Professional Documents
Culture Documents
DSC3705 Study Guide 2016
DSC3705 Study Guide 2016
MO001/4/2016
Financial Risk Modelling
DSC3705
Bar code
Open Rubric
university
of south africa
c UNISA 2015
Department of Decision Sciences
University of South-Africa
All rights reserved.
Printed and distributed by the
University of South-Africa
Mucleneuk, Pretoria.
DSC3705/001/4/2016
CONTENTS
DSC3705
Contents
1 Study unit 1
Introduction
2 Study unit 2
Mean-variance portfolio theory
3 Study unit 3
The capital asset pricing model
25
CONTENTS
DSC3705
37
46
. . . . . . . . . 47
47
. . . . . . . . . . . . . . . 55
58
CONTENTS
DSC3705
Certainty equivalent . . . . . . . . . . . . . . . . . . . 66
6.2.2
Risk profile . . . . . . . . . . . . . . . . . . . . . . . . 68
6.3 Arrow-Pratt . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
6.4 Optimisation . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
7 Solutions: Exercises in study unit 2
80
85
90
97
11 References
104
DSC3705
In this module we follow the textbook, Investment Science by David Luenberger closely. This study guide indicates the chapters covered in this module
and the exercises in the textbook that you should do. It also includes extra
study notes, study outcomes and study hints to help you. The notes clarify
some of the difficulties you may experience in using the textbook.
The references and exercises refer to the textbook and it is indicated with Luenberger. After studying each chapter (including the examples), you should
work through the exercises indicated in the study guide and make sure you
understand the solutions.
You should spend approximately 120 hours on this module (it is the time
spent on problem solving, reading and studying).
Study unit 1
Introduction
DSC3705
dynamics of investments
risk aversion
hedging versus pure investments
simple interest
compound interest
continuous compounding interest
present and future values of cash flows
Study unit 2
Mean-variance portfolio theory
Luenberger: Chapter 6
Purpose
We apply mean-variance analysis to model uncertainty in financial decisionmaking.
2.1
Study hints
At the end of this section you should be able to do the following:
Distinguish between total return R and the rate of return r.
Explain the process of short selling.
Explain why short selling is considered to be risky.
DSC3705
Activity
Work through the calculations of the portfolio weights and the expected rate
of return of the portfolio in Table 6.1(Luenberger, page 141).
Apply the formulas on page 140 with:
w1 = 0,25, w2 = 0,50 and w3 = 0,25
r1 = 0,17, r2 = 0,13 and r3 = 0,23
R1 = 1,17, R2 = 1,13 and R3 = 1,23
X0 = 16 000
X01 = w1 X0 = ............
X02 = w2 X0 = ............
X03 = w3 X0 = ..........
Then
R =
= ..................................................
= ..................................................
Also determine R using the formula:
R=
P3
i=1
wi Ri = ............................
Determine r =
P3
i=1
wi ri = ...................
2.2
DSC3705
Revision notes
See Appendix A, page 475 (Luenberger) and the OPS/QMS/OMG202 or
DSC2602 study guide. You may also consult the textbook by Winston used
in the other modules.
Study hints
At the end of this section you should be able to do the following:
Explain what a random variable is.
Define a probability function p(x).
Define and calculate the expected value, variance and standard deviation of a random variable.
Explain independent random variables.
Define and calculate the covariance of two random variables.
Derive equation 6.3.
Derive equation 6.4.
Construct a mean-standard deviation diagram (or r diagram).
Study note
Properties of the covariance between random variables are often used in
the module. For revision consult Winston or the QMG/OPS/QMS202 or
DSC2602 study guide.
The definition for covariance is given on page 144.
The textbook uses the notation xi to refer to a random variable. However,
it is sometimes useful to indicate a random variable with a tilde and write
xi to distinguish it from real numbers.
DSC3705
Let x and y and z be real-valued random variables and a and b real numbers.
Suppose E[
x] = x and E[
y ] = y. Then the following holds:
cov(
x,
y ) = E[(
x x)(
y y)] (definition)
cov(
x,
y ) = E(
x y) x y
if x and y are independent, then cov(
x,
y) = 0
cov(
x,
x) = var(
x)
cov(
x,
y ) = cov(
y,
x)
cov(a
x,b
y ) = ab cov(
x,
y)
cov(a
x + b
y ,
z ) = a cov(
x,
z ) + b cov(
y ,
z)
Activity
Suppose the following possible rates of return (in percentage) for two assets
r1 and r2 are available.
market condition
good
average
poor
probability
p1
p2
p3
asset 1
15
9
3
asset 2
1
10
19
Suppose the alternative return on each asset are assumed equally likely (thus
the probability for each market condition is the same). Then pi = 31 for
i = 1; 2; 3.
Express the % values in decimal form. Determine the mean rate of return
(or expected rate of return) for each asset:
r1 = E[
r1 ] = 31 (0,15) + 31 (0,09) + 13 (0,03) = 0,09
r2 = E[
r2 ] = ...............................................
DSC3705
var(
r1 ) = E[
r12 ] (
r1 ) 2
1
1
1
(0,15)3 + (0,09)2 + (0,03)2 (0,09)2
=
3
3
3
= 0,0024
Then the standard deviation 1 =.....................
Determine (in the space below) 2 , the standard deviation of asset 2:
12 = cov(
r1 ,
r2 ) = E[(
r1 r1 )(
r2 r2 )]
1
1
=
(0,15 0,09)(0,01 0,10) + (0,09 0,09)(0,10 0,10)
3
3
1
+ (0,03 0,09)(0,19 0,10)
3
= .......................
Calculate (in the space below) the correlation coefficient using the formula
12
12 =
:
1 2
10
DSC3705
2.3
Function: noun
Inflected Form(s): plural portfolios
Etymology: Italian portafoglio, from portare to carry (from Latin)
+ foglio leaf, sheet, from Latin folium
Date: 1722
1: a hinged cover or flexible case for carrying loose papers, pictures, or pamphlets
2 [from the use of such a case to carry documents of state] : the
office and functions of a minister of state or member of a cabinet
3: the securities held by an investor : the commercial paper held
by a financial house (as a bank)
4: a set of pictures (as drawings or photographs) usually bound
11
DSC3705
12
DSC3705
Study note 1
Suppose a portfolio consists of n assets with random rates of return ri . The
textbook uses the notation ri without a tilde on pages 150 and 151: however
we want to emphasize that a assets rate of return is uncertain or a random
variable.
The expected rate of return of asset i is E[
ri ] = ri . Suppose we form a
portfolio of these n assets, with weight wi for asset i . Then thePoverall rate of
return of the portfolio r is also a random variable with r = ni=1 wi ri . The
expected rate of return or mean of the portfolio is indicated by E[
r ] or r.
Then the expected rate of return (or mean) of the portfolio satisfies:
E[
r] =
=
n
X
i=1
n
X
wi E[
ri ]
wi ri
i=1
var(
r) = =
n
X
wi wj ij
i,j=1
Activity
The results of the activity on page 6 are summarised in the next table:
asset 1 asset 2
0,09
0,10
0,0024 0,0054
13
DSC3705
1
2
Then:
rp = E[
r ] = w1 r1 + w2 r2 = ...................
p2
2
X
wi wj ij
i,j=1
= w1 w1 11 + w1 w2 12 + w2 w1 21 + w2 w2 22
= w12 11 + 2w1 w2 12 + w22 22
= ............................................
= ............................................
Study note 2
Suppose a portfolio is constructed by taking equal portions of the n assets.
Assume the assets are mutually uncorrelated. Suppose that the expected
rate of return (or mean) of each asset is equal to m and the variance of each
asset is equal to 2 .
Then the weights are wi = n1 , i = 1; . . . ; n and the expected rate of return
(or mean) of the portfolio satisfies:
14
DSC3705
r = E[
r]
n
1X
E[
ri ]
=
n i=1
n
1X
=
ri
n i=1
n
1X
=
m
n i=1
1
=
(nm)
n
= m
The variance of the portfolio satisfies:
var(
r) = 2 =
n
X
wi wj ij
i,j=1
n
X
11
ij
n
n
i,j=1
n
1 X 2
n2 i=1
1
(n 2 )
n2
2
=
n
=
15
DSC3705
Activity
To understand the proof of the portfolio diagram lemma consider a portfolio
with the following two assets:
asset mean return ri
i=1
0,10
i=2
0,15
standard deviation i
0,20
0,25
weight wi
(1 )
(a) Plot the coordinates of asset 1 (0,20; 0,10) and asset 2 (0,25; 0,15) on a
r diagram. Note that r is on the y-axis and on the x-axis. Compare
with points 1 and 2 in figure 6.8 (Luenberger).
(b) Calculate the mean return (
r) of the portfolio and the standard deviation
() of the portfolio return. Then
r = w1 r1 + w2 r2
= (1 )(0,10) + (0,15) = 0,10 + 0,05
Note that the textbook uses the notation r() to indicate that the expected
return is a function of the weight , hence we may write r() = 0,10 + 0,05.
Case 1: = 1
=
12
.
