
it is reasonable to assume that the revenues from two mutually exclusive projects
will be identical. For example, this is the usual assumption in machine replacement
COMPARING RISKY COST STRUCTURES 415
problems. The revenues of the firm will be invariant to the choice of equipment.
Therefore the usual capital budgeting process simply discounts the various incre-
mental costs associated with the mutually exclusive alternatives and chooses the
project with the lowest discounted cost.
Assuming that the multiperiod risk-adjusted rate of return is the appropriate
technique for capital budgeting, how should it be used to compare cash outflows on
a risk-adjusted basis? As we shall see, the correct approach discounts the expected
costs at a lower rate when the cash outflows have greater variance. As long as the
absolute value of the cash outflows is positively correlated with the market portfolio,
greater variance in the outflows will imply lower overall project risk; i.e., there will
be a lower project beta. To develop this result let us look at a simple one-period
case. At the end of the period, the after-tax cash flows from operations may be written
as follows:
CF = (Rev — VC)(1 — r c) + r cdep,
where
CF = after-tax cash flows for capital budgeting purposes,
Rev = end-of-period revenues,
VC = end-of-period variable cash costs,
= the corporate tax rate,
dep = depreciation.
The rate of return, on the project is the return on investment (where investment
is 10):
(Rev — VC)(1 — t) + Tc dep — I,
If we assume that the project is fully depreciated during the period, then / 0 = dep,
and we have
T ) (1 T )
(1 ` Rev c VC — (1 — t c ).
/0 Jo
Using the properties of random variables derived in Chapter 6, we can write the
covariance between the return on the project and the return on the market portfolio
as
(1 — c) (1 — t e)
COV(Fi, fm) = COV(Rev, fin) COV(VC, Fin). (12.30)
Io /0
Equation (12.30) shows that the covariance risk of a project can be partitioned into
two parts: the covariance risk of its revenue stream and the covariance risk of its
416 CAPITAL BUDGETING UNDER UNCERTAINTY: THE MULTIPERIOD CASE
cost stream. Note that if the costs have positive covariance with the market in the
sense that they are high when the market return is high, and vice versa, then their
contribution to the project's covariance risk will be large and negative. This implies
that the cost streams of riskier projects should be discounted at lower (and even
negative) discount rates in order to properly adjust for risk. Note also that Eq. (12.30)
can be rewritten in terms of systematic risk by dividing both sides by the variance
of market return. This yields
(1 —
Tc) R (1 —
tc) u
16 j = /0
YjRev
/0
PjVC, (12.31)
where
= the systematic risk of the project where j = 1, 2,
fijRev = the systematic risk of the revenue stream,
F. = • - =0
•
Io
By inspecting the cash flows in Table 12.2, we see that the revenue streams are posi-
tively correlated with the market return, and so are the variable cost streams of the
projects. However, the correlation between the variable cost stream of project 1 and
the market return is such that it causes the project's net cash flow, CF, to be neg-
atively correlated with the market. The higher variance and the positive correlation
between (the absolute value of) the cash outflows of the first project and the market
return require that its cash costs be discounted at a lower rate. Project 1 has a lower
/3 than project 2.
Table 12.3 shows the results of computations of various statistics necessary for
the calculation of the risk-adjusted rate of return. By using the CAPM and the statis-
State 1 .33 .26 .04 610 500 105 .05 610 495 107.5 .075
State 2 .33 .14 .04 600 470 115 .15 600 500 100.0 0
State 3 .33 .20 .04 610 520 95 —.05 610 505 102.5 .025
COMPARING RISKY COST STRUCTURES 417
Project 1 .050 606.67 496.67 —.0020 .20 .60 —.833 83.33 250.00
Project 2 .033 606.67 500.00 .0015 .20 —.10 .625 83.33 —41.67
Market return .200 — .0024 — 1.000 —
tics in Table 12.3, we can calculate the appropriate discount rates for the projects'
net cash flow streams, or their cost streams. Using project 1 as an example, we employ
the CAPM
and use Eq. (12.31) to compute the correct adjusted betas: for the project, f3J , its
revenues, /3jRev and its costs, , B jvc:
( 1 — ic) n
7
1 — t e ),
jfi = /0 1
'R
Rev
/0
PjVC•
Substituting data from Table 12.3, we see that the systematic risk for the project is
a weighted average of the betas for the revenue and cost streams.'
(1 — .5 (83.33) (1 — .5)
— .833 = (250)
100 ) 100
= .4167 — 1.25.
Note that the beta for the cost stream is negative because the costs are outflows. The absolute value of
the cost stream is positively correlated with the market, but the signed cash costs are negatively correlated
with the market.
8 It may seem unusual to discount cash flows at negative rates of return. Normally, this would not be the
case because real-world projects are almost always positively correlated with the market. However, in the
artificially constructed example above, it is perfectly consistent with the CAPM to require negative rates
on projects with negative betas.
418 CAPITAL BUDGETING UNDER UNCERTAINTY: THE MULTIPERIOD CASE
E(Ri )
20
10 Project 1 revenues
(10.67%, .4167)
—10
Project 1 costs
(-16%, —1.25) —20
Figure 12.3
Systematic risk and required rate of return for project 1.
Suppose you are asked to compare the risk of two compensation schemes. In
the first, salespeople are paid on a commission basis with no salary. The second
scheme is straight salary with no commission. The expected revenues (annual sales
figures) and expected sales costs (including compensation) are identical for both
schemes. Which compensation plan has greater variance? Which has greater risk for
shareholders? The intuition should be clear. The commission scheme has greater cost
variance but results in less risk for shareholders because it smooths profits. When
revenues are low, so are commissions, and vice versa. Hence profits have less variability
than they otherwise might.
A word of caution to the practitioner is appropriate at this point. It is never ad-
visable to totally ignore revenues or the riskiness of the revenue stream (even though
this is frequent pi.actice). In the above example the market required rate of return
on project 2 is 14'%, and the expected rate of return (the mean return in Table 12.3)
is only 3.3%. Therefore project 2 is unacceptable under any circumstances. This could
have been true for both projects. The practitioner who ignores revenues and chooses
the project with the lower discounted cost may easily accept a project with negative
NPV. Costs tell only half the story. Decision making on the basis of cost comparisons
alone is inappropriate unless the decision maker is absolutely sure that the mutually
exclusive projects all have positive net present value.
Given the results of Fama [1977] and Constantinides [1980], which were dis-
cussed in the previous section, it is possible to extend the one-period cost comparison
procedure into a multiperiod framework. So long as the market parameters are as-
sumed ex ante to be stationary over time, we still have the result that cost streams
with higher variance should be discounted at lower rates.
F. ABANDONMENT VALUE
9
For example, an accident at a nuclear power plant or a large chemical spill may leave management in
the position where they wish they could simply walk away from the problem. Ex ante the NPV of such
projects is reduced because, by law, they cannot be abandoned.
420 CAPITAL BUDGETING UNDER UNCERTAINTY: THE MULTIPERIOD CASE
Year I Year 2
Initial Conditional
Probability Probability
p(1) Cash Flow p(211) Cash Flow
chances are that the second year will also bring high profits. To obtain the proba-
bility that a particular first-year outcome and a particular second-year outcome will
both occur, we must multiply the initial probability by the conditional probability
to obtain what is termed the joint probability.
These concepts are applied to the data of Table 12.5 to construct Table 12.6.
The project is not expected to have any returns after the second year. The cost of
capital relevant to the project is assumed to be 12 percent. To indicate the role of
abandonment value, we first calculate the expected net present value of the invest-
ment and the expected standard deviation of the project's internal rate of return with-
out considering abandonment value. In the calculation made in Table 12.6, we find
the expected NPV to be $201.
Next, in Table 12.7, we calculate the standard deviation of the project's rate of
return, finding that a = 33.5%. Then, we can expand this analysis to take abandon-
ment value into account. Suppose the abandonment value of the project at the end
of the first year is estimated to be $250. This is the amount that can be obtained by
liquidating the project after the first year, and the $250 is independent of actual first-
year results." If the project is abandoned after one year, then the $250 will replace
any second-year returns. In other words, if the project is abandoned at the end of
year 1, then year 1 returns will increase by $250 and year 2 returns will be zero. The
present value of this estimated $250 abandonment value is, therefore, compared with
the expected present values of the cash flows that would occur during the second
I°
In other words, we assume that the exercise price of the put option at the end of the first year is known
with certainty. It is not a random variable.
Table 12.6 Calculation of Expected Net Present Value
Year 1 Year 2 Probability Analysis
From J. F. Weston and T. E. Copeland, Managerial Finance, 8th edition, Hinsdale, Ill., Dryden Press, 1986, 515. © 1986 CBS Publishing.
1-+
Table 12.o Calculation of éate of éeturn Standard Deviation
Cash Flow
Joint
êea] 1 êea] 2 Iéé Iéé - Iéé (Iéé - Iéé)2 x P]obability = pi(Iéé - Iéé)2
$<00 $=00 0.0% -.P94 .3P< .09 .03=8
<00 <00 <=.P -.379 .=44 .=P .0<=P
<00 300 38.7 -.<07 .043 .06 .00<6
300 <00 4P.7 -.=37 .0=9 .=< .00<3
year if abandonment did not take place. To make the comparison valid, however,
we must use the second-year flows based on the conditional probabilities only, rather
than the joint probabilities that were used in the preceding analysis. This calculation
is shown in Table 12.8.
We next compare the present value of the $250 abandonment value, $250 x
0.8929 = $223, with the branch expected present values for each of the three pos-
sible cash flow patterns (branches) depicted in Table 12.8. If the $223 present value
of abandonment exceeds one or more of the expected present values of the possible
branches of cash flows, taking abandonment value into account will improve the in-
dicated returns from the project. The $223 does exceed the $152 expected PV shown
in Table 12.8 for second-year cash flows when the first-year cash flow is $200. In
Table 12.9, therefore, abandonment after year 1 is assumed for the $200 case and
the new NPV is calculated; the $250 abandonment value is added to the $200 cash
flow to obtain a $450 year 1 cash flow, and the year 2 cash flow becomes $0. The
new calculation of the standard deviation of returns is shown in Table 12.10.
We may now compare the results when abandonment value is taken into account
with the results when it is not considered. Including abandonment value in the calcu-
lations increases the expected net present value from $201 to $223, or by about 10%.
It reduces the expected standard deviation of returns from 33.5% to 22.3%. Thus for
this problem, abandonment value improves the attractiveness of the investment.
Abandonment value is important in another aspect of financial decision making:
the reevaluation of projects in succeeding years after they h'ave been undertaken. The
decision to continue the project or to abandon it sometime during its life depends
Present
Year 1 Year 2 Value
Cash PV Cash PV of Total Joint Expected
Flow x Factor PV Flow x Factor = PV Cash Flow x Probability = Value
(1) (2) (3) (4) (5) (6) (7) (8) (9)
$450 0.8929 $402 $ 0 0.7972 $ 0 $402 0.30 $121
{200 0.7972 159 427 0.12 51
300 0.8929 268 300 0.7972 239 507 0.20 101
400 0.7972 319 587 0.08 47
'300 0.7972 239 596 0.09 54
400 0.8929 357 400 0.7972 319 676 0.12 81
500 0.7972 399 756 0.09 68
1.00
Expected present value = $523
Expected net present value = $223
From J. F. Weston and T. E. Copeland, Managerial Finance, 8th edition, Hinsdale, Ill., Dryden Press, 1986, 517. © 1986 CBS Publishing.
Taîle =<.=0 Calculation of Rate of Return Standard
Deviation with Aîandonment Value Included
Cash Flow
Joint
êea] 1 êea] 2 IRR IRR - IRR (IRR - IRR)< x P]obability = pi(IRR - IRR)<
$4P0 $ 0 P0.0% -.=88 .03P .09 .003<
4P0 0 P0.0 -.=88 .03P .=P .00P3
4P0 0 P0.0 -.=88 .03P .06 .00<=
300 <00 4P.7 -.<3= .0P3 .=< .0064
300 300 6=.8 -.070 .00P .<0 .00=0
300 400 7P.8 .070 .00P .08 .0004
400 300 86.9 .=8= .033 .09 .0030
400 400 =00.0 .3=< .097 .=< .0==6
400 P00 ==<.0 .43< .=87 .09 .0<49
afrIVA INITAINOCINV£1V
Sum =.00 VAR(IRR) = .0498
Note- IRR = E AIRR i , where pi = joint proîaîility, IRR = 68.8%
[VAR(IRR)]"< = a(IRR) --- .<<3< or <<.3<%.
From J. F. Weston and T. E. Copeland, Manage]ial Finance, 8th edition, Hinsdale, Ill., Dryden Press, =986, P=8. c =986 CBS
Puîlishing.
426 CAPITAL BUDGETING UNDER UNCERTAINTY: THE MULTIPERIOD CASE
on which branch occurs during each time period. For example, suppose that during
year 1 the cash flow actually obtained was $200. Then the three possibilities asso-
ciated with year 2 are the three that were conditionally dependent upon a $200 out-
come in year 1. The other six probabilities for year 2, which were considered in the
initial evaluation, were conditional upon other first-year outcomes and are thus no
longer relevant. A calculation (Table 12.11) is then made of the second-year net cash
flows, discounted back one year.
At the end of the first year the abandonment value is $250. This is compared
with the expected present value of the second-year net cash flow series discounted
one year. This value is determined to be $171, so the abandonment value of $250
exceeds the net present value of returns for the second year. Therefore the project
should be abandoned at the end of the first year.
In summary it is sometimes advantageous to abandon a project even though the
net present value of continued operation is positive. The basic reason is that the pres-
ent value of abandonment after a shorter time may actually be greater than the
present value of continued operation.
Another, perhaps better, way to analyze the abandonment decision is to compute
the NPV of the project without the option to abandon, then add to it the value of
the abandonment put option. Thus we have
NPV (with abandonment) = NPV (without abandonment)
+ Value of abandonment put option.
The greater the variance of returns on the project, the greater will be the value of
she abandonment option. In Table 12.7 we saw that the standard deviation of returns
was 33.5% for the project without the abandonment option. This is the correct stan-
dard deviation to use in the Black-Scholes formula because it is an estimate of the
standard deviation of returns on the underlying asset. The way our example has been
structured, we also know that the put option may be exercised only at the end of
the first year. Therefore it is a European put option with one year to maturity and
an exercise price of $250. The present value of the underlying asset is the present
value of the project without abandonment, i.e., $501. We assume the risk-free rate
is 5%.
Year 2 Joint
Cash Flow Probability PV Factor PV
Note that if the project is abandoned at the end of the first year, we will abandon
it only after receiving the first year's cash flows. Therefore we must compute the asset
value without these cash flows in order to value the abandonment put option. The
calculation is shown in Table 12.12.
The value of the abandonment put option can be found by using the Black-
Scholes formula to value the corresponding call, then put-call parity to compute the
put value. The Black-Scholes call value is
c = SN(d 1 ) - X e -r N(d 2),
where
ln(S/X) + r fT + 1 -
di = o
a \FT 2
d 2 = d i - o- 11.
Substituting in the numbers from our example, we have
ln(233.57/250) + .05(1) 1
di = + (.335)
.335/i 2
ln(.9343) + .05
+ .168
335
- .068 + .05
+ .168
.335
= -.0537 + .168 = .1143
d 2 = .1143 - .335 - .2207.
428 CAPITAL BUDGETING UNDER UNCERTAINTY: THE MULTIPERIOD CASE
Using the table of cumulative normal probabilities (Table 8.7) at the end of Chapter
8, we find that
N(d i ) = .5 + .0455 = .5455
and
N(d 2) = .5 — .0874 = .4126.
Thus the value of the call option is
c = 233.57(.5455) — 250(.4126)e-.°5('
= 127.41 — 250(.4126)(9512)
= 127.41 — 98.12 = 29.29.
Finally, we can use Eq. (8.10), put-call parity, to find the value of the European put
that is implied by the option to abandon:
- T
co — P o = So — Xe ri (8.10)
P o = c o — So + Xe'f'
.°5( )
= 29.29 — 233.57 + 250e '
= 29.29 — 233.57 + 250(.9512)
= 29.29 — 233.57 + 237.80
= $33.52.
The decision tree approach gave an abandonment value equal to $22 (i.e., $223,
the value with abandonment, minus $201, the value without abandonment). The op-
tion pricing approach gave an abandonment value of $33.52. We obtained different
answers because the assumptions of the Black-Scholes OPM and the decision tree
approach are different. For example, Black-Scholes assumes a lognormal distribution
of outcomes, whereas the decision tree only crudely approximates the continuum of
possibilities. It is hard to say which assumption is more realistic for project abandon-
ment decisions.
2. Deferred Abandonment
The traditional abandonment decision rule is that the project should be aban-
doned (or sold) in the first year that the abandonment value exceeds the present value
of the remaining expected cash flows from continued operation. Unfortunately, aban-
donment at the first opportunity may not be optimal because deferred abandonment
may result in an even greater net present value." For example, consider a truck with
two years of remaining useful life. The present value of continued use is, say, $900,
11
See Dyl and Long [1969], Robichek and Van Home [1967], and Joy [1976].
ABANDONMENT VALUE 429
but the current market value of the truck is $1000. Clearly, if the proceeds from the
sale can be invested to earn at least the applicable cost of capital, the better decision
would be to sell the truck now. However, there is one option that has not been con-
sidered, which is to operate the truck for another year and collect the cash flow
from one year's operations (which have a present value of $500) and then abandon it
(assuming the present value of abandonment in a year is $600). Thus the present value
of this alternative is $1100. In this case the truck should be used for one year and
then sold.
The optimal abandonment decision rule is to determine the combination of re-
maining operating cash flows and future abandonment that has the maximum ex-
pected net present value. This decision rule is, unfortunately, difficult to implement,
especially when the project life is long and there are numerous opportunities for aban-
donment over time. If a piece of equipment can be used for 20 years or abandoned at
the end of any year, then 20 different net present value calculations might be required
to determine the optimum pattern that will result in maximum expected net present
value.
Option pricing is difficult to apply for two reasons. First, the resale price or aban-
donment value (analogous to the exercise price of the implied American put) is not
constant across time and may be dependent on variables such as the state of the
economy. Second, the underlying asset (the project) pays cash flows ("dividends") to
those who hold the project and the implied American put on it.
At present the "best" solution to deferred abandonment decisions is to consider
n different mutually exclusive alternatives for an n year project and choose from
among them the one that has the greatest NPV.
and Pindyck [1987] value the flexibility option to accelerate or delay the construction
of a project. When uncertainty about the value of the completed project increases, so
does the value of delaying. McDonald and Siegel [1986] model the value of waiting
to invest in an irreversible project—a similar problem. It is likely that option pricing
will provide a better way to think about research and development. Roberts and
Weitzman [1981] have written an early paper on the topic. Paddock, Siegel, and
Smith [1984] have applied the option pricing methodology to offshore oil leases.
The application of option pricing to the problem of modeling flexibility for stra-
tegic investment decisions is exciting because it is the first major advance in capital
budgeting in decades. It shows great promise for solving a wide set of interesting and
useful problems.
SUMMARY
By considering the relations between periods recursively, Bogue and Roll [1974]
develop an equation for multiperiod capital budgeting. In addition to the usual beta
risk, the resulting equation contains two additional risk premia: the covariation risk
of the intermediate values of the project and the risk premium related to the risk
of interest fluctuations. Fama [1977] points out that in a world in which securities
are priced according to the CAPM, the only admissible form of uncertainty is the
expected cash earnings in time t assessed one period earlier. The risk-free rate, the
covariance, and the risk-adjusted rate are then constant through time, and the tradi-
tional capital budgeting model can be employed.
Gehr [1981] shows how the arbitrage pricing model can be employed to solve
multiperiod capital budgeting problems. The technique avoids all the difficulties of the
multiperiod CAPM but unfortunately requires estimates of state-contingent future
security prices.
In most equipment replacement problems it is plausible that revenues from two
mutually exclusive types of equipment will be identical. However, the riskiness of the
cost streams may be different. We are accustomed to increasing the amount of risk
adjustment in the capitalization factor applied to riskier net cash flow streams. This
has the effect of penalizing riskier investments. When the streams under comparison
are costs alone, we need to have higher present values for riskier cost streams. This
is accomplished by lowering the discount rate, rather than increasing it.
Another important aspect of the assessment of risky investments is to estimate
the abandonment value of an asset. The initial criterion proposed abandonment in
the first year that abandonment value exceeded the present value of continued use
of the asset. Later studies pointed out the necessity of consideration of abandonment
possibilities in future years to obtain the maximum present value from the selection
of the optimal time for abandonment. Finally, we saw that temporary shutdown is
often a viable alternative to outright abandonment, and that option pricing may
also be applied to the option to delay or accelerate investment and to research and
development problems.
PROBLEM SET 431
PROBLEM SET
12.1 Which of the following types of uncertainty are inadmissible in a multiperiod model if
we are using the CAPM? Why?
a) COV(g„ d) COV(f7,_ , _)
b) rft e) [E(k„,,) — r f dlo-2(k- „„)
c) E/-1(X1)
12.2 The Ramsden Company is installing 10 new forklift trucks. Electric trucks cost $8000
each, whereas gas-powered trucks cost $5000 each. The operating costs for the electric trucks
would be $5200 per truck per year compared with $6000 per gas truck for the eight-year ex-
pected life of each vehicle. Expected salvage value is zero and straight-line depreciation is to
be used. Ramsden will apply a 10% discount factor for analysis of the electric trucks and a 2%
differential to the gas trucks owing to their higher operating risks. The applicable tax rate is
40%. Should electric or gas forklift trucks be purchased by Ramsden?
12.3 You are asked to perform a capital budgeting analysis of two projects. Both will require
an immediate cash outlay of $1000. Both projects last one year and they both produce revenues
at the end of the year amounting to $1500 with certainty. Cash outflows at the end of the year,
however, are risky. They are given below, along with the market rate of return, Rm :
End-of-Period Outflows
Since you are given the cash flows, there is no need to worry about taxes, depreciation, or
salvage value. Note that the cash outflows of Project B have higher variance than those of
Project A. Which project has greater NPV? Show your work and explain your reasoning.
12.4 Your firm is trying to choose between two mutually exclusive projects. Both cost $10,000
and have a five-year life and no salvage value. The company uses straight-line depreciation and
the corporate tax rate is 40%. Over the life of the project the annual expected rate of return
on the market portfolio is 15% and the risk-free rate is 5%. The first project has expected
revenues of $5,000 per year with an adjusted 13„v of 1.5 and expected variable costs of $2,333
per year with an adjusted f3vc of — 1.4 (i.e., both the revenues and variable costs are positively
correlated with the market). The second project has expected revenues of $6,067 per year with
an adjusted fiRev of 1.3 and expected variable costs of $2,400 with an adjusted 13v, of .3 (i.e.,
revenues are positively correlated with the market, and variable costs are negatively correlated
with the market). Which project should the firm accept? [Note: Adjusted fls are discussed in
section E of Chapter 12.]
12.5 Figure Q12.5 gives the cash flows, CF, for a two-period project that requires a $300 initial
outlay. Also shown are the state-contingent risk-free rates of return, r f , and the state-contingent
prices of a comparison risky asset. What is the NPV of the project?
432 CAPITAL BUDGETING UNDER UNCERTAINTY: THE MULTIPERIOD CASE
CF = $190, S = $25
State C
S = $22, CF = $180,
r f = 18%
State A
State D
CF = $160, S = $20
S = $20, r f = 15%
CF = $160, S = $22
State E
S = $20, CF = $150,
r f =10%
State B State F
CF= $130,S= $18
Figure Q12.5
12.6 The following investment decision is being considered by Citrus Farms. For $7000 the
company can acquire ownership of 10 acres of 15-year-old orange trees and a 15-year lease on
the land. The productive life of an orange tree is divided into stages as follows:
There is a market for decorative orange trees. Suppliers will buy trees and remove them accord-
ing to a schedule based on the age of the tree. Expected prices that can be obtained for the
10-acre harvest are: $9,000 at end of age 20, $12,000 at age 25, and $8,000 at age 30. Citrus
Farms has a 10% cost of capital.
a) What is the present value of each alternative? Since the land and anything on it will belong
to the lessor in 15 years, assume that once the trees are harvested the land will not be
replanted by Citrus.
b) As an alternative to this investment, Citrus can use the $7,000 to buy a new orange-sorting
machine. The machine would reduce sorting expenses by $1300 a year for 15 years. Which
investment would you make? Why? Assume that all other investment opportunities for
the next 15 years will earn the cost of capital.
c) In the tenth year Citrus discovers that everyone else with 25-year-old trees has sold them.
As a consequence the price the firm can get for the trees is only $8000. Since so many
trees have been sold for decoration, small orange crops are expected for the next 5 years.
As a result the price of oranges will be higher. Your acreage will yield $1200 a year. The
selling price of your trees in another 5 years is expected to be still depressed to $6000. What
should you do?
PROW EM SET 433
d) Given the situation in (c), what was the NPV of your actual investment over the 15-year
period?
e) What would the NPV be if the trees had been sold in year 10 for $8000?
12.7 Southern Electric Power Company'
The State Public Service Commission (PSC) is charged with the responsibility of regulating
the two large public utility companies in the state. The PSC is composed of seven persons;
four elected, one gubernatorial appointee, and one representative each from the two public
utilities.
The PSC is currently considering a capacity expansion program submitted by Southern
Electric Power (SE). SE operates three power plants, as shown in Table Q12.7a. The population
in the area served by SE has been growing 2% per year. This trend is expected to continue for
at least the next 10 years. Industrial growth is even more rapid about 4% per year. As a result,
demand for electricity is growing at approximately 3% per year. Peak-load demand is very
near SE's capacity, so it is necessary to begin some expansion within the year.
Percentage of
Date Current Power Remaining
Type of Plant Constructed Needs Supplied Life, in Years
Coal 1946 25 30
Hydroelectric 1958 10 35
Natural gas generating 1963 65 45
Another problem the company is confronted with is the growing shortage of natural gas.
Last year SE was unable to buy enough natural gas to meet the full electric power needs of
the area. The company was able to buy electricity generated by a company in an adjoining
state that has coal, but there is no assurance that this source of power can be relied on in
the future.
Given the uncertainty of future natural gas supplies, and the fact that hydroelectric power
is already being fully utilized, SE feels the only sensible course is to expand its own coal-
generating capacity. This would be especially desirable since it would use the large soft-coal
deposits in the northern part of the state.
There are two ways to increase the capacity for coal generation. For $40 million it would
be possible to double the generating capacity of the existing plant. This would fully satisfy the
growing demand for electricity for the next 6 to 8 years, after which time additional capacity
would be needed. The expansion would have a useful life of 30 years. The existing plant is
located 20 miles from the center of the largest city in the state. In 1946, when the plant was
originally built, the location was selected specifically to be at a considerable distance from any
populated area so that the smoke produced by burning coal would not be an environmental
nuisance.
12
This is really a short case, rather than a problem in the usual sense of the word. It does not necessarily
have a simple, cut-and-dried solution.
434 CAPITAL BUDGETING UNDER UNCERTAINTY: THE MULTIPERIOD CASE
In 30 years the city has grown considerably. Heavily populated suburbs are now located
within four miles of the plant, and pollution created by SE is a growing political issue. The
company believes it probable that within the next 5 years political pressure will require that
smoke scrubbers be installed at the existing plant (Table Q12.7b). Scrubbers can be installed
any time at the existing plant for $4 million If plant capacity is doubled and scrubbers are
installed at that time, the total cost would be $5.5 million. (This would be in addition to the
base cost of the new plant discussed below.) If the installation of scrubbers is postponed, and
carried out as a separate capital investment, the total cost is estimated to be $8 million, subject
to the same probability that installation might never occur (see Table Q12.7b).
In year 2 0%
In year 3 10
In year 4 50
In year 5 40
The alternative to expanding the capacity of the old plant is to build an entirely new plant.
Such a plant would cost $110 million and would have a maximum capacity of three times the
existing coal plant. With the new capacity in addition to the existing plant, energy needs could
be met for the next 18 to 20 years. The useful life of the plant would be 40 years. If desired,
the capacity of the new plant could be increased to the point that the old plant could be aban-
doned and capacity would still be adequate for the same period. The additional cost of the
extra capacity would be $22,082,000, which would be depreciated over the full life of the plant.
If the old plant is abandoned, equipment worth $7 million could be used in the new plant, and
sale of the land would provide additional capital of $3 million.
Annual operating costs of the old plant are $2 million. If its capacity is doubled, operating
costs will rise by $1.5 million Cost of operating the new plant will be $4 million per year. If the
old plant is abandoned, operating costs of the new plant will increase by $450,000 per year for
the remaining life of the old plant. SE uses straight-line depreciation to zero salvage value on
all capital investments discussed here. The cost of scrubbers is amortized over the remaining
life of the plant in which they are installed. Book value of the existing coal plant is $10 million.
SE has an effective tax rate of 40%. SE's weighted average cost of capital is 10%. Revenues
and other costs will be the same under either alternative. Excess capacity can be sold outside
the state at the same rate as within the state.
1. As a representative of the utility company, you are concerned with maximizing the present
value of the project. Analyze the alternatives and indicate your recommendation.
2. Keeping in mind that you are working only with costs, how would you adjust the discount
rate to account for: (a) the extra uncertainty for any alternative that includes delayed
installation of smoke scrubbers; (b) the pollution that would result from operating the old
plant without smoke scrubbers?
3. How would these factors affect your recommendation? As an elected member of the PSC,
what factors would you include in your analysis of the project? How would each of these
factors bear on your decision?
REFERENCES 43P
REFERENCES
Beedles, W., "Evaluating Negative Benefits," Jou]nal of Financial and Quantitative Analysis,
March =978, =73-=76.
Bogue, M. C., and R. R. Roll, "Capital Budgeting of Risky Projects with 'Imperfect' Markets
for Physical Capital," Jou]nal of Finance, May =974, 60=-6=3.
Booth, L., "Correct Procedures for the Evaluation of Risky Cash Outflows," Jou]nal of Finan-
cial and Quantitative Analysis, June =98<, <87-300.
Breeden, D., "An Intertemporal Asset Pricing Model with Stochastic Investment Opportuni-
ties," Jou]nal of Financial Economics, Septemîer =979, <6P-<96.
Brennan, M., "An Approach to the Valuation of Uncertain Income Streams," Jou]nal of
Finance, June =973, 66=-674.
, and E. Schwartz, "Evaluating Natural Resource Investments," Jou]nal of Business,
April =98P, =3P-=P7.
Celec, S., and R. Pettway, "Some Oîservations on Risk-Adjusted Discount Rates: A Comment,"
Jou]nal of Finance, Septemîer =979, =06=-=063.
Constantinides, G., "Market Risk Adjustment in Project Valuation," Jou]nal of Finance, May
=978, 603-6=6.
, "Admissiîle Uncertainty in the Intertemporal Asset Pricing Model," Jou]nal of Finan-
cial Economics, March =980, 7=-86.
Dyl, E. A., and H. W. Long, "Aîandonment Value and Capital Budgeting: Comment," Jou]nal
of Finance, March =969, 88-9P.
Fama, E. F., "Multiperiod Consumption-Investment Decisions," Ame]ican Economic Review,
March =970, =63-=74.
, "Risk-Adjusted Discount Rates and Capital Budgeting under Uncertainty," Jou]nal of
Financial Economics, August =977, 3-<4.
, and J. D. MacBeth, "Tests of the Multiperiod Two-Parameter Model," Jou]nal of
Financial Economics, May =974, 43-66.
Gehr, A., "Risk-Adjusted Capital Budgeting Usin Arîitrage," Financial Management, Winter,
=98=, =4-=9.
Joy, 0. M., "Aîandonment Values and Aîandonment Decisions: A Clarification," Jou]nal of
Finance, Septemîer =976, =<<P-=<<8.
Lewellen, W., "Some Oîservations on Risk-Adjusted Discount Rates," Jou]nal of Finance,
Septemîer =977, =33=-=337.
, "Reply to Pettway and Celec," Jou]nal of Finance, Septemîer =979, =06P-=066.
Long, J. B., Jr., "Stock Prices, Inflation and the Term Structure of Interest Rates," Jou]nal
of Financial Economics, July =974, =3=-=70.
McDonald, R., and D. Siegel, "The Value of Waiting to Invest," Qua]te]ly Jou]nal of Economics,
Novemîer =986, 707-7<7.
Majd, S., and R. Pindyck, "Time to Build, Option Value, and Investment Decisions,"
Jou]nal of Financial Economics, March =987, 7-<8.
Mason, S. P., and R. C. Merton, "The Role of Contingent Claims Analysis in Corporate Fi-
nance," in Recent Advances in Co]po]ate Finance, E. I. Altman, and M. G. Suîrahmanyam,
eds., Richard D. Irwin, Homewood, Illinois, =98P, 7-P4.
<Ç6 CAPITAL BUDGETING UNDER UNCERTAINTY: THE MULTIPERIOD CASE
Merton, R. C., "An Intertemporal Capital Asset Pricing Model," Economet]ica, Septemîer =973,
867-887.
Miles, J., and D. Choi, "Comment: Evaluating Negative Benefits," Jou]nal of Financial and
Quantitative Analysis, Decemîer =979, =09P-=099.
Miller, M., and C. Upton, "A Test of the Hotelling Valuation Principle," Jou]nal of Political
Economy, Feîruary =98P, =-<P.
Myers, S. C., "Procedures for Capital Budgeting under Uncertainty," Indust]ial Management
Review, Spring =968, =-<0.
, and S. Majd, "Calculating Aîandonment Value Using Option Pricing Theory," Work-
ing Paper, Sloan School of Management, MIT, =983.
Myers, S. C., and S. M. Turnîull, "Capital Budgeting and the Capital Asset Pricing Model:
Good News and Bad News," Jou]nal of Finance, May =977, 3<=-33<.
Paddock, J., D. Siegel, and J. Smith, "Option Valuation of Claims and Physical Assets: The
Case of Offshore Petroleum Leases," working paper, MIT Energy Laîoratory, Camîridge
Mass., =984. Septemîer =973, 867-887.
Roîerts, K., and M. Weitzman, "Funding Criteria for Research, Development and Exploration
Projects," Economet]ica, Septemîer =98=, =<6=-=<88.
Roîichek, A., and S. Myers, "Conceptual Proîlems in the Use of Risk-Adjusted Discount
Rates," Jou]nal of Finance, Decemîer =966, 7<7-730.
Roîichek, A., and J. C. Van Home, "Aîandonment Value and Capital Budgeting," Jou]nal
of Finance, Decemîer =967, P77-P90.
Ross, S., "A Simple Approach to the Valuation of Risky Streams," Jou]nal of Business, July
=979, <P4-<86.
Stapleton, R., and M. Suîrahmanyam, "A Multiperiod Equiliîrium Asset Pricing Model,"
Economet]ica, Septemîer =978, =077-=096.
, "Multiperiod Equiliîrium: Some Implications for Capital Budgeting," TIMS Studies
in the Management Sciences, Vol. ==, =979, <33-<48.
Weitzman, M., W. Newey, and M. Roîin, "Sequential R & D Strategy for Synfuels," Bell
Jou]nal of Economics, Autumn =98=, P74-P90.
13
The average cost of capital to any firm is completely independent of
its capital structure and is equal to the capitalization rate of a pure
equity stream of its class.
F. Modigliani and M. H. Miller, "The Cost of Capital, Corporation
Finance, and the Theory of Investment," American Economic Review,
June 1958, 268
Funds for investment are provided to the firm by investors who hold various types
of claims on the firm's cash flows. Debt holders have contracts (bonds) that promise
to pay them fixed schedules of interest in the future in exchange for their cash now.
Equity holders provide retained earnings (internal equity provided by existing share-
holders) or purchase new shares (external equity provided by new shareholders). They
do so in return for claims on the residual earnings of the firm in the future. Also,
shareholders retain control of the investment decision, whereas bondholders have no
direct control except for various types of indenture provisions in the bond that may
constrain the decision making of shareholders. In addition to these two basic cate-
gories of claimants, there are others such as holders of convertible debentures, leases,
preferred stock, nonvoting stock, and warrants.
Each investor category is confronted with a different type of risk, and therefore
each requires a different expected rate of return in order to provide funds to the firm.
The required rate of return is the opportunity cost to the investor of investing scarce
resources elsewhere in opportunities with equivalent risk. As we shall see, the fact
that shareholders are the ones who decide whether to accept or reject new projects
is critical to understanding the cost of capital. They will accept only those projects
that increase their expected utility of wealth. Each project must earn, on a risk-adjusted
437
438 CAPITAL STRUCTURE AND THE COST OF CAPITAL: THEORY
E(Ri) Marginal
cost of
capital
Investment dollars
Marginal
efficiency of
investment
Figure 13.1
Demand and supply of investment for projects of equal risk.
basis, enough net cash flow to pay investors (bondholders and shareholders) their
expected rates of return, to pay the principal amount that they originally provided,
and to have something left over that will increase the wealth of existing shareholders.
The cost of capital is the minimum risk-adjusted rate of return that a project must
earn in order to be acceptable to shareholders.
The investment decision cannot be made without knowledge of the cost of capital.
Consequently, many textbooks introduce the concept of the cost of capital before
they discuss investment decisions. It probably does not matter which topic comes
first. Both topics are important and they are interrelated. Figure 13.1 shows the in-
vestment decision as the intersection of the demand and supply of investment capital.
All projects are assumed to have equivalent risk. Also, fund sources have equal risk
(in other words, in Fig. 13.1 we make no distinction between equity and debt).
Chapters 2, 3, and 12 discussed the ranking of projects assuming that the appropriate
cost of capital was known. The schedule of projects with their rates of return is some-
times called the marginal efficiency of investment schedule and is shown as the demand
curve in Fig. 13.1. The supply of capital, represented as the marginal cost of capital
curve, is assumed to be infinitely elastic. Implicitly, the projects are assumed to have
equal risk. Therefore the firm faces an infinite supply of capital at the rate E(R;)
because it is assumed that the projects it offers are only a small portion of all
investment in the economy. They affect neither the total risk of the economy nor
the total supply of capital. The optimal amount of investment for the firm is /1, and
the marginally acceptable project must earn at least E(R1). All acceptable projects, of
course, earn more than the marginal cost of capital.
Figure 13.1 is an oversimplified explanation of the relationship between the
cost of capital and the amount of investment. However, it demonstrates the inter-
relatedness of the two concepts. For a given schedule of investments a rise in the
THE VALUE OF THE FIRM GIVEN CORPORATE TAXES ONLY 439
cost of capital will result in less investment. This chapter shows how the firm's mix
of debt and equity financing affects the cost of capital, explains how the cost of
capital is related to shareholders' wealth, and shows how to extend the cost of capital
concept to the situation where projects do not all have the same risk. If the cost of
capital can be minimized via some judicious mixture of debt and equity financing,
then the financing decision can maximize the value of the firm.
Whether or not an optimal capital structure exists is one of the most important
issues in corporate finance—and one of the most complex. This chapter begins the
discussion. It covers the effect of tax-deductible debt on the value of the firm, first
in a world with only corporate taxes, then by adding personal taxes as well. There
is also a discussion of the effect of risky debt, warrants, convertible bonds, and call-
able bonds.
Chapter 14 continues the discussion of optimal capital structure by introducing
the effects of factors other than taxes. Bankruptcy costs, option pricing effects, agency
costs, and signaling theory are all discussed along with empirical evidence bearing
on their validity. Also, the optimal maturity structure of debt is presented. Corporate
financial officers must decide not only how much debt to carry but also its duration.
Should it be short-term or long-term debt?
It goes without saying that many of these assumptions are unrealistic, but later
we can show that relaxing many of them does not really change the major conclusions
of the model of firm behavior that Modigliani and Miller provide. Relaxing the
440 CAPITAL STRUCTURE AND THE COST OF CAPITAL: THEORY
assumption that corporate debt is risk free will not change the results (see section
D). However, the assumptions of no bankruptcy costs (relaxed in Chapter 14) and
no personal taxes (relaxed in section B of this chapter) are critical because they change
the implications of the model. The last two assumptions rule out signaling behavior—
because insiders and outsiders have the same information; and agency costs--because
managers never seek to maximize their own wealth. These issues are discussed in
Chapter 14.
One of the assumptions requires greater clarification. What is meant when we
say that all firms have the same risk class? The implication is that the expected risky
future net operating cash flows vary by, at most, a scale factor. Mathematically this
is
CF, = 2CFJ,
where
CF = the risky net cash flow from operations (cash flow before interest and taxes),
2 = a constant scale factor.
This implies that the expected future cash flows from the two firms (or projects) are
perfectly correlated.
If, instead of focusing on the level of cash flow, we focus on the returns, the perfect
correlation becomes obvious because the returns are identical, as shown below:
— CFi,t-i ,
CFi,t -1
and because CF,,, = i1CF,,„ we have
-~CF J , t =
K ,t
Therefore if two streams of cash flow differ by, at most, a scale factor, they will have
the same distributions of returns, the same risk, and will require the same expected
return.
Suppose the assets of a firm return the same distribution of net operating cash
flows each time period for an infinite number of time periods. This is a no-growth
situation because the average cash flow does not change over time. We can value
this after-tax stream of cash flows by discounting its expected value at the appropriate
risk-adjusted rate. The value of an unlevered firm, i.e., a firm with no debt, will be
E(FCF)
= (13.1)
p
where
V U = the present value of an unlevered firm (i.e., all equity),
E(FCF) = the perpetual free cash flow after taxes (to be explained in detail below),
p = the discount rate for an all-equity firm of equivalent risk.
THE VALUE OF THE FIRM GIVEN CORPORATE TAXES ONLY 441
This is the value of an unlevered firm because it represents the discounted value of
a perpetual, nongrowing stream of free cash flows after taxes that would accrue to
shareholders if the firm had no debt. To clarify this point, let us look at the following
pro forma income statement:
Rev Revenues
—VC Variable costs of operations
— FCC Fixed cash costs (e.g., administrative costs and real estate taxes)
—dep Noncash charges (e.g., depreciation and deferred taxes)
NOI Net operating income
—1(,,D Interest on debt (interest rate, times principal, D)
EBT Earnings before taxes
—T Taxes = Tc(EBT), where T e is the corporate tax rate
NI Net income
It is extremely important to distinguish between cash flows and the accounting defi-
nition of profit. After-tax cash flows from operations may be calculated as follows.
Net operating income less taxes is
ti
NOI — Tc(NOI).
Rewriting this using the fact that NOI = Rev — VC — FCC — dep, we have
ti
(Rev — VC — FCC — dep)(1 — Te).
This is operating income after taxes, but it is not yet a cash flow definition because
a portion of total fixed costs are noncash expenses such as depreciation and deferred
taxes. Total fixed costs are partitioned in two parts: FCC is the cash-fixed costs, and
dep is the noncash-fixed costs.
To convert after-tax operating income into cash flows, we must add back depre-
ciation and other noncash expenses. Doing this, we have
(Rev — VC — FCC — dep)(1 — -c c ) + dep.
Finally, by assumption, we know that the firm has no growth; i.e., all cash flows
are perpetuities. This implies that depreciation each year must be replaced by invest-
ment in order to keep the same amount of capital in place. Therefore dep = I, and
the after-tax free cash flow available for payment to creditors and shareholders is
FCF = (Rev — VC — FCC — dep)(1 — -c c ) + dep — I,
ti
FCF = (Rev — VC — FCC — dep)(1 — 'r e) since dep = I.
The interesting result is that when all cash flows are assumed to be perpetuities, free
cash flow (FCF) is the same thing as net operating income after taxes, i.e., the cash
flow that the firm would have available if it had no debt at all. This is shown below:
NOM — T c ) = FCF = (Rev — VC — FCC — dep)(1 —
442 CAPITAL STRUCTURE AND THE COST OF CAPITAL: THEORY
Note also that this approach to cash flows is exactly the same as that used to define
cash flows for capital budgeting purposes in Chapter 2. The reader should keep in
mind that in order to determine the value of the firm correctly, the definition of cash
flows and the definition of the discount rate, i.e., the weighted average cost of capital,
must be consistent. The material that follows will prove that they are.
Given perpetual cash flows, Eq. (13.1), the value of an unlevered firm, can be
written in either of two equivalent ways:1
E(FCF) E(NOI)(1 — r e)
Vu = or vu = • (13.1)
From this point forward we shall use the net operating income definition of cash
flows in order to be consistent with the language originally employed by Modigliani
and Miller.
Next assume that the firm issues debt. The after-tax cash flows must be split up
between debt holders and shareholders. Shareholders receive NI + dep — I, net cash
flows after interest, taxes, and replacement investment; and bondholders receive
interest on debt, kd D. Mathematically, this is equivalent to total cash flow available
for payment to the private sector:2
ti
NI + dep — / + k d D = (Rev — VC — FCC — dep — kd D)(1 — T e) + dep — / + k d D.
Given that dep = I, for a nongrowing firm we can rearrange terms to obtain
NI + k d D = (Rev — VC — FCC — dep)(1 — r e) — k d D + k d DT C + k d D
= NOI(1 — re) + k d Dre. (13.2)
The first part of this stream, NOI(1 — r e), is exactly the same as the cash flows for
the unlevered firm [the numerator of (13.1)], with exactly the same risk. Therefore
recalling that it is a perpetual stream, we can discount it at the rate appropriate for
an unlevered firm, p. The second part of the stream, k d DT C , is assumed to be risk free.
Therefore we shall discount it at the before-tax cost of risk-free debt, k b . Consequently,
the value of the levered firm is the sum of the discounted value of the two types of cash
flow that it provides:
E(NOI)(1 — r e) kdDr
(13.3)
kb
Note that k d D is the perpetual stream of risk-free payments to bondholders and that
k b is the current before-tax market-required rate of return for the risk-free stream.
Therefore since the stream is perpetual, the market value of the bonds, B, is
B = k d D/k b . (13.4)
The present value of any constant perpetual stream of cash flows is simply the cash flow divided by the
discount rate. See Appendix A at the end of the book, Eq. (A.5).
The government receives all cash flows not included in Eq. (13.2); i.e., the government receives taxes
(also a risky cash flow).
THE VALUE OF THE FIRM GIVEN CORPORATE TAXES ONLY 443
The value of the levered firm, V'', is equal to the value of an unlevered firm, V u, plus
the present value of the tax shield provided by debt, T,13. Later on we shall refer
to the "extra" value created by the interest tax shield on debt as the gain from le-
verage. This is perhaps the single most important result in the theory of corporation
finance obtained in the last 30 years. It says that in the absence of any market
imperfections including corporate taxes (i.e., if -r, = 0), the value of the firm is com-
pletely independent of the type of financing used for its projects. Without taxes, we
have
vL = if = 0. (13.5a)
5%) and buying Company A. In effect, we create homemade leverage in the following
way:
1. We sell our stock in B (say we own 1%, then we sell $772.72).
2. We borrow an amount equivalent to 1% of the debt in B, i.e., $300 at 5% interest.
3. We buy 1% of the shares in A.
Before arbitrage we held 1% of the equity of B and earned 11% on it, i.e.,
.11($772.72) = $85.00. After arbitrage we hold the following position:
1 % of A's equity and earn 10%, i.e., .10($1000.00) = $100.00
pay interest on $300 of debt, i.e., .05($300) = — 15.00
85.00
This gives us the same income as our levered position in Company B, but the amount
of money we have available is $772.72 (from selling shares in B) plus $300 (from
borrowing). So far, in the above calculation, we have used only $1000.00 to buy shares
of A. Therefore we can invest another $72.72 in shares of A and earn 10%. This
brings our total income up to $85 + $7.27 = $92.27, and we own $772.72 of net worth
of equity in A (the bank "owns" $300). Therefore our return on equity is 11.94% (i.e.,
$92.27/$772.72). Furthermore, our personal leverage is the $300 in debt divided by
the equity in A, $772.72. This is exactly the same leverage and therefore the same
risk as we started with when we had an equity investment in B.
The upshot of the foregoing arbitrage argument is that we can use homemade
leverage to invest in A. We earn a higher rate of return on equity without changing
our risk at all. Consequently we will undertake the arbitrage operation by selling
shares in B, borrowing, and buying shares in A. We will continue to do so until the
market values of the two firms are identical. Therefore Modigliani-Miller Proposition
I is a simple arbitrage argument. In a world without taxes the market values of the
levered and unlevered firms must be identical.
However, as shown by Eq. (13.5), when the government "subsidizes" interest pay-
ments to providers of debt capital by allowing the corporation to deduct interest
payments on debt as an expense, the market value of the corporation can increase as
it takes on more and more (risk-free) debt. Ideally (given the assumptions of the
model) the firm should take on 100% debt.'
We shall see later in this chapter that this result is modified when we consider a world with both
corporate and personal taxes, or one where bankruptcy costs are nontrivial. Also, the Internal Revenue
Service will disallow the tax deductibility of interest charges on debt if, in its judgment, the firm is using
excessive debt financing as a tax shield.
THE VALUE OF THE FIRM GIVEN CORPORATE TAXES ONLY 445
that original shareholders' wealth increase. From Eq. (13.3) we see that the change
in the value of the levered firm, AV L, with respect to a new investment, Al, is'
AVL (1 — T c ) AE(NOI) AB
+I (13.6)
Al Al ' Al
L
If we take the new project, the change in the value of the firm, AV , will also be
equal to the change in the value of original shareholders' wealth, AS°, plus the new
equity required for the project, ASH, plus the change in the value of bonds outstanding,
AB°, plus new bonds issued, AB":
AVL = AS° + AS + AB° + AB". (13.7a)
Because the old bondholders hold a contract that promises fixed payments of interest
and principal, because the new project is assumed to be no riskier than those already
outstanding, and especially because both old and new debt are assumed to be risk
free, the change in the value of outstanding debt is zero (AB° = 0). Furthermore, the
new project must be financed with either new debt, new equity, or both. This implies
that'
Al = AS + AB". (13.8)
Using this fact (13.7b) can be rewritten as
AVL AS° AS + AB"_ AS°
AI + 1. (13.9)
A/ A/ Al
which is equivalent to the requirement that AVL /AI > 1. Note that the requirement
that the change in original shareholders' wealth be positive, i.e., AS°/AI > 0, is a be-
havioral assumption imposed by Modigliani and Miller. They were assuming (1) that
managers always do exactly what shareholders wish and (2) that managers and share-
holders always have the same information. The behavioral assumptions of Eq. (13.10)
are essential for what follows.
5
Note that T, and p do not change with Al. The cost of equity for an all-equity firm does not change
because new projects are assumed to have the same risk as the old ones.
6 Note that (13.8) does not require new issues of debt or equity to be positive. It is conceivable, e.g., that
the firm might issue $4000 in stock for a $1000 project and repurchase $3000 in debt.
446 CAPITAL STRUCTURE AND THE COST OF CAPITAL: THEORY
When the assumptions of inequality [(13.10)] are imposed on Eq. (13.6) we are
able to determine the cost of capital'
A V L (1 — t) AE(NOI) AB"
Al — AI AI
or, by rearranging terms, we have
(1 — i c ) AE(NOI) AB
P
Al (1 T
AI ).
The left-hand side of (13.11) is the after-tax change in net operating cash flows brought
about by the new investment, i.e., the after-tax return on the project.' The right-hand
side is the opportunity cost of capital applicable to the project. As long as the antic-
ipated rate of return on investment is greater than the cost of capital, current share-
holders' wealth will increase.
Note that if the corporate tax rate is zero, the cost of capital is independent of
capital structure (the ratio of debt to total assets). This result is consistent with Eq.
(13.5a), which says that the value of the firm is independent of capital structure. On
the other hand, if corporate taxes are paid, the cost of capital declines steadily as the
proportion of new investment financed with debt increases. The value of the levered
firm reaches a maximum when there is 100% debt financing (so long as all of the debt
is risk free).
Two questions arise in the interpretation of the leverage ratio, AB/AI. First, is the
leverage ratio marginal or average? Modigliani and Miller, in the above quote, set
the marginal ratio equal to the average by assuming the firm sets a long-run target
ratio, which is constant. Even if this is the case, we still must consider a second issue,
namely: Is the ratio to be measured as book value leverage, replacement value leverage,
or reproduction value leverage? The last two definitions, as we shall see, are both mar-
ket values. At least one of these three measures, book value leverage, can be ruled out
immediately as being meaningless. In particular, there is no relationship whatsoever
between book value concepts, (e.g., retained earnings) and the economic value of
equity.
The remaining two interpretations, replacement and reproduction value, make
sense because they are both market value definitions. By replacement value, we mean
the economic cost of putting a project in place. For capital projects a large part of
this cost is usually the cost of purchasing plant, equipment, and working capital. In
the Modigliani-Miller formulation, replacement cost is the market value of the invest-
ment in the project under consideration, Al. It is the denominator on both sides of the
cost of capital inequality (13.11). On the other hand, reproduction value, A V, is the
total present value of the stream of goods and services expected from the project.
The two concepts can be compared by noting that the difference between them is the
NPV of the project, that is,
NPV = A V — Al.
For a marginal project, where NPV = 0, replacement cost and reproduction value
are equal.
Haley and Schall [1973, 306-311] introduce an alternative cost of capital defini-
tion where the "target" leverage is the ratio of debt to reproduction value, as shown
below:
If the firm uses a reproduction value concept for its "target" leverage, it will seek to
maintain a constant ratio of the market value of debt to the market value of the firm.
With the foregoing as background, we can now reconcile the apparent conflict in
the measurement of leverage applicable to the determination of the relevant cost of
capital for a new investment project. Modigliani and Miller define the target L* as
the average, in the long run, of the debt-to-value ratio or B*/V*. Then regardless
of how a particular investment is financed, the relevant leverage ratio is dB/dV. For
example, a particular investment may be financed entirely by debt. But the cost of that
particular increment of debt is not the relevant cost of capital for that investment.
The debt would require an equity base. How much equity? This is answered by the
long-run target B* IV*. So procedurally, we start with the actual amount of invest-
ment increment for the particular investment, dI. The L* ratio then defines the amount
of dB assigned to the investment. If the NPV from the investment is positive, then dV
will be greater than dI. Hence the debt capacity of the firm will have been increased
by more than dB. However, the relevant leverage for estimating the WACC will still
be dB/dV, which will be equal to B* IV*. We emphasize that the latter is a policy target
448 CAPITAL STRUCTURE AND THE COST OF CAPITAL: THEORY
= 9000 + .4(5625)
= 11250.
Clearly, the value of the firm is higher if we use the reproduction value definition of
leverage. But as a practical matter, what bank would lend $5625 on a project that
has a $1000 replacement value of assets? If the bank and the firm have homogeneous
expectations this is possible. If they do not, then it is likely that the firm is more opti-
mistic than the bank about the project. In the case of heterogeneous expectations
there is no clear solution to the problem. Hence we favor the original argument of
Modigliani and Miller that the long-run target debt-to-value ratio will be close to
dB/dI; i.e., use the replacement value definition.
deît in section D.) The cost of equity capital is the change in the return to equity
holders with respect to the change in their investment, AS° + AS". The return to
equity holders is the net cash flow after interest and taxes, NI. Therefore their rate
of return is ANI/(AS° + AS"). To solve for this, we îegin with identity (=3.<),
NI + kd D = NOI(= — T c ) kdftr e.
Next we divide îy AI, the new investment, and oîtain
ANI A(kdD) A(kd D) ANOI
— (= Tc) (=3.=4)
Al Al A/ A/
Suîstituting the left-hand side of (=3.=4) into (=3.6), we get
4=7=' ANI/A/ + (= — i ° )4(kd D)/4/ AB
+T (=3.=P)
Al . A/
The change in new equity plus old equity equals the change in the total equity of
the firm (AS = AS° + AS"). Therefore the cost of equity, lc, = ANI/AS, is written
AB
ks = p + (1 ÿ t c )(p ÿ kb ) As • (=3.=8)
The implication of Eq. (=3.=8) is that the opportunity cost of capital to shareholders
increases linearly with changes in the market value ratio of deît to equity (assuming
450 CAPITAL STRUCTURE AND THE COST OF CAPITAL: THEORY
that AB/AS = B/S). If the firm has no debt in its capital structure, the levered cost of
equity capital, k„ is equal to the cost of equity for an all-equity firm, p.
(a) (b)
Figure 13.2
The cost of capital as a function of the ratio of debt to equity; (a) assuming 2, = 0;
(b) assuming T, > 0.
THE VALUE OF THE FIRM 451
We can see that this is the same as the Modigliani-Miller definition, Eq. (13.12) by
substituting (13.18) into (13.19) and assuming that B/S = AB/AS. This is done below:
B1
WACC =- (1 — i c )kb B + s + [p (1 -] c )(p kb )
B+S
B
— „ B S B S
= (1 B + s
T e )K b P (1 Tc)k,
B + S + (1 Tc)P S B + S SB+S
B + B B
= (1 ÿ tc)Kb +P (1 tc)Kb
B+S (B+S B+S B + S B+S
B
= p (1 "C c QED
B +S )
There is no inconsistency between the traditional definition and the M-M definition
of the cost of capital [Eqs. (13.12) and (13.19)]. They are identical.
G= ÿ = T e13. (13.20)
Miller [1977] modifies this result by introducing personal as well as corporate taxes
into the model. In addition to making the model more realistic, the revised approach
adds considerable insight into the effect of leverage on value in the real world. We
do not, after all, observe firms with 100% debt in their capital structure as the original
Modigliani-Miller model suggests.
Assume for the moment that there are only two types of personal tax rates: the
rate on income received from holding shares, and the rate on income from bonds,
Tps,
TpB •The expected after-tax stream of cash flows to shareholders of an all-equity firm
would be (NOI)(1 — t)(1 — -cps). By discounting this perpetual stream at the cost of
equity for an all-equity firm we have the value of the unlevered firm:
VU = • (13.21)
Alternatively, if the firm has both bonds and shares outstanding, the earnings stream
is partitioned into two parts. Cash flows to shareholders after corporate and personal
452 CAPITAL STRUCTURE AND THE COST OF CAPITAL: THEORY
taxes are:
Payments to shareholders = (NOI — k d D)(1 — -0(1 — 2 ps ),
where B = k d D(1 — pB)/kb , the market value of debt. Consequently, with the intro-
duction of personal taxes, the gain from leverage is the second term in (13.23):
TY1 TP
G =[1 (1 1B. (13.24)
(1 — Õ pB)
Note that when personal tax rates are set equal to zero, the gain from leverage in
(13.24) equals the gain from leverage in (13.20), the earlier result. This finding also
obtains when the personal tax rate on share income equals the rate on bond income.
In the United States it is reasonable to assume that the effective tax rate on common
stock is lower than that on bonds.' The implication is that the gain from leverage
when personal taxes are considered [Eq. (13.24)] is lower than T13 [Eq. (13.20)].
If the personal income tax on stocks is less than the tax on income from bonds,
then the before-tax return on bonds has to be high enough, other things being equal,
to offset this disadvantage. Otherwise no investor would want to hold bonds. While
it is true that owners of a levered corporation are subsidized by the interest deduct-
ibility of debt, this advantage is counterbalanced by the fact that the required interest
payments have already been "grossed up" by any differential that bondholders must
pay on their interest income. In this way the advantage of debt financing may be lost.
9
The tax rate on stock is thought of as being lower than that on bonds because of a relatively higher
capital gains component of return, and because capital gains are not taxed until the security is sold.
Capital gains taxes can, therefore, be deferred indefinitely.
THE VALUE OF THE FIRM 453
1
rp = demand = ro ,rpiB
Supply _ ro
—
ro
1 —T,
Dollar amount
of all bonds
Figure 13.3
Aggregate supply and demand for corporate bonds (before tax
rates).
individual with a marginal tax rate on income from bonds equal to will not hold
i
Tp B
corporate bonds until they pay r 0 /(1 — -r piB), i.e., until their return is "grossed up."
Since the personal income tax is progressive, the interest rate that is demanded has
to keep rising to attract investors in higher and higher tax brackets." The supply
of corporate bonds is perfectly elastic, and bonds must pay a rate of r 0 /(1 — r e) in
equilibrium. To see that this is true, let us recall that the personal tax rate on stock
is assumed to be zero 0) and rewrite the gain from leverage:
(Tps =
G
=(
1 (1 — Tc) 13.
(1 - T,B)
) (13.26)
If the rate of return on bonds supplied by corporations is I., = 1'0 /(1 — ic), then the
gain from leverage, in Eq. (13.26), will be zero. The supply rate of return equals the
1° Also, as will be shown in Chapter 16, it is possible to shield up to $10,000 in dividend income from taxes.
" Keep in mind the fact that the tax rate on income from stock is assumed to be zero. Therefore the
higher an individual's tax bracket becomes, the higher the before-tax rate on bonds must be in order for
the after-tax rate on bonds to equal the rate of return on stock (after adjusting for risk).
454 CAPITAL STRUCTURE AND THE COST OF CAPITAL: THEORY
Consequently,
( 1ÿ = (1 ÿ pB),
and the gain from leverage in (13.26) will equal zero. If the supply rate of return is
less than r 0 /(1 — r,), then the gain from leverage will be positive, and all corporations
will try to have a capital structure containing 100% debt. They will rush out to issue
new debt. On the other hand, if the supply rate of return is greater than r 0 /(1 — te ),
the gain from leverage will be negative and firms will take action to repay outstanding
debt. Thus we see that in equilibrium, taxable debt must be supplied to the point
where the before-tax cost of corporate debt must equal the rate that would be paid
by tax-free institutions grossed up by the corporate tax rate.
Miller's argument has important implications for capital structure. First, the gain
to leverage may be much smaller than previously thought. Consequently, optimal
capital structure may be explained by a tradeoff between a small gain to leverage
and relatively small costs such as expected bankruptcy costs. This tradeoff will be
discussed at greater length in Chapter 14. Second, the observed market equilibrium
interest rate is seen to be a before-tax rate that is "grossed up" so that most or all
of the interest rate tax shield is lost. Finally, Miller's theory implies there is an equi-
librium amount of aggregate debt outstanding in the economy that is determined by
relative corporate and personal tax rates.
Dollar amount
of all îonds
Figure 13.4
Aggregate debt equilibrium with heterogeneous corporate
and personal tax rates.
zero or negative earnings will increase, thereby causing the interest tax shield to
decline in expected value. They further show that if there are positive bankruptcy
costs, there will be an optimum tradeoff between the marginal expected benefit of
interest tax shields and the marginal expected cost of bankruptcy. This issue will be
further discussed in the next chapter.
The CAPM discussed in Chapter 7 provides a natural theory for the pricing of risk.
When combined with the cost of capital definitions derived by Modigliani and Miller
[1958, 1963], it provides a unified approach to the cost of capital. The work that we
shall describe was first published by Hamada [1969] and synthesized by Rubinstein
[1973].
The CAPM may be written as
E(R;) = R f
[E(Rm)
R f ][3 j , (13.27)
where
E(R J ) = the expected rate of return on asset j,
R f = the (constant) risk-free rate,
E(Rm ) = the expected rate of return on the market portfolio,
= COV(Ri , Rm)/VAR(R m ).
Recall that all securities fall exactly on the security market line, which is illustrated
in Fig. 13.5. We can use this fact to discuss the implications for the cost of debt, the
cost of equity, and the weighted average cost of capital; and for capital budgeting
when projects have different risk.
Figure 13.5 illustrates the difference between the original Modigliani-Miller cost
of capital and the CAPM. Modigliani and Miller assumed that all projects within
456 CAPITAL STRUCTURE AND THE COST OF CAPITAL: THEORY
R(R i
• B
E(R F „,, WACC (Firm)
RA
E(R A
PA 1é Finn
Figure 13.5
The security market line.
the firm had the same business or operating risk (mathematically, they assumed that
CF i = /ICE). This was expedient because in 1958, when the paper was written, there
was no accepted theory that allowed adjustments for differences in systematic risk.
Consequently, the Modigliani-Miller theory is represented by the horizontal line in
Fig. 13.5. The WACC for the firm (implicitly) does not change as a function of
systematic risk. This assumption, of course, must be modified because firms and
projects differ in risk.
default risk is completely diversifiable (fi b = 0). We shall temporarily maintain the
assumption that k b = R f , then relax it a little later in the chapter.
The M-M definition of the cost of equity for the unlevered firm was tautological,
i.e., p = p, because the concept of systematic risk had not yet been developed in 1958.
We now know that it depends on the systematic risk of the firm's after-tax operating
cash flows, /3,. Unfortunately for empirical work, the unlevered beta is not directly
observable. We can, however, easily estimate the levered equity beta, }é L. (This has
also been referred to as }é 5 elsewhere.) If there is a definable relationship between the
two betas, there are many practical implications (as we shall demonstrate with a
simple numerical example in the next section). To derive the relationship between
the levered and unlevered betas, begin by equating the M-M and CAPM definitions
of the cost of levered equity (line 3 in Table 13.2):
R f + [E(R,b —RAfl,=p+ (p — l3 h (} —T —
s.
R + [E(R bi — Rf]}é, = p + (p — R f (} — t
Then substitute into the right-hand side the CAPM definition of the cost of unlevered
equity, p:
Rf \E(R m) Rf]13| = R f + [E(Rm — R
fi = [ } + —
c S
H&J. (13.28)
The implication of Eq. (13.28) is that if we can observe the levered beta by using
observed rates of return on equity capital in the stock market, we can estimate the
unlevered risk of the firm's operating cash flows.
2. A Simple Example
The usefulness of the theoretical results can be demonstrated by considering the
following problem. The United Southern Construction Company currently has a
market value capital structure of 20% debt to total assets. The company's treasurer
believes that more debt can be taken on, up to a limit of 35% debt, without losing
the firm's ability to borrow at 7%, the prime rate (also assumed to be the risk-free
458 CAPITAL STRUCTURE AND THE COST OF CAPITAL: THEORY
rate). The firm has a marginal tax rate of 50%. The expected return on the market
next year is estimated to be 17%, and the systematic risk of the company's equity,
h, is estimated to be .5.
• What is the company's current weighted average cost of capital? Its current cost
of equity?
• What will the new weighted average cost of capital be if the "target" capital
structure is changed to 35% debt?
• Should a project with a 9.25% expected rate of return be accepted if its systematic
risk, 13,, is the same as that of the firm?
To calculate the company's current cost of equity capital we can use the CAPM:
ks = R f + [E(Rm) ÿ R f ]fl,
= .07 + [.17 — .07].5 = .12.
Therefore the weighted average cost of capital is
WACC = (1 T )R
c f B+S B+S
= (1 ÿ .5).07(.2) + .12(.8) = 10.3%.
The weighted average cost of capital with the new capital structure is shown in Fig.
13.6. =< Note that the cost of equity increases with increasing leverage. This simply
reflects the fact that shareholders face more risk with higher financial leverage and
that they require a higher return to compensate them for it. Therefore in order to
calculate the new weighted average cost of capital we have to use the Modigliani-
Miller definition to estimate the cost of equity for an all-equity firm:
B
WACC = (1 B+s ),
WACC .103
P= = 11.44%.
1 — T c [BAB + S)] 1 ÿ .5(.2)
As long as the firm does not change its business risk, its unlevered cost of equity
capital, p, will not change. Therefore we can use p to estimate the weighted average
cost of capital with the new capital structure:
WACC = .1144[1 — .5(.35)] = 9.438%.
Therefore, the new project with its 9.25% rate of return will not be acceptable even
if the firm increases its ratio of debt to total assets from 20% to 35%.
A common error made in this type of problem is to forget that the cost of equity
capital will increase with higher leverage. Had we estimated the weighted average
=<
Note that if deît to total assets is <0%, then deît to equity is <P%. Also, 3P% converts to P3.8P% in Fig. =3.6.
INTRODUCING RISK A SYNTHESIS OF M-M AND CAPM 459
12.64%
12%
(1 — T e )P
(1-7- ) R f
25% 53.85%
Figure 13.6
Changes in the cost of capital as leverage increases.
cost of capital, using 12% for the old cost of equity and 35% debt as the target capital
structure, we would have obtained 9.03% as the estimated weighted average cost of
capital and we would have accepted the project.
We can also use Eq. (13.28) to compute the unlevered beta for the firm. Before
the capital structure change the levered beta was )6', = .5; therefore
fiL = 1
+ ( 1 — rc) -§ flu,
.5 = [1 + (1 — .5)(.25)16 u ,.
fl u = .4444.
Note that the unlevered beta is consistent with the firm's unlevered cost of equity
capital. Using the CAPM, we have
p= R1 + [E(Rm) — Rf ]fl u
(31, = [ 1 + (1 — -cc ) -1 NU
= [1 + (1 — .5).5385].4444
= .5641.
460 CAPITAL STRUCTURE AND THE COST OF CAPITAL: THEORY
Hence the increase in leverage raises the levered equity beta from .5 to .5641, and
the cost of levered equity increases from 12% to 12.64%.
B
WACC = p(1
B+S
= .19[1 — .5(.2)] = 17.1%.
E(Ri )
E(Ri) = 19%
E(R m ) = 17%
Rf =7%
Figure 13.7
Using the CAPM to estimate the required rate of return
on a project.
INTRODUCING RISK A SYNTHESIS OF M-M AND CAPM 461
And if the leverage is increased to 35%, the required return falls to 15.675%:
WACC = .19[1 — .5(35)] = 15.675%.
Firms seek to find projects that earn more than the project's weighted average
cost of capital. Suppose that, for the sake of argument, the WACC of our firm is
17%. Project B in Fig. 13.7 earns 20%, more than the firm's WACC, whereas project
A in Fig. 13.7 earns only 15%, less than the firm's WACC. Does this mean that B
should be accepted while A is rejected? Obviously not, because they have different
risk (and possibly different optimal capital structures) than the firm as a whole. Proj-
ect B is much riskier and must therefore earn a higher rate of return than the firm.
In fact it must earn more than projects of equivalent risk. Since it falls below the
security market line, it should be rejected. Alternately, project A should be accepted
because its anticipated rate of return is higher than the rate that the market requires
for projects of equivalent risk. It lies above the security market line in Fig. 13.7.
The examples above serve to illustrate the usefulness of the risk-adjusted cost of
capital for capital budgeting purposes. Each project must be evaluated at a cost of
capital that reflects the systematic risk of its operating cash flows as well as the finan-
cial leverage appropriate for the project. Estimates of the correct opportunity cost
of capital are derived from a thorough understanding of the Modigliani-Miller cost
of capital and the CAPM.
the project has greater value as the financial leverage of the firm increases. As a
practical matter, this problem is usually "handled" by assuming that the firm
has decided on an "optimal" capital structure and that all projects are financed,
at the margin, with the optimal ratio of debt to equity. The relevant factors that
may be used to determine the optimal capital structure are discussed in the fol-
lowing chapter. However, assuming that an optimum does exist, and assuming
that all projects are financed at the optimum, we may treat the investment deci-
sion as if it were separable from the financing decision. First the firm decides
what optimal capital structure to use, then it applies the same capital structure
to all projects. Under this set of assumptions the decision to accept or reject a
particular project does not change the "optimal" capital structure. We could use
Eq. (13.12) with B/(B + S) set equal to the optimal capital structure in order to
determine the appropriate cost of capital for a project. This is precisely what we
did in the example.
But suppose projects carry with them the ability to change the optimal capital
structure of the firm as a whole." Suppose that some projects have more debt capac-
ity than others. Then the investment and financing decisions cannot even be "handled"
as if they were independent. There is very little in the accepted theory of finance that
admits of this possibility, but it cannot be disregarded. One reason that projects may
have separate debt capacities is the simple fact that they have different collateral
values in bankruptcy. However, since the theory in this chapter has proceeded on
the assumption that bankruptcy costs are zero, we shall refrain from further discus-
sion of this point until the next chapter.
So far it has been convenient to assume that corporate debt is risk free. Obviously
it is not. Consideration of risky debt raises several interesting questions. First, if debt
is risky, how are the basic Modigliani-Miller propositions affected? We know that
riskier debt will require higher rates of return. Does this reduce the tax gain from
leverage? The answer is given in section D.1 below. The second question is, How can
one estimate the required rate of return on risky debt? This is covered in section D.2.
13
This may be particularly relevant when a firm is considering a conglomerate merger with another firm
in a completely different line of business with a completely different optimal capital structure.
THE COST OF CAPITAL WITH RISKY DEBT 463
change the total value of the cash stream provided by the productive investments of
the firm. Therefore so long as there are no costs of bankruptcy (paid to third parties
like trustees and law firms), it should not make any difference whether debt is risk
free or risky. The value of the firm should be equal to the value of the discounted
cash flows from investment. A partition that divides these cash flows into risky debt
and risky equity has no impact on value. Stiglitz [1969] first proved this result, using
a state-preference framework, and Rubinstein [1973] provided a proof, using a mean-
variance approach.
Risky debt, just like any other security, must be priced in equilibrium so that it
falls on the security market line. Therefore if we designate the return on risky debt
as Rbi, its expected return is
= R f + [E(R m — R f ] Bhp (13.29)
-
where 1é 5j = COV(k, ici)/o ni . The return on the equity of a levered firm, ks , can be
written (for a perpetuity) as net income divided by the market value of equity:
(NOI — TZ biB)(1 — Te)
(13.30)
Recall that NOI is net operating income, iZbi B is the interest on debt, t is the firm's
tax rate, and SL is the market value of the equity in a levered firm. Using the CAPM,
we find that the expected return on equity will be'
E(l3s = R f + X*COV(ks, (13.31)
The covariance between the expected rate of return on equity and the market index
is
[(NOT — R bLB (} — E ((NOT — R, B (} — 3
COV(ks, Rm = E
x [R m — E(Rm ]}
} — -r e (} — L 3 B
= COV(NOI, R m) COV(R b j , R m). (13.32)
S
Substituting this result into (13.31) and the combined result into (13.30), we have the
following relationship for a levered firm:
R f S L + 2û(} — 3 COV (NOT, R m — 2û(} — 3 B[COV(R b j , R m ]
14
In this instance 7* [E(R„, — R
464 CAPITAL STRUCTURE AND THE COST OF CAPITAL: THEORY
Substituting (13.34) for E(NOI)(1 — t in the right-hand side of (13.33) and using the
fact that V L = S L + B, we have
R f V U — [R f + 2ûCOV (R bi , R m ]B(} — , ,
VL = + T eB .
This is exactly the same Modigliani-Miller result that we obtained when the firm
was assumed to issue only risk-free debt. Therefore the introduction of risky debt
cannot, by itself, be used to explain the existence of an optimal capital structure. In
Chapter 14 we shall see that direct bankruptcy costs such as losses to third parties
(lawyers or the courts) or indirect bankruptcy costs (disruption of services to custom-
ers or disruption of the supply of skilled labor) are necessary in conjunction with
risky debt and taxes in order to explain an optimal capital structure.
15
All accumulated interest on zero coupon bonds is paid at maturity; hence B(T , the current market
value of debt with maturity T, must be less than its face value, D, assuming a positive risk-free rate of
discount.
THE COST OF CAPITAL WITH RISKY DEBT 465
Payoffs at Maturity
Stakeholder Positions If V < D IfV>D
Shareholders' position:
Call option, S 0 V — D
Bondholders' position:
Default-free bond, B
Minus a put option, P — (D — V) 0
Value of the firm at maturity V V
Algebraically this is
V + P = B + S,
or rearranging,
V = (B — P) + S. (13.35)
Equation (13.35) illustrates that the value of the firm can be partitioned into two
claims. The low-risk claim is risky debt that is equivalent to default-free debt minus
a put option, i.e., (B — P). Thus, risky corporate debt is the same thing as default-
free debt minus a put option. The exercise price for the put is the face value of debt,
D, and the maturity of the put, T, is the same as the maturity of the risky debt. The
higher-risk claim is shareholders' equity, which is equivalent to a call on the value
of the firm with an exercise price D and a maturity T. The payoff to shareholders at
maturity will be
S = MAX[0, V — (13.36)
Table 13.3 shows both stakeholders' payoffs at maturity. If the value of the firm
is less than the face value of debt, shareholders file for bankruptcy and allow the
bondholders to keep V < D. Alternately, if the value of the firm is greater than the
face value of debt, shareholders will exercise their call option by paying its exercise
price, D, the face value of debt to bondholders, and retain the excess value, V — D.
The realization that the equity and debt in a firm can be conceptualized as options
allows us to use the insights of Chapter 8 on option pricing theory. For example, if
the equity, S, in a levered firm is analogous to a call option then its value will increase
with (1) an increase in the value of the firm's assets, V, (2) an increase in the variance
of the value of the firm's assets, (3) an increase in the time to maturity of a given
amount of debt with face value, D, and (4) an increase in the risk-free rate. The value
of levered equity will decrease with a greater amount of debt, D, which is analogous
to the exercise price on a call option.
Next, we wish to show the relationship between the CAPM measure of risk, i.e.,
fl, and the option pricing model. First, however, it is useful to show how the CAPM
and the OPM are related. Merton [1973] has derived a continuous-time version of
466 CAPITAL STRUCTURE AND THE COST OF CAPITAL: THEORY
This equation says that the change in the stock price is related to the change in the
value of the firm, dV, movement of the stock price across time, dt, and the instan-
taneous variance of the firm's value, a 2. Dividing by S, we have, in the limit as dt
approaches zero,
dS OS dV OS dV V
lim = (13.39)
dt- 0 S
, aV S 0V V S
We recognize dS/S as the rate of return on the common stock, rs , and dV/V as the
rate of return on the firm's assets, r v, therefore
OS V
rs aVSr (13.40)
If the instantaneous systematic risk of common stock, fi s, and that of the firm's assets,
fly, are defined as
COV(rs, I'm) COV(r v, r.)
fi]s — fl y
(13.41)
VAR(r m) VAR(r m)
then we can use (13.40) and (13.41) to rewrite the instantaneous covariance as
OS V COV(r v,r„,) OS V n
Qs = =—— PT , (13.42)
017 S VAR(r,i) 0V S
Now write the Black-Scholes OPM where the call option is the equity of the firm:
S = VN(di) — e -rfT DN(d2), (13.43)
THE COST OF CAPITAL WITH RISKY DEBT 467
where
S = the market value of equity,
V = the market value of the firm's assets,
r f = the risk-free rate,
T = the time to maturity,
D = the face value of debt (book value),
N(• = the cumulative normal probability of the unit normal variate, d },
ln(VID) + r f T
di = + a \,mf , d 2 = — o- N IT
6 NIT 2
Finally, the partial derivative of the equity value, S, with respect to the value of the
underlying asset is
OS
= i , where 0 < N(d i < }. (13.44)
OV N(d
Substituting this into (13.42) we obtain
This tells us the relationship between the systematic risk of the equity, fls, and the
systematic risk of the firm's assets, fly. The value of S is given by the OPM, Eq. (13.43);
therefore we have
=
V N(d } w
Ps
vN(d i — De - rf T N(d 2
1
=
av . (13.46)
1 — (DIV e -r f T [N(d 2 }N(d i n
We know that DmV < 1, that e -rf r < 1, that N(d 2 < N(d i , and hence that }é s > fé > 0.
This shows that the systematic risk of the equity of a levered firm is greater than the
systematic risk of an unlevered firm, a result that is consistent with the results found
elsewhere in the theory of finance. Note also that the beta of equity of the levered firm
increases monotonically with leverage.
The OPM provides insight into the effect of its parameters on the systematic
risk of equity. We may assume that the risk characteristics of the firm's assets, fiy,
are constant over time. Therefore it can be shown that the partial derivatives of
(13.46) have the following signs:
Most of these have readily intuitive explanations. The systematic risk of equity falls
as the market value of the firm increases, and it rises as the amount of debt issued
increases. When the risk-free rate of return increases, the value of the equity option
increases and its systematic risk decreases. The fourth partial derivative says that as
468 CAPITAL STRUCTURE AND THE COST OF CAPITAL: THEORY
the variance of the value of the firm's assets increases, the systematic risk of equity
decreases. This result follows from the contingent claim nature of equity. The equity
holders will prefer more variance to less because they profit from the probability that
the value of the firm will exceed the face value of the debt. Therefore their risk actually
decreases as the variance of the value of the firm's assets increases." Finally, the fifth
partial says that the systematic risk of equity declines as the maturity date of the
debt becomes longer and longer. From the shareholders' point of view the best
situation would be to never have to repay the face value of the debt. It is also possible
to use Eq. (13.45) to view the cost of equity capital in an OPM framework and to
compare it with the Modigliani-Miller results.
Substituting fl, from (13.45) into the CAPM, we obtain from Eq. (13.37) an
expression for k s, the cost of equity capital:
ks=Rf +(Rm —
Rf)N(di) --s- fly. (13.47)
Note that from Eq. (13.45), as = N(di)(V/S)fiv. Substituting this into (13.47) yields
the familiar CAPM relationship k s = R f + (Rm — Rf)/3s. Furthermore, the CAPM
can be rearranged to show that
a — Rf
PV
Rm — R f
which we substitute into (13.47) to obtain
)V
k, = R f + N(d i )(R, — R1 (13.48)
Equation (13.48) shows that the cost of equity capital is an increasing function of
financial leverage.
If we assume that debt is risky and assume that bankruptcy costs (i.e., losses to
third parties other than creditors or shareholders) are zero, then the OPM, the CAPM,
and the Modigliani-Miller propositions can be shown to be consistent. The simple
algebraic approach given below was proved by Hsia [1981].
First, note that the systematic risk, /3,, of risky debt capital in a world without
taxes can be written in an explanation similar to Eq. (13.42) as'
OB V
iqB = fly B•
(13.49)
We know that in a world without taxes the value of the firm is invariant to
changes in its capital structure. Also, from Eq. (13.44), we know that if the common
stock of a firm is thought of as a call option on the value of the firm, then
OS
OV=N(c11).
" Note that since the value of the firm, V, and the debt equity ratio DIV are held constant, any change
in total variance, a 2, must be nonsystematic risk.
17
See Galai and Masulis [1976, footnote 15].
THE COST OF CAPITAL WITH RISKY DEBT 469
In other words, any change in the value of equity is offset by an equal and opposite
change in the value of risky debt.
Next, the required rate of return on risky debt, k b , can be expressed by using the
CAPM, Eq. (13.37):
k b = R f + (R. — R JOB. (13.51)
Substituting Eqs. (13.49) and (13.50) into (13.51), we have
k b = R f + (R m — R f flyN( — d l B.
Note that Eq. (13.52) expresses the cost of risky debt in terms of the OPM. The
required rate of return on risky debt is equal to the risk-free rate, R}, plus a risk
premium, <, where
<= (p — R f N( — d l
A numerical example can be used to illustrate how the cost of debt, in the absence
of bankruptcy costs, increases with the firm's utilization of debt. Suppose the current
value of a firm, V, is $3 million; the face value of debt is $1.5 million; and the debt
will mature in T = 8 years. The variance of returns on the firm's assets, 6<, is .09; its
required return on assets is p = .12; and the riskless rate of interest, R f , is !M. From
the Black-Scholes option pricing model, we know that
ln(VID) + R f T }
dl= j'
\IT"] +2
ln(3/1.5) + .05(8)
+ .5(.3)\
.3 \/T
.6931 + .4
+ .4243 = 1.7125.
.8485
470 CAPITAL STRUCTURE AND THE COST OF CAPITAL: THEORY
• Figure 13.8
The cost of risky debt.
.08 —
.07
.06
R f = .05
.04
D
0.5 1.0 V
From the cumulative normal probability table (Table 8.7), the value of N(— 1.7125)
is approximately .0434. Substituting into Eq. 13.52, we see that the cost of debt is
increased from the risk-free rate, 5%, to
3
k b = .05 + (.12 — .05)(.0434)
1.5
Figure 13.8 shows the relationship of the cost of debt and the ratio of the face value
of debt to the current market value of the firm. For low levels of debt, bankruptcy
risk is trivial and therefore the cost of debt is close to the riskless rate. It rises as
DmV increases until k b equals 6.3% when the face value of debt, due eight years from
now, equals the current market value of the firm.
To arrive at a weighted average cost of capital, we multiply Eq. (13.52), the cost
of debt, by the percentage of debt in the capital structure, BmV, then add this result
to the cost of equity, Eq. (13.48) multiplied by S/V, the percentage of equity in the
capital structure. The result is
= Rf ( B v
+ S
+ (p — RAN(—3l l + N(3}, ]
= Rf (p - 1
R4[ — N(d} + N(d 1)]
= P. (13.53)
= + (}< — kb —
• (13.54)
s
THE MATURITY STRUCTURE OF DEBT 471
P + (p—kb)A
WACC = p
k b = R f + (p—R f )N(—d 1 )
B
1.0 B+S
(a) No taxes (b) Only corporate taxes
Figure 13.9
The cost of capital given risky debt.
This is exactly the same as Eq. (13.18), the Modigliani-Miller definition of the cost
of equity capital in a world without taxes. Therefore if we assume that debt is risky,
the OPM, the CAPM, and the Modigliani-Miller definition are all consistent with
one another.
This result is shown graphically in Fig. 13.9(a). This figure is very similar to Fig.
13.2, which showed the cost of capital as a function of the debt to equity ratio, B/S,
assuming riskless debt. The only differences between the two figures are that Fig.
13.9 has the debt to value ratio, B/(B + S), on the horizontal axis and it assumes
risky debt. Note that [in Fig. 13.9(a)] the cost of debt increases as more debt is used
in the firm's capital structure. Also, if the firm were to become 100% debt (not a
realistic alternative) then the cost of debt would equal the cost of capital for an all-
equity firm, p. Figure 13.9(b) depicts the weighted average cost of capital in a world
with corporate taxes only. The usual Modigliani-Miller result is shown, namely, that
the weighted average cost of capital declines monotonically as more debt is employed
in the capital structure of the firm. The fact that debt is risky does not change any
of our previous results.
Optimal capital structure refers not only to the ratio of debt to equity but also to
the maturity structure of debt. What portion of total debt should be short term and
what portion long term? Should the firm use variable rate or fixed rate debt? Should
long-term bonds pay annual coupons with a balloon payment, or should they be
fully amortized (equal periodic payments)?
There are three approaches to answering the maturity structure problem. The
earliest, by Morris [1976], suggests that short-term debt or variable rate debt can
472 CAPITAL STRUCTURE AND THE COST OF CAPITAL: THEORY
reduce the risk to shareholders and thereby increase equity value if the covariance
between net operating income and expected future interest rates is positive. This
cross-hedging argument is based on the assumption that unexpected changes in in-
terest rates are a priced (undiversifiable) factor in the arbitrage pricing model. It does
not rely directly either on bankruptcy costs or on interest tax shields. However, the
argument for cross-hedging is only strengthened if it increases debt capacity by re-
ducing the risk of bankruptcy and thereby allowing a greater gain from leverage.
Smith and Stulz [1985] support this point of view.
A second approach to optimal debt maturity is based on agency costs. Myers
[1977] and Barnea, Haugen, and Senbet [1980] argue that if the shareholders' claim
on the assets of a levered firm is similar to a call option, then shareholders have an
incentive to undertake riskier (higher variance) projects because their call option
value is greater when the assets of the firm have higher variance. If the firm with
long-term risky debt outstanding undertakes positive net present value projects, share-
holders will not be able to capture the full benefits because part of the value goes to
debt holders in the form of a reduction in the probability of default. Short-term debt
may alleviate this problem because the debt may come due before the firm decides
to invest. Hence the theory suggests that firms with many investment opportunities
may prefer to use short-term debt (or callable debt).
Brick and Ravid [1985] provide a tax-based explanation. Suppose the term struc-
ture of interest rates is not flat and there is a gain to leverage in the Miller [1977]
sense. Then a long-term maturity is optimal because coupons on long-term bonds
are currently higher than coupons on short-term bonds and the tax benefit of debt
(the gain to leverage) is accelerated. If the gain to leverage is negative, then the result
is reversed.
Although none of these theories has been adequately tested, each of them has
some merit for potentially explaining cross-sectional regularities in the maturity struc-
ture of debt. This remains a fruitful area for further research.
Other than straight debt and equity, firms issue a variety of other securities and con-
tingent claims. The number of different possibilities is limited only by your imagina-
tion. However, the actual number of alternative financial instruments is fairly small
and their use is limited. A possible explanation for why corporations tend to use
only straight debt and equity has been offered by Fama and Jensen [1982]. They
argue that it makes sense to separate the financial claims on the firm into only two
parts: a relatively low-risk component (i.e., debt capital) and a relatively high-risk re-
sidual claim (i.e., equity capital). Specialized risk bearing by residual claimants is an
optimal form of contracting that has survival value because (1) it reduces contracting
costs (i.e., the costs that would be incurred to monitor contract fulfillment) and (2)
it lowers the cost of risk-bearing services.
THE EFFECT OF OTHER FINANCIAL INSTRUMENTS 473
For example, shareholders and bondholders do not have to monitor each other.
It is necessary only for bondholders to monitor shareholders. This form of one-way
monitoring reduces the total cost of contracting over what it might otherwise be.
Thus it makes sense that most firms keep their capital structure fairly simple by using
only debt and equity.
The theory of finance has no good explanation for why some firms use alterna-
tive financial instruments such as convertible debt, preferred stock, and warrants.
1. Warrants
A warrant is a security issued by the firm in return for cash. It promises to sell
m shares (usually one share) of stock to an investor for a fixed exercise price at any
time up to the maturity date. Therefore a warrant is very much like an American
call option written by the firm. It is not exactly the same as a call because, when ex-
ercised, it increases the number of shares outstanding and thus dilutes the equity of
the stockholders.
The problem of pricing warrants has been studied by Emanuel [1983], Schwartz
[1977], Galai and Schneller [1978], and Constantinides [1984]. The simplest ap-
proach to the problem (Galai and Schneller [1978]) assumes a one-period model. The
firm is assumed to be 100% equity financed, and its investment policy is not af-
fected by its financing decisions. For example, the proceeds from issuing warrants are
immediately distributed as dividends to the old shareholders. Also the firm pays no
end-of-period dividends, and the warrants are assumed to be exercised as a block."
These somewhat restrictive assumptions facilitate the estimation of the warrant value
and its equilibrium rate of return.
Galai and Schneller show, for the above-mentioned assumptions, that the returns
on a warrant are perfectly correlated with those of a call option on the same firm
without warrants. To obtain this result let V be the value of the firm's assets (without
warrants) at the end of the time period, i.e., on the date when the warrants mature.
Let n be the current number of shares outstanding and q be the ratio of warrants to
shares outstanding.' Finally, let X be the exercise price of the warrant. If the firm
had no warrants outstanding, the price per share at the end of the time period would
be
V
S = —•
n
With warrants, the price per share, assuming that the warrants are exercised, will be
V + nqX S + qX
S= •
n(1 + q) (1 + q)
8
Block exercise is, perhaps, the most restrictive assumption.
19The amount of potential dilution can be significant. For example, in July 1977 there were 118 warrants
outstanding. Of them 41% had a dilution factor of less than 10%, 25% had a dilution factor between 10
and 19%, 13% between 20 and 29%, and 21% a factor of over 50%.
474 CAPITAL STRUCTURE AND THE COST OF CAPITAL: THEORY
End-of-Period Payoffs
IfS<X If S > X
S + qX 1
Warrant on firm with warrants, W 0 X = (S X)
1+q 1+q
Call on firm without warrants, C 0 S—X
Of course, nqX is the cash received and n(1 + q) is the total number of shares out-
standing if the warrants are exercised. The warrants will be exercised if their value
when converted is greater than the exercise price, i.e., if
S + qX
S= > X.
1+q
This condition is exactly equivalent to S > X. In other words the warrant will be
exercised whenever the firm's end-of-period share price without warrants exceeds the
warrant exercise price. Therefore the warrant will be exercised in exactly the same
states of nature as a call option with the same exercise price. Also, as shown in Table
13.4, the payoffs to the warrant are a constant fraction, 1/(1 + q), of the payoffs to
the call written on the assets of the firm (without warrants).
Therefore the returns on the warrant are perfectly correlated with the dollar re-
turns on a call option written on the firm without warrants. To prevent arbitrage
the warrant price, W, will be a fraction of the call price, C, as shown below:
W= C. (13.55)
1+q
Because the warrant and the call are perfectly correlated, they will have exactly the
same systematic risk and therefore the same required rate of return.' This expected
'From Eq. (13.45) we know that the beta of an option is related to the beta of the underlying asset as
follows:
= N(c1 1) —
c fis
From Eq. (13.55) we know that the warrant s perfectly correlated with a call option written on the shares
of the company, ex warrants; therefore
fia, —
Consequently, it is not difficult to estimate the cost of capital for a warrant because we can estimate
= fin, and then employ the CAPM.
)3,
THE EFFECT OF OTHER FINANCIAL INSTRUMENTS 475
return is the before-tax cost of capital for issuing warrants and can easily be estimated
for a company that is contemplating a new issue of warrants.
One problem with the above approach is that warrants are not constrained to
be exercised simultaneously in one large block. Emanuel [1983] demonstrated that
if all the warrants were held by a single profit-maximizing monopolist, the warrants
would be exercised sequentially. Constantinides [1984] has solved the warrant valua-
tion problem for competitive warrant holders and shown that the warrant price,
given a competitive equilibrium, is less than or equal to the value it would have given
block exercise. Frequently the balance sheet of a firm has several contingent claim
securities, e.g., warrants and convertible bonds, with different maturity dates. This
means that the expiration and subsequent exercise (or conversion) of one security
can result in equity dilution and therefore early exercise of the longer maturity con-
tingent claim securities. Firms can also force early exercise or conversion by paying
a large cash or stock dividend.
2. Convertible Bonds
As the name implies, convertible debt is a hybrid bond that allows its bearer to
exchange it for a given number of shares of stock anytime up to and including the
maturity date of the bond. Preferred stock is also frequently issued with a convertible
provision and may be thought of as a convertible security (a bond) with an infinite
maturity date.
A convertible bond is equivalent to a portfolio of two securities: straight debt
with the same coupon rate and maturity as the convertible bond, and a warrant
written on the value of the firm. The coupon rate on convertible bonds is usually
lower than comparable straight debt because the right to convert is worth something.
For example, in February 1982, the XYZ Company wanted to raise $50 million by
using either straight debt or convertible bonds. An investment banking firm informed
the company's treasurer that straight debt with a 25-year maturity would require a
17% coupon. Alternately, convertible debt with the same maturity would require only
a 10% coupon. Both debt instruments would sell at par (i.e., $1000), and the con-
vertible debt could be converted into 35.71 shares (i.e., an exercise price of $28 per
share). The stock of the XYZ Company was selling for $25 per share at the time.
Later on we will use these facts to compute the cost of capital for the convertible
issue. But first, what do financial officers think of convertible debt?
Brigham [1966] received responses from the chief financial officers of 22 firms
that had issued convertible debt. Of them, 68% said they used convertible debt be-
cause they believed their stock price would rise over time and that convertibles would
provide a way of selling common stock at a price above the existing market. Another
27% said that their company wanted straight debt but found conditions to be such
that a straight bond issue could not be sold at a reasonable rate of interest. The
problem is that neither reason makes much sense. Convertible bonds are not "cheap
debt." Because convertible bonds are riskier, their true cost of capital is greater (on
a before-tax basis) than the cost of straight debt. Also, convertible bonds are not
476 CAPITAL STRUCTURE AND THE COST OF CAPITAL: THEORY
deferred sale of common stock at an attractive price.' The uncertain sale of shares
for $28 each at some future date can hardly be compared directly with a current share
price of $25.
Brennan and Schwartz [1977a] and Ingersoll [1977a] have analyzed the valua-
tion of convertible bonds, assuming that the entire outstanding issue, if converted,
will be converted as a block. Constantinides [1984] has extended their work to study
the value of convertible debt if conversion does not occur all at once. The reader is
referred to these articles for the derivations that show that the market value of con-
vertible debt, CV, is equal to the market value of straight debt, B, and a warrant, W:
CV = B + W.
Suppose you want to compute the cost of capital for the convertible debt being
considered by the XYZ Company as mentioned above. You already know that the
maturity date is 25 years, similar straight debt yields 17% to maturity, the convertible
bond coupon rate is 10% (with semiannual coupons), the conversion price (exercise
price) is $28 per share, the bond will sell at par value, i.e., $1000, and the current
stock price is $25. In addition you need to know that: (1) if converted the issue would
increase the firm's outstanding shares by 5%, i.e., the dilution factor, q, is 5%; (2) the
standard deviation of the firm's equity rate of return is a = .3; (3) the risk-free rate
is 14.5% for 25-year Treasury bonds; (4) the expected rate of return on the market
portfolio is 20.6%; (5) the firm's equity beta is 1.5; and (6) the firm pays no dividends.
Given these facts, it is possible to use the capital asset pricing model and the option
pricing model to estimate the before-tax cost of capital, k„, on the firm's contem-
plated convertible bond issue as a weighted average of the cost of straight debt, k b ,
and the cost of the warrant, k„„ 22
k„ = kb +
•
B+W B+W
The value of the straight debt, assuming semiannual coupons of $50, a principal
payment of $1000 twenty-five years hence, and a 17% discount rate, is B = $619.91.
Therefore the remainder of the sale price namely, $1000 619.91 = $380.09 is
the value of the warrant to purchase 35.71 shares at $28 each. The cost of straight
debt was given to be k b = 17% before taxes. All that remains is to find the cost of
the warrant. From section F.1 we know that the warrant implied in the convertible
bond contract is perfectly correlated with a call option written on the firm (without
warrants outstanding) and therefore has the same cost of capital. The cost of capital,
k„ for the call option can be estimated from the CAPM:
/cc = R f• + [E(R7b) — R f ][3„
21
From the theory of option pricing we know that S + P = B + C; i.e., a bond plus a call option is the
same thing as owning the stock and a put option. Thus one could think of a convertible bond as roughly
equivalent to the stock plus a put.
22
Throughout the analysis we assume that there is no tax gain to leverage. Therefore the conversion of
the bond will decrease the firm's debt-to-equity ratio but not change the value of the firm.
THE EFFECT OF OTHER FINANCIAL INSTRUMENTS 477
where
k, = the cost of capital for a call option with 25 years to maturity,"
R f = the risk-free rate for a 25-year Treasury bond = 14.5%,
E(R,,,) = the expected rate of return on the market portfolio = 20.6%,
/3, = N(d„)(S/C)/3, = the systematic risk of the call option,
f3 = the systematic risk of the stock (without warrants) = 1.5,
N(d 1) = the cumulative normal probability for option pricing in Chapter 8,
C = the value of a call option written on the stock, ex warrants.
ln(S/X) + R f T 1
d, =
o- NrT +26
Substituting in the appropriate numbers, we find that d 1 = 3.09114 and N(d 1) = .999.
And using the Black-Scholes version of the option pricing model, we find that C =
$24.74. Therefore
25.00
24.74 (.999)(1.5) = 1.514,
The cost of capital for the warrant is slightly above the cost of equity for the firm.
Actually, the warrant is not much riskier than the equity because its market value
is almost equal to the market value of the firm's equity, given a 25-year life and only
a $3 difference between the exercise price and the stock price.
Taken together, these facts imply that the before-tax cost of capital for the con-
vertible issue will be
619.91 + .2374 380.09
k c, = .17
1000.00 1000.00
= 19.56%.
This answer is almost double the 10% coupon rate that the convertible promises to
pay, and it shows that the higher risk of convertible debt requires a higher expected
rate of return.
23
If the firm pays dividends which are large enough, then the convertible debentures may be exercised if
the implied warrants are in-the-money. Exercise would occur just prior to the ex dividend date(s). We are
assuming, for the sake of simplicity, that the firm pays no dividends.
