View original document

The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.

M A R C H / A P R I L

1 9 9 5



7 7 ,



The FOMC in 1 9 9 3 and
1 9 9 4 : M o n e ta ry P o licy
in T ran sitio n
R e g u la tio n , M a rk e t
Structure an d the B a n k
F a ilu re s of the G re a t
D e p ressio n
U .S . O fficia l Fo recasts of
G - 7 Eco n o m ies, 1 9 7 6 - 9 0
An O u tsid e r's G u id e to
R e a l B u sin ess Cycle
M o d elin g


Tho m as C. M e lz e r
Director o f Research

W illia m G . D e w ald
A ssociate Director o f Research

Cletu s C. Coughlin
Research Coordinator and
Review Editor

W illia m T. G a v in


R. A lton G ilb e rt
D a v id C. W h e elo ck

C h risto p h er J . N e e ly
M ich a el R. P a k k o
P a tric ia S . P o lla rd
Macroeconom ics

D o n ald S . A lle n
R ich ard G . A n d erso n
Ja m e s B. B u llard
M ich a el J . D u e k e r
Jo sep h A . R itter
D a n ie l L. Thornton
P e te r Yoo

M ich e lle A . C la rk
K e v in L. K lie s e n
A d am M . Z a r e ts k y

Director o f Editorial Services

D a n ie l P. B ren n an
Managing Editor

C h a rle s B. H enderson
Graphic Designer

B ria n D. E b ert

Review is published six times per year by the Research
Department o f the Federal Reserve Bank of St. Louis.
Single-copy subscriptions are available free of charge.
Send requests to: Federal Reserve Bank o f St. Louis,
Pu blic Affairs D epartm ent, P.O. Box 4 4 2 , St. Louis,
Missouri 63166-0442.
The views expressed are those o f the individual authors
and do not necessarily reflect official positions of the
Federal Reserve Bank o f St. Louis, the Federal Reserve
System , or the Board o f Governors.

A rticles may be

reprinted or reproduced if the source is credited. Please
provide the Public Affairs Department with a copy o f the
reprinted materials.




rates during the Great Depression,
providing new insight into the effects
of governm ent policies and m arket
structure on the perform ance o f the
banking system.
Volum e 7 7 , N um ber 2

39 U .S . O fficial Forecasts of
G - 7 Eco nom ies, 1 9 7 6 - 9 0

The FOMC in 1 9 9 3 and
1 9 9 4 : M o n e tary P o licy
in Transition
M ichael R . P a k k o

The Federal Reserve’s m onetary policy
actions in 1994 might appear to represent
an abrupt departure from those o f the
previous year. Six highly publicized
increases in short-term interest rates
followed a period of relative stability in
short-term rates in 1993. M ichael R.
Pakko reviews the actions o f the the
Federal O pen M arket Com m ittee
(F O M C ), the Federal Reserve’s primary
policym aking body, over the last two
years. He views the FO M C ’s actions in
the context o f the Fed’s support for a
policy o f long-term price stability.

M ichael U la n , W illia m G . D ew ald and
Ja m e s B . B ullard

Decisions concerning monetary and fiscal
policy often depend in part on forecasts
o f the future course o f real activity. But
how accurate are those forecasts? Michael
Ulan, W illiam G. Dewald and Jam es B.
Bullard exam ine the relative accuracy
o f forecasts for inflation and econom ic
growth made by the U.S. governm ent
for G -7 econom ies from 1 9 7 6 to 1990.
They com pare the official forecasts to
those made by other m ajor private and
public groups, including the Organization
for E conom ic Cooperation and Develop­
m ent (O E C D ) and the Federal Reserve.

49 An O u tsid er's G u id e to R e al
B u sin ess Cycle M o d eling
Jo sep h A . R itter

27 R eg u la tio n , M a rk e t
Structure and the B an k
F a ilu re s of the G re a t
D ep ressio n
D a vid C. W h eelo ck

The bank and S & L debacle o f the 1980s
focused attention on the possibility that
government regulation and deposit insur­
ance can m ake depository institutions
more prone to failure. Branch banking
restrictions, for example, limit geographic
diversification, leaving banks m ore vul­
nerable to localized econ om ic distress.
David C. W heelock shows that state and
federal banking policies contributed to
interstate differences in bank failure

During the last decade, real business
cycle (R BC ) m odeling has w on a large
market share in m acroeconom ic research.
Though the econom ics are familiar, the
com putational techniques still appear
inaccessible to many econom ists. Joseph
A. Ritter exposes som e o f the “Secrets of
the RBC Temple” by describing— from
an outsider’s vantage— the specification,
calibration, solution and evaluation of
one such model.




M ich ael R. P a k k o is an econom ist a t th e Fed eral R e se rv e B a n k of St. Louis. K elly M. M orris provided research a ssistan ce.


The FOM C in
1993 and 1 9 9 4 :
M o n e ta ry Policy
in Transition

ening o f consensus regarding the ultim ate
goals and lim itations o f m onetary policy has
continued to develop: E conom ists and
Federal Reserve policym akers increasingly
agree that price stability should be the over­
riding long-run concern o f the central bank,
serving as a foundation for m aintaining
econom ic growth.
M onetary policy in 1993 and 1 9 9 4 might
therefore be characterized as a period o f tran­
sition. Only in hindsight will we know the
ultim ate outcom e o f this process. This article
seeks to describe the nature o f the evolution
to date. The next section describes the nature
of the m ulti-stage policy process embodied
in the fram ew ork o f interm ediate targeting.
Subsequent sections are organized within the
structure of this framework, focusing on the
way in which the intermediate targeting strate­
gy has evolved during the past two years. In
particular, the article exam ines the growing
consensus for price stability as the ultim ate
objective o f m onetary policy, describes the
continuing de-emphasis of monetary aggregate
targeting, and discusses som e issues relating
to the characterization o f short-run policy in
1993 and 1994.'

M ichael R. P akko
n the surface, an analysis o f monetary
policy in 1993 and 1 9 9 4 would seem to
be a study in contrasts. During 1993,
there were no changes in the policy directives of
the Federal Open Market Committee (FO M C),
and short-term interest rates rem ained steady
throughout the year. In 1 9 9 4 , on the other
hand, the FOM C announced six separate policy
changes, each associated with highly publicized
increases in short-term interest rates. From a
broader perspective, monetary policy over the
past two years could be characterized as
reflecting an evolution o f the Federal Reserve’s
instrum ent settings in response to strength­
ening econom ic growth. Policy remained
deliberately stimulative during 1993, as m em ­
bers o f the FO M C cautiously evaluated the
robustness o f the ongoing econom ic recovery.
A careful reading o f the policy record o f the
FO M C and statem ents by its m em bers, how ­
ever, reveals that a shift to a m ore-neutral
policy stance was viewed as quite likely, though
the tim ing o f the policy adjustm ents was in
Beyond the short-term adjustments in the
Federal Reserve’s policy settings during 1994,
a num ber o f additional them es characterize
m onetary policy in 1 9 9 3 and 1994. Over the
course of these two years, the relationships of
the m onetary aggregates to econom ic activity
continued to depart from historical patterns.
As a result, the strategy o f using m onetary
aggregates as intermediate targets has becom e
less im portant in the process o f form ulating
policy and com m unicating its intent to the
public. At the same time, however, a broad­


Since at least the 1970s, the Federal
Reserve’s m onetary policy has followed a
m ulti-stage process, often referred to as the
“Interm ediate Targeting” approach.2 The
underlying presum ption o f this strategy is
that some set of observable econom ic variables
can serve as indicators or operational targets
o f m onetary policy in a way that provides
inform ation about the links betw een specific
policy actions and the ultimate goals of policy.
Typically, the interm ediate targeting
strategy is presented in the econom ic literature
as a sequence o f four levels o f policy. At the
m ost basic level are the tools o f m onetary
policy— the fundam ental instrum ents over
which the Federal Reserve exerts direct control.


1 As a committee, the FOMC repre­
sents a range of individual view­
points. This article does not seek
to characterize the views of any
particular member, nor does it rep­
resent official positions of the
Committee. Rather, it reflects one
interpretation of recent events and
decisions of the FOMC.
2 See Meulendyke (1990) for an
account of the historic evolution of
the Fed's operating procedures and
intermediate targeting strategy.


3 Public Law 95-523, The Full
Employment and Balanced Growth
Act of 1978, Sec. 108.

These tools include reserve requirem ents, the
d iscount rate and open m arket operations.
The next stage, often referred to as the
operating instrum ents or proxim ate targets of
policy, consists of measures w hich are directly
affected by policy actions, but w hich are not
under the direct control of the Federal Reserve.
Included in this category are those variables
that provide inform ation on the m arket for
b ank reserves. The actions o f the Federal
Reserve’s open m arket operations directly
affect the supply o f reserves available to the
banking system. Hence, readings on conditions
in the reserve market can be drawn by observing
either the quantity o f reserves (m easured by
som e reserve aggregate or its growth rate) or
the interest rate in the m arket for inter-bank
reserve lending (the federal funds rate).
O n the n ext level are the interm ediate
targets o f policy. Theoretically, interm ediate
targets should have two key attributes: They
m ust be affected by the actions o f m onetary
policy and have a predictable relationship to
the ultim ate goals o f policy. An ideal interm e­
diate target would therefore serve to provide
tim ely inform ation on the im plications of
policy actions, allowing policymakers to make
mid-course corrections in response to readings
on the interm ediate target. As the m onetary
policy process has evolved over recent decades,
the interm ediate targeting strategy has devel­
oped around the notion o f using m onetary
aggregates as intermediate targets. In fact, the
use o f m onetary aggregates is reflected in the
congressional mandate given to the Fed to
guide the conduct o f policy, requiring that the
Fed report “objectives and plans...with respect
to the ranges o f growth or dim inution o f the
m onetary and credit aggregates... ,”3
Finally, at the end o f the spectrum are the
ultimate goals of monetary policy. The success
or failure o f policy can only be meaningfully
judged by its ability to achieve these goals.
Yet the particular criteria for m aking such a
ju d gm ent have not always been apparent.
Congress has legislated a number of objectives
for the Fed to pursue, which include econom ic
growth, high em ploym ent, stable prices and
low long-term interest rates. If the various
objectives seem , at tim es, to be incom patible
with one another, the legislation leaves unclear
how conflicts should be resolved.


W ithin the interm ediate targeting frame­
w ork, the policymaking process can be thought
o f as involving strategic and tactical decisions
relating the settings at the various levels.
As m andated by the Full Em ploym ent and
Balanced Growth Act o f 1 9 7 8 (otherw ise
know n as the H umphrey-Hawkins A c t), the
FO M C evaluates its longer-term objectives
twice per year, reporting to the Congress on
its projections for econ om ic activity, and pre­
senting its intermediate targeting objectives in
terms o f m onetary aggregate growth ranges.
This bi-annual exercise can be thought o f as
establishing the objectives and strategy o f
policy. At each o f its eight m eetings per year,
the FO M C m akes this strategy operational by
providing a “directive” to the M anager of the
System Open M arket A ccount at the Federal
Reserve Bank o f New York. This directive
specifies a short-term operating objective, cast
qualitatively in term s o f a “degree o f pressure
on [bank] reserve p o sition s.” The directive
also suggests the Com m ittee’s inclination
toward modification of policy during the inter­
meeting period. T he officials at the Open
M arket Desk then carry out the tactical aspects
o f the policy, arranging day-to-day purchases
or sales o f Treasury securities to achieve the
Com m ittee’s objectives for proxim ate targets
(for example, the federal funds rate and reserve
aggregate growth), in some cases adjusting the
instrument settings in response to incoming
information regarding the intermediate target
Figure 1 illustrates the process in a stepby-step m anner in a way w hich indicates the
links am ong the various stages and suggests
the type o f feedback rules with w hich policy is
evaluated and modified. The strategic decisions
of the FOM C are represented by the directional
arrows running from ultim ate objectives back
toward the tools o f policy, while the tactical
decisions o f short-run policy im plem entation
run in the opposite direction.
As the structure of the econom y and
econom ists’ understanding of its m echanism s
change over tim e, the Federal Reserve’s
approach to policym aking has evolved to meet
new challenges. This evolution of the structure
of policymaking is perhaps m ore significant
than the day-to-day and m onth-to-m onth
adjustm ents o f the Fed’s policy instrum ents,




b u t is often overlooked in analyses o f m one­
tary policy. Subsequent sections o f this arti­
cle exam ine som e o f the emerging trends in
the adaptation o f the FO M C ’s policym aking
approach, organized within the context of the
intermediate targeting framework.