1 2
16
DSC3705
Then 0,20(1 ) 0,25 = 0 implies = 0,44. You may also apply the
1
formula = 1+
to obtain 0,44. Verify the value 0,44 in the space below:
2
If > 0,44, the values r of the portfolio will be on the straight line
) = (0; 0,122) and (2 ; r2 ) (or point
connecting (0; A) = (0; (0,10)(0,24)+(0,15)(0,20)
0,20+0,25
2), thus the upper straight line in figure 6.8. If < 0,44, the portfolio mean
and standard deviation will be on the straight line between (0; A) and (1 ; r1 )
(or point 1), thus the lower line in figure 6.8. Note that the boundaries of
the triangle between A, point 1 and point 2 in figure 6.8 cover all possible
portfolio outcomes if = 1.
Choose any value for and determine the mean and standard deviation of
the portfolio:
Plot the coordinates on the diagram in (a) to verify that this coordinate is
either on the upper straight line or the lower straight line.
Case 3
For other values the portfolios mean and standard deviation will not be
on a straight line. The coordinates will be on a curve connecting points 1
17
DSC3705
r =
0
0,25
0,5
0,75
1,0
Then the portfolio values for = 0,0; 0,25; 0,5; 0,75; 1 are (0,20; 0,10), (0,1737; 0,1125),
(0,1750; 0,1250), (0,1737; 0,1375), (0,25; 0,15). If these values are represented
on the diagram in (a) and are connected a curve-shaped graph is obtained.
One of the problems in Assignment 1 is a practical problem to understand
figure 6.8 (the portfolio diagram). Please do this problem and construct the
diagram with two assets. Although this is quite time-consuming you will
benefit from the time spent on this problem.
18
DSC3705
the three assets. Show that the variance of the portfolio return satisfies the
following equation.
0,7 2
var(r) =
+ 0,3 2
3
2.4
Study hints
At the end of this section you should be able to answer the following:
What is a feasible set?
What is a minimum-variance set?
Which portfolio will a risk-averse investor prefer?
Where is the efficient frontier on a r diagram?
Which is the minimum-variance point (MVP)?
2.5
Study hints
At the end of this section you should be able to do the following:
Formulate the mathematical problem that leads to minimum-variance
portfolios for n assets.
Apply the Lagrangean method to derive the equations that should be
solved to determine the weights of portfolios that lead to minimumvariance portfolios. Note that different weights are obtained for each
specific r. The weights are used to determine the standard deviation
of the portfolio. The coordinates (; r) are points on the minimumvariance set.
19
DSC3705
1
[w1 w1 (0,0225) + 2w1 w2 (0,0045) + w2 w2 (0,09)]
2
Subject to
0,10w1 + 0,18w2 = r
w1 + w2 = 1
Derive the Lagrangean equations to find the minimum-variance for a given
r:
1
1
L = (0,0225)w12+w1 w2 (0,0045)+ (0,09)w22(0,10w1 +0,18(w2
r )(w1 +w2 1)
2
2
Hence,
L
= 0,0225w1 + 0,0045w2 0,10
w1
L
= 0,0045w1 + 0,09w2 0,18 .
w2
20
DSC3705
Then the set of equations to determine an efficient portfolio with mean rate
of return r is:
0,0225
0,0045
0,10
1
solve:
0,0045
0,09
0,18
1
0,10
0,18
0
0
w1
1
1 w2
0
0
0
= .
r
1
For instance, if r = 0,12, the solution to the previous system of equations is:
0,75
0,25
0,0984
0,0082
Activity
Use Maxima/Mupad/Matlab/Derive to verify the solution.
In this special case where n = 2, one need only solve the last two equations
for every r. (This will only be possible if n = 2.)
Thus for r = 0,12 we should solve:
0,10w1 + 0,18w2 = 0,12
w1 + w2 = 1.
21
DSC3705
Activity
Show that the solution to
0,12
w1
0,10 0,18
=
1
w2
1
1
is
0,75
0,25
For each r, one determines the weights of a portfolio on the minimumvariance set. Then the variance, and hence the standard deviation, of the
portfolio are calculated.
The following table is a summary of a few possible portfolios on the minimumvariance set. Verify some of the values, to make sure you know how to do
the calculations.
r
w1
w2
0,10
1
0
0,15
0,11 0,875 0,125 0,1401
0,12
0,75
0,25 0,1413
0,13 0,6250 0,3750 0,1513
0,14
0,5
0,5 0,1743
0,15 0,375 0,625 0,2011
0,16
0,25
0,75 0,2318
0,18
0
1
0,30
Activity
Plot the standard deviation and expected return values on a r diagram
to obtain the minimum-variance set.
Note that the minimum-variance point (MVP) will be between r = 0,10 and
r = 0,12. (Compare the values and remember the minimum should be
calculated.) How do we determine the coordinates of the minimum-variance
point? Since this is an example with n = 2, the problem can be reduced to
a single-variable calculus problem.
22
DSC3705
Thus
M V P 0,1392.
Indicate the point, MV P = (0,1392; 0,1139), on the previous r diagram.
Alternative method to find the minimum-variance point:
See the comment before Example 6.11 on page 163 for another method to
obtain the weights of the minimum-variance set.
Hence, set = 1 and = 0 is the following equations (see equation 6.5a in
23
DSC3705
P2
i=1
wi ,
24
2.6
DSC3705
Study hints
At the end of this section you should be able to do the following:
Formulate and apply the two-fund theorem.
Study note
Verify all the steps in the following example:
Suppose there are two mutual funds w1 and w2 available on the minimumvariance set with:
Fund
w1
w2
Weights
(w11 ; w21; . . . ; wn1 )
(w12 ; w22; . . . ; wn2 )
Expected return
r1
r2
Standard deviation
1
2
NB: Note that the superscripts 1 and 2 in this problem indicate fund 1 and
fund 2, thus r1 is the expected return of fund 1 and r2 is the expected return
of fund 2. Then 1 is the standard deviation of fund 1 and 2 is the standard
deviation of fund 2. In this problem 2 does not indicate . See also the
notation on page 162.
Since portfolios 1 and 2 are on the minimum-variance set, both satisfy equations 6.8a, 6.8b and 6.8c on page 162.
Activity: Write down the equations for the two portfolios:
Then a new portfolio with weights (w11 ; w21; . . . ; wn1 )+(1)(w12; w22; . . . ; wn2 )
is also a minimum-variance portfolio for any R. To prove that the
new portfolio is a minimum-variance portfolio we should show that the new
portfolio satisfies equations 6.8a to 6.8c.
Suppose the weights of the new portfolio p are w1p ; . . . ; wnp .
25
DSC3705
wip =
i=1
n
X
i=1
[wi1 + (1 )wi2]
n
X
i=1
wi1
+ (1 )
n
X
wi2
i=1
= 1 + (1 ) 1
= 1
P
P 2
and equation 6.8c is satisfied. ( wi1 = 1 and
wi = 1 since portfolios 1
and 2 satisfy equation 6.8c.)
To show that equationP
6.8b is satisfied, P
remember the n assets have return
2
1
1
r1 ; r2 ; . . . ; rn and r =
wi ri and r = wi2 ri .
Then:
n
n
X
X
wip ri =
(wi1 + (1 )wi2)
ri
i=1
i=1
n
X
i=1
wi1 ri
+ (1 )
n
X
wi2 ri
i=1
=
r 1 + (1 )
r2
= rp
Next we show that the new portfolio satisfies the minimum-variance condition
(equation 6.8a).
The following holds for each i (i = 1; 2; . . . ; n):
n
X
j=1
ij wjp
rip
n
X
ij [wj1 + (1 )wj2 ]
rip
j=1
n
X
j=1
[
ri1
=
= 0
ij wj1
+ (1 )
] + (1
n
X
ij wj2
rip
j=1
)[
ri2
] (
ri1 + (1 )
ri2]
26
DSC3705
also on the minimum-variance set. The return of this new portfolio will then
be r =
r 1 + (1 )
r 2 . Hence, if you want a minimum-variance portfolio
2.7
Study hints
At the end of this section you should be able to:
Define the rate of return of a portfolio with one risk-free asset.
Define the standard deviation of a portfolio with one risk-free asset.
Formulate the one-fund theorem and illustrate the theorem on a r
diagram.
Apply the one-fund theorem in a simplified case.
2.8
27
DSC3705
Study unit 3
The capital asset pricing model
Luenberger: Chapter 7
Purpose
We discuss the capital asset pricing model (CAPM) and apply it to investment decision problems. The CAPM describes the relationship between the
expected return of a asset and its risk relative to the market risk. It is used
in the pricing of risky securities.
3.1
28
DSC3705
Study hints
At the end of this section you should be able to do the following:
Discuss the concept of the market portfolio.
Define capitalisation weights.
3.2
Study note
The capital market line represents the set of risk-return combinations available by combining the market portfolio with risk-free borrowing or lending.
It is represented on a r diagram (or a mean-standard deviation diagram.)
The capital market line relates the expected rate of return (
r ) for an efficient
portfolio to a given risk of return (measured by the standard deviation ()).
It is based on the idea that an investor demands additional expected return
(called the risk premium) if he/she is asked to accept additional risk. For a
given level of risk, the portfolio located on this line offers the best return. The
capital market line is graphically tangent to the efficient frontier containing
the market portfolio.
The capital market line passes through the risk-free point and the market
portfolio. Hence, to determine the equation of the capital market line, you
should determine the equation of the straight line that passes through (0; rf )
and (M ; rM ), with rf the risk free rate, rM the market rate of return and
M the market standard deviation of return.