478 CAPITAL STRUCTURE AND THE COST OF CAPITAL: THEORY
The final point of discussion is why convertible debt is used if financial officers
understand its true cost. It certainly is not a cheap form of either debt or equity.
Another irrational explanation is that until the accounting rules were changed to re-
quire reporting earnings per share on a fully diluted basis, it was possible for an
aggressive firm to acquire another company via convertible debt financing. The lower
interest charges of convertible debt meant that earnings of the merged company were
often higher than the sum of premerger earnings of the separate entities. Also, the
actual number of shares outstanding was lower than the number that would be
reported if the conversion were to occur. Given all the evidence in Chapter 11 on
the efficiency of markets, it is hard to believe that the market was fooled by the
accounting conventions. A possible reason for issuing convertibles is that they are
better tailored to the cash flow patterns of rapidly growing firms. The lower coupon
rate during the early years keeps the probability of bankruptcy lower than straight
debt; then, if the firm is successful, more cash for growth will be available after con-
version takes place. Brennan and Schwartz [1986] suggest an alternative rationale
namely, that because of the relative insensitivity of convertible bonds to the risk of
the issuing company, it is easier for the bond issuer and purchaser to agree on the
value of the bond. This makes it easier for them to come to terms and requires no
bonding or underwriting service by investment bankers. Green [1984] shows that
agency costs between equity and bondholders are reduced by issuing convertible debt
or straight debt with warrants. Bondholders are less concerned about the possibility
that shareholders may undertake risky projects (thereby increasing the risk of bank-
ruptcy) because their conversion privilege allows them to participate in the value
created if riskier projects are undertaken. Finally, convertible debt may be preferred
to straight debt with warrants attached because convertible debt often has a call pro-
vision built in that allows a firm to force conversion. The call feature is discussed in
the next section.
3. Call Provisions
Many securities have call provisions that allow the firm to force redemption.
Frequently, ordinary bonds may be redeemed at a call premium roughly equal to 1
year's interest. For example, the call premium on a 20-year $1000 face value bond
with a 12% coupon might be $120 if the bond is called in the first year, $114 if called
in the second year, and so on.
The call provision is equivalent to a call option written by the investors who
buy the bonds from the firm. The bonds may be repurchased by the firm (at the
exercise price, i.e., the call price) anytime during the life of the bond. If interest rates
fall, the market price of the outstanding bonds may exceed the call price, thereby
making it advantageous for the firm to exercise its option to call in the debt. Since
the option is valuable to the firm, it must pay the bondholders by offering a higher
interest rate on callable bonds than on similar ordinary bonds that do not have the
call feature. New issues of callable bonds must often bear yields from one quarter to
one half of a percent higher than the yields of noncallable bonds.
THE EFFECT OF OTHER FINANCIAL INSTRUMENTS 479
Brennan and Schwartz [1977a] show how to value callable bonds. If the objec-
tive of the firm is to maximize shareholders' wealth, then a call policy will be estab-
lished to minimize the market value of callable debt. The value of the bonds will be
minimized if they are called at the point where their uncalled value is equal to the
call price. To call when the uncalled value is below the call price is to provide a
needless gain to bondholders. To allow the uncalled bond value to rise above the
call price is inconsistent with minimizing the bond value. Therefore the firm should
call the bond when the market price first rises to reach the call price. Furthermore,
we would never expect the market value of a callable bond to exceed the call price
plus a small premium for the flotation costs the firm must bear in calling the issue.
Almost all corporate bonds are callable and none are puttable. Why? A plausible
answer has been put forth by Boyce and Kalotay [1979]. Whenever the tax rate of
the borrower exceeds the tax rate of the lender, there is a tax incentive for issuing
callable debt. Since corporations have had marginal tax rates of around 50% while
individuals have lower rates, corporations have had an incentive to issue callable
bonds.' From the firm's point of view the coupons paid and the call premium are
both deductible as interest expenses. The investor pays ordinary income taxes on
interest received and capital gains taxes on the call premium. If the stream of payments
on debt is even across time, then low and high tax bracket lenders and borrowers
will value it equally. However, if it is decreasing across time, as it is expected to be
with a callable bond, then low tax bracket lenders will assign a higher value because
they discount at a higher after-tax rate. Near-term cash inflows are relatively more
valuable to them. A high tax bracket borrower (i.e., the firm) will use a lower after-tax
discount rate and will also prefer a decreasing cash flow pattern because the present
value of the interest tax shield will be relatively higher. Even though the firm pays
a higher gross rate, it prefers callable debt to ordinary debt because of the tax ad-
vantages for the net rate of return.
Brennan and Schwartz [1977a] and Ingersoll [1977a] both examined the effect
of a call feature on convertible debt and preferred. Unlike simpler option securities,
convertible bonds and preferred stocks contain dual options. The bondholder has
the right to exchange a convertible for the company's common stock while the com-
pany retains the right to call the issue at the contracted call price. One interesting
implication of the theory on call policies is that a convertible security should be
called as soon as its conversion value (i.e., the value of the common stock that would
be received in the conversion exchange) rises to equal the prevailing effective call
price (i.e., the stated call price plus accrued interest). Ingersoll [1977b] collected data
on 179 convertible issues that were called between 1968 and 1975. The calls on all
but 9 were delayed beyond the time that the theory predicted. The median company
waited until the conversion value of its debentures was 43.9% in excess of the call
price.
'4 Interestingly, the opposite is true when the government is lending. The government has a zero tax rate
and holders of government debt have positive rates. Consequently, the government has incentive to offer
puttable debt and it does. Series E and H savings bonds are redeemable at the lender's option.
480 CAPITAL STRUCTURE AND THE COST OF CAPITAL: THEORY
Mikkelson [1981] discovered that, on average, the common stock returns of com-
panies announcing convertible debt calls fell by a statistically significant — 1.065%
per day over a two-day announcement period. These results are inconsistent with
the idea that optimal calls of convertible debt are beneficial for shareholders.
Harris and Raviv [1985] provide a signaling model that simultaneously explains
why calls are delayed far beyond what would seem to be a rational time and why stock
returns are negative at the time of the call. Suppose that managers know the future
prospects of their firm better than the marketplace—i.e., there is heterogeneous in-
formation. Also, assume that managers' compensation depends on the firm's stock
price, both now and in future periods. If the managers suspect that the stock price will
fall in the future, conversion will be forced now because what they receive now, given
conversion, exceeds what they would otherwise receive in the future when the market
learns of the bad news and does not convert. Conversely, managers' failure to convert
now will be interpreted by the market as good news. There is incentive for managers
not to force conversion early because the market views their stock favorably now, and
it will also be viewed favorably in the future when the market is able to confirm the
managers' good news. A paper of similar spirit by Robbins and Schatzberg [1986]
explains the advantage of the call feature in nonconvertible long-term bonds.
4. Preferred Stock
Preferred stock is much like subordinated debt except that if the promised cash
payments (i.e., the preferred coupons) are not paid on time, then preferred shareholders
cannot force the firm into bankruptcy. All preferred stocks listed on the New York
Stock Exchange must have voting rights in order to be listed. A high percentage of
preferred stocks have no stated maturity date and also provide for cumulative divi-
dend payments; i.e., all past preferred dividends must be paid before common stock
dividends can be paid. Approximately 40% of new preferred stocks are convertible
into common stock.
If preferred stock is not callable or convertible, and if its life is indefinite, then
its market value is
coupon
P=
kp
where k, is the before-tax cost of preferred. Of course, the before- and after-tax costs
are the same for preferred stock because preferred dividends are not deductible as an
expense before taxes. The nondeductibility of preferred dividends has led many com-
panies to buy back their preferred stock and use subordinated debt instead. It is a
puzzle why preferred stock is issued at all, especially if there is a gain to leverage from
using debt capital as a substitute.
SUMMARY
The cost of capital is seen to be a rate of return whose definition requires a project
to improve the wealth position of the current shareholders of the firm. The original
Modigliani-Miller work has been extended by using the CAPM so that a risk-adjusted
cost of capital may be obtained for each project. When the expected cash flows of the
project are discounted at the correct risk-adjusted rate, the result is the NPV of the
project.
In a world without taxes the value of the firm is independent of its capital structure.
However, there are several important extensions of the basic model. With the intro-
duction of corporate taxes the optimal capital structure becomes 100% debt. Finally,
when personal taxes are also introduced, the value of the firm is unaffected by the
choice of financial leverage. Financing is irrelevant! The next chapter takes a more
careful look at the question of optimal capital structure and summarizes some of the
empirical work that has been done.
PROBLEM SET
13.1 The Modigliani-Miller theorem assumes that the firm has only two classes of securities,
perpetual debt and equity. Suppose that the firm has issued a third class of securities preferred
stock—and that X% of preferred dividends may be written off as an expense (0 < X < 1).
a) What is the appropriate expression for the value of the levered firm?
b) What is the appropriate expression for the weighted average cost of capital?
13.2 The Acrosstown Company has an equity beta, fiL , of .5 and 50% debt in its capital structure.
The company has risk-free debt that costs 6% before taxes, and the expected rate of return on
the market is 18%. Acrosstown is considering the acquisition of a new project in the peanut-
raising agribusiness that is expected to yield 25% on after-tax operating cash flows. The Carter-
nut Company, which is the same product line (and risk class) as the project being considered, has
an equity beta, )3,, of 2.0 and has 10% debt in its capital structure. If Acrosstown finances the
482 CAPITAL STRUCTURE AND THE COST OF CAPITAL: THEORY
new project with 50% debt, should it be accepted or rejected? Assume that the marginal tax
rate, for both companies is 50%.
13.3 The XYZ Company has a current market value of $1,000,000, half of which is debt. Its
current weighted average cost of capital is 9%, and the corporate tax rate is 40%. The treasurer
proposes to undertake a new project, which costs $500,000 and which can be financed com-
pletely with debt. The project is expected to have the same operating risk as the company and
to earn 8.5% on its levered after-tax cash flows. The treasurer argues that the project is desirable
because it earns more than 5%, which is the before-tax marginal cost of the debt used to finance
it. What do you think?
13.4 Given a world with corporate taxes, r c , a personal tax rate paid on bonds, and a per-
TpB ,
sonal tax rate on income from equity, T ps, what would be the effect of a decrease in the corporate
tax rate on
a) the aggregate amount of debt in the economy, and
b) the optimal capital structure of firms?
13.5 Congress has proposed to eliminate "double taxation" on dividends by reducing the per-
sonal tax on dividend income. At the same time, a compensating increase in taxes on capital
gains (traditionally taxed at a much lower percentage than dividend income) has been proposed.
a) What effect would this joint proposal have on the optimal capital structure of a firm, ac-
cording to the Miller model?
b) What effect would it have on the aggregate amount of corporate debt outstanding?
13.6 Consider firm B as an unlevered firm and firm C as a levered firm with target debt-to-equity
ratio (B/ = 1. Both firms have exactly the same perpetual net operating income, NOI = 180,
before taxes. The before-tax cost of debt, k b , is the same as the risk-free rate. The corporate
tax rate = .5. Given the following market parameters,
a) Find the cost of capital and value for each firm. [Ignore any effect from personal income
taxes.]
b) Evaluate the following four projects to determine their acceptance (or rejection) by firms
B and C. What do the results of this evaluation tell you about leverage in a world with
corporate taxes but no personal taxes? [Note: rim is the correlation between the unlevered
free cash flows of each project and the market.]
r j.
E(NOLO) Correlation of j
Project ] Cost j (after-tax) aj with the Market
1 100 9 .10 .6
2 120 11 .11 .7
3 80 9 .12 .8
4 150 18 .20 .9
13.7 A firm with $1,000,000 in assets and 50% debt in its capital structure is considering a
$250,000 project. The firm's after-tax weighted average cost of capital is 10.4%, the marginal
PROBLEM SET 483
Table Q13.9
cost of debt is 8% (before taxes), and the marginal tax rate is 40%. If the project does not
change the firm's operating risk and is financed exclusively with new equity, what rate of return
must it earn to be acceptable?
13.8 The firm's cost of equity capital is 18%, the market value of the firm's equity is $8 million,
the firm's cost of debt capital is 9%, and the market value of debt is $4 million. The firm is
considering a new investment with an expected rate of return of 17%. This project is 30%
riskier than the firm's average operations. The riskless rate of return is 5%; the variance of
the market return is .08. Is the project profitable? [Assume a world without taxes.]
13.9 Susan Varhard, treasurer of the Gammamax Company, has proposed that the company
should sell equity and buy back debt in order to maximize its value. As evidence, she presents
the financial statements given in Table Q13.9. The company currently has a price/earnings
ratio of 50. Before the change in capital structure it has 10 shares outstanding; therefore its
earnings per share are $1.00, and the price per share is $50. If 10 new shares are issued at $50
each, $500 is collected and used to retire $500 of debt (which pays a coupon rate of 8%). After
the capital structure change, earnings per share have increased to $1.50 (since there are now
20 shares outstanding); and with a price/earnings ratio of 50, presumably the price per share
will increase from $50 before the capital structure change to $75 afterward. Given your under-
standing of modern finance theory, discuss the above proposal.
13.10 Community Bank must decide whether to open a new branch. The current market value
of the bank is $2,500,000. According to company policy (and industry practice), the bank's
capital structure is highly leveraged. The present (and optimal) ratio of debt to total assets is
.9. Community Bank's debt is almost exclusively in the form of demand, savings, and time de-
posits. The average return on these deposits to the bank's clients has been 5% over the past
five years. However, recently interest rates have climbed sharply, and as a result Community
Bank presently pays an average annual rate of 64% on its accounts in order to remain compet-
itive. In addition, the bank incurs a service cost of n% per account. Because federal "Regula-
tion Q" puts a ceiling on the amount of interest paid by banks on their accounts, the banking
industry at large has been experiencing disintermediation—a loss of clients to the open money
484 CAPITAL STRUCTURE AND THE COST OF CAPITAL: THEORY
market (Treasury bills, etc.), where interest rates are higher. Largely because of the interest
rate situation (which shows no sign of improving), Community Bank's president has stipulated
that for the branch project to be acceptable its entire cost of $500,000 will have to be raised
by 90% debt and 10% equity. The bank's cost of equity capital, k,, is 11%. Community Bank's
marginal tax rate is .48. Market analysis indicates that the new branch may be expected to
return net cash flows according to the following schedule:
Year 0 1 2 3 4 5 6 to co
— 500,000 25,000 35,000 45,000 45,000 50,000 50,000
REFERENCES
Barnea, A.; R. Haugen; and L. Senbet, "A Rationale for Debt Maturity Structure and Call
Provisions in the Agency Theory Framework," Journal of Finance, December 1980,
1223-1243.
Baxter, N. D., "Leverage, Risk of Ruin and the Cost of Capital," Journal of Finance, September
1967, 395-403.
Beranek, W., "The WACC Criterion and Shareholder Wealth Maximization," Journal of
Financial and Quantitative Analysis, March 1977, 17 -32.
Bierman, H., "The Cost of Warrants," Journal of Financial and Quantitative Analysis, June
1973, 499-503.
Black, F., and J. Cox, "Valuing Corporate Securities: Some Effects of Bond Indenture
Provisions," Journal of Finance, May 1976, 351-367.
Boyce, W. M., and A. J. Kalotay, "Tax Differentials and Callable Bonds," Journal of Finance,
September 1979, 825-838.
Brennan, M. J., and E. S. Schwartz, "Convertible Bonds: Valuation and Optimal Strategies
for Call and Conversion," Journal of Finance, December 1977a, 1699-1715.
, "Savings Bonds, Retractable Bonds and Callable Bonds," Journal of Financial
Economics, Aug4st 1977b, 67-88.
, "Corporate Income Taxes, Valuation, and the Problem of Optimal Capital Structure,"
Journal of Business, January 1978, 103-114.
, "Analyzing Convertible Bonds," Journal of Financial and Quantitative Analysis,
November 1980, 907-929.
, "The Case for Convertibles," in Stern and Chew, eds., The Revolution in Corporate
Finance. Basil Blackwell, Oxford, England, 1986.
Brick, I. E., and A. Ravid, "On the Relevance of Debt Maturity Structure," Journal of Finance,
December 1985, 1423-1437.
Brigham, E. F., "An Analysis of Convertible Debentures," Journal of Finance, March 1966,
35-54.
Chen, A. H. Y., "A Model of Warrant Pricing in a Dynamic Market," Journal of Finance,
December 1970, 1041-1060.
486 CAPITAL STRUCTURE AND THE COST OF CAPITAL: THEORY
Jennings, E. H., "An Estimate of Convertiîle Bond Premiums," Jou]nal of Financial and
Quantitative Analysis, January =974, 33-P6.
Kim, H., "A Mean-Variance Theory of Optimal Capital Structure and Corporate Deît
Capacity," Jou]nal of Finance, March =978, 4P-64.
Kraus, A., and R. Litzenîerger, "A State-Preference Model of Optimal Financial Leverage,"
Jou]nal of Finance, Septemîer =973, 9==-9<<.
Marshall, W., and J. Yawitz, "Optimal Terms of the Call Provision on a Corporate Bond,"
Jou]nal of Financial Resea]ch, Summer =980, <0<-<==.
Mason, S. P., and S. Bhattacharya, "Risky Deît, Jump Processes, and Safety Covenants,"
Jou]nal of Financial Economics, Septemîer =98=, <8=-307.
Merton, R. C., "An Intertemporal Capital Asset Pricing Model," Economet]ica, Septemîer
=973, 867-887.
Mikkelson, W. H., "Convertiîle Calls and Security Returns," Jou]nal of Financial Economics,
Septemîer =98=, <37-<64.
Miller, M. H., "Deît and Taxes," Jou]nal of Finance, May =977, <6=-<7P.
, and F. Modigliani, "Some Estimates of the Cost of Capital to the Electric Utility
Industry =9P4-P7," Ame]ican Economic Review, June =966, 333-348.
Modigliani, F., and M. H. Miller, "The Cost of Capital, Corporation Finance, and the Theory
of Investment," Ame]ican Economic Review, June =9P8, <6=-<97.
, "Corporate Income Taxes and the Cost of Capital," Ame]ican Economic Review, June
=963, 433-443.
Morris, J. R., "On Corporate Deît Maturity Policies," Jou]nal of Finance, March =976a,
<9-37.
, "A Model for Corporate Deît Maturity Decisions," Jou]nal of Financial and Quanti-
tative Analysis, Septemîer =976î, 339-3P7.
Myers, S. C., "Determinants of Corporate Borrowing," Jou]nal of Financial Economics, Novem-
îer =977, =47-=76.
, "The Capital Structure Puzzle," Jou]nal of Finance, July =984, P7P-P9<.
Pye, G., "The Value of a Call Option on a Bond," Jou]nal of Political Economy, April =966,
<00-<0P.
Roîîins, E. H., and J. D. Schatzîerg, "Callaîle Bonds: A Risk-Reducing Signalling Mech-
anism," Jou]nal of Finance, Septemîer =986, 93P-949.
Roîichek, A., and S. Myers, "Proîlems in the Theory of Optimal Capital Structure," Jou]nal
of Financial and Quantitative Analysis, June =966, =-3P.
Ruîinstein, M. E., "A Mean-Variance Synthesis of Corporate Financial Theory," Jou]nal of
Finance, March =973, =67-=8=.
Samuelson, P. A., "Rational Theory of Warrant Pricing," Indust]ial Management Review,
Spring =96P, =3-3<.
Schneller, M., "Taxes and the Optimal Capital Structure of the Firm," Jou]nal of Finance,
March =980, ==9-=<7.
Schwartz, E. S., "The Valuation of Warrants: Implementing a New Approach," Jou]nal of
Financial Economics, January =977, 79-93.
Smith, C., and R. Stulz, "The Determinants of a Firm's Hedging Policies," Jou]nal of Financial
and Quantitative Analysis, Decemîer =98P, 39=-406.
<88 CAPITAL STRUCTURE AND THE COST OF CAPITAL: THEORY
In Chapter 13, we analyzed several aspects of the cost and value of financial instru-
ments. Another aspect of the valuation of securities is the sensitivity of their present
values to unexpected changes in interest rates. We can conceptualize this relationship
as the elasticity o1 the price of securities (particularly bonds) to interest rates. For
bonds the elasticity is the ratio of percentage changes in prices to percentage changes
in market rates of interest for a given coupon and face value.
A. DURATION
489
♦
where
I t = dollar value of coupon payment in period t,
F = dollar value of maturity payment,
T = maturity period.
Next we take the derivative of the bond price to the change in the market yield rate:
dB, - (7 +1)
= _1(1 + r) - — 21(1 + r) - — • • • — TI(1 + ]) ' ÿ TF(1 + ])'
d(1 + r)
I +
2/ TI TF 1.
(1 + r)2 (1 + ]) 3 ã • • (1 ã + + (1 ã +
[
Divide both sides by Bo and (1 + r):
dB0 /130 dBo(1 +
d(1 + r)/(1 + r) d(1 + ])B 0
2/ TI TF
+•••+
Bo Li + r (1 + r) 2
+ T
(1 + r)T (1 + r)
Let C represent the appropriate cash flows, and express the result in summation form:
T tC,
(1 + ])t
Di = ♦(A13.3)
Bo
This result can also be expressed as in Eq. (A13.3a):
T tit TF
t T
;1. (1 + ]) + (1 + ])
D (A13.3a)
It F
T
t= (1 + r)
t +
(1 + r)
From Eq. (A13.3a) we can see that duration is not the same as the time to maturity
of the payment stream. Unlike maturity, duration considers all cash flows and gives
some weight to their timing.
Thus duration is calculated as the weighted average of the lengths of time prior
to the last cash flows, by using the ratios of the present values of each coupon pay-
ment to the present value of the bond as the weights. It identifies the "actual" weighted
length of time needed to recover the current cost of the bond. For example, assume
a five-year $1000 bond has a coupon payment of $25 each six months (5% coupon
rate), pays $1000 at the end of the fifth year, and has a yield rate of 12%. The dura-
tion of the bond can be calculated as shown in Table A13.1.
DURATION 491
$6,508.0775
D= = 8.7662 (semiannual)
$742.4050-
= 4.3831 years.
R QR T(1 Q - QR)
D= (A13.4)
R-1 R T - - Q QR
The new terms in Eq. (A13.4) are
R = (1 r) = 1.06,
Q = (Fli t) = $1000/$25 = 40,
T = 10.
Only for zero coupon bonds is duration the same as maturity. For other payment
streams, duration is shorter than maturity. For a given par value of a bond, the higher
the coupon payments and the higher the yield to maturity, the shorter is duration.
The relationship between the bond price and its duration is more complicated.
For bonds selling at or above par, the duration increases with maturity, but at a de-
creasing rate, and is bounded by (r + p)/rp years, where r is the yield to maturity and
p is the number of times per year interest is paid and compounded. For discount
bonds, duration increases with maturity to a maximum point before it matures, and
then declines. For shorter-term bonds, the differences between duration and maturity
are small. However, as maturity increases, the differences will be substantially larger.
492 APPENDIX TO CHAPTER 13: DURATION AND OPTIMAL MATURITY
B. IMMUNIZATION
Next we calculate the price that would be received when the bond is sold at the
end of the five-year planning horizon. At the end of 5 years, or 10 periods, the bond
had 1.7 years, or 3.4 periods, remaining. Its value and the capital gain (loss) at 10%
are shown below:
Vb = $45(3.0572) + $1000(.8471)
= $137.57 + $847.10
= $984.67
Capital gain = $984.67 — $1000 = — $15.33.
Thus we see that the amount of the capital gain or loss is a negative $15. This is slightly
different from the $13 gain on interest because we have used approximations for a
duration of, e.g., 5 years when the exact duration was 5.17 years.
Next we can verify that a 9% yield will actually be achieved and that the current
market value of $1000 is immunized against the rise in the required market yield.
In the analysis we use
So we have
The example illustrates that by choosing the duration of the payment stream to
be equal to the length of the planning horizon, the present value of the payment stream
is immunized against changes in the market rate of interest, and the initial yield on
the stream is preserved.
Another example of immunization uses the total balance sheet position of a firm.
Consider a bank portfolio manager with the following initial position:
Initial Position
Assets Liabilities
Portfolio value $800,000 $800,000
Portfolio yield 12.0% 8.0%
Portfolio duration 8 yrs 3 yrs
494 APPENDIX TO CHAPTER 13: DURATION AND OPTIMAL MATURITY
With a 1% rise in yields, we can use Eq. (A13.1) to calculate the change in the equity
position of the bank. For the assets side we have
dr
dPa = — Da (— )Pa
r
.1
= 8 $800,000
(1.12)
= —$57,142.
For the liabilities (or claims on assets) side, we have
dP, = 3 .01
$800,000
(1.08)
= — $22,222.
So the decline in the value of assets exceeds the decline in the value of claims on assets
by $34,920. By changing the duration of the asset portfolio, the bank can be immu-
nized against a change in the interest rate levels. The required duration for the asset
portfolio is
. 1
dPa = D $800,000 = — $22,222.
a (1.12)
Solving, we have Da is 3.11, indicating the shorter duration required for immuniza-
tion of the portfolio.
The examples above illustrate the mechanics of immunization procedures. In
actual application immunization involves a wide range of assumptions in connection
with its use in immunizing bond portfolios. One assumption is parallel shifts in the
yield curve [Yawitz and Marshall, 1981]. The practical effects of this are small [Kolb
and Gay, 1982, 83].
Also, there may be multiple rate changes during the planning horizon. This prob-
lem is dealt with by rebalancing the portfolio to maintain a duration matching the
remaining life of the planning period. In addition, there are a wide variety of more
aggressive approaches to managing the bond portfolio developed under the concept
of contingent immunization [Leibowitz and Weinberger, 1981, 1982]. Contingent
immunization is a form of active portfolio management. It generally involves some
degree of rate anticipation reflecting the portfolio manager's judgments about the
future direction of interest rates. The procedures are too detailed to be covered in this
brief treatment.
C. APPLICATION OF DURATION TO
DEBT MATURITY STRUCTURE
Redington [1952] applied the duration concept to the analysis of investment decisions
made by insurance companies. He proposed an immunization rule under which the
weighted durations of asset and liability streams are made equal so that the firm's
REFERENCES TO APPENDIX 495
net worth is hedged against interest rate movements. Redington noted that even for
insurance companies the concept would be complex in its implementation. Grove
[1974] analyzed immunization in a portfolio-choice model in a framework of uncer-
tainty of income stream patterns and of interest rate changes.
Morris [1976a, 1976b] sought to apply the duration concept to the problem of
corporate debt maturity strategies. Here the problem is even more complex than that
faced by insurance companies. In theory insurance companies can formulate income
and payment streams that are highly predictable by investing in fixed income secu-
rities, with payment streams related to the operation of life expectancy tables (which
can be further hedged by combining death policies with annuity policies). For the
industrial firm the liability structure can be fixed by its corporate debt maturity pat-
tern. However, the income stream varies with the impact of the economy and com-
petition on the firm's revenues and costs.
Morris observes that when the covariance of interest costs with the firm's net
operating income is high, a short-term borrowing policy will reduce the variation of
net income even though it increases the uncertainty of interest costs in future periods.
Thus for a weighted asset life with long duration, immunization through the choice
of the duration of the debt structure is not necessarily the least risky maturity policy
because of the variability of the income streams from the assets. A shorter debt matu-
rity policy may decrease the uncertainty of net income derived from the asset when
there is a high covariance between net operating income and interest costs. If the
duration of the asset structure is short, immunization calls for a weighted maturity
of short-duration debt. But if long-term debt with a longer duration were employed,
and if interest rates were negatively correlated with the firm's net operating income,
a long-term borrowing policy could reduce the variance of net income. In addition,
the level of interest costs would be fixed and certain over the life of the debt. Thus the
concept of duration appears to have some potential for developing corporate debt
maturity strategies. But the problem is more complex than that encountered in man-
aging portfolios of financial assets and claims.
REFERENCES TO APPENDIX
Fisher, L., and R. L. Weil, "Coping with the Risk of Interest Rate Fluctuations: Returns to
Bondholders from Naive and Optimal Strategies," Journal of Business, October 1971,
111-118.
Grove, M. A., "On 'Duration' and the Optimal Maturity Structure of the Balance Sheet,"
Bell Journal, Autumn 1974, 696-709.
Hicks, J. R., Value and Capital. Clarendon Press, Oxford, 1946.
Hopewell, M. H., and G. G. Kaufman, "Bond Price Volatility and Term to Maturity: A Gen-
eralized Respecification," American Economic Review, September 1973, 749-753.
Hsia, C. C., and J. F. Weston, "Price Behavior of Deep Discount Bonds," Journal of Banking
and Finance, September 1981, 357-361.
Kolb, R. W., Interest Rate Futures: A Comprehensive Introduction. Dame, Richmond, Va.,
1982.
496 APPENDIX TO CHAPTER 13: DURATION AND OPTIMAL MATURITY
, and G. D. Gay, "Immunizing Bond Portfolios with Interest Rate Futures," Financial
Management, Summer 1982, 81-89.
Leibowitz, M. L., "Specialized Fixed Income Security Strategies," in Altman, ed., Financial
Handbook, 5th ed. Wiley, New York, 1981, section 19.
, and A. Weinberger, Contingent Immunization: A New Procedure for Structured Active
Management. Salomon, New York, 1981.
, Risk Control Procedures under Contingent Immunization. Salomon, New York, 1982.
Macaulay, F. R., Some Theoretical Problems Suggested by the Movements of Interest Rates,
Bond Yields, and Stock Prices in the U.S. since 1856. National Bureau of Economic Re-
search, New York, 1938.
Morris, J. R., "On Corporate Debt Maturity Strategies," Journal of Finance, March 1976a,
29 —37.
, "A Model for Corporate Debt Maturity Decisions," Journal of Financial and Quanti-
tative Analysis, September 1976b, 339-357.
Redington, F. M., "Review of the Principles of Life Office Valuations," Journal of the Institute
of Actuaries, 1952 (vol. 78, part 3, no. 350), 286-340.
Samuelson, P. A., "The Effect of Interest Rate Increases on the Banking System," American
Economic Review, March 1945, 16-27.
Weil, R. L., "Macaulay's Duration: An Appreciation," Journal of Business, October 1973,
589-592.
Yawitz, J. B., and W. J. Marshall, "The Shortcomings of Duration as a Risk Measure for Bonds,"
Journal of Financial Research, Summer 1981, 91-101.
14
One kind of evidence in favor of the traditional position is that
companies in various industry groups appear to use leverage as if there
is some optimal range appropriate to each group. While significant
intercompany differences in debt ratios exist within each industry,
the average use of leverage by broad industrial groups tends to follow
a consistent pattern over time.
E. Solomon, The Theory of Financial Management, Columbia
University Press, New York, 1963, p. 98.
Capital St]uctu]e-
Empi]ical Evidence and
Applications
A. INTRODUCTION
The theories presented in the previous chapter provide some fairly unsettling con-
clusions about capital structure. On one hand it is argued that capital structure has
no effect on the value of the firm (Modigliani and Miller [1958] or Miller [1971),
and on the other hand it is suggested that the firm carry 100% debt (Modigliani and
Miller [1963]). Neither result is consistent with what seem to be cross-sectional re-
gularities in the observed capital structures of U.S. firms. For example, the electric
utility and steel industries have high financial leverage, whereas service industries
like accounting firms or brokerage houses have almost no long-term debt.
The first part of this chapter looks at some possible explanations for why there
might be such a thing as an "optimal" capital structure that contains both debt and
equity. Recall that in the previous chapter we examined the argument that risky debt
(in a world without bankruptcy costs) may be the cause of an optimal capital struc-
ture and concluded that it has no effect. The first issue to be studied here is the
effect of bankruptcy costs. It turns out that if they are nontrivial, then it is possible
that an optimal capital structure can be obtained as the tax advantage of debt is
traded off against the likelihood of incurring bankruptcy costs. Second, we consider
497
498 CAPITAL STRUCTURE: EMPIRICAL EVIDENCE AND APPLICATIONS
the possibility that changes in financial structure may be interpreted by the market-
place as signals of the future health (or sickness) of the firm. Based on the assumption
that managers are better informed than outsiders, the signaling theories lead to a
"pecking" order concept of capital structure where retained earnings is preferred to
debt and debt is preferred to new equity. Third, we discuss the implications of option
pricing theory for the capital structure of the firm and for the valuation of risky debt.
Finally, we discuss agency costs as determinants of capital structure. Agency theory
is novel because it can explain optimal capital structure without relying on taxes or
bankruptcy costs.
Next, the discussion turns to the empirical evidence. The central issue is whether
or not the value of the firm is affected by changes in its debt-equity ratio. Important
related questions are: (1) Does the cost of equity increase as financial leverage does?
(2) Are bankruptcy costs really nontrivial? (3) Which theory of optimal capital structure
is best supported by the evidence? (4) How does the market interpret various leverage
changing signals?
The third part of the chapter examines some practical issues dealing with changes
in the liabilities side of the balance sheet. We discuss a wrong reason for carrying
no debt, and how to make the debt refunding decision.
(b)
Figure 14.1
Optimal capital structure in the presence of bankruptcy costs: (a) the cost of capital; (b) the
value of claims.
bondholders [the solid line, k b , in Fig. 14.1(a)] increases with leverage. This in turn
results in a "U-shaped" weighted average cost of capital (solid line WACC') and an
optimal capital structure, [B/(B + S)]*. The optimal ratio of debt to equity is deter-
mined by taking on increasing amounts of debt until the marginal gain from leverage
is equal to the marginal expected loss from bankruptcy costs. The optimal capital
structure minimizes the weighted average cost of capital and maximizes the value of
the firm.
An important question for the existence of optimal capital structure brought
about by the deadweight losses of bankruptcy is: Exactly how large are bankruptcy
costs? If they are trivial, some other explanation for optimal capital structure is
needed. Warner [1977b] collects data for 11 railroad bankruptcies that occurred
between 1933 and 1955. He measures only direct costs, such as lawyers' and ac-
countants' fees, other professional fees, and the value of managerial time spent in
administering the bankruptcy. He does not estimate indirect costs to creditors, such
as the opportunity cost of funds tied up during bankruptcy proceedings and losses
in asset value due to forced capital structure changes, or indirect costs to share-
holders, such as lost profits created by decreased sales in anticipation of bankruptcy
or from disruptions in production during reorganization.' The evidence indicates
that direct costs are trivial, averaging about 1% of the market value of the firm seven
years prior to bankruptcy, and rising to 5.3% of the market value immediately prior
to bankruptcy. Furthermore, direct costs as a percentage of value seem to decrease
as a function of the size of the bankrupt firm. This would suggest that the direct
1
Some of the agency costs that are discussed in section B.4 are closely related to bankruptcy and might
also be considered indirect costs. For example, see Titman [1984].
500 CAPITAL STRUCTURE: EMPIRICAL EVIDENCE AND APPLICATIONS
costs of bankruptcy are less important for the capital structure decisions of large
firms than of small firms. Although Warner's evidence is inconclusive because indirect
costs are not measured, it does suggest that direct bankruptcy costs may not be suffi-
ciently large to be important determinants of optimal leverage.
In a second paper, Warner [1977a] examines the effect of bankruptcy on the
market returns of 73 defaulted bonds of 20 separate railroads, each of which was in
bankruptcy at some time during the period from 1930 through 1955. None of the
railroads was actually liquidated, although there were mergers, spinoffs, and aban-
donments of operations. The following effects were observed:
• The systematic risk of bonds increased prior to filing for bankruptcy. A reason-
able explanation is that as the market value of equity, relative to bonds, falls
prior to bankruptcy, the bondholders' claim becomes more like that of the share-
holders of an all-equity firm, and hence their risk rises.
• After adjusting for risk, the performance of the bonds in the month of the
bankruptcy petition was significantly negative.
• Investors who purchased a portfolio of bonds on the date of the bankruptcy
petition appeared to earn significant risk-adjusted abnormal returns in the post-
bankruptcy period. One possible explanation is that the courts approved capital
structure simplifications that had been unanticipated and that benefited
bondholders.
It is hard to argue that conclusions based on a study of the railroad industry
can be generalized because federal regulations encourage the continued operation of
railroad service even though the firm is in bankruptcy. Nevertheless, if one views the
bondholders' position as being the residual claim at the time of bankruptcy, the
significant negative return to bondholders on the date of the bankruptcy petition is
evidence of nontrivial bankruptcy costs.
Evidence on indirect bankruptcy costs is provided by Altman [1984]. 2 Admitedly,
because indirect costs are opportunity costs (what might have happened in the absence
of bankruptcy proceedings), they are difficult to estimate. Altman provides an estimate
(for a sample of 19 firms, 12 retailers and 7 industrials, that went bankrupt between
1970 and 1978) that compares expected profits, computed from time series regressions,
with actual profits. The arithmetic average indirect bankruptcy costs were 8.1% of
firm value three years prior to bankruptcy and 10.5% the year of bankruptcy. A
second method uses unexpected earnings from analysts' forecasts for a sample of 7
firms that went bankrupt in the 1980-1982 interval. Average indirect bankruptcy
costs were 17.5% of value one year prior to bankruptcy. Although more research
needs to be done on this topic, Altman's evidence suggests that total bankruptcy
costs (direct and indirect) are sufficiently large to give credibility to a theory of optimal
capital structure based on the trade-off between gains from leverage-induced tax
shields and expected bankruptcy costs.
2
Kalaba, Langetieg, Rasakhoo, and Weinstein [1984] discuss a potentially useful methodology for esti-
mating the expected cost of bankruptcy from bond data but provide no empirical results.
POSSIBLE REASONS FOR AN "OPTIMAL" MIX OF DEBT AND EQUITY 501
2. Signaling Hypotheses
If we assume that financial markets are not fully aggregating in the sense that
market prices do not reflect all information, especially that which is not publicly
available, then it is possible that managers may elect to use financial policy decisions
to convey information to the market. Changes in capital structure are an obvious
candidate for a signaling device, and as we shall see in Chapter 15, so is dividend
policy.
The first application of signaling to finance theory has been put forth by Ross
[1977]. He suggests that implicit in the Miller-Modigliani irrelevancy proposition
is the assumption that the market knows the (random) return stream of the firm and
values this stream to set the value of the firm. What is valued in the market place,
however, is the perceived stream of returns for the firm. Putting the issue this way
raises the possibility that changes in the capital structure (or dividend payout) may
alter the market's perception. In the terminology of Modigliani and Miller, by chang-
ing its financial structure (or dividend payout) the firm alters its perceived risk class
even though the actual risk class remains unchanged.
Managers, as insiders who have monopolistic access to information about the
firm's expected cash flows, will choose to establish unambiguous signals about the
firm's future if they have the proper incentive to do so. To show how this incentive-
signaling process works, let us assume that managers are prohibited (perhaps by SEC
regulations) from trading in the securities of their own firm. This keeps them from
profiting by issuing false signals, such as announcing bad news and selling short even
though they know the firm will do well.
In a simple one-period model the manager's compensation, M, paid at the end
of the period may be expressed as
V, if Vi > D,
M = (1 + r)T 0 V0 + y i { vi _ 3 (14.1)
if Vi < D,
where
Yo' ))1 = positive weights,
r = the one-period interest rate,
V0, Vi = the current and future value of the firm,
D = the face value of debt,
C = a penalty paid if bankruptcy occurs, i.e., if V < D.
This result can be used to establish a signaling equilibrium if we further assume that
investors use D, the face value of debt, to tell them whether a firm is successful (type
A) or unsuccessful (type B). Assume that D* is the maximum amount of debt that
an unsuccessful firm can carry without going bankrupt. If D > D*, the market per-
ceives the firm to be successful, and vice versa. For the signaling equilibrium to be
established, (1) the signals must be unambiguous (i.e., when investors observe D > D*,
the firm is always type A), and (2) managers must have incentive to always give the
appropriate signal. If the end-of-period value of a successful type-A firm is Via and
502 CAPITAL STRUCTURE: EMPIRICAL EVIDENCE AND APPLICATIONS
is always greater than the value of an unsuccessful type-B firm, V„, then the com-
pensation of the management of a type-A firm is
Vl a
yo(1 r) + y 1 171a if D* < D < Via (tell the truth),
1+r
Ma = (14.2)
Vib
y o(1 + r) + Via if D < D* (lie).
r
+r
In order for management of a type-B firm to have incentive to signal that the firm
will be unsuccessful, the payoff from telling the truth must be greater than that
produced by telling lies. Mathematically,
Yo Via ))1(Vlb C) < YoVib 71V1b,
which can be rewritten as
o( Via Vi b < (14.4)
This condition says that management will give the correct signal if the marginal gain
from a false signal, Via — Vib, weighted by management's share, y o , is less than the
bankruptcy costs incurred by management, C, weighted by its share, )),.
The incentive-signaling approach suggests that management might choose real
financial variables such as financial leverage or dividend policy as the means of
sending unambiguous signals to the public about the future performance of the firm.
These signals cannot be mimicked by unsuccessful firms because such firms do not
have sufficient cash flow to back them up and because managers have incentives to
tell the truth. Without management incentives to signal truthfully there would be
no signaling equilibrium.
The concept is easily applied to dividend policy as well as to financial structure.
A firm that increases dividend payout is signaling that it has expected future cash
flows that are sufficiently large to meet debt payments and dividend payments without
increasing the probability of bankruptcy. Therefore we may expect to find empirical
evidence that the value of the firm increases, because dividends are taken as signals
that the firm is expected to have permanently higher future cash flows. Chapter 16
reviews the empirical evidence relevant to dividend policy.
Ross's paper suggests that greater financial leverage can be used by managers
to signal an optimistic future for the firm. Another signaling paper, by Leland and
POSSIBLE REASONS FOR AN "OPTIMAL" MIX OF DEBT AND EQUITY 503
Pyle [1977], focuses on owners instead of managers. They assume that entrepreneurs
have better information about the expected value of their venture projects than do
outsiders. The inside information held by an entrepreneur can be transferred to
suppliers of capital because it is in the owner's interest to invest a greater fraction
of his or her wealth in successful projects. Thus the owner's willingness to invest in
his or her own projects can serve as a signal of project quality, and the value of the
firm increases with the percentage of the equity held by the entrepreneur relative to
what he or she otherwise would have held given a lower-quality project. An empirical
implication of this signaling argument is that if the original founders of a company
going public decide to keep a large fraction of the stock, then these firms should
experience greater price earnings multiples. A second implication is that if the firm's
value is positively related to the fraction of the owner's wealth held as equity in the
firm, then the firm will have greater debt capacity and will use greater amounts of
debt. Although debt is not a signal in this model, its use will be positively correlated
with the firm's value.
Myers and Majluf [1984] present a signaling model that combines investment
and financing decisions and that is rich in empirical implications. Managers, better
than anyone else, are assumed to know the "true" future value of the firm and of
any projects that it might undertake. Furthermore, they are assumed to act in the
interest of "old" shareholders, i.e., those who hold shares in the firm at the time a
decision is made. Finally, "old" shareholders are assumed to be passive in the sense
that they do not actively change their personal portfolios to undo the decisions of
management.' To keep things simple, assume that interest rates are zero and that
there are no taxes, transactions costs, or other market imperfections.
To begin the analysis, consider a situation where there are two equally likely
states of nature (good news and bad news). The firm has liquid assets, L i, and tangible
assets in place, A1, that can take the values illustrated in Table 14.1. It has no positive
net present value projects for the time being. (We shall examine the effect of positive
NPV projects next.) Also, there is no debt (that will be the third case.) Information
asymmetry is created by the fact that insiders are assumed to know which state, good
or bad, will turn up for the firm. The market, however, knows nothing except what
the value of the firm would be in each state of nature. If the firm does nothing, the
market (i.e., outsiders) will compute the current value of the firm as the expected value
of its payouts,
V, = E p1(L 1 + = .5(250) + .5(130) = 190.
This is equal to the value of the "old" shareholders' claim.
To establish a rational expectations signaling equilibrium let us look at the pay-
offs to "old" shareholders in each state of nature given each of two possible actions:
(1) do nothing or (2) issue $100 of new equity to new shareholders. We will see that
although "old" shareholders have the incentive to issue new shares when the firm is
overvalued i.e., when they know the bad news is coming the very fact that they try
to issue shares will signal their information to the market and consequently destroy
their informational advantage. If "old" shareholders know good news (state 1) will
occur, their wealth conditional on doing nothing is
(Vo lgood news, do nothing) = L 1 + A l = 250.
Alternately, they can issue $100 of new equity, E, and their value is
V, 190
(Vo lgood news, issue equity) — (L 1 + A l E) = (350) = 229.31.
V, + E 290
Their fraction of the firm if they issue new equity is their current value, 190, divided
by 190 plus the cash received from the new equity issue, 100. If "old" shareholders
know bad news (state 2) will occur, their payoff from doing nothing is
(V0 lbad news, do nothing) = L 2 + A 2 = 130,
and if they issue new equity it is
190
( V,Ibad news, issue equity) = V0 (L 2 + A 2 + E) = (230) = 150.69.
V, + E 290
The payouts to original shareholders are summarized in Table 14.2. It seems that
the optimal actions of the informed "old" shareholders (i.e., the payouts with asterisks)
are to do nothing if they think the good news state will occur and to issue equity if
the bad news state will occur, because the firm is currently overvalued. Outsiders,
however, will not be fooled. When the firm issues new equity they know the firm
believes the bad news state will occur, and they impute the bad news value, 130, to
the firm. Therefore the expected payout to old shareholders, given that they issue
new equity and that the outsiders infer bad news, is
V2
(V0 issue equity) = E (V2 + E) = V2 = 130.
Invest and
Do Nothing Issue Equity
Liquid assets, L, 50 50 50 50
Assets in place, A, 200 80 300 180
NPV of new project, b, 0 0 20 10
Value of firm, V 250 130 370 240
The upshot of this argument is that original shareholders cannot take advantage of
their inside information because the very act of issuing new shares (when they think
the firm is overvalued) will reveal their information to the market. Hence they are
indifferent between doing nothing and issuing new equity, and the market will attach
no significance to new equity issues.
Next, let us complicate the model slightly by assuming that the firm has a positive
net present value project that requires an initial cash outlay of $100, and that has
the state-dependent net present values, b 1, illustrated in Table 14.3. Going through
the same type of computations as before, we first compute V0 , the unconditional value
of original shareholders' wealth, assuming they do nothing.
V0 = E pi(Li + A 1) = .5(250) + .5(130) = 190.
Alternately, if they issue and invest, their unconditional expected wealth is
V', = E p,(L, + A, + b 1) = .5(270) + .5(140) = 205.
Now let us examine their wealth, contingent on each state of nature. If they issue
$100 of new equity and invest the proceeds in the new positive NPV project, their
wealth in the good news state is
are better off doing nothing in the good state because the positive NPV of the pro-
ject (given good news) is not large enough to offset the fraction of ownership that
they must sacrifice by issuing new shares.' Hence they desire to issue new equity and
invest only if they know the bad state will occur. As before, the market is not fooled.
As soon as insiders announce their intention to issue and invest, the market learns
that the bad state is forthcoming, and in the bad state the firm is worth only 240,
with 100 going to outsiders and the remaining 140 going to original shareholders.
The rational expectations equilibrium payoffs are illustrated in Table 14.5. In equilib-
rium, given the set of numbers we have chosen, the firm issues and invests in the
bad news state but not in the good news state.' This surprising result implies that
the value of the firm may fall when new equity issues are announced an important
empirical implication.
So far we have examined two cases. First, when the firm had no new projects
and the market knew it, then issuing new equity was an unambiguous financial signal
that the market could use to discover the inside information held by managers.
Hence it was impossible for managers to benefit from issuing new equity when they
knew the future prospects of the firm were dismal. Second, when positive NPV pro-
jects (good news) were financed with equity issues (bad news), the signal became
mixed. The market could not separate information about new projects from infor-
mation about whether the firm is under- or overvalued. If there were some way to
provide two separate signals—one for investment decisions and another for financing
decisions—the problem would vanish. If project outcomes were uncorrelated with
states of nature (e.g., if the project had the same NPV in both states of nature), the
problem would vanish. Or if the firm were to use financing that is not subject to the
information asymmetry problem, the problem would vanish.
It is important to realize that outsiders pay nothing for the expected NPV of the new project. The
entire NPV accrues to "old" shareholders.
It is puzzling to understand why old shareholders do not provide all investment funds if they know
the good state will occur. These funds can come from cash (what Myers and Majluf call slack) or via
a rights offering.
POSSIBLE REASONS FOR AN "OPTIMAL" MIX OF DEBT AND EQUITY 507
Myers and Majluf point out that if the firm uses its available liquid assets, L1,
to finance positive NPV projects, then all positive NPV projects would be under-
taken because no new equity is issued and the information asymmetry problem is
thereby resolved. They suggest that this may be a good reason for carrying excess
liquid assets. They also suggest that debt financing, which has payoffs less correlated
with future states of nature than equity, will be preferred to new equity as a means
of financing. Myers [1984] suggests a pecking o]de] theo]y for capital structure. Firms
are said to prefer retained earnings (available liquid assets) as their main source of
funds for investment. Next in order of preference is debt, and last comes external
equity financing. Firms wish to avoid issuing common stock or other risky securities
so that they do not run into the dilemma of either passing up positive NPV projects
or issuing stock at a price they think is too low.
The pecking order theory is a dynamic story. The observed capital structure of
each firm will depend on its history. For example, an unusually profitable firm in
an industry with relatively slow growth (few investment opportunities) will end up
with an unusually low debt-to-equity ratio. It has no incentive to issue debt and
retire equity. An unprofitable firm in the same industry will end up with a high debt
ratio.
6
All accumulated interest on zero coupon bonds is paid at maturity; hence B(T), the current market value
of debt with maturity T, must be less than its face value, D, assuming a positive risk-free rate of discount.
508 CAPITAL STRUCTURE: EMPIRICAL EVIDENCE AND APPLICATIONS
call option,
S = MAX[0, V — D],
and bondholders' wealth is
B = MIN[V , D].
The option pricing model offers a great deal of insight into the way that capital
structure changes may affect shareholders and bondholders. Although the arguments
that follow are made intuitively, the reader who is interested in a mathematical
presentation is referred to Galai and Masulis [1976].
In each of the following cases we rely on the Modigliani-Miller result that in
the absence of transactions costs, information heterogeneity, or taxes the value of the
firm remains constant regardless of the financial decisions made by management.
Furthermore, we assume that any changes that affect the systematic risk of various
securities, or their expected rate of return, are unanticipated changes. To the extent
that changes in the value of securities are unanticipated, it is possible that there may
be a redistribution of wealth from one class of security holders to another.
We also assume that two-fund separation does not apply. Two-fund separation
implies, among other things, that all individuals hold the same portfolio of risky
assets, namely, the market portfolio. For individuals holding both the equity and
risky debt of a firm, any offsetting change in the market value of the debt and equity
claims against the firm will not change their wealth position. Therefore they would
be indifferent to the redistribution effects that we are about to discuss. It is necessary,
then, to rule out two-fund separation and discuss the wealth of shareholders and
bondholders as if they were separate and distinct. If shareholders are not constrained
by the indenture provisions of debt from issuing new debt with an equal claim on
the assets of the firm, then current bondholders will experience a loss of wealth when
new debt is issued. It is possible to increase the book value debt-to-equity ratio by
issuing new debt and using the proceeds to repurchase equity. In this way the assets
of the firm remain unchanged. If the new debt has equal claim on those assets, then
the current bondholders end up with only a partial claim to the assets of the firm,
whereas before the new debt was issued, they had complete claim on the assets.
Clearly, this approach puts current bondholders in a riskier position, and they are
unable to charge more for the extra risk because the discounted value of their bonds
has already been paid (i.e., they cannot raise their coupon payments once the bonds
have been issued). Consequently, the market value of their bonds will fall. At the
same time, the value of the firm remains unchanged, and new bondholders pay a
fair market price for their position. Therefore the value that is expropriated from
current bondholders must accrue to shareholders, who are the residual claimants of
the firm. Their wealth increases. This is called the bondholder wealth expropriation
hypothesis.
The theory of option pricing argues that in a world with no transactions costs
or taxes the wealth of shareholders is increased by greater financial leverage. In
Chapter 13, the Modigliani-Miller propositions argue that under the same set of
POSSIBLE REASONS FOR AN "OPTIMAL" MIX OF DEBT AND EQUITY 509
.5 $ 9,000 $ 2,000
.5 11,000 18,000
510 CAPITAL STRUCTURE: EMPIRICAL EVIDENCE AND APPLICATIONS
Figure 14.2
Optimal capital structure determined
by minimizing total agency costs.
Total agency
costs
■
such covenants may well be nontrivial. Debt holders must charge higher ex ante
yields to compensate them for possible wealth expropriation by shareholders. Further-
more, these costs may increase with the percentage of financing supplied by bond-
holders as illustrated in Fig. 14.2.
On the other hand, there are agency costs associated with external equity. Sup-
pose we begin with a firm owned exclusively by a single individual, the owner-
manager (0-M). The O-M will obviously take every action possible to increase his
or her own wealth. For example, if he or she decides to take Wednesday afternoon
off, then as owner-manager he or she bears the full cost of doing so. However, if
the O-M sells a portion of the ownership rights by selling external equity to new
shareholders, there will arise conflicts in interest. Now the O-M is co-owner with
the new shareholders. If the O-M can maximize his or her wealth at the expense
of the new shareholders (e.g., by purchasing an executive jet and taking long vaca-
tions), then he or she will do so. Co-ownership, of equity implies agency problems.
The new shareholders will have to incur monitoring costs of one form or another in
order to ensure that the original owner-manager acts in their interest. It is assumed,
as illustrated in Fig. 14.2, that the agency costs of external equity increase as the
percentage of financing supplied by external equity goes up. The agency costs of
external equity may be reduced if the management and shareholders agree to hire
an independent auditor. For an interesting exposition of this idea the reader is re-
ferred to an article by Watts and Zimmerman [1979].
Jensen and Meckling suggest that, given increasing agency costs with higher
proportions of equity on the one hand and higher proportions of debt on the other,
there is an optimum combination of outside debt and equity that will be chosen
because it minimizes total agency costs. In this way it is possible to argue for the
existence of an optimum capital structure even in a world without taxes or bank-
ruptcy costs.
Figure 14.2 illustrates the Jensen-Meckling argument for an optimal capital
structure based on the agency costs of external equity and debt (in a world without
taxes). Agency costs of external equity are assumed to decrease as the percentage
of external equity decreases, and the agency costs of debt are assumed to increase.
Figure 14.2 illustrates a case where total agency costs are minimized with an optimal
capital structure between 0% and 100%—an interior solution. If the agency costs of
POSSIBLE REASONS FOR AN "OPTIMAL" MIX OF DEBT AND EQUITY 511
external equity are very low, as may be the case for a widely held firm, then op-
timal capital structure can result as a trade-off between the tax shelter benefit of
debt and its agency cost.
The discussion of agency costs need not be limited to costs associated with pro-
viding debt and equity capital. For example, Titman [1984] suggests that agency
costs are important for contracts (whether implied or explicit) between the firm and
its customers or between the firm and its employees. Consider the relationship be-
tween a firm and its customers. If the firm's product is a durable good and requires
future services such as parts and repair, the customer is paying not only for owner-
ship of the product but also for the availability of an expected future stream of
services. If the firm goes bankrupt, its customers lose their anticipated services without
any hope of being compensated. Consequently, they must assess the probability of
bankruptcy and weigh it in their decision to purchase durable goods. Firms that
produce durable goods will have lower demand for their products if they increase
their probability of bankruptcy by carrying more debt. This is an example of an
indirect cost of bankruptcy. Ceteris paribus, we would expect durable goods pro-
ducers to carry less debt.' Agency costs in labor contracts are also important. If a
firm's labor force has acquired specialized skills that cannot easily be transferred to
alternate employment, then laborers bear nontrivial costs if a firm goes bankrupt.
They have to search for new jobs and learn new skills. There is no hope that the
bankrupt firm will compensate them for their loss. Consequently, if labor markets
are competitive, then laborers will charge lower wages to work for a firm that has a
lower probability of bankruptcy. Thus we should expect to find that firms that use a
larger percentage of job-specific human capital will also tend to carry less debt,
ceteris paribus.
An exception might be regulated firms. They can carry more debt because regulatory commissions are
expected to "guarantee" a reasonable rate of return. Consequently, bankruptcy is very unlikely.
512 CAPITAL STRUCTURE: EMPIRICAL EVIDENCE AND APPLICATIONS
monitoring costs. For example, many bonds require periodic payments of both prin-
cipal and interest and are fully amortized. This is a crude way of tying the market
value of the debt to the projected liquidation (or resale) value of the underlying assets.
An even more extreme method for reducing agency costs, one commonly employed
by small firms, is leasing. Leased assets are fully secured because they are literally
the property of the lessor and can be repossessed in the event of default on the
lease payments.
Stulz and Johnson [1985] prove that secured debt actually increases the value
of the firm, rather than merely redistributing wealth among various claim holders.
There are two reasons. First, secured debt can decrease debt holders' monitoring
costs because the firm cannot sell the collateral to pay dividends, because the col-
lateral cannot be exchanged for a more risky asset, and because secured creditors
are less likely to require restrictive covenants about what the firm can or cannot
do later on. The second, and perhaps more important, reason is that secured debt
makes it more advantageous for shareholders to undertake positive net present value
projects. Thus secured debt will be generally preferred to unsecured debt, a result
consistent with the Myers-Majluf [1984] story.
income and interest charges, or current assets and current liabilities (working capital
tests). There may also be "cleanup" provisions that require the company to be debt
free for limited periods.
If there is any advantage to the firm that holds debt in its capital structure,
bondholders can benefit by allowing new debt, but only under the condition that
acquiring this obligation does not increase the riskiness of their position. Hence an
outright prohibition of new debt under any condition is rare.
Other techniques that are used to protect bondholders against subsequent finan-
cing include restrictions on rentals, leases, and sale-leaseback agreements; sinking fund
requirements (which roughly match the depreciation of the firm's tangible assets);
required purchase of insurance; required financial reports and specification of ac-
counting techniques; and required certifications of compliance by the officers of the
firm.
Bond covenants that restrict dividend payments are necessary if for no other
reason than to prohibit the extreme case of shareholders voting to pay themselves
a liquidating dividend that would leave the bondholders holding an empty corporate
shell. Kalay [1979] reported that in a random sample of 150 firms every firm had a
dividend restriction in at least one of its debt instruments. Restrictions on dividend
policy are relatively easy to monitor, and they protect debt holders against the un-
warranted payout of the assets that serve as collateral. Appropriately, most indentures
refer not only to cash dividends but to all distributions in respect to capital stock,
whether they be dividends, redemptions, purchases, retirements, partial liquidations,
or capital reductions, and whether in cash, in kind, or in the form of debt obliga-
tions to the company. Without such general provisions the firm could, e.g., use cash
to repurchase its own shares. From the bondholders' point of view, the effect would
be the same as payment of cash dividends.' No matter what the procedure is called,
once cash is paid out to shareholders, it is no longer available for collateral in the
event of reorganization or bankruptcy.
Most restrictions on the payout of the firm's assets require that dividends increase
only if the firm's earnings are positive, if the firm issues new equity capital, or if
dividends paid out since the bonds were issued have been kept below a predefined
minimum level. Mathematically, the "inventory" of funds allowable for dividend pay-
ment, DivI, in quarter T, can be expressed as
T-1
E NI, + E St + F E
T T
DivT, = K — Dive,
t= o t=o t=o
where,
NI = net earnings in quarter t,
K = predetermined constant, 0 < K < 1,
St = net proceeds from issue of new equity,
F = number fixed over life of bonds, known as the "dip,"
Dive = dividends paid out in quarter t.
8 See Chapter 16 for empirical evidence regarding the effect of share repurchases on the wealth of bond-
holders and shareholders.
514 CAPITAL STRUCTURE: EMPIRICAL EVIDENCE AND APPLICATIONS
Thus the dividend covenant does not restrict dividends per se; rather, it restricts the
financing of the payment of dividends with new debt or by sale of the firm's existing
assets. This arrangement is in the interest of stockholders because it does not restrict
the payment of earned income. It is also in the interest of bondholders because it
prevents any dilution of their claim on the firm's assets.
Bond covenants that restrict merger activity prohibit many mergers. More
often, though, they will allow mergers, provided certain conditions are met. The
effect of a merger on bondholders can be beneficial if the cash flows of the merged
firms are not perfectly correlated. Offsetting cash flow patterns can reduce the risk
of default, thereby bettering the positions of the bondholders of both firms. Merger
can also be detrimental to bondholders. For example, if firm A has much more debt
in its capital structure than firm B, the bondholders of B will suffer increased risk
after the merger.' Or if the maturity of debt in firm A is shorter than for firm B,
the bondholders of B will (for all practical purposes) become subordinate to those
of firm A after the merger.
To protect against the undesirable effects that can result from a merger, it is
possible to require bond covenants that allow merger only if the net tangible assets
of the firm, calculated on a postmerger basis, meet a certain dollar minimum or are
at least a certain fraction of long-term debt. The merger can also be made contingent
on the absence of default of any indenture provision after the transaction is com-
pleted.
Bond covenants that restrict production or investment policies are numerous.
They are frequently difficult to enforce, however, given the impossibility of effectively
monitoring the investment decisions that the managers of the firm decide not to
undertake. Myers [1977] suggests that a substantial portion of the value of a firm
is composed of intangible assets in the form of future investment opportunities. A
firm with outstanding debt may have the incentive to reject projects that have a
positive net present value if the benefit from accepting the project accrues to the
bondholders without also increasing shareholders' wealth.
Direct restrictions on investment-disinvestment policy take the following forms:
(1) restrictions on common stock investments, loans, extensions of credit, and ad-
vances that cause the firm to become a claim holder in another business enterprise,
(2) restrictions on the disposition of assets, and (3) covenants requiring the mainte-
nance of assets. Secured debt is an indirect restriction on investment policy. Assets
that provide surety cannot be disposed of under the provisions of the indenture
agreement. Collateralization also reduces foreclosure expenses because the lender
already has established title via the bond covenant.
Even though covenants are designed to protect bondholders from various actions
that can diminish the surety of their position, no set of covenants can eliminate all
risk. Consequently, there is considerable interest in accurate information about
changes in the riskiness of corporate debt on a firm-by-firm basis.
The usual method for determining the default risk of corporate long-term debt
is to refer to the bond ratings supplied by various agencies. Major bond rating
9
For a more complete exposition, see Shastri [1981].
POSSIBLE REASONS FOR AN "OPTIMAL" MIX OF DEBT AND EQUITY 515
agencies are Moody's Investors Service, Inc., Standard & Poor's Corp., and Fitch
Investor Service. Moody's bond rating has seven classifications, ranging from Aaa,
which is the highest quality bond, down to Caa, the lowest quality. Weinstein [1977]
collected data on 179 new bond issues between 1962 and 1964. Table 14.7 shows
the distribution by risk class. About 40% of the new bonds qualified for the two
highest quality ratings. Figure 14.3 shows the yields on bonds of different risk. Just
as expected, the high-quality, low-risk bonds have lower promised yields than do
the low-quality, high-risk bonds. A common-sense way of estimating the marginal
cost of new debt for a firm (assuming that the new debt will not change the firm's
bond rating) is to compute the yield to maturity on other bonds with maturities
and bond ratings similar to the new issue.
20 Baa
A
Aaa
18
16
1
I
15 I
'i I A
\
14 ,/ \A I
. \ \
, i \
7.. 1 3 /
1 c.v
,....\ /-.
9
8
7
75 76 77 78 79 80 84
Bi-monthly beginning January 1975
Figure 14.3
Comparison of bond yields for bonds of different risks. (From Moody's Bond Record,
various issues.)
516 CAPITAL STRUCTURE: EMPIRICAL EVIDENCE AND APPLICATIONS
Of the roughly 2000 major corporations that are evaluated by the agencies,
approximately 500 are rerated quarterly because they issue commercial paper, an-
other 500 are rerated annually (most of the utilities), and the remaining 1000 have
no established review date but are usually reviewed annually.
From an investor's point of view, one might ask the following question: Do the
agencies determine the prices and interest rates paid for bonds or do investors in the
capital markets? The evidence collected by Wakeman [1978] and Weinstein [1978]
shows that changes in bond ratings are not treated as new information by capital
markets. In fact, changes in ratings usually occur several months after the capital
markets have already reacted to the fundamental change in the bond's quality.
Changes in agency ratings do not cause changes in required yields to maturity. It is
the other way around. However, this does not imply that bond ratings are without
value. On average the ratings provide unbiased estimates of bond risk and are there-
fore a useful source of information.
C. EMPIRICAL EVIDENCE
ON CAPITAL STRUCTURE
1. Cross-Sectional Studies
Modigliani and Miller [1958] use cross-section equations on data taken from
43 electric utilities during 1947-1948 and 42 oil companies during 1953. They esti-
mate the weighted average cost of capital (WACC) as net operating cash flows after
taxes divided by the market value of the firm." When regressed against financial
leverage (measured as the ratio of the market value of debt to the market value of
the firm), the results were:"
where d is the financial leverage of the firm and r is the correlation coefficient. These
results suggest that the cost of capital is not affected by capital structure and there-
fore that there is no gain to leverage.
Weston [1963] criticizes the Modigliani-Miller results on two counts. First, the
oil industry is not even approximately homogeneous in business risk (operating
leverage); second, the valuation model from which the cost of capital is derived
assumes that cash flows are perpetuities that do not grow. When growth is added
to the cross-section regression, the result for electric utilities becomes
where A is the book value of assets (a proxy for firm size), and E is the compound
growth in earnings per share (1949-1959). Since WACC decreases with leverage,
Weston's results are consistent with the existence of a gain to leverage, i.e., that the
tax shield on debt has value.
Later on, Miller and Modigliani [1966] also found results (based on a sample
of 63 electric utility firms in 1954, 1956, and 1957) that were consistent with a gain
from leverage. Table 14.8 summarizes their results. The value of the firm is attributed
to the present value of the operating cash flows generated by assets in place, by the
tax subsidy on debt, by growth potential, and by firm size. For our purposes the im-
portant result is that the empirical evidence indicates that the tax subsidy (i.e., the
gain from leverage) on debt does contribute a significant amount to the value of
the firm, about 26% on average. This is consistent with the notion that the firm's
WACC falls as leverage increases.
10
Net operating cash flows after taxes were actually estimated as net income after taxes plus interest
payments on debt. This assumes that there is no growth in earnings and that replacement investment
equals depreciation expense.
11
Standard errors are in parentheses.
518 CAPITAL STRUCTURE: EMPIRICAL EVIDENCE AND APPLICATIONS
From M. Miller and F. Modigliani, "Some Estimates of the Cost of Capital to the Electric Utility Industry,
1954-57," American Economic Review, June 1966, 373. Reprinted by permission of the authors.
Cordes and Sheffrin [1983] use Treasury Department data to examine cross-
sectional differences in effective tax rates that may be caused by tax carry-backs and
carry-forwards, by foreign tax credits, by investment tax credits, by the alternate tax
on capital gains, and by the minimum tax. They found significant differences across
industries with the highest effective rate for tobacco manufacturing (45%) and the
lowest rate (16%) for transportation and agriculture. This tends to support the
DeAngelo-Masulis [1980] contention that the gain from leverage-induced tax shields
can be positive.
An important part of the Modigliani-Miller theory is that the cost of equity
capital increases with higher leverage. Hamada [1972] tests this proposition empiri-
cally by combining the Modigliani-Miller theory and the CAPM. He finds that on
the average the systematic risk of the levered firm is greater than that for the unle-
vered firm:
.91, fiu _ . 70.
This, of course, is consistent with the increased risk associated with higher leverage.
However, in order to construct the return on equity for an unlevered firm, Hamada
had to assume that the Modigliani-Miller theory was correct. Suppose that it is not
correct. Namely, what would happen if the return on equity (i.e., the cost of equity
capital) did not increase with increasing leverage? We would expect that for a sample
of firms with the same operating risk there would be no increase in systematic risk
with higher financial leverage. Because it is almost impossible to find firms with
identical operating risk, Hamada suggests that within an industry if the fl u-values of
individual firms are closer or less scattered than their /3L-values, then the Modigliani-
Miller theory would be supported. Greater variability in the flu-values implies that
the cost of equity changes with financial leverage. In nine industries examined, /3L
u
was greater than fl u in all cases, and the standard deviation of the fl -values was
u
greater than eight out of nine of the fl -values. This may be taken as indirect evidence
that the cost of equity increases with higher financial leverage.
Recent cross-sectional work has been done by Bradley, Jarrell, and Kim [1984],
Long and Malitz [1985], and Titman and Wessels [1985]. Bradley, Jarrell, and Kim
EMPIRICAL EVIDENCE ON CAPITAL STRUCTURE 519
regressed leverage against (1) earnings volatility as a proxy for bankruptcy risk, (2)
the ratio of depreciation plus investment tax credits to earnings as a proxy for
non-debt tax shields, and (3) the ratio of advertising plus research and development
expenditures to net sales as a proxy for noncollateralizable assets. The first and third
variables were significantly negative, supporting the importance of bankruptcy costs
and collateral; but the second variable was significantly positive, seeming to be
inconsistent with debt as a tax shield. Long and Malitz estimate a similar regression
but add several additional variables. They obtain results similar to Bradley, Jarrell,
and Kim but find non-debt tax shields to be negatively related to leverage (although
not significant).
Titman and Wessels employ linear structural modeling to explicitly accommo-
date explanatory variables as proxies for their theoretical counterparts. Their results
show that asset uniqueness and profitability were significantly negatively related to
leverage. This result supports the Myers-Majluf [1984] pecking order theory, because
more profitable firms will tend to use less external financing. It also supports the
Titman [1984] idea that firms with unique assets can carry less debt owing to agency
costs.
12
Returns for leverage-decreasing offers were multiplied by —1.0 and added to the returns of leverage-
increasing offers.
EMPIRICAL EVIDENCE ON CAPITAL STRUCTURE P<=
Leverage effect
Figure 14.4
Two dimensions of announcement effects.
COST OF CAPITAL: APPLICATIONS 523
is the new issue of debt securities where Dann and Mikkelson [1984], Eckbo [1986],
and Mikkelson and Partch [1986] found negative but insignificant announcement
effects. This result is also consistent with the pecking order theory. The majority of
events with no leverage change had insignificant announcement effects.
Announcements with favorable (unfavorable) implications for the future cash
flows of the firm such as investment increases (decreases) and dividend increases
(decreases) were accompanied by significant positive (negative) effects on sharehold-
ers' wealth.
With these results in mind, it is tempting to try to place each type of corporate
event into the two dimensions of Fig. 14.4. Not all events fit neatly. Yet there does
seem to be a convincing pattern. Events that both increase leverage and provide a
favorable signal about the future prospects of the firm, common share repurchases,
and debt-for-common exchange offers seem to have the largest positive announce-
ment effects.
Even when one is very familiar with the theoretical concept of the cost of capital, it
is not a straightforward or easy task to apply the theory to practice. Too often one
is confronted with questions not made explicit because theoretical expositions are
deliberately oversimplified. For example, how do unusual liabilities such as accruals,
preferred debt, convertibles, or accounts payable affect the cost of capital? How should
the market value weights of various sources of capital be estimated? How can one
tell whether or not the firm is at its long-run target capital structure? This section
further develops some of the cost of capital example calculations given in Chapter
13. First, however, a simple example shows how firms with high price/earnings ratios
may choose to carry no debt for the wrong reason. Second, a detailed cost of capital
calculation is given for Bethlehem Steel. Third, the debt refunding and defeasance
decisions are analyzed. And finally, the effect of flotation costs on the cost of capital
is discussed.
the price per share is $50. If 10 new shares are issued at $50 each, $500 is collected
and used to retire $500 of debt (which pays a coupon rate of 8%). After the capital
structure change, earnings per share have increased to $1.50 per share (since there
are now 20 shares outstanding), and with a price/earnings ratio of 50, presumably
the price per share will increase from $50 before the capital structure change to $75
afterward. The pro forma income statements and balance sheets are shown in Table
14.10.
The above example shows that if a firm with a high price/earnings ratio seeks
to maximize earnings per share, it will carry little or no debt. The problem of course
is that shareholders care about net cash flows (and the gain from leverage), not earn-
ings per share. In the above example, the change in operating cash flows is zero.
Before debt is retired, cash flow (neglecting depreciation, which is not given anyway)
is equal to net income, $10, plus interest on debt after taxes (1 — r e) A(kdD), $40, or a
total of $50.' After the swap of stock for bonds the cash flow is still $50. If there
were no gain from leverage, the value of the firm would be unchanged. However,
because there is a leverage effect in a world with corporate taxes, the value of the
firm will actually fall, because leverage declines.
Using the Modigliani-Miller valuation model, Eq. (13.3), we can calculate the de-
crease in the value of the firm. The value of the firm before the change is
E(I\10I)(1 — -cc) + (1kp
Vi = , (13.3)
13 This method for computing cash flows is covered in Chapter 2, footnote 15.
COST OF CAPITAL: APPLICATIONS 525
where
17' = the market value of debt, B, plus the market value of equity, S,
E(NOI) = the expected operating income, $100,
= the marginal corporate tax rate, 50%,
p = the cost of equity for an all-equity firm,
k, = the coupon rate on bonds, 8%,
k b = the market rate on new debt, 8%,
D = the book value of debt, $1000.
Using the numbers in the Betamax example, we have
100(1 — .5) [.08(1000)1
500 + 1000 = .5.
P .08
Solving for p, the cost of equity for the unlevered firm, we have'
p = 5%.
We can use the fact that p does not change when capital structure changes to deter-
mine the value of the firm after the repurchase of debt. Substituting the new, lower
amount of debt into Eq. (13.3), we have
17' = 100(1 — .5) [.08(500)1
.5 = $1250.
.05 .08
Therefore the value of the firm has fallen from $1500 to $1250. Since 17' = B + S,
the new value of equity is $1250 — 500 = $750.
We might also ask what the new price per share will be. It is common to assume
that new shares can be issued at the preswap price of $50. If this is the assumption,
then 10 new shares are issued, bringing the total number of shares outstanding to
20 and the implied price per share to $37.50.
Another (better) approach requires that we compute the new equilibrium price
per share in order to determine the number of shares that must be issued in order
to supply the $500 needed for the proposed exchange offer. We know that the value
of equity after the exchange, So + AS, is equal to the number of original shares,
n ° = 10, times the new price per share, P", plus the number of new shares nn times
the new price per share:
S o + AS = n° P" + nnPn = $750.
We also know that the number of new shares times the price per share, n"P", must
equal the dollar amount of bonds repurchased, AB" < 0. Therefore
—AB' = nnPn = $500.
14
The astute reader will recognize that p, the cost of equity for the unlevered firm, is less than k,„ the
before-tax cost of debt. This relationship is feasible in the presence of risky debt.
526 CAPITAL STRUCTURE: EMPIRICAL EVIDENCE AND APPLICATIONS
Combining these two expressions, we can solve for the price per share after the
exchange offer:
So + AS = n°13" + AY,
S o + AS — ABn pn,
$750 — $500
= $25 per share.
10
Hence the firm will need to issue 20 new shares at $25 each in order to raise the
$500 needed for the exchange. After the deal the total number of shares outstanding
rises to 30, the price per share will fall from $50 to $25, earnings per share is $1, not
$1.50, and the price/earnings ratio declines from 50 to 25.
where
p = the cost of equity for an all-equity firm,
= the marginal corporate tax rate,
AB = the market value of new debt,
Al = the replacement cost of new investment.
Given that the firm is at its long-run "optimal capital structure," we can say that an
equivalent definition of the weighted average cost of capital is the weighted average
of the marginal costs of various sources of capital, Eq. (13.19):
B
.
WACC = (1 — tc)kb B s + ks (13.19)
B +
To compute the weighted average cost of capital by means of Eq. (13.19), we must
know (1) the marginal cost of debt and equity, k b and ks, and (2) the market value
capital structure used by the firm.
Although we use the term weighted average cost of capital to mean the cost of a
mixture of sources of funds, it is important to emphasize that the costs of these funds
must be measured as marginal costs. Hence the weighted average cost of capital is a
weighted average of the marginal costs of the firm's various sources of capital. In this
context the word marginal has two meanings. Foremost is that marginal cost means
COST OF CAPITAL: APPLICATIONS 527
the cost of new financing at current market equilibrium rates of return—not historical
cost. Second, and implied in the rate of return required by the market, is the impact
of new financing on the perceived capital structure of the firm. For example, if the
percentage of debt is perceived to be rising, the marginal cost of debt must include
its impact on the weighted average cost of capital. We shall ignore this second pos-
sibility by adopting the convention that the firm establishes a target capital structure
and sticks with it. Given this assumption, Eqs. (13.12) and (13.19) are identical. Con-
sequently, there are no changes in leverage to complicate matters.
Some of the complexities that arise while we estimate the cost of capital are illus-
trated by the following example, the cost of capital for Bethlehem Steel Corporation
during December 1976. Because market-equilibrium conditions change from day to
day, so does the cost of capital. Therefore, the estimate that is given below is only
an historical number, valid at the end of 1976.
Bethlehem Steel is the second-largest producer of steel in the United States. It
manufactures steel products for markets in construction, transportation, service
centers, and machinery. It also produces minerals and plastic products for industrial
uses. Table 14.11 provides a simplified pro forma balance sheet.' To estimate a
weighted average cost of capital we need the marginal cost of each capital source
and the appropriate weighting scheme.
A. THE COST OF LONG-TERM DEBT. The first problem is estimating the current
market rate of return that would be required if the firm issued new long-term debt.
(See Table 14.12.) Almost all of Bethlehem Steel's long-term debt has had a maturity
of 25 to 30 years when first issued. Therefore we will assume that any new long-term
debt will also be issued with a 30-year maturity and an Aa bond rating. What rate
of return would the market require for a new issue with this risk? The "yield" pro-
vided by Wall Street Journal is not useful because it assumes an infinite maturity
for the debt.
15
Data taken from Moody's Industrial Manual.
528 CAPITAL STRUCTURE: EMPIRICAL EVIDENCE AND APPLICATIONS
Table 1<.12 Composition of Long-term Debt (in thousands of dollars) in Table 1<.11
êea]
Call Recent of
Issue Rating Amount P]ice P]ice êield Issue
Consol. Mtge. S.F. 3s, K, =979 Aa <=,800 =008 NA NA =949
Deîenture 3=s, =980 Aa 3,=00 =00 89< 3.6 =9PP
Deîenture P.40s, =99< Aa =09,<00 =0<* 848 6.4 =967
Deîenture 6is, =999 Aa 8P,800 =044 944 6.6 =969
Deîenture 9s, <000 Aa =44,000 =0P< =06< 8.P =970
Deîenture 8.4Ps, <00P Aa <P0,000 =07.4P =03< 8.< =97P
Deîenture 88s, <00= Aa <00,000 =06.63 =0P+ 7.9 =976
Suîord. Deî. 44s, =990 A 94,P00 =0<.40 764 P.9 =96P
Notes payaîle 30,000 NA NA NA
Suîsidiary deît 3,<00 NA NA NA
Revenue îonds P4s-6s, <00< =00,000 NA NA NA
NA = not applicaîle.
For example, take the 3+% debentures that were due in 1980. The "yield" is the
coupon divided by the market price:
32.50
Yield = = 3.6%.
895.00
This number is completely unrealistic. The required market rate of return is the rate
that equates the discounted value of the expected future cash flows with the current
market price of the security. We can find this rate by solving Eq. (14.5) for the before-
tax cost of debt, kb :
T
E(coupon), E(face value)
Bi = (14.5)
t = i (1 + kbY (1 + kbf
where
E(coupon) = the expected coupon payment in year t (assumed to be $32.50),
E(face value) = the expected face value (assumed to be $1000),
T = number of years to maturity,
Bi = market value of the jth debt issue.
Note that we have assumed that the bond rated Aa will actually pay its full face
value of $1000 when it matures. This may not always be a valid assumption. If the
bond is very risky, the expected payout may be less than $1000. (Or if the bond is
callable, the expected payout may be more than $1000 and the time to maturity may
be less than T if the bond is called early.) Given the above assumptions about the
expected payout of the 3+% debentures, the current required rate of return from them
is approximately 8%. At the end of 1976, this was the appropriate interest rate for an
COST OF CAPITAL: APPLICATIONS 529
Market Value
(in thousands of dollars) Percent Cost (in percent)
Aa-rated debenture due to mature in around four years.' Bonds with the same rating
but longer maturities are slightly riskier and therefore yield higher rates. For exam-
ple, the 8.45s maturing in the year 2005 yield approximately 8.3%. In our judgment,
if Bethlehem Steel had decided to issue new long-term debt in December 1976 with
an Aa bond rating and a maturity of between 25 and 30 years, the company would
have had to pay approximately 8.3%. We shall use this as the before-tax cost, k b ,
of long-term debt, because it is the best estimate of the marginal cost of new debt.
Next we need an estimate of the percentage of long-term debt, B/(B S), used by
Bethlehem in its target capital structure. Table 14.13 shows the market value weights
of various capital sources.
B. MARKET VALUE WEIGHTS. The market value weights of long-term debt and
equity are obvious. For long-term debt we use the market values from Table 14.13,
and when they are unavailable, we use book value. The market value of equity is
simply the number of shares outstanding multiplied by the price per share. Short-
term liabilities are calculated at book value. This is not an unreasonable assumption
because, under normal circumstances, the market value of short-term debt rarely
deviates much from its book value. We assume that the current market weights are
a reasonable estimate of the firm's long-run "optimal" target capital structure. A
quick, but somewhat unreliable, way of checking this assumption is to observe that
the book value of long-term debt is currently 27.5% of total long-term debt plus
equity. Its high between 1970 and 1976 was 27.5%, the average was 23.4%, and the
low was 20.67%. Therefore the current capital structure is not too far from the
normal pattern observed in recent years.
" Coupon payments of $16.25 per $1000 are made every May first and November first. The issue matured
on May 1, 1980. Therefore the actual computation is more complicated than Eq. (14.5). Students familiar
with the actuarial complexities of discounting formulas will find the exact calculation to be
895.00 32.50 , _ (1 + kb",)-6] + 1016.25
(1 + k0 116 2 [
r
ki,j/2 (1 + kki/2) 1
where k5/2 is a semiannual nominal rate. The solution to this problem provides an annual effective rate
of approximately 8%.
530 CAPITAL STRUCTURE: EMPIRICAL EVIDENCE AND APPLICATIONS
D. THE COST OF EQUITY. We can use the CAPM to estimate the cost of equity,
but two of the important parameters are judgmental. Equation (14.6) gives the CAPM
equation for the cost of equity:
ks = R f [E(R,n) — R s. (14.6)
The expected future rate of return on the market cannot be measured. However, a
good way of guessing what it might be is to add three components: (1) the real rate
of growth of the economy, 2% to 3%, (2) an adjustment for inflation next year (in
December 1976 a good guess might have been 5% to 6%), and (3) a risk premium for
Market Value
(in thousands) Percent Cost
the riskiness of the market portfolio (say 6% to 7%). Using the mean of each of these
components, we estimate the expected rate of return as 14.5%. The risk-free rate
may be approximated by the 90-day rate on U.S. government Treasury bills, which
in December 1976 was 4.67%. Finally, we need an estimate of the systematic risk
of the common stock of the company, which was .9.1' This is really an estimate of
the future systematic risk of Bethlehem Steel, and it is as much a guess as is the future
rate of return on the market index. Substituting these parameter estimates into Eq.
(14.6), we estimate the cost of equity capital to be 13.52%.
17
There are several companies that publish estimates of the systematic risk of individual firms. We used
an estimate provided by Wilshire Associates, Santa Monica, California.
18
The accounting treatment of bond repurchase when bonds are selling at a discount can cause extremely
perverse behavior. Suppose that interest rates have risen, thereby causing fixed rate outstanding debt to
sell at a discount. Repurchase of this debt at market value has no economic gain or loss before taxes
because the firm is paying the fair market price. However, the accounting treatment allows the firm to
record the difference between the face value and the market value as profit. Therefore firms that desire
to increase their reported earnings per share may decide to repurchase discounted debt. Unfortunately,
this is a negative present value decision because the firm has to pay (1) ordinary income taxes on its paper
gain and (2) higher coupons on replacement debt.
532 CAPITAL STRUCTURE: EMPIRICAL EVIDENCE AND APPLICATIONS
To provide a focus for the analysis, consider the following facts. The L and S
Company has a $10,000,000 bond issue outstanding with five years to maturity and
a coupon rate of 16%, and with annual interest payments. The current market rate
on debt of equivalent risk is 10%. The call price on the $1,000 face value debt is
$1,050, and the firm has a 40% tax rate. What is the net present value of calling
the outstanding bonds and replacing them with new bonds of equivalent maturity
and risk? Assume no transactions costs.
This is a capital budgeting decision and the usual procedures apply. The changes
in the after-tax cash flows should be discounted at the appropriate risk-adjusted
after-tax discount rate. In this case the appropriate rate is the after-tax rate on debt
of equivalent risk, i.e., (1 — Te)k, = (1 — .4).10 = .06. The after-tax cash flows have
three components (in a world without transactions costs).
1. The call premium will cost $500,000. It is deductible as an expense; therefore
its after-tax cost is $500,000(1 — .4) = $300,000.
2. The new debt issue will have to be $10,300,000, and it will pay a 10% coupon
(r 2 = 10%); thus the new interest will be $1,030,000 per year. The old interest
was 16% (r 1 = 16%) of $10,000,000, or $1,600,000 per year. Therefore the after-
tax interest saving amounts to (1 — te)(r i — r 2)B — (1 — AB, where B is the
original book value and AB is the change in book value. Numerically the annual
interest savings is
(1 — .4)(.16 — .10)(10,000,000) — (1 — .4)(.10)(300,000) = 342,000.
3. Incremental principal on the new debt, namely, $300,000, must be paid off at
maturity.
The net present value of these components is
N
v (= - tc)(r i — r 2)B N (1 — T c)r, AB AB
NPV = t
[1 + (1 — 'Oka t [1 + (1 — Tc)k b] + (1 — tc)kbr
5
(1 — .4)(.16 — .10)10,000,000 5
( = - .4).10(300,000)
1- 1 [1 + (1 — .4)(.10)T r [1 + (1 — .4)(.10)T
300,000
[1 + (1 — .4)(.10)r
= 1,516,464 — 75,823 — 224,190
= 1,216,451.
Therefore the debt should be refunded.
Two additional issues are pertinent. First, we should consider the fact that the
debt refinancing will slightly alter the capital structure of the firm because the market
value of the new debt issue exceeds the market value of the outstanding debt. This may
create a tax gain from leverage. The second issue is that callable debt should never
sell for more than the call price plus a small premium approximately equal to the
flotation costs of exercising the call. No investor would rationally pay $1100 for a
bond that might be called (any minute) at $1050.
COST OF CAPITAL: APPLICATIONS 533
Another method for refunding is called debt defeasance. It can be used to retire
noncallable outstanding debt without actually purchasing the debt in the open
market. The corporation purchases U.S. Treasury bonds (or other government-
guaranteed obligations) with cash inflows that match the cash outflows of the debt
to be defeased. The government securities are placed in an irrevocable trust, and the
trustee is charged with making principal and interest payments on the corporate debt
as they come due. The defeased debt may then be removed from the firm's balance
sheet (and the government securities are not added).
In 1982 Exxon defeased six issues of long-term debt that had a total face value
of $515 million and that had yields of 5.8 to 6.7%. To do the job, it purchased $312
million in U.S. Treasury bonds yielding 14%. The after-tax difference between the
face value of the corporate bonds and the cost of the government bonds was about
$132 million Exxon added this amount to its second quarter income, increasing its
per share earnings from about $0.87 to $1.02.'
Financial reporting for defeasance is governed by FASB No. 76, issued in 1983.
The defeased debt is removed from the balance sheet and the U.S. government secu-
rities are not added. A footnote is required, giving the general terms of the defeasance
and the amount being defeased. Only debt with specific maturities and fixed payment
schedules may be defeased. Convertible securities, floating rate debt, and redeemable
preferred stock cannot be defeased. Defeasance is nontaxable at the time it takes place.
The difference between the face value of the corporate debt and the cost of the
defeasance trust drops directly to the company's bottom line. The interest on the
government securities in the defeasance trust is subject to an ordinary income tax.
Finally, the difference between the cost of the government bonds (if they were bought
at a discount) and their face value is subject to a capital gains tax when they mature.
Bierman [1985] and Peterson, Peterson, and Ang [1985] have analyzed the de-
feasance decision. Consider the following example. The XYZ Corporation has $500
million of low coupon debt outstanding that cannot be repurchased because it is a
private placement. The debt has five years left before maturity and pays an (annual)
coupon rate of 6%, with the face value due at maturity. You are considering defeasance
with U.S. Treasury bonds that pay coupons on the same dates that your corporate
debt payments come due. The T-bonds sell for par, yield 14%, and mature in five
years. Your corporate tax rate is 40% (and the capital gains rate is half of the ordi-
nary rate). How do you defease and what is the net present value of the defeasance
decision?
To defease, it will be necessary to purchase two U.S. government bond issues.
First, we need a five-year 14% bond that pays $30 million in interest each year. If
D 1 is the face value of this bond, we know that
.14D 1 = $30,000,000
D 1 = $214,285,710.
This issue will provide cash flows to meet the scheduled coupon payments on our
corporate debt, but it falls short of providing the full $500 million face value due
19 The increase in earnings does not necessarily mean that this was a positive NPV decision. Remember,
at maturity.' For this we need a second U.S. government bond, a zero coupon bond,
with a face value of D 2 , where
D 2 = $500,000,000 — $214,285,710
= $285,714,290.
The present value of this bond is
285 714 290
, ,
B2 = = $148,391,049.
(1.14)5
The total cash outlay needed to build a defeasance trust is
B 1 + B2 = 214,285,710 + 148,391,049
= 362,676,759
Note that this has exactly the same market value as the debt being defeased. Aside
from tax consequences, we are using $362.68 million of U.S. government debt to
defease $362.68 million of corporate debt.
The cash flows from the decision are (1) a current outlay of $362.68 million that
is used to retire $362.68 million of debt, (2) interest saved of $30 million per year
and an equal amount of interest lost, (3) principal of $500 million not paid in year
5 and an equal amount not received, and (4) capital gains tax on the zero coupon
bond, paid in year 5. All cash flows, except for the last, net out. Therefore the NPV
of the decision is
—(capital gain on zero coupon bond) (r c /2)
NPV =
[1 + 141 — -c c )] N
—(285,714,290 — 148,391,049)(.4/2)
[1 + .14(1 — .4)] 5
= —18,349,643.
This is a negative net present value decision because of the capital gains tax liability
on the zero coupon government bonds. Note also that reported earnings would in-
crease by the amount ($500.0 million — $362.68 million = $137.32 million).
20
Note that this bond yields the market rate exactly; therefore its present value is B, = $214,285,710.
21
For an excellent comparison of underwritten new issues versus rights issues, see Smith [1977].
COST OF CAPITAL: APPLICATIONS 535
managerial finance textbooks to include flotation costs in the cost of capital is not
generally consistent with value maximization. The normally followed procedure has
been to adjust the required rate of return for flotation costs, F. For example, if
one were using the Gordon growth model to estimate the cost of equity, the incor-
rect formulation would be
Div,
ks =
F
}— S
where
Div, = the end-of-period expected dividends (aggregate dollar value),
S = the current market value of equity,
F = the dollar amount of flotation costs,
g = the expected long-term growth in dividends.
The above approach is incorrect because it implicitly adjusts the opportunity cost
of funds supplied to the firm. Yet the true market-determined opportunity cost is
unaffected by the flotation costs of a particular firm.
The correct method is to begin with a definition of net present value that explicitly
recognizes flotation costs. One approach, which uses reproduction cost leverage, has
been provided by Ezzell and Porter [1976]. Another approach, which we prefer, uses
replacement value leverage and is described below. Let F be the dollar amount of
flotation costs for a project financed with AS dollars of equity and AB dollars of
debt. Thus the firm's target leverage, L, is
AB
L= (14.7)
AS + AB
The total investment outlay for a project is AI dollars for capital invested and F
dollars of flotation costs. Therefore we have
A/ + F = AS + AB = AS + L(AI + F . (14.8)
If the firm's flotation costs are a% of the new equity issued and y% of new debt, then
F= AS + T AB, (14.9)
and the weighted percentage flotation cost, 3, is
= 3i3(} — L + yL. (14.10)
Substituting (14.7) and (14.8) into (14.9) yields
This definition explicitly recognizes that flotation costs are cash flows incurred at
the present time, not stretched over the life of the project. Finally, substituting (14.11)
into (14.12) we obtain
NCF, 1
NPV = A l. (14.13)
t= (1 + WACC)t 1 — e
Thus the correct procedure for the economic analysis of flotation costs does not alter
the weighted average cost of capital, WACC. Rather, it grosses up the investment
outlay by [1/(1 — E)], a factor that considers the weighted average flotation costs of
debt and equity.
SUMMARY
The cost of capital, the capital structure of the firm, and the capital budgeting deci-
sion are all inextricably linked The theory of finance provides equations that may
be applied to the solution of the weighted average cost of capital under a variety
of assumptions. Although no completely satisfactory theory has yet been found to
explain the existence of optimal capital structure, casual empiricism suggests that
firms behave as though it does exist. Therefore, for the time being, suggested tech-
niques for estimating the weighted average cost of capital usually assume that each
firm has a target capital structure. This target is then applied to the cost of capital
formulas.
A major empirical issue is the impact of capital structure on the value of the firm
and on the weighted average cost of capital. Although the evidence is mixed, the
articles that were summarized seem to indicate that there is in fact a gain from lever-
age that seems to be between 10% and 20%. Of course, this is less than the marginal
corporate tax rate (34%), and we could have used a tax rate of 10% to 20% in
estimating the cost of capital for Bethlehem Steel. If we had done so, the weighted
average cost of capital would have been 11.34%.
PROBLEM SET
14.1 The puzzle of optimal capital structure is that there appear to be cross-sectional regu-
larities in the observed ratios of debt to equity of U.S. firms. For example, the steel industry
appears to carry a higher percentage of debt than the public accounting industry does. These
same regularities appeared even before the existence of corporate income taxes. How can op-
timal leverage be explained without relying on the tax shield of debt or bankruptcy costs?
14.2 What are the empirical problems involved in testing for the effect of capital structure
on the value of the firm?
PROBLEM SET 537
14.3 Assume a world without taxes. How would the weighted average cost of capital vary with
the ratio of debt to total assets, B/(B + S), if the cost of equity remained constant, i.e.,
k s =p>kb?
14.4 During recent years your company has made considerable use of debt financing, to the
extent that it is generally agreed that the percentage of debt in the firm's capital struc-
ture (either in book or market value terms) is too high. Further use of debt will likely lead
to a drop in the firm's bond rating. You would like to recommend that the next major capital
investment be financed with a new equity issue. Unfortunately, the firm has not been doing
very well recently (nor has the market). In fact the rate of return on investment has been
just equal to the cost of capital. As shown in the financial statement in Table Q14.4, the
market value of the firm's equity is less than its book value. This means that even a profitable
project will decrease earnings per share if it is financed with new equity. For example, the firm
is considering a project that costs $400 but has a value of $500 (i.e., an NPV of $100), and
that will increase total earnings by $60 per year. If it is financed with equity the $400 will
require approximately 200 shares, thus bringing the total shares outstanding to 1200. The new
earnings will be $660, and earnings per share will fall to $0.55. The president of the firm
argues that the project should be delayed for three reasons.
a) It is too expensive for the firm to issue new debt.
b) Financing the project with new equity will reduce earnings per share because the market
value of equity is less than book value.
c) Equity markets are currently depressed. If the firm waits until the market index improves,
the market value of equity will exceed the book value and equity financing will no longer
reduce earnings per share.
Critique the president's logic.
14.5 Southwestern Electric Company.' John Hatteras, the financial analyst for Southwestern
Electric Company, is responsible for preliminary analysis of the company's investment projects.
22 This problem is really a short case. It has a definite answer but requires knowledge of cash flows, dis-
counting, the CAPM, and risky cost of capital.
538 CAPITAL STRUCTURE: EMPIRICAL EVIDENCE AND APPLICATIONS
Table Q14.5
1 250 10 7.5
2 250 20 15.0
3 250 25 22.5
4 250 60 30.0
5-30 0 110 30.0
31-40 0 80 30.0
41 0 40 0
He is currently trying to evaluate two large projects that management has decided to consider
as a single joint project, because it is felt that the geographical diversification the joint project
provides would be advantageous.
Southwestern Electric was founded in the early 1930s and has operated profitably ever
since. Growing at about the same rate as the population in its service areas, the company has
usually been able to forecast its revenues with a great deal of accuracy. The stable pattern in
revenues and a favorable regulatory environment have caused most investors to view South-
western as an investment of very low risk.
Hatteras is concerned because one of the two projects uses a new technology that will be
very profitable, assuming that demand is high in a booming economy, but will do poorly in
a recessionary economy. However, the expected cash flows of the two projects, supplied by
the engineering department, are identical. The expected after-tax cash flows on operating in-
come for the joint project are given in Table Q14.5. Both projects are exactly the same size,
so the cash flow for one is simply half the joint cash flow.
In order to better evaluate the project, Hatteras applies his knowledge of modern finance
theory. He estimates that the beta of the riskier project is .75, whereas the beta for the less
risky project is .4375. These betas, however, are based on the covariance between the return
on after-tax operating income and the market. Hatteras vaguely recalls that any discount rate
he decides to apply to the project should consider financial risk as well as operating (or
business) risk. The beta for the equity of Southwestern is .5. The company has a ratio of debt
to total assets of 50% and a marginal tax rate of 4071. Because the bonds of Southwestern
are rated Aaa, Hatteras decides to assume that they are risk free. Finally, after consulting his
investment banker, Hatteras believes that 1871 is a reasonable estimate of the expected return
on the market.
The joint project, if undertaken, will represent 10% of the corporation's assets. South-
western intends to finance the joint project with 50% debt and 50% equity.
Hatteras wants to submit a report that answers the following questions:
a) What is the appropriate required rate of return for the new project?
b) What are the cost of equity capital and the weighted average cost of capital for South-
western Electric before it takes the project?
c) Should the joint project be accepted?
d) What would the outcome be if the projects are considered separately?
e) If the joint project is accepted, what will the firm's new risk level be?
REFERENCES P39
REFERENCES
Altman, E., "A Further Empirical Investigation of the Bankruptcy Cost Question," Jou]nal of
Finance, Septemîer =984, =067-=089.
American Bar Association, Commenta]ies on Model Debentu]e Indentu]e P]ovisions, ABA,
Chicago, =97=.
Asquith, P., and D. Mullins, Jr., "Equity Issues and Offering Dilution," Jou]nal of Financial
Economics, January—Feîruary =986, 6=-90.
Barges, A., The Effect of Capital St]uctu]e on the Cost of Capital. Prentice-Hall, Englewood
Cliffs, N.J., =963.
Barnea, A.; R. Haugen; and L. Senîet, Agency P]oblems and Financial Cont]acting. Prentice-
Hall, Englewood Cliffs, N.J., =98P.
Baron, D., and B. Holmstrom, "The Investment Banking Contract for New Issues under
Asymmetric Information: Delegation and the Incentive Proîlem," Jou]nal of Finance,
Decemîer =980, ===P-==38.
Baxter, N., "Leverage, Risk of Ruin and the Cost of Capital," Jou]nal of Finance, Septemîer
=967, 39P-403.