F ig u r e 1

The Fed's Interm ediate Targeting Strategy
< -----------------


----------- ►

By tradition and legislation, the Federal
Reserve is charged with considering a number
o f objectives in the form ulation o f m onetary
policy. For exam ple, the Federal Reserve Act
as amended in 197 7 specifies that the Fed is
to “prom ote effectively the goals o f maxim um
em ploym ent, stable prices, and m oderate
long-term interest rates.”4 Federal Reserve
policym akers also seek to m aintain “orderly”
financial m arkets, w hich operationally has
meant an apparent tendency to smooth interest
rate changes.
The existence o f m ultiple goals raises
the possibility that two or m ore objectives
may com e into conflict. A lthough congres­
sional legislation specifies a num ber o f goals,
it gives no clear guidance how potential con ­
flict among objectives should be resolved.
Flistorically, Federal Reserve policym akers
have tended not to specify the relationships
among the goals explicitly or how potential
conflicts are to be resolved, preferring instead
to defer to the need to retain flexibility in the
im plem entation o f policy. As M aisel (1 9 7 3 )
noted: “Frequently, m em bers o f the FO M C
argued over the m erits o f a policy w ithout
ever having arrived at a m eeting o f the minds
as to what m onetary policy was and how it
worked. These problem s were, and still are,
neither recognized nor clarified.”5
Recently, however, public statem ents by
FO M C m em bers have tended to emphasize
the long-run consistency betw een the ob jec­
tives of “price stability” and econom ic
growth, recognizing that the trade-off w hich
was once com monly thought to exist between
inflation and real econom ic growth does not
exist in the long ra n (see the shaded insert
titled, “Statem ents by FO M C M em bers on
Price Stability”). This view has been shaped
both by theoretical advances in m acroeco­


nom ics and the experience o f the 1 9 7 0 s in
particular. As Federal Reserve Chairm an
Alan Greenspan has noted:
“...the experience o f the past three decades
has dem onstrated that what appears as
a tradeoff betw een unem ploym ent and
inflation is quite ephem eral and m is­
leading. Over the longer run, no such
tradeoff is evident...Experience both here
and abroad suggests that lower levels of
inflation are conducive to the achievement
of greater productivity and efficiency and,
therefore, higher standards o f living.”6
From this perspective, the trade-offs
among the various goals o f m onetary policy
appear less in conflict w ith one another than
they are often perceived to be: In the long
ra n , the pursuit o f price stability is consistent
with— perhaps even necessary for— the mainte­
nance of econom ic growth and low long-term
interest rates. This view also stands in contrast
to many characterizations o f recent m onetary
policy by the media, w hich suggest that the
Fed’s policy is to deliberately impede econom ic
growth in order to subdue inflation.7 Never­
theless, there is often pressure from outside
the Federal Reserve to pursue policies w hich
prom ise to provide short-term gains in ou t­
put and em ploym ent, but at the expense of
potential inflationary consequences in the
longer term .8
Despite the general support for price
stability, such broad statem ents o f purpose
rem ain som ewhat vague as operational objec­
tives. As Chairman Greenspan described the
issue: “...price stability does n ot require that
measured inflation literally be zero but rather
is achieved when inflation is low enough that
changes in the general price level are insignif-


4 Federal Reserve Reform Act,
Section 2A, November 1 6 ,1 97 7
(91 Slot. 1387).
5 Maisel (1973), p. 78.
‘ Statement before the Joint
Economic Committee, U.S.
Congress, January 31, 1994.
Federal Reserve Bulletin (March
1994, p. 232).
7 For a critique of this view, see
Jordan (1994).
! See, for example, Papadimitriou
and Wroy (1994).


W h ile their views often differ in em phasis and w ith regard to specific policy recom m en­
dations, m em bers o f the FO M C display a broad unanim ity o f opinion regarding the u lti­
mate long-run objectives o f m onetary policy:
“...the Federal Reserve seeks to foster
m axim um sustainable econom ic growth
and rising standards o f living. And in that
endeavor, the m ost productive function
the central bank can perform is to achieve
and m aintain price stability.”

“I believe that the prim ary goal o f
policy is to prom ote econom ic growth and
em ploym ent and that the Federal reserve
can best pursue this goal by fostering a
stable aggregate price level over tim e.”
- J . Alfred Broaddus, Jr., President,
Federal Reserve Bank
of Richmond

- Alan Greenspan, Chairman,
Federal Reserve Board

“Inflation has to, by default, take pri­
macy because that is what we can control
in the long ru n .”

“W e now know that m axim um
sustainable econom ic growth is achieved
w hen changes in the price level cease to
be a factor in econom ic d ecision-m aking.”

- Alan Blinder, Vice Chairman,
Federal Reserve Board

- Thomas C. Melzer, President,
Federal Reserve Bank
of St. Louis

“T he Federal Reserve is com m itted to
keeping inflation down n ot for its own
sake, b u t because it is im portant for long­
term econom ic growth for this country.”

“ the long run the m ost significant
contribution m onetary policy can m ake to
achieving m axim um sustainable growth

- Lawrence B. Lindsey, member,
Federal Reserve Board

in real output is to foster price stability.”
- Gary H. Stern, President,
Federal Reserve Bank
of Minneapolis

“I think in general we’ve made good
progress on price stability but it’s not
som ething where you can say, ‘W e’ve won
the battle and we can go hom e.’ It’s som e­
thing we always have to pay attention to .”

“1 think we all agree that the goal o f
m onetary policy is to prom ote m axim um
sustainable growth over tim e...Bu t ju s t as
im portant, and consistent w ith this goal,
the Federal Reserve m ust w ork toward
ensuring an environm ent of price stability.”

- Susan M. Phillips, member,
Federal Reserve Board

“Keeping inflation low is a necessary
ingredient for m axim izing sustainable
econom ic and jo b grow th.”

- Thomas M. Hoenig, President,
Federal Reserve Bank
of Kansas City

- Edward G. Boehne, President,
Federal Reserve Bank
of Philadelphia

“ the long run the m ost significant
contribution we can m ake to econom ic
growth is by providing a low -inflation
environm ent, and we have made progress
in that area...”

“W hat m onetary policy can do to
prom ote long-run econom ic efficiency is
to stabilize the aggregate price level and to
create a clim ate o f confidence about the
ou tlook for price stability.”

- Robert T. Parry, President,
Federal Reserve Bank
of San Francisco

-Je r ry L. Jordan, President,
Federal Reserve Bank
of Cleveland


T a b le 1

FOMC Central Tendency Projections



Feb. 93

N om inal GDP
Real GDP
Un em p lo ym en t rate

514 — 6
3 -3 /4
6 /! 4 - 7

N om inal G D P
Real G D P
Un em p lo ym en t rate

July 93

Feb. 94

July 94

5 .0
2 .7
6 .5

5 - 5 /<
2 / 4- 2 34
3 - 3/ 4
5 - 6/2
2 ' A - 3/4
3 - /3 2
6 / 2- 6 / ,


5/ 2 -6

5/ 2 -6

3 -3 /4
About 3
6 / 2- 6 / 4

3 -3 /4
234 - 3
6 - 6/4

6 .5
2 .7
5 .6

5 - 5 /
2 / , -2 34
2 / 4 - 3/2
6 - 6/4

N om inal G D P
Real G D P
U n em p lo ym en t rate

Note: Unemployment rote refers to the overage level for the fourth quarter. All other dato represent fourth quorter-to-fourth quarter percentage changes.

econom ic outcom es over this tim e horizon,
the distinction betw een Com m ittee mem bers’
expectations and objectives are som ewhat
unclear in these projections.
The notion o f price stability as the ulti­
mate goal of m onetary policy and the recog­
nition o f the im portance o f long-term plan­
ning horizons have led som e to advocate the
introduction of som e form o f explicit longrange price level or inflation target to the
m onetary policy process.1 R ecent policy
reform s in New Zealand, Canada and Great
Britain have moved in this direction, w ith
apparent success to date. Advocates o f such
a policy for the U nited States emphasize the
im portance o f credibility in m onetary policy;
that is, individuals and businesses are more
likely to have faith in the Fed’s ability to
m aintain price stability when there is a clear
com m itm ent to a specific objective."
Indeed, the im portance o f inflation
expectations and the role o f Fed credibility
in the form ation o f those expectations are
issues w hich have been emphasized in recent
statem ents by Chairm an Greenspan: “The
effects o f policy on the econom y critically
depend on how m arket participants react to
Federal Reserve actions as well as on expec­
tations o f our future actions.”1 As an example
of the im portance o f expectations, Greenspan
has suggested that a significant feature o f the
econom y’s slow em ergence from the 19 9 0 -9 1

icant for econom ic and financial planning.”’
However, many questions rem ain unre­
solved: W hat price index should be used for
measurem ent? W hat rate o f inflation corre­
sponds to “price stability”? Should the Fed
pursue objectives stated in terms o f price
levels or inflation rates? W hat operating
procedures should be used to achieve the
objective? W hat is the relevant tim e frame
for achieving and m aintaining price stability?
The em phasis on the benefits o f longrun price stability suggests the potential effi­
cacy o f establishing long-run objectives for
m onetary policy. Under the present stru c­
ture of policy form ulation, the only quantita­
tive m ethod o f com m unicating long-term
expectations regarding the objectives o f poli­
cy is the bi-annual econom ic projections of
the Com m ittee, w hich are presented by the
Chairm an o f the Board o f G overnors in each
o f the Humphrey-Hawkins reports to
Congress. Table 1 reports the central ten­
dency m easures o f these projections reported
in 1993 and 1994. Note that these projec­
tions extend only 12 to 18 m onths. More
importantly, it is unclear how the forecasts
subm itted by FO M C m em bers incorporate
anticipated m onetary policy actions. Tinsley
and others (1 9 8 1 ) refer to the nature o f these
projections as “econom ic w eather forecasts
with provisions for cloudseeding.” To the
extent that the Fed’s policy actions effect


9 Statement before the Committee
on Banking, Mousing, and Urban
Affairs, U.S. Senate, February 19,
1993. Federal Reserve Bulletin
(April 1993, p. 300).
10 See, for example, 1994 Annual
Report, Federal Reserve Bonk of
St. Louis. For an example of a
specific operational plan for achieving
price stability, see Gavin and
Stockman (1992).
1 See, for example, Angell (1994)
ond Jordan (1993).
1 Statement before the Committee
on Banking, Housing, and Urban
Affairs, U.S. Senate, February 19,
1993. federal Reserve Bulletin
(April 1993, p. 293).


one o f the crucial issues in m onetary policy
in 1995 and beyond.

M onetary A ggregate O bjectives
Ranges (percentage growth rates)
Date of Meeting
Feb . 2 - 3 , 1 9 9 3
Ju ly 6 - 7 , 1 9 9 3
Feb . 2 2 , 1 9 9 4
Ju ly 2 0 , 1 9 9 4

13 Statement before the Committee
on Bonking, Housing, and Urban
Affairs, U.S. Senate, February 19,
1993. Federal Reserve Bulletin
(April 1993, p. 293).
14 Statement before the Joint
Economic Committee, U.S. House
of Representatives (December 7,
15 During the first half of 1993, both
M2 and M3 were running below
the growth ranges originally speci­
fied by the Committee in February
1993. Reflecting uncertainty
regarding the factors distorting the
aggregates' growth rates, the
ranges were lowered in July of
1993 (see Table 2).
16 See Poole (1994).
17 Moisei (1973) provides an insid­
er's view of this period of monetary