Activity
Suppose the All-Shares index for a country is used as a proxy for the market
of the country. Suppose the expected rate of return of the index is 13% and
the standard deviation of the rates of return is 19%. Suppose the current
risk-free rate is 9%.
Determine the equation of the capital market line. Represent the capital
market line graphically.
29
DSC3705
Solution:
Given: rf = 0,09; rM = 0,13; M = 0,19
Equation of the capital market line (see equation 7.1 p 176, Luenberger):
rM rf
r = rf +
M
= ...............................
= ................................
Did you obtain the equation r = 0,09 + 0,2105?
Draw a graph of this straight line with r on the y-axis and on the x-axis (or
use Maxima to plot the function). Note that it is important to indicate the
coordinates (or position) of the risk-free asset and the market on the graph
and to use a proper scale.
Study hints
At the end of this section you should be able to do the following:
Derive equation (7.1) of the capital market line.
Represent the capital market line on a r diagram.
Define, calculate and interpret the price of risk.
30
DSC3705
Understand that the capital maket line relates the expected rate of
return of an efficient portfolio to its standard deviation.
3.3
This section discusses the theory of the capital asset pricing model.
Study hints
At the end of this section you should be able to do the following:
Formulate the capital asset pricing model (CAPM).
Prove the capital asset pricing model (CAPM).
Define and interpret the beta () of an asset.
Apply the CAPM formula.
Determine the beta of a portfolio.
Understand that the expected rate of return of an asset is proportional
to its covariance with the market.
Study notes
Make sure you understand the following:
For an individual asset, is a good indication of risk.
For a portfolio, is a good indication of risk.
rf is the risk-free rate of return.
ri rf is the expected excess rate of return of asset i.
rM rf is the expected excess return of the market.
The expected excess return of the individual asset i is proportional to
the expected excess return of the market. The value of measures this
proportion.
31
3.4
DSC3705
Study hints
At the end of this section you should be able to do the following:
Determine the equation of a security market line in terms of the covariance of assets and the market (thus cov(r,rM ) is on the x-axis).
Determine the equation of a security market line in terms of the beta
of assets (thus is on the x-axis).
Represent the security market line on a graph.
Define systematic risk.
Explain nonsystematic risk.
Explain diversification.
Know that efficient (minimum-variance) assets have systematic risk
only.
Work through the explanation in the following study notes.
Study note 1
The CAPM model can be represented as a security market line. There are
two possible equations to describe this security market line.
Equation 1
The security line can be expressed in terms of the expected asset return ri
and the covariance of the asset and the market iM . The equation is:
rM rf
iM + rf .
ri =
2
M
Hence, the risk-reward is expressed in terms of the covariance of the return
with the market. To sketch this equation, note that the covariance iM is
32
DSC3705
represented on the x-axis and the expected return is represented on the yaxis. See figure 7.3a (Luenberger). Note that if the asset is the market, then
2
the expected return is rM and iM = M M = M
. This value is indicated
with an M in figure 7.3a (Luenberger).
Equation 2
The security line can also be expressed as
ri = (
rM rf )i + rf .
In this case, the risk-reward is expressed as a linear relationship between the
expected return of asset i and the beta of the asset. Thus the risk of an asset
is a function of its beta. To sketch this equation note that the i value is on
the x-axis, and the expected return on the y-axis.
Activity
Suppose the All-Shares index for a country is used as a proxy for the market
of the country. Suppose the expected rate of return of the index is 13% and
the standard deviation of the rates of return is 19%. Suppose the current
risk-free rate is 9%.
1. Determine the equation of the security market market line in terms of the
covariance of asset i and the market M (iM ). Represent the equation on a
(iM ; r) -diagram.
Solution
Apply equation 1. The equation of the security market line is:
iM
(
rM rf )
2
M
= ..............................
ri = rf +
= ..............................
Did you obtain r = 0,09 + 1,1080iM ?
Draw a graph of this straight line with r on the y-axis and iM on the x-axis
33
DSC3705
(or use Maxima to plot the function). Indicate the positions of the risk-free
asset and the market on the graph and use a proper scale.
2. Determine the equation of the security market line in terms of i . Represent the equation of the security market line on a (i ; r)-diagram.
Solution:
Apply equation 2. The equation of the security market line is:
ri = rf + i (
rM rf )
= .............................
= .............................
Did you obtain r = 0,09 + 0,04i ?
Draw a graph of this straight line with r on the y-axis and i on the x-axis
(or use Maxima to plot the function). Indicate the positions of the risk-free
asset and the market and use a proper scale.
34
DSC3705
ri rf
E[ri ] E[rf ]
ri rf
E[i ]
=
=
=
=
i (rm rf ) + i
i (E[rM ] E[rf ]) + E[i ]
i (
rm rf ) + E[i ]
0
cov(ri ,rM ) =
=
=
=
35
DSC3705
Hence,
i2 = cov(ri ,ri )
= cov(rf + i (rM rf ) + i ,rf + i (rM rf ) + i )
= cov(rf ,rf + i (rM rf ) + i ) + i cov(rM rf ,rf + i (rM rf ) + i )
+cov(i ,rf + i (rM rf ) + i )
= i2 cov(rM ,rM ) + cov(i ,i ) (the other cov values are zero)
2
= i2 M
+ var(i )
= systematic risk + specific/nonsystematic risk
3.5
Read and understand this paragraph. Remember the CAPM is a model that
may be applied in investments. Do you think that all investors apply this
model?
3.6
Study hints
At the end of this section you should be able to do the following:
Note that r is the expected return of an asset obtained from a sample.
Define Jensens index.
Determine and interpret Jensens index.
Define the Sharpe index.
Determine and interpret the Sharpe index.
Study note: Note that the covariance in table 7.3 has been expressed as a
170
percentage with cov = (100)(100)
= 0,0107.
36
DSC3705
AA
17
34
5
32
10
S&P T-bills
16,8
6,4
31,5
8,4
3,2
7,8
30,6
5,6
7,7
3,5
37
3.7
DSC3705
Study hints
At the end of this section you should be able to do the following:
Formulate the pricing form of the CAPM.
Explain the relationship between the linearity of the pricing formula
and arbitrage opportunities.
Apply the pricing formula to verify the value of a share.
Study note
The following are comments on example 7.5(Luenberger):
The beta of the fund is:
=
r rf
(0,1)(0,07) + (0,9)(0,15) 0,07
=
= 0,9
rM rf
0,15 0,07
38
DSC3705
Activity
If the expected value of the share after one period is R120, what should you
pay at the start of the period?
Answer:
P =
3.8
120
105,08
1,142
Read and understand this paragraph. It summarises all the important concepts in the chapter.
39
DSC3705
Study unit 4
Models and data
Luenberger: Chapter 8
Purpose
In this unit models to approximate the parameter values (such as mean values
and covariance values) for stock returns are discussed. The factor model is
explained and the arbitrage pricing theory derived.
Read chapter 8.1.
4.1
Study hints
At the end of this section you should be able to do the following:
40
DSC3705
f
ai + ei
= return on asset i
(ri is a random variable)
= change in the return on asset i per unit change in the
return in factor f
(f is a random variable)
= value of the factor
= the portion of the return on asset i not related to the
factor
ai is a known expected value and ei unknown residual
(or error) component.
41
DSC3705
42
DSC3705
=
=
=
=
E[a + e] + bE[f]
a + bf ( since E[
e] = 0)
a + (0,5)(7)
6,5
Thus the assets non-factor mean return is 6,5% and the factor-related mean
return is 3,5%.
43
DSC3705
Study note 3
Derive equations 8.2b to 8.2d:
i2 =
=
=
=
cov(ri ,ri )
cov(ai + bi f + ei ,ai + bi f + ei )
b2i cov(f,f ) + 2bi cov(f,ei ) + cov(ei ,ei )
b2i f2 + e2i
The factor-related value b2i f2 is also called the nondiversifiable risk, whereas
e2i refers to the diversifiable risk.
Justification:
Since ai is not random cov(ai ,ai ) = 0, cov(ai ,f ) = 0 and cov(ai ,ei ) = 0.
The assumptions of the factor model imply that the following hold:
cov(f,ei ) = 0
cov(ei ,ej ) = 0 for i 6= j
The following holds for i 6= j:
ij =
=
=
=
cov(ri ,rj )
cov(ai + bi f + ei ,aj + bj f + ej )
bi bj cov(f,f ) + cov(ei ,ei ) (the other covariances are zero)
bi bj f2 + 0
= bi bj f2
Also:
cov(ri ,f ) =
=
=
=
cov(ai + bi f + ei ,f )
cov(ai ,f ) + bi cov(f,f ) + cov(ei ,f )
0 + bi cov(f,f ) + 0
bi f2
44
DSC3705
Study note 4
This relates to estimating historical factor loadings. Suppose we have known
data return values for a stock (thus not a stochastic model) then we apply
linear regression to determine the factor loading values.
Suppose the returns on a stock for the past n weeks are r 1 ; r 2 ; . . . ; r n . Suppose the factor return values for the n weeks are f 1 ; f 2 ; . . . ; f n . For example
the factor values may be the interest rates, market values or index values.
Thus f i may be the interest rate value in week i.
Then the factor loading rate b value, measuring the expected change in r,
given the change in f , can be determined using regression analysis.
Suppose r =
1
n
Pn
k=1
r k and f =
1
n
Pn
k=1 f
Pn
k
)(f k f)
k=1 (r r
Pn
k
2
k=1 (f f )
a = r bf
45
DSC3705
(a) Determine the equation of the factor model for each stock.
