The Business Cycle: Theories and Evidence

The Business Cycle:
Theories and Evidence

The Business Cycle:
Theories and Evidence
Proceedings of the Sixteenth Annual Economic Policy
Conference ofthe Federal Reserve Bank of St. Louis
edited by
Michael T. 8elongia
Federal Reserve 8ank of St. Louis
and
Michelle R. Garfinkel
Universityof California at Irvine
"
~.
Springer Science+8usiness Media, LLC

Ubrary of Congress CataIoging-in-Publication Data
Economic Pc>icy Conlerence of the Federal Reserve Bank 01 SI. Louis

(16th: 1991: Federal Reserve Bank of SI. Louis)
lhe business cycle: theories and evidence: proceedings 01 the
Sixteenth Annual Economic Policy Conlerence 01 the Federal Reserve
Bank of SI. Louis I edited by Michael T. Belongia and Michelle R.
Garfinkel.
p. cm.
Includes bibliographical references.
ISBN 978-94-010-5312-9 ISBN 978-94-011-2956-5 (eBook)
DOI 10.1007/978-94-011-2956-5
1. Business cycles-United Slates-Congresses. 1. Belongia,
MichaelT. II. Garfinkel, MichelleR., 1960- III. Title.
HB3743.E24 1991
338.5' 42'0973-dc20 92-11616
CIP
Copyright © 1992 Springer Science+Business Media New York

Originally published by Kluwer Academic Publishers in 1992
Softcover reprint of the hardcover 1st edition 1992
AII rights reserved. No pari of this publication may be reproduced,

stored in a retrieval system ar transmitted in any form or by any means,
mechanical, photocopying, recarding, OI otherwise, without the prior
written permission 01 the publisher,Springer Science+Business Media,
LLC.
Printed on ac;d-free pape,.
Set in 1O/I2pt Times Roman

by Graphicralt Typesetters Ud .. Hong Kong
Contents
Contributing Authors vii
President's Message by Thomas C. Melzer ix
Preface xi
Acknowledgments xxi
SESSION I
1
What Is a Business Cycle? 3
Victor Zarnowitz
Commentary by James H. Stock 73

2
The Cycle Before New-Classical Economics 85
David Laidler
Commentary by Ben S. Bernanke 113
SESSION II 119
3
For a Return to Pragmatism 121
Olivier Jean Blanchard
v
vi CONTENTS
4
The Cowles Commission Approach, Real Business Cycle Theories,
and New-Keynesian Economics 133
RayC. Fair
Commentary by Arnold Zellner 148
SESSION III 159
5
How Does It Matter? 161
Benjamin M. Friedman
Commentary: Whatever Happened to Contracyclical Policy? 179

by Michael R. Darby
CONFERENCE OVERVIEW 187
Commentary: Deja Vu All Over Again by Alan S. Blinder 189
Commentary: Business Cycle Developments and the Agenda for Business

Cycle Research by Herschel f. Grossman 197
Commentary: Where Do We Stand? by Michael Parkin 202

Contributing Authors
Ben S. Bernanke Herschel I. Grossman

Princeton University Brown University
Robertson Hall BoxB
Princeton, NJ 08544-1013 Providence, RI 02912
David Laidler
Olivier Jean Blanchard Department of Economics
Department of Economics University of Western Ontario
E52-391 Social Science Centre
Massachusetts Institute of Technology London, Ontario N6A 5C2
Cambridge, MA 02139 Canada
Michael Parkin
Alan S. Blinder Department of Economics
Department of Economics University of Western Ontario
Princeton University Social Science Centre
Princeton, NJ 08544-1021 London, Ontario N6A 5C2
Canada
Michael R. Darby
Andersen School of Management James H. Stock
University of California Harvard University
Los Angeles, CA 90024 79 John F. Kennedy Street
Cambridge, MA 02138
RayC. Fair Victor Zarnowitz

Yale University University of Chicago
Cowles Foundation Graduate School of Business
P.O. Box 2125 Yale Station 1101 East 58th Street
New Haven, CT 06520-2125 Chicago, IL 60637
Arnold Zellner
Benjamin M. Friedman University of Chicago
Harvard University Graduate School of Business
Littauer Center 127 11 01 East 58th Street
Cambridge, MA 02138 Chicago, IL 60637
Vll
President's Message
Asking what we know about business cycles seems, in one sense, a curious
topic for a conference of professional economists. After all, it is not as
though the cycle is a new phenomenon. From the less well documented ups
and downs of the colonial economy through the various booms and busts of
the 1800s, the turbulent decade of the 1920s, and the nine complete cycles
of the postwar period, the U.S. economy has had ample experience with
sustained economic expansion followed by periods of decline. The sum
total of this experience, however, has produced few firm conclusions about
the "whys" or "hows" ofthe business cycle.
To get a little perspective on the issue, I looked back to those days of
the New Frontier when Walter Heller and Co. sat down "to the levers of
control" as a Time magazine cover story described it. The first impression
to catch my eye was the seeming harmony within the economics profession.
Time asserted that "by broadening the areas of fact, the professionalization
[of economics] has narrowed the areas of theory, of disagreement, and
blurred old boundaries between liberals and conservatives." Indeed, the
accepted wisdom of the day seemed to be that the cycle, though a bit of
an annoyance, could be dealt with by the enlightened application of the
appropriate mix of tax and budget policies. This state of affairs is in sharp
contrast to the factional disputes that have dominated at least the last
decade.
This confidence was based, again according to the Time story, not on
the insights of Keynes's General Theory but on Wesley Clair Mitchell's
Business Cycles. This book, labeled as "the most important of all U.S.
contributions to economics," and Mitchell's call for "a body of statistical
information on how the economy actually behaves under the impact of
various policies" were cited as key to transforming economics from "the
IX
x PRESIDENT'S MESSAGE
dismal science" to one of optimism. The new generation of economists was

the first to use econometric models to evaluate the effects of pulling on
different policy levers. They genuinely believed that it was possible to
"tinker successfully" with the economy and "do a lot more good than harm
in the process."
Kenneth Arrow summed up the prevailing view by claiming "you have
to find a real crackpot to get an economist who doesn't accept the principle
of government intervention in the business cycle." If this is what it means
to be a crackpot, I know now why members of our staff here may have
readily accepted the alternative designation, "maverick." They have
argued, and I tend to agree, that the cycle is littered with all kinds of well-
intentioned policies that have had all sorts of important, deleterious, and
quite unanticipated effects.
The present time certainly is a different world from the confident
optimism of the early 1960s. Moreover, the distinctions between liberal
and conservative or activist and laissez faire economists, if they ever were
blurred, now have widened to the point that splits among theories seem to
preclude conversation between economists of different camps.
This is why we chose to devote our conference this year to a discussion
of the cycle. I cannot envision the quality of policy making moving forward
until there is agreement on a common language and until a path to resolv-
ing conflicts among competing theories is identified. Exactly what is this
cycle we are trying to explain? Precisely where do our theories differ? And,
on what evidence can we defend the idea that we are doing more good than
harm by activist meddling with market forces?
Thirty years ago, in the same Time story I mentioned earlier, a panel of
17 economists were asked their views on the then-current recession and the
prospects for recovery. Sixteen gave authoritative statements on when and
why a recovery would begin. The seventeenth, John Kenneth Galbraith,
said: "I do not believe anybody in this Administration makes the preten-
tious mistake of thinking he knows what is going to happen. One of the
greatest pieces of economic wisdom is to know what you do not know." I
hope that these proceedings offer some guidance both on what we know,
and on what we don't know, about business cycles. The former will tell us
how far we have come in the past 30 years; the latter will remind us how
much further we have yet to go.
Thomas C. Melzer
President
Federal Reserve Bank of St. Louis
Preface
These proceedings, from a conference held at the Federal Reserve Bank of

St. Louis on October 17-18, 1991, attempted to layout what we currently
know about aggregate economic fluctuations. Identifying what we know
inevitably reveals what we do not know about such fluctuations as well.
From the vantage point of where the conference's participants view our
current understanding to be, these proceedings can be seen as suggesting
an agenda for further research.
The conference was divided into five sections. It began with the formu-
lation of an empirical definition of the "business cycle" and a recitation of
the stylized facts that must be explained by any theory that purports to
capture the business cycle's essence. After outlining the historical develop-
ment and key features of the current "theories" of business cycles, the
conference evaluated these theories on the basis of their ability to explain
the facts. Included in this evaluation was a discussion of whether (and how)
the competing theories could be distinguished empirically. The conference
then examined the implications for policy of what is known and not known
about business cycles. A panel discussion closed the conference, high-
lighting important unresolved theoretical and empirical issues that should
be taken up in future business cycle research.
What Is a Business Cycle?
Before gaining a genuine understanding of business cycles, economists

must agree and be clear about what they mean when they refer to the cycle.
In the keynote chapter, "What Is a Business Cycle?" Victor Zarnowitz
takes on the difficult task of defining a business cycle. Approaching this
Xl
xu PREFACE
task from several perspectives, Zarnowitz argues that the business cycle is
"pervasive." Although business cycles historically have differed in their
duration and intensity, they are all generally characterized by a decline and
contraction and a subsequent rise and expansion of aggregate economic
activity, as measured by total employment, output, real income, and real
expenditures. National in scope and typically lasting several years, business
cycles manifest themselves in the co-movements of and interactions among
many economic variables. Not all variables are perfectly synchronized,
however. Some lead and others follow the cycle. In addition, while most
economic variables are procyclical, they do not generally move with the
cycle to the same degree; other variables are countercyclical.
Zarnowitz also finds that the characteristics of business cycles have
changed over time and differ across nations. For example, Zarnowitz finds
that, in the United States, postwar cycles have become milder: Business
contractions have become shorter and less severe, and business expansions
have become longer relative to the cycles experienced before the 1930s. As
possible reasons for this observed change, Zarnowitz suggests "the shift of
employment to production of services, automatic stabilizers, some financial
reforms and avoidance of crises, greater weight and some successes of
government actions and policies, and higher levels of public confidence."
He also finds that the postwar recessions in France, Italy, West Germany,
and Japan were even milder and attributes this difference to the relative
strength of their economic growth.
Thus, in trying to explain cycles, we must be guided not only by the
features common to all business cycles but also by their diversity and
evolution. By limiting the focus to features shared by all cycles, economic
analyses potentially fail to gain a comprehensive understanding of aggre-
gate economic tluctutations. The processes by which shocks affect eco-
nomic activity can depend on a nation's institutions and stage of economic
development and thus could be changing also.
James Stock, in his comments on this chapter, commends Zarnowitz for
his useful and thorough overview of the business cycle, including his ana-
lysis of international business cycles. In his discussion, Stock focuses on
only two main points. First, by drawing on recent research, he argues that
Zarnowitz's evidence of the shortening of postwar recessions and lengthen-
ing of postwar expansions might be "biased" by the National Bureau of
Research's (NBER) dating chronology, which is based on different time
series depending on the historical period: "The prewar dating relied on
series with more cycles, longer contractions, and shorter expansions than
the series used for the postwar dating." Stock notes that, alternatively,
the differences in the series used might accurately capture fundamental
PREFACE xiii
changes in economic activity, so that the NBER's dating methodology does

not create a bias. While there is evidence that weakens the credibility of
this alternative interpretation, Stock argues that additional research is
needed to resolve the issue satisfactorily.
Stock's second point builds on one of Zarnowitz's main themes-that
historical business cycles are not identical but, rather, have evolved over
time. Stock suggests that case study analyses of individual cycles are likely
to yield new and significant insights into the sources of cycles and their
propagation mechanisms as they have changed with the evolution of busi-
ness cycles. The usefulness of this approach is illustrated by examining the
1990 recession and highlighting how it was unlike any other postwar
recession. Stock's evidence indicates that "historical correlations between
leading indicators and overall economic activity were not good guides
to this episode." In contrast to other recent U.S. recessions, monetary
indicators played a small role in the 1990 episode. Rather, the data suggest
an important role for "unprecedented shifts in consumer expectations."
What Are the Sources of Business Cycles?
Despite the voluminous research on fluctuations in aggregate economic

activity, there appears to be nothing resembling a consensus among
economists and policy makers about the sources of such fluctuations.
Although a consensus about macroeconomic fluctuations and their impli-
cations emerged in the 1950s, that consensus was short-lived. Differing
opinions were sparked by theoretical and empirical advances in the late
1960s, culminating in a debate between the Keynesians and monetarists.
Subsequent theoretical research attempting to establish the micro-
foundations of macroeconomic phenomena eventually shifted the focus of
the debate to one between the new-Keynesians and the real business cycle
theorists. The next two chapters trace the evolution of these debates and
summarize their contributions to our understanding of why cycles occur.
In "The Cycle Before the New-Classical Economics," David Laidler
describes the evolution of the debate between the Keynesians and
monetarists. The consensus that remained intact in the 1950s and early
1960s downplayed the relevance of monetary factors in explaining macro-
economic fluctuations, emphasizing instead autonomous fluctuations
in investment. Building on this general acceptance of Keynesian econo-
mics, business cycle research continued to make progress, incorporating
the notion of the accelerator into the Keynesian framework to explain not
only how fluctuations in investment affect aggregate demand but also how
xiv PREFACE
the economy's productive capacity affects investment and vice versa. As

Laidler points out, the theories produced by this research interpret busi-
ness cycles as real rather than monetary phenomena, in the spirit of
Keynesian economics. Laidler argues that the eventual demise of this view
of business cycles, which had integrated growth theory with business cycle
theory, was driven by subsequent advances in the mainstream of macro-
economic research that undermined some of the theory's key assumptions.
Monetarism, which had gained popularity as a theory about cycles
mainly because it filled the gap in macroeconomics opened by the demise
of the multiplier-accelerator models of the business cycle, revitalized
the empirical search for exogenous impulses of cycles. At the same time,
Keynesians focused on developing and implementing large econometric
models to identify the propagation mechanisms of cycles. The specific
economic relationship that became the focal point of the debate was the
Phillips curve, depicting a trade-off between nominal wage growth and
unemployment-a relationship that was quickly absorbed into mainstream
Keynesian economics. Monetarists, however, argued that the Phillips curve
was inadequate as a long-run representation and that what matters in labor
decisions is not the nominal wage but the real wage. This criticism was
reinforced and extended by subsequent theoretical and empirical research,
later referred to as "new-classical" economics, stressing the importance
of microfoundations and internal consistency. Laidler argues that, while
new-classical economics, in essence, drew attention away from the debate
between the monetarists and Keynesians, the latter two views began to con-
verge, with the remaining differences being quantitative, not qualitative.
In Laidler's view, this tum of events has proved to be unfortunate.
Researchers adopted new modeling strategies that they considered
theoretically appealing-that is, the application of general equilibrium
theory to explain macroeconomic phenomena-without conforming to the
facts (for example, price and wage stickiness), almost as though the beauty
of the theory (derived from first principles) was sufficient to justify this
approach. Laidler suggests, for future research, moving away from that
strategy and toward an empirical challenge to the existing theories of busi-
ness cycles.
In his discussion of this chapter, Ben Bernanke first focuses on Laidler's
list of contributions by the monetarists. Agreeing with Laidler, he notes
that one of the major contributions was to document the empirical
relevance of money in economic fluctuations. Bernanke adds emphasis
to the monetarist recognition of the now well-accepted notion that, in the
long run, the unemployment rate equals the natural rate, and stresses the
PREFACE xv
monetarist prescription for nonactivist policy or rules over activist policy or

discretion. Second, Bernanke highlights the similarities and differences
between the monetarists' and the real business cycle theorists' challenges
to the Keynesian paradigm, as well as their respective contributions. To
add a favorable note to the real business cycle approach, while admitting
that it has not been overwhelmingly fruitful, Bernanke suggests that, in
time, this general equilibrium approach could offer much insight to our
understanding of the cycle. Thus, in contrast to Laidler, Bernanke reserves
judgment until this methodology matures sufficiently to address macro-
economic issues.
In "For a Return to Pragmatism," Olivier Blanchard describes the pro-
gress made by macroeconomic researchers during the past two decades,
taking the new-classical challenge to "macroeconomics, circa 1970" as a
starting point. 1 According to Blanchard, the new-classical challenge, calling
for theoretical reconstruction and thereby producing a "back-to-basics"
mentality, was met in two phases. First, researchers incorporated the
rational expectations hypothesis and the potential importance of supply
shocks into their analyses. Second, starting roughly a decade ago, research-
ers began to analyze the structure of different markets more closely in an
effort to provide the mainstream view of macroeconomics (circa 1970)
with a stronger theoretical foundation. This effort was directed primarily
to learning more about the processes of wage and price determination.
Blanchard questions whether the advances made here constitute
important contributions to our understanding. Although Blanchard is
impressed by the list of advances made, he is disillusioned by their lack of
cohesiveness, by the general failure of researchers to extend their analyses
to capture the increasing integration of national economies, and by the
scarcity of attempts to test the theories empirically or extend them to
explain the cyclical "facts." Like Laidler, Blanchard argues that the wave
of research in the past two decades, in its narrow "back-to-basics" mode,
has provided little insight into contemporary real-world issues and policy
problems. His prescription for the profession is to "adopt a more flexible,
more pragmatic approach to research."
Like Laidler, Blanchard argues that the wave of the new-Keynesian
and real business cycle research has provided little insight into contempor-
ary real-world policy problems. He also agrees with Laidler's prescription
for the profession to get back to basics. In addition, he says, the wave has
pushed econometrics in the wrong direction, away from the fundamental
issues about characterizing the data structurally while handling simul-
taneity of the determination of economic variables. Blanchard encourages
xvi PREFACE
macroeconomists to adopt a common framework, one that sees demand

shocks from home and abroad as dominant in explaining macroeconomic
fluctuations, accounts for the changing nature of the shocks that generate
the fluctuations, and explicitly allows for real if not nominal rigidities. Such
a framework should permit an analysis of the interaction among techno-
logical change, recessions, and unemployment. Blanchard also draws atten-
tion to three areas of research that he believes should be addressed in this
framework: the functioning of labor, goods, and credit markets.
How Might the Debate Be Resolved?
One of the overriding concerns among many conference participants was

the movement away from empirical research in macroeconomics during
the past two decades. In the absence of rigorous testing of the empirical
validity of each of the competing theories, the debate concerning the
sources and implications of business fluctuations appears to be vacuous. In
"The Cowles Commission Approach, Real Business Cycle Theories, and
New-Keynesian Economics," Ray Fair articulates this concern, using the
Cowles Commission approach as a standard for judgment. This approach
translates theoretical models into econometric ones that, in turn, enable
estimation, testing, and prediction. He argues that the Lucas critique,
which generated excitement about searching for the deep, structural
parameters that are invariant to policy changes, might have led the
profession astray, making it less able to deal with more difficult problems
than those posed by the critique itself. While admitting that the critique is
correct, Fair points out that its empirical significance remains to be seen.
Testing of real business cycle models is difficult at best and thus will not
fare well in competing with other models. Similarly, he criticizes the new-
Keynesian approach for its failure to combine theories to produce a full
model of the economy. This failure precludes the line of research from
model estimation, testing, and prediction.
Given the lack of empirical content in both approaches, which have
dominated macroeconomic research for nearly two decades, Fair is pessim-
istic that the debate will be resolved satisfactorily. He does, however,
believe that empirical testing to distinguish the theories is possible. To do
so would require more sophisticated and thorough testing and more eco-
nometric model building, by both the new-Keynesians and the real busi-
ness cycle theorists, taking the approach of the Cowles Commission.
Arnold Zellner expresses general agreement with most of Fair's points.
He is skeptical, however, about the fruitfulness of Fair's proposal to resolve
the debate between the new-Keynesians and the real business cycle
PREFACE xvii
theorists empirically. Zellner points out, for example, that the actual struc-
ture of the economy is subject to much uncertainty. In addition, the
estimated coefficients obtained from the Cowles Commission approach
are not constant over time. Moreover, they are likely to change not only
because of changes in policy, but also because of adaptive optimization
by economic agents responding to large aggregate shocks. Zellner also
expresses concern about the problems of aggregation, regime changes,
measurement errors, and seasonal adjustments in the data.
He suggests, as an alternative to the Cowles Commission approach and
fixed-parameter models, what he has dubbed the "structural econometric
modeling, time series analysis" (SEMTSA) approach. With this approach,
the researcher posits a tentative structural model based on theory and, from
that model, derives "transfer" functions. In going back and forth between
the structural formulation of the model and the transfer functions in
estimation, the researcher ensures compatibility, thoroughly tests the
structural model's components, and compiles them in a reasonable way
to formulate a model. The result is a final model that can be tested further.
Zellner believes that the SEMTSA approach might be quite successful in
producing an empirical model that can explain the essential characteristics
of business cycles and predict future outcomes with satisfactory precision.
What Are the Policy Implications?
Most economists would agree with the notion that understanding the
causes and mechanisms of business cycles is necessary in making informed
policy decisions. The issue receiving considerable attention in recent policy
debates, however, is more fundamental-namely, whether macroeconomic
policy could possibly enhance social welfare. Without sufficient empirical
evidence to resolve the debate among competing camps of different
normative (as well as positive) prior beliefs, one naturally wonders whether
macroeconomics has anything to offer policy makers in the way of useful
advice.
In "How Does It Matter?" Benjamin Friedman expresses his doubt that
recent research on business cycles can be of much help to policy makers.
His doubt is not driven by the lack of consensus among economists, however.
Indeed, he argues that the principal distinction between new-Keynesian
economics and real business cycle theories has less relevance for policy
than is commonly thought. That is to say, whether the sources of business
cycles originate on the demand side or the supply side of the economy
might not matter in deciding whether a policy response would be welfare-
improving. The reason is that exogenous shocks to the supply side typically
xviii PREFACE
generate shocks to the demand side, which often cannot be disentangled

empirically or theoretically. Instead, Friedman argues, the distinction that
is important for policy is whether the key assumptions of the real business
cycle theories are consistent with the facts. Because the economy is better
characterized by Walrasian rigidities and market imperfections rather than
perfectly flexible prices and wages, frictionless markets, and nondistor-
tionary taxes, Friedman argues, "there is room for macroeconomic policy
to respond to even the purest of "supply shocks." Even then, Friedman
argues, the existing theories have little to say about what policies should be
implemented in response to exogenous shocks because they fail to incor-
porate heterogeneity of the population and thereby abstracts from the
distributional questions that it raises.
Echoing both Laidler and Blanchard, Friedman also feels that research
in business cycles has sacrificed relevance in its search for elegance.
Specifically, the theories are incapable of addressing pragmatic policy
issues about the differing effects on the welfare of various economic agents.
Given that aggregate shocks generate "winners" and "losers," the notion
of Pareto improvement "loses its practical relevance." Friedman argues
that, to be able to provide reasonable guides for policy, theories of business
cycles must account for heterogeneity in the population. In addition,
Friedman suggests that economists should examine more closely the role of
policy to influence expectations. In models exhibiting multiple equilibria,
policy's influence on expectations could aid economic <!gents to coordinate
on the best equilibrium.
While appreciating the strength of Friedman's arguments, Michael
Darby challenges Friedman's view that monetary policy should pursue
"contracyclical" or stabilization policies. In Darby's view, contracyclical
policy promises a "cure that in actuality makes things worse." To be effec-
tive, such a policy requires a guide to determine the magnitude, direction,
and timing of the policy response. Without such a guide and with insti-
tutional constraints that limit the role of fiscal policy, policies intended to
stabilize aggregate income are likely to have destablilizing effects. Darby
emphasizes that policy should avoid the temptation to stabilize income and
aim for longer-term strategies and goals. such as price-level stability.
Where Do We Stand?
With so many unsettled issues in the study of business cycles, it might
appear that the foundation on which future research should build is weak.
That there is much room for further progress is hardly arguable. In the
panel discussion closing this conference the participants offered their views
PREFACE xix
on where we now stand and what areas of inquiry would likely prove most
fruitful in future research.
Alan Blinder, in "Deja Vu All Over Again," expresses much agreement
with Laidler, Fair, and especially Blanchard in arguing that macro-
economic research since 1972 has made little progress in resolving the
controversial issues about how the economy works. In addition, Blinder
emphasizes, the "fact that such issues remain so controversial for so long
is a condemnation of our profession." Although he is not inclined to
condemn the new-Keynesians for the "nonempirical flavor" of their
research, Blinder does question the real business cycle models for their
lack of empirical basis. After reviewing the historical development of
macroeconomic models, starting with the early Keynesians, he offers a
"new, post-RBC consensus," which is hardly distinguishable from that
reached in 1972 with the resolution of the positive issues of concern in the
Keynesian-monetarist debate. The primary differences include the recog-
nition of the importance of supply shocks and the general acceptance of
rational expectations by macroeconomists. Blinder argues, in agreement
with Fair and Laidler, that macroeconomists should return to empirical
work, "the task that was so unfortunately abandoned in 1972."
In "Business Cycle Developments and the Agenda for Business Cycle
Research," Herschel Grossman discusses how the evolution of research in
this area was influenced by historical business cycle developments. These
developments, in turn, suggest where it should go in the future. Before the
mid-1970s, business cycle research centered on understanding the determi-
nation of nominal aggregate demand, its relation to real aggregate demand,
and policies to stabilize nominal aggregate demand. The first development
that influenced the direction of research, in Grossman's mind, was the oil
price shock of 1973, which suggested that a possibly important source of
economic fluctuations could be exogenous changes in resource endow-
ments. He interprets the U.S. recessions of 1974-1975 and 1981-1982,
as well as the recession of 1991, with their uneven regional and sectoral
effects, as suggesting that research also should examine the determina-
tion of economic activity on regional and sectoral levels. In addition, exper-
ience in recent years suggests that the Federal Reserve System is able and
willing to stabilize nominal aggregate demand without attempting to go
beyond the nation's resource constraints. Grossman points out, neverthe-
less, that there is little to ensure that the Fed will continue to "give priority
to this objective" indefinitely, for neither the preferences of our policy
makers nor the institutions that constrain their choices have changed.
Thus, he emphasizes, the events of the past two decades suggest that
research efforts should be directed not only toward the implications of
xx PREFACE
aggregate supply shocks and the cyclical aspects of economic activity from
a regional and sectoral perspective, but also toward enhancing our "under-
standing of the political-economic structure underlying monetary and fiscal
policy."
Michael Parkin, in "Where Do We Stand?" offers a characterization of
the current debate quite different from that of Blanchard, suggesting that
the current path of macroeconomic research might not be just a dead
end. Parkin argues that the criticism against the new-classical research
program-that it merely aims to "build beautiful models"-is not well
founded. The research program currently being pursued by young macro-
economists aims to resolve economic puzzles by building models, with
the ultimate goal of sUbjecting the models to empirical verification. In his
view, the important distinctions between the new-Keynesians and the
new-classical economists do not concern with methodology and ideology.
Rather, they are "based on views about which abstractions might turn
out to be useful and which might not." The new-classical economists
emphasize intertemporal substitution to explain economic fluctuations.
new-Keynesians, while accepting the importance of intertemporal sub-
stitution, emphasize problems of coordination and rigidities to explain
economic fluctuations. But the new-classical economists do not deny the
existence of these problems, and their modeling strategy does not preclude
the incorporation of these features of the economy. Parkin suggests,
in addition, that the empirical evidence offered by the new-classical
economists should not be dismissed because of its lack of conformity with
traditional empirical analyses, for that evidence is consistent with the view
that "policy is a process, not an event." But, without a feasible commit-
ment technology, this process is a discretionary one. Thus, Parkin believes
that macroeconomists in the future should study alternative institutional
arrangements to determine which would best stabilize the economy.
Over all, these proceedings offer many reasons for macroeconomists to
feel humble about their work. At the same time, they contain a great deal
of constructive criticism about which avenues of future research are most
likely to produce useful results. Macroeconomists from all schools of
thought should find plenty of stimulus in this volume for their own research
agendas.
Note
1. It should be noted that Blanchard's chapter in this volume differs considerably from the
paper he presented at the conference. This earlier paper, entitled "The New Classicals and
the New Keynesians: The Long Pause" included comments on a variety of topics omitted in
the published version. We regret that, due to time constraints, not all of conference par·
ticipants were able to incorporate these changes in their own contributions to this volume.
Acknowledgments
The Bank's annual economic policy conference is organized by its Research

and Public Information Department. Special thanks are due, however,
to several people who did the lion's share of the work. Linda Moser co-
ordinated all local arrangements, ranging from the mailing list to menus
and hotel accommodations. Daniel Brennan helped edit the proceedings
and worked closely with Zachary Rolnik and Judy Pereira of Kluwer
Academic Publishers to produce this volume. Finally, Peggy Dooley
typed, and kept organized, all of the conference correspondence in her
usual conscientious and timely manner.
XXI
SESSION I
1 WHAT IS A BUSINESS CYCLE?
Victor Zarnowitz
Sober men, whose projects have been disproportioned to their capitals, are
as likely to have neither wherewithal to buy money, nor credit to borrow it,
as prodigals whose expence has been disproportioned to their revenue.
Before their projects can be brought to bear, their stock is gone, and their
credit with it. ... When the profits of trade happen to be greater than ordi-
nary, overtrading becomes a general error both among great and small
dealers.
-Adam Smith (1776, p. 406)
When prices fall, production is arrested until the expences of production fall
in equal degree; and whilst production is thus arrested, consumption is
also diminished . .. the inducements to employ labour . .. are diminished
... and the prices of labour fall. The consumption of labour is thus dimin-
ished., and the prices of property again fall, and again act in depressing
labour, and in crippling production. ... It is the deficiency of money
which has occasioned the depression ofprices. ...
-Thomas Attwood (1817, pp. 99, 101)
We find {the state of trade] subject to various conditions which are

periodically returning: it revolves apparently in an established cycle.
First we find it in a state of quiescence,-next improvement,-growing
3
4 THE BUSINESS CYCLE
confidence, -prosperity, -excitement, -overtrading, -convulsion, -pres-

sure,-stagnation,-distress,-ending again in quiescence.
-Lord Overstone (1857, p. 44)
But though men have the power to purchase they may not choose to use it.
For when confidence has been shaken by failures, capital cannot be got to
start new companies or extend old ones. ... In short there is little occu-
pation in any of the trades which make Fixed capital. ... Other trades,
finding a poor market for their goods, produce less; they earn less, and
therefore they buy less. ... Thus commercial disorganization spreads. ...
The chief cause of the evil is a want of confidence.
-Alfred and Mary Marshall (1881, pp. 154-155)
1 The Emergence of Business Cycles in History and

Economic Thought
1. 1 Early ObselVations and Interpretations
Well before the concept of a business (or trade) cycle originated, serious
episodes of commercial and financial instability were observed repeatedly
by contemporaries. The preceding quotations, from Adam Smith (1776) to
Alfred Marshall (1881), illustrate the reactions of some of the classical
economists and politicians in England. Smith's brief mention of "over-
trading" is the only one in The Wealth of Nations'! Attwood, a banker
and politician, blamed reductions in the money supply under the gold stan-
dard for the resulting deflation and interacting declines in spending and
incomes (see Link, 1958, pp. 6-35; Backhouse, 1988, pp. 134-34). The
"famous words" of Lord Overstone, who may have been the first to write
about a multistage "cycle of trade," were cited with approval by Marshall a
quarter-century later. Marshall's stress of the confidence factor recalls not
only Pigou of 1929 but also Keynes of 1936 (Marshall and Marshall, 1881).
There is much in these and other early theories of crises and cycles that
deserves to be rediscovered and reconsidered today.2
Because of the predominance of classical tradition, problems of long-
term equilibrium constituted the principal concern of the prominent
theorists in the nineteenth century, and short-term business cycle problems
were at best a secondary interest. Yet throughout that century, from
Sismondi and Malthus to Marx and Hobson, intense controversies
prevailed about the validity of Say's law, that supply creates its own
demand, and about the possibility of a "general glut" (Sowell, 1972). There
WHAT IS A BUSINESS CYCLE? 5
can be little doubt that this controversy was so great because of the pres-
sure of events: the need to account for the recurrent "crises." the related
decreases in sales and profits. and the increases in unemployment.
A study of monetary statistics for France and England (since 1800) and
the United States (since 1836) led Juglar (1889) to believe that crises are
merely stages in recurrent business cycles. Although the title of his work
refers to the "periodic return" of the crises. the durations of his cycles vary
considerably.3 According to Juglar. cycles are principally a feature of eco-
nomies with highly developed commerce and industry. division of labor.
external trade, and the use of credit. This idea was accepted and developed
further by most of the leading scholars in the field.
The association of business cycles with modem capitalism is reflected in
the National Bureau of Economic Research (NBER) reference chrono-
logies. which begin (for annual dates of peaks and troughs) in 1792.1834.
1840, 1866 for Great Britain, the United States. France, and Germany,
respectively.4 There is substantial agreement that wars, poor harvests, and
other episodic disturbances played a greater relative role in the preindustri-
alization era than afterward, and the endogenous cyclical influences a
smaller role. The main reason, presumably, is the pronounced cyclicality of
investment in "fixed" capital-plant, machinery, and equipment. It was
only after the great technological innovations of the latter part of the eight-
eenth century were adopted that the size of such investment, and its share
in output, began to increase rapidly, in England, then in Western Europe
and in the United States. In earlier times, processing of raw materials by
labor must have represented a much larger part of total production. It is
likely, therefore, that fluctuations of trade and inventory investment
accounted for most of the overall economic instability prior to the Indus-
trial Revolution. But these movements tend to be shorter in duration and
more random in nature than the fluctuations related to investment in busi-
ness plant and equipment.5
Differences of opinion still persist on when modem business cycles came
into being. Schumpeter (1939. vol. 1, chap. 6B) argues that capitalism goes
"as far back as the element of credit creation" and that "there must have
been also prosperities and depressions of the cyclical type" in the seven-
teenth and eighteenth centuries. 6 Indeed. there is a modest amount of
evidence on how harvests, grain prices, exports, imports, and sales and
profits of a small sample of enterprises changed from year to year in Great
Britain between 1720 and 1800. Four economic historians compiled chrono-
logies of commercial crises for this period that show a fair amount of con-
sensus. There may have been six or eight major contractions of credit and
several more downturns of profit. 7 The problem is not with the existence
of some substantial fluctuations but with their nature. How comparable are
they with the later business cycles? Unfortunately, there is no conclusive
answer because of severe limitations in the available data.
1.2 Popular Meaning
Unlike a theoretical concept such as that of equilibrium, a business cycle is

principally an empirical phenomenon founded upon historical experience.
People engaged in business and public affairs have long observed that
economic conditions are generally satisfactory or good much of the time
but weak or bad some of the time. An old term for the good times was
prosperity, an old term for the bad times was depression. The transition
from the former to the latter used to be called a crisis (and occasionally a
recession, but this designation came to be reserved for a mildly depressed
period and is so applied presently). The transition from depression to pros-
perity was called revival (today the term recovery is more common). It
seems fair to say that these broad descriptions soon acquired a rather good
intuitive meaning and wide acceptance in use.
The evidence comes from hundreds of official documents and contem-
porary reports in which such characterizations of the prevailing business
situations are found over long sequences of years for market economies
at various stages of development and industrialization. The terms denote
phases of what later came to be called business (in England, trade) cycles,
and their context was from the outset an economy with increasing popu-
lation, labor force, productive capacity, and output. In the nineteenth
century, long-term economic growth trends, like the fluctuations around
them, became a widely recognized empirical phenomenon. Thus, pros-
perity soon acquired the connotation of an expansion and depression that
of a contraction, although these terms came into use only later and were
given a more explicit technical meaning by economists. Depressions were
seen clearly as distressful interruptions of growth that was necessary to
maintain prosperity. The dynamic nature of the perceived processes is well
documented by the qualifiers frequently attached to the descriptions of the
state of the economy. For example, a depression might be "mild" or
"severe," a revival might be "slow" or "rapid," and so on.
Business annals based on a large collection of such materials were
compiled by the NBER as one of its earliest projects (Thorp, 1926). They
provide comprehensive qualitative records of general economic changes in
a number of countries between 1790 and 1925.8 The construction of the
historical NBER business cycle chronologies for the United States, Great
Britain, France, and Germany was accomplished by a complementary use

of the business annals and quantitative information drawn from a large
collection of time series on aspects of general economic activity, price
indexes, monetary and financial statistics, and the like (Bums and Mitchell,
1946). On the whole, a good agreement was found between the dates
of the business cycle peaks and troughs suggested by the annals and the
dates based on the statistical time series (for evidence, see Mitchell, 1927,
pp.20-31).
That many contemporaries were able to diagnose business conditions
rather well in the face of great limitations of public data on the economy is
a remarkable fact and a testimony to the persistence and pervasiveness
of business cycles. It can take many months for downturns and declines
to spread through the economy and for their effects to grow and recede.
Such developments hurt the well-being of people in many occupations,
industries, and regions and are therefore generally apprehended by
them, although often with lengthy lags and great uncertainty. A similarly
extended recognition process applies to economic upturns and rises.
Business people track their sales and profits and learn when they are high
or low, increasing or decreasing. Workers learn when jobs and wages are
easy and when they are hard to get. Consumers know how much money they
are able and willing to spend. One would expect sharp turning points
and strong, sustained expansions or contractions to be recognized more
promptly and reliably than gradual transitions and weak, erratic upward or
downward movements. 9
1.3 Economists'Definitions
It may be easy to recognize some manifestations of a business cycle,

particularly the more extreme ones, but the history of events and ideas
suggests that it is anything but easy to define what a business cycle actually
is. lD In recording sharp economic fluctuations, contemporaries and his-
torians alike observe first the diverse disturbances, from bad weather to
political upheavals and speculative manias and panics, that seem to account
for crises and downturns. The concentration on isolated episodes and
outside shocks precludes a definition of the entire fluctuation or cycle as an
economic phenomenon and analytical target.
In contrast, economic theorists who accept the existence of general
economic fluctuations are interested in the concept of "the" business cycle
and its main cause or causes. In some cases the theory appears to rule the
definition. For example, according to Cassel, "A period of boom is one of
special increase in the production of fixed capital; a period of decline or a

depression is one in which this production falls below the point it had
previously reached .... This means that the alternation between periods of
boom and slump is fundamentally a variation in the production of fixed
capital, but has no direct connection with the rest of production" (see 1932,
pp. 550, 552). Cassel, like Tugan-Baranovskii and Spiethoff before him,
believed that changes in cost and value of capital goods are the main force
driving the cyclical motion of the economy.
As another example, consider Hawtrey's statement that "the trade cycle
is above all a periodical fluctuation in manufacturing activity and in the
price level, the two fluctuating together" (1927, p. 471). Hawtrey's theory
stressed the role of movements in bank credit, inventory investment, and
prices in the relatively regular cycles of the pre-World War I gold standard
era.
The search of literature for definitions of business cycles is on the whole
frustrating. Most authors do not formulate such definitions by identifying
empirically the principal "stylized facts" about what happens during expan-
sions, downturns, contractions, and upturns. However, Mitchell proceeded
along this way and arrived at a tentative working definition near the end of
his 1927 volume (p. 468). With "some modifications suggested by experi-
ence in using it," that definition was restated by Bums and Mitchell (1946,
p.3):
Business cycles are a type of fluctuation found in the aggregate economic
activity of nations that organize their work mainly in business enterprises: a
cycle consists of expansions occurring at about the same time in many economic
activities, followed by similarly general recessions, contractions, and revivals
which merge into the expansion phase of the next cycle; this sequence of
changes is recurrent but not periodic; in duration business cycles vary from more
than one year to ten or twelve years; they are not divisible into shorter cycles of
similar character with amplitudes approximating their own.
The main point here is the co-movement of many economic variables or
processes, which occurs with a rough synchronism (allowing for leads and
lags of mostly moderate length) in the course of any business cycle. What
matters is that many diverse activities tend to expand and contract
together; also, it should be added, that they evolve over time and cannot
be reduced to any single aggregate (Moore and Zarnowitz, 1986, p. 737).
Hence, the question of what precisely constitutes "aggregate economic
activity" is purposely and properly left open. The nature of business cycles
depends on, and changes with, the major characteristics of the economy,
society, and polity. The most common and salient feature of business cycles
is their pervasiveness and persistence (the high cyclical conformity or
coherence of numerous variables and their pronounced serial correlation).

It is not the fluctuation of any single aggregate, however important.
In support of this argument, note that in the peacetime cycle before
World War II, prices as well as quantities of industrial products tended to
increase in expansions and decrease in contractions, which implies that
nominal values of total output and income would have had larger
amplitudes of procyclical movement than the real values. In contrast, in
more recent times the overall level of prices continued to increase during
contractions, albeit often at a slower pace (lower rates of inflation). Hence,
real GNP is now a much more sensitive measure of cyclical changes in
"aggregate economic activity" than nominal GNP is-whereas historically
the opposite must have often been the case.
Of course, no statistics on GNP in either current or constant dollars
were available before World War I, and none appeared on any regular
basis in the interwar period. In fact, no single comprehensive measure
of aggregate economic activity, with sufficient comparability of coverage
and validity of estimation, exists for a long stretch of historical time.
This applies to annual data and, still more so, to monthly and quarterly
series that are much needed for the study of business cycles (as well as
of short-term economic changes in general). It is therefore not only
conceptually desirable but also statistically necessary to use a number of
indicators rather than any single one. To be sure, they should include the
most reliable and comprehensive time series data that can be had. In prac-
tice, for periods before 1914 or 1929, series representing various activities
of production, consumption, investment, trade, and finance had to be
collected and used as composite indexes of business activity, along with
price indexes and other useful aggregates such as bank clearings and
debits.ll
The Burns-Mitchell definition refers to a broad range of durations (1-12
years), thereby accommodating both short and long cycles. It includes no
strict amplitude criterion, but it rules out accepting movements smaller
than usual as cyclical (this is roughly the intended effect of the last clause
of the definition; see Burns and Mitchell, 1946, pp. 7-8). No distinction
between major and minor cycles is made. The definition admits equally
vigorous and weak expansions, severe and mild contractions.
1.4 Past Views on the Scope and Nature of Business Cycles
All descriptions and definitions of business cycles are guided in part by

some theoretical ideas and in part by some observations and measures
of what is going on in the economy. The data available to the early

generations of students of the subject were meager indeed. JugJar, in the
first large-scale work on the history of commercial crises and cycles (1862),
assembled mainly monetary, banking, and price statistics. He showed that
discounts and deposits were high in crisis (peak) years and low in liqui-
dation (trough) years and that the reverse was true for bullion reserves.
Prices rose before and fell after a crisis. Later, those economists who
stressed the cyclical role of business investment in plant and equipment
used various proxies, such as output of pig iron for production of capital
goods and output of coal for total nonagricultural production (Tugan-
Baranovskii, 1901; Cassel, 1932).
Note that the emphasis on some selected factors and data need not
imply a view of business cycles as movements that are narrowly defined
in spatial or temporal terms. On the scope and nature of the cycles, some
economists have held widely different opinions. Referring to "cyclical
fluctuations during the period 1870-1914," Cassel stated that they "have
no absolute necessity, but are to a great extent caused by factors which
represent passing phenomena of economic history, or which may be, if not
eliminated, at least to a great extent controlled" (1932, p. 538). In contrast,
according to Robertson (1937, p. 171), "In industrial fluctuation we are up
against a problem very deep-seated in the nature of capitalist industry
-perhaps of all modem industry-perhaps of man himself. I do not be-
lieve myself that we can solve it."
Overall, a perusal of the literature leaves a strong impression that most
of the principal writers on business cycles recognized the large dimensions
and importance of their problem and task. The interdependence of all parts
of the economy, which essentially accounts for the wide diffusion of cycli-
cal movements from any initial source, was well understood long before the
advent of modem macroeconomic analysis. Indeed, the latter in some ways
narrowed the approach to the study of business cycles, making it generally
more aggregative and often also less dynamic. The pioneers in this field had
a lively appreciation of the importance of endogenous cyclical processes
and their connection with long-term economic growth and development.
Many later contributions show less scope and vision in these respects. To
be sure, the more recent work also displays the major benefits of using much
more abundant, comprehensive, and accurate data as well as new tech-
niques and insights of modern stochastic and econometric approaches. 12
Those scholars who engaged in large-scale empirical research on the
subject have been particularly insistent on seeing business cycles as com-
plex phenomena of great importance: not only the main form of economic
instability but also the uneven path of economic growth. Thus, to Burns and
Mitchell, "The problem of how business cycles come about is ... insepar-
able from the problem of how a capitalist economy functions." They were
"not content to focus analysis on the fluctuations of one or two great
variables, such as production or employment" but "sought to interpret the
system of business as a whole ... to penetrate the facade of business
aggregates and trace the detailed processes-psychological, institutional,
and technological-by which they are fashioned and linked together.,,13
Similarly, Schumpeter's opening statement in his 1939 treatise that
"analyzing business cycles means neither more nor less than analyzing the
economic process of the capitalist era" conveys the sense of a most inclus-
ive conception. Although "economic fluctuations properly so called [are]
those economic changes which are inherent in the working of the economic
organism itself," the effects of "external factors" or disturbances are also
numerous and important; indeed in many instances they "entirely over-
shadow everything else" (1939, voL 1, pp. 7, 12). But Schumpeter has no
single definition of a business cycle. Rather, his "really relevant case" is
that of interaction of "many simultaneous waves" (p. 212). The scheme
involves 3-4, 7-10, and 48-60 year cycles (named after their investigators
Kitchin, Juglar, and Kondratieff, respectively; see chap. 5). But there is
little support from the data for periodicities in the occurrence of groups of
major and minor cycles, although it is certainly true that individual cycles
vary greatly in amplitude, duration, and diffusion (Burns and Mitchell,
1946, chap. 11; Zarnowitz and Moore, 1986, pp. 522-523).
1.5 New Developments and Ideas
Momentous developments bearing on thinking about business cycles

occurred during the 1930s in the worlds of both real events and ideas. The
Great Depression was not only an entirely unexpected calamity, it seemed
to contradict the established ways and means of economic analysis. Seeking
new explanations and remedies, Keynes claimed "to have shown ... what
determines the volume of employment at any time" and that therefore
"our theory must be capable of explaining the phenomena of the Trade
Cycle." Any particular cycle, he believed, "is highly complex and ... every
element in our analysis will be required for its complete explana-
tion." But "the essential character of the Trade Cycle and, especially, the
regularity of time-sequence and of duration which justifies us in calling it a
cycle, is mainly due to the way in which the marginal efficiency of capital
fluctuates" (1936, p. 313). Investment depends on the MEC as compared
with the rate of interest, which recalls Wicksell's (1936) account of the
relation between the value and the cost of capital goods, and Fisher's
(1907) concept of the "rate of return over cost." But Keynes adds the ex
ante discrepancies between investment and saving, the multiplier, and the
cycle-induced changes in the liquidity preference and the propensity to
consume. He stresses the instability of investment due to sharp fluctuations
in business confidence under uncertainty about long-run returns.
Keynes' leading role in the development of the macroeconomics of
income determination is beyond doubt, but his analysis of business cycles
was quite fragmentary and his influence on the subsequent work in this
area proved rather limited.1 4 Formal models of dynamic disequilibrium
based mainly on an interaction of the investment accelerator and the
consumption multiplier enjoyed considerable popUlarity between the late
1930s and the early 1950s (Samuelson, 1939; Metzler, 1941; Hicks, 1950).15
But it soon became apparent that these models are limited to a mechanical
treatment of relations between a few aggregates and neglect potentially
important factors of prices, money, finance, and expectations. A more last-
ing impact is attributable to the general idea that random shocks and
changes in exogenous factors provide impulses that are propagated into
cyclical movements by the dynamics of the interdependent market econ-
omy. This conception had some earlier origins but was first formalized by
Frisch (1933).
Starting in the 1930s, econometric models came into use as vehicles
for testing business cycle theories (Tinbergen, 1939). Their early postwar
versions were heavily influenced by the Frisch impUlse-propagation model
and the Keynesian I-S model of Hicks (1937). The development of macro-
econometric models depended critically on the growth of modern statistics,
and particularly the system of national income accounts.
Hansen (1964, p. 4) defines the business cycle as "a fluctuation in (1)
employment, (2) output, and (3) prices." Although he refers to Mitchell's
1927 definition, he is content to use only the aggregative measures of
output and employment, plus the indexes of consumer and wholesale
prices. More recent studies tend to restrict the criteria further to measures
of real income and output, dropping the comprehensive price indexes
because of their continued rise during the business contraction of the past
30 to 40 years. Other measures, such as those of the diffusion of cyclical
movements, attract little attention. However, the dates of business cycle
peaks and troughs as identified by the NBER continue to be generally
accepted.
Views on the nature and scope of business cycles remain disparate. The
once prevalent theories of endogenous or self-sustaining cycles have been in
retreat for some time, though surviving in various nonlinear macrodynamic
models and lately experiencing some resurgence. The predominant type is

theories of cyclical response to various exogenous disturbances. The main
controversies concern the source of the originating shocks: Are they real or
monetary? To aggregate demand or aggregate supply? Monetarists stress
the instability of growth in money supply due to errors in monetary policy.
Keynesians stress the instability of private spending on capital goods and
consumer durables. Both groups see the main shocks as arising on the
demand side, but some new equilibrium theorists give more attention to
supply shocks that impact the relative prices of oil and other major inputs,
technology, and productivity.
This is no place to discuss in any detail the recent evolution and present
state of business cycle theories (for a treatment of this subject, see
Zarnowitz, 1985, 1991). Only a few remarks are in order here, but more will
be made later in the context of a descriptive analysis of past and recent
business cycles.
The range of contemporary views is indeed remarkably wide. Some
theories predict the recurrence of serious financial crises and business
depressions (Minsky, 1982). Others see the "cycles" as phenomena of
moving equilibrium, no more than adjustments to random shocks (to tech-
nology, for example,) and of low social cost (Plosser, 1989). Most of the
writings can be located somewhere between these extremes.
The debate increasingly concerns the underlying institutional and
market conditions. Increases in the weight of the public sector and the role
of government actions first resulted in rising beliefs that business cycles can
be not only reduced but even entirely eliminated by proper stabilization
policies. The focus of such expectations shifted over time from fiscal to
monetary policies, and from purely discretionary to more rules-oriented
policies. But in the process the high hopes of banning all boom-and-
bust sequences were repeatedly disappointed. The celebrated inftation-
unemployment tradeoff had seemed to vanish altogether in the 1970s,
but this was a temporary eclipse reflecting the worsening inflation (actual
and expected) and the novel supply shocks (huge oil price jumps). What
became clear is that the tradeoff is neither stable nor predictably exploit-
able for purposes of stabilization policy but is itself dependent on the
varying sources and characteristics of business cycles.
A related issue is the cyclical sensitivity of prices and wages. The new
classical economists believe that prices in general are flexible, reacting
promptly to clear the markets. The new-Keynesian economists believe that
industrial prices and wages are sticky, reacting only slowly. In this view,
both nominal and real (or relative) rigidities exist, and they are presumed
to have good reasons, that is, to be consistent with considerations of
14 TIlE BUSINESS CYCLE
self-interest of individuals and more or less organized groups. These

reasons, however, are not yet well understood, so the causes of price and
wage stickiness are in urgent need of much further research.
1.6 Some Basic Reflections and Further Steps
The preceding sections have shown that the documented history of busi-
ness cycles goes back at least two hundred years. The popular awareness of
recurrent crises and alternations of prosperous and depressed times is
reflected in similarly long records. Serious thought about the nature and
causes of major economic fluctuations accompanied serious thought about
the nature and causes of industrial development and economic growth
since the early writings on political economy.
If this is so, should it then not be clear by now what a business cycle is?
Why the need to raise this question time and again? The reason is that
business cycles are changing as well as complex. A process that repeats
itself with substantial regularity over long periods of time will not withhold
its secrets from inquiring minds over many decades of intensive study, even
if it were quite complicated. But business cycles involve many interacting
processes (economic, political, and broadly social) whose roles vary and
evolve. They are most probably caused in part by uncontrollable outside
disturbances and in part by errors of public policy and private decision
makers that may be avoidable, but they are also, just as plausibly, to a large
extent self-sustained and self-evolving. They are certainly not the same at
all times and in all developed market economies, although they show a
great deal of international interdependence.
To improve our understanding of business cycles, it is necessary to learn
from their past as well as present and to study both their common features
and diversity, continuity and change. In what follows, I shall consider the
main lessons from the chronologies of business cycles in the United States
and abroad and from research on the indicators and measures of cyclical
and related processes. The discussion will include examples of stable and
changing behavior and of some international similarities and differences.
2 The Changing DimenSions of BUSiness Cycles
2.1 The Historical Reference Cycle Chronologies of the NBER
Annual chronologies of "reference cycles," that is, business cycle peaks

and troughs, as estimated by Bums and Mitchell (1946) from time series
studies and Thorp's annals, are available for the United States and Great
Britain between 1790 and 1858. Monthly (as well as quarterly and annual)
reference chronologies from the same source cover the United States and
Britain for 1854-1938, France for 1865-1938, and Germany for 1879-
1932. Table 1-1 sums up the record of durations of business expansions,
contractions, and full cycles as derived from these dates. Part A lists the
means and standard deviations for the overall periods covered and
consecutive segments of 3-6 (mostly 4 and 5) cycles each. Part B lists the
shortest (S) and the longest (L) durations over the same periods and some-
what fewer successive subperiods.
It is clear that the durations have varied greatly, as shown by the huge
S-L ranges in part B. For example, the U.S. expansions ranged from 10
to 72 months, contractions from 8 to 72 months, and full cycles from 24
to 108 months (trough-to-trough) or 17 to 101 months (peak-to-peak). The
results for the other countries are very similar: slightly larger S-L differ-
ences still for the British cycles, slightly smaller for the shorter histories of
the French and German cycles.
These are comparisons between outliers, however, that can greatly
overstate the more common differences between the duration measures.
Indeed, U.S. business expansions in peacetime averaged 22-26 months in
each of the four segments of 1854-1938 and 25 months in the period as
a whole}6 Remarkably, the corresponding contractions were not much
shorter, averaging 19-20 months in three segments, 27 months in one, and
21 months over the total 1854-1938 period. Expansions varied relatively
much less than contractions, as shown by the standard deviations attached
to these averages. For full cycles, most of the mean durations are close to
four years, with overall standard deviations of 18-20 months. These stat-
istics disclose no systematic changes in phase and cycle durations. They
surely justify the phrase "recurrent but not periodic" in the NBER
definition. Still, the central tendency of the duration figures suggests a
modest degree of regularity. Thus, 15 of the 19 peacetime expansions of
1854-1938 (79%) fall into the range of 1~ to 21h. years, and the same
proportion applies to contractions in the range of 1 to 2 years.
The cycles in the other countries show more variability in durations both
across and within subperiods. The U.S. cycles tended to be more numerous
and shorterP They numbered 21 in 1854-1938, averaging 4 years, while
the British cycles numbered 16 and averaged 5 % years. For the longest
common period, 1879-1938, the comparison is shown in the following
table. Figure 1-1 compares the timing of business cycle peaks and troughs
in the four countries by means of a schematic diagram. It suggests a fairly
high degree of correspondence between the chronologies. Of the total of
Table 1-1. Durations of Business Expansions and Contractions in Four Countries, Selected Periods Between
1790 and 1938
A. Phase and Cycle Durations, in Months: Means and Standard Deviations (S.D.)
Full Cycle Full Cycle