Beranek, W., The Effect of Leve]age on the Ma]ket Value of Common Stock. Bureau of Business
Research and Service, Madison, Wis., =964.
Bhattacharya, S., "Imperfect Information, Dividend Policy, and 'The Bird in Hand' Fallacy,"
Bell Jou]nal of Economics, Spring =979, <P9-<70.
Bierman, H., "Defeasance Is Not a Free Lunch," Jou]nal of Co]po]ate Finance, Spring =98P,
=3 —=6.
Black, F., and M. Scholes, "The Pricing of Options and Corporate Liaîilities," Jou]nal of
Political Economy, May—June =973, 637-6P4.
Boness, A. J., and G. M. Frankfurter, "Evidence of Non-Homogeneity of Capital Costs within
`Risk-Classes,'" Jou]nal of Finance, June =977, 77P-787.
Booth, J., and R. Smith, "Capital Raising, Underwriting and the Certification Hypothesis,"
Jou]nal of Financial Economics, January—Feîruary =986.
Bradley, M.; G. Jarrell; and E. H. Kim, "On the Existence of an Optimal Capital Structure:
Theory and Evidence," Jou]nal of Finance, July =984, 8P7-878.
Brennan, M., and E. Schwartz, "Savings Bonds, Retractaîle Bonds and Callaîle Bonds,"
Jou]nal of Financial Economics, August =977, 67-88.
, "Optimal Financial Policy and Firm Valuation," Jou]nal of Finance, July =984,
P93-607.
Cordes, J., and S. Sheffrin, "Estimating the Tax Advantage of Corporate Deît," Jou]nal of
Finance, March =983, 9P-=0P.
Dann, L., and W. Mikkelson, "Convertiîle Deît Issuance, Capital Structure Change and
Financing-Related Information: Some New Evidence," Jou]nal of Financial Economics,
June =984, =P7-=86.
Darrough, M., and N. Stoughton, "Moral Hazard and Adverse Selection: The Question of
Financial Structure," Jou]nal of Finance, June =986, P0=-P=3.
DeAngelo, H., and R. Masulis, "Optimal Capital Structure under Corporate and Personal
Taxation," Jou]nal of Financial Economics, March =980, 3-30.
5<0 CAPITAL STRUCTURE: EMPIRICAL EVIDENCE AND APPLICATIONS
Livingston, M., "Bond Refunding Reconsidered: Comment," Jou]nal of Finance, March =980,
=9=-=96.
Long, M., and I. Malitz, "Investment Patterns and Financial Leverage," in Friedman, ed.,
Co]po]ate Capital St]uctu]e in the United States. University of Chicago Press, Chicago,
=98P.
Masulis, R., "The Effects of Capital Structure Change on Security Prices: A Study of Exchange
Offers," Jou]nal of Financial Economics, June =980, =39-=78.
, "The Impact of Capital Structure Change on Firm Value, Some Estimates," Jou]nal
of Finance, March =983, =07-=<6.
, and A. Korwar, "Seasoned Equity Offerings: An Empirical Investigation," Jou]nal of
Financial Economics, January—Feîruary =986, 9=-==8.
Mayor, T., and K. McCorn, "Bond Refunding: One or Two Faces?" Jou]nal of Finance, March
=978, 349-3P3.
McConnell, J., and G. Schlarîaum, "Evidence on the Impact of Exchange Offers on Security
Prices: The Case of Income Bonds," Jou]nal of Business, January =98=, 6P-8P.
Mikkelson, W., and M. Partch, "Valuation Effects of Security Offerings and the Issuance
Process," Jou]nal of Financial Economics, January—Feîruary =986, 3=-60.
Miller, M., "Deît and Taxes," Jou]nal of Finance, May =977, <6=-<7P.
, and F. Modigliani, "Some Estimates of the Cost of Capital to the Electric Utility
Industry, =9P4-P7," Ame]ican Economic Review, June =966, 333-39=.
, "Some Estimates of the Cost of Capital to the Electric Utility Industry, =9P4-P7:
Reply," Ame]ican Economic Review, Decemîer =967, =<88-=300.
Modigliani, F., and M. Miller, "The Cost of Capital, Corporation Finance and the Theory of
Investment," Ame]ican Economic Review, June =9P8, <6=-<97.
, "Taxes and the Cost of Capital: A Correction," Ame]ican Economic Review, June =963,
433-443.
Myers, S. C., "Determinants of Corporate Borrowing," Jou]nal of Financial Economics, No-
vemîer =977, =47-=76.
, "The Capital Structure Puzzle," Jou]nal of Finance, July =984, P7P-P9<.
, and N. Majluf, "Corporate Financing and Investment Decisions When Firms Have
Information That Investors Do Not Have," Jou]nal of Financial Economics, June =984,
=87-<<=.
Ofer, A., and R. Taggart, Jr., "Bond Refunding: A Clarifying Analysis," Jou]nal of Finance,
March =977, <=-30.
, "Bond Refunding Reconsidered: Reply," Jou]nal of Finance, March =980, =97-<00.
Peavy, J., and J. Scott, "The Effect of Stock-for-Deît Swaps on Security Returns," Financial
Review, Novemîer =98P, 303-3<7.
Peterson, P.; D. Peterson; and J. Ang, "The Extinguishment of Deît through In-Suîstance
Defeasance," Financial Management, Spring =98P, P9-67.
Pinegar, J., and R. Lease, "The Impact of Preferred-for-Common Exchange Offers on Firm
Value," Jou]nal of Finance, Septemîer =986, 79P-8=4.
Riley, J., "Competitive Signalling," Jou]nal of Economic Theo]y, April =97P, =74-=86.
Roîichek, A.; J. McDonald; and R. Higgins, "Some Estimates of the Cost of Capital to the
Electric Utility Industry, =9P4-P7: Comment," Ame]ican Economic Review, Decemîer
=967, =<78-=<88.
P4< CAPITAL STRUCTURE: EMPIRICAL EVIDENCE AND APPLICATIONS
Rogers, R., and J. Owers, "Equity for Deît Exchanges and Shareholder Wealth," Financial
Management, Autumn =98P, =8-<6.
Ross, S. A., "The Determination of Financial Structure: The Incentive Signalling Approach,"
Bell Jou]nal of Economics, Spring =977, <3-40.
Rothschild, M., and J. Stiglitz, "Equiliîrium in Competitive Insurance Markets," Qua]te]ly
Jou]nal of Economics, Novemîer =976, 6<9-6P0.
Ruîinstein, M., "A Mean-Variance Synthesis of Corporate Financial Theory," Jou]nal of
Finance, March =973, =67-=8=.
Schipper, K., and A. Smith, "A Comparison of Equity Carve-outs and Seasoned Equity
Offerings: Share Price Effects and Corporate Restructuring," Jou]nal of Financial Eco-
nomics, January—Feîruary =986, =P3-=86.
Scott, J. H., Jr., "A Theory of Optimal Capital Structure," Bell Jou]nal of Economics, Spring
=976, 33-P4.
; G. H. Hempel; and J. Peavy III, "The Effect of Stock-for-Deît Swaps on Bank Holding
Companies," Jou]nal of Banking and Finance, June =98P, <33-<P=.
Shastri, K., "Two Essays Concerning the Effects of Firm Investment/Financing Decisions on
Security Values: An Option Pricing Approach," Ph.D. dissertation, UCLA Graduate
School of Management, =98=.
Smith, C., Jr., "Suîstitute Methods for Raising Additional Capital: Rights Offerings versus
Underwritten Issues," Jou]nal of Financial Economics, Decemîer =977, <73-307.
, "Investment Banking and the Capital Acquisition Process," Jou]nal of Financial Eco-
nomics, January—Feîruary =986, 3-30.
, and J. B. Warner, "On Financial Contracting: An Analysis of Bond Covenants,"
Jou]nal of Financial Economics, June =979, ==7-=6=.
Solomon, E., The Theo]y of Financial Management. Columîia University Press, New York,
=963.
Spence, A. M., "Joî Market Signaling," Qua]te]ly Jou]nal of Economics, August =973, 3PP-379.
Stiglitz, J., "A Re-Examination of the Modigliani-Miller Theorem," Ame]ican Economic Review,
Decemîer =969, 784-793.
, "Some Aspects of the Pure Theory of Corporate Finance: Bankruptcies and Take-
overs," Bell Jou]nal of Economics and Management Science, Autumn =97<, 4P8-48<.
, "On the Irrelevance of Corporate Financial Policy," Ame]ican Economic Review,
Decemîer =974, 8P=-866.
Stutz, R., and H. Johnson, "An Analysis of Secured Deît," Jou]nal of Financial Economics,
Decemîer =98P, P0=-P<=.
Titman, S., "The Effect of Capital Structure on a Firm's Liquidation Decision," Jou]nal of
Financial Economics, March =984, =37-=P=.
, "Determinants of Capital Structure: An Empirical Analysis," Working paper, UCLA
Graduate School of Management, =98<.
, and R. Wessels, "The Determinants of Capital Structure Choice," Working paper,
UCLA, =98P.
Wakeman, L. M., "Bond Rating Agencies and Capital Markets," Working paper, Graduate
School of Management, University of Rochester, Rochester, N.Y., =978.
Warner, J., "Bankruptcy, Aîsolute Priority, and the Pricing of Risky Deît Claims," Jou]nal
of Financial Economics, May =977a, <39-<76.
REFERENCES P43
Is the value of shareholders' wealth affected by the dividend policy of the firm? This
is another variation on the basic question, Can any financing decision affect the value
of the firm? The previous chapters looked at the relationship between capital structure
and the value of the firm, using a fairly simple valuation model that assumed a non-
growing stream of cash flows from investment. Capital structure theory shows that
in a world without taxes, agency costs, or information asymmetry repackaging the
firm's net operating cash flows into fixed cash flows for debt and residual cash flows
for shareholders has no effect on the value of the firm. This chapter develops valuation
models that include growth opportunities, thereby adding a greater element of realism.
Even so, we shall show that, in a world without taxes, it makes no difference whether
shareholders receive their cash flows as dividends or as capital gains. Thus in the
absence of taxes, agency costs, or information asymmetry, dividend policy is irrelevant.
It does not affect shareholders' wealth. The argument is then extended to a valuation
model that includes growth and corporate taxes, but the result does not change.
Dividend payout does not affect the value of the firm. However, in a world with
personal as well as corporate taxes the possibility arises that dividends may affect
value. Also, agency costs and information heterogeneity are proposed as possible
explanations for dividend policy. Empirical tests of dividend policy and applications
of corporate valuation are discussed in Chapter 16.
544
THE IRRELEVANCE OF DIVIDEND POLICY IN A WORLD WITHOUT TAXES 545
Miller and Modigliani [1961] present a cogent argument for the fact that the value
of the firm is unaffected by dividend policy in a world without taxes or transactions
costs. They begin by assuming that two firms are identical in every respect except
for their dividend payout in the current time period. Their streams of future cash
flows from operations are identical, their planned investment outlays are identical,
and all future dividend payments from the second time period on are also identical.
We can represent this mathematically as follows:
NOI,(t) = NOI,(t), t = 0, 1, , co,
1 t = 0, 1,..., co,
2(t),
Div JO = Div 2 (t), t = 1, . . . , GO,
Div,(0) Div2(0),
where
NOI,(t) = the random future cash flows from operations for the ith firm in time
period t,
Ti(t) = the variable investment outlay for the ith firm in time period t,
ti
Div i(t) = the random dividend payout for firms in period t,
Div i(0) = the dividend payout for the ith firm during the current time period.
I
For the sake of simplicity, we assume that both firms are 100% equity. This avoids the problem of
confusing capital structure effects with possible dividend policy effects.
546 DIVIDEND POLICY: THEORY
If the numerator and denominator of (15.1) are multiplied by the current number of
shares outstanding, n i(t , then by rearranging terms, we have
Div,(t + 1) + n i(t P i(t + }
V(t) = (15.2)
1 + p(t + }
where
Divi(t + 1) = total dollar dividend payment = n i(t divi(t + } ,
= the market value of the firm = n i(t P i(t .
Hence the value of the firm is seen to be equal to the discounted sum of two cash
flows: any dividends paid out, Divi(t + 1), and the end-of-period value of the firm.
To show that the present value of the firm is independent of dividend payout, we
shall examine the sources and uses of funds for the two firms in order to rewrite
(15.2) in a way that is independent of dividends.
2
This argument assumes, for the sake of convenience, that sources and uses of funds from balance sheet
items (e.g., changes in inventory or accounts receivable) are negligible.
THE IRRELEVANCE OF DIVIDEND POLICY IN A WORLD WITHOUT TAXES 547
where V(t + 1) = ni(t + 1):F",(t + 1). Therefore the valuation equation (15.2) may be
rewritten
NOI,(t + I) — + 1) + + 1)
V(t) =
+ p(t + 1)
• (15.8)
NO12 = N---1
Vo N011—/1 NM]. + r 1I1—I2 ••• NOI1 +E I', I, —IN
0,
►
to t1 t2
Figure 15.1
Time pattern of cash flows for a growing firm.
The Miller-Modigliani argument that the value of the firm is independent of dividend
policy also extends into a world with corporate taxes but without personal taxes. In
this section the valuation model [Eq. (15.8)] is extended to include corporate taxes
and a growing stream of cash flows. The result is a valuation model that has realistic
features and hence may be usefully applied to real-world valuation problems. Chapter
16 will expand on the usefulness of the valuation model by means of an example.
A reasonable assumption is that in any time period the value of the firm, V, is finite.'
Therefore, given a model with an infinite horizon, we have
N NOI, — I,
V, = Ern E • (15.10)
N-,Do t=1 (1 ± a
The tildes (-) are dropped for notational convenience. Also note that
NOI2 = NOI, + r111.
After all, no one has observed a firm with infinite value as yet.
VALUATION, GROWTH, AND DIVIDEND POLICY 549
Table 15.1
Time Cash
Pe]iod Cash Inflow Outflow
1 NOI1 —I1
2 NO12 = NOI1 + r1I1 -}2
3 NOT, = NOI, + r1I1 r212 -=3 (15.11)
N-1
N NOI N = NOI1 + —
IN
Equation (15.10) is the same formula used in Chapter 2 on capital budgeting. The
present value of the firm is the sum of the discounted cash flows from operations less
the new investment outlays necessary to undertake them.
Referring to Fig. 15.1, we can see that the average return on investment, ] 1 , is
assumed to continue forever at a constant rate. This is a perfectly reasonable assump-
tion because if the capital budgeting decision is made correctly, each project will re-
turn enough cash to cover payments to suppliers of capital and to recover the initial
investment. Thus the cash flows are sufficient to provide any needed replacement
investment to sustain the project at a constant level forever. The stream of cash flows
for the growing firm in Fig. 15.1 is given in Table 15.1 [also Eq. (15.11)].
Substituting (15.11) into (15.10), we can express the present value of the growing
firm as
=
N01 1 -11 N011 +1' 111 ÿ 12 N011 + 1' 1 _1 1 + ] 212 ÿ 3
V0
p (1 + p) 2 (1 + p)3
N- 1
NOI1 E ] t I, ÿ N
=1 (15.12)
+•••ã 1
( +
This extended equation can be simplified greatly. First, rewrite it by rearranging
terms as follows:
NOI, +
NOI 1 +
NOI1
Vo = 1 + p 2 • +
(1 + p) 1
( +
] 1
+ I1 •• + 1 N
[(1 + p) 2 + (1 + p)3 + (1 +p) +
+ 1.2 [ ]2 r2 ]2 1 +
(1 + p) 3 (1 p) 4 (1 + p) N (1 + p) 2
We can simplify Eq. (15.13) by recognizing that the first term is an infinite annuity
with constant payments of NOI, per period. Therefore
N NOI, NOT,
lim 1
• (15.14)
N co t= 1 ( ã P)t P
Next, the second term in (15.13) can be simplified as follows:
N ]t 1 Nÿt ]t
1 1
+
1( + O ]
( Pê (1
(15.15)
1 ]t 1 ]t
im
l
Substituting (15.14) and (15.15) back into (15.13), we obtain a simplified expression
for the present value of the firm:
Table 15.2
Another feature of Eq. (15.16) is that it is derived directly from Eq. (15.8), and
in both we have the result that dividend policy is irrelevant in a world without taxes,
information asymmetry, or transactions costs. All that counts is cash flows from
investment.
5
However, do not underestimate the usefulness of Eq. (15.16). It is the basis for most commonly used
valuation models, e.g., ALCAR, which is a personal computer—based model designed by Professor Al
Rappaport of Northwestern.
552 DIVIDEND POLICY: THEORY
Note that ]K is the same as the rate of growth, g, for cash flows. In other words, NOI
in the tth time period is the future value of NOI in the first time period, assuming
that cash flows grow at a constant rate, g-
NOT, = NOI =(1 +
By substituting (15.17) into (15.16) and maintaining the assumption that ] t = ], we
have
NOI, KNOIt(r — p)
Vo = + , • (15.19)
p t= i pki + /9/
Then by using (15.18) in (15.19) we obtain
NOI1
r — p °° (1 + ]K)t-1
Vo = + KNOI,
C
p t=1 (1+ P)t
— NOI, [ K(] — p) (1 + rK
= + (15.20)
1 + ]K t =i 1 + p
If ]K = p, then the last term in (15.20) will have a finite limit'
N
(1 + ]K)t 1 + ]K
p M ]K. (15.20a)
1 p p ÿ ]K
Substituting (15.20a) into (15.20) and simplifying, we have an equation for the value
of the firm, assuming infinite growth at a rate less than the market rate of return, p-
NOT, [ K(] ÿ p) 1 + Kr]
Vo = = +
p 1 + K] p ÿ ]K
NOI =(1 — K)
(15.21)
p ÿ K]
Equation (15.21), rewritten in a somewhat different form, is frequently referred
to as the Go]don g]owth model. Note that since K is the investment rate (although
K need not be less than one), the numerator of (15.21) is the same as dividends paid
at the end of the first time period:
NOI =(1 — K) = Div,.
Also, as was shown earlier, the product of the investment rate and the average rate
of return on investment is the same as the growth rate, g, in cash flows; therefore
K] = g.
6
For proof let (= + ]K)/(1 + p) = U. This can îe written as
S = U + U 2 + • • • + U N.
Multiplying this îy U and suîtracting the result from the aîove, we have
S = U/(1 ÿ U) ÿ U N+ 7(1 ÿ U).
The second term approaches zero in the limit as N approaches infinity. By suîstituting îack the definition
of U, we get (=P.<0a).
VALUATION, GROWTH, AND DIVIDEND POLICY 553
Given these facts and the necessary condition that g < p, the infinite growth model,
Eq. (15.21), can be rewritten as
Div,
Vo = (15.21a)
g
which is the Gordon growth model.
By substituting the approximation (15.23) into the valuation equation (15.22), we have
an approximate valuation formula for finite supernormal growth:8
NOI, K(r — p) (p — Kr NOI,
Vo = T
p p — Kr 1+p)p
NOI, r — p 1.
+ K(NOI,)T (15.24)
_p(1 + p)
The binomial expansion can be used to derive the approximation in the following way. Let (1 + K])/
(1 + p) = 1 + A. Then, recalling that K] = g, we have
1
( ± g )T = (1 + A) T = )(1) T K A K
T
(
1+p K
T
= 1 + TA + (T)AK 1 + TA.
K=2 K
Solving for A, we have
1 + K] K] ÿ p
A= 1
1+ p 1+p
Therefore the correct approximation is
(p ÿ K]
= + TA = = T
1+p
To simulate the validity of the approximation, assume that the investment rate, K, is 50%, the rate of
return on investment, ], is 20%, and the market-required rate of return is 15%. Figure 15.A is a plot of
[(1 + K])/(1 + p)]. We can see visually that the linear approximation is reasonable.
(1 + K])T
1+p T
.9565 1
.9149 2
.8752 3
.8371 4
.8007 5
Figure 15.A
The linear approximation of the growth term.
556 DIVIDEND POLICY: THEORY
Up to this point the models of firms that have been introduced assume a world with
only corporate taxes. What happens when personal taxes are considered? In partic-
ular, how is dividend policy affected by the important fact that in the United States
the capital gains tax is less than the personal income tax?9 An answer to this question
9
The 1986 tax code nominally makes the capital gains rate equal to the ordinary income rate. However,
capital gains taxes are still less than ordinary taxes in effect, because capital gains can be deferred in-
definitely, whereas taxes on ordinary income cannot.
DIVIDEND POLICY IN A WORLD WITH PERSONAL AND CORPORATE TAXES 557
The first term within the brackets is the after-tax cash flow of the firm, which is
(NOI — rD (} — t a). All of this is assumed to be paid out as dividends. The before-tax
income to the shareholder is the dividends received minus the interest on debt used
to buy shares. After subtracting income taxes on this income, we are left with Eq.
(15.26).
Alternatively, the firm can decide to pay no dividends, in which case we assume
that all gains are realized immediately by investors and taxed at the capital gains
rate.' In this event the after-tax income of a shareholder is
Y9 = (NOI — rD c (} — t e (} — tgi) rD pi(} — T pi , (15.27)
where
= the uncertain income to the ith individual if corporate income is received as
capital gains,
T gi = the capital gains rate for the ith individual.
I° More recently Miller and Scholes [1978] have also considered a world with dividends and taxes. The
implications of this paper are discussed later on in this chapter.
Obviously there is the third possibility that earnings are translated into capital gains and the capital
gains taxes are deferred to a later date. This possibility is considered in Farrar and Selwyn [1967]; it does
not change their conclusions.
558 DIVIDEND POLICY: THEORY
Now the individual pays a capital gains tax rate on the income from the firm and
deducts after-tax interest expenses on personal debt. The corporation can implement
the policy of translating cash flows into capital gains by simply repurchasing its
shares in the open market.
We can rewrite Eq. (15.27) as follows:
= [(NOI — rDc)(1 — ]) — ]D pi y1 ÿ -c gi ) ]D p ,(-c p , ÿ T (15.28)
From Eqs. (15.26) and (15.28) the advantage to investors of receiving returns in the
form of capital gains rather than dividends should be obvious. So long as the tax
rate on capital gains is less than the personal tax rate (Tpi < d pi), individuals will pre-
fer capital gains to dividends for any positive operating cash flows, rate of interest,
and level of debt (personal or corporate). The ratio of the two income streams,
[(NOI — r1),)(1 ÿ -c e) ÿ rD p j(1 ÿ T gi ) rD pi(T pi T gi ) M 1,
7 (15.29)
17. [(NOI — ]D c )(1 ÿ ]) ÿ ]D pi1(1 — T pi )
is greater than one if T gi < c pi . In general the best form of payment is the one that
is subject to least taxation. The implication, of course, is that corporations should
never pay dividends. If payments are to be made to shareholders, they should always
be made via share repurchase. This allows shareholders to avoid paying income tax
rates on dividends. Instead, they receive their payments in the form of capital gains
that are taxed at a lower rate.
What about debt policy? Again the same principle holds. The debt should be
held by the party who can obtain the greatest tax shield from the deductible interest
payments. This is the party with the greatest marginal tax rate. If the firm pays out
all its cash flow in the form of dividends, the favorable tax treatment of capital gains
is irrelevant. In this case we have the familiar Modigliani-Miller [1963] result that
the value of the firm is maximized by taking on the maximum amount of debt (see
Chapter 13). Proof is obtained by taking the partial derivative of Eq. (15.26) with
respect to personal and corporate debt and comparing the results.
Debt policy becomes more complex when the corporation repurchases shares in-
stead of paying dividends. Taking the partial derivatives of the capital gains income
equation, (15.27), we obtain
further. He shows that if the borrowing rate on debt is "grossed up" so that the
after-tax rate on debt equals the after-tax rate on other sources of capital, the marginal
investor will be indifferent between personal and corporate debt.'
Empirical evidence about the existence of debt clienteles is discussed in Chapter
16. Some clientele effects are obvious. For example, high tax bracket individuals hold
tax-free municipal bonds, whereas low tax bracket investors like pension funds (which
pay no taxes) prefer to invest in taxable corporate bonds. A much more subtle ques-
tion, however, is whether investors discriminate among various corporate debt issues,
i.e., do high tax bracket investors choose low-leverage firms?
Brennan [1970] extends the work of Farrar and Selwyn into a general equilibrium
framework where investors are assumed to maximize their expected utility of wealth.
Although this framework is more robust, Brennan's conclusions are not much differ-
ent from those of Farrar and Selwyn. With regard to dividend payout Brennan con-
cludes that "for a given level of risk, investors require a higher total return on a
security the higher its prospective dividend yield is, because of the higher rate of tax
levied on dividends than on capital gains." As we shall see in the next chapter, this
statement has empirical implications for the CAPM. It suggests that dividend payout
should be included as a second factor to explain the equilibrium rate of return on
securities. If true, the empirical CAPM would become
6 0 = a constant,
S i = influence of systematic risk on Rat,
5 2 = influence of dividend payout on Rat ,
flit = the systematic risk of the jth security,
divit /Pit = the dividend yield of the jth security,
= a random error term,
R f , = the risk-free rate.
If the dividend yield factor turns out to be statistically significant, then we might con-
clude that dividend policy is not irrelevant. Direct empirical tests of the relationship
between dividend yield and share value are discussed in Chapter 16.
A paper by Miller and Scholes [1978] shows that even if the tax on ordinary
personal income is greater than the capital gains tax, many individuals need not pay
more than the capital gains rate on dividends. The implication is that individuals
will be indifferent between payments in the form of dividends or capital gains (if the
firm decides to repurchase shares). Thus the firm's value may be unrelated to its divi-
dend policy even in a world with personal and corporate taxes.
12
The reader is referred to Chapter 13 for a complete discussion of this point.
560 DIVIDEND POLICY: THEORY
Table 15.3
Opening Balance Sheet Closing Balance Sheet
Assets Liabilities Assets Liabilities
2,500 shares 2,500 shares
at $10 = 25,000 Loan 16,667 at $10.60 = 26,500 Loan 16,667
Insurance 16,667 Net worth 25,000 Accrued Accrued
41,667 41,667 dividends 1,000 interest 1,000
Insurance 16,667 Net worth 26,500
44,167 44,167
Ordinary income Capital gains
Dividends received $1,000 Sale of 2,500 shares at $10.60 = $26,500
Less interest expense 1,000 Less original basis 25,000
0 1,500
Nontaxable income 1,000
1,000
To clarify their argument, Miller and Scholes provide the following simple ex-
ample. Let us suppose we have an initial net worth of $25,000, which is represented
wholly by an investment of 2500 shares worth $10 each in a company that earns
$1.00 per share. At the end of the year the company pays $0.40 per share in dividends
and retains $0.60. Consequently, its end-of-year price per share is $10.60. In order to
neutralize our dividend income for tax purposes, we borrow $16,667 at 6% and invest
the proceeds in a risk-free project (such as life insurance or a Keogh account) that
pays 6% of tax-deferred interest. Our opening and closing balance sheets and our
income statement are given in Table 15.3. Note that by investing in risk-free assets
we have not increased the risk of our wealth position. The riskless cash inflows from
insurance exactly match the required payments on debt. Our true economic income
could be $1,500 in unrealized capital gains plus the $1,000 of tax-deferred interest
from life insurance or our Keogh account.
Of course, federal tax laws are complex and these transactions cannot be carried
out without some transactions costs.' Nevertheless, the above argument is a clever
way to demonstrate the fact that ordinary income taxes on dividends can be avoided.
The 1986 tax code eliminated interest deductions on all forms of personal debt except
housing, where the amount of debt is limited to the original purchase price plus im-
provements. This shift in the tax code has caused a reorganization of the consumer
debt market and made home equity loans a growth business.
13
Also the maximum amount of dividends that can be sheltered in this way is $10,000. See Feenberg
[1981].
TOWARD A THEORY OF OPTIMAL DIVIDEND POLICY 561
14
Prior to the 1986 tax code the dividend exclusion rate was 85%.
562 DIVIDEND POLICY: THEORY
Figure 15.2
Corporate investment and dividend decisions with differing personal tax rates.
15
Given an individual tax rate of 40%, and a 15% before-tax rate on investment, the after-tax rate on a
dollar paid out as dividends should be
The firm has two categories of investment opportunity. First are investments in
real assets, represented by line segment AB and assumed to have diminishing returns
to scale. Second are investments in securities of other firms. These securities invest-
ments have constant returns to scale as illustrated by line segment BC. The before-tax
return on investments in securities of other firms is defined as I' s . There is a virtually
infinite amount of security investment (in assets of equivalent risk), but their after-tax
rate of return to the firm is affected by the fact that it must pay corporate taxes on
20% of the dividends it receives from ownership of other firms. Thus the after-tax
return on security investments is
To reach its optimal investment/dividend decision the firm in Fig. 15.2(a) uses internal
funds to undertake all investments in real assets, IA , and then invests in securities of
other firms up to the amount /*. At this point it stops because the after-tax return
on investing in securities is less than the opportunity cost of capital for externally
supplied equity, and we see that the investment in real assets, I A , is less than total
investment, I*. Since all internal funds have been used, dividends will not be paid
out. The high tax bracket shareholders, in Fig. 15.2(a), prefer low (or zero) dividend
payout.
In Fig. 15.2(b), which has the same investment schedule, low tax bracket share-
holders have a higher opportunity cost for internally generated funds. They will want
investment in real assets to stop at I A = I*. At this point, not all internally generated
capital has been spent on real investment and dividends are paid out. For low tax
bracket shareholders the cost of deferring dividends is sufficiently high that they pre-
fer dividend payout.
One of the implications of this model is that shareholders with different tax rates,
Tdi , will not unanimously agree on the firm's investment/dividend decision. High tax
bracket shareholders would prefer the firm to invest more, whereas low tax bracket
shareholders would prefer less investment and more dividend payout. This lack of
unanimity can be diminished somewhat if investors self-select into clienteles with low
tax bracket individuals purchasing shares of high-dividend firms and vice versa. Em-
pirical evidence on dividend clienteles is reviewed in Chapter 16.
There are (at least) five other implications of the Masulis-Trueman model. (1)
Firms are predicted not to externally finance security purchases for investment pur-
poses. However, they are likely to purchase marketable securities with internally
generated funds that remain after financing their own profitable production oppor-
tunities. (2) Firms with many profitable production opportunities (high-growth firms)
will use up all their internally generated funds without paying dividends, but older,
more mature firms will pay dividends because not all internally generated funds will
be exhausted by investment opportunities. (3) Mergers are predicted between firms
where one is internally financing its profitable investments and the other is externally
financing. (4) While a decrease in current earnings should leave unchanged the in-
vestment expenditures of externally financed firms, it is likely to decrease investment
expenditures of firms that initially planned to internally finance all their investments
564 DIVIDEND POLICY: THEORY
rather than to make up the shortfall of funds through external financing. (5) Share-
holder disagreement over internally financed investment policy will be more likely
the greater the amount of internally generated funds relative to the firm's investment
opportunities. In these cases, firms are more likely to experience takeover attempts,
proxy fights, and efforts to "go private." Given these tax-induced shareholder con-
flicts, diffuse ownership is more likely for externally financed firms than for internally
financed firms.
in order to pay the cash dividend.' Hence the signaling value of dividends is posi-
tive and can be traded off against the tax loss associated with dividend income (as
opposed to capital gains). Even firms that are closely held would prefer to pay divi-
dends because the value induced by the signal is received by current owners only
when the dividend message is communicated to outsiders. One of the important im-
plications of this signaling argument is that it suggests the possibility of optimal divi-
dend policy. The signaling benefits from paying dividends may be traded off against
the tax disadvantages in order to achieve an optimal payout.
Hakansson [1982] has expanded the understanding of informative signaling to
show that in addition to being informative at least one of three sufficient conditions
must be met. Either investors must have different probability assessments of dividend
payouts, or they must have differing attitudes about how they wish to allocate con-
sumption expenditures over time, or the financial markets must be incomplete. All
three of these effects may operate in a complementary fashion, and all three are
reasonable.
Miller and Rock [1985] develop a financial signaling model founded on the con-
cept of "net dividends." It is the first theory that explicitly combines dividends and
external financing to show that they are merely two sides of the same coin. The
announcement that "heads is up" also tells us that "tails is down." As was pointed
out in the original Miller-Modigliani [1961] article, every firm is subject to a sources
and uses of funds constraint:
Recall that sources of funds are NOI, the firm's net operating income; mP, the pro-
ceeds from an issue of external equity (the number of new shares, m, times the price
per share, P); and AB, the proceeds from new debt. Uses of funds are investment, I,
and dividends, Div. The sources and uses constraint can be rearranged to have net
cash flows from operations on the left-hand side and the firm's "net dividend" on the
right-hand side:
Now imagine a model where time 1 is the present, time 0 is the past, and time 2 is
the future. The present value of the firm, cum dividend, is the value of the current
dividend, Div,, plus the discounted value of cash flows (discounted at the appropriate
risk-adjusted rate, k):
E( NO12)
V, = Div, + (15.38)
1+k
16
This suggests that dividend payout and debt level increases are interrelated signals. A firm that simul-
taneously pays dividends and borrows may be giving a different signal than if it had made the same
dividend payment without borrowing.
566 DIVIDEND POLICY: THEORY
Original shareholders' wealth is the value of the firm minus the market value of debt
and new equity issued:
E( NOI<)
S1 = — AB' =. + 1 + ABI— mP1. (15.39)
k
E(NOI 2)
Si = NOT, — + (15.40)
1+k
Without any information asymmetry, this is just the original Miller-Modigliani prop-
osition that dividends are irrelevant. All that counts is the investment decision.
If there is information asymmetry, Eq. (15.40) must be rewritten to show how
market expectations are formed. If future earnings depend on current investment,
then we can write that net operating income is a function of the amount of investment
plus a random error term,
NOI, = f( 1o) + e„
NO12 = f(11 ) + 8 2 ,
where e l and 8 2 are random error terms with zero mean, i.e., E(8 1 ) = E(e 2 ) = 0. We
also adopt the special assumption that the expectation of 6 2 given 8, is not necessarily
zero:
E( 8 2I 61) =
If y is interpreted as a persistence coefficient, 0 < y < 1, the market is assumed to
only partially adjust to new information (the first-period error). If we use the notation
E 0 to remind us that the current value of the firm is based on preannouncement
information, then the current expected value of shareholders' wealth is
E 0[ f(I ,)]
E(S 1 ) = E 0 (N011) — E 1) +
1+k
f (I1)
= f(10) — 11 + 1 + k (15.41)
E 1( N01 2 )
S = NOI, — / 1 +
1+k
f( 11) + 8 <
= f(10)+ el— +
1+k
1
= f 10) + e ÿ f( 1) + E (15.42)
1+ k •
TOWARD A THEORY OF OPTIMAL DIVIDEND POLICY 567
S, — E(S 1) = e i [l +
1+k
= [NOT, — E o(NOI,)] [1 + 1 ±
'? (15.43)
Equation (15.43) says that the announcement effect on shareholders' wealth will de-
pend on the "earnings surprise." Thus we would expect that unexpected changes in
earnings will be correlated with share price changes on the announcement date.
Miller and Rock go on to show that the earnings, dividend, and financing an-
nouncements are closely related. Assuming that the expected and actual investment
decisions are at an optimum level, and are therefore equal, then the difference between
the actual and net dividends is
Div, — AB, — tn,P, — E o(Div, — AB, — m 1 P 1) = NOI, — I, — [E(NOI,) — I I ]
= NOI, — E(NOI,).
Thus the earnings surprise and the net dividend surprise can convey the same in-
formation. The financing announcement effect is merely the dividend announcement
effect, but with the sign reversed. An unexpected increase in dividends will increase
shareholders' wealth, and an unexpected issue of new equity or debt will be inter-
preted as bad news about the future prospects of the firm.
The Miller-Rock signaling approach shows that announcement effects (including
earnings surprises, unexpected dividend changes, and unexpected external financing)
emerge naturally as implications of the basic valuation model rather than as ad hoc
appendages.
One problem that the above theories have in common is that although they
explain how an optimal dividend policy may arise, none of them can successfully
explain cross-sectional differences in dividend payouts across firms.'
17
A possible exception is the work of Miller and Rock [1985], which suggests that the next theory shows
better promise in this regard.
18
The payout ratio is the ratio of dividends to net income.
568 DIVIDEND POLICY: THEORY
owners. The outsiders will charge, ex ante, for the potential problem that owner-
managers may increase their personal wealth at the expense of outsiders by means of
more perquisites or shirking. To decrease the ex ante charge, owner-managers will
find it in their own interest to agree to incur monitoring or bonding costs if such
costs are less than the ex ante charge that outsiders would be forced to request.
Thus a wealth-maximizing firm will adopt an optimal monitoring/bonding policy that
minimizes agency costs.
Dividend payments may well serve as a means of monitoring or bonding man-
agement performance. Although greater dividend payout implies costly external fi-
nancing, the very fact that the firm must go to the capital markets implies that it will
come under greater scrutiny. For example, banks will require a careful analysis of the
creditworthiness of the firm, and the Securities and Exchange Commission will re-
quire prospectus filings for new equity issues. Thus outside suppliers of capital help
to monitor the owner-manager on behalf of outside equity owners. Of course, audited
financial statements are a substitute means for supplying the same information, but
they may not be a perfect substitute for the "adversary" relationship between the firm
and suppliers of new capital.
Because of the transactions costs of external financing, Rozeff also argues that
the variability of a firm's cash flows will affect its dividend payout. Consider two
firms with the same average cash flows across time but different variability. The firm
with greater volatility will borrow in bad years and repay in good. It will need to
finance externally more often. Consequently, it will tend to have a lower dividend
payout ratio.
Rozeff [1982] selected a sample of 1000 nonregulated firms in 64 different in-
dustries and examined their average dividend payout ratios during the 1974-1980
interval. Five proxy variables were chosen to test his theory. The results are shown
in Table 15.4. The independent variables GROWL and GROW2 are an attempt to
measure the effect of costly external financing. Firms that grow faster can reduce
their need to use external financing by paying lower dividends. GROWL measures the
growth rate in revenues between 1974 and 1979, whereas GROW2 is Value Line's
forecast of the growth of sales revenue over the five-year period 1979-1984. Both
variables are negatively related to dividend payout and are statistically significant.
The variables INS and STOCK are proxies for the agency relationship. INS is the
percentage of the firm held by insiders. Dividend payout is negatively related to the
percentage of insiders because given a lower percentage of outsiders there is less need
to pay dividends to reduce agency costs.' On the other hand, if the distribution of
outsider holdings is diffuse, then agency costs will be higher; hence one would expect
STOCK, the number of stockholders, to be positively related to dividend payout.
Both INS and STOCK are statistically significant and of the predicted sign. Finally,
the variable BETA measures the riskiness of the firm. The prediction that riskier firms
have lower dividend payout is verified by the regression.
' 9 This relationship is also consistent with the tax argument that assumes that high tax bracket insiders
prefer to take their return in the form of capital gains rather than dividends.
OTHER DIVIDEND POLICY ISSUES 569
(=) 47.8= -0.090 -0.3<= -0.P<6 - <6.P43 <.P84 0.48 =.88 =8P.47
(=<.83) ( -4.=0) (-6.38) ( -6.43) ( -=7.0P) (7.73)
(<) <4.73 -0.068 -0.474 -0.7P8 <.P=7 0.33 =.79 =<3.<3
(6.<7) ( - <.7P) ( -8.44) ( -8.<8) (6.63)
(3) 70.63 -0.40< -0.603 - <P.409 0.4= =.88 <3=.46
(40.3P) ( -7.P8) ( -6.94) ( - =P.3P)
(4) 39.P6 -0.==6 - 33.P06 3.=P= 0.39 =.80 <=8.=0
(=0.0<) ( -4.9<) ( -<=.<8) (8.8<)
(P) =.03 -0.=0< 3.4<9 0.=< =.60 69.33
(0.<4) ( - 3.60) (7.97)
2
t-statistics are shown in parentheses under estimated values of the regression coefficients. R is adjusted
for degrees of freedom. D.W. is Durîin-Watson statistic.
From M. Rozeff, "Growth, Beta, and Agency Costs as Determinants of Dividend Payout Ratios," Jou]nal
of Financial Resea]ch, Fall =98<, <49-<P9. Reprinted with permission.
The best regression in Table 15.4 explains 48% of the cross-sectional variability
in dividend payout across individual firms. Although the results cannot be used to
distinguish among various theories of optimal dividend policy, they are consistent
with Rozeff's predictions. Furthermore, the very existence of strong cross-sectional
regularities suggests that there is an optimal dividend policy.
thereby leaving bondholders with a claim to nothing. For this very reason, most bond
indentures explicitly restrict the dividend policy of shareholders. Usually dividends
cannot exceed the current earnings of the firm, and they cannot be paid out of retained
earnings.
Other types of capital distributions are share repurchase and spinoffs. Share re-
purchase has exactly the same effect as dividend payment except that the form of
payment is capital gains instead of dividend income. The conventional procedure for
a spinoff is to take a portion of a firm's assets, often a division relatively unrelated
to the rest of the firm, and create an independent firm with these assets. The impor-
tant fact is that the shares of the new entity are distributed solely to the shareholders
of the parent corporation. Therefore, like dividend payment or share repurchase, this
may be used as a technique for taking collateral from bondholders. Empirical evi-
dence on the effects of repurchases and spinoffs is covered in Chapter 16.
It is an interesting empirical question whether or not any dividend payment, no
matter how large it is, will affect the market value of bonds. One would expect that
the market price of bonds would reflect the risk that future dividend payments would
lower the asset base that secures debt.' However, as changes in the dividend pay-
ments are actually realized, there may be changes in the expectations of the bond-
holders, which in turn would be reflected in the market price of bonds. All other
things being equal, we may expect that higher dividend payments or share repurchases
will be associated with a decline in the market value of debt. However, rarely do we
have a situation where all other things are equal. For example, if announcements
about dividend changes are interpreted as information about future cash flows, then
a dividend increase means that current debt will be more secure because of the antic-
ipated higher cash flows, and we would observe dividend increases to be positively
correlated with increases in the market value of debt.
20
Dividend payments do not necessarily change the assets side of the balance sheet. When cash balances
are reduced in order to pay dividends, there is an asset effect. However, it is not necessary. Dividends can
also be paid by issuing new debt or equity. In this case, assets remain unaffected, and the dividend decision
is purely financial in nature.
SUMMARY 571
SUMMARY
Several valuation models with or without growth and with or without corporate taxes
have been developed. Dividend policy is irrelevant in all instances. It has no effect
on shareholders' wealth. When personal taxes are introduced we have a result where
dividends matter. For shareholders who pay higher taxes on dividends than on capital
gains, the preferred dividend payout is zero; they would rather have the company
distribute cash payments via the share repurchase mechanism. Yet corporations do
pay dividends. The Rozeff [1982] paper suggests that there appear to be strong cross-
sectional regularities in dividend payout. Thus there may be optimal dividend policies
that result from a trade-off between the costs and benefits of paying dividends. The
list of possible costs includes (1) tax disadvantages of receiving income in the form
of dividends rather than capital gains, (2) the cost of raising external capital if divi-
dends are paid out, and (3) the foregone use of funds for productive investment. The
possible benefits of dividend payout are (1) higher perceived corporate value because
572 DIVIDEND POLICY: THEORY
of the signaling content of dividends, (2) lower agency costs of external equity, and
(3) the ability of dividend payments to help complete the markets.
The next chapter presents empirical evidence that examines the relationship
among the level of dividend payout, the value of the shares, and the amount of in-
vestment undertaken by the firm. Also, changes in dividend payout are studied in
relation to their effect on share value. These studies provide greater insight into the
empirical validity of the theories discussed here.
PROBLEM SET
15.1 The chairman of the board of Alphanull Corporation has announced that the corporation
will change its dividend policy from paying a fixed dollar dividend per share. Instead, dividends
will be paid out as a residual. That is, any cash flows left over after the firm has undertaken
all profitable investments will be paid out to shareholders. This new policy will obviously
increase the variability of dividends paid. How do you think it will affect the value of the firm?
15.2 The XYZ Company (an all-equity firm) currently has after-tax operating cash flows of
$3.00 per share and pays out 5011 of its earnings in dividends. If it expects to keep the same
payout ratio, and to earn 20% on future investments forever, what will its current price per
share be? Assume that the cost of capital is 15%.
15.3 The Highrise Investment Co. (an all-equity firm) currently pays a dividend of $2.00 per
share, which is 75% of its after-tax cash flows from operations. It is currently selling for $16
a share and earns 40% on invested capital. Its equity /3 is 2.0, the expected market rate of
return is 12.5%, and the risk-free rate is 57g. How long will its supernormal rate of growth
last before it levels off to equal the normal rate for a company with its risk?
15.4 The balance sheet of the Universal Sour Candy Company is given in Table Q15.4. Assume
that all balance sheet items are expressed in terms of market values. The company has decided
to pay a $2000 dividend to shareholders. There are four ways to do it:
1) Pay a cash dividend.
2) Issue $2000 of new debt and equity in equal proportions ($1000 each) and use the proceeds
to pay the dividend.
3) Issue $2000 of new equity and use the proceeds to pay the dividend.
4) Use the $2000 of cash to repurchase equity.
What impact will each of the four policies above have on the following?
a) The systematic risk of the portfolio of assets held by the firm,
b) The market value of original bondholders' wealth,
vL = E(NOI,)(1 — VU
WACC
15.7 Calculate the value of a company that earned $50,000 this year, has a 40% investment
rate, K, and a tax rate of 40%; has $200,000 in debt outstanding and a weighted average cost
of capital of 12%; and is expected to earn 40% on invested capital for the next five years, then
25% for the following five years, before the rate of return declines to a normal rate of growth.
15.8 How does an increase in the investment (retention) rate affect the anticipated stream of
investments that a company will undertake?
15.9 It was suggested that if a firm announces its intention to increase its dividends (paid from
cash), the price of common stock increases, presumably because the higher dividend payout
represents an unambiguous signal to shareholders that anticipated cash flows from investment
are permanently higher. A higher level of cash flows is also beneficial to bondholders because
it diminishes the probability of default. If dividends are paid from cash, what does the OPM
suggest will happen to the market value of debt? How does this contrast with the prediction
in the above paragraph?
REFERENCES
Bar-Yosef, S., and R. Kolodny, "Dividend Policy and Market Theory," Review of Economics
and Statistics, May 1976, 181-190.
Bhattacharya, S., "Imperfect Information, Dividend Policy, and 'The Bird in the Hand'
Fallacy," Bell Journal of Economics, Spring 1979, 259-270.
Black, F., "The Dividend Puzzle," Journal of Portfolio Management, Winter 1976, 5-8.
; M. Jensen; and M. Scholes, "The Capital Asset Pricing Model: Some Empirical Tests,"
in Jensen, ed., Studies in the Theory of Capital Markets. Praeger, New York, 1972, 79-124.
Black, F., and M. Scholes, "The Effects of Dividend Yield and Dividend Policy on Common
Stock Prices and Returns," Journal of Financial Economics, May 1974, 1-22.
Brennan, M., "Taxes, Market Valuation and Corporate Financial Policy," National Tax Jour-
nal, December 1970, 417-427.
, and T. Copeland, "A Model of Stock Split Behavior: Theory and Evidence," Working
paper, Anderson Graduate School of Management, UCLA, April 1987.
Brittain, J. A., Corporate Dividend Policy. Brookings Institute, Washington, D.C., 1966.
P74 DIVIDEND POLICY: THEORY
Darling, P., "The Influence of Expectations and Liquidity on Dividend Policy," Jou]nal of
Political Economy, June =9P7, <09-<<4.
Doîrovolsky, S., The Economics of Co]po]ation Finance. McGraw-Hill, New York, =97=.
Fama, E., "The Empirical Relationships îetween the Dividend and Investment Decisions of
Firms," Ame]ican Economic Review, June =974, 304-3=8.
, and H. Baîiak, "Dividend Policy: An Empirical Analysis," Jou]nal of the Ame]ican
Statistical Association, Decemîer =968, ==3<-==6=.
Fama, E., and M. H. Miller, The Theo]y of Finance. McGraw-Hill, New York, =97=.
Farrar, D., and L. Selwyn, "Taxes, Corporate Financial Policy and Return to Investors,"
National Tax Jou]nal, Decemîer =967, 444-4P4.
Feenîerg, D., "Does the Investment Interest Limitation Explain the Existence of Dividends,"
Jou]nal of Financial Economics, Septemîer =98=, <6P-<70.
Friend, I., and M. Puckett, "Dividends and Stock Prices," Ame]ican Economic Review, Sep-
temîer =964, 6P6-68<.
Gordon, M., "Dividends, Earnings, and Stock Prices," Review of Economics and Statistics, May
=9P9, 99-=0P.
, "The Savings, Investment and Valuation of a Corporation," Review of Economics and
Statistics, Feîruary =96<, 37-P=.
Hakansson, N., "To Pay or Not to Pay Dividends," Jou]nal of Finance, May =98<, 4=P-4<8.
John, K., and J. Williams, "Dividends, Dilution and Taxes: A Signalling Equiliîrium," Jou]nal
of Finance, Septemîer =98P, =0P3-=070.
Kaplanis, C., "Options, Taxes, and Ex-dividend Day Behavior," Jou]nal of Finance, June =986,
4==-4<4.
Khoury, N., and K. Smith, "Dividend Policy and the Capital Gains Tax in Canada," Jou]nal
of Business Administ]ation, Spring =977.
Kim, E. H.; W. Lewellen; and J. McConnell, "Financial Leverage Clienteles: Theory and
Evidence," Jou]nal of Financial Economics, March =979, 83-==0.
Lintner, J., "Distriîution of Incomes of Corporations among Dividends, Retained Earnings
and Taxes," Ame]ican Economic Review, May =9P6, 97-==3.
, "Optimal Dividends and Corporate Growth under Uncertainty," Qua]te]ly Jou]nal
of Economics, Feîruary =964, 49-9P.
, "The Valuation of Risky Assets and the Selection of Risky Investments in Stock Port-
folios and Capital Budgets," Review of Economics and Statistics, Feîruary =96P, =3-37.
Long, J., Jr., "Efficient Portfolio Choice with Differential Taxation of Dividends and Capital
Gains," Jou]nal of Financial Economics, August =977, <P-P4.
Masulis, R., and B. Trueman, "Corporate Investment and Dividend Decisions under Differen-
tial Personal Taxation," Working paper, Anderson Graduate School of Management,
UCLA, =986.
Miller, M., "Deît and Taxes," The Jou]nal of Finance, May =977, <6=-<7P.
, and F. Modigliani, "Dividend Policy, Growth and the Valuation of Shares," Jou]nal
of Business, Octoîer =96=, 4==-433.
Miller, M. and K. Rock, "Dividend Policy Under Asymmetric Information," Jou]nal of Finance,
Septemîer =98P, =03=-=0P=.
Miller, M., and M. Scholes, "Dividends and Taxes," Jou]nal of Financial Economics, Decemîer
=978, 333 —364.
REFERENCES 575
Our discussion first deals with models that simply explain the behavior of corporate
dividend policy over time. Evidence indicates that U.S. corporations behave as if
they had some target dividend payout in mind and that they move toward it with a
lag. They also show reluctance to lower dividends.
Second, we look at the possibility of clientele effects. Do people in high tax
brackets avoid investing in high-dividend companies in order to escape higher income
taxes on dividend income? On this question, the empirical evidence is mixed, although
it does lean toward the existence of a clientele effect.
Third, the information content of dividend increases is tested. There is reasonably
strong evidence that leans toward validation of the signaling hypothesis.
Fourth, we focus on the relationship between dividend yield and the market
value of equity. The best empirical evidence indicates that dividend yield is at most
weakly related to the value of the firm. Several explanations are given for why this
result is plausible, given our current tax system.
Fifth, the empirical literature relating to share repurchases via tender offer is
reviewed. The announcement of share repurchases tends to be interpreted as unan-
ticipated favorable news about the value of the company.
576
BEHAVIORAL MODELS OF DIVIDEND POLICY P77
A. BEHAVIORAL MODELS OF
DIVIDEND POLICY
One can conclude that U.S. corporations seem to increase dividends only after
they are reasonably sure that they will be able to maintain them permanently at the
new level. However, this does not help to answer the question of why corporations
pay dividends in the first place.
The clientele effect is a possible explanation for management reluctance to alter es-
tablished payout ratios because such changes might cause current shareholders to
incur unwanted transactions costs.
Elton and Gruber [1970] attempt to measure clientele effects by observing the
average price decline when a stock goes ex-dividend. If we were current shareholders
and sold our stock the instant before it went ex-dividend, we would receive its price,
PB, and pay the capital gains rate, tg , on the difference between the selling price and
the price at which it was purchased, Pc . Alternatively, we could sell the stock after
it went ex-dividend. In this case we would receive the dividend, div, and pay the
ordinary tax rate, t o , on it. In addition, we would pay a capital gains tax on the
difference between its ex-dividend price, P A , and the original purchase price Pc . To
prevent arbitrage profits, our gain from either course of action must be the same,
namely,
PB— t g(PB — P
c) = P
A — tg( P A — Pc) + diva — t 0). (16.2)
Using 4148 observations between April 1, 1966, and March 31, 1967, Elton and
Gruber discovered that the average price decline as a percentage of dividend paid
was 77.7%. This implied that the marginal tax bracket of the average investor was
36.4%. They continued by arguing that
. . . the lower a firm's dividend yield the smaller the percentage of his total return that a
stockholder expects to receive in the form of dividends and the larger the percentage he
expects to receive in the form of capital gains. Therefore, investors who held stocks which
have high dividend yields should be in low tax brackets relative to stockholders who hold
stocks with low dividend yield. [Elton and Gruber, 1970]
Table 16.1 shows the dividend payout ranked from the lowest to highest deciles
along with (1) the average drop in price as a percentage of dividends and (2) the implied
tax bracket. Note that the implied tax bracket decreases when dividend payout in-
creases. Elton and Gruber conclude that the evidence suggests that Miller and
Modigliani were right in hypothesizing a clientele effect.
A possible counterargument to this interpretation, provided by Kalay [1977,
1982], is that arbitrage may also be carried out by traders who do not own the stock
initially. They would not receive favored capital gains treatment but would have to
pay ordinary income taxes on short-term gains. Their arbitrage profit, it, may be
stated mathematically as
7C = - PB div - t odiv + PA tO(PB - PA). (16.4)
PB - PA
div
Probability
div/P Standard Z True Mean Is Implied
Decile Mean Mean Deviation Value One or More Tax Bracket
* Spearman's rank correlation coefficient between div/P and (PB - PA)/div is .9152, which is significant
at the 1% level.
t Indeterminant.
From E. J. Elton and M. J. Gruber, "Marginal Stockholders' Tax Rates and the Clientele Effect,"
reprinted with permission from Review of Economics and Statistics, February 1970, 72.
580 DIVIDEND POLICY: EMPIRICAL EVIDENCE AND APPLICATIONS
They spend P E to acquire the stock before it goes ex-dividend, then receive the div-
idend and pay ordinary income taxes on it, and finally sell the stock after it goes ex-
dividend (receiving P A dollars) and receive a tax shield from their short-term loss.
Rearranging (16.4), we see that their profit is
it = (} - t o)[P A — P E + div]. (16.5)
To prevent arbitrage profits, the price decline must equal the amount of dividend
payout, i.e., Pg - PA = div.
The above condition is completely different from Eq. (16.3) proposed by Elton
and Gruber. Of course neither model has taken transactions costs into account.
Eades, Hess, and Kim [1984] replicate the Elton and Gruber work but report their
results in the form of rates of return. If the price decline on the ex-date is less than
the amount of the dividend, then the ex-date return,
Pt+i — Pt + div,
=
Pt
will be positive. For the time period July 2, 1962, to April 30, 1975, they find the
average ex-date excess return to be .176% (statistically significant). This time interval
predates the era of negotiated commissions. On May 1, 1975, all brokerage com-
missions were competitively negotiated, and presumably transactions costs fell. For
the time interval May 1, 1975, to December 31, 1980, Eades, Hess, and Kim found
the ex-date return to be significantly lower—only .064%. This result suggests that,
given lower transactions costs, it was easier for short-term traders to arbitrage, as
was suggested by Kalay [Eq. (16.5)].
Eades, Hess, and Kim [1984] also examine ex-date dividend returns for a non-
convertible preferred stock sample, characterized by a relatively large preferred
dividend yield. During the sample period, January 1, 1974, to December 31, 1981,
these securities had a total of 708 ex-days that occurred on 493 trading days. The
average excess return was a significantly negative — .141%. This implies that the stock
price fell by more than the amount of the dividend. These results are consistent with
tax-induced clienteles if the marginal purchasers are corporations. Corporations are
able to exclude 85% of any dividends (80% following the 1986 tax code) received as
taxable income, whereas capital gains are taxable at rates as high as 46% (less following
the 1986 tax code) if they are short-term capital gains To see how the price might
fall by more than the dividend, suppose the preferred stock is worth $40 and it pays
a $4 dividend. If the marginal purchaser is a corporation it receives the following
returns:
[div — .46(.15)div] + (PB — PA W — .46).
The first term is the dividend, minus the taxable portion (15% of the dividend times
the tax rate); the second term is the tax shelter from the short-term capital loss (taxed
at 46%). If we set this return equal to zero and solve, we have
PE — PA —[1 — .46(.15)]
172.4%.
div 1 — .46
CLIENTELE EFFECTS AND EX DATE EFFECTS 581
Thus the stock price could fall by as much as $4(1.724) = $6.90 in our example before
the corporation would not profit. For large dividends and on preferred stock we tend
to see security prices bid up prior to ex-dates.
Table 16.2 shows the pattern of excess returns, and t-statistics for all taxable dis-
tributions on NYSE common stocks. The puzzle here is that abnormal returns are
not uniquely associated with the ex-day. No good explanation for this result has yet
been proposed.
Finally, Eades, Hess, and Kim report on the ex-date behavior of nontaxable cor-
porate distributions. They find significant positive returns for stock splits and stock
dividends (later confirmed by Grinblatt, Masulis, and Titman [1984]) and significant
negative returns for nontaxable cash dividends (primarily of high-yielding utilities).
Although there is no explanation for the abnormally positive split ex-date returns, we
may conjecture that tax arbitrage (short-term capital gains shelters) may explain the
negative returns on nontaxable cash dividends.
Lakonishok and Vermaelen [1986] test the hypothesis of tax arbitrage by study-
ing trading volume around the ex-date. If there is tax arbitrage, then volume should
be abnormally high around ex-dates, and it should be positively related to dividend
yield and negatively related to transactions costs. Their results show that trading
volume does increase significantly around ex-dates and that it is more pronounced
for high-yield, actively traded stocks and during the period following the introduction
of negotiated trading commissions.
582 DIVIDEND POLICY: EMPIRICAL EVIDENCE AND APPLICATIONS
Pettit [1977] has tested for dividend clientele effects by examining the portfolio
positions of approximately 914 individual accounts handled by a large retail broker-
age house between 1964 and 1970. He argues that stocks with low dividend yields will
be preferred by investors with high income, by younger investors, by investors whose
ordinary and capital gains tax rates differ substantially, and by investors whose port-
folios have high systematic risk. His model is
Dê = a, + a,fl i + a 3 AGE, + a 4INC, + a 5 DTR, + e i , (16.6)
where
DY = dividend yield for the ith individual's portfolio in 1970,
13 i = the systematic risk of the ith individual's portfolio,
AGE, = the age of the individual,
INC, = the gross family income averaged over the last three years,
DTR, = the difference between the income and capital gains tax rates for the ith
individual,
e i = a normally distributed random error term.
He finds that'
DY = .042 — .021/3, + .031AGE, — .037INC, + .006DTR,.
(11.01) ( — 16.03) (6.15) ( -2.25) (1.57)
The evidence suggests that there is a clientele effect because a significant portion of
the observed cross-sectional variation in individual portfolio dividend yields can be
explained. However, the study in no way suggests that the market price of a security
is determined by the dividend policy followed by the firm.
A second study by Lewellen, Stanley, Lease, and Schlarbaum [1978] was drawn
from the same data base as the Pettit study but reached different conclusions. They
ran a multiple regression to explain the dividend yields of investor portfolios as a
function of various investor characteristics. Although the tax rate variable was nega-
tively related to dividend yield and was statistically significant, it implied that a 10%
increase in an inve.stor's marginal (imputed) tax bracket was associated with only a
.1% decline in the yield of securities held. This suggests only a very weak dividend
clientele effect.
The numbers in parentheses are t-statistics. The r 2 was .3 for 914 observations.
CLIENTELE EFFECTS AND EX DATE EFFECTS 583
debt by individuals in high marginal personal tax brackets, and low-income investors
will prefer to invest in firms with high leverage.
Kim, Lewellen, and McConnell [1979] and Harris, Roenfeldt, and Cooley [1983]
tested for leverage-related clientele effects. The first study used a data set consisting
of 1140 companies whose stock was owned by at least three investors from among a
group of 887. Questionnaires provided demographic and income data for the inves-
tors during a three-year period from 1969 to 1971. From these data the marginal tax
rates of the investors were determined Table 16.3 shows the results of a multiple
regression that explains the corporate total debt to total capital ratio as a function
of various shareholder characteristics. The coefficients of five of the seven independent
variables are statistically significant as is the overall significance of the regression
(F = 7.51). The coefficient of the investor tax rate is of the correct sign and is sta-
tistically significant, but its magnitude indicates that an increase in an investor's per-
sonal tax rate from zero to 70% is associated with an increase of only 5% in the
corporate leverage ratio. Kim, Lewellen, and McConnell [1979] suggest, "A relation-
ship that slight has to be interpreted as somewhat less than strongly supportive of
the financial leverage clientele hypothesis."
Harris, Roenfeldt, and Cooley [1983] estimate investor-implied tax rates by using
the Elton and Gruber procedure, Eq. (16.3). They then examine a sample of large
firms, 1968 to 1976, to see if firms with high leverage have investors with low tax
rates, and vice versa. They find that implied tax rates are strongly negatively cor-
related with corporate financial leverage, thereby lending further support to the lever-
age clientele hypothesis.
584 DIVIDEND POLICY: EMPIRICAL EVIDENCE AND APPLICATIONS
C. DIVIDEND ANNOUNCEMENT
EFFECTS ON THE VALUE OF THE
FIRM: THE SIGNALING HYPOTHESIS
Most firms that pay dividends exhibit behavior that results in constant dividend
payouts that are increased only when management is relatively certain that the higher
dividend payout can be maintained indefinitely. Given this type of management be-
havior, it is likely that investors will interpret an increase in current dividend payout
as a message that management anticipates permanently higher levels of cash flows
from investment. We may therefore expect to observe an increase in share prices
associated with public announcement of a dividend increase. The dividend per se does
not affect the value of the firm. Instead it serves as a message from management that
the firm is anticipated to do better. If dividend changes are to have an impact on
share values, it is necessary that they convey information about future cash flows, but
it is not sufficient. The same information may be provided to investors via other
sources.' Therefore it becomes an empirical question whether or not announcements
of dividend changes actually affect share value.
The first study to look at this issue was the stock split study of Fama, Fisher,
Jensen, and Roll [1969], which was discussed in Chapter 11. They found that when
splits were accompanied by dividend announcements, there was an increase in ad-
justed share prices for the group that announced dividend increases and a decline in
share prices for the dividend decrease group. More recent studies of the effect of un-
expected dividend changes on share prices have been made by Pettit [1972], Watts
[1973], Kwan [1981], and Aharony and Swary [1980].
Watts finds a positive dividend announcement effect but concludes that the infor-
mation content is of no economic significance because it would not enable a trader
with monopolistic access to the information to earn abnormal returns after transac-
tions costs. On the other hand, Pettit finds clear support for the proposition that
the market uses dividend announcements as information for assessing security values.
Their methodologies are also different. Watts proceeds in two stages. First, he de-
velops a model to predict dividend changes. It is the same model that Fama and
Babiak [1968] found to provide the best prediction of next period's dividends. It may
be written as follows:
ADiv, = i Div, _ = + + )6 31\11P _ = + Z,, (16.7)
where
ADiv, = the change in dividends in period t,
Div,_, = the previous period's dividends,
NI, = this period's earnings,
Nit -1 = last period's earnings,
Z, = unanticipated dividend changes (the error term).
<
Ross [=977] argues that an increase in dividend payout is an unamîiguous message îecause (=) it cannot
îe mimicked îy firms that do not anticipate higher earnings and (<) management has an incentive to
"tell the truth."
DIVIDEND ANNOUNCEMENT EFFECTS ON THE VALUE OF THE FIRM 585
Using this equation, we are able to estimate unanticipated dividend changes, Z. Next,
an abnormal performance index that measures departures from the risk-adjusted rate
of return can be constructed from the market model,
R j, = a + fi 1R„„ + ei„ (16.8)
where
R 1, = the total return (dividends and capital gains) on the common stock of the jth
firm,
a = a constant term,
fi 1 = systematic risk,
R., = the rate of return on a market index,
e j, = the abnormal performance of the jth security
The abnormal performance index (API) for a security is computed as the product of
its one-month abnormal returns:
API = 11 (1 + eit), T = 1, . , N.
t=i
Watts looked at the abnormal performance index averaged across 310 firms. The
abnormal performance index for 24 months around the dividend announcement for
the subsamples of firms that had unanticipated dividend increases or decreases is
given in Table 16.4. The performance of firms with dividend increases is better than
that of firms with dividend decreases, but the greatest difference between the two
samples in the 6 months around the dividend change is only .7% in the month of the
dividend. This is a trivial difference.
Pettit used both monthly and daily data to investigate the abnormal performance
index of firms that had dividend changes of — 1% to — 99%, 1% to 10%, 10% to 25%,
and over 25%. Figure 16.1 shows the cumulative abnormal performance index using
daily data for 135 firms. Most of the price adjustment takes place very quickly either
on the dividend announcement date or on the following day. Furthermore, the price
changes appear to be significant. This leads Pettit to conclude that substantial infor-
mation is conveyed by the announcement of dividend changes.
Pettit's results have been criticized because he used the observed dividend changes
rather than the unexpected dividend changes. Kwan [1981] has improved on Pettit's
design by forming portfolios based on unexpected dividend changes, and he finds
statistically significant abnormal returns when firms announce unexpectedly large
dividend changes. A study by Aharony and Swary [1980] separates the information
content of quarterly earnings reports from that of unexpected quarterly dividend
changes. They examine only those quarterly dividend and earnings announcements
made public on different dates within any given quarter. Their findings strongly sup-
port the hypothesis that changes in quarterly cash dividends provide useful informa-
tion beyond that provided by corresponding quarterly earnings numbers. Kane, Lee,
and Marcus [1984] also select a set of firms whose quarterly dividend and earnings
announcements are separated by at least 10 days, build models to predict expected
586 DIVIDEND POLICY: EMPIRICAL EVIDENCE AND APPLICATIONS
earnings and dividends, and then test to see if unexpected dividend and earnings an-
nouncements corroîorate each other-in other words, is there an interaction effect?
Their empirical results confirm the earlier studies that found that îoth earnings and
dividend announcements have a significant effect on share price, and in addition they
find a significant corroîoration effect.
Woolridge [=983] studies the effect of dividend announcements on nonconvertiîle
îonds and nonconvertiîle preferred stock in an attempt to separate expropriation
effects from announcement effects. If dividend payouts to shareholders are viewed
as payments of collateralizaîle assets, and if deît covenants are imperfect protec-
tion, then deît holders and preferred shareholders would view dividend increases as
îad news and the market value of their claims on the firm would fall upon the an-
DIVIDEND ANNOUNCEMENT EFFECTS ON THE VALUE OF THE FIRM 587
1.04
- --' 30 firms 110% to
39 firms )25%
1.02 36 firms t 1% to
39 firms J 10%
—
1Almwelgill
- -.111110
."6000. ' 13 firms i Over
Lena,......
1.00 '4p-a li
4marun jigia,
lr,A1
..................---* --. .....--- ../
••• 9 firms J 25%
... . 7( .... .../...- - -.• \ 10 firms
r.' ■ .... ..-.
b
0.90 1
-9 -6 -3 0 3 7 10
Day relative to announcement day
Figure 16.1
Abnormal performance index for dividend announcement effects, using daily data. (From
R. R. Pettit, "Dividend Announcements, Security Performance and Capital Market
Efficiency," reprinted with permission of Jou]nal of Finance, December 1972, 1004.)
nouncement of dividend increases. On the other hand, if dividend increases are signals
about higher future cash flows, then bondholders and preferred stockholders should
feel more secure and the market value of their claims should increase. Woolridge's
empirical results support the signaling hypothesis (or at least the conclusion that the
signaling effect dominates any expropriation effect). Announcement date abnormal
returns are positive given unexpected dividend increases and negative given unex-
pected dividend decreases. Handjiinicolaou and Kalay [1984] find that for a sample
of 255 nonconvertible bonds, prices are unaffected by unexpected dividend increases
but react negatively to dividend reductions. They interpret this as evidence consistent
with the dividend signaling hypothesis.
Asquith and Mullins [1983] and Richardson, Sefcik, and Thompson [1986] study
the effect on shareholder wealth of the initial dividend announcement the firm's
first dividend (most firms had never paid a dividend, although a few had not paid a
dividend for 10 years). Both studies find large, statistically significant two-day an-
nouncement abnormal returns for initial dividend announcements, 3.7% to 4.0%. In
addition, Richardson, Sefcik, and Thompson (and Asquith and Krasker [1985]) study
trading volume around the announcement date, and between the announcement and
ex-dates. Unusual trading volume may be evidence of clientele changes induced when
588 DIVIDEND POLICY: EMPIRICAL EVIDENCE AND APPLICATIONS
high tax bracket shareholders sell out to low tax bracket investors when the higher
dividend payout is announced. Both studies find statistically significant abnormal
volume increases during the announcement week that are related to the information
content of dividends. There is only weak evidence for higher volume following the
announcement and hence only weak support for clientele adjustments.
Brickley [1983] studies the announcement effect of specially designated divi-
dends—those labeled by management as "extra," "special," or "year-end," and com-
pares them to surrounding regular dividend increases. Specially designated dividends
are interesting because they are not intended to be a part of continuing higher divi-
dend payout and may therefore not be interpreted by the market as a signal about
higher future cash flows. Brickley's results support the opposite conclusion namely,
that the market does react positively to the information content of specially desig-
nated dividends but that dollar-for-dollar regular dividends convey more information.
In sum the evidence in support of the informational content of dividends is over-
whelming Unexpected dividend changes do convey information to the market about
expected future cash flows.
In Chapter 15 we saw that in a world with only corporate taxes the Miller-Modigliani
proposition suggests that dividend policy is irrelevant to value. However, when per-
sonal taxes are introduced with a capital gains rate that is less than the rate on ordi-
nary income, the picture changes. Under this set of assumptions the firm should not
pay any dividends. One way to test these theories is to look directly at the relation-
ship between dividend payout and the price per share of equity.
Friend and Puckett [1964] use cross-section data to test the effect of dividend
payout on share value. Prior to their work, most studies had related stock prices to
current dividends and retained earnings, and reported that higher dividend payout
was associated with higher price/earnings ratios. The "dividend multiplier" was found
to be several times the "retained earnings multiplier." The usual cross-section equa-
tion was
suggests that riskier firms have both lower dividend payout and lower price/earnings
ratios. Consequently, the omission of a risk variable may cause an upward bias in
the dividend coefficient in Eq. (16.9). Second, there is almost no measurement error
in dividends, but there is considerable measurement error in retained earnings. It is
well known that accounting measures of income often imprecisely reflect the real
economic earnings of the firm. The measurement error in retained earnings will cause
its coefficient to be biased downward. Third, Friend and Puckett argue that even if
dividends and earnings do have different impacts on share prices, we should expect
their coefficients in (16.9) to be equal. In equilibrium, firms would change their divi-
dend payout until the marginal effect of dividends is equal to the marginal effect of
retained earnings. This will provide the optimum effect on their price per share.
No theory had been developed to allow the pricing of risk when they wrote their
paper, but Friend and Puckett were able to eliminate the measurement error on re-
tained earnings by calculating a normalized earnings variable based on a time series
fit of the following equation:
( NIIP)it
= a, + b it + e,„ (16.10)
(NI/P),
where
(NI/P),, = the earnings/price ratio for the firm,
(NI/P)„ = the average earnings/price ratio for the industry,
t = a time index,
8,, = the error term.
When normalized retained earnings were calculated by subtracting dividends from
normalized earnings and then used in Eq. (16.9), the difference between the dividend
and retained earnings coefficients was reduced. Unfortunately, no test was performed
to see whether the differences between the impact of retained earnings and dividends
were significant after Friend and Puckett had normalized earnings and controlled
for firm effects.
A study by Black and Scholes [1974] uses capital asset pricing theory to control
for risk.' Their conclusion is quite strong. "It is not possible to demonstrate, using
the best empirical methods, that the expected returns on high yield common stock
differ from the expected returns on low yield common stocks either before or after
taxes." They begin by pointing out that the assumption that capital gains tax rates
are lower than income tax rates does not apply to all classes of investors. Some classes
of investors might logically prefer high dividend yields. They include: (a) corporations,
because they usually pay higher taxes on realized capital gains than on dividend in-
come (because of the 80% exclusion of dividends); (b) certain trust funds in which
one beneficiary receives the dividend income and the other receives capital gains;
(c) endowment funds from which only the dividend income may be spent; and (d) in-
vestors who are spending from wealth and may find it cheaper and easier to receive
See Chapter 7 for a complete development of the capital asset pricing model.
590 DIVIDEND POLICY: EMPIRICAL EVIDENCE AND APPLICATIONS
dividends than to sell or borrow against their shares. Alternatively, investors who
prefer low dividend yield will be those who pay higher taxes on dividend income
than on capital gains. With all these diverse investors, it is possible that there are
clientele effects that imply that if a firm changes its dividend payout, it may lose some
shareholders, but they will be replaced by others who prefer the new policy. Thus
dividend payout will have no effect on the value of an individual firm.'
The Black and Scholes [1974] study presents empirical evidence that the before-
tax returns on common stock are unrelated to corporate dividend payout policy.
They adjust for risk by using the CAPM.
The CAPM predicts that the expected return on any asset is a linear function
of its systematic risk:
E(R;) = R f [E(1L) — R f ]ai. (16.11)
However, it is derived by assuming, among other things, that there are no differential
tax effects that would affect investors' demands for different securities. Brennan [1970]
has shown that if effective capital gains tax rates are lower than effective rates on
dividend income, then investors will demand a higher rate of return on securities
with higher dividend payout. Using annual data, Black and Scholes test this hypothe-
sis by adding a dividend payout term to an empirical version of the CAPM:
= + [1(Z ni — y o ]fli + y i (DYi — + ei, (16.21)
where
= the rate of return on the jth portfolio,
y o = an intercept term that should be equal to the risk-free rate, R f ,
according to the CAPM,
Rm = the rate of return on the market portfolio,
= the systematic risk of the jth portfolio,
y l = the dividend impact coefficient,
DYi = the dividend yield on the jth portfolio, measured as the sum of dividends
paid during the previous year divided by the end-of-year stock price,
DYn, = the dividend yield on the market portfolio measured over the prior
12 months,
= the error term.
If the coefficient, y l , of the dividend yield is significantly different from zero, we would
reject the null hypothesis that dividend payout has no impact on the required rate
of return for securities. The results of Black and Scholes are summarized in Table
16.5. Note that the dividend impact coefficient, yi , is not significantly different from
This does not rule out the possibility that in aggregate there is a desired equilibrium amount of dividend
payout. For example, in the United States there are obviously a far greater number of companies with
generous dividend payout than without.
THE RELATIONSHIP BETWEEN DIVIDENDS AND VALUE P9=
zero (since the t-test is less than the level required to make it significant at the 9P%
confidence level) across the entire time period, =936 through =966, or in any suîperiod.
This means that the expected returns on high-yield securities are not significantly
different from the expected returns on low-yield securities, other things îeing equal. 5
The Black-Scholes study has îeen criticized îecause their test is not very pow-
erful. Had the null hypothesis îeen that dividend yield does matter, it could not have
îeen rejected either. Their test is inefficient îecause they grouped stocks into port-
folios instead of using individual stock returns and, perhaps, îecause they used an-
nual data.
Litzenîerger and Ramaswamy [=979] also test the relationship îetween dividends
and security returns. They use the Brennan [=970] model, Eq. (=6.=3) îelow, with
monthly data for individual securities:
E(kit ) - R ft = a t + a 2 / 1 + a 3(Dêit - R ft ), (=6.=3)
The lower half of Taîle =6.P shows that yo is significantly different from the risk-free rate. This is not
important for the conclusions aîout dividend policy îut is consistent with other empirical work (e.g.,
Black, Jensen, and Scholes [=97<]) that shows that the intercept term in the CAPM is different from what
theory would predict.
592 DIVIDEND POLICY: EMPIRICAL EVIDENCE AND APPLICATIONS
where
Litzenberger and Ramaswamy conclude that risk-adjusted returns are higher for
securities with higher dividend yields. The implication is that dividends are undesir-
able; hence higher returns are necessary to compensate investors in order to induce
them to hold high dividend yield stocks.
There are (at least) three serious problems with testing for the dividend effect
predicted by Eq. (16.13). The first is that investors use dividend announcements to
estimate expected returns, E(Rit ); i.e., there is an information effect. The second is
that measures of systematic risk, flf , are subject to a great deal of error. And the third
is that individual security returns (rather than portfolio returns) are needed to obtain
statistically powerful results. Litzenberger and Ramaswamy [1979] largely solved the
second and third problems but have been criticized by Miller and Scholes [1982]
for their handling of the information effect of dividend announcements. When using
monthly data, about two thirds of the firms in the sample will have a zero yield be-
cause most firms pay dividends on a quarterly basis. Of the firms that pay their divi-
dend (i.e., go ex-dividend) in month t, about 30% to 40% also announce the dividend
in the same month. When the announcement date and the ex-dividend date occur in
the same month, the monthly return will contain both the information effect and the
tax effect (if any). To avoid confusing these effects, Litzenberger and Ramaswamy
computed dividend yields in the following way:
• If a firm declared its dividend prior to month t and went ex-dividend in month
t, then the dividend yield, D Yi„ was computed using the actual dividend paid in
t divided by the share price at the end of month t — 1.
• If a firm both declared and went ex-dividend in month t, then the yield, DYit ,
was computed using the last regular dividend, going back as far as one year.
Table 16.6 shows the results of regressions run by Miller and Scholes [1982] using
Eq. (16.13). Regressions using the actual dividend yield in month t show that the
dividend variable has a coefficient of .317 and it is highly significant, but recall that
the actual yield confuses announcement effects with dividend tax effects. When the
Litzenberger-Ramaswamy measure of dividend yield (called the level-revised yield)
was duplicated by Miller and Scholes, the dividend coefficient dropped from .317 to
.179 and also dropped in significance.
THE RELATIONSHIP BETWEEN DIVIDENDS AND VALUE 593
The third regression in Table 16.6 corrects for a bias not contemplated in the
two prior regressions. Namely, that some firms are expected to pay a dividend in
month t, but for some reason, the board of directors suspends the dividend. Miller
and Scholes call this the case of the "dog that didn't bark." Suppose that a $10 stock
has a 50/50 chance of either announcing a $2 dividend (in which case the stock price
doubles to $20) or suspending the dividend (thereby causing the stock price to fall
to $5). The ex ante rate of return (and the average ex post return) is 35%, and the
ex ante dividend yield is 10%.6 However, if the level-revised measure of dividend
yield is used, then if the firm actually pays the $2 dividend the yield is 20% and the
return is 120%. But if the dividend is passed, the yield is 0% and a -50% return is
recorded. Thus the regressions with the level-revised measure tend to show what
appears to be a positive association between returns and dividend yields. However,
the correlation is spurious. A simple way to correct for the problem is to use the div-
idend yield of 12 months ago. Shown in the third regression in Table 16.6, the results
indicate a small, statistically insignificant relationship between dividend yields and
returns.
Another approach, shown in the fourth regression in Table 16.6, is to drop from
the sample all firms except those that both paid dividends in month t and announced
them in advance. Again the dividend coefficient is insignificant.
Litzenberger and Ramaswamy [1982] have responded to the Miller-Scholes cri-
ticism by rerunning their regressions. Table 16.7 shows their results. The level-revised
dividend yield gave the highest coefficient (aé ), and it is slightly higher than the
Miller-Scholes estimate. Instead of using a dividend 12 months ago, Litzenberger
6
The ex ante return is computed as
2
(20 ) + .5(5
.5 - .35
- 10 + 10
10 1010)
and the ex ante dividend yield is
5( 2 + .5 ( 0
10
1_13) 1_13)
594 DIVIDEND POLICY: EMPIRICAL EVIDENCE AND APPLICATIONS
and Ramaswamy built a more sophisticated model to predict dividends. Their "pre-
dicted dividend yield" model avoids the Miller-Scholes criticism and continues to
give a statistically significant estimate of the dividend effect. So, too, does a restricted
subsample designed to avoid the Miller-Scholes criticism. Thus the empirical evidence,
at this point in time, points toward the conclusion that shareholders express their
displeasure with corporate dividend payments by requiring a higher risk-adjusted re-
turn (i.e., by paying a lower price) for those stocks that have higher dividend yields.
The Friend and Puckett, Black and Scholes, and Miller and Scholes studies tend
to support the conclusion that the value of the firm is independent of dividend yield.
The Litzenberger and Ramaswamy study supports the conclusion that dividends are
undesirable. The next study to be discussed concludes that dividends are desirable
to shareholders; i.e., they will require a lower rate of return on shares that pay a high
dividend yield.
Long [1978] provides a detailed analysis of two classes of shares issued by
Citizens Utilities Company in 1956. They are virtually identical in all respects except
for dividend payout. Series A shares pay only stock dividends that are not taxable
as ordinary income (due to a special Internal Revenue Service ruling granted to
Citizens Utilities). Series B shares pay only cash dividends. Series A shares are freely
convertible into Series B shares (on a one-for-one basis) at any time. However, the
opposite is not true. Series B shares may not be converted into Series A shares.
Historically the directors of Citizens Utilities have semiannually declared stock div-
idends that (with a high degree of certainty) are 8% to 10% larger than the cor-
responding Series B cash dividends (paid on a quarterly basis).
Figure 16.2 shows the natural logarithm of the ratio of the prices of Series A to
Series B shares on a monthly basis between 1956 and 1977. Note that the price of
Series A shares, P A , never falls significantly below the price of Series B shares, P B. If
it did, then Series A shareholders could immediately profit by converting to Series
B. Figure 16.2 also shows, q, the ratio of Series A stock dividends to Series B cash
dividends (illustrated by the circles). In a world without taxes the price per share of
Series A "should" always equal q times the price per share of Series B stock, i.e.,
THE RELATIONSHIP BETWEEN DIVIDENDS AND VALUE 595
0.25
0
000
0.20
-0 0.15
.4 0
0
1 o
tf) 0
0
0 /
0
0 0 0 0
0.10 b 0 0 o
0 I
000 0 0
o °
co JI
< 0, 0
0.05 0
0
-c
ù
h.0
0
0.00
Jti
—0.05 I I I I I I I I I I I I I I I I I I I I I I
56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77
Years
The natural log of PA /PB (the connected monthly observations) and the natural log of
the semi-annual ratio of Series A to Series B dividends (the unconnected 0's) for the period
1956-1976. /3.8 is the price per share of Series B stock with dividends reinvested during
each half-year prior to payment of the semi-annual Series A dividend. The unconnected
points representing the log of the dividend ratio are placed in the figure at the end of the
half-years to which they refer.
Figure 16.2
The natural log of P AI P B plotted monthly for the period April 1956 to December 1976.
(From J. Long, Jr., "The Market Valuation of Cash Dividends: A Case to Consider,"
reprinted with permission from Journal of Financial Economics, June—September 1978, 254).
© 1978 North-Holland.
P At = qPBt (t 0, 1, . • • , N). Figure 16.2 indicates that 80% of the dividend ratios, q,
between 1962 and 1976 fall in the range 1.07 to 1.137. The ratio of stock prices,
however, shows much more variability and is usually below the dividend ratio. How
can this be explained?
In a world with taxes, cross-sectional heterogeneity in investor tax rates will
cause the price ratio to differ from the dividend ratio. If very few investors face strictly
higher taxes on stock dividends than on cash dividends, then the ratio of prices,
P A/PB, should be greater than the dividend ratio, q. The equilibrium price ratio P A/P B
will be less than q if there are sufficient numbers of investors who have a strict tax-
induced preference for cash dividends over capital gains, and this is what is observed.
Therefore Long [1978] concludes that "claims to cash dividends have, if anything,
commanded a slight premium in the market to claims to equal amounts (before taxes)
of capital gains." The Citizens Utilities case is anomalous because it is the only major
study that indicates an investor preference for cash dividends. Furthermore, the
strength of Long's conclusion has been weakened by Poterba [1986] who finds (1)
that the relative price of the stock dividend shares has been higher since 1976 (the
end of Long's sample); (2) that the ratio of stock dividend to cash dividend share
prices averaged 1.134 during the 1976-1984 period; and (3) that the cash dividend
596 DIVIDEND POLICY: EMPIRICAL EVIDENCE AND APPLICATIONS
shares' ex-day price decline is less than their dividend payment, whereas stock dividend
shares fall by nearly their full dividend. These results are more consistent with investor
preference for capital gains rather than dividends.
According to the Miller-Modigliani irrelevancy proposition, it is also important
to know whether or not dividend policy can affect the investment decisions made by
managers of the firm. This is a particularly difficult empirical question because the
Miller-Modigliani theorem requires only that dividend payout not affect investment
decisions. However, the opposite causality is not ruled out by Miller and Modigliani.
That is, investment decisions can affect dividends. For example, the firm may simply
choose to treat dividends as a residual payout after all profitable investment projects
have been undertaken. This would not be inconsistent with the Miller-Modigliani
proposition that the value of the firm is unaffected by dividend policy.
Fama [1974] uses a sophisticated two-stage least squares econometric technique
in order to determine the direction of causality, if any, between dividend and in-
vestment decisions. Because a description of two-stage least squares is beyond the
scope of this book, we refer the interested reader to Fama's article for a detailed
exposition. His conclusion, however, is consistent with the Miller-Modigliani as-
sumption that the period-by-period investment decisions of the firm are separable
from its dividend decisions. There appears to be no causality in either direction. The
data could not reject the hypothesis that investment and dividend decisions are
completely independent from each other. Fama's conclusion that investment and
dividend decisions are independent is supported by Smirlock and Marshall [1983]
who employ causality tests using both firm-specific and aggregate data for 194 firms
between 1958 and 1977.
Although the foregoing studies appear to support the Miller-Modigliani irrele-
vancy proposition from the point of view of an individual firm, they do not neces-
sarily rule out the possibility that there may exist an aggregate equilibrium supply
of dividends that will increase if the difference between the ordinary income rate and
the capital gains rate declines. This type of situation is implicit in Miller's [1977]
paper "Debt and Taxes," which was discussed at length in Chapter 13.
Some empirical evidence that is consistent with the thesis that the aggregate
supply of dividends is sensitive to the differential between the ordinary income and
capital gains rates is contained in a study by Khoury and Smith [1977]. They observed
that Canadian corporations significantly increased their dividend payout after a
capital gains tax was introduced for the first time in 1972. A more recent paper by
Morgan [1980] finds that the change in the Canadian tax code in 1972 affected the
cross-sectional relationship between dividends and capital gains. Prior to 1972 they
were imperfect substitutes, but afterward they became more or less perfect substitutes.
Corporations can repurchase their own shares in either of two ways: on the open
market or via tender offer. Open market repurchases usually (but not always) involve
gradual programs to buy back shares over a period of time. In a tender offer the
CORPORATE EQUITY REPURCHASES VIA TENDER OFFER 597
The reader is also referred to studies by Woods and Brigham [1966], Bierman and West [1966], Young
[1967], Elton and Gruber [1968], Stewart [1976], Coates and Fredman [1976], and Lane [1976].
According to Vermaelen [1981] only 3 out of 105 tender offers that he studied actually were subject to
ordinary income taxes.
598 DIVIDEND POLICY: EMPIRICAL EVIDENCE AND APPLICATIONS
Share
Price
P
T
P
Average A 5 I Average expiration date
tender I } price decline 3%
premium Average
PE
23% post—announcement
price change
15%
Time
Announcement Expiration
date date
Figure 16.3
Schematic representation of average price changes surrounding tender
offers for repurchase.
above the pre-tender stock price, some shareholders may voluntarily decide not
to tender their shares.
A great deal can be learned about these hypotheses if we focus on the price effects
on shares, bonds, and preferred stock. Figure 16.3 shows the average pattern of share
price changes around the tender announcement date and the expiration date. More
or less the same results were reported by Masulis [1980], Dann [1981], and Vermaelen
[1981]. The average tender price, PT, is roughly 23% above the preannouncement
price, Po . If all shares tendered were actually purchased by the firm, then the tender
price, P T , would equal the average postannouncement price, But because of pro
rata repurchases given oversubscribed tenders, we observe that on average PA < P T .
The postannouncement price, PA , averages 15% above the preannouncement price,
P o . Finally, note that the average postexpiration price, PE, is only 3% below the
average postannouncement price, P A , and is above the preannouncement price, P o .
This suggests that the tender offer may have increased the market value of the firm's
equity.
Unfortunately, the difference between the preannouncement price and the post-
expiration price does not measure the information effect of the tender offer. We have
to look deeper. Begin by noting that the market value of the firm's equity after ex-
piration, PE N E , is equal to the preannouncement value, P o N 0 , minus the cash paid
out in the tender, PAN ° — N E), plus the tender offer effect, W:
PE N, = Po N 0 — PAN, — N E) + AW, (16.14)
CORPORATE EQUITY REPURCHASES VIA TENDER OFFER 599
where
PE = the postexpiration share price,
N E = the number of shares outstanding after repurchase,
P o = the preannouncement share price,
N o = the preannouncement number of shares outstanding,
P T = the tender price,
A W = the shareholder wealth effect attributable to the tender offer.
Note that the change in value attributable to the tender, A W, may be caused by (1)
personal tax savings, (2) a leverage effect, (3) expropriation of bondholder wealth, or
(4) the reassessment of the firm's earnings prospects.
If we define the fraction of shares repurchased, F P, as
NE
FP =1
(16.15)
No
and divide Eq. (16.14) by N o , we have
W
PE( 1 — F P) = Po — P T F P + (16.16)
No
Solving for A W/N o and dividing by P o gives
AW —Po
= (1 F p) PE PO
+ Fp PT • (16.17)
NoPo P
O P
O
Thus the rate of return created by the tender offer has two components. First is the
rate of return received by nontendering shareholders weighted by the percentage of
untendered shares, 1 — F P, and second is the rate of return received by tendering
shareholders weighted by the percentage of shares purchased, F.
Vermaelen [1981] found that the average wealth effect, AW/N o Po , was 15.7%
and that only 10.7% of the tender offers experienced a wealth decline. On average,
both nontendered shares and tendered shares experienced a wealth increase, although
not by equal amounts.
What causes the average 15.7% wealth gain from tender offers? Personal tax
savings are a possibility but seem too small to explain the large wealth gain. For
example, if 20% of the value of the firm is repurchased and if the marginal investor's
tax rate is 40%, then the tax savings would imply a 4% rate of return. This is too
small to explain the wealth gain.
The leverage hypothesis suggests that if the repurchase is financed with debt,
and if there is a tax gain from leverage, then the shareholders will benefit. Both
Masulis [1980] and Vermaelen [1981] find evidence consistent with a leverage effect.
Masulis divided his sample into offers with more than 50% debt financing where the
average announcement return was 21.9%, and offers with less than 50% debt where
the average announcement return was only 17.1%. Vermaelen finds similar results
and concludes that while it is not possible to reject the leverage hypothesis outright,
600 DIVIDEND POLICY: EMPIRICAL EVIDENCE AND APPLICATIONS
it is possible to conclude that it is not the predominant explanation for the observed
abnormal returns following the tender offer. Also, if leverage is a signal, then it is
not possible to separate the leverage signaling effect from the leverage tax effect.'
The best explanation for the shareholder wealth gain from the tender offer is
that the offer represents a favorable signal. Vermaelen [1981] finds that the per-share
earnings of tendering firms are above what would have been predicted by a time
series model using preannouncement data. Thus the tender offer may be interpreted
as an announcement of favorable earnings prospects. Also, the size of the tender pre-
mium, the fraction of shares repurchased, and the fraction of insider holdings are all
positively related to the wealth gain, 4W, and explain roughly 60% of its variance.
These results are also consistent with interpreting the tender offer as a signal.
Evidence on the bondholder wealth expropriation hypothesis is provided by
looking at bond price changes around the announcement date. Dann [1981] found
122 publicly traded debt and preferred stock issues for 51 tender offers. There were
41 issues of straight debt, 34 issues of convertible debt, 9 issues of straight preferred
stock, and 38 issues of convertible preferred stock. An analysis of abnormal returns
around the announcement date revealed significant positive rates of return for the
convertible securities and rates that were insignificantly different from zero for straight
debt and preferred. Furthermore, the correlation between common stock returns and
straight debt (and preferred) returns was positive. Thus the evidence seems to con-
tradict bondholder expropriation as the dominant effect and seems to support the
signaling hypothesis.
Repurchases via tender offer represent an interesting and significant corporate
event. The empirical evidence, although not rejecting leverage effects or dividend tax
avoidance effects, seems to most strongly support the hypothesis that the tender offer
for repurchase is interpreted by the marketplace as favorable information regarding
future prospects of the firm.
The theory of finance clearly demonstrates that in the absence of taxes dividend
policy has no effect on the value of the corporation. However, if taxes are introduced,
then the firm can maximize the value of shareholders' wealth by paying no dividends
so long as the personal tax rate on dividend income is higher than that on capital
gains. However, this is not true for all types of shareholders; hence there exists the
possibility of clientele effects with shareholders choosing the firm with the payout
they prefer. The evidence supporting dividend clientele effects is much stronger than
the evidence for capital structure clientele effects. Another issue is the relationship
between dividend yield and equity values. The preponderance of empirical evidence
seems to favor the conclusion that dividend yield has no strong effect on the required
9
See the discussion of debt for common exchange offers in Chapter 14 for evidence consistent with the
hypothesis that higher leverage (repurchases of equity with debt) is a favorable signal about the future
prospects of the firm.
VALUATION AND CORPORATE POLICY 601
rate of return on equity; however, if there is any effect it is in favor of capital gains
over dividends. The Litzenberger and Ramaswamy [1979, 1982] studies found that
higher dividend yields required higher rates of return to compensate investors for
the disadvantage of dividend payout. The only study that found that dividends were
desirable was the Citizens Utilities case [Long, 1978], which has been questioned by
Poterba [1986]. On the other hand, changes in dividends paid out are interpreted
as new information about the future cash flows of the firm. There is strong evidence
to support a dividend signaling effect. There appears to be no causal relationship
between investment and dividend policy. Finally, empirical evidence on share repur-
chase via tender offer indicates that the announcement effect is predominantly viewed
as favorable news regarding the firm's future cash flows. Little or no expropriation
of bondholders' wealth was observed.
The valuation models developed in Chapter 15 have many useful applications. For
example, fundamental analysis techniques use them as a tool for estimating the im-
pact of various types of new information on the value of the firm. Although the valua-
tion models cannot aid in attempts to "beat" the market, they do provide a useful
framework for concentrating on relevant information. When a corporate president
asks, "How much should we pay to acquire Company X?" the valuation model is a
useful tool. When a company is going public for the first time or when it becomes
necessary to value a privately held company, again the valuation model provides a
relevant framework for analysis.
In the next section we continue the Bethlehem Steel example that was started in
Chapter 14. The object is to see how to apply the Miller-Modigliani finite growth
model. In so doing, it becomes possible to analyze six factors that affect corporate
value.
10
Remember that we derive this version of the model by assuming, among other things, that corporate
taxes are the only form of government levy. Personal taxes do not exist.
602 DIVIDEND POLICY: EMPIRICAL EVIDENCE AND APPLICATIONS
where
V L = the market value of the levered firm, i.e., the market value of equity, S,
plus the market value of debt, B,
E(NOI 1) = the expected net cash flows from operations during the current year,
net of capitalized maintenance,
p = the cost of equity for the firm if it had no debt, i.e., the cost of equity
for an all-equity firm,
= the marginal effective corporate tax rate,
* During 1976 this includes current taxes (federal, foreign, and state) of -$9,000,000 and deferred taxes
(federal and state) of $35,000,000. Note that deferred taxes are a noncash charge.
11
Recall that we have assumed that F = dep.
604 DIVIDEND POLICY: EMPIRICAL EVIDENCE AND APPLICATIONS
to obtain the correct definition of cash flows, one further adjustment must be made.
Recall that the model assumes that depreciation is adequate to replace worn-out
equipment. In other words, cash outflows used to replace equipment each year are
exactly equivalent to depreciation. The two cancel each other. Therefore, because
depreciation is used for replacement investment, it is not part of the free cash flow
available to investors. Consequently, it is necessary to subtract depreciation from
both sides of (16.20). The result is the correct definition of after-tax free cash flows
from operations that can be maintained into perpetuity, NOLO — tc):
NOI 0 (1 — Tc) =- (Rev — VC)(1 — te) — dep(1 — T c.) = NI + k d D(1 — r e). (16.21)
For Bethlehem Steel the estimates are given in Table 16.10. Note that for 1976 esti-
mates, we also have to add deferred taxes of $35 million, a noncash charge (discovered
in the footnotes to the income statement). This makes our estimate of NOI 0 (1 — t c)
add up to be $243,404 instead of $208,404. Note that r, = .48.
B. THE INVESTMENT RATE. The reason for estimating the investment rate, K,
is that it is useful in estimating two more fundamental parameters: (1) the expected
dollar amount of new investment, I = K(NOI,)(1 — r e) and (2) the expected rate of
growth in the firm, g = Kr. There is no reason that the investment rate must be less
than 100% of cash flows. A higher investment rate simply means that the firm will
be issuing new equity in order to undertake new investment in excess of current cash
flows. This is certainly feasible for a firm with r > WACC.
From Table 16.10 we can estimate the investment rate by subtracting the dividend
payout [dividends ÷ NOI 0(1 — re)] from 100%. Bethlehem did not issue new equity
during the period 1973-1976. The average investment rate is 63.8%.
We assume that Bethlehem Steel will take on new investments with approximately
the same risk as its existing projects. Therefore we can use the weighted average cost
of capital, estimated in Chapter 14. It was WACC = 10.61%. In addition, we need
to know the cost of equity capital if Bethlehem Steel were an all-equity firm. This
can be computed by using the Modigliani-Miller definition of cost of capital, Eq.
(13.12):
B
WACC =p B + s ),
WACC (13.12)
= B \'
(1
B+S
where
p = the cost of equity for the all-equity firm,
T = the corporate tax rate = .48,
c
B 1(B + S) = the percentage of debt in the capital structure of the firm = 31.5%
(using market value weights),
WACC = the weighted average cost of capital = 10.61%.
Solving for p, we have
P=
.1061
= 12.50%.
1 — .48(.315)
G. THE VALUE OF BETHLEHEM STEEL. All that remains is to use the estimates
of the six valuation parameters in the valuation equation. For convenience it is re-
written below:
[ r WACC
E(NOI1)(1 — "cc)
V=
= + -c/3 + K[E(NOI,)(1— T e)] T (16.18)
p WACC(1 + WACC)
VALUATION AND CORPORATE POLICY 607
Note that the expected 1977 after-tax cash flows from operations, E(NOI1)(1 — Te),
equal the estimated 1976 figure multiplied by the estimated growth rate (g = Kr):
E(NOI1)(1 — r c) = NO4(1 — + Kr)
= 243,404[1 + .638(.0602)]
= 243,404(1.0384076) = $252,753.