Target Period


1 9 9 2 :Q 4 — 1 9 2 3to 64
9 :Q





'/2 to 4 '/2

4 '/ jt o 8 '/2

0 to 4
0 to 4

4 to 8
4 to 8

Perhaps the m ost fundam ental m odifica­
tion to the interm ediate targeting strategy
witnessed over the past two years has been the
1 9 9 3 :Q 4 — 1 9 9 4 :Q 4 1 to 5
0 to 4
4 to continuing de-emphasis of the monetary targets
as operational objectives. This is n ot to say
1 9 9 3 :Q 4 — 1 9 1 4 :Q54
9 to
0 to 4
4 to 8
that the aggregates are now disregarded alto­
1 9 9 4 : Q 4 - 1 9 9 5 : Q 4 1 to 5
0 to 4
3 to 7
gether as indicators o f policy, b u t rather that
recession was a need to restructure balance
the prom inence they once held in discussions
sheets, w h ich in turn was partly attributable
o f policy has dim inished significantly.
to inflation expectations: “...households and
Table 2 reports the ranges specified by
businesses apparently were skeptical that infla­
the FO M C for m oney and credit growth in
tion would continue to decline and...may even
1 9 9 3 and 1 9 9 4 , and Figure 2 displays actual
have expected it to rebound. As a consequence,
m easures o f the m onetary aggregates relative
many may have shaped their investm ent deci­
to these target ranges. Despite the lessened
sions im portantly based on expectations of
em phasis on attaining m onetary aggregate
inflation-induced appreciation of asset prices
targets as a policy objective (discussed further
rather than on m ore fundam ental econom ic
below ), the aggregates finished both years
considerations.” 1
w ithin the specified growth ranges.1
The idea that m onetary policy should
There is good reason to consider measures
be charged w ith a single specific objective
o f the m oney stock as im portant indicators
o f price stability is not new. In 1 9 8 9 , Repre­
o f the thrust of m onetary policy. B oth theo­
sentative Steven Neal, D .-N orth Carolina,
retically and empirically, the growth rate of
introduced legislation w hich would have
money and the rate of inflation are known to
mandated such a framework. At the time,
be closely related— at least over long periods.
the Neal proposal was m et by favorable reac­
This relationship can be clearly observed in
tions from Chairm an G reenspan and other
com parisons o f inflation rates and money
FO M C mem bers. More recently, Senator
growth rates across countries, and considera­
Connie M ack, R .-Florida, has indicated his
tion o f trends w ithin a single country over
inten tion to introduce legislation in 1 9 9 5 for
extended periods o f tim e.1 Over shorter time
the purpose o f modifying the Humphreyhorizons, however, the relationship is m uch
Hawkins fram ew ork to elim inate references
less apparent. T his is at least partly attribut­
to em ploym ent and interest rate objectives,
able to the fact that m easured m onetary
and to direct the Federal Reserve to lim it CPI
aggregates are, at best, an approxim ation of
inflation to less than 2 percent per year.
econom ists’ conceptual notion o f “money.”
Chairm an Greenspan has responded that he
From m onth to m onth or quarter to quarter,
favors such legislation in principle, but has
substitution am ong various assets often
dem urred on the issue o f a num erical o b jec­
m akes the growth rates o f the aggregates dif­
tive: “I have always argued that it would be
ficult to interpret. Moreover, the rapid pace
useful for be required to focus crucially,
o f financial innovation in recent years has
if not solely, on domestic price inflation... [but]
changed the nature o f the aggregates, further
I would be m ore inclined to go to a more
com plicating their interpretation.
T he FO M C began to consider m onetary
general type o f requirem ent for the central
b a n k ...1
aggregates as operational objectives o f policy
T he issue o f w hether to charge the
explicitly in its directives in 1 9 7 0 .1 The status
Federal Reserve with a specific long-run
of the aggregates took on m ore prom inence
price stability mandate is likely to rem ain as
over the years, and their role as interm ediate
1 9 9 2 :Q 4 — 1 9 9 3 :Q 4 1 to 5
1 9 9 3 :Q 4 — 1 9 9 4 :Q 4 1 to 5





F ig u r e 2

M o n etary A g g reg a te s
M l A g g reg ate

M 2 A g g re g a te

M 3 A g g re g ate

D om estic N o n fin a n cia l Debt


m onetary policy. Som e econom ists, em pha­
sizing the transactions role o f money, have
typically favored narrower m easures such as
M l. O thers, who stress the additional role of
money as a store of value, suggest that broader
aggregates like M 2 are m ore appropriate as
indicators o f available purchasing power.
W hile such theoretical considerations are
certainly considered by policym akers, the
FO M C ’s choice o f interm ediate target has
typically appeared to be guided m ore by
observations on the consistency and stability
o f relationships betw een the aggregates and
econom ic activity.
In the absence o f reliable inform ation
from M l after the m id -1980s, the broader
aggregate M 2 naturally took on greater
prom inence. In the early 19 9 0 s, however,
the relationship betw een M 2 and overall
econom ic activity also began to show signs
o f deterioration. O ne way o f sum m arizing
this relationship is to consider the velocity of
M2— the ratio o f the total dollar value o f GDP
to M 2. Figure 3 illustrates the historical
behavior o f M 2 velocity. U ntil recently, this
measure has show n little tendency to display

targets was w ritten into the congressional
mandates beginning w ith House C oncurrent
R esolution 133 in 1975 and later in the 1978
H umphrey-Hawkins legislation. The use of
m onetary aggregates as interm ediate targets
in the United States reached a high point
during the period from O ctober 1979 until
the autum n o f 1982 , when the FO M C placed
greater emphasis on m onetary growth in an
effort to establish a credible policy of halting
and reversing the rising trend of inflation.
W hile both M l and M 2 were cited as opera­
tional objectives, primary attention at the
tim e was focused on the narrow aggregate
M l. By the m id -1980s, however, the rela­
tionship o f M l to overall econom ic activity
had apparently changed so m uch that it
seemed less desirable as an interm ediate tar­
get. In fact, reference to M l was removed
from the FO M C ’s policy directives starting
w ith the O ctober 1 9 8 2 m eeting, and the
Com m ittee stopped reporting annual growth
objectives for M l in 1 9 8 7 .1
T h e issue o f w hich m onetary aggregate
is appropriate for guiding policy has a long
and controversial history in discussions of


18 A discussion of the problems
encountered with M l is beyond
the scope of this article. Some of
these issues ore explored in Stone
and Thornton (1987).

F ig u r e 3

velocity: They pointed out that the use of
the three-m onth T-bill yield to m easure
opportunity cost may n o t fully capture the
range o f alternative yields relevant to the
public’s demand for M 2 ." E xperim enting
with a broader range o f opportunity cost
measures, Feinm an and Porter found that
using loan rates and longer-term Treasury
yields helped to explain the behavior o f M 2
velocity in the 1 9 8 0 s and 1990s. This expla­
nation seem ed particularly relevant given the
steepness o f the yield curve that prevailed in
1992: If longer-term assets were, in fact,
good substitutes for the com ponents o f M 2,
then the relatively high rates o f return on
long-term instrum ents may well have been
attracting funds out o f M 2, depressing its
growth rate.
Although this insight helped to account
for unusually slow M 2 growth to som e
extent, its explanatory power was apparently
insufficient to sustain FO M C m em bers’ con ­
fidence in M 2 as an interm ediate target
beyond m id -1993. In discussing the de­
em phasis o f M 2 at the Humphrey-Hawkins
hearings in Ju ly 1 9 9 3 , Chairm an Greenspan
described how quickly this confidence had
evaporated: “T he evidence as of, say the end
o f last year, would suggest that it was proba­
bly correct to assume that M 2 was becom ing
increasingly faulty Six m onths later, it’s
becom ing extraordinarily persuasive.””
As further observations becam e avail­
able, the yield curve explanation becam e
even m ore untenable as a significant expla­
nation for the rapid growth o f M 2 velocity.
In particular, the sharp flattening o f the yield
curve in 1 9 9 4 appeared to be associated w ith
little if any slowdown in velocity growth.
A lthough the slow growth o f M 2 in
recent years is n ot fully explicable, at least
two additional transitory special factors have
contributed to the weakness. First, the large
decline in interest rates during 1991 and
1 9 9 2 stim ulated extensive refinancing of
long-term debt, particularly m ortgages. In
the refinancing process, m ortgage servicers
tended to hold funds in highly liquid
deposits prior to transferring the balances to
investors holding the underlying mortgagebacked securities. The large volum e o f funds
moving through liquid deposit accounts

V elo city a n d O p p ortunity Cost of M 2
M2 Velocity (ratio)

M2 Opportunity Cost (percent)





19 The FOMC's discussion of the
Feinman and Porter study was
reported by the American Banker
(see Cummins, 1992).
20 Statement before the Committee on
Banking, Mousing, and Urban Affairs,
U.S. Senate (July 22, 1993, p. 34).

any trend rate o f growth, fluctuating around
a constan t value o f approxim ately 1.65.
As show n in Figure 3, m uch o f the vari­
ability o f M 2 velocity around its trend can be
related to a m easure o f opportunity cost.
The opportunity cost m easure illustrated in
Figure 3 is defined as the difference betw een
the interest rate on three-m onth Treasury
bills (considered to be an alternative to hold­
ing M 2 assets) and a weighted average o f the
rates o f return on assets included in M 2.
W hen the opportunity cost o f holding M 2 is
high (th at is, when the return on M 2-type
assets is relatively low ), the growth rate of
M 2 tends to be lower than it otherwise
would be as people take advantage o f other,
higher-yielding alternatives. H ence, the
velocity of M 2 tends to rise above its trend.
It is clear from Figure 3 that this rela­
tionship has recently departed from its typi­
cal historical pattern. W h ile the measured
opportunity cost o f holding M 2 has been
quite low during the early 1 9 9 0 s, M 2 growth
has been uncharacteristically slow so that
velocity has risen far above its average level.
As the breakdow n o f this relationship
becam e apparent in the early 1990s, Federal
Reserve officials searched for explanations.
In a staff study presented to the FO M C at the
m eeting o f Novem ber 17, 1 9 9 2 , Feinm an
and Porter (1 9 9 2 ) suggested one possible
explanation w hich seem ed to account for
som e o f the anom alous behavior o f M2



associated with this activity had the effect of
boosting the measured growth rate o f the
m onetary aggregates. As this refinancing
activity declined in 1 9 9 4 , the associated run­
off o f liquid funds tended to dampen m one­
tary growth rates.2 This factor is clearly
temporary in nature, and it is likely that
inter-aggregate flows associated with refi­
nancing activity had subsided by m id -1994.
A second possible factor contributing to
uncharacteristically slow M 2 growth is the
recent surge in popularity o f bond and equity
mutual funds. Figure 4 illustrates net assets
of these instrum ents over the past several
years. A significant portion o f these funds
appeared to be flowing from tim e deposits
and m oney m arket m utual funds, w hich are
b oth included in M 2. Figure 4 illustrates the
correspondence betw een the recent period of
sharply rising M 2 velocity and the period of
dram atic m utual fund growth. To the extent
that portfolio shifts from M 2 assets into
m utual funds has accounted for the anom ­
alous behavior o f M 2 growth, an aggregate
that includes m utual funds along w ith M 2
assets (called M 2 Plus) could potentially per­
form better than the conventional M 2 defini­
tion. Researchers at the Board o f Governors
investigated this possibility, and the m atter
came up for d iscussion at the FO M C m eet­
ing o f Ju ly 6 -7 , 199 3 . This research suggest­
ed that although the velocity o f M 2 Plus was
som ewhat less anom alous than that o f M 2,
the inclusion o f mutual funds did n ot fully
elim inate the recent velocity puzzle.2
Accordingly, the m inutes o f the FO M C
report that “after exam ining the properties of
this measure and reviewing its past behavior
in relation to key indicators o f econom ic per­
form ance, the m em bers concluded that it
would not enhance the form ulation or
im plem entation o f m onetary policy, at least
at this p o in t.”2
Subsequent experience appeared to have
borne out the C om m ittee’s assessm ent. As
illustrated in Figure 4 , flows into m utual
funds have slowed dram atically during 1 9 9 4
as interest rates have risen. M 2 growth,
however, has remained uncharacteristically
slow, w ith its velocity reaching record highs
toward the end o f the year.
In recognition o f these unusual factors


F ig u r e 4

M 2 V elo city an d N et A ssets of
Bend an d Equity M utual Funds
M2 Velodty (ratio)

Mutual Fund Assets (bfliions of dollars)



- 1400



I- 400


affecting the growth rate o f the m onetary
aggregates, the FO M C lowered its growth
objectives for M 2 and M 3 at its July, 1993
m eeting (see Table 2 ). In his subsequent
report to Congress, Chairm an Greenspan
indicated that the Com m ittee had also decid­
ed to de-em phasize its consideration o f M 2
as a policy target: “At least for the time
being, M 2 has been downgraded as a reliable
indicator o f financial conditions in the econ­
omy, and no single variable has yet been
identified to take its place.”2
T he lack o f any particular m easure to fill
the role previously played by m onetary
aggregates has fundam entally altered the
ostensible control strategy o f the interm edi­
ate targeting approach, w ith m em bers o f the
Com m ittee left to r e ly on “ongoing assess­
m ents o f the totality o f incom ing inform a­
tion and appraisals o f the probable outcom es
and risks associated w ith alternative poli­
cies.”2 M ore importantly, the ability o f poli­
cym akers to com m unicate long-term policy
intentions is greatly dim inished by the
absence o f meaningful m onetary targets.
A lthough m onetary targets have never been
the sole guide for FO M C policy decisions,
they did provide a useful fram ew ork for
assessing short-run policy adjustm ents in a
con text o f longer-run objectives. In the
absence o f interm ediate targets, public atten­
tion has becom e m ore focused on short-term
adjustm ent in the federal funds rate and dis­
cou nt rate.



2 See Anderson (1993).
22 See Collins ond Edwards (1994)
and Orphanides, Reid and Small
23 Minutes of the Federal Open Market
Committee Meeting of July 6-7,
1993. Federal Reserve Bulletin
(October 1993, pp. 944-5).
21 Statement before the Subcommittee
on Economic Growth and Credit
Formation of the Committee on
Banking, Finance and Urban Affairs,
U.S. House of Representatives,
July 2 0 ,1 9 9 3 . Federal Reserve
Bulletin (September 1993, p. 852).
25 Statement by Alan Greenspan,
Chairman, Federal Reserve Board,
before the Subcommittee on
Economic Growth and Credit
Formation of the Committee on
Bonking, Finance and Urban Affairs,
U.S. House of Representatives,
February 22, 1994. Federal Reserve
Bulletin (April 1994, p. 304).