Period (No. of Business Cycles)· Expansion b Contraction C (Tto T) (P to P)
Tto T PtoP Mean S.D. Mean S.D. Mean S.D. Mean S.D.
Line (1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
United States
Excluding Civil War & WWI
1 1854-1879 (4) 1857-1882(4) 26 7 27 26 53 31 61 27
2 1879-1897 (5) 1882-1899 (5) 25 7 19 11 44 17 41 11
3 1897-1915 (5) 1899-1913 (4) 22 8 20 5 42 4 40 11
4 1919-1938 (5) 1918-1937 (5) 26 15 20 13 46 16 45 29
5 1854-1938 (19) 1857-1937 (18) 25 9 21 14 46 18 46 21
All cycles d
6 1854-1938 (21) 1857-1937 (20) 26 11 22 14 48 18 48 20
7 1790-1938 (35) 1790-1938 (33) 28 14 23 16 50 24 50 20
Great Britain
Excluding WWI
8 1854-1879 (5) 1857-1882 (5) 40 9 36 28 76 37 79 32
9 1879-1914 (5) 1882-1912 (5) 43 17 25 16 68 26 67 36
10 1919-1938 (5) 1918-1937 (5) 26 24 20 10 47 21 45 33
11 1854-1938 (15) 1857 -1937 (15) 36 18 27 19 63 29 64 34
All cyclesd
12 1854-1938 (16) 1857 -1937 (15) 37 18 26 19 63 29 64 33
13 1790-1938 (29) 1796-1937 (28) 39 18 22 16 60 24 62 28
France
Excluding WWl
14 1865-1887 (5) 1867-1881 (5) 22 3 30 23 52 25 56 31
15 1887-1914 (5) 1881-1913 (4) 39 21 26 14 65 28 67 32
16 1919-1938 (6) 1920-1937 (6) 24 11 15 8 39 14 38 10
17 1865-1938 (16) 1867-1937 (15) 28 15 23 17 51 24 52 26
All cycles
18 1865-1938 (17) 1867-1937 (16) 29 15 22 16 51 23 52 25
Germany
Excluding WWl
19 1879-1914 (6) 1882-1913 (5) 39 16 32 21 71 26 75 34
20 1919-1932 (3) 1918-1929 (3) 29 12 23 15 53 25 43 8
21 1879-1932 (9) 1882-1929 (8) 36 15 29 18 65 26 63 31
All cycles
22 1879-1932 (10) 1882-1929 (9) 37 14 27 18 64 24 63 29
Table 1-1. (Continued)
B. Shortest (S) and Longest (L) Durations, in Months
No. of Business Full Cycle Full Cycle

Cycles Expansion b Contraction c (Tto T) (PtoP)
Period e Tto T PtoP S L S L S L S L

Line (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11)
United States
23 1790-1855 14 13 12 72 12 72 24 108 24 84
24 1854-1899 10 10 18 461 8 65 30 99 30 101
25 1899-1938 11 10 10 50 71 43 28 64 17 93
Great Britain
26 1792-1858 14 14 24 72 12 36 36 84 36 108
27 1854-1900 6 6 30 64 6 81 39 135 36 123
28 1900-1938 10 9 8 61 61 37 26 79 17 98
France
29 1865-1895 6 5 19 41 11 68 34 95 33 109
30 1900-1938 11 11 8 52 8 30 24 92 24 110
Germany
31 1879-1904 4 3 17 61 18 61 35 102 41 122
32 1904-1932 6 6 16 52 12 40 28 77 34 69
Source: National Bureau of Economic Research. For details, see Moore and Zamowitz (1986, tables A.2 and A.3).
aThe years are those of the initial and terminal business cycle troughs (T) in column 1, and those of the initial and terminal business cycle
peaks (P) in column 2, according to the NBER monthly chronology. The numbers in parentheses refer to complete T to T or P to P cycles in
the given period. The entries in columns 7 and 8 correspond to the dates in column 1; the entries in columns 9 and 10 correspond to the dates
in column 2. The wartime cycles consist of expansions during the Civil War (for the United States) and World War I (WWI) for all four
countries, and of the immediately following contractions (for the T to T cycles) or the immediately preceding contractions (for the P to P
cycles).
b Measured from troughs (T) to peaks (P).
C Measured from peaks (P) to troughs (T).
d The reference cycle chronologies contain annual segments for the United States, 1790-1855, and Great Britain, 1792-1858. In Part A,
lines 7 and 13, observations based on these dates but converted from annual to monthly durations are combined with the observations based
on the monthly reference chronologies that begin in 1854 for both the United States and Great Britain.
e Years refer to the earliest and latest reference dates in each segment. Entries in lines 23 and 26, columns 4-10, are converted from
annual to monthly durations (see note d).
f Refers to a wartime cycle.
Figure 1-1. Timing of reference cycles for four countries and matched turning
pOints, 1854-1938
Note: For each country, the lines connect the dates of business cycle peaks (upper turning
points) and troughs (lower turning points). Thus, the upward-sloping segments of each country
line represent expansions, the downward-sloping segments, contractions. The dashed links
between the country lines connect the matched peaks or troughs for two or more countries.
The sign x denotes an unmatched tum. For France before 1865 and for Germany before 1879,
the reference dates are annual. They are plotted at the midpoint of the given calendar year
and connected with dashed and dotted lines. All other reference dates are monthly and
are connected with solid lines. Question mark (?) denotes a dubious turning point (see text).
Sources: German annual turning points 1855-1963 estimated from Hoffman by Rostow
(1980, pp. 38-39). All other dates are from Bums and Mitchell (1946, pp. 78-79).
146 turning points shown, 92 (63%) match for all four countries and
another 36 (25% ) for three countries. Four turning points can be matched
for two countries, and only 14 (10%) are unmatched. Practically the same
proportions apply to peaks and troughs taken separately.
United States Great Britain France Germany
Number of cycles (T to T) 17 13 14 10
Mean duration (months) 48 65 53 64
Standard deviation (months) 18 31 25 24
When examined more closely, figure 1-1 shows that during the first
quarter-century covered, French and German cycles diverged consider-
ably, probably in part because of annual dating. The U.S. cycles conformed
better to the British and German than to the French cycles. In the period
1879-1914-the heyday of the pre-World War I gold standard-the
conformity between the business cycles in the European countries was
particularly close, but the U.S. economy followed a different pattern
of shorter and more frequent fluctuations. The timing of the wartime
expansions (1914-1918) and contractions (1918-1919) was very similar in
all four countries. In the 1920s and 1930s the European economies were
much less in phase with each other than in the preceding 40 years, but the
degree of conformity between their cycles and those in the United States
increased (see Morgenstern, 1959, chap. 2).1 8
2.2 A Cautionary Note

The limitations of data, particularly for the early times, might have
produced various errors in the identification of the reference cycles. A
particularly likely type of error is that of mistaking a serious slowdown
(phase of a low, that is, below-trend but still mainly positive, growth) for a
contraction (phase of an absolute decline in overall economic activity).
One reason to suspect an error is that the reference contractions are so
much longer and more frequent in the pre-World War II era than for the
more recent times about which there is much better and more complete
statistical information (see section 2.3). Another reason is that business
annals and indexes representing deviations from "normal" business
conditions (where the "normal" levels or trends themselves are not
quantified) were extensively used in determining the U.S. reference dates
for the period before 1873, and these sources may be especially susceptible
to the suspected bias (Zarnowitz, 1981, pp. 494-499).
A reassessment of the evidence must consider that the historical
contractions may represent declines in real or nominal aggregates or both,
depending on the strength and persistence of the concurrent movements in
the general price level. Deflations greatly aggravated some past depressions,
and wholesale price indexes are among the oldest important cyclical
indicators. Both real and nominal representative time series must therefore
be considered carefully, and when this is done the NBER chronology is
found to be in good overall agreement with dating by other economic
historians and generally well supported by the available evidence (Bums
and Mitchell, 1946, pp. 107-113; Zarnowitz, 1981, pp. 494-505; Moore and
Zarnowitz, 1986, pp. 755-764).
22 mE BUSINESS CYCLE
This does not mean, however, that all the identified cycles (let alone
their precise dates) are equally well confirmed. A few episodes are doubt-
fuI. They include four NBER contractions: 1845-1846,1869-1870,1887-
1888, and 1899-1900, of which the first one is most uncertain. (The peaks
and troughs of the last three are denoted by a question mark (?) in figure
1-1.) The information on hand seems to me insufficient to make a con-
clusive determination of whether these were periods of actual declines
or retardations below average growth. There are some problems with a few
other minor contractions as well, but they are easier to resolve (Zarnowitz,
1981).
If these episodes were treated as slowdowns instead of contractions,
the average durations of expansion (£) would be considerably increased,
both absolutely and relative to those of contractions (C). The following
tabulations shows this for the two periods principally affected. 19
1834-1855 1854-1919
Number of cycles (T to T) 5 4 16 13
Expansion, months (E) 26 36 27 37
Contraction, months (C) 24 27 22 23
Ratio (Elt) 1.1 1.3 1.2 1.6
Some of the reference dates for the other countries may also be
questioned. Burns and Mitchell (1946, p. 113) mention explicitly the
German 1903-1905 contraction20 and France "after 1932." Consequently,
in figure 1-1, question marks denote these French and German contrac-
tions (as well as the three pairs of U.S. turning points already noted).
In addition, the authors of the NBER dating methodology warn that
"the chronology for France in the 1860s and 1870s requires careful con-
sideration." It is therefore possible that some of the several discrepancies
between the French and the other turning-point dates in this period are
spurious (but, for lack of more specific information, these dates are left
unquestioned in the diagram).
Figure 1-1 indicates that elimination of such dubious phases as U.S.
1887-1888 and France 1933-1935 would clearly improve the intercountry
correlation of cyclical movements. But in other cases the opposite effects
are suggested: Note U.S. 1899-190021 and Germany 1903-1905. Treating
the U.S. 1869-1870 period as a slowdown rather than a contraction would
improve the conformity with Great Britain but worsen the conformities
with Germany and France (the dates for which, however, are themselves
uncertain).
2.3 The Last Half-Century: Long Expansions, Short

Contractions
The NBER chronology identifies 15 business cycles measured from peak to

peak in 1857-1918, that is, on average one every four years. In 1918-1945
there were six cycles with a mean duration of about 4Yz years. Between
1945 and 1990, nine cycles occurred, averaging approximately five years.
Consequently, according to these estimates (see table 1-2, lines 7-9), the
length of U.S. business cycles increased and their frequency diminished
gradually. However-, these differences are small and could be due largely to
errors in a few early reference dates. 22 The predominant feature of the
average full-cycle durations, when compared across such long eras, is their
relatively high degree of stability.
What table 1-2 definitely demonstrates is that business expansions
tended to be much longer, and contractions much shorter, after World War
II than before. In the 1854-1919 period, expansions averaged 27 months
and contractions 22 months; for 1919-1945 the mean durations of the
respective phases are 35 and 18 months; and for 1945-1990 they are 50 and
11 months. Excluding wartime episodes would make the averages slightly
smaller for expansions and slightly larger for contractions, but it would not
alter significantly the changes between the periods.
Even allowing for the largest plausible errors in the early dating would still
leave a large contrast between the pre-1945 and post-1945 cycles with respect
to their relative divisions by phase. Thus the EIC ratio for 1854-1919 can
be estimated at 1.2-1.6 (see tabulation on page 22). The EIC ratio for 1919-
1938 is 1.9; that for 1945-1990 is 4.5 (based on the entries in lines 8 and 9). In
other words, in the last 50 years the economy has been in expansion phases
about 80% of the time; before then, perhaps only 55% to 60% of the time.
Business expansions have become not only longer but also more vari-
able, hence less predictable with regard to their durations. The standard
deviations of the latter, in months, increased from 10 before 1918 to 26 in
1919-1945 and 31 in 1945-1990. The range of peacetime expansions was
10 to 50 months before 1945 and 12 to 92 months thereafter (the longest
wartime expansions lasted 80 months in 1938-1945 and 106 months in
1961-1969; see lines 7-14, columns 3 and 4).
In contrast, business contractions have become not only shorter but
also more uniform in length, hence in this sense more predictable. Their
standard deviations, in months, dropped from 13 and 14 in 1857-1919
and 1920-1945, respectively, to 4 in 1945-1982. The range of peacetime
contractions was 8 to 65 months before 1945 and 6 to 16 months thereafter
(see lines 7-14, columns 5 and 6).
Table 1-2. Durations of Business Expansions and Contractions in the United States, Selected Periods between 1938
and 1990 and between 1854 and 1990
Full Cycle Full Cycle

Period (No. of Business Cycles)" Expansion b Contraction" (Tto T) (PlOP)
Tto T PtoP Mean S.D. Mean S.D. Mean S.D. Mean S.D.
Line (1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
1938-1990, excl. wartime cycles

1 1938-1961 (3) 1945-1990 (3) 33 8 10 2 43 8 42 9
2 1961-1982 (3) 1960-1990 (4) 50 34 13 6 48 18 62 38
3 1938-1982 (6) 1945-1990 (7) 43 26 11 4 46 13 53 30
1938-1990, all cycles
4 1938-1961 (5) 1945-1960 (5) 45 21 9 1 54 20 55 23
5 1961-1982 (4) 1960-1990 (5) 61 39 12 5 65 38 73 41
6 1938-1982 (9) 1945-1990(10) 53 31 11 3 59 28 64 33
1854-1990, all cycles
7 1854-1919 (16) 1857-1918 (15) 27 10 22 14 48 19 49 18
8 1919-1945 (6) 1918-1945 (6) 35 26 18 13 53 22 53 32
9 1945-1982 (8) 1945-1990 (9) 50 31 11 4 56 27 61 33
10 1854-1982 (30) 1857 -1990 (30) 35 23 18 12 51 22 53 26
Shortest (S) and Longest (L) Durations, in Months
S L S L S L S L
Excluding wartime cycles
11 1938-1982 (6) 1945-1990 (7) 12 92 6 16 28 64 18 108
12 1854-1938 (19) 1857-1945 (19) 10 50 8 65 30 99 17 101
All cycles
13 1938-1982 (9) 1945-1990 (9) 12 106d 6 16 28 117 d 18 116d
14 1854-1938 (21) 1857-1945 (21) 10 80d 7d 65 28 99 17 101
a T denotes initial and terminal troughs, P denotes initial and terminal peaks. Entries in columns 7 and 8 correspond to the dates in
column 1; the entries in columns 9 and 10 correspond to the dates in column 2. The wartime cycles consist of expansions during the Civil War,
World Wars I and II, the Korean War, and the Vietnam War, and of the immediately following contractions (for the T to T cycles) or the
immediately preceding contractions (for the P to P cycles).
b Measured from troughs (T) to peaks (P).
C Measured from peaks (P) to troughs (T).
d Refers to a wartime cycle.

Both expansions and contractions, and hence full cycles, increased in

length and variability between the early part of the recent era (the 1940s
and 1950s) and the later part (the 1960s, 1970s, and 1980s). This is true for
both peacetime and all cycles, as shown in some detail in table 1-2, lines
1-6.
2.4 Phase Durations in Eras of Inflationary and Deflationary

Trends
Through the great contraction of the early 1930s, prices in the United
States and Europe followed alternating upward and downward trends
(table 1-3). This is best seen in the wholesale or producer price indexes,
which also have the longest records, but even the less flexible consumer
price indexes show alternating periods of inflation and deflation. The
complete upswings in wholesale prices lasted generally between 20 and 30
years; the downswings varied more, from 9 to 32 years. Over long stretches
of time the deflations nearly offset the inflations, even when the war
periods are included, so that the level of wholesale prices at the bottom of
the depression in 1932 was not much higher than in the early times of the
Republic, more than 140 years earlier. However, no significant deflationary
trends appeared in the nearly 60 years since, only brief price declines in the
recessions of 1937-1938 and 1948-1949. Ours is the age of the most
persistent and pervasive inflation on record.
The long-period averages in table 1-3, column 2, are instructive, even
though they conceal much variability over shorter time intervals. Before
1950 comprehensive price indexes had generally, in addition to the longer
trends, clear procyclical movements, up in expansions and down in
contractions. Thereafter, disinflation replaced deflation as a frequent
concomitant of business recessions and some major slowdowns. Inflation
became a grave and stubborn problem primarily because the deflationary
slumps have disappeared, not because the booms have grown more
inflationary. This is a very important but relatively neglected point.
Periods of rising trends in prices witnessed relatively long business
expansions and relatively short contractions, whereas periods of declining
trends in prices witnessed the opposite. This was observed as early as 1926
by Mitchell (in his introductory chapter to Thorp, 1926, pp. 65-66) and
confirmed by Bums and Mitchell in 1946 (pp. 437-438,538). Table 1-3
provides a summary of their findings and of the generally consistent later
results (see also Moore, 1983, chap. 15; Zarnowitz and Moore, 1986,
pp. 525-531). In times of rising prices, the expansion-contraction duration
ratios EIC averaged 2.1-2.5; in times of falling prices, WC averaged 0.8-

1.0. These are large and systematic differences based on U.S. data for
1789-1990 and on European data for periods beginning 1854-1895 and
ending in the 1930s. The documented relationship is robust: It does not
depend on the inclusion of the highly inflationary wartime expansions, for
example, or on our use of average rather than total phase durations.23
The longest and most severe contractions in the history of U.S. business
cycles were associated with major deflations, including 1873-1879, 1882-
1885, 1895-1897, 1920-1921, and 1929-1933. This last world depression
was ascribed by Irving Fisher (1932, 1933) mainly to "debt-deflation," that
is, an overexpansion of debts in the preceding expansion followed directly
by sharp downturns in prices and profits. 24 Similar ideas on the role of
overindebtedness, credit restrictions, and liquidation reappear in some
recent theories that also stress, along the lines of Keynes, the instability of
real investment and are adapted to the present environment of large
government and persistent inflation (as in Minsky, 1982). But one does not
need to accept all these ideas to agree that the avoidance of major
deflations greatly helped the postwar economies to avoid major depres-
sions as well. Many monetarists and Keynesians alike would probably
accept this statement.
Let it be made immediately clear that to argue against deflation does not
mean at all to argue for inflation. Recent experiences with high inflation in
developing countries and with stagflation in North America and Europe
leave no doubt that any long and sizable inflation is harmful to both real
growth and economic stability. Once allowed to develop, inflation, like
deflation, is difficult to control and can be self-accelerating by generating
perfectly reasonable expectations about its own future. But it is also true
that a small rate of inflation really amounts to the desired state of stability
in the general price level, given the inevitable technical limitations of
price-index measurement. Further, expansions favor price raises, while con-
tractions favor cost and price cuts. 25
2.5 Some Other Analytical Uses of the Duration Data
Are business expansions, contractions, and full cycles more likely to end or
less likely to end as they grow older; that is, do they exhibit a positive or
a negative duration dependence? Diebold and Rudebusch (1990, 1991)
attempt to answer this question by fitting hazard functions to the NBER
duration data. 26 Using an exponential-quadratic hazard model, which
is relatively flexible yet parsimonious, they find evidence of a positive
Table 1-3. Trends in Wholesale Prices and the Business Cycle Phase Durations, United States 1789-1990 and Three
European Countries between 1854 and 1935
Trend in Wholesale Prices Business Expansions c Business Contractionsc

Ratio
First to Last Average Average of Average
Direction and Year Change Duration (E) Duration (C) Durations,
Dates in Years· (percent p.a.)b Number in Months Number in Months bc
Line (1) (2) (3) (4) (5) (6) (7)
United States
Rising
1 1789-1814 3.1 5 42 4 22 1.9
2 1843-1864 4.7 6 (5) 32 (30) 5 15 2.1 (2.0)
3 1896-1920 5.1 7 (6) 23 (20) 6 18 1.3(1.1)
4 1932-199Qd 4.1 11 (8) 53 (44) 10 11 4.6 (3.8)
5 Total or Averaged 4.2 29 (24) 40 (34) 25 16 2.5 (2.1)
Falling
6 1814-1843 -3.1 6 27 7 27 1.0
7 1864-1896 -3.2 7 25 8 26 1.0
8 1920-1932 -6.9 3 23 4 22 1.0
9 Total or average -3.9 16 25 19 25 1.0
Great Britain
Rising
10 1854-1873,1896-1920 3.6 10 30 8 17 2.2
Falling
11 1873-1896,1920-1933 -2.3 5 30 7 38 0.8
France
Rising
12 1865-1873,1896-1926 6.4 11 31 9 15 2.1
Falling
13 1873-1896,1926-1935 -2.6 4 30 6 34 0.9
Germany
Rising
14 1895-1923 2.3 6 40 6 18 2.2
Falling
15 1922-1933 -4.2 4 32 4 42 0.8
• Through 1932 based on Burns and Mitchell (1946, p. 432). For 1932-1982, see Moore (1983, p. 240).
b For the United States 1789-1982, see Zarnowitz and Moore (1986, p. 527). For other countries, see the sources listed in Burns and
Mitchell (1946, p. 432). Author's calculations are based on the data for the initial and terminal years; the averages are weighted by the
durations in years of the periods covered.
C The NBER monthly business cycle chronologies are used. See Zarnowitz and Moore (1986, pp. 528-529). Entries in parentheses
exclude wartime expansions (Civil War, World Wars I and II, Korean War, and Vietnam War).
d Updated by author's calculations.
duration dependence for pre-World War II business expansions in the

United States, Great Britain, France, and Germany. Thus, the longer these
expansions lasted beyond their minimum durations, the more likely they
were to end. On the other hand, the pre-1940 business contractions show
no duration dependence at all in any of these four countries. For full cycles
(both T-to-T and P-to-P) there is evidence of some positive duration
dependence but less than for expansions.
The U.S. business cycles of the post-World War II era present just the
opposite situation, according to the estimates by Diebold and Rudebusch.
Here the expansions show no duration dependence, and neither do the full
cycles whose length reflects largely that of the expansions. In contrast, the
contractions display a strong positive duration dependence.
These results are definitely associated with the findings reported earlier
in section 2.2 of this chapter: Prewar expansions and postwar contractions
have both stable average durations and positive duration dependence;
prewar contractions and postwar expansions have neither. These tendencies
are of great interest but difficult to explain. They are not consistent with
models of purely random "Monte Carlo" fluctuations that would show no
systematic duration dependence for either upward or downward move-
ments. They are also not consistent with models that would produce positive
duration dependencies in both expansions and contractions.
It is important to note that a strong periodicity of expansions implies
that business cycle peaks, in large measure, should be predictable. But they
are not: Even the best reputed professionals seldom make good forecasts
of an approaching downturn, and the public often recognizes a recession
tardily, with a surprise. Moreover, if the timing of economic downturns
were predictable, strong efforts would probably be made to prevent them,
or at least to diminish them greatly, so as to avoid the large individual
losses and social distress they cause. If such efforts fully succeeded, busi-
ness cycles as they are known now would no longer be with us. If they were
repeatedly frustrated, we would try to learn why. In any event, the experi-
ment remains only a mental one; were it actually performed, the working
of the economy itself could well be significantly altered. The upshot of the
argument is simply that strong periodicities of expansion are not to be
expected. This is quite consistent with indications of a significant positive
duration dependence in pre-World War II expansions.
For contractions, stable durations have different implications and are
not so difficult to imagine. That contractions have become much shorter
in the postwar age can be explained in part by changes in the structure
of the economy and in part by the direct and indirect effects of much
increased efforts of government policy makers to combat recessions and
unemployment. The experience of more persistent growth helped to gener-

ate expectations that a recession, once known to be under way, will prove
to be over soon. A business recovery in the near future is always a highly
welcome event (worries about an inflationary boom can wait), and
forecasts of a recurrence of such an event are always popular and apt to
spread.
2.6 On Theories and Evidence Relating to Periodic Business

Fluctuations
The duration data based on the reference cycle chronologies give no

encouragement to any theories that imply strong phase and cycle period-
icities, but they do not exclude the possibility of some weak and latent
periodicities.27 The class of strongly periodic theories includes the early
deterministic versions of multiplier-accelerator interaction and nonlinear
limit cycles (Samuelson, 1939; Kaldor, 1940; Metzler, 1947; Hicks, 1950;
Goodwin, 1951). Adding shocks can relax the periodicities of the cycles
generated by these models. Adding relations involving money, finance, and
expectations can enrich and diversify the models. None of these
refinements, however, is likely to produce any systematic asymmetries
and shifts in phase durations.
Nor is it probable that the answer lies in any single impulse or propa-
gation mechanism. The models of "forced oscillations" driven by factors
that are themselves periodic are the simplest and also the easiest to dismiss.
The oldest example here is that of weather and harvest cycles that used to
be attributed to variations in sunspots or other extraterrestrial influences
(Jevons, 1884; Moore, 1914). There is no doubt that changes in weather
can and do cause major and recurrent fluctuations of production, prices,
wealth, and welfare in poor agricultural countries. Some of these may be
strongly or weakly periodic and occasionally are very large. 28 But today the
weather and other "seasonal" hypotheses have hardly any plausibility
as the explanation of business cycles in modem industrialized countries
(including their farm sectors).
A more timely and currently much debated hypothesis is that of a
"political business cycle" linked to the electoral cycle in democratic
countries. Here, too, there are elements of forced oscillations, though not
ordained by nature but due to events with a calendar rhythm imposed by
a nation's laws and institutions. That economics and politics interact in
various ways is important and widely acknowledged. But the notion of a
regular political business cycle encounters serious theoretical problems,
being based on assumptions that would enable the government to pursue a

purely self-interested and myopic but highly effective fine-tuning policy.
Not surprisingly, good evidence in support of this hypothesis is hard to
find. 29
Cycles with a positive duration dependence, and indeed a substantial
degree of periodicity, can be produced by damped linear models with small
and dense white-noise shocks and the required structural properties. This
hypothesis, in the form developed by Frisch (1933), has strongly influenced
the theory and, particularly, the econometric modeling of business cycles.
It belongs to a class of models of "free oscillations," which explain the
cycles by the way the economic system responds to the forces that impinge
upon it, not by the cyclicality of these forces. This is a useful approach
when given a sufficiently broad interpretation. But the empirical validity of
models with purely random shocks was never adequately demonstrated.
Indeed, the conclusion from simulation studies that used macro-
econometric models driven only by such shocks is that the so generated
movements lack the dimensions and properties observed in the actual
data on business cycles (Hickman, 1972).
One of the needed extensions is to large shocks, which contribute
to macroeconomic instability, along with the small ones (Blanchard and
Watson, 1986). Large shocks presumptively occur at irregular intervals but
are often serially correlated: Think of wars, shifts in policy, embargoes,
price-level disturbances, speculative waves, financial crises, major bank-
ruptcies, and strikes. In general such events and their direct and indirect
effects are likely to increase the diversification and irregularity of business
cycles over time.
2.7 The Moderation of U.S. Business Cycles
Business contractions in the United States have become not only much
shorter but also much milder in the post-World War II era compared with
earlier times. Severe depressions like those of the 1870s, 1880s, and 1930s
appear to have disappeared for good. Table 1-4 compares measures of
depth and diffusion for the five contractions of 1920-1938 and the eight
contractions of 1948-1982. The former set includes two major depressions
(1920-1921 and 1937-1938), one disastrous depression (1929-1933), one
mild recession (1926-1927), and one more severe recession (1923-1924).
The latter set includes three mild recessions (1960-1961, 1969-1970, and
1980) and five more severe recessions (1948-1949, 1953-1954, 1957-1958,
1973-1975, and 1981-1982). The presently available, incomplete data
indicate that the U.S. recession of 1990-1991 (not covered here) will qual-
ify as another mild one.
Table 1-4 leaves no doubt about the relative shallowness of even the
most severe of the recent recessions (1973-1975 and 1981-1982), which
contrasts sharply with the depth of either of the two interwar depressions,
let alone that of the 1929-1933 collapse. Despite the upward trend in
joblessness during the recent period, the maximum unemployment rate in
the early 1980s was still less than half those reached in the 1930s (column
5). Consistent evidence is also provided by diffusion measures, although
here the differences are less striking, with the proportions of industries
with declining employment ranging from 71 % to 100% (column 7).
For the pre-World War I period, comprehensive estimates of economic
activity are scarce and their quality leaves much to be desired. Quarterly
data used in table 1-5, lines 1-3, suggest that output was H2 times more
volatile in 1919-1945 than in 1875-1918 and twice as volatile in 1919-1945
as in 1946-1983 (column 3).30 Although the pre-1919 figure may be an over-
estimate (Romer, 1986), there is little doubt about the ranking of the three
periods in terms of both quarter-to-quarter and cyclical variability. In
percentage terms and on average, the trough-to-peak increases were the
largest in 1919-1945 and somewhat larger after 1945 than before 1919
(column 6). The peak-to-trough decreases were very large in 1919-1945,
much smaller in 1875-1918, and the smallest by far in 1946-1983 (column 9).
Before 1945 wholesale prices tended to rise in expansions and fall in
contractions, as shown clearly in table 1-5 (see lines 4-6, columns 6 and 9).
The trend in prices was just moderately up in 1875-1918 and mildly down
in 1919-1945, despite the big inflations of World Wars I and II, but it was
strongly up in 1946-1983 (column 2).
Even though economic growth decreased and cyclical instability
increased in the latter part of the postwar era (since the early 1970s),
business cycles remained moderate in comparison to the pre-World War I
and, a fortiori, the interwar periods. In other major countries with higher
growth rates, business contractions became much less frequent and much
milder still (see section 3). Several not mutually exclusive hyptheses may
explain the observed reduction of the business cycles in the United States.
I can only list them briefly here (Zarnowitz, 1991, offers an extended
discussion).
1. Employment in service-producing industries and government, which

is only weakly procyclical or acyclical, increased greatly relative
to employment in goods-producing industries, which is strongly
procyclical.
Table 1-4. Depth and Diffusion of Business Contractions in the United States, 1920-1982
Percentage Changeb in Unemployment Rate Nonfarm

Employment, %
Business Cycle Real Industrial Nonfarm Percentage of Industries
Contraction" GNP Production Employment Maximum< Increase < Contracting d
Line (I) (2) (3) (4) (5) (6) (7)
A. Selected Business Contractions

1 111920-7/1921 (D) n.a. -32.4 n.a. 11.9 10.3 97
2 10/1926-11/1927 (M) -2.0 -7.0 n.a. 4.4 2.4 71
3 811929-311933 (GD) -32.6 -53.4 -31.6 24.9 21.7 100
4 5/1937-611938 (D) -13.2 -32.4 -10.8 20.0 9.0 97
5 1111948-1011949 (S) -1.5 -10.1 -5.2 7.8 4.5 90
6 811957-4/1958 (S) -3.3 -13.5 -4.3 7.3 3.7 88
7 10/1969-1111970 (M) -1.0 -6.8 -1.5 5.9 2.6 80
8 1111973-3/1975 (S) -4.9 -15.3 -2.9 8.8 4.3 88
9 111980-7/1980 (M) -2.3 -8.5 -1.4 7.7 2.2 77
10 711981-1111982 (S) -3.0 -12.3 -3.1 10.7 3.6 79
B. A verages by Degree of Severity
11 Two major depressions e -13.4 -32.4 -10.6 16.0 9.6 97
12 Six severe recessions! -3.3 -13.1 -3.8 7.7 3.8 88
13 Four mild recessionsg -1.7 -3.8 -1.7 6.2 2.3 77
Source: Moore and Zamowitz (1986, Table A.7).

n.a. = not available.
a
Symbols in parentheses classify the contractions as follows: GD, Great Depression of1929-1933; D, depression; S, severe recession;
M, mild recession.
b Measured from peak to trough in quarterly data for real GNP and monthly data for industrial production and nonfarm employment.
C Entries for 1920-1933 (lines 1-3) are annual averages (monthly data not available).
d Before 1948 based on cyclical changes in employment in 41 industries. Since 1948 based on changes in employment over six-month
spans in 30 industries (1948-1959); 172 industries (1960-1971); and 186 industries (1972-1982).
e 1920-1921 and 1937-1938 (as marked D above).
f Includes the recessions of5/1923-7/1924 and 7/1953-5/1954 in addition to the four marked S above.
g Includes the recession of 4/1960-211961 in addition to the three recessions marked M above.
Table 1-5. Variabilities of Relative Change and Amplitudes of Cyclical Movement, Real GNP and Prices, 1875-1918,
1919-1945,and1946-1983
Log Differences Expansions b Contractions b
Mean Mean Mean Mean

Duration Amplitude Duration Amplitude
Period· Mean S.D. Number (quarters) (percent) Number (quarters) (percent)
Line (1) (2) (3) (4) (5) (6) (7) (8) (9)
Real Gross National Product

1 1875-1918 1.0 2.4 12 8.9 17.3 12 3.2 -5.1
2 1919-1945 0.8 3.6 4 11.0 30.1 5 6.0 -14.1
3 1946-1983 0.8 1.2 7 16.3 20.9 8 2.6 -2.5
Wholesale (Producer) Price Index
4 1875-1918 0.3 3.3 11 6.6 12.1 10 6.9 -11.8
5 1919-1945 -0.2 4.1 4 8.2 15.4 5 8.6 -21.6
6 1946-1983 1.1 2.3 4 12.0 8.7 5 5.0 -4.8
Source: Calculated from data in Balke and Gordon (1986, pp. 788-810) (see note 30).
a Number of quarterly observations per series: 1875-1918, 175; 1919-1945, 107; 1946-1983, 151. The annual dates refer to the first and
last turning points of the series during each period.
b Identified by specific cycle peaks and troughs in the series. Expansions are measured from peaks to troughs, contractions from troughs
to peaks. Only complete upward and downward movements are counted.
2. Fiscal automatic stabilizers relating to procyclical income taxes and

countercyclical transfer payments became quite effective, particu-
larly before the rise in inflation and its distortionary tax effects.
3. Federal deposit insurance prevented general banking panics,
although lately at high cost to the taxpayers.
4. The rates of change in money supply were more volatile in 1875-
1917 and, particularly, 1919-1939 than after 1946.
5. The record of discretionary fiscal and monetary policies, although
quite mixed over time, was on balance sufficiently positive to help
avert depressions and persistence of high unemployment.
6. The gradual recognition that business contractions have become
shorter and milder strengthened the confidence of business people,
workers, and consumers and made them adopt behavior that
promoted growth (but this was partially offset in more recent times
by the expectational effects of rises in inflation and unemployment).
3 Some International Comparisons of Cyclical

Movements and Growth Trends
3. 1 The Major Market Economies After World War /I
World War II devasted the economies of Continental Europe and the Far
East. The physical wealth and capital of once rich nations---cities, factories,
machinery-lay in ruins and in dire need of reconstruction. But human
capital was much better preserved; that is, people retained their high
productive potential-education, skills-and the backlog of effective
demand was huge. Monetary, fiscal, and political reforms enabled both the
defeated nations and the nations liberated by the Allies to make a rela-
tively smooth transition from closed-war to open-market economies, and
from totalitarian and oppressive to democratic and free social systems.
Further, foreign aid, mainly from the United States, was made available to
friend and former foe alike.
The result of this historically rare, perhaps unique, combination of
circumstances was that France, Italy, West Germany, and Japan (as well as
a number of smaller countries) soon came to enjoy extraordinarily high
rates of real economic growth. Of course, the initial activity levels were
very low, which helps explain the long persistence of very high and only
gradually declining growth rates, but nevertheless the progress achieved
in the 1950s and 1960s was spectacular, particularly in the case of West
Germany and Japan.
As long as the steep upward trends in employment and output (much of

it exported) continued, real growth in Continental Europe and the Far
East was interrupted infrequently and mostly by slowdowns rather than
absolute declines in overall economic activity. Thus, the benefits of high
growth were augmented by those of high cyclical stability.
Figure 1-2 presents the evidence. The data, collected and processed
by the Center for International Business Cycle Research (CIBCR) at
Columbia University Business School, are composite indexes of coincident
indicators that combine monthly and quarterly measures of aggregate
activity (total output, industrial production, employment, real sales,
inverted unemployment) and have trends made equal to those of the
corresponding series for real GNP or GDP. Note that West Germany had
its first postwar recession in 3/1966-5/1967, Japan (for which the index
begins in 1954) in 11/1973-2/1975 (see figure 1-2A). For France (since
1955) the first postwar contraction shown is in 4/1958-4/1959, for Italy it is
in 10/1963-3/1965 (figure 1-2B). Meanwhile, Canada and the United
Kingdom (figure 1-2C) and the United States and Australia (figure 1-2D),
all countries that suffered much smaller direct wartime damages, had less
need of domestic reconstruction, slower growth rates, and earlier and more
frequent recessions.
Economic growth slowed markedly everywhere during the 1970s and
1980s, and recessions became both more common and more severe. They
spread worldwide in 1974-1975 after the oil embargo and the huge
increases in the OPEC cartel oil prices, and again in 1980-1982 after new
price shocks and strong counterinflationary policy moves in the United
States. However, the index for Japan shows only a brief decline in late
1980; its growth after the mid-1970s was much slower than before but also
remarkably steady. The index for France shows frequent but only short and
shallow declines throughout, except in 1958-1959 and 1974-1975 (its sharp
decline in the spring of 1968 largely reflects the concurrent political unrest).
3.2 Estimated Dimensions of Business Cycles in Eight

Countries
There is no NBER-type international reference chronology of business

cycle peaks and troughs for the post-World War II period. Perhaps the
main reason is the belief that recessions have been generally replaced by
mere retardations of growth in most countries. But this is a vague view or
expectation, not a well-established fact. What is certainly true is that
contractions of total output and employment have been short and few and
far between in developed and developing countries that achieved high

average rates of real growth.
The laborious and data-intensive task of identifying and dating business
cycles in different economies cannot be attempted here. However, much
can be learned from the rich data bank of CIBCR, and particularly the
coincident indexes, covered in figure 1-2.
The black dots in figure 1-2 mark peaks and troughs of specific cycles in
the indexes. The so-identified declines last several months (as a rule, six or
more) and have amplitudes of several or more percentage points (as a rule,
at least 1%). They presumably reflect business contractions. The open dots
mark shorter andlor smaller declines that do not qualify as recessions
but may be associated with significant retardations in general economic
activity. However, substantial business slowdowns can occur without any
noticeable decreases in the levels of the index series, particularly in places
and times of very high overall growth (see, for example, West Germany,
1956-1957, and Japan, 1957 and 1964).
The charts show some important similarities between the country
indexes, notably the concentration of long expansions in the 1960s and
1980s, of mild recessions or slowdowns in the mid- or later 1960s, and
of more severe contractions in the mid-1970s and early 1980s. But the
differences are even more pronounced, and it is clear that they are primar-
ily related to the longer growth trends. The main point here is that high
growth helps to reduce the frequency and depth of business contractions.
Thus periods of rapid and recession-free growth are observed for West
Germany and Italy before 1962, for France in most of the 1960s and again
after 1985, for Japan before 1973, for Canada between 1961 and 1981, and
for Australia between 1961 and 1973. The economies of the United States
and United Kingdom had relatively low rates of real growth and most cycli-
cal variability. In sharp contrast, Japan had both the fastest growth and
least cyclicality; indeed, one can say that business cycles as we know them
were essentially absent there. 31
Figure 1-2 suggests that the postwar business cycles were shorter and
more numerous in the United States than in any other of the seven
countries. In 1955-1990, the longest common period covered, the numbers
of complete peak-to-peak cycles in the indexes were as follows: United
States, 6; United Kingdom and Italy, 5 each; France and Australia, 4 each;
Canada and West Germany, 3 each; and Japan, 2. Comparisons between
fewer countries over longer periods yield generally consistent results.
Further, inspection of the charts suggests that the greater frequency of U.S.
cycles was mainly due to the shorter durations of U.S. expansions. On the
whole, it was the expansions that differed greatly across the countries in
Japan W. Germany
'T -
log-scale log-scale
2.25
2.00
2.25
i West Gennany 1.75
2.00
1.50
'$L
1.25
1.75·j
1.00
Japan 0.75
....--
1.25 0, j j r j j -,------,- -. j
050
1948 52 56 60 64 6B 72 76 so 84 sa 1992
A
France Italy
log-scale log-scale
-
2.50 2.25
----.--~--
Italy
2.25 2.00
•
2.00 0
0
.• 1.75
1.75 1.50
1.50 1.25
1.25 1.00
l
1948 52 56 60 64 68 72 76 80 84 sa 1992
B
Canada U.K.
log-scale log-scale
250 , - - - - - - - - - - - - - - - - - - - - - - - - - - - , 2.25
•
..
2.25 2.00
2.00
-
United Kingdom o
o
o
1.75
o
t.75 1.50
1.50
--. .
Canada
•
1.25
1.25 'r----,-----,----,-----,----,----.----,---,----,-----,-----r 1.00
Australia U.S.
log-scale log-scale
2.50 - - - - - - - - - - - - - - - - - - - , 2.25
2.25 2.00
United States
2.00 1.75
1.75 1.50
1.50 -
Australa
1.25
1.25 'r----,----,---,.----,----.----,---,----r----,----,-----,l 1.00
Figure 1-2. Coincident indexes (adjusted). A: Japan and West Germany; B:

France and Italy; C: Canada and United Kingdom; 0: Australia and United States
both length and amplitude, whereas the contractions varied much less in
both dimensions.
Table 1-6 presents a summary based on the data plotted and the
specific-cycle turning points identified in figure 1-2. It demonstrates that
the cyclical rises in the U.S. index were on average much shorter than the
rises in the foreign indexes (column 6), and also much smaller in percent-
age terms than the indexes for the other countries, with the sole exception
of the United Kingdom (column 7). As for the cyclical declines in the U.S.
index, they were on average considerably deeper than the declines in the
other indexes, but not particularly long (columns 9 and to). The entries on
mean durations and mean amplitudes quantify the impression from the
charts that expansions differed greatly across the countries (even apart
from the extreme cases of West Germany and Japan), while contractions
differed only moderately.32
A comprehensive assessment of how business cycles in the different
countries compare would require an analysis of the performance of many
individual indicators, which is not possible here. However, I examined the
industrial production and employment components of the coincident
indexes under consideration and found that a few general observations
deserve to be made. The industrial production indexes cover sectors of
high cyclical sensitivity (manufacturing everywhere, mining and/or public
utilities in most countries).33 Therefore they have at least as many specific
cycles as the coincident indexes (which cover less sensitive sectors and
processes as well), often significantly more. Thus the industrial production
series for Japan (mining and manufacturing) clearly shows seven
expansions and six contractions in the 1953-1990 period.
In the United States, although nonfarm employment has a much
broader coverage than industrial production, the cyclical profiles of the two
indicators are quite similar. This is definitely not the case in some of the
other countries, however. In Japan total employment of "regular workers"
had only one cyclical contraction (4/1975-5/1976) between 1954 and 1990.34
In the United Kingdom employment in production industries followed a
gradual downward trend since 1966, and the same applies to West German
employment in manufacturing and mining from 1971 until mid-1984 as well
as to French nonfarm employment between mid-1974 and mid-1985. In
contrast, in the United States an upward trend in employment prevailed
throughout. It is well known that, after having stayed very low in the earlier
postwar years, unemployment in Western Europe increased rapidly
(quadrupled) between 1970 and 1985, from well below to well above the
U.S. levels. Yet, despite the downtrends in employment and clear cyclical
movements around them, business cycles apparently remained milder in
Table 1-6. Average Durations and Amplitudes of Expansions and Contractions in Composite Indexes of Coincident
Indicators, Eight Countries, 1948-1991
Dates of
Turning Points Expansions Contractions
Mean Mean Mean Mean

Period First" Last> Duration Amplitude Duration Amplitude
Country Covered (Por T) (Por T) Number (months) (percent) Number (months) (percent)
Line (1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
1 U.S. 1948-1991 48(P) 90(P) 8 49 30.4 8 13 -9.1

2 Canada 1948-1991 51(P) 90(P) 5 82 57.8 5 12 -3.4
3 u.K. 1952-1991 52(T) 90(P) 6 63 29.2 5 15 -4.1
4 France 1955-1991 58(P) 91(P) 4 88 47.6 4 11 -5.8
5 Italy 1948-1991 48(T) 91(P) 6 77 60.2 5 18 -4.9
6 W. Germany 1950-1991 50(T) 91(P) 4 101 110.8 3 26 -6.5
7 Japan 1954-1991 54(T) 91(P) 3 140 430.7 2 10 -4.6
8 Australia 1950-1991 51(P) 90(P) 5 81 65.1 5 12 -7.5
Source: Center for International Business Cycle Research, Graduate School of Business, Columbia University.
• The following initial or early low values are assumed to represent trough dates: Italy, 111948; West Germany, 111950; Japan, 6/1954 (see
figures 1-2A and 1-2B).
b The following late or terminal values are assumed to represent peak dates: France, 111991; Italy, 111991; West Germany, 4/1991; Japan,
211991 (see figures 1-2A and 1-28).
Europe than here. This suggests that the persistent rise in European unem-
ployment was largely noncyclical in nature, which is consistent with most
recent hypotheses. 35
3.3 Possible Reasons for the Observed Differences
There are no tested hypotheses, let alone accepted knowledge, on why the
United States may be subject to more frequent and deeper recessions than
Japan, West Germany, or France. However, there is much debate about
the sources of differences in longer-term growth rates between the major
industrial and trading countries. Insights from this literature bear on the
present problem insofar as higher growth is in fact conducive to greater
cyclical stability.
The central argument here is simple. Investment in human and physical
capital produces advances in knowledge, technological progress, and
increased rates of growth in factor productivity, output, and real income.
What is desired, therefore, is higher maintained rates (and shares in total
output) of saving-cum-investment in real terms. 36 There is no need here to
cite the voluminous theoretical and empirical work on sources of economic
growth, which generally supports this basic position. As a practical out-
come, there is considerable consensus that the most successful economies
are those that have the highest average long-term rates of capital invest-
ment. For some time, the United States has had relatively low shares of
private saving and total productive investment; hence it is widely and vari-
ously urged to perform better in this respect, as its main competitors in
Europe and the Far East do.
On the other hand, fluctuations in business and consumer capital outlays
have long been recognized to be a major source of cyclical instability in
aggregate demand. The preceding argument seems to ignore or assume
away the problem. Since investment is much more cyclical than consump-
tion, is it not necessarily true that a larger share of investment will increase
the fluctuations of the economy, along with its growth?
The short answer is no. Investment can be both high and stable,
provided it is a part of, and a response to, growth in aggregate demand that
is sufficient to keep the economy near full employment. As the classical
long-run version of the accelerator theory has it, net investment is in the
end only justified by growth in the demand for the product of the new capi-
tal (which ought to be given the broadest interpretation). But, as noted
before, it is equally true that growth itself depends positively on prior rates
of investment. Increases in real capital, physical and human, generate
improvements in productivity that can reduce costs and prices. This fact is
particularly important in the competitive world of open economies.
In sum, higher rates of saving, investment, and growth can coexist with,
and indeed may favor, greater cyclical stability. The problem is how to
maintain them. But several heavily export-oriented countries in Europe
and Asia did very well on this score, and their experiences may hold useful
lessons.
The other side of the argument that more growth promotes more stab-
ility is that economic fluctuations can have costly long-term consequences
in the form of suboptimal investment and growth. In a recent model, firms
must make technology commitments in advance, and unanticipated vola-
tility causes errors in these decisions (for example, on the scale of a new
plant or size of the work force) that have substantial negative output
effects (Ramey and Ramey, 1991).37
To be sure, there are other factors to be considered. Business cycles
have historically been more frequent and shorter in the United States than
in Europe, as shown in section 2.1 and table 2-1. This difference points in
the same direction as that observed in recent times, but it is much weaker
and there are no large and systematic disparities in country growth trends
to account for it. To my knowledge, it was never explained. Possible
reasons include a larger role of short inventory fluctuations in the United
States and more frequent financial disturbances.
3.4 Growth Cycles
In countries where growth persisted at high rates and business contractions

occurred rarely and remained mild for a considerable time, even mere
slowdowns cause much public concern and indeed are often treated as
actual recessions are treated elsewhere. The reason is that the slow-
downs are themselves of significant duration, result in some rise in unem-
ployment and weak business conditions generally, and are usually
associated with absolute declines in some more cyclically sensitive activi-
ties and sectors. In the 1960s, when confidence about the long-lasting era of
high growth and rising prosperity reached its peak, business cycles seemed
increasingly "obsolete" and interest shifted to "growth cycles," that is,
fluctuations around the upward trend in a nation's economic activity
expressed in real terms (Bronfenbrenner, 1969; Mintz, 1969, 1974).
The concept of growth cycles is an old one, as illustrated by the fact
that early indexes of general business conditions and trade were more
often than not available only in the form of percentage deviations from
estimated trend or "normal" curves (see Zarnowitz, 1981, for a brief

history of this approach and references). Implicitly, the idea of growth
cycles has also long been popular with textbook authors, who refer vaguely
to business cycles being "fluctuations around the trend" and stress the
transitory nature of departures of the actual from the potential (full em-
ployment) output. But it is very important to distinguish clearly between
business cycles and growth cycles because the two differ qualitatively
and not just in degree. A slow expansion is still an expansion; the problems
posed by a contraction are quite distinct. Moreover, trends interact with
cycles, vary over time, and are difficult to isolate and measure.
In spite of these difficulties, much interesting work has been done in the
last 20 years on the postwar growth cycles in many countries, particularly
by Mintz at the NBER and by Moore and his associates at the CIBCR.38
When cautiously interpreted, the results can yield lessons on when and how
expansions speed and slow and retardations do or do not develop into
contractions.
The chronology of growth cycles is derived from the observed consensus
of the corresponding turning points in series of deviations from trend, just
as the chronology of business cycles is derived from the consensus of turn-
ing points in series of levels. In both cases, the same basic set of data
is used, namely, a country's principal coincident indicators-the main
comprehensive measures of GNP or GDP, personal income, sales, employ-
ment, and industrial production (all deflated or in physical units). In both
cases, too, peaks and troughs mark specific cycles in the indicators used
(whether they relate to the original levels or those in the trend-adjusted
series). The trends should be flexible and as free of any cyclical
components as possible (in the CIBCR studies they are determined by
interpolation between segments of the series determined with the help of
selected long-term moving averages).
This chapter is mainly concerned with business cycles, not growth cycles,
but a brief review of how the two units of observation are related is in
order. Figure 1-3 presents in schematic form the timing of growth cycle
peaks and troughs in six countries. The initial and terminal dates of
slowdowns that ended in renewed expansions are marked by ordinary dots,
the dates of slowdowns that ended in contractions are marked by heavier
black dots. The following table indicates that slowdowns without recession
prevailed strongly in Japan, less so in West Germany and Canada, while
slowdowns with recession were more common in the other three countries,
most notably the United States. 39 Growth cycles include both types of
slowdown; hence they are much more numerous than business cycles that
are defined by the presence of absolute decreases in aggregate activity
UNITED STATES P~IP

T
- -. --~- - - T
CANADA
UNITEDKNGOOM ~l-_=-~l~
WESTGERMANY ~L-....~_J~
FRANCE ~1 ~=~=_J~
.v.PAN ~I~ ~~I~
1948 50 52 54 56 58 60 62 64 66 68 70 72 74 76 76 80 82 84 86 88 90
Figure 1 -3. Timing of growth cycles in six countries, 1948-1980

Note: In each schematic diagram, the drawn-out lines connect the dates of the consecutive
peaks (P) and troughs (T) in growth cycles for the given country. Heavy black dots identify
slowdowns that became recessions.
Sources: For the U.S. business cycle chronologies---National Bureau of Economic
Research, Inc. For the growth cycle chronologies of Canada, United Kingdom, West
Germany, France, and Japan-Center for International Business Cycle Research, Graduate
School of Business, Columbia University.
(recessions). In the trend-adjusted indicator series, all major retardations

are reflected in specific-cycle declines that make up growth cycle contrac-
tions; some of these movements descend into the negative region, others
stop short of it.
U.S. Canada U.K. West Germany France Japan
Slowdown without recession 4 7 3 5 3 6

Slowdown with recession 7 5 5 3 4 2
Slowdowns occur either in the late stages of business expansions (some

time before peaks) or they interrupt long expansions. As a result,
expansions are shorter and contractions longer in growth cycles than in
business cycles. Growth cycles are more nearly symmetrical and less vari-
able than business cycles with respect to both durations and amplitudes of
their phases. Growth cycle peaks tend to occur before the corresponding
business cycle peaks, while the troughs of the matching growth cycles and
business cycles tend to be roughly coincident.
Figure 1-3 suggests a high degree of international correspondence in
the timing of growth cycles. Thus much (though not all) of the time the
diagrams for the United States and Canada, and those for the United King-
dom and West Germany, display nearly synchronous movements. Indeed,
these chronologies are considerably more similar than those of business

cycles that can be inferred from figure 1-2. This presumably reflects the
effects of the elimination or reduction of the divergencies in the national
trends.
4 The Observed Tendencies, Disturbances, and