Substituting the values of the remaining parameters into (16.18), we have
252 753 0602 — .1061]
V= ' + .48(815,945) + .638(252,753)3[
.125 .1061(1.1061)
= 2,022,024 + 391,653 — 189,209
= $2,224,468 (thousands of dollars).
The actual market value of Bethlehem Steel at the end of December 1976 was
$2.590 billion.12 Therefore the estimate provided by the valuation model is around
$0.366 billion too low, an error of around 14.1%. The discrepancy may be due to
any number of reasons. Most likely our estimates of the anticipated operating cash
flows and the rate of return on future investment (which were based on historical
accounting data) were too low.
The valuation model does serve to point out the contribution of the three major
components. Most of the market value of Bethlehem Steel comes from the stream
of cash flows provided by investments that are already in place. Some extra value
is provided by the tax shield from debt. And almost no value or more accurately,
negative value—results from the profitability of anticipated growth opportunities.
12 The market value of equity was $40i per share, and 43.665 million shares were outstanding. Thus
the total value of equity was $1.774 billion. Adding this to the $0.816 billion of long-term debt, we obtain
a market value of $2.590 billion.
608 DIVIDEND POLICY: EMPIRICAL EVIDENCE AND APPLICATIONS
PROBLEM SET
16.1 Under what conditions might dividend policy affect the value of the firm?
16.2 According to federal tax law, corporations need not pay taxes on 80% of dividends
received from shares held in other corporations. In other words, only 20% of the dividends
received by a corporate holder are taxable. Given this fact, how much must the price of a stock
fall on the ex-dividend date in order to prevent a corporate holder from making arbitrage
profits? Assume that the capital gains rate equals the corporate tax rate, i, = .5.
16.3 Empirical evidence supports the existence of a clientele effect. This implies that every
time a company revises its dividend policy to pay out a greater (or smaller) percentage of
earnings, the characteristics of its shareholders also change. For example, a firm with a higher
payout ratio may expect to have more shareholders in lower tax brackets. Suppose that lower-
income people are also more risk averse. Would this have an effect on the value of the firm?
16.4 Miller and Scholes [1978] suggest that it is possible to shelter income from taxes in
such a way that capital gains rates are paid on dividend income. Furthermore, since capital
gains need never be realized, the effective tax rate will become zero. Why would this scheme
not be used to shelter all income, instead of just dividend income? The implication would be
that no one has to pay taxes ever!
16.5 The Pettit study suggests an increase in the price per share of common stock commen-
surate with an increase in dividends. Can this be taken as evidence that the value of the firm
is in fact affected by dividend policy?
REFERENCES 609
16.6 Assume that the XYZ firm has the following parameters in a world with no taxes:
a) What will be the market value of equity cum dividend (i.e., before any dividend is paid)?
b) If the shareholders decide to pay themselves a $500 dividend out of cash, what will be
the ex-dividend wealth of shareholders? [Note- The dividend payment will have two ef-
fects. First, it will decrease the market value of the firm to $1500. Second, since cash has
little or no risk, the instantaneous standard deviation of the firm's assets will increase to .25.]
REFERENCES
Aharony, J., and I. Swary, "Quarterly Dividend and Earnings Announcements and Stock-
holders' Returns: An Empirical Analysis," Jou]nal of Finance, March 1980, 1-12.
Asquith, P., and W. Krasker, "Changes in Dividend Policy, and Stock Trading Volume,"
Working paper, Harvard Business School, Cambridge, Mass., October 1985.
Asquith, P., and D. Mullins, Jr., "The Impact of Initiating Dividend Payments on Shareholders'
Wealth," Jou]nal of Business, January 1983, 77-96.
Austin, D., "Treasury Stock Reacquisition by American Corporations 1961-1967," Financial
Executive, May 1969, 40-61.
Bar-Yosef, S., and R. Kolodny, "Dividend Policy and Market Theory," Review of Economics
and Statistics, May 1976, 181-190.
Bierman, H., and R. West, "The Acquisition of Common Stock by the Corporate Issuer,"
Jou]nal of Finance, December 1966, 687-696.
Black, F.; M. Jensen; and M. Scholes, "The Capital Asset Pricing Model: Some Empirical
Tests," in Jensen, ed., Studies in the Theo]y of Capital Ma]kets. Praeger, New York, 1972,
79-124.
Black, F., and M. Scholes, "The Effects of Dividend Yield and Dividend Policy on Common
Stock Prices and Returns," Jou]nal of Financial Economics, May 1974, 1-22.
Blume, M., "Stock Returns and Dividend Yields: Some More Evidence," Review of Economics
and Statistics, November 1980, 567-577.
Booth, L., and D. Johnston, "The Ex-Dividend Day Behavior of Canadian Stock Prices: Tax
Changes and Clientele Effects," Jou]nal of Finance, June 1984, 457-476.
Brennan, M., "Taxes, Market Valuation and Corporate Financial Policy," National Tax
Jou]nal, December 1970, 417-427.
Brickley, J., "Shareholder Wealth, Information Signaling and the Specially Designated Div-
idend," Jou]nal of Financial Economics, August 1983, 187-210.
610 DIVIDEND POLICY: EMPIRICAL EVIDENCE AND APPLICATIONS
Brittain, J. A., Co]po]ate Dividend Policy. Brookings Institution, Washington, D.C., =966.
Charest, G., "Dividend Information, Stock Returns and Market Efficiency," Jou]nal of Financial
Economics, June—Septemîer =978, <97-330.
Coates, C., and A. Fredman, "Price Behavior Associated with Tender Offers to Repurchase
Common Stock," Financial Executive, April =976, 40-44.
Dann, L., "Common Stock Repurchases: An Analysis of Returns to Bondholders and Stock-
holders," Jou]nal of Financial Economics, June =98=, ==3-=38.
Darling, P., "The Influence of Expectations and Liquidity on Dividend Policy," Jou]nal of
Political Economy, June =9P7, <09-<<4.
Dielman, T.; T. Nantell; and R. Wright, "Price Effects of Stock Repurchasing: A Random
Coefficient Regression Approach," Jou]nal of Financial and Quantitative Analysis, March
=980, =7P-=89.
Eades, K.; P. Hess; and E. H. Kim, "Market Rationality and Dividend Announcements,"
Jou]nal of Financial Economics, Decemîer =98P, P8=-604.
, "On Interpreting Security Returns during the Ex-Dividend Period," Jou]nal of Financial
Economics, March =984, 3-34.
Ellis, C., "Repurchase Stock to Revitalize Equity," Ha]va]d Business Review, July—August =96P,
==9-=<8.
Elton, E. J., and M. J. Gruîer, "The Effect of Share Repurchase on the Value of the Firm,"
Jou]nal of Finance, March =968, =3P-=49.
, "Marginal Stockholders' Tax Rates and the Clientele Effect," Review of Economics
and Statistics, Feîruary =970, 68-74.
Fama, E., "The Empirical Relationship îetween the Dividend and Investment Decisions of
Firms," Ame]ican Economic Review, June =974, 304-3=8.
, and H. Baîiak, "Dividend Policy: An Empirical Analysis," Jou]nal of the Ame]ican
Statistical Association, Decemîer =968, ==3<-==6=.
Fama, E.; L. Fisher; M. Jensen; and R. Roll, "The Adjustment of Stock Prices to New Informa-
tion," Inte]national Economic Review, Feîruary =969, =-<=.
Farrar, D., and L. Selwyn, "Taxes, Corporate Financial Policy and Return to Investors,"
National Tax Jou]nal, Decemîer =967, 444-4P4.
Friend, I., and M. Puckett, "Dividends and Stock Prices," Ame]ican Economic Review, Sep-
temîer =964, 6P6-68<.
Gonedes, N., "Corporate Signaling, External Accounting, and Capital Market Equiliîrium:
Evidence .on Dividends, Income, and Extraordinary Items," Jou]nal of Accounting Resea]ch,
Spring =978, <6-79.
Gordon, M., "Dividends, Earnings, and Stock Prices," Review of Economics and Statistics, May
=9P9, 99-=0P.
, "The Savings, Investment and Valuation of a Corporation," Review of Economics and
Statistics, Feîruary =96<, 37-P=.
Gordon, R., and D. Bradford, "Taxation and the Stock Market Valuation of Capital Gains
and Dividends," Jou]nal of Public Economics, Octoîer =980, =03-=36.
Griffin, P., "Competitive Information in the Stock Market: An Empirical Study of Earnings,
Dividends and Analysts Forecasts," Jou]nal of Finance, May =976, 63=-6P0.
Grinîlatt, M., R. Masulis, and S. Titman, "The Valuation Effects of Stock Splits and Stock
Dividends," Jou]nal of Financial Economics, Decemîer =984, 46=-490.
REFERENCES 6==
Guthart, U, "Why Companies Are Buying Their Own Stock," Financial Analysts Jou]nal,
March—April =967, =0P-==0.
Handjiinicolaou, G., and A. Kalay, "Wealth Redistriîutions or Changes in Firm Value: An
Analysis of Returns to Bondholders and Stockholders around Dividend Announcements,"
Jou]nal of Financial Economics, March =984, 3P-64.
Harris J., Jr.; R. Roenfeldt; and P. Cooley, "Evidence of Financial Leverage Clienteles," Jou]nal
of Finance, Septemîer =983, ==<P-==3<.
Hess, P., "The Ex-Dividend Behavior of Stock Returns: Further Evidence on Tax Effects,"
Jou]nal of Finance, May =98<, 44P-4P6.
Higgins, R., "The Corporate Dividend-Saving Decision," Jou]nal of Financial and Quantitative
Analysis, March =97<, =P<7-=P4=.
Hite, G., and J. Owers, "Security Price Reactions around Corporate Spin-off Announcements,"
Jou]nal of Financial Economics, Decemîer =983, 409-436.
Kalay, A., "Essays in Dividend Policy," Ph.D. dissertation, University of Rochester, =977.
, "The Ex-Dividend Behavior of Stock Prices: A Re-examination of the Clientele Effect,"
Jou]nal of Finance, Septemîer =98<, =0P9-=070.
, and U. Lowenstein, "Predictaîle Events and Excess Returns: The Case of Dividend
Announcements," Jou]nal of Financial Economics, Septemîer =98P, 4<3-4P0.
, "The Informational Content of the Timing of Dividend Announcements," Jou]nal of
Financial Economics, July =986, 373-388.
Kane, A.; Y. K. Lee; and A. Marcus, "Signaling, Information Content, and the Reluctance to
Cut Dividends," Jou]nal of Financial and Quantitative Analysis, Novemîer =980, 8PP-870.
, "Earnings and Dividend Announcements: Is There a Corroîoration Effect?" Jou]nal
of Finance, Septemîer =984, =09=-=099.
Keim, D., "Dividend Yields and Stock Returns: Implications of Aînormal January Returns,"
Jou]nal of Financial Economics, Septemîer =98P, 473-490.
Khoury, N., and K. Smith, "Dividend Policy and the Capital Gains Tax in Canada," Jou]nal
of Business Administ]ation, Spring =977.
Kim, E. H.; W. Lewellen; and J. McConnell, "Financial Leverage Clienteles: Theory and Evi-
dence," Jou]nal of Financial Economics, March =979, 83-==0.
Kwan, C., "Efficient Market Tests of the Informational Content of Dividend Announcements:
Critique and Extension," Jou]nal of Financial and Quantitative Analysis, June =98=,
=93-<06.
Lakonishok, J., and T. Vermaelen, "Tax Reform and Ex-Dividend Day Behavior," Jou]nal of
Finance, Septemîer =983, ==P7-==79.
, "Tax-induced Trading around Ex-Dividend Days," Jou]nal of Financial Economics,
July =986, <87-3<0.
Lane, W., "Repurchase of Common Stock and Managerial Discretion," Ph.D. dissertation,
University of North Carolina, =976.
Lauî, P. M., "On the Informational Content of Dividends," Jou]nal of Business, January =976,
73-80.
Lewellen, W.; K. Stanley; R. Lease; and G. Schlarîaum, "Some Direct Evidence on the Dividend
Clientele Phenomenon," Jou]nal of Finance, Decemîer =978, =38P-=399.
Lintner, J., "Distriîution of Incomes of Corporations among Dividends, Retained Earnings
and Taxes," Ame]ican Economic Review, May =9P6, 97-==3.
6=< DIVIDEND POLICY: EMPIRICAL EVIDENCE AND APPLICATIONS
Richardson, G.; S. Sefcik; and R. Thompson, "A Test of Dividend Irrelevance Using Volume
Reactions to a Change in Dividend Policy," Journal of Financial Economics, December
1986, 313-334.
Rosenberg, B., and V. Marathe, "Tests of the Capital Asset Pricing Hypothesis," Research in
Finance, January 1979, 115-223.
Ross, S. A., "The Determination of Financial Structure: The Incentive-Signalling Approach,"
Bell Journal of Economics, Spring 1977, 23 -40.
Smirlock, M., and W. Marshall, "An Examination of the Empirical Relationship between the
Dividend and Investment Decisions: A Note," Journal of Finance, December 1983,
1659-1667.
Stewart, S., "Should a Corporation Repurchase Its Own Stock?" Journal of Finance, June 1976,
911-921.
Van Home, J., and J. G. McDonald, "Dividend Policy and New Equity Financing," Journal
of Finance, May 1971, 507-520.
Vermaelen, T., "Common Stock Repurchases and Market Signalling: An Empirical Study,"
Journal of Financial Economics, June 1981, 139 -183.
Walter, J. E., Dividend Policy and Enterprise Valuation. Wadsworth, Belmont, Calif., 1967.
Watts, R., "The Information Content of Dividends," Journal of Business, April 1973, 191-211.
, "Comments on 'The Impact of Dividend and Earnings Announcements: -A Recon-
ciliation,'" Journal of Business, January 1976a, 97-106.
, "Comments on the Informational Content of Dividends," Journal of Business, January
1976b, 81-85.
Woods, D., and E. Brigham, "Stockholder Distribution Decisions: Share Repurchases or Divi-
dends?" Journal of Financial and Quantitative Analysis, March 1966, 15 -28.
Woolridge, J., "Dividend Changes and Security Prices," Journal of Finance, December 1983,
1607-1615.
Young, A., "The Performance of Common Stocks Subsequent to Repurchase," Financial Ana-
lysts Journal, September—October 1967, 117-121.
, "Financial, Operating and Security Market Parameters of Repurchasing," Financial
Analysts Journal, July—August 1969, 123-128.
17
It is estimated that from 15 to 20 percent of all new capital put in use
by business each year is leased.
Peter Vanderwicken, "The Powerful Logic of the Leasing Boom,"
Fortune, November 1973, 136.
A. INTRODUCTION
614
THE LEGAL AND ACCOUNTING TREATMENT OF LEASES 615
1. Types of Leases
Leases take several different forms, the most important of which are sale and
leaseback, service or operating leases, and straight financial leases. These three major
types of leases are described below.
Under a sale and leaseback arrangement, a firm owning land, buildings, or equip-
ment sells the property to a financial institution and simultaneously executes an
agreement to lease the property back for a certain period under specific terms.
Note that the seller, or lessee, immediately receives the purchase price put up by
the buyer, or lessor. At the same time, the seller-lessee retains the use of the property.
This parallel is carried over to the lease payment schedule. Under a mortgage loan
arrangement the financial institution receives a series of equal payments just sufficient
to amortize the loan and to provide the lender with a specified rate of return on invest-
ment. Under a sale and leaseback arrangement the lease payments are set up in the
same manner. The payments are sufficient to return the full purchase price to the
financial institution in addition to providing it with some return on its investment.
Operating (or service) leases include both financing and maintenance services.
IBM is one of the pioneers of the service lease contract. Computers and office copying
machines, together with automobiles and trucks, are the primary types of equipment
covered by operating leases. The leases ordinarily call for the lessor to maintain and
service the leased equipment, and the costs of this maintenance are either built into
the lease payments or contracted for separately.
Another important characteristic of the service lease is that it is frequently not
fully amortized. In other words the payments required under the lease contract are
not sufficient to recover the full cost of the equipment. Obviously, however, the lease
contract is written for considerably less than the expected life of the leased equipment,
and the lessor expects to recover the cost either in subsequent renewal payments or
on disposal of the equipment.
A final feature of the service lease is that it frequently contains a cancellation
clause giving the lessee the right to cancel the lease and return the equipment before
the expiration of the basic agreement. This is an important consideration for the
lessee, who can return the equipment if technological developments render it obsolete
or if it simply is no longer needed.
A strict financial lease is one that does not provide for maintenance services, is
not cancellable, and is fully amortized (i.e., the lessor contracts for rental payments
equal to the full price of the leased equipment). The typical arrangement involves the
following stqps:
1. The user firm selects the specific equipment it requires and negotiates the price
and delivery terms with the manufacturer or distributor.
2. Next, the user firm arranges with a bank or leasing company for the latter to buy
the equipment from the manufacturer or distributor, simultaneously executing
616 THE ECONOMICS OF LEASING
an agreement to lease the equipment from the financial institution. The terms
call for full amortization of the financial institution's cost, plus a rate of return
on investment. The lessee generally has the option to renew the lease at a reduced
rental on expiration of the basic lease but does not have the right to cancel the
basic lease without completely paying off the financial institution.
Financial leases are almost the same as sale and leaseback arrangements, the
main difference being that the leased equipment is new and the lessor buys it from
a manufacturer or a distributor instead of from the user-lessee. A sale and leaseback
can thus be thought of as a special type of financial lease.
2. Tax Treatment
The full amount of the annual lease payments is deductible for income tax
purposes provided the Internal Revenue Service (IRS) agrees that a particular
contract is a genuine lease and not simply an installment loan called a lease. This
makes it important that the lease contract be written in a form acceptable to the
IRS. Following are the major requirements for bona fide lease transactions from the
standpoint of the IRS:
1. The term must be less than 30 years; otherwise the lease is regarded as a form
of sale.
2. The rent must represent a reasonable return to the lessor.
3. The renewal option must be bona fide, and this requirement can best be met by
giving the lessee the first option to meet an equal bona fide outside offer.
4. There must be no repurchase option; if there is, the lessee should merely be given
parity with an equal outside offer.
3. Accounting Treatment
In November 1976 the Financial Accounting Standards Board (FASB) issued its
Statement of Financial Accounting Standards No. 13, Accounting for Leases. Like
other FASB statements, the standards set forth must be followed by business firms
if their financial statements are to receive certification by auditors. FASB Statement
No. 13 has implications both for the utilization of leases and for their accounting
treatment. Those elements of FASB Statement No. 13 that are most relevant for
financial analysis of leases are summarized below.
For some types of leases, this FASB statement requires that the obligation be
capitalized on the asset side of the balance sheet with a related lease obligation on
the liability side. The accounting treatment depends on the type of lease. The classi-
fication is more detailed than the two categories of operating and financial leases
described above.
From the lessee's point of view the two accounting categories are capital leases
and operating leases. A lease is classified in Statement No. 13 as a capital lease if it
meets one or more of four Paragraph 7 criteria:
THE LEGAL AND ACCOUNTING TREATMENT OF LEASES 617
1. The lease transfers ownership of the property to the lessee by the end of the lease
term.
2. The lease gives the lessee the option to purchase the property at a price sufficiently
below the expected fair value of the property that the exercise of the option is
highly probable.
3. The lease term is equal to 75% or more of the estimated economic life of the
property.
4. The present value of the minimum lease payments exceeds 90% of the fair value
of the property at the inception of the lease. The discount factor to be used in
calculating the present value is the implicit rate used by the lessor or the lessee's
incremental borrowing rate, whichever is lower. (Note that the lower discount
factor represents a higher present value factor and therefore a higher calculated
present value for a given pattern of lease payments. It thus increases the likeli-
hood that the 90% test will be met and that the lease will be classified as a
capital lease.)
From the standpoint of the lessee, if a lease is not a capital lease, it is classified as an
operating lease.
From the standpoint of the lessor, four types of leases are defined: (1) sales-type
leases, (2) direct financing leases, (3) leveraged leases, and (4) operating leases repre-
senting all leases other than the first three types. Sales-type leases and direct financing
leases meet one or more of the four Paragraph 7 criteria and both of the Paragraph 8
criteria, which are:
1. Collectability of the minimum lease payments is reasonably predictable.
2. No important uncertainties surround the amount of unreimbursable costs yet to
be incurred by the lessor under the lease.
Sales-type leases give rise to profit (or loss) to the lessor the fair value of the leased
property at the inception of the lease is greater (or less) than its cost-of-carrying
amount. Sales-type leases normally arise when manufacturers or dealers use leasing
in marketing their products. Direct financing leases are leases other than leveraged
leases for which the cost-of-cariying amount is equal to the fair value of the leased
property at the inception of the lease. Leveraged leases are direct financing leases in
which substantial financing is provided by a long-term creditor on a nonrecourse
basis with respect to the general credit of the lessor.
The actual bookkeeping for lessees is set up in the following way. For operating
leases, rentals must be charged to expense over the lease term, with disclosures of
future rental obligations in total as well as by each of the following five years. For
lessees, capital leases are to be capitalized and shown on the balance sheet both as a
fixed asset and as a noncurrent obligation. Capitalization represents the present value
of the minimum lease payments minus that portion of lease payments representing
executory costs such as insurance, maintenance, and taxes to be paid by the lessor
(including any profit return in such charges). The discount factor is [as described in
Paragraph 7(4)] the lower of the implicit rates used by the lessor or the incremental
borrowing rate of the lessee.
618 THE ECONOMICS OF LEASING
Current:
Leased property under capital Obligations under capital leases
leases less accumulated Noncurrent:
amortization Obligations under capital leases
The asset must be amortized in a manner consistent with the lessee's normal
depreciation policy for owned assets. During the lease term, each lease payment is
to be allocated between a reduction of the obligation and the interest expense to
produce a constant rate of interest on the remaining balance of the obligation. Thus
for capital leases the balance sheet includes the terms in Table 17.1.
In addition to the balance sheet capitalization of capital leases, substantial addi-
tional footnote disclosures are required for both capital and operating leases. These
include a description of leasing arrangements, an analysis of leased property under
capital leases by major classes of property, a schedule by years of future minimum
lease payments (with executory and interest costs broken out for capital leases), and
contingent rentals for operating leases.
FASB Statement No. 13 sets forth requirements for capitalizing capital leases
and for standardizing disclosures by lessees for both capital leases and operating
leases. Lease commitments therefore do not represent "off—balance sheet" financing
for capital assets, and standard disclosure requirements make general the footnote
reporting of information on operating leases. Hence the argument that leasing repre-
sents a form of financing that lenders may not take into account in their analysis of
the financial position of firms seeking financing will be even less valid in the future
than it is now.
It is unlikely that sophisticated lenders were ever fooled by off—balance sheet
leasing obligations. However, the capitalization of capital leases and the standard
disclosure requirements for operating leases will make it easier for general users of
financial reports to obtain additional information on firms' leasing obligations. Hence
the requirements of FASB Statement No. 13 are useful. Probably the extent to which
leasing is used will remain substantially unaltered since the particular circumstances
that have provided a basis for its use in the past are not likely to be greatly affected
by the increased disclosure requirements.
arbitrary. Anyone who owns an asset also may decide to lease it. For example, sup-
pose one owns a car or a house. There is always the choice between owning it and
using it for one's own purposes, thereby gaining a direct (nontaxable) stream of con-
sumption of transportation or housing services. Or alternatively, one can lease the
asset to a second party. In return one then receives a (taxable) stream of income that
can be used for the consumption of transportation and housing (among other things).
From the lessee's point of view, the choice to own the asset in the first place was an
investment decision. At the same time, there is a separate decision to make namely,
the financing decision. Should use of the asset be financed with debt and equity, or
should it be leased? How much of the lease financing can be considered to be debt?
How much is equity?
In the analysis that follows it is convenient to divide lease contracts into major
categories: (1) strict financial leases and (2) operating leases. Strict financial leases,
along with sale and leaseback arrangements, will be characterized as perfect substi-
tutes for debt capital; in other words, they have exactly the same risk.' A lessee may
not cancel a strict financial lease, the failure to meet lease payments can force the
lessee into bankruptcy (or reorganization), and the lease is fully amortized (i.e., the
lessor receives lease payments that repay the full cost of the leased asset). Operating
leases are riskier than financial leases. We assume that they may be cancelled at the
option of the lessee, are usually not fully amortized, and require that the salvage value
go to the lessor. Finally, either type of lease may involve a separable contract for
various types of maintenance on the leased asset, e.g., automobile servicing. Because
the maintenance contract is economically separable we shall not discuss it in this
chapter.
Financial leases and operating leases involve very different risks to the lessor and
must therefore be discussed separately. We shall defer a discussion of operating leases
until later in the chapter and focus on the much simpler financial lease for the time
being. Suppose that the lessor is a commercial bank. Recall that any commercial bank
will hold a well-diversified portfolio of corporate debt as its major asset. Obviously
it requires that this portfolio earn (at least) the bank's after-tax weighted average cost
of capital. On the other hand, what we call the rate of return to the bank is also the
cost of debt to the borrowing firm. Therefore if we designate the bank's after-tax
weighted average cost of capital as WACCB and the firm's before-tax cost of debt
as k b , then
WACCB
=
(1 — T c)
This means that debt capital lent to the ith firm and leased to the same firm have the same risk. Of
course, lending to different firms may have different risks.
2
See Chapter 13, Eq. 13.12.
620 THE ECONOMICS OF LEASING
cost of capital is
If the required rate of return on the lease project is 9.375% and the leasing firm
(the bank) uses 90% debt in its optimal capital structure, and has a 40% marginal
tax rate, then
WACCB = .09375[1 — .4(9)] = .06.
Thus from the lessor's point of view the lease project will have a 6% after-tax
weighted average cost of capital. From the lessee's point of view the before-tax cost
of leasing will be
WACCB —
kb — -10%.
1 Te 1 — .4
As mentioned before, each dollar in a pure financial lease is a perfect substitute
for one dollar of debt in the capital structure of the lessee firm. Thus if a lessee is at
its optimal capital structure prior to signing a lease contract, and wishes to maintain
that structure, then it must displace one dollar of debt for each dollar in the lease
contract.
What lease fee should the lessor charge for a pure financial lease? Assume the
cost of the leased asset is $1, the lessor's tax rate is T c, and the annual (straight-line)
depreciation write-off on the leased asset is dep t.' Also, assume that there is no sal-
vage value. If the lessor charges an annual lease payment of L„ then the net present
value of the lease to the lessor is'
N
Li.(1 — T C ) 2,dept
NPV(to lessor) = —I + t (17.1)
t=1 (1 + WACCB)
The numerator of Eq. (17.1) is the standard definition of after-tax cash flows from an
investment, including the depreciation tax shield.' To provide a numerical example,
let I = $10,000, -r, = 40%, WACCB = 6%, and the life of the project, N, be 5 years.
Given these facts and assuming that the lease fee is competitively determined so that
NPV (to lessor) = 0, then the minimum lease fee, L„ is
4(1 — .4) + .4(2000)
0 = —10,000 +
t=1 (1 + .06)t
0 = —10,000 + 4(6)(4.212) + .4(2000)(4.212),
10,000 — .4(2000)(4.212)
= L, = $2624.
.6(4.212)
3
Although we have assumed straight-line depreciation for convenience, there is usually an optimal de-
preciation schedule that maximizes the present value of the depreciation tax shield. Our analysis will not
change if, in practice, both the lessor and lessee use the same optimal depreciation schedule.
For convenience, we have assumed that the stream of lease payments is an annuity with the first payment
at the end of the first year. Most lease contracts require the first payment to be made immediately.
5
If there were an investment tax credit it would also be counted as a cash inflow.
THE THEORY OF LEASING 621
If the lessor charges $2624, then it will earn a rate of return that just compensates
it for taking a debt position in the lessee firm. A higher lease fee would result in a
positive NPV.
Note that this weighted average cost of capital for owning the project is the same
no matter who owns it, so long as their marginal tax rates are the same. We assume
that the optimal capital structure is project specific. For example, a commercial bank
with 90-95% debt in its capital structure should not apply the same leverage to a
wholly owned computer division. Presumably, the computer division has its own
optimal leverage, different from (less than) the commercial bank's.
To continue with the numerical example, assume that the unlevered cost of capi-
tal, p(project), is 14% for the project and that its optimal capital structure is one third
debt to total assets, i.e., B/(B + S) = .33. The required rate of return on the wholly
owned project is
WACC(project) = .14[1 — .4(.33)] = 12.152%.
If the project has a positive net present value when its after-tax operating cash flows
are discounted at the appropriate WACC, then it is a good investment.
To add realism to the investment decision, assume that the investment project
has expected annual sales revenues of St = $20,000, and expected annual cash costs
of C, = $16,711. Then the NPV of the investment project is
(S,— C,)(1 — t) + tedep,
NPV(investment) = —I + E
(1 + WACC)t
6(20,000 — 16,711) + .4(2,000)
= —10,000 +
t= (1.12152)t
= — 10,000 + .6(20,000 — 16,711)(3.5912) + .4(2,000)(3.5912)
= — 10,000 + 7,087 + 2,873
= — 40.
Under these assumptions, the project should be rejected.
622 THE ECONOMICS OF LEASING
1. A cash saving amounting to the dollar amount of the investment outlay, I, which
the firm does not have to incur if it leases.
2. A cash outflow amounting to the present value of the after-tax lease payments,
PV[L,(1 — .
3. The present value of the opportunity cost of the lost depreciation tax shield,
PV[Tcdept] (and lost investment tax credits, which were relevant prior to the
1986 tax code).
4. The present value of the change in the interest tax shield on debt that is displaced
by the lease financing, PV [r, A(k d Dt)], where D, is the remaining book value of
debt outstanding in period t.
We have assumed for strict financial leases that debt and lease financing are perfect
substitutes. Therefore the fourth term in Eq. (17.3) will reflect a dollar-for-dollar sub-
stitution of debt tax shield for leasing tax shield for the portion of the asset that
would be debt financed at the project's optimal capital structure. Furthermore, be-
cause leasing and debt are perfect substitutes, the cash inflows to the lessor and the
cash outflows from the lessee have the same risk. Therefore the appropriate before-tax
THE THEORY OF LEASING 623
discount rate for the cash flows in Eq. (17.3) is k b , the borrowing rate. Eq. (17.3) may
be rewritten as
Table 17.2 illustrates how the cash flow definitions of Eq. (17.4) can be used to
compute the present value of a lease contract. The example in Table 17.2 uses the
same numbers we have developed in this chapter. The annual lease fee (assumed to
be paid at the end of each year) is $2624. Except for a rounding error, the NPV of
the lease is $0. This is to be expected in a competitive market where the tax rates
of the lessor and lessee are identical. There is no advantage to leasing over borrowing
in these circumstances.
One problem in applying Eq. (17.4) is that the remaining debt balance, D„ declines
each year as the lease fees amortize the principal. This makes computations more
cumbersome than they need be. A much simpler way of looking at the problem was
derived by Levy and Sarnat [1979]. Eq. (17.4) puts the tax effect of displaced debt in
the numerator and discounts at the before-tax rate, k b . However, it has been our
practice throughout the text to account for the tax effect of financing costs by dis-
counting at the after-tax rate in the denominator and writing the numerator in terms
of after-tax cash flows net of financing effects. If this is done, Eq. (17.4) is equivalent
to Eq. (17.5) below:
— T e + T edept
NPV(to lessee) = I — (17.5)
[1 + (1 — T k b ] t
Also, recall that (1 — -3 e k b = WACCB. Substituting the numbers from our previous
example in order to determine the NPV of the lease, and assuming that the tax rate
of the lessee is the same as the lessor, namely, 40%, we have
The numerical result shows very clearly that the risk and cash flows to the lessor and
lessee are identical if they have the same marginal tax rates. There is an equilibrium
between the lessor market and the user market. Each term in Eq. (17.5) is identical
to the corresponding term in Eq. (17.1) except that the signs are reversed. In other
words a cash outflow to the lessee is a cash inflow to the lessor, and vice versa.
ONISVTI JO SOINONODI MI,
Table 17.2 NPV (to lessee) Using Eq. (17.4)
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11)
Year depi -r,depi (1 — '01., D1 _ 1 Ak a D, T,Ak,,D, AD, CE,
-
(1 + k b) ' PV(CF,)
1 2,000 800 1574.4 10,000 1,000 400.0 1,638 2774.4 .909 2,521.93
2 2,000 800 1574.4 8,362 836 334.4 1,802 2708.8 .826 2,237.47
3 2,000 800 1574.4 6,560 656 262.4 1,982 2636.8 .751 1,980.24
4 2,000 800 1574.4 4,578 458 183.2 2,180 2557.6 .683 1,746.84
5 2,000 800 1574.4 2,398 240 96.0 2,398 2470.4 .621 1,534.12
10,000 10,020.60
L, = the annual lease fee on a fully amortized lease, i.e., no salvage value,
der), = the annual (straight-line) depreciation write-off,
T, = the lessee's marginal tax rate,
D,_ 1 = the face value of debt displaced by the lease in the previous time period,
k„ = the before-tax cost of the displaced debt capital
Ak d D, = the change in the interest payments on debt displaced by the lease,
T, Alc,D, = the change in the interest tax shield displaced by the lease,
AD, = the repayment of principal on debt,
CF, = the cash flow for the lease contract = columns (3) + (4) + (7),
(1 + kb) -` = the present value factor.
THE THEORY OF LEASING 625
will depend on the specific asset life (lease period), depreciation schedule, capitaliza-
tion rates, and leverage policies involved.
See their article for a more complete description of the many nontax determinants of leasing. Only a few
are mentioned here.
THE THEORY OF LEASING 627
interval considerably shorter than the asset's economic life, then a demand for short-
term leasing will arise. If enough people have uncorrelated demand, then the volume
of business will be sufficient for someone to specialize in renting the asset, thereby
creating a supply.
It is conceivable that even though you want to use a tuxedo for one day, you
could buy it at the beginning of the day and sell it at day's end. If so, why is leasing
the preferred contract for obtaining short-term use of the asset? A supplier of short-
term leases, e.g., a hotel owner, could arrange a one-day sale and buyback of a room
rather than renting it out. There are, of course, some obvious transactions cost savings
from short-term leasing. For example, the transactions demand for money is much
less if people only have to exchange currency worth one day's use of an asset rather
than its full value. Also, the need for the user to separately contract for insurance and
financing is reduced.
Another consideration that favors short-term leasing over sale and buyback is the
cost of evaluating the quality of the leased asset. A lease is a contract for the use of
an asset, not its ownership. A potential owner of an asset, e.g., an automobile, will
wish to have a detailed inspection of the car's quality. On the other hand, a lessee
will only need to perform a less costly inspection because the potential loss from using
a low-quality asset for one day is less than the loss from owning it. Also, from the
owner's point of view the gain from deceiving a customer about an asset's quality is
not as great for a short-term lease as for long-term ownership. Thus the lower cost of
quality evaluation favors lease contracts over ownership. However, this advantage
must be weighed against higher average usage costs. For example, a lessee has less
incentive to take proper care of an asset than an owner. Lessors know this and include
a "moral hazard" cost as part of the lease fee. Still, leasing will be preferred if the
moral hazard cost is less than the cost of the frequent detailed inspections that would
be necessary if ownership were exchanged.
Thus whenever the desired period of usage is less than the economic life of an
asset, short-term leasing may be preferred to an ownership market such as a second-
hand market.
6. Leveraged Leasing
A leveraged lease is one where the lessor borrows a substantial portion of the
purchase price of the asset. Figure 17.1 is a schematic representation of the parties
to a leveraged lease. The lender typically holds a first mortgage on the asset. Also,
the lessor assigns the lease payments to the lender (or a trustee). The debt interest
and principal are deducted by the lender, who then returns the balance of the lease
payment which is kept by the equity holder. Equity may be supplied either directly
by the lessor or indirectly by third parties. The loan arrangement is called a non-
recourse loan because its effect is to indemnify the lessor in the event of default. The
lessor benefits from the investment tax credit (if any) created when the asset is pur-
chased, the depreciation tax shield, the residual value of the equipment (if any), the
interest tax deduction, and the equity payments (if not turned over to third-party
equity investors).
628 THE ECONOMICS OF LEASING
ILease fees
Equity Debt
Figure 17.1
A schematic representation of a leveraged lease.
The existence of leveraged leases may be explained by various institutional considerations and agency
costs. For example, for very large leveraged leases a single bank may be constrained from doing the deal
because regulations prohibit it from lending more than a small percentage of its equity to a single firm.
Also, leveraged leases match long-term borrowing against long-term lending (i.e., the lease), whereas most
of a bank's other loans and deposits are short or intermediate term.
THE THEORY OF LEASING 629
may be supplied, e.g., by an insurance company. We can get an idea of the rate of
return that will be required on debt by noting that the lease payments must be riskier
than the debt. In the event of default on a leveraged lease the debt holders are some-
what protected because they own the mortgage on the leased assets. By way of con-
trast the equity holders, being residual claimants, can lose everything. The lessor
charges a 10% before-tax rate of return on the lease. This results in an 8% before-tax
cost of borrowing, k b , which is the rate at which the debt holders will supply capital
to the leveraged lease.
The cost of equity depends on the amount of leverage used in the lease. For
"blue-chip" leases the ratio of debt to total assets might be 90-95%. We assumed the
ratio was 90%. Given these facts, we can employ the Modigliani-Miller cost of capital
definitions (from Chapter 13) to compute the cost of equity in the leveraged lease.
Recall that we assumed, earlier in the chapter, that the unlevered cost of equity, p,
was 9.375%. The Modigliani-Miller definition of the, cost of equity is
B
= p + (p k b)(1 — tc) —
where
p = the unlevered cost of equity = 9.375%,
kb = the before-tax cost of borrowing = 8%,
2, = the marginal effective tax rate = 40%,
B
= the market value debt-to-equity ratio = 9.
Time
Figure 17.2
Changes in replacement value over time.
For a paper on how to deal with uncertain salvage values the reader is referred to Lee, Martin, and
Senchack [1980] or to McConnell and Schallheim [1983].
THE THEORY OF LEASING 631
depreciate an equal amount each year (i.e., the replacement value declines in a straight
line at the rate of 1 — 0 each year, where 0 = .667). However, its value may be 50%
higher or lower than expected at the end of a given year. Also, the lessor has a tax
rate of 40% and will write a two-year lease. If the lease contract were a straight
financial lease, it would require a 10% before-tax rate of return (i.e., kb = 10%). The
salvage value is uncertain and requires a 16% risk-adjusted rate of return.' Using
Eq. (17.1) we can write the competitive net present value of the lease, if it were non-
cancelable, as follows:
<
Tcdep, + E(MV)
0 = —I + E (= - TAL1 t
(17.6)
t=i [1 + ( — "Oka (1 + k1)2 '
where
E(MV) = the expected market value of the asset after two years,
k = = the risk-adjusted discount rate for the salvage value.
Substituting in the numbers, we have
<
.4)L, + .4(3333)
(= - 3333
0 = —10,000 + E +
(1.16)2
t=i [1 + (1 — .4).10]t
Solving, we find that the competitive lease fee is $4617.
Next, we want to determine the competitive lease payments assuming that the
above contract is a cancellable operating lease. Equation (17.6) must be modified by
subtracting out the present value of the American put option, P. The new valuation
equation is
<
(= - T A t t edep, + E(MV)
0= —I +E P.
t=i [1 + (1 — tc)kb]t (1 + k =) 2
The value of the put (see Copeland and Weston [1982]) is determined to be $850.
Solving for the cancellable lease fee we find that it increases to $5392. The lease fee
has increased substantially to reflect the extra risk of possible early cancellation of
the operating lease."
The example serves to illustrate that the replacement cost uncertainty borne by
the lessor in a cancellable operating lease can have a profound impact on the
competitive lease fee. In general the cancellation clause may be thought of as an
American put purchased by the lessee. Its cost will increase with (1) greater uncertainty
in the replacement value of the leased asset and (2) decreases in the risk-free discount
rate.
9
For simplicity we ignore capital gains taxation on the salvage value as well as investment tax credits.
=0
If a lessee takes the lease fee of $5392 as an input and tries to compute an internal rate of return
on the contract using Eq. (17.6), the noncancellable lease formula, the IRR will be approximately 14%.
However, the lessee would be mistaken to compare the 14% return on a cancellable lease with 10% on a
straight financial lease (or comparable debt financing).
632 THE ECONOMICS OF LEASING
The empirical evidence on leasing-related issues is scant. There are (at least) four
topics of interest. First is whether or not the theory of leasing is sufficiently rich to
predict when a firm will use leasing instead of debt financing. Although debt and
leasing are very similar, they are not necessarily perfect substitutes. Otherwise the
form of financing being used would be randomly distributed across firms. Casual
empiricism suggests that patterns of choice do exist. Factors such as tax shield utiliz-
ation, economies of scale in service contracts, and comparative advantages in resale
of equipment may explain the use of lease financing.
A second issue is the extent to which lease financing is a substitute for debt
financing within a given firm. The theory of leasing logically assumes that each dollar
of leasing utilized by the firm will replace one dollar of debt capacity; i.e., they are
perfect substitutes. Whether or not firms actually behave in this manner is an empi-
rical issue. Bowman [1980] collected a 1973 sample of 92 firms in seven different
industries (according to the two-digit SIC code) where both lease and debt financing
were reported. A second sample of 158 firms that did not use leasing was also col-
lected. Bowman then ran a cross-section regression to explain the systematic risk of
the nonlease sample as a function of the accounting beta and the debt-to-equity ratio.
The results are given below, with t-statistics in parentheses:
= 1.223 + 0.880 + .104(D/S),, (R2 = .29),
(2.51) (4.19)
where
Qi = the systematic risk of the ith firm estimated using the last 60 months of data,
= an estimate of the accounting beta for the ith firm. Two versions were tested
without obtaining different results. One version used the first difference of net
income before extraordinary items and the second used EBIT.
(D/S)i = the book value of debt over the market value of equity.
The association between systematic risk and lease utilization was then tested by first
adjusting the beta of the ith firm by using the above estimated coefficients, then
regressing the result against a lease utilization variable as shown below:
where
(L/S), = the book value of leasing over the market value of equity,
ei = the residual.
This research design controls for measurement error and multicollinearity. The rela-
tionship between systematic risk, )3,, and the use of leasing, (L/S)i, was positive and
statistically significant. This led Bowman to conclude that debt and lease financing
EMPIRICAL EVIDENCE ON LEASING 633
both affected the market's estimate of the systematic risk of firms. Consequently,
leasing and debt were recognized as close substitutes.
Ang and Peterson [1984] use cross-sectional data for approximately 600 nonre-
gulated and nonfinancial firms each year between 1976 and 1981. The book value
lease to equity ratio was explained as a function of the book value debt to equity
ratio, operating leverage, the coefficient of variation of sales, return on net fixed plant,
the price/earnings ratio, the book value of assets, and the current ratio. The results
indicated a significant positive relationship between leasing and debt. The conclusion
is that debt and leases appear to be complements, rather than perfect substitutes as
assumed in the theory of leasing.
A third area of empirical interest is whether or not the mandated disclosure of
off–balance sheet leasing had any measurable impact on the real decisions of firms.
For example, in order to comply with the accounting disclosure changes, firms had
to capitalize their lease obligations and report them on their balance sheets. When
this was done, some firms found themselves in technical violation of their debt
covenants. Ro [1978] tested whether the disclosure of lease information had an impact
on security prices and concluded that it had an adverse impact. Abdel-Khalik,
Thompson, and Taylor [1978] looked at the impact of lease disclosure on bond risk
premia and found no impact on security prices. Thus the evidence is mixed. As more
empirical research is reported we will obtain a better understanding of how the
market reacts (if at all) to the initial disclosure of off–balance sheet financing.
Finally, studies by Sorensen and Johnson [1977], McGugan and Caves [1974],
Gudikunst and Roberts [1975], and Crawford, Harper, and McConnell [1981] have
empirically estimated the internal rates of return (i.e., the "yields") on commercial
bank leases. They all report that the estimated yields were higher than debt of equi-
valent risk. If leases and debt are perfect substitutes in straight financial leases, this
should not be so. Schallheim, Johnson, Lease, and McConnell [1986] provide a
potential explanation for this puzzle. They find that the higher yields on financial
leases are related to the discounted value of the leased asset's residual value covariance
risk (or to the residual value itself). Franks and Hodges [1986] use a sample of
English leasing data and conclude that low tax firms are lessees and high tax firms
are lessors, but interestingly most of the tax shelter value was captured by the lessor
rather than the lessee. Their result implies that the estimated yields on leases would
be higher than the debt rate because taxable earnings (owned by lessors) was a scarce
resource and lessees had to pay a premium for using it.
SUMMARY
We have examined the leasing problem from the point of view of the lessee and the
lessor. If they have identical tax rates, then a competitive lease fee will have zero net
present value to both parties. Yet if their leasing analysis inputs are different—e.g.,
the lessee may have a higher tax rate than the lessor—then it is possible that some
negotiated lease fee can have a positive net present value to both parties. They can
share the net present value of the tax shield from leasing (if any).
634 THE ECONOMICS OF LEASING
Strict financial leases are assumed to be perfect substitutes for debt capital. There
is no such thing as 100% lease financing, just as there is no such thing as 100% debt
financing. For each dollar of leasing employed by a firm, one dollar of debt capacity
is displaced.
Leveraged leasing is really no different from straight financial leasing. Both
employ debt and equity that are used by the lessor to purchase the asset for leasing.
If one knows the financing mix that will be used by the lessor, the rate of return on
the lease payments, and the (lower) rate of return received by debt holders in the
contract, then it is possible to estimate the required rate of return on equity invested
in leveraged leases.
Cancellable leases contain a put option sold to the lessee. Often valuable, this
option may considerably raise the implied lease cost.
The character of short-term leases is quite different from that of long-term finan-
cial leases. In particular the transactions costs and agency costs of short-term leases
are important in explaining why we rent hotel rooms instead of buying them for one
day and then selling them back the next.
The empirical evidence on leasing is scant. What little there is provides mixed
evidence on whether or not leasing is viewed by the market as a close substitute for
debt financing.
PROBLEM SET
17.1 Your firm is considering lease financing for a computer that is expected to have a five-year
life and no salvage value (it is a strict financial lease). You have the following facts:
• Your firm's tax rate is 30%. There is no investment tax credit.
• If purchased, the project would require a capital outlay of $100,000.
• The project will be depreciated using the straight-line method.
• Debt of equivalent risk costs 10% before taxes.
• The annual lease fee is $32,000 paid at the beginning of each year for five years.
• The optimal capital structure for the project is 50% debt to total assets.
Should you use lease financing or not?
17.2 Giveaway State Teacher's College is trying to decide whether to buy a new computer or
to lease it from Readi Roller Leasing. The computer costs $500,000. Giveaway has a zero tax
rate, whereas Readi Roller enjoys a 40% tax rate. There is no investment tax credit. The
computer is expected to last five years and have no salvage value. It will be depreciated using
the straight-line method. The college can borrow at a 15% interest rate. If the five annual lease
fees are $147,577 paid at the end of each year,
a) What is the NPV of the lease for Readi Roller Leasing Co.?
b) What is the NPV of the lease for Giveaway State?
c) What do the results tell you about the lease/buy decision for tax-free institutions?
17.3 This question involves a more realistic set of facts and therefore requires a more detailed
analysis of cash flows than contained in the chapter. Your company is going to negotiate a
REFERENCES 635
lease contract for manufacturing equipment. You have the following facts:
• The equipment cost $100,000 and is expected to have a five-year life with an expected
salvage value of $10,000; however, it can be fully depreciated in four years using the sum-
of-the-years digits method of accelerated depreciation.
• Whether leased or owned, the equipment will be sold for $1 at the end of the fifth year,
and the owner will pay a capital gains tax equal to one half of the ordinary income tax
rate on the difference between $1 and the book value.
• Your firm will pay no taxes for the next two years and then will return to its normal 48%
tax rate.
• The leasing firm will require equal annual lease payments paid at the beginning of each
year. The leasing firm's tax rate is 48%.
• Your firm can borrow at the prime rate plus 1%, i.e., at 17%.
a) What is the maximum lease payment that your firm can afford to offer in the negotiations?
b) What is the minimum lease payment that you think the leasing company can accept in
the negotiations?
17.4 Your firm has been approached to become an equity participant in a leveraged leasing
deal. You need to estimate the minimum rate of return on equity that is acceptable. You have
collected the following facts:
• The asset to be leased will cost $100,000,000; 90% will be financed with debt and the
remaining 10% with equity.
• The debt portion of the financing is to receive a 14% rate of return before taxes.
• Your tax rate is 40%. The lessor's tax rate is 48%.
• The before-tax rate of return that the lessee will be paying is 18%.
Use the Modigliani-Miller cost of capital assumptions to make your analysis (i.e., assume a
world with corporate taxes only).
17.5 The Mortar Bored Company was considering whether to buy a new $100,000 reduction
machine or to lease it. It was estimated that the machine would reduce variable costs by
$31,000 per year and have an eight-year life with no salvage value. The machine will be depre-
ciated on a straight-line basis, and there is no investment tax credit. The firm's optimal capital
structure is 50% debt to total assets, its before-tax costs of debt and equity are 15% and 25%,
respectively, and it has a 40% tax rate. If it were to lease, the fees would be $21,400 per year
paid at the end of each year.
a) What is the NPV of the project if the firm owns the project?
b) What is the NPV of the lease to the company?
c) Should the company lease the project? Why or why not?
REFERENCES
Abdel-Khalik, A.; R. Thompson; and R. Taylor, "The Impact of Reporting Leases Off the
Balance Sheet on Bond Risk Premiums: Two Explanatory Studies," Accounting Research
Center Working Paper # 78-2, University of Florida, February 1978.
6Ç6 THE ECONOMICS OF LEASING
Ang, J., and P. Peterson, "The Leasing Puzzle," Jou]nal of Finance, Septemîer =984, =0PP-=06P.
Athanosopoulos, P., and P. Bacon, "The Evaluation of Leveraged Leases," Financial Manage-
ment, Spring =980, 76-80.
Bower, R., "Issues in Lease Financing," Financial Management, Winter =973, <P-33.
Bowman, R., "The Deît Equivalence of Leases: An Empirical Investigation," Accounting Re-
view, April =980, <37-<P3.
Brealey, R., and C. Young, "Deît, Taxes and Leasing A Note," Jou]nal of Finance, Decemîer
=980, =<4P-=<P0.
Copeland, T., and J. F. Weston, "A Note on the Evaluation of Cancellaîle Operating Leases,"
Financial Management, Summer =98<, 60-67.
Cox, J.; S. Ross; and M. Ruîinstein, "Option Pricing: A Simplified Approach," Jou]nal of
Financial Economics, Septemîer =979, <<9-<64.
Crawford, P.; C. Harper; and J. McConnell, "Further Evidence on the Terms of Financial
Leases," Financial Management, Autumn =98=, 7-=4.
Dyl, E., and S. Martin, Jr., "Setting Terms for Leveraged Leases," Financial Management,
Winter =977, <0-<7.
Elam, R., "The Effect of Lease Data on the Predictive Aîility of Financial Ratios," Accounting
Review, January =97P, <P-43.
Financial Accounting Standards Board, Statement of Financial Accounting Standa]ds no. 13,
Stamford, CT., =976.
Flath, D., "The Economics of Short-Term Leasing," Economic Inqui]y, April =980, <47-<P9.
Franks, J., and S. Hodges, "Valuation of Financial Lease Contracts: A Note," Jou]nal of
Finance, May =978, 647-669.
"Lease Valuation When Taxaîle Earnings Are a Scarce Resource," Working paper,
London Business School, =986.
Gudikunst, A., and G. Roîerts, "Leasing: Analysis of a Theoretic-Pragmatic Dilemma," paper
presented at the annual meeting of the Financial Management Association, Kansas City,
Octoîer =97P.
Heaton, H., "Corporate Taxation and Leasing," Jou]nal of Financial and Quantitative Analysis,
Septemîer =986, 3P=-3P9.
Idol, C., "A Note on Specifying Deît Displacement and Tax Shield Borrowing Opportunities
in Financial Lease Valuation Models," Financial Management, Summer =980, <4-<9.
Johnson, R., and W. Lewellen, "Analysis of the Lease-or-Buy Decision," Jou]nal of Finance,
Septemîer =97<, 8=P-8<3.
Kim, E.; W. Lewellen; and J. McConnell, "Sale-and-Leaseîack Agreements and Enterprise
Valuation," Jou]nal of Financial and Quantitative Analysis, Decemîer =978, 87=-883.
Lee, W.; J. Martin; and A. J. Senchack, "An Option Pricing Approach to the Evaluation of
Salvage Values in Financial Lease Arrangements," Working paper, University of Texas
at Austin, Septemîer =980.
Levy, H., and M. Sarnat, "Leasing, Borrowing and Financial Risk," Financial Management,
Winter =979, 47-P4.
Lewellen, W.; M. Long; and J. McConnell, "Asset Leasing in Competitive Capital Markets,"
Jou]nal of Finance, June =976, 787-798.
Long, M., "Leasing and the Cost of Capital," Jou]nal of Financial and Quantitative Analysis,
Novemîer =977, P79-P98.
REFERENCES 637
McConnell, J., and J. Schallheim, "Valuation of Asset Leasing Contracts," Jou]nal of Financial
Economics, August =983, <37-<6<.
McGugan, V., and R. Caves, "Integration and Competition in the Equipment Leasing Indus-
try," Jou]nal of Business, July =974, 38<-396.
Miller, M., and F. Modigliani, "Some Estimates of the Cost of Capital to the Electric Utility
Industry, =9P4-P7," Ame]ican Economic Review, June =966, 333-39=.
Miller, M., and C. Upton, "Leasing, Buying and the Cost of Capital Services," Jou]nal of
Finance, June =976, 76=-786.
Myers, S., "Interactions of Corporate Financing and Investment Decisions—Implications for
Capital Budgeting," Jou]nal of Finance, March =974, =-<P.
; D. Dill; and A. Bautista, "Valuation of Financial Lease Contracts," Jou]nal of Finance,
June =976, 799-8=9.
Ofer, A., "The Evaluation of the Lease versus Purchase Alternatives," Financial Management,
Summer =976, 67-7<.
Perg, W., "Leveraged Leasing: The Proîlem of Changing Leverage," Financial Management,
Autumn =978, 47-P=.
Ro, B., "The Disclosure of Capitalized Lease Information and Stock Prices," Jou]nal of
Accounting Resea]ch, Autumn =978, 3=P-340.
Roîerts, G., and A. Gudikunst, "Equipment Financial Leasing Practices and Costs: Comment,"
Financial Management, Summer =978, 79-8=.
Schall, L., "Asset Valuation, Firm Investment, and Firm Diversification," Jou]nal of Business,
January =97<, ==-<8.
, "The Lease-or-Buy and Asset Acquisition Decisions," Jou]nal of Finance, Septemîer
=974, =<03-=<=4.
Schallheim, J.; R. Johnson; R. Lease; and J. McConnell, "The Determinants of Yields on
Financial Leasing Contracts," Working paper, University of Utah, =986.
Smith, C. Jr., and L. Wakeman, "Determinants of Corporate Leasing Policy," Jou]nal of
Finance, July =98P, 896-908.
Snyder, N., "Financial Leases," in Weston and Goudzwaard, eds., The T]easu]e]'s Handbook.
Dow-Jones Irwin, Homewood, Ill., =976, 783-8<<.
Sorensen, I., and R. Johnson, "Equipment Financial Leasing Practices and Costs: An Empirical
Study," Financial Management, Spring =977, 33-40.
Stickney, C.; R. Weil; and M. Wolfson, "Income Taxes and Tax Transfer Leases," Accounting
Review, April =983, 439-4P9.
Taylor, D., "Technological or Economic Oîsolescence: Computer Purchase vs. Lease,"
Management Accounting, Septemîer =968.
Vancil, R., "Lease or Borrow—New Method of Analysis," Ha]va]d Business Review, Septem-
îer–Octoîer =96=.
Vanderwicken, P., "The Powerful Logic of the Leasing Boom," Fo]tune, Novemîer =973,
=36 –=40.
18
There is nothing on the label of a tool which says whether its intended
use is harmful or beneficial, only how to use it, and sometimes that it
should be used with caution.
Anonymous
Applied Issues in
Co]po]ate Finance
This chapter covers miscellaneous corporate finance topics that are (or should be)
of particular interest to chief financial officers. We start with an analysis of pension
fund management—a topic that requires prior knowledge of option pricing (Chap-
ter 8), portfolio theory (Chapter 6), and Modigliani-Miller tax effects (Chapter 13).
Then we focus on a variety of shorter topics. They include swaps of fixed for variable
rate debt, going-private transactions, leveraged buyouts, and executive compensation
schemes.
638
PENSION FUND MANAGEMENT 639
employer and must be disclosed in the firm's financial statements. A pension fund
is established on behalf of employees and is managed by a trustee, who collects cash
from the firm, manages the assets owned by the fund, and makes disbursements to
retired employees. The firm is able to expense pension fund contributions for tax
purposes. The fund pays no taxes on its earnings. However, beneficiaries must pay
personal taxes upon receiving retirement payments from the fund. Hence pension
funds are a tax-favored form of employee compensation because taxes are deferred
until retirement.
The composition of pension fund assets is given in Table 18.1. Most pension
funds hold their assets in the form of marketable securities: money market accounts,
bonds, and stocks. Because pension fund earnings are not taxed, it never pays to hold
municipal bonds because their low tax-exempt interest rates are always dominated
by the higher interest paid by taxable bonds. Direct investment in real estate (with
the possible exception of undeveloped land) is also not advisable because most real
estate investments are priced such that the investor must be in a relatively high tax
bracket in order to receive a positive after-tax return. Pension funds are in a zero
tax bracket.
The most striking change in pension fund portfolio composition over the past
four decades is the decline in the proportion invested in bonds from 74.3% in 1950
to 35.9% in 1980, and the increase in stocks (common and preferred) from 17.1% in
1950 to 50.5% in 1980.
Later on in the chapter we shall discuss some of the possible influences that may
affect the composition of assets in pension fund portfolios. Although the pension
fund can profitably hold taxable securities, it is not immediately clear what percentage
should be held in the form of interest-bearing securities (money market funds and
bonds) or common stock.
Table 18.2 gives the format for a typical pension fund income statement and
balance sheet. Cash inflows to the fund are provided by corporate contributions,
employee contributions, dividends and interest earned by the fund's stocks and bonds,
and capital gains. Cash outflows are management fees, brokerage expenses, disburse-
ments to beneficiaries, and capital losses. The change in the net fund balance is the
difference between inflows and outflows. The fund's profit is not taxable. Marketable
640 APPLIED ISSUES IN CORPORATE FINANCE
securities is the only item in the pension fund balance sheet that is not the result of a
present value calculation. The present value of future contributions to the fund is the
other major asset. Contributions are received in two forms: cash from the firm and
earnings on the fund's assets. A major issue is: What rate of return will be generated
from the fund's assets? If the return is high, then the firm can reduce the amount of
cash it puts into the fund. As we shall see, later in the chapter, the rate of return as-
sumption is a tricky decision.
Liabilities are subdivided into two categories. The present value of benefits from
past service is handled one of two ways. Some companies calculate the present value
of vested benefits only. These are the benefits that would be paid if all employees left
the firm immediately. However, it is typical that employees become vested in the pen-
sion plan only after accumulating a minimum period of seniority, say five years. If
they leave prior to five years, they receive none of their promised pension benefits. An
alternative procedure is to calculate the present value of all benefits accrued for past
service whether employees are fully vested or not. Hence accrued benefits will usually
be larger than vested benefits because not all employees are fully vested. Regardless
of how the present value of benefits from past service is handled, total pension liabil-
ities remain unchanged. If only vested benefits are included in the present value of
benefits for past service, then unvested benefits are included in the second liability
category.
The second major liability item is the present value (PV) of benefits for future
service. Its computation is complex and depends on actuarial assumptions about
the amount of employee turnover, the age and seniority of retiring employees, their
life expectancy, and the choice of a discount rate for present value computations.
Of major concern to all parties is the size of the pension fund deficit or surplus.
An unfunded deficit is an asset of the pension fund (as shown in Table 18.2) and a
liability of the firm, and it can be enormous. For example, had the pension liabilities
of Du Pont been included, its balance sheet for the end of its 1984 fiscal year would
have looked like Table 18.3. The $7.6 billion pension liability represents the vested
PENSION FUND MANAGEMENT 641
liabilities of Du Pont, i.e., the liability that would be incurred if all the employees left
the firm at the end of 1984. Du Pont's pension was overfunded by $800 million In
principle, this money "belongs" to shareholders. Even though the pension was over-
funded, the addition of pension assets and liabilities to the balance sheet raised Du
Pont's debt-to-total-assets ratio from 49% to 60%.' Clearly, pension fund liabilities
are important enough to require full disclosure.
1
The effect of the pension fund on the balance sheet is to increase assets by $8.4 billion, to increase pension
liabilities by $7.6 billion, and to increase equity by $0.8 billion (the amount of overfunding). Note that
Table 18.3 is purely hypothetical and does not conform to the generally accepted accounting practices that
are discussed later in the chapter.
642 APPLIED ISSUES IN CORPORATE FINANCE
The provisions of ERISA are many. No employee older than 25 years and with
more than 1 year of service with a company, or hired more than 5 years before
normal retirement age, may be excluded from participation in that company's pen-
sion plan. Prior to ERISA, unusual vesting practices resulted in many injustices. For
example, some plans required 20 or more years of uninterrupted service before an
employee became vested. Sometimes workers would be fired in their nineteenth year
simply to prevent vesting them in a pension plan. With the advent of ERISA, all
plans must choose from one of three minimum vesting schedules for the corporate
portion of the contributions to the pension plan:
All employee contributions to a pension fund, and investment returns on such contri-
butions, are fully vested from the beginning.
ERISA legislates the minimum funding of defined benefit plans, whereas the IRS
(Internal Revenue Service) sets limits on the maximum corporate contribution. Ac-
cording to ERISA, the minimum contribution is determined as follows: (1) all normal
costs attributable to benefit claims deriving from employee services in a given year
must be paid that year; (2) any experience losses (caused by a decline in the value of
the securities in the fund, by unexpected changes in employee turnover, or by changes
in actuarial assumptions about the discount rate) must be amortized over a period
not to exceed 15 years; and (3) supplemental liabilities resulting from increased bene-
fits or unfunded past service costs must be amortized over a period not to exceed 30
years (40 years for companies with pre-ERISA supplemental liabilities). On the other
hand, the IRS defines the maximum corporate pension contribution as the actuarially
determined normal cost of the plan plus any amount necessary to amortize supple-
mental and experience losses over a 10-year period. The ERISA and IRS restrictions
limit corporate discretion over the amount of funds contributed to a plan.
One of the most important provisions of ERISA was the creation of the Pension
Benefit Guaranty Corporation (PBGC). It is a pension insurance fund operated under
the supervision of the U.S. Department of Labor. Corporations must pay the PBGC
a fixed annual premium (currently $2.60) for each employee in a pension plan. This
central fund is then used to guarantee pension benefits even if a plan fails. A pension
plan may be terminated voluntarily by the corporation or involuntarily by the PBGC
upon court order. The PBGC may terminate a plan (1) if the plan fails to meet
minimum funding standards, (2) if the plan is unable to pay benefits when due, (3) if
the plan is administered improperly, or (4) if the liability of the PBGC for fulfilling
claims deriving from the plan is likely to increase unreasonably.
If a plan is terminated because it is underfunded, the company is liable for 100%
of the deficit up to 30% of the company's net worth. Furthermore, the PBGC may
PENSION FUND MANAGEMENT 643
place a lien on corporate assets that has the same priority as federal taxes. Hence
unfunded pension liabilities are equivalent to the most senior debt. A bankrupt firm
may have few assets to pay to the PBGC; hence a worthy public policy question is
whether the PBGC has enough resources of its own to adequately insure pensioners
of major corporate bankruptcies. In July 1987 the PBGC had total assets of $3 billion
and faced total obligations of $7 billion for people currently retired or who would
retire under plans of which it was trustee. The agency also faced cash flow problems
because the premium income that it collected from corporations and the dividends
from its investments were less than the benefits that it had already undertaken to
pay. These were the realized liabilities. The present value of the PBGC potential
liabilities is much larger. Marcus [1985] estimates the present value of the PBGC
insurance liability for a sample of 87 of the Fortune 100 companies based on their
1982 annual reports. His estimates range from $6.7 to $14.8 billion.
other purposes; then, during years of high profitability, payment can be accelerated
(up to the limits established by the IRS) as a way of sheltering cash flows from income
taxes. Defined contribution plans do not allow similar flexibility because they are
established as a fixed percentage of employee compensation. We might expect to see
defined benefit plans used by corporations, especially those that can benefit from tax
planning, and defined contribution plans are more likely to be used by nontaxable
entities and by partnerships.
FASB Statement No. 36 allows companies to use different interest rate assumptions for disclosure in
the annual report and for funding purposes; for example, Bethlehem used 7% for funding and 10% for
disclosure. See Regan [1982].
PENSION FUND MANAGEMENT 645
There are two major pension fund assets. First is the current market value of
the stocks, bonds, mortgages, and so forth, held by the pension fund. Second is the
present value of the expected pension fund contributions, which are multiplied by
one minus the corporate tax rate (1 — r) in order to reflect the fact that pension
fund contributions are tax deductible by the firm. As long as the firm is making
profits, then pension contributions are "shared" with the government because more
contributions mean lower taxes.' Balancing the pension fund assets is the pension
fund liability, the present value of expected pension fund benefits to be paid to
employees.
The main difference between the book value of the pension fund deficit and its
market value, or true economic value, PFL, is reflected in the rates of return (discount
rates). Equation (18.3) further elaborates Eq. (18.2) by showing the present value of
the pension fund along with the appropriate market-determined discount rates:
PFL = — Market value of pension fund assets
n
, E(contributions in year t)(1 —
t=i [1 + k b(1 — TA'
E(benefits in year t)
(18.3)
t=1 (1 + k b)t
The expected pension benefits are discounted at the pretax cost of senior debt, k b ,
because ERISA has made the payment of pension benefits a senior obligation of the
firm, second only to tax liabilities.' Pension contributions are also discounted at the
rate k b but on an after-tax basis. Prior to ERISA, the expected benefits would have
been discounted at the cost of junior, or subordinated, debt, kJ, which is higher than
k b , the cost of senior debt. One of the major effects of ERISA was to transfer wealth
from shareholders to pension beneficiaries by increasing the present value of pen-
sion deficits, PFL. The transfer was especially large for plans that were seriously
underfunded.
The real effect of a change in pension plan actuarial assumptions depends on
the cash flow consequences. If the actuarial discount rate assumption is raised, then
the present value of accumulated benefits in book value terms decreases, as do the
normal costs that have to be paid into the fund. This has the effect of decreasing the
annual expected contributions into the fund and hence decreasing their present value
in Eq. (18.3) because expected contributions decrease, whereas the market-determined
discount rate, k b, does not change. The present value of expected benefits, however,
3
If one considers Social Security to be a pension plan, then recent changes in the Social Security tax
law that require nonprofit organizations to pay Social Security for their employees are burdensome. Be-
cause nonprofit organizations have no tax shelter, they must bear the full cost of Social Security expenses.
Some have argued that promised pension benefits are subordinated to other debt claims in spite of
ERISA because other debt comes due before pension obligations. Pension beneficiaries cannot force the
firm into bankruptcy, whereas debt holders can. The existence of large unfunded pension deficits will, in
our opinion, cause debt holders to force bankruptcy sooner than they might if there were no pension
obligations. Nevertheless, pension liabilities will still be senior claims at the time of bankruptcy.
646 APPLIED ISSUES IN CORPORATE FINANCE
remains unchanged. The net effect is to increase the market value of pension liabil-
ities, PFL. There is usually no effect on the firm as a whole because the cash flow
not put into pension fund contributions may be used either to decrease other liabil-
ities or to increase assets. Either way, the increased pension liability is exactly offset.'
Thus we see that, from the shareholders' point of view, changing the actuarial as-
sumptions in order to change pension contributions is usually an exercise in futility.
Even worse, if the funds generated by cutting pension contributions are used for a
purpose that is not expensed (e.g., repaying the principal on debt), the effect is to
increase taxable income and decrease net cash flows to shareholders. Accounting
profits have increased, but the firm has sacrificed the pension contribution tax shield.
The net effect (assuming the firm is paying taxes) is to benefit the IRS at the expense
of shareholders. Finally, changing actuarial assumptions for disclosure in the annual
report but not for funding purposes is chicanery at best and stupid at worst. If taxes
are based on actual contributions, then, at best, managers think they can somehow
fool the marketplace.
C. CHOOSING THE MIX Of PENSION PLAN ASSETS. As with any other portfolio
decision, the choice of assets for a pension plan involves a selection of risk and return.
Furthermore, tax considerations and pension fund insurance through ERISA are
paramount.
Modeling pension plan payoffs. Before turning to the effect of ERISA and taxes
on pension fund investments, let us build a more complete understanding of their risk
and return characteristics. Prior to the passage of ERISA, corporate pension liabilities
were analogous to risky debt, and the shareholders' position was equivalent to a call
option on a leveraged firm.' To illustrate this, assume a one-period framework, an all-
equity firm that has an uncertain end-of-period market value, V1 , and a world with
no taxes. The pension fund holds some risky assets with an end-of-period value, A t ,
and the pension beneficiaries have been promised an end-of-period benefit, B.
Figure 18.1 shows the end-of-period payoffs to the pension beneficiaries, assum-
ing that the pension fund is uninsured. Along the horizontal axis, we have the market
value of the firm plus the market value of the pension assets, V + A, whereas dollars
of end-of-period payoff are graphed along the vertical axis. The pension beneficiaries
will receive the full promised amount if the market value of total assets, V + A,
exceeds the promised benefits, B. But if not, the pension beneficiaries receive V +
A < B. The solid line OXB in Fig. 18.1 shows the pension beneficiaries' payoff. Be-
cause we have assumed the firm has no debt, the shareholders' payoff is simply the
5
One sometimes hears that pension contributions can be legitimately cut if the funds are alternatively
used to invest in positive net present value projects. This argument confuses the investment decision (take
the profitable project) with the way it is financed (cut pension fund contributions). The project can be
financed either by cutting pension contributions, which increases pension liabilities, or by borrowing,
which increases debt liabilities. Either way, the effect on shareholders' wealth is the same.
6
For a more complete presentation of pension fund liabilities as options, see Sharpe [1976] and Treynor,
Priest, and Regan [1976].
PENSION FUND MANAGEMENT 647
Figure 18.1
End-of-period pension fund payoffs.
V+A
V+A<B V+A>B
Figure 18.2
The pension beneficiaries' position is
IX Buy a bond B equivalent to risky debt (long in a
riskless bond and short in a put option).
Sell a put, —P
Given that this is a one-period model and that pension benefits are not payable until employees retire
at the end of the period, all options in the model are European options. They cannot be exercised before
maturity.
648 APPLIED ISSUES IN CORPORATE FINANCE
the put). Using our current notation, the put-call parity expression becomes
(V0 + A 0 ) + P, = B, + S,,
S, = (V0 + A0 ) — (B o — P 0 ). (18.5)
The shareholders' position, So , is equivalent to a call option on a levered firm. On
the right-hand side of Eq. (18.5), we note that (V0 + A o) is the present value of the
firm and pension fund assets and that (Bo — P o) is the present value of risky debt,
i.e., the pension benefits.
Considerable insight into pension fund asset mix can be provided by this simple
option pricing approach. For example, what happens to shareholders' wealth if the
pension trustees change the mix of pension assets from a well-diversified portfolio
of equity to being 100% invested in shares of the firm?' The effect would be to in-
crease the correlation between V, the value of the firm, and A, the value of the pension
assets. Consequently, the variance of the underlying portfolio of assets increases, and
the value of shareholders' wealth, S,, which is a call option on the assets, will also
increase. Thus the effect of any decision that unexpectedly increases the risk of
(V + A) is to shift wealth to shareholders and away from pension beneficiaries. The
only mitigating circumstance, which was pointed out by Sharpe [1976], is that em-
ployees may be able to demand higher wages to compensate them for the higher
risk they must bear when pension assets are invested in the firm's own stock. Or
they might require pension fund insurance.'
The effects of ERISA and the PBGC. Now let us look at the effect of government
pension fund insurance on the pension fund asset mix but maintain our assumption
that there are no taxes. As was mentioned earlier, the Pension Benefit Guaranty Cor-
poration (PBGC) ensures pension fund liabilities. Corporations contribute into PBGC
a fixed insurance premium per employee each year. In the event that an underfunded
pension plan is terminated, the firm is liable up to 30% of its net worth, and the PBGC
guarantees the remainder of the pension fund liability.
If the PBGC were a privately owned insurance company, it would charge pre-
miums based on the probability of corporate default on a pension fund. However, as
a government organization, it charges all firms exactly the same insurance premium
regardless of the extent of pension plan underfunding or the likelihood of bankruptcy.
One implication, of course, is that firms with overfunded pension funds are paying
too much to the PBGC relative to those with badly underfunded pension plans. An-
other implication is that firms threatened with bankruptcy can decide to change their
pension plan asset mix to maximize the value of the call option that represents their
shareholders' wealth. If they go bankrupt, shareholders receive nothing, and although
the PBGC can claim 30% of each firm's net worth, 30% of nothing is still nothing.
a
This situation is not unusual. For example, at one time, the Sears pension fund had over 50% of its
assets invested in its stock.
9
For more on the economics of insuring portfolios of risky assets, see Gatto, Geske, Litzenberger, and
Sosin [1980].
PENSION FUND MANAGEMENT 649
The PBGC claim on equity is worthless in both Chapter 7 bankruptcy and Chapter
11 reorganization. Consequently, the optimal strategy from the point of view of share-
holders is to put all the pension assets into very risky stocks. If they are lucky the
risky portfolio may do well and even result in overfunding of the pension fund. If
they are unfortunate, then they end up with nothing, which is where they would have
been anyway, and the PBGC has to pay off the pension beneficiaries.
Given that the PBGC undercharges for pension fund insurance for underfunded
plans, then there is the distinct possibility that corporations facing potential bank-
ruptcy can game the PBGC by shifting pension plan assets to being 100% invested
in risky stocks.
An interesting case history of a company in trouble is International Harvester.
In May 1982, the Wall Street Journal reported that International Harvester Company's
pension fund abruptly switched at least $250 million of stock holdings into bonds,
chiefly U.S. government issues. Pension industry executives suggested that the com-
pany was pursuing a strategy that would let it reduce pension contributions. As of
October 31, 1981, Harvester's combined pension assets totaled $1.35 billion.
What are the real economic consequences of Harvester's decision? First, since
the company had negative earnings, it is not likely that the tax consequences of the
decision were important." Second, by changing the actuarial assumptions of the
plan either (1) by realizing gains on the stocks that were sold or (2) by raising the
fund rate of return assumption due to the shift from stocks to bonds, Harvester could
reduce its planned cash contributions to the fund. We have already seen (in the pre-
vious material in this chapter) that the change in actuarial assumptions has no effect
on shareholders' wealth at best and a negative effect at worst. Finally, the analysis
in this section of the chapter suggests that a shift from stock to bonds (in the absence
of tax benefits) decreases shareholders' wealth and benefits pension beneficiaries (and
debt holders) of the firm. Although we have insufficient information to draw a definite
conclusion about the Harvester decision, it looks like the net effect was to diminish
shareholders' wealth.
The effect of taxes. For most firms, pension fund contributions reduce taxes be-
cause they are immediately deductible. At the same time, the pension plan pays no
taxes on its earnings. Hence the rapid growth of pensions is largely attributable to
the fact that they are a form of tax-deferred compensation.
The pension assets should be invested in those securities that have the most fa-
vorable pretax rates of return. Obvious examples of securities that pension managers
should not invest in are those that are used as tax shelters by investors with high
marginal tax rates, such as municipal bonds or real estate with depreciable assets
like buildings.
Perhaps the most interesting tax implication for the pension fund asset mix is
that pension plans should be fully funded and invested totally in bonds as opposed
" The next section of this chapter provides the only rational tax explanation for why Harvester share-
holders may have benefited from switching pension assets to bonds.
650 APPLIED ISSUES IN CORPORATE FINANCE
to equities.' The logic is developed in two parts. The first argument is that the return
on debt held in a corporate pension fund is passed through the firm to its share-
holders in the form of higher share prices because an overfunded pension plan is an
asset of the firm.' The implication is that the return on debt held in the pension
fund is ultimately taxed at the lower personal tax rate on equities. Shareholders will
pay less tax than if the debt were held in their personal portfolios. Consequently,
shareholders are better off if the pension funds of corporations are invested in bonds,
whereas their personal portfolios are invested in equities. This conclusion is based
on the fact that pension plan earnings are not taxed and that bond income is taxed
at a higher rate than capital gains.' It does not depend on any theoretical gain to
leverage (e.g., Chapter 13).
The second reason for investing pension assets in bonds is the potential value
of the tax shelter involved when the firm borrows to invest pension assets in bonds.
The following example compares two pension investment strategies, the first with all
pension assets in stock and the second with all assets in bonds. For the sake of sim-
plicity, we assume a one-period world with two equally likely states of nature. If the
economy is good, stocks will yield a 100% rate of return, whereas bonds will yield
10%. If the economy is bad, stocks yield — 50% and bonds yield 10%. The risk-free
rate is 10%. Note that the expected (or average) return on stocks is 25%, whereas
bonds are expected to yield only 10%. Even so, we will see that the bond investment
strategy is better for shareholders.
Table 18.4 shows a beginning-of-period market value balance sheet that com-
bines the firm and pension fund assets and liabilities for each of the two pension
investment strategies: all stock and all bonds. The firm's defined benefit pension plan
promises to pay $220 million at the end of the period. The present value of this
liability is 8200 million, and it appears on the liabilities side of the corporate balance
sheet. On the assets side, the current market value of the pension assets is $200 million
(either in stock or bonds). The pension plan is fully funded because the present value
of its assets equals that of its liabilities.
If we employ the 100% stock investment strategy for our pension plan, the end-
of-period payoffs are as shown in Table 18.5. Using the "good economy" as an exam-
ple, we see that the pension fund stocks can be sold for $400 million at the end of
the year. After paying the $220 million of pension benefits, shareholders are left with
$180 million pretax and $90 million after taxes. In the "bad economy," they suffer a
$60 million loss. The expected gain in shareholders' wealth is $15 million, but they
are exposed to a great deal of risk. The alternate pension investment strategy is to
invest $200 million in bonds. If that is all we did, the end-of-period payoff would be
exactly $220 million in either economy, the pension benefits would be paid off, and
11
For proof of this proposition, the reader is referred to Tepper and Affleck [1974], Black [1980], and
especially to Tepper [1981].
12
The next section of this chapter discusses ways that shareholders can gain access to the assets of over-
funded pension plans.
13
Even though the 1986 tax code makes the scheduled capital gains rate equal to the ordinary income
rate, the effective capital gains rate is still lower because of the tax-timing option implicit in the realization
of capital gains.
PENSION FUND MANAGEMENT 651
there would îe no gain or loss to shareholders. Their expected gain is zero, îut they
take no risk at all.
To present a valid comparison of the stock and îond strategies, we need to keep
shareholders' risk constant. Then we can compare after-tax expected returns to see
which strategy is îetter, given equivalent risk. Taîle =8.4 shows îalance sheets that
have the same risk for shareholders." On the assets side, $<00 million of îonds is
less risky than $<00 million of stock. Therefore to offset the decline in risk caused
=4
It really does not make any difference, in our example, how risk is measured. Shareholders' risk is
equivalent whether one uses the range, the variance, or the îeta to measure risk.
6P< APPLIED ISSUES IN CORPORATE FINANCE
îy the =00% îond strategy, we increase the firm's financial leverage îy îorrowing
$=00 million and using the proceeds to repurchase $=00 million in equity.' The
resulting payoffs are given in Taîle =8.6.
In the "good economy," the îonds are sold for $<<0 million and the proceeds
are used to pay off the defined îenefits. Next the $=00 million of repurchased equity
is reissued for $<00 million (îecause the stock has appreciated îy =00% in the good
economy). Half of the $<00 million is used to repay the $=00 million of îorrowing, and
$=0 million pays the required interest. Note that the interest payments are tax de-
ductiîle. If the firm is in a P0% tax îracket, then taxes are reduced $P million îelow
what they otherwise would have îeen. Net cash availaîle to shareholders in the
favoraîle state of nature is $9P million with the =00% îond strategy îut is only $90
million with the =00% equity strategy. The îond strategy also dominates the equity
strategy in the unfavoraîle state of nature (— $PP million versus — $60 million). Hence
our example demonstrates the superiority of the îond strategy from the shareholders'
point of view. We have increased their return in îoth states of nature without changing
their risk îecause the range of payoffs is $=P0 million in either case. Regardless of
whether the actual return on stock investments is higher or lower than on îonds,
the îond strategy is preferaîle.
Summarizing, we have seen that investing all pension fund assets in îonds îenefits
shareholders in two ways. First, the pretax îond rate of return is passed through the
firm to its shareholders in the form of higher share prices, which are in turn taxed
at the lower capital gains rate. This argument applies even if there is no gain to
leverage. The second reason for favoring îonds over equity is that there may îe a
=P
In practice, it is not necessary for corporations to actually repurchase shares in order to implement
the =00% îond pension investment plan. What is important is that when pension assets are invested in
îonds rather than stock, the risk of the corporate asset portfolio is lower. Hence, from the point of view
of lenders, there is greater deît capacity. More îorrowing provides a deît tax shield.
PENSION FUND MANAGEMENT 653
Level of
pension funding
Figure 18.3
Corner solutions to the pension funding and asset
mix problem.
gain to firms that can carry more debt without increasing shareholders' risk a gain
to leverage. We have seen that firms that choose to invest pension assets in bonds
actually experience lower total asset risk than firms that put pension assets into stock.
The lower risk means a greater debt capacity. If the firm uses this debt capacity and
if there is a valuable tax shield created by the deductibility of interest payments, then
there is a gain to leverage from investing pension assets in bonds while borrowing
to hold shareholders' risk constant."
Empirical evidence by Landsman [1984] covering a large sample of firms for the
years 1979 through 1981 shows that on average each dollar contributed to the assets
of defined benefit pension funds results in a $1.12 increase in the value of shareholders'
equity, other things being held constant. That the increase is significantly greater
than $1 means that there may, in fact, be a clear tax advantage resulting from pension
assets held by the firm.
Combining the insurance and tax effects. The insurance and tax effects of pension
funding on shareholder wealth seem to suggest corner solutions for the choice of the
level of funding and the type of asset mix. If a firm is successful and is paying high
taxes, then it should make full use of the tax shield provided by pension plans it
should overfund to the maximum extent permitted by law and invest primarily in
bonds. Alternatively, if it is losing money, or if it is not paying taxes for other reasons,
it cannot benefit from the pension fund tax shield and should therefore underfund to
the maximum extent permitted by law and put all the pension assets into risky equities.
This result is illustrated in Fig. 18.3. Note, however, that if the line representing the
present value of the tax shield were steeper, then the plan would be overfunded.
Bicksler and Chen [1985] and Westerfield and Marshall [1983] suggest that this
conclusion may be too strong. Rarely do firms actually go to either extreme. And
they usually have a mix of debt and equity in their pension fund portfolios. There
16
The gain to leverage is most likely to be valuable for those firms that have higher effective tax rates
because their tax shelters from other sources (such as depreciation, research and development expenses,
or tax carryback and carryforward) are limited.
654 APPLIED ISSUES IN CORPORATE FINANCE
Figure 18.4
An interior solution for funding and the choice of
asset mix. (From J. Bicksler and A. Chen, "The
Integration of Insurance and Taxes in Corporate
Pension Strategy," Journal of Finance, July 1985,
951, reprinted with permission.)
are two reasons, and they are illustrated in Fig. 18.4. First, there are deadweight
losses associated with involuntary pension plan termination due to underfunding.
These costs have the effect of reducing the value of pension fund insurance provided
by the PBGC from point X to X'. Examples of such costs are legal expenses and
higher employee wage demands to offset the probability that their defined pension
benefits may not be paid. On the tax side, Bicksler and Chen suggest that the present
value of the tax shield diminishes because of the asymmetric structure of the U.S.
tax code. Firms pay taxes when their income is positive but cannot count on using
all tax credits from carryforwards and carrybacks if income is negative. As illustrated
in Fig. 18.4, the joint effects of termination deadweight losses and a diminishing tax
shelter result in an interior optimum. If the tax effect dominates, the level of pension
funding will be higher and more of the fund assets will be in bonds than stocks, and
vice versa.
7
For a more complete exposition, the reader is referred to Bulow, Scholes, and Manell [1982].
656 APPLIED ISSUES IN CORPORATE FINANCE
riskless return. Market rates of interest of 14% on long-term bonds were not unusual,
but those companies whose pension assets were committed to insurance company
contracts found they were locked into an 8% return. This is the hard way to learn
about opportunity cost (although it is still a fair game).
If your company is large enough to provide its own pension fund accounting for
employees, then there is no difference between contributing pension funds to an in-
surance company plan and directly investing in 8% 10-year bonds yourself. Just bear
in mind that long-term bonds are riskier than short-term bonds or money market
assets. Some companies have decided to immunize their pension liabilities by pur-
chasing long-term bonds that mature with the same pattern as employee retirements.
They know for sure that maturing bonds will pay promised benefits.
It has been estimated that a total of at least $150 billion in interest rate swaps had
been completed in the United States by the end of 1985. A fairly recent phenomenon,
interest rate swaps are a rapidly growing activity.
An interest rate swap is a contract between firms in which interest payments are
based on a notational principal amount that is itself never paid or received. Instead
the parties agree to pay each other the interest that would be due on the notational
principal. Swaps are usually between fixed and floating rate instruments, although
floating for floating and fixed for fixed are also possible. The most common swap is
where one interest stream, the floating payment, is tied to a short-term money market
rate such as the U.S. Treasury bill rate or to LIBOR (the London Interbank Offer
Rate). The other payment stream is fixed for the life of the swap. Both fixed and floating
interest payments start accruing on the swap's effective date and cease on the swap's
maturity date.
INTEREST RATE SWAPS 657
Figure 18.5
An interest rate swap.
Figure 18.5 illustrates a swap. Suppose that an Aaa-rated firm can borrow five-
year fixed rate debt at 10% and floating rate debt at the T-bill rate, 7%, plus 1%. At
the same time, a Baa-rated firm can borrow five-year fixed rate debt at 11.5%, and
floating rate debt at the T-bill rate plus 4%. Thus the short-term quality premium
is only 4%, whereas the long-term premium is 1.5%.
Suppose the Aaa-rated firm has borrowed $100 million of five-year fixed rate
debt at 10%. A swap can be arranged whereby the Baa-rated firm agrees to pay
104% on the five-year debt and the Aaa-rated firm pays the T-bill rate plus 1%
(i.e., 7r.). The net position of the Aaa-rated firm is a gain of 1%, the extra 4% re-
ceived on the five-year note less the extra 4% paid on the variable rate loan. The
Baa-rated firm borrows $100 million at the T-bill rate plus receives the same rate
from the Aaa-rated firm, and agrees to pay 104% on the fixed rate debt. Since the
rate on the fixed debt is 4% less than the Baa-rated firm would otherwise have to pay,
it comes out 4% ahead as a result of the swap. In this example the presumption is
that both firms benefit by splitting the difference in the quality spread on short-term
variable rate debt and longer-term fixed rate debt.
In the absence of market imperfections and comparative advantages among
different classes of borrowers, there would be no reason for interest rate s waps.
However, in less than perfect markets there are a number of possible motivations
for engaging in a swap. Henderson and Price [1984], Bicksler and Chen [1985],
and Smith, Smithson, and Wakeman [1986] discuss the more frequently mentioned
reasons.
I. Duration matching. Firms with variable rate assets and fixed rate liabilities (or vice
versa) are exposed to interest rate risk. If the rates that they receive on assets fall,
but they continue to pay high fixed rates on their liabilities, they end up in a
losing position. A swap of fixed rate for variable rate debt can help to match the
duration of their assets and liabilities and to reduce their interest rate risk.
2. Quality spread arbitrage. When the quality spread between short- and long-term
debt gets far enough out of line, it may be possible to engage in the quality spread
arbitrage, as was illustrated in Fig. 18.5.18
18
One of the reasons why the quality yield spread on short-term notes is less than the spread on long-
term notes is that the probability of default on a low-quality bond is less in the short run than in the long
run. Therefore parties who agree to lock into a long-term swap may be fooling themselves into thinking
they are arbitraging the quality spread when in fact the spread on a one-year position rolled over N times
should be the same as that on an N-year position.
658 APPLIED ISSUES IN CORPORATE FINANCE
fr yen
Dual currency îond yen I 4;$,
yen =, yen
plus
yen
Zero coupon yen îond yen 4;
plus
z'\.$ I yen yen yen yen
Fixed currency swap
\7 yen $ $
plus
Spot currency transaction < <
yen
equals I
Figure =8.6
Creating a synthetic deep discount dollar—denominated îond. (From
C. Smith, Jr., C. Smithson, and L. Wakeman, "The Evolving Market for
Swaps," Midland Co]po]ate Finance Jou]nal, Winter =986, <P, reprinted
with permission.)
3. Refunding debt. When debt is noncallable, or is privately held, swaps may be used
to refund the debt and to simultaneously convert fixed rates to variable.
4. Tax and regulatory arbitrage. The introduction of a swap allows an "unbundling"
of currency and interest rate exposure from regulation and tax rules in very cre-
ative ways. For example, until recently, zero coupon bonds in Japan were not
taxed until maturity and then only at the capital gains rate. Furthermore, the
Japanese Ministry of Finance limited the amount a pension fund could invest
in non-yen-denominated foreign issues to 10% of the fund's portfolio. U.S. firms
issued zero coupon yen bonds plus dual currency bonds (with interest payments
in yen and principal in dollars) and were able to capitalize on the superior tax
treatment of zero coupon bonds in Japan as well as the Japanese funds' desire
to diversify their funds internationally because the ministry qualified dual cur-
rency bonds as yen-denominated. To transfer their yen exposure back to a U.S.
dollar exposure, the U.S. firms used currency swaps in conjunction with spot
currency transactions. See Fig. 18.6 for an illustration of the cash flow pattern.
The economic evaluation of swap decisions requires that we compare the present
values of fixed and variable rate instruments not an easy task because of differences
in default risk and the difficulty in modeling the term structure of interest rates. Cox,
Ingersoll, and Ross [1980], and Ramaswamy and Sundaresan [1986] have shown how
to value variable rate debt given various assumptions. The following example assumes
no default risk and a monotonic term structure of interest rates.
The first task, and by no means the easiest, is to model the term structure of interest
rates. To keep things simple, assume a three-period world where one-period risk-free
INTEREST RATE SWAPS 659
17.2'8%
12.96%
11.11%
9.72%
7.29%
Figure 18.7
A binomial model of one-period riskless interest rates for three
time periods.
interest rates are modeled as binomial trials." The interest rate can move up by a
factor of 1.2 or down by a factor of .9. Figure 18.7 shows the term structure assuming
that last period's rate was 11.11% and that this period's rate is 10%. Since the up
movements are larger than the down movements, the term structure is upward slop-
ing in this example. Once one has an adequate model of the term structure, the next
step is to model the actual payouts on both instruments. If the floating rate bond
always paid the current one-period rate, it would adjust perfectly to changes in the
interest rate and would always sell for par. Unfortunately, this is not usually the case.
Suppose the floating rate bond promises to pay a coupon based on the arithmetic
average of the current one-period rate and last period's rate. Coupons based on mov-
ing average schemes of this type are common. Figure 18.8 illustrates the coupons that
would be paid, contingent on the path of one-period riskless rates. Since there is no
default risk involved, we can discount the coupon payments at the one-period riskless
rate. The iterative pricing formula is
q(coupon t,„) + (1 — q)(coupont, a)
Bt,, = 1
(18.6)
+ r f,t,s
where t is the time period; s is the state of nature (up or down); coupon,,, is the coupon
in period t and state s; q is the probability of an upward movement; and is the
riskless rate in period t and state s. To illustrate the use of Eq. (18.6) take the second
highest payoff (state 2) in the third time period (see Fig. 18.8). It was reached in one
of three ways: after two upward movements followed by one downward movement
in the interest rate, or after one downward and two upward movements, or via an up,
a down, and then an upward movement. With the first path, the coupon payment is
$136.80 and the face value is $1000, resulting in a total payment of $1136.80. Via the
second and third paths, the coupon is $118.80 and the face value is still $1000. Thus
the coupon payment is path dependent. To compute the value of the bond in state
19
See Chapter 8 on option pricing for a complete exposition of the binomial model.
660 APPLIED ISSUES IN CORPORATE FINANCE
3,u3 = 1158.40
14.4%
Note: 129.6 + 144
B2 0,2 = 1003.15 B3,uud = 1000
2
12% 14.4% 129.6 + 108
B3,udu + 1000
2
.81, u = 1005.17
B3 ,u2d = 1136.80
10% or 1118.80
B3 udz = 1089.10
Bid = 1004.04 or 1102.60
9%
B2, dd = 1001.87 8.1%
B3,d3 = 1076.95
Figure 18.8
Payments on a floating rate bond where the coupon is an average of rates for the
last two periods.
1 of period 2, one averages the payout in state 1 of period 3, $1158.40, with the payout
in state 2 of period 3 which was reached via 2 upward and 1 downward movements,
$1136.80; and discounts at 14.4%. The result is B2 ,u2 = $1003.15. The present value
of the bond in state 1 of time period 2 is the average of the total payoffs (since they)
are equally likely) discounted at the one-period rate, 14.4%. Given the numbers in our
example, the present value of the floating rate bond (at time 0) is $1006.46.
Next, suppose that you are the treasurer of an Aaa-rated company that has $100
million of floating rate debt with exactly the same terms as the instrument we just
valued. You are approached by a company that has the same default risk (assume no
default risk at all for an Aaa-rated company). They want to swap their three-year
fixed rate, which pays a 12% coupon, with your floating rate debt. What should you
do? The solution of course is to analyze the fixed rate debt and compare its value with
the value of your floating rate debt. If your debt is worth less than the fixed rate debt,
you would take the offer.
Figure 18.9 shows how to use Eq. (18.6) to value the fixed rate debt. The procedure
is much the same as before. The value of the bond cum coupon is discounted each
period at the risk-free rate in the appropriate state of nature. The present value of the
12% fixed rate bond turns out to be $1038.41. Since it is worth more than your floating
rate bond, you would be willing to undertake the proposed swap.
Interest rate and foreign exchange swaps are a rapidly growing business. There
are many reasons. One is that there are capital market inefficiencies that allow arbi-
trage using swaps. But another is that the parties involved in the swaps are being
fooled because they do not understand how to price the complex instruments involved
in the deals especially floating rate notes with default risk and with complex terms.
LEVERAGED BUYOUTS AND GOING PRIVATE 661
B3, u3 = 1120
14.4%
B2,0 = 1099.02
12% 14.4%
BL it = 1115.47
B3, 0d —1120
10%
10.8%
B0 = 1038.41 2,ud = 1130.83
10.8%
3, ud 2 = 1120
131,d = 1169.04
B3, d3 = 1120
Figure 18.9
The valuation of three-year fixed rate debt paying a 12% coupon.
There is little empirical research yet to help answer the question of whether market
inefficiencies are driving the growth in swaps. Partly this is so because of the lack of
good data and partly because of the difficulty of theoretical models to predict yield
premia. More research in both areas is needed.
Leveraged buyouts involve management purchase of the entire public stock interest
of a firm, or division of a firm. If the shares are owned exclusively by management the
transaction is called going private, and there is no market for trading its shares. If
ownership in the subsequent private firm is shared with third-party investors and
financed heavily with debt, the transaction is called a leveraged buyout. The issues
raised by leveraged buyouts are many. Why do they happen in the first place? Are they
motivated as an attempt by incumbent management to expropriate wealth from
minority shareholders a minority freezeout? Are they done for tax reasons? Who
benefits? Who loses and why? How are the deals structured?
1. How to Go Private
There are four commonly used techniques for implementing the going-private
transaction. Management may form a shell corporation that combines with the firm
via merger. Usually merger approval is required by shareholders of the original firm,
and the shell corporation may pay with cash or securities. Asset sales are similar in
that a vote is required, and assets are purchased by a shell corporation owned by
662 APPLIED ISSUES IN CORPORATE FINANCE
management. A tender offer does not require a vote and does not require minority
shareholders to surrender their shares involuntarily. In a tender offer the firm buys
back its own shares, either with cash, debt, or convertible securities. Least common
among the methods is a reverse stock split. Holders of fractional shares are usually
required to sell their ownership back to the corporation. DeAngelo, DeAngelo, and
Rice [1984] found that of a sample of 81 going-private proposals between 1973 and
1980, 27 were mergers, 3 were sale of assets, 16 were tenders (or exchange offers),
and 1 was a reverse split. The remaining 34 were either leveraged buyouts (LBOs)
with third party participation (28) or unclassified acquisitions (6).
The distinction between a pure going-private transaction and a leveraged buyout
with the involvement of third parties is important for leverage changes. DeAngelo,
DeAngelo, and Rice [1984] report that for those firms where the proxy statement had
a forecast of leverage changes, the leveraged buyout book ratio of debt to total assets
increased from 11% to 86%, but for the pure going-private transactions it changed very
little-26% versus 30%.
Figure 18.10
Senior debt Strip financing in an LBO.
-\-CT-'- l
Subordinate debt
Mezzanine
level
financing
"`Convertible debt/'
rI
Preferred stock Ade
. A
Common stock
strip financing in an LBO. The senior claim is usually bank debt financing. Sub-
ordinate to it are the "mezzanine" securities including subordinate debt, convertible
debt, and preferred stock. Third-party financiers typically hold strips of the mezzanine
as well as equity. Venture capitalists may hold up to 80% of the equity, with man-
agement holding the remainder. Because venture capitalists are generally the largest
shareholder and control the board of directors, they have both greater ability and
stronger incentives to monitor managers than directors representing diffuse public
shareholders in the typical public corporation.
LBO targets are frequently firms with relatively stable cash flows and unutilized
debt capacity. It may be possible for management to benefit from the greater tax
shield provided when the buyout is financed with debt. Thus the value gained from
leverage may be an additional motivation for leveraged buyouts.
An example of how transactions might work out for a target company is shown
in Table 18.7. Table 18.7 shows how the earnings before interest and taxes grow at
10% a year. The interest payments reflect the amortization of principal that takes
place each year. A 40% tax rate is assumed. After deduction of taxes, net income is
shown in row 5. Depreciation is added back to obtain the usual definition of cash
flow shown in row 7. Row 8 illustrates an amortization schedule for the debt. This
results in the cash flow cushion depicted in row 9. With the amortization schedule
shown in row 8, we can also indicate how the debt/equity position changes over time.
This is illustrated by Table 18.8.