T a b le


FOMC D irectives and M e asu res of M o n etary P olicy Stance
Intermeeting Stance Toward
M eeting

D irective for
R e se rv e Pressure

G reater
R estraint

Le sse r Restraint

Result from Change in Reserve Pressure.
Fund s rate
"targ e t" *

D ate of C hang e

Discount rate

1 9 9 3
Feb. 2 -3

m ain tain


Mar. 2 3

m ain tain


N /A

3 .0 0
N /A

3 .0 0

3 .0 0

3 .0 0

M ay 18

m ain tain

m ight


N /A


J u l. 6-7

m ain tain

m ight


N /A


m m ight

Aug. 17

m ain tain

Sep . 21

m ain tain

m ight

m ight

N /A0
3 .0

3 .0 0

Nov. 16

m ain tain

m ight

m ight

N /A0
3 .0

3 .0 0

m ain tain

m ight

m ight

N /A0
3 .0

3 .0 0

D ec. 21

N /A

3 .0 0

3 .0 0

1 9 9 4
m ight
m ight

Feb. 3 -4

in cre ase slightly

M ar. 22

in crease slightly m ight

m ight

2 /4/94

m ight

m ight

3 .2 5
3 / 2 2 /39.540 &

4 /18/94
M ay 17

in cre ase som ew hat

Ju l. 5-6

m ain tain

m ight

3 .7 5


in crease so m ew h at would



m ain tain

m ight

w ould

N /A 5
4 .7

Dec. 2 0

would ould

m ain tain

m ight


3 .5 0
N /A

Sep. 27

in crease sig n ifican tly

3 .0 0

5 / 1 7 / 954
4 .2

A ug . 1 6
Nov. 15

3 .0 0
3 .0 0

4 .2 5

3 .5 0
4 .7 5
4 .0 0

5 .5 0

4 .7 5

N /A 0
5 .5

4 .7 5

* F e d e ra l fu n ds rate expected to be consistent with desired re se rv e restrain t.


occasions, w ith the cum ulative effect of these
actions reflected in an increase o f around
2 lh percentage points in the federal funds rate.
Table 3 sum m arizes the actions taken by the
FOM C at its meetings over the 1993-94 period.
The appendix to this article sum m arizes the
discussions that took place at those meetings.
As is always the case w hen the FO M C ’s
policy decisions are associated with interest
rate increases, the Fed has been criticized in
1994 for hampering the econom y by pursuing
overly “tight” m onetary policy. However, it is
not clear that the policy moves taken in 1 9 9 4
should be considered particularly restrictive.
Rather, the stated intentions of the Committee
have been to move the stance o f policy from
one of “accom m odation” to “a m ore neutral
The steady policy pursued in 1993 was
recognized by Com m ittee m em bers as being
purposefully accom m odative. Lacking any

Although developments in the evolution of
the FO M C’s policy framework described above
are significant, it is the m eeting-to-m eeting
actions o f the Com m ittee and their effect on
short-term interest rates w hich have attracted
the attention of the public. From 1989 through
19 9 2 , the FO M C endorsed a policy o f easing
reserve restraint, perm itting rapid growth
o f bank reserves and facilitating 25 distinct
declines in short-term interest rates, cu m u­
lating in nearly a 7 percentage point drop from
previous peaks. During 1993, the Com m ittee
called for “m aintaining the existing degree of
pressure on reserve positions” at each of its
m eetings, and the federal funds rate remained
fairly constant at around 3 percent. In 1 9 9 4 ,
on the other hand, the Fed announced actions
to increase reserve pressure on six separate




MEMBERS OF THE FOMC IN 1 9 9 3 AND 1 9 9 4
At any given time, the Federal O pen M arket Com m ittee consists o f 12 voting mem bers.
The Com m ittee includes all seven m em bers o f the Board o f G overnors o f the Federal
Reserve System, as well as five o f the 12 presidents o f the regional Federal Reserve banks.
Reflecting the im portance o f the Federal Reserve Bank o f New York in policy im plem enta­
tion, the president o f that Reserve Bank is always a voting m em ber and is, in fact, elected as
Vice Chairm an o f the Com m ittee (the Chairm an o f the Board o f G overnors is elected as
Chairm an o f the FO M C ). The rem aining four positions rotate am ong the presidents o f the
other 11 Federal Reserve banks. Although only a lim ited num ber o f Federal Reserve Bank
presidents are voting m em bers o f the Com m ittee, all 12 attend the m eetings and participate
in the discussions.
In addition to the usual rotation o f Federal Reserve Bank presidents as voting m em bers
o f the Com m ittee, the Com m ittee’s com position in 1 9 9 3 and 1 9 9 4 changed due to changes
in the m em bership o f the Board of G overnors and the presidency o f the New York Fed.
Listed below are the voting m em bers o f the FO M C in 1993 and 1994.



Alan Greenspan
C h airm an , B oard of G o vern o rs

Alan Greenspan

E. Gerald Corrigan/William J. McDonough*
President, Fed eral R e se rv e B an k of New York

William J. McDonough
P resident, F e d e ra l R e se rv e B a n k of New York

David W. Mullins, Jr.
Vice C h airm an , B oard of G o verno rs

Alan S. Blinder!
Vice C h airm an , B oard of G o vern o rs

Wayne D. Angell
m em ber, B oard of G o vern o rs

Janet L. Yellent
m em ber, B oard of G o verno rs

John P. LaWare
m em ber, B oard of G o verno rs

John P. LaWare
m em ber, B oard of G o verno rs

Lawrence B. Lindsey
m em ber, B oard of G o vern o rs

Lawrence B. Lindsey

Susan M . Phillips
m em ber, B oard of G o verno rs

Susan M . Phillips
m em ber, B oard of G o verno rs

Edward W. Kelley, Jr.
m em ber, B oard of G o verno rs

Edward W. Kelley, Jr.

Edward G. Boehne
P resident, F e d e ra l R e se rv e B an k of P hilad elphia

J. Alfred Broaddus, Jr.
P resident, Fed eral R e se rv e B an k of Richm ond

Silas Keehn
P resident, F e d e ra l R e se rv e B an k of Chicago

Robert P. Forrestal

Robert D. McTeer, Jr.
President, Fed eral R e se rv e B an k of D allas

Jerry L. Jordan

Gary H. Stern
President, Fed eral R e se rv e B a n k of M inneapolis

Robert T. Parry
President, Fed eral R e se rv e B an k of S an Francisco

C h airm an , B oard of G o vern o rs

m em ber, B oard of G o verno rs

m em ber, B oard o f G o verno rs

P resident, F e d e ra l R e se rv e B an k of A tlanta
P resident, F e d e ra l R e se rv e B a n k of C leveland

* The lost FOMC meeting attended by Mr. Corrigan was May 18, 1993. Mr. McDonough began his tenure wilti the FOMC at the meeting of August 17, 1993.
At the meeting of July 6-7, 1993, Mr. Janies H. Oilman, First Vice President of the New York Fed, served as alternate for Mr. Corrigan.
t Mr. Blinder's tenure on the FOMC began with the meeting of July 5-6,1994.
t Ms. Yellen's tenure on the FOMC began with the meeting of August 16, 1994.




26 Statement by Alan Greenspan,
Chairman, Federal Reserve Board,
before the Committee on Bonking,
Housing, and Urban Affairs, U.S.
Senate, July 20, 1994.Federal
Reserve Bulletin (September 1994,
pp. 793-4).
27 Minutes of the Federal Open Market
Committee Meeting of July 6-7,
1993. Federal Reserve Bulletin
(October 1993, p. 946).
28 Statement before the Joint Economic
Committee, United States Congress,
January 31, 1994. Federal Reserve
Bulletin (March 1994, p. 233).
29 Minutes of the Federal Open Market
Committee Meeting of February 3-4,
1994. Federal Reserve Bulletin
(May 1994, p. 408).
30 Statement by Alan Greenspan,
Chairman, Federal Reserve Board,
before the Committee on Banking,
Housing, and Urban Affairs, U.S.
Senate, July 20, 1994.Federal
Reserve Bulletin (September
1994, p. 794).
3 Minutes of the Federal Open Market
Committee Meeting of August 16,
1994. Federal Reserve Bulletin
(November 1994, p. 996).
32 Statement before the Committee
on Banking, Housing, and Urban
Affairs, U.S. Senate, February 19,
1993. Federal Reserve Bulletin
(April 1994, p. 294).

clear guidance from the behavior of m onetary
aggregates, this characterization o f policy was
based on the observation that short-term interest
rates remained extraordinarily low, particularly
in relation to the underlying rate o f inflation.
W ith short-term interest rates and inflation
b oth holding at about 3 percent, short-term
real interest rates (inflation adjusted) were
close to zero. The Committee members viewed
the m aintenance o f such low levels of interest
rates as being an unwise and unsustainable
policy over the long run: “...history strongly
suggests that m aintenance o f real short-term
rates at levels prevailing [in 1993] ultim ately
would have fueled inflationary pressures.”2
This policy was m aintained in an effort
to alleviate a num ber o f special factors w hich
appeared to be inhibiting a strengthening of
the econom ic recovery. In his testimony before
Congress in February 1 9 9 3 , G reenspan cited
a need for balance sheet restructuring by
households and firms, difficult adjustm ents
associated w ith business restructuring, and
the contractionary effects o f cuts in federal
defense spending.
Throughout 1 9 9 3 , the m em bers o f the
Com m ittee were circum spect regarding the
accom m odative nature o f policy. At both
the May and Ju ly m eetings, the Com m ittee
endorsed a policy w hich— although calling
for no im m ediate change in the stance of
policy— specified a bias toward the possibility
o f increasing the degree o f reserve restraint
(see the appendix). T he M inutes o f the Ju ly
m eeting reveal that som e m em bers “com ­
m ented that while the need for any policy
adjustm ent during the period ahead seemed
som ew hat rem ote, the n ext policy move was
more likely to be in the direction o f some
firm ing than toward easing.”2
At a hearing before the Jo in t E conom ic
Com m ittee o f Congress in late Janu ary 1994,
Chairm an Greenspan clearly indicated that a
move toward greater reserve restraint was not
a m atter o f if, but o f w hen: “At som e point,
absent an unexpected and prolonged w eak­
ening o f econom ic activity, we will need to
move [short-term interest rates] to a more
neutral stance.”2 At the next m eeting o f the
FO M C , the first step in that direction was

199 5

the Committee agreed to a proposal to have the
Chairm an announce the move. “T he purpose
of such an announcem ent, w hich would be a
departure from past Committee practice, was to
avoid any misinterpretation o f the Com m ittee’s
action and its purpose.”2 (See the shaded
insert titled “Policy D isclosure.”)
As the year progressed, further increases
in the degree o f reserve pressure were under­
taken on five additional occasions, three of
w hich were accom panied by increases in the
discount rate (see Table 3 and the appendix).
The Committee proceeded with the tightening
in this step-by-step m anner in recognition o f
the difficulty of knowing precisely what trading
range for the federal funds rate was appropriate:
“ is an open question whether our actions
to date have been sufficient to head off infla­
tionary pressures and thus m aintain favorable
trends in the econom y.”3
At the same tim e, Com m ittee m em bers
expressed a desire to move decisively enough
to head off emerging inflationary expectations.
In the discussion surrounding the % percentage
point increase in the fed funds rate and dis­
cou nt rate taken on August 16, m em bers
noted that “a more decisive policy move might
reduce the need for further tightening later...
by helping to curb inflationary expectations
more effectively.”3
In fact, the objective o f subduing infla­
tionary expectations was a prom inent con ­
sideration in the FO M C ’s policy deliberations
in 1993 and 1994. One o f the indicators used
to discern these expectations is the slope of
the yield curve, the steepness o f w hich had
been o f concern to policym akers for som e
time. As early as February 1 9 9 3 , Greenspan
had pointed out that “T he steep slope o f the
yield curve and the expectations about future
interest rates that the slope im plies suggest
that investors rem ain quite concerned about
the possibility o f higher inflation... ,”3
Although conclusions about inflationary
expectations embedded in the yield curve
should be interpreted cautiously, the reaction
o f the term structure o f interest rates to the
policy moves taken in 1 9 9 4 provides an
interesting perspective on those events. As
illustrated in Figure 5, the initial increases in
short-term interest rates during 1 9 9 4 were
accom panied by shifts in the entire term

taken. Because this move was the first tight­
ening o f policy to be undertaken in some time,