(b) For each stock plot the values (f k ; rik ) on a graph and draw the equation
of the factor model for the stock on the same graph. (Thus you should have
two graphs similar to figure 8.1, p199 Luenberger).
(Hint: The problem is similar to example 8.1. You may use a spreadsheet
package (Excel) if you prefer.)
4.2
Study hints
At the end of this section you should be able to do the following:
Adapt the factor model to model the relationship between the variable
(ri rf ) and factor (rM rf ).
46
DSC3705
Determine the equation 8.5 of the characteristic line and illustrate the
line graphically.
State which assumptions ei satisfies.
Distinguish between the characteristic line and the CAPM model. (Read
and understand the paragraph before example 8.2 on page 206).
Study note
Note that the single-factor model makes no assumptions about efficiency.
Make sure you understand the difference between the CAPM model and the
single-factor model. The least-squares line (regression line) fitted through
data points (rM rf ; ri rf ) for a specific stock in such a way that E[ei ] = 0,
is called a characteristic line. If i = 0, this characteristic line gives the
CAPM model.
The following exercise will help you to understand the concept of a characteristic line.
Exercise 4.2.1
1. Apply equation 8.5 to determine the equation of the characteristic line for
the return (in percentages) of the following stock.
Year
1
2
3
4
5
k
Stock (rik ) Market (rM
) Riskless (rfk )
27,24
23,0
7,1
17,05
15,3
6,9
10,2
11,5
6,8
20,26
18,5
7
19,84
18,1
6,8
P
Determine rf = 51 5k=1 rfk . Determine the equation of the characteristic line.
Represent the values (rM rf ; ri rf ) and the equation of the characteristic
line on the same diagram.
47
4.3
DSC3705
Study hints
At the end of this section you should be able to do the following:
Discuss the assumptions of the APT model.
Explain the APT model in the one-factor model environment.
Give the conditions under which the APT model and the CAPM model
are identical.
Suppose there are n assets governed by two factors. Formulate the
APT model for the two factors. Thus, apply the APT theorem with
m = 2.
Explain the relation between the APT model and CAPM model in the
two-factor case (see page 211 at the bottom).
4.4
48
DSC3705
Study unit 5
General principles
Luenberger: Chapter 8
Purpose
In this unit, investment decision-making in the utility framework is discussed.
The utility criterion and the mean-variance criterion are compared. Linear
pricing, one of the fundamental properties in security pricing, is explained.
Appendix 6 (in this study guide) contains extra notes on utility decisionmaking to enhance the understanding of chapters 9.1 to 9.5.
5.1
Study hints
At the end of this section you should be able to do the following:
What is a utility function?
Define and calculate the expected utility of a random variable.
Apply the expected maximum utility criterion (see appendix 6 in this
guide).
Sketch a utility function.
49
5.2
DSC3705
Study hints
At the end of this section you should be able to do the following:
Define and calculate the certainty equivalent of a random variable or
lottery.
Define and apply risk aversion.
Define and apply the Arrow-Pratt absolute risk aversion coefficient.
Distinguish between risk-averse, risk-neutral and risk-seeking investor
behaviour.
5.3
Study hints
Read through this section. It explains methods to determine an investors utility function.
5.4
Study hints
At the end of this section you should be able to do the following:
Define a quadratic utility function.
Compare the mean-variance criterion with the expected utility criterion.
50
DSC3705
Note that the mean-variance criterion may be applied if the returns are
normal random variables (thus the normal distribution holds).
Activity
Make sure you have worked through appendix 6 (in this study
guide).
5.5
Study hints
At the end of this section you should be able to do the following:
Explain type A arbitrage.
Explain type B arbitrage.
Describe the concepts of linear pricing.
Determine the price of a portfolio based on the linear pricing principle.
5.6
Study hints
At the end of this section you should be able to do the following:
State the portfolio choice problem (9.3a-9.3d).
Formulate the portfolio choice theorem.
Use a utility function to construct a portfolio that gives maximum
satisfaction (utility) for a given wealth level.
Apply Lagrangean multipliers to solve equations 9.3a to 9.3d.
51
DSC3705
52
DSC3705
Maximise E[u(
2
X
i di )] = max E[u(1 d1 + 2 d2 )]
i=1
3
31 + 1,22
1
+ 0,4
1 + 22
L
1,2
1,2
1,2
= 0,3
+ 0,4
+ 0,3
2
31 + 1,22
1 + 22
1,22
Hence, the following set of nonlinear equations should be solved for fixed W :
0,4
0,9
+
=
31 + 1,22 1 + 1,22
0,36
0,48
0,36
+
+
=
31 + 1,22 1 + 1,22 1,22
1 + 2 = W
Using the first two equations we obtain:
0,08
0,36
0,54
=0
31 + 1,22 1 + 1,22 1,22
53
DSC3705
In this example, since we have only two unknowns, we could apply the following trick to solve the system of equations.
Suppose one unit of wealth is invested, thus W = 1 and 1 + 2 = 1. Then:
0,54
0,08
0,36
= 0
3(1 2 ) + 1,22 (1 2 ) + 1,22 1,22
0,54
0,08
0,36
= 0 (II)
3 1,82 1 + 0,22 1,22
Solve this equation using MUPAD/Maxima or any other computer package
with a suitable function to find the root of a nonlinear function. You may also
apply a numerical technique such as the secant method or Newton-method to
approximate the zero (or root). Your pocket calculator may also be used if it
has a suitable function (see the reference manual of your pocket calculator).
The solution is 2 0,911 (the other zero of the equation is a negative value).
Then 1 0,089.
Thus for any value W , 1 0,089W and 2 0,911W and =
1
.
W
P
Suppose x = 2i=1 i di is the optimal solution. Assume W = 1 unit, then
the expected return of the optimal portfolio is:
E[x ] = E[0,089d1 + 0,911d2]
= 0,3(0,089 3 + 0,911 1,2) + 0,4(0,089 1 + 0,911 1,2) + 0,3(0 + 0,911 1,2)
Would you advice the investor to invest in the film venture?
Alternative solution:
Since this portfolio problem has only two unknowns the problem could be
solved by expressing the original portfolio problem in one variable.
Since 1 = W 2 , the portfolio problem is equivalent to:
Maximise V (2 ) = 0,3 ln(3(W 2 ) + 1,22 ) + 0,4 ln((W 2 ) + 1,22 ) + 0,3 ln(1,22 )
= 0,3 ln(3W 1,82 ) + 0,4 ln(W + 0,22 ) + 0,3 ln(1,22 )
54
DSC3705
5.7
Study hints
At the end of this section you should be able to do the following:
Explain states (sometimes called states-of-nature).
Define elementary state securities.
Express the price of a security in terms of the elementary state security
prices.
Formulate the positive state price theorem.
Determine the state prices if no arbitrage opportunities exist.
Study note
In the following note, an arbitrage portfolio and state prices are discussed
using the examples and terminology used in the textbook Mathematics for
Economics and Finance by Martin Anthony and Norman Biggs. This textbook is prescribed for the Mathematical modelling II module.
Suppose an investor can choose between a number of different assets. After a
specified time period, the assets in the portfolio will have a new value. This
55
DSC3705
future value of the portfolio will depend on a number of factors that may
change in the time period. This uncertainty is modelled by assuming that
there are a number of possible states. If the future portfolio has the same
value at the end of the time period, whichever state has occurred, it is called
a riskless portfolio. A portfolio that costs nothing, cannot lose money and
yields a profit in at least one state is called an arbitrage portfolio.
We illustrate an arbitrage portfolio with an example.
Suppose the citizens of an island can invest in three assets namely land,
bonds (which yield a fixed return) and stocks (which yield an uncertain
return). A general election is due. After the election, one of two parties,
the conservative party (CP) or the liberal party (LP), will be in power. The
winning partys economic decisions will influence the value of the assets. An
economist predicts the following possible total returns of each asset in the
two states.
Land
Bonds
Stocks
CP LP
1,25 0,95
1,05 1,05
0,90 1,15
1,25 0,95
Then R = 1,05 1,05 is the matrix of total returns.
0,90 1,15
If an investor decides to invest $5 000 in land, $1 000 in bonds and $4 000 in
stocks, the row vector
Y = [5 000 1 000 4 000]
represents the portfolio.
Suppose the CP wins the election, then the value of the portfolio at the end
of the time-period is 5 000(1,25) + 1 000(1,05) + 4 000(0,90) = 10 900.
Then the product Y R is the vector [10 900 10 400], where the 10 900 is the
final value of the portfolio if CP wins and 10 400 is the final value if LP wins.
If the portfolio of the investor is Y = [5 000 10 000 5 000] the investor
borrows $10 000 from the bank and uses the money to buy land and stocks.
At the end of the year he/she owes the amount 10 000 1,05 dollars, but the
56
DSC3705
...
R = u
with
1
2
..
.
n
and u =
1
1
..
.
1
where u is a (m 1) matrix and represents the final price of 1 unit for each
asset.
To determine state prices the system R = u should be solved.
If R = u is inconsistent (i.e. the system has no solution), state prices do
not exist and there must be an arbitrage portfolio. Also if the system has a
57
DSC3705
solution, with one of the i values negative or zero, i.e. i 0 for at least
one i, there must be an arbitrage portfolio.