Regularities
4.1 Endogenous Cyclical Variables
As we have seen, business cycles vary greatly in duration and intensity, less
in diffusion. They are not only diverse but also evolving. What they have
in common is not their overall dimensions but the makeup, features, and
interaction of their many constituent processes.
Thus in each cycle-whether long or short, large or small-production,
employment, real incomes, and real sales tend to expand and contract
together in many industries and regions, though at uneven rates. Other
variables-for example, hours worked per week, real new orders for
manufactured goods, and changes in prices of industrial raw materials-
rise and fall correspondingly but earlier, with variable leads. Still others
-for example, inventory-sales ratios, real business loans outstanding,
and changes in unit labor cost-also rise in business expansions and fall
in contractions but somewhat later, with variable lags. These sequential
movements usually recur in each successive cycle. They all are significantly
persistent and pervasive.
Systematic differences exist not only in the timing of cyclical movements
in different variables but also in their relative size and conformity or coher-
ence (that is, correlation with business cycles). Among the earliest and most
important observations in this area is that activities relating to durable
(producer and consumer) goods have particularly large and well-
conforming cyclical fluctuations. Other variables long and rightly viewed
as highly cyclical are business profits; investment in plant, equipment,
and inventories; and cost and volume of bank credit used to finance such
investments.
More than any other sector of the modem economy, manufacturing
has historically been central to both economic growth and fluctuations.
Mining, construction, transportation, communication, and public utilities
have varying but significant degrees of cyclical sensitivity. Other nonfarm
sectors, which produce largely services rather than goods, tend to be much
less responsive to business cycles. Employment in broadly defined service
industries (including government) followed a strong and smooth upward

trend. Agriculture, which underwent a long and strong downward trend in
employment and a huge secular rise in productivity, is but weakly cyclical,
except for prices. In sum, business cycles, which developed in the age of
industrialization, still affect most strongly industries producing goods,
especially durables (Zarnowitz and Moore, 1986, pp. 536-539).
Given the weight of these "stylized facts," it is not surprising that the
predominantly endogenous theories of business cycles, old and new, stress
the role of those variables that have clear patterns of regular participation
in the motion of the economy at large. These variables, which may be
called "cyclical," and the corresponding theories, follow;40
• Business fixed investment (overinvestment, vertical maladjustments,

accelerator).
• Inventory investment (multiplier-accelerator models with feedback
effects).
• Business profits (cost-price imbalances, profit margins and expecta-
tions).
• Credit and interest rates (unstable supply of bank credit; discre-
pancies between, and changes in, the money, real, and natural rates
of interest).
The cyclical variables are endogenous-generated by the system of

relationships within which they interact. For these variables to be primarily
responsible for business cycles, the system must have the required dynam-
ics in the form of some essential nonlinearities, or leads and lags, or both.
Such elements are undoubtedly important, but we are still far from really
understanding how they work and hence how far the endogenous models
can go in explaining business cycles.
4.2 Exogenous Variables and the Role of Money
Recent theories view economic fluctuations as resulting mainly from

changes in observable exogenous factors (variables) or unobservable ran-
dom shocks (errors). Here the cyclical variables, although still important
as propagators of business cycles, are no longer seen as the central part of
a system that can produce self-sustaining cycles.
Important policy variables such as the monetary base, tax rates, and
federal government spending have been traditionally treated as exogenous,
although this cannot be strictly true since they are clearly influenced by the
economy as well as influencing it. Money supply variables are of even more
mixed nature in this regard. The extent to which a central bank controls the
stock of money or (which is more relevant) its rate of growth depends on
factors that vary across countries and over time. They include the national
and international monetary systems; the powers, objectives, and perform-
ance of the bank; and the definition and composition of the supply of
money.
In the United States, narrowly and broadly defined monetary aggregates
normally trend upward in both expansions and contractions, though often
at reduced rates before downturns. Long absolute declines in Ml or M2 are
rare and as a rule associated with business depressions or stagnations.
Monetary growth rates tend to lead at business cycle peaks and troughs but
by intervals that are highly variable and on average long (Friedman and
Schwartz, 1963a, 1963b). Cyclical changes in the deposit-reserve ratio and
particularly the deposit-currency ratio, which reflect the chain of influence
that runs from business activity to money, contribute on average strongly
to the patterns of movement in money growth during business cycles
(Cagan, 1965; Plosser, 1991). Money shows a systematic tendency to grow
faster in business expansions than in contractions, and so does the domestic
nonfinancial credit, but the stability of monetary relationships in the cycli-
cal context is subject to much doubt and debate (Friedman, 1986; Meltzer,
1986). Certainly, there is more regularity in the long-term relation between
money growth and changes in the price level, and in the procyclical
behavior of interest rates and the income velocity of money (allowing for
its long trends).
For all of this, there is no denying that business cycles have important
monetary and financial aspects. A six-variable, four-lag quarterly vector
autoregressive (VAR) model applied to postwar and earlier U.S. data by
Zarnowitz and Braun (1990) shows that the rate of change in real GNP was
significantly affected by lagged rates of change in the monetary aggregates
(base, Ml, and particularly M2) but much more strongly yet by lagged
values of short-term interest rates. However, the influence of changes in
the planned volume of fixed and inventory investment and purchases of
durable goods (represented by series of new orders and contracts) proved
to be strong as well. 41
4.3 Monetary, Real, and Expectational Shocks
In the original monetarist theory, the money supply was treated as the
main exogenous factor driving the business cycle. In the more recent
equilibrium version with rational expectations, anticipated money growth

can influence only prices, not output; that is, only unanticipated money
shocks have real effects. Tests of this hypothesis have produced evidence
that is mixed but mostly unfavorable in the sense of not confirming the
importance of the distinction between the effects of the anticipated and
unanticipated components of monetary change. 42
Full exogeneity of money is one extreme and unrealistic assumption;
full endogeneity of money is another, at the opposite end of the spectrum.
The latter view posits that changes in inside money accommodate the
money stock to the level of economic activity; it is accepted both in
some Keynesian disequilibrium models and the real business cycle (RBC)
models in which fluctuations are strictly equilibrium phenomena. 43
In the RBC model, stochastic oscillations in the economy's real growth
path result from a mixture of transitory and permanent shocks to pro-
ductivity (Kydland and Prescott, 1982). There is a long gestation lag in the
production of capital goods, which imparts some persistence to output
movements. As productivity gains fluctuate, so do real wages (or, more
accurately, real returns on the work effort) to which labor supply responds
very elastically.
But the cyclical sensitivity of real wages is low, and tests of the hypo-
thesis of high intertemporal substitution of leisure are mostly negative (see,
for example, Altonji, 1982). To explain booms and busts, sufficiently large
and frequent shocks to productivity would be necessary, yet the evidence
for them is hard to find (or nonexistent: Has there been any technological
regress of a large-scale or recurrent nature?). When interpreted as
productivity shocks, the procyclical movement of the Solow residual
supports the RBC hypothesis, but it overlooks the high probability that
much of this movement reflects the effects of labor hoarding and measure-
ment errors (as noted in several critiques; see, for example, Eichenbaum,
1990). Indeed, a recent study, whose authors include two RBC theorists,
finds that productivity disturbances account for no more than 35% to 44%
of total output fluctuations in a cointegrated V AR model with both real
and nominal variables (King et aI., 1991).
Much of the recent work that seems to be concerned with explaining
business cycles is actually preoccupied with such perennial theoretical
issues as the rationality of economic expectations and behavior and the
neutrality of money. The predominant view is that stable growth at full
employment would prevail in the absence of outside shocks; moreover, the
shocks tend to be reduced to one or two broad categories per model. Thus,
as already noted, the equilibrium theorists concentrate either on unanti-
cipated changes in money and prices or on shifts in technology. Some
52 mE BUSINESS CYCLE
attention is given to shifts in preferences and desired consumption (Hall,

1986). One study singles out sectoral demand shifts (Lilien, 1982), another
oil price shocks (Hamilton, 1983). The role of shocks to, and imperfections
of, credit markets is stressed by some authors (Wojnilower, 1980; Green-
wald, Stiglitz, and Weiss, 1984), while the direct effects of independent
shifts in monetary policy are central to others (Romer and Romer, 1989).
However, there is also considerable and perhaps increasing recognition
that macroeconomic fluctuations may have many important sources.
Blanchard and Watson (1986) distinguish four major categories of
shocks-to aggregate demand, aggregate supply, fiscal policy, and money
supply-and conclude that each made a roughly equal contribution to U.S.
postwar cycles. Occasional large disturbances are at work along with
frequent small shocks, so that business cycles are quite differentiated.
Recent papers by Blanchard and others generally assume that the effects of
supply shocks are more persistent than those of demand shocks.
That a variety of factors can and do serve as proximate causes of busi-
ness cycles is an old idea embodied in the synthetic theories of several early
writers, such as Aftalion, Mitchell, Schumpeter, and very explicitly Pigou
(1927). Evidence from recent macroeconometric models supports this
concept. Unlike the early Keynesian models dominated by demand factors
and fiscal policy, the present models give considerable attention to supply
factors as well and actually have monetary policy matter more than fiscal.
In large multisector models the distinction between demand and supply
shocks is rather blurred anyway.
Since the incisive and detailed analysis of the monetary history of
the United States by Friedman and Schwartz (1963a), the monetarist
interpretation of business cycle developments gained and still retains a
wide acceptance, despite critical countercurrents within the profession. It
is simply very difficult to deny that strong restrictive monetary measures
designed to combat inflation have contributed to subsequent economic
downturns at least in some well-known episodes (notably the Federal
Reserve increases of the discount rate in 1920 and of reserve requirements
in 1936-1937). Similarly, the shift to a policy of lower money growth and
sharply higher interest rates, first signaled in October 1979, was followed by
the "Volcker disinflation" and the recessions of 1980 and 1981-1982. In
each case, however, these monetary shocks were accompanied by changes
in factors unrelated to monetary policy that must have also contributed to
the recessions that followed (fiscal stringencies in 1920 and 1936-1937, an
oil shock and financial deregulation in the late 1970s, a fall in money
velocity in the early 1980s). So monetary and real disturbances (or unusual
developments) coincided, and their effects are difficult to disentangle.
Nor are these the only types of disturbances to be considered, since

there are also the "psychological" factors emphasized early by Pigou, that
is, the shifts in expectations that may at times spread widely enough to
become self-fulfilling. In contemporary models of financial crises or specu-
lative price bubbles, multiple equilibria exist, many of them involving
expectations that appear arbitrary ex post yet are rational ex ante: Many
act on them because they believe that others will (see Woodford, 1987, for
a brief review).
The controversy about real versus monetary theories of business cycles
is now very old; although it has been instructive, prolonging it further is apt
to produce more heat than light. It is time to recognize that short-term
variations in economic activity are influenced by forces of each of the three
types-real, monetary, and expectational. Their mix varies, and the really
important question is how they interact and why.
The 1990-1991 recession presents an interesting example, although
more time and data will be needed to assess it. It was preceded by a long
slowdown starting early in 1989 following a sharp decrease in monetary
growth and increase in interest rates. Both monetary policy and credit
conditions tightened. Efforts to counteract the persistently large govern-
ment deficits made fiscal policy quite tight, too. Real disposable income
and real domestic final sales flattened and then turned down early. So both
monetary and real factors worked to depress enconomic activity. They
were finally joined by a sharp drop in consumer expectations and business
confidence, which occurred after Iraq's invasion of Kuwait raised oil prices,
the threat of war, and the general level of uncertainty about inflation and
unemployment. Thus the last blow was delivered to the old and by then
seriously weakened expansion.
4.4 Leads and Lags
Presumably, one important reason for the differentiation of business

cycles is that they are affected by large disturbances of all kinds, including
shifts in policies. Also, changes in the economy's structure contribute to an
evolution of cyclical behavior that is more predictable, though not without
some important exceptions. 44 Still, business cycles are like individuals
of a species: diverse yet also alike in many essential respects (think of
other events of which the same can be said, wars, for example). The
main common element, as we already know, consists of the properties
of co-movements in cyclical variables.
Table 1-7 covers 32 U.S. monthly indicators for which there are avail-
able both the traditional measures of cyclical timing and variability
Table 1-7. Selected Measures of Average Cyclical Timing, Variability, and Correlation with an Index of Coincident Indicators,
32 Monthly Series, 1948-1980 and 1959-1989
Median Timing (months)b Average Change (%)C Correlation wlNew IC]d
Series Title Series Cyclical Lag (k)

(BCD Number)" Peaks Troughs All Turns (s.a.) Component Transformation (months) r(k)
Line (1) (2) (3) (4) (5) (6) (7) (8) (9)
1 Average workweek, mfg (*1) -11 -1 -4.5 .5 .2 N -1 .79

2 Unempl. insur. claims, inv. (*5) -12 0 -5.5 5.4 2.4 DLN -2 -.87
3 Help wanted index (46) -7 +2 -2.5 2.9 2.2 DLN -1 .95
4 Capacity utilization, mfg (82) -11 +1 -1.5 1.6' 1.4' D -1 .96
5 New orders, duro goods, c$ (7) -8 -1 -3 3.4 1.4 DLN -1 .90
6 N.D., cons, goods & mtls, c$ (*8) -12 -1 -4.5 2.8 1.3 DLN -2 .89
7 Vendor performance (*32) -6 -4 -6 3.8X 2.4x N -2 .67
8 Contr. & orders, plant & equip ("'20) -8 -1 -3.5 6.2 1.6 DLN -1 .76
9 New bldg permits, priv. housing -13 -3 -9.5 5.1 2.4 N -6 .59
(*29)
10 New bus. incorporations, no. (13) -10 -1.5 -5 2.5 1.0 DLN -3 .55
11 Change in mfg & trade invent. (31) -6 -2 -3.5 7.1' 1.4x DLN 6 .79
12 Stock price index, S&P 500 (*19) -9.5 -4 -5.5 2.7 1.6 DLN -6 .55
13 Change in sensitive mtls prices (*99) -9.5 -3.5 -5.5 .3x .2x DLN 0 .55
14 Change in money stock, Ml (85) -15 -3 -14 .3' .03x DLN -3 .25
15 Money supply M2, c$ (*106) -16 -3 -10 .42 .37 DLN -6 .68
16 Change in bus. & cons. credit (*111) -11 -2 -7 2.9' .8x N -2 .67
17 Liab. of bus. failures, inv. (14) -14 -2 -10 27.8 2.5 DLN -3 -.40
18 Employees on nonag. payrolls (*41) 0 0 0 .32 .28 DLN 1 .92
19 Pers. income - transfer pmts, c$ 0 -0.5 0 .5 .4 DLN 0 .87
(*51)
20 Mfg & trade sales, c$ (*57) -3 -1 -1.5 1.0 .6 DLN -1 .94
21 Index of industrial production (*47) -2 0 0 .9 .7 DLN 0 .99
22 Indus. prod., duro mfg (73) -3 0 0 1.4 1.0 DLN 0 .98
23 Indus. prod., nondur. mfg (74) -1 -1 -1 .7 .6 DLN -1 .92
24 Retail sales, c$ (59) -4 -2 -2.5 1.1 .5 DLN -2 .69
25 UnempL, aver. duration, inv. (*91) 0 +8 +3.5 3.6 1.6 N 6 -.88
26 Invent.lsales, mfg & trade, c$ (*77) +8 +13 +10 .02x .01' N -2 -.79
27 Unit labor cost, mfg, detrend. (*62) +5 +13 +10 .6x .3 x DLN -2 -.69
28 Comm. & indus. loans outst., c$ +4 +6 +4.5 .9 .7 DLN 10 .63
(*101)
29 Cons. instal. credit/pers. inc., % +4 +7 +6 .07x .05x DLN 6 .57
(*95)
30 Treasury bill rate. 3 mo . .(114) +1.5 +3.5 +2 .28x .17x D 1 .52
31 Federal funds rate (119) +2 +5 +2.5 .38x .26' D 1 .59
32 Yield, high grade corp. bonds (116) +1 +2 +1.5 .18x .09x D -12 -.45
Table 1-7. (Continued)
Sources: U.S. Department of Commerce (1984, Tables 8 and 9A, pp. 167 -68, 172-68, 172-75 [for columns 1-6)); Stock and Watson (1990, table
1 [for columns 7-9]).
a Series titles are abbreviated. The BCD numbers in parentheses are series numbers used by the U.S. Department of Commerce in its monthly
publications Business Conditions Digest (BCD, through March 1990) and Survey of Current Business (SCB, current). * identifies a series included in
the composite index of leading or coincident or lagging indicators (referring to the indexes of the Bureau of Economic Analysis in the Commerce
Department as of 1984).
b The measures cover the seven pairs of business cycle peaks and troughs between 1948 and 1980. The median is the middle value in an odd-
numbered array and the average of the two middle values in an even-numbered array. Minus (-) signs are leads at business cycle turning points; plus
(+) signs are lags.
C Entries in column 5 are average changes, without regard to sign, between consecutive values in a seasonally adjusted (s.a.) series. Entries in
column 6 are average changes, without regard to sign, between consecutive values in the cyclical component, which is a smooth, flexible moving
average of the seasonally adjusted series. Entries marked x are average actual changes in the series, in its original units of measurement; all other
entries are average percentage changes.
d The transformation codes (column 7) are N = no transformation; D = first differences of the series; DLN = first differences of the natural logs
of the series (Le., growth rates). The transformations are generally from the original levels of the (s.a.) series. Entries in column 8 (k) are leads (-) or
lags (+), in months, which are associated with the maximum correlations r(k), as listed in column 9. These statistics are based on the cross-
correlogram between filtered log ICI and the filtered series, using the 24-month moving average filter a24 (L). ICI is the new experimental index of
coincident indicators by Stock and Watson; it covers the period January 1959 to May 1989.
Abbreviations: rnfg = manufacturing; duro = durable; insur. = insurance; mtls = materials; invent. = inventories; contr. = contracts; comm. =
commercial; cons. = consumer; instal. = installment; pers. = personal; inc. = income; outst. = outstanding; pmts. = payments; nonag. =
nonagricultural; / = ratio (inventory/sales in Line 26; credit/income in Line 29); c$ = in constant dollars; inv. = inverted (peaks in the series are
matched with business cycle troughs, troughs in the series are matched with business cycles peaks); detrend. =adjusted for (deviations from) trend.
(columns 2-6) and new measures of filtered cross correlation with an index
designed to approximate aggregate economic activity in real terms. The
first set of statistics, from the Bureau of Economic Analysis (Commerce
Department), refers to the 1948-1980 period; the second, from an NBER
study by Stock and Watson (1990), refers to the 1959-1989 period. The two
are entirely independent and based on very different methods, yet they are
generally consistent and complementary.
The first 17 series listed in table 1-7 are classified as leading at both
peaks and troughs of business cycles and include ten components of the
BEA index of leading indicators as of 1984 (marked with an asterisk, "*").
They reflect marginal employment adjustments, which under uncertainty
are made ahead of decisions that result in changes of employment and
unemployment (lines 1-3); adjustments of delivery periods and activities
marking the early stages of investment processes, which tend to lead
production, shipments, construction, and installation of equipment (lines
4-10); changes in inventories and sensitive prices (lines 11-13); and
changes in money and credit conditions (lines 14-17).
Such series have a long history of leading at business cycle peaks and
troughs. They are heavily represented in a group of 75 series whose median
timing at 15 peaks and 16 troughs of the 1885-1937 period was -6 (5) and
-5 (3), respectively (that is, leads of six and five months, with standard
deviations of five and three months; for sources and detail, see Zarnowitz
and Moore, 1986, pp. 565-571). In 1948-1980 the means (s.d.), in months,
of the median leads were -11 (3) at peaks and -2 (2) at troughs (as
calculated from the entries in table 1-7, lines 1-17, columns 2 and 3). Thus,
the relative timing of these indicators remained remarkably consistent
over the past century, although the leads at peaks have become longer
and the leads at troughs shorter since 1948.45
Leads maximize correlations with an index of coincident indicators in
1959-1989 for all but two of the series used in table 1-7, lines 1-17,
columns 7-9.46 This is a strong confirmation of the tendency of these
series to move ahead of aggregate economic activity (output, employment,
real income, and real sales). But because these leads (column 8) average -
2 (3) months, they are considerably shorter than the corresponding "all
turns" leads (column 4), which average -6(3) months.
This would be expected for the following reasons. The differences in
timing are likely to be larger and more systematic for the major cyclical
movements than for the many small and short variations that may often
be caused by random influences affecting many variables more or less
simultaneously. The cyclical measures in table 1-7, columns 1-3, refer
mostly to the indicator levels (except where the title specifies change), but
the correlation measures in columns 7 -9 refer mostly to smoothed rates of

change, including those in the coincident index itself (note the prevalence
of DLN = f:, In in column 7).47 Thus the former relate directly to business
cycle turning points, whereas the latter relate directly to all observations
for growth cycles. The series that lead at business cycle turns also tend to
lead at growth cycle turns, in the United States and elsewhere (Klein and
Moore, 1985). But the leads at business cycle peaks are particularly long,
reflecting high levels of demand and capacity utilization, large backlogs
of unfilled orders, and extended delivery periods. Growth cycle peaks
typically occur several months before business cycle peaks and have
shorter leads and signals (Zarnowitz and Moore, 1982).
It should be noted that table 1-7 covers monthly indicators only. Hence
the table omits one particularly important class of leaders available in
quarterly form only, namely, corporate profits (totals, rates, and margins,
before and after taxes, in current and constant dollars; also net corporate
cash flows). Profits decline in late stages of expansion well before sales
do because costs start rising faster than selling prices, depressing profit
margins in many businesses. Labor markets get tight and wages rise even
as productivity slackens; interest rates and the cost of holding inventories
rise; but product prices tend to increase less, being held back by prior
commitments and domestic and international competition for market
shares. (Mitchell hypothesized such developments as a major cause of busi-
ness cycles in his earliest treatise of 1913). Stock price indexes tend to have
strong cyclical leads (line 12) because they anticipate or reflect the early
movements in profits and also in inverted interest rates, as well as probably
other changes such as those in money and credit (lines 15 and 16).
Roughly coincident indicators of employment, production, real personal
income, and real sales are listed in lines 18-24 of table 1-7 (the four
marked with an asterisk are components of the BEA coincident index).
Here coincidences (0) or very short leads (-1) dominate both the all-turns
and maximum-correlation timing (columns 4 and 8), but in a few cases
longer leads appear, mainly at peaks, for industrial production and sales.
All of these series have pronounced growth trends and are used in log
differences for the correlation measures.
The indicators listed in lines 25-32 of table 1-7 are classified as lagging
in the historical NBER studies and BEA reviews of cyclical indicators (the
five marked with an asterisk are components of the BEA lagging index).
The common feature of these series is their association with the cost of
doing business. Increases in the duration of unemployment increase the
private and social burdens of joblessness and insurance against it. Labor
cost per unit of output is the largest component of average costs on an
economywide scale. Interest rates measure the costs of credit and financing
investments of various types. The larger the inventories are relative to
sales, and the larger the volume of credit is relative to income, the heavier
are the costs of finance, given the level of interest rates. Accordingly, the
lagging increase in these cost factors around the business cycle peak works
to discourage economic activity, and their lagging decrease around the
trough works to encourage it. Indeed, when inverted (divided into one), the
lagging index shows the earliest cyclical signals that anticipate even the
turning points in the leading index (Moore, 1983, chap. 23).
The eight series in the last section of table 1-7 have sizable median lags,
averaging in months about +3(3) at peaks, +7(4) at troughs, and +5(3) at all
turns (columns 2-4). It is broadly consistent with the lagging nature of
these indicators that five of them also show the best correlations for lags
(which are long, except for the two series of first differences in short-term
interest rates; see columns 7 -9). For three series that tend to lag at busi-
ness cycle turns (inventory-sales ratio, unit labor cost, and corporate bond
yields), the correlations with the coincident index refer to leads, but they
are negative. This is consistent with the old and well-established finding
that the inverted lagging indicators lead. 48
4.5 Co-movements and Amplitudes
The typical leads and lags of the major economic variables constitute
important and relatively enduring features of cyclical behavior. Another
class of such characteristics relates to how close the co-movements of the
variables tend to be in the course of business expansions and contractions.
Still another consists of relative amplitudes of cyclical change in the differ-
ent processes.
The correlations in the last column of table 1-7 provide estimates of
the co-movements of the selected indicators with an index that approxi-
mates well aggregate economic activity on a monthly basis. The Stock-
Watson index of coincident indicators (denoted leI for simplicity)
resembles the BEA coincident index rather closely but has a formal
probabilistic interpretation. Of the leading series (lines 1-17), capacity
utilization, the help-wanted index, and the two aggregates of new orders
show the highest positive r(k) of .89-.96, while unemployment insurance
claims has -.87 (lines 2-6). Average workweek, contracts and orders for
plant and equipment, and inventory change show r(k) correlations of
. 76-.79; seven other indicators, including in descending order real money
supply, change in credit, vendor performance, housing permits, and stock
prices, have an r(k) of .55-.68; and for two series these measures are very
low indeed (-.40 for failure liabilities and.25 for M1 growth).
Not surprisingly, the correlations of the individual coincident indicators
with the ICI are high (lines 18-24). They range from .92 to .99 for five of
the series, are .87 for real personal income less transfer payments, but are
only .69 for real retail sales. Industrial production indexes for total and
durable manufacturing show the highest r(k) coefficients.
The co-movements of the lagging indicators with the ICI are not very
close. The highest correlations here are the negative ones for unemploy-
ment duration, inventory-sales ratio, and unit labor cost (lines 25-27). Two
series on credit and two on short-term interest rates have an r(k) of .52-
.63, and the bond yield has one of only -.45 (lines 28-32).
The conventional method of analyzing time series for the study of busi-
ness cycles is to estimate and then eliminate their seasonal components.
Working with seasonally adjusted series Gust as working with trend-
adjusted or heavily smoothed series) is often covenient but not without
risks. 49 Even after seasonal adjustment, the month-to-month change in
most series contains a large component of short, random variations; the
cyclical component represents a much longer and smoother movement that
is often much smaller than the total change on a per-month basis (compare
the paired entries in columns 5 and 6).50 Some leading indicators show
great volatility of monthly change, which obscures the relatively small
trend- cycle movement (inventory investment and business failures provide
extreme examples; see lines 11 and 16). In contrast, some other early-
moving series are remarkably smooth to begin with and are dominated by
cyclical fluctuations (for example, the help-wanted index, capacity
utilization, and vendor performance, lines 3, 4, and 7) or by the trend-cycle
(for example, real money supply, line 15). Most of the leading indicators
fall somewhere in the broad intermediate range; that is, they display great
overall sensitivity and have both clear, specific cycles and many smaller
random oscillations.
The coincident indicators have generally much smaller, and also more
uniform, amplitudes of monthly total and cyclical change than the leading
indicators. The following tabulation, which uses only percentage entries
from columns 5 and 6, brings out the contrast between the two groups.
Behind it are important particulars. Thus consider manufacturing of dur-
able goods, where much of the production is to order. Here new orders
move in large swings that are followed with variable but significant lags and
much smaller fluctuations in output and shipments. The resulting changes
in backlogs of unfilled orders and average delivery lags are also large, lead-
ing, and procyclical. Production of nondurable goods is both less cyclical
and less volatile than that of durable goods (lines 22 and 23), and
production of services (not shown) is least, again on both counts.
Total Change Trend-Cycle Change
Leading (11 series) 5.4 (7.6) 1.5 (0.8)

Coincident (7 series) 0.8 (0.4) 0.6 (0.2)
Other related or similar examples may be mentioned. 51 New housing

permits (and starts) lead residential investment, which is somewhat
smoother. New orders and contracts for plant and equipment lead business
capital outlays, which fluctuate much less and lag behind considerably,
particularly at troughs. Other commitments and activities marking the
early stages of investment processes are new capital appropriations, new
business formations and incorporations, and new bond and equity issues
(all strongly procyclical and leading), whereas other late-stage processes
are completions of industrial and commercial construction projects,
shipments, and installations of machinery.
Interest rates tend to lag at business cycle turns, especially troughs, but
they are both strongly influenced by, and themselves strongly influence,
general business conditions. All vary procyclically, but the short-term rates
have larger cyclical movements than do long-term rates (lines 30-32).
Their changes are measured in their own units, as is also the case for some
other series in this group (lines 25-26), so no amplitude comparisons can
be made here with measures expressed in terms of percent changes. But let
us note that several important lagging indicators have strongly predomin-
ant trend-cycle components and are very smooth (notably so for commer-
cial and industrial loans outstanding and the ratio of consumer installment
credit to personal income; see lines 28 and 29).
5 General Conclusions
In this chapter I have considered the short but not-so-easy question, "What
is a business cycle?" from several angles. Section 1 looked at economic
history and at the development of thinking about business cycles. Section
2 reviewed the chronologies of business expansions and contractions,
the lessons from the duration data, the concepts of periodicity of cycles
and phases, and the evidence and reasons for the apparent moderation
of macroeconomic fluctuations in the post-World War II era. Section 3
compared the postwar business cycles and growth cycles in several major
industrialized, market-oriented countries. Section 4 discussed the behavior

of endogenous and exogenous variables; the role of monetary, real, and
expectational shocks; the systematic timing sequences (tendencies to lead
or lag); and the regularities of cyclical co-movements and amplitudes.
Understanding business cycles can be aided by each of these modes of
analysis.
Some sections of this study interpret history, others literature; some
discuss old, others new, findings and ideas. Many of the results refer to
particular aspects of theory and evidence, and there is no need to restate
any of them. Hence only a few general conclusions are collected here.
1. A business cycle includes a downturn and contraction followed by an
upturn and expansion in aggregate economic activity, which ideally should
be represented by comprehensive and reliable measures of total employ-
ment, output, real income, and real expenditures. (But nominal income
aggregates and price indexes can serve as good criteria, provided that the
fluctuations extend to a broad class of sufficiently flexible prices, as was the
case in the now rather distant past.)
2. A business cycle is pervasive in the sense that it consists of co-
movements and interactions of many variables. The regularity, magnitude,
and timing of the fluctuations vary across the variables, and these
differences are in part systematic. Thus most activities are procyclical,
mildly or strongly, but some are countercyclical. Some variables tend to
have approximately coincident timing, others tend to lead, and still others
tend to lag. These patterns might or might not be symmetrical; for example,
timing might be systematically different at peaks and troughs.
3. A business cycle is at least national in scope, that is, it involves most
industries and regions of a country, though again with variations in inten-
sity and timing. It can attain much larger dimensions when transmitted
across countries through channels of international trade and finance.
4. As a rule, a business cycle lasts several years and so is sufficiently
persistent for serially correlated as well as intercorrelated movements in
many variables to develop sequentially in the downward as well as upward
direction. The movements tend to cumulate before reversing themselves.
5. In most but not all business cycles, prices in general move pro-
cyclically, at least apart from their long trends (in the last half-century,
upward). This indicates a long and large role for the fluctuations of, and
disturbances to, aggregate demand. These can be of real, monetary, or
expectational origin and can well involve interactions between any or all
such factors. But supply shifts are also part and parcel of business cycles,
and they can be dominant in some. (The two major oil price shocks in the
1970s have been associated with inflationary recessions.) The cyclical
instability of profits, investment, and credit has a long history and is well-
documented.
6. Business cycles have varied greatly over the past 200 years in length,
spread, and size. They included vigorous and weak expansions, long and
short; mild and severe contractions, again some long, some short; and many
moderate fluctuations of close-to-average duration (about three to five
years). But since the 1930s the United States suffered no major depression.
Business expansions have become longer, recessions shorter and milder.
The probable reasons include the shift of employment to production of
services, automatic stabilizers, some financial reforms and avoidance of
crises, greater weight and some successes of governmental actions and
policies, and higher levels of public confidence.
7. The postwar recessions are much fewer and generally milder still in
France, Italy, and particularly West Germany and Japan. These countries
also had much higher average rates of real economic growth than the
United States, especially in the early reconstruction phase of the post-
1945 era. In the 1970s and 1980s growth decreased everywhere and
the recessions became more frequent and serious. All this suggests that
countries and periods with stronger growth trends are less vulnerable to
cyclical instability. This is a potentially important proposition in need of
explanation and testing (with possible extensions to regions, industries, and
the like).
8. In sum, business cycles make up a class of varied, complex, and
evolving phenomena of both history and economic dynamics. Theories or
models that try to reduce them to a single causal mechanism or shock seem
to me altogether unlikely to succeed.
Notes
1. In a letter of June 27, 1772, David Hume informed Smith that "We are here in a very
melancholy situation: Continual bankruptcies, universal loss of credit. ... Do these events
any-wise effect your theory?" Smith's answer is unknown (see Mirowski, 1985, pp. 15-18).
2. See Haberler (1964), Zarnowitz (1985), Moore and Zarnowitz (1986), Backhouse
(1988), and Sherman (1991).
3. Thus, for England he listed 14 crises in 1803-1882 at intervals averaging 6.1 years, with
a standard deviation of 2.6 and a range of 1-10 years (see also Burns and Mitchell, 1946,
p.442).
4. See Burns and Mitchell (1946, table 16, pp. 78-79).
5. For arguments and some evidence consistent with the previous discussion, see
Backhouse (1988, chs. 2 and 5).
6. Schumpeter refers to Mitchell and Spiethoff as disagreeing with him on this point.
7. See Mirowski (1985, pp. 201-213). The other sources of the crisis chronologies in ques-
tion are Bouniatian (1908), Ashton (1959), and Deane (1967). All authors agreed on two
dates, three agreed on six, and two agreed on ten (out of the 13 dates proposed by Ashton).
For 1700-1802, Ashton suggested a sequence of 16 "cycles," with durations averaging about
five years and concentrated mainly in the three-to-six-year range (Moore and Zarnowitz,
1986,pp.741-743).
8. The coverage of the annals is United States and England, 1790-1925; France, 1840-
1925; Germany, 1853-1925; Austria, 1867-1925; and 12 other countries on four continents
(various periods between 1890 and 1925).
9. This is certainly consistent with the recent evidence on the performance of business
outlook analysts, policy makers, and forecasters (see Fels and Hinshaw, 1968; Zarnowitz,
1967,1974).
10. Of course, many seemingly familiar facts or events are complex or controversial
enough to elude a tight formulation of meaning, say, in a definitional or legal sense. One is
reminded, for example, of Justice Potter Stewart's writing, in the 1964 case of Iacobellis vs.
Ohio, that obscenity was indeed difficult to define, but "1 know it when I see it."
11. For some detail on the historical statistics applied in the construction of the business
cycle chronology for the United States, see Zarnowitz (1981) and Moore and Zarnowitz
(1986), and references therein.
12. For a further analysis along these lines, see Zarnowitz (1985, 1992).
13. From the introduction by Burns to the posthumous book by Mitchell (1951, pp. VII-
VIII).
14. Something similar could be said as well about some earlier important influences,
notably that of Marx.
15. Early accelerator models made investment a function of changes in output (Aftalion,
1913; Clark, 1917). The more general and satisfactory formulation equates investment to some
fraction of the gap between the desired and actual capital stock.
16. This excludes the expansions during the Civil War and World War I periods, which
were longer than the average and different in some respects. Business contractions that
followed these expansions had shorter than average durations. But these distinctions are
blurred by the fact that some peacetime expansions were even longer and some peacetime
contractions were shorter. Compare lines 5 and 6 in table 1-1.
17. The rest of this section draws on material in Moore and Zarnowitz (1986, pp. 754-
758).
18. During the short but sharp and widespread 1920-1921 decline, Germany was insulated
by hyperinflation and the associated floating exchange rate (Friedman and Schwartz, 1963a,
p. 362). During the similarly diffuse and painful 1937-1938 contraction, Germany was no
longer a free market economy but a controlled economy under the Nazi dictatorship and
heavily engaged in arming for the war, annexing Austria, invading Czechoslovakia, and
threatening other neighboring countries.
19. The first of these is covered mainly by annual data (the duration measures are
expressed in monthly terms for comparability). Combining the two periods would add little to
the story.
20. This contraction is omitted from the chronology of Spiethoff (1955, vol. 1., p.147).
21. However, Friedman and Schwartz do not recognize the 1901 trough and the 1903 peak
included in the NBER chronology for Britain (1982, p. 74).
22. Thus, assuming that 1869-1870, 1887-1888, and 1899-1900 represented major slow-
downs rather than marginal recessions, the 1857-1918 period would contain 12 (not 15) busi-
ness cycles with an average length of 5 (not 4) years.
23. Indeed, using the latter would only strengthen the case, since the relative frequency of
expansions has been somewhat greater in the long upswings of prices, that of contractions in
the long downswings (Zarnowitz and Moore, 1986, p. 530).
24. The results of this combination include a sharp increase in real debt burdens; distress
selling of assets to payoff the debts; contraction of deposits; declines in production, trade, and
employment; depressed confidence; and lower nominal but higher real interest rates. When
the liquidation process and the debts and costs are reduced sufficiently in real terms, however,
confidence will gradually return, hoarding will give way to new buying and lending, and
reflation will pave the way to another recovery.
25. For more on the association between the price trends and the relative duration of
business cycle phases, see Zarnowitz and Moore (1986, pp. 530-531).
26. A hazard function shows the dependence of the termination or "failure" probability
on the duration of the process; in this case, the probability of a peak (trough) as a function of
the length-to-date of the preceding expansion (contraction). Let F(t) = Pr(T < t) be the prob-
ability that the duration random variable T is less than some value t; f(t) = dF(t)ldt be the
corresponding density function; and Set) = 1 - F(t) be the corresponding "survivor function."
Then the general form of the hazard function is A(t) = f(t)/S(t) (see Kiefer, 1988).
27. The following discussion draws in part on material from Zarnowitz (1985, 1987).
28. Consider the consequences of such natural disasters as cyclones and storm surges in
the Bay of Bengal (for India and Bangladesh) and hurricanes and shifts in ocean currents (for
fishing off the coast of Peru).
29. See Alt and Chrystal (1983). For some more positive appraisals and recent
contributions, see Willett (1988).
30. The measures are based on estimates by Balke and Gordon (1986, pp. 788-810) that
use the historical annual series by Gallman and Kuznets and more recent series of the Depart-
ment of Commerce. Persons' index of industrial production and trade (1875-1918) and the
Federal Reserve System index of industrial production (1919-46) were used as quarterly
interpolators.
31. Similarly, the index for Taiwan available since 1961 shows only one serious decline, of
9.1 %, in 12173-2175 (a minor decline of 1.3% occurred in 1190-8/90). Unlike Taiwan and
Japan, the economy of South Korea escaped a recession in 1973-1975 but had one in 1979-
1980. The Korean index declined by 10.3% in 3179-11180, its only lapse from growth since its
beginning in 1963.
32. Taking into account the divergences between the periods covered by the indexes does
not alter this conclusion.
33. For a description of the international indicators, see Moore and Moore (1985).
34. The Japanese series refers to the number of employees on payrolls in all non-
agricultural establishments (private and government-owned). The series for the United States,
Canada, France, and Italy cover nonfarm employment; for the United Kingdom, manu-
facturing, mining, construction, and public utilities; and for West Germany, mining and manu-
facturing.
35. The theories of hysteresis explain the persistence of high unemploymcnt by making
long-run equilibrium depend on history. Adverse shocks reduce demand for physical or
human capital or both as well as the demand for labor. Or "insider" workers keep real wages
high and "outsider" workers out of jobs (Blanchard and Summers, 1986).
36. This assumes that investment is predominantly productive in nature, well allocated by
market forces, and supported rather than hindered by government policies.
37. In this model, volatility is associated with productivity shocks, but I suspect that the
analysis of the Rameys has broader applicability.
38. See Klein and Moore (1985), Klein (1990), and Lahiri and Moore (1991).
39. This count does not include the 1990-1991 recessions in the United States, Canada,
and the United Kingdom or the concurrent slowdowns or possibly recessions in any of the
other three countries. Although the initial dates (peaks) of some of these movements have
been established, at least tentatively (see figure 1-3), their severity and duration cannot as yet
be known.
40. The listing uses broad labels and offers only selected types of theory. The theories are
in general not mutually exclusive and are often used in combinations, with varying emphases.
See Haberler ([1937]1964) and Zamowitz (1985) for surveys of the literature.
41. Still other, weaker effects accounted for in this model are those of changes in a fiscal
policy variable and in the price level. One should remember that a V AR model shows the
lead-lag interactions among all the selected variables, allowing also for the effects of their
own lagged values (for the serial correlations that are high in many important economic
aggregates).
42. On the monetary policy ineffectiveness in the new classical macroeconomics, see
Lucas (1972, 1973) and Sargent and Wallace (1975); on tests, see Barro (1978), Boschen and
Grossman (1982), and Mishkin (1983).
43. That is, prices clear markets, expectations are rational and adjust promptly, and all
opportunities for mutually beneficial transactions are used up.
44. Thus, the shift from production of goods to production of services made employment
much less cyclical, as already noted in section 2.7 (see also Zarnowitz and Moore, 1986,
pp. 536-538). But the 1990-1991 recession, although relatively mild, was unusually harsh in
causing a shrinkage of jobs in many service-producing industries. It is much too early to tell,
however, whether this constitutes a reversal of the trend previously observed, and if so to what
extent.
45. It can also be shown that the opposite change occurred for the lagging indicators,
where the lags have become shorter at peaks and longer at troughs (see Zarnowitz and
Moore, 1986, pp. 567-71). These changes are explained by the shift toward longer business
expansions and shorter recessions in the postwar era.
46. The two exceptions are the change in sensitive materials prices (line 14), where the
indicated k max timing is 0 (coincident), and the change in inventories (line 11), where it is a lag
of six months. This last result seems puzzling, since inventories tend to lag in levels but lead in
changes.
47. Only the average workweek, the diffusion index of slower deliveries (vendor perform-
ance), housing permits, and change in business and consumer credit-all series with little
trend-are used in level form (N) for the entries in columns 7-9.
48. The leads are particularly long in the case of inverted bond yields (line 32) or, which
amounts to much the same, indexes of corporate bond prices (Zamowitz, 1990). Note also
that the average duration of unemployment, like other series on unemployment, is naturally
taken in inverted form to match like turns in business activity. This indicator has a roughly
coincident timing at peaks but substantial and regular lags at troughs (line 25).
49. The big difference between seasonal and business cycles is that seasonal movements
are much more periodic (more truly "cyclical") and much more capable of being anticipated.
Over short spans within the year, seasonal fluctuations dominate the changes in many
variables, but over longer spans of several years it is the movements associated with business
cycles (the "specific cycles") that are predominant. The risk is that seasonal and cyclical
movements may interact so that their workable separation is impeded.
50. The cyclical (actually, trend-cycle) component is estimated as a weighted moving
average (Henderson curve) chosen on the basis of the relative amplitude of the irregular and
cyclical movements. Most series are smoothed with a 13-month device, but a nine-month
formula is used for relatively smooth and a 23-month formula for very volatile series. See
Zarnowitz and Boschan ([1975]1977, p. 173).
51. Any documentation would require more data, mainly in quarterly form.
References
Aftalion, A 1913. Les Crises Periodiques de Surproduction. Paris: Riviere.

Alt, J.E., and K.A Chrystal. 1983. Political Economics. Berkeley: University of
California Press.
Altonji, J.G. 1982. "The Intertemporal Substitution Model of Labour Market
Fluctuations: An Empirical Analysis." Review of Economic Studies 49: 783-824.
Ashton, T.S. 1959. Economic Fluctuations in England, 1700-1800. Oxford: Claren-
don Press.
Attwood, T. 1817. "A Letter to Nicholas Vansittart, on the Creation of Money, and
on Its Action upon National Prosperity." Birmingham: R Wrightson.
Backhouse, R 1988. Economists and the Economy: The Evolution of Economic
Ideas 1600 to the Present Day. Oxford: Basil Blackwell.
Balke, N.S., and R.I. Gordon. 1986. "Historical Data." In The American Business
Cycle Today, pp. 781-850. In Gordon (1986).
Barro, R.I. 1978. "Unanticipated Money, Output, and the Price Level in the United
States." Journal of Political Economy. 86: 549-580.
Blanchard, O.J., and L.H. Summers. 1986. "Hysteresis and European Unemploy-
ment Problem." In S. Fischer, ed. NBER Macroeconomics Annual 1986.
Cambridge, MA: MIT Press.
Blanchard, O.J., and M.W. Watson. 1986. "Are Business Cycles All Alike?" In The
American Business Cycle Today, pp. 123-156 and 178-179. See Gordon (1986).
Boschen, J.F., and H.1. Grossman. 1982. "Tests of Equilibrium Macroeconomics
Using Contemporaneous Monetary Data." Journal of Monetary Economics. 10:
309-33.
Bouniatian, M. 1908. Geschichte der Handelskrisen in England, 1640-1840.
Munich: Ernst Reinnhardt.
Bronfenbrennner, M., ed. 1969. Is the Business Cycle Obsolete? New York: Wiley-
Interscience.
Burns, AF., and W.e. Mitchell, 1946. Measuring Business Cycles. New York:
National Bureau of Economic Research.
Butler, W.F., RA Kavesh, and RB. Platt, eds. 1974. Methods and Techniques of
Business Forecasting. Englewood Cliffs, N.J.: Prentice-Hall.
Cagan, P. 1965. Determinants and Effects of Changes in the Stock of Money, 1875-
1960. New York: Columbia University Press for NBER
Cassel, G. [1923] 1932. Theory of Social Economy. Harcourt, Brace and Company.
Clark, J.M. 1917. "Business Acceleration and the Law of Demand: A Technical
Factor in Economic Cycles." Journal of Political Economy. 25: 217-235.
Deane, P. 1967. The First Industrial Revolution. Cambridge: Cambridge University

Press.
Diebold, FX., and G.P. Rudebusch. 1990. "A Nonparametric Investigation of
Duration Dependence in the American Business Cycles." Journal of Political
Economy 98(3): 596-616.
- - - . 1991. "Turning Point Predictions with the Composite Leading Index: An
Ex ante Analysis." Chapter 14 in K. Lahiri and G.H. Moore, eds., Leading
Indicators: New Approaches and Forecasting Records. Cambridge: Cambridge
University Press.
Eichenbaum, M. 1990. "Real Business Cycle Theory: Wisdom or Whimsy?" NBER
Working Paper no. 3432, (September) Cambridge, MA.
Fels, R, and C.E. Hinshaw. 1968. Forecasting and Recognizing Business Cycle
Turning Points. New York: NBER
Fisher, I. 1907. The Rate of Interest. New York: Macmillan Co.
- - - . 1932. Booms and Depressions: Some First Principles. New York: Adelphi
Co.
---.1933. "The Debt-Deflation Theory of Great Depressions." Econometrica 1:
337-357.
Friedman, B.M. 1986. "Money, Credit, and Interest Rates in the Business Cycle."
In Gordon (1986, pp. 395-438,456-458).
Friedman, M., and A.J. Schwartz. 1963a. A Monetary History of the United States,
1867-1960. Princeton, N.J.: Princeton University Press for NBER
- - - . 1963b. "Money and Business Cycles." Review of Economics and Statistics
(Supplement) 45: 32-64.
- - - . 1982. Monetary Trends in the United States and the United Kingdom: Their
Relation to Income, Prices; and Interest Rates; 1867-1975. Chicago: University of
Chicago Press for NBER
Frisch, R 1933. "Propagation Problems and Impulse Problems in Dynamic
Economics." Economic Essays in Honor of Gustav Cassel. London: George
Allen & Unwin.
Goodwin, RM. 1951. "The Non-linear Accelerator and the Persistence of Business
Cycles." Econometrica 19: 1-17.
Gordon, RJ. 1986. The American Business Cycle Today: Continuity and Change.
Chicago: University of Chicago Press for NBER
Greenwald, B.C., J.E. Stiglitz, and A. Weiss. 1984. "Informational Imperfections in
the Capital Market and Macroeconomic Fluctuations." American Economic
Review 74: 194-199.
Haberler, G. [1937] 1964. Prosperity and Depression. League of Nations. Reprinted
Cambridge, MA: Harvard University Press.
Hall, RE. 1986. "The Role of Consumption in Economic Fluctuations." In RJ.
Gordon (1986, pp. 237-266).
Hamilton, J.D. 1983. "Oil and the Macroeconomy Since World War II." Journal of
Political Economy 91: 228-248.
Hansen, A.H. [1951] 1964. Business Cycles and National Income (expanded
edition). New York: W.W. Norton (first edition 1951).
Hansen, A.H., and R.V. Clemence. 1953. Readings in Business Cycles and National
Income. New York: W.W. Norton & Co.
Hawtrey. RG. 1927. "The Monetary Theory of the Trade Cycle and Its Statistical
Test." Quarterly Joumal of Economics. 671-686. Reprinted in Hansen and
Clemence (1953, pp. 139-149).
Hickman, B.G., ed. 1972. Econometric Models of Cyclical Behavior (2 vols.).
Studies in Income and Wealth, no. 36. New York: Columbia University Press for
NBER
Hicks, J.R 1937. "Mr. Keynes and the 'Classics': A Suggested Interpretation."
Econometrica 5: 147-159.
- - - . 1950. A Contribution to the Theory of the Trade Cycle. Oxford: Clarendon
Press.
Hoffman, W. 1965. Das Wachstum der deutschen Wirtschaft seit der Mitte des 19.
Jahrhunderts. Berlin: Springer-Verlag.
Jevons, W.S. 1884. Investigations in Currency and Finance. London: Macmillan &
Co.
Juglar, C. [1862] 1889. Des Crises Commerciales et leur Retour Periodique, en
France, en Angleterre et aux Etats-Unis. Paris: Guillaumin.
Kaldor, N. 1940. "A Model of the Trade cycle." Economic Journal 50: 78-92.
Keynes, J.M. 1936. The General Theory of Employment, Interest and Money.
London: Macmillan.
Kiefer, N. 1988. "Economic Duration Data and Hazard Functions." Journal of
Economic Literature 26: 646-679.
King, RG., c.1. Plosser, J.H. Stock, and M.W. Watson. 1991. "Stochastic Trends
and Economic Fluctuations." American Economic Review 81: 819-840.
Klein, P.A., ed. 1990. Analyzing Modern Business Cycles: Essays Honoring
Geoffrey H. Moore. Armonk, N.Y.: M.E. Sharpe.
Klein, P.A., and G.H. Moore. 1985. Monitoring Growth Cycles in Market-Oriented
Countries: Developing and Using International Economic Indicators. Cambridge,
Mass.: Ballinger Publishing Company for NBER
Kydland, F., and E.C. Prescott. 1982. "Time to Build and Aggregate Fluctuations."
Lahiri, K., and G.H. Moore, (Eds.). 1991. Leading Economic Indicators: New
Approaches and Forecasting Records. Cambridge, MA: Cambridge University
Press.
Lilien, D.M. 1982. "Sectoral Shifts and Cyclical Unemployment." Journal of
Political Economy 90: 777 - 793.
Link, R.G. 1959. English Theories of Economic Fluctuations, 1815-1848. New
York: Columbia University Press.
Lucas, RE., Jr. 1972. "Expectations and the Neutrality of Money." Journal of
Economic Theory 4: 103-124.
- - - . 1973. "Some International Evidence on Output-Inflation Tradeoffs."
American Economic Review 63: 326-334.
Marshall, A., and M. Marshall. 1881. The Economics of Industry. (2nd ed.) London:
Macmillan.
Meltzer, AH. 1986. "Comment (on B.M. Friedman, 1986)." In Gordon (1986,
pp.441-450).
Metzler, L.A 1941. "The Nature and Stability of Inventory Cycles." Review of
Economic Statistics. 23: 113-129.
- - - . 1947. "Factors Governing the Length of Inventory Cycles." Review of
Economic Statistics 29: 1-15.
Minsky, H.P. 1982. Can 'It' Happen Again? Essays on Instability and Finance. New
York: M.E. Sharpe.
Mintz, I. 1969. Dating Postwar Business Cycles: Methods and Their Application to
Western Germany, 1950-67. New York: NBER.
- - - . 1974. "Dating United States Growth Cycles." Explorations in Economic
Research 1(1): 1-113.
Mirowski, P. 1985. The Birth of the Business Cycle. New York & London: Garland.
Mishkin, F.S. 1983. A Rational Expectations Approach to Macroeconometrics:
Testing Policy Ineffectiveness and Efficient-Markets Models. Chicago: University
of Chicago Press for NBER.
Mitchell, W.C.1913. Business Cycles. Berkeley: University of California Press.
---.1927. Business Cycles: The Problem and Its Setting. New York: NBER.
- - - . 1951. What Happens During Business Cycles: A Progress Report. New
York:NBER.
Moore, G.H. 1983. Business Cycles, Inflation, and Forecasting (2nd ed.; 1st ed.,
1980). Cambridge, Mass.: Ballinger for NBER.
Moore, G.H., and M.H. Moore. 1985. International Economic Indicators: A
Sourcebook. Westport, Conn.: Greenwood Press.
Moore, G.H., and V. Zarnowitz, 1986. "The Development and Role of the National
Bureau's Business Cycle Chronologies." In Gordon (1986).
Moore, H.L. 1914. Economic Cycles: Their Law and Cause. New York: Macmillan
Co.
Morgenstern, O. 1959. International Financial Transactions and Business Cycles.
Princeton: Princeton University Press for NBER.
Overstone, S.J.L. 1837. "Reflections Suggested by a Perusal of Mr. J. Horsley
Palmers' Pamphlet on the Causes and Consequences of the Pressure on the
Money Market."
Pigou, AC. 1927. Industrial Fluctuations. London: Macmillan.
Plosser, c.l. 1989. "Understanding Real Business Cycles." Journal of Economic
Perspectives 3: 51-77.
---.1991. "Money and Business Cycles: A Real Business Cycle Interpretation."
In M.T. Belongia, ed., Monetary Policy on the 75th Anniversary of the Federal Re-
serve System. Boston: Kluwer Academic Publishers.
Ramey, G., and V.A Ramey. 1991. "Technology Commitment and the Cost of
Economic Fluctuations." NBER Working Paper no. 3755.
Robertson, D.H. 1937. "The Trade Cycle: An Academic View." Lloyds Bank
Review (September): 504-510. Reprinted in Hansen and Clemence (1953,
pp.166-174).
Romer, c.n. 1986. "The Prewar Business Cycle Reconsidered: New Estimates of
Gross National Product, 1869-1918." NBER Working Paper no. 1969.
Romer, c.n., and D.H. Romer, 1989. "Does Monetary Policy Matter? A New Test
in the Spirit of Friedman and Schwartz." In OJ. Blanchard and S. Fischer,
eds., Macroeconomics Annual 1989. Cambrige, Mass.: MIT Press, pp. 121-
170.
Rostow, W.W. 1980. "Cycles in the Fifth Kondratieff Upswing." In The Business
Cycle and Public Policy, 1929-80. A Compendium of Papers Submitted to the
Joint Economic Committee, Congress of the United States, Washington, D.C.:
Government Printing Office.
Samuelson, P.A. 1939. "Interactions Between the Multiplier Analysis and the Prin-
ciple of Acceleration." Review of Economic Statistics 21: 75-78.
Sargent, T. J., and N. Wallace. 1975. "Rational Expectations, the Optimal Monet-
ary Instrument, and the Optimal Monetary Rule." Journal of Political Economy
83(2): 241-254.
Schumpeter, J.A. 1939. Business Cycles: A Theoretical, Historical, and Statistical
Analysis of the Capitalist Process. New York: McGraw-Hill.
Sherman, H.J. 1991. The Business Cycle: Growth and Crises Under Capitalism.
Princeton, N.J.: Princeton University Press.
Smith, A.I776. The Wealth of Nations. Cannon ed., Modern Library, 1937.
Sowell, T. 1972. Say's Law: An Historical Analysis. Princeton, N.J.: Princeton
University Press.
Spiethoff, A. 1955. Die wirtschaftlichen Wechsellagen (2 vols.). Tiibingen: Mohr.
Stock, J.H., and M.W. Watson. 1990. "Appendixes to Business Cycle Properties of
Selected U.S. Economic Time Series, 1959-1988." NBER Working Paper no.
3376.
Thorp, W.L.1926. Business Annals. New York: NBER.
Tinbergen, J. 1939. Statistical Testing of Business-Cycle Theories (2 vols.). Geneva:
League of Nations, Economic Intelligence Service.
Tugan-Baranovskii, M.I. [1894, 1901] 1913. Les Crises Industrielles en Angle-
terre (2nd ed.). Paris: Giard & Briere. (Published first in Russian, then in
German)
U.S. Department of Commerce. Bureau of Economic Analysis. 1984. Handbook of
Cyclical Indicators: A Supplement to the Business Conditions Digest. Washington
D.C.: U.S. Government Printing Office.
Wicksell, K. [1896] 1936. Interest and Prices: A Study of the Causes Regulating the
Value of Money. London: Macmillan.
Willett, T.n. 1988. Political Business Cycles: The Political Economy of Money,
Inflation, and Unemployment. Duke University Press.
Wojnilower, A.M. 1980. "The Central Role of Credit Crunches as Recent Financial
History." Brookings Papers on Economic Activity 2: 277 -339.
Woodford, M. 1987. "Three Questions About Sunspot Equilibria in an Ex-
planation of Economic Fluctuations." American Economic Review 77: 43-
48.
Zarnowitz, V. 1967. An Appraisal of Short-Term Economic Forecasts. New York:

NBER.
- - - . 1974. "How Accurate Have the Forecasts Been?" In W.F. Butler, R.A.
Kavesh, and R.B. Platt, eds., Methods and Techniques of Business Forecasting.
Englewood Cliffs, N.J.: Prentice-Hall.
- - - . 1981. "Business Cycles and Growth: Some Reflections and Measures."
In W.J. Mucld and A.E. Ott, eds. Wirtschaftstheorie und Wirtschaftspolitik:
Gedenkschriftfur Erich Preiser. Passau: Passavia UniversiHitsverlag.
---.1985. "Recent Work on Business Cycles in Historical Perspective." Journal
of Economic Literature 23: 523-580.
--.1987. "The Regularity of Business Cycles." NBER Working Paper no. 2381
(September). Reprinted as Chapter 8 in Zarnowitz (1992).
- - - . 1990. "Corporate Bond Prices as a Leading Indicator." In G.H. Moore
Leading Indicators for the 1990s. Homewood, IL: Dow-Jones Irwin.
- - - . [1989]1992. "Facts and Factors in the Recent Evolution of Business Cycles
in the United States." NBER Working Paper no. 2865. Reprinted as Chapter 3
in Zarnowitz (1992).
- - - . 1992. Business Cycles: Theory, History, Indicators, and Forecasting.
Chicago: University of Chicago Press for NBER.
Zarnowitz, V., and C. Boschan. [1975] 1977. "Cyclical Indicators: An Evaluation
and New Leading Indexes." Business Conditions Digest (May): v-xxii.
Reprinted in U.S. Department of Commerce, Bureau of Economic Analysis,
Handbook of Cyclical Indicators. Washington, D.C.: U.S. Government Printing
Office, pp. 170-184.
Zarnowitz, V., and P. Braun. 1990. "Major Macroeconomic Variables and Leading
Indexes: Some Estimates of their Interaction." In P.A. Klein, ed., Analyzing
Modern Business Cycles. Armonk, N.Y.: M.E. Sharpe, pp. 177-205.
Zarnowitz, V., and G.H. Moore. 1982. "Sequential Signals of Recession and Recov-
ery." Journal of Business 55: 57-85.
---.1986. "Major Changes in Cyclical Behavior." In Gordon (1986).
Commentary
by James H. Stock
Victor Zarnowitz has provided us with an authoritative and valuable

overview of the business cycle: its definition, the history of thought asso-
ciated with business cycle analysis, and key facts about the evolution of the
business cycle, both in the United States and abroad. His summary at once
provides a handy reference for many of the key empirical facts about busi-
ness cycles and some new comparisons of cycles since World War II across
countries.
Given the limited scope of a discussant, it would be impossible to
address most of Zarnowitz's points. These comments are therefore limited
to two of the main themes in the chapter. The first concerns changes over
time in the U.S. business cycle, more precisely the finding that contractions
are shorter, and expansions longer, after World War II than before. The
second topic is Zarnowitz's observation that there can be no single theory
that can explain the business cycle but that rather each cycle can differ in
its sources and perhaps even in its propagation mechanisms. This obser-
vation in tum implies that careful studies of individual cyclical episodes can
in principle yield new insights on the workings of the macroeconomy and
its response to various types of shocks. My second set of remarks builds on
this observation by considering a concrete example: the lessons that we as
macroeconomists can learn from the 1990 recession.
1 The U.S. BUSiness Cycle, Prewar and Postwar
The argument that the U.S. economy has stabilized since World War II
has two parts: first, that the volatility of output (say, its standard deviation)
has decreased, and second, that postwar contractions are shorter, and ex-
pansions longer, than they were before the Depression. The first part of
this argument has come under serious attack by Romer (1986, 1989).
Her argument is that the prewar output figures were constructed using
inherently more volatile series: The available series cover more volatile
sectors of the economy. Although the particulars of her findings are still
debated, the case for stabilized postwar output is certainly much weaker than
previously thought. Zarnowitz lays out the evidence for the second part of the
argument. Drawing on previous careful analyses of the dating of certain
suspect prewar cycles, he concludes that, even if several prewar recessions
73
were deleted from the chronology, the empirical evidence remains that
postwar cycles have longer expansions and shorter contractions.
My purpose in this part of the discussion is to review some recent
research by Watson (1991) that questions this conclusion. As Moore and
Zarnowitz (1986) point out, different economic time series were used to
construct the NBER reference cycle chronology, depending on the histori-
cal period. This raises the possibility that systematic differences in the
series might induce an apparent shift in the properties of the reference
cycles. Watson addresses this possibility in two ways: first, by comparing
pre- and postwar phase durations for the same or comparable series and
second, by constructing an aggregate postwar index of industrial produc-
tion that has a similar coverage and definition to the available prewar
industrial production (IP) series. He concludes that specific cycles (cycles
based on individual series) appear not to have changed when they are
computed for series that are consistently defined prewar and postwar.
Selected evidence on this point is presented in table 1C-l. Panel A
presents average expansion and contraction lengths for the prewar and
postwar data based on the NBER chronology, and also based on the
NBER chronology as modified by Zamowitz by dropping some suspicious
early recessions. As Zamowitz emphasizes, postwar the recessions are
shorter and the expansions longer, whether the suspicious early recessions
are included or not.
Panel B presents evidence for specific cycles for series with comparable
definitions pre- and postwar. The specific cycle dates were computed using
the turning point identification algorithm developed by Bry and Boschan
(1973).1 For each series with comparable definitions, the differences in
prewar and postwar phase durations are modest or negligible. For example,
contractions in stock prices averaged 19 months prewar and 17 months
postwar. The most noteworthy results are for industrial production. Prewar
contractions and expansions, measured using the historical IP index
recently constructed by Miron and Romer (1990), averaged 16 and 22
months, respectively, both shorter but comparable to the average durations
based on the official NBER chronology in panel A. Similarly, the postwar
expansions and contractions based on the official IP series (produced by
the Federal Reserve Board) are essentially the same as those based on the
official NBER chronology. To examine the hypothesis that this difference
arises because of differences in the composition of series used to construct
the FRB and Miron-Romer IP indexes, Watson constructed a postwar
extension of the Miron-Romer series. His approach was to apply the
Miron-Romer weights to component series that were similar to those used
by Miron and Romer. As the results in the final rows of table 1C-1 indicate,
COMMENTARY 75
Table 1C-1. Durations of Historical Expansions and Recessions, in Months
A. Durations Based on Reference Cycles
Period Dating Method
1854-1919 27 22 NBER
1854-1919 37 23 Modified NBER*
1945-1990 50 11 NBER
Source: Zamowitz (1992). E and t respectively denote average expansion and contraction
lengths, in months.
* Modified NBER: drop 1869-1870, 1887 -1888, and 1899-1900 recessions.
B. Durations for Selected Individual Series
Series Period E (;
S&P composite 1871-1940 24 19

S&P composite 1945-1990 26 17
Pig iron production 1887-1940 28 13
IP-iron & steel 1947-1990 22 20
Plans for new buildings 1868-1937 19 18
Building permits 1947-1990 20 18
IP-Miron-Romer 1884-1940 22 16
IP-Miron-Romer (approx.) 1947-1990 22 14
IP-FRB 1947-1990 50 13
Source: Watson (1991). Peaks and troughs were dated using the Bry-Basehan (1973)
algorithm.
Watson's approximate Miron-Romer IP series has phase durations very

close to its prewar counterpart-and in particular has longer recessions and
much shorter expansions than those measured using the FRB IP series or
the postwar NBER chronology.
These results suggest that the postwar shifts in duration, as measured
using the FRB IP series or using the NBER postwar chronology (which
itself relies in large part on the specific cycles in IP), are a consequence of
the chronologies having been constructed using different series in the two
periods. Of course, as Zarnowitz emphasizes, the sectoral composition of
the U.S. economy has evolved over this period, so the use of a different
mix of series to date turning points might be appropriate. Watson (1991)

examines this possibility but argues that it does not really resolve the large
differences between the dates obtained from the official IP series and his
approximate Miron-Romer series. The reason, it seems, is that the older
series were not representative of historical economic activity and em-
phasized too heavily materials, metals, financial series, and extractive
industries.
In summary, this evidence suggests that the postwar shortening of
recessions and lengthening of expansions using the NBER chronology can
largely be attributed to the use of different mixes of series to date the
prewar and postwar recessions; the prewar dating relied on series with
more cycles, longer contractions, and shorter expansions than the series
used for postwar dating. Whether this imparts a "bias" to the dating chron-
ology is, however, a matter of interpretation. One possibility is that the
shifting mix of series accurately reflected compositional changes in the
United States; despite Watson's preliminary negative evidence on this
point, the issue remains unresolved. A second possibility is that the under-
lying time series behavior of the U.S. economy has not changed prewar
and postwar, but rather the conception of a recession has changed by
the NBER dating authorities; that is, the definition of a recession has inten-
tionally changed. This view, however, is at odds with the the NBER
researchers' emphasis on continuity and the importance of having a chron-
ology that can be compared over time. A third possibility is that the intent
of the NBER has been the same but that the postwar use of a broader set
of indicators has spuriously resulted in fewer and shorter recessions being
identified. Learning which of these explanations prevails must await
further research.
2 The 1990 Recession: Initial I mpl ications for

Economic Theory
A second theme in Zarnowitz's chapter is that business cycles can have

different causes and propagation mechanisms. The implication is that each
business cycle provides new and different information about the workings
of the economy and that each cycle therefore merits special study. This
general observation is consistent with the behavior of the economy
entering and during the 1990 recession. In brief, current evidence suggests
that relationships among economic variables that held on average over the
postwar period, and held during the recessions of the 1970s and 1980s, did
not hold while the economy was entering the 1990 recession. The modest
COMMENTARY 77
Table 1C-2. Monthly Percentage Changes in Coincident Economic Indicators,

April 1990 to July 1991
Month IP PersInc Emp-Hours MT Sales XCI
1990: 4 -.1% .1% -.6% -.9% -.2%

90: 5 .5 .0 .7 .9 .4
90: 6 .6 .1 .5 .6 .5
90: 7 .3 .2 -.2 -.7 .1
90: 8 .1 -.4 -.3 1.2 .0
90: 9 .1 -.4 .4 -1.8 -.1
90:10 -.6 -1.1 -1.3 -.2 -.7
90:11 -1.5 .2 .3 -1.6 -.8
90: 12 -1.0 .4 .4 -2.1 -.5
1991: 1 -.6 -1.6 -1.2 -.9 -.9
91: 2 -.8 .0 .2 .4 -.4
91: 3 -.7 .3 -.4 -.4 -.2
91: 4 .5 -.3 -.5 1.8 .3
91: 5 .8 .3 .6 .8 .7
91: 6 .8 .5 .5 .3 .8
91: 7 .7 -.2 -1.0 .8 .2
Notes: 1990: 4 denotes April 1990, etc.

Series:
IP: Industrial production (total).
Pers Inc: Real personal income less transfer payments.
Emp-Hours: Employee-hours in nonagricultural establishments.
MT Sales: Real manufacturing and trade sales, total.
XCI: Stock-Watson (1991) experimental index of coincident indicators.
purpose of this part of my comments, then, is to illustrate Zamowitz's

general observations by listing some of the characteristics that distinguish
the 1990 recession from the preceding three recessions.
Although the NBER dated the cyclical peak as July 1990, overall eco-
nomic activity was stagnant from June through September. Monthly growth
rates of four main economic indicators-industrial production, real per-
sonal income less transfers, employee-hours in nonagricultural establish-
ments, and real manufacturing and trade sales-are presented in table
1C-2 for this period. Although employee-hours peaked in June, industrial
production did not peak until September. The striking feature in this
recession is not the flat growth over the summer but rather the precipitous
declines in the main indicators in the fourth quarter of 1990, in particular

the sharp drops in IP, personal income, and employee-hours in October.
My argument is that the "explanations" for recent postwar U.S.
recessions do not really work for the 1990 experience. The evidence for this
is simple (perhaps too simple): A large number of time series forecasting
relations that worked over the previous 30 years performed extremely
poorly during the fall of 1990. This assertion is documented in detail in
Stock and Watson (1991); here, I present some of the key evidence on this
matter.
One way to examine whether "typical" behavior was present during the
onset of the 1990 recession is to ask whether the correlations between a
broad set of leading economic indicators and overall economic activity
appeared to be stable during this episode. More concretely, this can be
examined by using leading indicators to construct forecasts of coincident
indicators or an index of coincident indicators. Such forecasts were
constructed in Stock and Watson (1991) by regressing future three-month
growth of an experimental index of coincident indicators (the "XCI") on
current values and two lags of XCI growth and on current and five lags of a
candidate indicator. Such forecasts were constructed for more than 45 lead-
ing indicators. The regressions were estimated using data from 1959:1 to
1988:9, and the out-of-sample performance was examined in each of two
periods: for forecasts made during 1989:10-1990:4, and forecasts made
during 1990:5-1990:8. (The three-month ahead forecast made in 1990:7,
say, forecasted growth from July through October 1990.) The performance
of the forecasts can be compared by examining the root mean square errors
(RMSEs) of the forecasts over the various subsamples.
Selected results from this out-of-sample forecasting exercise are
presented in Table 1C-3. Two general sets of observations are apparent.
First, the forecasts for most series were seriously off track during the
summer and early fall of 1990, with three-month-ahead forecast errors
as high as 8.0% at an annual rate, even though the forecasting relations
actually performed better during the initial out-of-sample period than
during the in-sample period. Only a few leading indicators had perform-
ance comparable to the 1959:1-1988:9 sample: real M2, the index of
help wanted advertising, building permits, and the index of consumer expec-
tations. Second, those series that had relatively good average performance
during the 1962-1988 period, particularly indicators of monetary activity
such as the 10-yearll-year Treasury bond spread or the fed funds rate,
performed particularly poorly during the 1990 episode. 2 Conversely, the
few series that performed well during the summer of 1990 had in fact
performed poorly during the 1962-1988 sample. Interestingly, two series
COMMENTARY 79
that one might think would have performed well during the summer of
1990---stock prices and oil prices-did not produce reliable forecasts when
these forecasts were made using the historically estimated forecasting
equation.
These observations suggest that the historical correlations between
these leading indicators and overall economic activity were not good
guides to this episode. More concretely, one can reasonably argue that dis-
inflationary monetary policy was an important proximate cause of the 1975,
1979, and 1981 recessions. The poor performance of the monetary
indicators in the 1990 recession suggest that restrictive monetary policy, or
an associated credit crunch, played at most a limited role in inducing the
sharp contractions of October and November.
The results instead are consistent with an alternative mechanism based
on unprecedented shifts in consumer expectations. The timing was right:
Expectations dropped precipitously in August; manufacturing and trade
sales fell sharply in September; and production, employee-hours, and
income all dropped in October. One interpretation is that the recession
transpired as a self-fulfilling prophecy. For example, the drop in consumer
expectations induced firms to restrict hiring and to expect lower sales; with
lower incomes, sales did in fact drop. Another interpretation is that
consumer expectations played only a passive role and reflected increased
uncertainty in the face of rising oil prices and the threat of war in the Gulf,
which together induced consumers to adopt more conservative spending
patterns. If these explanations are right, they illustrate Zarnowitz's point
that each cycle can have different sources: Over the previous 30 years,
consumer expectations had in fact been of only limited forecasting value. 3
Whether or not consumer expectations played a causal role in the contrac-
tion, the circumstances of the 1990 downturn differed from their recent
predecessors, in which indicators of monetary policy provided reasonably
reliable forecasts.
3 Conclusions
These observations only touch on the rich set of facts and theories
reviewed in Zarnowitz's chapter. In particular, his work on international
business cycles warrants careful reading and raises an interesting set of
research questions concerning international phase durations, cyclical stab-
ility, and cross-country patterns of growth. One of the implications of this
chapter, which I have attempted to illustrate and which I hope is taken
up by subsequent researchers, is that there are good reasons to study
Table 1C-3. Performance of Selected Leading Indicators as Forecasts of the XCI Coincident Index During the Onset of
the 1990 Recession (three-month-ahead forecast horizon)
RMSE of Forecasts of XCI Growth Over:
Series 62:1-88:9 88:10-90:4 90:5-90:8
Financial Indicators
S&P composite 4.30 2.61 4.72
RealMl 4.28 1.92 5.06
RealM2 4.08 2.02 3.57
Monetary base 4.51 1.68 6.40
Federal funds rate 4.09 1.39 5.67
lO-year T-bond rate, smoothed 4.36 2.18 5.47
Commercial paperrr-bill spread, six months 3.66 1.51 8.00
10-year T-bondll-year T-bond spread 4.12 1.38 6.63
Employment Indicators
New claims, unemployment insurance 4.46 1.83 5.34
Persons unemployed < five weeks 4.55 2.11 5.95
Index of help wanted ads, newspapers 4.05 2.52 3.07
Part-time work, economic reasons 4.35 2.03 5.38
Average weekly hours, manufacturing production workers 4.54 1.88 5.16
Inventories and Orders
Real unfilled orders, durable goods industries 4.48 2.01 5.31
Real manufacturing & trade inventories 4.55 2.37 6.21
Additional Indicators
Real retail sales 4.53 1.82 5.27
Building permits-private housing 4.15 1.56 4.13
Consumer expectations (University of Michigan) 4.27 2.01 3.00
PPI, crude petroleum 4.56 2.04 6.82
Weighted average of nominal exchange rates 4.42 2.10 7.83
Capacity utilization, manufacturing 4.55 1.86 5.08
Index of vendor performance 4.40 1.73 6.57
Notes: All results were computed using the data as revised through 91:02. RMSEs were computed for forecasts of x, = In(c'+3/c,) on
current values plus two lags of In(c/ct-J) and current values plus five lags of the (suitably transformed) individual leading indicator, where c, is
the Kalman smoother estimate of the Stock-Watson (1989) experimental coincident index (the XCI) computed on data revisions through
1991:2. The dates in the column headings are the months in which the three-month forecasts were made. Units are annual rate percent
growth rates.
Source: Stock and Watson (1991).
individual cycles in detail. Perhaps because this sort of analysis is antitheti-

cal to the econometric methodology developed by the Cowles Commission
and taught in modern textbooks, it has had limited applications in recent
years and therefore might yield important new insights.
Notes
1. The Bry-Boschan (1973) algorithm is an early computerized expert system developed

at the NBER to automate the dating of specific cycles in the large number of series that
NBER business cycle researchers have collected over the years. This algorithm succeeded
in closely replicating specific cycle dates ascertained by the NBER experts. Watson (1991)
demonstrates that the algorithm, applied to the Miron-Romer IP series, closely replicates the
prewar NBER chronology.
2. Although real M2 produced relatively small forecast errors, recent research (for
example, Bemanke and Blinder, 1991, and Friedman and Kuttner, 1989) suggests that other
financial market indicators provide more accurate measures of monetary policy than does M2.
These alternative indicators include, in particular, interest rates such as the fed funds rate, the
slope of the Treasury bond yield curve, and matched maturity public-private spreads such as
that between the commercial paper and the Treasury bill rate. Moreover, forecasts based on
M1 and the monetary base during the 1990 episode perfonned poorly. It is therefore unclear
how to interpret the relatively satisfactory performance of M2 during this episode.
3. The poor historical forecasting perfonnance of consumer expectations is supported by
its relatively large in-sample root mean squared error (RMSE) in table lC-3. In a more
thorough study, Garner (1991) also found a weak historical relationship between consumer
expectations and consumer spending.
References
Bernanke, B., and A. Blinder. Forthcoming. "The Federal Funds Rate and the
Channels of Monetary Transmission." American Economic Review.
Bry, G., and C. Boschan. 1973. Cyclical Analysis of Time Series: Selected Computer
Procedures and Computer Programs. New York: Columbia University Press.
Friedman, B.M., and K.N. Kuttner. 1989. "Another Look at the Evidence on
Money-Income Causality." Manuscript, Department of Economics, Harvard
University. Journal of Econometrics, forthcoming.
Gamer, C. Alan. 1991. "Forecasting Consumer Spending: Should Economists Pay
Attention to Consumer Confidence Surveys?" Economic Review of the Federal
Reserve Bank of Kansas City (May/June): 57-71.
Miron, J .A., and C.D. Romer. 1990. "A New Index of Industrial Production, 1884-
1940." Journal of Economic History 50: 321-337.
Moore, G.H., and V. Zarnowitz. 1986. "The Development and Role of the NBER's
Business Cycle Chronologies." In RJ. Gordon, ed., The American Business
Cycle: Continuity and Change. Chicago: University of Chicago Press.
COMMENTARY 83
Romer, C.D. 1986. "Spurious Volatility in Historical Unemployment Data."

Journal of Political Economy 94: 1-37.
---.1989. "The Prewar Business Cycle Reconsidered: New Estimates of Gross
National Product, 1869-1908." Journal of Political Economy 97: 1-38.
Stock, James H. and M.W. Watson. 1989. "New Indexes of Coincident and Leading
Economic Indicators." NBER Macroeconomics Annual, 1989. Cambridge,
Mass.: MIT Press, pp. 351-393.
Stock, J.H., and M.W. Watson. 1991. "Predicting Recessions." In J.H. Stock and
M.W. Watson, eds., Business Cycles, Indicators and Forecasting. Chicago: Univer-
sity of Chicago Press for the NBER.
Watson, M.W. 1991. "Business Cycle Durations and Postwar Stabilization of the
U.S. Marcroeconomy." Manuscript, Northwestern University.
Zarnowitz, V. 1992. "What is a Business Cycle?" In M.T. Belongia and M.R.
Garfinkel, eds., The Business Cycle: Theories and Evidence. Norwell, Mass.:
Kluwer Academic Publishers.
2
THE CYCLE BEFORE
NEW-CLASSICAL ECONOMICS
David Laidler
If an idea is to be influential in economics, it is not enough that it be a good

idea. Economists must also be convinced of its quality, and this is not
achieved simply by demonstrating theoretical rigor and empirical relev-
ance. It also involves the persuasive use of language. The formal study of
"The Rhetoric of Economics," of which the use of persuasive language is
but one component, is of more recent origin than the phenomenon itself.
It was, after all, Keynes, not Donald McCloskey (1983) who said, "Words
ought to be a little wild-for they are the assault of thought upon the
unthinking." Even so, recent developments in business cycle theory
provide a striking example of the power of the right words to draw atten-
tion to an idea. 1 Who, when all is said and done, could resist the appeal of a
"real" theory of the business cycle? How could it fail to be better than,
shall we say, a "mythical" theory? And if it wins hands down in this
comparison, might it not also prove to be more satisfactory than a "monet-
ary" theory? That is surely Kydland and Prescott's (1990) underlying mess-
age, and to judge by the literature built upon their original (1982) work,
they have been remarkably successful in getting it across.
But the language used to promote modern business cycle theory is, in
the context of this chapter, misleading. The major innovation of Kydland
and Prescott has been to build a plausible model of the cycle in which the
85
dominant impulse (to borrow a phrase from Brunner and Meltzer, for
example, 1978) arises from shocks to the aggregate production function,
rather than to aggregate demand. Though it is true that monetary shocks
provide a genuine alternative to those hypothesized by Kydland and
Prescott, it is also the case that, except in the literature based on the twin
assumptions of (Walrasian) clearing markets and rational expectations,
monetary-impulse models have always been something of a minority taste.
In the literature of the 1950s and 1960s, and long before that, "real"
theories of the cycle predominated, though they were very different from
anything that currently carries that label.
Modem business cycle theory is based on Walrasian general equilibrium
analysis. In such a system, markets always clear and the distinction
between "demand" and "supply" shocks is not all that helpful: Say's law
holds, supply creates its own demand, and in general any shock will affect
both simultaneously. The demand-supply shock distinction does make
sense in a system in which the fact of monetary exchange permits aggregate
demand and supply to move independently of each other and in which,
therefore, a situation of general excess demand or supply can emerge.2 The
appropriate contrast, then, between Kydland and Prescott's work and that
of the 1950s and 1960s is not between "real" and "monetary" models; it is
between those that locate the shocks initiating the cycle in the economy's
production sector and in which both supply and demand behavior
interdependently respond, and those that locate them on the demand side
and treat production as passively adjusting to demand fluctuations whose
origins can be either real or monetary.
It is also worth noting that some advocates of "real business cycle
theory"-for example, Plosser (1991)-have lately taken to classifying
anything but a change in the monetary base, including a change in the
required reserve ratio, as a "real shock." They cannot be prevented from
using the word real this way, but they should make it clear to their readers
that it is then appropriate to use the adjective nominal rather than monetary
to characterize an alternative approach. This matter is important, because
every monetary theory of the cycle of which I am aware has paid careful
attention to, and indeed usually accorded prime importance to, endogenous
fluctuations in money multipliers (and in the case of 1937-1938 to changes
in reserve requirements as well) when confronting empirical evidence. No
one, to the best of my knowledge, ever advanced a purely nominal theory
of the cycle as a serious explanation of any actual historical experience.
None of this would matter very much if inability to cope with empirical
evidence was a characteristic of all demand side models of the cycle. But in
fact the models that found themselves in trouble from the mid-1960s onward
THE CYCLE BEFORE NEW-CLASSICAL ECONOMICS 87
were real (albeit demand-side) models, and these were already in difficulty
as a result of a re-examination of previous empirical evidence. The historical
studies of Friedman and Schwartz had, by the mid 1960s, established what
we would nowadays call the monetarist explanation of the cycle as some-
thing to be taken very seriously. Subsequent experience with inflation and
unemployment, which was so damaging to the orthodox Keynesian demand-
side macroeconomics of the time, did nothing to undermine monetarism.
Nor is the stylized fact that the price level has moved in a generally counter-
cyclical fashion in the postwar years, of which Kydland and Prescott (1990)
make so much, damaging to monetarist models of this vintage.
It is indeed true that assertions to the effect that the price level is a
procyclical variable are to be found in the literature. 3 It is also true, as
Kydland and Prescott (1990, p. 7) remark, that "if these perceptions are not
in fact the regularities, then certain lines of research are misguided," and
that, as they further suggest, the equilibrium monetary models of the 1970s
are vulnerable to such criticism. Monetarist models of the 1960s to early
1970s, however, were not constructed to predict any such fact. Nor, incid-
entally, did they rely on the postulate that "the money stock, whether
measured by the monetary base or by Ml, leads the cycle" (1990, p. 5).
Hence Kydland and Prescott's exposure of this postulate as yet another
"misperception" also leaves monetarist models unscathed.
In the following pages, I shall first give an account of how business cycle
theory fit into the macroeconomic consensus of the 1950s and early 1960s,
and then discuss monetarism's role in disturbing this aspect (along with
many others) of the consensus in question, paying particular attention to
the predictions of monetarist models about the cyclical behavior of money
and prices. I shall go on to argue that the main strength of later new-
classical economics, in the Lucas-Sargent-Wallace mode, was not so much
its explanatory power, as its conformity to particular a priori methodological
criteria, and that modem real business cycle theory also conforms to those
criteria, whose appeal is understandable in the light of those earlier
debates. I shall argue that the criteria in question are insufficient to estab-
lish the superiority of modem real business cycle theory and that it there-
fore remains open to empirical challenge from an alternative approach to
which the monetarist economics of the 1960s made a vital contribution.
1 The Rise and Fall of Multiplier-Accelerator Theory
Before the publication of The General Theory, the business cycle occupied
a central position in what we would nowadays call macroeconomics.
Though it shared the stage with inflation, the fact that many monetary
models of the cycle treated inflation as an integral characteristic of the
cycle's upswing meant that divisions here were not sharply marked. The
essentially comparative static analysis of The General Theory, and of the
IS-LM model that came to be regarded as embodying its central theoretical
contribution, shifted the focus of macroeconomics away from cycle theory,
but there was more to the "Keynesian Revolution" than that. 4
Whatever Keynes may have meant, Keynesian economics downgraded
the importance of monetary factors as an explanation of macroeconomic
phenomena. The empirical postulates that the demand for money was
highly interest elastic, and probably unstable too, and the belief that expen-
diture, again as an empirical matter, was insensitive to interest rates,
implied that the real side of the economy, as described by the IS curve,
must be the source of economic disturbances. An alleged empirically stable
marginal propensity to save underpinned a stable multiplier. Hence
fluctuations in investment came to be regarded as the main factor driving
movements in real income and employment, with variations in interest
rates playing at most a marginal and complicating role in the process.
Although, with the onset of the Keynesian revolution, models of the
cycle that located the impulses driving it in the monetary sector therefore
vanished from the mainstream of the literature, cycle theory continued to
flourish as an important component of the subdiscipline, requiring the
inclusion of a chapter or two in any intermediate textbook and the pro-
duction of supplementary textbooks and collections of readings for more
advanced students. The comparative statics of real income determina-
tion, embodied in IS-LM analysis, had to be grasped as ends in themselves,
but, for serious students, mastering them was also an essential preliminary
to coming to grips with "multiplier-accelerator" models, which treated the
cycle as the outcome of systematic rightward and leftward shifts of the IS
curve along an essentially horizontal LM curve. s
The pre-General Theory literature had contained a number of "real"
theories of the cycle. Jevons' speculations about the importance of fluctu-
ations in agricultural output stemming from sunspot activity continued to
attract attention well into the 1920s, operating, in the hands of Pigou, in com-
bination with the notion that business investment was subject to influence
from waves of optimism and pessimism. Schumpeter argued that invest-
ment fluctuated because of irregularities in the pace of innovation and that
the cycle was an integral component of the process of capitalist growth.
Though all of these approaches found a role for monetary factors, it was
a strictly supporting one; but viewed from the perspective of IS-LM analysis,
they were analytically uninteresting. In that framework they provided little
more than reasons--endogenous to the economy perhaps, but exogenous

to the model-why the IS curve might shift back and forth over time.
The accelerator hypothesis was also well developed in that earlier
literature, usually being deployed to explain the relatively large amplitude
of cyclical fluctuations in investment, but it had not found a home in
any complete model of endogenous (cyclical) fluctuations. Keynesian
economics provided the requisite accommodation, and the accelerator
proved to be the perfect partner for the multiplier in a union forged by
exponents of the analytic techniques (then newly available to economists)
of difference and differential equations. The idea of the accelerator as it
appeared in the resulting models was simple enough. Suppose that the
economy is characterized by a fixed coefficients production function, that
capital is fully utilized, but that labor is freely available. Then any increase
in the demand for real output will require firms to acquire new capital
equipment to produce it. The rate of investment will thus be linked to the
rate of change of output by a parameter given by the capital-output ratio.
Investment, however, will have a multiplier effect on output, which in tum
will feed back through the accelerator to investment. In discrete time,
this interactive process may be made to yield a second-order difference
equation, whose dynamic properties might be monotonic, cyclical, damped,
explosive, and so on, depending upon the precise values of the parameters
of the multiplier and accelerator relationships.6
But the incorporation of the accelerator into Keynesian economics was
not just a matter of the application of technique for technique's sake. The
General Theory had made much of the influence of investment on the level
of aggregate demand for goods and services but had carefully avoided deal-
ing with its other role in changing the economy's productive capacity. As
Harrod (1948) and Domar (1947) each showed, a time path for aggregate
demand generated by a particular time path of investment could be an
equilibrium path only if investment was simultaneously providing just the
right amount of capacity to meet the demand in question. The first of these
two relationships involved the multiplier, and the second the economy's
capital-output ratio and hence the accelerator relationship. Multiplier-
accelerator interaction thus seemed to provide the key to extending
Keynesian economics into a system that could deal with both growth and
cycles. The economy's growth path would, within a limit given by popu-
lation growth, be the steady state of the same process driving the cycle, and
of course the policy implications of Keynesian analysis pointed to the
possibility both of stabilizing the cycle and manipulating the economy's
steady state, or warranted, growth path so that it coincided with that of the
labor force and hence the economy's natural growth path.
Now multiplier-accelerator interaction is capable of generating virtually

any type of dynamic behavior, depending on the size of the relevant
coefficients, but it seemed to some, notably Hicks (1949, 1951), that the
accelerator was, as an empirical matter, powerful enough to produce a
monotonically explosive time path for income's deviation from its steady
state path. This presumption formed the basis of his famous "nonlinear"
cycle model, whose qualitative characteristics are easily described. Begin
in the middle of an upswing, with output moving along an explosive
expansionary path, and note that output must eventually reach a level at
which its subsequent rate of growth will be constrained by that of the labor
force. When output growth is thus constrained, investment will fall and a
downward multiplier effect will set a downswing in motion. Since, with
durable capital, the flow of net investment cannot fall below the rate of
depreciation of the actual capital stock, the multiplier alone will carry the
economy to a floor whose value is determined by some exogenously given
level of autonomous expenditure. Once at the floor, existing surplus capital
will wear out, and when it eventually does, an injection of replacement
investment demand will set an explosive upswing going again.
A number of points are worth making about this model. First, it treats
the cycle as an entirely "real" phenomenon, with the demand side of the
economy dominating the upswing and downswing phases of the cycle but
with supply side phenomena providing the constraints that generate its
turning points. Second, the cycle it produces is completely self-generating
and poses no questions about what kind of impulse might set it in motion.
Third, difficult though it is to defend now, until the mid-1960s this model
did dominate the chapters devoted to the cycle in intermediate textbooks.
Its demise, however, was not the result of the rise of monetarism. It is more
accurate to argue that the popularity of monetarist analysis of the cycle
resulted from its ability to fill a gap in macroeconomics created by the fact
that the Hicks cycle model, and the many variations on it in the literature,
failed to survive developments within mainstream macroeconomics during
the 1950s.1t will suffice to enumerate these.
First, the "new" theories of the consumption of function of Friedman
(1957) and Modigliani and Brumberg (1954) suggested that though the
marginal propensity to consume out of permanent, or life cycle, income
might be stable, this did not imply that the marginal propensity to save out
of current income, and therefore the mUltiplier, was also stable. Second,
the neo-classical model of Meade (1961), Solow (1956), and Swan (1956)
soon took over from the Harrod-Domar model as the centerpiece of
growth theory, using a variable proportions production function. Hence
the capital-output ratio became an endogenous variable, whose long-run
equilibrium value was determined as part of the outcome of a process

whereby the economy converged smoothly to a full employment growth
path determined by the rate of growth of the labor force. In the light of this
analysis it became indefensible to treat it as the key structural parameter
linking the growth process to an endogenous cycle. Third, work on the
microeconomics of the accelerator, as surveyed by Knox (1952), reduced
it to little more than a proposition that investment would be some dis-
tributed lag function of the difference between the desired and actual
capital stock. Here, too, variable proportions were the order of the day,
and in the slightly later work of Jorgenson (e.g., 1963), a variable linking
the real and monetary sectors, namely the rate of interest, re-emerged
as a component of the cost of capital and hence as a determinant of the
desired capital stock and an argument in the investment function.
What came to be called the "flexible accelerator," derived from the
capital stock adjustment principle, involved the inclusion of forward-
looking variables in investment equations to capture the expected profit-
ability of new capital equipment; and, analytically closely related, the new
theories of the consumption function modeled consumer expenditure
decisions as being inherently forward looking. But expectations were
usually modeled as distributed lag functions of the variables they con-
cerned, while adjustment processes gave rise to similar functional forms.
As early as 1962, then, a new textbook (Bailey, 1962), in my view the
best (but probably not best selling) of its time, had replaced the obliga-
tory chapter on multiplier-accelerator interaction with one entitled
"Expectations and Adjustment to Change: A Key Link in Cumulative
Movements of the Economy." But expenditure functions based on these
ideas were much less prone to generate explosive behavior. Rather they
tended to produce persistent but damped fluctuations that required
exogenous shocks to get and keep them going. And models containing
them were no longer simple multiplier-accelerator systems.
The aim of modem real business cycle theory is to produce an integrated
account of growth and cycles, firmly based on sound microeconomic
foundations. We have now seen that the integration of growth and cycle
theory was also the aim of multiplier-accelerator theorists. It is ironic,
therefore, that their work foundered on the results generated by the first
round of efforts to produce more satisfactory microfoundations for
macroeconomics. Though multiplier-accelerator models lingered on in
many textbooks for another decade or more, they had ceased to be the
center of active research effort by the early 1960s. The next round of work
on the cycle moved in two directions, at first largely separate and distinct
from one another, each dealing with a separate component of Ragnar
Frisch's (1933) agenda for this area. First, the demise of the self-perpetuating
cycle models meant that the temporarily suspended search for exogenous
impulses had to be taken up again. Second, the study of propagation
mechanisms moved away from the analysis of low-order difference and
differential equation systems that had characterized multiplier-accelerator
theory into the domain of the large econometric models whose complex
distributed lag structures required numerical simulation exercises to reveal
their dynamic properties. 7 These two endeavors were not logically incom-
patible with one another, but, as a matter of fact, monetarism came to
dominate the first while an eclectic type of Keynesian economics was
perpetuated in the second.
2 Monetarism and the Cycle
The literature of what later came to be called monetarism is as complex as

its intellectual roots are varied. It dealt with inflation and monetary policy,
as well as the business cycle. That part of it dealing with the positive theory
of the cycle drew heavily on the National Bureau tradition of Bums and
Mitchell (e.g., 1946), whose work Koopmans (1947) disparagingly char-
acterized as "Measurement Without Theory." In this case, though, a more
accurate description might be "measurement before theory." Friedman's
earliest published studies concentrated on describiIlg certain persistent
quantitative relationships among the behavior of the money stock, money
income, real income, prices, and so on. Only later, partly in response to
critics, did he reveal the theoretical framework in which those observations
could be explained and interpreted, a framework, as is by now well under-
stood, differentiated from that of the then-prevailing Keynesian orthodoxy
more by its quantitative than its qualitative characteristics. 8
Friedman's (1958) study for the Joint Economic Committee of Congress
was mainly concerned with describing both the secular and cyclical
relationships between money and prices revealed in U.S. data, and it was
there that he pointed to the fact that the rate of change (not the level) of the
stock of money displayed "well marked cycles that match closely those in
economic activity in general and precede the latter by a long interval"
(p. 181). Though Friedman was careful to deny unidirectional causation
here-because "the character of our monetary and banking system means
that an expansion of income contributes to an expansion of the money
stock. ... Similarly a contraction of income contributes to a reduction or
slower rate of rise in the money stock" (p. 179)-he nevertheless insisted
that regardless of their source, changes in the rate of money growth would
have subsequent effects on economic activity in general and on the time

path of prices in particUlar. Hence, monetary changes affected the econ-
omy with a long and variable time lag. Subsequent monetarist work on
money and the cycle elaborated these basic themes.
Friedman and Schwartz's (1963a) A Monetary History of the United
States, among many other things, documented the preceding relationships
on a cycle-by-cycle basis and established that the predominance of money
growth as a causative factor was more easily discerned in the violent
fluctuations that marked the interwar years than at other times. Drawing
on the then-unpublished work of Phillip Cagan (1965), it also established
that changes in the stock of money were predominantly the result of
variations usually endogenous, but sometimes, notably in 1937, were the
result of regulatory changes in the banking system's demand for reserves
relative to deposits, and in the public's currency-to-deposit ratio, hence
confirming the cumulative and interactive nature of money's relationship
to the cycle. And Friedman (1961) and Friedman and Schwartz (1963b)
explained the transmission mechanism through which variations in money
growth affected the economy, something that had been very much missing
from Friedman's first account of this matter, and an omission for which he
was frequently criticized long after he had repaired it. Two matters are
important here: First, why is the rate of money growth, rather than the
level of the money stock, the important impulse variable? Second, how can
such an impulse lead to the fluctuations that we call the cycle?
Friedman stressed more than once that one element in the case for
paying attention to money's rate of change was statistical. The stock of
nominal money is a heavily trended variable, and its percentage rate of
change (log first difference) is not. But the process of first differencing
moves the monetary impulse forward in time and, as John Culberston
(1961) pointed out, opens up the possibility that the "long and variable"
time lag is purely a statistical artifact. The level of the nominal money
supply was (as Kydland and Prescott have recently reminded us) not so
much a leading as a coincident variable. There was a good theoretical
reason for treating changes in nominal money growth as the critical monet-
ary impulse, however, and Friedman's (1961) account of it, offered in
response to Culbertson, is worth quoting at some length.
Consider a hypothetical long-run moving equilibrium in which both output and
the stock of money are rising at constant percentage rates, the rise being fully
anticipated so that actual, expected, and desired stocks of money are equal. The
result would be a roughly constant percentage rate of change of prices .... The
percentage rate of change of prices itself is the opportunity cost of holding
money rather than goods, so a constant rate of change in the stock of money
corresponds to a constant opportunity cost of holding money .... An unanti-

cipated change in the rate of change of the stock of money would then produce
a deviation of the actual from the desired stock of money for two reasons:
initially, it would make the actual stock deviate from the expected stock and there-
fore from the desired stock; subsequently, by altering the cost of holding money,
it would change the desired stock itself. These discrepancies will set up adjust-
ments that may very well be cyclical. ... It is therefore theoretically appealing
to regard the "normal" or secular monetary base [sic] around which cyclical fluc-
tuations occur as described by a constant percentage rate of change in the stock
of money and to regard changes in the percentage rate of change as the feature
of monetary behavior that contributes to the generation of cycles. [po 247]
For Friedman, that is to say, a monetary surprise constituted the cyclical
impulse, and a change in the rate of nominal money growth marked its
occurrence. But, although his monetarism had this much in common with
the later new-classical analysis of Lucas and others, its view of the sub-
sequent propagation process, what used to be called the "transmission
mechanism," was very different. Indeed, it was very much in line with what
we nowadays usually think of as Keynesian economics, as the following
summary of Friedman and Schwartz's (1963b, pp. 229-234) account of the
matter makes clear. 9
Beginning in precisely the moving long-run eqUilibrium just described,
they postulate a once and for all change in the rate of monetary growth,
brought about by "an increased rate of open-market purchases by a central
bank." They then point out that "although the initial sellers of the securi-
ties ... were willing sellers, this does not mean that they want to hold the
proceeds in money indefinitely." They sold because they were offered a
good price, not because their cash balances were deficient, and hence will
"seek to readjust their portfolios." To the extent that these original
purchasers were commercial banks, moreover, this effort will involve a
further round of credit and hence money creation. Be that as it may, "It
seems plausible that ... holders of redundant balances will turn first to
securities similar to those which they have sold, say, fixed interest, low-risk
obligations;" but this will bring about a rise in the price of such securities
which will spread to other assets:
... higher-risk fixed-coupon obligations, equities, real property and so forth ....
These effects raise demand curves for current productive services, both for
producing new capital goods and for purchasing current services. The monetary
stimulus is, in this way, spread from the financial markets to the markets for
goods and services.
Output and prices will begin to rise, and money income will tend to over-
shoot its new equilibrium time path because, as higher inflation becomes
anticipated, the economy's demand for real balances falls and also because
second and subsequent round influences on deposit-currency ratios will
affect the behavior of money growth relative to its ultimate equilibrium.
In new-classical analysis, a monetary surprise affects output because,
although the price level moves instantaneously to restore equilibrium be-
tween the supply and demand for nominal money, that movement is misper-
ceived by agents who react by increasing their supply of goods and services,
and its effects are spread out over time because of costs of adjusting
quantities supplied. In stark contrast, for Friedman and Schwartz, persistent
portfolio disequilibrium and its effects on flows of expenditure is of the
essence:
The central element in the transmission mechanism ... is the concept of cyclical
fluctuations as the outcome of balance sheet adjustments, as the effects of flows
on adjustments between desired and actual stocks. It is this interconnectedness
of stocks and flows that stretches the effect of shocks out in time, produces a
diffusion over different economic categories, and gives rise to cyclical reaction
mechanisms. The stocks serve as buffers or shock absorbers of initial changes in
rates of flow, by expanding or contracting from their "normal" or "natural" or
"desired" state, and then slowly alter other flows as holders try to regain that
state. [po 234]
Since among the stocks to which Friedman and Schwartz explicitly drew
attention were those of consumer durables and fixed capital, there was no
reason at all why their view of the monetary impulse could not have been
combined with a real propagation mechanism of the type, already referred
to, incorporating a flexible accelerator mechanism, to produce an eclectic
account of cyclical fluctuations. Matters did not at first develop in this way,
however. While the quantitative analysis of these real propagation mechan-
isms advanced in the context of work on large-scale econometric models,
a separate literature debated the monetarist view of the primacy of monet-
ary impulses.
The debate about the monetary origins of the cycle was untidy and
drawn out. It touched on general issues concerning the difficulty of deriving
conclusions about the direction of causation between variables from data
on timing, and about whether Friedman's (1959) view of the empirical
nature of the demand for money function as being essentially devoid of
significant interest rate effects had led him and his associates to pay
insufficient attention to the possibility that nonmonetary impulses might
sometimes be important. lO But when these issues had been settled in favor
of the critics-data on timing might indeed be misleading, and velocity was
after all interest-sensitive so money income could fluctuate even if money
did not-that still left the core of the monetarist case undamaged. It had
been advanced, not as a set of theorems that were true of logical necessity
under any conceivable circumstances, but as a collection of falsifiable
hypotheses about how real-world data had been generated. Much, though
not all, of the debate about these hypotheses concentrated upon a particu-
lar episode, namely, the Great Depression, and for a number of very good
reasons.
First, the episode in question had generated sufficiently large data
swings that it seemed likely that they might permit discriminating tests to
be carried out. Second, and closely related, Friedman and Schwartz had
used their interpretation of the Depression (and of one or two other rela-
tively major fluctuations) to establish an a priori presumption of the
importance of money for milder cycles, where the data made it much
harder to allocate cause and effect. Finally, and decisively, though the
actual history is a good deal more complicated, in the mythology of
the mainstream economics of the 1950s and 1960s, the Depression was the
episode that had given birth to the Keynesian revolution in economic
thought. To argue that the causes of the Great Depression were monetary,
then, was to mount a challenge to that mythology that could not be ignored
by Keynesians, in whose rhetoric it supplied the example par excellence of
the irrelevance of money in general, and the impotence of monetary policy
in particular.
But undermining Friedman and Schwartz's account of the Depression
proved to be hard indeed. They attributed some causative significance for
the initial downturn, which more or less coincided with the 1929 stock
market crash, to a preceding slowdown in the growth of the monetary base
and hence of the money supply. However, their key claim was that a sharp
recession turned into a major catastrophe because the Federal Reserve
failed to prevent a series of bank failures. The first round of these occurred
in 1930, precipitating large increases in reserve and currency deposit ratios,
which led to a collapse of money growth. In Friedman and Schwartz's view,
this collapse was the proximate cause of falling prices and output. From
1931 to 1933 this "great contraction" fed upon the further bank failures,
which a continued absence of lender of last resort action on the part of
the Federal Reserve permitted to occur. After 1933, the monetary base
continued to grow (as it had since late 1930) but was mainly absorbed into
bank reserves; and, according to Friedman and Schwartz, this showed not
that the Fed was "pushing on a string," but rather that surviving banks had
a well-developed, and easily justified, appetite for free reserves. This
hypothesis found support in the fact that, when reserve requirements were
increased in 1937 to absorb what the Fed believed to be excessive bank
liquidity, the banks responded by restoring their free reserve levels, in
the process causing a money growth slowdown that brought on the 1938
contraction. I I
In short, endogenous, cumulative interaction between the quantity of
money and the cycle, which the authorities could have, but did not, offset,
rather than some mechanical one-way link between exogenous money and
endogenous income, was at the heart of Friedman and Schwartz's story.
And indeed, in their original exposition, they were careful not to identify
the mild monetary slowdown that began in 1928 as the only plausible ex-
planation for the initial onset of the contraction. Thus, though other
scholars could, and did, point to the Smoot-Hawley tariff, bank failures in
Europe, problems in the U.S. housing market, and so forth, as other
important contributing factors, such arguments left the core of the
monetarist case untouched. Only Peter Temin (1976) attempted to rebut
that core. He pointed out that, after the bank failures of 1930, real balances
continued to grow, and nominal interest rates to fall, and argued that these
facts were prima facie evidence that monetary contraction was a purely
passive response to a downturn whose fundamental cause was real, namely,
a downward shift of the consumption function.
There is not space here to survey the debate that Temin provoked.
Suffice it to note his arguments failed to convince the majority of his
colleagues, who wondered, for example, whether the supply of real
balances might not have expanded less than the demand for them between
1929 and 1931, or whether nominal rates of interest were a reliable sole
measure of the opportunity cost of holding money during a period of great
financial uncertainty and rapidly falling prices. Mayer (1978) showed that
Temin's postulated downward shift of the consumption function was very
sensitive to his choice of data. And this is not to mention the difficulty
involved in attempting to reinstate a Keynesian interpretation of the
Depression on the basis of instability of the consumption function, of all
relationships! All in all, the criticism their work provoked failed to under-
mine Friedman and Schwartz's reinstatement of the importance (though
not the sole importance) of monetary influences on the Great Depression.
Certainly, as Karl Brunner (1981, p. 316) remarked, "The story of a power-
less Central Bank acting valiantly but in vain in order to contain the eco-
nomic storm" was effectively disposed of by the controversies of the 1960s
and 1970s. The status of the monetary explanation of the cycle was thus
considerably enhanced, even though, as Brunner also noted, these contro-
versies stopped well short of establishing monetary shocks as the sole
impulse driving the cycle during the 1930s, let alone at other times.
Nor, by and large, did the econometric studies of the time convinc-
ingly establish money's dominance during the Depression. Friedman and
Meiselman's celebrated (1963) Commission on Money and Credit paper,

whose central thrust was to compare the explanatory power over the time
path of nominal consumer spending of variations in autonomous expendi-
ture on the one hand, and the money supply on the other, and which found
that the money supply seemed to perform much better on average, never-
theless attributed an important influence to variations in autonomous
expenditure during subperiods dominated by the 1930s. 12 And, as Ando
and Modigliani (1965) and DePrano and Mayer (1965) soon showed, the
apparent dominance of money in other periods was very sensitive to the
way in which autonomous expenditure was defined. It is true that Andersen
and Jordan's (1968) methodologically closely related study of the relative
effects of money and fiscal variables found little role for the latter, but
fiscal policy is not quite the same thing as autonomous expenditure in
general, and Andersen and Jordan did not study the 1930s, but only the
postwar period.
In short, though the debates of the 1960s and early 1970s did re-establish
monetary factors as being far more important than the almost exclusively
real business cycle models of the preceding two decades had allowed, a
dispassionate reading of their outcome surely supports an eclectic, rather
than monocausal, monetary interpretation of the cycle.13 Interest in the
work that seemed to be leading to this conclusion faded, however, before
such a result became firmly established, as macroeconomics took up and
debated a specific relationship that was to be of central importance to the
evolution of cycle theory during the 1970s and 1980s. I refer, of course, to
the Phillips curve.
3 The Phillips Curve and StiCky Prices
Friedman and Schwartz (1963b) were more precise about the way in which
monetary changes would affect the nominal demand for goods and services
than about how real income and prices would separately react to demand-
side shocks. In this their analysis was completely representative of its time.
The Keynesian economics of the postwar period was above all an econo-
mics of real income determination, and as we have seen the multiplier-
accelerator theory of the cycle that went along with it was a completely
real model. Though, in their work of the 1950s, Friedman and his associates
were far from the Keynesian mainstream in their attention to price level
behavior, this was in contexts in which ouput changes could be ignored. The
macroeconomics of the 1950s, that is to say, could deal with real output
fluctuations, or inflation, but not the two simultaneously and separately.
A.W. Phillips had encountered just this problem in constructing an