A nother issue w ith w hich the Com m ittee grappled throughout 1993 and 1 9 9 4 was
the tim ing of, and extent to w hich, policy decisions should be announced to the public.
Traditionally, the policy decisions o f the FO M C have been closely guarded secrets, w ith
m inutes o f each m eeting issued only after the subsequent m eeting had concluded (so that
the current operational directive was never made public). T he purpose o f this confidentiality
was to avoid the possibility of financial m arket instability in the wake o f policy changes, as
well as to give the FO M C m ore flexibility in the im plem entation o f policy. This practice has
always been controversial, and criticism o f the Fed’s traditional secrecy had recently intensi­
fied, particularly am ong som e m em bers o f Congress. The FO M C reconsidered the disclo­
sure issue during 1 9 9 3 , and experim ented w ith announced policy changes during 1994.
The issue o f public disclosure was discussed at the first FO M C m eeting o f 1 9 9 3 , as the
Com m ittee considered a prelim inary report o f a subcom m ittee that had “been established to
exam ine various issues relating to the release o f inform ation about Com m ittee m eetings and
d ecisions.” The m em bers agreed that the public should be fully inform ed about policy
decisions, but expressed con cern that “release o f inform ation should n ot be allowed to com ­
prom ise the overriding objective o f m aking and im plem enting the best possible d ecisions.”
At the Ju ly 6 -7 , 1 9 9 3 , m eeting, the issue arose again in the context o f “media reports of
the purported results of the May m eeting before the Com m ittee had made public any infor­
m ation about that m eeting.” On that occasion, the m em bers “agreed that particular care
needed to be taken for som e period before and after each o f its m eetings” to prevent leaks.
An extended discussion o f alternatives for releasing detailed inform ation on the delib­
erations o f the Com m ittee took place at the m eeting o f N ovem ber 16, 1993. The
Com m ittee agreed to authorize “lightly edited” transcripts o f past m eetings and to release
the transcripts to the public five years after the m eetings, “su b ject to the redaction o f espe­
cially sensitive m aterials.”
The issue o f public announcem ents took on greater prom inence at the first m eeting of
1994, w hen the Com m ittee decided to announce the short-term policy decision prom ptly
after the m eeting. T he purpose o f this announcem ent was to “avoid any m isinterpretation
o f the Com m ittee’s action and its purpose. Because this would be the first tightening policy
action... [since early 19 8 9 ,] it was likely to attract considerable atten tion.” Com m ittee
m em bers were careful to point out that they did not consider this announcem ent to set any
precedents for future announcem ents.
Nevertheless, each o f the subsequent changes in policy during 1 9 9 4 were followed by a
b rief announcem ent at the conclu sion o f the meeting. T he announcem ents were generally
brief, but gave qualitative inform ation regarding the nature o f the policy decisions, and also
gave an indication as to the m agnitude o f federal funds rate alterations that would be asso­
ciated w ith the changes. For instance, a statem ent follow ing the m eeting o f August 16,
19 9 4 , com bined the announcem ent o f a V percentage point increase in the discount rate
w ith an announcem ent that the FO M C had decided that “this increase would be allowed to
show through com pletely into interest rates in reserve m arkets”.
At the m eeting o f Ju ly 5 -6 , 1 9 9 4 , the Com m ittee addressed the issue o f announcing the
outcom e o f a d ecision to leave policy unchanged. T he m em bers agreed to “provide a brief
and informal indication that the m eeting had ended and that there would be no further
announcem ents.” A sim ilar announcem ent o f “no further announcem ents” was released at
the conclusion o f the Septem ber and D ecem ber m eetings. In early 1 9 9 5 , the Com m ittee
endorsed the practice o f having the Chairm an issue a b rief statem ent describing policy
actions after each m eeting as a regular practice.

1All quotes in this shaded insert ore
token from various issues of the
Federal Reserve Bulletin.




F ig u r e 5

on reserve positions” on six separate occasions
in 1 9 9 4 , after leaving policy unchanged over
the course o f 1 9 9 3 . T he m ost readily observ­
able response to these developm ents has been
the rise in short-term interest rates. It is often
asserted that the Fed is responsible for pushing
short-term interest rates higher, and that b oth
the intent and the effect of these rate increases
is to slow econom ic growth. On the other
hand, interest rates reflect the balance of supply
and dem and in credit m arkets. H ence, when
econom ic activity is accelerating and credit
demands rising, m arket forces should be
expected to push interest rates higher and
the Fed’s actions in the m arket for bank
reserves could be interpreted as allowing
those m arket forces to w ork. F o r econom ists
evaluating the im pact o f the FO M C ’s policy
decisions, the distinction betw een these two
perspectives is o f great im portance. Is the
Fed actively attem pting to m anipulate the
course o f the economy, or merely adjusting
the settings o f its policy instrum ents to m eet
evolving econom ic conditions?
O ne can take several approaches to
addressing this question, none o f w hich is
entirely satisfactory. Perhaps the sim plest
approach is to exam ine the behavior o f the raw
data sum m arizing FO M C policy actions— the
instrum ents or proxim ate targets o f policy.
Figure 6 illustrates the recent behavior o f the
federal funds rate, the m ost widely m onitored
measure o f the stance o f m onetary policy.
After declining from 1 9 8 9 through 1 9 9 2 , the
funds rate rem ained fairly stable at around
3 percent during 1993 and then gradually rose
to 5 .5 percent during 1 9 9 4 . It is the increase
in this key short-term rate w hich m ost
observers point to as a m easure o f the delib­
erate tightening o f m onetary policy in 1994.
However, a perusal o f the behavior of
longer-term interest rates illustrated in Figure 7
show s that m any interest rates began to rise
before the FO M C ’s first policy adjustm ent in
February 1 9 94 . Long-term interest rates
reached their lows during Septem ber and

Y ie ld C u rves



structure. By April 2 9 , after the first three
3 point increases in the federal funds rate,
the three-m onth and 30-year yields had risen
by roughly equivalent magnitudes. Yields on
interm ediate-term m aturities had risen by
som ew hat greater am ounts, suggesting that
investors expected that further increases in
short-term rates were likely in the near future.
By early November, the yield curve had shifted
further, but with the spread betw een long-term
and short-term yields narrowing: From the
beginning o f the year, three-m onth yields
had risen by m ore than 2 percentage points,
while the yields on 30-year bonds had risen
about 13 percentage points. The response of
the term structure to the 75 basis point increase
in the federal funds rate on Novem ber 15 is
even more striking. Longer m aturity yields
actually declined follow ing that change, and
continued to trend downward until the end
o f the year. This unusual pattern o f rate
m ovem ents— and the flattening o f the yield
curve that they represent— suggest that by
the end o f 1 9 9 4 , inflationary expectations
were responding favorably to the cum ulative
im pact o f the FO M C ’s policy moves.


O ctober o f 1 9 9 3 , and rose through m ost o f
19 9 4 . Figure 8 suggests a reason for the
upward pressure on rates: Dem and for credit,
represented by the volum e o f com m ercial
bank loans, picked up dram atically during
the latter part o f 1 9 9 3 . Taking these devel-

As described in the previous section, the
FO M C acted to “increase the degree of pressure

N K OP S T . L O U I S



opments into consideration, the FO M C’s policy
approach in 1 9 9 4 m ight be m ore accurately
described as one o f n o t preventing a natural
increase in interest rates, rather than one o f
deliberately pushing rates higher.
A nother raw m easure o f the thrust o f
m onetary policy is the growth rate o f non­
borrowed reserves. Although the FO M C
itself does not presently define its policies in
term s o f reserve growth, the supply o f non ­
borrowed reserves is directly affected by the
Fed’s open m arket operations. Changes in the
demand for reserves largely reflects fluctuations
o f the checkable deposits com ponent of M l.
As illustrated in Figure 9 , a reserve-based view
suggests that policy in 1 9 93 was n ot as static
as suggested by the stability o f the fed funds
rate. Rather, to m aintain a stable funds rate,
reserve growth was allowed to fluctuate rather
widely throughout the year. Figure 9 also
show s that by recent historical standards, the
average growth rate o f reserves in 1993 was
quite rapid. Reserve growth dropped off
sharply during 1 9 9 4 , turning negative in the
latter part o f the year.
E conom ists w ho take a narrow -m oney
approach view the rapid growth in non bor­
rowed reserves, the m onetary base and M l in
1993 as suggesting that policy was not neutral.
Such a view is based on the notion that rapid
growth in m oney w ill, w ith a lag, cause rapid
growth in aggregate demand. According to
this view, policy m ight be characterized as
being highly expansionary in 1 9 9 2 and 1993,
w ith an abrupt reversal in 1994. Such stopand-go policy, illustrated in Figure 9 by wide
fluctuations in the growth o f nonborrow ed
reserves, is thought by som e to exacerbate
the business cycle.
Given these som ew hat disparate indica­
tions from the proxim ate targets o f monetary
policy and the am biguity o f distinguishing
betw een deliberate changes in the stance of
policy from endogenous responses to broader
econom ic developm ents, econom ists have
sought to develop m ore specific m ethods o f
identifying m ajor Federal Reserve policy shifts
and distinguishing them from m inor policy
adjustm ents. O ne approach, pioneered by
Fried m an and Schwartz (1 9 6 3 ) and recently
extended by Rom er and R om er (1 9 8 9 ), is the
“narrative approach.” This approach seeks to


F ig u r e 6

F e d e ra l Funds R ate and Discount R ate

F ig u r e 7

Selected In terest R ates

identify discrete shifts in policy by exam ining
qualitative m easures o f policy: for instance,
the statem ents issued by the FO M C and its
members. Romer and Romer developed criteria
for distinguishing turning points in policy,
w hich identify a policy “sh o ck ” as a situation
“in w hich the Federal Reserve attem pted to
exert a contractionary influence on the econ­
om y in order to reduce in flation.”3 These
events can be thought o f as deliberate changes
in the overall thrust o f policy.
Although many observers m ight charac­
terize the FO M C ’s policy decisions in 1 9 9 4 as
constituting such a policy sh ock , it is hard to
ju stify this conclu sion using the R om er and
Rom er criteria. In particular, the Rom ers
exclude from their classification episodes in
w hich the FO M C acted to prevent the em er-

33 Romer ond Romer (1989, p 134).

Another approach to characterizing policy
follows a statistical m ethodology to isolate
w hat is know n as a Federal Reserve reaction
function. By exam ining historical data, this
approach attem pts to identify com ponents o f
FO M C policies w hich are predicable reactions
to emerging econom ic data. After controlling
for these factors, the m ovem ents in the Fed ’s
instrum ents w h ich rem ain unexplained are
interpreted as constituting policy innovations
or shocks.
Figure 10 illustrates the identification
of innovations using a m odel suggested by
Bernanke and Blinder (1 9 9 2 ). In the model
used to generate the shocks illustrated in
Figure 10, the federal funds rate is used as the
measure o f policy, and the Fed’s reaction func­
tion is assumed to depend on a measure of infla­
tion (as m easured by the C PI) and a m easure
of real econom ic activity (th e unem ploym ent
rate for m ales ages 2 5 -5 4 ). The FO M C ’s
reaction function is estim ated to depend on
inflation and unem ploym ent over the prior
six m onths, w ith the rem aining m ovem ents
of the fed funds rate taken to b e exogenous
policy innovations (th at is, deliberate acts by
the FO M C , rather than standard responses to
em erging econom ic developm ents).3
T h e series o f innovations illustrated in
Figure 10 is m uch m ore variable than one
m ight ordinarily associate with deliberate
FO M C policy changes.3 T h e innovations
appear more to reflect random variability in the
series than deliberate, discrete policy changes.
It could be argued, however, that it is the
cumulative effect o f sm all innovations to the
funds rate w hich are im portant in evaluating
the overall thrust o f m onetary policy.
Figure 11 show s the cum ulative im pact
o f the innovations identified in Figure 10.
W h en the sh ocks are added up over time,
they provide a m ore readily interpretable
accou n t o f the thrust o f m onetary policy,
w ith easily identifiable turning points. F o r
exam ple, this cum ulative m easure suggests
that policy was roughly constan t from 1985
through 1 9 8 7 (a period w hen the fed funds
rate itself was generally falling), then tightened
rather dram atically during 1 9 8 8 . T h e period
from 1 9 8 9 through 1 9 9 2 is characterized by
a gradual easing o f policy. Note, however, that
the stability o f the fed funds rate during 1993

F ig u r e 8

Com m ercial B a n k Loans
Consum er, an d Com m ercial & In d u strial Loans
Billions of dollars

Billions of dollars

34 Romer and Romer (1989, p. 138).
35 Romer ond Romer (1989, p. 138).
36 Technically, the shocks in Figure 10
are the innovations to the funds rote
equation in a three-variable, unre­
stricted vector autoregression (VAR),
estimated using six lags of monthly
37 This criticism of the VAR approach
to estimating policy functions has
been made before; see, for example,
Cecchetti (1994).

gence o f inflationary pressures (as opposed
to responding to current inflation). F or
exam ple, they specifically exclude an episode
in 1966, in which “the Federal Reserve’s stated
intent was clearly not to reduce aggregate
demand, but rather to prevent outward shifts
in aggregate demand that it believed would
otherw ise have occurred .”3 T he sim ilarity to
recent events is evidenced by the widespread
interpretation of the Fed’s 1 9 9 4 policy actions
as being preemptive in nature. Ju s t as the
Rom ers describe for 1 9 66 , “the perception of
the econom y’s strength was based not ju s t on
current data bu t also on projection s... .”3 E x
post, analysts who follow a narrative approach
to characterizing Fed policy m ight consider
1 9 9 4 to constitute a policy shock, b u t it is
n ot apparent that it is w hen one applies the
ex ante criteria o f the Rom ers.



is not associated w ith an unchanging policy
thrust using this measure. Rather, the relatively
low level o f the funds rate in 1993 is associ­
ated with a series o f negative shocks w hich,
w hen cum ulated, suggest that the policy was
increasingly stim ulative throughout the year.
In 1994, innovations to the funds rate
were generally positive. N ote, however, that
the cumulative im pact o f innovations in 1 9 94
did not nearly approach the level o f restraint
im plied by this m easure o f policy for 1987,
when the federal funds rate itself rose by far
less than it did last year. It also should be
noted that Figure 11 provides little inform a­
tion on w hich to ju d g e the absolute position
o f a neutral policy stance. Although the
cum ulative position o f the shocks end near
zero, this level should be interpreted as rep­
resentative o f the average degree of reserve
restraint over the estim ation period, 19 5 9 -9 4 .
(In fact, that the cum ulated residuals end the
period at zero is true by con stru ction .) This
period was characterized by an average inflation
rate of 4.75 percent and an unemployment rate
o f m ore than 6 percent. Hence, this level of
cum ulative adjustm ent in the funds rate is
neutral only if these outcom es are deemed
desirable (and if the estim ated equations
are, in fact, stable over tim e).