If the system R = u has more than one solution or has a unique solution
(i.e. the system is consistent) and i > 0 for all i, then positive state prices
do exist and arbitrage is not possible.
In example 9.9 (Luenberger), there are three states and three assets and the
system of equations R = u has a unique solution. In this example
3
1
0
R = 1,2 1,2 1,2 .
6
0
0
Thus we should solve:
3
1
0
1
1
1,2 1,2 1,2 2 = 1
6
0
0
3
1
Activity
Verify that the solution to this system is 1 = 61 , 2 = 21 and = 16 . Because
i > 0, for i = 1; 2; 3, we can say there is no arbitrage.
5.8
Study hints
At the end of this section you should be able to do the following:
Derive the risk-neutral pricing formula (equation 9.15).
How do you determine the risk-neutral probabilities in the risk-neutral
pricing formula?
58
5.9
DSC3705
probability
0,3
0,4
0,3
B
rate of return
0
0,2
0,6
probability
0,3
0,2
0,5
C
rate of return
0,2
0,3
0,4
probability
0,45
0,1
0,45
6. Thobela has a utility function u(x) = ln(x). Suppose she has initial wealth
w and is offered a gamble. If she bets x her final wealth will be w + x with
59
DSC3705
probability p (0 < p < 1), otherwise her final wealth will be w x. Determine
the value x as a function of p if she is an expected utility maximiser.
7. Luenberger p257, 9.9
8. Luenberger p257, 9.10
9. Luenberger p257, 9.11
10. Luenberger p257, 9.14
60
DSC3705
Appendix
Utility and Risk Preferences
61
6.1
DSC3705
Expected utility
n
X
pi xi
i=1
If u(x) is the utility function, the expected utility of the x is given by:
E[u(
x)] =
n
X
pi u(xi )
i=1
n
X
pi xi
i=1
In the same way the expected utility of the lottery P is given by:
E[u(P )] = E[u(
x)] =
n
X
pi u(xi )
i=1
In the context of this section you may model uncertain decisions either with
monetary outcomes in terms of lotteries or with random variables.
Let the collection of possible outcomes be given in: Z = {0; 10; 20; 30; 40; 50}.
A decision-maker must choose between lotteries P and Q. Apply the expected maximum utility criterion.
s
0,2
Ps
A
A
A 0,5
A
As
s
0,3
20
50
s
0,3
s
A
A
A 0,3
A
s As
0,4
30
40
62
DSC3705
The utility function is given in the first two columns of the following table:
Outcome
zi
0
10
20
30
40
50
p(z)u(z)
0,64
0,0
1,65
0
0
3,85
6,14
prob. lottery Q
q(zi )
0,3
0
0
0,4
0,3
0
1,0
q(z)u(z)
0,96
0
0
2,52
2,13
0
5,61
E[u(P )] =
X
zZ
E[u(Q)] =
X
zZ
63
DSC3705
we have:
E[u(P )] =
p(z)u(z)
zZ
64
DSC3705
s
@
@
@Y es
@
@s
@
@
0,45
0,30
0,25 @
@
s
s
@s
No
x 5 000
x + 10 000
We calculate the expected utility of the investment and compare it with the
expected utility of no investment.
The expected utility of no investment:
65
6.2
DSC3705
John Von Neumann and Oscar Morgenstern showed in 1944 that any consistent, rational decision-maker should choose between lotteries or gambles
according to the rules of utility theory. Since then, decision analysts have
developed methods to assess the utility function of an individual. The relative attractiveness of different outcomes is measured and is then captured in
a utility function over a range of possible consequences. Initially the concept
of utility functions was developed to apply in economic decisions. Between
1960 and 1970 the mathematical properties of utility functions describing attitude towards money and risk were worked out in detail. It was recognized
that there are families of utility functions that represent different attitudes
towards risk. Nowadays social sciences and behaviour sciences also apply the
utility concept.
Why should we apply risk preferences when choosing between lotteries? Consider the following lotteries for example:
Qs
Ps
s
0,5
A
A
A 0,5
A
As
30
s
0,5
A
A
1 900
A 0,5
A
As
2 000
66
DSC3705
If you were to make decisions on the basis of expected value, you would
choose lottery Q. However, which lottery is riskier? If we play each lottery
10 times, the worst we could do in lottery P is to lose R10. If we play lottery
Q ten times, the worst that can happen is to lose R19 000! We could say
that lottery Q is more risky. A risk-averse person will prefer lottery P , while
a risk-seeking person might enjoy lottery Q! Expected payoff should not be
the only criteria used to choose between lotteries.
Assume you are obliged to play lottery Q, but you do not want to lose R1 900.
Your optimistic friend is interested in taking your place and is willing to pay
R30 for the opportunity. If you are prepared to accept the R30, but not
less, for the lottery to avoid playing, your certainty equivalent for the lottery
would be R30. We write CE[Q] = 30. The certainty equivalent is the least
you would accept to avoid playing the lottery. You will receive R30 for sure,
whether your friend wins or loses.
The utility function of a decision-maker gives information on his/her attitude
towards risk. In the textbook methods to determine a decisionmakers utility
function are described.
The first restriction placed on a utility function is that it should be consistent
with more being preferred to less. The utility of (x+1) rand should be greater
than the utility of x rand. Thus an utility function is an increasing function
with a positive first derivative. Whether a utility function is concave or
convex is the second important property of a utility function. This property
describes a decision-makers attitude towards risk.
In the definitions in this section X is a subset of the real line.
Definition 6.2.1 A function u : X R is strictly increasing whenever
u(x1 ) < u(x2 ) if x1 < x2 and x1 ; x2 X.
Note that if the function u is strictly increasing and it is differentiable then
u (x) > 0 for all x X. If u (x) 0, then u is increasing on the domain.
A function u(x) is said to be strictly concave if every pair of two distinct
points on the curve can be joined with a straight line entirely below the
curve. A concave function opens downward. The following criteria may be
used to determine whether a function is concave.
67
DSC3705
Definition 6.2.2 A twice differentiable function u : X R is strictly concave if and only if u (x) < 0 for all x X.
Note that if u (x) < 0 for all x X, then u(x) is strictly concave.
If u (x) 0, u is concave.
Example of a concave function:
y
..........
..............
u(x)
..............
.............
.........
..........
straight line below
.
...........
.........
..........
..........
.........
.
.
.
.
.
.
.
......
......
......
......
.....
.
.
.
.
.....
....
....
....
...
.
.
.
....
....
...
...
...
.
..
..
...
..
....
..
..
...
..
...
.
..
..
..
...
.
..
..
..
..
...
..
..
A function u(x) is strictly convex if any two points on the curve y = u(x) can
be joined by a straight line above the curve. A convex curve opens upward.
Definition 6.2.3 A twice differentiable function u : X R is strictly convex if and only if u (x) > 0 for all x X
68
DSC3705
Convex function
100
90
80
70
u(x)
60
50
40
30
20
10
6.2.1
5
x
10
Certainty equivalent
69
DSC3705
Example 6.2.1
Determine the expected value, the expected utility and the certainty equivalent of the following lottery, if the utility function is u(x) = 4x x2 , where
0 x 2.
Ps
0,5
s
A
A
A 0,5
A
As
y
6
u(x)
...
.......................................
.................
.............
..........
.
.
.
.
.
..
......
.....
......
.....
....
.
.
....
...
....
....
...
.
.
...
...
...
..
.
...
...
...
..
.
...
...
...
....
.
..
..
..
..
...
..
..
..
EU=2 . . . . .
.
.
.
.
.
.
.
.
.
.
CE=?
*
EV=1
-x
70
DSC3705
with EU = 2.
4x x2 = 2
x2 + 4x 2 = 0
p
4 (4)2 4(1)(2)
x =
2(1)
x 3,414 or x 0,585
Hence x 0,585 [ because 3,414 is not in the domain.]
The certainty equivalent is CE 0,585.
Note that the function u is monotone increasing and concave on the interval
[0 ; 2]. To verify this, determine u(x) and u (x).
Definition 6.2.5 The risk premium RP of a lottery P is the difference between the expected value and the certainty equivalent.
RP [P ] = EV [P ] CE[P ]
Example 6.2.2
Maria has a utility function u(x) = x. What would she be willing to pay
for a lottery that pays R 1 000 000 with a probability of 0,0015 and zero
otherwise?
Solution:
Determine the expected utility EU = E[u(
x)] of the lottery. Solve
x = EU.
6.2.2
Risk profile
Consider the choice between a prospect that offers an 80% chance of winning
R1000 and a 20% chance of winning nothing, and the alternative of receiving
71
DSC3705
R700 for sure. Many people prefer the certain amount over the gamble,
although the gamble has a higher expected value (R800). This preference for
a sure amount instead of taking part in the gamble is called risk-aversion. If
an individual prefers taking part in this lottery over taking the sure amount,
his/her behaviour is called risk-seeking. Different people have different risk
attitudes and are willing to accept different levels of risk. A utility function
is used to represent a decision-makers attitude towards risk. Different utility
functions represent different qualitative attitudes towards risk.