analytic cycle model for the study of stabilization policy as an application
of control theory. He had "solved" it in 1954 with an equation in which the
rate of price inflation varied with the excess demand for output. In 1958, in
a paper that was essentially a by-product of the already mentioned research
agenda, he showed that in almost a century of United Kingdom data there
appeared to have existed a rather stable relationship between the rate of
money wage inflation on the one hand and the level of unemployment and
its rate of change on the other. 14 This "Phillips curve" provided the proto-
type for equations that could be inserted in econometric models to allocate
fluctuations in nominal income between real output and the price level; and
in an apparently natural policy application, it also permitted the effects of
stabilization policy on the inflation rate, as well as the level of income and
employment, to be discussed. It was quickly absorbed into the mainstream of
Keynesian macroeconomics, therefore, but remained an object of consider-
able suspicion among exponents of a monetary approach to cycle theory who
were enamored of neither large econometric models nor stabilization policy.
Analysis of the Phillips curve's theoretical foundations soon uncovered
difficulties. Richard Lipsey (1961), elaborating on Phillips' own very
sketchy suggestions, argued that the unemployment rate should be
interpreted as a proxy variable for the excess demand for labor, and wage
inflation's responsiveness to it as nothing more than a manifestation of the
general tendency of any price to respond to excess demand. As Edmund
Phelps (1967) and Friedman (1968) soon pointed out, however, the
relevant price here, according to standard economic theory, was the real
and not the money wage; so that the apparently permanent inflation-
unemployment tradeoff promised by the Phillips curve ought to be at best
a temporary phenomenon, stemming from a tendency for agents' expec-
tations of inflation to lag behind experience. Though this point was quickly
and widely accepted as theoretically correct, the apparent weight of empiri-
cal evidence to the contrary led many commentators-Solow (1968) and
Tobin (1972) provide examples-to argue, as a practical matter, either that
agents did not fully adjust their expectations to experience or that the
frictions of a less-than-perfect labor market prevented them from acting
upon those expectations. Hence the first impact of the Phelps-Friedman
critique of the Phillips curve was to persuade its proponents to distinguish
between long- and short-run inflation-unemployment tradeoffs and to
accept that the former might be less favorable than the latter. It did not,
however, convince them that it was nonexistent.
The catastrophic breakdown of large-scale econometric models during
the inflation of the 1970s was almost solely due to the fact that they
contained equations implying the existence of a long-run inflation-

unemployment tradeoff. That the failure of one equation (or block of
equations), only recently developed, should lead, as Lucas and Sargent
(1978) were later to claim, to the demise of Keynesian economics is surpris-
ing. To begin with, there was more to Keynesian economics than large
forecasting models. Moreover, the failure of one equation did not logically
require the radical transformation of the structure of macroeconomic
analysis that the proponents of new-classical economics set in motion.
Indeed, the monetary model of the cycle that they proposed as the alterna-
tive to the Keynesian analysis embodied in the large econometric models
of the 1970s was from the outset empirically suspect in its own way: That
model has output fluctuations occurring as a response to misperceived
price levels and hence cannot cope with the fact that output fluctuations
lead those in the inflation rate at cyclical turning points. The new-classical
revolution was, however, brought about more by the attractive elegance of
its theoretical models, not to mention its proponents' persuasive use of
language, than by the empirical content of their case.
Lucas (1977) argued that the Phelps-Friedman critique of the Phillips
curve had amounted to much more than a suggestion that an extra variable,
carrying a particular coefficient, should be added to the equation. Rather,
he pointed out, that critique had been based on the contention that the
equation should be constrained to reflect equilibrium behavior. In this
Lucas was quite correct, and as things subsequently transpired, the collapse
of the Phillips tradeoff involved a victory for those who had insisted on the
validity of equilibrium theory over those who had relied on empirical
evidence to discipline their work. IS Hence the episode seemed to vindicate
one methodological criterion and weaken the status of another. In this
context, the word equilibrium can be construed only to refer to the
behavior of maximizing agents whose plans concern real (not nominal)
magnitudes and are executed. Since it is hard to see how economics could
have any predictive content if the plans of individuals were not executed
and therefore did not produce observable consequences, it is difficult to
regret that vindication.
It is, however, a nonsequitur to go on to argue that, because macro-
economics is concerned with the economy as a whole, the same methodo-
logical criterion requires that it should concern itself only with situations
of general equilibrium in which the plans of all agents are rendered compat-
ible with one another by the operation of flexible prices. Nevertheless, that
is how new-classical economists did, and continue to, argue; and the history
of macroeconomics over the past 15 years bears eloquent witness to the
persuasiveness of their nonsequitur and to the attractive tractability of the
theoretical models which adherence to it enabled them to build. 16 I have

written elsewhere (e.g., Laidler 1990a, chap. 4) about this more recent
history, and at a recent conference it was dealt with by Olivier Blanchard
and Robert King. For purposes of this chapter it is sufficient to point out
that the practice of treating the expectations-augmented Phillips curve as
an aggregate supply relationship, forced upon new-classical economics by
its adoption of Walrasian foundations, led it to develop, not a more refined
version of an already existing monetary impulse theory of the cycle, but a
radically different theory; it is this radically different theory that has been
found empirically wanting in the last decade, rather than the general
hypothesis that monetary impulses are important.
It was, and is, possible to construct a model that captures the essential
insights of a late 1960s monetarist view of macroeconomic processes,
including those concerning the absence of any long-run inflation unemploy-
ment tradeoff, but that is also compatible with the basic stylized facts of the
cycle as summarized by, say, Kydland and Prescott (1990). This author
constructed and estimated the prototype of just such a model in 1973. It
consisted of three equations: an income velocity formulation of the quan-
tity theory, an expectations-augmented Phillips curve linking price inflation
to real output's deviation from trend, and-dare one still mention it-a
first-order error-learning characterization of the formation of inflation
expectations. It yielded three key predictions: (1) that the critical exogen-
ous monetary variable on the right-hand side of the reduced form equation
for output's deviation from trend was indeed, as Friedman had suggested,
the rate of growth of the nominal money supply; (2) that it was inflation's
rate of change that varied with the deviation, and rate of change of that
deviation, of output from trend; and (3) that there would be a damped
cyclical response to a monetary shock, with the rate of inflation lagging a bit
less than one-quarter cycle behind output, so that the price level was an
inherently countercyclical variable.
Now the model of which I am here referring was in no sense an outlier in
the literature of its time. It was, in fact, a "bare bones" version of a frame-
work that could easily be extended so as to (1) incorporate the kind of
financial stock-flow interaction discussed by Friedman and Schwartz
(1963b) or Karl Brunner (1961), as well as endogenous fluctuations in
nominal interest rates; (2) test for, and confirm, the presence of other-than-
monetary shocks as a source of real fluctuations; and (3) cope with balance
of payments phenomena in an open economy settingP In short, a small
monetarist model of early 1970s vintage proved to be a limiting case of a
more eclectic Keynesian model. With rather minor extension, moreover, it
was capable of tracking some data as well as, and other data better than,
say Barro's (1978) new-classical system. From the point of view of the
model's mechanics, there is no mystery about why this proved to be so.
First, and crucially different from any monetary new-classical model,
it incorporated the idea that in response to a monetary shock, ouput
changes preceded those in the inflation rate; second, the inclusion of a
geometrically declining weighted average of past values of the inflation
rate on the right-hand side of the Phillips curve guaranteed considerable
inertia to the predicted behavior of inflation. Since these are properties of
the data, the model could not help but fit them. 18
4 The Lessons of the Debates
Two points about the type of model just described are worth stressing.
First, it can accommodate the idea that monetary changes are an important
cyclical impulse and can simultaneously predict the key stylized facts about
the business cycle that readers of Kydland and Prescott (1990) might think
disqualify a monetary explanation from serious consideration. Second, its
capacity to do so stems from its assuming price stickiness. This assumption
ensures that quantities begin to move before prices and that the inflation
rate displays inertia. But, of course, in the 1990s price stickiness is the hall-
mark of so-called "new-Keynesian economics." Monetarism seemed, two
decades ago, to provide a distinct approach to understanding the cycle.
Nowadays it looks more like a particular version of the very orthodoxy to
which it had at one time been diametrically opposed. This appearance is
not just a matter of the vantage point from which we now view the macro-
economics of the 1960s and early 1970s. Friedman's debate with his critics,
published in Gordon (1974), showed that most of their differences, though
deep and important, were nevertheless more quantitative than qualitative
and that further movement toward a synthesis of monetarist and Keynesian
views was in fact implicit in the debates described earlier in this chapter.
Those debates seemed to establish that monetary impulses had been far
more important in generating economic instability than the Keynesian
economics of the 1950s would have had it. They also seemed to establish
that a long-run inflation-unemployment tradeoff, which could be treated
as defining a policy menu, did not exist. Keynesian economics had to
absorb these lessons if it was to continue as an empirically viable and
policy-relevant body of doctrine-and it did so. But the story is not of the
universal triumph of monetarist ideas. To establish monetary instability as
an important source of cyclical impulse and propagation mechanisms does
not also establish its general dominance in either role. To begin with, any
cycle theory has to address the fact of the relatively large amplitude of
variations in investment, and a purely monetary approach seems to have
little to say about the matter. Moreover, the debates that I have described
stopped far short of establishing the unimportance of nonmonetary cyclical
impulses. Anyone still in doubt about the lessons of the historical evidence
on this matter in the early 1970s was soon confronted with the effects of
OPEC's activities on the economy's behavior. Monetarists had to absorb
that lesson, as they did the evidence produced in the 1970s about the sensi-
tivity of the demand for money function to institutional change and its
implications for the effectiveness of a legislated money growth rule as a
built-in stabilizer.1 9
This is not to say that there would have been nothing to argue about by
the mid-1970s had new-classical economics not come upon the scene, but it
is to say that the differences between that body of theory and what had
gone before it were far more profound than any that still needed settling
between monetarists and Keynesians. The new-classical economists did not
advance an alternative set of empirical hypotheses formulated within an
already existing theoretical framework whose validity could be judged by
applying already accepted criteria. Rather they put forward an alternative
theoretical framework whose superiority they defended with recourse to
alternative methodological criteria. In their hands, cycle theory became an
application of general equilibrium analysis, and the approach was defended
with reference, not so much to superior empirical content, as to its con-
formity with the theoretical principles of eqUilibrium modeling.
Even so, empirical evidence did play a role, if only a subsidiary one, in
the literature of new-classical economics. Thus Lucas and Sargent (1978)
made much of the breakdown of the permanent inflation-unemployment
tradeoff in their premature obituary of Keynesian economics, arguing,
misleadingly, that only a model incorporating rational expectations could
cope with the evidence of the 1970s. I have already commented on this and
on Kydland and Prescott's more recent efforts to use stylized facts as a way
of distinguishing between various approaches to cycle theory. The acid test
of new-classical economics' methodological priorities arises, however, with
respect to its treatment of money wage and price stickiness. That many
money wages and prices are subject to contractual arrangements that render
them hard to change is beyond question, as it is that the incorporation
of such a fact into any macroeconomic model gives it distinctly Keynesian
characteristics. If we give priority to the facts in the selection of models, we
should let models that assume price stickiness stand or fall by their
predictions. But because we do not by any means fully understand the
nature of the underlying maximizing behavior that might lead to money
wage and price stickiness, new-classical economists refuse to take seriously

the predictions of models that assume it. 2o An important theme to emerge
from the history recounted in this chapter is, nevertheless, that it is danger-
ous to ignore microfoundations. We have seen that models of consumption
and investment behavior based on maximizing principles undermined
multiplier-accelerator analysis in the 1950s and 1960s, and that the appli-
cation of simple micro theory produced the expectations-augmented Phillips
curve, at a time when a simple permanent tradeoff version of the relation-
ship seemed confirmed by the facts. One can understand the attractiveness
of equilibrium modeling in the light of this experience. But, to repeat a
point made earlier, Walrasian general equilibrium modeling is a rather
special case of that approach.
Only someone who believes that Walrasian general equilibrium theory
embodies a set of first principles whose correct application will always
guarantee the generation of scientific truth would refuse to entertain the
possibility that the postulate of wage and price stickiness might not only
generate better predictions but might also, at some time in the future, be
discovered to be quite compatible with maximizing behavior after all. To
insist on explicit microfoundations as a condition for taking an empirical
hypothesis seriously is to indulge in what Karl Brunner, who discussed
these matters extensively (for example in Brunner, 1989), used to call the
"Cartesian Fallacy," namely, the belief that there exist certain fundamen-
tal postulates from which scientific truth can be deduced, and, crucially,
that we can know when such postulates have been discovered. In the
ongoing debate about business cycles, let us hope that this conference will
help to reinstate empirical evidence as a factor more important than a
priori principles reinforced by the deployment of persuasive language.
Acknowledgments
I am grateful for the research support of the Bradley Foundation. I am also

grateful to Andreas Hornstein, Peter Howitt, Michael Parkin, and Steve
Williamson for comments on an earlier draft.
Appendix A: A Generic Model of Multiplier-Accelerator

Interaction
Where A is real autonomous expenditure, C is real consumption, I is real

investment, Y is real income, s is the marginal propensity to save, and v is
the desired capital-output ratio, we can write
y= C + 1 =A + (1-s)Y_1 + v(Y- 1 - Y- 2 ), (1)

so that
Y=A +(1-s+v)L1 -vL2 • (2)
Since steady state income, Y*, is given by
y*=_l_ A , (3)
l-s
the dynamics of income's deviation from this steady state, Y, are described
by
Y= (l-s + v)Y_ 1 + vY- z. (4)
In general, a difference equation of the form
x - a1x_1 + azx_z = 0
is stable if and only if
a1 - a2 -1 < 0 -a1 -az -1 < 0 and
and will display cyclical characteristics if and only if
at -4«z <0.
If v > 1 and s < 1, such a multiplier-accelerator system mayor may not
generate cycles, but it will be explosive. This is the system that forms the
starting point for Hicks' (1951) cycle model.
Appendix B: A Generic Monetarist Model
When m is the log of the nominal money supply, y the deviation of the log
of real income from its steady state value y*, p the log of the price level, and
Ape the expected inflation rate, we write the quantity theory equation as
Am = bAy + bAy* + Ap, (5)
the expectations-augmented Phillips curve as
Ap = gy-1 + Ap:'l> (6)
and the error-learning hypothesis as
Ape = dAp + (1 - d)Ap:'1. (7)
This model yields, as a reduced form for y,
Y = 1-
b
(il2m _ il2y*) + 2b - g y _ b - (1 - d)g y
b -1 b -2'
(8)
Application of the conditions for stability and cyclical behavior set out in
Appendix A will confirm that it is likely to be cycle prone, but that only if g
is "large" relative to b will this model fail to converge on its steady state.
The relationship between inflation and the cycle may be written as
il2p = gilY_1 + dgilY_2' (9)
With the rate of acceleration of prices thus a function of the rate of growth
of income's deviation from trend and the lagged value of this variable, the
price level clearly lags well behind income's deviation from trend, and
hence the "cycle."
This model is described in detail, and estimated, in Laidler (1973).
Notes
1. The quotation from Keynes comes from a New Statesman article of 1933. I have
borrowed it from Elizabeth Johnson (1978, p. 20). The reader's attention is drawn to the fact
that I do not here use the phrase "persuasive use of language" as synonymous with the word
"rhetoric." As readers of McCloskey (1983) will understand, there is much more to the way
in which economists converse with one another than their choice of words. This matter is,
however, sometimes important-indeed to someone like myself who clings to the rather old-
fashioned stance of attaching primary importance to the persuasive power of empirical
evidence, it is occasionally too important 1
2. Even "real demand side" models-of the multiplier-accelerator variety, for example
-require that the monetary system operate so as to prevent the rate of interest moving
to equilibrate savings and investment at full employment if they are to generate cyclical
fluctuations.
3. But, the data for the first half of this century do show prices to move procyclically. See
Haberler (1956, p. 133), Matthews (1959, p. 3), and indeed Lucas (1977, p. 3) for assertions
about this characteristic of the data. It is only in the postwar period that this has ceased to be
the case. The postwar reversal of the "loops" around the Phillips curve, discussed in note 15,
reflects this change in the stylized facts.
4. Haberler (1937) still provides the best available survey of interwar cycle theory. See
Patinkin (1990) for a persuasive defense of IS-LM as a valid, though simplifying, account of
the central theoretical content of The General Theory.
5. The American Economic Association produced two sets of readings in the area of
business cycles-Haberler (1944) and Gordon and Klein (1%5). A random inspection of old
intermediate macroeconomics textbooks confirms that those of Day (1957), Demberg and
McDougall (1960), and Brooman (1962) all contained chapters on the cycle. Matthews (1959)
was a widely used supplementary text of the same period. There is no reason to believe that
this small sample of what students might have been asked to read three decades ago is unrep-
resentative, and in each case the cycle theory presented was of the multiplier-accelerator
type.
6. Harrod (1936) and Samuelson (1939) should be credited with pioneering the analysis of
multiplier-accelerator interaction as a key to cycle theory. Curious readers, whose training
has been recent enough to leave them unfamiliar with the general nature of this class of
models, are referred to Appendix A, where a simple generic example is briefly described.
7. The pioneering paper was surely that of Adelman and Adelman (1959).
8. The relevant papers of Friedman are reprinted as chapters 9-12 in his 1969 collection,
The Optimum Quantity of Money. See also R.I. Gordon (1974).
9. Though it is convenient to cite and quote Friedman and Schwartz's account of the trans-
mission mechanism here, the reader should not be misled into believing that their account was
in any way unique. Johnson (1962) noted that a variety of essentially similar accounts of the
mechanism were emerging in the literature at that time, as much from Keynesians as from
quantity theorists (as monetarists were then called). He quoted a lengthy section from
Brunner (1961) to illustrate its properties.
10. In particular, Friedman's downplaying of the importance of the interest sensitivity
of velocity, most notable in (1959) but also apparent in Friedman and Schwartz (1963b),
attracted much comment before being finally laid to rest as an issue by Laidler (1966) and
Friedman himself (1966). The question of interpreting the timing of data was most notably
debated by Tobin and Friedman in their 1970 exchange in the Quarterly Journal of
Economics.
11. Even so, this 1933-1937 episode still seems to me to be the most difficult one for a
monetary explanation of the 1930s to cope with. That surviving banks might want to build up
free reserves in the wake of the bank failures of 1930-1933 is not surprising, but that they
should do so slowly over a four-year period is much harder to understand. Morrison's (1966)
explanation of the phenomenon is cast in terms of backward-looking expectations and would
be well worth some new analysis. The reader's attention is drawn to the fact that the em-
phasis of monetarist analysis of the 1930s is on the behavior of the money supply. Other
commentators (e.g., Bernanke, 1983), while focusing on events in the banking system, stress
the destruction of the information-processing and decision-coordinating capacity of the
financial sector, rather than the mere contraction of the money supply, as an important
element in the propagation of the Great Depression.
12. And, as Johnson (1962) noted at the time, a Keynesian could take considerable
comfort from the fact that a theory designed to explain fluctuations in income during periods
of chronic unemployment seemed capable of doing just that.
13. This is the view of Brunner and Meltzer (1987).
14. Two points are worth making about Phillips' contribution. First, the major finding
of his research on stabilization policy, which utilized his engineer's knowledge of control
mechanisms, was that feedback rules had to be extremely carefully designed and implemented
if they were not to be actively destabilizing; so Phillips himself was very nearly as skeptical as
Friedman about the feasibility of "fine-tuning." Second, according to Conrad Blyth (1987),
the Phillips curve itself began life as an attempt to meet Sir Dennis Robertson's criticism that
Phillips' earliest work concentrated too exclusively on real variables. That the analytical
centerpiece of the Keynesian econometric models that were later to be so confidently used to
design stabilization policies should have such a pedigree is ironic, to say the least.
15. But it should be noted that Phillips' original empirical results still need to be
explained. They show procyclical behavior of the inflation rate, rather than the price level, but
the presence of counterclockwise loops around the relationship in earlier years (they are
clockwise in postwar data) tends to make inflation lead the cycle, and hence to make the price
level itself procyclical, thus reflecting the stylized facts referred to in note 3. Presumably the
relatively constant inflation expectations existing in Britain under the gold standard, and
under the Bretton Woods system, go a long way toward explaining the appearance of a stable
inflation-unemployment tradeoff before 1914 and during the 1950s. The interwar years never
did generate so well-defined a relationship. During the 1960s, and with considerable percep-
tiveness, G.c. Archibald referred to this phenomenon in conversations as "the econometric
consequences of Mr. Churchill." And, of course, the disappearance of the Phillips curve in the
1970s came immediately after the breakdown of the Bretton Woods system.
16. This is not the place to rehearse well-known arguments about the appropriateness of
assuming price stickiness in the absence of a widely accepted microeconomic explanation of
the phenomenon. The argument that the behavior of prices ought to be explained as the
outcome of maximizing behavior is unexceptional but does not suffice to defend new-
classical analysis, since, in a competitive model, it is well known that all endogenous agents
are price-takers and that such a model cannot explain how prices are set. Since I know of
no well-articulated microeconomic explanation of money price flexibility, and since stickiness
of many money prices does seem to be a well-documented fact, I agree wholeheartedly with
Brunner (1989) on the appropriateness of making the assumption in question.
17. The reader interested in the formal properties of this type of model is referred to
Appendix B, where these are briefly described. The most complete extension of this line of
research into a rnacroeconometric model is surely to be found in the publications of Peter
Jonson and his associates at the Reserve Bank of Australia (see Jonson, Moses, and Wymer,
1976, and Jonson and Trevor, 1980). These models were differentiated from contemporary
Keynesian models by their relatively small size « 30 behavior equations), their emphasis on
the importance of monetary impulses, and their lack of a long-run inflation-unemployment
tradeoff. With hindsight, their similarities to Keynesian models, which stem from the assump-
tion that wages and prices adjust slowly to market disequilibrium over real time, are more
striking than their differences. It is worth noting that these models are not simply systems of
independently estimated equations, but incorporate the cross-equation restrictions implied
by the theoretical analysis of stock-flow interaction that lies at the heart of the monetarist
analysis of the transmission mechanism. It is also worth noting that they represent a product
of the research agenda begun by Phillips (1954) and continued by Rex Bergstrom and
Clifford Wymer (1974). Finally, it should be pointed out that small empirical models were
not a complete monetarist monopoly. Benjamin Friedman (1977), for example, presents a five-
equation model of a very Keynesian type.
18. For a small model of the U.S. economy with these characteristics, and results that are
directly comparable (and are compared) to those obtained by Barro (1978), see Laidler and
Bentley (1983).
19. The relevant literature as far as the United States is concerned was ably surveyed by
Judd and Scadding (1982). My own views on the significance of the evidence on the instability
of the demand for money function that emerged in the 1970s have most recently been set out
in Laidler (1990b).
20. And given that, from the 1880s onward, money wage stickiness was the explanation of
cyclical unemployment accepted by those whom we would nowadays classify as "classical"
economists-for example, Alfred Marshall and A.C. Pigou among others-and that, in Chap-
ter 19 of The General Theory, Keynes made it quite clear that he did not believe that his own
explanation of the phenomenon depended on such stickiness, the inappropriateness of the
current usage of the adjectives "new-classical" and "new-Keynesian" could not be more total.
It is, however, far too late to do anything about it.
THE CYCLE BEFORE NEW·CLASSICAL ECONOMICS 109
References
Adelman, 1., and F.L. Adelman. 1959. "The Dynamic Properties of the Klein-
Goldberger Model." Econometrica. Reprinted in Gordon and Klein (1965).
Andersen, L.e., and J.L. Jordan. 1968. "Monetary and Fiscal Actions: A Test of
Their Relative Importance in Economic Stabilization." Federal Reserve Bank
of St. Louis Monthly Review 50: 11-23.
Ando, A, and F. Modigliani. 1965. "The Relative Stability of Monetary Velocity
and the Investment Multiplier." American Economic Review 55: 693-728.
Bailey, M.J. 1962. National Income and the Price Level. New York, McGraw-Hill.
Barro, RJ.1978. "Unanticipated Money, Output, and the Price Level in the United
States." Journal of Political Economy 86: 549-58l.
Bergstrom, Rand C.R Wymer. 1974. "A Model of Disequilibrium Neo-Classical
Growth and Its Application to the United Kingdom." LSE International Monet-
ary Economics Research Program (mimeo).
Bernanke, B. 1983. "Non-Monetary Effects of the Financial Crisis in the Propa-
gation of the Great Depression." American Economic Review 73: 257 -276.
Blyth, e. 1987. "AW. Phillips." In J. Eatwell, M. Millgate, and P. Newman, eds.,
The New Palgrave. London: Macmillan.
Brooman, F.1962. Macroeconomics. London: George Allen and Unwin.
Brunner, K. 1961. "The Report ofthe Commission on Money and Credit." Journal
of Political Economy 69: 605-620.
- - - . 1981. "Epilogue: Understanding the Great Depression." In K. Brunner,
ed., The Great Depression Revisited. The Hague: Martinus Nijhoff Publishing.
- - - . 1989. "The Disarray in Macroeconomics." In F. Capie and G.E. Wood,
Monetary Economics in the 1990s. London: Macmillan.
Brunner, K., and AH. Meltzer, eds. 1978. The Problem of Inflation. Carnegie
Rochester Conference Series, vol. 8. Amsterdam: North-Holland.
- - - . 1987. "Money and the Economy: Issues in Monetary Analysis." The 1987
Raffaele Mattioli Lectures, Carnegie-Mellon University (mimeo).
Burns, AF., and W.e. Mitchell. 1946. Measuring Business Cycles. New York:
NBER
Cagan, P. 1965. Determinants and Effects of Changes in the Stock of Money 1870-
1960. New York: Columbia University Press for the NBER
Culbertson, J.M. 1961. "Friedman on the Lag in Effect of Monetary Policy."
Journal of Political Economy 68: 617-621.
Day, AC.L. 1957. Outline of Monetary Economics. Oxford: Clarendon Press.
DePrano, M., and T. Mayer. 1965. "Tests of the Relative Importance of Auto-
nomous Expenditure and Money ." American Economic Review55: 729-752.
Dernberg, T.F., and D.M. McDougall. 1960. Macroeconomics. New York:
McGraw-Hill.
Domar, E. 1947. "Expansion and Employment." American Economic Review 37:
34-55.
Friedman, B.M. 1977. "The Inefficiency of Short-Run Monetary Targets for
Monetary Policy." Brookings Papers on Economic Activity 2: 293-346.
Friedman, M. 1957. A Theory of the Consumption Function. Princeton, N.J.:

Princeton University Press for the NBER
- - - . 1958. "The Supply of Money and Changes in Prices and Output." JEC of
85th Congress. Reprinted as chap. 9 of Friedman (1969).
---.1959. "The Demand for Money: Some Theoretical and Empirical Results."
Journal of Political Economy. Reprinted as chap. 6 of Friedman (1969).
- - - . 1961. "The Lag in Effect of Monetary Policy." Journal of Political Eco-
nomy. Reprinted as chap. 11 of Friedman (1969).
- - _ . 1966. "Interest Rates and the Demand for Money." Journal of Law and
Economics. Reprinted as chap. 7 of Friedman (1969).
- - - . 1968. "The Role of Monetary Policy." American Economic Review.
Reprinted as chap. 5 of Friedman (1969).
- - - . 1969. The Optimum Quantity of Money and Other Essays. London:
Macmillan.
---.1970. "Comment on Tobin." Quarterly Journal of Economics 84: 318-327.
Friedman, M., and D. Meiselman. 1963. "The Relative Stability of Monetary
Velocity and the Investment Multiplier in the United States 1898-1958." In
Commission on Money and Credit, Stabilization Policies. Englewood Cliffs,
N.J.: Prentice-Hall.
Friedman, M., and A.J. Schwartz. 1963a. A Monetary History of the United States
1867-1960. Princeton, N.J.: Princeton University Press for the NBER
- - - . 1963b. "Money and Business Cycles." Review of Economics and Statistics.
Reprinted as chap. 10 of Friedman (1969).
Frisch, R 1933. "Propagation Problems and Impulse Problems in Dynamic
Economics." In Economic Essays in Honour of Gustav Cassel. Reprinted in
Gordon and Klein (1965).
Gordon, RA., and L.R. Klein, eds. 1965. Readings in Business Cycles. Homewood,
m.: Richard Irwin, for the AEA.
Gordon, R.J., ed. 1974. Milton Friedman's Monetary Framework. Chicago: Univer-
sity of Chicago Press.
Haberler, G. 1937. Prosperity and Depression. Revised edition reprinted (1964).
Cambridge, Mass.: Harvard University Press.
- - - . ed. 1944. Readings in Business Cycle Theory. Homewood, Ill.: Richard
Irwin for the AEA.
- - - . 1956. "Monetary and Real Factors Affecting Economic Stability: A
Critique of Certain Tendencies in Modem Economic Theory." Banca Nazionale
del Lavoro Quarterly Review. Reprinted in Gordon and Klein (1965).
Harrod, RF. 1936. The Trade Cycle. Reprinted 1965. New York: Augusts Kelley.
- - - . 1948. Towards a Dynamic Economics. London: Macmillan.
Hicks, J.R 1949. "Mr. Harrod's Dynamic Theory." Economica. Reprinted in
Gordon and Klein (1965).
- - - . 1951. A Contribution to the Theory of the Trade Cycle. London: Oxford
University Press.
Johnson, E.S. 1978. "Keynes as a Literary Craftsman." In E.S. Johnson and H.G.
Johnson, The Shadow of Keynes. Chicago: University of Chicago Press.
Johnson, H.G. 1962. "Monetary Theory and Policy." American Economic Review
52: 335-384.
Jonson, P.D., E.R. Moses, and e.R. Wymer. 1976. "A Minimal Model of the
Australian Economy." Reserve Bank of Australia Discussion Paper 7601,
Sydney (mimeo).
Jonson, P.D., and R.G. Trevor. 1980. "Monetary Rules: A Preliminary Analysis."
Reserve Bank of Australia Discussion Paper 7903 (revised 1980), Sydney
(mimeo).
Jorgenson, D.W. 1963. "Capital Theory and Investment Behavior." American
Economic Review 53, Papers and Proceedings, 247-259.
Judd, J., and J. Scadding. 1982. "The Search for a Stable Money Demand Function:
A Survey of the Post-1973 Literature." Journal of Economic Literature 20: 993-
1023.
Knox, A.D. 1952. "The Acceleration Principle and the Theory of Investment: A
Survey." Economica NS. 19: 269-297.
Koopmans, T.e. 1947. "Measurement Without Theory." Reprinted in Gordon and
Klein (1965).
Kydland, F., and E. Prescott. 1982. "Time to Build and Aggregate Fluctuations."
---.1990. "Business Cycles: Real Facts and a Monetary Myth." Federal Reserve
Bank of Minneapolis Quarterly Review (Spring): 3-18.
Laidler, D. 1966. "The Rate of Interest and the Demand for Money: Some Empiri-
cal Evidence." Journal of Political Economy 74: 545-555.
- - - . 1973. "The Influence of Money on Real Income and Inflation: A Simple
Model with Some Empirical Tests for the United States 1953-72." Manchester
School 41: 367 -395.
- - - . 199Oa. Taking Money Seriously. Cambridge, MA: MIT Press.
- - . 1990b. "Understanding Velocity: New Approaches and Their Policy
Relevance-Introduction." Journal of Policy Modeling 12: 141-164.
Laidler, D., and B. Bentley. 1983. "A Small Macro-Model of The Post-War United
States." Manchester School 51 : 317 - 340.
Lipsey, RG. 1961. "The Relationship Between Unemployment and the Rate of
Change of Money Wage Rates in the United Kingdom 1862-1957." Economica
NS. 27: 115-138.
Lucas, RE., Jr. 1977. "Understanding Business Cycles." In K. Brunner and
A.H. Meltzer, eds., Stabilization of the Domestic and International Economy.
Carnegie Rochester Conference Series, vol. 5. Amsterdam: North-Holland.
Lucas, RE. Jr., and T.J. Sargent. 1978. "After Keynesian Economics." In After the
Phillips Curve: Persistence of High Inflation and High Unemployment. Boston:
Federal Reserve Bank of Boston.
Matthews, RC.D. 1959. The Business Cycle. Chicago: University of Chicago Press.
Mayer, T. 1978. "Consumption in the Great Depression." Journal of Political
Economy 86: 139-146.
McCloskey, D. 1983. "The Rhetoric of Economics." Journal of Economic Litera-
ture 21: 481-517.
112 lHE BUSINESS CYCLE
Meade, J.E. 1961. A Neoclassical Theory of Economic Growth. London: Allen and
Unwin.
Modigliani, F., and R. Brumberg. 1954. "Utility Analysis and the Consumption
Function: An Interpretation of the Cross-Section Data." In K. Kurihara. ed.,
Post Keynesian Economics. New Brunswick, N.J: Rutgers University Press.
Morrison, G. 1966. Liquidity Preferences of Commercial Banks. Chicago: Univer-
sity of Chicago Press.
Patinkin, D. 1990. "In Defense of IS-LM." Banca Nazionale del Lavoro Quarterly
Review 172: 119-134.
Phelps, E.S. 1967. "Phillips Curves, Expectations of Inflation, and Optimal Unem-
ployment over Time." Economica NS 34: 254-281.
Phillips, A.W. 1954. "Stabilization in a Gosed Economy." Economic Journal 64
(June): 290-323.
- - - . 1958. "The Relation Between Unemployment and the Rate of Change of
Money Wages Rates in the United Kingdom." Economica NS 25: 283-299.
Plosser, C. 1991. "Money and Business Cycles: A Real Business Cycle Inter-
pretation." In M. Belongia, ed., Monetary Policy on the 75th Anniversary of the
Federal Reserve System. Boston: Kluwer Academic Publishers.
Samuelson, P.A. 1939. "Interactions Between the Multiplier Analysis and the Prin-
ciple of Acceleration." Review of Economics and Statistics 21: 75-78.
Solow, R.M. 1956. "A Contribution to the Theory of Economic Growth." Quar-
terly Journal of Economics 70: 65-94.
- - - . 1968. "Recent Controversies in the Theory of Inflation: An Eclectic View."
In S.W. Rousseas, ed., Proceedings of a Symposium on Inflation, Its Causes,
Consequences and Control. Wilton, Conn.: Calvin K. Kazanjian Economics
Foundation.
Swan, T.W. 1956. "Economic Growth and Capital Accumulation." Economic
Record 32: 334-361.
Temin, P. 1976. Did Monetary Forces Cause the Great Depression? New York:
Norton.
Tobin, J. 1970. "Money and Income: Post Hoc Ergo Propter Hoc?" Quarterly
Journal of Economics 84: 301-317; "Rejoinder," 328-329.
- - - . 1972. "Inflation and Unemployment." American Economic Review 62:
1-18.
Commentary
by Ben S. Bernanke
David Laidler's chapter is a very informative and broad-minded discussion

of the traditional Keynesian-monetarist debate-so broad-minded in fact
as to make one wonder how, with participants like Laidler, the debate
could have been as rancorous as it was. Along the way, Laidler also
offers some views on the more recent debate, the controversy between
the Keynesian-monetarist synthesis and the so-called real business cycle
(RBC) school.
In my commentary I will do two things: First, I will list what I believe
to be the principal contributions of monetarism, with a few words about
recent developments. Unlike the author, I was not a participant in the
Keynesian-monetarist debate as it happened, so perhaps the differences
in our perceptions will be sociologically if not economically illuminating.
Second, I would like to say a bit about the theme suggested by the chapter's
title, namely, the parallels and differences between the monetarist and
RBC assaults on the Keynesian paradigm.
In retrospect, I believe monetarism made three major contributions.
The first was to establish the empirical importance of money in cyclical
fluctuations (Laidler and I agree here). Friedman and Schwartz's Monetary
History (1963), the first comprehensive documentation of the role of money
in the business cycle, is one of the most persuasive books ever written on
economics. The findings of the "narrative approach" used by Friedman
and Schwartz (as described and updated by Romer and Romer, 1989), in
conjunction with postwar episodes such as the Volcker and Thatcher
disinflations, are the main reason that purely nonmonetary models of the
business cycle remain today a minority taste.
During the last two decades, considerable research has confirmed the
empirical importance of money in economic fluctuations, real business
cycles notwithstanding. For example, Friedman and Schwartz's monetarist
interpretation of the Great Depression has been substantially supported by
a developing consensus among economic historians. According to this new
consensus, the Depression was caused by the gross mismanagement of the
international gold standard, which in turn led to falling money supplies
in the gold standard countries (see, for example, Eichengreen and Sachs,
1985; Hamilton, 1987; Temin, 1989; and Bernanke and James, 1991). Much
recent research has also found evidence for the role of money in postwar
cycles. For example, attempts to measure the effects of monetary policy
113
and its channels of transmission have tended to find that contractionary

policy played a role in nearly every recession of the last 30 years (Romer
and Romer, 1989; Bemanke and Blinder, forthcoming).1
The second central contribution of monetarism that I would point to is
the idea that in the long run the unemployment rate equals the natural rate.
Laidler does not emphasize this point; however, the natural rate idea was
clearly stated by Friedman (for example, in his Presidential Address and
Nobel lecture), and it underlies some of his policy recommendations (such
as the constant-growth-rate rule for money).
Currently, the notion of a natural rate is an integral part of the macro-
economic orthodoxy, and a vertical long-run Phillips curve is implied by all
the leading theories of monetary nonneutrality. It is interesting that, empir-
ically, the natural rate hypothesis may not be on such firm ground, a point
that Laidler makes. For example, at least some people interpret the persis-
tence of the Great Depression and the' 'hysteresis" in European unemploy-
ment as evidence against the natural rate view. More narrowly, empirical
studies such as that of Romer and Romer (1989) find that the real effects of
a monetary shock decay extremely slowly. This issue awaits further research.
The third contribution of monetarism, also not much emphasized by
Laidler, was to make the case for nonactivist policy and in particular
the use of rules (such as the famous "k percent rule" for money growth).
The non activist conclusion follows from the combination of the natural
rate theory, monetarists' empirical findings about long and variable outside
lags, and the plausible belief that policy makers may not always have
incentives to act in the interest of the society as a whole. The advocacy of
rules by monetarists over the years has been an important check on excess-
ive optimism about what stabilization policy could realistically accomplish.
More recent research has found a powerful new theoretical rationale for
rules, namely, that the use of rules can help policy makers achieve credi-
bility and avoid the problems of dynamic inconsistency. Nevertheless, in
practice, policy making these days seems pretty activist (on the day I write
this, the Fed has just lowered the discount rate in response to cyclical
conditions). Why does nonactivism seem to be losing out in policy circles?
The problem is that we are no longer confident that we can design a rule
that can deliver at least reasonable macroeconomic performance almost all
of the time. Without such confidence, policy makers cannot commit cred-
ibly to a rule. In particular, money growth rules are perceived as having
failed to be robust in situations where there is significant financial inno-
vation or other instability in money demand.
Having reviewed the very substantial contributions of monetarism, I
tum now to a discussion of the current "cycle" in macroeconomics-the
COMMENTARY 115
competition between the Keynesian-monetarist synthesis and the RBC

approach. In his chapter, Laidler vacillates to some degree both in his
assessment of RBC contributions and in his explanation of why this
approach has had significant success at winning adherents among re-
searchers. In the end, I suspect that the following three statements, which
describe my own views, are not too far from what Laidler is suggesting
in his chapter:
1. The principal criteria that determine whether a new approach is

accepted in any scientific field are demonstrated empirical success,
theoretical or methodological advantages, and persuasive rhetoric.
2. In varying degrees, the monetarist approach of the 1950s and 1960s
met all three of these criteria and thus became part of the main-
stream synthesis.
3. The real business cycle approach has some clear methodological
attractions and some effective rhetoric to commend it. What is miss-
ing at this point-and this is the reason that Laidler is less than
enthusiastic-is a demonstration that the RBC models produce
better empirical fits.
Ray Fair's chapter goes into more detail about the empirical
shortcomings of the RBC approach. Here I would like to focus
on the question of methodology. What exactly is distinctive and appeal-
ing about the RBC methodology, enough so to bring the approach a
considerable following even in the absence of convincing empirical
successes?
Let me start by listing some features of RBC models that are often
pointed to but are not, in my view, the defining aspects of those models.
First, although much has been made of the productivity or supply shock
aspect of early RBC models, the focus on supply shocks is not essential to
that approach. Witness, for example, the recent work on fiscal and monet-
ary shocks in an RBC context. Nor is the emphasis on the representative
agent, the attempt to base behavior on rigorous microfoundations, or even
the assumption of instantaneous market-clearing essential. It is not difficult
to find recent examples of RBC-style research missing each of these
characteristics.
Rather, what is distinctive and attractive about the RBC methodology is
its emphasis on general equilibrium modeling as the right way to do macro-
economics. The strict adherence to general equilibrium modeling has a
number of advantages. Greatest among these is logical coherence-the
general equilibrium approach helps us see how the various aspects of the
economy fit together. For example, in his chapter Ben Friedman correctly
points out the difficulty of isolating "pure" demand or supply shocks to
the macroeconomy. Rather than being an argument against the RBC
approach, however, Friedman's observation is an argument in favor of it: It
is precisely a general equilibrium approach that is needed to work out both
the demand and supply implications of, for example, a rise in the price of
imported oil.
In principle, the general equilibrium approach also has important
empirical advantages. At a minimum, the general equilibrium approach
eliminates the spectacle of independent research teams working on the
different equations of the macro model, each with its own set of mutually
inconsistent identifying assumptions. More positively, the full-information
estimation of a general equilibrium model increases our ability to identify
parameters or test specifications, since each paramenter value or specifica-
tion choice has implications for the behavior of the entire model, not just
its "own" equation.
Given these advantages, why hasn't the RBC approach produced more
solid empirical results? The main problem, it seems to me, is the com-
putational complexity associated with solving general equilibrium models
that embody even modest degrees of realism. At this point the cost and
technical difficulty of working with these models are too great for them to
be the workhorses of quantitative macroeconomic analysis that their
proponents would like them to be. This may change in the future, of
course, as we learn more about how to work with these models and as
computing costs continue to decline.
This brief discussion helps clarify, I hope, the difference between the
monetarist and RBC attacks on the Keynesian paradigm-the old cycle and
the new cycle. While the monetarist approach involved a methodological
component, its main contributions-the empirical relevance of money
to the business cycle, the natural rate hypothesis, and the case for non-
activism-are straightforward propositions about how the world works. In
contrast, the RBC innovation is to introduce an appealing but-at
this point-not fully developed methodology for systematically studying
macroeconomic questions. This methodology has not delivered very much
so far, it is true; but would we rule out further investigation of supercon-
ductivity (for example) because of its limited practical value thus far?
I think a fair evaluation of the RBC approach will have to wait a few
more years. In the meantime, the practical knowledge delivered by the
Keynesian-monetarist synthesis will continue to be of great value in practi-
cal debates about the economy and economic policy.
COMMENTARY 117
Note
1. However, these authors do not necessarily equate tight monetary policy with slow
money growth per se. Romer and Romer use the minutes of FOMe meetings to try to
measure policy intentions, while Bernanke and Blinder use interest rate indicators of monet-
ary policy.
References
Bernanke, Ben, and Alan Blinder. Forthcoming. "The Federal Funds Rate and the
Channels of Monetary Transmission." American Economic Review.
Bernanke, Ben, and Harold James. 1991. "The Gold Standard, Deflation, and
Financial Crisis in the Great Depression: An International Comparison." In
R. Glenn Hubbard, ed., Financial Markets and Financial Crises. Chicago:
University of Chicago Press for NBER.
Eichengreen, Barry, and Jeffrey Sachs. 1985. "Exchange Rates and Economic
Recovery in the 1930s." Journal of Economic History 45: 925-946.
Friedman, Milton, and Anna J. Schwartz. 1963. A Monetary History of the United
States, 1867-1960. Princeton, N.J.: Princeton University Press for NBER.
Hamilton, James. 1987. "Monetary Factors in the Great Depression." Journal of
Monetary Economics 19: 145-169.
Romer, C. and Romer, D. 1989. "Does Monetary Policy Matter? A New Test in the
Spirit of Friedman and Schwartz." NBER Macroeconomics Annual 1989.
Cambridge, MA: MIT Press, pp. 121-170.
Temin, Peter. 1989. Lessons from the Great Depression. Cambridge, MA: MIT
Press.
SESSION II
3 FOR A RETURN TO
PRAGMATISM
Olivier Jean Blanchard
For the last 20 years, mainstream macroeconomics has been facing a strong
challenge. For a while, it looked as though it would not survive, as though
it would be reconstructed along drastically new lines. As the challenge is
now clearly fading, it is a good time to assess what has been done and
where to go from here. This is what this chapter does. The message is not
one of self-congratulation. There has been progress but also much time
wasted. Put simply, to capitalize on what we have done, we need to return
to pragmatism.
1 Macroeconomics, Circa 1970
Behind the heated exchanges between Keynesians and monetarists, there

was, by 1970, a largely common view both of the structure of the economy
and of the sources and nature of fluctuations: Shocks were seen to come
mostly from shifts in aggregate demand, from either IS or LM shifts. The
adjustment process, which required the adjustment of nominal prices and
wages, was slow and unreliable, with both stabilizing and destabilizing
mechanisms at work. Without policy, the effects of shocks were prolonged.
121
Policy could in principle reduce their effects and their duration. But given
the realities of the political process, and the lags in the transmission pro-
cess, whether fiscal and monetary policies were likely to actually improve
things was very much a matter of debate.
This common view of structure and basic mechanisms allowed for
division of labor, with work on the parts proceeding in parallel with work
on an integrated structure. Work on the pieces-consumption, invest-
ment, and so on-was proceeding through the simultaneous development
and interaction of theoretical models, construction of the appropriate
aggregates, and estimation of aggregate relations. Those making the theor-
etical breakthroughs-Modigliani, Friedman, Jorgenson and Hall-were
often those making empirical progress as well. But the empirical speci-
fications were inspired from, rather than dictated by, theoretical work.
The belief was that aggregation problems were such that individual
microoptimizing relations should serve as a guide, not as a corset to the
estimation of aggregate relations.
Work on an integrated structure was proceeding in parallel. Small
analytical models, with an IS-LM backbone, but paying attention to a
particular aspect of general equilibrium, were used to identify channels and
strengthen the intuition. The strength of the best ones, those built by the
likes of Tobin and Mundell, came from the smart shortcuts they used to
deliver a high ratio of relevant insights to machinery. At the same time,
larger empirical models such as the MPS provided, by integrating the
different estimated components, a much richer dynamic structure. They
gave a sense of the time dimension of the relevant dynamics, something
small analytical structures could not do. When simulated, they would often
reveal unexpected dynamics; if those dynamics appeared implausible,
research would identify their source and possibly modify the models. The
models could be used to study medium-run issues, which were hard to
handle analytically. Most importantly, they provided a common practical
and mental structure in which the implications of new contributions could
be assessed, alternative formalizations compared, and so on.
Not everything was perfect. There were two main problems. The first
would trigger the crisis and the new-classical challenge. The second would
become obvious only in retrospect.
While optimization was heavily used to think about the behavior of con-
sumers and firms, from consumption to investment to portfolio decisions,
this reliance on optimization did not carry over to the explanation of price
and wage determination. This was not based on the belief that markets
could be thought of as perfectly competitive, so that there was no need to
think about who was setting prices and how. Indeed, a central aspect of the
FOR A RETURN TO PRAGMATISM 123
adjustment to shocks was the small and slow adjustment of nominal prices
and wages. And the belief was that goods markets were largely oligo-
polis tic, that complex bargaining was taking place in the labor market, and
that credit rationing was a pervasive feature of credit markets. Neverthe-
less, descriptions of price and wage evolutions were mostly data deter-
mined, and theories were used rather casually to justify the presence of a
variable in an equation. For example, imperfect competition was used to
justify markup over standard cost in the "price equation." Labor markets
were in effect treated as a black box, with Phillips curve relations providing
an empirical, atheoretical "wage equation."
The second problem was that the basic model, and U.S. macroeconomic
thinking in general, was very much that of a closed economy. While
perhaps a decent approximation for the U.S. economy at that time, this
was already a serious problem in thinking about fluctuations in other
countries. More importantly, these models were very much "one-sector,"
"one-country" models, leaving U.S. macroeconomists with little training
in thinking about "two-sector/two-country models" and thus about such
issues as changes in the prices of imported raw materials, changes in compet-
itiveness, and the like, issues that were to dominate the macroeconomic
scene for the following 20 years.
2 1970-1990: Back to Basics
In 1970 the main claim to success of macroeconomics was success itself,

the ability to explain events and help policy through progressively more
sophisticated conceptual and empirical models. But in the early 1970s
macroeconomics, finding itself confronted with issues it had not thought
about, from supply shocks to stagflation, was having a tough empricial
time. The challenge posed by the new-classical attack, and their argument
that empirical failure stemmed directly from theoretical flaws, was thus a
tough one to counter. The challenge was met badly at first. Initial fights
were about the appropriateness of the assumption of rational expectations,
as the assumption seemed so damaging to mainstream macroeconomics.
But, by the late 1970s, regrouping had occurred, and progress happened in
two phases:
2.1 Integrating Supply Shocks and Rational Expectations
The first phase, in the late 1970s and early 1980s, was based on the
realization that supply shocks and rational expectations could be fruitfully
integrated in mainstream macroeconomics. The development of a richer

supply side and the analysis of the effects of changes in the price of oil were
major research projects of the late 1970s. And by the early 1980s, books
such as those by Bruno and Sachs showed how it could be done. How one
could reconcile rational expectations and slow adjustment of prices and
wages to changes in nominal demand was shown in contributions by
Fischer and Taylor. With the techniques required to solve models under
rational expectations now available, small and large models were expanded
to allow for a richer role of expectations in consumption, investment, and
portfolio choices, to study the implications of forward-looking prices in
asset markets and slowly adjusting prices and wages in other markets. And
empirical models with rational expectations were quickly developed and
used by Fair, Taylor, and others to forecast as well as to study policy
choices and implications.
2.2 Looking at the Structure of Markets
These developments were important. But they were the easy part. The
hard part was to understand the structure of markets, to understand how
prices and wages were set. This is where most of the work in U.S. main-
stream macroeconomics has taken place for the last ten years. Much prog-
ress has been made.
We have shown how small individual delays in changing prices or wages
can be individually rational yet lead to substantial aggregate price inertia
and lasting effects of demand on output. In the process, we have learned
about the trickiness of aggregation issues, about how, for example, indi-
vidual price inertia can disappear or instead be amplified in the process of
aggregation from individual prices to the price level. We have learned about
the effects of inflation on the distribution and the information content of
prices and clarified the allocation costs of inflation. I feel that this part of
the research agenda has largely been fulfilled and that decreasing returns
are setting in. But there is no longer any doubt that slow adjustment of
the price level could be given and has been given solid foundations.
We have explored alternative theories of wage setting, with the primary
purpose of explaining how fluctuations in labor demand lead to fluctua-
tions in unemployment rather than in wages. We have explored various
structures of individual and collective bargaining, asking what role the
~nemployed play in the process and how they affect the outcome. Follow-
ipg up on work that had started with the Phelps volume, we have explored
rhode Is of "efficiency wages," models in which firms use wages to elicit
effort, motivate workers, or recruit better workers. All these models can
explain the presence of unemployed workers who would rather be work-
ing at the prevailing wage. I do not believe that they yet provide very con-
vincing reasons why fluctuations in labor demand translate mostly in
employment variations rather than in wages. I return to this later.
We have explored alternative theories of imperfect competition and
price setting in the goods markets, with the aim of explaining why firms are
willing to satisfy changes in demand at roughly constant markups over
wages. We have examined the most natural explanation, that marginal cost
is fiat, and concluded it probably is not. Following again lines first drawn in
the Phelps volume, we have explored the idea that firms concerned about
customers may decide not to increase their prices in response to an increase
in demand. Using developments from industrial organization and game
theory, we have examined whether oligopolists are also likely to smooth
prices relative to marginal cost. On these issues, I believe, we still have a
long way to go; I also return to this later.
We have explored the role of asymmetric information, moral hazard,
and adverse selection in credit markets, with the purpose of understanding
the interactions among money, credit, and activity. We have looked at the
role of banks and made progress in thinking about credit rationing. We
have studied the role of internal finance and thus the role of current cash
flow in investment decisions. In other financial markets, Shiller nearly
single-handedly has forced us to think about deviations from arbitrage and
the efficient market hypothesis. This has led to work on how interactions of
speculators and noise traders in asset markets can generate large deviations
from fundamentals, which can in turn feed back on consumption and
investment decisions.
2.3 Why AllIs Not Well
The preceding description of progress sounds impressive, and to a large