F ig u r e 9

N onborrow ed R e se rv e s G row th
Percent Change Over Previous 12 months

F ig u r e 1 O

F e d e ra l Funds R ate Inno vatio ns

In spite o f the m arked contrast betw een
the character o f policy actions o f the FO M C
in 1993 versus 199 4 , in a broader context the
actions of the Com m ittee can be interpreted
as part o f a continuing process o f evolution
in the strategy and tactics o f the conduct o f
m onetary policy. T he two-year period was
characterized by a continuing rhetoric sup­
porting the pursuit o f long-term price stability,
w ith policy actions taken in the context of
this objective.
The changing nature o f the U.S. financial
structure and the general econom ic environ­
m ent, however, have made the rigorous pursuit
o f m onetary aggregate objectives m ore diffi­
cu lt to justify; and the past two years have
witnessed the Committee grappling with issues
regarding the appropriate conduct o f policy
in the absence o f reliable signals from various
m oney stock measures.

F ig u r e 1 1

C um ulative Funds R ate In n o vatio n s


Gavin, William T., and Alan C. Stockm an. "A Price Objective for
Monetary Policy," Federal Reserve Bank of Cleveland
Commentary (April 1 , 1 9 9 2 ) .

In the process o f making tactical decisions,
the members o f the Com m ittee reached a con­
sensus early in 1 9 9 4 that the existing policy
stance was one w h ich had rem ained overly
accommodative, and policy actions during 1994
have been made in the context o f adjusting
policy to a less-accommodative posture. Unlike
many other periods w hen the Committee
reacted in response to emerging price pressures,
the policy adjustm ents in 1 9 9 4 were intended
to be preem ptive m oves, designed to head off
what the members viewed as a substantial risk
o f rising inflation. H ence, the lack o f visible
signs o f an increase in inflation should not be
taken as a lack o f ju stification for the FO M C ’s
recent stance, bu t as evidence o f its success.

Joint Economic Committee. Hearing To Examine M onetary Policy and the
Economic Outlook. U .S . House of Representatives, December 7 , 1 9 9 4 .
Jordan, Jerry L. "M ust the Fed Fight G row th?" Federal Reserve Bank of
Cleveland Economic Commentary (July 1 5 , 1 9 9 4 ) .
_ _ _ _ _ _ _ _ "Credibility Begins with a Clear Commitment to Price
Stability" Federal Reserve Bank of Cleveland
Economic Commentary
(October 1 , 1 9 9 3 ) .
M aisel, Sherm an J.
Managing the Dollar. Norton, 1 9 7 3 .
Meulendyke, Ann-Marie. "A Review of Federal Reserve Policy Targets
and Operating Guides in Recent D ecades," in Federal Reserve Bank of
New York, Intermediate Targets and Indicators for Monetary Policy: A
Critical Survey ( 1 9 9 0 ) , pp. 4 5 2 -7 3 .
Orphanides, Athanasios, Brian Reid and Dovid Sm all. "The Empirical
Properties of a Monetary Aggregate That Adds Bond and Stock Mutual
Funds to M 2 ," thisReview (Novem ber/Decem ber 1 9 9 4 ) , pp. 3 1 -5 1 .

Anderson, Richard G. "The Effect of Mortgage Refinancing on Money
Demand and the M onetary Aggregates," this
Review (July/August
1 9 9 3 ) , pp. 4 9 -6 3 .

Papodimitriou, Dimitri B ., and L. Randall Wray. "M onetary Policy
Uncovered," Public Policy Brief. The Jerome Levy Economics Institute
of Bard College, 1 9 9 4 .

Angell, W ayne. "A Single Goal for the FedThe Wall Street Journal
(November 1 6 , 1 9 9 4 ) , p. A 2 8 .

Poole, W illiam , "Keep the M in M onetary Policy," and Capital
Bernanke, Ben S ., and Alan S . Blinder. "The Federal Funds Rate and the (winter 1 9 9 4 ) , pp. 2-4.
Channels of Monetary Transmission," American Economic Review
Ritter, Joseph A., "The FOMC in 1 9 9 2 : A M onetary Conundrum," this
(Septem ber 1 9 9 2 ) , pp. 9 0 1 -2 1 .
Review (M ay/Ju n e 1 9 9 3 ) , pp. 3 1 -4 9 .
Cecchetti, Stephen G. "Distinguishing Theories of the Monetary
Romer, Christina D., and David H. Romer. "Does M onetary Policy
Transmission M echanism ," Economic Policy Conference, this
M atter? A New Test in the Spirit of Friedman ond S chw artz," NBER
Macroeconomics Annual. MIT Press, 1 9 8 9 , pp. 1 2 1 -7 0 .
Collins, S e an , and Cheryl L. Edwards. "An Alternative Monetary
Stone, Courtenay C ., ond Daniel L. Thornton. "Solving the 19 8 0 s '
Aggregate: M 2 Plus Household Holdings of Bond and Equity Mutual
Velocity Puzzle: A Progress Report," this
Review (August/Septem ber
Funds," thisReview (Novem ber/Decem ber 1 9 9 4 ) , pp. 7 -29.
1 9 8 7 ), pp. 5 -23.
Committee on Banking, Housing, and Urban Affairs. Federal Reserve's
Tinsley, P., J . Berry, G. Fries, B. Gorrett, A. Norman, P.A.V.B. S w am y ond
Second Monetary Policy Report for 1 9 9 3 . U .S. Senate, July 2 2 , 1 9 9 3 .
P. Von Zur Muehlen. "The Impact of Uncertainty on the Feasibility of
Humphrey-Hawkins O bjectives," Journal of Finance (M ay 1 9 8 1 ) ,
Cummins, Claudia. "Can Adjustments Help Lagging M 2 to Measure U p ?"
pp. 4 8 9 -9 6 .
American Banker, Vol 1 5 7 , No. 2 2 8 , (November 3 0 , 1 9 9 2 ) , p. 1.
Federal Reserve Bank of St. Louis. "A Price Level Objective for Monetary
Policy," 1994 Annual Report (forthcoming).
Federal Reserve Bulletin. Various issues.

Federal Reserve Reform Act Public Low 9 5 -1 8 8 .
Feinman, Joshua N., and Richard D. Porter. "The Continuing W eakness in
M 2 ," Finance and Economics Discussion Series
Working Paper No. 209,
(Board of Governors of the Federal Reserve System , September 1 9 9 2 ).
Friedm an, Milton, and Anna Jacobson Schwortz. Monetary History of
the United States, 1867-1960. Princeton University Press, 1 9 6 3 .
Full Employment and Balanced Growth Act of 1 9 7 8 , Public Law 9 5 -5 2 3
(H.R. 5 0 ) , October 2 7 , 1 9 7 8 .




A p p e n d ix

February 2-3, 1993

able n ot only to arrest the possible emergence
of greater inflation b u t especially to prom ote
further disinflation.”

At the outset o f 1 9 9 3 , the inform ation
available to the FOM C suggested that economic
activity had picked up sharply toward the end
o f the previous year. Com m ittee m em bers
cited num erous conditions that led them to
believe that the expansion would continue
throughout 1993: “Structural im pedim ents
to the expansion seem ed to be dim inishing
as the financial cond ition o f households,
business firm s, and financial institutions
continued to im prove.” D eficit reduction
program s, expected to be announced by
President C linton, were additional signs
that interest rates could fall.
Nevertheless, it was noted that “the
ou tlook rem ained su b ject to a good deal o f
uncertainty.” The Com m ittee agreed, unani­
mously, that im m ediate policy should be “to
m aintain the existing degree o f pressure on
reserve positions.” The directive left open
the possibility o f accepting either greater or
lesser reserve restraint, should conditions
during the interm eeting period warrant.

May 18, 1993
At the May 18 m eeting, the FO M C was
presented w ith staff projections w hich sug­
gested that “econom ic activity would grow at
a m oderate pace and that such growth would
foster a gradual reduction in margins of unem ­
ployed labor and capital.” T his analysis
included portions o f the C linton A dm inistra­
tion’s fiscal package pertaining to the long run.
The Committee also saw “evidence o f a slower
econom ic expansion and a higher rate of
inflation sin ce late 1 9 9 2 ... .”
“In the view o f a m ajority o f the members,
wage and price developm ents over recent
months were sufficiently worrisome to warrant
positioning policy for a move toward restraint
should signs of intensifying inflation continue
to multiply.” Nevertheless, “som e m em bers
preferred to retain a directive that did not
incorporate a presum ption about the likely
direction o f a change in policy...during the
interm eeting period. They were concerned
that adopting a biased directive m ight prove
to be an overreaction to tem porary factors
and to a short-lived upturn in inflationary
sentim ent that was not warranted by under­
lying econom ic cond itions.” In the end, the
Com m ittee adopted an asym m etric directive
w hich suggested an inclination toward
greater reserve restraint rather than lesser.
There were two dissents from this decision,
which reflected widely divergent perspectives.
Mr. Boehne saw the adoption o f a biased
directive as being unwarranted, since “under­
lying econom ic conditions did n o t point
toward an extended period o f higher inflation.”
In contrast, Mr. Angell dissented “because he
believed that the persisting indications of rising
inflation...called for a prompt move to tighten
m onetary policy.”

March 23, 1993
A review o f recent econom ic activity at
the M arch m eeting found the expansion con ­
tinuing at a m oderate pace in the first few
m onths of 1993, after strong gains during the
latter part of 1992. Short-term market interest
rates remained relatively unchanged though
long-term rates fell substantially. It was noted
that in early M arch, “Treasury bonds and
conventional fixed-rate m ortgages reached
their low est levels sin ce 1 9 7 3 .”
The policy directive adopted by the
Committee called for “maintaining the existing
degree o f pressure on reserve positions.”
Again, the Com m ittee decided upon a sym­
m etric directive for guiding policy during
the interm eeting period.2
Governors Angell and Lindsey dissented
from the Com m ittee’s decisions because they
were concerned about the inflation outlook.
They favored “an im m ediate m ove to tighten
reserve conditions...Such an action was desir­

July 6-7, 1993
In the discussion o f short-term policy
at the Ju ly m eeting, m ixed signals on the


1 All direct quotations cited in this
appendix are drawn from the
"Minutes of the Federal Open
Market Committee," os reported in
various issues of the Fecfera/
Reserve Bulletin.

2 A "symmetric" directive is worded
even-handedly with respect to pos­
sible modifications to policy during
the intermeeting period. A socalled "asymmetric" directive is
one which suggests a preferred
direction for policy changes, and is
indicated by the use of the words
"might" and "would," with
"would" considered to be the
stronger of the two terms. For
example, a directive which states
"somewhat greater reserve
restraint would be acceptable, and
somewhat lesser reserve restraint
might be acceptable" leans in favor
of greater reserve restraint. See
Ritter (1993).


perform ance o f the econom y and questions
about fiscal policy “contributed to considerable
uncertainty about the ou tlook .” Some m em ­
bers were concerned, however, that “despite
the very sluggish behavior o f the broad m ea­
sures o f m oney thus far this year, m onetary
policy was relatively expansive as evidenced
by a variety o f other indicators including the
growth in narrow m easures o f m oney and
reserves and the very low levels o f m oney
m arket interest rates.” Several mem bers went
on to point out that in the face o f w orsening
inflation expectations, an unchanged policy
could be more accom m odative than intended.
M ost m em bers indicated that there was
“little or no reason to change m onetary policy
in either direction.” Consequently, the degree
o f pressure on reserve positions was left
unchanged. T he Com m ittee also retained the
asymmetric bias towards restraint that was
adopted at the previous m eeting. Mr. Angell
dissented, preferring an im m ediate tightening
o f reserve restraint.

but also appeared to have improved confidence
in financial m arkets.”
A num ber o f factors were cited as sources
o f concern. New taxes associated w ith the
deficit reduction legislation and uncertainties
about health care reform w ere said to have
generated “cautious attitudes among business
executives.” T h e ou tlook for n et exports was
also “cited as a negative factor.”
The Com m ittee decided to m aintain the
short-term policies o f the August m eeting.