Table 18.8 reflects the amortization program agreed upon. Over the five-year
period, debt is reduced from 90% to 0%. We then assume that the company is sold
at its book value at the end of year 5. This is a conservative assumption because,
with the record it has established, the firm might well sell for a premium over book
value. The ratio of the price received to the initial equity investment is 476.5/50,
which equals 953%. This represents a five-year annual compounded rate of return
on the initial $50,000 investment of 57%. The plausibility of these results is indicated
by some published statistics. A Fortune magazine article [Ross, 1984] stated that
one of the leveraged buyout specialist companies, Kohlberg, Kravis, Roberts and
Co., has earned an average annualized return of 62% on the equity it has invested
in its transactions. Another buyout specialist, Carl Ferenbach, has stated that his
firm expects an annual return of 50% on its equity investment [DeAngelo, DeAngelo,
and Rice, 1984].
Much has been made of the conflict of interest that managers, as insiders, have
when engaging in a buyout. If they obtain the best deal for themselves, it may be at
the expense of minority shareholders. One would expect managers to try a buyout
when they have inside information that indicates that the future of the firm is better
than previously expected. Rational expectations theory, however, suggests that share-
holders know that insiders have an informational advantage and are not fooled.
Therefore they demand a higher price in order to sell their ownership claims. Fur-
thermore, minority shareholders must approve the transaction and often have veto
power. A countervailing force is that management frequently has a majority owner-
ship position prior to the transaction. For example, DeAngelo, DeAngelo, and Rice
[1984] report that the median management ownership was 51% of the stock in pure
going-private transactions and 33% in LBOs with third-party participation. This
makes it difficult for raiders to compete to bid away potential management gains.
Therefore we might predict that gains from going private are shared between manage-
ment and minority shareholders.
DeAngelo, DeAngelo, and Rice (DDR) report a highly significant average 2-day
abnormal return of 22.4% upon the initial announcement of a going-private or LBO
transaction. If the announcement period is extended to include information leakage
during the 40 trading days prior to the announcement, the cumulative abnormal
return increases to 30.4%. DDR also report that the initial market reaction is well
below the 56.3% average premium offered by management. The difference is explained
by the relatively high percentage of offers that are withdrawn (18 firms). The 2-day
abnormal return at the time of the withdrawal announcement was a significantly
negative 8.97.. These facts clearly indicate that minority shareholders received signifi-
cant gains from the transaction. They do not shed any light on what percentage of
the total gain was received by management and by minority shareholders. For those
buyouts that were consummated, both parties certainly believed they were gaining.
There are many unanswered questions regarding going-private and LBO trans-
actions. For example, how does one evaluate the transaction on a net present value
basis before actually taking action? What determines the best structure for the deal?
And how does one estimate the required rates of return on the securities involved?
Future research is needed to help answer these thorny problems.
wealth. For example, it is conceivable that a firm can underperform relative to its
competition but still experience an increase in share price simply because the
market went up.
6. The tax efficiency of plans should be compared. If two plans are alike in most
regards, but one is better designed to minimize the tax liabilities of the firm and
its management, then its tax efficiency may become the decisive factor.
The popular press has often criticized management as overpaid. Owners worry
that managers do not bear the full cost of shirking, or of excessive perquisites, and
may choose to be more conservative in risk taking than owners would desire. Execu-
tive compensation schemes are a deliberate attempt to modify managerial behavior
to more closely conform to shareholder objectives.
Empirical research into the relationship between the level of executive compensa-
tion and changes in shareholder wealth is just beginning Murphy [1985] reports that
a 10% change in the equity value of the firm is associated with only about a 2%
increase in total executive compensation. Although the correlation is weak, it is statis-
tically significant and positive. Coughlan and Schmidt [1985] also find a significant
positive relationship between executive compensation and changes in shareholder
wealth. These studies imply that executive compensation plans in the United States
are not complete nonsense, but one can hardly infer that they are optimally designed.
1. Bonus Plans
Most executive compensation plans have three components: salary, stock options,
and bonuses. The rough proportions for 1966, 1971, and 1981 are given in Brindisi
[1985] (see Table 18.9). The proportion represented by option plans has been rising
at the expense of bonus plans. Table 18.10 shows the measures that are typically used
for bonuses. It is based on a sample of 80 firms reported by Brindisi [1985]. Most
bonus plans are based on accounting data. Earnings per share, or the growth in
earnings per share, is the most common measure. Earnings per share has been criti-
cized because it is a short-term goal that can be easily manipulated and because it
sometimes moves in the opposite direction from cash flows.
Table 18.9 Proportion of Executive Table 18.10 Performance Measures Used for
Compensation from Various Sources Chief Executive Officer Bonus Plans
Source 1966 1971 1981 Growth in income or earnings per share 85%
Return on investment 33
Salary 43% 58% 53%
Individual objectives 17
Bonuses 22 19 6
Sales growth 3
Options 35 23 41
Discretionary 27
From L. Brindisi, Jr., "Creating Share-
holder Value: A New Mission for Execu- From L. Brindisi, Jr., "Creating Shareholder Value: A
tive Compensation," Midland Corporate New Mission for Executive Compensation," Midland
Finance Journal, Winter 1985, 62, re- Corporate Finance Journal, Winter 1985, 61, reprinted
printed with permission. with permission.
EXECUTIVE COMPENSATION PLANS 667
Healy [1985] studied the relationship between year-end accounting changes and
executive compensation plans. A typical funding formula for the yearly bonus pool
establishes both a minimum threshold and a maximum payout. For example, no
bonuses are paid unless net income exceeds 57. of total capital employed; and the
maximum bonus cannot exceed some target percentage of executive salaries. Healy
found that when firms had bonus plans, there was a tendency to defer the recognition
of revenues and accelerate expenses when profits fell near the bonus threshold level.
Although this behavior does not result in a bonus this year, it does increase the prob-
ability of receiving a bonus next year. When profits fell between the minimum and
the ceiling, there was a tendency to accelerate the recognition of revenues and defer
expenses. It is hard to believe that short-term behavior of this sort is intended to maxi-
mize shareholders' wealth.
Another problem with bonus plans based on earnings or earnings growth is that
they can distort investment decisions. Many positive net present value projects lose
money during their gestation phase. Managers who myopically pursue short-term
profit objectives may decide to underinvest to milk the firm for short-term profits.
In an effort to modify bonus plans to provide incentives to adopt longer horizons,
many firms have developed bonus plans based on a moving average of three to seven
years of profits. This is an improvement, but it does not solve the problem that it is
still based on accounting measures of income that may or may not have a direct cor-
respondence with shareholder wealth creation. We shall also see, later on, that bonus
plans are typically inferior from a tax point of view.
Another important dimension to executive compensation is risk. If all compensa-
tion were salary, unrelated to performance, then executives (as a group) would have
more or less the same risk as bondholders in a levered firm. They could not do much
to increase their income if the firm did well, and they would lose their salary only
if the firm went bankrupt.2° Consequently, many scholars of executive compensation
believe that in order to encourage managers to take more risk, there must be option-
like features to their compensation. As mentioned earlier, most bonus plans have this
feature, especially if no bonus is earned unless a minimum threshold is exceeded and
if there is no maximum. Executive stock options and stock appreciation rights (SARs)
have exactly this type of payout.
2°
Fama [1980] presents an alternative point of view, arguing that the labor market for managers makes
their salaries vary wi.,h their productivity, and that they do an excellent job of monitoring each other.
Furthermore, salaries and bonuses can be adjusted ex post to eliminate the agency problem.
668 APPLIED ISSUES IN CORPORATE FINANCE
since 1981 the tax laws govern two types of plan: incentive stock options (ISOs) and
stock appreciation rights (SARs).
Neither ISOs nor SARs have any tax consequences for either the firm or the
executive at the time the options are issued. At the time of exercise, the ISO has no
tax consequences; however, the SAR allows the firm to deduct the difference between
the share price on the exercise date, Se , and the exercise price, X, whereas the execu-
tive must pay ordinary income taxes on Se — X. Finally, when the shares purchased
by exercising the option are eventually sold, the executive can qualify for capital
gains tax treatment. For ISOs the tax is on the difference between the sale price, S„
and the exercise price, X; and for SARs it is on the difference between the sale price
and the purchase price, St — Se .
ISOs must be exercised sequentially, whereas SARs can be exercised in any or-
der. There is a definite drawback to the sequential exercise requirement. Suppose the
firm's share price was $50, fell to $30, then rose again to $40. If options were issued
at each stage with the exercise price equal to the stock price, executives would be
required to exercise the earlier $50 options (that are $10 out-of-the-money) before
they could exercise the options issued at $30 (that are $10 in-the-money).
Both ISOs and SARs have a maximum life of 10 years from the date of issue.
The exercise price of ISOs must be greater than or equal to the stock price at the
time of issue, but SARs can have an exercise price as low as 50% of the stock price.
There is an increasing tendency to complement executive stock option plans with
SARs. In 1970 none of the plans for the largest 100 companies in the United States
had SARs, but according to Smith and Watts [1982], by 1980, 68 out of the top 100
firms had SARs.
Two reasons for using executive stock options at all are (1) that they augment
salaries with a call option so that managers' total compensation pattern is more like
that of shareholders and (2) that stock options are a more tax efficient form of com-
pensation than straight salaries.
Miller and Scholes [1982] demonstrate that when compared with salaries SARs
are tax neutral from the firm's point of view and tax dominant from the manager's
point of view. To illustrate their argument, suppose that a firm expects $140 million
of cash flows in a good state of nature before paying management salaries, and $120
million in a bad state. The firm's tax rate is 507, and management salaries amount
to $100 million. Table 18.11 shows their net payoffs after taxes and management
salaries.
Assets $15 Debt $12.5 Assets $15 Debt $15 Assets $10 Debt $10
Equity 2.5 Equity 0 Equity 0
Total $15 Total $15.0 Total $15 Total $15 Total $10 Total $10
If the probability of the good state is .5 and if we assume the discount rate is
zero, then the present value of the firm in this one-period example is $15 million.
Assume that there are 10 million shares outstanding so that the price per share is
$1.50. An at-the-money call option will sell for $0.25, the average of its good and
bad state payouts, as shown in Table 18.12.
Now, suppose the company offers managers an SAR plan that is tax neutral from
the company's point of view. What will it look like? The plan will have a present
value of $2.5 million and will involve the issue of 10 million at-the-money options
to management. In return, management agrees to reduce its salary by $2.5 million
an equal dollar amount.
It is assumed that the firm established a subsidiary to handle the potential li-
ability created by the SAR plan. The subsidiary will have $2 5 million in equity (the
salary expense reduction) and will borrow $12.5 million via a zero coupon bond with
a face value of $15 million. The $15 million in cash is used to purchase 10 million
shares of stock. Table 18.13 shows the current balance sheet of the subsidiary as well
as its balance sheets given the good and bad states of nature. If the good state occurs,
the options will be exercised whereupon the subsidiary receives $15 million in cash
and delivers stock worth $20 million. The $15 million in cash is used to pay off the
loan, and the equity in the subsidiary is worthless, leaving the parent firm with a
$2.5 million tax-deductible loss. If the bad state occurs, the options will not be ex-
ercised, and the stock held by the subsidiary will be worth $10 million and turned
over to the bank. The equity in the subsidiary will be worthless, once again leaving
the parent with a $2.5 million tax-deductible loss. The bank lends $12.5 million, and
its expected payoff is $12.5 million. The parent firm ends up with a $2.5 million loss
either way.
Table 18.14 shows the firm's expected cash flows given the terms of the option
plan, and its payoffs from the subsidiary. Comparing the net cash flow of Tables 18.14
and 18.11 we see that the firm is completely indifferent between the two alternatives.'
21 Note that we have not argued that the firm will use the tax shelter created if it can write off the difference
between the stock price and the exercise price in the good state of nature.
670 APPLIED ISSUES IN CORPORATE FINANCE
value of the firm as it is at a given point of time, then separate changes in the valuation
parameters caused by management from those caused by external factors, it would
be possible to measure the marginal product of management and reward management
appropriately. Although there have been some attempts to accomplish this task, we
still have no accurate guidelines. 22
SUMMARY
This chapter has used the knowledge base provided earlier in the book to examine
a number of topical issues. It was necessary to understand discounting, the risk-return
trade-off, option pricing, and capital structure before tackling the practical issues
covered here—pension fund management, interest rate swaps, going-private and
leveraged buyout transactions, and executive compensation schemes. None of these
topics has been completely researched, and therefore, all leave us with unanswered
questions. However, they are all timely and of current interest to practitioners.
PROBLEM SET
18.1 Table 18.2 shows a pension fund income statement and balance sheet. Where in the
balance sheet would you place the following items:
a) $500 million in U.S. government bonds that are held in trust for the members of the pen-
sion plan?
b) An amount of $200 per month that the firm plans to put aside to pay the pension of an
existing employee?
c) A $15,000 per year pension that you expect to pay to an employee who is currently not
vested?
d) A $10,000 per year pension that the plan is currently paying to a retired employee?
18.2 Suppose that your firm is Aaa-rated and can borrow five-year fixed rate debt at 12%
and floating rate debt at the T-bill rate plus i.e., at 9+%. You are approached by a Baa-
rated firm that can borrow five-year fixed rate debt at 13% and floating rate debt at the
T-bilI rate plus 1%. The Baa firm wants to swap $100 million of its floating rate debt for an
equivalent amount of your fixed rate debt. It is willing to pay 121% to you on the fixed rate
debt and asks that you pay the T-bill rate plus 1% on its floating rate debt. Should you do
the deal?
18.3 Assume a binomial process for changes in the risk-free rate, with annual movements of
u = 1.3 and d = .7. This period's rate is 8% and has just moved down from 11.4286%. What
is the market value of a three-year variable rate loan where the variable rate is set as the arith-
metic average of this period's rate and last period's rate? How would your answer change if the
interest rate had just moved up from 6.1538% instead of down?
22
For example, see Rappaport [1986, Chapter 8].
REFERENCES 673
18.4 You are trying to structure a leveraged buyout (LBO) of an all-equity firm estimated to
be worth $100 million Its equity beta is .8 and the CAPM market risk premium is 6.2%. The
deal is expected to be financed with 80% debt and will require a 50% premium above market.
In other words, the firm will cost $150 million and you expect to borrow $120 million The firm's
tax rate is 413%. The first $30 million is a five-year zero coupon loan designed to yield 10% to
maturity. The next $90 million is zero coupon subordinated debt. If the standard deviation of
return on the firm's assets is 34% per year, the five-year risk-free rate is 9%, and you expect zero
dividend payout, what yield to maturity will be required on the subordinated debt if its face value
is $267.065 million? What will the cost of equity be?
REFERENCES
American Institute of Certified Public Accountants, Accounting Trends and Techniques, N.Y.:
AICPA, 1986.
Arnold, T., "How to Do Interest Rate Swaps," Harvard Business Review, September—October
1984, 96-101.
Bagehot, W. (pseud.), "Risk and Reward in Corporate Pension Funds," Financial Analysts
Journal, January—February 1972, 80-84.
Bank Administration Institute, Measuring the Investment Performance of Pension Plans. BAI,
Park Ridge, Ill., 1968.
Bicksler, J., and A. Chen, "The Integration of Insurance and Taxes in Corporate Pension
Strategy," Journal of Finance, July 1985, 943-955.
, "An Economic Analysis of Interest Rate Swaps," Journal of Finance, July 1986, 645-655.
Black, F. "The Tax Consequences of Long-run Pension Policy," Financial Analysts Journal,
July—August 1980, 21-28.
Brickley, J.; S. Bhagat; and R. Lease, "The Impact of Long-range Managerial Compensation
Plans on Shareh\olders' Wealth," Journal of Accounting and Economics, April 1985, 115-129.
Brindisi, L., Jr., "Creating Shareholder Value: A New Mission for Executive Compensation,"
Midland Corporate Finance Journal, Winter 1985, 56-66.
Bulow, J.; M. Scholes; and P. Manell, "Economic Implications of ERISA," Working paper,
Graduate School of Business, University of Chicago, March 1982.
Copeland, T., "An Economic Approach to Pension Fund Management," Midland Corporate
Finance Journal, Spring 1984, 26-39.
Cornell, B., "Pricing Interest Rate Swaps: Theory and Empirical Evidence," Working paper,
Anderson Graduate School of Management, UCLA, April 1986.
Coughlan, A., and R. Schmidt, "Executive Compensation, Management Takeover, and Firm
Performance: An Empirical Investigation," Journal of Accounting and Economics, April
1985, 43-66.
Cox, J.; J. Ingersoll; and S. Ross, "An Analysis of Variable Rate Loan Contracts," Journal of
Finance, May 1980, 389-403.
Davidson, S.; C. Stickney; and R. Weil, Intermediate Accounting. Dryden Press, Hinsdale, Ill.,
1980.
DeAngelo, H., and L. DeAngelo, "Management Buyouts of Publicly Traded Corporations,"
Financial Analysts Journal, May—June 1987, 38-48.
DeAngelo, H., L. De Angelo, and E. Rice, "Going Private: Minority Freezeouts and Stockholder
Wealth," Journal of Law and Economics, October 1984, 367-401.
674 APPLIED ISSUES IN CORPORATE FINANCE
Narayanan, M., "Managerial Incentives for Short-term Results," Jou]nal of Finance, Decemîer
=98P, =469-=484.
Pesando, J., "The Usefulness of the Wind-up Measure of Pension Liaîilities: A Laîor Market
Perspective," Jou]nal of Finance, July =98P, 9<7-940.
Ramaswamy, K., and S. Sundaresan, "The Valuation of Floating Rate Instruments," Jou]nal
of Financial Economics, Decemîer =986, <P=-<7<
Rappaport, A., C]eating Sha]eholde] Value. Free Press, Macmillan Inc., New York, =986.
Regan, P., "Reasons for the Improving Pension Fund Figures," Financial Analysts Jou]nal,
March—April =98<, =4-=P.
Rich, J., and E. Bergsma, "Pay Executives to Create Wealth," Chief Executive, Autumn =98<.
Ross, I., "How the Champs Do Leveraged Buyouts," Fo]tune, January <3, =984, 70-7<, 74, 78.
Sharpe, W., "Corporate Pension Funding Policy," Jou]nal of Financial Economics, June =976,
=83-=94.
Smith, C., Jr., C. Smithson; and L. Wakeman, "The Evolving Market for Swaps," Midland
Co]po]ate Finance Review, Spring =986, =6-3=.
Smith, C., Jr., and R. Watts, "Incentive and Tax Effects of Executive Compensation Plans,"
Aust]alian Jou]nal of Management, =98<, =39-=P7.
Stowe, J., and M. Walker, "The Effect of Executive Stock Options on Corporate Financial
Decisions," Jou]nal of Financial Resea]ch, Spring =980, 69-83.
Tehranian, H., and J. Waegelein, "Market Reaction to Short-term Executive Compensation
Plan Adoption," Jou]nal of Accounting and Economics, April =98P, =3=-=44.
Tepper, I., "Taxation and Corporate Pension Policy," Jou]nal of Finance, March =98=, =-=3.
, "The Future of Private Pension Funding," Financial Analysts Jou]nal, =98<, <P-3=.
, and A. Affleck, "Pension Plan Liaîilities and Corporate Financial Strategies," Jou]nal
of Finance, Decemîer =974, =P49-=P64.
Treynor, J., "The Principles of Corporate Pension Finance," Jou]nal of Finance, May =977,
6<7-638.
; W. Priest; and Regan, P., The Financial Reality of Funding unde] ERISA, Dow Jones—
Irwin Inc., Homewood, Ill., =976.
Warner, J., "Stock Market Reaction to Management Incentive Plan Adoption: An Overview,"
Jou]nal of Accounting and Economics, April =98P, =4P-=49.
Westerfield, R., and W. Marshall, "Pension Funding Decisions and Corporate Shareholder
Value: A New Model and Some Empirical Results," mimeo, Septemîer =983.
19
Corporate takeovers offer a field day for speculators and traders in our
"Casino Society"; they strip away equity, build up shaky debt, and
destabilize markets.
F. G. Rohatyn, "Needed: Restraints on the Takeover Mania,"
Challenge, May—June 1986, 30.
Mergers, Restructuring,
and Corporate Control:
T heory
A. INTRODUCTION
Takeovers and related activities in the 1980s are much broader in scope and raise
more fundamental issues than previous merger movements. The daily newspapers
are filled with a series of case studies of mergers and acquisitions (M&A's), tender
offers (both friendly and hostile), spinoffs and divestitures, corporate restructuring,
changes in ownership structures, and struggles for corporate control. In recent years,
leverage ratios for some companies have increased, and newer forms of financing
have proliferated, including an increase in the use of bonds with ratings below the
first two levels of A and B (usually a C rating, referred to as junk bonds). Thus the
traditional subject of M&A's has been expanded to include takeovers and related
issues of corporate restructuring, corporate control, and changes in the ownership
structure of firms, as suggested by the changed title of this chapter. An overview of the
nature of the various activities will first be presented, then some will be analyzed
in greater depth.
676
CORPORATE RESTRUCTURING AND CONTROL 677
B. CORPORATE RESTRUCTURING
AND CONTROL
Table 19.1 presents a listing of the many forms of corporate activities that have been
filling the pages of both the popular and academic literature in recent years. The nature
of each of the forms of activity will be briefly summarized to convey what is involved.
1. Expansion
Under expansion, we include mergers, tender offers, and joint ventures. Mergers
and acquisitions have long played an important role in the growth of firms. Growth is
generally viewed as vital to the well-being of a firm. Among other reasons, growth is
needed for a firm to compete for the best managerial talent by offering rapid promo-
tions and broadened responsibilities. Without a continued inflow of able executives,
the firm is likely to decline in efficiency and value.
From a legal standpoint there are many distinctions between types of mergers
and combinations. Most generally, merger means any transaction that forms one
economic unit from two or more previous ones.
II. Sell-offs
Spinoffs
Splitoffs
Splitups
Divestitures
Equity carve-outs
III. Corporate control
Premium buybacks
Standstill agreements
Antitakeover amendments
Proxy contests
2. Selloffs
Several distinct types of selloffs should be distinguished. The two major types are
(1) spinoffs and (2) divestitures. A spinoff creates a separate new legal entity; its shares
are distributed on a pro rata basis to existing shareholders of the parent company.
Thus existing stockholders have the same proportion of ownership in the new entity
as in the original firm. There is, however, a separation of control, and over time, the
new entity as a separate decision-making unit may develop policies and strategies dif-
ferent from those of the original parent. Note that no cash is received by the original
CORPORATE RESTRUCTURING AND CONTROL 679
3. Corporate Control
The third group of activities in Table 19.1 we have referred to as corporate control.
Premium buybacks represent the repurchase of a substantial stockholder's ownership
interest at a premium above the market price (called greenmail). Often in connection
with such buybacks, a standstill agreement is written. These represent voluntary con-
tracts in which the stockholder who is bought out agrees not to make further invest-
ments in the company in the future. When a standstill agreement is made without a
buyback, the substantial stockholder simply agrees not to increase his or her owner-
ship, which presumably would put that individual in an effective control position.
Antitakeover amendments are changes in the corporate bylaws to make an ac-
quisition of the company more difficult or more expensive. These include (1) super-
majority voting provisions, requiring a percentage (e.g., 80%) of stockholders to
approve a merger; (2) staggered terms for directors, which can delay change of control
for a number of years; (3) golden parachutes, which award large termination pay-
ments to existing management if control of the firm is changed and management is
terminated; and (4) "poison pill" provisions, which give present stockholders the right
to buy at a substantial discount the shares of a successor company formed by a stock
takeover.
In a proxy contest an outside group seeks to obtain representation on the firm's
board of directors. The outsiders are referred to as "dissidents" or "insurgents," who
seek to reduce the control position of the "incumbents" or existing board of directors.
Since the management of a firm often has effective control of the board of directors,
proxy contests are often regarded as directed against the existing management.
680 MERGERS, RESTRUCTURING, AND CORPORATE CONTROL: THEORY
C. RECENT DEVELOPMENTS IN
M&A ACTIVITY
Merger activity has been characterized by dramatic bidding wars involving tender
offers and mergers of very substantial size. Notable was du Pont's November 1981
acquisition of Conoco for over $7.5 billion, in which du Pont outbid Seagram and
Mobil. In November 1982 U.S. Steel outbid Mobil to obtain Marathon Oil for about
$6.5 billion. Notable deals in 1984, some of which involved bidding contests, were the
Chevron takeover of Gulf Oil for $13.2 billion and the Texaco acquisition of Getty
Oil for $10.1 billion (the litigation with Pennzoil over this transaction had reached
the U.S. Supreme Court in June 1986). Several notable transactions in 1984 did not
RECENT DEVELOPMENTS IN M&A ACTIVITY 681
involve oil companies. Nestle (a Swiss company) acquired the Carnation Company
for $2.9 billion. Beatrice bought Esmark for $2.5 billion and then itself was the subject
of a leveraged buyout with a value of $5.4 billion by a group led by Kohlberg-Kravis.
General Motors bought Electronic Data Systems in 1984 for $2.5 billion and Hughes
Aircraft in 1985 for $5.0 billion. Other dramatic deals in 1985 included General Elec-
tric's purchase of RCA for $6.0 billion, Philip Morris of General Foods for $5.6 billion,
R. J. Reynolds of Nabisco for $4.9 billion, Allied Chemical of Signal for $4.5 billion,
and Baxter Travenol of American Hospital Supply for $3.7 billion.
Table 19.2 places these individual deals in broader perspective. Column (1) shows
the total dollar value paid where a purchase price was disclosed; this would of course
We have now viewed the various forms of corporate restructuring and control. We
have also presented data on the amount of merger-acquisition activity. We will now
seek to move on to understand the theories or rationale for mergers, tender offers,
THEORIES OF M&A ACTIVITY 683
and joint ventures. We have grouped these theories into five major areas:
1. Efficiency,
2. Information,
3. Agency problems,
4. Market power, and
5. Taxes.
Each of these five groups of theories will be briefly discussed.
1. Efficiency Explanations
Efficiency theories are the most optimistic about the potential of mergers for
social benefits. The most general theory involves differential efficiency. In theory, if
the management of firm A is more efficient than the management of firm B, and if
after firm A acquires firm B, the efficiency of firm B is brought up to the level of
efficiency of firm A, efficiency is increased by merger. Note that this would be a social
gain as well as a private gain. The level of efficiency in the economy would be raised
by such mergers.
One difficulty in the differential efficiency theory is that if carried to its extreme,
it would result in only one firm in the economy, indeed in the world—the firm with
the greatest managerial efficiency. Clearly, problems of coordination in the firm would
arise before that result was reached. Hence another formulation of the differential
efficiency theory of mergers is that there are always many firms that exhibit below-
average efficiency or that are not operating up to their potentials, however defined.
It is further suggested that firms operating in similar kinds of business activity would
be most likely to be the potential acquirers. They would have the background for
detecting below-average or less-than-full-potential performance and have the man-
agerial know-how for improving the performance of the acquired firm. The latter
scenario is plausible, but in practice the acquiring firms may be overoptimistic in
their judgments of their impact on the performance of the acquired firms. As a conse-
quence, they may either pay too much for the acquired firm or fail to improve its
performance to the degree reflected in the acquisition value placed upon it. (See Roll's
[1986] hubris hypothesis under "Agency Problems" below.)
The inefficient management theory may be difficult to distinguish from the
differential efficiency theory discussed above or the agency problem treated below.
In one sense inefficient management is simply not performing up to its potential.
Another control group might be able to manage the assets of this area of activity
more effectively. Or inefficient management may simply represent management that
is inept in some absolute sense. Almost anyone could do better. If so, this would
provide a rationale for conglomerate mergers. The differential efficiency theory is
more likely to involve management that is superior because it has experience in a
particular line of business activity. The differential efficiency theory is more likely to
be a basis for horizontal mergers. The inefficient management theory could be a basis
for unrelated mergers.
684 MERGERS, RESTRUCTURING, AND CORPORATE CONTROL: THEORY
Efficiency theories also include the possibility of achieving some form of synergy.
If synergy occurs the value of the combined firm, VAB , exceeds the value of the in-
dividual firms brought together by the mergers. With synergy,
VAB VA + VB•
Because much of the theoretical literature was stimulated in response to the
dramatic conglomerate merger movement of the 1960s and because true social gains
from conglomerate mergers were not readily perceived, much of the formal analysis
in the scholarly literature assumed no synergy:
VAB VA ± VB•
Myers [1968], Schall [1972], and Mossin [1973], in the spirit of the earlier
Modigliani-Miller (M-M) [1958] formulation, have all argued that value is conserved
(value additivity obtains) under addition of income streams (mergers). Nielsen [1974]
pointed out that they utilized a partial equilibrium approach in which prices and
other parameters are assumed to be constant. Nielsen also called attention to the
underlying assumption that the capital markets are complete and perfect, and to the
further assumption that the equilibrium allocation of income (hence also the marginal
utilities of all investors) is invariant with respect to a change in the number of trading
instruments due to merger. The conditions for value additivity are identical to those
required for the M-M propositions. Thus at the theoretical level the conditions under
which value is conserved (even assuming the absence of synergy) are more restrictive
than generally acknowledged.
Nevertheless, the dominant theme in the theoretical literature is that the value-
additivity principle holds in still another setting, the area of combining business
entities through mergers or acquisitions. Much of the journal literature has sought
to determine whether, in the absence of synergy, any theoretical justification for a
merger could be found. However, the justification for mergers given by the executives
engaging in mergers and by some of the general literature as well was that there
were at least potential real gains from combining business firms. These arguments
are stated in various forms, but implicitly they all depend on one condition—that
operating economies of scale may be achieved. This theory is based on a number of
major assumptions. It assumes that economies of scale do exist in the industry and
that prior to the merger, the firms are operating at levels of activity that fall short
of achieving the potentials for economies of scale.
Basically, economies of scale involve "indivisibilities," such as people, equipment,
and overhead, which provide increasing returns if spread over a larger number of
units of output. Thus in manufacturing operations, heavy investments in plant and
equipment typically produce such economies. For example, costly machinery such
as the large presses used to produce automobile bodies requires optimal utilization.
Similarly, in the area of financing, data on flotation costs indicate that the cost of
floating a larger issue amounts to a smaller percentage of the issue floated because
the fixed costs of investigation and compliance with SEC regulations are spread over
a larger dollar amount. (Financial synergy also encompasses the potential for achiev-
ing a lower cost of capital as a result of the reduced risk of bankruptcy when im-
THEORIES OF M&A ACTIVITY 685
perfectly correlated cash flow streams are joined.) And the same principle applies to
the research and development (R & D) departments of chemical and pharmaceutical
companies, which often have to have a large staff of highly competent scientists who,
if given the opportunity, could develop and oversee a larger number of product areas.
Finally, in marketing, having one organization cover the entire United States may
yield economies of scale because of the increase in the ratio of calling-on-customer
time to traveling time, which in turn is due to the higher density of customers who
can be called on by the same number of salesmen.
One potential problem in merging firms with existing organizations is the ques-
tion of how to combine and coordinate the good parts of the organizations and
eliminate what is not required. Often the merger announcement will say that firm A
is strong in research and development but weak in marketing, whereas firm B is
strong in marketing but weak in R & D, and the two firms combined will complement
each other. Analytically, this implies underutilization of some existing factors and
inadequate investment in other factors of production. (Since the economies are jointly
achieved, the assignment of the contributions of each firm to the merger is difficult
both in theory and in practice.)
Economies in production, research, marketing, or finance are sometimes referred
to as economies in the specific management functions. It has also been suggested
that economies may be achieved in the generic management activity such as the plan-
ning and control functions of the firm. It is argued that firms of even moderate size
need at least a minimum number of corporate officers. The corporate staff with capa-
bilities for planning and control is therefore assumed to be underutilized to some
degree. Acquisitions of firms just approaching the size where they need to add cor-
porate staff would provide for fuller utilization of the corporate staff of the acquiring
firm and avoid the necessity of adding such staff for the other firm.
A third area in which operating economies may be achieved is in vertical inte-
gration. Combining firms at different stages of an industry may achieve more efficient
coordination of the different levels. The argument here is that costs of communication
and various forms of bargaining can be avoided by vertical integration [Williamson,
1971; Arrow, 1975].
Efficiency theories provide a basis for mergers to achieve strategic planning goals
in response to a rapidly changing environment. The literature on long-range strategic
planning has exploded in recent years. This literature is related to diversification
through mergers. The emphasis of strategic planning is on areas related to firms'
environments and constituencies, not just operating decisions [Summer, 1980].
Earlier the emphasis of long-range strategic planning was on doing something
about the so-called gap. When it is necessary to take action to close a prospective
gap between the firm's objectives and its potential based on its present capabilities,
some difficult choices must be made. For example, shall the firm attempt to change
its environment or capabilities? What will be the costs of such changes? What are
the risks and unknowns? What are the rewards if successful? What are the penalties
of failure? Because the stakes are large, the iterative process is employed. A tentative
decision is made. The process is repeated, perhaps from a different management func-
tion orientation, and at some point the total-enterprise point of view is brought to
686 MERGERS, RESTRUCTURING, AND CORPORATE CONTROL: THEORY
bear on the problem. Ultimately, decisions are made and must involve entrepreneurial
judgments.
Alternatively, the emphasis may be on broader orientations to the effective align-
ment of the firm with its environments and constituencies. Different approaches may
be emphasized. One approach seeks to choose products related to the needs or mis-
sions of the customer that will provide large markets. A second approach focuses on
technological bottlenecks or barriers, the solution of which may create new markets.
A third strategy chooses to be at the frontiers of technological capabilities on the
theory that some attractive product fallout will result from such competence. A fourth
approach emphasizes economic criteria including attractive growth prospects and
appropriate stability.
Other things being equal, a preferred strategy is to move into a diversification
program from the base of some existing capabilities or organizational strengths. Guid-
ance may be obtained by answers to the following questions: Is there strength in the
general management functions? Can the company provide staff expertise in a wide
range of areas? Does the firm's financial planning ando control effectiveness have a
broad carryover? Are there specific capabilities such as research, marketing, and
manufacturing that the firm is seeking to spread over a wider arena? [Chung and
Weston, 1982].
It appears that the strategic planning approach to mergers implies either the pos-
sibilities of economies of scale or utilizing some unused capacity in the firm's present
managerial capabilities. Another rationale is that by external diversification the firm
acquires management skills for needed augmentation of its present capabilities. This
still leaves some questions unanswered. New capabilities and new markets could be
developed internally. It may be less risky to buy established organizations, but a
competitive market for acquisitions implies that the net present value from such in-
vestments is likely to be zero. However, if these investments can be used as a base
for still additional investments with positive net present values, the strategy may
succeed.
2. Information Theories
The information, or signaling, hypothesis refers to the revaluation of the ownership
shares of firms owing to new information that is generated during the merger nego-
tiations, the tender offer process, or the joint venture planning. Alternative forms
of the information hypothesis have been distinguished by Bradley, Desai, and Kim
[1983]. One is the kick-in-the-pants explanation. Management is stimulated to im-
plement a higher-valued operating strategy. A second is the sitting-on-a-gold-mine
hypothesis. The negotiations or tendering activity may involve the dissemination of
new information or lead the market to judge that the bidders have superior informa-
tion. The market may then revalue previously "undervalued" shares.
Another aspect of the undervaluation theory is the difference in the position of
a control group versus an individual investor. For example, in recent years the q-ratio
has been running between .5 and .6. The q-ratio is the ratio of the market value of
the firm's shares in relation to the replacement costs of the assets represented by
THEORIES OF M&A ACTIVITY 687
3. Agency Problems
In their seminal paper, Jensen and Meckling [1976] formulated the implications
of agency problems. An agency problem arises when managers own only a fraction
of the ownership shares of the firm. This partial ownership may cause managers to
work less vigorously than otherwise and/or to consume more perquisites (luxurious
offices, company car, membership in clubs) because the majority owners bear most
of the cost. Furthermore, the argument goes, in large corporations with widely dis-
persed ownership, there is not sufficient incentive for individual owners to expend
the substantial resources required to monitor the behavior of managers. A number
of compensation arrangements and the market for managers may mitigate the agency
problem [Fama, 1980].
The agency problem theory of mergers has two aspects. On the one hand the
threat of takeover may mitigate the agency problem by substituting for the need of
individual shareholders to monitor the managers. The agency theory extends the
previous work by Manne [1965]. Manne emphasized the market for corporate control
and viewed mergers as a threat of takeover if a firm's management lagged in perfor-
mance either because of inefficiency or because of agency problems.
On the other hand, mergers may be a manifestation of the agency problem rather
than the solution. The "managerialism" explanation for mergers was set forth most
fully by Mueller [1969]. Mueller hypothesizes that managers are motivated to in-
crease the size of their firms further. He assumes that the compensation to managers
is a function of the size of the firm, and he argues therefore that managers adopt a
lower investment hurdle rate. But in a study critical of earlier evidence, Lewellen and
Huntsman [1970] present findings that managers' compensation is significantly cor-
related with the firm's profit rate, not its level of sales. Thus the basic premise of the
Mueller theory is doubtful.
Roll's [1986] hubris hypothesis also suggests that the agency problem is not
checked by the control mechanisms above. Hubris is a Greek word meaning "animal
688 MERGERS, RESTRUCTURING, AND CORPORATE CONTROL: THEORY
spirits," with connotations of overexuberance and excess pride. Roll suggests that
managers commit errors of overoptimism in evaluating potential merger candidates,
thus bidding more than they should and transferring virtually all gains from the
transaction to the target shareholders.
4. Market Power
One reason often given for a merger is that it will increase a firm's market share,
but it is not clear how increasing the market share will achieve economies or synergies.
If increasing the firm's market share simply means that the firm will be larger, then
we are essentially talking about economies of scale, which we have already discussed.
Increasing market share really means increasing the size of the firm relative to other
firms in an industry. But it is not made clear why increasing the firm's relative size
will provide economies or other social gains.
Indeed, this poses a challenge to the arguments for merger emphasizing econ-
omies of scale and vertical integration. These could also be achieved by the internal
expansion of the firm. Why is the external acquisition of another firm necessary to
achieve these economies if indeed they do exist? A number of possible explanations
may be offered, such as acquiring a larger volume of operations sooner. But it is not
clear whether the price required by the selling firm (the firm to be acquired) will
really make the acquisition route the more economical method of expanding a firm's
capacity either horizontally or vertically.
An objection that is often raised against permitting a firm to increase its market
share by merger is that the result will be "undue concentration" in the industry.
Indeed, public policy in the United States holds that when four or fewer firms account
for 40% or more of the sales in a given market or line of business, an undesirable
market structure or undue concentration exists. The argument in brief is that if four
or fewer firms account for a substantial percentage of an industry's sales, these firms
will recognize the impact of their actions and policies upon one another. This
recognized interdependence will lead to a consideration of actions and reactions to
changes in policy that will tend toward "tacit collusion." As a result, the prices and
profits of the firms will contain monopoly elements. Thus if economies from mergers
cannot be established, it is assumed that the resulting increases in concentration may
lead to monopoly returns. If economies of scale can be demonstrated, then a com-
parison of efficiencies versus the effects of increased concentration must be made.
In 1982 and 1984, the Department of Justice announced new merger guidelines
to supersede those that had been issued in May 1968. The new merger guidelines
adopt the Herfindahl index (H index), which takes into consideration the market
shares of all the firms in the industry. The theory behind the use of the Herfindahl
index is that if one or more firms have relatively high market shares, this is of even
greater concern than the share of the largest four firms. An example presented with
the announcement of the new merger guidelines illustrates this point.
In one market four firms each hold a 15% market share and the remaining 40%
is held by 40 firms, each with a 1% market share. Its H index would be
H = 4( 15)2 + 40(1)2 = 940.
THEORIES OF M&A ACTIVITY 689
In another market one firm has a 57% market share and the ,remaining 43% is held
by 43 firms, each with a 1% market share. Like the first market, the four-firm con-
centration ratio here would be the same 60%. However, the H index would be
H = (57)2 + 43(1)2 = 3292.
Thus the H index registers a concern about inequality as well as degree of con-
centration. The economic basis for either concern has not been well established.
The regulatory authorities also continue to give heavy weight to the percentage
share of the market held by individual firms. It was on the basis of individual firm
market shares that the Federal Trade Commission (FTC) raised objections in June
1986 to Coca-Cola's plan to buy Dr. Pepper and the offer of Philip Morris to sell
its Seven-Up soft-drink unit to Pepsi. As a consequence of FTC objections Philip
Morris withdrew its offer, but Coca-Cola and Dr. Pepper said they would challenge
the FTC's attempt to block their merger. Similarly, in 1984 the FTC challenged a
proposed merger between Warner Communications Inc. and Polygram Records Inc.,
principally owned by N. V. Philips, a Netherlands-based telecommunications and
electronics concern. After some two years of legal maneuvering, in June 1986 both
companies agreed to seek FTC approval before merging with, or acquiring a stake
in, other large record companies. The basis for the original FTC challenge was that
if Warner and Polygram were joined, it would become the largest recorded music
distributor in the world, accounting for 26% of the U.S. market.
While some economists hold that high concentration, however measured, causes
some degree of monopoly, other economists hold that increased concentration is
generally the result of active and intense competition. They argue further that the
intense competition continues among large firms in concentrated industries because
the dimensions of decision making over prices, outputs, types of product, quality of
product, service, etc., are so numerous and of so many gradations that collusion
simply is not feasible. This is an area where the issues continue to be unresolved.
5. Tax Considerations
Tax considerations are also involved in mergers. One such tax consideration is
to substitute capital gains taxes for ordinary income taxes by acquiring a growth
firm with a small or no dividend payout and then selling it to realize capital gains.
Also, when the growth of a firm has slowed so that earnings retention cannot be
justified to the Internal Revenue Service, an incentive for sale to another firm is
created. Rather than pay out future earnings as dividends subject to the ordinary
personal income tax, an owner can capitalize future earnings in a sale to another
firm. One often sees a small proprietor who develops an ongoing business (say, Com-
pany A) and begins to have a significant flow of net income. If he or she takes the
net income as dividends, it is subject to a high personal income tax rate. Also, given
potential competition from other firms, the ability of Company A to earn the income
in the future is uncertain from the owner's standpoint. The certainty of that income
flow would be increased if Company A were part of a larger firm that had the neces-
sary full complement of management capabilities. The owner of Company A converts
690 MERGERS, RESTRUCTURING, AND CORPORATE CONTROL: THEORY
E. THEORIES OF RESTRUCTURING
The first group of activities in Table 19.1, expansion, involves combining assets. The
remaining groups involve uncombining assets (sell-offs); establishing and defending
rights to assets (corporate control); and altering the format of asset control (changes
in ownership structure).
Merger theories imply that 2 + 2 can be greater than 4. The rationale for other
types of restructuring seems to be that 4 — 2 can be more than 2; or that 3 + 1 can
be more than 2 + 2. Thus a new math has evolved to explain both mergers and other
restructuring activities.
The various forms of divestiture can be rationalized as transferring business assets
to a higher-valued use or to a more efficient user. A divestiture may create value by
slicing off a business that was a poor fit with the remaining operations. If so, good
divestiture programs may increase the market values of both the buying and selling
companies.
There is another aspect of the second to fourth categories of activities in Table
19.1 that deals with corporate control and rearranging the ownership structure. Here
the hypothesis is that improvements in managerial accountability and a strengthening
of incentives may be achieved by separating unrelated business activities. Spinoffs
CONGLOMERATE MERGERS 691
F. CONGLOMERATE MERGERS
One reason for the study of conglomerate mergers is that after the tightening of the
laws with respect to mergers that took place in 1950, most merger activity until the
late 1970s was conglomerate because major horizontal and vertical mergers were sub-
ject to legal challenge. In addition the economic rationale for conglomerate mergers
seemed to be the weakest. If by definition conglomerate mergers brought together
unrelated activities, how could there be economic benefits? Some possible benefits of
conglomerate mergers may be considered.
1. Financial Gains
Several financial gains from conglomerate mergers have been proposed. Argu-
ments claiming financial gains from mergers other than the possible advantage of
economies in financing are of dubious validity. As presented in most basic finance
texts, it can readily be demonstrated that a differentially higher price/earnings ratio
can achieve gains in earnings per share for acquiring firms. If merger terms represent
a differentially higher price/earnings ratio for the acquiring firm, that firm's earnings
per share after the acquisition will be higher and the earnings per share of the acquired
firm will be lower. The acquiring firm achieves earnings accretion while the acquired
firm suffers earnings dilution.
But it can be readily demonstrated that the higher price/earnings ratio must
reflect differentially more favorable earnings growth prospects. The effect of acquiring
firms with low price/earnings ratios and lower earnings growth prospects will be to
depress the average rate of future growth in earnings of the combined firm. The new
price/earnings ratio should reflect this change in the outlook for the growth in earnings
of the merged firm so that there should be no gain in the value of the combined firm
over the values of the two separate firms. Hence the differential price/earnings ratio
theory and its immediate effects on earnings per share have no validity as a theory
for measuring the potential gains of a merger, since it is valuation, not earnings per
share, that is the relevant test.
692 MERGERS, RESTRUCTURING, AND CORPORATE CONTROL: THEORY
Table 19.4*
(1) (2) (3)
State (s i ) P(si) YI
1 .1 100
2 .2 250
3 .7 500
It is assumed that each firm has incurred borrowings to the point that its annual
cash contractual obligation amounts to $240. The probability, P(D), that one or both
firms will be unable to meet their debt service obligations of $240 each is shown
below:
P(D) = P(YA < 240) + P(YB < 240) - P(YA < 240, YB < 240)
= .1 + .1 - (.1)(.1) = .19.
If the two firms merge, the distribution of their joint returns can be calculated
as shown in the matrix in Table 19.5. The distribution of their joint returns would
therefore become
Y„, 200 350 500 600 750 1,000
P(Ym) .01 .04 .04 .14 .28 .49
Since their aggregate debt burden would become $480 per annum, the probability
of default now drops to .05 as compared with .19 before the merger.
The foregoing was based on the assumption of zero correlation between the two
returns. If the correlation were - 1, the gains from merger would be even greater. If
the correlation were + 1, reducing the probability of default would require other dif-
ferences such as differences in the size of debt obligations. Thus if we assume the
same distribution of returns for the two firms as before, but assume a correlation of
+ 1 between the two and the debt obligations of A and B to be $255 and $240, respec-
tively, the before-merger probability of default would be
P(D) = .3 + .1 - .1 = .3.
After merger the total debt obligations would be $495, and they would be related
to the following combined cash flow pattern:
Table 19.5
.1 .2 .7
.1 Joint probability .01 .02 .07
Amount 200 350 600
.2 Joint probability .02 .04 .14
Amount 350 500 750
.7 Joint probability .07 .14 .49
Amount 600 750 1,000
694 MERGERS, RESTRUCTURING, AND CORPORATE CONTROL: THEORY
Levy and Sarnat [1970, 801] set forth a similar argument. They state, "A some-
what stronger case can be made for conglomerate mergers when economies in capital
costs are considered. . . . large firms have better access to the capital markets and
also enjoy significant cost savings when securing their financing needs. . . . These cost
savings presumably reflect, at least in part, the reduction in lenders' risk achieved
through diversification."
Galai and Masulis [1976] point out the confusion involved between the value
of the merged firm and the positions of the debt and equity holders. They argue that
the value of the merged firm is the simple sum of the constituent firms. "This can
be seen once one recognizes that investors in the marketplace could have created an
identical financial position by purchasing equal proportions of the debt and equity
of the two firms" [1976, 68]. The OPM establishes that the relative position of the
creditors and the equity holders of the firms will be changed. If the correlation be-
tween the returns of the merging firms is less than 1, the variance in the rate of return
of the merged firm will be lower than the variance of the rates of return of the merging
firms (assumed to be equal). (For numerical illustrations see Problems 19.6 to 19.8.)
It follows from the OPM that the value of the equity of the merged firm will be
less than the sum of the constituent equity values and the value of the debt will be
higher. According to the OPM, increased variability increases the value of the option,
and conversely. Since the equity is an option on the face value of the debt out-
standing, its value will fall with a decrease in volatility. "What is taking place, as
Rubinstein points out, is that the bondholders receive more protection since the
stockholders of each firm have to back the claims of the bondholders of both com-
panies. The stockholders are hurt since their limited liability is weakened" [Galai
and Masulis, 1976, p. 68].
Thus a pure diversification rationale for conglomerate mergers would not seem
valid. Reducing the risk to bondholders represents a redistribution of value from
shareholders, leaving the total value of the firm unchanged.
However, a number of alternatives could be used to return the wealth of different
classes of security holders to the original position they held prior to the merger. One
solution would be to increase the amount of the face value of debt and use the
proceeds to retire equity. This process is continued until the original bondholders'
holdings have a market value equal to their constituent sum prior to the merger.
The debt-to-equity ratio of the merged firm can be increased to offset the decrease
in the volatility of the merged firm's rate of return. The increased amount of debt
implies that the total value of the firm is increased through merger due to the tax
deductibility of interest payments. Galai and Masulis suggest that this may explain
some conglomerate mergers.
k — (S A + S8)
Bond A Bond B Common Stock*
6 A > 6C > an >0 <0 >0
<
Variance
(TA < ac < us <0 >0 <0
effect
6
A as > ac >0 >0 <0
under these more general conditions are presented in Table 19.6. Because the correla-
tion between the cash flow streams of the two firms can be either positive or negative,
the resulting combined firm variance may be less than the variance of the individual
firms or greater than one of the firms and less than the other firm. However, the
leverage effects and maturity effects are simply weighted average effects. The com-
bined firm will simply have a weighted average of the leverage or maturity pattern
of the combining firms.
The other relationship that stands out in the table is that the value of the common
stock of the combined firm may under some conditions be larger than the addition
of the premerger stock values of the two firms even in a pure conglomerate merger,
assuming no synergy. The reasons for Shastri's conclusions are reviewed next.
There are three possibilities under the variance effect. The variance of the com-
bined firm may be less than the variance of either of the individual firms before the
merger. This is the G-M result where the variance effect is positive for the bonds of
both firm A and firm B and negative for both stocks. However, the variance of the
combined firm may be greater than one of the firms and lower than the other. For
example, when ciA > o-c > cr,, the securities of firm B become riskier with the merger,
conversely with firm A. So there is a positive impact on the bonds of A and the stock
of B, negative on the bonds of B and the stock of A. The effects on the value of the
stock of the combined firm compared with the sum of the premerger stock values is
ambiguous depending on the relative magnitudes of the premerger stock A and stock
B values.
Shastri defines the leverage ratio as the ratio of the face value of debt to firm
value. If the leverage ratio of firm A is greater than the ratio for B, the combined
firm has a leverage ratio less than that of firm A and greater than that of firm B.
The merger results in a decrease in the leverage-ratio related risk for bond A, with
the opposite result for bond B. This implies an increase in the value of bond A and
a decrease in the value of bond B. When the leverage ratio of A is lower than that of
696 MERGERS, RESTRUCTURING, AND CORPORATE CONTROL: THEORY
S aj = (1 — t)MAX[Qaj — C aj — R a , 0].
Ra if Q aj > (R a + C aj
Ydai
}MIN[R a , Q aj — B aj ] otherwise.
This states that a conglomerate merger of all-equity firms can never be profitable.
The intuition is that the merger is unprofitable because the limited liability protec-
tion of a merged firm is weaker than that of two unmerged firms. This is a familiar
result, but Scott's is the first paper to obtain this result so easily because of the state-
preference approach he employs. Scott next analyzes the merger of levered firms.
Evaluation relationships are as follows:
Using the valuation relationships above, Scott works out a table of the effects
of pure conglomerate mergers. He explicitly considers all possible combinations of
solvency and bankruptcy for the two merging firms and for the merged firm. He
assumes economies of scale in bankruptcy costs. His table presents the incremental
cash flows to security holders as a result of the conglomerate merger. His results are
reproduced as our Table 19.7.
698 MERGERS, RESTRUCTURING, AND CORPORATE CONTROL: THEORY
AB Solvent AB Bankrupt
A, B solvent* 0
A, B bankrupt MIN[Q„, + Q bj — k bj , Ra + R b ]
—MIN[Q aj — k J , R„]
—MIN[Q bi — Bhp Rb]
* This should be interpreted as follows: If A and B had remained unmerged, these are the states in
which neither would have gone bankrupt. The other entries are interpreted in a similar fashion.
From J. H. Scott, Jr., "On the Theory of Conglomerate Mergers," Journal of Finance, September
1977, 1244.
To aid in the interpretation of the table, consider the case where firm B would
be bankrupt while firm A and the combined firm would remain solvent. The proof
of his result is as follows:
Z, = Yab j Ya j Yb j
Z i = [(} — t (Q aj Qbj — C aj — C hin t(R, + R b — — t (Qaj — C a — tR a
— MIN[(Q b , — Rb ]
Z i = Q bi C hi tQ bi tC bi tR h — MIN[(Q bi — B bj , Rd
From J. H. Scott, Jr., "On the Theory of Conglomerate Mergers," Journal of Finance, September 1977,
1248.
Scott assumes further that the tax rate is 50% and that the price of the state-
contingent claims for either state is .4. The computations under alternative assump-
tions are as follows:
With no merger:
If Ra = Qai = 20
If State 1 Yap = (1 — .5)(20) + .5(20) 20 x .4 = 8
If State 2 Ya 2 = — .5)(30) + .5(20) 25 x .4 = 10
V); = 18
700 MERGERS, RESTRUCTURING, AND CORPORATE CONTROL: THEORY
If Ra = Qa2 = 30
If State = Ya= = MIN(30, <0 - 4) =6 x .4 = 6.4
If State < Ya< = (= - .P)(30) + .P(30) 30 x .4 = =<.0
Va = =8.4
So R.- = Q a2 = 30.
If R b = Qbi = <P
If State = Yîi = (= - .P)(<P) + .P(<P) <P x .4 = =0.0
If State < Yî< = MIN[<P, (<0 - <)] =8 x .4 = 7.<
Vî = =7.<
If R b - Q b2 - <0
If State = Yîi = (= - .P)(<P) + .P(<0) <<.P x .4 = 9
If State < Yî< = (= - .P)(<0) + .P(<0) <0.0 x .4 = 8
V?r, = =7
So Rt = Qbl = <P.
With a me]ge]-
IfR aî =PP
If State = K î , = MIN[PP, (4P - =3)] 3< x .4 = =<.8
If State < Yaî< = MIN[PP, (P0 - ==)] 39 x .4 = =P.6
Vab = <8.4
If R,= 4P
If State = Yaîl = ( = - .P)(4P) + .P(4P) 4P.0 x .4 = =8.0
If State < Yaî< = (= - .P)(P0) + .P(4P) 47.P x .4 = =9.0
Vab = 37.0
If R aî = P0
If State = Ya„ = MIN[P0, (4P - =3)] 3< x .4 = =<.8
If State < Yaî< = ( = - .P)(P0) + .P(P0) P0 x .4 = <0.0
F!,'b = 3<.8
So R:b = Qabi = 4P.
The numerical results aîove illustrate Scott's propositions. Like the Kraus and
Litzenîerger results they demonstrate that it is always profitaîle to increase the
optimal level of deît, R*, to equal one of the îefore-tax cash flows. This is îecause
one can always increase the optimal deît level to the next highest cash flow, thereîy
decreasing the present value of future tax payments with no increase in the proîaîility
of incurring îankruptcy costs. To oîtain the optimal capital structure we calculate
the value of the firm when the deît level equals the pretax cash flow in state = and
again when it equals the pretax cash flow in state <. We then choose the deît level
that results in the highest value of the firm. For the numerical example the optimal
deît level is set equal to the cash flow of the firm that is highest for a given state.
CONGLOMERATE MERGERS 701
However, after the merger the highest value of the firm is achieved with the state 1
cash flow of the merged firm that is lower than the assumed state 2 cash flow of the
merged firm. If the merged firm were to maintain the capital structure of the un-
merged firms at a level of 55, it would result in a value of the merged firm of 28.4.
This is less than the value of the unmerged firms (18.4 + 17.2 = 35.6). Scott's example
illustrates that optimal debt may be reduced, therefore, as a consequence of the pure
conglomerate merger. Yet if the firm chooses an optimal level of debt after the merger,
the value of the firm is increased to 37. This exceeds the sum of the components,
which totaled 35.6.
These results are of course specific to the pattern of the numerical relationships
chosen for Scott's example. They are general, however, to the extent that they are
illustrative of the kind of possibilities that may conceivably occur. The general view
is that conglomerate mergers may increase "debt capacity" (however defined) but may
(or may not) increase firm value. Scott presents the possibility that the conglomerate
merger may actually reduce the optimal level of debt. On the other hand, choosing
an optimal level of debt after the merger may produce a value of the combined firm
that is greater than the sum of the individual values of the components. Thus Scott
concludes that the probability of bankruptcy is in general not an appropriate measure
of the profitability of a conglomerate merger. He argues also that as a result of a
conglomerate merger the resulting firm's optimal level of debt can either rise or fall.
These results illustrate a more general proposition about merger theory. Many
possible motives for mergers can exist under various models and assumptions. Various
types of complementarities may make mergers profitable. Alternative assumptions
about tax rules can make mergers profitable or unprofitable under alternative assump-
tions. This makes modeling theories for testing mergers and approaching empirical
data difficult to formulate.
u j = the mean of X j ,
j
With given values of the risk-free interest rate and the mean and standard devia-
tion of Xi for a given level of debt obligations incurred, Y, Eq. (19.1) can be used with
the additional constraint that VD.] = (1 + i) to calculate debt capacity (i is equal to
the effective yield to maturity on the debt).
To illustrate the calculation of debt capacity and some related relationships in
the Stapleton formulation, let us use the following illustrative values:
For Firm A ti A = 100, o-A = 25, r = 10%, YA = 68.
The solution by iteration for debt capacity utilizing Eq. (19.1) plus the constraint with
i = .12 is DA = 60.7. Hence YA = D'A'(1 + = (60.7)(1.12) = 68. The left-hand side (lhs)
of Eq. (19.1) is 1.1 times IPA because the riskless rate is 10%. Thus the left-hand side
of Eq. (19.1) is 1.1(60.7) = 66.77.
The right-hand side (rhs) of Eq. (19.1) is as follows:
rhs = 100F[(68 — 100)/25] — 25f[(68 — 100)/25] + 68(1 — F[(68 — 100)/25] )
= 100(1 — F[(100 — 68)/25] ) — 25f[(100 — 68)/25] + 68(F[(100 — 68)/25] )
= 100(.1003) — 25(1758) + 68(8997)
= 10.03 — 4.395 + 61.18 = 66.81.
Therefore the equality of rhs to lhs is verified. Finally, the probability that X will be
less than Y is just F(Y) = F(68), which is .1003 = 10.03%.
In a like manner, Stapleton calculates debt capacity for different firms with the
same expected cash flow but increasing standard deviations of cash flows. The calcu-
lation is also performed for different required yields. The results are summarized in
Table 19.9.
The data in Table 19.9 indicate that at any given required yield the amount of
debt obtainable declines at an increasing rate as the risk of the underlying cash flows
increases. Debt capacity increases at a diminishing rate as the promised coupon
increases.
The analysis is next extended to an application to pure conglomerate mergers in
which no synergy is assumed. The only influence is the effect on the standard devia-
tion of the combined firm in relation to the component firms. The results in Table
19.10 depend on the correlation between cash flows of the merging firms. This, in turn,
produces the magnitude of the standard deviation of returns for the combined firm.
The new sigma in Table 19.10 is obtained using the expression in Eq. (19.2):
<
, (19.2)
'BD = ' D ''FBD"B"D•
Table 19.9 Debt Values: Risk Neutrality
YA ni -, ,\ v- , YB YC YD
percent YA DA CAA A < }A }B -.]B n P(2B < YB Y3 D 3 P(2 3 < Y3 YD D D P(2 D < YD
DA -..° B P3 DD
Debt level 1 12 68 60.7 1.12 .10 57 50.9 1.12 .08 42 38.4 1.12 .05 23 20.5 1.12 .03
Debt level 2 14 79 69.3 1.14 .20 70 61.4 1.14 .16 59 51.8 1.14 .12 45 39.5 1.14 .08
Debt level 3 16 86 74.1 1.16 .27 78 67.2 1.16 .23 70 60.3 1.16 .20 58 50.0 1.16 .15
From R. C. Stapleton, "Mergers, Debt Capacity and the Valuation of Corporate Loans," reprinted by permission of the publisher, from Mergers and A3quisitions,
edited by Michael Keenan and Lawrence J. White, Lexington Books, D. C. Heath & Co., Lexington, Mass. Copyright © 1982, D. C. Heath & Co.
704 MERGERS, RESTRUCTURING, AND CORPORATE CONTROL: THEORY
The results in Table 19.10 follow directly from the application of Eq. (19.1) just
as the computations were made for Table 19.9. For example, we simply verify that
the left-hand side equals the right-hand side of Eq. (19.1) for the data in Table 19.10.
At a yield of 12% the left-hand side is
lhs = 1.1 x 57/1.12 = 55.98.
(100 x .0764) — (30 x .1435) + (57 x .9236) = 7.64 — 4.30 + 52.65 = 55.99.
Thus, again, it is demonstrated that the values shown in Table 19.10 satisfy Eq.
(19.1). The other results in Table 19.10 are calculated in a similar manner. The spread
of results in Table 19.10 seeks to illustrate that the effect of merger on debt capacity
is dependent on the relative risks of the merging firms as well as on the correlation
between the earnings of the firms. The sum of the individual debt capacities of firms
B and D at a required yield of 12% is $80. Table 19.10 then shows the effect on debt
capacity of merging firms, assuming that the cash flows of the firm are uncorrelated,
correlated, and perfectly correlated. Even in a perfect correlation case, debt capacity
rises from $80 to $84. The specific numerical examples in Stapleton are special cases
of the general propositions set forth by Shastri.
This contrasts with the results for Lewellen [1971] where debt capacity is defined
as the level of promised payments such that the probability of Xi being less than Y
CONGLOMERATE MERGERS 705
is less than a. In the example in Table 19.10, a is chosen so that the total debt of
firms B and D is the same as that generated by a 12% interest yield requirement.
Thus for the Lewellen formulation we have the requirement, using the data of Table
19.10, that
Pr(X B < 48.6) = Pr(X D < 31.4).
Now we can test the equality of the expression as follows:
lhs = 1 — F[(100 — 48.6)/30] = 1 — F(1.713) = 1 — .9564 = .0436,
rhs = 1 — F[(100 — 31.4)/40] = 1 — F(1.715).
Thus we have verified that lhs = rhs and that
Pr(X B < 48.6) = Pr(X D < 31.4) = .0436.
For the merged firm, when the correlation coefficient between the cash flows is
zero, the standard deviation of the merged firm is 50. We then seek to verify that
the debt capacity for the merged firm is 114.3. This is demonstrated by the following:
Pr(X BD < 114.3) = 1 — F[(200 — 114.3)/50] = 1 — F(1.714) = .0436
as above.
The results for the calculations for correlation coefficients of .46 and 1.0 are also
shown in the table. The results illustrate that debt capacity is higher under the OPM
formulation than under the limited formulation of Lewellen. Indeed, with perfect cor-
relation between the cash flows in the Lewellen formulation, debt capacity is not in-
creased. However, in the OPM model, debt capacity is increased as long as the sigmas
of the merging firms are different, even when the correlation coefficient between the
returns is 1.
Similar calculations of the effects of mergers on debt capacity are also developed
by Stapleton for the risk-averse case, using the results of Brennan [1979]. Here we do
not have risk-neutral pricing; rather we have risk-averse pricing (greater covariances
with the market lead to lower prices). Equation (19.1) is used again in Table 19.11;
however, the location parameter for the distribution of cash flows (the mean return)
has to be transformed to a certainty equivalent (rV). This can be calculated using
the CAPM, since we assume exponential utility and a normal distribution. The deter-
mination of rV is given by Eq. (19.3):
rV = pi — A(ori pini). (19.3)
Thus for a required yield of 12% and with 2 = 3 and a correlation coefficient
with the market of 1 for firm B (whose o- = 30), the certainty equivalent rV = 100 —
[4(30 x 1)] = 90. This is substituted into Eq. (19.1) using rV instead of Thus
lhs = 1.1 x (44/1.12) = 43.21.
For the right-hand side we first calculate the standardized variable, which is
(44 — 90)/30 = —1.53, F(— 1.53) = 1 — F(1.53) = 1 — .9370 = .0630.
706 MERGERS, RESTRUCTURING, AND CORPORATE CONTROL: THEORY
From R. C. Stapleton, "Mergers, Debt Capacity and the Valuation of Corporate Loans," reprinted by
permission of the publisher, from Mergers and Acquisitions, edited by Michael Keenan and Lawrence J.
White, Lexington Books, D. C. Heath & Co., Lexington, Mass. Copyright © 1982, D.C. Heath & Co.
Firm BD
Firm B Firm D p = 0.46 p =1
Mean p 100 100 200 200
Standard deviation a 30 40 60 70
Correlation with market io n, 0.75 0.75 0.875 0.75
i percent YB YD YB + YD YBD YBD
Debt capacity 12 47 0 47 91 56
Debt capacity 14 60 22 82 116 92
Debt capacity 16 68 41 109 133 112
From R. C. Stapleton, "Mergers, Debt Capacity and the Valuation of Corporate Loans," reprinted by
permission of the publisher, from Mergers and Acquisitions, edited by Michael Keenan and Lawrence J.
White, Lexington Books, D. C. Heath & Co., Lexington, Mass. Copyright © 1982, D. C. Heath & Co.
CONGLOMERATE MERGERS 707
.46 the increase in debt capacity is larger in both absolute and relative terms, even
though the initial debt capacities were smaller. This is a logical result because with
risk aversion and the reduction in risk through merger, the impact on debt capacity
is greater.
Thus the main results of the Stapleton formulation are quite strong. Using the
OPM, the effects of merger on debt capacity are shown to be positive even when
earnings of the component firms are perfectly correlated. By comparison, the Lewellen
definition of debt capacity understates the effects of mergers.
Although debt capacity is increased, we should recall the results of Galai-Masulis
and of Shastri. Their results suggest that in general the increase in debt capacity will
benefit the creditors but have a negative effect on the value of the common stock.
However, if the increase in debt capacity is utilized after merger, the value of the
common stock may be increased owing to tax savings on interest payments.
which complete data of monthly prices and dividends were available for the 10 years
from 1964 through 1973. Similar data were available for 35 nonconglomerate firms
that were part of a larger sample (matched by major industry) as developed by
Melicher and Rush [1973]. Standard and Poor's Composite Stock Price Index was
used as a surrogate for the overall stock market. For comparisons of managed port-
folios, they compiled data for 104 mutual funds and also for 17 closed-end investment
companies.
The risk-adjusted performance of conglomerates was found to be significantly
better than that of the mutual funds. The higher-beta conglomerates performed better
during the rising market but less well during the flat market. However, on theoretical
grounds, the risk-adjusted performance measure should not show better performance
for higher-risk securities during an up-market or worse performance during a down-
market situation. As discussed by Friend and Blume [1970] and Black, Jensen, and
Scholes [1972], a possible reason for the early differentially better performance of
conglomerates is that the CAPM from which the risk-adjusted performance measures
are derived is misspecified. An alternative explanation is expectation errors coupled
with institutional changes. The attitudes toward conglomerates changed considerably
over time, exhibiting overoptimism about their potential during 1964-1968. During
the second period, 1969-1973, some unfavorable institutional changes took place.
Accounting rules were changed, adverse tax treatment was legislated, and antitrust
suits were filed by the. Department of Justice. Also, the aerospace industry, which
spawned many of the conglomerates as a form of defensive diversification, suffered
from excess capacity and sharp product shifts with the escalation of the Vietnam war.
Tests of operating effectiveness suggest an initial overoptimism about the potentials
for management performance of conglomerates. Following 1969, conglomerates began
to be viewed with considerable pessimism. These expectation changes are consistent
with the risk-adjusted performance exhibited by the conglomerates.
As experience with conglomerates grew, investors were able to develop a more
dependable basis for forming expectations with respect to their performance. We
would expect conglomerates to continue to exhibit high betas because of the charac-
teristics of the product markets of conglomerate firms. Risk-adjusted measures of con-
glomerate performance are not likely to be significantly different from those of other
firms and portfolios. Further tests of merger performance, using residual analysis, are
covered in the following chapter.
SUMMARY
Most generally mergers form one economic unit from two or more previous ones.
In recent years increased use has been made of tender offers in which the bidder
makes an offer directly to the stockholders of the firm it is seeking to control (the
target). After obtaining control the bidder may merge the companies. Other recent
developments include increased emphasis on so-called reverse mergers (selloffs), cor-
porate control activities (takeover defenses), and changes in ownership structure.
Alternative theories of mergers have been formulated. Efficiency arguments in-
clude both the differential efficiency of acquiring firms over targets, and just plain
SUMMARY 709
those future incomes. Thus a capital gains tax may be substituted for a tax on ordinary
income. In addition, if stock is received in payment, realization of the capital gain
can be deferred to a time period selected by the sellers.
Other types of restructuring activities must also be assumed to have some ra-
tional explanation. Divestitures may create value by transferring assets to a higher-
valued use. Corporate control and changes in structure may improve managerial
efficiency by more directly linking accountability, performance evaluation measures,
and compensation.
Conglomerate mergers have been an area of particular study because they were
the only type of merger permitted after the tightening of the legal rules in 1950 until
new theories and research findings influenced regulatory agencies and the courts in
the 1970s to begin to depart from a strict concentration doctrine (market share is a
measure of monopoly power) to consider international competition and industry
dynamics in judging the effects of mergers. One potential source of synergy in con-
glomerate mergers was the application of the general management functions of plan-
ning and control to a wide range of companies, particularly those companies where
these general functions had not been performed well. The pure financial theory of
conglomerate mergers emphasized reduction in the probability of bankruptcy. If
bankruptcy costs are large, debt capacity may be increased by conglomerate mergers.
However, this simply transfers wealth from shareholders to debt holders in the firm.
Under alternative assumptions of the relative size of bankruptcy costs, the degree
of interdependence or correlation between cash flows of the constituent companies,
and different investor attitudes toward risk, a wide range of alternative models can
be developed. These models predict the resulting effects on debt capacity.
In the option pricing model framework, the variance of the combined firm is
decreased, which reduces the value of the combined stock and increases the value of
the combined debt. Assuming that the value of the combined firm is still the sum of
the individual constitutents, the transfer from owners to creditors can be redressed
by increasing the debt ratio. These effects are more complex than originally set forth
because of variations in the relationships among the variances and leverage ratios
of the individual firms as well as in the maturity patterns of their debt structures.
Another explanation for conglomerate mergers is that they represent an out-
growth of the computer age. Improved information systems have expanded the ge-
neric managerial capabilities of some firms, leading them to view conglomerate
mergers as a natural extension of their existing capabilities. Studies of market rela-
tionships have found that the performance of conglomerate firms is not significantly
different from the performance of other firms on average over extended periods of
time.
PROBLEM SET
19.1 Discuss the assumptions and implications of the proposition that value is conserved
(value additivity obtains) under the addition of income streams (mergers).
19.2 Summarize the sources of synergy or operating gains from mergers that have been pre-
sented in the literature. Evaluate the validity of the arguments for synergy.
PROBLEM SET 711
19.3 Explain and illustrate how differential price/earnings ratios reflected in the terms of merg-
ers result in increases or decreases in earnings per share of the merging firms. Do such effects
on price/earnings ratios also have effects on the valuation of the firms resulting from mergers?
19.4 Explain and evaluate the managerialism theory of conglomerate mergers set forth by
Mueller.
19.5 Explain the pure financial theories of conglomerate firms and evaluate their validity.
For Problems 19.6 and 19.7 assume the following:
i) We are dealing with a world where there are no taxes.
ii) The changes in the parameters affecting value are unanticipated; therefore redistribution
effects are possible.
iii) Firms A and B initially have the following parameters:
6
A= = .2 Instantaneous standard deviation
T A = T B = 4 years Maturity of debt
VA = VB = $2000 Value of the firm, V = B + S
R1 = .06 Risk-free rate
D A = DB = $1000 Face value of debt
19.6 The correlation between the cash flows of firms A and B is .6. If they merge, the resultant
firm will be worth $4000 = VA VB, but its new instantaneous variance will be
u AB = . 179.
What will the market value of debt and equity in the merged firm be? If there were no other
merger effects, would shareholders agree to the merger?
19.7 Given the results of Problem 19.6, suppose that the merged firm has 1000 shares out-
standing. Furthermore, suppose that the shareholders decide to issue $1000 of new debt (which
is not subordinate to outstanding debt), maturing in four years, and invest the proceeds in
marketable securities, so that the new value of the merged firm is $5000. What will be the
new price per share? Assume the merged firm's instantaneous variance is unchanged by this
investment.
19.8 You are given the following information:
Firm A Firm B
Value prior to merger $1000 $1000
Face value of debt 500 500
In addition, the value of equity for firm A equals the value of equity for firm B, and the vari-
ance of returns for firm A and firm B are also equal. Using a risk-free rate of 8%, an appropriate
time horizon of five years, and a variance for each firm of 10%, apply the OPM to calculate
712 MERGERS, RESTRUCTURING, AND CORPORATE CONTROL: THEORY
the value of equity of the two firms before the merger. Under the further assumption that the
correlation between the percentage returns on firms A and B is zero, calculate the value of
equity and the value of debt of the merged firm, using the OPM.
a) How does the new market value of equity and debt of the merged firm compare with the
sum of the values of equity and debt of the constituent firms that combined in the merger?
b) How much additional debt would the merged firm have to issue to restore equity holders
to their original position?
19.9 Empirical studies have established that the betas of conglomerate firms have been signif-
icantly above 1. What does this imply about diversification as a strong motive for conglomerate
mergers?
19.10 Over a long period of time would you expect the risk-adjusted performance of conglom-
erate firms to be significantly different from the risk-adjusted performance of a broad market
index? Explain.
19.11 Galai and Masulis argue that if two firms merge and thus decrease the probability of
default on their debt along the lines of Lewellen's scenario, then the stockholders are actually
hurt, since they have assumed some of the risk previously borne by the bondholders. Why
might nonowner managers of a firm be motivated to transfer risk from bondholders to stock-
holders in this manner?
REFERENCES*
Alchian, A., and H. Demsetz, "Production, Information Costs, and Economic Organizations,"
American Economic Review, December 1972, 777-795.
Arrow, K. J., "Vertical Integration and Communication," Bell Journal of Economics, Spring
1975, 173-183.
Beckenstein, A. R., "Merger Activity and Merger Theories: An Empirical Investigation," Anti-
trust Bulletin, Spring 1979, 105-128.
Beman, L., "What We Learned from the Great Merger Frenzy," Fortune, April 1973, 70ff.
Black, F.; M. Jensen; and M. Scholes, "The Capital Asset Pricing Model: Some Empirical
Tests," in Jensen, ed., Studies in the Theory of Capital Markets. Praeger, New York, 1972.
Blume, M., "Portfolio Theory: A Step toward Its Practical Application," Journal of Business,
April 1970, 152-173.
Bock, B., Statistical Games and the "200 Largest" Industrials: 1954 and 1968. The Conference
Board, Inc., New York, 1970.
Bradley, M.; A. Desai; and E. H. Kim, "The Rationale behind Interfirm Tender Offers: Infor-
mation or Synergy?" Journal of Financial Economics, April 1983.
Brennan, M., "The Pricing of Contingent Claims in Discrete Time Models," Journal of Finance,
March 1979, 53-68.
Celler, E., Investigation of Conglomerate Corporations Hearings before the Antitrust Subcom-
mittee of the Committee on the Judiciary, House of Representatives, 1969-1970, Parts
1-7. U.S. Government Printing Office, Washington, D.C., 1971.
* Since Chapters 19 and 20 are closely related in subject matter, the reader is urged to consult the reference
sections for both chapters.
REFERENCES 713
Lewellen, W. G., "A Pure Financial Rationale for the Conglomerate Merger," Jou]nal of
Finance, May =97=, P<=-P4P.
, and B. Huntsman, "Managerial Pay and Corporate Performance," Ame]ican Economic
Review, Septemîer =970, 7=0-7<0.
Lintner, J., "Conglomerate and Vertical Responses to Market Imperfection: Expectations,
Mergers and Equiliîrium in Purely Competitive Markets," Ame]ican Economic Review,
May =97=, =0=-===.
Livermore, S., "The Success of Industrial Mergers," Qua]te]ly Jou]nal of Economics, Novemîer
=93P, 63-96.
Lynch, H. H., Financial Pe]fo]mance of Conglome]ates. Harvard Graduate School of Business
Administration, Boston, =97=.
Manne, H. G., "Mergers and the Market for Corporate Control," Jou]nal of Political Economy,
April =96P, ==0-=<0.
Markham, J. W., Conglome]ate Ente]p]ises and Public Policy. Harvard Graduate School of
Business Administration, Boston, =973a.
, "Market Structure and Decision-Making in the Large Diversified Firm," in Weston
and Ornstein, eds., The Impact of La]ge Fi]ms on the U.S. Economy, Chapter =4. D. C.
Heath, Lexington, Mass., =973.
Mason, R. H., and M. B. Goudzwaard, "Performance of Conglomerate Firms: A Portfolio
Approach," Jou]nal of Finance, March =976, 39-48.
Melicher, R. W., and T. H. Harter, "Stock Price Movements of Firms Engaged in Large
Acquisition," Jou]nal of Financial and Quantitative Analysis, March =97<, =469-=47P.
Melicher, R. W., and D. F. Rush, "The Performance of Conglomerate Firms: Recent Risk and
Return Experience," Jou]nal of Finance, May =973, 38=-388.
, "Evidence on the Acquisition-Related Performance of Conglomerate Firms," Jou]nal
of Finance, March =974, =4=-=49.
Merton, R., and M. Suîrahmanyan, "The Optimality of a Competitive Stock Market," Bell
Jou]nal of Economics, Spring =974, =4P-=70.
Modigliani, F., and M. H. Miller, "The Cost of Capital, Corporation Finance, and the Theory
of Investment," Ame]ican Economic Review, June =9P8, <6=-<97.
Mossin, J., Theo]y of Financial Ma]kets. Prentice-Hall, Englewood Cliffs, N.J., =973.
Mueller, D. C., "A Theory of Conglomerate Mergers," Qua]te]ly Jou]nal of Economics, No-
vemîer =969, 643-6P9. (See also comment îy Dennis E. Logue and Philippe A. Naert,
Novemîer =970, 663-667; comment îy David R. Kanerschen, 668-673; and reply îy
Dennis C. Mueller, 674-679.)
, "The Effects of Conglomerate Mergers," Jou]nal of Banking and Finance, =977, 3=P-
347.
Myers, S. C., "Procedures for Capital Budgeting under Uncertainty," Indust]ial Management
Review, Spring =968, =-=9.
Nielsen, N. C., "The Firm as an Intermediary îetween Consumers and Production Functions
under Uncertainty," Ph.D. dissertation, Graduate School of Business, Stanford University,
Palo Alto, Calif., =974.
Reid, S. R., Me]ge]s, Manage]s and the Economy. McGraw-Hill, New York, =968.
Reinhardt, U. E., Me]ge]s and Consolidations- A Co]po]ate Financial App]oach. General Learn-
ing Press, Morristown, N.J., =97<.
REFERENCES 715
Rohatyn, F. G., "Needed: Restraints on the Takeover Mania," Challenge, May—June 1986,
30-34.
Schall, L. D., "Asset Valuation, Firm Investment, and Firm Diversification," Jou]nal of Busi-
ness, January 1972, 11-28.
Scott, J. H., Jr., "On the Theory of Conglomerate Mergers," Jou]nal of Finance, September
1977, 1235-1250.
Shastri, K., "Valuing Corporate Securities: Some Effects of Mergers by Exchange Offers,"
Working Paper #S17, University of Pittsburgh, revised January 1982.
Shick, R. A., "The Analysis of Mergers and Acquisitions," Jou]nal of Finance, May 1972,
495-502.
Smith, K. V., and J. F. Weston, "Further Evaluation of Conglomerate Performance," Jou]nal
of Business Resea]ch, March 1977, 5-14.
Stapleton, R. C., "Mergers, Debt Capacity, and the Valuation of Corporate Loans," in Keenan
and White, eds., Me]ge]s and Acquisitions, Chapter 2. D. C. Heath, Lexington, Mass.,
1982.
Summer, C., St]ategic Behavio] in Business and Gove]nment. Little, Brown and Company,
Boston, 1980.
U.S., Federal Trade Commission, Staff Report: Economic Repo]t on Conglome]ate Me]ge] Pe]-
fo]mance. Government Printing Office, Washington, D.C., November 1972.
, Staff Report: Economic Repo]t on Co]po]ate Me]ge]s. Government Printing Office,
Washington, D.C., August 28, 1969.
, Statistical Repo]t on Me]ge]s and Acquisitions. Bureau of Economics, November 1977.
Westerfield, R., "A Note on the Measurement of Conglomerate Diversification," Jou]nal of
Finance, September 1970, 909-914.
Weston, J. F., "The Determination of Share Exchange Ratios," in Alberts and Segal, eds., The
Co]po]ate Me]ge], 117-138. University of Chicago Press, 1966.
, "Discussion," Ame]ican Economic Review, May 1971, 125-127.
, "ROI Planning and Control," Business Ho]izons, August 1972, 35-42.
, Me]ge]s and Economic Efficiency, Proceedings of a Workshop and Supplementary
Papers, Vol. 1. U.S. Government Printing Office, Washington, D.C., November 1980.
, Me]ge]s and Economic Efficiency, Industrial Concentration, Mergers and Growth,
Vol. 2. U.S. Government Printing Office, Washington D.C., June 1981.
, and S. K. Mansinghka, "Tests of the Efficiency Performance of Conglomerate Firms,"
Jou]nal of Finance, September 1971, 919-936.
Weston, J. F., and E. M. Rice, "Discussion," Jou]nal of Finance, May 1976, 743-747.
Weston, J. F.; K. V. Smith; and R. E. Shrieves, "Conglomerate Performance Using the Capital
Asset Pricing Model," Review of Economics and Statistics, November 1972, 357-363.
Williamson, 0. E., Co]po]ate Cont]ol and Business Behavio]. Prentice-Hall, Englewood Cliffs,
N.J., 1970.
, "The Vertical Integration of Production: Market Failure Considerations," Ame]ican
Economic Review, May 1971, 112-123.
20
Co]po]ate takeove]s a]e the logical outg]owth of competitive st]uggles
in the f]ee ma]ket. . . . Me]ge]s and acquisitions make sense because
they inc]ease the value of the sha]es held in the ta]get company. . . .
The takeove] ma]ket also p]ovides a unique, powe]ful, and impe]sonal
mechanism to accomplish the majo] ]est]uctu]ing and ]edeployment of
assets continually ]equi]ed by changes in technology and consume]
p]efe]ences. . . . Scientific evidence indicates that activities in the
ma]ket fo] co]po]ate cont]ol almost unifo]mly inc]ease efficiency and
sha]eholde]s' wealth. êet the]e is an almost continuous flow of unfavo]able
publicity and calls fo] ]egulation and ]est]iction of unf]iendly takeove]s.
Many of these appeals a]ise f]om manage]s who want p]otection f]om
competition fo] thei] jobs and othe]s who desi]e mo]e cont]ols on
co]po]ations.
Reprinted îy permission of the Harvard Business Review, "Takeovers:
Folklore and Science" îy Michael C. Jensen (Nov./Dec. =984).
Copyright © =984 îy the President and Fellows of Harvard College;
all rights reserved.
Me]ge]s and
Rest]uctu]ing- Tests and
Applications
The preceding chapter dealt with theories of mergers and restructuring and descriîed
some empirical tests of conglomerate merger performance. In the first part of the
present chapter we review tests of the theories îy a consideration of empirical studies
in four areas: (A) tests of merger and tender offer studies; (B) studies of antitrust cases;
(C) studies of corporate governance; and (D) studies of other forms of restructuring.
We then draw (E) generalizations from the studies. In the second part of the chapter
we discuss managerial aspects of mergers, covering (F) terms of mergers and (G)
managerial policies in a valuation framework.
Recall that in Chapter ==, descriîing tests of the efficient market hypothesis,
three îasic types of empirical models were employed. These were Eqs. (7.3<), (7.36),
and (==.=):
7=6
TESTS OF MERGER AND TENDER OFFER RETURNS 717
R j, = R ft + [Rmt
R ft ]fli + ei„ (7.32)
Rft = 1.1ot + 1'1113 + (7.36)
Rit = ai biR mt eft . (11.1)
Equation (7.32) uses the capital asset pricing theory. Equation (7.36) is the
empirical market line derived from the CAPM. However, it does not require the
intercept term to equal the risk-free rate. Equation (11.1) is the market model. Under
appropriate definitions and assumptions, one can move from one model to another
[Ellert, 1975, Appendix A]. However, as noted in Chapter 11, in empirical work the
implications are different.
For example, the market model employs time series analysis, developing pa-
rameters for periods before and after a time interval influenced by the "event"
under analysis. The meaning of the intercept term in the model is not defined, and
the nature of the event may influence its measurement. Thus if firms that have not
been "performing up to their potential" are the ones that become target or acquired
firms, the intercept term will be negative. If firms that have had excellent and improving
performance in managing assets become acquiring firms, their intercept term will be
positive. In the calculation of the residuals, the intercept for the control period is
deducted and hence this form of selection bias will influence the results.
The CAPM form of the model involves the use of a risk-free return. Empirical
tests have yielded estimates of the intercept term that are systematically greater than
the return on a riskless asset.
Since the implicit empirical assumptions of the alternative models differ, the
results of studies using different models require interpretation related to the char-
acteristics of the model employed. In Chapter 11 the three models were used to
test for market efficiency. In the present chapter the models are used to test whether
mergers or tender offers and other unique events produce temporary "abnormal"
returns. Our discussion of the empirical studies of merger performance will take into
account the implications of the particular statistical methodology employed.
Mergers and tender offers achieve similar ends; successful tender offers are followed
by formal merger in more than two thirds of the cases. However, the means and
motives may differ, resulting in unequal impacts on the shareholders of the firms
involved. In tender offers, there is less likely to have been previous negotiations
between the parties. In general there is likely to be less information leakage before
the formal announcement. There is some plausibility also that synergy is more likely
to be involved in mergers where previous discussions about fitting two companies
together take place. In tender offers the likelihood is greater that the bidder sees an
opportunity for improving the management of the target firm or that its shares are
undervalued for one of a number of possible reasons.
718 MERGERS AND RESTRUCTURING: TESTS AND APPLICATIONS
1. Early Studies
Early empirical studies of mergers (pre-1973) used comparative studies of firm
performance to test for synergy in mergers. Kelly [1967] was the first to study merger
profitability using measures including security price changes. His sample consisted
of 42 firms matched in 21 pairs of one merging and one nonmerging firm. He com-
pared pre- and postmerger performance based on five measures of profitability (per-
centage changes in stock price, P/E ratio, EPS, sales per share, and profit margin)
and concluded that mergers had little impact on acquiring firm shareholders.
Hogarty [1970] constructed indexes of investment performance based on changes
in stock prices. His sample consisted of 43 acquiring firms whose indexes were com-
pared with similarly constructed indexes of their respective industries. He concluded
that mergers resulted in negative synergy; investment performance of acquiring firms
was 5% less (significant at a 10% level) than their industries' performance.
Lev and Mandelker [1972] faced a similar problem of selecting a standard against
which to compare merging firms' performance. Measuring profitability by the annual
stock market return on each of 69 acquiring firms, they calculated the average return
for the five premerger and five postmerger years for each firm; they then deducted
the respective pre- and postmerger average returns of 69 matching firms to control
for factors (other than the merger) presumed to identically affect each pair of firms.
They found that the market value of acquiring firms rose an average of 5.6% (significant
at the 10% level) more than that of the matching control firms.
To this point the evidence on security price changes resulting from mergers is
conflicting and is confounded by legitimate criticisms of sample size, methodologies,
and control devices used to screen out other influences. It is in this setting that the
studies of Halpern [1973] and Mandelker [1974] appear. These two analyses begin
the use of asset pricing models.
2. Halpern
Halpern [1973] attempts to directly measure buyer and seller premiums in mergers
in a sample of approximately 75 acquisitions. Basically his method is to adjust the
observed market prices of acquiring and acquired firms for general market variations
during the period when merger information affects their share prices: the price change
that remains unexplained by market variations is that attributable to the merger.
Two estimates are required to make the proposed share price adjustments and
measurements. First, Halpern needs a base period, the interval before the announce-
ment date during which merger information is reflected in the stock prices. To
determine this, he employs the "residual technique" developed by Fama, Fisher,
Jensen, and Roll and runs the following time series regression:
j R,,t = ct i + /31;Rm.,,. + / 21R1, t. + (20.1)
where ;R,,, is the price relative for company i in industry j during the month t,
is the industry price relative, and iR m,,*, the market price relative. The regression
yields estimates of cici, )6' k , and Substituting these, he rewrites the regression equation
in the form
ei,t = jRi,t } jRm,tûRI,tû $2} — (}‘ti• (20.2)
TESTS OF MERGER AND TENDER OFFER RETURNS 719
Halpern then notes that if all of company i's price relative in month t could be
explained by industry or market price relatives, the value of the merger, as measured
by 8,, t , would be zero. Thus a nonzero ei,, indicates that firm i's share price is not
entirely accounted for by "normal" factors. Since mergers are unusual events, we
would expect the estimated residuals to display unusual behavior during the merger
adjustment period. A base date can then be chosen as that date before which
"abnormal" residuals were observed. Because of extraneous influences, cross-sectional
residuals (over merging firms) were calculated over time relative to the announcement
date.
Halpern's analysis indicates that, on the average, merger information is available
for seven months before the announcement date. From the twenty-third until the
eighth month prior to announcements, the cumulative average residuals are randomly
increasing and decreasing. From the seventh month onward, they increase steadily.
Average residual values are small and vary in size until the large positive value is
encountered in the seventh month. Also, a large proportion of positive residuals in
the seventh month seems to strengthen this result.
The second element of Halpern's analysis is calculating adjusted security prices.
Excluding data from the 12 months prior to the merger, Halpern runs the following
equation for each firm using five years of monthly observations:
Rk, = + + /3b2 R,,,
b
+ E. (20.3)
From the estimates ot lab and il2 and the actual price relative in the market and
industry for the adjustment interval, Halpern calculates the unbiased estimates of the
expected price relative for each firm. Multiplying this relative by the base period
price and adjusting for dividends paid during the adjustment period, he then obtains
an unbiased estimate of the adjusted price. After calculating the gains to "buyers"
and "sellers," the firms were classified as "larger" or "smaller" on the basis of their
equity value at the base date. Halpern finds that the mean gain prior to dividend ad-
justment of larger firms exceeded those of smaller firms by a factor of 4. After the
dividend adjustment, the gains were smaller for both firm types and approximately
equal in absolute amount. Adjustments also had a tendency to make negative gains
less negative and at times turned negative "gains" into positive ones.
Subsequently Halpern calculates price premiums, the gain relative to a base price,
for both acquiring and acquired firms. While the premium accruing to smaller firms
was significantly greater than zero (at 5%), the premium accruing to larger firms was
not. His results suggest synergy or improvement in the performance of the smaller
firms that is reflected in prices paid by acquirers.
3. Mandelker
Mandelker [1974] used the Fama and MacBeth [1973] methodology to examine
the testable implications of what we have termed the empirical market line,
Rit —1.1ot + 0itflit + eft, (7.36)
presented in Chapter 7.
720 MERGERS AND RESTRUCTURING: TESTS AND APPLICATIONS
First, Fama and MacBeth estimate betas with seven years of monthly data for
individual securities using the regression analog of our Eq. (7.32):
Rit — R ft = (R,„ — R ft )fl + ei„ (7.32)
which is the ex post form of the CAPM equation. The individual betas are ranked
and placed in 20 portfolios with maximum dispersion of systematic risk in the attempt
to minimize the measurement error in the beta estimates.
Second, they use the next five years to recalculate the betas and average them
to obtain the portfolio betas. The betas for individual securities are updated every
10 months on the basis of the return data for the preceding five years.
Third, using monthly data for the next four years, they run cross-sectional re-
gressions for each month across the 20 portfolios utilizing Eq. (7.36), the empirical
market line. These provide estimates of y o and y 1, which have been used as measures
of the empirical market line parameters in a number of other studies.
Mandelker used the Fama and MacBeth procedures, adding two additional steps
for his study of mergers.
Fourth, he estimated the betas for individual firms involved in mergers, using
the ex post form of the asset pricing relation, Eq. (7.32). The time period measured
covered months prior to the merger as well as months following it.
Fifth, he calculated residuals for each month, using the gamma values from Eq.
(7.36) and utilizing the values of returns and betas calculated in the fourth step.
Mandelker tested two hypotheses. One was that acquisitions took place in a
market under conditions of perfect competition. The other was the hypothesis of ef-
ficient capital markets with respect to information on acquisitions. The stockhold-
ers of the acquired firms received cumulative average residuals that were positive, in-
dicating that they earned abnormal gains from the mergers. This suggests that the
acquired firms may have had unique resources whose values are realized to a greater
degree by mergers. Alternatively, the acquired firms may have been operating at
below their optimal levels of efficiency, and the mergers were seen as increasing the
effectiveness of their operations. The possible benefits to be derived from the acquired
firms are apparently perceived by a number of potential acquirers. The competition
in the market for acquisitions results in competitive prices for the acquired firms.
The acquiring firms appear to operate in a competitive market so that the prices they
pay for the acquired firm's stock result in normal returns on the acquisitions. The ac-
quiring firms earn a rate of return equal to other investment or production activities
of similar risks. The average residuals for the acquiring firms are generally positive
but not statistically significant. This finding controverts the argument that acquiring
firms overpay and lose from mergers.
With respect to the hypothesis of efficient capital markets, Mandelker's findings
are consistent with the view that the stock market operates efficiently with respect
to information on mergers. The price movements that take place at the time of
the merger announcement and even before reflect all valuable information about the
merger preceding the effective date of the transaction. The stock prices of the constit-
uent firms at the date when the merger is consummated reflect the economic gains
expected. The stock prices of the merged firms do not undergo postmerger adjustments.
TESTS OF MERGER AND TENDER OFFER RETURNS 721
While significant changes in betas were observed, the rates of return adjusted efficiently
to the changes in risk. Stockholders were not misled by accounting manipulations in
mergers or by the artificial increase in earnings per share resulting from the differential
price/earnings ratio game played by acquiring firms. The views that mergers reflect
the desires of managers to control larger firms and their emphasis on growth maximi-
zation imply losses to the acquiring stockholders. But the finding of positive average
residuals for the acquiring firms is inconsistent with the managerialism hypothesis.
4. Ellert
Ellert [1975, 1976] also employs the Fama-MacBeth methodology, updating their
estimates of T o,. Ellert's studies, like the earlier findings of Halpern [1973] and
Mandelker [1974], indicate that the impact on the market prices of merging firms
takes place 7 to 12 months prior to the actual merger. The announcement necessarily
precedes the merger, and there are leaks even before the public announcement. For
acquiring firms, his evidence indicates that while the cumulative average residual
(CAR) is generally positive during this period, it is either not significant or the
magnitude of the change in CAR during this period is small. This evidence is incon-
sistent with the managerial theory and its related growth maximization prediction. It
is also inconsistent with the monopolistic exploitation theory, at least in providing
monopoly gains to acquiring firms.
Both Mandelker and Ellert find that very substantial increases in the CAR of
acquiring firms take place during the period from four to eight years prior to the
merger activity. This is consistent with the hypothesis set forth by both Mandelker
and Ellert that the differentially higher efficiency of acquiring firms prior to mergers
leads to their subsequent expansion both externally and internally. With respect to
acquired firms, both Mandelker and Ellert find that their CARs are significantly nega-
tive in the years and months running up to the period when information about their
upcoming acquisition by others becomes available. In the subsequent 7 to 12 months
through the actual merger date, the CARs turn strongly positive and are highly sig-
nificant by statistical tests.
The finding of positive residuals for the acquired firms in the months preceding
the merger is consistent with the theory of monopoly control with all the gains going
to the acquired firms. However, the evidence of negative CARs in the period running
up to the time information on the mergers becomes available is consistent with an
alternative explanation. The inefficient utilization of economic resources by the man-
agement of the firms prior to the merger leads to their acquisition by firms with a
record of above-average performance. Hence the evidence leans in the direction of
efficiency and/or synergy as the explanation rather than the managerial or monopoly
theories.
If differential efficiency between acquired and acquiring firms were the explana-
tion for mergers, it is likely that there would be a number of acquiring firms bidding
for an individual acquired firm. Competition among acquiring firms would, on the
average, eliminate abnormal gains to the acquiring firm from the merger activity. This
is consistent with the general findings of Mandelker and Ellert. If the acquiring firms
722 MERGERS AND RESTRUCTURING: TESTS AND APPLICATIONS
had a unique synergistic relationship with an individual acquired firm, the gains from
the merger would be attributable to both and could not be allocated individually.
This is consistent with Halpern's finding that the absolute dollar amount of the
gains is equally divided between acquiring and acquired firms. However, Halpern,
Mandelker, and Ellert find that while the absolute gains of acquired firms were sta-
tistically significant, neither the percentage premium returns nor CARs for the ac-
quiring firms were statistically significant.
will be new suitors. Hence the near-term performance for both acquired and acquiring
firms under the internal efficiency or undervalued assets theory would be unaffected
by the unsuccessful tender offer. If the acquiring firm had spent substantial sums of
money in the negotiations, this might cause a decline in the CARs because such
investments in the effort to acquire other firms would not come to fruition. The
Dodd-Ruback data for unsuccessful tenders in the near term following the event date
show zero CARs for unsuccessful bidders as well as zero CARs for the target firms.
This pattern is consistent with the internal efficiency or undervalued assets theory
but not with the other theories.
Dodd and Ruback point out that a very substantial proportion of unsuccessful
target firms merge within five years of the aborted tender offer [1977, 370]. This leads
them to state that their conclusion that the results support the efficiency hypothesis
is qualified by the possibility that the subsequent mergers may have been anticipated
by the market. If so, the nondecline in the CAR after a tender is unsuccessful could
be due to expectations of the future monopoly or synergistic gains from mergers
[1977, 370]. However, since their sample period begins in 1958, and since most of the
observations in their sample occur in the 1960s and 1970s, the monopoly explanation
does not apply. The Celler-Kefauver Amendment, which greatly strengthened the
Clayton Act of 1914, was enacted in 1950 and greatly reduced the relative number of
horizontal and vertical mergers [Stigler, 1966]. Between 1950 and 1980, which en-
compasses the data for their study, the largest proportion of mergers had been con-
glomerate in nature where the issue has not been monopoly in individual markets
[see also FTC Staff Report, 1972]. Therefore the Dodd and Ruback findings are con-
sistent with the internal efficiency or undervalued asset theory as well as with the
synergy theory but not with the monopoly power or managerialism theories.
For bidder firms the cumulative average residual rose to about 107 during the
period (— 28, — 11). Another gain of 6 percentage points was achieved by shareholders
of bidder firms during the period ( — 10, — 1). An additional 5% abnormal gain was
received during the month in which the tender offer was announced.