Decision-makers may have one of three risk profiles.
risk-averse
risk-neutral
risk-seeking
Intuitively, we think of a risk-averse person as one who prefers to behave
conservatively. Say a decision-maker must choose between the 50-50 lottery
P with outcomes 0 and 2, so the expected value EV [P ] = 1. If the decisionmaker prefers the expected value of the lottery for certain, and does not want
to participate in the lottery, the decision-maker is risk-averse. The riskaverse decision-maker would rather prefer receiving the R1 and will avoid
taking part in the lottery.
Definition 6.2.6 A decision-maker is risk-averse if she/he prefers the expected value of the lottery to the lottery.
For a risk-averse decision-maker the utility of the expected value of the lottery
must be greater than the expected utility of the lottery. Suppose the lottery
is represented by a random variable x with EU the expected utility of the
lottery and EV the expected value of the lottery. Then:
u(EV ) > EU
hence u(E[
x]) > E[u(
x)]
72
DSC3705
Example 6.2.3
In the previous example, with u(x) = 4x x2 where 0 x < 2, and P the
50-50 lottery with outcomes 0 and 2, we have:
Utility of the expected value:
u(EV ) = u(E[
x]) = u(1) = 3
Expected utility of the lottery:
EU = E[u(
x)] = 2
Hence: u(E[
x]) > E[u(
x)] or u(EV ) > EU.
The risk premium of the lottery is:
RP = EV CE 1 0,586 = 0,414
Note the risk premium is positive. If a decision-maker is risk-averse, the
expected value of the lottery should be larger than the certainty equivalent
of the lottery (thus the risk premium should be positive).
The following theorem can be proved:
Theorem 6.2.1 A decision-maker is risk-averse if his or her utility function
is concave.
A risk-averse decision-maker has a (strictly) concave utility function if u (x) >
0 and u (x) < 0 for all x. The utility function of a risk-neutral decision-maker
is a straight line, and a risk-seeking decision-maker has a convex utility function.
73
DSC3705
Utility function
Risk premium
Risk-averse
concave
u (x) < 0
positive
RP > 0
Risk-neutral
straight line
u (x) = 0
zero
RP = 0
Risk-seeking
convex
u (x) > 0
negative
RP < 0
74
DSC3705
0,5
s
s
A
A
A 0,5
A
As
30
u (x) = e
= ex/35
35
35
Then u (x) > 0 for all x, thus u is an increasing function.
The function is concave because:
u (x) =
1 1 x/35
e
< 0 for all x.
35 35
75
DSC3705
p1 u(x1 ) + p2 u(x2 )
0,5u(30) + 0,5u(1)
0,5(1 e30/35 ) + 0,5(1 e(1)/35 )
0,5(0,5756) + 0,5(0,0289)
0,2733
35
x
=
=
=
=
0,2733
0,2733 1
0,7267
ln(0,7267)
0,3192
11,17
RP = EV CE
14,5 11,17
= 3,33
76
DSC3705
Note that the risk premium is positive. This confirms that the decision-maker
is risk-averse.
Activity
The following exercise will help you to understand the previous example.
1. Determine u(x) = 1 ex/35 for x = 0; 5; 10; 15; 20; 25; 30 accurate
to 2 decimal places.
2. Sketch the function u(x) and indicate on the x axis the values CE,
EV , RP .
3. Indicate the value EU (the expected utility) on the y axis. Note that
u(CE) = EU.
4. Indicate the value of u(EV ) (thus the utility of the expected value) on
the y-axis. Note that u(EV ) > EU.
Example 6.2.5
You are a risk-averse decision-maker with an utility function
u(x) = 1 ex/35 .
Which of the following lotteries would you prefer?
Qs
Ps
s
0,5
A
A
A 0,5
A
As
30
s
0,4
A
A
A 0,6
A
As
23,5
77
DSC3705
6.3
Arrow-Pratt
For a given baseline wealth w and a random variable x, an interesting question is what happens to the risk premium if the wealth w increases? For
greater amounts of w, is the risk premium of the decision-maker larger or
smaller?
Let us consider the risk premium of a risk-averse person with a given utility
function u. Let u be the risk premium function, depending on the initial
wealth w and the random variable x, and we write u (w,
x). The function
u will be positive for all values of w. It seems to be empirically true for
many people that, as their assets (wealth) increase, they are willing to pay a
smaller risk premium for a given risk. The reasoning is that if they become
richer, they can better afford to take a specific risk. (An example: a student with pocket money of R500 per month can afford a loss of R10 better
than a student with pocket money of R100.) We say the decision-maker is
decreasingly risk-averse.
Definition 6.3.1 A decision-maker is said to be decreasingly risk-averse if
she/he is risk-averse and his risk-premium u (w,
x) for any lottery x, is decreasing as the wealth amount w increases.
Remember: The derivative of a (strictly) increasing differentiable function
is always positive and the derivative of a (strictly) decreasing function is
78
DSC3705
negative.
Definition 6.3.2 The Arrow-Pratt measure of absolute risk-aversion for a
utility function u(x), is defined by
u (x)
RA (x) =
u (x)
The next theorem can be proved:
x12
1
x
1
x
Note: x1 is a decreasing function for all x > 0, hence the decision-maker has
decreasing risk-aversion.
Decreasing risk-aversion is a reasonable assumption. It seems reasonable that
you should become less risk averse as your wealth level increases. A person
who can afford larger losses, may be prepared to undertake greater risks.
Definition 6.3.3 A decision-maker is constantly risk-averse if she/he is
risk-averse and the risk premium u for any lottery x and wealth w is a
constant positive value.
1
Decisions with Multiple Objectives: Preferences and Value Tradeoffs by Keeney and
Raiffa, 1976
79
DSC3705
and 0
80
6.4
DSC3705
Optimisation
s
p
Ps
A
A
2y
A1p
A
As
The speculator with initial capital x can invest only an amount between 0
and x. Suppose the amount x, 0 1, is invested. The speculators
final wealth will either be (x x) + 2x = x + x if the gamble is successful,
or (xx) if the gamble is lost. We represent the final wealth in the following
tree diagram:
s
p
Ps
A
A
x + x
A1p
A
As
x x
Assume the utility function of the speculator is given by u(x) = ln(x). Hence,
the expected utility of his final wealth will be:
81
DSC3705
V () = E[u(
x)] = pu(x + x) + (1 p)u(x x)]
= p ln(x + x) + (1 p) ln(x x)
To find the optimal value of V , we differentiate V with respect to to get:
V () = p
x
x
+ (1 p)
x + x
x x
Solve V () = 0. Then
px
x + x
px(x x)
simplify then:
(2p 1)x2
since x 6= 0 we have
(1 p)x
x x
= (x px)(x + x)
= x2
= 2p 1
Thus the speculator should invest (2p 1)100% of his initial wealth. Note
that, if p 12 , then < 0 and the optimal amount to be invested is 0.
Suppose the probability to double the amount is 0,6. The speculator should
invest (2 0,6 1)(100) = 20% of his initial wealth. Thus, if the investor has
an initial wealth of R1 000, she/he should consider investing R200. You may
verify that, if she/he invests 20% of his initial wealth, the expected utility of
the investment will be 7,04. If she/he invests any other amount the expected
utility will be less.
Note that you can differentiate V () to prove that V () is indeed negative
for all possible values. Thus V () will have a maximum value if = 2p1.
You should obtain:
V () =
(1 p)x2
px2
(x + x)2 (x x)2
if
[0; 1]
82
DSC3705
EXERCISE 2.2.1
Question 1
Derive equation 6.3
Solution
cov(x1 ,x2 ) =
=
=
=
=
E[(x1 x1 )(x2 x2 )]
E[x1 x2 x1 x2 x1 x2 + x1 x2 ]
E[x1 x2 ] x1 E[x2 ] E[x1 ]
x2 + x1 x2
E[x1 x2 ] x1 x2 x1 x2 + x1 x2
E[x1 x2 ] x1 x2
EXERCISE 2.3.1
Question 1
This question refers to the example on page 152.
Suppose each asset i = 1; 2; 3 has a rate of return with mean m and variance
2 . Assume each return pair has a covariance 0,3 2 for i 6= j. Suppose a
portfolio consists of equal portions of the three assets. Show that the variance
of the rate of return of portfolio satisfies the following equation.
0,7 2
+ 0,3 2
var(r) =
3
83
DSC3705
Solution
var(r) =
n
X
wi wj ij
i,j=1
= w1 w1 11 + w1 w2 12 + w1 w3 13 + w2 w1 21 + w2 w2 22 + w2 w3 23 +
+w3 w1 31 + w3 w2 32 + w3 w3 33
1 1
= ( )( ) 2 + 0,3 2 + 0,3 2 + 0,3 2 + 2 + 0,3 2 + 0,3 2 + 0,3 2 + 2
3 3
1
1
(3 2 ) + (6 0,3 2 )
=
9
9
1 2
2
=
+ 0,3( ) 2
3
3
1
1 2
+ 0,3 2 (1 )
=
3
3
1 0,3 2
=
+ 0,3 2
3
0,7 2
+ 0,3 2
=
3
Compare this last equation with the results for the general case (n assets)
on page 152.
EXERCISE 2.8.1
Question 1
Luenberger p170, 6.3
Solution
=
AB
A B
= 1, thus AB = 0,0045.
d
(p2 )
d
= 0.
0,8261.
M V P 0,1392.
84
DSC3705
rp 0,1139.