November 16, 1993
Inform ation reviewed at the Novem ber
meeting continued to suggest the maintenance
o f a sustained, but moderate expansion. Some
evidence of strengthening was cited. The
Com m ittee, however, noted that “econom ic
activity clearly rem ained sluggish or even
depressed in som e parts o f the country and
overall business attitudes could still be described
as cautious.” Fiscal policy developm ents— in
particular, uncertainty regarding health care
reform and the ongoing retrenchment of defense
spending— continued to be cited as factors
w hich were likely “to inhibit the expansion
over the year ahead.” In hindsight, this view
m ight be characterized as being overly pes­
sim istic: The fourth quarter o f 1 9 9 3 turned
out to be one o f the strongest quarters for
econom ic growth in recent memory, w ith real
GDP rising at a rate o f 6 .3 percent. Data
revealing this strength, however, were n ot
generally available un til early 1994.
In this context, the m em bers o f the
Com m ittee unanim ously agreed to support a
directive w hich called for “m aintaining the
existing degree o f pressure on reserve posi­
tions and that did not include a presum ption
about the likely direction o f any adjustm ent
to policy during the interm eeting period.”

August 17, 1993


At the August m eeting, the m em bers o f
the FO M C saw little inform ation in recent
developments w hich would alter the “outlook
for moderate and sustained growth in economic
activity.” Although m any m em bers noted
that current policy was associated w ith very
low short-term interest rates, there was also
“no com pelling evidence that current m one­
tary policy was fostering credit flows usually
associated with speculative excess or impending
increases in price pressures.”
W ith these considerations in mind,
Committee members agreed “to the desirability
of a steady policy course.” Accordingly, the
Committee voted to “maintain the existing
degree o f pressure on reserve positions.” The
directive gave no indication o f a preference for
altering this stance in either direction during
the interm eeting period.

December 21, 1993
By Decem ber, indicators were beginning
to suggest that econom ic activity had picked
up in recent m onths, w ith strength observed
in consum er spending, durable equipm ent
purchases, con struction and industrial pro­
duction (particularly in the automotive sector).
M eanw hile, price indexes “pointed to little
change in inflation trends.” In their comments
about recent developm ents, Com m ittee

September 21, 1993
At the Septem ber m eeting, Com m ittee
members noted that general econom ic activity
remained moderate at best, with considerable
disparities existing across locales and indus­
tries. D eficit-reduction legislation that was
passed in Ju ly “implied increased fiscal restraint


m v i t w
members observed that the positive signs “had
fostered appreciable improvement in business
and consum er sentim ent... .” M em bers also
recognized, however, that the strengthening
was not geographically uniform and that a
num ber o f factors continued to exert con ­
straining influences. Particular concerns cited
included “balance-sheet rebuilding, business
restructuring and downsizing activities, and
the downtrend in defense spending.”
In discussing the directive for the
upcom ing period, m ost m em bers “indicated
that they could support a directive that called
for maintaining the existing degree of pressure
on reserve positions,” w ith no bias toward
adjusting conditions one way or the other
during the interm eeting period.
Messrs. Angell and Lindsey both dissented,
citing the belief that current policy was overly
accom m odative, and “needed to be adjusted
promptly toward a m ore neutral stan ce.” Mr.
Angell also stressed that the Committee should
focus on “forw ard-looking indicators such as
the price of gold and the estimate of the natural
rate o f interest provided by the yield on fiveyear Treasury notes. He favored an immediate
increase o f 5 0 basis points in the federal
funds rate...Mr. Lindsey com m ented further
that a modest policy move now would appro­
priately signal the Com m ittee’s concern
about the potential for inflation.”

the proposed slight policy adjustm ent at
this point, but m any observed that additional
firming probably would be desirable later.”
T he directive adopted by the Com m ittee at
this tim e, however, retained an unbiased
instruction w ith regard to possible inter­
m eeting adjustm ents.
During the subsequent interm eeting
period, federal funds traded at a rate of
around 3 ‘A percent— approxim ately *A per­
cent higher than the rate that had prevailed
throughout 1993.

March 22, 1994
Information reviewed at the March meeting
indicated that the econom y “expanded appre­
ciably further in the early m onths o f 1994,
despite unusually severe w inter w eather.”
In discussing policy for the upcom ing
period, “all the m em bers supported a further
move toward a less accom m odative policy
stan ce.” As a conceptual objective, it was
agreed that policy should strive toward
reaching a “m ore neutral p osition .” The
m em bers generally concluded that “such a
policy stance was still som e distance away,
and the key issue facing the Com m ittee was
not w hether but by how prom ptly the neces­
sary adjustm ent should be com pleted.”
After a discussion o f the possible m agni­
tude o f policy adjustm ents for the upcom ing
period, the Committee decided to duplicate its
previous policy move, seeking to increase
slightly the existing degree o f pressure on
reserve positions, w ith no explicit asymmetry
in the intermeeting stance. “Messrs. Broaddus
and Jordan dissented because they preferred a
stronger move toward a m ore neutral policy
stan ce.” They viewed recent increases in
long-term interest rates as indicating rising
inflation expectations, and perceived that “the
principal policy risk had become one of remain­
ing accom m odative for too long a period .”
Subsequently, incom ing data suggested
considerable strength in the econom y and

February 3-4, 1994
By the tim e o f the FO M C ’s first m eeting
o f 1994, incom ing econom ic data revealed
that a sharp increase in econom ic activity
had taken place in late 1993 and that the
data available for the early weeks o f the year
“suggested appreciable further gains.”
D uring the Com m ittee’s discussion,
“m em bers generally expressed concern about
a buildup in inflationary pressures...especially
if what they currently viewed as a very accom ­
modative m onetary policy were m aintained.”
W ith regard to policy for the upcom ing peri­
od, mem bers “favored an adjustm ent toward
a less accom m odative policy stance, though
views differed to som e extent w ith regard to
the am ount o f the ad ju stm ent.”
After discussing options involving the
magnitude o f possible policy adjustm ents, “all
the mem bers indicated that they could accept

“indications that financial m arkets were less
likely to be destabilized by a further policy
action.” Against this background, on April 18,
“the degree o f accom m odation in reserve
pressures was reduced a little further.” Each
of the policy moves resulted in federal funds
rate increases o f about ‘A percent.



May 17, 1994

The Com m ittee generally agreed that “a
prom pt further tightening m ove was needed
to provide greater assurance that inflationary
pressures in the econom y would rem ain sub­
dued.” Consequently, the FO M C approved a
directive which called for “increasing somewhat
the degree o f pressure on reserve p o sition s.”
It was agreed that if the Board o f G overnors
approved a lh percentage p oint increase in the
discount rate (as was expected), that action
should be allowed to be reflected fully in
reserve market conditions. Given that members
generally expected “that a further policy action
was not likely to be needed for som e tim e,”
the directive adopted by the Com m ittee
included a sym m etric instruction regarding
possible interm eeting adjustm ents.

At the meeting of May 1 7 ,1 9 9 4 , the
Committee reviewed evidence “o f consider­
able m om entum in the econom ic expansion.”
M em bers noted that “the expansion over the
first half o f the year was likely to be a little
stronger than had been expected at the time
o f the M arch m eeting.”
In the context of current policy, Committee
m em bers “favored prom pt further action to
remove much of the remaining accommodation
in the stance o f m onetary policy, at least as
measured by real short-term interest rates.”
Consequently, the Committee adopted a direc­
tive w hich called on the O pen M arket Desk
to “increase som ew hat the existing degree of
pressure on reserve positions.” A sym m etric
policy toward intermeeting period adjustments
was adopted. It was agreed that “the adjust­
m ent should fully reflect the '/i percentage
point increase in the d iscount rate that the
Board o f G overnors was expected to approve
later in the day.”3

September 27, 1994
Data reviewed at the Septem ber m eeting
suggested that “the pace o f econom ic expan­
sion remained substantial, though it appeared
to have moderated slightly in recent m onths.”
Moreover, staff projections “suggested that
growth in econom ic activity would slow appre­
ciably over the next several quarters.” Previous
policy moves were seen to have “elicited only
a m ild response thus far in interest-sensitive
sectors o f the econom y,” and output growth
was “near m axim um sustainable levels.” It
was judged that “the risks o f som e rise in
inflation rates probably had increased.”
Nevertheless, m ost o f the Com m ittee
mem bers felt “that the recent evidence did
not warrant an immediate further tightening,”
given that there had been an “appreciable
tightening o f policy approved in A ugust.”
It was expected that incom ing inform ation
during the intermeeting period might “provide
a firmer basis for ju d ging the course o f the
econom y and the risks o f greater inflation.”

July 5-6, 1994

3 Changes in the discount rate are ini­
tiated by the individual Federal
Reserve banks, but must be approved
by the Board of Governors before
becoming effective. This division
of responsibility underlies the rather
odd circumstances surrounding
increases in the federal funds rate
and discount rates in 1994. In
each case, the FOMC endorsed a
policy which incorporated expected
changes in the discount rate, which
had already been proposed by the
Federal Reserve banks and which
wete approved by the Board of
Governors as part of the overall
policy change.


Inform ation reviewed at the Ju ly meeting
indicated that the econom y grew substantially
in the second quarter, but that expansion was
expected to slow som ew hat over the balance
o f the year. Given uncertainty regarding the
extent o f the economy’s slowing and the effects
o f previous policy moves, m ost FO M C m em ­
bers considered “that it would be prudent for
the Com m ittee to assess further developments
before taking any actio n .”
Consequently, the policy directive adopted
for the upcoming period called for “maintaining
the existing degree o f pressure on reserve
position s,” although it also included a bias
toward the possibility o f increasing the degree
o f reserve pressure prior to the next m eeting.
“Mr. Broaddus dissented because he believed
that additional near-term tightening was n ec­
essary to contain inflation.”

Consequently, the Com m ittee approved a
directive that called for “m aintaining the
existing degree o f pressure on reserve posi­
tion s,” b u t w hich also included “a shift from
the sym m etry in the August directive to
asymmetry toward restraint.”
Mr. Broaddus dissented from this directive,
believing “that a prom pt move to som ewhat
greater m onetary restraint was needed at this
p o in t,” given “signs o f increasing price pres­
sures and rising inflationary expectation s.”

August 16, 1994
Although the pace o f the econom ic
expansion remained substantial, inform ation
reviewed by the Com m ittee in August sug­
gested som e slowing. Staff forecasts suggested
“that the econom y was operating close to its
long-run capacity.”




November 15, 1994

slowing in econom ic activity over the next
few quarters, b u t this outlook was predicated
on the assumption “that monetary policy would
not accom m odate any continuing tendency
for aggregate dem and to expand at a pace
that could foster sustained higher inflation.”
In their discussion of econom ic developments,
Com m ittee m em bers “saw scant evidence at
this point o f any m oderation in the growth
o f overall econom ic activity.”
In their discussion o f policy for the
interm eeting period ahead, many m em bers
anticipated that “the need for further monetary
restraint was highly likely.” A majority, how­
ever, advocated no change in policy, at least
through the beginning o f 1 9 9 5 , preferring a
pause in order “to assess the underlying
strength of the econom y and the im pact of
previous m onetary restraint.” Given the
probable need for further tightening at some
point, a m ajority agreed that the directive
should express an asym m etry “tilted toward
Mr. La W are dissented from this directive,
favoring an im m ediate policy tightening. He
cited “high and increasing levels o f utilization
in labor and capital m arkets” as indicating a
risk o f rising inflation, and feared that inaction
by the Committee “could heighten inflationary
expectations by raising concerns about the
System ’s com m itm ent to the objective o f sus­
tainable, noninflationary econ om ic grow th.”

By the Novem ber m eeting, incom ing
inform ation suggested that “growth o f the
economy remained substantial,” and “mem­
bers com m ented on widespread statistical
and anecdotal indications of considerably
greater strength in the business expansion
than they had anticipated earlier.” In this
context, m em bers “saw a considerable risk
o f higher inflation.”
In their discussion of near-term policy, “all
the members agreed that the current stance of
m onetary policy presented unacceptable risks
o f embedding higher inflation in the economy.”
A lthough m em bers “acknow ledged the diffi­
culty o f ju d ging the precise degree o f m one­
tary restraint that would be needed to attain
the Com m ittee’s objectives,” m ost mem bers
advocated “an unusually sizable firming of
m onetary policy.” O thers were reported to
have “preferred a less forceful policy m ove,”
taking a m ore “cautious approach.”
Ultimately, all m em bers ended up sup­
porting a directive calling for a “significant
increase” in reserve pressure, w hich was to
take account of a 3/4 percentage point increase
in the discount rate. Given the relative forcefulness o f this move, the Com m ittee adopted
a directive that was sym m etric w ith regard to
interm eeting adjustm ents, although it was
noted that “a symmetric directive would n ot
prevent an interm eeting adjustm ent if nearterm developm ents differed substantially
from expectations.”