extent it is. But all is not well. From closer, these developments look
more scattershot, a point made recently by Barro in his critique of "new-
Keynesian" macroeconomics. An analogy here may be to a team of
engineers asked to build a new car, spending their time designing many
new state-of-the-art catalytic converters, disk brake systems, and so on,
but never bothering to check which one works best and whether and how
the car will run. In labor markets, for example, a bewildering array of
theories, efficiency wages, and insider-outsiders search, coexist side by side,
with few systematic attempts to relate them or even to explore whether
they can explain the large set of cyclical facts that characterize labor
markets. The point goes beyond labor markets. We have developed a large
collection of theories or, more often, sketches of theories; we have not
explored whether and how they add up to a macro theory. This failure
can be traced to two causes. The first has been our "back to basics" mode.
The second has been our too frequent adoption of the wrong methodology.
Going "hack to basics" meant going away from complex issues and
returning to the study of the basic elements of our models. It also meant
keeping an open mind and exploring whatever new directions the models
we were developing to shore up foundations would lead us. Thus con-
frontation with broad sets of facts and integration were clearly not
at the top of the list of priorities. What was a healthy intellectual attitude at
the beginning has now become a disease. And part of the reason we never
proceeded to the next step-namely, systematic empirical explora-
tion and integration-is that along the way we got stuck with the wrong
methodology.
While we were rejecting the new-classical view of the world, we have too
often adopted one of its methodological ukases, namely, the quasi-religious
insistence on micro foundations and the derivation of results from "tastes
and technology." (This point echoes a theme developed by Larry Summers
a few years back, in which he denounced the dangers of "scientific illusion"
in macroeconomics.) In doing so, we have constructed too many monsters,
too many heavy models with few interesting results. But, worse than that,
those rules have de facto prevented both systematic empirical exploration
and integration. Exploring how a theory of the labor market can explain
not the time series behavior of one or two specific time series but rather the
broad set of facts about the cyclical behavior of the structure of wages, the
composition of unemployment, and so on, just cannot be done when one
plays by such rules: The taste and technology machine that would
be required would be monstrous, if it could be built at all. Exploring how
various theories of labor, goods, and asset markets combine into a macro
theory also requires shortcuts, the ability to capture the essence and ignore
the details of the various theories. Deciding what is of the essence is neither
easy nor uncontroversial, but this is precisely what shortcuts are about.
Absent the willingness to take them, systematic empirical explorations and
integration have simply not happened.
What we have seen instead of smart new Tobin-Mundell-like machines
has been an explosion of "fully articulated" one-explanation-models of
"the business cycle." The premium has too often been on intellectual
excitement per se, on epater les bourgeois rather than on relating a model
to the facts and to the existing body of knowledge. Think of the typical
NBER Economic Fluctuations conference, circa 1990, in which six papers,

all starting from scratch, would present six unrelated explanations of the
business cycle. Excesses have been made worse by the nature of what we
were exploring: Imperfect competition, increasing returns, externalities,
can all easily lead to multiple equilibria, limit cycles, and so on. It is all the
more crucial to confront the assumptions and implications of those models
with the data. Think of the many models of multiple equilibria that depend
crucially on a testable, but untested, nonlinearity in the data.
Lack of systematic exploration and of integration has not been the only
cost. Another has been an increasing separation between research and
actual macroeconomic developments and policy issues. By being in the
"back to basics" mode, the field has developed mostly according to its
own internal logic, and we have shunned external stimulus too much. Too
many of us have a knowledge of the economy reduced to a number
of correlations and statistical regularities. And, unfortunately, one of the
aspects that we have borrowed from the pre-1970 mode of thinking has
been a nearly complete ignorance of the open economy, sharply reducing
our field of vision.
For all these reasons-ignorance of the world and of the policy issues,
reluctance or inability to handle messy issues-we have missed on many
if not most of the important macro policy events, debates, and decisions
of the past decade. Events elsewhere in the world, or events with a clear
open economy aspect, the hyperinflations of Latin America, European
integration-what labor economists would call "natural experiments" and
examine eagerly-have been left to international macroeconomists, who
were not embarrassed at the idea of using variations of Mundell-Fleming,
Cagan, and the like. They have not done badly, but it is clearly a suicidal
strategy for U.S. macroeconomists to follow. Complex happenings, such as
the determinants of worldwide interest rates, prolonged unemployment in
Western Europe, the macroeconomics of transition in Eastern Europe,
which require pragmatic analysis, have also been left largely untouched by
U.S. macroeconomists. (Or, more precisely, in the case of Eastern Europe,
those macroeconomists who have been most involved are among those
who questioned the evolution of macro research in the 1980s and have
followed a more applied, policy-oriented, research agenda.) And even with
respect to the worldwide shift in macroeconomic policy away from activism
and toward simple rules, an evolution to which we would have had a lot to
contribute, our contributions have often been of little direct use. Our
debates about extreme forms of Ricardian equivalence, for example, have
provided little intellectual help to governments trying to chart a path of
debt and deficits or thinking about the best path for fiscal consolidation.
3 Where We Should Go from Here
Much of what I believe we should do is obvious from the preceding

discussion. The main order of business is, I believe, to identify both what
we have and have not explained and to see what macro theory has emerged
from our research. And to do so, we must adopt a more flexible, more prag-
matic appoach to research.
Let me sketch my preferred path, starting with my view of fluctuations
and moving on to a number of specific research topics.
1. We should get rid of another new-classical heritage, the focus
on "business cycles." "Business cycles" is too reductionist a view of what
short- and medium-term macroeconomics should be about. Most of the
time, economies are indeed subject to regular fluctuations. Understanding
those, and thinking about the role of policy in that context, is obviously
desirable. But there are too many other "irregular" fluctuations, from
which we can learn much and for which the issue of whether and how
policy can help is often of crucial political and social importance. The list
is a long one, from those without an obvious trigger, such as the Great
Depression, the long period of high unemployment in Europe, and the
inability of Latin American countries to stabilize, to those with clear
triggers, such as the integration of Germany and the stabilization and
restructuring process in Eastern Europe. Focusing on "business cycles" has
led too many of us to think of those as freak events and to leave them aside.
We should go back to using "fluctuations": Words do matter.
2. The view of fluctuations pre-1975 was one of dominant demand
shocks and weak stabilizing mechanisms. This view is still fundamentally
right. The deep u.K. and U.s. recessions of the early 1980s are testimony
to the real effects of monetary contraction. But the nature of the shocks has
changed and is likely to keep changing in the future. And so have changed
and will change the relevant stabilizing mechanisms.
The traditional view emphasized the importance of shifts in domestic
demand, such as shifts in the propensity to save coming from changes
in consumer confidence. Of increasing importance, however, are shifts in
the demand for domestic goods. As countries become more open and
integrated, they also become more specialized, and there is an increas-
ing discrepancy between the basket of goods they produce and the basket
they consume. The relative importance of the two types of shocks ob-
viously varies from country to country, depending on their degree of
specialization. U.S. states, European countries, and the United States
nicely cover the range. U.S. states produce and consume very different
baskets of goods. In the medium run, the evolution of their employment is
dominated by the relative demand for the particular basket of goods they
produce. European countries are, for the moment, much less specialized
than U.S. regions, but Europe 1992 is likely to lead to higher specialization.
The United States as an economy is still much less specialized, so that
shocks to domestic demand still dominate. But even the United States
is becoming increasingly specialized. Like most OECD countries, it is
specializing in products using skilled labor while consuming products which
we both skilled and unskilled labor.
The natures of these two shocks are quite different. While consumer
confidence, for example, can swiftly reverse direction, consumers who shift
their demand from one product to another are unlikely to make a quick
turnaround. And the adjustment mechanisms are also quite different. With
respect to shifts in aggregate demand, the basic stabilizing mechanism
involves the effects of changes in employment on nominal prices and wages
and in turn on real money balances. Of central importance in the process
are "nominal rigidities," to which we have devoted so much work. But for
employment to return to normal in the face of shifts in relative demand,
relative prices, and thus real consumption wages must adjust, or over a
longer horizon, factors must move, or the basket of goods produced must
change. "Real rigidities," in particular the speed and strength of the effect
of unemployment on real wages, are central to that process; even absent
nominal rigidities, changes in relative demand can generate sustained
fluctuations in unemployment. And, indeed, in a number of European
countries, nominal rigidities appear to play only a small role in generating
fluctuations in unemployment. In a rather ironic twist, what we are observ-
ing are increasingly "real business cycles," or, following my semantic edict,
"real fluctuations."
3. One aspect of those fluctuations is the set of interactions among
recessions, technological change, and unemployment. I am struck at how
these issues keep coming up and how primitive our thinking about them is.
Do recessions trigger or hinder technological progress and the introduction
of new products? Do they leave permanent scars on the labor force? Or do
they instead lead to cleansing, to the elimination of weak firms, to the elim-
ination of fat in others? From the work on the apparent disenfranchising of
the long-term unemployed in Europe to the work on the cyclical behavior
of job creation and destruction in the United States, we have tantalizing
hints of the presence of both sets of effects. Will the politics of austerity
followed by many governments in Western Europe make or break their
economies, will they lead to rationalization and improved competitiveness,
or will they lead instead to inefficient bankruptcies and prolonged unem-
ployment? Closely related issues are also at the center of discussions in
130 lHE BUSINESS CYCLE
Eastern Europe. How much unemployment will the transition to market

economies involve, and how much unemployment should those countries
accept? It is clear that, on these issues, help will come from the resurgence
of growth theory, from the work on trade and growth, from the models
that formalize the introduction of new products, the process of creation
and destruction. They are too sketchy yet to be of direct use, but they will
eventually help.
4. Within that general framework, we should go back to our work on
labor, goods, and credit markets; take stock of what we have and have not
explained; and see how the pieces may fit. Having not worked on credit
markets, I shall limit myself here to labor and goods markets. In so doing, I
shall develop themes I touched on earlier.
First, in labor markets, we should go beyond sketches. We have a large
number of good sketches, all of them probably relevant, and all with the
basic implication that the wage exceeds the marginal utility of not working
and that the wedge depends on the state of the labor market. But we have
not explored seriously whether those theories can explain the observed
dynamic relation between wages and unemployment, in unusual times as in
Europe in the 1980s, but even in more normal times as in the United States
during the same decade. As a result, our empirical specifications have
remained embryonic. We also have not explored the theories in sufficient
detail to explore their more micro implications, for the distribution of
unemployment, the dynamics of unemployment duration, the cyclical
movement of relative wages, and the like. On those issues, there is a large
body of work by labor economists on both sides of the Atlantic that is
highly relevant and has largely gone unexploited by U.S. macro-
economists. The returns to making contact with those findings are high.
In goods markets, the problem is not one of excess of good theories.
More than 20 years after the Nordhaus survey and its conclusion that
we did not have a clue about pricing behavior, we are still not very
far from ground zero, theoretically and empirically. The basic fact to
be explained there is why firms, which appear to face upward-sloping mar-
ginal cost, reduce their markup as demand expands. The two main fully
articulated theories we have-customer markets and supergames between
oligopolists-have a strong air of implausibility as general explanations of
price behavior. And the price setting behavior of firms during regular
fluctuations is not the only thing we do not understand. There has been a
very large increase in profit shares in many OEeD countries since the mid-
1980s. We have few clues as to why. Goods markets may be the exception
to my earlier proposition that what we do not need are new theories.
Alternative theories, more directly aimed at explaining cross-sectional
differences and similarities in behavior among types of firms, across

sectors, and across countries, are very much needed here.
5. I have emphasized work on the functioning of markets because this is
where we must capitalize on the work of the past decade. But the pieces of
the old agenda-consumption, investment, the transmission mechanisms of
monetary policy, and fiscal policy-are also ready for a thorough going
over. An encyclopedic review would be boring. Let me pick a few topics
I believe to have high social and private returns. Some are primarily em-
pirical, requiring either panel or aggregate data. Others are primarily
synthetic, requiring the development of simple analytical machines, a new
generation of Mundell-Tobin models.
We have worked hard on consumption. But we still have a very poor
idea of how much foresight people use in making consumption/saving
decisions. Separating decisions about retirement income from others,
taking into account the increasing importance of pension plans in allowing
people to in effect delegate their retirement saving decisions, may be a
productive way to go. On investment, we have shown that aggregate time
series for investment do not satisfy the tight constraints imposed by any of
our standard theories, but, in the end, the rejections have not been very
informative. Despite useful work using panel data, we still do not know
how much of the strong relation between aggregate profit and investment
comes from profit as an internal source of funds rather than as a signal of
future profitability. We also suspect that some of the variations in invest-
ment spending come from changes in uncertainty and the changing option
value of waiting to invest. But we have little idea of the importance of
those effects. Our belief that user costs affect investment is still very much
based on faith rather than on hard evidence. All these issues should be high
priority items. Recent theoretical work on irreversibility and aggregation
may prove of much value here.
The transmission mechanism of monetary policy, as summarized by
Modigliani, had many parts. We have worked on many of those, from the
behavior of financial intermediaries, to the role of internal finance and
bank credit, to the behavior of the term structure of interest rates. It is time
to see how the parts fit. Many of the pieces we have worked on, from price
setting to the behavior of financial intermediaries, suggest asymmetric
effects of monetary expansions and recessions. This is a topic just ripe for
new research. Finally, while we have become much more aware of some of
the medium-term costs of deficit finance, we have not taken the analysis to
the point of deciding whether, when, and how much governments should
rely on deficit finance.
6. Let me end on an optimistic note. Mainstream macroeconomics, circa
1970, was not perfect. But it had solid foundations and was basically right.
Because of that, it has survived the new-classical challenge. Much progress
has been made; much of it, however, is still latent. Realizing that progress
will require more systematic confrontation with the data and more system-
atic integration. This task is far from impossible. But it will require a return
to more pragmatic, data-oriented research than we have relied on for the
past decade.
Acknowledgments
I thank Larry Ball, Robert Barra, Roland Benabou, Ricardo Caballero,

Peter Diamond, Stanley Fischer, Bob Gordon, Herschel Grossman, Chad
Jones, Anil Kashyap, Bob Solow, Philippe Weil, and especially Rudiger
Dornbusch and Julio Rotemberg, for comments, suggestions, and
objections.
4
THE COWLES COMMISSION
APPROACH, REAL BUSINESS CYCLE
THEORIES, AND NEW-KEYNESIAN
ECONOMICS
Ray C. Fair
I have been given the daunting task of discussing how the debate among
the various schools of business cycle theorists might be resolved. People
obviously differ in how they think the macroeconomy works. Will there
ever be a winner? I am optimistic enough to think so, although I view the
last two decades as making only modest progress in this direction. One
problem is that there is too little testing of alternative models. There has
been no systematic attempt to find the model that best approximates the
macroeconomy. As disturbing, however, is the fact that macroeconomic
research appears to be moving away from its traditional empirical em-
phasis. I will elaborate on both points in this chapter.
From Tinbergen's (1939) model building in the late 1930s through the
1960s, the dominant methodology of macroeconomics was what I will call
the "Cowles Commission" approach.l Structural econometric models were
specified, estimated, and then analyzed and tested in various ways. One of
the major macroeconometric efforts of the 1960s, building on the earlier
work of Klein (1950) and Klein and Goldberger (1955), was the Brookings
model (Duesenberry et aI., 1965, 1969). This model was a joint effort
of many individuals, peaking at nearly 400 equations. Although much was
133
learned from this exercise, the model never achieved the success initially
expected and was laid to rest around 1972.
Two important events in the 1970s contributed to the decline in popu-
larity of the Cowles Commission approach. pte first was the commer-
cialization of macroeconometric models. 2 'J1his changed the focus of
research on the models. Basic research gave wj1y to the day-to-day needs of
updating the models, of subjectively adjustiq,g the forecasts to make them
"reasonable," and of meeting the special ineeds of clients. The second
event was Lucas's (1976) critique, which argued that the models are not
likely to be useful for policy purposes.
The Lucas critique led to a line of research that has cumulated in the
real business cycle (RBC) theories. This in tum has generated a counter
response in the form of new-Keynesian economics. I will argue that neither
the RBC approach nor new-Keynesian economics is in the spirit of the
Cowles Commission approach and that this is a step backward. The Cowles
Commission approach is discussed in section 1; the RBC approach, in
section 2; and new-Keynesian economics, in section 3. Suggestions for the
future are then presented.
1 The Cowles Commission Approach
1. 1 Specification
Some of the early macroeconometric models were linear, but this soon
gave way to the specification of nonlinear models. Consequently, only the
nonlinear case will be considered here. The model will be written as
(i=1, ... ,n) (t = 1, ... , T), (1)
where Yt is an n-dimensional vector of endogenous variables, X t is a vector
of predetermined variables (including lagged endogenous variables), (Xi is a
vector of unknown coefficients, and Uit is the error term for equation i for
period t. For equations that are identities, Uit is identically zero for all t.
Specification consists of choosing (1) the variables that appear in each
equation with nonzero coefficients, (2) the functional form of each
equation, and (3) the probability structure for Uit. (In modem times one has
to make sufficient stationarity assumptions about the variables to make
the time series econometricians happy. The assumption, either explicit or
implicit, of most macro econometric model building work is that the vari-
ables are trend stationary.) Economic theory is used to guide the choice
of variables. In most cases there is an obvious left-side variable for the
THE COWLES COMMISSION APPROACH 135
equation, where the normalization used is to set the coefficient of this vari-
able equal to one. This is the variable considered to be "explained" by the
equation.
It will be useful to consider an example of how theory is used to specify
an equation. Consider the following maximization problem for a represen-
tative household. Maximize
(2)
subject to
St = W t (H - L t) + rtA t- l - PtCt (3)
At = At-l + St
AT=A.,
where C is consumption, L is leisure, S is saving, W is the wage rate, H is the
total number of hours in the period, r is the one-period interest rate, A is
the level of assets, P is the price level, A is the terminal value of assets, and
t = 1, ... ,T. Eo is the expectations operator conditional on information
available through time o. Given Ao and the conditional distributions
of the future values of W, P, and r, it is possible in principle to solve for
the optimal values of C and L for period 1, denoted Ci and Li- In
general, however, this problem is not analytically tractable. In other words,
it is not generally possible to find analytic expressions for Ci and Li.
The approach that I am calling the Cowles Commission approach can be
thought of as specifying and estimating approximations of the decision
equations. In the context of the present example, this approach is as
follows. First, the random variables, WI' Pt , and 't, t =1, ... , T, are replaced
by their expected values, EoWt, EoPt , and Eorl' t = 1, ... , T. Given this
replacement, one can write the expressions for Ci and Li as
Ci= gl(Ao,A., EOW1,·· ., EoWT' EoP1> . .. , EoP T, E or1> . .. ,Earn~) (4)
Ll* = gz(Ao,A,
- EoWj, . .. , EOWT' EoP!> .. . ,EoP!> E O'1> . .. , EorT' ~), (5)
where f3 is the vector of parameters of the utility function. Equations (4)
and (5) simply state that the optimal values for the first period are a func-
tion of (1) the initial and terminal values of assets; (2) the expected future
values of the wage rate, the price level, and the interest rate; and (3) the
parameters of the utility function. 3 The functional forms of equations (4)
and (5) are not in general known. The aim of the empirical work is to try
to estimate equations that are approximations of equations (4) and (5).
Experimentation consists in trying different functional forms and in trying
different assumptions about how expectations are formed. Because of the
large number of expected values in equations (4) and (5), the expectational
assumptions usually restrict the number of free parameters to be estimated.
For example, the parameters for EOWb . .. ,EOWT might be assumed to lie
on a low-order polynomial or to be geometrically declining. The error
terms are usually assumed to be additive, as specified in equation (1), and
they can be interpreted as approximation errors.
It is often the case when equations like (4) and (5) are estimated that
lagged dependent variables are used as explanatory variables. Since Co and
Lo do not appear in equations (4) and (5), how can one justify the use of
lagged dependent variables? A common procedure is to assume that C! in
equation (4) and Li in equation (5) are long-run "desired" values. It is then
assumed that because of adjustment costs, there is only a partial adjust-
ment of actual to desired values. The usual adjustment equation for
consumption would be
0< A < 1, (6)
which adds Co to the estimated equation. This procedure is ad hoc in the
sense that the adjustment equation is not explicitly derived from utility
maximization. One can, however, assume that there are utility costs to
large changes in consumption and leisure and thus put terms like (Cl - CO)2,
(C2 - C l )2, (Ll - Lo)2, (L 2 - Ll)2, ... in the utility function, equation (2).
This would add the variables Co and Lo to the right-hand side of equations
(4) and (5), which would justify the use of lagged dependent variables in
the empirical approximating equations for (4) and (5).
This setup can handle the assumption of rational expectations in the
following sense. Let E 1_l Y21+l denote the expected value of Y21+1o where the
expectation is based on information through period t - 1, and assume that
Et-lY2I+l appears as an explanatory variable in equation (1). (Equation 1
might be the equation explaining consumption, and Y2 might be the wage
rate.) If expectations are assumed to be rational, equation (1) can be
estimated by either a limited information or a full information technique.
In the limited information case E t- 1Y21+1 is replaced by Y21+b and the
equation is estimated by Hansen's (1982) generalized method of moments
(GMM) procedure. In the full information case the entire model is
estimated at the same time by full information maximum likelihood, where
the restriction is imposed that the expectations of future values of variables
are equal to the model's predictions of the future values. 4 Again, the
parameters of the expected future values might be restricted to lessen the
number of free parameters to be estimated.
The specification just outlined does not allow the estimation of "deep
structural parameters," such as the parameters of utility functions, even
under the assumption of rational expectations. Only approximations of the

decision equations are being estimated. The specification is thus subject
to the Lucas (1976) critique. More will be said about this later. The
specification also uses the certainty equivalence procedure, which is strictly
valid only in the linear-quadratic setup.
1.2 Estimation
A typical macroeconometric model is dynamic, nonlinear, simultaneous,

and has error terms that may be correlated across equations and with their
lagged values. A number of techniques have been developed for the esti-
mation of such models. Techniques that do not consider the correlation of
the error terms across equations (limited information techniques) include
limited information maximum likelihood and two-stage least squares.
Techniques that do account for this correlation (full information tech-
niques) include full information maximum likelihood and three-stage least
squares. It is straightforward to modify these techniques to handle the case
in which the error terms follow autoregressive processes. Also, as already
noted, the techniques can be modified to handle the assumption of rational
expectations.
Computational advances, both in hardware and software, have made the
application of these techniques fairly routine. Even full information maxi-
mum likelihood when expectations are assumed to be rational appears
computationally feasible for most models.
1.3 Testing
Testing has always played a major role in applied econometrics. When

an equation is estimated, one examines how well it fits the data, if its
coefficient estimates are significant and of the expected sign, if the proper-
ties of the estimated residuals are as expected, and so on. Equations are
discarded or modified if they do not seem to approximate very well the
process that generated the data.
Complete models can also be tested, but here things are more compli-
cated. Given (1) a set of coefficient estimates, (2) values of the exogenous
variables, (3) values of the error terms, and (4) lagged values of the
endogenous variables, a model can be solved for the endogenous variables.
If the solution (simulation) is "static," the actual values of the lagged
endogenous variables are used for each period solved; if the solution is
"dynamic," the values of the lagged endogenous variables are taken to be

the predicted values of the endogenous variables from the previous
periods. If one set of values of the error terms is used, the simulation is said
to be "deterministic." The expected values of the error terms are usually
assumed to be zero, and so in most cases the error terms are set to zero for
the solution. A "stochastic" simulation is one in which (1) the error terms
are drawn from estimated distributions, (2) the model is solved for each set
of draws, and (3) the predicted value of each endogenous variable is taken
to be the average of the solution values across the sets of draws.
A standard procedure for evaluating how well a model fits the data is to
solve the model by performing a dynamic, deterministic simulation and
then compare the predicted values of the endogenous variables with the
actual values using the root mean squared error (RMSE) criterion. Other
criteria include mean absolute error and Theil's inequality coefficient. If
two models are being compared and model A has lower RMSEs for most
of the variables than model B, this is evidence in favor of model A over
modelB.
There is always a danger in this business of "data mining," which means
specifying and estimating different versions of a model until a good fit has
been achieved (say in terms of the RMSE criterion). The danger with this
type of searching is that one finds a model that fits well within the esti-
mation period but is in fact a poor approximation of the economy. To
guard against this, predictions are many times taken to be outside of the
estimation period. If a model is poorly specified, if should not predict well
outside the period for which it was estimated, even though it may fit well
within the period. 5
One problem with the RMSE criterion is that it does not take account
of the fact that forecast-error variances vary across time. Forecast-error
variances vary across time because of nonlinearities in the model and
because of variation in the exogenous variables. Although RMSEs are in
some loose sense estimates of the averages of the variances across time,
no rigorous statistical interpretation can be placed on them: They are not
estimates of any parameters of a model.
A more serious problem with the RMSE criterion as a means of
comparing models is that models may be based on different sets of exogen-
ous variables. If one model takes investment as exogenous and a second
does not, the first model has an unfair advantage when computing RMSEs.
I have developed a method, based on stochastic simulation, that
accounts for the RMSE difficulties (Fair, 1980). The method accounts
for the four main sources of uncertainty of a forecast from a model:
uncertainty due to (1) the error terms, (2) the coefficient estimates, (3) the
exogenous variables, and (4) the possible misspecification of the model.

The forecast-error variance for each variable and each period estimated
by the method accounts for all four sources of uncertainty, and so it can be
compared across models. The estimated variances from different structural
models can be compared, or the estimated variances from one structural
model can be compared to those from an autoregressive or vector
autoregressive model. If a particular model's estimated variances are
in general smaller than estimated variances from other models, this is
evidence in favor of the particular model.
A by-product of the method is an estimate of the degree of mis-
specification of a model for each endogenous variable. Any model is likely
to be somewhat misspecified, and the method can estimate the quantitative
importance of the misspecification.
The method can handle a variety of assumptions about exogenous-
variable uncertainty. One polar assumption is that there is no uncertainty
attached to the exogenous variables. This might be true, for example, of
some policy variables. The other polar assumption is that the exogenous
variables are in some sense as uncertain as the endogenous variables. One
can, for example, estimate autoregressive equations for each exogenous
variable and add these equations to the model. This would produce
a model with no exogenous variables, which could then be tested. An
in-between case would include estimating the variance of an exogenous-
variable forecast error from actual forecasting errors made by a forecast-
ing service-say the errors made by DRI in forecasting defense spending.
Another method comparing models is to regress the actual value of an
endogenous variable on a constant and forecasts of the variable from two
or more models. This procedure is explained in Fair and Shiller (1990) and
is related to the literature on encompassing tests-see, for example,
Davidson and MacKinnon (1981), Hendry and Richard (1982), and Chong
and Hendry (1986). Again, one can use autoregressive or vector auto-
regressive models as comparisons for structural models using these tests.
Testing models in the ways described here seems clearly in the spirit of
the Cowles Commission approach. A model to the Cowles Commission
was a null hypothesis to be tested.
1.4 A Digression on Monetarism
Laidler (1992) has written an interesting and useful chapter on the history
of monetarism. From the perspective of the Cowles Commission approach,
I have no complaints about this chapter. The Laidler and Bentley (1983)
model, for example, is a standard macroeconometric model and can be

tested in the ways already discussed. In fact, in Laidler's last sentence he
states that he hopes to "reinstate empirical evidence as a factor more
important than a priori principles" in the debate about business cycles. I
would interpret this as an argument for the Cowles Commission approach.
This is not to say, however, that I would argue in favor of the Laidler
and Bentley model. Although in simple tests the model does about the
same as Barro's (1978) model, I doubt that it would hold up well against
larger structural models or even against autoregressive and vector auto-
regressive models. But the main point is that the model can be tested. The
debate is simply between which macroeconometric model-a model in the
monetarist spirit or some other model-best explains the data.
2 The Real Business Cycle Approach
In discussing the RBC approach, it will be useful to begin with the utility
maximization model already considered. The RBC approach to this model
would be to specify a particular functional form for the utility function
in equation (2). The parameters of this function would then be either
estimated or simply chosen ("calibrated") to be in line with parameters
estimated in the literature.
Although there is some parameter estimation in the RBC literature,
most of the studies calibrate rather than estimate, in the spirit of the semi-
nal article by Kydland and Prescott (1982). If the parameters are estimated,
they are estimated from the first-order conditions. A recent example is
Christiano and Eichenbaum (1990), where the parameters of their model
are estimated using Hansen's (1982) GMM procedure. Altug (1989)
estimates the parameters of her model using a likelihood procedure. Chow
(1991) and Canova, Finn, and Pagan (1991) contain interesting discussions
of the estimation of RBC models. There is also a slightly earlier literature
in which the parameters of a utility function, as in equation (2), were
estimated from the first-order conditions-see, for example, Hall (1978),
Hansen and Singleton (1982), and Mankiw, Rotemberg, and Summers
(1985).
The RBC approach meets the Lucas critique; deep structural para-
meters are being estimated (or calibrated). It is hard to overestimate the
appeal this has to many people. Anyone who doubts this appeal should
read Lucas's 1985 lahnsson lectures (Lucas, 1987), which is an elegant
argument for dynamic economic theory. The tone of these lectures is an
exciting sense of progress in macroeconomics and hope that in the end
there will be essentially no distinction between microeconomics and macro-

economics. There will simply be economic theory applied to different
problems.
Once the coefficients are chosen, by whatever means, the overall model
can be solved. In the earlier example, one could solve the utility maxi-
mization problem for the optimal consumption and leisure paths. The
properties of the computed paths of the decision variables are then
compared to the properties of the actual paths of the variables. If the
computed paths have similar properties to the actual paths (for example,
similar variances, covariances, and autocovariances), this is judged to be a
positive sign for the model. If the parameters are chosen by calibration,
there is usually some searching over parameters to find that set that gives
good results in matching the computed paths to the actual paths in terms of
the particular criterion used. In this sense the calibrated parameters are
also estimated.
Is the RBC approach a good way of testing models? At first glance it
might seem so, since computed paths are being compared to actual paths.
But the paths are being compared in a very limited way in contrast to the
way that the Cowles Commission approach would compare them. Take the
simple RMSE procedure. This procedure would compute a prediction
error for a given variable for each period and then calculate the RMSE
from these prediction errors. This RMSE might then be compared to the
RMSE from another structural model or from an autoregressive or vector
autoregressive model.
I have never seen this type of comparison done for a RBC model. How
would, say, the currently best-fitting RBC model compare to a simple first-
order autoregressive equation for real GNP in terms of the RMSE
criterion? My guess is very poorly. Having the computed path mimic the
actual path for a few selected moments is a far cry from beating even a first-
order autoregressive equation (let alone a structural model) in terms
of fitting the observations well according to the RMSE criterion. The
disturbing feature of the RBC literature is there seems to be no interest in
computing RMSEs and the like. People generally seem to realize that the
RBC models do not fit well in this sense, but they proceed anyway.
If this literature proceeds anyway, it has in my view dropped out of the
race. The literature may take a long time to play itself out, but it will
eventually reach a dead end unless it comes around to developing models
that can compete with other models in explaining the economy observation
by observation.
One of the main reasons that individuals proceed anyway is undoubt-
edly the Lucas critique and the general excitement about deep structural
parameters. Why waste one's time in working with models whose co-
efficients change over time as policy rules and other things change? The
logic of the Lucas critique is certainly correct, but the key question for
empirical work is the quantitative importance of this critique. Even the
best econometric model is only an approximation of how the economy
works. Another potential source of coefficient change is the use of aggre-
gate data. As the age and income distributions of the population change,
the coefficients in aggregate equations are likely to change, and this is
a source of error in the estimated equations. This problem may be
quantitatively much more important than the problem raised by Lucas.
Put another way, the representative agent model that is used so much in
macroeconomics has serious problems of its own, which may completely
swamp the problem of coefficients changing when policy rules change. The
RBC literature has focused so much on solving one problem that it is likely
in the process to have exacerbated the effects of a number of others.
There are a number of reasons the RBC models probably do not fit the
data well. The RBC approach requires that a particular functional form
for, say, the utility function be chosen, and errors made in this choice may
lead to large prediction errors. Remember that this function represents the
average of the utility functions of all the households in the economy, and
it is unlikely that one is going to get this quite right. The advantage of
estimating approximations of the decision equations, as discussed in section
1, is that it allows more flexibility in estimating functional forms. The data
are allowed more play, if you will. Using the approach of estimating approx-
imations of decision equations, one trades off estimating deep structural
parameters for less sensitivity to functional-form errors and the like.
When deep structural parameters have been estimated from the first-
order conditions, the results have not always been very good even when
judged by themselves. The results in Mankiw, Rotemberg, and Summers
(1985) for the utility parameters are not supportive of the approach. In a
completely different literature-the estimation of production-smoothing
equations-Krane and Braun (1989), whose study is based on quite good
data, report that their attempts to estimate first-order conditions was
unsuccessfuL It may be that one is asking too much of the aggregate data to
force them into estimating what one thinks are parameters from some
postulated function.
Finally, one encouraging feature regarding the Lucas critique is that
it can be tested. Assume that for an equation or set of equations the
parameters change considerably when a given policy variable changes.
Assume also that the policy variable changes frequently. In this case the
model is obviously misspecified, and so methods like that discussed in
section 1 should be able to pick up this misspecification if the policy vari-

able has changed frequently. If the policy variable has not changed or
changed very little, the model will be misspecified but the misspecification
will not be given a chance to be picked up in the data. But otherwise,
models that suffer in an important way from the Lucas critique ought to be
weeded out by various tests.
3 The New-Keynesian Economics
After reading or rereading a number of new-Keynesian articles for this

chapter, I came away feeling uneasy. It's like coming out of a play that
many of your friends liked and feeling that you did not really like it, but not
knowing quite why. Given my views of how the economy works, many of
the results of the new-Keynesian literature seem reasonable, but something
seemed missing. One problem is that it is hard to get a big picture. There
are many small stories, and it's hard to remember each one. In addition,
many of the conclusions do not seem robust to small changes in the models.
On further reflection, however, I do not think this was my main source
of uneasiness. The main problem is that this literature is not really em-
pirical in the Cowles Commission sense. This literature has moved macro-
economics away from its econometric base. Consider, for example, the
articles in the two volumes of New Keynesian Economics, edited by
Mankiw and Romer (1991). By my count, of the 34 chapters in these two
volumes, only 8 have anything to do with data. 6 Of these 8, one (Carlton,
"The Rigidity of Prices") is more industrial organization than macro and
one (Krueger and Summers, "Efficiency Wages and the Interindustry
Wage Structure") is more labor than macro. These two studies provide
some interesting insights that might be of help to macroeconomists, but
they are not really empirical macroeconomics.
It has been pointed out to me 7 that the Mankiw and Romer volumes
may be biased against empirical papers because of space constraints
imposed by the publisher. Nevertheless, it seems clear that there is very
little in the new-Keynesian literature in the nature of structural modeling
of the kind outlined in section 1. As in the RBC literature, one does not
see, say, predictions of real GNP from some new-Keynesian model
compared to predictions of real GNP from an autoregressive equation
using a criterion like RMSE. But here one does not see it because no
econometric models of real GNP are constructred! So this literature has
dropped out of the race not because it is necessarily uninterested in serious
tests but because it is uninterested in constructing econometric models.
I should hasten to add that I do not mean by these criticisms that there is
no interesting empirical work going on in macroeconomics. For example,
the literature on production smoothing, which is largely empirical, has
produced some important results and insights. It is simply that literature of
this type is not generally classified as new-Keynesian. Even if one wanted
to be generous and put some of this empirical work in the new-Keynesian
literature, it is surely not the essence of new-Keynesian economics.
One might argue that new-Keynesian economics is just getting started
and that the big picture (model) will eventually emerge to rival existing
models of the economy. This is probably an excessively generous inter-
pretation, given the focus of this literature on small theoretical models,
but unless the literature does move in a more econometric and larger-
model direction, it is not likely to have much long-run impact.
4 Looking Ahead
So I see the RBC and new-Keynesian literatures passing each other like
two runners in the night, both having left the original path laid out by the
Cowles Commission and its predecessors. To answer the question posed to
me at the conference, I see no way to resolve the debates between these
two literatures. The RBC literature is only interested in testing in a very
limited way, and the new-Keynesian literature is not econometric enough
to even talk about serious testing.
But I argue there is hope. Models can be tested, and there are
procedures for weeding out inferior models. The RBC literature should
entertain the possibility of testing models based on estimating deep struc-
tural parameters against models based on estimating approximations
of decision equations. Also, the tests should be more than just observing
whether a computed path mimics the actual path in a few ways. The new-
Keynesian literature should entertain the possibility of putting its various
ideas together to specify, estimate, and test structural macroeconometric
models.
Finally, both literatures ought to consider bigger models. I have always
thought it ironic that one of the consequences of the Lucas critique was to
narrow the number of endogenous variables in a model from many (say a
hundred or more) to generally no more than three or four. If one is worried
about coefficients in structural equations changing, it seems unlikely that
getting rid of the structural detail in large-scale models is going to get one
closer to deep structural parameters.
Acknowledgments
I am indebted to my discussant, Arnold Zellner, and other participants at

the conference for many helpful comments.
Notes
1. See Arrow (1991) and Malinvaud (1991) for interesting historical discussions of eco-
nometric research at the Cowles Commission (later Cowles Foundation) and its antecedents.
2. It should be noted that the commercialization of models has been less of a problem in
the United Kingdom than in the United States. In 1983 the Macroeconomic Modeling Bureau
of the Economic and Social Research Council was established at the University of Warwick
under the direction of Kenneth F. Wallis. Various u.K. models and their associated databases
are made available to academic researchers through the bureau.
3. If information for period 1 is available at the time the decisions are made, then EoW!,
EaPl' and EOrl should be replaced by the actual values in equations (4) and (5).
4. Sec Fair and Taylor (1983) for a description of this procedure. This procedure is based
on the assumption of certainty equivalence, which is only an approximation for nonlinear
models.
5. This is assuming that one does not search by (1) estimating a model up to a certain
point, (2) solving the model for a period beyond this point, and (3) choosing the version that
best fits the period beyond the point. If this were done, then one would have to wait for more
observations to provide a good test of the model. Even if this type of searching is not formally
done, it may be that information beyond the estimation period has been implicitly used in
specifying a model. This might then lead to a better-fitting model beyond the estimation
period than is warranted. In this case, one would also have to wait for more observations to
see how accurate the model in fact is.
6. One might argue nine. Okun's article, "Inflation: Its Mechanics and Welfare Costs,"
which I did not count in the eight. presents and briefly discusses data in one figure.
7. By Olivier Blanchard.
References
Altug, Sumru 1989. "Time-to-Build and Aggregate Fluctuations: Some New

Evidence." International Economic Review 30: 889-920.
Arrow, Kenneth 1.1991. "Cowles in the History of Economic Thought." In Cowles
Fiftieth Anniversary. New Haven, Conn.: Yale University.
Barro, Robert 1. 1978. "Unanticipated Money, Output, and the Price Level in the
United States." Journal of Political Economy 86: 549-581.
Canova, F., M. Finn, and A.R. Pagan. 1991. "Econometric Issues in the Analysis of
Equilibrium Models." Paper given at the Canadian Econometric Study Group
meeting, Quebec City, September.
Chong, Yock Y., and David F. Hendry. 1986. "Econometric Evauation of Linear
Macro-Econometric Models." Review of Economic Studies 53: 671-690.
Chow, Gregory C. 1991. "Statistical Estimation and Testing of a Real Business

Cycle Model" August 30, mimeo.
Christiano, Lawrence J., and Martin Eichenbaum, 1990. "Current Real Business
Cycle Theories and Aggregate Labor Market Fluctuations." Discussion Paper
24, Institute for Empirical Macroeconomics, January.
Davidson, Russell, and James G. MacKinnon. 1981. "Several Tests of Model
Specification in the Presence of Alternative Hypotheses." Econometrica 40:
781-793.
Duesenberry, James S., Gary Fromm, Lawrence R. Klein, and Edwin Kuh, eds.
1965. The Brookings Quarterly Econometric Model of the United States.
Chicago: Rand McNally & Company.
- - - . 1969. The Brookings Model: Some Further Results. Chicago: Rand McNally
& Company.
Fair, Ray C. 1980. "Estimating the Expected Predictive Accuracy of Econometric
Models." International Economic Review 21: 355-378.
Fair, Ray c., and Robert J. Shiller. 1990. "Comparing Information in Forecasts
from Econometric Models." American Economic Review 80: 375-389.
Fair, Ray c., and John Taylor. 1983. "Solution and Maximum Likelihood Esti-
mation of Dynamic Rational Expectations Models." Econometrica 51: 1169-
1185.
Hall, Robert E. 1978. "Stochastic Implications of the Life-Cycle Permanent Income
Hypothesis: Theory and Evidence." Journal of Political Economy 86: 971-989.
Hansen, Lars Peter. 1982. "Large Sample Properties of Generalized Method of
Moments Estimators." Econometrica 50: 1029-1054.
Hansen, Lars Peter, and Kenneth Singleton. 1982. "Generalized Instrumental
Variables Estimation of Nonlinear Rational Expectations Models." Econome-
trica 50: 1269-1286.
Hendry, David F., and Jean-Francois Richard. 1982. "On the Formulation of
Empirical Models in Dynamic Economics." Journal of Econometrics 20: 3-33.
Klein, Lawrence R. 1950. Economic Fluctuations in the United States, 1921-1941.
New York: Wiley, Cowles Monograph No. II.
Klein, Lawrence R., and Arthur S. Goldberger. 1955. An Econometric Model of the
United States 1929-1952. Amsterdam: North-Holland.
Krane, Spencer D., and Steven N. Braun. 1989. "Production Smoothing Evidence
from Physical-Product Data," mimeo.
Kydland, F .E., and E.C. Prescott. 1982. "Time to Build and Aggregate Fluctuations."
Laidler, David. 1992. "The Cycle Before New-Classical Economics." In M.T.
Belongia and M.R. Garfinkel, eds., The Business Cycle: Theories and Evidence.
Boston: Kluwer Academic Publishers.
Laidler, David, and B. Bentley. 1983. "A Small Macro-Model of the Post-War
United States." Manchester School 51: 317-340.
Lucas, Robert E., Jr. 1976. "Econometric Policy Evaluation: A Critique." In
K. Brunner and A.H. Meltzer, eds., The Phillips Curve and Labor Markets.
Amsterdam: North-Holland.
- - - . 1987. Models of Business Cycles. New York: Blackwell.

Malinvaud, Edmond. 1991. "Econometric Methodology at the Cowles Commission:
Rise and Maturity." In Cowles Fiftieth Anniversary. New Haven, Conn.: Yale
University.
Mankiw, N. Gregory, and David Romer, eds. 1991. New Keynesian Economics,
vols. 1 and 2. Cambridge, MA: MIT Press.
Mankiw, N. Gregory, Julio J. Rotemberg, and Laurence H. Summers. 1985.
"Intertemporal Substitution in Macroeconomics." Quarterly Journal of Eco-
nomics 100: 225-251.
Tinbergen, J. 1939. Statistical Testing of Business Cycle Theories. Geneva: League
of Nations.
Commentary
by Arnold Zellner
Ray Fair is to be congratulated for his thoughtful discussion not only on

how the debate between the real business cycle (RBC) and new-Keynesian
(NK) theories might be resolved but also on how macroeconomic research
can be made more productive. I am in agreement with much that he has
written but, as is to be expected in such a controversial area, I have several
points of disagreement and shall put forward a different prescription in the
last paragraphs of my comment, one that we are actively pursuing.
Fair cogently observes, "I view the last two decades as making only
modest progress in this direction [of resolving the debate among various
business cycle theorists]. One problem is that there is too little testing of
alternative models. There has been no systematic attempt to find the model
that best approximates the macroeconomy. As disturbing, however, is the
fact that macroeconomic research appears to be moving away from its
traditional empirical emphasis" (p. 133). There can be no doubt but that
Fair is correct in these observations. He could even be more critical by
pointing to mystical elements in recent business cycle terminology, namely,
"laws of motion" to denote autoregressive processes, "deep structural
underlying parameters" to denote alphas and betas, "stylized facts" to
denote uncertain facts, "new-Keynesian" to invoke images of Keynes'
approval from the grave, and so on.
Fair's chapter starts with an introduction in which he points to modest
progress in the last two decades and a movement away from an empirical
emphasis, as indicated in the preceding quotation. He then describes the
Cowles Commission approach and the reasons for its decline. In section
1, a modern version of the Cowles Commission approach is presented
along with a discussion of selected technical issues and a digression on
monetarism. In section 2 attention is directed at the RBC approach, while
section 3 treats the NK approach. Finally, in section 4 he appraises the
future by looking ahead. In this section he suggests that "there is hope"
if all parties were to become involved in serious comparative testing of
alternative theories and if they were "to consider bigger models." While
the proposition that bigger is necessarily better is controversial (see a later
consideration of this issue), it is hard to disagree with the need for more
strenuous, empirical testing of alternative theories. However, here the
issue is how to approach the testing problem most fruitfully.
148
COMMENTARY 149
Perhaps the most depressing aspect of Fair's chapter is that after all
these years of research on the business cycle, we still do not have a depend-
able, thoroughly tested model that explains the past well, predicts
satisfactorily, and is useful for policy purposes. He recognizes this fact but
is still optimistic, as am I. However, I believe that what is missing is an
honest admission of our ignorance in important areas. Let's be frank and
admit that we don't know enough to formulate the n nonlinear equa-
tions in Fair's equation (1) with much confidence. The forms of consump-
tion equations, investment equations, production functions, and many other
equations are highly uncertain. Further, it is not at all clear that the
parameter vectors, the a/s, are constant for reasons given by Fair-namely,
Lucas effects and aggregation effects-and, I add, adaptive optimization
on the part of firms, consumers, and policy makers, changes in production
techniques induced by large-factor price changes, wars, oil crises, and
consumer changes in tastes. It is shocking that the underlying deep struc-
tural parameters may be variable. Further, who knows how to prove that
this n-equation nonlinear system has a unique solution and what its global
dynamic properties are? With a large n, there is a substantial probability
that equations, error term properties, restrictions, and the like are
incorrectly formulated. And with all of these problems, who knows how
to estimate parameters well and test hypotheses with confidence? Frank
answers to these questions are not reassuring. And indeed, simulation
experiments reported by the Adelmans (1959), Hickman (1972), Zellner
and Peck (1973), and others have revealed important, unusual properties
of existent models unknown to the model builders and users. Further,
various forecasting tests of Christ (1951), Cooper (1972), Nelson (1972),
Litterman (1986), and McNees (1986) reveal that Cowles Commission-
type model forecasts have not been satisfactory. Results such as these and
the poor performance of these models led McNees (1986, p. 15) to state in
the last paragraph of his paper that "macroeconometric models may have
been 'oversold' in the 1960s and early 1970s, leading to disappointment and
rejection in the late 1970s and 1980s."
Let's admit that the economic theory underlying the equations and over-
all structure of macroeconometric models, be they RBC or NK models, is
at best tentative and probably inadequate. Fair's chapter gives a reasonable
overview of how economic theory has been employed in model building.
Let me add a few critical remarks to indicate how shaky this theoretical
structure may be. First, with respect to equation (2), there is an expec-
tation operator E in the equation with no remarks on the underlying prob-
ability measure. Knight might say, in view of the near ignorance of experts,
that consumers face uncertainty, not risk, and the difficulty in formulating
a probability measure in this circumstance is formidable. However, if we

go along with Savage and Friedman, who disagree with Knight's distinc-
tion, and give the typical consumer a probability measure, what is it and
what axiom system supports the rationality of dynamic maximization
of expected utility? I know of none. But even if we allow the consumer
an appropriate probability measure, an exact solution to the consumer's
dynamic optimization problem is difficult to compute, as Fair points out.
He opts for a "certainty equivalent" approximation by replacing random
variables by their means. This "solution" has been shown to be a poor
approximation even in very simple problems by Fisher (1962), Zellner
(1971), and others. Further, the "solution" of a typical consumer, who is
endowed by the rational expectations hypothesis with a knowledge of the
"true" model, seems quite irrelevant to situations in which there is model
uncertainty and dispersion of beliefs. Also, Lovell's (1986) summary of
the empirical evidence relating to the rational expectations hypothesis
indicates that it is not as strong as could be desired.
To patch up the approximate equations of theory, a partial adjustment
equation is introduced in equation (6). Fair comments that "the adjustment
equation is not explicitly derived from utility maximization" (p. 136). Thus
it could and probably is at odds with the expected utility maximization
solution, theoretically quite disturbing. But more practically, equations
such as (6) probably reflect discreteness in durable goods and other types
of investment expenditures. Studies of panel data indicate that not every-
one buys a car or other large durables each period. On aggregation over
buyers and nonbuyers, the aggregate data appear to be generated by an
equation like (6) with the parameter lambda not a constant but intimately
related to the proportion of individuals making a purchase in a given
period. For analyses of such problems involving discreteness for consumers
and firms, see, for example, Chau (1967), Levedahl (1969), Laub (1971,
1972), and Peck (1973, 1974).
Further, models such as those described by Fair are deficient in that they
don't allow firms to shut down. Indeed in Muth's (1961) fundamental paper
on rational expectations, he modeled a competitive industry under the
assumption that the number of firms is constant, that is, no net entry or
exit. Marshall probably turned over in his grave on learning of this assump-
tion. See Veloce and Zellner (1984, 1985) for analysis of a demand, supply,
and entry model for a competitive industry. In most RBC, NK, and other
aggregate models, there does not appear to be allowance for shutdowns
and startups by plants and firms, a phenomenon extremely important
in major recessions and depressions. In RBC models, if the "represent-
ative firm" shuts down, does the whole economy shut down? No firm
COMMENTARY 151
heterogeneity has been introduced to determine which firms shut down,

and most models furthermore only indirectly and probably inaccurately
predict the number of plants and firms in operation in a given period.
Shutdowns and startups involve discrete changes and behavior on the part
of entrepreneurs and labor that are different from behavior in "normal"
times, another reason for parameters to change over the business cycle.
Also, in the latter part of expansions, when firms' profits are abnormally
high, there may be systematic inefficiencies afflicting production, another
reason for production parameters to vary over the cycle. Some of these
effects are being analyzed by Min (1991) in the context of a generalized
real business cycle model.
Add to these problems the problem of modeling policy-makers'
behavior, and one indeed becomes humble in connection with the theor-
etical underpinnings of current macroeconometric models. Faced with so
much ignorance on the theoretical side, there appears to be a tendency for
workers to follow Arrow's analysis of rational behavior under ignorance,
namely, to take extreme positions. One is that one or another of the
current theories, involving "laws of motion," "deep underlying structural
parameters," "rational" expectations, "true" data-generating processes,
Keynesian "truths," and so on, is "obviously" correct. Only ignorant
barbarians such as Ray Fair, Milton Friedman, and other misguided
individuals fail to believe the beautiful truths embodied in these theories
and suggest the crass need to test them empirically. At the other extreme
are those who reject all theories and retreat to a position of extreme
empiricism-this includes almost all of the time series statisticians and
many applied forecasters who have trouble understanding the current
jargon. The best of them, along with some econometricians, recognize that
multivariate time series models, including V ARs, have too many free
parameters, which means that usually estimates and forecasts are imprecise
and tests of hypotheses have low power. SUbject matter judgment and
theoretical knowledge are needed to restrict the number of free para-
meters, just as Fair indicates in discussion of his equation (1). However, if
the subject matter theory and knowledge isn't there, what's to be done?
This problem is irritating to both economic theorists and econometricians.
A suggested approach in this situation is described later. Before discussing
this problem, let me take up a few technical statistical issues that Fair has
raised.
In discussing statistical issues, I'm going to be as frank as I was in
discussing economic theory. First of all, the aggregate quarterly data are
afflicted with all kinds of systematic and random measurement errors and
then subjected to mysterious, horrendous seasonal adjustment procedures
with their peculiar assumptions about seasonal, cyclical, trend, and noise
components. All of these problems, which are swept under the rug in most
econometric studies, not only can vitally affect statistical estimation,
testing, and predictive procedures but also undermine the economic ration-
ale underlying statistical tests. Yet they are rarely mentioned. Perhaps
engineers' state space models involving measurement equations, reflecting
both systematic and random measurement errors and state equations with
state equation errors, are worth considering. Also, when error-ridden,
current preliminary estimates of GNP and other variables are employed,
who knows what results?
Abstracting from these serious measurement error and seasonal adjust-
ment problems, what are we to make of a model with n uncertain nonlinear
equations to be implemented with one set of data? There are issues of
parameter identification, testing overidentifying restrictions, estimation,
diagnostic checking, and prediction. In all of these areas there are serious
problems as I discussed years ago (Zellner, 1979). I shall just remark here
that there are, as yet, no secure and tested statistical methods for
formulating an n-equation nonlinear model from a single set of data.
Formulation of such a model is still an art. In the process, the data is often
used over and over again in tests of alternative variants of equations, and
thus subsequent estimates are afflicted with pretest biases and standard
errors are hardly satisfactory. Also, there is the well-known problem of
data mining, overfitting, and the like. For example in testing a valid null
hypothesis at the 10% level in independent trials, the probability of
accepting the null on one try is .9; on two, .81; on three, .73; and so on.
Thus after just three tests, the probability is .27 of rejecting the true null
hypothesis. Effects like this are well known to be present when a single
set of data is employed to screen many hypotheses-for a discussion of
how to correct for such "selection" effects in a simple case, see Jeffreys
(1967, p. 253). Also, the discussion in Friedman and Schwartz (1991) is
relevant. When many alternative hypotheses are considered with a single
set of data, conclusive results are hard to obtain, a fact that is probably well
known. Getting more data is helpful in such situations as is out-of-sample
predictive testing, as Fair indicates.
Fair remarks, perhaps in jest, "In modern times one has to make
sufficient stationarity assumptions about the variables to make time series
econometricians happy." With the presence of Lucas effects, regime
changes, aggregation effects, adaptive optimization, and so on, it is prob-
ably the case that the stationarity assumptions, which can be tested, under-
lying most macroeconometric analyses are violated. These include spectral
and cross-spectral analyses, impulse response function analyses, method of
COMMENTARY 153
moments analyses, and many analyses based on the assumption of fixed

parameters. There is a need to utilize flexible time-varying parameter
models that can adapt quickly to sudden parameter changes or gradually to
more gradual parameter changes. Also, theorists who theorize in terms of
fixed parameter models probably have to broaden their sights.
In our past work (Min and Zellner, 1990) we have developed and
applied posterior odds for fixed versus time-varying parameter versions of
a forecasting equation for the growth rate of real output. The posterior
odds are useful not only for model selection and/or combination purposes
but also for forecast selection and/or combination purposes-see Palm and
Zellner (1990) for further discussion of these issues. Posterior odds reflect
relative predictive performance and other factors-see Poirier (1991) for a
number of applications of posterior odds to various problems as well as
references to the earlier literature, which includes Geisel (1974), who used
posterior odds to compare simple Quantity Theory and Keynesian models.
The posterior odds approach to model selection deserves more attention in
macroeconometrics as does the Bayesian estimation and testing procedures
for equations of structural econometric models in Zellner, Bauwens, and van
Dijk (1988). These procedures yield finite sample posterior densities for para -
meters and finite-sample analogues of Wu-Hausman specification error
tests. Also, Bayesian predictive densities have been found useful in general
forecasting and turning-point forecasting problems-see, for example,
Zellner, Hong, and Min (1991), in which good turning-point forecasts
were obtained using a relatively simple "autoregressive-leading indicator"
CARLI) model in the form of a simple transfer function equation.
Now to return to the difficult problem of initial model formulation, in
earlier work Zellner and Palm (1974, 1975) and Zellner (1979), a structural
econometric modelling, time series analysis, (SEMTSA) approach has been
described. In this approach, a tentative structural equation model may be
formulated. Then the model's transfer functions can be derived, along with
other equation systems. The transfer functions so derived are usually
highly restricted and can be estimated and evaluated in extensive
forecasting testing. One by one the transfer functions or the components
can be evaluated. Once the components have been tested and found to
perform well, the problem of finding structural models compatible with the
transfer functions can be addressed. Here, rather than start with some fully
specified system of n-equations, we construct individual transfer functions
that forecast reasonably well and then consider structural equation systems
compatible with them that can be tested further in forecasting experiments
using as much new data as possible. Currently, we are using data for 18
countries in our work and data for many more are available.
On the issue of how many equations to employ, of course the objectives

of an analysis are important in this regard. If our objectives are just to
explain and forecast broad aggregates, then probably a small, sophisticatedly
simple model will suffice. On the other hand, if our objectives involve
explaining the behavior of many markets and sectors in detail, a large,
hopefully sophisticatedly simple, model will be needed. In either case,
the advice KISS, keep it sophisticatedly simple, is recommended. In U.S.
industry, KISS denotes keep it simple stupid. Since some simple models are
stupid, I prefer the first interpretation. Many scientists emphasize the
importance of sophisticated simplicity in science-see, for example, the
"simplicity postulate" in Jeffreys (1967).
On forecasting aggregates, one can model an aggregate variable in a
transfer function and forecast it. On the other hand, it is possible to break
the aggregate variable up into its components, forecast the components,
and add up the forecasts to obtain a forecast of the aggregate variable.
Many times, but not always, the latter procedure works better according to
some analyses that Enrique de Alba and I have underway. Further, too
much disaggregation can obviously lead to diminishing returns.
To illustrate some of the above points, suppose that our tentative struc-
tural equation system is given by HlYt = H2~t + F~t, where H l , H 2, and F
are matrix lag operators, Yt is a vector of endogenous variables, lSt a vector
of exogenous variables, and ~t a zero-mean white noise error vector. Given
that Hl is invertible, y, = H]lH2?St + H]lF~t or with H]l = Hjl1H11, where
Ht is the adjoint matrix,
(1)
is the transfer equation system. Note that the same polynomial operator,
IHll, hits each element of Yt. Let yf = Ylt + Y2t + ... + Ymt = j.'Yc. where j.' =
(1,1, ... , 1). Then from equation (1), -
(2)
It is possible to estimate equation (2) and use it for forecasting or to
estimate the equations in (1), forecast the components, and add up the
forecasts to obtain a forecast of yf.
Further, the equations actually fitted in (1) should be compatible with a
restricted version of the structural equation system. Thus we work back
and forth between the structural formulation and the transfer functions
that have been found to be successful in forecasting using as much data as
possible, that is, data for many countries and reasonably long time periods.
See, for example, Hong (1989), who has compared the predictive perform-
ance of transfer function (TF) models for the rate of growth of real output
COMMENTARY 155
incorporating lagged leading indicator variables with that of a version

of Barro's monetary surprise model, the Nelson-Plosser (0,1,1) ARIMA
model, and various naive models using data for 18 countries, 1951-1973
for fitting and 1974-1984 for forecasting. Hong found that the 1F or
autoregressive-leading indicator CARLI) model performs best. He has also
rationalized the ARLI model by showing that it can be obtained as a
relation implied by a simple structural econometric model. By having
relations like equation (1) or equation (2), or time-varying parametric
versions of them, implemented with data for many countries, it is possible
to use modem shrinkage estimation and prediction procedures to get
improved results.
In our recent work, we have been pursuing this approach, which differs
in important respects from the Cowles and V AR approaches, both of
which involve linking results to tenuous, complicated maintained models,
say a large highly restricted simultaneous equation model or a high-order
V AR. Since these maintained models are often untried and untested, there
is no assurance that results based on them are reliable, an obvious point.
Our SEMTSA approach involves testing the components of models
thoroughly and then putting them together sensibly to form a model that
can then be tested further in simulation experiments, forecasting, and other
uses using as much new data as possible. Hopefully, the SEMTSA
approach will yield a model capable of explaining the main features of
business cycles and of predicting future outcomes with satisfactory
precision.
References
Adelman, I., and F.L. Adelman. 1959. "The Dynamic Properties of the Klein-
Goldberger Model." Econometrica 27: 569-625.
Chau, L.c. 1967. "Stock Adjustment, Liquidity, and Normal Income in the Deter-
mination of Household Expenditures on Durable Goods." Doctoral disser-
tation, University of Wisconsin, Madison.
Christ, c.F. 1951. "A Test of an Econometric Model for the United States, 1921-
1947." In Conference on Business Cycles. New York: National Bureau of
Economic Research, pp. 35-107.
Cooper, R.L. 1972. "The Predictive Performance of Quarterly Econometric
Models of the United States." In B.G. Hickman, ed., Econometric Models of
Cyclical Behavior. New York: Columbia University Press, pp. 813-926.
Fisher, W.D. 1962. "Estimation in the Linear Decision Model." International Eco-
nomic Review 3: 1-29.
Friedman, M., and A.I. Schwartz. 1991. "Alternative Approaches to Analyzing
Economic Data." American Economic Review 81: 39-49.
Geisel, M.S. 1974. "Bayesian Comparisons of Simple Macroeconomic Models."