Question 2
Luenberger p170, 6.5
Solution
Final wealth (profit of loss) without insurance:
state
rain
no rain
probability
0,5
0,5
outcome
106
(3 1)106
Rate of return=
probability
outcome
0,5
u 106 12 u
0,5
(3 1)106 12 u
0,5u 106
1 2(106) 0,5u
0,5(106 )
+
=
.
0,5u + 106
2
0,5u + 106
0,5u + 106
To minimise the risk he should buy full insurance, thus he should buy 3
million units of insurance with a initial payment of 23 million rand. He will
then have 21 million rand at the end of the event, whether it rains or not.
The expected rate of return in this case is
0,5(106 )
.
106 + 32 (106 )
85
DSC3705
Question 3
Luenberger p170, 6.7
Solution
(a) To determine the minimum-variance portfolio you may apply one of the
following methods.
Solution method 1
Solve
P3
j=1
ij wj
ri = 0 for i = 1; 2; 3.
Let = 0 and = 1.
Solve:
2 1 0
w1
1
1 2 1 w2 = 1 .
0 1 2
w3
1
Solution: w1 = 21 , w2 = 0 and w3 = 12 .
Solution method 2
Let Z = p2 =
P3
i,j=1 wi wj ij .
Substitute the sigma (ij ) values (from the covariance matrix), remember
wi wj = wj wi and simplify the expression.
Then Z = 2w12 + 2w1 w2 + 2w2 w3 + 2w22 + 2w32 .
Let w1 = w3 (see hint) then w2 = 1 w1 w3 = 1 2w1 .
Then Z = 2(2w12 2w1 + 1).
Differentiate to w1 , then Z (w1 ) = 2(4w1 2).
Solve Z (w1 ) = 0. Then Z (w1 ) = 0 if w1 = 21 . Z ( 21 ) > 0, thus Z has a
minimum if w1 = 12 .
Hence, p2 has a minimum if w1 = 12 , w2 = 0 and w3 = 21 .
86
DSC3705
2 1 0
w1
0,4
1 2 1 w2 = 0,8 .
0 1 2
w3
0,8
w1
w2 =
w3
1
6
2
6
3
6
P3
i=1
wi = 0,6.
P3
i=1
ki vi = rk rf for k = 1; 2; 3.
Then:
2 1 0
v1
0,4 0,2
0,2
1 2 1 v2 = 0,8 0,2 = 0,6 .
0 1 2
v3
0,8 0,2
0,6
The solution is v1 = 0, v2 = 0,2 and v3 = 0,2.
Normalise the weights (
Then
P3
i=1
vi = 0,4).
v1
0
v2 = 1 .
2
1
v3
2
87
DSC3705
EXERCISE 3.6.1
Question
The following problem is similar to example 7.4 in the textbook.
Suppose the following rates of return percentages are available:
year
1
2
3
4
5
AA
17
34
5
32
10
S+P T-bills
16,8
6,4
31,5
8,4
3,2
7,8
30,6
5,6
7,7
3,5
Rates of return
Standard deviation
AA S&P T-bills
17,6 16,68
6,34
16,165 14,91
1,94
240,24
233,317
1,08,
88
DSC3705
0,230,07
0,32
= 0,07 + 0,5.
89
DSC3705
Solution
2
M
2
X
wi wj ij
i,j=1
1 1
1 1
1 1
1 1
= ( )( )A2 + ( )( )AB + ( )( )AB + ( )( )B2
2 2
2 2
2 2
2 2
= 0,02
1
cov(A,M) = cov(A, (A + B))
2
1
1
= cov(A, A) + cov(A, B)
2
2
1
1
cov(A,A) + cov(A,B)
=
2
2
1 2 1
+ AB
=
2 A 2
= 0,025
A =
cov(A,M)
0,025
= 1,25
=
2
M
0,02
0,015
0,02
= 0,75.
90
DSC3705
rM = rW + (1 ) rV
= (0,12) + (1 )(0,08)
= 0,08 + 0,04
The weights of market portfolio are
(w1M ; w2M ; w3M ) = (0,6; 0,20; 0,20) + (1 )(0,80; 0,20; 0,40)
= (0,8 0,2; 0,4 0,2; 0,20 + 0,40)
The market portfolio has positive weights for all three assets.
Hence,
0,8 0,2 0 4;
0,4 0,2 0 21 ;
0,20 + 0,40 0 2.
Thus
1
2
2.
300
450
150
450
= 13 .
= 32 .
91
DSC3705
0,0105
= 1,296
(c) cov(rA ,rM ) = 0,0105 and A = 0,0081
1
2
1
(cov(rA ,rM ) = cov(rA , 3 ra + 3 rB ) = 3 cov(rA ,rA ) + 32 cov(rA ,rB )
Thus cov(rA ,rM ) = 13 (0,15)2 + 32 31 0,15 0,09)
92
DSC3705
P =
100
= 93,5
(1 + 0,04) + 0,75(0,08 0,04)
Question 6
Prove that the beta of the market is always equal to one.
Solution
i =
iM
2
M
Let i = M, then
M =
2
M M
M
=
= 1.
2
2
M
M
93
DSC3705
The records of the rates of return (in percentages) for two stocks and an
index of industrial prices are given in the following table.
Stock 1 (r1k ) Stock 2 (r2k ) Index (f k )
23,27
27,24
12,3
19,47
17,05
5,5
6,58
10,20
4,3
15,08
20,26
6,7
16,24
19,84
9,7
year k
1
2
3
4
5
(a) Determine the equation of the factor model for each stock.
(b) For each stock i plot the values (f k ; rik ) k = 1; . . . ; 5 on a graph. Draw
the equation ri = a + bf on the same graph.
You may use a spreadsheet package (Excel) if you prefer. Remember when
applying the covariance function of Excel you should multiply the answer by
n
(n is the number of data values in the sample).
n1
Solution
Apply the following formulas for stock i, i = 1; 2.
ri =
1
5
P5
k
k=1 ri
P
f = 15 5k=1 f k
var(ri ) = cov(ri ,ri ) = r2i =
var(f ) = cov(f,f ) = f2 =
cov(ri ,f ) =
bi =
1
4
P5
k
i=1 (ri
1
4
1
4
P5
k
k=1 (ri
P5
k=1 (f
ri )(f k f)
cov(ri ,f )
var(f )
ai = ri bi f
e var = var(ri ) b2i var(f )
ri )2
f)2
94
DSC3705
Answers:
f = 7,7, cov(f,f ) 10,64
ri
var(ri )
cov(ri ,f )
bi
ai
e var
i=1 i=2
13,496 18,918
136,07 37,85
25,995 18,133
2,44
1,7
5,3
5,8
72,57
6,95
EXERCISE 4.2.1
Question 1
Apply equation 8.5 to determine the equation of the characteristic line for
the return (in percentages) of the following stock.
k
year (k) stock (rik ) market (rM
)
1
27,24
23,0
2
17,05
15,3
3
10,2
11,5
4
20,26
18,5
5
19,84
18,1
Determine rf =
1
5
riskless (rfk )
7,1
6,9
6,8
7
6,8
P5
k
k=1 rf .
Let (ri rf ) be represented on the y-axis and (rM rf ) on the x-axis. Reprek
sent the coordinates (rM
rf ; rik rf ) and the equation of the characteristic
line on the same diagram.
Solution
Note that in this example i = 1, since we have only one stock. (Compare
with table 8.2, with 4 stocks.)
Let rf =
1
5
P5
k
k=1 rf .
Then rf 6,92.
k
Calculate rik rf and rM
rf for k = 1; . . . ; 5. Note the value rf is the
95
DSC3705
expected riskless value for the given period and is kept constant in the calculation of the other values.
The mean values ri rf and rM rf are 11,998 and 10,36 respectively.
Then:
1X k
((r rf ) ri rf )2 37,84992
vari =
4 k=1 i
5
varM
1X k
=
((rM rf ) ri rf )2 18,052
4 k=1
with
rf 6,92.
1X k
k
((r rf ) ri rf )((rM
cov(i,M) =
rf ) ri rf ) 26,02095
4 k=1 i
cov(i,M)
26,02095
=
1,441
varM
18,052
1X k
rM =
rM
5
k=1
5
ri =
1X k
r
5 k=1 i
96
DSC3705
5
varM =
2
M
1X k
(rM rM )2
=
4 k=1
5
cov(i,M) = i,M
1X k
k
=
(ri ri )(rM
rM )
4
k=1
You would obtain the same results since the subtraction of a constant
value rf does not influence the covariances.
k
Use Excel to fit a regression line through the values (rM
rf ; rik rf ).
Enter the values rM rf (the x-axis values) and ri rf (the y-axis values) in two columns on a spreadsheet. Choose the option [Tools] on the
task bar, then choose [Data Analysis] and [Regression]. Complete
the pop-up menu. The regression line you obtain should be similar to
the equation of the characteristic line.
EXERCISE 4.4.1
Question 1
Luenberger p224, 8.1
Solution
rM = 0,12
M = 0,18
rf = 0,05
Then rp = 0,2
rA + 0,5
rB + 0,3
rC , with rA the expected rate of return of stock
A; rB the expected rate of return on stock B and rC the expected rate of
return on stock C.
If CAPM conditions are satisfied the i values should be zero. One of the
criticisms against the CAPM model is that it assumes all investors hold the
same portfolio (i.e. market portfolio). In practice some of the assumptions
are often violated. Remember the CAPM model and the factor model are
both merely models trying to explain market behaviour.