7. Langetieg
Langetieg [1978] reexamined 149 mergers using alternative performance indices
as well as a matched nonmerging control group of firms and found that the results
using alternative performance tests were generally consistent with the results of Man-
delker. But the use of the control group changes the interpretation somewhat. For
example, the acquired firms had a significantly negative cumulative average residual
of 12.6% over the time interval (-72, —19) months and a significantly positive stock-
holder gain of 10.6% over the time interval ( — 6, — 1). But the nonmerging control
firms also exhibited negative excess returns in the time interval ( — 72, —19). In two
of the three paired-difference tests, Langetieg concludes that some external influence
affected both the merging firms and the control firms in a similar way over the similar
time periods. For the shareholders of acquiring firms, if the starting point is viewed
as 18 months before the merger, the net benefit is 6.1%, which is statistically signif-
icant. But the stockholder gain for acquiring companies is substantially less than that
for acquired firms.
Langetieg also found that the postmerger performance of a consolidated firm
over the period (1, 24) was a negative 12.9 percent. However, in the paired-difference
analysis, no significant cumulative difference existed. The postmerger performance
of the consolidated companies reflected influences that affected similar nonmerging
companies in the same way.
8. Firth
Firth's [1979] study of performance covers data for companies in the United
Kingdom for the period 1972-1974. As in U.S. studies, the residuals for the acquired
companies were slightly negative in the period ( —36, —12) months. The residuals
had no strong movement until 1 month before the takeover announcement. Halpern
and Mandelker reported that generally 7 months elapsed between the beginning of
merger negotiations and the merger announcement. Tender offers, however, are less
often associated with communications between the companies involved until very
shortly before the public offer announcement.
The CARs in Firth's sample reached 38% for acquired firms within 2 months
of the bid announcement. For the acquiring firms the CARs were slightly positive
before the announcement but turned to a negative 4 to 5% within 24 months of the
tender offer. In absolute amounts the shareholders of acquired firms gained about
650 million British pounds, whereas the shareholders of acquiring firms lost about
the same amount. This contrasts with the earlier Halpern study for the United States
in which there was a total gain of about $28 million that was split almost equally
among the acquired and acquiring firms. Firth comments that previous research for
the United Kingdom showed that acquired firms on average had poor profitability
TESTS OF MERGER AND TENDER OFFER RETURNS 725
records but that the recovery potential was offset by the premiums paid by acquirers.
He observes that the relative incidence of mergers in the United Kingdom has been
higher than in the United States and suggests that acquired firms had the benefit of
a stronger sellers' market for mergers. Also, Firth's data covered 1972-1974, whereas
the U.S. studies covered longer periods starting in the 1950s. Analysis of mergers
and tender offers in the United States for the years following 1970 might give results
similar to those found by Firth for the United Kingdom.
9. Bradley
Bradley [1980] studied 258 cash tender offers that took place during the period
July 1962 through December 1977. For calculating the residuals he uses the abnormal
price indices compiled by the Center for Research in Security Prices (CRSP) of the
University of Chicago. Their procedure is to group securities into 10 equal control
portfolios according to their estimated risk. Using the CAPM, an estimate of the daily
excess return for an individual security is the difference between the realized return
to the security less its control portfolio's return.
For the 161 successful tender offers the mean premium is 49%. The postexecution
market price of a target share is 36% higher than its preannouncement level. Hence
there is a loss of 13% per target share to the acquirer, and target shareholders realized
a significant 36% capital gain on the shares not purchased in the offer. Bradley com-
ments that these data do not support a corporate raiding interpretation of tender
offers. Despite the significant premium paid, the acquiring stockholders realize an
excess capital gain of 5% within 40 days of the offer with an excess capital gain of 4%
taking place within 5 trading days of the offer. Bradley comments that these results
are consistent with synergistic gains but cannot distinguish whether they come from
market power or cost reductions.
In 33 unsuccessful tender offers the postexecution price level of a target share is
67% above the preannouncement level. This increase in the value of the target shares
is higher than the rejected offer premium by 15%. The foregoing results hold for tender
offers where all shares were bid for. For fractional tender offers the target shareholders
realized a 36% capital gain relative to the preannouncement price. This mean price
index is 19% above the rejected tender offer. For the unsuccessful bidders the market
price of their shares on the fortieth day after the offer is 4% below their preannounce-
ment level. The negative returns for unsuccessful bidders are attributed by Bradley to
the search and administrative costs of conducting a tender offer. Further analysis of
unsuccessful takeovers in relation to single-bid versus multiple-bid situations would
help sharpen the analysis. Our hypothesis is that positive returns are likely to be pre-
served with multiple bids but to fall in the single-bid cases. The more fundamental
question is to explain the reasons for single bids versus multiple bids.
Bradley also calculates the results in absolute dollar terms. Over the period from
two months before the announcement to two months after the announcement the
average acquirer's equity increased by $7.7 million even though the average offer
premium paid to tendering stockholders was also $7.7 million Over the same period
the value of the outstanding shares of target firms increased an average of $31 million
726 MERGERS AND RESTRUCTURING: TESTS AND APPLICATIONS
10. Dodd
Dodd [1980] is the first of a number of studies of mergers, tenders, and other
aspects of corporate governance to employ the market model expressed in our Eq.
(11.1). Like Halpern, he uses the announcement date rather than the completion of
the merger as the "event." However, studies indicate that there is leakage of informa-
tion even before the public announcement date. Since Dodd seeks to point up the
importance of the role of the announcement date, he uses daily returns rather than
the monthly data employed in the previous merger studies.
Daily returns were studied for 151 merger proposals announced in the Wall
Street Journal from 1971 to 1977. Of these, 71 were eventually completed, and 80
were canceled by either target or bidder management. The market reaction to the
merger proposal over Day —1 and Day 0 was significantly positive for target firms
(approximately 13%) and significantly negative (though smaller) for bidder firms
(approximately — 1%).
Dodd concludes that the gains from mergers go to target firms and not to bidders.
Bidder CARs for the period 10 days before the proposal to 10 days after approval are
negative ( — 7.22%1; the bidder CAR is also negative for canceled mergers regardless
of which firm terminates negotiations.
Target shareholders earn large significant positive returns at the proposal an-
nouncement (13.43%). For completed mergers the CAR from 10 days before the pro-
posal to 10 days after approval is 33.96%. For canceled mergers the CAR from 10 days
before the proposal to 10 days after cancellation is 3.68% for the entire 80 cancella-
tions. For target-initiated cancellations the CAR is 10.95%. For bidder cancellations
the target CAR over the same period is 0.18%; i.e., returns revert to their preproposal
level. Dodd observes that since the net effect of target-canceled mergers is positive,
it cannot be concluded that managers are necessarily acting against the best interests
of shareholders when they veto a proposed merger. Our further judgment would be
that cancellations by targets are taken as a signal by the market that the bidder has
uncovered a profitable opportunity, or that there is a possibility of a return bid from
someone else. Bidder cancellations may be taken as a signal that the target does not
represent a profitable investment opportunity.
because of the resolution of uncertainty. Schipper and Thompson explain the negative
postmerger performance found in some prior studies by the fact that many of the
postmerger months in these studies fall in calendar time after the date of the regula-
tory changes.
15. Malatesta
Malatesta [1983] uses the ex post form of the CAPM model expressed in our
Eq. (7.32). However, he does not constrain the intercept term of the model to be
zero. Rather, "it enters as a component of predicted normal returns" [21, emphasis
in the original]. A distinctive variation in the Malatesta study is the calculation of
the wealth effects of mergers by the abnormal dollar returns in addition to percentage
returns. Malatesta also formulates a distinction between alternative definitions of
"event" dates. He defines t 1 as the time when investors learn that the firm will invest
resources in merger activity. A later time, t 2 , is a resolution date at which a decision
is made that the merger will take place or be abandoned. The announcement date is
taken as a proxy for t 1, but Malatesta observes that for his sample in more than half
the cases the first public announcement was that a merger had essentially been agreed
upon. For t 2 his proxy is "substantial board/management approval of the merger"
[20]. A problem here is that until stockholder approval takes place and the legal
details completed, some risk remains that a merger will not actually take place. Some
confounding of influences is involved for both reference dates.
Malatesta's findings based on percentage returns are generally consistent with
those of previous studies. However, his results for dollar abnormal returns are dif-
ferent. Portfolios of firms experiencing an information event in the same month were
formed both with equally weighted investments in each firm and with investments
weighted in proportion to the market values of equity one month prior to the event.
The dollar value cumulative excess return was calculated using the value-weighted
portfolio forecast error and the total market value of equity of the firms in the port-
folio at one month prior to the event. Monthly returns were studied over the period
—24 to + 12 months.
The first published reference to the merger in the Wall Street Journal was taken
to be a proxy for t 1. For acquiring firms, percentage returns in the 24 months prior
to t 1 were insignificantly different from zero, and postannouncement returns were
significantly negative. For acquired firms, percentage returns in months — 24 to — 4
were negative but only marginally significant; then, however, they began to rise. The
market reaction to the announcement was significantly positive, and postannounce-
ment returns were also positive.
The first published announcement of substantial board approval was taken as a
proxy for t 2 . Acquiring firms exhibited inferior performance following t, (the cumula-
tive average forecast error dropped 8% over months +2 to + 12). Acquired firms
experienced significant share price appreciation at t,. (All the above results were
essentially the same for both equally weighted and value-weighted portfolios.)
Dollar returns for acquiring firms (relative to the date of board approval) indi-
cated cumulative excess negative returns of $49 3 million per firm in the 24 months
prior to t 2, but this figure was only barely significant. Acquired firms had significant
positive abnormal returns of $19.2 million in the 4 months prior to t 2 . The effect
of mergers seems to be a net reduction in shareholder wealth; the gain to acquired
shareholders in the period — 5 to t 2 is $5 million less than the loss to acquiring
shareholders.
730 MERGERS AND RESTRUCTURING: TESTS AND APPLICATIONS
When Malatesta recalculates the postmerger results with a risk adjustment, the
negative postmerger performance of large acquirers is eliminated. The overall result
becomes insignificantly positive.
The statistical methodology takes on even greater significance for the Malatesta
study. As noted at the beginning of this chapter, the use of the CAPM as expressed
in Eq. (7.32) in excess return form implies no intercept term. The intercept term in
Malatesta's model is taken as a component of the normal return. It may reflect prob-
lems of measuring the risk-free return as well as the behavior of the intercept term
calculated for the prior control period. One possibility is that the acquired firms had
negative intercept terms while acquiring firms had positive intercept terms for periods
in years before the merger "event." If so, the net present value of merger activity
adjusted could well be positive for Malatesta's study rather than negative.
Other studies that provide information on the role of mergers are those that analyze
antitrust episodes. This was the major emphasis of the Ellert [1975, 1976] studies
referred to above. Ellert analyzed the data for 205 defendants in antimerger com-
plaints initiated by the Justice Department and Federal Trade Commission (FTC)
under Section 7 of the Clayton Act for the period 1950-1972. Of the complaints,
121 were issued by the Justice Department and 84 by the FTC. Ellert observes that
the government had not lost a single Supreme Court merger case after the 1950
revision to the Clayton Act. In 60% of the cases studied, defendants canceled merger
plans or were ordered to divest part or all of the assets previously acquired. The
average duration of litigation measured by the interval between the filing of the com-
plaint and the entry of the last judicial order was 34 months.
Ellert analyzes the behavior of the data for the 205 defendants and also the two
groups broken into 123 defendants ordered to divest acquired assets and 82 defen-
dants not required to divest assets. For both groups of defendants for a period pre-
ceding the filing of a merger complaint by four years, the residual performance was
positive and statistically significant for both groups. The cumulative average residual
was over 18% for defendants ordered to divest and about 13% for defendants not
required to divest. For the. 48 months prior to the filing of the merger complaint,
defendants required to divest achieved a further positive residual that was statistically
significant. For the same 48 months proximate to the filing of the merger complaint,
defendants not required to divest had returns that were not statistically different from
the average for the market. In the 12 months preceding the merger complaint the
residual performance was not statistically significant for either class of defendants.
On the filing of the merger complaint the average portfolio residual declines by
1.86% for defendants ordered to divest and 1.79% for defendants not required to
divest. While these percentages are small, they represent substantial absolute amounts
when applied to the large dollar amount of assets involved. In the postwar period
these percentages translate into an average dollar loss per respondent of about $7.5
million
STUDIES OF ANTITRUST CASES 731
Following the filing of the merger complaint the behavior of the residuals is not
statistically significant. This is true for the period between the filing of the merger
complaint and the settlement of the litigation. Also, in the 48-month period following
the settlement of the litigation, the behavior of the returns to the companies is not
statistically different from the market as a whole. It is of interest to note also that
final settlements that resolve future uncertainties are typically followed by statistically
insignificant behavior of the residuals. On the other hand, settlements such as nolo
contendere, which are more likely to be followed by private triple damage suits, result
in a persistence of negative residuals that are statistically significant.
It is somewhat surprising that the magnitude of the negative movement in the
residuals upon the filing of a complaint is relatively small. The $7 million to $8 million
involved is mainly accounted for by litigation costs. A possible explanation may be
that there was no basis for bringing the complaint in the first place and that, therefore,
nothing significant could be expected to be accomplished in the final action taken by
the antitrust authorities. Ellert observes that these results "are also consistent with
nonmonopolistic hypotheses of merger motivation. If merger is viewed as a means of
expanding the operations of an efficient management group, a specific divestiture of
assets would not be expected to constitute a large threat to stockholders, particularly
if the firm may substitute internal expansion or external acquisitions in other markets
in the future" [1976, 727].
A similar explanation is related to case selection by the FTC and the Department
of Justice. Ellert [1975, 66-67] observed the following:
Thus acquiring firms had a record of effective management of assets in the years
preceding their merger activity. Ellert also observes that companies acquired were
typically those whose premerger performance was consistent with ineffective manage-
ment of assets. This is consistent with the role of mergers as performing a useful
economic function in reallocating resources from less efficient to more efficient users.
Both Stillman [1983] and Eckbo [1981] analyze the residuals of the rivals to
firms participating in mergers. They sought to distinguish between the possible effi-
ciency vs. monopolization effects of mergers. The problem is somewhat complex
because at the theoretical level alternative hypotheses can be formulated. The com-
plexity is illustrated by Table 20.1.
732 MERGERS AND RESTRUCTURING: TESTS AND APPLICATIONS
In Table 20.1 three event dates are identified. They are announcement dates with
respect to the merger, to its challenge by the antitrust authorities, and to the an-
nouncement of a government decision. For firms participating in the merger the sec-
ond column of Table 20.1 indicates that the predicted signs of the CARs are the same
for the collusion vs. efficiency hypotheses at the times of the announcements of the
merger and its challenge. It was for the purpose of sharpening the analysis that the
residuals for rival firms were analyzed. But as the final column of Table 20.1 indicates,
STUDIES OF ANTITRUST CASES 733
there is considerable overlap in the predicted signs for the alternative collusion and
efficiency hypotheses. Nevertheless, the pattern of these relationships is such that
some judgments can be formulated.
Specifically, at the merger announcement date, if the sign of the effect on rivals
is not positive, this is consistent with efficiency because efficiency can have a negative
or zero sign. Similarly, at the date of the merger challenge, if the effect on the abnormal
returns of rivals is not negative, this also is consistent with efficiency, which can also
take on a positive or zero sign. In general, when the effect on the CARs is the same
for both participants and rivals, it is not possible to distinguish between collusion
and efficiency. If the signs differ, then the two groups are affected differently, which
tends to rule out the collusion hypothesis. However, at the decision date the results
are more difficult to interpret. This is because the effects on rivals can take any of the
three possible signs for either outcome. In the light of the general framework presented,
let us now look at the empirical results that have been compiled.
Stillman [1983] found that the effect of 30 major challenged horizontal mergers
on the residuals of rivals was not statistically significant. The concentration-collusion
theory argues that positive residuals should have been observed when the merger was
in process and negative residuals when it was challenged. Lack of significance was
observed both in relation to the original merger proposal and when it was challenged.
Since the effect on rivals was not statistically significant, this casts doubt on the
concentration-collusion theory that the mergers were in fact viewed as opportunities
for increased possibilities of collusion among the firm's major rivals in the industry.
Eckbo [1981] extended the Stillman study, using a larger sample and a "control"
sample of vertical mergers. Eckbo finds that on the announcement of the mergers
there are positive residuals both for the participants and their major rivals. This ap-
pears to be consistent with the monopoly theory. It is not unambiguous though, be-
cause one could also argue that the announcement of the proposed merger conveys
information to rivals of opportunities for increased efficiency by expanding scale.
Eckbo further finds that at the announcement of the filing of a suit by the antitrust
authorities, there is not much effect on the residuals of either the participants or their
rivals; in fact, in cases brought by the Federal Trade Commission the effect on rivals
is slightly positive. This is consistent with the explanation that the merger partners
would have been more efficient and the rivals are protected from this increased effi-
ciency by the Federal Trade Commission suit blocking the merger. Eckbo concludes
that the positive performance of rivals of challenged mergers at the time of the original
merger announcement reflects information conveyed by the proposed merger that
efficiencies can be achieved by expanding scale either internally or externally.
In an extension of Eckbo's earlier work, Eckbo and Wier [1985] paid particular
attention to mergers challenged after 1978 i.e., following passage of the Hart-Scott-
Rodino Antitrust Improvements Act of 1976, which was intended to improve the
likelihood that mergers selected for prosecution were truly anticompetitive. Eckbo
and Wier found little evidence of improvement; the 17 post-1978 mergers in their
study were "economically efficient" and "apparently would not have harmed com-
petition" [139]. They lay the blame for this failure on the case selection criteria, i.e.,
inappropriate application of the Department of Justice Merger Guidelines of 1968
734 MERGERS AND RESTRUCTURING: TESTS AND APPLICATIONS
C. CORPORATE GOVERNANCE
A number of other studies deal with corporate governance aspects of mergers and
tender offer activity. In our judgment, the main significance of such studies is to em-
phasize the role of mergers and tender offers in dealing with agency problems. If the
firm is not being managed in the best interests of shareholders, another group may
see opportunities for increasing the value of the firm. The studies that will be treated
next deal with this and related issues.
Concern has been expressed that older views of antitrust still unduly influence U.S. policy. A case in point
is the offer by PPG Industries to acquire Swedlow Plastics, a specialized Defense Department—oriented glass
manufacturer with advanced technology. Swedlow's founder was desirous of selling his firm because of his
advancing age and the risks of further investments to stay abreast of rapidly advancing technology. PPG
made an offer because it was willing to invest financial resources to maintain and perhaps advance Swedlow's
technological leadership. But the Federal Trade Commission voted to subject the proposed merger to a
lengthy administrative review and evaluation. Swedlow then accepted an offer from a major British glass
firm because of the delays and uncertainties imposed by the U.S. antitrust regulatory agency, the FTC.
CORPORATE GOVERNANCE 735
shares, thus "free-riding" on the value added by the bidder activity. The reduction in
returns to bidder activity may cause underinvestment in acquisition activity. Gross-
man and Hart analyze how to deal with the free-rider problem. They seek to formulate
a set of rules of the game so that if some of the shareholders do not tender, there will
be some risk that they will have something to lose. The rules, it is hoped, will result
in an optimal amount of acquisition activity.
Grossman and Hart observe that it is possible to avoid the free-rider problem
by providing for a contract written into the charter of the firm itself that permits the
bidder to "expropriate" the shareholders remaining after the takeover to the extent
of a dilution factor they call 0. This expropriation may be accomplished by selling
assets or output of the firm to another firm owned by the new control group at prices
below market value.
What Grossman and Hart are seeking to achieve is best understood in the light
of socially desirable mergers, acquisitions, and takeovers. From a social standpoint,
they should have a positive NPV. For value to be created the bidding firm does some-
thing after the target firm is acquired. It changes the target firm in ways that will add
value. Usually this involves the expenditure of resources by the bidding firm; at a mini-
mum, it involves the expenditures required to research and to take over a company.
The problem is that after the acquisition all shareholders share proportionately in any
gains that are created. Because the old shareholders participate proportionately in
the gains, this reduces the incentive to bidder firms to make investments that create
additional value. From a social viewpoint this may lead to underinvestment in acquisi-
tion activity. The goal is to formulate rules of the game so that those who spend the
resources are enabled to capture a portion of the gains commensurate with the re-
sources they have committed and the risks they have incurred.
One of the implications of the Grossman and Hart model is the use of two-tier
tender offers. A higher price and better terms, such as permitting a choice of cash or
stock to fit the tax needs of the tendering shareholder, may be used on the first block
of shares required for control. For the remaining shares the price is lower and the form
of payment may be less attractive, such as the debt or preferred stock of the acquiring
company. So shareholders have a strong incentive to tender in the first group rather
than in the residual group. The first-tier offer is likely to be oversubscribed and there-
fore handled on a pro rata basis by the acquiring company. The real price realized by
tendering shareholders is an average of the first-tier and second-tier arrangements.
Thus differential prices can be used to increase the probability of success of the tender
offer. The possibility of such an arrangement permits the dilution factor 0 that will
be set in the original charter to be much lower than otherwise.
An illustration of this two-tier bidding is provided in the U.S. Steel tender for
Marathon Oil, which took place in early 1982. The first step was the bid by U.S. Steel
on November 5, 1981, for 30 million Marathon common shares at a price of $125
per share. The second step was the provision that on the effective date of the merger
each remaining shareholder of Marathon would be entitled to receive for each com-
mon share $100 principal amount of 122 percent guaranteed notes due 1994. Some
dissatisfaction was expressed with the second step because the market value of the
736 MERGERS AND RESTRUCTURING: TESTS AND APPLICATIONS
notes offered appeared to be in the $70 range. This offer represented an average value
of $106 for Marathon shares. This was relatively close to Mobil's competing bid,
which was valued at $108. When the U.S. Steel bid was approved by shareholders on
March 11, 1982, the value of the notes was $76, and the tender was valued at $101 per
Marathon share.
However, it is more appropriate to view the two steps as the components of a
single price [Brudney and Chirelstein, 1978]. The two steps should be recognized
as a useful social device for increasing the probability of success of tender offers. Thus
it is a useful method of dealing with the managerial agency problem. Furthermore,
competition among firms that make tender offers will push up the level of the average
price in the two-part transaction. Grossman and Hart conclude that U.S. government
policy on takeover bids from the Williams Act in 1968 to the present may have un-
desirable consequences. The disclosure provisions of the Williams Act require a bidder
to make an announcement of intentions after buying 5% of the company. This en-
courages competition from other bidders but also aggravates the free-rider problem.
managers, giving them incentives to pursue long-run maximization goals rather than
short-term objectives.
In DeAngelo and Rice's empirical tests the mailing date of the proxy announce-
ment is used as the best estimate of the date the amendments are compounded in
the stock price. The analysis of daily returns indicates a weakly negative price impact,
which is statistically insignificant. The methodology was varied in a number of ways
without altering these results. DeAngelo and Rice observe that their tests may be
biased against the managerial-entrenchment hypothesis because the proposal of anti-
takeover amendments may signal an increased probability that the firm will be ap-
proached with a merger or tender offer. Since stockholders in acquired firms typically
earn positive abnormal returns, such a signal should impart an upward bias to the
announcement return. The signaling hypothesis would imply positive returns; the
management-entrenchment hypothesis implies negative returns. The net effect of
the two influences is uncertain. Their observance of nonsignificant negative returns
implies that without the positive signaling influence the returns would be more
strongly negative, consistent with the management-entrenchment motive for anti-
takeover amendments.
are prtinent: (1) the first announcement of the contest in the Wall Street Journal;
(2) the stockholder meeting and election; and (3) the announcement of the election
outcome.
Contrary to the improved-management hypothesis, there is little evidence of poor
performance in the months and years preceding the contest. There is evidence of sig-
nificant positive abnormal performance in the five months preceding the contest an-
nouncement. Daily returns show cumulative returns of + 11.9 percent over the 60
days up to and including the announcement date. This is almost certainly related
to the contest rather than to another simultaneous event: (1) the abnormal perfor-
mance is concentrated in the days very close to the announcement; and (2) the Wall
Street Journal announcement often comes relatively late in the proxy contest process,
and preannouncement leakage is great. Positive performance persists in both control
and participation contests (higher in control contests), and regardless of the eventual
outcome (although it is higher for successful dissident groups).
Significant negative abnormal returns are observed in the period from the an-
nouncement through the election outcome; the mean cumulative residual is —4.3%.
These returns are concentrated in control contests, but participation contests also
exhibit insignificantly negative returns during this interim period. Again, these results
persist in spite of outcome. Although these negative returns indicate that at least some
of the preannouncement gains are not permanent, the net result over the entire period
surrounding the contest remains positive for the entire sample.
Dodd and Warner conclude that proxy contests are associated with significant
share price appreciation of an average magnitude of 5 to 10%, consistent with the
improved management of corporate resources hypothesis. That the positive perfor-
mance persists regardless of contest outcome indicates that it is the challenge per se
that results in improved corporate performance. The erosion of early share price gains
can be partly explained by the prior overvaluation caused by increased demand for
the stock before the holder-of-record date.
strajnts on management's ability to resist control transfer would reduce the incentive
to invest in firm-specific human capital inputs, with corresponding negative effects
on shareholders.
inferior-vote common Although the samples are small, statistical tests indicate that
the results are significant and of economic importance in terms of dollar value. The
authors conclude that superior voting rights must provide the potential for incre-
mental benefits (of necessity indirect since explicit cash flow benefits are identical);
one possibility is additional compensation and perquisites for employee/shareholders
of the superior-vote stock. The incremental value of voting rights appears to be
diluted when control is shared with voting preferred stock, indicating that control
involves costs as well as benefits. The distribution of costs and benefits may, in part,
be dependent on the complexity of the control structure. DeAngelo and DeAngelo
confirmed, to some extent, the Lease et al. finding of a price premium for superior-
vote stock in their analysis of four mergers in which the superior-vote common com-
manded a rare explicit price premium over inferior-vote common stock with an
identical cash flow claim.
9. McDaniel
McDaniel [1986] emphasizes bondholder interests in his article. He examines
the relationship between bondholders and shareholders overall, how bondholders are
affected by mergers, spinoffs, etc., and the need for bondholder protection. While the
traditional focus has been on shareholder wealth maximization and the differences
between shareholder and bondholder claims, McDaniel suggests that the distinctions
between the two securities are blurring. He holds that the average shareholder in a
large public corporation with a nominal ownership interest is essentially a passive
property owner not unlike the bondholder. (Junk bonds even have risk characteristics
very similar to equity, and the changing nature of the bond markets has increased
the risk volatility of all bonds.) Therefore, McDaniel suggests, the emphasis on share-
holder wealth maximization may be misguided if such wealth comes at the expense
of bondholders.
Firm value maximization is the preferred goal and is synonymous with share-
holder wealth maximization only if bondholder interests are protected. There is some
evidence that at least some of the shareholder gains in mergers and voluntary spinoffs
are the result of bondholder expropriation; McDaniel argues that any degree of
expropriation is unacceptable. With respect to voluntary spinoffs, e.g., Schipper and
Smith [1983] indicate that bondholder expropriation is less likely under certain con-
ditions (where parent debt is allocated to the spun-off subsidiary, or where the sub-
sidiary has its own debt). But McDaniel suggests that bondholders are less protected
than many previous authors have supposed.
Bond indentures with their protective covenants are regarded as a strong bulwark
against expropriation. But McDaniel analyzes the indenture covenants of Fortune
magazine's 1984 100 largest industrial corporations and finds virtually no restrictions
on risk alteration or asset disposition (i.e., restrictions on investment risks and dis-
investment), which can be damaging to bondholders. He feels that the trend over
time has been toward fewer rather than more protective covenants. Perfect covenants
require perfect foresight and costless contracting. If covenants are too lenient, share-
holders may benefit at bondholder expense; if too strict, bondholders exercise virtual
veto power over the firm's actions.
744 MERGERS AND RESTRUCTURING: TESTS AND APPLICATIONS
The bond trustee, another purported defender of bondholder interests, has almost
no duties prior to default; his or her postdefault duties are of little help to bond-
holders since most of the damage has already occurred. State and federal statutory
constraints do little more than restrict dividends and asset transfers that would result
in outright insolvency, whereas alteration of investment risks can be very harmful to
bondholders' positions well short of insolvency.
Stock exchange listing criteria are not very stringent, and in any case, bonds are
increasingly unlisted; even listed bonds trade mostly on the over-the-counter market.
Market forces that could discipline managers with higher interest rates for a reputa-
tion for harming bondholders are only effective if there is a "next time out." In merger
contests the beleaguered target is in an "end-game" situation, which increases the
incentive to exploit bondholders by, e.g., a scorched earth policy.
McDaniel argues that the costs and frictions in the market for corporate control
reduce its disciplinary effectiveness. Bond market liquidity can protect bondholders
against gradual deterioration in risk (by allowing bondholders the possibility of selling
out quickly) but offers little protection against sudden changes in risk. Another pos-
sibility for bondholder protection is buying both stocks and bonds of the same firm
(diversification in an all-bond portfolio would be ineffective, since the possibility of
bondholder expropriation would be a systematic risk in that case).
McDaniel advocates applying the fiduciary duty principle of corporate law as a
substitute for costly contracting to protect bondholder interests. This would give
management a responsibility to look out for bondholders as well as shareholders.
McDaniel would give bondholders the same right as shareholders to bring deriva-
tive suits (on behalf of the corporation) as well as direct suits (against the corporation),
which is all that is permitted to bondholders under current law. He places great faith
in the courts to decide whether management is acting according to its fiduciary duty
and which type of lawsuit is appropriate in each case.
McDaniel also proposes a number of methods to protect bondholders in take-
over contests, among them: (1) due-on-sale clauses, which would make debt payable
upon sale of the firm; (2) requiring that bidders make tender offers for bonds as well
as for shares, or enacting a conditional self-tender for bonds triggered by takeover;
(3) higher interest rates and/or voting rights for bondholders triggered by takeover;
(4) giving target bondholders a first mortgage on target assets, thus effectively sub-
ordinating all other debt.
value decisions, and thus their announcement should result in upward stock price
movement. They use the mean-adjusted returns, or comparison period, technique for
daily returns around the event date (the date of the Wall Street Journal in which the
event is first announced). That is, the expected return on a security is assumed to be
constant, and the analysis focuses on the error term, or the difference between the
expected return and the actual observed return.
Three procedures for estimating the expected return and its standard deviation
are discussed. The pre-only procedure uses data from a preevent estimation period;
however, if either the expected return or its standard deviation or both shift to new
levels following the event, the pre-only procedure will result in a biased estimator of
postevent standardized residuals. A second procedure, the both-combined method,
calculates estimates of expected return and standard deviation after combining pre-
event and postevent estimation periods; the resulting standardized residual estima-
tor may be biased for both pre- and postevent residuals. Alexander, Benson, and
Kampmeyer propose a both-but-separate technique and calculate two separate stan-
dardized residual estimators; the preevent estimation period provides an unbiased
estimator of preevent standardized residuals, and a postevent estimation period pro-
vides an unbiased estimator of postevent standardized residuals.
Three possible motives for voluntary selloffs are each associated with potentially
unfavorable information: (1) firms seek to sell unprofitable ventures; (2) firms seek to
become less diverse; and (3) firms seek to generate needed cash. Thus the selloff
announcement itself may become part of a "bad news/good news" scenario. Since it
is unlikely that a firm would release the good news before the bad news, the re-
maining alternatives are to release the bad news first or to release it simultaneously
with the good news. Thus the expected stock price effects might follow a negative-
normal-positive sequence or might be offsetting to some degree.
Alexander, Benson, and Kampmeyer's sample consisted of 53 voluntary selloff
announcements by NYSE firms over 1964-1973; data consisted of 301 daily returns
centered on the announcement date. Results were calculated using all three estimation
procedures and using the market-adjusted returns technique as well as the mean-
adjusted returns technique. The pattern and magnitude of cumulative average resid-
uals (CARs) was found to be similar regardless of the estimation procedure, although
there was some difference in significance of the results. Selloffs appear to be announced
following a period of generally negative returns. The study found that voluntary
selloff announcements had a very slight positive price impact; whether the size of the
positive impact was offset by the simultaneous release of negative information, re-
flected a near-perfect market for selloffs, or was just indicative that the NPV of the
selloff was small relative to all other cash flows of the firm is impossible to distinguish.
2. Jain
Jain [1985] also examines the effect of voluntary selloffs on shareholder wealth
but extends the work of Alexander et al. by examining stock price reactions of both
buying and selling firms. Jain identifies four possible motives for the selling firm: (1)
unforeseen circumstances, such as poorer-than-expected earnings; (2) the selloff as a
746 MERGERS AND RESTRUCTURING: TESTS AND APPLICATIONS
partial merger when certain parts of the firm may be more valuable to another user
than to the current owners; (3) desire to specialize in a limited number of business
activities, to become less diverse; and (4) desire to transfer wealth from bondholders
to shareholders, the agency problem. Buyers are viewed as investors in new projects,
and thus their shareholders are expected to experience zero or positive stock price
effects since investors, on average, invest in at least nonnegative NPV projects.
Jain's sample consisted of 1062 event dates for sellers (date of first published
information about a selloff in the Wall Street Journal, whether or not a buyer was
identified) and 304 event dates for buyers (the date the possible buyer was initially
identified in the Wall Street Journal not necessarily the same as the seller's event
date). The methodology consisted of estimating the expected return for a given secu-
rity on a given date using beta-controlled portfolios of firms formed in event time,
and a preevent estimation period.
Jain found that sellers earned significant positive excess returns of 0.70% over
day — 5 to day — 1. Buyers were found to experience significant positive excess returns
of 0.34% on day — 1, although returns over days — 5 to —1 were insignificantly dif-
ferent from zero. Thus selloff announcements appear to be good news for both buyers
and sellers. The fact that abnormal returns to sellers are much smaller in absolute
terms than those earned by target firms in mergers may reflect the relative importance
of the two types of events or the fact that the selloff is initiated by the seller, whereas
the merger is usually initiated by the buyer. The results may indicate that the selloff
market is less than fully competitive only rarely did more than one buyer come
forward to competitively negotiate with the seller or the seller's desire for a quick
ti ansaction for needed liquidity.
Examining the preevent period revealed significant negative abnormal returns for
selling firms of — 10.8% from day — 360 to day —11. Thus selloffs may be one kind
of action taken by managers in adversity to enhance their shareholders' welfare.
significantly positive, 2.5%, with a 3.3% positive abnormal return over a two-day
announcement period. 2
The authors test for the effect of spinoff size by splitting their sample into two
subsamples: a large spinoff group for which the divested unit has an equity value of
at least 10% as large as the parent's equity value, and a small spinoff group, consisting
of 34 and 21 firms, respectively. The effect of minor spinoffs was, not surprisingly,
significantly smaller relative to the effect of large spinoffs.
For the sample as a whole the abnormal return over the entire 181-day observa-
tion period (day —120 through day + 60) was positive and significant at 22.1%, with
more than half this total (13.6%) occurring over day —120 through day —20, leading
Miles and Rosenfeld to conclude that spinoff announcements, on average, follow a
period of abnormally positive returns, which is in marked contrast to the generally
negative abnormal returns preceding selloff announcements. Thus while both spin-
offs and selloffs are similar events and both result in positive share price effects, they
appear to be motivated by very different underlying circumstances.
Miles and Rosenfeld denote as day 0 the day preceding the Wall Street Journal publication date since
this is the most likely date of the parent firm's press release announcing the spinoff; the significant positive
returns on day 1 probably reflect the fact that press releases are often not issued until after the day's
trading has closed.
748 MERGERS AND RESTRUCTURING: TESTS AND APPLICATIONS
trust); fewer government restrictions (e.g., spinning off a regulated utility or a foreign
subsidiary); or relief from onerous labor union contracts (i.e., a spun-off subsidiary
may not be obliged by collective bargaining obligations of its former parent). For 18
sample firms where the parent specifically stated regulatory or tax motives in its spin-
off announcement, the two-day announcement period return was 5.07% (significant).
Thus this type of contract revision is shown to be a source of shareholder wealth
increase in cases where it is available. However, it cannot explain the gains for the
entire sample.
Finally, Schipper and Smith consider the effect of spinoffs on the contract between
shareholders and managers. Gains in this case could result from reducing the size
and diversity of the asset base under a given management's control to eliminate or
reduce diseconomies of scale and improve management incentives and productivity.
The authors examine firm growth in the five years prior to the spinoff for evidence
of possible diseconomies of size and find average annual sales growth, fixed asset
growth, and number of employees growth of 20%, 30%, and 19%, respectively, for the
sample as a whole. For a subsample of 30 firms with a stated motive of reducing size
and diversity, some of this rapid growth was clearly the result of merger activity;
65 acquisitions were completed in the two years preceding the year of the spinoff,
and 17 of the spinoffs were reversals of prior acquisitions.
The effect of diversity on diminishing returns to management is measured by the
industry membership of parent and spun-off subsidiary. In 72 of 93 cases the sub-
sidiary was classified in a different industry. A second measure of diversity seeks evi-
dence of structural shifts in parent firm stock returns following the spinoffs. For a
subsample of 62 firms for which sufficient data were available, 22 exhibited statisti-
cally significant shifts in stock returns, more than could be expected by chance, rein-
forcing the evidence of industry membership.
The authors conclude that spinoff announcements benefit parent firm share-
holders and that the source of the gains is a combination of relaxed regulatory or
tax constraints and increased productivity due to reduced size and diversity of assets
under a single management. Bondholder expropriation was not judged to be a sig-
nificant source of shareholder gain.
prior to the press date and extends to a completion date, the date the spinoff dividend
declaration is reported in the Wall Street Journal, an average of 62 days from the press
date. They used the market model with a preevent estimation period, and their results
are consistent with those of other studies—a significant abnormal two-day announce-
ment period return of + 3.3% for the sample as a whole. However, the results were
not uniformly positive; 38 announcements resulted in a negative two-day return.
Similarly, there was evidence of good price performance in the preannouncement
period. For the entire sample the abnormal return over the event period is +7.7%
(significant); postannouncement abnormal returns are insignificant here, as elsewhere,
indicating semistrong market efficiency.
The expropriation hypothesis was tested on a subsample of 31 firms that had
53 publicly traded senior issues. Since infrequent trading can be a problem in analysis
of senior security price changes, the authors used the period "around" the announce-
ment date rather than a strict two-day announcement period return. However, there
was no evidence of any significant price effect on senior security prices. Further,
matching pairs of senior security price reactions and common stock price reactions
resulted in price changes of the same sign in 29 of 53 cases. The authors reject the
expropriation hypothesis.
Hite and Owers also tested for the effect of relative spinoff size, with results
consistent with Miles and Rosenfeld. Large spinoffs led to a greater stock price re-
action than small spinoffs (5.2% versus 0.8%, respectively, for the two-day announce-
ment period return). They find that, overall, spinoffs generate gains roughly equal to
the value of the divested unit; parent firm values were not significantly diminished
by distributing assets to subsidiaries.
Finally, Hite and Owers break their sample into four subgroups according to
the stated rationale for the spinoff. The first group consisted of 12 spinoffs designed to
facilitate mergers (either the acquirer wanted only the subsidiary or only the parent)
and resulted in event period excess returns of 11.6%. These so-called partial mergers
thus produced shareholder gains analogous to the takeover premia paid to target
firms in full mergers, but of a lesser magnitude.
For 27 spinoffs the stated motive was specialization, or "undiversification," to
"get back to basics." This group showed the largest excess returns over the event
period, 14.5% .
A third group of 19 spinoffs were undertaken as a defense against anticipated
external interference, e.g., from regulators or antitrust authorities. Although the two-
day announcement period return for this group was + 3.4%, this barely overcame
negative preannouncement performance of — 3.2% , and the group experienced nega-
tive returns of —4.7% for the entire event period.
A catchall group of 63 spinoffs without explicitly stated motives not surpris-
ingly showed price performance equivalent to a portfolio of the other three groups
combined.
Reexamining their initial hypotheses, Hite and Owers conclude that there is little
support for the expropriation hypothesis but also that the magnitude of abnormal
returns is too large to be fully explained by unique optimal contracting for parent
and subsidiary. Rather, they suggest that there must be some fundamental change
750 MERGERS AND RESTRUCTURING: TESTS AND APPLICATIONS
in the underlying opportunity set facing parent and/or subsidiary that leads to the
spinoff and is announced simultaneously. This is supported by the differences in ab-
normal returns for subsamples grouped by spinoff motive.
E. GENERALIZATIONS FROM
THE STUDIES
While the numerous studies reviewed are not always in complete agreement, some
generalizations can be formulated on the basis of the dominant patterns observed.
These generalizations relate to an evaluation of the role of mergers and related re-
structuring activities on a number of important issues of corporate financial theory
and policy. The empirical results are summarized in Table 20.2.
For acquired firms, from the period just before the merger announcement date,
the shareholders achieve significant positive gains. In an earlier period the cumulative
average residuals are negative, indicating that acquired firms had not been performing
up to their potentials.
On average for the period just before the merger announcement date the share-
holders of acquiring firms neither gained nor lost. But in tender offers where the ne-
gotiation period is shorter, gains for the shareholders of acquiring firms tend to be
significantly positive. The conclusion indicated is that legislative enactments that have
delayed the completion of tender offers have also reduced the gains for acquiring firms.
In earlier periods the CARs for acquiring firms are positive, indicating that ac-
quiring firms have had a record of managing asset growth successfully. This leads to
the further hypothesis that in pure conglomerate mergers the prospective depletion
of internal investment opportunities leads acquiring firms to look to other industries
where growth prospects are more attractive and to seek out firms that appear less
capable of responding to these favorable prospects [Chung, 1982].
Some studies indicate that the CARs decline after the merger or tender offer.
However, this appears to be accounted for by the nature of the industries in which
the acquiring firms mainly operate. Netting out the industry effect, the CARs do not
754 MERGERS AND RESTRUCTURING: TESTS AND APPLICATIONS
A. Merger studies
1. Acquired firms 20%
2. Acquiring firms 2-3'
C. Sell-offs
1. Spinoffs 2-4
2. Divestitures
Sellers .5-1
Buyers .34
3. Equity carve-outs 2
D. Premium buybacks
1. Single blocks from outsiders —2
2. From insiders or small shareholdings 1.2
3. Sellers of single blocks 1.5
E. Standstill agreements
(nonparticipating stockholders) —4
decline after the merger. These results contradict both the managerialism theory and
the theory that mergers are not based on valid economic or business reasons.
Analysis of unsuccessful tender offers permits a further sharpening of the analysis
among competing theories. The monopoly theory would hold that an unsuccessful
tender offer should result in negative residuals for shareholders of both acquiring and
acquired firms, since the gains from the monopoly rents will be lost. The manageri-
alism theory predicts that shareholders of the acquiring firm would gain, since the
proposed tender offer was uneconomic in the first place; aborting the combination
therefore would be a blessing to shareholders of the acquiring firm. To the extent
that these acquiring firms were ready to overpay the shareholders of the targets, the
latter shareholders would lose if the tender offer does not succeed. An alternative
theory would be that since the potentials for improving the management of the ac-
quired firm or the possibilities of synergy between an acquiring and an acquired firm
GENERALIZATIONS FROM THE STUDIES 755
have been discovered and brought to light, no market reaction will occur when a
particular tender offer is unsuccessful. The empirical results are that after unsuccessful
tender offers there is no significant change in the CARs of the acquired firms. This is
consistent with the new-information hypothesis. The effect on the bidding firms would
be predicted to be negative because they have spent resources in search and costs of
formulating the offer, and now these investments have no fruition. The data confirm
that the residuals are negative, but they are not significant [Dodd and Ruback, 1977,
368]. One reason why the negative residuals are not significant is that the market
may view bidder firms as engaged in a series of tender offer attempts, because they
have a track record of managing assets effectively. Hence the market reacts negatively
if search costs for specific tender offers fall through. But taking a longer view [a la
Schipper and Thompson, 1983], the market judges that this experience will be put to
good use in other tender and merger activities.
The differences in results obtained may be explained in part by the signaling effect
of the type of compensation used in an acquisition [DeAngelo and DeAngelo, 1987].
If a firm uses stock in the acquisition, this may imply that it considers its stock to
be overvalued. If a firm uses cash, this may reflect the firm's judgment that its own
stock is too valuable to use—it is undervalued. (Cf. Myers and Majluf [1984] for a
similar hypothesis with respect to internal investment decisions.) The empirical studies
are consistent with this information hypothesis. The Bradley study is of cash tender
offers, Dodd and Ruback study both cash and stock exchanges, and the Dodd study
is on mergers with mainly stock swaps. Bradley found positive returns to acquiring
firms, Dodd negative returns, and Dodd and Ruback no significant effects. When cash
is used, this may signal that the stock of the acquiring firm is undervalued, and con-
versely when stock is used for the purchase. When both cash and stock are used, the
two effects may be counterbalancing.
The studies of the effects of merger activity on rivals seek to distinguish be-
tween monopoly and efficiency motives for mergers. The concentration-collusion
theory holds that there would be positive effects on rivals upon the announcement
of major horizontal mergers. The evidence is somewhat conflicting on this point. The
concentration-collusion theory also holds that when the government authorities make
a final decision to prevent a merger, the effects should be negative residuals for the
rivals as well as for the participant firms. However, there appear to be no 'significant
effects on any of the groups. This is inconsistent with the monopoly hypothesis. It
may be that, regardless of the actions taken by the antitrust authorities, the underlying
economic forces of the industry have now been identified. It would appear that op-
portunities for efficiencies will be effectuated internally if they are not permitted to
take place externally. If there are no monopoly effects from the horizontal mergers,
and if the economic results are ultimately the same, then the antitrust activities rep-
resent extra costs in the economic resource allocation process that delay the achieve-
ment of potential improvements in efficiency.
The corporate governance studies portray an important role for merger and take-
over activity (including proxy contests) in utilizing the market for corporate control
to deal with agency problems. The literature is not in agreement on whether the market
for managerial labor and various forms of settling up in compensation contracts are
756 MERGERS AND RESTRUCTURING: TESTS AND APPLICATIONS
sufficient to deal with the agency problem. However, it would be plausible that the
market for corporate control, the threat of takeovers, and the market for managers
all operate to deal with the agency problem. Grossman and Hart treat the free-rider
problem in the functioning of the market for takeovers. Charter provisions for dilution
after the completion of a merger or two-tier merger proposal are among the mecha-
nisms for dealing with the free-rider problem.
Some studies indicate that antitakeover provisions, standstill agreements, and
buyback provisions may have a negative, but not significant, influence on share prices.
Other tests find a positive but not significant influence. Structural defenses in response
to hostile takeover bids had a negative effect on target share prices, although over the
entire contest period, target shareholders were not harmed. Overall, the studies sug-
gest no significant effect. This result could be consistent with an efficiently operating
market for managers and a competitive market for corporate acquisitions. The Dodd-
Warner study suggests that proxy contests represent an alternative at the margin for
making the market for corporate control more effective. Their results are consistent
with the hypothesis that proxy contests are not simply power struggles but are asso-
ciated with changes that result in the improvement of share price performance of the
firm. These studies highlight the importance of the agency problem introduced into
the literature by Jensen and Meckling [1976].
Studies of takeover defenses and the extent of managerial vote control appear
to indicate that the agency problem inherent in the shareholder-manager relationship
is not a serious cause for concern. Rather, management's control over the board
and its ability to resist takeover attempts may actually benefit shareholders by in-
creasing manager's incentives to make firm-specific human capital investments and/or
by driving up the price at which the firm ultimately sells.
The conflict between shareholders and bondholders may be more troublesome,
since, as McDaniel [1986] points out, bond covenants offer bondholders very little
protection against expropriation in takeovers. A 1986 study by Dennis and Mc-
Connell of senior security price reactions to merger announcements, however, finds
little evidence of negative effects on bond prices.
Consistent with the above studies, analyses of various other forms of restructuring
including selloffs, spinoffs, equity carve-outs, and going private were also found to
benefit shareholders. Selloffs, which bring in additional funds, were found to follow
periods of generally negative returns, whereas spinoffs, which generate 'no influx of
cash to the parent firm, were found to follow periods of above-normal performance.
Reasons for the positive stock price reaction to selloff and spinoff announcements
include the elimination of negative synergies, improvements in managerial incentives
and organizational design, and relaxation of regulatory or tax constraints. Expro-
priation of other security holders was ruled out as a primary source of gain.
In equity carve-outs the parent firm maintains control of the carved-out subsid-
iary while bringing in cash to finance its growth opportunities without the negative
stock price effects generally associated with additional offerings of seasoned parent
firm stock.
Going-private and leveraged buyout transactions resulted in gains to public
shareholders almost as high as gains to target shareholders in tender offers, primarily
TERMS OF MERGERS 757
F. TERMS OF MERGERS
Larson and Gonedes [1969] set forth a model of exchange ratio determination. They
reject the effect of mergers on earnings per share as a test and argue that the effects
of mergers on market value would be an appropriate test. The market values of
common stocks are used to determine the precombination Wealth positions of the
parties involved in a merger, and the result is compared with the postmerger wealth
positions of the parties. Since the Larson-Gonedes decision horizon is the immediate
postmerger market values, the postmerger price/earnings ratio of the merged firm
receives their greatest emphasis. They observe that the earnings multiple of a com-
bined firm will be a weighted average of the earnings multiples of the constituent
firms if (1) the growth rate of the combined entity is a weighted average of its constit-
uents' growth rates, and (2) the riskiness of earnings stream of the combined entity
758 MERGERS AND RESTRUCTURING: TESTS AND APPLICATIONS
is a weighted average of the earnings streams [1969, 722]. The earnings multiple of
a merged firm will exceed the average of its constituents' earnings multiples if the
new growth rate exceeds the average of the constituents' growth rates. The earnings
multiple of a combined entity will also be affected by its risk. The risk of the combined
firm will be decreased, increased, or remain the same depending upon whether the
covariance of the earnings stream with the market is decreased, increased, or remains
the same.
The Larson-Gonedes model holds that the exchange ratio will be determined by
each firm's assessment of the postmerger price/earnings multiple and postulates that
each firm requires that its equivalent price per share be at least maintained as a result
of the merger. Their model is summarized algebraically and graphically by Conn and
Nielsen [1977], using the following symbols:
ER = exchange ratio,
P = price per share,
EPS = earnings per share,
PE = price/earnings multiple,
E = earnings,
S = number of common shares outstanding,
AER = actual exchange ratio.
In the formulations that follow, subscripts 1, 2, and 12 are used to refer to the ac-
quiring, acquired, and combined firms, respectively.
An exchange ratio is determined depending upon postmerger advantages expected
from the combined firm subject to the stockholder wealth constraint. This is expressed
as follows:
P12 >P1. (20.4)
We use the equality relationship. The market price per share for the combined firm
is defined in terms of earnings and the price/earnings ratio as
Table 20.3
Number of Market
Total Shares of Earnings Price/Earnings Price per
Earnings, Common Stock, per Share, Ratio, Common Share,
E S EPS PE P
An example will illustrate the nature of Eq. (20.8a) and some subsequent rela-
tionships. Let us assume that firm 1 and firm 2 are contemplating a merger in which
firm 1 will acquire firm 2. Table 20.3 contains the information gathered about the
two firms.
Using the data from Table 20.3 in Eq. (20.8a), we obtain
ER, = gg + 2400000 PE 12, (20.8b)
which is expressed in simplified form as
ER1 = —1 + IPE,„. (20.8c)
We then use some illustrative values of PE12 to indicate the required maximum
exchange ratio that firm 1 may offer the shareholders of firm 2 if the wealth constraint
for the shareholders of firm 1 is to be satisfied:
PE12 0 7 10 11 12 15 20 30
Max ER1 —1 0.4 1 1.2 1.4 2.0 3.0 5.0
These data, of course, plot on a straight line, as illustrated in Fig. 20.1. We have
discussed the maximum ER1. Let us now consider the minimum ER, that may be
ER
10 — •1
9—
8—
7—
6—
III II ~fER1
5— •
4—
A
3— KI F D „op
-TR ———
2
1 ER 2
0 PE1 2
10 12 14 16 18 20 22 24 26 28 30
--1
Figure 20.1
Influence of AER and PE 12 on merger gains and losses.
760 MERGERS AND RESTRUCTURING: TESTS AND APPLICATIONS
Using the equality form of Eq. (20.9), we rewrite it utilizing the accounting
determination of P 1. This is simply the price/earnings ratio times the earnings per
share for the combined firm,
P2 (20.10a)
P12 = (PE12)(EPS12) =
ER2
and substituting for the earnings per share from Eq. (20.6),
P2 (PE12)(E1 E2) (20.10b)
P12
ER 2 S i + S2(ER2)
We then solve for ER 2 :
P 2 S1 + P2S2(ER2)
= (PE12)(E1 + E2),
ER 2
P 2 S,
ER2 =(PE (20.11a)
,2 )(E, + E 2 ) P 2 S2
Substituting the numerical values of our example, we obtain
ER2 = 4000 = 10
(20.11b)
400PE 12 — 4000 PE12 — 10
Again, we show the nature of the relationship conveyed by Eq. (20.11b) by cal-
culating values of ER 2 for a range of values of PE 12 :
PE12 11 12 15 20 30
Min ER 2 10 5 2 1 0.5
The above data are plotted in Fig. 20.1; as indicated, ER 2 is in the form of a hyper-
bola. Two relationships of significance are portrayed in the figure. One is the inter-
section of the two ER lines. The other is the four quadrants formed by the two lines.
The two ER lines intersect at a PE 12 of 15, and ER, = ER 2 = 2. Thus at an exchange
ratio of 2 and a PE 12 of 15, neither firm will have immediate gains or losses from
the merger. The PE 12 that brings about this result represents a weighted average of
the two PE ratios. The weighted average can be expressed in two forms as shown
below:
P 1 S 1 + P 2 S2 2000 + 4000
PE1z (20.12a)
Ei E2 400 15
(PE1)S1 + (PE 2)S2 1000 + 2000
= 15. (20.12b)
#l + S2 200
TERMS OF MERGERS 761
Premium Merger
or (discount) Premium to
Conn and Nielsen make the calculations with reference to announcement date,
to merger consummation date, and for the month following the consummation of
the merger. The results are summarized in Table 20.5.
Two statistical tests were employed by Conn and Nielsen. A nonparametric
chi-square test of the hypothesis of equal likelihood that a merger will fall in any
one of the four quadrants is rejected at the I% level. A first-difference test of the
Larson-Gonedes model was also statistically significant. Conn and Nielsen observed
that the statistical tests supported the Larson-Gonedes model, but they also noted
and were concerned about the relatively large 40% of the mergers that did not fall
into quadrant I. But rarely did the stockholders of the acquiring firm gain while the
acquired stockholders lost. In any event the pattern suggests a competitive market
among acquiring firms in merger activity.
Conn and Nielsen noted that a large number of mergers fell into quadrant III;
in these mergers the postmerger market valuation of the combined entity was less
Consummation
I 72 64 7 0 1
II 36 2 32 2 0
III 19 1 1 15 2
IV 4 0 0 1 3
Month following
consummation
I 67
II 40
III 18
IV 6
* The roman numerals I, II, III, IV refer to the quadrants in which the ex post price/earnings
ratios fall based on the prior decision rules.
From R. L. Conn and J. F. Nielsen, "An Empirical Test of the Larson-Gonedes Exchange
Ratio Determination Model," Journal of Finance, June 1977, 754.
MANAGERIAL POLICIES IN A VALUATION FRAMEWORK 763
than the sum of the valuations of the acquiring and acquired firms prior to the merger
announcement. However, Conn and Nielsen acknowledged that leakages of merger
information might occur even earlier than the three to five months lead time they
used in their study. They also recognized that analyzing the effects of the merger only
one month after the consummation date might be too restrictive. It would be of
interest to analyze the results over longer periods of time subsequent to the merger.
Finally, Conn and Nielsen did not provide any numerical measures of possible changes
in risk resulting from the merger.
G. MANAGERIAL POLICIES IN A
VALUATION FRAMEWORK
In the perspective of alternative merger theories and empirical tests, the foundation
has been provided for guides to managerial policies toward merger and acquisition
decisions. From an operational standpoint, mergers and acquisitions should be re-
lated to a firm's general planning framework. These requirements have been set forth
in detail in other studies [Weston, 1970]. Here we focus on merger policies in a val-
uation framework. We make the concepts explicit by using an illustrative case example
to convey the ideas.
The Allison Corporation is a manufacturer of materials handling equipment with
heavy emphasis on forklift trucks. Because of a low internal profitability rate and
lack of favorable investment opportunities in its existing line of business, Allison is
considering a merger to achieve more favorable growth and profitability opportuni-
ties. It has made an extensive search of a large number of corporations and has
narrowed the candidates to two firms, for a number of considerations. The Connors
Corporation is a manufacturer of agricultural equipment and is strong in research
and marketing. It has had high internal profitability and substantial investment op-
portunities. The Dorden Company is a manufacturer of plastic toys. It has a better
profitability record than Connors.
Some relevant data on the three firms are summarized in Table 20.6.
Additional information on market parameters includes a risk-free rate, r f , of 6%
and an expected return on the market, E(R,,), of 11%. Each firm pays a 10% interest
rate on its debt. The tax rate, T„ of each is 507.. A period of 10 years is estimated
for the duration of supernormal growth, T.
From the information provided we can first formulate the accounting balance
sheets for the three firms (Table 20.7).
Dividing the internal profitability rate ] by (1 - -c c ) and multiplying by total
assets, we get the net operating income. From the net operating income we can obtain
the market price per share and the total market value that would have to be paid
for each of the three companies (Table 20.8).
We now have earnings per share, market values per share, and total market
values of equity for use in the subsequent analysis.
One popular criterion for evaluating desirability of making acquisitions from the
standpoint of the acquiring company is to determine the effect on its earnings per
share. Table 20.9 illustrates these effects based on the data in the present example.
It can be seen that the merger effects on Allison's earnings per share is a sub-
stantial decline. The percentage dilution in Allison's earnings per share would be
47% if Connors were acquired, and 39% if Dorden were acquired. We believe that
this widely used criterion is in error. It is the effect on market value that is relevant,
not the effect on earnings per share.
In a valuation framework it is necessary to make a forecast of the key variables
affecting value after the merger has taken place. This requires an in-depth business
analysis of each proposed merger in terms of its impact on the key valuation factors.
From the background provided, we observe that Allison is a manufacturer of mate-
rials handling equipment. Connors is a manufacturer of agricultural equipment with
strength in research and marketing. Dorden is a manufacturer of plastic toys. While
Dorden has a better profitability record than Connors, the toy industry is under the
pressure of continuously creating new ideas and concepts if growth and profitability
of a toy manufacturing firm are to continue. In addition, there seems to be less
potential for favorable interaction of management capabilities in a merger between
Allison and Dorden than there would be in a merger between Allison and Connors.
Connors is known to have a strong research organization that may be able to develop
new products in Allison's area of materials handling equipment. This merely sketches
the kind of favorable carryover of capabilities that may be achieved in a merger be-
tween Allison and Connors. Reflecting these qualitative considerations, the following
estimates are made of the new financial parameters of the combined firms:
NOI
We can now proceed to evaluate the two alternative acquisition prospects, using
a valuation analysis. First we calculate the new beta for the merged company under
the two alternatives. We assume the beta for the combined companies is a market
value—weighted average of the betas of the constituent companies. We use the new
betas in the security market line equation to obtain the cost of equity capital for
each of the two combined firms:
( 10 1.4 ( 66
flAc ---- 1.2
+ 66) + 1() + 66)
= .1579 + 1.2158 = 1.374 = 1.37.
ks(AC) = R f + [E(Rm) — R f igAc
= .06 + [.05]1.37 = .1285 = 12.85%.
( 10 1.6 66
fiAD — 1.2 +
10 + 66) 10 + 66)
= .1579 + 1.3895 = 1.547 = 1.55.
ks(AD) = .06 + .05(1.55)
= .1375 = 13.75%.
Given the debt cost of 10% and the cost of equity capital as calculated, we can
then proceed to determine the weighted average cost of capital for the two combined
firms:
AC AD
Debt, B 27 27
Equity, S 76 76
Value, V L 103 103
where
K = the percentage of earnings invested in new assets, K 1,
r = the tax-adjusted rate of return on new assets, r > WACC,
T = the number of years during which r > WACC (i.e., the number of years during
which the firm experiences a rate of growth greater than the economy).
For ease of computation, we shall combine the first two terms in the equation.
We can define V L as
E(NOI)(1 — c3 )
= +
as
E(NOI)(1 — T 3 ) E(NOI)(1 — r e)
VL =
WACC k,
The two are equivalent formulations of V L, so we can use the simpler, latter equa-
tion. We can now insert the numerical values to determine the value of the combined
firm if Allison merges with Connors (AC) or alternatively with Dorden (AD). The
results are shown below:
Merger AC:
[ .14 — .108 1
18(.5)
VL = + 1[18(5)] 10
.108 .108(1 + .108)
( .032 )
= 83.33 + 9 10
.119664
= 83.33 + 90(.2674)
= 83.33 + 24.07
= 107.40 107.
Merger AD:
[.13 — .115 1
16(.5)
V L = .115 + 1[16(.5)] 10
.115(1 + .115)
( .015 )
= 69.57 + 8 10
.128225
= 69.57 + 80(.1170)
= 69.57 + 9.36
= 78.93 z 79.
Using the results obtained, we make a summary comparison of the gains or
losses from the two alternative mergers. The data are summarized in Table 20.10.
768 MERGERS AND RESTRUCTURING: TESTS AND APPLICATIONS
Allison/Connors Allison/Dorden
(millions) (millions)
Postmerger value, V $107 $79
Less amount of debt, B — 27 27
Value of equity, S 80 52
Less Allison's premerger market value 10 10
Gain in equity value 70 42
Cost if acquired at market price 66 66
Gain in value (loss) 4 (24)
The data show that based on estimates of the key parameters a gain in value of
$4 million would result from a merger between Allison and Connors. However, the
merger between Allison and Dorden would result in a loss in valuation amounting
to $24 million We believe that comparing the effects on value represents the con-
ceptually correct way of approaching merger analysis from a managerial standpoint.
The results of this comparison permit some margin of error yet clearly indicate that
a merger between Allison and Connors is preferable to a merger between Allison and
Dorden. Indeed, the gain in value of $4 million could be divided between the share-
holders of Allison and those of Connors. Allison could pay a premium over the cur-
rent market price of Connors and still achieve a gain in net value that would go
to its shareholders.
The foregoing illustrative case example provides a general methodology for the
management analysis of merger activity, which utilizes a number of principles: The
acquiring firm is considering other firms as alternative merger candidates. To come
up with a rational basis for analysis, prospective returns and risk from alternative
merger combinations must be estimated. While historical data may be used as inputs,
a forecast or estimate must be made of the returns and risk that may arise after
alternative merger combinations have taken place.
Thus the forecast of the variables that measure prospective returns and risk for
alternative postmerger combinations is critical to a sound evaluation of merger alter-
natives. The estimates of net operating earnings and of their potential growth may
or may not reflect synergy between the combining firms, depending on the nature
and potential of the combined operations. Analysis in depth of the relevant product
markets and the results of combining the organizations of the two firms is required.
The resulting forecasts are subject to prediction errors that are sometimes of sub-
stantial magnitude.
We may obtain the measures of risk by market value—weighted averages of the
betas (the systematic risk) of the combining firms. With the estimates of the new
betas, along with a selection of market parameters, we can calculate the new relevant
cost of capital for the merged firm, utilizing the security market line relationship. We
must also estimate the effect of alternative merger combinations on the cost of debt.
With estimates of the cost of equity capital and the cost of debt, we must formulate
appropriate capital structure targets for the combined firm and use these to estimate
a cost of capital.
PROBLEM SET 769
Having obtained an estimate of the applicable cost of capital and the estimates
of returns discussed earlier, we can apply valuation principles to formulate estimates
of the value of alternative merger combinations. From these, we deduct the value of
the acquiring firm in the absence of the merger to determine the total value remaining,
which we next compare with the cost of acquiring the firm or firms with which a
merger is being considered. If the value contributed by the merger exceeds the cost
of the acquisition, the acquiring firm has a basis for making an offer that includes a
premium to the shareholders of the acquired firm, yet still provides an increase in
value for the shareholders of the acquiring firm.
Merger analysis thus involves the application of the same basic principles of cost
of capital and valuation discussed in earlier chapters. However, the problems of appli-
cation are more difficult. While merger analysis is fundamentally a form of capital
budgeting analysis, the magnitudes of the alternative projects are usually quite large.
Unlike standard capital budgeting projects in fields related to the firm's past experi-
ence, some mergers involve the analysis of industrial activities quite different from
the firm's own experience. Prediction errors can therefore be substantial. Our empha-
sis has been to avoid compounding large potential errors of forecasting with unsound
valuation procedures. What we have done in the foregoing illustration has been to
utilize the predictions of the relevant return and risk data and apply sound valuation
procedures to the process of determining whether the costs of a prospective acquisi-
tion result in commensurate increases in prospective values.
SUMMARY
The first part of this chapter was summarized in section E. Only the second part of
the chapter will be covered here. Larson and Gonedes constructed a model specifying
a region of mutually beneficial mergers. Conn and Nielsen used this model to measure
market value changes in order to test the hypothesis that mergers were undertaken
to increase values. They found some support for the hypothesis. They also observed
that information about the intended merger appeared to be capitalized in the market
well in advance of the actual merger announcement.
In the perspective of alternative merger theories and tests, we developed a frame-
work for managerial analysis of prospective mergers. Basically, good forecasts of
postmerger returns and risks are required as a starting point. Standard capital budget-
ing procedures, cost of capital analysis, and valuation principles presented in the
preceding chapters are then applied. The aim is to determine whether the value of
the merged firm exceeds the value of the constituent firms. If it does, the merger has
a valid social and private justification.
PROBLEM SET
20.1 Firm 1 and firm 2 are contemplating a merger in which firm 1 will acquire firm 2. The
information in Table Q20.1 has been developed on the two firms. In the questions and answers
related to this case, let subscripts 1 and 2 refer to the individual firms. Let ER, stand for the
770 MERGERS AND RESTRUCTURING: TESTS AND APPLICATIONS
Table Q20.1
Number of Market
Total Shares of Earnings Price/Earnings Price per
Earnings, Common Stock, per Share, Ratio, Common Share,
E S EPS PE P
Firm 1 $400 100 $4.00 15 $60
Firm 2 200 400 0.50 30 15
exchange ratio from the standpoint of firm 1 and ER2 stand for the exchange ratio from the
standpoint of firm 2. AER is the actual exchange ratio. PE12 will be the price/earnings ratio
for the merged firm, after the merger.
a) The managements of the two firms are negotiating the terms of the merger, specifically
the number of shares of firm 1 that will be exchanged for one share of common stock of
firm 2. Three alternative criteria are under consideration:
i) The effect on each firm's earnings per share after the merger.
ii) The expected market value of the merged firm's common stock per one original share
immediately after the merger.
iii) The expected market value of the holdings per one original share after synergistic effects
have been developed, e.g., three years after the merger.
Of the three criteria, which would it be most rational for the management and shareholders
of the firms to emphasize?
b) Using the estimates of the following range of postmerger PE12 values, determine the ER1
that will equate P1 to P12:
Make a graph on which ER1 is plotted against PE12 and label the curve ER1.
c) Calculate the ER 2 that will equate P2 to P12 for the PE12 estimates given in part (b).
Plot the curve ER 2 on the graph begun in part (b).
d) At what PE12 do the two curves intersect? What is the significance of this point of inter-
section?
e) For the following combinations, calculate and graph the premium or loss to each firm:
Actual exchange ratio (AER) 0.4 0.4 0.4 0.4 0.2 0.1
Postmerger price/earnings ratio (PE12) 30 26 22 18 16 18
20.2 The Watro Personal Computer Company is considering merger to achieve better growth
and profitability. It has narrowed potential merger candidates to two firms. The Alber Com-
pany, a producer of PBXs, has a strong research department and a good record of internal
profitability. The Saben Corporation operates a chain of variety stores and has a very high
PROBLEM SET 771
Each firm pays 15% interest on its debt and has a 40% tax rate. Ten years of supernormal
growth are forecast, followed by no growth.
a) What are the total assets of each firm?
b) What is each company's NOI if it earns its before-tax r on total assets?
c) What is the indicated market value of each firm?
d) Compare Watro's increase in value as a result of merger at market value with the cost
of acquiring Alber or Saben if the combined firms have the following financial parameters:
Watro-Alber Watro-Saben
20.3 The Jordan Corporation is a manufacturer of heavy-duty trucks. Because of a low internal
profitability rate and lack of favorable investment opportunities in the existing line of business,
Jordan is considering merger to achieve more favorable growth and profitability opportunities.
It has made an extensive search of a large number of corporations and has narrowed the can-
didates to two firms. The Konrad Corporation is a manufacturer of materials handling equip-
ment and is strong in research and marketing. It has had higher internal profitability than the
other firm being considered and has substantial investment opportunities.
The Loomis Company is a manufacturer of food and candies. It has a better profitability
record than Konrad. Data on all three firms are given in Table Q20.3. Additional informa-
tion on market parameters includes a risk-free rate of 6% and an expected return on the
market, E(R„,), of 11%. Each firm pays a 10% interest rate on its debt. The tax rate, 'C c, of each
is 40%. Ten years is estimated for the duration of supernormal growth.
a) Prepare the accounting balance sheets for the three firms.
b) If each company earns the before-tax r on total assets in the current year, what is the net
operating income for each company?
772 MERGERS AND RESTRUCTURING: TESTS AND APPLICATIONS
Table Q20.3
Book
Value Price! Internal Invest-
per Earnings Number Debt for Profit- ment Growth
Share, Ratio, of Shares Ratio, Existing ability Rate, Rate,
PE (millions) B/S Leverage Rate, r
c) Given the indicated price/earnings ratios, what is the market price of the common stock
for each company?
d) What will be the immediate effects on the earnings per share of Jordan if it acquires
Konrad or Loomis at their current market prices by the exchange of stock based on the
current market prices of each of the companies?
e) Compare Jordan's new beta and required return on equity if it merges with Konrad with
the same parameters that would result from its merger with Loomis.
f) Calculate the new required cost of capital for a Jordan-Konrad combination and for a
Jordan-Loomis combination, respectively.
g) Compare the increase in value of Jordan as a result of a merger at market values with
the cost of acquiring either Konrad or Loomis if the combined firms have the following
financial parameters:
NOI r WACC
20.4 The Sentry Company and the Wong Corporation have the same market value of $100
million. Both firms have the following identical parameters:
a) What is the initial market value of debt and equity for Sentry and Wong?
b) The correlation between the two firms' cash flows is 0.5. If the two firms merge, the
surviving firm will be worth $200 million What will the market value of debt and equity
in the merged firm be? If there were no other merger effects, would shareholders agree
to the merger?
REFERENCES 773
REFERENCES
, "Going Private: Minority Freezeouts and Stockholder Wealth," Jou]nal of Law and
Economics, October 1984b, 367-401.
Dennis, D., and J. McConnell, "Corporate Mergers and Security Returns," Jou]nal of Financial
Economics, June 1986, 143-187.
Dodd, P., "Merger Proposals, Management Discretion, and Stockholder Wealth," Jou]nal of
Financial Economics, June 1980, 105-137.
, and R. Ruback, "Tender Offers and Stockholder Returns: An Empirical Analysis,"
Jou]nal of Financial Economics, December 1977, 351-374.
Dodd, P., and J. B. Warner, "On Corporate Governance: A Study of Proxy Contests," Jou]nal
of Financial Economics, April 1983, 401-438.
Easterbrook, F. H., and D. R. Fischel, "Proper Role of a Target's Management in Responding
to a Tender Offer," Ha]va]d Law Review, April 1981, 1161-1204.
Eckbo, B. E., "Examining the Anti-Competitive Significance of Large Horizontal Mergers,"
Ph.D. dissertation, University of Rochester, 1981.
, and P. Wier, "Antimerger Policy under the Hart-Scott-Rodino Act: A Reexam-
ination of the Market Power Hypothesis," Jou]nal of Law and Economics, April 1985,
119-149.
Ellert, J. C., "Mergers, Antitrust Law Enforcement and Stockholder Returns," Jou]nal of
Finance, May 1976, 715-732.
, "Antitrust Enforcement and the Behavior of Stock Prices," Ph.D. dissertation, Grad-
uate School of Business, University of Chicago, 1975.
Fama, E.; L. Fisher; M. Jensen; and R. Roll, "The Adjustment of Stock Prices to New Infor-
mation," Inte]national Economic Review, February 1969, 1-21.
Fama, E., and J. MacBeth, "Risk, Return and Equilibrium," Jou]nal of Political Economy, May
1973, 607-636.
Firth, M., "The Profitability of Takeovers and Mergers," Economic Jou]nal, June 1979,
316-328.
Goldschmid, H. J.; H. M. Mann; and J. F. Weston, Indust]ial Concent]ation- The New Lea]ning.
Little, Brown and Company, Boston, Mass., 1974.
Grossman, S., and 0. Hart, "Takeover Bids, the Free-Rider Problem and the Theory of the
Corporation," Bell Jou]nal of Economics, Spring 1980, 42-64.
, "The Allocational Role of Takeover Bids in Situations of Asymmetric Information,"
Jou]nal of Finance, May 1981, 253-270.
Halpern, P. J., "Empirical Estimates of the Amount and Distribution of Gains to Companies
in Mergers," Jou]nal of Business, October 1973, 554-575.
Hite, G. L., and J. E. Owers, "Security Price Reactions around Corporate Spin-off Announce-
ments," Jou]nal of Financial Economics, December 1983, 409-436.
Hogarty, T., "The Profitability of Corporate Mergers," Jou]nal of Business, July 1970, 317-327.
Jain, P. C., "The Effect of Voluntary Sell-off Announcements on Shareholder Wealth," Jou]nal
of Finance, March 1985, 209-224.
Jarrell, G. A., and M. Bradley, "The Economic Effects of Federal and State Regulations of
Cash Tender," Jou]nal of Law and Economics, October 1980, 371-407.
Jensen, M. C., "Takeovers: Folklore and Science," Ha]va]d Business Review, November—
December 1984, 109-120.
REFERENCES 775
, and W. H. Meckling, "Theory of the Firm: Managerial Behavior, Agency Costs and
Ownership Structure," Journal of Financial Economics, October 1976, 305-360.
Kelly, E. M., The Profitability of Growth through Mergers, Pennsylvania State University,
University Park, Pa., 1967.
Keown, A. J., and J. M. Pinkerton, "Merger Announcements and Insider Trading Activity: An
Empirical Investigation," Journal of Finance, September 1981, 855-870.
Krouse, C. G., "A Test of Competition in the Capital Market," manuscript of a presentation
to UCLA, AT&T Seminar, August 13, 1975.
Kummer, D. R., and J. F. Hoffmeister, "Valuation Consequences of Cash Tender Offers,"
Journal of Finance, May 1978, 505-516.
Langetieg, T. C., "An Application of a Three-Factor Performance Index to Measure Stock-
holder Gains from Merger," Journal of Financial Economics, December 1978, 365-383.
Larson, K. D., and N. J. Gonedes, "Business Combinations: An Exchange Ratio Determination
Model," Accounting Review, October 1969, 720-728.
Lease, R. C.; J. J. McConnell; and W. H. Mikkelson, "The Market Value of Control in Publicly
Traded Corporations," Journal of Financial Economics, April 1983, 439-471.
Lev, B., and G. Mandelker, "The Microeconomic Consequences of Corporate Mergers," Jour-
nal of Business, January 1972, 85-104.
Linn, S. C., and J. J. McConnell, "An Empirical Investigation of the Impact of Antitakeover
Amendments on Common Stock Prices," Journal of Financial Economics, April 1983,
361-399.
Malatesta, P. H., "The Wealth Effect of Merger Activity and the Objective Functions of
Merging Firms," Journal of Financial Economics, April 1983, 155-181.
, "Measuring Abnormal Performance: The Event Parameter Approach Using Joint
Generalized Least Squares," Journal of Financial and Quantitative Analysis, March 1986,
27-38.
, and R. Thompson, "Partially Anticipated Events," Journal of Financial Economics,
June 1985, 237-250.
Mandelker, G., "Risk and Return: The Case of Merging Firms," Journal of Financial Economics,
December 1974, 303-335.
Manne, H. G., "Mergers and the Market for Corporate Control," Journal of Political Economy,
April 1965, 110-120.
McDaniel, M. W., "Bondholders and Corporate Governance," Business Lawyer, February
1986, 413-460.
Miles, J. A., and J. D. Rosenfeld, "The Effect of Voluntary Spin-off Announcements on Share-
holder Wealth," Journal of Finance, December 1983, 1579-1606.
Myers, S. C., and N. S. Majluf, "Corporate Financing and Investment Decisions When Firms
Have Information that Investors Do Not Have," Journal of Financial Economics, June
1984, 187-222.
Roll, R., "The Hubris Hypothesis of Corporate Takeovers," Journal of Business, April 1986,
197-216.
, "Empirical Evidence on Takeover Activity and Shareholder Wealth," in Copeland, ed.,
Modern Finance and Industrial Economics, Chapter 5. Basil Blackwell, New York, 1987.
Rumelt, R. P., Strategy, Structure, and Economic Performance. Graduate School of Business
Administration, Harvard University, Boston, Mass., 1974.
776 MERGERS AND RESTRUCTURING: TESTS AND APPLICATIONS
A. THE IMPORTANCE OF
INTERNATIONAL FINANCE
International finance in recent years has taken on great significance. Widely fluctu-
ating exchange rates have affected not only profits and losses from changes in foreign
currency values but also the ability to sell abroad and to meet import competition.
For example, suppose that a Japanese auto producer needs to receive 1.2 million yen
per car to cover costs plus a required return on equity. At an exchange rate of 200 yen
to the dollar, a rate that existed in the late 1970s, the Japanese producer would have
to receive $6000 for an automobile sold in the United States. When the exchange rate
is 265 yen to the dollar, as existed in early 1985, dividing the 1.2 million yen by 265
tells us that the dollars required now are $4528. Thus the Japanese producer is in a
position to either reduce his dollar price by approximately 25% and still receive the
same number of yen or take higher profit margins on sales in the United States. By
early 1987 the yen had moved to about 150 to the dollar. The Japanese car would then
have to be priced at $8000, or profit margins (for the Japanese) would suffer. Of
course, the success of Japanese auto companies in the United States has not been
completely due to changes in foreign exchange rates alone. Auto producers in Japan
have achieved improved production processes that have resulted in greater produc-
tivity and high-quality cars. But exchange rate movements have also been a factor, as
the above example illustrates.
777
778 EXCHANGE RATE SYSTEMS AND PARITY CONDITIONS
From the standpoint of U.S. companies selling products abroad, the rising value
of the dollar in relation to foreign currencies has the opposite consequences. For
example, suppose that an American producer is selling a product in the United
Kingdom, and to meet competition, it has to be sold for 500 pounds. When the pound
had a value of $2.25, as it did in the late 1970s, the dollar amount received by the U.S.
seller would be $1125. By March 1985 the value of the pound had fallen to $1.07. If
the U.S. producer continued to sell the product for 500 pounds, he or she would now
receive $535. If the original $1125 represented a dollar price necessary to earn its cost
of capital, the U.S. firm would find it difficult to survive with a price that had declined
by more than 50 percent. Or alternatively, to continue to realize a dollar price of
$1125 at the exchange rate of $1.07 to the pound would require a new selling price
of 1051 pounds, a price increase of over 100% expressed in pounds. By early 1987
the British pound had risen to the $1.50 range about midway between the two ex-
tremes. This moderated the unfavorable impact on U.S. sales in the United Kingdom.
From a general economic standpoint the main issues in international finance
relate to efficient trading rules, the adjustment processes for achieving international
equilibrium, and the issue of whether there are segmented international markets. Our
emphasis in this book will be on the implications of adjustment processes for corpo-
rate financial policy. These in turn have implications for the activities of firms, such
as sales, purchases, and investment policies.
The nature of risks in an international financial setting takes on new dimensions.
We shall focus on corporate financial policies to manage these risks. In addition to
the pattern of cash inflows and cash outflows that the firm develops, we shall examine
the changes in its balance sheet in terms of monetary versus nonmonetary net posi-
tions. We shall also examine the use of the forward market for dealing with foreign
exchange fluctuations and analyze the use of money and capital markets for man-
aging foreign exchange risks. Issues here are whether outlays to limit the risks of ex-
change rate fluctuations are worth the cost. These activities will also be examined for
their influence on keeping the international financial markets performing as efficient
markets.
Our presentation will emphasize some basic propositions in international finance
that are the key to measuring returns and costs in international financial activities.
These basic relations are best understood after background material on the adjust-
ment processes in international finance have been developed.
savings that increase productivity and provide opportunities for saving surplus units
to postpone consumption and for deficit units to increase the output of real goods
by utilizing these savings of (surplus) units willing to consume less now and more
later.
X
Country A $3 $1
Country B M12 M6
At more than 6 marks per dollar, both goods X and Y would be cheaper in country
B. Country B would export both X and Y and import neither. For example, at
M8 = $1, B could sell X in A for $1.50 and Y in A for $0.75. On the other hand, a
rate of less than 4 marks per dollar would make both goods cheaper in country A.
Country A would export both and import neither. For example, at M2 = $1, A could
sell X in B for M6 and Y in B for M2. To achieve equilibrium between the two
countries, the exchange rate would have to fall somewhere between 6 and 4 marks
per dollar. For example, at M5 = $1, B can sell X in A for $2.40, and A can sell Y
in B for M5. A basis for cross trade would now exist.
In the example the relative prices of X and Y are 3:1 in country A and 2:1 in
country B. With a large number of products the cheaper one country's currency is in
relation to other currencies, the larger the range of that country's products that are
underselling foreign products of the same type. As a result the importing country will
need greater amounts per unit of time of the exporting country's currency in order
to buy the latter's relatively cheap goods. Differences in the patterns of relative prices
result from differences among countries in resources, skills, and tastes and in social
and political conditions, which in turn lead to comparative advantages in different
kinds of activities. As a consequence, there will be profit incentives for businesspeople
780 EXCHANGE RATE SYSTEMS AND PARITY CONDITIONS
to engage in trade. These private benefits will, in turn, lead to social gains as the
theory describes. Exchange rates alone could bring the trade between countries to
levels on which the exports and imports of individual countries will be in balance in
their own currency.
However, the overall balance is affected by the existence of short- and long-term
capital flows associated with borrowing and lending activities, shifts in the compara-
tive rates of development of individual industries in different countries, and differences
in the domestic economic development programs among individual countries. Pos-
sible imbalances in international activities can be restored to relative equilibrium by
these adjustment processes. We discuss the adjustment processes for international dis-
equilibrium under two alternative exchange mechanisms. One is a gold standard with
fixed exchange rates. The other postulates flexible exchange rates.
The reverse process takes place for country B, the deficit area.
THE INTERNATIONAL FINANCIAL MECHANISM 781
Under the gold standard the exchange rates are said to remain "fixed" through
this entire adjustment process, because of the inherent assumption that gold would
flow to prevent exchange rates from moving beyond the "gold points." For example,
let us suppose that the British pound contains four times as much gold as the U.S.
dollar: the dollar contains .05 ounce of gold, whereas the British pound contains .2
ounce of gold. A U.S. trade deficit vis-à-vis the United Kingdom would increase the
demand for pounds to pay for the imports from the United Kingdom, and the dollar
price of pounds would therefore rise, say to $4.10. We could take $1000 to the U.S.
Treasury and get 50 ounces of gold since $1 contains .05 ounce of gold. Next we could
transport the 50 ounces of gold to the United Kingdom, where we would get £250 in
exchange: 50 ± .2 = 250. With the rate of exchange of $4.10 to £1 less the cost of
transporting the gold amounting to $.02 for insurance and interest loss, we would net
$4.08 for each pound. Multiplying $4.08 times £250 yields $1020, or a profit of $20.
Thus at any rate of exchange above $4.02 or below $3.98, by the actions of the gold
arbitrageurs, the rate of exchange would be checked from falling outside the gold
points.
To the extent that a gold standard with fixed exchange rates worked, it was
because maintaining two-way convertibility between a nation's monetary unit and a
fixed amount of gold was a policy goal that received great emphasis and high priority.
So long as it was recognized that convertibility was a major policy goal, speculative
capital movements were likely to be stabilizing rather than destabilizing. In other
words, the general expectation that the convertibility of the currency would be main-
tained was so strong that when a gold standard currency did weaken almost to its
gold export points, one could reasonably assume that it would not drop much lower
and indeed would probably rise. Speculators would then take positions based on the
expectations that a rise in the value of the currency was imminent, and this would
of course strengthen the currency.
During the post—World War II period the international monetary system was not
a gold standard but rather a gold exchange standard. In addition to gold, nations
held claims on the currency of other countries. Most countries simply related their
exchange rates to one of the "key" or "reserve" currencies, holding part of their official
reserves in that currency. Dollars and sterling were key currencies immediately after
the end of World War II. The monetary authorities outside the United States had the
assurance of the U.S. Treasury that the two-way convertibility between dollars and
gold would be maintained.
Because the growing volume of international trade required an increase in the
supply of international reserves, the initial balance of payments deficits of the United
States during the 1950s performed the role of augmenting the supply of key currencies.
The rate at which new gold was being produced was far too low to meet the rate of
growth in demand for official reserves to support the growing volume of international
trade. Problems developed with the gold exchange standard in the early 1970s, giving
rise to a substantial shift to the use of flexible exchange rates. In the next section,
we describe the nature of the adjustment process under flexible exchange rates. This
provides a foundation for the discussion of the broader issues of achieving interna-
tional equilibrium under current conditions in the international economy.
782 EXCHANGE RATE SYSTEMS AND PARITY CONDITIONS
point in the right direction if the reader remembers the general maxim that arbi-
trageurs will seek to "sell high and buy low." First we discuss the consistency of spot
rates. Let us suppose the dollar value of the pound sterling is $2 in New York City
and $1.90 in London. The following adjustment actions would take place: In New
York City, we sell £190 for $380. We sell the pounds in New York because this is
where the pound value is high. In London, we sell $380 for £200. In London the
dollar value is high in relation to the pound. Thus £190 sold in New York City for
$380 can be used to buy £200 in London at a gain of £10. The sale of pounds in
New York causes the value of the pound to decline, and the purchase of pounds in
London causes their value to rise until no further arbitrage opportunities are left.
Assuming minimal transportation costs, we postulate that the same foreign exchange
prices obtain in all locations.
The relationship between the currencies of two individual localities can be gener-
alized across all countries and is referred to as the consistent cross rate. It works in
the following fashion: Assume that the equilibrium relation between the dollar and
the pound is $2 to £1, and the dollar and the franc is $0.25 to Fr 1. Now, suppose
that in New York City, £.10 = Frt. The following adjustment process will take place:
We sell $200 for £100 and use these to obtain Fr1000, which in turn will buy $250.
We thus make a profit of $50 over our initial $200. In general, then, we sell dollars
for pounds and pounds for francs since the pound is overvalued with respect to the
dollar-to-pound and dollar-to-franc relationships. The dollar will fall in relation to
the pound, and the pound will fall in relation to the franc until consistent cross rates
are reached. If the relation were Fr 1 = £.125, the cross rates would be consistent. We
check using the following relation:
$1 = £.5, £1 = Fr8, Fri = $.25.
The product of the right-hand sides of all three relationships must equal 1. Check.•
.5 x 8 x .25 = 1. We have thus established consistency between foreign exchange
rates.
Table 21.1
of fixed exchange rates, which call for shrinkage in the domestic money supplies and
incomes of countries with external deficits. This shrinkage occurs through contraction
in the reserves of member (of the Federal Reserve System) banks that must make
payments to foreigners in foreign currency when the home country imports more than
it exports. The process is readily summarized by a series of journal entries that would
take place. For an excess of imports into the United States from Japan the journal
entries are as shown in Table 21.1.
On the assumption that credit balances are not expanded, the journal entries
reflect the demand of the Japanese exporter (Mr. Toyama) to be paid. The U.S. com-
mercial bank obtains yen from the U.S. central bank (FED) by reducing its reserves
with the Federal Reserve Bank (FED). The FED had purchased the yen by paying
with gold certificates. The yen received by Mr. Toyama in Japan became primary
deposits in a commercial bank in Japan. The reserves of the commercial banks in
Japan increase and so do the holdings of gold certificates by the central bank of
Japan.
The net result of the journal entries is a decrease in the deposits of the U.S.
importer at that individual's commercial bank accompanied by a reduction in the
commercial bank's reserves with the FED. For the U.S. central bank, member bank
deposits decrease as does its gold certificate account. The opposite net transactions
take place in Japan.
For a normal volume of transactions the exchange rate remains unchanged. How-
ever, there is a contraction in the effective money supply in the United States and
an expansion in Japan. In utilizing its gold or other foreign exchange reserves, the
central bank is covering the country's balance of payments deficits. The shrinkage of
domestic bank reserves would bring about a deflationary adjustment of the classic
gold standard kind. The resulting multiple contraction in the domestic money supply
would deflate incomes and prices, reduce spending on imports, and stimulate efforts
to achieve exports.
THE SHIFT FROM FIXED TO FLEXIBLE EXCHANGE RATES 785
180
Trade-weighted averages
170
160
22 OECD currencies
150
140
130
120
110
100
1983 1984 1985 1986
1. Australia, Austria, Belgium-Luxembourg, Canada, Denmark, Finland, France, Germany, Greece,
Iceland, Ireland, Italy, Japan, Netherlands, New Zealand, Norway, Portugal, Spain, Sweden, Switzer-
land, Turkey, and United Kingdom.
Data: U.S. Department of the Treasury. End- of-month rates. Index rebased by BEA.
2. Belgium, Canada, France, Germany, Italy, Japan, Netherlands, Sweden, Switzerland, and United
Kingdom.
Data, Federal Reserve Board. Monthly average rates. Index rebased by BEA.
U.S. Department of Commerce, Bureau of Economic Analysis
Figure 21.1
Indices of foreign currency price of the U.S. dollar (1977 = 100).
(From A. J. Dilullo, "U.S. International Transactions, Third
Quarter 1986," Survey of Current Business, December 1986, 25.)
1986]. It is the economic policies of individual countries that determine the variability
of the exchange values of currencies. It is becoming recognized that it is the degree
of coordination of economic policies among countries that determines the degree of
fluctuations in foreign exchange rates rather than a fixed rate versus flexible exchange
rate regime.
Nevertheless, proposals have been made to reduce the use of the U.S. dollar as an
international reserve currency. Proposals have been made to extend the experience of
the European monetary system more generally through the International Monetary
Fund and the use of its Special Drawing Rights (SDRs). The nature of each of these
will be briefly described.
central par value (except Italy, which is allowed 6% deviations). An attempt is made
to moderate fluctuations in the foreign exchange values of the currencies of the par-
ticipating countries through intervention in the foreign exchange market. But period-
ically when the strains become large, revaluations of relative curency values formally
take place.
These efforts have been viewed as successful by some, and it has been recom-
mended that they be extended more generally under the aegis of the International
Monetary Fund to the broader world economy [Triffin, 1986]. Fundamentally, this
involves using the IMF as a kind of central bank. The beginning was made in 1970
when the IMF created Special Drawing Rights (SDRs).
Beginning July 1974 the value of the SDR is determined daily by the IMF on
the basis of a basket of currencies, with each currency assigned a weight in the deter-
mination of that value. The currencies of the basket are valued at their market
exchange rates for the U.S. dollar, and U.S. dollar equivalents of each of the cur-
rencies are summed to yield the rate of the SDR in terms of the U.S. dollar. Beginning
January 1, 1986, the SDR valuation basket consists of the currencies of the five
member countries having the largest exports of goods and services during the 1980-
1984 period. The weights for the five currencies are: U.S. dollar, 42%; deutsche mark,
19%; Japanese yen, 15%; French franc and pound sterling, 12% each.
The SDRs are book entries that are credited to the accounts of IMF-member
countries according to specified quotas. They can be used to help meet payments
imbalances or to add to the stock of foreign exchange reserves of the individual
countries. Interest is paid to the countries that hold SDRs and by those that draw
on their SDRs, with the rate based on an average of money market interest rates in
the United States, the United Kingdom, Japan, Germany, and France.
Thus the IMF operates as a kind of central bank to the countries of the world.
It makes provision for the countries to have either balances or borrowings from the
IMF. It is proposed that the central bank status of the IMF be extended so that it
could moderate what some regard as unduly wild currency fluctuations under flexible
exchange rates. The aim would be for "stable but flexible" exchange rate movements.
It is recognized, however, that the basic cause of the fluctuations is lack of coordi-
nation among the policies of the individual countries. Nevertheless, a recent review
of the European monetary system concludes that it has contributed toward economic
convergence among EMS-member countries [Ungerer, Evans, Mayer, and Young,
1986]. In short, it is argued that increasing the use of the IMF as a central bank
and as an international economic supervisory mechanism could potentially lead to
better coordination among the monetary, fiscal, and other economic policies of the
countries of the world.'
The previous analysis has used a balance of payments framework to appraise
the need for international adjustment processes, including the devaluation or depre-
ciation of the dollar. We next consider in somewhat more detail the role of a balance
A full examination of these issues is made in a number of papers (with commentaries) originally presented
at a meeting on exchange rates of the Royal Economic Society and published in the March 1987 issue of
the Economic Journal.
788 EXCHANGE RATE SYSTEMS AND PARITY CONDITIONS
of payments analysis in defining the kinds of policies and adjustments that may
have to be made in corporate financial policy in order to deal with changes in the
international financial environment.
In recent years the U.S. Department of Commerce has stopped using the term balance
of payments in favor of the more general expression U.S. international transactions.
However, the compilations are still a balance of payments analysis and will be dis-
cussed in those terms. The balance of payments of a nation is a double-entry account-
ing statement of its transactions with the rest of the world during a specified time
period, usually one quarter or one year. Inflows are recorded as a plus. (Sometimes
the words receipts and credits are used, but the terms plus and minus are preferred
as being more neutral in their implications.) Outflows are a minus. (Sometimes called
debits or payments, but again, minus is preferred.)
The basic entries in the balance of payments statement can be summarized into
four categories of items as shown in Table 21.2. This summary indicates how the
adjustment process can be complicated by a number of relationships going in different
directions. Thus if a given country runs a deficit by having an excess of imports over
exports, instead of settling the balance by payments in the foreign currency, a number
of other offsets may take place. As shown in Table 21.2, the offsets can take the
form of increasing liabilities to foreigners, decreasing claims on foreigners, liquidating
assets, or decreasing other foreign investments. Consequently, there will be a lag in
the pressures that would result if payments in foreign currencies had to be made im-
mediately. This lag will postpone the contraction in the money supply or incomes
of the deficit country.
Similarly, a country that is increasing its investments abroad will improve its
investment position, but it will be creating minus entries in its current balance of
payments statement. The long-run outlook for a country that is making substantial
foreign investments, however, may be favorable as a result of the future income that
may be generated from those investments.
The U.S. Department of Commerce follows the format for the balance of pay-
ments statement shown in Table 21.3. It has been modified to deemphasize individual
balances.
In the traditional balance of payments statement, lines 1 and 4 give "the balance
on goods and services." This is considered significant as an indicator of the basic
trade position of the United States, which was a negative $103 billion in 1985.
Another balance also emphasized is the current account balance obtained by
adding line 7 (U.S. government grants) and line 8 (remittances, pensions and other
transfers). This was the negative $103 billion plus a negative $15 billion, giving a
total of a negative $118 billion for 1985. This is the balance that will have to be
offset by capital flows and changes in government official reserve assets. U.S. assets
abroad also increased by $32 billion, so the total amount to be offset is $150 billion.
This is exactly equal to the increase in foreign assets in the United States plus a
statistical discrepancy. Basically, in 1985 (and in other recent years) a substantial
current account deficit had to be offset by a substantial flow of foreign investment
into the United States.
In summary, while government reports on U.S. international transactions avoid
the concept of balance of payments, it is nevertheless there. The pattern in recent
790 EXCHANGE RATE SYSTEMS AND PARITY CONDITIONS
years for the United States has been a very substantial negative balance on current
account. In terms of Table 21.2, the negative balance has been further aggravated by
accumulating U.S. private assets abroad. However, a very substantial buildup of
foreign private investments in the United States has taken place to offset the very
large negative U.S. balance on trade and services.
The significance of alternative measures of "the balance of payments" depends
on the circumstances of an individual country and the pattern of international eco-
nomic developments taking place. To determine all contributing factors, one must
analyze as many components of information as possible, including the more general
economic developments taking place in individual countries. The balance of pay-
ments along with other information sources may be analyzed within a more general
economic framework to develop judgments about adjustment processes taking place
that will have implications for changes in foreign exchange rates as well. It is this
kind of analysis that is required for formulating corporate financial policies of firms
substantially affected by changes in foreign exchange rates. We next analyze some
of the fundamental relationships that reflect the broad economic adjustment forces
reflected in balance of payments statements.
E. FUNDAMENTAL EQUILIBRIUM
RELATIONSHIPS
Al. Financial markets are perfect. (Numerous buyers and sellers; no taxes, no in-
formation or transactions costs, no controls.)
A2. Goods markets are perfect. (Numerous buyers and sellers; no transportation costs
or time, no barriers to trade.)
A3. There is a single consumption good common to all.
A4. The future is known with certainty.
A5. The competitive markets are in equilibrium.
The fundamental exchange relationships require one or more of the above as-
sumptions. The following equilibrium relationships can then be established.
where
X, Eo X0 = FC units per dollar now,
—0 =
E, X1 = FC units per dollar one period later,
1
E0 = = dollars per FC unit now,
X0
1
E, = = dollars per FC unit one period later,
X,
X,
CX = = change in exchange rate,
X0
P/P
RPC = f ° = change in relative prices = ratio of inflation rates,
Pd11Pd0
Pf0 = initial price level in the foreign country,
Pf , = foreign country price level one period later,
T1= foreign country inflation rate,
Pao = initial domestic price level,
Pdi = domestic price level one period later,
'
I d = domestic inflation rate.
Some numerical examples will illustrate some of the implications of the purchasing
power parity doctrine. Let us assume that for a given time period foreign price levels
have risen by 32%, whereas domestic price levels have risen by 20%. If the initial
exchange rate is FC 10 to $1, the subsequent new exchange rate will be
1.32 X,
= , X, = 1.1(10) = 11.
1.20 10
It will now take 10% more foreign currency units to equal $1 because the relative
inflation rate has been higher in the foreign country. Alternatively, with an exchange
rate of FC 10 to $1, let us assume that foreign prices have risen by 17% while domestic
CX b Q Figure 21.2
Alternative parity relationships.
RPC
FUNDAMENTAL EQUILIBRIUM RELATIONSHIPS 793
prices have risen by 30%. The expected new exchange rate would be
1.17 X,
= X, = .9(10) = 9.
1.30 10 '
In the present instance the number of foreign currency units needed to buy $1 would
drop by 10%. Thus the value of the foreign currency has increased by 10% due to
the differential rates of inflation in domestic vs. foreign prices.
The approximation relationship is
X1- X0
Tr.
X0
Using the data from our above example, we would have
11 — 10
.32 — .20
10
.12 = .10.
The nature of the relationships can vary from the precision suggested by the
numerical examples. Some of the alternatives are suggested in Fig. 21.2. Line a implies
that both the average and marginal relations are one to one. Line b implies that the
marginal relation is one to one, but the average relation is not. Line c implies a
relationship, but not a one-to-one relationship.