In S.E. Fienberg and A. Zellner, eds., Studies in Bayesian Econometrics and
Statistics in Honor of Leonard J. Savage. Amsterdam: North-Holland, pp. 227-
256.
Hickman, B.G., ed. 1972. Econometric Models of Cyclical Behavior. New York:
Columbia University Press.
Hong, C. 1989. "Forecasting Real Output Growth Rates and Cyclical Properties
of Models: A Bayesian Approach." Doctoral dissertation, Department of
Economics, University of Chicago.
Jeffreys, H. 1967. Theory of Probability (3rd ed.). Oxford: Oxford University Press.
Laub, P.M. 1971. "The Dividend-Earning Relationship: A Study of Corporate
Quarterly Panel Data, 1947-65." Doctoral dissertation, Graduate School of
Business, University of Chicago.
- - - . 1972. "Some Aspects of the Aggregation Problem in the Dividend-Earning
Relationship." Journal of the American Statistical Association 67: 552-559.
Levedahl, W.J. 1969. "The Effects of Households' Permanent and Transitory
Income on Both Their Timing of Automobile Purchases and Their Net Outlay
for Automobile Services for 1968, with Predictions for 1969." Doctoral disser-
tation, University of Chicago.
Litterman, R.B. 1986. "A Statistical Approach to Economic Forecasting." Journal
of Business and Economic Statistics 4: 1-4.
Lovell, M.e. 1986. "Tests of the Rational Expectations Hypothesis." American
Economic Review 76: 110-124.
McNees, S.K. 1986. "Forecasting Accuracy of Alternative Techniques: A Compari-
son of U.S. Macroeconomic Forecasts." Journal of Business and Economic
Statistics 4: 5-15.
Min, e. 1991. "Economic Analysis and Forecasting of International Growth Rates
Using a Bayesian Approach." Thesis proposal, Department of Economics,
University of Chicago.
Min, e., and A. Zellner. 1990. "Bayesian and Non-Bayesian Methods for Com-
bining Models and Forecasts with Applications to Forecasting International
Growth Rates." Journal of Econometrics, Annals.
Muth, J. 1961. "Rational Expectations and the Theory of Price Movements."
Nelson, e.R. 1972. "The Prediction Performance of the FRB-MIT-Penn Model of
the U.S. Economy." American Economic Review 62: 902-917.
Palm, F.e., and A. Zellner. 1990. "To Combine or Not to Combine? Issues of
Combining Forecasts." Journal of Forecasting.
Peck, S. 1973. "A Test of Alternative Theories of Investment Using Data from the
Electric Utilities Industry." Doctoral dissertation, Graduate School of Business,
University of Chicago.
- - - . 1974. "Alternative Investment Models for Firms in the Electric Utilities
Industry." Bell Journal of Economics and Management Science 5: 420-458.
Poirier, D.J., ed. 1991. Bayesian Empirical Studies in Economics and Finance,
Journal of Econometrics, Annals 49 (July/August).
COMMENTARY 157
Veloce, W., and A Zellner. 1984. "Modeling a Competitive Industry with Entry."
Economics Letters 16: 71-75.
- - - . 1985. "Entry and Empirical Demand and Supply Analysis for Competitive
Industries." Journal of Econometrics 30: 459-471.
Zellner, A. 1971. An Introduction to Bayesian Inference in Econometrics. New
York: John Wiley and Sons, reprinted by Krieger Publishing Co., 1987.
- - - . 1979. "Statistical Analysis of Econometric Models." Journal of the
American Statistical Association 74: 628-65l.
Zellner, A, L. Bauwens, and H.K. van Dijk. 1988. "Bayesian Specification Analysis
and Estimation of Simultaneous Equation Models Using Monte Carlo
Methods. " Journal of Econometrics 38: 39- 72.
Zellner, A, C. Hong, and C. Min. 1991. "Forecasting Turning Points in Inter-
national Output Growth Rates Using Bayesian Exponentially Weighted
Autoregression, Time-Varying Parameter, and Pooling Techniques." Joumal of
Econometrics 49: 275- 304.
Zellner, A., and F.e. Palm. 1974. "Time Series Analysis and Simultaneous
Equation Econometric Models." Journal of Econometrics 2: 17-54.
- - - . 1975. "Time Series Analysis of Structural Monetary Models of the U.S.
Economy." Sankhya 37 (Series C): 12-56.
Zellner, A., and S. Peck. 1973. "Simulation Experiments with a Quarterly
Macroeconometric Model of the U.S. Economy." In A.A. Powell and R.A.
Williams, eds., Econometric Studies of Macro and Monetary Relations.
Amsterdam: North-Holland, pp. 149-168; reprinted in A. Zellner. Basic Issues
in Econometrics. Chicago: University of Chicago Press, 1984.
SESSION III
5 HOW DOES IT MATTER?
Benjamin M. Friedman
Making economic policy typically requires positive as well as ethical

judgments. Any kind of public policy inevitably involves fundamental
presumptions that classify some aspects of human experience as desirable
and others not, and that value some of these desirables more highly than
others. (It is no accident that "policy" has the same root as "politics.") At
the same time, if the actions taken are to be at all effective in promoting
the ends sought, they must reflect a not wholly inaccurate perception of
how they relate to the aspects of experience they are supposed to influence.
In the case of economics, an essentially behavioral field of inquiry, the
central question for policy purposes is how (if at all) individuals and
institutions will alter the conduct of their affairs in response to any of the
vast variety of changes that economic policy can bring about in the
environments they face.
Macroeconomic policy is certainly no exception in this regard. Here,
too, the notion of which outcomes are desirable and which are not-most
obviously, the inherent desirability of a higher rather than lower standard
of living, appropriately defined, over time-must ultimately be assumed,
not established. And here, too, positive questions about the connection
between public policy actions and specific developments bearing on the
161
achievement of these objectives-Do large government deficits depress

private capital formation? Does tight monetary policy dampen overall
economic activity?-are central.
Moreover, many of the positive questions that bear most importantly on
policy making in the macroeconomic sphere are inherently quantitative,
and not merely in the sense that empirical evidence is necessary to judge
among competing theories of the same phenomenon. Simply asking which
theory "explains" unemployment, for example, misses the point that with
anywhere from 6 to 11 million people unemployed at any given time (in the
United States, during the last decade), there is ample room for different
sets of circumstances and responses to underlie the condition of different
would-be workers. But conversely, simply asking whether any specific
theory of unemployment is "true," in the sense that it accounts for the
condition of at least some empirically detectable number of would-be
workers, also misses what policy making is, or ought to be, all about. The
right question from a policy perspective is which theory or theories, if any,
can account for enough of the observed unemployment to serve as a plaus-
ible guide to useful policy.
Business cycles-the subject of this conference, and hence of this chap-
ter-have traditionally figured prominently in macroeconomic research,
although it is important not to confound the study of business cycles with
the entirety of policy-relevant macroeconomics. Price inflation often varies
with the business cycle, to be sure; but inflation also can and does occur on
a time scale different from that which defines most business cycle research,
and inflation can and does arise from causes not directly related to the busi-
ness cycle. Economists have also recently taken up with renewed vigor the
question of what makes economies grow and whether countries that start
out either richer or poorer than one another will converge economically,
over even longer spans of time. These matters are also part of macro-
economics, and they too bear importantly on key issues of public policy.
Even so, the study of business cycles-what accounts for the irregu-
larities surrounding whatever growth trend aggregate economic activity
follows, and in particular what causes occasional episodes of apparent
underutilization of the society's available economic resources-is the
heart of macroeconomics. After all, macroeconomics, as a distinguish-
able subfield within the discipline, was born of economists' perplexity at
the events of the 1930s, which not only challenged existing economic
orthodoxies but also, for a time, threatened the integrity of several
countries' democratic political structures. The positive question of why so
many people were out of work and factories idle, and the corollary issue of
what public policy could or should do to ameliorate the resulting human
HOW DOES IT MATTER? 163
hardship, were, in the first instance, what the subj ect was all about. Although
the aggregate-level fluctuations observed in the world's advanced industri-
alized economies in the post-Depression era have been both less severe
and less distressing, in human as well as political terms, the same questions
have largely framed this line of study ever since.
Modem research on business cycles revolves around several important
distinctions, distinctions that appear to correspond to policy prescriptions
no less than to positive economics. The line of research that grew most
directly out of the Depression experience, and that dominated the first
quarter-century of thought on the subject following World War II,
emphasized fluctuations in the public's demand for goods and services. The
implicit assumption behind this focus on demand was either that the eco-
nomy's ability to produce goods and services was highly elastic or that that
ability, if less elastic, at least did not experience sharp movements over
business cycle horizons. The remaining question was then why the demand
for goods and services fluctuated, including fluctuations that at times
carried demand in the aggregate below what the available resources could
readily supply. The main distinction this line of research came to em-
phasize was that between "monetary" disturbances, which disrupted the
equilibrium between the public's desire to hold cash balances and the
banks' ability to create those balances, and such nonmonetary factors
as changes in government spending and revenues or "autonomous"
shifts in households' desire to consume or in firms' desire to invest. (See
chapter 2.)
The economic events of the 1960s and 1970s, however, undermined
confidence in the assumptions behind this exclusive focus on disturbances
to the demand for goods and services as the source of business fluctuations.
First, the emergence of rapid price inflation, as a by-product of the effort in
many countries to stimulate demand so as to achieve ever higher levels of
employment, made the corresponding supply appear less elastic than policy
makers in those countries had hoped. Then the quadrupling of petroleum
prices by the OPEC cartel (and subsequent redoubling, several years later)
showed that aggregate supply could be not only inelastic but also subject to
disturbances just as abrupt, and apparently just as important for business
fluctuations, as those highlighted in the earlier demand-oriented research.
The emergence of a new set of business cycle theories primarily em-
phasizing disturbances to aggregate supply, in contrast to either strand of
the earlier demand-based theories, framed yet another important distinc-
tion from which continuing research drew normative as well as positive
conclusions. Over time, the ensuing "real business cycle" approach came
to encompass not merely disturbances to aggregate supply but also those
disturbances to aggregate demand that entered the story in an analytically

parallel fashion. (See chapter 3.)
The object of this chapter is to question whether the distinctions among
different theories of the business cycle actually have the force usually
assumed in their behalf in contemporary discussions of macroeconomic
policy. To anticipate, these distinctions do carry important implications
about which macroeconomic policies are likely to be more efficacious than
others in a business cycle context and about how best to carry out these
policies. The distinction most popularly associated with the debate over
"real business cycle" theories, however-that is, the distinction between
aggregate demand and aggregate supply as the principal location of the
disturbances that drive business cycles-is, from a policy perspective, less
important than is commonly believed. The policy prescriptions that follow
from most "real business cycle" models have more to do with the kinds of
assumptions that these models incorporate about how markets function
than with whether the chief disturbances to which the economy is subject
work through demand or supply. At the same time, a further set of
distinctions not customarily addressed in the business cycle literature,
mostly revolving around the definition of "income," turns out to be sur-
prisingly important. Finally, yet further issues, which traditionally receive
too little attention from economists, arise from the fact that the people
and the business institutions that make up the private sector of a modern
industrialized economy are vastly heterogeneous and that democratic
forms of government, for all their virtues, have not been very effective in
arranging appropriate transfers from one group to others as the need arises.
1 Aggregate Supply Versus Aggregate Demand:

A Conventional View
In analysis along the lines of the earlier postwar business cycle literature, in
which the focus is on aggregate demand and the question at issue amounts
to whether the chief disturbances to that demand are monetary or
nonmonetary, the analysis nearly always points to at least the potential
usefulness of one or another kind of corrective policy intervention. If
increased concerns about the risks embodied in nonmoney assets lead
investors to want to hold a larger share of their portfolios in money
balances, the central bank in a fractional reserve banking system should
expand the quantity of bank reserves, so that banks can accommodate
the larger demand for deposit creation. If a decline in stock prices leads
households to want to cut back on their consumption spending, the fiscal
authority should either increase its own spending or stimulate private

spending by reducing taxes, or the central bank should stimulate interest-
sensitive elements of private spending, such as home building or auto-
mobile purchases, by increasing the quantity of bank reserves sufficiently
to reduce market interest rates. And so the story goes.
Needless to say, a host of assumptions about the behavior of households
and businesses, and about the economic environment that they face,
underlies these policy inferences. As is well known, a nonaccommodated
disturbance to the demand for money--or, in the same vein, an action by
the central bank to increase the supply of bank reserves-would not affect
real economic activity if prices and wages were perfectly flexible, if credit
markets were subject to neither ordinary frictions nor failures due to asym-
metric information, and if none of the other familiar sources of monetary
nonneutrality were present. Increased government spending would not
stimulate real economic activity if whatever the government purchased
were perfectly substitutable for private consumables. Nor would tax
changes affect economic activity if all taxes were lump-sum, if credit
markets were perfect (so that, for example, no one faced liquidity
constraints), and if no such factors as childlessness or immigration or
uncertainty over future incomes rendered the appropriately discounted
value of taxes to be levied at some undefined future date less than that of
taxes to be paid forthwith. But in light of the readily apparent gulf between
these sets of rarified conditions and the economies in which actual
households and businesses carry out their affairs, the assumption that
money is not neutral, or that government spending and tax changes do
affect economic activity, is certainly plausible enough.
Within the demand-oriented approach, the distinction between policy
interventions that are likely to be useful versus only potentially useful turns
importantly on assumptions about how much knowledge policy makers
have. Early contributions to this literature established that uncertainty
about forces or events that will affect the economy in the same way regard-
less of what actions policy makers take need not impair the effectiveness of
policy interventions. By contrast, uncertainty about the magnitude and
timing of the consequences of policy actions themselves clearly blunts the
ability of such interventions to do any good at all. Indeed, without at least
some minimum knowledge about the how and when of such effects, an
interventionist policy might well be destabilizing. These concerns are
especially relevant for monetary policy in light of the familiar finding that
the lags by which central bank actions affect economic activity are both
long and variable. As a result, much of the debate within this literature
over the relative merits of a more versus less activist stance for monetary
policy has in fact hinged on the uncertainty issue, more so than on any
disagreement over such behavioral questions as whether money might be
neutral.
Given the assumption that macroeconomic policy actions can stimulate
or retard aggregate demand with at least some modicum of reliability on
average over time, the guiding presumption of the demand-oriented busi-
ness cycle literature is that policy actions should do so in such a manner as
to even out, insofar as possible, disturbances affecting aggregate demand.
The rationale underlying this presumption is simply that in the best of
all worlds-which by the economist's standard assumption is, of course,
a world free of all impediments to market-clearing equilibrium, so that
all relevant marginal tbis's always equaled the appropriate marginal that's
-disturbances to aggregate demand would not affect real economic out-
comes anyway. In that world the allocation of all economic resources would
depend solely on considerations associated with aggregate supply (and
on more fundamental aspects of demand, like the prevailing economy-wide
rate of time preference, which are unlikely to vary much over business cycle
horizons). In an actual economy not blessedly free of all such imper-
fections, the role of macroeconomic policy is therefore to nullify the impact
of disturbances to aggregate demand whenever possible and thus to restore
real economic activity to its pristine supply-determined equilibrium.
Against the background of this general philosophy of the demand-
oriented approach to business fluctuations, the realization that aggregate
supply also might be subject to sudden disturbances naturally created the
appearance of a sharp analytical contrast. Unusually good or bad harvests,
new technologies that increase productivity, or changes in an open eco-
nomy's terms of international trade, all imply changes in just the mar-
ginal this's and that's that are ideally supposed to determine how the
society deploys its resources. When supply considerations shift, therefore,
the standard presumption is not to offset their impact on real activity
but instead to interpret that impact as the requisite movement to a fresh
supply-determined equilibrium. This new equilibrium may be either
superior or inferior to the one that preceded it, depending on whether the
disturbance that brought about the change was favorable or adverse, but in
either case it is superior to any other allocation available in the new post-
disturbance environment.
Hence the chief policy implication of the view that the business cycles
actually observed in modern industrialized economies are "real business
cycles," in the sense of movements driven by disturbances to aggregate
supply (or, less likely for short-run fluctuations, disturbances to the fun-
damental underpinnings of aggregate demand), is that no macroeconomic
HOW DOES IT MAITER? 167
policy response is needed or appropriate. Indeed, any effort by policy makers

to resist such a "real business cycle" would only impede the economy's
progress toward its newly appropriate optimal allocation of resources.
This general philosophy-in an important but usually unstated sense, a
direct corollary of that underlying the earlier demand-oriented research
-also was well understood, at least implicitly, in that earlier literature.
What was new from this perspective in the "real business cycle" approach
was simply the idea that disturbances to aggregate supply might play a
major role in short-run business fluctuations, as distinct from phenomena
observed on a more secular time scale. After the experience of 1973-1975,
however, in which the then-largest decline in u.s. output and employment
since the 1930s quickly followed a fourfold increase in the price of a major
input to industrial production, and other oil-importing economies around
the world experienced analogous problems, the possibility of business
cycles caused by supply disturbances no longer seemed remote.
2 Aggregate Supply and Aggregate Demand:

A Closer Look
Notwithstanding its intuitive attractiveness, the practical usefulness of this

supply-versus-demand dichotomy as an organizing principle for macro-
economic policy making is less than the underlying logic suggests, and
for several reasons. The most straightforward of these is the difficulty, in
an actual policy-making context, of drawing the requisite distinction. Many
occurrences that initially seem to represent disturbances to aggregate
supply likewise cause disturbances to aggregate demand, and vice versa.
For example, if the United States used only domestically produced oil,
an oil cartel's sudden decision to restrict production (in the spirit of the
Texas Railroad Commission of earlier days) would simply correspond to an
adverse shock to the economy's ability to supply the many goods, such as
petrochemicals and services, such as transportation and heating, that use
oil as inputs. From a demand perspective, the incomes of oil users facing
higher prices would initially fall by just the amount that the incomes of oil
producers rose, so that aggregate demand would not shift. The economy's
new equilibrium would represent the intersection of the adversely shifted
aggregate supply schedule and the same aggregate demand schedule that
prevailed beforehand. To resist the decline in overall real output by stimu-
lative monetary or fiscal policy would only generate inflation.
In the actual circumstances surrounding the OPEC price increases of
1973 and 1979, however-circumstances that persist today and are likely to
do so for years to come-the United States imports one-half or more of its

petroleum. An increase in price imposed by a foreign cartel constitutes a
disturbance to both aggregate supply and aggregate demand. With U.S. oil
imports running at some 7.5 million barrels per day (as of the time of
writing), a doubling of the current price of $21 per barrel would immedi-
ately impose on American oil users the equivalent of an excise tax of
$57 billion per year, or about 1% of the national income, to be paid
to foreigners. It is always possible, of course, that the right choice for
macroeconomic policy would be to make no response, and so to accept the
equilibrium represented by the intersection of the economy's adversely
shifted aggregate supply schedule and the dampened aggregate demand
schedule. But there is no reason why that need always be so.
A second, more fundamental reason why the distinction between supply
disturbances and demand disturbances is of less value than meets the eye is
that in many circumstances the dichotomy breaks down for reasons not of
institutional fact (like reliance on foreign oil) but basic economic logic. A
standard example of a contractionary disturbance to aggregate demand is
a decline in households' willingness to spend on consumables at a given
price, caused by any of a variety of trauma to consumer confidence. A stan-
dard example of an adverse disturbance to aggregate supply is a decline in
workers' willingness to sell their labor at a given wage, caused by any of a
variety of perceived changes in working conditions. But are these two
disturbances really distinct?
An analogy to a different application of economic analysis might clarify
the answer. In analysis of portfolio behavior, it is customary to take explicit
account of the interrelationships among any investor's demands for
different assets (or supplies, if the investor can also issue assets-say, by
borrowing). The root of these interrelationships is a balance sheet
constraint that allows the investor at any given time to hold assets that sum
to no more, and in most familiar circumstances no less, than the value
of the portfolio to be invested. At a given moment, therefore, increased
demand for one asset necessarily means reduced demand for (or, if the
investor can borrow, increased supply of) at least one other asset. The
point is not just that a change in expected returns that induces an investor
to hold less of one asset means he can then hold more of another. More
important, anything that shifts the investor's entire demand schedule for
one asset must also shift the demand schedule for at least one other asset.
In the absence of special assumptions, a shift in the demand schedule for
anyone asset will shift the demand schedules for all other assets.
In the case of household behavior determining flows such as consump-
tion demand and labor supply, the analog to the balance sheet constraint in
HOW DOES IT MATIER? 169
the portfolio case is a budget constraint. People earn incomes, by working

and from other sources, and they either spend their incomes on consumables
or save them. The sum of incomes earned from all sources must equal the
sum of incomes spent and incomes saved. If some change in circumstances
leads to a shift in the supply of labor, therefore, the budget constraint im-
plies that it must also shift either the demand for consumables or the dem-
and for saving, or both. Here, too, the point is not merely that a wage change
that leads someone to work less will then reduce the amount that indivi-
dual consumes or saves. More important, anything that shifts a workers's
entire labor supply schedule will necessarily shift either that person's
consumption demand schedule or saving schedule-or, in general, both.
In principle, it is still possible to distinguish supply shocks from demand
shocks by pushing the analysis back still further, to focus not on shifts
in labor supply schedules or consumption demand schedules but on the
specific events that trigger those shifts. For example, if a new President
chose not to enforce the regulations protecting the safety and healthfulness
of the workplace, and large numbers of Americans therefore became less
willing to work at any given real wage, the resulting shift in both labor
supply and consumer demand schedules might plausibly be attributed to a
supply disturbance. It is not clear what would be gained from that labelling,
however, and in any case the resulting distinction would not fully corre-
spond to the more conventional supply-versus-demand dichotomy with its
powerful ability to support inferences about the circumstances under which
macroeconomic policy should or should not attempt to counteract business
fluctuations.
There is yet a third reason, however, why even the conventional supply-
versus-demand dichotomy-and even when the identification of a supply
or demand disturbance is completely unambiguous-does not have the
force commonly associated with it in contemporary discussions of macro-
economic policy. From a policy perspective, what most importantly
distinguishes the "real business cycle" approach from the more traditional
analysis of either monetary or nonmonetary influences on aggregate
demand is not so much whether the relevant disturbance in the first
instance affects the supply or demand side of the economy (again, on the
assumption that it is possible to draw such aline), but whether the analysis
takes account of either the non-Walrasian market mechanisms or the non-
Pigouvian tax systems that characterize actual economies. The standard
presumption that the equilibrium allocation of resources established by
supply considerations is "right," and conversely that movements of aggre-
gate demand should not be left to interfere with it, rests on the assumption
that there are no impediments to the clearing of all markets via the price
mechanism and that the tax system embodies a set of penalties and
subsidies sufficient to internalize all relevant externalities. By contrast, if
wages and prices are not perfectly flexible, or if markets are subject to
frictions, or if the tax system leaves some externalities uncorrected, then
there is room for macroeconomic policy to respond to even the purest of
"supply" shocks.
For example, consider the circumstances ("imagine" might be more
appropriate in this case) if OPEC's leaders had been sincere in their pro-
testations throughout the 1970s that they were merely using a higher price
to encourage the world to economize on a scarce nonrenewable resource,
and thus had compensated each oil-importing country by remitting the
extra revenue that the cartel received, to be distributed on a lump-sum
basis to that country's popUlation. At least in conventional analyses-that
is, abstracting from the prior discussion of inherent interrelationships
among such aspects of behavior as labor supply and consumption demand
-this situation would strictly correspond to an adverse supply disturbance.
The higher price of a key imported input would reduce the economy's
overall ability to produce goods and services, but the lump-sum distribution
of remitted proceeds would restore the demand for goods and services (in
aggregate) to its prior level. In the face of a decline in productivity, the
economy's new equilibrium would presumably call for a lower real wage.
As long as labor supply exhibits at least some positive elasticity, this new
equilibrium would therefore involve lower levels of both output and
employment.
But can an economy characterized by sticky nominal wages reach that
new equilibrium without some assist from macroeconomic policy? Suppose
that in this circumstance the central bank simply kept the supply of bank
reserves fixed and that no further disturbance affected either the public's
demand for money balances or banks' willingness to create them from a
given reserve base. The adverse aggregate supply shift would involve both
reduced output and higher prices. The higher prices in turn would imply
lower real wages. So far, so good. But in general there is no reason why the
reduction of real wages brought about by the price rise needed to clear the
market for goods and services would be proportional to the reduction of
real wages required to clear the labor market. A plausible role for mone-
tary policy in that case would be to deliver either a greater or a smaller price
rise than would occur otherwise, so as to achieve whatever decline in real
wages were necessary to reach the new equilibrium despite the rigidity of
nominal wages.
The fact that monetary policy can play such a role, of course, does not
mean that it actually should do so. The implications of uncertainty, as
HOW DOES IT MAITER? 171
emphasized earlier in the context of disturbances to aggregate demand, are

pertinent here as well. In addition, inflation and deflation presumably
impose costs too, and these need to be weighed against the costs of the
incorrect allocation of resources that would ensue from too high or too low
a real wage. But like the analysis of the relevant uncertainty, that compari-
son would remain to be carried out and would require not only some
explicit conceptual view of the costs of price movements but also some
ability to quantify these costs (relative to the costs of continuing product
and labor market disequilibrium)-on neither of which fronts has the
economics profession achieved much progress. In any case, the prima facie
presumption that there would be no role for macroeconomic policy in these
events, simply because the initiating disturbance was a "supply shock,"
does not withstand scrutiny.
It is a merit of the real business cycle approach that most of the leading
contributions to it have been consistent and even explicit in this regard.
The models used to carry out the analysis typically describe explicitly
Walrasian market mechanisms with perfectly flexible wages and prices.
For the most part, they also assume away such other potential impedi-
ments to optimal Walrasian outcomes as transactions costs, information
asymmetries in credit markets, behavioral nonlinearities, multiple
equilibria (and hence the possibility of self-fulfiling expectations), and
heterogeneities that might explain the existence of "inside" financial assets
(importantly including most of what is conventionally called "money"). Of
such assumptions are sharp policy conclusions made.
By contrast, the demand-oriented approach has been neither explicit
nor consistent. To recall, the notion that macroeconomic policy should
change aggregate demand, to counteract disturbances to it, stems from the
presumption that the supply-determined equilibrium is always the right
one. But macroeconomic policy can affect aggregate demand only if some
kind of price rigidity or market imperfection (or incompleteness) renders
monetary policy nonneutral and fiscal policy non-Ricardian. Whatever
combination of rigidities and imperfections accounts for the potency of
macroeconomic policy therefore vitiates the automatic presumption in
favor of the optimality of the supply-determined equilibrium.
In the end, the dichotomy that remains genuinely compelling in its
implications for macroeconomic policy in a business cycle context is not
whether the disturbances underlying the fluctuations that confront policy
makers predominantly affect aggregate supply or aggregate demand, but
whether non-Walrasian rigidities and imperfections importantly charac-
terize the individual behaviors and market mechanisms that collectively
constitute "the economy." But that idea is harddly new.
3 Macro Policy for a Heterogeneous Economy
A large part of what distinguishes the modern world from the primitive is
its incredible richness of texture. Individuals differ along an infinity of
dimensions, of which many probably do not bear on economic behavior
but many probably do. People have not only different preferences for
this good or that, or for working more or less, but also whole different
approaches to organizing their existence in this world. People also differ in
what they bring to the economic table in ways that go far beyond such
familiar distinctions as who has had how much formal education or on-the-
job training or who owns what tradable assets. Institutions, to the extent
that they take on an organic aspect and therefore reflect more than just the
collective attributes of the individuals associated with them, likewise
exhibit enormous differences among one another.
Standard theories of the business cycle, be they of the aggregate demand
or real business cycle type, mostly ignore this heterogeneity. Most familiar
models at best distinguish the "representative household" from the "repre-
sentative firm," although some demand-oriented models also distinguish
those households that face liquidity constraints from those that do not.
Financial intennediaries usually exist in these models only in the fonn
of banks, which except for a stochastic element, unrelated to anything else
in the analysis, amount to no more than an extension of the central bank.
Much of the demand-oriented literature simply proceeds from mathemat-
ical statements describing the behavior of economywide aggregates, with
no explicit representation of either households or finns.
Simplification and categorization are both essential, of course, to fruitful
study of complex phenomena. Nevertheless, they also bear costs. From
the perspective of macroeconomic policy, one of the costs of the level of
abstraction at which the standard theories analyze business cycles is the
blurring, if not total elimination, of distinctions that experience suggests
matter importantly for actual policy decisions.
A hypothetical example can illustrate this point. In 1991, the govern-
ment of Japan made a $9 billion cash payment to the U.S. government in
consideration for the American role in the Persian Gulf War. (For
purposes of this discussion it is irrelevant whether one construes this trans-
action as a cost-sharing contribution in a joint endeavor or as a simple fee
paid for services rendered.) It is widely reported that this payment aroused
substantial political antipathy among Japanese voters. Suppose, therefore,
that instead of remitting $9 billion in cash, the government of Japan had
delivered 1 million Japanese-made automobiles, suggesting that the U.S.
government then sell them at an average price of $9,000 each.
The arrival in the United States of a million new foreign-made auto-

mobiles, free of charge to the economy as a whole, would probably have
a readily visible impact at the macroeconomic level. Moreover, because
that impact would presumably be short-lived, the resulting macroecono-
mic disturbance would be of the sort commonly addressed in the business
cycle literature. But would this event constitute a "supply shock" or a
"demand shock" in the context of the standard theories?
Although it is perhaps conceivable to argue that the U.S. economy's
ability to supply automobiles had thus been augmented by 1 million units,
any such argument would inevitably hinge on arbitrary conventions of
timing. If the gift cars all arrived in 1991, then this line of thinking would
hold that 1991 aggregate supply had received a favorable shock and that
aggregate supply in 1992 and beyond would remain unaffected. In a model
based on monthly time aggregation, however, the shock may have occurred
only in July. This kind of awkwardness is inevitable in trying to relate
discrete one-time quantities to concepts like production, which properly
represent flows per unit time, and they are one indication that in this case
the "supply shock" designation is inadequate. Instead, the receipt of the
million cars would more plausibly represent a one-time change in the eco-
nomy's stock of consumer durables, which under most standard theories of
consumption behavior would shift aggregate demand.
Given the identification of this event as a disturbance to aggregate
demand, the standard presumption is that macroeconomic policy should
respond. But how? Should the objective be to maintain levels of aggre-
gate spending, inclusive of consumers' purchase from the government
of the million Japanese-made automobiles? Should it be to maintain
levels of aggregate output (which would be like maintaining aggregate
spending exclusive of the new cars)? And what, if anything, should policy
do about the shift in the composition of demand for U.S. output,
which would involve a sizable decline in demand for American-made
automobiles?
If the U.S. economy consisted entirely of "representative agents," these
matters would be either straightforward or irrelevant. The appropriate new
equilibrium would involve a higher level of U.S. consumption and spend-
ing (inclusive of the million gift cars), and a lower level of U.S. production
and value added, than would otherwise be the case. The fact that demand
for American-made cars in particular would decline, while demand for
other consumer goods and services would rise, would not matter. Each
"representative agent" would be better off.
In the actual world of American economic policy making, these
questions would also be irrelevant, but for a different reason: The U.S.
government would decline the gift. The point is not merely that the govern-
ment would prefer the cash to the cars. If the cash were simply not in the
picture, and the only choice were to take the cars or receive nothing from
the Japanese in consideration of the American war effort, the U.S. govern-
ment would still decline the gift of the cars.
The reason why the actual policy choice would no doubt be to reject a
gift that would make the "representative agent" in the American economy
better off sheds light on two shortcomings of standard macroeconomic
analysis. To begin, the models used do not adequately distinguish between
income and output. In the hypothetical case under discussion, the heart of
the matter is that aggregate income should rise while aggregate output
should decline. If everyone in the economy were a "representative agent,"
with an equal share in aggregate income and aggregate output alike, there
would be no reason to focus macroeconomic policy on maintaining output
as distinct from income. Income is what would matter. (In fact, there are
good reasons why output matters along with income, but they arise from
dynamic considerations of international competition rather than economic
fluctuations in the ordinary sense.) But in an economy made up of
heterogeneous elements, many people's ability to earn income depends
directly on their opportunity to contribute to output. If output falls, their
income falls too, even if aggregate income rises.
Moreover-and from a practical perspective, more importantly-many
people's ability to earn income depends on their opportunity to contribute
to the output of a specific good or service. In the hypothetical example of
the Japanese gift cars, even if macroeconomic policy managed to maintain
aggregate u.s. output unchanged (so that aggregate U.S. spending rose
by the value of the million cars), the output of the American automobile
industry would be smaller and the output of many if not most other
American industries would be greater. This shift, even within a given level
of aggregate output, would leave some people worse off even if the
"representative agent" were better off.
To the extent that macroeconomic models address such issues at all,
as opposed to burying them under the abstraction of the "representative
agent," they typically do so in two ways. One is to assume that factors of
production, including labor as well as capital, are mobile among alternative
uses. The other is to assume that appropriate redistributions from those
individuals initially made better off by any change to those initially made
worse off can, in the end, leave everyone better off as long as the change is
welfare-improving in the aggregate. Both of these responses fall short, at
least in terms of what is relevant to macroeconomic policy in a business
cycle context.
It is readily apparent that both labor and capital are far from fully
HOW DOES IT MATIER? 175
mobile, even over horizons longer than any standard business cycle.
Individuals possess both industry-specific and employer-specific human
capital. Machines and buildings have limited functional adaptability or
geographical mobility. Institutions, including conventional businesses as
well as many in the "not for profit" sector, likewise acquire vested interests
in the continuation or expansion of quite specific economic activities. Even
people with no direct participation in a company or industry may acquire
analogous interests, in that the elimination of a business (or, for the same
reasons, a military base) would reduce the demand for their own services
or the market value of their nearby property. The costs of adjustment that
induce people and institutions to strive so hard to continue in their eco-
nomic activity rather than move to deployment elsewhere-witness the
1991 Washington "summit" meeting of some one hundred American
corporations to coordinate lobbying strategies to prevent the downscaling
or possible elimination of the B2 bomber project, for which each was a
supplier-are clearly great enough to preclude the immediate and full
transfer of resources in the wake of some kind of supply disturbance.
The absence of transfers by which the "winners" can so compensate
the "losers" as to leave everyone better off after a change that would be
welfare-improving in the aggregate also involves, conceptually, a kind of
transactions or adjustment cost. Here, however, the costs precluding what
the standard theory simply assumes will take place are not economic
but political. For all their virtues, the democratic political institutions
that govern the world's advanced industrialized economies have not been
very successful at compensating those individuals or businesses, or other
institutions, that are adversely affected by changes that benefit many
others. The economist's notion of a Pareto improvement (a change that
leaves at least some people better off and no one worse off) therefore loses
its practical relevance. Given the combination of enormous heterogeneity
and limited mobility that characterizes the modern economy, few changes
are likely to be Pareto-improving on initial impact. And given the inability
of the prevailing political institutions to achieve the requisite transfers, the
winners cannot compensate the losers so as to turn a change that is merely
welfare-improving in the aggregate into a Pareto improvement. Hence
"policy," including macroeconomic policy, becomes a matter of "politics"
in the classic sense.
A different example may further sharpen the point. Unlike the hypo-
thetical receipt of gift cars from Japan, the currently proposed free trade
agreement between the United States and Mexico would, if instituted, alter
considerations bearing on production in ways that properly constitute a
"supply shock." Most obviously, the agreement would expand opportu-
nities to produce goods, for sale in U.S. markets, from American capital
and Mexican labor. Like a rise or fall in oil prices, these changes in supply
considerations would persist over time. But also like a rise or fall in oil
prices, they would have a short-run impact that bears analysis in a business
cycle context.
That analysis suggests that the new equilibrium would-as in the case
of the gift cars-involve higher aggregate U.S. income and spending and
lower aggregate U.S. output and value added. For just the reasons already
discussed, this change would be welfare-improving in the aggregate. (That
is why economists typically favor free trade.) But here again, heterogeneity
importantly enters the story. The reason why U.S. income would be higher
despite lower U.S. output is that the additional income earned on
American-owned capital deployed in Mexico would outweigh the loss of
income earned at home by American workers. In this case, therefore, the
relevant distinction is not who works (or owns stock) in the automobile
industry and who works elsewhere, but who earns income from selling
labor and who earns income from owning capital.
The fact that the free trade agreement would be welfare-improving in
the aggregate for the United States means that owners of capital could, in
principle, compensate workers so as to render everyone better off. But
because there is little prospect that those transfers will occur, organizations
representing U.S. labor strongly oppose the free trade agreement. If the
agreement is enacted, the same groups will no doubt seek macroeconomic
policy action to offset the loss of U.S. employment. (The more than
10% unemployment that developed in Canada following the implemen-
tation of a parallel U.S.-Canadian free trade agreement elicited wide-
spread calls for monetary and fiscal expansion there.) Explaining that what
has happened constitutes a straightforward supply shock-and a favorable
one, at that-and that conventional business cycle theories therefore
mandate simply allowing the economy to go to its new (in the aggregate,
preferred) equilibrium, would be of little practical import.
The point of all this is that, for reasons wholly apart from the questions
of rigidities and market imperfections discussed earlier on, standard busi-
ness cycle theories fail to address issues that importantly bear on the
making of macroeconomic policy in a business cycle context. The upshot is,
again, to blunt the force of whatever policy implications these models have
to offer.
4 Concluding Thoughts
The main line of argument in this chapter has been that recent
developments in business cycle theory-specifically, the emergence of
"real business cycle" theories-offer less in the way of practical guidance

to macroeconomic policy than what the usual discussion of them normally
conveys. One set of reasons involves the difficulty, either practical or
conceptual, of drawing the required distinctions between aggregate supply
and aggregate demand as the focal point of any given disturbance to the
economy. Another stems from the dependence of these theories' policy
implications on the absence of price rigidities or other impediments to fully
Walrasian market outcomes, a key set of issues mostly resolved by assump-
tion (and, to judge by the evidence, counterfactual assumptions at that)
rather than analysis. A third set of reasons reflects the tension between the
aggegate level at which these theories study economies and the rich
heterogeneity characterizing actual economic behavior and interests in the
modern world.
While each of these three arguments is relevant to macroeconomic
policy in general, and therefore also to monetary policy in particular, there
is yet an additional difficulty in attempting to apply lessons from the real
business cycle approach to the practical conduct of monetary policy. The
injunction to leave monetary policy unchanged in the face of business
fluctuations, because they are presumed to reflect disturbances to aggre-
gate supply, is relevant in practice only when it is possible to define,
recognize, and implement the "no change" monetary policy in the first
place. Does it mean maintaining growth of the money stock at some
previously established rate? If so, which measure of the money stock is
that? Moreover, a straightforward extension of the argument drawn earlier
about the interdependence between supply shocks and consumption
demand suggests that there is no guarantee that supply shocks necessarily
leave portfolio demands unchanged-including demand for whatever is the
chosen measure of money. Is the "no change" monetary policy that real
business cycle models warrant then to make the money stock grow along
some new, appropriately adjusted path? If so, is monetary policy conducted
in this way really distinguishable from the kind of actively interventionist
policy that real business cycle models supposedly reject?
Questions like these are hardly unfamiliar, of course. They have
traditionally stood at the core of the debate over the proper role of activist
monetary policy within the demand-oriented approach. That they emerge
once more, even in the context of real business cycle models, shows that
this supply-oriented analysis, even with its full panoply of restrictive
Walrasian assumptions, still does not resolve the long-standing issues at
the cutting edge of how to conduct actual monetary policy. In the end,
there is no easy way to avoid grappling with such hard problems as which
measure(s) of money (or reserves, or credit, or interest rates) provides
the best guide to the effect of monetary policy on economic activity, and
how these quantitative relationships change over time in response to busi-

ness cycle phenomena as well as other influences.
Finally, one last issue also merits attention. Expectations have always
stood close to the center of thinking about business cycles. No reader of
Keynes' General Theory could miss the persistent emphasis on the role of
expectations, and in particular the role of changing expectations, as the
source of the "autonomous" shifts that play such a major role in demand-
oriented models of business fluctuations. The more recent real business
cycle models incorporate expectations in a more up-to-date, and therefore
more explicit, way.
Although early work with models of "rational" expectations appeared
not only to incorporate expectational elements explicitly but also to restrict
them so as to preclude the kinds of randomly shifting sentiments to which
previous work had attributed much of the observed aggregate fluctuation,
the development of models exhibiting mUltiple equilibria has shown that
this contrast was more one of style than of substance. In these models an
economy can have a high- or a low-activity equilibrium, either of which
its inhabitants may "rationally" expect. Moreover, in a manner strongly
reminiscent of the earlier literature of "animal spirits" that caused "auto-
nomous" shifts in demand, these expectations can be self-fulfilling. Which-
ever equilibrium people expect wilJ prevail, and so they are "rational" to
expect it.
What all this leaves open, of course, is the role that macroeconomic poli-
cies can or should play in influencing those expectations. Such a role for
authority is well established in other areas relevant to public policy. No one
any longer seriously suggests that the right to free speech includes the right
to shout "fire" in a crowded theater in which no danger is present but
the shout itself, and if somebody were to do that, no one would deny the
responsibility of the management to take whatever steps it could to restore
order. Similarly, the role of public policy institutions in arresting bank runs
that arise from no source other than the spread of false information is also
well accepted. Is there then a parallel role for macroeconomic policy
in fostering expectations that correspond to high-activity equilibria and
resisting the development of low-activity expectations? If there is, how
can either monetary or fiscal policy go about playing that role? Perhaps
the next wave of research on the business cycle will say.
Acknowledgments
I am grateful to Michael Belongia for helpful comments on an earlier draft.