Use the formula
r rf = (
rM rf )
to obtain
rA = 0,05 + 1,10(0,12 0,05) 0,127
97
DSC3705
e2 =
3
X
i=1
Then
p2 = (0,92)2(0,18)2 + 0,00034 0,02776.
Thus
p 16,67%.
Question 2
Luenberger p224, 8.2
Solution
Apply the theorem on page 209, then rf = 0,10, r1 = 0,15, r2 = 0,20 with
0 ,1 ,2 the APT factors. Then 0 = rf = 0,10 and
ri = 0,10 + bi1 1 + bi2 2 .
Thus
0,15 = 0,10 + 21 + 12
0,20 = 0,10 + 31 + 42 .
Solve the set of equations. Then 1 = 0,02, 2 = 0,10.
Example:
The equation r = 0,10 + 0,02b1 + 0,02b2 describes a plane in the three dimensional space (with b1 on the x-axis, b2 on the y-axis and r on the z-axis)and
98
DSC3705
describes all assets (and portfolios consisting of the assets) which are influenced by factors f1 and f2 . Thus all assets governed by factors f1 and f2
must lie on this plane. Each asset has its own b1 and b2 values. The price of
risk associated with factors f1 and f2 are indicated by 1 and 2 . When an
asset does not satisfy the equation it is not on the plane and it is mispriced.
For example, suppose P is a portfolio with r = 0,15, b1 = 0,8 and b2 = 1,2.
Thus the expected rate of return of the portfolio according to the market is
0,15. But according to the APT theorem the expected rate of return should
be
rp = 0,10 + 0,02(0,8) + 0,01(1,2) 0,128.
Since the the observed expected rate of return in the market is 0,15 arbitrage
is possible. Find the asset on the plane with the same b1 and b2 values as
the portfolio P (thus satisfying the equation of the plane). Short this asset
which has expected rate of return of 0,128. You receive an amount per unit
and invest this amount in portfolio P which has expected rate of return of
0,15. At the end of the investment period you must replace the shorted asset
at its then value.
Let us assume you receive an amount A from shorting the asset and you
invest this amount A in portfolio P . At the end of one time period the
expected amount you should receive is A(1 + 0,15) = 1,15A. The asset which
you should replace, has expected value A(1 + 0,128) = 1,128A.
The expected risk-free profit is 1,28A 1,15A > 0. More investors will follow
the same strategy of buying P and the rate of return of P will decrease since
the price will increase because of the demand of P .
According to Wikipedia( http://en.wikipedia.org/wiki/Arbitragepricingtheorem):
The APT describes the mechanism of arbitrage whereby investors will bring an asset which is mispriced, according to the
APT model, back into line with its expected price. Note that
under true arbitrage, the investor locks-in a guaranteed payoff,
whereas under APT arbitrage as described below, the investor
locks-in a positive expected payoff. The APT thus assumes arbitrage in expectations - i.e. that arbitrage by investors will
bring asset prices back into line with the returns expected by
the model.
99
10
DSC3705
EXERCISE 5.9.1
Question 1
Luenberger p255, 9.1
Solution
Solve u(x) = E[u(x)] to determine the certainty equivalent.
1
1
1
1
1
1
1
1
1
1
(80 000) 4 + (90 000) 4 + (100 000) 4 + (110 000) 4 + (120 000) 4
7
7
7
7
7
1
1
1
1
+ (130 000) 4 + (140 000) 4
7
7
18,1534
E[u(x)] =
100
DSC3705
Question 3
Luenberger p256, 9.3
Solution
The Arrow-Pratt coefficient of u(x) is a1 (x) =
u (x)
.
u (x)
bu (x)
u (x)
v (x)
=
=
.
The Arrow-Pratt coefficient of v(x) is a2 (x) =
v (x)
bu (x)
u (x)
Hence a1 (x) = a2 (x).
Question 4
Luenberger p 256, 9.4
Solution
Arrow-Pratt relative risk aversion coefficient if u(x) = ln x is:
=
x (x2 )
= 1.
x1
101
A
outcome
0,1
0,3
0,5
B
outcome
0
0,2
0,9
probability
0,3
0,4
0,3
probability
0,3
0,2
0,5
DSC3705
C
outcome
0,2
0,3
0,4
probability
0,45
0,1
0,45
Solution
(a) The utility function is a parabola. The utility function increases if u > 0
u(w) = 3 2w > 0 if 0 < w < 23 .
(b) Arrow-Pratt:
a(w) =
2
3 2w
u (w) dw =
3 2w dw = 3w w 2 + k
102
DSC3705
Solution
V (x) = E[u(
x)] = pu(w + x) + (1 p)u(w x)
= p ln(w + x) + (1 p) ln(w x)
1
1
Hence V (x) = p
+ (1 p)
w+x
wx
Solve V (x) = 0, then x = w(2p 1).
Note V (x) = p(w + x)2 (1 p)(w x)2 < 0 for all x, thus V has a
maximum if x = w(2p 1).
If p
1
2
103
DSC3705
Friend:
E[(a b x2 )di ] = i Pi
thus E[(a b x1 w)di ] = i Pi
then b = b w
b
b =
w
Question 8
Luenberger p257, 9.10
Solution
The random optimal payoff x satisfies
E[u (x )di ] = Pi
Suppose there is a risk-free asset with total return R then di = R and Pi = 1.
Then
= E[u (x)]R
and
E[u (x )di ] = (E[u (x )]R)Pi .
Assume R = 1+rf , with rf fixed. Pi is the current known price. Note that di
i
, thus di = Pi (1+ri ).
(the payoff) and ri are random variables, with ri = diPP
i
Then for each i:
=
=
=
=
=
=
=
0
0
0
0
0
0
0
104
DSC3705
Question 9
Luenberger p 257, 9.11
Hint: Use the state prices (1 ; 2 ; 3 ) in example 9.9 and apply equation
9.15.
Solution
State prices are 1 = 61 , = 21 and 3 = 16 . Suppose the initial price is P , if
state 1 (high success) occurs then outcome d1 = 3 000, if state 2 (moderate
success) occurs then outcome d2 = P , if state 3 (failure) occurs then outcome
d3 = P .
payoff = (d1 ; d2 ; d3 ) = (3 000; P ; P )
0 = 1 + 2 + 3 =
q1 =
1
6
5
6
= 0,2;
q2 =
3
6
5
6
5
1 3 1
+ + =
6 6 6
6
= 0,6;
q3 =
1
6
5
6
= 0,2
E(d)
= 0,2(3000) + 0,6P + 0,2P
Apply formula 9.15:
Assume 0 = price of security with payoff (1; 1; 1). The risk-free return
(risk-free bond) is R = 1,2.
Then
P =
1
1
E(d) =
(0,2(3000) + 0,6P + 0,2P )
R
1,2
and
P = 1 500.
Question 10
Luenberger p257, 9.14
105
DSC3705
Solution
For each $1 the probability is 14 to get $5, and the probability is 34 to receive
zero dollar. Suppose Gavins initial wealth is w. Suppose w is bet on the
horse. If the horse wins, Gavin receives 5w or else he receives nothing.
Thus the final wealth of Gavin is:
state
wins
loses
probability
1
4
3
4
wealth
w + 5w w = w + 4w
w w
1p
3p
w(1 + 4) +
w(1 )
4
4
3
1
w 1 + 4 +
w 1 .
Then V () =
4
4
1
1
1
1
3
1
V () =
w (1 + 4) 2 +
w (1 ) 2 (1)
4
2
4
2
E[u(x)] =
106
11
DSC3705
References
The following source provides an explanation of the lexicon of decisionmaking. Look for the menu Lexicon.
http://faculty.fuqua.duke.edu/daweb/
You could also use a search engine to find the lexicon. Then you should search
for Lexicon of decision-making. Please try to read through this explanation.
The author, Tom Spradlin, provides a short narrative that should clarify
some of the concepts we use in this module.
In the development of this tutorial letter I used the following references. Most
of these books are available in the library.
Prof B Swart, Lecture notes
Anthony M and Norman Biggs. 2000. Mathematics for Economics and
Finance. Cambridge University Press
Aliprantis, CD and Chakrabarti, SK. 2000. Games and Decision Making.
Oxford University Press.
Clemen, RT and Reilly, T. 2001. Making Hard Decisions. Pacific Grove,
Calif: Duxberry.
Elton, EJ and Gruber MJ. 1995. Modern Portfolio Theory and Investment
Analysis. John Wiley and Sons, Inc.
Keeney, R and Raiffa, H. 1976. Decisions with Multiple Objectives: Preferences and Value Tradeoffs New York: Wiley.
Luenberger, DG. 1998. Investment Science Oxford University Press
Ross, SM. 2003. An Elementary introduction to Mathematical Finance New
York: Cambridge University Press.
Tapiero, CS. 2004. Risk and Financial Management John Wiley and Sons,
Ltd.
Van Zandt, T. 2003. Introduction to the Economics of Uncertainty and Information. (This book has not been published yet, however the author gave
us permission to use the book.)
Varian, HR. 1990. Intermediate Microeconomics:A Modern Approach. New
York: WW Norton.
Winston, WL. 1993. Operations Research Applications and Algorithms Belmont, Calif: Duxberry Press.
http://www.stanford.edu/wfsharpe/mia/fac/mia fac2.htm
http://homepage.newschool.edu/het/essays/uncert/vnmaxioms.htm