Empirical studies indicate that while the purchasing power parity relationship
does not hold perfectly, it holds on average [Pippenger, 1986]. More fundamentally,
the doctrine predicts that an equilibrium rate between two currencies will reflect
market forces and that random deviations from the central tendency will tend to be
self-correcting; i.e., it suggests the existence of some strong equilibrating forces. Fur-
thermore, it argues that the relations between exchange rates will not be haphazard
but will reflect underlying economic conditions and changes in these conditions. The
relationships are not precise because of a number of factors. These include:
1. Differences in incomes or other endowments between the two countries.
2. Differences in tastes and/or market baskets consumed.
3. Changes in government policies.
4. Transportation costs.
5. Lags in market responses.
6. Differences between the two countries in the price ratios of internationally traded
goods to domestically traded goods.
7. The addition of a risk premium influence.
demand for money causes prices to rise by 10%, relative prices should remain un-
changed. Although change in the quantity of money may affect prices and exchange
rates, neutrality implies that the rate at which domestic goods are exchanged for
foreign goods should not be changed (PPPT). Neither should there be an effect on the
rate at which goods today are exchanged for goods in the future (the Fisher relation).
The Fisher relation states that nominal interest rates rise to reflect the anticipated
rate of inflation. The Fisher relation can be stated in a number of forms, as shown
below:
1
Po— + r (21.2a)
P1 1 +
o
1 + r = (1 + R„) , (21.2b)
Pt
1)
R„=[(1+ r)(1, 1, (21.2d)
where
We shall also show how the Fisher relation in an international setting can be
used to derive PPPT as well as the interest rate parity theorem presented next.
Xf Rf 0 E 0
— (21.3)
Xo 1 ± Rdo E f
where
X f = current forward exchange rate expressed as FC units per $1,
E f = current forward exchange rate expressed as dollars per FC 1,
X, = current spot exchange rate expressed as FC units per $1,
E 0 = current spot exchange rate expressed as dollars per FC 1,
R10 current foreign interest rate,
=
i.e., 10.45 units of foreign currency equal $1. Thus, viewed on an annual basis, the
foreign forward rate is seen to be at a discount of 4.5%. If we assume the time period
of a transaction to be 90 days, we must rework the problem accordingly. The first
step is to prorate the interest rates on a quarterly basis. Thus for 90 days,
1.0375
X = (10)
f 1.025
= 10.122;
i.e., the 90-day forward rate is 10.122, and it follows that on the quarterly basis the
discount on the 90-day forward rate is 1.22%.
The approximation for IRPT is
Xf — X0
R fi — Rdi 11
= •
Xo
796 EXCHANGE RATE SYSTEMS AND PARITY CONDITIONS
11
4. Earn Rf o (36 0 )interest
on AoX0 •
AO2O [} + Rf0
(1)]
n\
1. Borrow A o in $ in U.S. 7. Pay back Ao El!Rdo■3"\11 •
J
Figure 21.3
Covered interest arbitrage with investment outflows.
The data for the annual example are used to illustrate the approximation:
10.45 — 10
.15 — .10 =
— 10
0.05 = .045.
The IRPT can also be used to illustrate another general proposition for inter-
national finance. In the absence of market imperfections, risk-adjusted expected real
returns on financial assets will be the same in foreign markets as in domestic markets.
Equilibrium among the current exchange rate, the forward exchange rate, the domes-
tic interest rate, and the foreign interest rate is achieved through covered interest
arbitrage. Assume that the forward contract is for n days and that R f , and R„ are
annual rates. A sequence of seven transactions takes place when the interest rate dif-
ferential in X exceeds the forward exchange discount on the currency of X (Fig. 21.3).
Equilibrium occurs when the principal + interest earned in country X equals the
principal + interest paid in the United States, or when
A0)1(3[1 + Rf 0(1/360)] A 0 [1 + Rd0 (36 X f 1 + (n/360)R 10
X f
4' or X0 1 + (n/360)R,0
Alternatively, we would obtain the same equilibrium requirement if the interest
rate differential in X were less than the forward exchange discount in the currency
FUNDAMENTAL EQUILIBRIUM RELATIONSHIPS 797
1
3
Figure 21.4
Covered interest arbitrage with investment inflows.
of X, by borrowing in the foreign country and buying FC to cover the amount
borrowed plus interest owed (see Fig. 21.4).
Equilibrium occurs when the principal plus interest earned in the United States
equals the principal plus interest paid in country X, or when
,
( X f 1 + R fo (n/360)
{ PD
x-
'1 (360)1}Xj-= Fo[1 Rf9 360)1 or
[ I 4- R° X, 1 + Rao(n/360)
If the interest differentials are positive in the foreign country, the forward rate on
the foreign currency will be negative. If negative, the forward rate on the foreign
currency will be positive, as shown in Fig. 21.5.
Plus Minus
Forward Discount —1%, +2% IF
market rates —2%, +1% IH
(in foreign
currency Premium +1%, —2% IH
units, X4 +2%, —1% IF
Figure 21.5
Covered interest arbitrage.
798 EXCHANGE RATE SYSTEMS AND PARITY CONDITIONS
The pair of relations followed by "IF" indicate foreign investment in the covered
interest arbitrage operation. The pair of relations followed by "IH" leads to invest-
ment in the United States in the covered interest arbitrage operation.
As we have demonstrated, whether or not the covered interest arbitrage opera-
tion results in investment in a foreign country in preference to investment in the
United States, the resulting equilibrium interest and forward exchange parity relation-
ship is the same.
The above general relationships can also be illustrated in some numerical ex-
amples of covered interest arbitrage. First, an arbitrage outflow situation will be
described. The basic assumptions are as follows:
U.S. interest rate 5%,
German interest rate 7%,
Spot exchange rate $1 = DM4,
Forward exchange rate discount 1% per annum.
The following arbitrage transaction will take place. In New York, we borrow $100,000
for 90 days (one quarter) at 5%. The loan repayment at the end of 90 days is
$100,000[1 + (5% x = $101,250.
At the spot exchange rate, we convert the $100,000 loan into DM400,000. In Ger-
many, we invest the DM400,000 for 90 days at 7% and receive at the end of 90 days
DM400,000[1 + (7% x +)] = DM407,000.
An arbitrage inflow takes place when the forward exchange rate discount exceeds
the interest rate differential. The basic facts are now:
U.S. interest rate 5%,
German interest rate 6%,
Spot exchange rate DM4 = $1,
Forward exchange rate discount 2% per annum.
The arbitrage transaction involves borrowing in the foreign country. In Germany,
we borrow DM400,000 for 90 days at 6%. The loan repayment at the end of 90 days
is
DM400,000[1 + (6% x 4)] = DM406,000.
At the spot exchange rate, we convert the DM400,000 loan into $100,000. In New
York we invest the $100,000 for 90 days at 5% and receive at the end of 90 days
$100,000[1 + (5% x = $101,250.
Again, we would make a covering transaction. To ensure coverage for the loan
repayment, we buy DM406,000 forward. At a 2% forward exchange rate discount,
it costs DM4[1 + (2% x = DM4.02 to buy $1 forward. Thus repayment of prin-
cipal plus interest requires
DM406,000 4.02 = $100,995.
Thus we have
Arbitrage profits = investment receipts — loan repayments
= $101,250 — $100,995
= $255.
The arbitrage transaction increases the demand for deutsche marks and increases
the supply of dollars. The U.S. interest rate decreases and the German rate rises; thus
the differential increases. Covering transactions increase the demand for German
forward exchange and increase the spot supply of deutsche marks, thus decreasing
the discount on forward deutsche marks. The interest rate differential increases and
the forward exchange rate discount decreases until both rates are equalized, e.g.,
[1 + (.064/4)]/[1 + (.0481/4)] = (4.0158/4).
We can now show the relationships between the three fundamental equilibrium
relationships thus far set forth. We begin with the international Fisher relation. The
Fisher relation for country f can be written
1 +R f P fi
1)
+r Pfo
And for country d it would be
1 + Rd Pdl
2)
r Pd0
800 EXCHANGE RATE SYSTEMS AND PARITY CONDITIONS
Dividing (1) by (2) (and assuming the real rate of interest is equal across the two
countries), we have
P fl
1 + R f P10
1 + Rd Pd1 .
P d0
+ R1 X .
The left-hand side is the IRPT =
1 + Rd X0
If this investment earns at the domestic real rate, its end-of-period value will be
4) E0F0(1 + r d ).
One plus the real return is the ratio of the end-of-period value of the position in
foreign currency units to the end-of-period value of the domestic investment:
E i (P,IP 1)[F,(1 + R10)]
5) 1 +T1 = •
E o F 0(1 + rd)
If the ratio on the right-hand side is equal to one, there is neither a benefit nor a loss
from having a long position in foreign currency units as compared to having a long
FUNDAMENTAL EQUILIBRIUM RELATIONSHIPS 801
position in domestic currency units. We may next make use of the Fisher relation in
the form
6) 1 + rd = (1 + Rdo )(P o /P,).
We then substitute the right-hand side of the above expression for (1 + rd) in Eq. (f,)
and obtain Eq. (7a). The F o's cancel, and
(E}/E0 )(P0 /P})(1 + Rfo )
7a) 1 + }'1 = •
(} + Rdo)(Po/P1)
The price ratios cancel to give us Eq. (7b):
1/E 0)(1 + R ro)
7b) 1+ = (E •
1 + Rd0
The interest rate parity relationship is written as Eq. (8):
E 0 1 + R}0
8)
Er — Rao •
The right-hand side of the equation is substituted in Eq. (7b), yielding Eq. (9).
E, E,
9) 1 = E.
Next we cancel the E, and solve for r }:
r} = (E)/E4 — 1. (21.4)
Thus the real return (gain or loss) from a long or short position in a foreign currency
or net assets denominated in foreign currency units can be determined by a relatively
simple expression. It is the ratio of the actual value of the foreign currency's future
spot rate in relation to the expected value of its future spot rate reflected in its current
forward rate.' We can illustrate the measurement of real gains or losses by a number
of simple examples. First let us assume that we are long in foreign currency units
and that the relative devaluation of the foreign currency units is more than anticipated
in the current spot rate (Er):
E, = $.20, Er = $.25.
The real gain or loss can therefore be calculated as follows:
1' } = (E,/Ef ) — 1 = (.20/.25) — 1 = —20%.
Thus if we are long in foreign currency units and the forward value of the foreign
currency unit was 5.25 but its actual spot value at the end of the period is $.20, there
has been a loss of 20%. On the other hand, if the spot rate one period later is $.30,
2 The insight that the real return on an investment in foreign currency over the real domestic rate of interest
can be calculated from the divergence between the anticipated and actual future spot rates of exchange is
presented in Farber, Roll, and Solnik [1977].
802 EXCHANGE RATE SYSTEMS AND PARITY CONDITIONS
If we have a short position in the foreign currency units, the signs of the above calcu-
lations will be reversed. Thus we have a powerful but simple measurement of gains or
losses from long or short positions in foreign currency units or in assets denominated
in foreign currency units. The practical implications are that a divergence between the
current forward rate and the future spot rate will produce a real gain or loss. This
means a violation of the forward parity theorem.
An issue raised with regard to the effect of the shift from fixed exchange rates
to flexible exchange rates concerns the possibility of an increase in uncertainty under
flexible exchange rates, which in turn could dampen the rate of growth in interna-
tional trade. But if hedging can be used to limit risks from exchange rate fluctuations,
increased fluctuations will be no problem unless the foreign exchange premium is
increased. Furthermore, if there were an increase in uncertainty and if it were accom-
panied by an increase in expected returns, the reward-risk ratio could remain un-
changed. The simple expression for measuring real gains or losses could thus be used
to measure the extent to which net returns from unhedged positions were increased
or decreased.
A similar model was formulated by Grauer, Litzenberger, and Stehle [1976].
Their model can be expressed as follows:
r 1 = (E i lE f ) — 1 = (X f IX",)— 1. (21.4)
In terms of our previous examples, if we are long in foreign currency units and
the current forward value of the foreign currency unit was $0.25 but its actual spot
value at the end of the period is $0.20, the future spot rate is at a discount from the
current forward rate. This implies that the purchasing power uncertainty reflected in
exchange rates is negatively correlated with world wealth so that the risk premium
represents a reduction rather than an addition to the real rate of return. Conversely,
if the future spot rate is higher than the current forward rate, we obtain a real gain.
SUMMARY 803
If we are in a short position in the foreign currency units, the determination of pre-
miums or discounts will be reversed.
R fi —
X,
In our example we had
10.45 — 10
.15 — .10 =
— 10
0.05 = .045.
This suggests that the interest differential should approximate the forward ex-
change rate differential percentage, measured as shown. The forward differential will
therefore approximate the expected change in the exchange rate. Thus under the effi-
cient market and perfect market assumptions postulated, forward exchange rates
should be unbiased predictors of future spot rates.
SUMMARY
International financial markets direct savings flow into countries that can best employ
the funds. The law of comparative advantage states that trade will be mutually ad-
vantageous since different countries have different relative efficiencies in producing
different goods. A comparative advantage means that even though one country could
produce some good more efficiently in absolute terms, it has other opportunities that
are so superior that it is preferable for some other country—less efficient and with
lower opportunity cost—to produce the good.
With a gold standard, countries with surplus exports receive gold in exchange.
Domestic prices in the country receiving gold rise, and this price increase in turn
stimulates imports to restore the trade balance. Income and employment may also be
affected by the adjustment process. Convertibility of currencies to gold at fixed rates
forces exchange rates to be fixed. With fixed exchange rates, disequilibrium forces all
adjustments to be on prices of individual goods or capital flows.
Floating exchange rates were introduced to allow currency value adjustments
based on the relative price changes of each country. The primary advantage is that
the exchange rate is a market rate and is more flexible than wages and prices in general.
But countries may establish programs to avoid or retard the automatic exchange rate
804 EXCHANGE RATE SYSTEMS AND PARITY CONDITIONS
adjustment processes. These include tariffs, duties, quotas, and domestic price controls.
Typically the impact is to force the adjustment into another direction, such as changes
in real output.
In balance of payments accounting, inflows are recorded as pluses and outflows
as minuses. A trade imbalance may trigger a number of offsetting transactions. The
subtotals of the balance of payments accounts provide much information about the
short- and long-run developments for the economy of any country. Analysis of balance
of payments trends can be useful for formulating corporate financial policy in manag-
ing exposure to exchange rate fluctuations and in the use of the international financial
markets.
Foreign exchange markets enable buyers and sellers in different countries to deal
in their own currencies. The foreign exchange rate is the price of one currency in terms
of another and depends on the supply and demand relationships between the two cur-
rencies. There are strong forces linking spot, forward, future spot, and interest rates
between countries. Arbitrage opportunities ensure these relationships against break-
downs. The consistency of cross rates means that given any three countries any one
exchange rate will be determined by the other two exchange rates. If the relationship
breaks down, an arbitrage opportunity will exist.
The purchasing power parity doctrine states that the change in exchange rates
between countries depends on the relative changes in their prices for the same period.
Thus the relative values of the currencies depend on what they will buy.
The Fisher relation holds that the nominal rate of interest is determined by the
real rate of interest and the rate of inflation. The rate of inflation is also equal to the
ratio of the expected future spot price to the current price. The current forward price
is an estimate of what the future spot price will be.
The IRPT incorporates the Fisher relation by recognizing that between any two
currencies the forward exchange rate should be the same as the spot exchange rate
except for the anticipated effect of different rates of inflation. Since anticipated in-
flation rates are incorporated in the nominal interest rates of each country, there is
a potential arbitrage possibility between current and forward exchange rates based
on the relative interest rates of the countries. The process is known as covered interest
arbitrage.
The real return from holding an investment denominated in a foreign currency
(over and above the domestic real return) is a simple relationship of the future spot
to the forward exchange rate. This relationship directs the international flow of invest-
able funds. The forward parity theorem holds that the forward exchange rates should
be unbiased predictors of future spot exchange rates.
The five fundamental equilibrium relationships imply the law of one price, which
holds that under the assumptions set forth the exchange-adjusted prices of identi-
cal tradable goods and financial assets must be within transactions costs of equality
throughout the world. Different rates of inflation and changing nominal interest rates
should not affect the rates at which today's goods are exchanged for future goods or
the rate at which domestic goods are exchanged for foreign goods. In the following
chapter, we summarize empirical tests of the above propositions and discuss their
implications for corporate international financial policies.
PROBLEM SET 805
PROBLEM SET
21.1 You are given the prices of products in two countries as shown below:
Product
X
Country A $3 $1
Country B M12 M6
At an exchange rate of 5 marks per dollar, describe the pattern of exports and imports between
countries A and B.
21.2 Country A and country B are each on a full gold standard with fixed exchange rates.
Country A runs an export surplus, whereas country B runs an export balance deficit. Describe
the adjustment process that will restore balance to the flow of trade between the two countries.
21.3 Country A and country B are on the gold standard. The currency of country A contains
.1 ounce of gold, whereas the currency of country B contains .025 ounce of gold. What will be
the par exchange rate between the two countries?
21.4 Consider two countries C and D operating in a world with complete flexible exchange rates.
Country C runs a substantial export surplus to country D, which experiences a substantial
trade deficit. Assuming no initial offsetting capital flows, explain the adjustment process to
bring the trade between the two countries into balance.
21.5 Keep in mind Table Q21.5 listing the effects of individual transactions on the balance of
payments.
1. Exports 1. Imports
2. Increase liabilities to foreigners 2. Decrease liabilities to foreigners
3. Decrease claims on foreigners 3. Increase claims on foreigners
4. Decrease investments; sell assets 4. Increase investments; buy assets
Indicate the plus entry and the minus entry for the following transactions. For example, the
country exports goods in the amount of $1000 paid for by the importer by a check on a foreign
bank. The entry would be:
P1 $1000 M3 $1000
a) Country A exports $10,000 of goods to country I paid for by the importer by a check on
his account with a bank in country A.
b) Country A imports $5000 worth of merchandise paid for by a check on a bank in country
A.
c) Direct investment income of $2000 was received by a firm in country A from a foreign
subsidiary, which paid by drawing a check on a bank in its own country F.
806 EXCHANGE RATE SYSTEMS AND PARITY CONDITIONS
REFERENCES
Adler, M., and B. Dumas, "International Portfolio Choice and Corporation Finance: A Syn-
thesis," Journal of Finance, June 1983, 925-984.
Adler, M., and B. Lehman, "Deviations from Purchasing Power Parity in the Long Run,"
Journal of Finance, December 1983, 1471-1487.
Aizenman, J., "Testing Deviations from Purchasing Power Parity," Journal of International
Money and Finance, March 1986, 25-36.
Aliber, R. Z., The International Money Game, 3rd ed. Basic Books, New York, 1979.
Allsopp, C., "The Rise and Fall of the Dollar: A Comment," Economic Journal, March 1987,
44-48.
Barthold, T. A., and W. R. Dougan, "The Fisher Hypothesis under Different Monetary Re-
gimes," Review of Economics and Statistics, November 1986, 674-679.
Batten, D. S., and D. L. Thornton, "Discount Rate Changes and the Foreign Exchange Mar-
ket," Journal of International Money and Finance, December 1984, 279-292.
Calvo, G. A., "Temporary Stabilization: Predetermined Exchange Rates," Journal of Political
Economy, December 1986, 1319-1337.
Caves, R. E., and R. W. Jones, World Trade and Payments: An Introduction, 2nd ed. Little,
Brown, Boston, 1977.
Chacholiades, M., International Trade Theory and Policy, 2nd ed. McGraw-Hill, New York,
1978.
REFERENCES 807
Chan, K. S., "The International Negotiation Game: Some Evidence from the Tokyo Round,"
Review of Economics and Statistics, August 1985, 456-464.
Cooper, R. N.; P. B. Kenen; J. Braga de Macedo; and J. van Ypersele, eds., The International
Monetary System under Flexible Exchange Rates. Ballinger, Cambridge, Mass., 1982.
Corbo, V., "International Prices, Wages and Inflation in an Open Economy: A Chilean Model,"
Review of Economics and Statistics, November 1985, 564-573.
Cornell, B., and M. R. Reinganum, "Forward and Future Prices: Evidence from the Foreign
Exchange Markets," Journal of Finance, December 1981, 1035-1045.
Dilullo, A. J., "U.S. International Transactions, Third Quarter 1986," Survey of Current Busi-
ness, December 1986, 23-27.
Dornbusch, R., Open Economy Macro-Economics. Basic Books, New York, 1980.
, "Exchange Rate Economics: 1986," Economic Journal, March 1987, 1-18.
Eiteman, D. K., and A. I. Stonehill, Multinational Business Finance, 4th ed. Addison-Wesley,
Reading, Mass., 1986.
Errunza, V., and E. Losq, "International Asset Pricing under Mild Segmentation: Theory and
Test," Journal of Finance, March 1985, 105-124.
Farber, A.; R. Roll; and B. Solnik, "An Empirical Study of Risk over Fixed and Flexible
Exchange," Journal of Monetary Economics, Supplementary Series, 1977, 235-265.
Feige, E. L., and K. J. Singleton, "Multinational Inflation under Fixed Exchange Rates: Some
Empirical Evidence from Latent Variable Models," Review of Economics and Statistics,
February 1981, 11-19.
Frenkel, J. A., "The Implications of Mean-Variance Optimization for Four Questions in Inter-
national Macroeconomics," Journal of International Money and Finance, March 1986,
853 —876.
, and H. G. Johnson, eds., The Economics of Exchange Rates. Addison-Wesley, Reading,
Mass., 1978.
Goodhart, C., "Exchange Rate Economics 1986: A Comment," Economic Journal, March 1987,
19-22.
Grabbe, J. 0., International Financial Markets. Elsevier, New York, 1986.
Grauer, F. L. A.; R. H. Litzenberger; and R. E. Stehle, "Sharing Rules and Equilibrium in an
International Capital Market under Uncertainty," Journal of Financial Economics, June
1976, 233-256.
Grubel, H. G., International Economics. Richard D. Irwin, Homewood, Ill., 1977.
Haberler, G., "The International Monetary System," Economist, July 1986, 1-8.
Hakkio, C. S., "Interest Rates and Exchange Rates What Is the Relationship?" Federal
Reserve Bank of Kansas City, Economic Review, November 1986, 33-43.
Hartman, D. G., "The International Financial Market and U.S. Interest Rates," Journal of
International Money and Finance, April 1984, 91-104.
Helpman, E., "A Simple Theory of International Trade with Multinational Corporations,"
Journal of Political Economy, June 1984, 451-471.
Huang, R. D., "The Monetary Approach to Exchange Rate in an Efficient Foreign Exchange
Market: Tests Based on Volatility," Journal of Finance, March 1981, 31-41.
, "Some Alternative Tests of Forward Exchange Rates as Predictors of Future Spot
Rates," Journal of International Money and Finance, August 1984, 153-168.
808 EXCHANGE RATE SYSTEMS AND PARITY CONDITIONS
Johannes, J. M., "Testing the Exogeneity Specification Underlying the Monetary Approach to
the Balance of Payments," Review of Economics and Statistics, February 1981, 29-34.
Koromzay, V.; J. Llewellyn; and S. Potter, "The Rise and Fall of the Dollar: Some Explana-
tions, Consequences and Lessons," Economic Journal, March 1987, 23-43.
Kreinin, M. E., International Economics: A Policy Approach, 3rd ed. Harcourt Brace Jovanovich,
New York, 1979.
Lessard, D. R., International Financial Management: Theory and Application, 2nd ed. John
Wiley & Sons, New York, 1985.
Levi, M., International Finance. McGraw-Hill, New York, 1983.
Lindert, P. H., and C. P. Kindleberger, International Economics, 7th ed. Richard D. Irwin,
Homewood, Ill., 1982.
Mark, N. C., "Some Evidence on the International Inequality of Real Interest Rates," Journal
of International Money and Finance, June 1985, 189-208.
Meltzer, H., "Monetary and Exchange Rate Regimes: A Comparison of Japan and the United
States," Cato Journal, Fall 1986, 667-683.
Overturf, S. F., "Interest Rate Expectations and Interest Parity," Journal of International Money
and Finance, March 1986, 91-98.
Pippenger, J., "Arbitrage and Efficient Markets Interpretations of Purchasing Power Parity:
Theory and Evidence," Federal Reserve Bank of San Francisco, Economic Review, Winter
1986, 31-47.
Rodriguez, R. M., and E. E. Carter, International Financial Management, 2nd ed. Prentice-Hall,
Englewood Cliffs, N.J., 1979.
Rolfe, S. E., and J. L. Burtle, The Great Wheel: The World Monetary System. Quadrangle/New
York Times Book Company, New York, 1973.
Roll, R., "Violations of the Law of One Price and Their Implications for Differentially De-
nominated Assets," in Marshall Sarnat and George Szego, eds., International Finance and
Trade, vol. 1, Ballinger, Cambridge, Mass., 1979.
Shapiro, A. C., Multinational Financial Management, 2nd ed. Allyn and Bacon, Boston, 1986.
Solnik, B. H., European Capital Markets. Lexington Books, Lexington, Mass., 1973.
, "International Arbitrage Pricing Theory," Journal of Finance, May 1983, 449-457.
Solomon, R., The International Monetary System, 1945-1976, An Insider's View. Harper &
Row, New York, 1977.
Takayama, A., International Trade. Holt, Rinehart and Winston, New York, 1972.
Triffin, R., "Correcting the World Monetary Scandal." Challenge, January—February 1986,
4-14.
Ungerer, H.; 0. Evans; T. Mayer; and P. Young, "The European Monetary System: Recent
Developments," International Monetary Fund Occasional Papers 48, International Mone-
tary Fund, Washington, D.C., 1986.
Yeager, L. B., International Monetary Mechanism. Holt, Rinehart and Winston, New York,
1968.
, International Monetary Relations: Theory, History and Policy, 2nd ed. Harper & Row,
New York, 1976.
ûû
The exchanges between countries are at par only whilst they have
precisely that quantity of currency, which in the actual situation of
things, they should have to carry on the circulation of their
commodities.
David Ricardo, Principles of Political Economy and Taxation,
Gonner Edition, G. Bell and Sons, Ltd., London, 1927, 213-214.
Inte]national Financial
Management- Tests and
Implications
809
810 INTERNATIONAL FINANCIAL MANAGEMENT: TESTS AND IMPLICATIONS
A. INTERNATIONAL DIVERSIFICATION
From the purely financial point of view, investors, whether they be firms or individ-
uals, ought to consider the possibility of expanding their investments beyond the
geographical limits of their own countries, if only because of the greater number
and diversity of investment possibilities available. If the universe of assets available
for investment is larger than just the assets in one country, even a country as large
as the United States, our discussion in earlier chapters should suggest that investors
may be able to reduce the risk of their investment portfolios by diversifying across
countries. Solnik [1974b] has some empirical estimates for the risk of an interna-
tionally diversified portfolio compared with a diversified portfolio that is purely
domestic. Using weekly data on stocks in eight major European countries and the
United States, Solnik found that an internationally diversified portfolio would be one
tenth as risky as a typical security and one half as risky as a well-diversified portfolio
of U.S. stocks alone. He used the variance of returns as a measure of risk. He also
found that interindustry diversification is inferior to intercountry diversification.
In a related study, Jacquillat and Solnik [1978] compared the performance of
multinational corporations with that of an internationally diversified portfolio. By
the same risk measure as above, a portfolio of U.S. multinational firms has about
90% of the standard deviation of a purely domestic U.S. portfolio. Internationally
diversified portfolios have only about 30% to 50% of the risk of the U.S. domestic
portfolio. This is suggestive evidence that the international dimension has not yet
been fully exploited by multinational corporations (MNCs). A portfolio of MNC
docks is a poor substitute to the investor for a truly internationally diversified
portfolio.
Two empirical studies relate to the above. Maldonado and Saunders [1983]
concluded that their empirical evidence "supports capital market efficiency and the
law of one price for internationally traded stocks" [23]. This casts doubt on the
benefits of international diversification. But Philippatos, Christofi, and Christofi
[1983] argue that the above study and an earlier one (Maldonado and Saunders
[1981]) were influenced by methodologies and sample employed. They agree with
other studies that find ex ante gains from international diversification.
Asset pricing models, similar in form and spirit to the CAPM, have been derived
for international financial assets. In a manner analogous to the standard CAPM
derivations in continuous time, Solnik develops an international asset pricing model
(IAPM) and tests it [1974a, 1974d]. A market portfolio, with properties similar to
that in the CAPM, is specified by:
1. Market value—weighted stocks,
2. A pure exchange risk portfolio of risk-free assets of the various countries in the
world. Relative weights in this portfolio depend on net foreign investment posi-
tions in each country and the relative risk aversions of citizens of each country.
ASSET PRICING MODELS 811
that prices of commodities and the exchange rate itself follow a stochastic Ito process.
All assets are assumed to be traded, but all commodities are not. The law of one
price (naive PPP) holds only for traded goods. A pricing relationship is derived,
which states that
,u — V ap = (T w ) -1{Vac Va pc,C}, where T w = T F + T D. (26)
Stulz [1981b, 396] comments on his Eq. (26) as follows:
He]e p is the vecto] of expected excess ]etu]ns fo] domestic investo]s. The excess ]etu]n
of an asset fo] a domestic investo] is defined as the ]etu]n in the domestic count]y of a
po]tfolio with a long holding of that asset financed by selling the nominal safe bond of
the home-count]y of the asset. T w is a measu]e of wo]ld ]isk-tole]ance. V ac is the vecto]
of cova]iances between home-count]y ]etu]ns of ]isky assets and changes in wo]ld
consumption in domestic cu]]ency. V a pa, and Vapc, a]e the vecto]s of cova]iances of
home-count]y ]etu]ns with d in Pm and d in Pa. An inc]ease in Pm means a dec]ease in
the ]eal value of a ma]ginal dolla] of consumption expenditu]es, whe]eas an inc]ease in
Pa means a fall in the ]eal value of total consumption expenditu]es. Eq. (26) is the
fundamental p]icing equation of this pape].
From his Eq. (26), a proposition is shown to follow: The expected excess real
return of a risky asset is proportional to the covariance between the home currency
rate of return of that asset and changes in world real consumption rate. Note that
the return on an asset does not depend on unexpected changes in the exchange rate,
so exchange rate risk is not rewarded.
As a further result, Stulz shows that the forward exchange rate is inversely pro-
portional to the covariance between the changes in the exchange rate and changes
in the real-world consumption rate. This is the first full model to permit the forward
rate to change through time, always remaining in equilibrium. Stulz finally notes that
this more general model of asset pricing in international financial markets may ex-
plain why some previous models seemed to find the international financial markets
to be segmented: If changes in the world consumption rate are not perfectly corre-
lated with the market indices used by previous researchers, spurious results may
Blow from the misspecification of the tested models. The Breeden and Stulz models
have not been systematically tested to date.
Fluctuating exchanges are often viewed as an additional source of risk facing the in-
vestor or trader importing and exporting goods. The prices of commodities, the yields
on assets, are usually fixed in nominal terms, i.e., in fixed amounts of foreign currency
at the time of transacting. An investor who buys a $100 pure discount bond, e.g.,
knows only that he or she will receive $100 at the maturity of the bond. But if the
investor happens to be a German who needs marks for consumption, he or she does
not know how many marks the dollars received will buy. That depends on the ex-
change rate between the dollar and the mark. With the highly fluctuating exchange
rates of recent years, this seems to be a substantial source of risk.
Because the value of the investor's return at the end of the period is uncertain
in terms of the investor's home currency, his or her consumption is uncertain. This
is readily interpretable as the risk that the variations in exchange rates over a period
will not exactly offset inflation differentials, which is another way of saying that PPP
will not hold. Thus deviations from PPP acquire practical significance in terms of
exchange risk.
The key to understanding departures from PPP, or equivalently, the phenomenon
of exchange risk, is to recognize that there are many consumption goods. Further,
people in different countries have different consumption baskets—they consume the
goods in different proportions. For example, the Japanese consume more sake than
wine and French consume more wine than sake. Such differences may be caused by
a variety of factors: social, cultural, historical—everything subsumed under the ex-
pression "different tastes."
The relative prices of goods can fluctuate over time because of shifts in supply
or demand conditions. Demand conditions may change with changes in tastes owing
to shifts in the demographic composition of the population, for example. Supply con-
ditions may change owing to bad weather or because of technological innovation.
814 INTERNATIONAL FINANCIAL MANAGEMENT: TESTS AND IMPLICATIONS
1. Inflation Risk
Suppose that we are investors in the United States who have a security in En-
gland whose return is fixed in pounds. Assuming that there is no inflation in the
United States, but that the inflation rate in England is uncertain, the dollar value of
our investment at the end of the period is uncertain and hence risky. Our position has
"exchange risk." It is clear that this exchange risk is simply inflation risk. With no
relative price fluctuations, arbitrage of physical goods will ensure that PPP holds,
because now, in a sense, there is only one consumption good "composite" and PPP
is the same as the law of one price, which always holds in perfect commodity markets.
The change in exchange rate reflects exactly the change in inflation rates and nothing
else.
However, the risk of our position is purely nominal. It can be eliminated by
indexing all contracts, tying the quantity of pounds from the investment to the price
index in England, for instance, or contracting in "real" terms. To the extent this is
not feasible or not done, it reflects information costs or transations costs of doing so.
In our present model of the world, it is perfectly feasible to eliminate nominal risk
completely. In any case the point is that it is a mistake to attribute the "risk" to
the exchange rate. It is caused by uncertainty about the foreign price level, not so
much by the phenomenon of fluctuating exchange rates. If all contracts were written
in real terms, all profits, all returns on investments of all kinds, would be independent
of the geographical location of the investor or the investment, and the exchange rate
would have no real role to play beyond the arithmetic of conversions. Grauer, Litzen-
berger, and Stehle [1976] have shown that, in the case where all individuals in the
world have the same tastes, if a real riskless asset exists, all individuals face the same
investment and consumption opportunity set—and so which country the investor
happens to be in is irrelevant. The exchange risk is then merely some sort of white
noise that has no effect on anything real, though it gives apparent fluctuations in
prices in nominal terms. Under the assumption of identical homothetic tastes (equiv-
alent to one consumption good when relative prices are constant), for the exchange
risk to be irrelevant is for goods arbitrage to be instantaneous and costless, so that
the real price of a good will be the same everywhere in the world.
Solnik [1974c] was the first to explore formally the implications to financial
markets of a world where individuals in different countries consume different baskets
of goods. We shall use Solnik's [1978] example of Japanese who consume more sake
than wine and French who consume more wine than sake. (We shall follow his
notation for ease of reference to the original.) Wine and sake are the only goods in
the world. It is not necessary for this story that the French produce wine and only
wine, or that the Japanese produce sake and only sake.
Suppose, to begin with, the price of sake equals the price of wine i.e., the rela-
tive price is 1, in both countries. The relative price must be the same in both countries,
else there will be goods arbitrage, since the transportation of goods is perfect, in-
stantaneous, and costless by the assumptions of the model. Suppose also that the
exchange rate, which we shall define to be the price of yen in francs, is also 1. Next
postulate an exogenous shock, e.g., something damages the wine crop. Then the
relative prices of the two goods will change at once to reflect the reduced supply of
wine. Wine will become more expensive. Again, because of the possibility of goods
arbitrage, the relative price between wine and sake must be the same in both coun-
tries. The question we seek to answer is: What will be the new exchange rate? As-
sume the following relationships:
Wine Sake
French consumption 9 units 1 unit
Japanese consumption 1 unit 9 units
Price of wine = price of sake: The exchange rate is 1:1.
Note that the consumptions have adjusted to the changed relative price. Now,
we need to separate the effect of a relative price change from that of an inflationary
change. We next assume that there is no general rise in the price level in either
country. The governments in both countries do whatever they have to do to ensure
that the prices of the consumption baskets remain the same in terms of their respec-
tive local currencies: adjusting the money supply, etc.
Let the price of sake and wine in France be S,7. and and in Japan be S?
C
WI"
and WY in francs and yen, respectively, before the shock. Let W,lf , S'}- ; 1/17 , be the
corresponding after-the-shock prices in nominal terms (i.e., in terms of the home cur-
rencies—francs in France and yen in Japan).
816 INTERNATIONAL FINANCIAL MANAGEMENT: TESTS AND IMPLICATIONS
We know that
= WY,
SY = WY; The relative price = 1.
= 1.5 Si- ,
= 1.5 The new relative price = 1.5.
Now we note that by the law of one price the price of sake must be the same
in both countries at a given time, after converting by the new exchange rate. Then
after the shock, =V, SY francs must buy a unit of sake in both France and Japan,
Eq. (22.1). Recall that SY = SY because the initial exchange rate was 1. By (22.2) we
also have
Syl = =
falls relative to the price of francs, which is what happened. Stockman [1980] makes
the same general point.
In the Solnik example, the exchange rate fluctuation is a "real" risk because we
cannot hedge it away merely by contracting in real terms. Those in France who have
invested in Japan get only .741 of what they expected to get, in francs, and can only
buy, in their own consumption baskets, .741 of what they had expected to buy if the
shock had not occurred and the exchange rate had not changed.
In what sense is this an "exchange risk"? The risk was caused by a change in
relative prices of goods. The risk did not come from the exchange rate directly. The
exchange rate only reflected something more fundamental that changed. In fact,
business firms always face the risk that there may be changes in the relative prices
of their product relative to its substitutes, of the product relative to the inputs, etc.
(So does any individual: a vegetarian, e.g., always faces the risk that the price of his
or her food will change relative to the price of meat.) The point is that it is important
to recognize that "exchange risk" is essentially the same as "business" risk. The fact
that the firm may be operating in two different countries causes it only to be mani-
fested, partly, through the exchange rate. To ascribe the investor's risk to the phe-
nomenon of exchange rate fluctuations, or to the fact of flexible exchange rates, is to
mistake the symptom for the malady.
To return this discussion to PPP, recall that in the extended example of the
wine and sake there was no inflation in either country. PPP would require that the
exchange rate remain the same. But it did not. Hence PPP was violated. Real ex-
change risk is to be found wherever PPP is violated.
be equal to the expected change in exchange rates plus a term depending on ex-
change rate covariances. Their findings indicate that exchange risk is substantial and
significant.
D. MARKET EFFICIENCY
Having considered some fundamental ways in which the international financial mar-
kets may differ from domestic financial markets, we may now proceed to ask the sorts
of questions we have been asking in our earlier studies of the domestic markets: To
what degree are exchange markets efficient? What is the empirical magnitude and
significance of exchange risk? Several possible approaches to the question of efficiency
of the exchange markets are first examined.
industrialized nations, they set up some simple filter rule—based trading strategies,
taking transactions costs into account.
In the moving average filter, when the exchange rate moves X% above the moving
average for a prespecified interval, the rule calls for the investor to move into the
foreign currency; when it falls X% below the moving average, the investor moves
back into dollars. Moving averages of 10, 25, and 50 days were tried. None of these
filter rules produced a profit of more than 4% per annum for the British pound,
Canadian dollar, or Japanese yen. However, the situation was quite different with the
German mark, the Dutch guilder, and Swiss franc. The 25-day moving average rule
consistently produced profits of more than 15% per annum for all three currencies,
with filters ranging from 0.2% to 2.0%. The simple filter did not do quite as well,
because of the larger transactions costs accruing to the larger number of transactions
it triggered. Still, it did produce profits of more than 10% for some filters.
The three currencies with the high trading-rule profits were also the three with
the highest variance. Hence the "profit" observed may well be merely compensation
for the greater risk of trading in these currencies. Since variance alone is not systematic
risk, Cornell and Dietrich attempted to separate the systematic from the unsystematic
portion of the total variation. Exchange rate variations were correlated with the S&P
500 index as the "market." The result was to eliminate almost entirely the risk of
these currencies, casting doubt on the proposition that the "excess profit" was a "risk
premium."
As with all filter tests, the main problem in interpretation of the results seems
to be that we do not know how to adjust for the different risks of the alternative
strategies. Also, it does not appear that there is any way to determine, ex ante, the
appropriate size of filter to test. They depend on historical relationships that are
likely to be unstable over time.
4. Forward Speculation
A fourth approach to testing the efficiency of markets is to examine the returns
to forward speculation. The basic assumption underlying the simpler versions of this
class of efficiency tests is that the forward rate should be an unbiased forecaster of
the future spot rate. But there may be good theoretical reasons for the forward rate
to be systematically different from the expected (corresponding) future spot rate. For
example, Stockman [1978] developed a simple model with only money supply un-
certainty that shows that the forward rate may have a risk premium associated with
it, which would lead it to be a biased estimator of the future spot rate. Another model
that makes the same point is that of Grauer, Litzenberger, and Stehle [1976]. A
forward bias can also come from the "real" exchange risk.
To turn to the empirical evidence itself, it appears that, on the whole, the forward
rate is very close to the expected future spot rate. Stockman [1978] found that the
risk premium is significant only for a couple of currencies (the British pound and
the Swiss franc). He interpreted it to mean that the risk premium was not important.
The risk premium, when significant, was not stable across subperiods. The question
of exchange market efficiency was also studied by Cornell [1977]. Using monthly
820 INTERNATIONAL FINANCIAL MANAGEMENT: TESTS AND IMPLICATIONS
on a forward contract depends on the same covariation, since forward contracts are
risky nominal assets. But this position depends on the underlying theories of the
relation between spot and future prices, as discussed in Chapter 9.
Huang [1984] reviews other studies that he argues use "more statistically power-
ful techniques" but are unable to confirm the hypothesis of forward rates as unbiased
predictors of future spot rates. His own study of nine countries in relation to the
United States yields results that "suggest support for the random walk model" [1984,
157]. He finds little evidence of a liquidity premium. When a second alternative
hypothesis of excessive or inadequate volatility for market efficiency is considered,
the evidence cannot reject the alternative hypothesis that forward rates are unbiased
predictors of future spot rates.
Using an experimental market approach, Forsythe, Palfrey, and Plott [1984] find
that the effect of a futures market is to increase the speed at which a rational expec-
tation equilibrium is reached. The combination of the market process with each in-
dividual's learning process drives the market to a rational expectations equilibrium
with a relatively small number of replications.
A careful analysis of the relation between forward rates and future spot rates,
containing many insights, is due to Fama [1984]. We summarize his presentation.
The forward exchange rate, e f , observed at time zero for an exchange at time 12
is the market-determined certainty equivalent of the future spot exchange rate, e 1.
We can split this certainty equivalent into an expected future spot rate and a premium:
In ef = E(ln e 1 ) + P o . ( 1)
From (1) the difference between the forward rate and the current spot rate is
In e f ÿ ln e, = P, + E(ln e 1 ÿ ln e 0 ). (2)
Fama then considers the regressions of (ln of ÿ ln e 1 ) and (ln e 1 ÿ ln e 0 ), both
observed at t = 1, on (ln e f ÿ ln e,), observed at t = 0:
(ln e f ÿ ln e 1 ) =a 1 + /31(ln e.ÿ ln e 0 ) +E 1 , t = 1, (3)
(ln e, ÿ ln e 0 ) = a 2 + fi,(1n e f ÿ ln e„)+ e 2 , t = 1. (4)
Since the dependent variables sum to the same independent variable in both (3)
and (4), the sum of the intercepts in (3) and (4) must be 0, the sum of the slopes must
be 1.0, and the disturbances, period by period, must sum to 0.0. Given an efficient
or rational exchange market, the deviation of /32 from 1.0 is a direct measure of the
variation of the premium in the forward rate. The complementary of the regression
coefficients in (3) and (4) help in the interpretation of the empirical results.
Since a major conclusion of his empirical findings is that variation in forward
rates is mostly variation in premia, Fama examines the economics of premia. He
shows that with three central conditions discussed in the previous chapter (PPP, IRP,
and IFR), the premium in the forward rate expression [Eq. (1)] is the difference
between the expected real returns on the nominal bonds of the two countries.
2
Since Fama subsequently expresses all exchange rates in dollars per unit of foreign currency, we shall
use our relatively simple notation but his equation numbering since he employs them in his explanations.
822 INTERNATIONAL FINANCIAL MANAGEMENT: TESTS AND IMPLICATIONS
The first issue is whether the parity conditions described in Chapter 21 preclude
the necessity of dealing with foreign exchange risk. Dufey and Srinivasulu [1983]
address this question of whether the parity conditions make it unnecessary to attempt
to manage foreign exchange risk. They point to a number of market imperfections
that must be taken into account incomplete securities markets, positive transac-
tions and information costs, the deadweight costs of financial distress, and agency
costs. Their existence makes it desirable for corporate management to seek to cope
with exchange risk.
future, these exposures could be hedged. They recognize that as exposures vary over
time, it would be necessary to seek to derive multiperiod hedging rules.
Johnson and Walther [1984] measure the effectiveness of portfolio hedges in
currency forward markets. To hedge the exposed cash position using portfolio theory,
the firm takes an offsetting position in the forward market. The portion of the spot
market holding to be held in the forward market equals the variance-minimizing
portfolio hedge ratio. This is the subjective covariance between the forward market
price changes and the spot market price changes to the variance of the forward mar-
ket price changes. The portfolio approach is compared with a naive hedge in which
the exposure is offset completely in the forward market and the gain or loss is deter-
mined by the difference between the forward market rate and the future spot rate of
the foreign currency. They conclude that the naive hedge is superior but that neither
achieves complete elimination of foreign currency price risk.
where
Cp = cost of net monetary position (NMP) due to exchange rate changes,
MA = monetary assets,
ML = monetary liabilities,
X, = exchange rate at the beginning = 1/E0 ,
X} = exchange rate a period later = 1/E1,
tu.s. = tax rate in the United States.
The effects of a decline in foreign currency value are:
a) Net monetary debtor gains.
b) Net monetary creditor loses.
MANAGERIAL ASPECTS OF FOREIGN EXCHANGE RISKS 825
We calculate the net monetary position to which the effect of the exchange rate shift
is applied:
1) NMP = MA — ML = 200,000 — 100,000 = 100,000;
C p = NMP(E 0 — E 1)(1 — tu.s.) = 100,000(25 — .20)(.5)
= $5000(.5) = $2500.
Our calculations show a decrease in the dollar value of our asset position i.e., a
loss of $2500.
We now let ML = FC 300,000. Then
2) NMP = MA — ML = 200,000 — 300,000 = —100,000;
Cp = NMP(E 0 — E 1)(1 — t u.s.) = — 100,000(25 — .20)(5)
= —$2500.
The cost is a negative $100,000, representing a gain in the value of the net monetary
position with revaluation upward in the FC currency.
2) Let MA = FC 6 million. Then
NMP = 6,000,000 — 8,000,000 = — 2,000,000,
Cp = NMP(E 0 — E 1 ) = — 2,000,000(.20 — .25) = + $100,000.
Both terms are negative, so their product is positive, indicating that revaluation up-
ward results in a positive cost (or a loss) to a firm in a negative net monetary position
(monetary liabilities exceed monetary assets). The FC values of its net obligations
have increased.
The impact of an exposed position is similar if the exposure results from an ex-
cess of receipts over payments due to be paid in the foreign currency. A firm in this
826 INTERNATIONAL FINANCIAL MANAGEMENT: TESTS AND IMPLICATIONS
A firm in such a position will lose from a devaluation (or decrease in value or
depreciation in the value) of foreign currencies.
2. The firm is exposed to an increase in the value of foreign currencies. Then
a) Expected future payments exceed expected future receipts.
b) The net monetary position is negative.
3
To simplify the discussion, we omit consideration of taxes here.
828 INTERNATIONAL FINANCIAL MANAGEMENT: TESTS AND IMPLICATIONS
Let us suppose that interest rates in the foreign market are 32%, whereas interest rates
in the domestic market are 10%. Then the expected ratio between the current spot
rate and the current forward rate is 1.2. If the current spot rate is .3, the equilibrium
forward rate is .25 (1.32/1.10 = .30/E f ).
If parity obtains, the cost of hedging in the forward market or of borrowing in
foreign currency and investing in the United States is the same. For example, if the
amount of foreign currency involved is 100,000 FC units, then the cost of hedging in
the foreign market, C f , is
C f = (E 0 — E f )F 0
= (.30 — .25)100,000
= $5,000.
In the money and capital markets the situation is different from the covered in-
terest arbitrage operation because an exposure of FC 100,000 already exists. If FC
100,000 is borrowed in the foreign country to neutralize the foreign exchange risk
exposure, the principal is covered by the future net receipts or net monetary assets
position of the firm. Hence the cost of using the money and capital markets by borrow-
ing foreign currency and investing in the United States is
Cbf = Cost of borrowing in foreign country
= E i F o R f — E,F o Rus. = (E1Rf — E o R u.s.)F 0
= [.32E 1 — .30(.10)]100,000.
Now the cost of borrowing foreign currency depends on what the future spot
rate will be, because interest on the foreign borrowing must be paid at the future spot
rate. The firm can remove this uncertainty by buying the foreign currency in the
amount of the required interest to be paid at the current forward rate. The cost of
borrowing foreign now becomes
Cbf = FC 32,000Ef — $3,000.
Next, we evaluate at the current forward rate:
Cbf = 32,000(.25) — $3,000 = $8,000 — $3,000
= $5,000.
The cost of borrowing foreign will now be $5,000, which is exactly the same as
in the hedged case in which the foreign currency was sold in the forward exchange
market. It should be emphasized that borrowing in the foreign market produces an
interest obligation that will become due in the future. Hence to protect against this
exposure, the action is to buy the foreign currency in the amount of the interest ob-
ligation in the forward market rather than sell it as in the hedging operation in the
forward market. Because this strategy requires having foreign currency on hand in the
future to pay the foreign interest, the effects on the position of the firm will be reversed
if the firm does not cover its future interest payments to be made in foreign currency
but permits matters to depend upon the future spot foreign exchange rate. For ex-
INTEREST RATE AND CURRENCY SWAPS 829
ample, if the future spot foreign exchange rate is higher than the current forward rate,
the cost of borrowing foreign will be higher:
Cbf = ( E 1 R1 — E o Ru .0100,000
[.27(.32) — .30(.10)]100,000
= (.0864 — .03)100,000 = .0564(100,000)
= $5,640.
With the future expected spot rate being $.27 rather than $.25, costs of borrowing
foreign are in excess of $5,000. However, if the future spot rate falls below the current
forward rate, the cost of borrowing foreign will be less than $5,000. Hence if the firm
does not immediately cover its future interest payments to be made in foreign currency,
it continues to be exposed to foreign exchange risk.
We next consider the position of the firm that is in a negative net monetary posi-
tion or that will be required, at some time in the future, to make net payments in excess
of future receipts. The situation is the reverse of the firm in a positive monetary posi-
tion. Instead of facing the risk of devaluation of FC units, the firm faces the risk of
appreciation in the value of FC units. Hence its protective action in the forward market
is to buy the FC units. If it does not obtain this protection, it will face an uncertain
cost if the future spot rate is higher than the current forward rate. If it uses the money
and capital markets, it will borrow in the United States and invest in the FC units in
which it will have to make future payments.
This analysis demonstrates that if a firm is in an exposed foreign exchange posi-
tion, it will incur some costs to obtain protection against that exposed Even
rearranging the firm's pattern of payments and receipts or monetary asL-is and liabil-
ities will represent a departure from normal operations and therefore involve some
costs. If the firm uses the forward market or borrows, it incurs some costs, but it will
know the exact amount of these costs.
Another important set of institutions has been developed to assist firms in managing
risks arising from both interest rate and exchange rate volatility. These are interest
rate and currency swaps that enable firms to lower financing costs and to provide
4
hedges in foreign currencies. We begin with the discussion of interest rate swaps.
4 This presentation uses illustrations from Hutchison [1985]. See also Whittaker [1987].
830 INTERNATIONAL FINANCIAL MANAGEMENT: TESTS AND IMPLICATIONS
In December 1981 the Financial Accounting Standards Board (FASB) issued FASB
No. 52, Foreign Currency Translation, superseding FASB No. 8, which had been
issued in 1976. In general, translation gains or losses are carried directly to the equity
FOREIGN CURRENCY TRANSLATION 831
account on the balance sheet and do not affect net income. Individual transactions
gains or losses net of hedging costs and net of translation gains or losses do enter into
the calculations that determine net income. Also, the method of translation is changed
from the temporal method to the use of the current exchange rate for all balance sheet
items and the use of the average exchange rate for the period for the income state-
ment. However, the temporal method will continue to be applied to operations in
highly inflationary economies, defined as those in which the price level doubles within
a three-year period of time.
Since the temporal method of currency translation continues to be applied in
the circumstances indicated, it will be useful to explain and illustrate both methods.
The temporal method was essentially the monetary-nonmonetary method with one
change. Under the monetary-nonmonetary method, the logic of defining exposure
by the net monetary position was followed. Monetary assets and liabilities were trans-
lated at current exchange rates, whereas nonmonetary assets and liabilities were trans-
lated at the applicable historical exchange rates. Under the monetary-nonmonetary
method of translation, inventories were treated as real assets to which the applicable
historical exchange rate was applied. The temporal method recognizes that alternative
inventory valuation methods may be used. For example, FIFO charges the income
statement for the historical costs of inventory flows, resulting in balance sheet values
that are closer to current values. The use of LIFO has the opposite effect. Thus if FIFO
is used, the current exchange rate should be applied to the balance sheet inventory
account; with LIFO an applicable historical rate should be used. This was the dis-
tinction recognized by the temporal method. The two methods are now illustrated by
a numerical example.5
In this example the Canadian subsidiary of a U.S. company with a Canadian
dollar functional currency started business and acquired fixed assets at the beginning
of the year when the Canadian $/U.S. $ exchange rate was .95. The average exchange
rate for the period was .90, the rate at the end of the period was .85, and the historical
rate for inventory was .91. The LIFO inventory valuation method is employed.
The different methods are illustrated in Table 22.1. The temporal method used
in FASB 8 and for inflationary economies in FASB 52 applies the current (end-of-
period) rate to monetary assets and liabilities. It uses the applicable historical rates
for the nonmonetary assets and liabilities. Since LIFO is used, the balance sheet in-
ventory account reflects historical costs, and the historical rate for inventories is used.
In the income statement the applicable average rates are applied to all items except
cost of goods sold and depreciation. Depreciation expense in the income statement
would employ the same rate as fixed assets on the balance sheet.
In contrast, FASB 52 applies the current rate to all balance sheet items except
common stock to which the historical rate is applied. The average rate is applied
to all income statement items. The net income figure that results is reflected in the
translated retained earnings account. Total assets and claims are brought into bal-
ance by a translation adjustment account.
Taken from Peat, Marwick, Mitchell & Co., Statement of Financial Accounting Standards No. 52, Foreign
Currency Translation, 1981.
83< INTERNATIONAL FINANCIAL MANAGEMENT: TESTS AND IMPLICATIONS
The use of the current method of FASB P< results in financial ratios that are
unchanged from their relationships in the foreign currency îefore translation. This
is claimed to îe an advantage of the new method. But if the underlying reality is a
change in the ratios, preserving them is a distortion. The logic of the temporal method
captures the underlying economic determinants of exposure as demonstrated in the
previous discussion of the net monetary creditor or deîtor position of the foreign
suîsidiary. It is difficult to discern the economic logic for the application of the cur-
rent method prescriîed îy FASB P<.
SUMMARY 833
SUMMARY
contracts in real terms. As GLS show, if a real riskless rate exists, and agents have iden-
tical homothetic tastes, this nominal exchange risk is irrelevant to real magnitudes.
As long as individuals consume different baskets of goods, relative prices will
vary. This relative price risk is due to changes in underlying preferences of individuals
and/or changes in supply conditions in various countries. Exchange rate changes due
to relative price movements represent real risks that can be hedged only at a cost. As
we demonstrated, exchange rate risk is a form of business risk resulting from changes
in relative prices, which in turn reflect changes in demand and supply conditions.
In an efficient market we would expect PPP to hold on an expected value basis.
Roll's [1979] study shows that PPP holds on average and that deviations from PPP
are more attributable to relative price fluctuations than to inflation fluctuations. In
examining other aspects of market efficiency, several different approaches have been
used. We briefly describe the methods and results.
One requirement of market efficiency is that arbitrage profits after transactions
costs do not exist. The preponderance of studies support this conclusion. Disagree-
ment exists on whether the forward rate is an unbiased predictor of the spot rate.
Some hold that a bias should exist as a premium for risk. But the empirical studies
are not unanimous in this area.
A final test of market efficiency that has been used is to see if forecasting services
can "beat" the market. Levich [1979] found that forecasting services may yield profits
consistently for some currencies. On the whole the literature on market efficiency
seems to suggest that international markets are weak-form and semistrong-form
efficient.
The exposure of a business firm to exchange risks is defined by its cash flow and
asset stock positions. These in turn depend on expected future receipts and payments
and the firm's net monetary position. Unless payments and receipts in relation to
the future net monetary position of the firm exactly balance, the firm is exposed to a
decline or increase in the value of foreign currencies. If the net monetary position
of a firm is positive and expected receipts exceed expected payments, the firm will
suffer from a devaluation. The opposite will hold with respect to an appreciation if
the firm has a negative net monetary position and expected payments exceed expected
receipts.
There are three methods of managing exchange risk. One alternative is to develop
information about and formulate expectations between expected future exchange rates
and the current forward rate. A second is to hedge by using the forward market. A
third is to use the money and capital markets. The firm will choose various combina-
tions of these methods that will tend to produce efficiency in the foreign exchange
markets. In comparing hedging versus money and capital markets it is important to
note that IRPT will ensure agents are indifferent between the two. The benefit to the
firm of using either the money and capital markets or hedging is that it knows what
costs are involved.
PROBLEM SET
22.1 Agrimex, S.A., a Mexican corporation, borrowed $1,000,000 in dollars at a 15% interest
rate when the exchange rate was 25 pesos per dollar. When the company repaid the loan plus
interest one year later, the exchange rate was 40 pesos to the dollar.
PROBLEM SET 835
a) What was the rate of interest on the loan based on the pesos received and paid back by
Agrimex?
b) Use the interest rate parity theorem to illustrate this result.
22.2 An American manufacturing company has imported industrial machinery at a price of
DM4.6 million. The machinery will be delivered and paid for in six months. For planning pur-
poses, the American company wants to establish what the payment (in dollars) will be in six
months. It decides to use the forward market to accomplish its objective. The company con-
tacts its New York bank, which provides the quotations given in Table Q22.2. The bank states
that it will charge a commission of 1% on any transaction.
Table Q22.2
DM Swiss Franc
Six-month Eurocurrency
rates (% p.a.) denominated
in the following currencies 8 7 9
Spot exchange rates
(currency/Swiss franc) 1.1648 .56
a) Does the American company enter the forward market to go long or short of forward
DM?
b) What is the number of DM/$? What is the dollar value of the deutsche mark?
c) What is the equilibrium forward rate for the deutsche mark expressed as DM/$?
d) Does the commission increase or decrease the number of DM/$ in the transaction?
e) What price in dollars can the American company establish by using the forward market
in deutsche marks?
22.3 A West German company buys industrial machinery from a U.S. company at a price of
$10 million The machinery will be delivered and paid for in six months. The German company
seeks to establish its cost in deutsche marks. It decides to use the forward market to accom-
plish its objective. The company contacts its Bonn bank, which provides the quotations listed
in Table Q22.3. The bank states that it will charge a commission of 1% on any transaction.
a) Does the German company enter the forward market to go long or short of forward
dollars?
Table Q22.3
DM £
Six-month Eurocurrency
rates (% p.a.) denominated
in the following currencies 8 9.5 9
Spot exchange rates
(currency/U.K. £) 3.878 $1.90
836 INTERNATIONAL FINANCIAL MANAGEMENT: TESTS AND IMPLICATIONS
b) What is the number of DM/$? What is the dollar value of the deutsche mark?
c) What is the equilibrium forward rate for the deutsche mark expressed as 8/DM?
d) Does the commission increase or decrease the dollar value of the deutsche mark?
e) What price in deutsche marks can the German company establish by using the forward
market in dollars?
22.4 Globalcorp makes a sale of goods to a foreign firm and will receive FC 380,000 three
months later. Globalcorp has incurred costs in dollars and wishes to make definite the amount
of dollars it will receive in three months. It plans to approach a foreign bank to borrow an
amount of local currency such that the principal plus interest will equal the amount Globalcorp
expects to receive. The interest rate it must pay on its loan is 28%. With the borrowed funds,
Globalcorp purchases dollars at the current spot rate that are invested in the United States
at an interest rate of 8%. When Globalcorp receives the FC 380,000 at the end of three months,
it uses the funds to liquidate the loan at the foreign bank. The effective tax rate in both countries
is 40%.
a) What is the net amount that Globalcorp will receive if the current spot rate is FC 1.90
to the dollar?
b) How much less is this than the amount Globalcorp would have received if the remittance
had been made immediately instead of three months later?
c) At what forward rate of exchange would the amount received by Globalcorp have been
the same as that it would have obtained using the capital markets? Would Globalcorp
have sold the FC forward short or long to hedge its position?
d) If a speculator took the opposite position from Globalcorp in the forward market for
FC, would the speculator sell long or short? If the speculator received a risk premium
for holding this position, would this place the current forward rate in FC above or below
the expected future spot rate in FC per dollar?
22.5 Transcorp has made a purchase of goods from a foreign firm that will require the payment
of FC 380,000 six months later. Transcorp wishes to make definite the amount of dollars it will
need to pay the FC 380,000 on the due date. The foreign firm is domiciled in a country whose
currency has been rising in relation to the dollar in recent years. The tax rate in both countries
is 40%. Transcorp plans to borrow an amount in dollars from a U.S. bank to immediately
exchange into FCs to buy securities in the foreign country, which, with interest, will equal
FC 380,000 six months later. The interest rate that will be paid in the United States is 12%; the
interest rate that will be earned on the foreign securities is 8%. When at the end of six months
Transcorp is required to make the payment in FC, it will use the funds from the maturing foreign
securities in FC to meet its obligation in FC. At the same time it will pay off the loan plus
interest in the United States in dollars.
a) What is the net amount that Transcorp pays to meet the obligation of FC 380,000 in six
months if the current spot rate is FC 2.00 to the dollar?
b) How much more is this than the amount Transcorp would have paid if payment had been
made immediately instead of six months later?
c) At what forward rate of exchange would the amount paid by Transcorp have been the
same as that it would have paid using the capital markets? Would Transcorp have taken
the long position in the forward FC or have sold the FC forward short to hedge its position?
d) If a speculator took the opposite position from Transcorp in the forward market for FCs,
REFERENCES 837
would the speculator be long or short? If the speculator received a risk premium for holding
this position, would this place the current forward rate in FC above or below the expected
future spot rate in FC per dollar?
REFERENCES
Adler, M., "The Cost of Capital and Valuation of a Two-Country Firm," Jou]nal of Finance,
March 1974, 119-132.
, and B. Dumas, "Optimal International Acquisitions," Jou]nal of Finance, March 1975,
1-20.
, "Exposure to Currency Risk: Definition and Measurement," Financial Management,
Summer 1984, 41-50.
Batten, S., and D. L. Thornton, "Discount Rate Changes and the Foreign Exchange Market,"
Fede]al Rese]ve Bank of St. Louis Review, December 1984, 279 -292.
Black, F., "International Capital Market Equilibrium with Investment Barriers," Jou]nal of
Financial Economics, December 1974, 337-352.
Breeden, D. T., "An Intertemporal Asset Pricing Model with Stochastic Consumption and
Investment Opportunities," Jou]nal of Financial Economics, September 1979, 265-296.
Cho, D. C.; C. S. Eun; and L. Senbet, "International Arbitrage Pricing Theory: An Empirical
Investigation," Jou]nal of Finance, June 1986, 313-329.
Cornell, B., "Spot Rates, Forward Rates, and Exchange Market Efficiency," Jou]nal of Financial
Economics, August 1977, 55-65.
, "Inflation, Relative Price Changes, and Exchange Risk," Financial Management,
Autumn 1980, 30-34.
, and K. Dietrich, "The Efficiency of the Market for Foreign Exchange under Floating
Exchange Rates," Review of Economics and Statistics, February 1978, 111-121.
Dufey, G., and S. L. Srinivasulu, "The Case for Corporate Management of Foreign Exchange
Risk," Financial Management, Winter 1983, 54-62.
Errunza, V. R., and L. W. Senbet, "The Effects of International Operations on the Market
Value of the Firm: Theory and Evidence," Jou]nal of Finance, May 1981, 401-419.
Fama, E. F., Foundations of Finance. Basic Books, New York, 1976.
, "Inflation, Output and Money," Jou]nal of Business, April 1982, 201-231.
, "Forward and Spot Exchange Rates," Jou]nal of Moneta]y Economics, 1984, 319-338.
, and A. Farber, "Money Bonds and Foreign Exchange," Ame]ican Economic Review,
September 1979, 639-650.
Fama, E. F., and M. R. Gibbons, "Inflation, Real Returns, and Capital Investment," Jou]nal
of Moneta]y Economics, May 1982, 297 -327.
Farber, A.; R. Roll; and B. Solnik, "An Empirical Study of Risk under Fixed and Flexible Ex-
change Rates," in Brunner and Meltzer, eds., Stabilization of the Domestic and Inte]national
Economy. North Holland, Amsterdam, 1977, 235-275.
Fieleke, N. S., "Foreign-Currency Positioning by U.S. Firms: Some New Evidence," Review of
Economics and Statistics, February 1981, 35 -42.
Forsythe, R.; T. R. Palfrey; and C. R. Plott, "Futures Markets and Informational Efficiency:
A Laboratory Examination," Jou]nal of Finance, September 1984, 955-981.
8Ç8 INTERNATIONAL FINANCIAL MANAGEMENT: TESTS AND IMPLICATIONS
Frenkel, J., and R. Levich, "Covered Interest Arîitrage: Unexploited Profits?" Jou]nal of
Political Economy, April =97P, 3<P-338.
, "Transactions Costs and Interest Arîitrage: Tranquil versus Turîulent Periods," Jou]-
nal of Political Economy, Decemîer =977, =<09-=<<6.
Grauer, F. L. A.; R. H. Litzenîerger; and R. E. Stehle, "Sharing Rules and Equiliîrium in an
International Capital Market under Uncertainty," Jou]nal of Financial Economics, June
=976, <33-<P6.
Gregory, A. W., and T. H. McCurdy, "The Unîiasedness Hypothesis in the Forward Foreign
Exchange Market," Eu]opean Economic Review, =986, 36P-38=.
Gruîel, H. G., "Internationally Diversified Portfolios: Welfare Gains and Capital Flows,"
Ame]ican Economic Review, Decemîer =968, =<99-=3=4.
Hekman, C. R., "A Model of Foreign Exchange Exposure," Jou]nal of Inte]national Business
Studies, Summer =98P, 8P —99.
Hodgson, J. S., and P. Phelps, "The Distriîuted Impact of Price-Level Variation on Floating
Exchange Rates," Review of Economics and Statistics, Feîruary =97P, P8-64.
Hodrick, R. J., and S. Srivastava, "An Investigation of Risk and Return in Forward Foreign
Exchange," Jou]nal of Inte]national Money and Finance, April =984, P-30.
Huang, R. D., "Some Alternative Tests of Forward Exchange Rates as Predictors of Future
Spot Rates," Jou]nal of Inte]national Money and Finance, August =984, =P3-=67.
Hutchison, M. M., "Swaps," FRBSF Weekly Lette], May 3, =98P.
Isard, P., "Exchange Rate Determination: A Survey of Popular Views and Recent Models."
Princeton Studies in International Finance, No. 4<, =978.
Jacquillat, B., and B. Solnik, "Multinationals Are Poor Tools for Diversification," Jou]nal of
Po]tfolio Management, Winter =978, 8-=<.
Johnson, L. J., and C. H. Walther, "New Evidence of the Effectiveness of Portfolio Hedges in
Currency Forward Markets," Management Inte]national Review, =984, =P-<3.
Kaufold, H., and M. Smirlock, "Managing Corporate Exchange and Interest Rate Exposure,"
Financial Management, Autumn =986, 64-7<.
Lee, C., "A Stock Adjustment Analysis of Capital Movements: The U.S. Canadian Case," Jou]-
nal of Political Economy, July =969, P=<-P<3.
Lee, W. Y., and K. S. Sachdeva, "The Role of the Multinational Firm in the Integration of
Segmented Capital Markets," Jou]nal of Finance, May =977, 479-49<.
Lessard, D. R., "International Portfolio Diversification: A Multivariate Analysis for a Group
of Latin-American Countries," Jou]nal of Finance, June =973, 6=9-633.
, ed., Inte]national Financial Management- Theo]y and Application, <nd ed. John Wiley
and Sons, New York, =98P.
Levich, R., "The Efficiency of the Market for Foreign Exchange: A Review and Extension,"
in Lessard, ed., Inte]national Financial Management. Warren, Gorham and Lamont,
Boston, =979, <43-<77.
Levy, H., and M. Sarnat, "International Diversification of Investment Portfolios," Ame]ican
Economic Review, Septemîer =970, 668-67P.
, "Devaluation Risk and the Portfolio Analysis of International Investment," in Elton
and Gruîer, eds., Inte]national Capital Ma]kets. North Holland, Amsterdam, =97P,
=77-<06.
Longworth, D., "Testing the Efficiency of the Canadian—U.S. Exchange Market under the
Assumption of No Risk Premium," Jou]nal of Finance, March =98=, 43-P=.
REFERENCES 839
Maldonado, R., and A. Saunders, "International Portfolio Diversification and the Inter-
Temporal Stability of International Stock Market Relationships, 1957-78," Financial
Management, Autumn 1981, 54-63.
, "Foreign Exchange Restrictions and the Law of One Price," Financial Management,
Spring 1983, 19-23.
Merton, R. C., "An Intertemporal Capital Asset Pricing Model," Economet]ica, September
1973, 867-887.
Mundell, R., "Inflation and Real Interest," Jou]nal of Political Economy, June 1963, 280-283.
Philippatos, G. C., and A. Christofi, "Liquid-Asset Management Modeling for Inter-Subsidiary
Operations of Multinational Corporations," Management Inte]national Review, 1984, 4-14.
, and P. Christofi, "The Inter-Temporal Stability of International Stock Market Re-
lationships: Another View," Financial Management, Winter 1983, 63-69.
Robichek, A. A., and M. R. Eaker, "Foreign Exchange Hedging and the Capital Asset Pricing
Model," Jou]nal of Finance, June 1978, 1011-1018.
Rodriguez, R. M., "Corporate Exchange Risk Management: Theme and Aberrations," Jou]nal
of Finance, May 1981, 427-439.
Roll, R., "A Critique of the Asset Pricing Theory's Tests; Part I: On Past and Potential Test-
ability of the Theory," Jou]nal of Financial Economics, March 1977, 129-176.
, "Violations of Purchasing Parity and their Implications for Efficient Markets," in
Sarnat and Szego, eds., Inte]national Finance and T]ade, Vol. 1. Ballinger, Cambridge,
Mass., 1979, 133-176.
, and B. Solnik, "A Pure Foreign Exchange Asset Pricing Model," Jou]nal of Inte]na-
tional Economics, May 1977, 161-179.
Shapiro, A. C., Inte]national Co]po]ate Finance- A Su]vey and Synthesis. Financial Manage-
ment Association, Tampa, Fla., 1986.
Solnik, B., "The International Pricing of Risk: An Empirical Investigation of the World Capital
Market Structure," Jou]nal of Finance, May 1974a, 365-378.
, "Why Not Diversify Internationally Rather Than Domestically?" Financial Analysts
Jou]nal, July—August 1974b, 48-54.
, "An Equilibrium Model of the International Capital Market," Jou]nal of Economic
Theo]y, August 1974c, 500-525.
, "An International Market Model of Security Price Behavior," Jou]nal of Financial
and Quantitative Analysis, September 1974d, 537-554.
, "Testing International Asset Pricing: Some Pessimistic Views," Jou]nal of Finance,
May 1977, 503-512.
, "International Parity Conditions and Exchange Risk," Jou]nal of Banking and Finance,
October 1978, 281-293.
Stehle, R., "An Empirical Test of the Alternate Hypotheses of National and International
Pricing of Risky Assets," Jou]nal of Finance, May 1977, 493-502.
Stockman, A., "Risk, Information, and Forward Exchange Rates," in Frenkel and Johnson,
eds., The Economics of Exchange Rates. Addison-Wesley, Reading, Mass., 1978, 159-179.
, "A Theory of Exchange Rate Determination," Jou]nal of Political Economy, August
1980, 673-699.
Stulz, R., "On the Effects of Barriers to International Investment," Jou]nal of Finance, Sep-
tember 1981a, 923-934.
8<0 INTERNATIONAL FINANCIAL MANAGEMENT: TESTS AND IMPLICATIONS
A. INTRODUCTION
In any economy, capitalist or socialist, we find positive rates of interest. This reflects
two underlying influences: the productivity of economic goods and time preference.
Capital goods are goods used in the production of other goods and services. Some
capital goods are specialized machinery and others are materials such as iron,
copper, or textiles—used in the production of machinery to produce other goods.
More basically, our productive efforts may be used to produce goods that we consume
immediately or to produce goods that will produce other goods and services for
future use. One reason to use some of our productive efforts to have goods that will
produce future goods is that the postponement of current consumption will enable us
to have more wealth in the future than we would otherwise have. This is true whether
we think of the use of actual goods or financial spending power used on current
consumption vs. goods that will produce future goods. For example, we can consume
grains now or plant them to harvest future crops that will represent larger quantities
than the seeds with which we started. Because of the productivity of goods, they have
a time value. A bushel of seeds today will become several bushels of grain in the
future. So productivity is one basis for the time value of money and positive rates
of interest.
A second basis is time preference. Would we rather have the use of an automobile
now or wait five years? Clearly, it is more advantageous to have the use of goods now
than to wait for them.
841
842 DISCOUNTING
values. For example, assume that if a company received funds immediately, it could
earn a 10% return on those funds. We could then state the problem as follows: Let
The result is the compound interest formula. In general terms it may be stated
as follows:
Sn = P(1 + (A.1)
We now have all the information needed to compute the value at the end of the fifth
year, using a compound interest table' (Table A.1):
S5 = $1000(1.10)5.
We then look in the compound interest table to find that at 10% a dollar over a
five-year period grows to $1.611. Since the amount we have is $1000, it is multiplied
times the interest factor:
S 5 = $1000(1.611) = $1611.
Therefore if the firm can earn 10% with the money, it is more worthwhile for it to
receive the $1000 today rather than at the end of the fifth year.
1
All tables mentioned in Appendix A are located at the end of Appendix A.
THE TIME VALUE OF MONEY: DISCRETE COMPOUNDING 843
instead of for Sn , which, in this situation, is known. By simple algebra the required
formula would be:
S„
= (A.2)
(1 + on
U + rr
A n, = r l, (A.3)
where
A = present value of an annuity,
a = amount of the periodic annuity payment,
r = interest factor,
n = number of annuity payments.
844 DISCOUNTING
= a + a + a a
+ • • • + (1 + r)"
1+r (1 + r)2 (1 + r)3
If we let u = [1/(1 + r)] this becomes
A = au + au 2 + au 3 + • • + au". (A.4)
Multiplying Eq. (A.4) by u and subtracting the result from Eq. (A.4) yields
An, — uA, = au — au" ± 1,
au(1 — un)
An , =
1—u
Substituting back the value of u we have:
a
1
(} + r [ 1 (} +1 r ni
A,—
= a[ 1 — (1 +
(A.3)
r
Note that if the number of payments is infinite, then the present value of the annuity
becomes
a
lim An,,. = —, (A.5)
n
-1
since we know that when r > 0, then limn ,co (1 + r) ` = 0. An example of an annuity
with an infinite number of constant payments is a consol bond. It pays a coupon at
the end of each time period (usually a year) and never matures.
The expression in brackets in Eq. (A.3) is rather cumbersome. For convenience,
then, instead of the cumbersome expression set out above, we shall use the symbol
Pn, r , where Pn ,=
r present value of an annuity factor for n years at r percent. Equation
(A.3) above can therefore be rewritten as
An„ = aP„. (A.3a)
Substituting actual numbers and using Table A.4 (the present value of an annuity
interest table), we would have the following for 10 years at 10%:
$6145 = $1000(6.145).
THE TIME VALUE OF MONEY: DISCRETE COMPOUNDING 845
$6145
Pio,to% — $1000 6.145.
In Table A.4, which shows the present value of periodic payments received annu-
ally, we look across the row for year 10 until we find the interest rate that corre-
sponds to the interest factor 6.145. This is 10%. We are earning a 10% return on our
investment.
Let us consider another situation. Suppose that we are going to receive a return
of $2000 per year for five years from an investment of $8424. What is the return on
our investment? This is generally referred to as the internal rate of return on the
investment, or it is also sometimes referred to as the DCF discounted cash flow
approach to valuing an investment.
We follow the same procedure as before:
$8424
P, =
$2000
= 4.212.
We look again in the present value of an annuity table (Table A.4) along the
row for the year 5 to find the interest factor 4.212. We then look at the interest rate
at the top of the column to find that it is 6%. Thus the return on that investment
is 6%. If our required rate of return were 10% we would not find this investment
attractive. On the other hand, if the required return on our investment were only 5%,
we would consider the investment attractive.
These relationships can be used in still another way. Taking the facts of the pre-
ceding illustration, we may ask the following question: Given an investment that yields
$2000 per year for five years, at an appropriate discount factor (or cost of capital) of
846 DISCOUNTING
6%, what is that investment worth today? What is the present value of a series of future
income flows? For example, if a firm were to make a sale of goods on an open account
with a down payment of $1000 plus yearly payments of $2000 for five years, what
would the present value of all the payments be at a 6% interest rate? From our pre-
vious calculations we know that the series of payments of $2000 for five years at a
6% interest rate are worth $8424 today. When we add the $1000 down payment
to this figure we would have a total of $9424.
[(1 + r)" —
S" r = a (A.6)
where
= the future sum to which an annuity will accumulate in n years at rate r,
a = the amount of the annuity payment.
Suppose the firm were to receive annual payments of $1000 a year for 10 years
and is earning an interest rate of 10%. What will be the amount that the firm will
have at the end of 10 years? We can solve this problem by consulting Table A.3.
Utilizing our equation, we would have
Sn,,. = $1,000(15.937)
= $15,937.
The 10 payments of $1,000 with interest would amount to $15,937 by the end
of the tenth year. Thus if we had to make a payment of $15,937 in 10 years, we would
be able to do it by annual payments of $1,000 per year into a fund that earns interest
at 10% per year.
2
Note that the present value of an annuity can be obtained by discounting the expression back to the
present:
[(1 + - =
L r(1 +
Now divide the numerator and the denominator by (1 + r)". We have
[1 — (1 + r)1
A,„ = a
r
This is now in the form of Eq. (A.3), the present value of an annuity.
THE TIME VALUE OF MONEY: DISCRETE COMPOUNDING 847
Using the present value interest table (Table A.2) at an interest rate of 15%, we
obtain the amounts indicated above. The interest factor is multiplied by the receipts
to provide the amounts in the present value column. The amounts for each year are
then summed to provide the present value of the investment, which in this example
is $804.60. What we are doing in this example is illustrating how an annuity of un-
equal payments that could not be computed directly from the present value of an
annuity table (Table A.4) can be handled by breaking the problem into a series of
one-year payments received at successively later time periods.
By multiplying Eq. (A.7) by u and subtracting the result from Eq. (A.7), we obtain
PV — uPV = ud o(1 — u"),
and solving for the present value of the growing annuity, we have
ud,(1 ÿ un)
PV =
1—u
Substituting back the value of u gives us
(1 + g )d,[1. ( 1
1+] 1+]
PV =
1
(1 + g)
1+]
By rearranging terms and recalling that d o(1 + g) = d 1 , we obtain
(11 + r
PV = d,[1 • (A.8)
]ÿg
Equation (A.8) is the present value of n annuity payments that start at a level of d,
and grow at a constant rate, g.
Note that if the number of payments is infinite, we can obtain a finite present
value if we assume that the growth rate in dividends, g, is less than the time value
of money, ]. If g = ], then the fraction in the numerator of Eq. (A.8) goes to zero
in the limit as n approaches infinity:
(1 )n g
hill = 0, g = ].
n —> co 1 ]
Therefore the present value of an infinite number of growing dividends is
d1
lim PV = (A.9)
n co ]ÿg
Equation (A.9) is used frequently in the text, where it is called the Go]don g]owth
model. It provides us with an estimate of the present value of a share of common
stock where the stream of dividends received from it is assumed to grow at a constant
rate that is assumed to be less than the discount rate (which in this case would be
the cost of equity capital, ks ).
value of A plus the present value of B plus the present value of C plus the present
value of D. In this earlier example, we had four cash flows (F 1 + F 2 + F 3 + F4). They
came in different time periods. We can obtain the present value of cash flows that
occur in different years by applying the appropriate discount factor and adding:
F, F, F3 F4
PV + + + (A.10)
(1 + r) (1 + (1 + r)3 (1 + r)4
If investments over a period of years are required to produce the cash flows,
their values are accumulated to the present:
F, F2 F„
NPV — / 0 — / 1(1 + r) - — / 2(1 + r)2 + + + • •••• +
(1 + r) (1 + r)2 (1 + r)n •
(A.13)
The net present value represents the addition to value created by an investment
project. The applicable discount factor is the rate of interest or cost of capital appro-
priate to the characteristics of a project. Hence the NPV criterion is a general rule
for allocating resources. By following the NPV rule the financial markets produce
the maximum amount of additions to the value of the economy. Individual investors
850 DISCOUNTING
can then buy and sell in the financial markets to obtain the use of funds for consump-
tion in early years or to postpone consumption to later years. Nor do managers of
individual firms need to consult the preferences of their shareholders or owners to
determine whether they prefer investments that come to fruition in early years or
later years. By following the NPV criterion, the maximum amount of additions to
wealth will be achieved. The financial markets provide opportunities to the individual
investors to arrange their consumption over time in years that they prefer. The NPV
rule leads to the fundamental principles of valuation for flows of different kinds and
different time patterns.
06 5(2)
S5/2 = $1000(1 ) .
2
Thus the new expression is equivalent to compounding the $1000 at 3% for 10 periods.
The compound interest table (Table A.1) for 10 years shows that the interest factor
would be 1.344. Our equation would therefore read:
10
S5 /2 = $1000(1 + .03) ,
= $1344.
It will be noted that with semiannual compounding the future sum amounts to
$1344 as compared with the $1338 we had before. Frequent compounding provides
compound interest paid on compound interest, so the amount is higher. Thus we
would expect that daily compounding, as some financial institutions advertise, or
THE TIME VALUE OF MONEY: CONTINUOUS COMPOUNDING 851
(A.14a)
V= Po ( 1 + — •
ky
Since we can multiply qt by k/k, we can set qt = (q/k)(kt) and rewrite Eq. (A.14a) as
(q/k)](k
t
V, = Po [(1 + k (A.15)
approaches the value 2.718. . . . The value e is defined as this limiting case:
1 )111
e = fim (1 + — = 2.718 . . . .
m-00 m
Thus if $2225 is due in eight years and if the appropriate continuous discount rate
k is 10%, the present value of this future payment is
$2225
PV = = $1000.
2.225
Using basic integral calculus (discussed in Appendix D), we can express the present
value of a constant payment stream as the discounted value of the payment function
given in Eq. (A.21):
Note that we have employed Eq. (A.20) to discount each payment. The stream of
payments is assumed to start immediately (t = 0) and continue for n time periods.
Hence the limits of integration in Eq. (A.22) are 0 to n. Following the applicable rules
of integral calculus to evaluate the definite integral, we obtain
PV = a 0 on e —kt d t
= ao [ e -kt n
- kn
—e — e0
a,[
[1 — ke -k ni .
= cto (A.23)
Equation (A.23) is the continuous-time analogue to Eq. (A.3), which was the discrete
time version of the present value of an annuity of constant payments. Note that the
continuous discount factor e'n in Eq. (A.23) is roughly equivalent to the discrete
-
discount factor (1 + r) n in Eq. (A.3).
If we want the present value of an infinite stream of constant, continuously com-
pounded payments we take the limit of Eq. (A.23) as n becomes infinite:
a,
lim PV = • (A.24)
PV = o ate' dt
= a _ fn egt e — kt dt.
°
854 DISCOUNTING
PV = a, f n e — (1' g)t
dt.
o
SUMMARY
Consumption is allocated over time by "the" interest rate. Positive rates of interest
induce people to postpone consumption and save part of their income. The pool of
savings at any given time is used for investments that yield output in the form of
goods that may be consumed at future dates.
The combined preferences of all members of society and the society's technology
combine to determine the pattern of interest rates that will allocate consumption
over time optimally. The structure of interest rates guides individuals into making
investment decisions that are most desired by the society as a whole.
Present value or future value calculations at appropriately chosen interest rates,
given the riskiness of the project, will tell an investor whether the future receipts are
sufficient to justify the current investment. Since the pattern of interest rates is deter-
mined by the behavior of all members of society, a positive present value means not
only that the project will yield a profit to the investor but also that no member of the
society has a superior use for the resources being invested. If many other investment
opportunities were to appear that were superior to the one in question, the interest
rate appropriate for the present value calculation would rise, and the present value
of the project might then appear to be negative.
To evaluate projects with cash flows distributed over time it is necessary to ex-
press all flows in terms of their value at one specific point in time. Expressing them
in terms of value today is discounting to net present value; expressing them at their
SUMMARY 855
DNIINflOOSICI
Year 1% 2% 3% 4% 5% 6% 7% 8% 9% 10% 11% 12% 13% 14% 15% 16%
1 ................ 1.010 1.020 1.030 1.040 1.050 1.060 1.0L0 1.080 1.090 1.100 1.110 1.120 1.130 1.140 1.150 1.160
2 ................ 1.020 1.040 1.061 1.082 1.102 1.124 1.145 1.166 1.188 1.210 1.232 1.254 1.2LL 1.300 1.322 1.346
3 ................ 1.030 1.061 1.093 1.125 1.158 1.191 1.225 1.260 1.295 1.331 1.368 1.405 1.443 1.482 1.521 1.561
4 ................ 1.041 1.082 1.126 1.1L0 1.216 1.262 1.311 1.360 1.412 1.464 1.518 1.5L4 1.631 1.689 1.L49 1.811
5 ................ 1.051 1.104 1.159 1.21L 1.2L6 1.338 1.403 1.469 1.539 1.611 1.685 1.L62 1.842 1.925 2.011 2.100
6 ................ 1.062 1.126 1.194 1.265 1.340 1.419 1.501 1.58L 1.6LL 1.LL2 1.8L0 1.9L4 2.082 2.195 2.313 2.436
L ................ 1.0L2 1.149 1.230 1.316 1.40L 1.504 1.606 1.L14 1.828 1.949 2.0L6 2.211 2.353 2.502 2.660 2.826
8 ................ 1.083 1.1L2 1.26L 1.369 1.4LL 1.594 1.L18 1.851 1.993 2.144 2.305 2.4L6 2.658 2.853 3.059 3.2L8
9 ................ 1.094 1.195 1.305 1.423 1.551 1.689 1.838 1.999 2.1L2 2.358 2.558 2.LL3 3.004 3.252 3.518 3.803
10 ................ 1.105 1.219 1.344 1.480 1.629 1.L91 1.96L 2.159 2.36L 2.594 2.839 3.106 3.395 3.L0L 4.046 4.411
11 ................ 1.116 1.243 1.384 1.539 1.L10 1.898 2.105 2.332 2.580 2.853 3.152 3.4L9 3.836 4.226 4.652 5.11L
12 ................ 1.12L 1.268 1.426 1.601 1.L96 2.012 2.252 2.518 2.813 3.138 3.499 3.896 4.335 4.818 5.350 5.936
13 ................ 1.138 1.294 1.469 1.665 1.886 2.133 2.410 2.L20 3.066 3.452 3.883 4.363 4.898 5.492 6.153 6.886
14 ................ 1.149 1.319 1.513 1.L32 1.980 2.261 2.5L9 2.93L 3.342 3.L9L 4.310 4.88L 5.535 6.261 L.0L6 L.988
15 ................ 1.161 1.346 1.558 1.801 2.0L9 2.39L 2.L59 3.1L2 3.642 4.1LL 4.L85 5.4L4 6.254 L.138 8.13L 9.266
16 ................ 1.1L3 1.3L3 1.605 1.8L3 2.183 2.540 2.952 3.426 3.9L0 4.595 5.311 6.130 L.06L 8.13L 9.358 10.L48
1L ................ 1.184 1.400 1.653 1.948 2.292 2.693 3.159 3.L00 4.328 5.054 5.895 6.866 L.986 9.2L6 10.L61 12.468
18 ................ 1.196 1.428 1.L02 2.026 2.40L 2.854 3.380 3.996 4.L1L 5.560 6.544 L.690 9.024 10.5L5 12.3L5 14.463
19 ................ 1.208 1.45L 1.L54 2.10L 2.52L 3.026 3.61L 4.316 5.142 6.116 L.263 8.613 10.19L 12.056 14.232 16.LLL
20 ................ 1.220 1.486 1.806 2.191 2.653 3.20L 3.8L0 4.661 5.604 6.L28 8.062 9.646 11.523 13.L43 16.36L 19.461
Table A.2 Present Value of $1 P = S.(1+ r) -"
Years Hence 1% 2% 4% 6% 8% 10% 12% 14% 15% 16% 18% 20% 22% 24% 25% 26% 28% 30% 35% 40% 45% 50%
1 ................ 0.990 0.980 0.962 0.943 0.926 0.909 0.893 0.877 0.870 0.862 0.847 0.833 0.820 0.806 0.800 0.794 0.781 0.769 0.741 0.714 0.690 0.667
2 ................ 0.980 0.961 0.925 0.890 0.857 0.826 0.797 0.769 0.756 0.743 0.718 0.694 0.672 0.650 0.640 0.630 0.610 0.592 0.549 0.510 0.476 0.444
3 ................ 0.971 0.942 0.889 0.840 0.794 0.751 0.712 0.675 0.658 0.641 0.609 0.579 0.551 0.524 0.512 0.500 0.477 0.455 0.406 0.364 0.328 0.296
4 ................ 0.961 0.924 0.855 0.792 0.735 0.683 0.636 0.592 0.572 0.552 0.516 0.482 0.451 0.423 0.410 0.397 0.373 0.350 0.301 0.260 0.226 0.198
5 ................ 0.951 0.906 0.822 0.747 0.681 0.621 0.567 0.519 0.497 0.476 0.437 0.402 0.370 0.341 0.328 0.315 0.291 0.269 0.223 0.186 0.156 0.132
6 ................ 0.942 0.888 0.790 0.705 0.630 0.564 0.507 0.456 0.432 0.410 0.370 0.335 0.303 0.275 0.262 0.250 0.227 0.207 0.165 0.133 0.108 0.088
7 ................ 0.933 0.871 0.760 0.665 0.583 0.513 0.452 0.400 0.376 0.354 0.314 0.279 0.249 0.222 0.210 0.198 0.178 0.159 0.122 0.095 0.074 0.059
8 ................ 0.923 0.853 0.731 0.627 0.540 0.467 0.404 0.351 0.327 0.305 0.266 0.233 0.204 0.179 0.168 0.157 0.139 0.123 0.091 0.068 0.051 0.039
9 ................ 0.914 0.837 0.703 0.592 0.500 0.424 0.361 0.308 0.284 0.263 0.225 0.194 0.167 0.144 0.134 0.125 0.108 0.094 0.067 0.048 0.035 0.026
10 ................ 0.905 0.820 0.676 0.558 0.463 0.386 0.322 0.270 0.247 0.227 0.191 0.162 0.137 0.116 0.107 0.099 0.085 0.073 0.050 0.035 0.024 0.017
11 ................ 0.896 0.804 0.650 0.527 0.429 0.350 0.287 0.237 0.215 0.195 0.162 0.135 0.112 0.094 0.086 0.079 0.066 0.056 0.037 0.025 0.017 0.012
12 ................ 0.887 0.788 0.625 0.497 0.397 0.319 0.257 0.208 0.187 0.168 0.137 0.112 0.092 0.076 0.069 0.062 0.052 0.043 0.027 0.018 0.012 0.008
13 ................ 0.879 0.773 0.601 0.469 0.368 0.290 0.229 0.182 0.163 0.145 0.116 0.093 0.075 0.061 0.055 0.050 0.040 0.033 0.020 0.013 0.008 0.005
14 ................ 0.870 0.758 0.577 0.442 0.340 0.263 0.205 0.160 0.141 0.125 0.099 0.078 0.062 0.049 0.044 0.039 0.032 0.025 0.015 0.009 0.006 0.003
15 ................ 0.861 0.743 0.555 0.417 0.315 0.239 0.183 0.140 0.123 0.108 0.084 0.065 0.051 0.040 0.035 0.031 0.025 0.020 0.011 0.006 0.004 0.002
16 ................ 0.853 0.728 0.534 0.394 0.292 0.218 0.163 0.123 0.107 0.093 0.071 0.054 0.042 0.032 0.028 0.025 0.019 0.015 0.008 0.005 0.003 0.002
17 ................ 0.844 0.714 0.513 0.371 0.270 0.198 0.146 0.108 0.093 0.080 0.060 0.045 0.034 0.026 0.023 0.020 0.015 0.012 0.006 0.003 0.002 0.001
18 ................ 0.836 0.700 0.494 0.350 0.250 0.180 0.130 0.095 0.081 0.069 0.051 0.038 0.028 0.021 0.018 0.016 0.012 0.009 0.005 0.002 0.001 0.001
19 ................ 0.828 0.686 0.475 0.331 0.232 0.164 0.116 0.083 0.070 0.060 0.043 0.031 0.023 0.017 0.014 0.012 0.009 0.007 0.003 0.002 0.001
20 ................ 0.820 0.673 0.456 0.312 0.215 0.149 0.104 0.073 0.061 0.051 0.037 0.026 0.019 0.014 0.012 0.010 0.007 0.005 0.002 0.001 0.001
21 ................ 0.811 0.660 0.439 0.294 0.199 0.135 0.093 0.064 0.053 0.044 0.031 0.022 0.015 0.011 0.009 0.008 0.006 0.004 0.002 0.001
22 ................ 0.803 0.647 0.422 0.278 0.184 0.123 0.083 0.056 0.046 0.038 0.026 0.018 0.013 0.009 0.007 0.006 0.004 0.003 0.001 0.001
23 ................ 0.795 0.634 0.406 0.262 0.170 0.112 0.074 0.049 0.040 0.033 0.022 0.015 0.010 0.007 0.006 0.005 0.003 0.002 0.001
24 ................ 0.788 0.622 0.390 0.247 0.158 0.102 0.066 0.043 0.035 0.028 0.019 0.013 0.008 0.006 0.005 0.004 0.003 0.002 0.001
25 ................ 0.780 0.610 0.375 0.233 0.146 0.092 0.059 0.038 0.030 0.024 0.016 0.010 0.007 0.005 0.004 0.003 0.002 0.001 0.001
26 ................ 0.772 0.598 0.361 0.220 0.135 0.084 0.053 0.033 0.026 0.021 0.014 0.009 0.006 0.004 0.003 0.002 0.002 0.001
27 ................ 0.764 0.586 0.347 0.207 0.125 0.076 0.047 0.029 0.023 0.018 0.011 0.007 0.005 0.003 0.002 0.002 0.001 0.001
28 ................ 0.757 0.574 0.333 0.196 0.116 0.069 0.042 0.026 0.020 0.016 0.010 0.006 0.004 0.002 0.002 0.002 0.001 0.001
29 ................ 0.749 0.563 0.321 0.185 0.107 0.063 0.037 0.022 0.017 0.014 0.008 0.005 0.003 0.002 0.002 0.001 0.001 0.001
Li8sTZSVZ
30 ................ 0.742 0.552 0.308 0.174 0.099 0.057 0.033 0.020 0.015 0.012 0.007 0.004 0.003 0.002 0.001 0.001 0.001
40 ................ 0.672 0.453 0.208 0.097 0.046 0.022 0.011 0.005 0.004 0.003 0.001 0.001
50 ................ 0.608 0.372 0.141 0.054 0.021 0.009 0.003 0.001 0.001 0.001
00
00
DNIINI103SICI
Taîle A.3 Sum of an Annuity for $= for n Years $1[11 + - 1]
Snr
=
6 ........... 6.152 6.308 6.468 6.633 6.802 6.975 7.153 7.336 7.523 7.716 7.913 8.115 8.323 8.536 8.754 8.977
7 ........... 7.214 7.434 7.662 7.898 8.142 8.394 8.654 8.923 9.200 9.487 9.783 10.089 10.405 10.730 11.067 11.414
8 ........... 8.286 8.583 8.892 9.214 9.549 9.897 10.260 10.637 11.028 11.436 11.859 12.300 12.757 13.233 13.727 14.240
9 ........... 9.369 9.755 10.159 10.583 11.027 11.491 11.978 12.488 13.021 13.579 14.164 14.776 15.416 16.085 16.786 17.518
10 ........... 10.462 10.950 11.464 12.006 12.578 13.181 13.816 14.487 15.193 15.937 16.722 17.549 18.420 19.337 20.304 21.321
11 ........... 11.567 12.169 12.808 13.486 14.207 14.972 15.784 16.645 17.560 18.531 19.561 20.655 21.814 23.044 24.349 25.733
12 ........... 12.683 13.412 14.192 15.026 15.917 16.870 17.888 18.977 20.141 21.384 22.713 24.133 25.650 27.271 29.002 30.850
13 ........... 13.809 14.680 15.618 16.627 17.713 18.882 20.141 21.495 22.953 24.523 26.212 28.029 29.985 32.089 34.352 36.786
14 ........... 14.947 15.974 17.086 18.292 19.599 21.051 22.550 24.215 26.019 27.975 30.095 32.393 34.883 37.581 40.505 43.672
15 ........... 16.097 17.293 18.599 20.024 21.579 23.276 25.129 27.152 29.361 31.772 34.405 37.280 40.417 43.842 47.580 51.659
Eln r =
[1 +,)--1
Table A.4 Present Value of $1 Received Annually
L -
Years (n) 1% 2% 4% 6% 8% 10% 12% 14% 15% 16% 18% 20% 22% 24% 25% 26% 28% 30% 35% 40% 45% 50%
1 .......... 0.990 0.980 0.962 0.943 0.926 0.909 0.893 0.877 0.870 0.862 0.847 0.833 0.820 0.806 0.800 0.794 0.781 0.769 0.741 0.714 0.690 0.667
2 .......... 1.970 1.942 1.886 1.833 1.783 1.736 1.690 1.647 1.626 1.605 1.566 1.528 1.492 1.457 1.440 1.424 1.392 1.361 1.289 1.224 1.165 1.111
3 .......... 2.941 2.884 2.775 2.673 2.577 2.487 2.402 2.322 2.283 2.246 2.174 2.106 2.042 1.981 1.952 1.923 1.868 1.816 1.696 1.589 1.493 1.407
4 .......... 3.902 3.808 3.630 3.465 3.312 3.170 3.037 2.914 2.855 2.798 2.690 2.589 2.494 2.404 2.362 2.320 2.241 2.166 1.997 1.849 1.720 1.605
5 .......... 4.853 4.713 4.452 4.212 3.993 3.791 3.605 3.433 3.352 3.274 3.127 2.991 2.864 2.745 2.689 2.635 2.532 2.436 2.220 2.035 1.876 1.737
6 .......... 5.795 5.601 5.242 4.917 4.623 4.355 4.111 3.889 3.784 3.685 3.498 3.326 3.167 3.020 2.951 2.885 2.759 2.643 2.385 2.168 1.983 1.824
7 .......... 6.728 6.472 6.002 5.582 5.206 4.868 4.564 4.288 4.160 4.039 3.812 3.605 3.416 3.242 3.161 3.083 2.937 2.802 2.508 2.263 2.057 1.883
8 .......... 7.652 7.325 6.733 6.210 5.747 5.335 4.968 4.639 4.487 4.344 4.078 3.837 3.619 3.421 3.329 3.241 3.076 2.925 2.598 2.331 2.108 1.922
9 .......... 8.566 8.162 7.435 6.802 6.247 5.759 5.328 4.946 4.772 4.607 4.303 4.031 3.786 3.566 3.463 3.366 3.184 3.019 2.665 2.379 2.144 1.948
10 .......... 9.471 8.983 8.111 7.360 6.710 6.145 5.650 5.216 5.019 4.833 4.494 4.192 3.923 3.682 3.571 3.465 3.269 3.092 2.715 2.414 2.168 1.965
11 .......... 10.368 9.787 8.760 7.887 7.139 6.495 5.937 5.453 5.234 5.029 4.656 4.327 4.035 3.776 3.656 3.544 3.335 3.147 2.752 2.438 2.185 1.977
12 .......... 11.255 10.575 9.385 8.384 7.536 6.814 6.194 5.660 5.421 5.197 4.793 4.439 4.127 3.851 3.725 3.606 3.387 3.190 2.779 2.456 2.196 1.985
13 .......... 12.134 11.343 9.986 8.853 7.904 7.103 6.424 5.842 5.583 5.342 4.910 4.533 4.203 3.912 3.780 3.656 3.427 3.223 2.799 2.468 2.204 1.990
14 .......... 13.004 12.106 10.563 9.295 8.244 7.367 6.628 6.002 5.724 5.468 5.008 4.611 4.265 3.962 3.824 3.695 3.459 3.249 2.814 2.477 2.210 1.993
15 .......... 13.865 12.849 11.118 9.712 8.559 7.606 6.811 6.142 5.847 5.575 5.092 4.675 4.315 4.001 3.859 3.726 3.483 3.268 2.825 2.484 2.214 1.995
16 .......... 14.718 13.578 11.652 10.106 8.851 7.824 6.974 6.265 5.954 5.669 5.162 4.730 4.357 4.033 3.887 3.751 3.503 3.283 2.834 2.489 2.216 1.997
17 .......... 15.562 14.292 12.166 10.477 9.122 8.022 7.120 6.373 6.047 5.749 5.222 4.775 4.391 4.059 3.910 3.771 3.518 3.295 2.840 2.492 2.218 1.998
18 .......... 16.398 14.992 12.659 10.828 9.372 8.201 7.250 6.467 6.128 5.818 5.273 4.812 4.419 4.080 3.928 3.786 3.529 3.304 2.844 2.494 2.219 1.999
19 .......... 17.226 15.678 13.134 11.158 9.604 8.365 7.366 6.550 6.198 5.877 5.316 4.844 4.442 4.097 3.942 3.799 3.539 3.311 2.848 2.496 2.220 1.999
20 .......... 18.046 16.351 13.590 11.470 9.818 8.514 7.469 6.623 6.259 5.929 5.353 4.870 4.460 4.110 3.954 3.808 3.546 3.316 2.850 2.497 2.221 1.999
21 .......... 18.857 17.011 14.029 11.764 10.017 8.649 7.562 6.687 6.312 5.973 5.384 4.891 4.476 4.121 3.963 3.816 3.551 3.320 2.852 2.498 2.221 2.000
22 .......... 19.660 17.658 14.451 12.042 10.201 8.772 7.645 6.743 6.359 6.011 5.410 4.909 4.488 4.130 3.970 3.822 3.556 3.323 2.853 2.498 2.222 2.000
23 .......... 20.456 18.292 14.857 12.303 10.371 8.883 7.718 6.792 6.399 6.044 5.432 4.925 4.499 4.137 3.976 3.827 3.559 3.325 2.854 2.499 2.222 2.000
24 .......... 21.243 18.914 15.247 12.550 10.529 8.985 7.784 6.835 6.434 6.073 5.451 4.937 4.507 4.143 3.981 3.831 3.562 3.327 2.855 2.499 2.222 2.000
25 .......... 22.023 19.523 15.622 12.783 10.675 9.077 7.843 6.873 6.464 6.097 5.467 4.948 4.514 4.147 3.985 3.834 3.564 3.329 2.856 2.499 2.222 2.000
26 .......... 22.795 20.121 15.983 13.003 10.810 9.161 7.896 6.906 6.491 6.118 5.480 4.956 4.520 4.151 3.988 3.837 3.566 3.330 2.856 2.500 2.222 2.000
27 .......... 23.560 20.707 16.330 13.211 10.935 9.237 7.943 6.935 6.514 6.136 5.492 4.964 4.524 4.154 3.990 3.839 3.567 3.331 2.856 2.500 2.222 2.000
28 .......... 24.316 21.281 16.663 13.406 11.051 9.307 7.984 6.961 6.534 6.152 5.502 4.970 4.528 4.157 3.992 3.840 3.568 3.331 2.857 2.500 2.222 2.000
6S8STRIVI
29 .......... 25.066 21.844 16.984 13.591 11.158 9.370 8.022 6.983 6.551 6.166 5.510 4.975 4.531 4.159 3.994 3.841 3.569 3.332 2.857 2.500 2.222 2.000
30 .......... 25.808 22.396 17.292 13.765 11.258 9.427 8.055 7.003 6.566 6.177 5.517 4.979 4.534 4.160 3.995 3.842 3.569 3.332 2.857 2.500 2.222 2.000
40 .......... 32.835 27.355 19.793 15.046 11.925 9.779 8.244 7.105 6.642 6.234 5.548 4.997 4.544 4.166 3.999 3.846 3.571 3.333 2.857 2.500 2.222 2.000
50 .......... 39.196 31.424 21.482 15.762 12.234 9.915 8.304 7.133 6.661 6.246 5.554 4.999 4.545 4.167 4.000 3.846 3.571 3.333 2.857 2.500 2.222 2.000
860 DISCOUNTING
N 0 1 2 3 4 5 6 7 8 9
1.0 0.00000 .00995 .01980 .02956 .03922 .04879 .05827 .06766 .07696 .08618
.1 .09531 .10436 .11333 .12222 .13103 .13976 .14842 .15700 .16551 .17395
.2 .18232 .19062 .19885 .20701 .21511 .22314 .23111 .23902 .24686 .25464
.3 .26236 .27003 .27763 .28518 .29267 .30010 .30748 .31481 .32208 .32930
.4 .33647 .34359 .35066 .35767 .36464 .37156 .37844 .38526 .39204 .39878
.5 .40547 .41211 .41871 .42527 .43178 .43825 .44469 .45108 .45742 .46373
.6 .47000 .47623 .48243 .48858 .49470 .50078 .50682 .51282 .51879 .52473
.7 .53063 .53649 .54232 .54812 .55389 .55962 .56531 .57098 .57661 .58222
.8 .58779 .59333 .59884 .60432 .60977 .61519 .62058 .62594 .63127 .63658
.9 .64185 .64710 .65233 .65752 .66269 .66783 .67294 .67803 .68310 .68813
2.0 0.69315 .69813 .70310 .70804 .71295 .71784 .72271 .72755 .73237 .73716
.1 .74194 .74669 .75142 .75612 .76081 .76547 .77011 .77473 .77932 .78390
.2 .78846 .79299 .79751 .80200 .80648 .81093 .81536 .81978 .82418 .82855
.3 .83291 .83725 .84157 .84587 .85015 .85442 .85866 .86289 .86710 .87129
.4 .87547 .87963 .88377 .88789 .89200 .89609 .90016 .90422 .90826 .91228
.5 .91629 .92028 .92426 .92822 .93216 .93609 .94001 .94391 .94779 .95166
.6 .95551 .95935 .96317 .96698 .97078 .97456 .97833 .98208 .98582 .98954
.7 .99325 .99695 .00063a .00430a .00796a .01160a .01523a .01885a .02245a .02604a
.8 1.02962 .03318 .03674 .04028 .04380 .04732 .05082 .05431 .05779 .06126
.9 .06471 .06815 .07158 .07500 .07841 .08181 .08519 .08856 .09192 .09527
3.0 1.09861 .10194 .10526 .10856 .11186 .11514 .11841 .12168 .12493 .12817
.1 .13140 .13462 .13783 .14103 .14422 .14740 .15057 .15373 .15688 .16002
.2 .16315 .16627 .16938 .17248 .17557 .17865 .18173 .18479 .18784 .19089
.3 .19392 .19695 .19996 .20297 .20597 .20896 .21194 .21491 .21788 .22083
.4 .22378 .22671 .22964 .23256 .23547 .23837 .24127 .24415 .24703 .24990
.5 .25276 .25562 .25846 .26130 .26413 .26695 .26976 .27257 .27536 .27815
.6 .28093 .28371 .28647 .28923 .29198 .29473 .29746 .30019 .30291 .30563
.7 .30833 .31103 .31372 .31641 .31909 .32176 .32442 .32708 .32972 .33237
.8 .33500 .33763 .34025 .34286 .34547 .34807 .35067 .35325 .35584 .35841
.9 .36098 .36354 .36609 .36864 .37118 .37372 .37624 .37877 .38128 .38379
4.0 1.38629 .38879 .39128 .39377 .39624 .39872 .40118 .40364 .40610 .40854
.1 .41099 .41342 .41585 .41828 .42070 .42311 .42552 .42792 .43031 .43270
.2 .43508 .43746 .43984 .44220 .44456 .44692 .44927 .45161 .45395 .45629
.3 .45862 .46094 .46326 .46557 .46787 .47018 .47247 .47476 .47705 .47933
.4 .48160 .48387 .48614 .48840 .49065 .49290 .49515 .49739 .49962 .50185
.5 .50408 .50630 .50851 .51072 .51293 .51513 .51732 .51951 .52170 .52388
.6 .52606 .52823 .53039 .53256 .53471 .53687 .53902 .54116 .54330 .54543
.7 .54756 .54969 .55181 .55393 .55604 .55814 .56025 .56235 .56444 .56653
.8 .56862 .57070 .57277 .57485 .57691 .57898 .58104 .58309 .58515 .58719
.9 .58924 .59127 .59331 .59534 .59737 .59939 .60141 .60342 .60543 .60744
(2
C = 2).