Commentary: Whatever Happened to
Contracyclical Policy?
by Michael R. Darby
One of the few advantages of reaching a certain age is the ability to put
things into the perspective of the history and breadth of one's science.
Obviously, I envy Ben Friedman's assignment to do just that, and I will
embroider my own license along the same lines.
I want to leave aside the issues of appropriate research agendas-we all
make our views clear in the agendas we pursue. Instead I want to focus on
the policy debate. Ben has challenged the now almost conventional wisdom
that the monetary authority's central goal should be price-level stability
and that it should eschew both contracyclical policy and unintentional
destabilization policy. He also challenges the view that fiscal policy is
pretty much hopeless. Finally, he proposes an agenda for future research
needed to better choose the correct set of policy instruments to address
particular business cycle phenomena. It is my intention to defend the
conventional views-a personally novel experience but perhaps that too
comes with middle age. Although I must differ with Ben on the desirability
of contracyclical policy, I applaud the clarity and vigor of his argument in
its favor.
Until living through the recent recession in a policy-making position in
Washington, I believed that, by and large, the economics profession had
outgrown its fascination with contra cyclical policy and fine-tuning eco-
nomic growth. Obviously I was wrong, and Ben is far from alone in arguing
for an activist approach to macroeconomic policy. Nonetheless, after
having seen the theory and practice of contracyclical policy thoroughly
tested in the 1960s and 1970s, the bulk of the academic segment of the
profession has-in my opinion-come to share the view that fine-tuning
the economy is beyond our collective wisdom and institutional capacity.
Still the urge to do good is strong; so I plan to devote the main portion of
my comments to explaining why that urge should properly be resisted.
These remarks will differ sharply from Ben's chapter in that I shall
not assume this shift or that shift as necessary to illustrate a particular
quandary-instead I shall try to present a bare-bones framework for look-
ing at policy strategies in the face of an economy in which macroeconomic
shocks occur randomly and with no clear, timely indication of their size and
direction. I shall concentrate on shocks to aggregate demand. The real
179
business cycle theorists do point to stochastic growth in equilibrium real

income which is largely orthogonal to the conduct of macroeconomic
policy, as will become apparent.
1 Three Approaches to Contra cyclical Policy
Let me begin here by stating that by "contracyclical policy" I mean policy

aimed at offsetting those demand shocks that move the economy away
from its secular equilibrium. I should add that I am using the terms
contracyclical and stabilization interchangeably. The goal of such policy is
to attenuate the business cycle, that is, to reduce fluctuations in real income
and unemployment.
Economists clearly disagree about the theoretical validity and practical
utility of contracyclical policy. The profession boasts at least three major
schools of thought on the subject. At the risk of oversimplification, I will
label these schools Keynesian, monetarist, and new-classical.
Keynesians and monetarists seem to agree, in theory, that contracyclical
policies can stabilize the economy. They disagree, however, on the extent
of variance reduction possible and on whether government has the tools or
the knowledge to implement such policies effectively and without aggra-
vating the conditions they are meant to correct. In contrast, the new-
classicals contend that contracyclical policy cannot stabilize the economy
because it is either systematic and thus has no effect on the economy at all
or is random and hence destabilizing.
The question is: Whose theory provides the most constructive basis for
economic policy? One way to address this question is to focus on how the
three schools view the effects of stabilization policies on the variance of
real income.
In a simplified example, two types of shocks-autonomous shocks and
policy-induced shocks-can perturb real income. Autonomous shocks arise
from any number of factors, such as shifts in money demand or shifts in invest-
ment demand, which in turn can arise from things like war in the Per-
sian Gulf, political unrest in the former Soviet Union, or sunspots. Policy
shocks, of course, are man-made changes in either monetary or fiscal policy.
Changes in real income (due to demand shocks) can be represented by
the sum of autonomous and policy disturbances. If no shocks were affecting
the system, real income would continually equal its growing equilibrium
value, leaving aside any adjustments to unexpected supply shocks. Govern-
ment policies would influence this secular equilibrium through the saving
rate, aggregate production function, and the level, quality, and growth of
COMMENTARY 181
the labor force. These policies have long-run rather than contracyclical
aims. Deviations from equilibrium real income due to demand shocks
-which I shall term cyclical income-would be zero if any autonomous
disturbance to the system were offset by a policy disturbance of equal
magnitude and opposite sign.
In the absence of stabilization policy, the variance in cyclical income is
equal to the variance of the autonomous disturbance. But, if we implement
stabilization policy, the variance of cyclical income equals the sum of the
variances of autonomous and policy shocks plus twice their covariance.
Thus a reduction in the variance of cyclical income is possible only if a
negative covariance between autonomous disturbances and policy dis-
turbances can be achieved. This means that when autonomous disturb-
ances are moving cyclical income in one direction, policy disturbances
on average are moving cyclical income in the opposite direction.
Where a negative covariance in the values of autonomous and policy
disturbances can be demonstrated, contracyclical policy may indeed be
stabilizing. There is no certainty, however, that this will be the case. The
assumed negative covariance has two components: First, a negative corre-
lation reflects the average ability of policy makers to move in an offsetting
direction. Second, a scale factor indicates the magnitude or forcefulness of
the policy moves taken. At the optimal scale factor, the variance of cyclical
income is reduced by the square of the correlation coefficient. For
example, if the correlation between autonomous and policy shocks were
-0.2, then at best the variance of cyclical income could be reduced by 0.04,
or 4%. If the scale of policy shocks supplied by government were too small,
the reduction in variance of cyclical income would be less than this amount.
If the policy shocks were too large, the result would be either a smaller
decrease or an actual increase in the variance of cyclical income.
We are now at the heart of the policy problem. If stabilization policy is
to be effective, policy makers must be able to determine the appropriate
direction, magnitude, and timing for the required policy shock. If direction
and timing can be determined so as to achieve a negative correlation,
restraint on the scale of policy is necessary for the policy to be stabilizing
rather than destabilizing. Conversely, a correlation of zero would indicate
that any active policy can only increase the variance of cyclical income and
hence destabilize the economy.
1. 1 What Keynesians Believe
Keynesians typically believe that the correlation between autonomous and

policy disturbances is substantially less than zero and therefore that the
potential gains from stabilization policy are substantial. This is implicit in

Ben's presentation, in which it is simply assumed that we can know what
shifts will occur in time for an offsetting policy to be implemented and have
an effect on the economy. Keynesians may acknowledge that contracyclical
policies have often been destabilizing in the past. But they believe that
better data and better techniques of analyzing data, as well as the seasoning
influence of past mistakes, have made us more adept at implementing
stabilization policy in the present.
1.2 What the New-Classica/s Believe
In contrast to Keynesians, a group that I shall term new-classicals believes

that the correlation between autonomous and policy disturbances is zero.
They argue that if policy makers react to autonomous disturbances in a
regular and predictable way, the public willieam to anticipate the policy
actions and thus they will have no effect on cyclical income. Therefore,
only random changes in policy will affect real income.
Furthermore, random policies-policies that are unrelated to the state
of the economy-are certain on average to destabilize the economy. Thus
their theory leaves no room at all for contracyclical policy. For new-
classicals, contra cyclical policy is a null set.
1.3 What Monetarists Believe
The monetarist approach to contracyclical policy lies somewhere between

the approaches of the Keynesians and the new-classicals on the negative
correlation condition, but it argues that contracyclical monetary policy
is likely to be destabilizing except as it stabilizes growth in monetary
aggregates.
Monetarists believe there can be some negative correlation between
autonomous disturbances and policy disturbances. Even in the new-
classical paradigm such a negative correlation could be related to superior
information or costly indexing. However, if the correlation is small in
magnitude, then the optimal scale of policy actions must be quite small.
The kind of bold actions that would be popularly acclaimed would also
destabilize the economy. As a result, monetarists have traditionally argued
that policy makers must tie their hands in advance to avoid intense political
pressure to oversteer.
COMMENTARY 183
A few decades ago, when the debate was limited to Keynesians and
monetarists--aren't the terms quaint to today's ears-I believed that
framing the debate in terms of slopes of IS and LM curves was completely
misleading. Here I have proposed a framework first presented by Milton
Friedman (1953) some four decades ago and hope that you will find it more
to the point.
Now let me summarize the points of view in terms of correlation and
scale: On correlation, Keynesians believe that they are able to recognize
autonomous shocks and take offsetting actions rapidly enough to achieve
a substantial negative correlation. New-classicals, at the other extreme,
believe that systematic actions are ineffective and therefore that only
random policy shocks uncorrelated with the state of the economy are poss-
ible. Monetarists fall in between, allowing for some effects of systematic
policy but emphasizing lags of recognition, decision, and effect so that the
correlation between autonomous and policy shocks is necessarily small
in absolute magnitude. On scale, Keynesians tend not to worry about find-
ing the right setting as well as direction, monetarists fear that political
pressures will lead to an overzealous and hence destabilizing scale of policy
actions, while any scale of random actions is destabilizing for the new-
classicals.
So what has happened to contracyclical policy? First, Keynesian
optimism has not fared at all well in the academic community and little
better in the practical world. By and large the 1960s and 1970s have been
read as demonstrating that the Keynesian policy means well but is badly
destabilizing. I must say that the theoretical arguments against contra-
cyclical policy, which Ben challenges, may be used to explain why it didn't
work, but have not been per se fatal to the Keynesian agenda. Second, and
perhaps as importantly, neither monetary nor fiscal policy has really been
available for attempting sophisticated contracyclical fine-tuning. Fiscal
policy has been dominated by longer-term clashes over the scale of taxes
and spending while the monetary authorities were facing a daunting task
just to avoid doing harm.
2 Monetary Policy
Monetary policy early in the Reagan administration can be characterized

as anti-inflation shock therapy that reduced the 13% inflation rate by one-
third. There was no contracyclical intent, and the resulting recession was
accepted by the monetary authorities and much of the public as the unfor-
tunate side effect of long overdue medicine. As we emerged from that
recession, regulatory changes in the financial service industry led to real

uncertainty as to what effect monetary policy actions would have on the
economy. Parametric uncertainty works to reduce the absolute magnitude
of the expected correlation of policy and autonomous actions, so there was
little scope for stabilizing monetary policy and a stable monetary policy
was a considerable achievement.
I have argued that both MIA (Mlless other checkable deposits) and M2
continued in the 1980s to be good indicators of the effect of monetary
policy on real income and inflation (Darby, Mascaro, and Marlow, 1989).
Knowing either monetary aggregate does not make one a perfect fore-
caster, but it does seem that either may serve as a reasonable indicator of
future effects of monetary policy. Chairman Greenspan's recent remarks
suggest that the Federal Reserve System's reading of the evidence has
come to correspond more closely with ours at least with respect to the M2
aggregate. Thus the recent lowering of interest rate targets can be under-
stood in terms of M2 growth persistently below Federal reserve targets. I
would characterize this as avoiding unintentional destabilizing as opposed
to contra cyclical policy in the conventional sense.
Should we expect that the reemergence of usable monetary aggregates
will lead the monetary authorities to undertake an active contracyclical
policy again? Probably not. Much has been learned since 1980, and it seems
clear that the current Federal Reserve believes it cannot stabilize real
income to any significant extent and should instead focus on an achievable
goal of price stability. In that world view, usable monetary aggregates
make it easier to avoid unintended policy shocks.
3 Fiscal Policy
Fiscal policy actions do not generate nearly as long-lasting effects on real

income as does monetary policy. This tends to increase the absolute magni-
tude of any negative correlation between policy and autonomous effects on
cyclical income. Despite this favorable comparison on the effect lag, even
two decades ago fiscalists faced a daunting problem trying to reduce the
decision and implementation lag to manageable proportions. The auto-
nomous shocks affecting the economy when the fiscal changes finally took
effect were unlikely to be the ones originally targeted and were, therefore,
unlikely to be appropriately offset by the fiscal actions.
In the 1980s, conflicts between the administration and Congress have led
to even longer decision and implementation lags. Fiscal policy too must be
inevitably aimed at long-term goals.
COMMENTARY 185
4 Conclusion
Stabilization policy was tried in the 1960s and 1970s. It failed. I believe that
it is that failure, and not any literal belief in the assumptions to which Ben
points, that explains its current disrepute. That does not mean that the
Federal Reserve's move since 1988 toward price stability did not affect the
real economy. Rather, those effects were viewed as a temporary side effect
of a long-run shift to a better equilibrium. While clearly there have been
internal and external differences of opinion about the speed with which the
Fed has moved toward price stability, those debates have not been in the
context of contracyclical policy.
I believe that the passing of contracyclical policy as a strategy is a good
thing for economists as a group. First, we are no longer promising a cure
that in actuality makes things worse. Second, as concentration shifts to
medium- and longer-term strategies, the systematic forces that economists
understand become more important relative to the random shocks that
elude us. It is perhaps inevitable that during what appears to be the early
stages of recovery from a recession, the siren song of contracyclical policy
might sound good. But nostalgia for a past which wasn't must not blind us
to the present reality. While I applaud the Fed's efforts to get M2 growing
again, that falls under eschewing procyclical policy or automatic destab-
ilizers. That, I think, we can all agree is a good idea.
References
Darby, Michael R., Angelo R. Mascaro, and Michael L. Marlow. 1989. "The
Empirical Reliability of Monetary Aggregates as Indicators: 1983-1987." Eco-
nomic Inquiry 27: 555-585.
Friedman, Milton. 1953. "The Effects of a Full-Employment Policy on Economic
Stability: A Formal Analysis." In Essays in Positive Economics. Chicago:
University of Chicago Press.
CONFERENCE OVERVIEW
Commentary: Deja Vu All Over Again
by Alan S. Blinder
Often, conferences acquire nicknames. I suggest we name this one either
the Michael 1. Fox conference or the Yogi Berra conference, for it has
provided the most reminiscences since my high school reunion. The
analogy is an apt one because the way David Laidler and Olivier Blanchard
review the two "debates"-a polite word for cockfights-reminds me of
puberty: It isn't much fun when you are going through it, but you look back
at it fondly-though inaccurately-as "the good old days."
Permit me a brief biographical detour. I received my undergraduate and
graduate educations in economics between 1963 and 1971, just the years
when the Keynesian-monetarist debate was in full swing, and began my
career as a macroeconomist on the eve of the new-classical counter-
revolution. In 1971, when I joined the cadre of practicing macroeconomists
-thereby Granger-causing the subsequent arrival of both Olivier Blanchard
and Bob King-I believed myself to be joining a process of Kuhnian
"normal science." Those of us trained in the 1960s knew there were
problems with both the empirics and, especially, the theoretical foun-
dations of the standard consumption, investment, and money-demand
functions-not to mention the wobbly Phillips curve. We also knew that
the treatment of inflationary expectations was a major problem for the
latter. But progress was being made. The macroeconomics of 1971 was
noticeably better than the macroeconomics of 1961 and 1951. And we had
every reason to think that the macroeconomics of 1981 would be better yet.
But those expectations were far from rational. What has happened since
has been neither normal nor science. Who would say that the macro-
economics of 1991 is demonstrably better than the macroeconomics of 1972?1
Not I, nor, I surmise, Blanchard. It turned out that we macroeconomists
spent the better part of two decades arguing-often from behind ideologi-
cal barricades (a point the authors in this volume delicately ignore)
-whether wages and prices move instantly to clear all markets. Not why
they do not, which is a good question, but whether they do. That, in retro-
spect, and probably even in prospect, was ridiculous.
1 The Conference Theme
Laidler, Blanchard, and Fair articulate a fairly clear, common theme in

their chapters. Their message is that we wasted a great deal of time and
189
energy arguing about theoretical basics (Blanchard) and that it is high time
we got back to empirics (Laidler, Fair).
Blanchard writes (p. 127) that "by being in the 'back to basics' mode,
the field has developed mostly according to its own internal logic, and we
have shunned external stimulus too much."
Laidler observes (p. 103) that "the new-classical economists did not
advance an alternative set of empirical hypotheses .... Rather they put
forward an alternative theoretical framework whose superiority they
defended ... with reference, not so much to superior empirical content, as
to its conformity with the theoretical principles of equilibrium modeling."
There is no point in providing a similar quotation from Fair, for Ray's
entire chapter carries this message. Let me just state, instead, that I agree
entirely.
However, those who endorse the pre-eminence of empirical evidence
over theory, as I do, must reckon with the Phillips curve episode of the
1960s. Laidler described it exactly correctly. The position of mainstream
Keynesian economists in, say, 1969 was that the Friedman-Phelps argu-
ments for a vertical long-run Phillips curve were theoretically correct but
that the empirical evidence said otherwise-and we should stick with the
empirical evidence until it was proven wrong. We did; and it was. So those
who placed their bets on a priori theorizing over empirics won that round.
Advocates of real business cycle theory maintain that its eventual triumph
will be a second case in point. Blanchard, Laidler, Fair, and I implicitly
take the other side of this bet. (In fact, I'll give odds.) But we should re-
member that it's a bet and duly note that our side was wrong once before.
There is another aspect of these debates-in addition to and related to
the ideological component-that should be brought out; but none of the
chapters have done so. Once a few positive issues (like the interest elas-
ticity of money demand) were resolved, the Keynesian-monetarist debate
was largely over normative, policy-related issues. Should the authorities
pursue an activist stabilization policy or follow nonreactive rules? Should
monetary or fiscal policy be the chief instrument of demand management?
Should policy makers (and their advisers) worry most about the short-run
unemployment problem or the long-run inflation problem?2 It is almost a
foregone conclusion that questions that begin with the word should can
never be given a definitive answer by research results alone. At minimum,
value judgments and political considerations enter. So these questions con-
tinue to be debated. That fact is no condemnation of economics as a science.
The new-classical debate was, in stark contrast, largely over positive
issues. Do all markets clear (approximately) instantly? Is anticipated
money neutral? Do shocks to aggregate demand principally move real
DEJA VU ALL OVER AGAIN 191
output or the price level in the short run? Are economic fluctuations
dominated by mainly demand or mainly supply shocks?3 This second list
of questions is not about what policy makers should do but about how
the economy works. Such questions ought to be resolvable by economic
research-by the process of normal science-and perhaps by now have
been. The fact that such issues remained controversial for so long is a
strong condemnation of our profession.
One final important point has not been brought out in the discussion but
should be. So-called new-Keynesianism has come in for a great deal of
criticism, but little praise, at the conference so far. One main reason is
its starkly nonempirical flavor. It is difficult even to imagine taking most
of these models to the data; and few (but not no) new-Keynesians seem
so inclined. But I think this criticism is unfair, for it entirely ignores
the historical context. New-classical counter-revolutionaries sacked the
Keynesian temple not because its empirical content was found wanting,
though some falsely argued that point,4 but because its theoretical
underbelly was so soft. The challengers did not offer superior empirical
predictions, but rather models that were smaller, cuter, and more consist-
ent with maximizing behavior. It was therefore rational for younger
economists seeking to defend the Keynesian tradition to work on strength-
ening the theory, not the empirics-which is precisely what they did.
Were they misguided? Consider this metaphor. You are the commander
of Fort Keynes, which guards the intellectual frontier from dangers
believed to lurk principally on the empirical east side. Though your fortress
is far from beautiful, it is functional-and functioning. You are aware of
both homely aspects and structural weaknesses in the consumption, invest-
ment, and money-demand towers arrayed along the east side, where most
of the fortifications have been built. And you are working hard to repair
them. In the meantime, you have allowed yourself to live with meager
defenses on the theoretical west side, believing that there is little danger of
attack from that quarter. Then events prove you wrong. Enemies from the
theoretical west mount an attack whose ferocity was previously unimagin-
able. Soon they are scaling the walls. You suffer heavy losses for a while,
but then fresh troops arrive. My question is simple: Where do you send
them? That, I believe, is the story behind new-Keynesian economics. It
both explains and rationalizes its nonempirical flavor.
2 A Time Line of Macro Models
Let me now try to illustrate the main message of the conference by turning
to the achievements, or rather lack thereof, of what is here being called the
second debate. To do so, consider the following simple macro "model;" it

is really no more than a schema, a way to organize thought:
yd=D(P,eb e2,'" ) (1)
yS = S(P, Ub Uz, ••• ) (2)
tV = aPe + A(yd _ yS) (3)
In this schema, equation (1) is the aggregate demand curve relating real
GNP negatively to the price level; the e's are demand shocks such as money,
government purchases, tax rates, and the like. Money presumably enters
the model as MIP, so the derivative Dp indicates (with sign reversed) the
strength of monetary impulses. Equation (2) is the aggregate supply curve,
whose positive slope derives from some sort of nominal rigidity (fixed
money wages, for example) and which can be perturbed by various
supply shocks, the u's. Equation (3), a Phillips curve, summarizes the evolu-
tion of the money wage, here modeled as a state variable. This schema is
incomplete since, among other things, it does not specify how P evolves (it
could, for example, have its own dynamic equation) nor how y is deter-
mined when yS is not equal to yd (Keynesians usually specify y = yd). But it
suffices to make my points.
Table C-1 presents a "time line" of macro models that encapsulates the
two debates. History here begins in the mid- to late 1960s. Reflecting my
age,5 the left-hand column is called "early" Keynesianism. Monetary
effects on aggregate demand were not denied, but they were believed to be
empirically minor (Dp small). Demand shocks from other sources-such as
"autonomous spending" or fiscal policy-were, on the other hand, believed
to have large variances. The aggregate supply curve was thought to be
quite flat, perhaps even horizontal, in the short run (Sp large), but it was not
thought to shift much over the business cycle (var(u) = 0). Finally, we come
to the Phillips curve. As previously noted, early Keynesians cited empirical
evidence that the sum of the coefficients of lagged inflation-which they
mistakenly identified as expected inflation, pe-was considerably less than
unity.6 These expectations were modeled as a mechanical distributed lag
("adaptive" for short). And the coefficient A was estimated-I repeat,
estimated, not assumed-to be quite small, indicating sluggish adjustment
of wages and prices to excess demand.
The second column of table C-1 represents the monetarist challenge,
and my main point-it is also Laidler's-is that the challenge was entirely
empirical. Monetarists argued that the effect of money on aggregate
demand (Dp) was much larger than Keynesians allowed. Some of the
zealots even argued that fiscal impulses were inconsequential, so that
TableC-1. A Time Line of Macro Models
Early Early Real Business New/Old

Keynesian Monetarist New-Classical Cycle Consensus
Dp Small Large Large Zero Sizable,

but erratic
var(e) Large, Large, Large;M Small Large;G
especially mostly M andM
C,J,G
Sp Large ? Zero if Zero Usually
(infinite?) anticipated large
var(u) =0 =0 Small Large Usually
small,
sometimes
large
a «1 =1 =1 =1 =1
X Small Small ao ao Small
pe "Adaptive" "Adaptive" Rational Rational Rational?
monetary impulses were the dominant source of macroeconomic instab-

ility. Here too their argument, though weak, was primarily empirical.
Like the Keynesians, the monetarists pretty much ignored supply shocks
(var(u) = 0). Since it is not clear that there was a "monetarist position" on
the slope of the aggregate supply curve, I have left a question mark in that
position, inviting Laidler or Michael Parkin to fill it in. Finally, an impor-
tant monetarist innovation was the insistence that a had to be 1.0. Other-
wise, monetarists more or less accepted the Keynesian position on the
Phillips curve-including, importantly, the small value of A.
The resolution of the Keynesian-monetarist debate, or rather the posi-
tive aspects thereof, can be thought of as a process of normal science.
Evidence accumulated that Dp was in fact sizable, that fiscal policy did
matter, and that a really was approximately unity. The monetarists won
some and lost some; in the process, Keynesianism became less crude and
monetarism became less monotheistic. These empirical debates were
essentially over when the "second debate" burst on the scene.
Those of us old enough to remember the early 1970s will recall that new-
classicism was originally dubbed "Monetarism II." My time line shows
why. The early models adopted the Quantity Theory of Money, thereby
elevating M to pre-eminence as a determinant of aggregate demand (Dp

large, fiscal shocks small). This, however, can and should be interpreted
allegorically; it was not essential to the new-classical approach and I will
not belabor it. The key differences all came on the supply side. Supply
shocks were recognized-the Lucas supply function did, after all, have a
disturbance term-but I think it fair to say that they were not emphasized.
The monetary misperception models, like their Keynesian and monetarist
forebears, were demand-driven.
Where new-classical economics departed radically from the macro-
economics that came before it was in its view of the Phillips curve. Expec-
tations were modeled as "rational," meaning consistent with the model,
and Awas asserted to be extremely large, if not infinite. Notice that the first
of these was not an empirical statement at all; no evidence for rational
expectations was offered, as if none were needed. The second innovation
was worse: It was patently antiempirical, contradicted by mountains of
empirical evidence. Consistent with this view, the aggregate supply curve
was assumed to be vertical for anticipated changes in money. Robert Barro
subsequently offered empirical evidence to support this assumption, but his
evidence did not withstand the intense scrutiny it received.'
After about a decade, monetary misperception models passed from the
scene and were replaced by real business cycle models, the fourth column
of table C-l. Taken literally (which is probably not appropriate), these
models deny any role for money as a determinant of real aggregate
demand (Dp = 0) and denigrate the importance of demand shocks more
generally (var(e) small). Need I point out that these assumptions are not
empirically motivated?
Since misperceptions are no longer a central issue, real business cyclists
simply adopt a vertical aggregate supply curve (Sp = 0). But this supply
curve is subject to substantial shocks, even at business cycle frequencies
(var(u) large). There is considerable controversy over the extent to which
this last aspect of the model is empirically based. Finally, RBC models
tacitly adopt the same Phillips curve specification as the early new-classical
models. In particular, that means accepting the nonempirical view of
expectations and the antiempirical view of price adjustment. Is it any
wonder that Laidler and Fair question the extent to which real business
cycle analysis is empirically based?
In the final column of table C-l, I am bold (foolish?) enough to
announce a new, post-RBC consensus. The title is meant to suggest that
this "new" consensus is not very different from the consensus we almost
had reached in 1972. In this new and unarticulated model, large effects of
monetary policy are accepted, but instability in money demand makes
them erratic. Both fiscal and monetary shocks are taken to contribute
significantly to the variability of aggregate demand. The aggregate supply
curve is usually taken to be quite flat in the short run, an exception being
when the economy is pressing up against capacity constraints. Supply
shocks, we learned in the 1970s, can on occasion be extraordinarily large
and disruptive. 8 But, for normal quarter-to-quarter economic fluctuations,
the contemporary consensus is that supply shocks are quite small.
And then there is the Phillips curve, which, despite much bad-mouthing,
has proven to be a remarkably sturdy empirical regularity once supply
shocks are appended. The contemporary consensus is that IX = 1 and A is
small, just as in the second column, and that expectations are probably
rational. I use the qualifying adverb because, while most economists
nowadays accept the rational-expectations hypothesis, the empirical
evidence in its favor is at best weak and at worst damning. In fact, my
personal assessment is that the weight of the evidence is against it. 9
3 Back to the Future
In conclusion, I invite you to compare the last column to the first two
in table C-1-thereby skipping over both versions of equilibrium macro-
economics. Other than recognizing the potential importance of supply
shocks-which was forced on us by events, not by the new-classical revol-
ution-and the wholesale adoption of rational expectations, perhaps for
unsound reasons, the "new" consensus is the same as the "old" consensus
that had been reached in 1972. Was that progress?
Well, that's the past. What of the future? Can macroeconomists,
equipped with this new/old consensus, get back to doing normal science?
Five and a half years ago, when I wrote an evaluation of the "second
debate," I thought the answer was no; we were not ready yet.lO Now I think
the answer may be yes. If we can break the habit of shouting "Lucas
critique" or "rational expectations" or "unexploited profit opportunities"
in a crowded theater, then workaday macroeconomists may be able to
return to the task that was so unfortunately abandoned in 1972.
Notes
1. I pick 1972 over 1971 advisedly. By 1972 the "vertical-in-the-long-run" view of the
Phillips curve had won the day.
2. It will be noted that this last question presupposes the existence of a short-run tradeoff.
Keynesians and monetarists did not and do not dispute this.
3. The grammatically awkward adverbs are in deference to Benjamin Friedman's admoni-

tion that most supply shocks affect aggregate demand and conversely.
4. See, for example, Robert E. Lucas, Jr., and Thomas J. Sargent, "After Keynesian
Economics."
5. Professionally speaking, I was age -2 in 1969.
6. It remained for Thomas Sargent to correct the error that identified the sum of the lag
coefficients with the effect of anticipated inflation. See his" A Note on the •Accelerationist'
Controversy. "
7. See Robert J. Barro (1977). Ultimately definitive criticisms were offered by Frederic
Mishkin (1982) and Robert J. Gordon (1982).
8. I must correct the mistaken intellectual history offered by Blanchard. He suggests that
supply shocks were not successfully incorporated into the Keynesian paradigm until "the late
1970s and early 1980s." In fact, this had been done by the first months of 1974, as I showed in
"The Fall and Rise of Keynesian Economics" (1988, note 16, p. 282). Given that the first
supply shock hit in October 1973, that was a very short lag.
9. See, for example, the summary of empirical results in Michael Lovell (1986).
10. Alan S. Blinder, "Keynes After Lucas," (1986).
References
Barro, Robert J. 1977. "Unanticipated Money Growth and Unemployment in the

United States." American Economic Review 67: 101-115.
Blinder, Alan S. 1986. "Keynes After Lucas." Eastern Economic Journal 12: 209-
216.
- - . 1988. "The Fall and Rise of Keynesian Economics." The Economic Record
64: 278-294.
Gordon, Robert J. 1982. "Price Inertia and Policy Ineffectiveness in the United
States, 1890-1980." Journal of Political Economy 90: 1087-1117.
Lovell, Michael. 1986. "Tests of the Rational Expectations Hypothesis." American
Economic Review 76: 110-124.
Lucas, Robert E., Jr. and Thomas J. Sargent. 1978. "After Keynesian Macro-
economics." In After the Phillips Curve: Persistence of High Inflation and High
Unemployment, Conference Series No. 19, Boston, MA: Federal Reserve Bank
of Boston.
Mishkin, Frederic. 1982. "Does Anticipated Monetary Policy Matter? An Econo-
metric Investigation." Journal of Political Economy 90: 22-51.
Sargent, Thomas J. 1971. "A Note on the 'Accelerationist' Controversy." Journal
of Money, Credit and Banking 24: 721-725.
Commentary: Business Cycle Developments and
the Agenda for Business Cycle Research
by Herschel!. Grossman
Important changes in the agenda for research into the business cycle in
modern industrial countries have occurred over the past two decades.
These changes have included the end of the single-minded concern of
researchers with the determination and management of aggregate demand
and a new emphasis instead on the cyclical implications of shocks to avail-
able resource endowments and on cyclical aspects of regional and sectoral
economic development.
Prior to the mid-1970s, conditioned by the cyclical experience of the
preceding hundred years and especially by the experience of the Great
Depression, we took it for granted that in modern industrial economies,
cycles in the economic activity reflect fluctuations in the utilization rate of
existing endowments of labor and capital and that these fluctuations in the
utilization rate are proximately the consequence of fluctuations in nominal
aggregate demand. It seemed obvious then that, with the exception of the
damage associated with major wars, the possibility of sharp changes in
existing resource endowments was not something worth worrying about, at
least not in the context of modern industrialized economies, which had
conquered the old problems of famine and pestilence. In addition, although
we were concerned with problems of economically depressed regions and
industries, we regarded these problems as involving secular issues of econ-
omic growth and stagnation and as distinct from the agenda of business
cycle research. In sum, we believed that, if we were able to understand
fluctuations in nominal aggregate demand, then we would have understood
economic fluctuations, and that, if we were able to prescribe how to
stabilize nominal aggregate demand and, thereby, to stabilize aggregate
economic activity at the national level, then we would have largely solved
the problem of economic fluctuations.
Accordingly, the business cycle research agenda prior to the mid-1970s
focused on the determination of nominal aggregate demand, on the
relation between nominal aggregate demand and real aggregate demand,
and on the management of economic policy to stabilize nominal aggregate
demand. This research agenda generated a variety of distinct, but related,
controversies. One controversy concerned the source of fluctuations in
nominal aggregate demand-specifically, the importance of disturbances to
197
tastes and technology, as manifested in changes in the marginal efficiency

of investment, in the propensity to consume, or in liquidity preference,
relative to the importance of perverse shifts in monetary and fiscal policy.
Another controversy concerned the problem of rationalizing the apparent
failure of disturbances to nominal aggregate demand to be translated
immediately into changes in the level of wages and prices and the resulting
impact, at least in the short run, of disturbances to nominal aggregate
demand on real aggregate demand. This controversy probably was most
fundamental, because, in addition to its practical implications for business
cycle research, it raised questions about the logical structure of neoclassical
economics. A third major controversy concerned the potential for actively
using monetary and/or fiscal policy to stabilize nominal aggregate demand.
Although these three controversies remain incompletely resolved and
continue to be the subject of active theoretical and empirical research, they
no longer monopolize the agenda for business cycle research. The agenda
now also emphasizes issues other than the determination of aggregate
demand. It seems to me that we can identify three historical developments,
beginning with the oil price shock of 1973, that have been largely respon-
sible for this evolution of the business cycle research agenda. First, the oil
price shock of 1973 suggested that even modem industrialized economies
could face sharp changes in available resource endowments and that it
might be relevant to think about such changes as a possible source of eco-
nomic fluctuations in modem industrialized economies. Second, in the
United States the recessions of 1974-1975 and 1981-1982, as well as the
last recession, were surprisingly uneven in their regional and sectoral
impact, with the implication that the business cycle research agenda should
include questions about the determination of regional and sectoral econ-
omic activity. Third, the experience of recent years has suggested-the
cyclical expansion and contraction of the late 1970s and the early 1980s as
well as the last recession notwithstanding-that the Federal Reserve, and
probably the monetary authorities of the other major industrialized
countries as well, is now both willing and able to conduct monetary policy
in such a way as to preclude large fluctuations in nominal aggregate
demand.
Let me elaborate on the last observation. In the mid-1960s President
Johnson claimed that recessions were a thing of the past. Given the then
ongoing economic expansion and the evolving belief, which probably was
correct, that we had learned how to use monetary and fiscal policy to
achieve effective control of nominal aggregate demand, Johnson's claim
seemed plausible at the time. In fact, one of my senior colleagues at that
time advised me that interest in business cycle research was dying and that
BUSINESS CYCLE DEVELOPMENTS 199
as a young assistant professor I would have little chance of making a repu-

tation by working on Keynesian problems like the determination of aggre-
gate demand.
As we know, subsequent cyclical experience made President Johnson's
claim seem overly optimistic and served to restimulate interest in business
cycle research. This experience included the inflationary surge of the late
1960s-which apparently resulted mainly from the political imperative
to try to finance the war on poverty and the war in Vietnam with both
minimal borrowing and minimal explicit tax increases-followed by the
recession of the early 1970s and the inflationary surge and recession of
the mid-1970s-both of which apparently resulted from the combination
of the oil price shock of 1973 and a policy response to the oil price shock
that probably mitigated its impact but aggravated its cyclical effects-and
the unprecedented inflation of the late 1970s, which apparently resulted
mainly from an unperceived increase in the natural rate of unemploy-
ment, followed by the recession of the early 1980s. But, this experience
notwithstanding, developments since the late 1970s, and two important
events in particular, have now made the view expressed by Johnson again
seem plausible, at least if we interpret Johnson as having in mind cycles in
the national economy associated with fluctuations in nominal aggregate
demand.
The first of the two events that I have in mind was the sharp reversal
of monetary policy midway through the Carter administration, marked
dramatically by President Carter's appointment of Paul Volcker to replace
his earlier appointee, William Miller, as chairman of the Federal Reserve
Board, together with similar policy reversals by the Callaghan Labor
government in Britain and by the Mitterand Socialist government in
France. These developments showed that by the late 1970s even the politi-
cal left was willing to accept the idea that resource endowments are a
relevant constraint on aggregate output and that, to paraphrase Prime
Minister Callaghan, "we cannot spend our way to prosperity." Given this
new consensus, it would seem reasonable to hypothesize that the sort
of policy decisions that in the past have generated unsustainably rapid
expansions of nominal aggregate demand and resulting cycles of inflation
and recession will not be a problem in the future.
The second event that I have in mind was the apparent confirmation for
this hypothesis provided by the performance of the American economy
during the 1980s. The Volcker Fed, after apparently having erred in
the early 1980s on the side of disinflating more rapidly and with a
sharper recession than either the Reagan administration or the Fed itself
wanted, was then able to manage a sustained noninflationary expansion of
aggregate demand through the rest of the decade. Only the mildest of
recessions as measured at the national level, probably resulting mainly
from a contractionary fiscal impulse, interrupted this expansion. And the
Fed seems sufficiently cautious in its reaction to this last recession that
we have little reason to fear an unsustainably fast expansion of nominal
aggregate demand that would lead to a major new cycle of inflation and
recession at the national level.
But a single observation does not provide a powerful test of an hypoth-
esis. Were the 1980s merely a lucky drawing from an unchanged economic
and political structure? Or are we actually living in a new era in which we
have no need to worry about major cyclical fluctuations in nominal aggre-
gate demand at the national level? To try to answer this question, we can
begin by trying to account for the two events that I have emphasized-the
policy reversal of the late 1970s and its apparent consequence, the macro-
economic stability of the 1980s. Two explanations, which are potentially
complementary, occur to me as good candidates for such an accounting.
The first explanation emphasizes the effects of the accumulation of new
knowledge about the economy during the 1970s. This new knowledge
included both new theories-especially the natural-rate hypothesis, and
also the rational-expectations hypothesis-as well as new empirical
evidence-especially, evidence that the natural rate of unemployment is
not a constant and, in particular, that by the late 1970s the natural rate of
unemployment was much higher than in the 1950s and 1960s. According to
this explanation, policy makers, who by the 1960s were confident of their
ability to manage nominal aggregate demand, learned during the 1970s the
limitations on the ability of aggregate-demand management to deliver high
levels of economic activity. The inflationary surges of the late 1960s, mid-
1970s, and late 1970s were all plausibly part of this learning experience.
The second explanation emphasizes that since the late 1970s the major
industrial countries have seen no exogenous economic or political disturb-
ances of the type that have led to inflationary monetary policies in the past.
Most obviously, there have been no adverse disturbances like the oil price
shocks of the 1970s. In addition, the large increase in military spending in
the 1980s, in contrast to spending for the war in Vietnam, had wide political
support, support that precluded the need to risk an inflationary surge to
finance it. The more recent easing of international tensions has reduced
further the strain on resources that potentially would generate political
pressures for inflation.
But, we have no guarantee that the future will not bring less favorable
developments. Most basically, there has been no obvious change in the
structure of preferences and political constraints underlying economic
BUSINESS CYCLE DEVELOPMENTS 201
policy, and especially monetary policy. It is one thing for policy makers to
know how to manage aggregate demand to mitigate fluctuations in aggre-
gate economic activity. It is another thing for them to give priority to this
objective. Even if policy makers have learned the dangers of under-
estimating the natural rate of unemployment and of trying to exploit the
short-run Phillips curve to drive unemployment below its natural rate,
without a change in preferences and political constraints we cannot
preclude the possibility that some future economic or political develop-
ment will cause policy makers knowingly to risk producing a possibly
unsustainable expansion of nominal aggregate demand and a consequent
cycle of inflation and recession. For example, can we be sure that the large
and growing fiscal obligations implied by current budget deficits and other
currently legislated spending commitments will not undermine future
monetary discipline?
Returning to the current agenda for business cycle research, it seems to
me that we still have to be concerned about fluctuations in nominal aggre-
gate demand. But the present priority does not seem to be to refine further
our knowledge either of the determination of nominal aggregate demand,
or of the relation between nominal aggregate demand and real aggregate
demand, despite the fundamental importance of this issue for economic
modeling, or of how to manage economic policy to stabilize aggregate
demand. Rather the experiences of the past two decades suggest that the
present priority, in addition to studying the cyclical implications of shocks
to available resource endowments and the cyclical aspects of regional and
sectoral economic development, should be to improve our understanding
of the political-economic structure underlying monetary and fiscal policy. I
leave a discussion of the directions for such a research agenda for another
occasion.
Acknowledgments
I am grateful to William Poole for constructive comments.

Commentary: Where Do We Stand?
by Michael Parkin
Albert Einstein is reputed to have said:

Creating a new theory is not like destroying an old barn and erecting a
skyscraper in its place. It is rather like climbing a mountain, gaining new and
wider views, discovering new connections between our starting point and its rich
environment. But the point from which we started still exists and can be seen,
although it appears smaller and fonus a tiny part of our broad view gained by
the mastery of the obstacles on our adventurous way Up.l
The mountain for macroeconomics is a particularly tough one. (Actu-

ally, it has three peaks-growth, inflation, and business cycles-and they
keep getting separated from each other's views by periodic dense clouds.
Here, I plan to remain on the business cycle's peak.) We stand on the side
of this mountain. In this position we are doing the best we can to achieve
macroeconomic stability and also making a series of attempts to climb
higher-to gain wider views. A variety of routes-research programs and
efforts-have brought us to the place where we now stand, and we are
exploring the promise of new routes-new research programs-the ultim-
ate utility of which we'll know only when we actually get to a higher
vantage point and can see them all in perspective-when there has been
another major theoretical advance.
The first debate was about how we got to here. Although it was a
debate, it was a debate among the crack mountaineers who were leading
the way up and their immediate followers. The rest of us watched, some-
times in awe and admiration, sometimes in dismay. But most of us made
our personal climbs up well-worn paths.
My own climb (as that of many of my generation) started out along the
traditional Keynesian track. Educated in England in the 1950s and early
1960s, I knew that the interest elasticity of investment demand was zero and
that the interest elasticity ofthe demand for money was infinity. These "facts"
had been established by solid empirical evidence, most notably survey data
on investment intentions and the experience with attempting to push interest
rates down to 2% in the immediate postwar years and finding it impossible
to do so. The "textbook" on which I was brought up was the Radcliffe Re-
port (1959), which endowed these views with an additional aura of authority.
Given this "knowledge," the IS-LM apparatus, while correct, was irrel-
evant. With a vertical IS curve and, lest there be any doubt about that fact,
202
WHERE DO WE STAND? 203
with a horizontal LM curve as well, the aggregate demand curve (the equi-
librium locus in price-output space) was vertical, so the entire apparatus
was irrelevant. All one needed to determine aggregate economic activity
was the 45-degree cross model of the consumption function and auto-
nomous expenditure.
Arnold Zellner has recalled that in the early 1960s while a member of
President Kennedy's Council of Economic Advisers, James Tobin was
asked whether he used large-scale econometric models to form his advice
to the President. 2 Tobin responded that he did not. Instead he used an
envelope on which he drew the 45-degree cross model to generate his
multiplier results. Remarked Tobin, "I don't believe the results but at least
I know what I am doing." The only difference between the 45-year-old
Tobin and the 25-year-old Parkin was that I did believe the results and
didn't know what I was doing!
Many of us left the traditional Keynesian track sometime in the 1960s.
For me the departure was dramatic. It was in the fall of 1967. I had just
arrived at the University of Essex, where I met David Laidler, fresh from
Chicago and Berkeley. He was as fast on his feet, well read, and well
equipped to do verbal battle as anyone I had ever met. After one or, at
most, two sessions in which I resoundingly lost all the arguments-because
I was ignorant of some crucial, by then well-known, pieces of empirical
evidence, especially Laidler's own work on the demand for money function
and Jorgenson's work on investment demand-I quickly recognized that
the IS curve indeed was not vertical and that the LM curve was not hori-
zontal. The IS- LM apparatus was not only elegant but useful.
Given this view of the IS-LM curves, the aggregate demand curve
sloped downward and to determine eqUilibrium output we needed an
aggregate supply curve. Thus I was ready to buy into the IS-LM, AD-AS
Phillips curve mainstream. Actually, accepting the Phillips curve part of
the story was really easy because I had already been convinced (on the
basis of the same ignorance that led me to my earlier conclusions on the IS
and LM curves) that the Phillips curve was the most robust of the empirical
relationships available!
As the mainstream flowed and the expectations augmentation of
the Phillips curve was absorbed into it, a great deal of constructive work
was done. Though the mainstream flowed strongly, it did not flow up
the mountain. (Karl Brunner was fond of reminding us-usually when
putting down some absurd statement based on factual error-that water
does not flow up hill, not even in Switzerland.) As the mainstream flows
it fills in the cracks and sorts out the details. That's what was happening in
the late 1960s.
But while the aggregate demand-aggregate supply-expectations aug-

mented Phillips curve mainstream was flowing, a few macroeconomists
were looking for ways of "gaining a wider view." Robert Lucas (the lead-
ing macro mountaineer of our generation) did most of this work (helped
of course by some extremely able colleagues). Lucas' work had two distinct
effects.
First, it enabled the near completion of the old way up the mountain.
The incorporation of rational expectations, combined with the institutional
features of overlapping wages on the supply side and the more sophisti-
cated, dynamic IS-LM model on the aggregate demand side, gave the
mainstream a respectable completeness on its own terms and enabled it to
form the basis of Cowles' style econometrics that has given us medium-
scale econometric models developed by, among others, Ray Fair and John
Taylor. The completion of the framework also informed and continues to
inform the selection of variables to use in time series/transfer function
work of the type described by Arnold Zellner in this volume. It yet further
enabled the tight formulation of the early versions of new classical macro-
economics and to the speedy rejection of those models.
And this is where we stand today. This is what we "know" about busi-
ness cycles. We know the rich fabric of fact displayed by Victor Zarnowitz
in this volume. We have a coherent rationalization of those facts in the
form of mediumscale econometric models or a series of single equation
(or V AR) time series models. And we think about these models with
the aggregate demand-aggregate supply-expectations augmented Phillips
curve framework.
But Lucas' work had a second effect. It offered some glimpses of what it
might be like up a particular mountain path. That path is not classical
macroeconomics made new. It is not any particular kind of economics. It is
a research method and one that could take us anywhere. It was given
important impetus by Kydland and Prescott (1982) in their work on real
business cycles, but it is not real business cycle theory.3 It has, however,
become associated with real business cycle theory and new-classical
macroeconomics, and so the two do get confused.
To clarify the distinction I want to quote (fairly extensively) from an
unpublished paper by Jean Pierre Danthine and John B. Donaldson (1991).
I use Danthine and Donaldson for three reasons. First, they are typical of
young macroeconomists who have no political or ideological position. They
are doing what they see as constructive work and the approach they are
taking to their work was, in my view, badly misrepresented in Olivier
Blanchard's paper on the second debate. 4 For me that was a one-sided and
misleading presentation of the nature of today's debate, and in part I want
to present what I believe to be a corrective to Olivier's characterization of

it. Third, I quote from this paper so extensively because I agree with it.
First, Danthine and Donaldson suggest:
The RBC [real business cycle] label is unfortunate as the major contribution of
this body of research is not a denial of any substantive role for money in
explaining business cycle phenomena, but rather the establishment of a new
research methodology for the study of the business cycle. The first component
of this proposed methodology is an "empirical reassessment" which calls for a
more systematic and complete statistical characterization of the economic
fluctuations to be explained. The second component is the recourse to what has
been called "quantitative theory"; i.e., the building of small, intuitive, comput-
able, general equilibrium dynamic models which can be evaluated not only
qualitatively but also quantitatively in terms of their ability to replicate the basic
business cycle stylized facts.
I want to emphasize the order in which the material is placed by the new
macroeconomists. Macroeconomics is fundamentally an empirical enter-
prise. Elegance is not its objective. We have not gone back to basics so that
we can build beautiful models. We've gone back to basics so that we can
build models that work and explain the phenomena of interest.
Danthine and Donaldson go on to say that three important conse-
quences for the whole of macroeconomic theory are likely to flow from this
research program:
First, ... a move toward a more inductive approach to macroeconomic research,
with the accent being placed on a more systematic qualitative and quantitative
description of the facts to be explained. Second, ... developing "quantitative
theory" will spread to other applications, mitigating the importance attributed
to purely qualitative results. Thirdly, added research discipline will come from
the view that a successful theoretical model must be one which not only explains
the stylized facts at its focus but which is also broadly consistent with other
accepted aspects of reality. Models calibrated successfully to explain one fact (or
set of facts) but which do so while contradicting other accepted empirical
findings will not be accorded much value.
They then go on to talk about the research strategy:
A natural strategy for the execution of such an ambitious program is first to
examine well known existing dynamic models to determine how well they
perform. Such logic fully justifies the attention given the stochastic growth
paradigm-a Walrasian model without money-by RBC authors to date.
Next they offer some instructive comments on the reception that these
efforts have received:
These initial attempts to construct a theory have, however, generated substantial
misunderstanding and dogmatic posturing while revealing how we macro-
economists have been accustomed to think. Indeed, the idea of even proposing
and "testing," in the above sense, a purely Walrasian model of the cycle has
generated heated objections, sometimes aggravated by the misguided claim that
these initial attempts conclusively demonstrated that business cycle phenomena
were nothing more than the optimal reaction of rational agents to exogenous
productivity shocks.
My final quotation from the Danthine-Donaldson paper is one that
shows how they view the enterprise and what they see as the possibility of
progress:
In reality, RBC methodology is by nature ideologically neutral in the sense that
it prefers the model or set of models that is (are) best able to replicate the
stylized facts independent of the hypotheses underlying it (them). The best RBC
model may thus ultimately be a demand-driven money model with substantial
non-Walrasian features. [My italic.] Such a convergence will ultimately occur,
however, not on the basis of prior views but as the outcome of a process of build-
ing increasingly richer models and confronting them with an increasingly richer
set of stylized facts .
. . . It is ... premature ... to claim victory of one model paradigm over
another given the modest set of facts which current models are able to repli-
cate .... However, a clear achievement of the RBC literature has been to free us
to reconsider what we know about the business cycle.
I believe the foregoing statements by Danthine and Donaldson
represent the point of view of the vast majority of young macroeconomists
working the field today. They certainly represent the views of my
colleagues who work in this area and of all the young macroeconomists
with whom I have discussed the subject in the past several years. Olivier
Blanchard's (1991) characterization of the research program on which
these young economists are embarked is severely at odds with the picture
that I have painted and, I think, is wrong.
The approach has already led to the development of models with enor-
mous diversity. There are papers that explore a variety of alternative types
of technology shocks,S international aspects of the business cycle,6 alterna-
tive interpretations of the Solow residual-seeking to understand the poss-
ible deeper sources of shocks7-and non-competitive market structures. 8
This new method of quantitative theory joins the other research methods
in macroeconomics-the Cowles method, qualitative models, and case
or episode studies-the outstanding example of which is Friedman and
Schwartz (1963) but a modern example of which is Romer and Romer
(1990).
If the new method of macroeconomics is doctrinally neutral, what has
become of the new-classical and new-Keynesian labels that are at the
centerpiece of the second debate and of Olivier Blanchard's chapter?

These classifications are orthogonal to the alternative research methods.
The distinctions between new-classicals and new-Keynesians are distinc-
tions based on views about which abstractions might turn out to be useful
and which might not. New-classical macroeconomics downplays problems
of coordination. It does not deny that they exist, and it does not deny that
they are important for some purposes. But it speculates that they are not
important for explaining the fluctuations in the economy. Instead, it
conjectures that intertemporal substitution is at the heart of aggregate
fluctuations. This may turn out to be a brilliant short-cut giving a huge ratio
of payoff to machinery (to paraphrase Blanchard). Or it may tum out to be
wrong. The research methods at our disposal-the traditional methods
supplemented by the new quantitative theoretical method-will help us
figure out whether it indeed is a brilliant simplification.
New-Keynesian macroeconomics accepts the importance of intertem-
poral substitution but adds coordination failures and rigidities to the list of
important factors influencing the fluctuations in the economy.9
New-Keynesians and new-classicals alike use all four of the research
methods that I have identified. Increasingly, new-Keynesian ideas are
being explored using quantitative theory. Typically they are not being done
by new-Keynesians (at the present time), but no doubt as the method begins
to payoff, new-Keynesians will start to use them. As we saw in the second
debate new-Keynesians, for the most part, are using qualitative theory (and
partial eqUilibrium theory at that) and thus are not really addressing the
quantitative issues that Ray Fair would have had them address. But nor
are new-classicals, with their use of quantitative theory, addressing quan-
titative issues in the traditional way. Thus far quantitative theorists have
concentrated on explaining data described in moment's space-means,
variances, covariances, and autocovariances. Traditional macroeconomists
have worked with time series and been concerned to make predictions
about next quarter's GNP, not the variance of GNP over some long period.
Concentration on moment's space is consistent with the broader view that
policy is a process, not an event, and that policy discussions can only use-
fully assess the consequences of employing well-defined policy "rules."lO
This view of policy, embraced by new-classical economists, is compatible
with the moment's characterization of data.
But it is not consistent with a view held by a large number of economists.
They want to get back to time-series predictions. Will new-classical (or for
that matter new-Keynesian) economists ever get around to couching their
predictions in terms of the more natural time series? I believe it is too early
to say. If they don't, it will be either because we all recognize that "policy is
a process, not an event"l1 and there is, therefore, no need to operate in the
time domain, or because other methods have superseded quantitative
theory.
Finally, what about policy itself? What does the current state of know-
ledge imply for policy choice? First, there does not exist a feasible commit-
ment technology that permits us to establish policy rules. All policy is
inevitably discretionary. In this I agree with Ben Friedman's assessment.
Second, given where we stand, policy can do no better than use the tradi-
tional econometric models rationalized through the aggregate demand-
aggregate supply-expectations augmented Phillips curve view of the world,
in its attempt to steer a course between recession on the one side and
inflation on the other. Third, given the political and institutional con-
straints on policy making, it is all but certain that policy actions will be
imperfect and will from time to time exacerbate the very cycles they seek
to smooth.
The more interesting and important policy questions, it seems to me, are
not those concerning the details about which instrument to assign to which
target, with what intensities and timing, but those concerning the way in
which policy makers react to the evolving economy and the ways in which
alternative institutional arrangements- central bank law and operating
procedures-operate to stabilize the economy. I recommend strongly to
our hosts and organizers that they consider this topic as a suitable one for a
future St. Louis conference.
Acknowledgments
I could not have prepared these comments without the help of my

colleagues Jeremy Greenwood, Andreas Hornstein, and David Laidler.
Although I do not hold them responsible for what I say, I acknowledge my
debt to them.
Notes
1. Attributed to Einstein in a letter by Oliver Sacks to The Listener 88, No. 2279, 30
November 1972, p. 756. (I have not been able to track down the original Einstein quotation.)
2. A remark made at the Sixteenth Annual Economic Policy Conference, Federal Reserve
Bank of St. Louis, October 17 -18,1991.
3. See Kydland and Prescott (1982).
4. The reader is referred to Olivier Blanchard (1991), not the revised version of his paper
published in this volume.
5. Such as Greenwood, Hercowitz, and Huffman (1988).

6. Enrique Mendoza (1991).
7. Mary Finn (1991).
8. Hornstein (1991) and Danthine and Donaldson (1991).
9. I believe this to be a correct characterization of the new-Keynesian research program.
When I coined the term new-Keynesian in 1980 I used it to describe a set of models then
developed by Edmund S. Phelps, Stanley Fischer, and John Taylor, models with rational
expectations but Keynesian implications-see Parkin (1982). I recognize that coining a term
does not give ownership to the way it is used and claim no more than that the characterization
I've just given does accord well with the way people align on these issues.
10. I use rules here in the sense of systematic policy reactions, not fixed rules. All policy is
discretionary in just the sense that consumer choice is discretionary but can be characterized
as rule-based in just the same way that consumer choice can be characterized by a set of
decision rules (demand functions).
11. I attribute this phrase and the initial insight to Neil Wallace.
References
Blanchard, Olivier. 1991. "New-Classicals and New-Keynesians. The Long Pause."

Boston: MIT.
Danthine, Jean Pierre, and John B. Donaldson. 1991. "Methodological and Empiri-
cal Issues in Real Business Cycle Theory." Discussion paper, Universite de
Lausanne and Columbia University.
Finn, Mary. 1991. "Energy Prices, Capacity Utilization and Business Cycle
Fluctuations." Unpublished Paper, Department of Economics, Michigan State
University.
Friedman, Milton, and Anna J. Schwartz. 1963. A Monetary History of the United
States 1867-1960. Princeton, N.J.: Princeton University Press.
Greenwood, J., Z. Hercowitz, and G. Huffman. 1988. "Investment, Capacity
Utilization and the Real Business Cycle." American Economic Review 78: 402-
417.
Hornstein, Andreas. 1991. "Monopolistic Competition, Increasing Returns to Scale
and the Importance of Productivity Changes." University of Western Ontario.
Kydland, Finn E., and Edward C. Prescott. 1982. "Time to Build and Aggregate
Fluctuations." Econometrica 50: 1345-1370.
Mendoza, Enrique. 1991. "Real Business Cycles in a Small Open Economy."
American Economic Review 81: 797-818.
Parkin, Michael. 1982. Modern Macroeconomics. Toronto: Prentice-Hall.
Radcliffe Committee. 1959. Committee on the Working of the Monetary System
(1959). Radcliffe Report, Cmnd. 827, HMSO, London.
Romer, Christina D., and David H. Romer, 1990. "New Evidence on the Monetary
Transmission Mechanism." Brookings Papers on Economic Activity I: 149-213.

The Business Cycle: Theories and Evidence

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

The Business Cycle: Theories and Evidence

Uploaded by

Copyright:

Available Formats

The Business Cycle:

Theories and Evidence

Springer Science+8usiness Media, LLC

Economic Pc>icy Conlerence of the Federal Reserve Bank 01 SI. Louis

Copyright © 1992 Springer Science+Business Media New York

AII rights reserved. No pari of this publication may be reproduced,

Printed on ac;d-free pape,.

Set in 1O/I2pt Times Roman

Contributing Authors vii

President's Message by Thomas C. Melzer ix

Commentary by James H. Stock 73

Commentary by Ben S. Bernanke 113

Commentary by Arnold Zellner 148

SESSION III 159

Commentary: Whatever Happened to Contracyclical Policy? 179

CONFERENCE OVERVIEW 187

Commentary: Deja Vu All Over Again by Alan S. Blinder 189

Commentary: Business Cycle Developments and the Agenda for Business

Commentary: Where Do We Stand? by Michael Parkin 202

Ben S. Bernanke Herschel I. Grossman

RayC. Fair Victor Zarnowitz

dismal science" to one of optimism. The new generation of economists was

These proceedings, from a conference held at the Federal Reserve Bank of

What Is a Business Cycle?

Before gaining a genuine understanding of business cycles, economists

changes in economic activity, so that the NBER's dating methodology does

What Are the Sources of Business Cycles?

Despite the voluminous research on fluctuations in aggregate economic

the economy's productive capacity affects investment and vice versa. As

monetarist prescription for nonactivist policy or rules over activist policy or

macroeconomists to adopt a common framework, one that sees demand

How Might the Debate Be Resolved?

One of the overriding concerns among many conference participants was

What Are the Policy Implications?

generate shocks to the demand side, which often cannot be disentangled

The Bank's annual economic policy conference is organized by its Research

We find {the state of trade] subject to various conditions which are

confidence, -prosperity, -excitement, -overtrading, -convulsion, -pres-

1 The Emergence of Business Cycles in History and

1. 1 Early ObselVations and Interpretations

1.2 Popular Meaning

Unlike a theoretical concept such as that of equilibrium, a business cycle is

Britain, France, and Germany was accomplished by a complementary use

It may be easy to recognize some manifestations of a business cycle,

special increase in the production of fixed capital; a period of decline or a

coherence of numerous variables and their pronounced serial correlation).

1.4 Past Views on the Scope and Nature of Business Cycles

All descriptions and definitions of business cycles are guided in part by

of what is going on in the economy. The data available to the early

1.5 New Developments and Ideas

Momentous developments bearing on thinking about business cycles

models and lately experiencing some resurgence. The predominant type is

self-interest of individuals and more or less organized groups. These

1.6 Some Basic Reflections and Further Steps

2 The Changing DimenSions of BUSiness Cycles

2.1 The Historical Reference Cycle Chronologies of the NBER

Annual chronologies of "reference cycles," that is, business cycle peaks

Full Cycle Full Cycle

B. Shortest (S) and Longest (L) Durations, in Months

No. of Business Full Cycle Full Cycle

Period e Tto T PtoP S L S L S L S L

2.2 A Cautionary Note