(2 x 2) 1
The matrix A is a 3 x 2 matrix because it has three rows and two columns. The
matrix B is a 3 x 4 matrix because it has three rows and four columns. The matrix
C is a 2 x 2 square matrix because it has two rows and two columns.
Each number in a matrix is called an element. The element on the ith row and
jth column of the matrix A is designated by alp . For example, in the matrix A above,
all = 1, a12 = 2, a 21 = 0, and so on. Similarly, in the matrix B, b12 = 3, b 3< = 1.1.
We say that two m x n matrices are equal if all their corresponding elements are
identical. In other words, if both R and S are m x n matrices, then R = S if and only
if r i; = so for all = 1, 2, ... , m and j = 1, 2, ... , n. For example,
1 2 1 2 1 1 1 0
=) = =) but (0
1 1 1
Vectors are matrices with only one row or one column. A 1 x m matrix is called
a row vector, and a m x 1 matrix is called a column vector. For example,
a = (1 —1 1), b = (1 2 0 1)
(1 x 3) (1 x 4)
861
862 MATRIX ALGEBRA
1
(2)
1
(2) (3 1 2) = (3 1 2),
but
/1
(1 2 I 2
(1) (1 2),
2 2 1 2
Addition and subtraction of two matrices A and B can be performed if A and B have
the same dimension—i.e., if the number of rows and the number of columns are the
same. Addition and subtraction is carried out on each corresponding pair of ele-
ments. If A + B = C, then a1 + = cu. For example,
(3
+ (-1 l) (1 + o (0 3
4) 2 —1) + 2 4 — 1) 3
(1 (-1 1) (1 — (-1) (2
3 4) 2 — 1) —2 4 — (-1)) 5)'
(21) 1
(1) O
(1)' (21) 4- (1)
1 2 .
(3)
If we multiply a matrix A by a scalar, the resultant matrix is obtained by multi-
plying each element of A by that scalar. So if
(1 2
A=
—1 1 2}'
then
6
( 2 4 —9
2A = —3A = (-3
—2 2 4)' 3 —3 — 6
THE OPERATIONS OF MATRICES 863
We can also multiply two matrices together provided that the number of col-
umns in the first matrix is equal to the number of rows in the second matrix. To
form the product AB of the two matrices A and B, the number of columns of A must
be equal to the number of rows of B. If we designate the result of the matrix multi-
plication AB by C, then C is again a matrix and C has the same number of rows as
A and the same number of columns as B. To summarize: if A is an m x n matrix
and B a p x q matrix, then the product AB can be formed only if n = p; further, if
C = AB, then C is an m x q matrix.
To complete our definition of matrix multiplication, we have to describe how
the elements of C are obtained. The following rule specifies cu, the element in the
ith row and jth column of the resultant matrix C, in terms of elements in A and B:
n
cii = aii b u + a12 b2J +. + ainbni = aik bki , (B.1)
k= 1
C11 = + a i2 b 21 + = 1 x (--1) + 2 x 2 + 3 x 1 = 6,
then
(2 1) (1 1) .
AB = but BA =
1 1 1
A system of linear equations can be expressed in matrix form. First, let us consider
one simple linear equation, say X= — 2X < + 2é 3 = 4. Using matrix multiplication,
the equation can be expressed as
Ix
(1 —2 2) = 1X = + (-2)X< + 2X3 = 4.
3/
Suppose we now have the following system of three equations:
X, — 2X, + 2X3 = 4,
X, + é, + X3 = 5,
— é 1 5X< — 3X 3 = 1.
Using matrix multiplication, this is equivalent to
—2 2 1 • X, — 2 • é, + 2 • X 3 \ 14\
1 1 X2 1 • é, + 1 • X 2 + 1 • X 3 5
5 =
VX3 ÿ 1 • X 1 + 5 • X 2 ÿ 3 • X 3/ /
3x3 3x1 3x1
And because of the equality definition of vectors, we must equate (1 • X = — 2 • X 2 +
2 • X 3) to 4 and (1 • é, + 1 • X 2 + 1 X 3 ) to 5 and (— 1 • é, + 5 • X 2 ÿ 3 • X 3 ) to 1,
which shows that the matrix formulation
1 ÿ2 2 /é,\ 4
1 1 1 X2 = 5 (B.2)
—1 5 —3 \é 3/ 1)
is equivalent to the system of linear equations. In general, Eq. (B.2) is written as
Ax = b, where
1 -2 2 /X = \ 4
1 1 1 , x= é, , b = (5).
—= 5 —3 \é 31 1
A is called the coefficient matrix, x is the vector of unknowns, and b is the vector of
constants. Finding the solution to a system of linear equations is equivalent to solving
for the unknown vector x in the matrix equation Ax = b. We will come back to
solving Ax = b in a later section.
MATRIX INVERSION DEFINED 865
D. SPECIAL MATRICES
There are several types of matrices that possess useful properties. Here we list some
of the more important ones.
The zero (or null) matrix is a matrix with all elements (or components) being zero.
For example,
(0 01
o 0 0\
0 0 0 0
0 0)'
\ 0 0 0 0/
are null matrices of dimensions 2 x 2 and 3 x 4. The zero matrix 0 possesses the
property that
A+0=0+A= A
for any matrix A of the same dimension.
A diagonal matrix is a square matrix whose elements are all zeros except on the
main diagonal—i.e., D is a diagonal matrix if d i; = 0 for i j. For example,
/1 0 0\
0 3 0
)' \0 0 4/
are diagonal matrices. The elements d11, d„, . . , d nn are called elements on the main
diagonal. Note that all diagonal matrices are square by definition.
The identity matrix, I, is a diagonal matrix that has ones on the main diagonal
and zeros everywhere else. For example,
(1 0 0 o\
/1 0 0\
(1 0 0 1 0 0
0 1 0 ,
1)' 0 0 1 0
0 0 1/
0 0 0 1/
are identity matrices of dimensions 2 x 2, 3 x 3, 4 x 4. The identity matrix has the
useful property that
AI = A, IB B
for all matrices A and B provided the matrix multiplication is defined; i.e., A and B
must be of appropriate dimensions.
Now, given a square matrix A, there may exist a matrix B, such that
AB = BA = I.
If such matrix B exists, then A is said to be nonsingular and the matrix B is called
the multiplicative inverse of A. We usually write B as A' to denote inverse. A -1
866 MATRIX ALGEBRA
F. MATRIX TRANSPOSITION
Before we describe how to compute A -= we must first define the transpose of a matrix.
For a given matrix A, the transpose of A, denoted by A', is obtained from A by writing
the columns of A as rows of A'. Formally, we have alb = aii . For example, if
1 2 3\ 1 3 4
3 2 1 , then A' = 2 2 3
4 3 2/ 3 1 2/
Finally, if A = A', then we say that A is a symmetric matrix. For example,
1 —1 4
A= —1 2 5
( 4 5 3
is a symmetric matrix. Notice that the entries of A are symmetric across the main
diagonal; hence all diagonal matrices are symmetric.
The class of symmetric matrices is very important and arises very often in many
real-life problems. The covariance matrix in portfolio theory is a symmetric matrix.
Furthermore, algebraic systems involving symmetric matrices are in general easier
to solve.
G. DETERMINANTS
Given x = A -=b, and A, the first step in finding A -= is to first determine if A is non-
singular. To do that, we make use of the determinant function that is defined for all
square matrices.
The determinant of a square matrix A, denoted by IA, is a unique number as-
sociated with that matrix. For a 2 x 2 matrix
For example,
2 1
= (2 x 4) — (1 x 3) = 8 — 3 = 5.
3 4
So the determinant of the matrix is 5.
The definition of a determinant of a 3 x 3 or higher-order square matrix involves
the notion of minors and cofactors of elements of the matrix. The minor of au, de-
noted by 111/1,;1, is the determinant of the submatrix of A obtained by deleting the ith
row and jth column of A. Suppose
a12 a13\
A= a21 a22 a23
\a31 a32 a 33/
then
a22 a23
= a22a33 — a23a32 ,
a32 a33
a12 a13
I m211 = = a12a33 — a13 a32 ,
a32 a33
a 32 a 33
+
and so on. The cofactor of au, denoted by ICi1I,is equal to (— 1)' j1M01. That is why
sometimes cofactors are called signed minors. Whenever i + j is even, IC i1i =
and whenever i + j is odd, ICJ = —111/J. Take the 3 x 3 matrix, A,
1 2
0 4
A= 3 0 4 , then 1 M111 = 1 = —4,
5
0 1 5/
3 4 , 3 0
1 M121 = 0 5 15,
i M131 = 0 1 = 3*
The reader may check that 11421 = 9, 11/1221 = 5, 1M231 = 1, 1M311 = 8, 1 M 321 = 1,
M33 = —6. Hence IC111 = —4, IC121 = —15, IC IC
1- 13I = 3 1-21.1 = —9, 1C221 = ,
5
= j=1
E M ijl (B.3a)
Although Eq. (B.3a) and Eq. (B.3b) may look rather complicated at first glance,
they are in fact quite simple. Each term of the sum in Eq. (B.3a) simply consists of an
element in the ith row and its cofactor (signed minor). An example will clarify this
further. Let
/1 2 1
A= 3 0 4 .
\O 1 5
From Eq. (B.3a), taking i = 1, and expanding by the ith row,
1
Al = a ll (-1) 1Mil l + a12( -1)
1+2
0/121 + a„(-1)1+31 M, 31
0 4 3 4 3 0
= 1 • ( — 1)2 • + 2( —1)3 +1 - (-1)4
1 5 0 5 0 1
= —4 + (-30) + 3 = —31.
In the above example l Al was evaluated through expansion by the first row. According
to Eq. (B.3b), we can also evaluate IA l through expansion of the jth column. Let us
take j = 2,
4- -1
lAl = a12( +2
1M121 + a22( — 1)2+2 0 221 + a32( )3+21 M321
3 4 1 1 1 1
= 2( — 1)3 + 0 • (-1)4 + 1( — 1) 5
0 5 0 5 3 4
= — 30 + 0 + (— 1) = —31,
Given a nonsingular square matrix A, construct a new matrix B of the same dimen-
sion with b i; = C u l, the cofactor of cl ip Then transpose B and call the resultant matrix
the adjoint of A, "adj A." That is, adj A = B'. It can be shown that
Since the nonsingularity of A implies 1A1 0 0, we can divide both sides by the scalar
(adj A)(A) = I;
AI
since A -1A = I, it is immediately evident that (1/1A1) (adj A) =
As an example take the 3 x 3 matrix A from the previous section. We have
already computed the determinant as well as all the cofactors, so
(4 9 —8 4 9 —8
31 31 31 31 31 31
/1 2 1 /1 0 0\ 1 2 1\
15 —5 1 15 —5 1
3 0 4 = 0 1 0 = 3 0 4
31 31 31 31 31 31
\0 1 5 \0 0 1/ 0 1 5/
—3 1 6 —3 1 6
31 31 31 31 31 31
870 MATRIX ALGEBRA
= , X2 = X3 = .
31 31 31
As a check on the solution, we insert the values into the original equation system:
—21 44 8
+ + = 1,
31 31 31
—63 32
+ 0+ = —1,
31
22 40
0+ + = 2.
31 31
J. CRAMER'S RULE
A direct but not obvious corollary to our derivation of A -= is Cramer's rule for the
solution of a linear equation. The rule states that
A.1
xi=
-
APPLICATIONS 871
where Ai is the matrix obtained from A by replacing the ith column with the constant
vector. Using the same example and applying Cramer's rule, we first substitute the
constant vector for the first column in the numerator and then expand by the first row.
Recall that the sign changes are the result of converting minors to cofactors.
1 2 1
—1 0 4 0 4 —1 4 —1 0
1• —2• +1•
2 1 5 1 5 2 5 2 1
x1 = 1A1 —31
—4+26-1 21
= .
—31 31
Next we replace the second column of the original numerator by the constant
vector and again expand by the first row:
1 1 1
3 —1 4 —1 4 3 4 3 —1
1 —1• +1•
0 2 5 2 5 0 5 0 2
X2 =
—31 —31
—13-15+6 22
—31 31'
and again for the third column,
1 2 1
3 0 —1 0 —1 3 —1 3 0
1• —2• +1•
0 1 2 1 2 0 2 0 1
X3 —
—31 —31
1 — 12 + 3 8
= .
—31 31
This agrees with the previous result. All the determinants above were evaluated by
expanding by the first row.
K. APPLICATIONS
In this section we present two applications of matrix algebra in the theory of finance.
5
12 = 621 be the covariance of return between X, and X 2, 613 = 631 the covariance
of return between X, and X 3 , 0 23 = 632 the covariance of return between X 2 and X,.
Constructing the covariance matrix A we wish to solve for the weight vector X that
will minimize the variance. Let
5
12
61 5
13 /x i \
A = 21 "2„.2
5 5 X = x2
23 5
(3 x 3) 5
31 532 '3 (3 x 1) \x 3 /
The variance of the portfolio with x, of X }, x 2 of X2 , x 3 of X 3 can be expressed in
matrix form as ap2 = X'AX. To minimize the variance of the portfolio is equivalent
to minimizing 6p2 subject to the weight constraint x, + x 2 + x 3 = 1. This constrained
optimization problem can be solved by the method of the Lagrange multiplier.1 Let
2
g(x}, x2 , x 3 , 2') = 0-1, + 2'(l — x, — x 2 — x 3)
= X'AX — x} — x2 — x 3 ),
ag
=0 (B.4a)
ax,
ag
0 (B.4b)
ax 2
ag
=0 (B.4c)
ax é
ag
02' = O. (B.4d)
The Eqs. (B.4a), (B.4b), and (B.4c) in matrix notation can be expressed as
/1\
AX = 2e, where e = 1 and 2 = 2'/2, (B.5)
11/
and (B.4d) is simply the reiteration of the constraint x, + x 2 + x 3 = 1. The solution
to the matrix Eq. (B.5) will give us the answer in terms of 2, and the constraint
condition will give us the value of 2, hence the complete solution.
1
Readers unfamiliar with the method of solving constrained optimization problems using Lagrange mul-
tipliers should consult Appendix D.
APPLICATIONS 873
/ 2 —1 0\
A = —1 2 —1
0 —1 2/
/ 2 —1 0 /X } \ 1\ A\
—1 2 —1 X 2 A 1 =
\ 0 —1 2 \X 3 / 1/ A/
Using Cramer's rule, substituting the A vector for the first column, and expanding
both numerator and denominator by the first column, we have
2 —1 —1 0 —1 0
A —A +A
—1 2 —1 2 2 —1 32 + 22 + 62 3
X1 = = = A.
2 —1 —1 0 6—2 4 2
2 +1 +0
—1 2 —1 2
A —1 A0
2A +1 +0 62 + 2A 8A
2 A2
= — — 22.
4 4 4
2A —1 A
2 +1 +0
—1 A —1 2 62 + 0 = 62 3
X3 = = 2 A.
4 4 4
2. Linear Regression
Very often, when we consider a security, we like to know how its return varies
as the market fluctuates. Suppose we have the data in Table B.1. Can we discern any
pattern or simple relation between the return on security A and the market? First,
874 MATRIX ALGEBRA
Table B.1
Return (in %)
STATE 1 2 3 4 5 6
RA
12 x
11 x
10
x
9 Rrn
8 9 10 11 12 13
We can see that an approximately linear pattern exists. The following question nat-
urally arises: What is the equation of the straight line that "best" fits the data points?
In mathematical terms, this is equivalent to finding the values of two constants a and
b such that whenever the return on the market is given, then the expression [a +
b x (return on market)] will give the "best overall" predictive value of the return on
security A. In symbols, let X 1, i = 1, 2, 3, 4, 5, 6, be the returns on the market given
the various states of the world; let Yi be the returns on security A; and let Y = a + bX
the predicted return on security A using the best-fitting straight line. We call
ei = Yi — Y , which is the difference between the observed value and the predicted value
of return on security A, the error term. These error terms can be positive or negative.
To find the best-fitting straight line is equivalent to minimizing the magnitude of the
error terms in a certain sense. The technique of minimizing the sum of the square of
the error terms is called linear regression. In our example above with six different
states, we have six error terms. Let
6 6 6
g(a, b) = E E fi)2 = E (Y— a — bX i) 2
= 1 e2 = ti =1 i= 1
be the sum of the squares of the error terms. The first-order conditions for mini-
mizing g(a, b) are
eg
=0 (B.6a)
ea
ag
0 (B.6b)
01)
APPLICATIONS 875
6 X, i Y
i=1 i=i
(b)
i Xi i X7 i Xi Yi
i =1 i=1 i=i
This is equivalent to two equations and two unknowns (a and b). As a numerical
example, let us take the data from Table B.1:
6 Xi = 8 + 9 + 10 + 11 + 12 + 13 = 63,
i=1
SO
6
( 5
63 679) ) 666
Therefore the equation of the best-fitting straight line is Y = 4.57 + 0.56X. Note that
b is the slope of the straight line. Both the sign and the magnitude of b contain
important information. If b is positive, we would expect that the return on security
A moves with the market, whereas a negative b implies that returns on security A
and the market generally move in opposite directions. The magnitude of b measures
the degree of volatility of security A. The larger the magnitude of b, the more volatile
the return on security A.
Appendix C An Introduction
to Multiple Regression
Business students are frequently confronted with journal articles that are riddled with
econometrics. On the other hand, econometrics courses assume prior knowledge of
matrix algebra and calculus and therefore present a formidable barrier to the curious.
This appendix is written to provide an overview of multiple regression techniques
that assumes only the rudimentary knowledge of calculus and matrix algebra provided
in the other appendixes. While not a substitute for a good econometrics course, this
appendix enables the reader to understand and interpret the computer output from
a typical multiple regression software package and to have an introductory level of
understanding of some of the typical errors made in econometric studies.
If we are trying to explain the distribution of sales revenue for the XYZ Company
given a forecast of gross national product, we might choose a linear model like
=a+ + (C.1)
where
= sales revenue in year t,
X, = forecast of gross national product for year t,
Et = error term (the difference between actual sales revenue and that predicted by
the model).
Linear relationships have the virtue that they are simple and robust. Many
natural phenomena are not linearly related, but linear approximations usually work
very well within a limited range.
The object is to find the set of weights (a and b) in Eq. (C.1) that provide the
best unbiased estimate of revenue given GNP. If GNP has any explanatory power
877
878 AN INTRODUCTION TO MULTIPLE REGRESSION
Yt DATA
• 1969 1970 1971 1972 1973 1974 1975 1976
Yt 125 110 220 170 240 225 195 290
300 Xt .86 .98 1.11 1.18 1.30 1.41 1.53 1.71
•
200
• Conditional
at distribution ( Yr Xt)
100 Unconditional
distribution of Yt
b = 174.8
I
Xt
.25 .50 .75 1.00 1.25 1.50 1.75
a = — 23.4
Figure C.1
Revenues vs. forecast GNP.
This result follows from the properties of random variables discussed in Chapter 6.
ORDINARY LEAST SQUARES LINEAR ESTIMATION 879
The intercept is determined by the fact that the line must pass through the mean
values for both Y and é. At that point, we have
= +
The estimated slope and intercept terms are computed in Table C.1.
Having obtained estimates of the slope and intercept that minimize the squared
error terms, we now have the following linear equation
= —23.42 + 174.84X t .
It can be used to predict sales revenue when given a forecast of GNP. The difference
between predicted revenue and actual revenue in any given year is the error of estimate,
often called the residual:
=
Note that in Table C.1 the average error term Et (except for rounding error in the
calculations) is zero. This implies that the linear estimate is unbiased. In other words,
on average, there is no error of estimate.
In linear regression the residual errors indicate the extent of movement in the
dependent variable (the variable on the left-hand side of the regression equation) that
is not explained by the independent variable(s) (the variable(s) on the right-hand
side). If the residuals are small relative to the total movement in the dependent vari-
able, then it follows that a major part of the movement has been explained. We define
<
the summary statistic known as the squared multiple correlation coefficient, r , as
the percentage of the total variation in the dependent variable that is explained by the
regression equation. The square of the correlation coefficient is
Variation explained by the regression equation
r2 -
Total variation of the dependent variable
~ ( Y — y )2 E (8 _ 02
= -
f)2
E (y
Note that E (Y — Y) < is the variance of the dependent variable, in our example.
Note also that E (e — 0< is the variance of the residuals. Furthermore, the average
00
Table C.1 Simple éegression Computations 00
ol, =
1 (ê - n
=
<P,<96.89
<
= 36=3.84 oi = E (é - XY .P668
N _ = 7 = .08097=4
N-1 7
,c -?, = V36=3.84 = 60.=<
ay =-\/7 ( 7 x = , /t]i = .<846
(Y - 1 )(é - X) 99.=0
7-
COV(Y, é) = - =4.=6
N1 7
. COV( Y, é) = =4.=6
î
= =
=74.88
6,<, .08097=4
a = Y - Lx = =96.876 - =74.84(=.<6) = -<3.4<
Symbol Definitions- Y, é = the means of revenue and GNP, respectively,
N = the numîer of oîservations in the sample,
= the variances of revenue and GNP, respectively,
COV(Y, é) = the covariance îetween revenue and GNP,
a, b = the intercept and slope estimates,
e = the error term.
SIMPLE HYPOTHESIS TESTING OF THE LINEAR REGRESSION ESTIMATES 881
error term, E, is always zero; therefore we can rewrite the squared correlation coeffi-
cient as2
2
2 (Y — Y)2 — E
E
r =
(IT
(C.5)
Y)2
Using the numbers from Table C.1,
2 25,296.89 — 7970.5932
r = .6849.
25,296.89
This means that about 68.5% of the variance in the dependent variable, sales revenue,
is explained by the independent variable, GNP. If all the variance were explained,
then the sum of the squared residuals, E e 2, would be zero and we would have r 2 = 1.
At the opposite extreme the equation would not reduce the variance of the dependent
variable at all, in which case we would have r 2 = 0.
Now that we know how to estimate the slope and the intercept, the next logical
question is whether or not we can reject the null hypothesis that they are equal to
2
An important relationship, which is used in Chapter 6, is that
COV(Y, X)
r=
urax
Proof follows from the definition of r 2 , of and of b. First, rewrite r 2 and the definition of b
X)
—2 COV(Y,
= b = !2
or X
[COV(Y, X)] 2
zero. To do this we can calculate t-statistics in order to test the significance of the
slope and intercept terms.
The t-statistics are defined as the estimates of the intercept, a, or the slope, b,
divided by their respective standard errors of estimate:3
a b
ta = , tb — •
(C.6)
se(a) se(b
We shall assume that the independent variable, 2, can be treated as a constant
in repeated samplings. In fact, this is where regression analysis derives its name. We
say that Y is regressed on 2. We also assume that the error terms are generated
by random selection from a stationary statistical distribution with a mean of zero
and a constant variance, Also, the error terms in successive samplings are indepen-
dent. This specification of the error-generating process may be stated as
E(8) = 0, (C.7)
2 2
VAR(e) = E[e — E(8 ] = E(8 = (C.8)
COV(e„ e t _ 1) = 0. (C.9)
To determine the standard error of estimate for b, recall the definition given in
Eq. (C.3):
COV(Y, 2
_2 (C.3)
X
We also know that the observed values of Y are
Y = a + b2 + e.
Rewriting Eq. (C.3), using the definitions of COV(Y, 2 and ( 2,
TX we have
V [(2 — 2 (Y — Y)]
b=
E [(2 — 2 (2 — 2 ]
Substituting in Y yields
_ E [(2 — 2 (a + b2 + e — — bX)]
E [(2 — 2 (2 — 2 ]
E [(X — g)(Y — P) + ( 2 — g)e]
E [(2 — 2 (2 — 2 ]
E [(2 — 2 (Y — Y ] E [(x — 2 e]
-
E [(x — g)(x — g)] E [(x — x (x — g)]
b E [(x — 2 e] (C.10)
E [(x — )02]
3
A good reference to the t-distribution is Hod [1954, 274-283].
SIMPLE HYPOTHESIS TESTING OF THE LINEAR REGRESSION ESTIMATES 883
Equation (C.10) tells us that the estimated slope, b, is equal to the true slope, b,
plus a term that depends on the variance of 2 (in the denominator) and the error
terms (in the numerator). The expected value of b is
E(b = b, since E(e = 0. (C.11)
Note that the expected value of the slope is equal to the true slope. Therefore we
can say that the slope estimate is unbiased. The variance of b is
[(Xi X)Ej
= E[b + 1 12
E ( 2 —— 2 2 "
'
E ]E [(X — 2 )ed 1
-2
[ (x )2 '
and since 2 is assumed to be a constant, we have
VAR(b) = (xi
l
x)2]EE (2 1
2
— 2 ei ]
Expanding the second term yields
EE — 2 8ir = ER2 — 2 ) 2 4 (X2 — X)2 £3 + " '
+ 2(X, — 2 (2 2 — 2 e i e 2 + ••• ]
= (2 , — 2 2 E(4 + (2 2 — 2 2 E(ei + • • -
+ 2(2 — 2 (2 2 — 2 E(e l e 2 + • • .
Using Eqs. (C.8) and (C.9), the above result can be reduced to
EE ( 2 i — 2 g ir = [1 ( 2 — 2 2](72
62
(C.12)
E ( 2 —2 2
We now have the result that the slope estimate, b, is distributed normally with
a mean of b and a variance of 6E /6X. The variance of the estimate of b provides a mea-
sure of the precision of the estimate. The larger the variance of the estimate, the more
widespread the distribution and the smaller the precision of the estimate.
A similar derivation would show that the intercept estimate, a, is also normally
distributed with a mean of
E(d = a (C.13)
884 AN INTRODUCTION TO MULTIPLE REGRESSION
and a variance of
X2)/6l
VAR(a) = se(a) = AR(d), (C.14)
N }(2 — X) 2
where N is the number of observations in the sample.
Using Eqs. (C.12) and (C.14) for the sample problem of Table C.1, and given
that the t-statistics defined in Eq. (C.6) have N — m degrees of freedom (where N =
8 = the number of observations and m = 2 = the number of independent variables
including the constant term), we can compute the appropriate significance tests for
the slope and intercept. The standard error for the slope term is
o-,2 7970.5932/(8 — 2)
se(b) = = 48.41,
(X — X)2 5668
and the t-statistic for b is
174.84
1
t(b = = 3.61.
se(b) 48.41
We refer to the table of t-statistics (Table C.2) for 8 — 2 = 6 degrees of freedom and
a 95% confidence interval (in a two-tail t-test). That table shows that the t-statistic
must be greater than 2.447 in order to reject the null hypothesis that the slope coeffi-
cient is not significantly different from zero. It is. Therefore we can say that predicted
GNP, the independent variable, is a significant explanatory variable for sales revenue,
given our sample data.
Next, compute the t-test to determine whether or not the intercept estimate, a, is
significantly different from zero. The standard error of a is
[ x2)01 i1/2
se( a) =
N E (X — X)2
[(13.2676)(7970.5932/6)r2
= 62.35
8(.5668)
and the t-statistic is
a - 23.42
t(a = = = .375.
se(a) 62.35
The t-statistic for a is less than 2.447, the required level for significance. Therefore
we cannot conclude that the intercept term is significantly different from zero.
Summarizing, up to this point we can write the results of the regression analysis
as follows:
Y, = —23.42 + 174.84X, r 2 = .6849
( — .38) (3.61) df = 6.
The numbers in parentheses are the appropriate t-statistics, and df designates the
degrees of freedom.
SIMPLE HYPOTHESIS TESTING OF THE LINEAR REGRESSION ESTIMATES 885
2. Sources of Bias
A. LEFT-OUT VARIABLES. One of the most frequently encountered problems of
regression analysis is that the empirical model is not founded on a sound theoretical
footing. When this happens we say that the model is misspecified. If an important
explanatory variable is left out of the regression equation, then the estimates of the
coefficients for the variables included in the equations can be biased. This was one
of the empirical difficulties in the early attempt to test for relationships between
capital structure and value (see Chapter 14). The empirical work was done before a
theoretical model of value had been derived. Therefore relevant variables were often
left out and the empirical results were biased.
Suppose that the true theoretical relationship is
= a + b i X„ + b 2 X 2, ± Et (C.17)
but that the researcher mistakenly estimates the following regression equation:
Y, = a + b1 Xit + U t . (C.18)
BIAS AND EFFICIENCY 887
E ( 2 1 — 2 1 (Y — Y)
1
(C.19)
E (x, — x 1 (x, — go .
By substituting the true relation [Eq. (C.17)] for Y into Eq. (C.19), we obtain
E ( 2,— 2 1 (a+b 1 2 1 + b 22 + 3 — a — b i 2 — b 22 2
b1
E (x, — x,)(x, — x i
E (x, — 2 , (a b,2 1 — a — b,2 1
E (x, — x i )2
E (x, x i )(b 2 x 2 b 2 x 2 +E (x, —
E (X 1 — x 1)2 (x, x))2
and because the error terms are assumed to follow Eqs. (C.7), (C.8), and (C.9), we
have
E (x, — x,)(x 2 — x2)
bl • (C.20)
= b1 + b2
E (x, — x i )2
Equation (C.20) shows that when a relevant variable is left out of the equation
specification the slope estimate, b, is biased. The direction of the bias depends on the
sign of b 2 (the relationship between Y and X 2 ) and on E (x, — x„)(x 2 — x 2) (the
relationship between the independent variables, X 1 and X 2). If X 1, X 2 , and Y are all
positively related, then b will be biased upward. In general the only way to eliminate
misspecification bias is to be sure that the empirical test is appropriately founded on
sound theory, rather than going on an ad hoc "fishing trip."
Dividends can be measured without any error whatsoever, but retained earnings (the
difference between accounting earnings and dividends paid) is only an estimate of
true economic retained earnings on which value is based. Thus retained earnings
possesses a great deal of measurement error. Consequently, the estimate of the effect
of retained earnings on the price per share was biased downward. This led earlier
researchers to incorrectly conclude that dividends had a greater effect on price per
share than retained earnings.
To demonstrate the effect of measurement error, suppose that both the inde-
pendent and dependent variables have sampling error. This may be written as
X = x + w, (C.21)
Y = y + v, (C.22)
where X and Y indicate the observations, x and y are the true values, and w and v
are the measurement errors. Suppose, further, that the true variables have the fol-
lowing relationship:
y = a + bx. (C.23)
We would like to have unbiased estimates of a and b.
Substituting Eqs. (C.21) and (C.22) into Eq. (C.23) gives
Y = a + bX + z,
where
z = v — bw.
From Eq. (C.3), the estimate of b is
E (X — X)(Y — Y)
b
E (X — X)(X — X)
E (x + w — )7 — vi,)(y + v — —
E (x + w — — 1,T-)2
E (x — 50(y — )7) + E (x + )7)(v — /7) + E (y — 37)(w — CC) + E (w — 113)(v — i7)
E (x — )7)2 + 2 E (x — )7)(w — cv-) + E (w — 1;02
Given that the measurement errors, w and v, are distributed independently of each
other and of the true parameters, then the last three terms in the numerator and the
middle term in the denominator vanish as the sample size becomes large. Therefore
the limiting value of b is
E (x — 5c)(y — )7)
E (x — )7)2 + E (w — iT))2
=
• (C.24)
1 + [E (w — fv-)2 /E (x — k)2]
BIAS AND EFFICIENCY 889
Equation (C.24) shows that even if the errors of measurement are assumed to be
mutually independent, independent of the true values, and have constant variance,
the estimate, b, will be biased downward. The greater the measurement error, the
greater the downward bias.
There are two generally accepted techniques for overcoming the problem of errors
in variables: (1) grouping and (2) instrumental variables. Grouping procedures can
reduce measurement error because when grouped the errors of individual observa-
tions tend to be cancelled out by their mutual independence. Hence there is less mea-
surement error in a group average than there would be if sample data were not
grouped. An instrumental variable is one that is highly correlated with the indepen-
dent variable but that is independent of the errors w and v [in Eqs. (C.21) and (C.22)].
This was the technique employed by Friend and Puckett in testing dividend policy.
Instead of using the accounting measurement of earnings, they used normalized
earnings (a time series estimate of predicted earnings) to eliminate most of the mea-
surement error bias.
3. Loss of Efficiency
where St is the sales revenue of a ski shop, R, and 4 are the sales of left and right
downhill skis, and 0, is the sales of other items. The estimated coefficient b 1 is sup-
posed to measure the impact of the sale of the right skis, holding all other variables
constant. Of course this is nonsense, since right and left skis are sold simultaneously.
The usual multicollinearity problem occurs when two independent variables are
highly, but not perfectly, correlated. And usually the effect is to reduce the efficiency
of estimates of b 1 and b 2 by increasing the standard error of estimate. The best rem-
edy for the problem is larger sample sizes.
but that the error term follows a first-order autoregressive scheme such as
at = Ket— +
890 AN INTRODUCTION TO MULTIPLE REGRESSION
= K(Ke,_ 2 U, _ 1 + U,
= U, + KU, 1 + U t _ 2 + • • • + K nUt,
= KtU
t=o
Equation (C.25) shows that the closer the relationship between et and et _ j , the closer
K is to unity and the greater will be the estimated error term and the loss of efficiency.
We can test for serial correlation by using the Durbin-Watson d-statistic. If e,
are the residuals from a fitted least squares equation, then d is defined as
n
2
(Et - Et_i)
t=i
Durbin and Watson have tabulated upper and lower bounds d u and d, for various
numbers of observations, n, and numbers of independent variables, K.
When the error terms are serially independent, the d-statistic has a theoretical
distribution with a mean of 2, but sampling fluctuations may lead to a different
estimate even when the errors are not autocorrelated. Table C.3 provides critical
values for the d-statistic. If the computed d is smaller than the lower critical value,
d 1 , or above the critical value (4 — d 1 , then the null hypothesis of serial independence
is rejected. When the statistic is larger than d u but smaller than (4 — d„ , then the
null hypothesis is accepted. When neither of these two cases is true, then the test is
inconclusive.
BIAS AND EFFICIENCY 89=
n d1 du d, du d, d d, du d, du
=P 0.9P =.<3 0.83 =.40 0.7= =.6= 0.P9 =.84 0.48 <.09
=6 0.98 =.<4 0.86 =.40 0.7P =.P9 0.64 =.80 0.P3 <.03
=7 =.0= =.<P 0.90 =.40 0.79 =.P8 0.68 =.77 0.P7 =.98
=8 =.03 =.<6 0.93 =.40 0.8< =.P6 0.7< =.74 0.6< =.93
=9 =.06 =.<8 0.96 =.4= 0.86 =.PP 0.76 =.7< 0.66 =.90
<0 =.08 =.<8 0.99 =.4= 0.89 =.PP 0.79 =.70 0.70 =.87
<= =.=0 =.30 =.0= =.4= 0.9< =.P4 0.83 =.69 0.73 =.84
<< =.=< =.3= =.04 =.4< 0.9P =.P4 0.86 =.68 0.77 =.8<
<3 =.=4 =.3< =.06 =.4< 0.97 =.P4 0.89 =.67 0.80 =.80
<4 =.=6 =.33 =.08 =.43 =.00 =.P4 0.9= =.66 0.83 =.79
<P =.=8 =.34 =.=0 =.43 =.0< =.P4 0.94 =.6P 0.86 =.77
<6 =.=9 =.3P =.=< =.44 =.04 =.P4 0.96 =.6P 0.88 =.76
<7 =.<= =.36 =.=3 =.44 =.06 =.P4 0.99 =.64 0.9= =.7P
<8 =.<< =.37 =.=P =.4P =.08 =.P4 =.0= =.64 0.93 =.74
<9 =.<4 =.38 =.=7 =.4P =.=0 =.P4 =.03 =.63 0.96 =.73
30 =.<P =.38 =.=8 =.46 =.=< =.P4 =.0P =.63 0.98 =.73
3= =.<6 =.39 =.<0 =.47 =.=3 =.PP =.07 =.63 =.00 =.7<
3< =.<7 =.40 =.<= =.47 =.=P =.PP =.08 =.63 =.0< =.7=
33 =.<8 =.4= =.<< =.48 =.=6 =.PP =.=0 =.63 =.04 =.7=
34 =.<9 =.4= =.<4 =.48 =.=7 =.PP =.=< =.63 =.06 =.70
3P =.30 =.4< =.<P =.48 =.=9 =.PP =.=3 =.63 =.07 =.70
36 =.3= =.43 =.<6 =.49 =.<0 =.P6 =.=P =.63 =.09 =.70
37 =.3< =.43 =.<7 =.49 =.<= =.P6 =.=6 =.6< =.=0 =.70
38 =.33 =.44 =.<8 =.P0 =.<3 =.P6 =.=7 =.6< =.=< =.70
39 =.34 =.44 =.<9 =.P0 =.<4 =.P6 =.=9 =.63 =.=3 =.69
40 =.3P =.4P =.30 =.P= =.<P =.P7 =.<0 =.63 =.=P =.69
4P =.39 =.48 =.34 =.P3 =.30 =.P8 =.<P =.63 =.<= =.69
P0 =.4< =.P0 =.38 =.P4 =.34 =.P9 =.30 =.64 =.<6 =.69
PP =.4P =.P< =.4= =.P6 =.37 =.60 =.33 =.64 =.30 =.69
60 =.47 =.P4 =.44 =.P7 =.40 =.6= =.37 =.6P =.33 =.69
6P =.49 =.PP =.46 =.P9 =.43 =.6< =.40 =.66 =.36 =.69
70 =.P= =.P7 =.48 =.60 =.4P =.63 =.4< =.66 =.39 =.70
7P =.P3 =.P8 =.P0 =.6= =.47 =.64 =.4P =.67 =.4< =.70
80 =.P4 =.P9 =.P< =.6< =.49 =.6P =.47 =.67 =.44 =.70
8P =.P6 =.60 =.P3 =.63 =.P= =.6P =.49 =.68 =.46 =.7=
90 =.P7 =.6= =.PP =.64 =.P3 =.66 =.P0 =.69 =.48 =.7=
=.P8 =.6< =.P6 =.6P =.P4 =.67 =.P< =.69 =.P0 =.7=
9P
=.P9 =.63 =.P7 =.6P =.PP =.67 =.P3 =.70 =.P= =.7<
=00
892 AN INTRODUCTION TO MULTIPLE REGRESSION
For the set of sample data in Table C.1, the estimated d-statistic is computed
below:
From Table C.3 the critical values for the Durbin-Watson test are d1 = .95 and
d u = 1.23. Since our computed value is neither below d1 = .95 nor above the critical
value of (4 - d1) = 3.05, the null hypothesis of serial independence cannot be rejected.
However, because d = 2.88 is greater than du = 1.23 but not smaller than (4 - du) =
2.77, we cannot accept the null hypothesis. Because serial correlation cannot be either
accepted or rejected, the test is inconclusive in this case.
SUMMARY
This has been an extremely brief overview of linear regression analysis. We have shown
how to estimate the slope, the intercept, the standard errors of each, their t-statistics,
and the correlation coefficient for a two-variable case. Multivariate estimates of the
same variables in a multiple regression equation have the same interpretations and
are provided by many different computer software packages. The summary statistics
for the example problem of Table C.1 would appear in a computer printout in some-
thing like the following form:
Y, = - 23.42 + 174.84X, r 2 = .6849 d = 2.88
(-.38) (3.61) df = 6
We can infer that sales revenue is significantly related to predicted GNP, with an in-
tercept term insignificantly different from zero and a significant slope term. Because
the Durbin-Watson test is inconclusive we cannot be sure whether or not serial cor-
relation has reduced the efficiency of our estimates. Furthermore, additional testing
would be necessary to determine whether or not left-out variables have caused a biased
estimate of b1.
REFERENCES 893
The mean square error criterion is one way of trading off bias and loss of effi-
ciency. It may be desirable, e.g., to accept a small bias in order to gain much greater
efficiency. Although we have not discussed all the causes of bias or inefficiency, a few
of the more important ones were covered. The interested reader should refer to an
econometrics text for a more rigorous and detailed presentation.
REFERENCES
Christ, C. G., Econometric Models and Methods. Wiley, New York, 1966.
Dhrymes, P. J., Econometrics: Statistical Foundations and Applications. Harper & Row, New
York, 1970.
Goldberger, A. S., Econometric Theory. Wiley, New York, 1964.
Hoel, P. G., Introduction to Mathematical Statistics, 3rd ed. Wiley, New York, 1954.
Johnston, J., Econometric Methods. McGraw-Hill, New York, 1963.
Rao, P., and L. Miller, Applied Econometrics. Wadsworth, Belmont, Calif., 1971.
Wonnacott, R. J., and T. H. Wonnacott, Econometrics. Wiley, New York, 1970.
Appendix D Calculus
and Optimization
A. FUNCTIONS
A fundamental notion used in finance is the concept of a function. There are three
ways to express functions: as (1) mathematical equations, (2) graphs, and (3) tables.
Example: Suppose a variable Y is related to a variable X by the following math-
ematical equation:
Y = 2X 2 — 3X + 6.
A shorthand way of expressing this relationship is to write Y = f(X), which is read
"Y is a function of the variable X" and where Y is the range and X is the domain of
the function. X is also called the independent variable and Y the dependent variable,
since Y's value [f(X)] is posited to depend on X's value.
We can also express the function in a tabular and graphical manner. Thus the
equation enables us to construct a range of Y values for a given table of X values.
The data in the table can then be plotted in a graph as in Fig. D.1.
Definition: The dimension of a function is determined by the number of indepen-
dent variables in the domain of the function.
Example: Y = f (X, Z) is a two-dimensional function
Y = f(X 1, X2 , . . . , Xn) is an n-dimensional function.
894
FUNCTIONS 895
I
•
Figure D.1
Example: From basic capital budgeting concepts (see Chapter 2), we know that the
net present value (NPV) of an investment project is equal to
N
NCF /0,
NPV = E
t=i (1 +
where
NCF, = net cash flow in time period t,
I0 = the project's initial cash outlay,
k = the firm's cost of capital,
N = the number of years in the project.
We can express this relationship functionally as
NPV = f(NCF„ k, N) t = 1, . . . , N.
Given values for the right-hand side independent variables we can determine the
left-hand dependent variable, NPV. The functional relationship tells us that for every
X that is in the domain of the function a unique value of Y can be determined.
1. Inverse Functions
How about going the other way? Given Y, can we determine X? Yes, we can.
Definition: The function that expresses the variable X in terms of the variable Y
is called the inverse function and is denoted X = f '(Y).
Example:
Y f(X) = 2X — 5.
896 CALCULUS AND OPTIMIZATION
Y = <X-P é = <.P + .P Y
Y é é Y
= 3 3 =
3 4 4 ► 4 3
P P Note that each é is associated P P
7 6 with the same unique Y. 6 7
9 7 • ► 7 9
== 8 8 ==
=< 12r-
=0 =0 —
8—
6—
4—
<—
I I I I I I I I I ►y
= < 3 4 P 6 7 8 9 =0 0 = < 3 4 P 6 7 8 9 =0
(a) (î)
Figure D.2
2. Linear Functions
An important type of function consists of linear functions of the form
Y = a i X, + a 2 X 2 + • • • + a„Xn .
These functions are used in regression and in the CAPM. In two dimensions a linear
function is a straight line, usually written as Y = a + bX, where a is the intercept
on the Y axis and b is the slope of the line:
Y1 — Y< AY
slope =
X, — X< AX
change in Y rise
change in X run
The CAPM is of the form Y = a + bX, where E(R J) = R f + Mi . This equation
plots like the relationship in Fig. D.3(a), as we see in Fig. D.3(b).
FUNCTIONS 897
Figure D.3
The slope of the security market line in Fig. D.3(b) is [E(R m ) — R f ], which is the
market risk premium, A..
The slope of a function is an important concept: it tells us the change in Y per
unit change in X. The various types of slopes are pictured in Fig. D.4.
Example: Straight-line depreciation is a simple linear function:
BV = c — ( — )X,
c
where
BV = book value of the asset,
c = original cost of the asset,
N = estimated economic life of the asset,
X = number of years that have elapsed,
so that the book value after two years is BV = c — (c/N)2.
positive or negative or infinite or
increasing slope decreasing slope zero slope undefined slope
Y Y Y Y
• • •
b positive
b=0
b> 0
• X ►X •X
(a) (b) (c) (d)
Figure D.4
898 CALCULUS AND OPTIMIZATION
3. Exponential Functions
As their names suggest, exponential functions are those in which the independent
variable é appears as an exponent. They are useful for describing growth and com-
pound interest. More formally:
(a) (b)
Figure D.5
FUNCTIONS 899
20 10
15 = m(1 + r)n
10
n
I 2 3 4 5 0 2 3 4 5
(a) (b)
Figure D.6
900 CALCULUS AND OPTIMIZATION
4. Logarithmic Functions
Definition: If N = b', where both n > 0, b > 0, then we define r = log, N, which
is read "r is the log to the base b of N."
In other words, log, N is the number to which b has to be raised exponentially in
order to equal N. So a log is simply an exponent.
Examples:
100 = 102
so log10 100 = 2
1 _ -1
2—
so loge i = —1.
The two most widely used bases for logarithms are base 10 and base e, where e is
an irrational number equal to 2.7182818....
Definition: The logarithm to the base 10 of N is called the "common logarithm
of N." It is usually designated log N.
Definition: The logarithm to the base e of N is called the "natural logarithm of
N." To differentiate it from the common log, the natural log is usually designated:
ln N.
Definition: The function Y = log, X is called a logarithmic function.
Y
Since by definition Y = log, X if and only if X = b , we see that the exponential and
logarithmic functions are inverse functions of each other:
X = 10Y <=> Y = log io X
X = eY Y = ln X.
The logarithmic function Y = ln X is graphed in Fig. D.7.
10 15 20
2.7182818. .
Figure D.7
DIFFERENTIAL CALCULUS 901
Y Figure D.8
SO Y = log1/2 X.
Since logarithms are simply exponents, the rules of logs simply mirror the rules of
exponents:
Exponents Logarithms
B. DIFFERENTIAL CALCULUS
1. Limits
The central idea in calculus is the concept of the limit of a function. Often we
want to know how the values of a function, f(X), behave as the independent variable
902 CALCULUS AND OPTIMIZATION
X approaches some particular point, a. If as X —> a (read "X approaches a"), f(X)
approaches some number L, then we say that the limit of f(X) as X approaches a
is L. This is written more compactly as
lim f(X) = L.
Intuitively, the existence of a limit L means that the function of f(X) will take on a
value as close to L as one may desire, given that the independent variable takes a
value that is sufficiently close to a.
Example: Many times we are interested in just what happens to a function as X
increases without bound, i.e., when X —> co (read "X approaches infinity"). For in-
stance, what is lim as X —> co of [(X + 1)/X]? The way to evaluate this limit is to
observe the behavior of f(X) as X gets larger and larger. From the table and the
graph in Fig. D.9 we see that f(X) approaches 1 as X —> co, so we can write lim as
X —> co of [(X + 1)/X] = 1.
Intuitively, as X gets very, very large, the fact that the numerator is greater by
one than the denominator does not matter "much," so we have 2m2 = 1.
Example: As we will see next, we are often interested in what happens to f(X) as
X gets very, very small i.e., when X —> 0. For instance, what is the lim as X —> 0
of (3X/X2)? Again, to evaluate this limit we see what happens to f(X) as X —> 0
(Fig. D.10).
Example: Generally we assume that compounding and discounting occur discretely
in annual periods. If the compounding is more than once a year the compound value
interest factor is changed from (1 + r)" to [(1 + r/m)]"m, where m is the number of
times per year compounding occurs. We can now see with limits what is the relation-
ship between the continuous compounding rate and the discrete compounding rate.
Continuous compounding means
rr
lim [(1 + — i= ern by definition of e.
m-00 m
Y
2 + }
Y = f(2 —
2
1 2
10 1.1 f(2
2 X1
100 1.01
1,000 1.001
1,000,000 1.000001
Figure D.9
DIFFERENTIAL CALCULUS 903
3X
X Y= f (X)--- —
X2
1 3
.05 60
.01 300
.001 3,000
.0001 30,000
3
3X
So we see that Ern —< = co
x—o X 0
Figure D.10
2. Derivatives
The rate of change of a function (the change in Y per unit change in X) is an
important concept in mathematics. It is referred to as the derivative of Y with respect
to X. In finance and economics the rate of change is called "marginal." For example,
the marginal cost of capital (MCC) is the rate of change of the total cost of capital
per change in new capital raised. Analytically, the marginal quantities are simply the
slopes of the total quantities.
The derivative is usually denoted as dY/dX, or f'(X). The advantage of the f'(X)
notation is that it reminds us that the derivative is itself a function: the value of the
derivative depends on where it is evaluated. Fortunately there are special rules of
differentiation that can be used to guide calculations.
3. Rules of Differentiation
1. f(X) = c (c is a constant), f'(X) = 0. This rule states that the slope of a horizontal
line is zero, since by definition Y does not change when X changes.
1
2. f(X) = X", f'(X) = nXn . In order to differentiate X", reduce the exponent by
one and multiply by n.
904 CALCULUS AND OPTIMIZATION
é. f(2 = g(2 • h(2 , f](2 = g](2 • h(2 + h](2 • g(2 . The derivative of g(2 • h(2
equals h(2 times the derivative of g(2 plus g(2 times the derivative of h(2 .
4. f(2 = g(2 mh(2 , [h(2 0], f](2 = [g](2 h(2 — g(2 h](2 ]m[h(2 ] 2.
!. f(2 = 3 • g(2 , [3 constant], f](2 = 3 • g](2 .
6. f(2 = g(2 + h(2 , f](2 = g](2 + h](2 .
7. f(2 = In 2, f](2 = 1/X.
8. f(2 = eg (x , f](2 = g](2 eg (x .
<. f(2 = 2, f](2 = 1.
}0. f(2 = a x , f](2 = a x • In a.
}}. f(2 = logs 2, f](2 = 1/(X In b .
}2. f(2 = log[g(X)], f](2 = g](2 mg(2 .
Examples:
} Y = 62 é — é2 2 + 42 + 7,
dY
= 6• (r 3• (22 + 4 • (2 + (7
d2 d2 d2 d2 d2
2) Y = 2 2( 2 + 3),
dY d
( 2 2 }(2 + 3) + [ ( 2 + 3)1X2
d2 = _d2 d2
2 2
= 22(2 + 3) + (1)X = 3X + 6X.
3) Y=
dY = 4x-! = x!4
d2
(2X2 + 6)
4) Y =
X3
2 4 2
dY 42(2 é — (22 2 + 6)(3X ) —2X — 18X
d2 = (2é 2
x6
2
5) Y= = 2X -1/ 2
-N/Y
dY } —} —}
=2. X -312 = X -312 = 3/2
d2 2 x ,/x3
DIFFERENTIAL CALCULUS 905
4. Chain Rule
An extremely useful and powerful tool in differential calculus is the chain rule, or
the function of a function rule. Suppose Y is a function of a variable Z:
Y = f(Z),
but Z is in turn a function of another variable X:
Z = g(X).
Because Y depends on Z, and Z in turn depends on X, Y is also a function of X.
We can express this fact by writing Y as a composite function (i.e., a function of a
function) of X: Y = f[g(X)].
To determine the change in Y from a change in X, the chain rule says:
dY dY dZ
dX dZ dX = f'(Z) • g'(X) Chain Rule
Intuitively the chain rule says, "Take the derivative of the outside (function) and
multiply it by the derivative of the inside (function)." The reason behind the name
"chain" rule is that there is a chain reaction relationship between X, Z, and Y:
vial,
AX =AZ AY.
In words, a change in X has an ultimate impact on Y by causing a change in Z via
function g, and this change in Z will in turn cause a change in Y by function f.
There is a temptation to look at the chain rule by canceling the intermediate dZ
term:
dY _ Y - dZ dY
= = .
dX dZ dX dX
=
3 09
— 0 9.
i3 1
The usefulness of the chain rule can best be seen by considering some examples in
which it is used.
Examples:
Suppose we wanted to differentiate
Y
(3 + 6x-2)1 o .
= 0 + 12é,
dZ = 12é.
dé
Likewise, if we let
ê = (Z)1° ,
d _,
(=/5é + 7) = (5é + 7)113
dé dé
=
= - (5é + 7) -213 P
3
P =
3
WPX + 7)2
P
3(PX + 7)</3
d (e 3x -4) = e 3X-4 3
dé
3e3X-4.
DIFFERENTIAL CALCULUS 907
5. Higher-Order Derivatives
In our development of derivatives we have emphasized that the derivative of a
function is also a function. That is, the value of the derivative depends on the point
X at which it is being evaluated. Like f(X), f'(X) is also a function of X.
Example: Consider the function
f(X) = — 10X 2 + 2400X — 8500, then
f'(X) = — 20X + 2400.
The value of this derivative depends on the point at which it is being evaluated:
f'(120) = — 20(120) + 2400 = 0,
[(60) = — 20(60) + 2400 = 1200.
Because it is also a function of X, we can take the derivative of f'(X). This new func-
tion, f"(X), is called the second derivative of the original function, f(X). The third
derivative is the derivative of the second derivative and is written f"'(X). In principle
we can go on forever and form derivatives of as high order as we like. Notationally
these higher-order derivatives are symbolized in the same manner as the second
derivative:
f'"(X), f (4)(X), f (5)(X), , fm(X),
d 4Y d 5 Y d'Y
dX ' dX 4' dX 5 '
3
Example:
Y = f(X) = — 7 X 2 + 6X — 5,
f'(X) = 3X 2 — 14X + 6,
f"(X) = 6X — 14,
f"'(X) = 6,
f (4)(X) = 0,
[ 5)(X) = 0,
f (n)(X) = 0.
As we shall see, higher-order derivatives play an important role in Taylor and Mac-
Laurin series (see section D of this appendix). The most important of the higher-order
derivatives is the second derivative. Understanding the meaning of the second
derivative is crucial. We know that the first derivative of a function, f'(X), is the slope
of a function or the rate of change of Y as a result of a change in X. The second
derivative, f"(X), is the rate of change of the slope of f(X); that is, it is the rate of
908 CALCULUS AND OPTIMIZATION
Table D.1
f '(X) f "(X) f (X) is
a) >0 >0 Increasing at an increasing rate
b) >0 <0 Increasing at a decreasing rate
c) <0 <0 Decreasing at an increasing rate
d) <0 >0 Decreasing at a decreasing rate
change of the rate of change of the original function, f(X). Table D.1 and Fig. D.11
show various combinations of signs for f'(X) and f"(X) and the implied shape of
the graph of f(X).
Example: In developing the theory of investor choice under uncertainty, cardinal
utility functions [U(W)] are used. These utility functions should have the following
f(X) f(X)
(a) (b)
f(X)
(e ) (d)
Figure D.11
DIFFERENTIAL CALCULUS 909
property: U'(W) > 0, U"(W) < 0. That is, they should look like Fig. D.11(b). Check
to see if and when the following four functions have this property.
1) U(W) = aw — bw2 (quadratic utility function),
U'(W) = a — 2bw > 0 when a > 2bw,
U"(W) = —2b < 0 when b > 0.
2) U(W) = In W (logarithmic utility function),
1
U'(W) = — > 0 W always > 0 by definition of log function,
w
-1
U "( W) = w2 < 0.
TR
MR, p
100
MR = 100 —20q —
m = slope = —20—
Figure D.12
So marginal revenue declines at a constant rate of —20 per unit increase in quantity
sold. Graphically the relationship between total, average, and marginal revenue is
shown in Fig. D.12.
Definition: An important class of functions are those functions whose first deriv-
ative is positive for all values of the independent variable. Such functions are
called monotonically increasing functions. Likewise, functions whose first deriv-
ative is negative for all values of the independent variable are monotonically
decreasing.
6. Differentials
Let Y = f(X); then the differential, dY, is defined as
dY = f/(X)dX.
If we regard dX = AX, a small increment in the independent variable X, then we
can see that dY is an approximation to AY induced by AX because f'(X) = lim as
AX —> 0 of (A Y/AX).
OPTIMIZATION 911
7. Partial Differentiation
So far we have only considered differentiation of functions of one independent
variable. In practice, functions of two or more independent variables do arise quite
frequently. Since each independent variable influences the function differently, when
we consider the instantaneous rate of change of the function, we have to isolate the
effect of each of the independent variables. Let W = f(X, Y, Z). When we consider
how W changes as X changes, we want to hold the variables Y and Z constant. This
gives rise to the concept of partial differentiation. Note that only the variable X is
changing, while both Y and Z remain constant. The rules for partial differentiation
and ordinary differentiation are exactly the same except that when we are taking
partial derivative of one independent variable, we regard all other independent vari-
ables as constants.
Examples:
1) W = XY YZ XZ,
OW
OX =Y+ 0 +Z=Y+Z,
OW
=X-I-Z+ 0 =X+Z,
OY
OW
OZ=O+Y+X=Y+X.
2) W = YZ3 + e x + ln YZ,
OW 3 x
= 2X YZ + e ,
OX
OW =
= X2Z 3 + YZ • Z = X 2 Z3 +
Y♦
0W 1 1
= 3X2 YZ 2 + Y • Y = 3X 2 YZ 2 + Z.
0Z Z
C. OPTIMIZATION
A company seeks to maximize its profit. A production unit seeks to minimize its cost
for a given level of output. An individual investor seeks to maximize his or her utility
when choosing among investment alternatives. Indeed, we are all engaged in big and
small optimization problems every day at work or at leisure. If we have a mathe-
matical objective function, then we can solve our optimization problem using calculus.
912 CALCULUS AND OPTIMIZATION
Figure D.13
To accomplish step 1, let us first consider the graph of a function f(2 that
appears in Fig. D.13. At the point 2 = a, the function f(2 is said to have a relative
maximum because f(a > f(Z for all Z sufficiently close to a. Similarly, f(2 has
a relative maximum at 2 = 3, and f(2 has relative minima at 2 = b and 2 = d. One
common characteristic those four points share is the slope of f(2 at those points.
If we draw tangent lines to f(2 at 2 = a, b, 3, d, then all the tangent lines must be
perfectly horizontal. In other words the slopes f](a = f](b = f](3 = f](d = 0. Thus
we have the following theorem:
Theorem: If f(2 has a relative maximum or minimum at 2 = a, then f](a = 0.
Note that the theorem does not say that if f'(a = 0, then 2 = a is a relative maximum
or minimum. It says that if f](a = 0, then 2 = a is a 3andidate for relative maximum
or minimum. There exist points for which the derivative of f(2 is zero, but the points
are neither relative maxima nor minima. Nevertheless, to locate all relative maxima
and minima, we differentiate f(2 , set the result to zero, and solve for 2. That is,
find all the solutions to the equation
f](2 = 0.
The above equation is called the first-order condition. The solutions are candidates
for relative maxima and minima. To determine which of these solutions are indeed
relative maxima or minima, we need the so-called second-order conditions. Consider
the relative maximum shown in Fig. D.14(a). The slope, f](2 , is zero at the top, posi-
Figure D.14
(a) (b)
OPTIMIZATION 913
tive to the left of the top, and negative to the right of the top. Therefore as 2 increases
from left to right, the slope, f](2 , is decreasing from positive to zero to negative.
We know from the previous section that if f](2 is decreasing, then the derivative of
f](2 , f"(2 , is negative. The condition f"(2 < 0 is called the second-order condition
for relative maxima. Similar reasoning would indicate that at a relative minimum,
f"(2 > 0. We can now summarize step one: Find all the 2 such that f](2 = 0; then
for each of those 2, if f"(2 > 0, it is a relative minimum; if f"(2 < 0, it is a rela-
tive maximum; if f"(2 = 0, we cannot tell (and have to use more sophisticated
techniques.)
Step two requires us to compare function value at the relative maxima and
minima and the boundary points to determine the global optimum. Boundary points
exist because we generally wish to optimize f(2 in some interval, say a < 2 < b,
then a and b are boundary points. Sometimes the global maximum or minimum
occurs at the boundary (see Fig. D.15). That is why we want to evaluate f(2 at the
boundary.
Example: A monopolist faces a downward-sloping demand curve given by p(2 =
100 — 22, where 2 is the quantity and p(2 is the price at that quantity. Suppose
the fixed cost of production is 10 and variable cost is constant at 8 per unit. How many
units should the monopolist produce to maximize profit?
Profit = total revenue — total cost
= price x quantity — (total variable cost + fixed cost),
x(2 = (100 — 22 2 — (82 + 10)
= 100X — 2X2 — 82 — 10
= 92X — 22 2 — 10,
mm(2 = 92 — 4X = 0 (first-order condition),
SO
92 — 4X or 2 = 23,
7E"(2 = — 4;
Figure D.15
914 CALCULUS AND OPTIMIZATION
hence
7r"(X) < 0 (second-order condition).
Therefore X = 23 is a relative maximum. Implicit in this problem is the boundary
X > 0. So X = 0 is a boundary. Obviously n(0) = 0 because this is the decision of not
getting into the business at all. n(23) = 92 x 23 — 2 x (23)2 — 10 = 1048 > n(0). The
solution to this problem is therefore X = 23. The monopolist should produce 23 units.
If we change the original problem by making the fixed cost 1060 instead of 10, then
X = 23 is still the only relative maximum. But now n(23) = —2 < n(0). So the mo-
nopolist is better off not getting into the business at all. Here the optimum point
occurs at the boundary.
If our objective function has two independent variables, then we have to resort
to partial derivatives. Suppose Z = f(X, Y), let
aZ
aX
0.Z
fY ay' 'fxx -x aZ
ax
(taking partial derivative twice with respect to X),
a (az) , fx f a (az
LY
OY aê ". ax (ay)
(taking partial derivative twice, first with respect to Y, then with respect to X). The
conditions for relative maxima and minima are
fx = 0
(first-order conditions).
fy = 0
In addition, if fxx fn, > f x2y , then the point is either a relative maximum or minimum.
To distinguish relative maximum and minimum, we have
Ly < 0 maximum
(second-order conditions).
LSI fyy > 0 minimum
An example of using partial derivatives to find the optimum point is given in the appli-
cation section of Appendix B.
1. Constrained Optimization
Very often, a business entity operates under certain constraints. They may be
budgetary, technological, or physical constraints. To solve this constrained optimi-
zation problem, we can use the method of Lagrange multipliers if the constraints are
given as equations. For example, the production function of a firm may be F(X, Y) =
2X Y, where X represents units of labor and Y represents units of capital. The budget-
ary constraint may look like
g(X, Y) = 100 — 2X — 10Y = 0,
OPTIMIZATION 915
where 100 represents the maximum amount of money to be spent on this production
and 2 and 10 represent unit costs of labor and capital, respectively. To use the method
of Lagrange multipliers, we first construct a new function of three independent
variables:
L(X, Y, )) = f(X, Y) + :tg(X, Y)
= 2X Y+ X(100 — 2X — 10Y),
where X is a new variable that is called the Lagrange multiplier. The constrained opti-
mum will appear as a solution to the first-order condition:
OL
= L =0
OX
01,
— = Ly = first-order conditions).
OY
01,
071
-
-
Let
0 gx gy
H = g x L„„ L xy and 1111 = determinant of H (defined in Appendix B);
g y L yx L yy l
then
IHI < 0 relative minimum
(second-order conditions).
IHI > 0 relative maximum
Example: Take the production function and the budgetary constraint above and
find the optimal combination of labor and capital.
L(X, Y, )) = 2X Y + .4100 — 2X — 10Y)
L x = 2Y — 2X =0
L y = 2X — 10:t = 0 (first-order conditions).
L A = 100 — 2X — 10Y = 0
Solving these equations simultaneously gives us X = 25, Y = 5, X = 5. (For a discus-
sion of methods of solving system of linear equations, see Appendix B.) Therefore under
the budgetary constraint the maximum output level is f(25, 5) = 2 x 25 x 5 = 250
when we employ 25 units of labor and 5 units of capital. We know that this must be
the maximum point before computing the second-order condition because a relative
maximum is the only sensible solution. The interested reader may check that
0 —2 —10
IHI = —2 0 2 = 80 > 0.
—10 2 0
916 CALCULUS AND OPTIMIZATION
2. The Meaning of
The solution of 2 also has a meaning. The magnitude of 2 measures how much
the optimum changes as we relax the constraint. In the above example the solution
of 2 is 5. That means if we relax the budgetary constraint 1 unit from 100 to 101, the
optimal level of output would increase approximately 5 units to 255.
If the solution to is equal to zero, then the constraint is not binding. That means
the constrained optimum is equal to the unconstrained optimum.
The Taylor and MacLaurin series are widely used in economics and finance. Their
most important use is to help evaluate the value of a function around a certain point.
Suppose we are interested in evaluating the function Y = f(X) around a point a in its
domain. Then we can make use of Taylor's theorem:
Taylor's theorem. In the one-dimensional case we can evaluate the function Y =
f(X) around the point a in terms of its derivatives as follows:
f"(a)(X — a)2
f(X) = f(a) + f'(a)(X — a) +
2!
)
f"'(a)(X — a)3 (a)(X — a)n
+ • • • + fo
3! n!
Alternatively, if we let h = (X — a), then the Taylor series is
f(a)h2
f (a + h) = f (a) + f(a)h +
2!
f"( a)11
a)h3 3
f (n)(a)h"
+•••+ (Pratt [1964] uses this.),
n!
where f(a) = value of the function at point a. This is called a Taylor series.
Definition: If we evaluate the function around zero (i.e., if a = 0 above), then we
have what is called a MacLaurin series:
f
" (0) f NO) f"
f (X) = f (0) + f'(0) • X + •V+ • X3 + • • + n!(0) • Xn.
2! 3!
Definition: The symbol n! [read "n factorial"] represents the product of all posi-
tive integers from 1 to n (or vice versa). That is,
n! = n • (n — 1) • (n — 2) (n — 3) • (n — 4) • • 4 • 3 2 • 1.
TAYLOR AND MACLAURIN SERIES 917
Examples:
!! = ! • 4 • 3 • 2 1 = 1 • 2 • 3 • 4 • 5 = 120,
10! = 10 • 9 • 8 7 • 6 • 5 • 4 • 3 • 2 • 1 = 3,628,800,
(n — r ! = (n — r • (n — r — 1) • (n — r — 2) • (n — r — 3) • • • 4 • 3 • 2 • 1.
By definition 1! = 0! = 1.
Intuitively, what the Taylor series is trying to do is to approximate the function
f (2 with the following polynomial:
} f(2 T o + T i( 2 — a + T 2( 2 — a)2 + T 3( 2 — a)3 + • • • . The problem is to
find the values of the coefficients (the Ts) of this polynomial. To find them, take
the higher-order derivatives of (1):
2 f](2 = + T2 2(X — a + T3 • é(2 — a 2 + T4 • 4(X — a 3 + • •
( 2 = 2T 2 + T3 • 2 • 3(X — a + T4 • 4 • 3(X — a) + • • •
2
é
4 f]"(2 = 2 • 3T3 + T4 • 4 • 3 2(X — a + T 5 • ! • 4 • 3(X — a)2 + • • •
Plugging these values of the Ts into (1) results in the Taylor series we stated earlier.
The usefulness of Taylor series can best be seen with the help of a numerical example.
Example: Expand the function f (2 = }m2 around 1, for n = 0, 1, 2, 3. Computing the
derivatives:
f(2 = X so f(} = 1= },
—} —}
f](2 = 22 so } (} = = },
(1)2
2
2
f"(2 = —
so f"(1) = = 2,
X3 ( 1) 3
—6 —6
f]"(2 =
X4
so f" ](} = (1) 4 = —6.
918 CALCULUS AND OPTIMIZATION
2 2
= f(} + (-} (2 — } + (2 — }
2 1
= } — (2 — 1) + ( 2 — 1)2.
d) The approximation when n = 3 is
2(a f](a
2 é
Té(2 = f(a + f](a (2 — a + (2 a + (2 — a
3!
6
= 1 — ( 2 — 1) + ( 2 — 1)2 + (2 1)3
3•2 1
= 1 — ( 2 — 1) + ( 2 — 1)2 — ( 2 — 1)3.
Expanding and rearranging the polynomials:
T o( 2 = 1 (constant),
T }( 2 = } — (2 — } = —2 + 2 (straight line),
2
T 2( 2 = 1 — ( 2 — 1) + ( 2 — } = 2 2 — é2 + 3 (parabola),
2 é é
T é( 2 = } — (2 + } + (2 — } — (2 — } = —2 + 42 2 — 62 + 4 (cubic).
Figure D.16 graphs the function f(2 = }m2 and each of the approximating
polynomials:
dT 2( 2
= 22 — 3 = 0, 2 = 1.5 = min,
d2
d T3(X)
— 3X2 + 8X — 6 = 0
d2
( — é2 + 2 (2 — 3) inflection point > 1.
From the graph we see that each successive Taylor series does a better job of approxi-
mating f(2 = }m2 in the vicinity of 1; see Fig. D.17.
TAYLOR AND MACLAURIN SERIES 919
2
Y = T 2 (X) = X -3X + 3
3
1 Y T o (X) = 1
.75
Y=
x
3
Y ( X)=-X+ 2
3 2
Y = T 3 ( X)= -X + 4X - 6X + 4
Figure D.16
T1 ( X) -= -X + 2
Figure D.17
920 CALCULUS AND OPTIMIZATION
Example: Pratt (1964) uses Taylor series to derive a measure of absolute relative risk
aversion. Let
X = amount of wealth,
U = acceptable utility function,
7E = risk premium, it is a function, n(2, 2 ,
2 = a gamble (and a random variable),
2 - E(Z - T3( 2, 2 .
} E{U(2 + 2 } = U[2 + E(2 — n(2, 2 ]
by choosing an actuarially neutral risk E(Z = 0. So
1') E{U(2 + 2 } = U[2 — n(2, 2 ].
0
E(2
= E{U(2 } + E{Z}U'(X) + 2 U"(2 .
2!
But
E{U(2 } = U(2 not a random variable,
E{2} = 0,
E{2<} = a since u = Z p i [Z i — E(•,7 ] 2 = E pi Z? = E(Z? .
0
So
U(é) rr
U (X) = U(2 - 7En2 .
2
INTEGRAL CALCULUS 921
1 ,< U"(2 }}
= Z[ ur(x)
a function of 2 and 2.
-
E. INTEGRAL CALCULUS
1. Indefinite Integrals
Integration is the reverse process to differentiation. Given a function f(2 , the
indefinite integral of f(2 , denoted by .1 f(2 d2, is a function whose derivative is
f(2 . In other words,
A peculiar feature regarding the indefinite integral of f(2 is that it is not unique.
Observe the following fact: if F](2 = f(2 , so is [F(2 + C]] = F](2 + 0 = f(2
where C is an arbitrary constant. Therefore both F(2 and F(2 + C can be an in-
definite integral of f(2 . So, in general, we write
f(2 d2 = F(2 + C
2. Rules of Integration
1. f X' dX — x n+i 3. (n —}
n + 1
1
2. f — d2 = ln X + C. ( 2 > 0)
X
3. f ex d2 = e x + C.
6. ax dX = ax + C.
ln a
922 CALCULUS AND OPTIMIZATION
1) 5 2X 2 + 3X+1 dX= 2 X 2 dX + 3 f X dX + f 1 dX
X2
+3
=2 +X+C
3 2
= 4X 3 + X + C.
s 2X + 1
2) dX. Here we have to use the method of substitution again.
+X
Let u = X 2 + X, then du = (2X + 1) dX.
c 2X + 1 1
dX= f du = ln u + C = ln(X 2 + X) + C.
X2 + X
1 1 1 U3/2
= 5U 1 /2 du = +C
2 2 3/2
= — (X 2 + 1)3 / 2 + C.
3
3. Definite Integrals
A typical definite integral looks like f 'a f(X)dX [read "integral of f(X) from a
to b"]. Here f(X) is called the integrand, a is called the lower limit, and b is called
the upper limit of integration. The main difference between an indefinite and a definite
integral is that the result of indefinite integration is a function, whereas the result of
definite integration is a number. The meaning of that number is as follows.
Let f ab f(X)dX = A. If f(X) > 0, then A is simply the area under the curve
Y = f (X) from a to b, shown in Fig. D.18(a). That area is equal to f ab f (X) dX. Suppose
INTEGRAL CALCULUS 923
(a) (b)
Figure D.18
now f(2 is both positive and negative in the range of a < 2 < b, then A, the result
of c'ci f( 2 d2, is the signed area "under" f(2 from a to b. By "signed area," we mean
that the area above the X-axis is assigned a positive sign and the area below the
X-axis is assigned a negative sign. Then A is the sum of all the positive and nega-
tive area [see Fig. D.18(b)]. If the curve of f(2 is the one in Fig. D.18(b), then
b f(2 d2 = A = area 1 — area 2 + area 3.
a
The link between the definite integral and the indefinite integral is given by the
next theorem, which is called the Fundamental Theorem of Calculus.
Theorem: Let F(2 be an indefinite integral of f(2 , then
sab
f(2 d2 = F(b — F(a .
The theorem shows us a way to evaluate the definite integral. We need only to find
the indefinite integral of the integrand and then substitute the upper and lower limits
and find the difference. Although the indefinite integral of a function is not unique,
the theorem says that any one will do, as long as it is the same one in which you
substitute the upper and lower limits.
Examples:
2
12
(X + 2) d2 = ( x,y 2 + 22
j12
=[2 + 2(2)] — 2[- + 2 . 1]
= (2 + 4) — (-21 + 2) = 31,
101
2) e x
d2 = e x 1
= _ _ e — }.
0
I. faa f(2 d2 = O.
!. fab [f(2
+ g(2 ] d2 = f(2 d2 + f b g(2 d2 .
4. Applications
Example 1: Let the fixed cost of production be 100 and let marginal cost be 10/Nrk
per unit. What is the total cost function for producing q units?
Total cost = fixed cost + total variable cost
= 100 + f g 10/ NN d2
}
= 100 + 101/ X'2
2 0
1 _ —
.05 x 0]
= 10,000 x e
.05
= —200,000[0.6065 — 1]
= 78,700.
Note that the formula for the present value of continuous discounting of a continuous
flow is ft, (CF)e -"tdt, where CF = cash flow per time unit, T = time when cash flow
ends, and r = discount rate.
5. Improper Integrals
Sometimes the limits of integration may be — co or + co. Such a definite integral
is called an improper integral. To evaluate an improper integral, we do not substitute
— oo or + oo into the indefinite integral, but rather we substitute a variable b in place
REFERENCE 9<P
f_ . f (é) d é = lim f — co
b
a
f (é) d é ,
Example- Suppose the income stream in the previous example is perpetual; then
the present value would îe
PV = fo =0,000e dt
PV= lim f b
10,000e — 5
dt
=
= lim =0,000 _ [e -.056 e -.05 x 0]
—
b—, 00 .0P
= =0,000 x (-<0)[0 — =], since e — 05b
' --> 0 as b ÿM
PV = $<00,000.
Note that for a perpetual stream, we also have
PV CF = =0,000
$<00,000 where CF is the cash flow.
.0P
This gives the same result as the integral calculus method.
REFERENCE
Pratt, J. W., "Risk Aversion in the Small and in the Large," Economet]ica, January—April =964,
=<<-=36.
Author Index
Aîdel-Khalik, A., 633 Bradley, M., P=8-P=9, 686, 7<P, 7<6-7<7, 739,
Aîramovitz, M., 300 7PP
Adler, M., 8<3 Branch, B., 39=
Affleck, A., 6P0 Brauer, G., 390
Aharony, J., 369, P84, P8P Breeden, D. T., =3P, 8==
Alchian, A., <0 Brennan, M. J., <==, <=<, <77, 306, 307, 3=7,
Alexander, G. J., 744-74P 3<<, 383, 4<9, 476, 478, 479, PP7, PP9,
Alexander, S. S., 349 P6=, P70, P90, P9=, 70=, 70P
Altman, E., 9P, P00 Brick, I. E., 47<
Anderson, R., 307, 3=3 Brickley, J., P88, 670
Ang, J., P33, 633 Brigham, E., 47P, P97
Arrow, K. J., 88, ===, 68P Brindisi, L., 666
Asquith, P., P<<, P87, 7<8 Brown, L., 367-369, 38P
Brudney, V., 736
Baîiak, H., P77, P84 Bruner, R. F., 7<8
Ball, R., 3P0, 367-369 Bulow, J., 6PP
Banz, R. W., =<7, =3P-=39, =40, <=4, <=P
Barnea, A., 47< Capozza, D., 3=P, 3=6
Bartter, B., <P7 Caraîini, C., 3=6
Basu, S., <=4, <=P Carleton, W., P8
Baumol, W. S., P8 Cass, D., =<3
Bautista, A., 6<< Caves, R., 633
Bawa, V. J., 9P Chang, E., 3<=
Baxter, N., 498 Chen, A., 6P3-6P4, 6P7
Beckers, S., <88 Chen, N. F., <<8, <<9, <30, 386
Benson, P. G., 744-74P Chew, I. K., 6=
Bernhard, A., 38P-387, 394 Chirelstein, M. A., 736
Bernhard, R. H., P8 Christofi, A., 8=0
Bhagat, S., 670 Christofi, P., 8=0
Bhattacharya, M., <8P, <86, P64 Chung, K. S., 686, 69<, 7P3
Bicksler, J., 6P3-6P4, 6P7 Coase, R. H., <0
Bierman, H., P33, P97 Coates, C., P97
Black, F., =93, <=4, <=P, <=6, <4P, <P6, <66, Conn, R. L., 7P8-763, 769
<67-<69, <83, <87, 307, 309, 3==, 3P=, Constantinides, G. M., 308, 40<, 4==, 4=8,
38P, 386, 466, P07, P89, P90, P9=, 6P0, 708 473, 47P, 476
Blattîerg, R., <87 Cooley, P. L., 6=, P83
Blume, M., 90, <=4, 349, 383, 708 Copeland, T. E., 38<, 383, 38P, 386, 4<0-4<7,
iroale76), 3<0, 3<= P70, 630, 63=, 7P=
Bogue, M. C., 40<, 403, 406, 407, 4=0, 4==, Cordes, J., P=8
430 Cornell, B., 308, 3=P, 3=6, 344, 346, 3P<, 38P,
Bowman, R., 63< 8=6, 8=8-8<0
Boyce, W. M., 479 Coughlan, A., 666
9<7
9<8 AUTHOR INDEX
Cox, J. C., <40, <P<, <P7, <P9, <68, <69, <77, Fisher, L., 380-38<, P84, 7=8
<80, <8=, <87, 4=3, 6P8, 70= Fisher, S., <8<
Cramer, H., =P0 Flath, D., 6<6
Crawford, P., 633 Forsythe, R., 339-343, 8<=
Crockett, J., 383 Foster, G., 369
Culîertson, J. M., 69 Foster, T., 38<
Cyert, R. M., <0 Frankle, A. W., 6=, 63
Franks, J., 633
Dann, L., 374-37P, P<3, P97, P98, 600, 738, Fredman, A., P97
739, 740-74< French, K., 308, 3=6, 3=7, 3<=, 3<<, 390-39=
DeAngelo, H. C., =<6, =<7, 4P4, 46=, P=8, Frenkel, J., 8=8
66<, 664, 66P, 736, 737, 738, 739, 740- Friend, I., 90, <=4, 383, P88, 708, 887
74<, 74<-743, 7P<-7P3, 7PP
DeAngelo, L., 66<, 664, 66P, 74<-743, 7P<- Gagnon, J. M., <P
7P3, 7PP Galai, D., <7P, <84-<8P, 468, 473, P08, 694,
Deîreu, G., === 707
Demsetz, H., <0 Garman, M., =3P
Dennis, D., 7P6 Gatto, M., 648
Dertouzos, J. N., 690 Gay, G., 303, 494
Desai, A., 686 Gehr, A., <<8, 4==, 430
Dietrich, J. R., P<< Geijsîeek, W., 48
Dietrich, K., 8=8-8=9 Geske, R., <P<, <7P, <79, <8=, <88, 648
Dill, D., 6<< Giîîons, M. R., <=<, <=4, <=P, <=6, 8<<
Dilullo, A. J., 786, 789 Givoly, D., 376
Dimson, E., <=< Goldschmid, H. J., 734
Dodd, P., 7<<-7<3, 7<6, 739-740, 7PP, 7P6 Gonedes, N. J., <87, 7P7-763, 769
Dufey, G., 8<3 Goudzwaard, M. B., 707
Dumas, B., 8<3 Grauer, F. L. A., 80<, 8==, 8=4, 8=9, 8<<, 834
Dusak, K., 3<0, 3<= Green, R., 478
Dyîvig, P., 307 Griffin, P., 369
Dyl, E., 39=, 4<8 Grinîlatt, M., 38<, 383, P8=
Grossman, S. J., 343, 344, 38P, 734-73P, 7P6
Eades, K., P80, P8= Grove, M. A., 49P
Eckîo, B. E., P<3, 73=-734 Gruîer, M. J., P78, P79, P97
Ellert, J. C., 7<=-7<<, 730-73= Gudikunst, A., 633
Elton, E. J., P78, P79, P97 Gultekin, M., 39=
Emanuel, D. C., 473, 47P Gultekin, N. B., 39=
Evans, 0., 787
Ezzell, J., P3P Hakansson, N., P6P
Haley, C. W., 447
Fama, E. F., 69, 79, =68, =8P, =86, =94, <=0, Halpern, P. J., 7=8-7=9, 7<=, 7<6
<=4, <=7, <=8, <=9, 3=7, 3<=, 3<<, 33<, Hamada, R. S., 4PP, P=8
338, 343, 349, 3P=, 380-38<, 40<, 407, Handjiinicolaou, G., P87
4=0, 4==, 4=8, 430, 47<, P77, P84, P96, Hanoch, G., 9P
667, 687, 7=8, 7=9, 8<=, 8<< Harper, C., 633
Farîer, A., 80=, 8<< Harris, J., P83
Farrar, D. E., 383, PP7, PP9, P6= Harris, M., 480
Feller, W., <6P Hart, 0., 734-73P, 7P6
Ferguson, D. L., 9P Hartzmark, M. L., 306
Findlay, M. C., 6=, 63 Hatfield, K., 377
Finnerty, J. E., 376 Haugen, R., 47<
Firth, M., 7<4-7<P Hausman, W., 38P
Fisher, A. K., 377 Hawkins, G. D., 48=
Fisher, I., 3, 46 Hayek, F. H., 343
AUTHOR INDEX 9<9
Healey, P., 667 Keown, A. J., 7<8-7<9
Hekman, C. R., 8<3 Kessel, R. A., 68
Heminler, M., 303 Keynes, J. M., 68, 309, 339
Henderson, S., 6P7 Khoury, N., P96
Henricksson, R., 383 Kim, E. H., 498, P=8-P=9, P80, P8=, P83, 686,
Hess, P., P80, P8= 7<8
Hicks, J. R., 68, 309 Kira, D., 9P
Higgins, R. C., 698 Klammer, T., 47
Hirshleifer, D., 309 Klemkosky, R., <86
Hirshleifer, J., 49, 68, =09, =<4, 330, 33< Kolî, R. W., 494
Hite, G. L., 748-7P0 Kolodny, R., P<<
Hodges, S., 633 Korwar, A., P<<
Hodrick, R. J., 8<0 Krasker, W., P87
Hoel, P. G., 88< Kraus, A., =<7, 37=, 373, 374, 498, 699-700
Hoffmeister, J. F., 7<3-7<4 Kroll, Y., 9P
Hogarty, T., 7=8 Kummer, D. R., 7<3-7<4
Holloway, C., 38P Kwan, C., P84, P8P
Hong, H., 36P, 366
Houthakker, H. S., 300 Lakonishok, J., P8=
Hsia, C. C., 464, 468 Lamîert, R., 67=
Huang, C. C., 336 Landsman, W., 6P3
Huang, R. D., 8<= Lane, W., P97
Huntsman, B., 687 Langetieg, T., P00, 7<4
Husick, F., <=4 Larcker, D., 670, 67=
Hutchison, M. M., 8<9 Larson, A. B., 39<
Larson, K. D., 7P7-763, 769
Iîîotson, R. G., =8P, =86, 377-379, 380 Latane, H., <74
Ingersoll, J., 390, 476, 478, 6P8 Latham, M., 33<
Isard, P., 8== Lease, R., P8<, 633, 670, 74<
Lee, W., 630
Jacquillat, B., 8=0 Lee, W. H., P<0
Jaffe, J., 376 Lee, Y. K., P8P
Jain, P. C., 74P-746 Leiîowitz, M. L., 49<, 494
Jarrell, G., P=8-P=9, 7<6-7<7 Leland, H., P0<-P03
Jean, W., 9P Lemgruîer, E., 670, 67=, 7P=
Jensen, M., <0, <=4, <=P, 3P=, 380-38<, 383, Lev, B., 7=8
384, 386, 47<, P09, P84, P9=, 66<, 687, Levich, R., 8=8, 8<<, 834
708, 7=6, 7=8, 7P6 Levy, H., 9P, 6<3, 694
Joehnk, M. D., 707 Lewellen, W., P8<, P83, 6<P, 687, 69<, 704
Johnson, H., <79, <8<, P=< Linn, S. C., 737
Johnson, L. J., 8<4 Lintner, J., =93, <==, 3P=, 4==, P77
Johnson, R., 633 Litzenîerger, R., =<7, =3P, <=4, <=P, 369, 498,
Joy, 0. M., 369, 4<8 P9=, P9<, P93, 60=, 648, 699-700, 80<,
8==, 8=4, 8=9, 8<<, 834
Kahneman, D., =0< Logue, D. E., 377
Kalaîa, R., P00 Long, H. W., 407, 4=0, 4<8
Kalay, A., P=3, P79, P80, P87 Long, J., P94, P9P, 60=
Kalotay, A. J., 479 Long, M., P=8, P=9, 6<P
Kampmeyer, J., 744-74P Lorie, J. H., P8
Kane, A., P8P
Kaplan, R. S., 36P, 366, 38P Macaulay, F. R., 49=
Kaufold, H., 8<3 Macîeth, J., <=4, <87, <88-<89, 3P=, 7=9
Keim, D., <=P, 39= Machlup, F., <0
Kelly, E. M., 7=8 Mains, N. E., 383, 384
930 AUTHOR INDEX
Taylor, R., 633 Warner, J., 499, P00, P=<, 739, 7P6
Tehranian, H., 67= Wart, J. R., 9P
Teichroew, D., 3P Watts, R., 369, P=0, P84, P8P, P86, 668
Telser, L., 307, 309, 3=7 Weil, R., 38P
Tepper, I., 6P0 Weinîerger, A., 494
Thompson, R., 388, 389, 390, P87, 633, 7<7- Weingartner, H. M., P8, P9, 60
7<8, 7PP Weinstein, M., 380, P00, P=P, P=6
Thorpe, K. E., 690 Weitzman, M., 430
Titman, S., 38<, 383, 499, P==, P=8, P=9, P8= Wessels, R., P=8, P=9
Toîin, J., 96, 8<< West, R., P97
Treynor, J., =93, =96, 383, 646 Westerfield, R., 6P3
Triffin, R., 78P, 787 Weston, J. F., 4<0-4<7, P=7, 630, 63=, 686,
Trueman, B., P6=, P63 707, 734, 763
Tversky, A., =0< Whaley, R., <7P
Whitmore, G. A., 9P
Ungerer, H., 787 Whittaker, J. G., 8<9
Wier, P., 733-734
Vanderwicken, P., 6=4 Williams, J., <=<
Van Home, J. C., 6=, 4<8 Williamson, 0. E., <0, 68P
Vermaelen, T., P8=, P97, P98, P99, 600 Wolfson, M., 370
Vertinsky, I., 336 Woods, D., P97
Vickers, D., 383 Woodward, S., 68
Vickrey, D., 38< Woolridge, J. R., 38<, P86, P87
Vickson, R. G., 9P Working, H., 309
Von Neumann, J., 77, =0<
Yawitz, J. B., 494
Waegelein, J., 67= Young, A., P97
Wakeman, L., P=6, 6<6, 6P7-6P8, 739 Young, P., 787
Walker, C. E., 69
Walter, J. E., 83, 84, P44 Ziemîa, W. T., 9P, 336
Walther, C. H., 8<4 Zimmerman, J., P=0
S&–ject Index
933
934 SUBJECT INDEX
constant elasticity of, <8=, <87 WACC. See Weighted average cost of capital
of expected spot price, 3== Warrants, 473-476
of futures price, 3=< Weak form market efficiency, 33<, 348, 3P0
instantaneous, <74 Wealth expropriation, îondholder, P07-P09,
marginal rate of suîstitution îetween mean P=9, 600, 743-744
and, 96 Weekend effect, 390-39<
matrix definition, =74 Weighted average cost of capital, <9, 39, 40<,
minimum variance opportunity set, =6P-=66, 444-446, 4P0, 4P8, 470, 499, P=7, P<6-
=67, =70, =94 P3=, 60<
minimum variance portfolio, =6=-=6<, 87=- Williams Act of =968, 7<6-7<7
873 Working capital, 4=
minimum variance zero-îeta portfolio, <06,
36< Year-end effect, 390-39<
of a portfolio, =6=, =73-=74, =84 Yield curve, twists in, <30
of a two-asset portfolio, =PP Yield to maturity, 6P-66
Variance-covariance matrix, 87<
Venture capitalists, 663 Zero-îeta, <<0
Vertical integration, 68P Zero-îeta portfolio, <06, <=8, 36<
Vertical mergers, 678 Zero coupon îonds, 66, <48, <P3, 464
Vested îenefits, 640
Voluntary selloffs, 744-746
Voluntary spinoffs, 746-7P0