December 20, 1994
Inform ation reviewed at the D ecem ber
m eeting suggested “a further pickup in eco­
nom ic growth in recent months.” The forecast
presented by the staff suggested a marked






D avid C. W heelock is a sen ior econom ist at the Fed eral R e se rv e B a n k of St. Louis. H eidi L. B ey e r provided research assistan ce.


R egulation,
M a rk e t
and the Bank
Failures of
the G re a t

em pirical study o f the effects o f banking
m arket structure and regulation on failures
during this period, a gap w hich this article
attem pts to fill.
Previous studies have taken little notice
o f the wide interstate variation in the num ber
o f failures and failure rates during the Great
Depression. This article investigates whether
this variation can be explained solely by dif­
ferences in the exten t to w h ich incom e
declined, or w hether various state banking
policies or differences in m arket structure
contributed to interstate variation in failure
rates. It also investigates why banking market
structures differed across states. The analysis
indicates that, after controlling for the extent
to w h ich econom ic activity declined, the
proportion o f deposits in failed banks was
low er in states where branch banking was
more prevalent. In addition, both the bank
failure rate and proportion o f deposits in
failed banks varied inversely w ith the relative
num ber o f federally chartered (national)
banks in a state. Finally, the study shows
that the state deposit insurance system s of
the 1 9 2 0 s had lingering effects on banking
m arket structures even after insurance had
ended. Thus, as researchers have found for
the 19 8 0 s, governm ent policies, such as
branching restrictions and deposit insurance,
appear to have had m easurable im pacts on
m arket outcom es and bank failures during
the Great Depression.

D avid C. W heelock
he surge o f bank failures in the United
States during the 1980s focused the atten­
tion o f policym akers and researchers on
the causes o f failure, especially on the role
o f governm ent policy. D eposit insurance had
left the banking industry m ore leveraged
than it would otherw ise have been, and
encouraged individual banks to take greater
risks as losses eroded their n et worth. In
response, regulators im posed risk-adjusted
capital requirem ents and Congress enacted
the Federal D eposit Insurance Corporation
Im provement Act o f 1991 (FD IC IA ), w hich
mandated risk-based deposit insurance
premium s and refined the risk-based capital
standards.1 Similarly, the enactm ent o f The
Interstate Banking and Branching Efficiency
A ct o f 1 9 9 4 , w hich perm itted interstate
branching, stem m ed from the view that
branching restrictions ham per geographic
diversification and had contributed to the
high num ber o f failures in regions suffering
econom ic downturns.
The United States last experienced high
num bers o f bank failures during the Great
Depression, w hen som e 9 ,0 0 0 banks failed.
Researchers have blamed various governm ent
policies, especially branching restrictions, for
contributing to banking instability during the
D epression. There has, however, b een little


From 1 9 2 9 to 1 9 3 3 , U .S. gross national
product declined 2 9 percent (in constant
d ollars), the price level fell 2 5 percent, the
unem ploym ent rate reached 25 percent,
and som e 9 ,0 0 0 banks suspended operations
because o f financial distress. A bank that
suspended operations need not have “failed,”
in that a receiver need not have been appointed
to liquidate the bank. Suspended banks,
however, include only those that closed on


1 FDICIA also limited the discretion of
regulators to permit insolvent banks
from continuing to operate. Some
researchers argue that the closure
policy known as "too-big-to-fair hod
encouraged excessive risk-taking
because it had the effect of expand­
ing deposit insurance coverage
beyond $100,000 per account at
banks that regulators deemed too
large to close. See Keeley (1990)
for further analysis of the role of
deposit insurance during the 1980s.


accou nt o f financial difficulty. Follow ing
m uch o f the literature, I use the term s “sus­
pension” and “failure” interchangeably.
E conom ists have debated the causes of
the D epression since the 1930s. In the past
3 0 years, this debate has focused on the role
o f bank failures. In Monetary History o f the
United States, Friedm an and Schw artz (1 9 6 3 )
argue that banking panics in the autum n o f
1 9 3 0 , and the spring and autum n o f 1931
sharply reduced the supply o f money, w hich,
in turn, caused econom ic activity to decline.
O ther researchers, however, such as Temin
(1 9 7 6 ), contend that bank failures occurred
largely as a result o f falling national incom e.
In Temin’s view, the econom ic dow nturn
reduced the dem and for money, and bank
failures were the m eans by w hich the m oney
supply fell to accom m odate that decline.
T h e debate over the role o f bank
failures and m onetary forces in causing the
Great D epression continues to sim mer, and
is reviewed by W h eelock (1 9 9 2 b ). A recent
view, originating w ith Bernanke (1 9 8 3 ), pro­
poses a non-m onetary explanation o f how
bank failures contributed to the Depression.
Bernanke argues that apart from their im pact
on the m oney supply, bank failures depressed
output by raising the cost o f credit interm e­
M uch o f the research on the causes
and consequences o f bank failures during the
Depression has had a m acroeconom ic orien­
tation, w ith little em phasis on the role o f reg­
ulation or m arket structure. Some
researchers, however, have argued that the
prevalence o f un it banking left the U.S.
banking system especially vulnerable to
failures during the Depression, and that
nationw ide branching helped lim it failures
and banking panics in other countries.
F o r exam ple, the conventional view is that
nationwide branching protected the Canadian
banking system during the Depression (for

lations or m arket structure contributed to
interstate differences in b ank failure rates.
Regional variation in failures has largely been
ignored or sim ply attributed to differences in
the extent to w hich incom e declined.
Several studies have attem pted to
determine whether the causes o f bank failures
during the D epression were like those o f fail­
ures during the 1920s. F o r exam ple, Temin
(1 9 7 6 ) finds that, like the 1 9 2 0 s, declining
agricultural incom e explains m any o f the
failures o f 1 9 3 0 and 1931. W h ite (1 9 8 4 )
shows that the characteristics o f banks that
failed in 1 9 3 0 were like those o f previous
failures. Calomiris and M ason (1 9 9 4 ) present
similar findings for failures during the Chicago
banking panic o f Ju n e 1 9 3 2 . On the other
hand, W ick er (1 9 8 0 ) show s that m any fail­
ures in 1 9 3 0 stem m ed from the collapse of
one Southern financial institu tion, Caldwell
and Company, w hich he concludes was largely
independent o f the decline in econom ic
activity. Stauffer (1 9 8 1 ) offers further evidence
that bank failures were independent o f the
decline in activity by show ing that in the
11 cotton-producing states w ith significant
declines in output, bank failures were m ore
closely related to banking m arket structure
than to changes in local incom e. W h eth er
this was also true o f other states, however,
is unclear.

This article investigates the interstate vari­
ation in bank failures during 1 9 2 9 -3 2 . The
failures o f 1 9 3 3 are not studied here because
the bank holiday in March 1 9 3 3 , and subse­
quent institutional changes, substantially
altered the timing and likely causes of failures.
All banks were shut during the bank holiday,
and only those licensed by regulators were
permitted to reopen. Not all banks that would
reopen had done so by the end o f 1 9 3 3 , and
some that did were later found to be insolvent.

exam ple, see W hite, 1 9 8 4 ; or Grossm an,
1994), though Kryznowski and Roberts (1 9 9 3 )
estim ate that on a m arket value basis, all
Canadian banks were insolvent at som e
point during the Depression. T his focus on
branching vs. unit banking has been national,
with little consideration of whether differences
in state branching laws, other banking regu­


T his suggests that the determ inants o f bank
failures in 1933 should be studied apart from
those o f other Depression years. Similarly, I
leave for future research the causes o f failures
during the rem ainder o f the 1930s.



Figure 1 shows the distribution of bank
failures across the United States during 1929-32
(see the appendix for data sources). Rhode
Island escaped the period w ithout any bank
failures. No other state had fewer than two
failures. O ther states with fewer than 10 bank
failures include Vermont, M aine, New Hamp­
shire, Delaware, New M exico and W yom ing.
Generally, M idwestern states suffered the
highest numbers o f bank failures. Illinois had
6 02 failures, the m ost of any state. Only three
other states had more than 3 0 0 failures: Iowa
with 4 7 6 , Nebraska w ith 3 5 8 and M issouri
with 328. The mean number of failures across
all states was 120, and the m edian was 91
failures. For com parison, from 1980 to 1989,
the two states with the most bank failures were
Texas w ith 3 5 0 and O klahom a w ith 105.
The num ber of failures can, o f course, be
a m isleading statistic because the num ber of
banks varies widely across states. Figure 2
maps the distribution o f bank failure rates
during 1 9 2 9 -3 2 , in w hich the annual failure
rate is defined as the total num ber o f suspen­
sions during a year divided by the num ber of
banks operating at mid-year. Even though
Illinois had the m ost failures, it did n ot have
the highest failure rate. That dubious dis­
tinction w ent to Nevada, w hich had a yearly
average failure rate o f more than 16 percent,
despite having ju s t 19 bank failures during
the period. Illinois, other M idwestern and
Southern states w ith high num bers o f bank
failures, however, generally also had high
failure rates. Besides Nevada, other states with
high failure rates included South Carolina,
Florida and Arkansas, each w ith a rate of
15 percent. At the other extrem e, five New
England states, plus New M exico, W yom ing,
New York, M assachusetts and New Jersey, all
had failure rates under 3 percent. The mean
failure rate am ong all states was 6 .6 percent,
while the m edian was 5 .5 percent. F or com ­
parison, betw een 1 9 8 0 and 1 9 8 9 , the average
annual bank failure rate in the United States
was 0 .7 7 percent. Eight states had no failures
during the period, while Alaska, Oregon and
Texas had failure rates o f 6.3 percent, 2 .4 per­
cent and 2.3 percent, respectively, the m ost
o f any states.
Figure 3 maps the average annual rate of
deposits in failed banks during 1929-32, where


F ig u r e 1

Number of Bank Suspensions, 1 9 2 9 - 3 2

Bank Suspensions P er A ctive B an k ,
1 9 2 9 -3 2

the annual rate o f deposits in failed banks is
the sum o f deposits in failed banks during a
year divided by the volum e o f deposits in all
banks at mid-year. A state could have had a
low num ber o f bank failures, or a low failure
rate, but a high rate o f deposits in failed banks
if those banks that did fail held a high share
o f the state’s bank deposits. On the other
hand, a high num ber o f failures, or a high
failure rate, did n ot necessarily produce a
high rate o f deposits in failed banks if failing
banks held a com paratively low share o f a
state’s deposits. Moreover, there is no reason
to expect that the determ inants o f the bank
failure rate and rate o f deposits in failed
banks will be the same.
During 1 9 2 9 -3 2 , the rate o f deposits in
failed banks and the bank failure rate were
highly correlated (a correlation coefficient




able to expect that failure rates were higher
in states suffering the largest incom e declines.
Conceivably, incom e fell m ore in som e
states because o f a high rate o f b ank failures.
Some researchers argue that banking panics
triggered the decline in national incom e, while
others contend that bank failures merely
reflected falling incom e caused by other
forces. Ideally, an econom etric analysis
of the determ inants o f b ank failure would
treat the change in incom e as sim ultaneously
determ ined w ith bank failures. The specifi­
cation o f such a system in this con text pre­
sents a num ber o f challenges and is therefore
left to future research. Am ong the difficulties
is a lack of suitable variables to serve as instru­
ments for state-level changes in per capita
incom e. For exam ple, state incom e estim ates
prior to 1 9 2 9 are not available. Readers are
cautioned that the models presented here may
be su b ject to sim ultaneous-equations bias.
As in the 1920s, the m ajority o f banks
that failed during the D epression were small
rural banks whose prosperity depended largely
on agriculture. A lston, Grove and W heelock
(1 9 9 4 ) show that differences in farm foreclo­
sure rates explain m uch o f the interstate vari­
ation in bank failure rates during 1926-29.
O ther studies, including Friedm an and
Schwartz (1 9 6 3 ) and Temin (1 9 7 6 ), note a
relationship betw een agricultural distress
and bank failures in the 1930s, bu t do n ot
exam ine w hether falling agricultural incom e
dom inates other possible explanations o f
failures. An exception is Stauffer (1 9 8 1 ), but
he focuses exclusively on 11 Southern states.
M oreover, even though falling agricultural
incom e m ight explain a high num ber o f bank
failures during the D epression, because the
rural banks that failed in such large num bers
were typically quite small, agricultural distress
m ight not explain the proportion o f deposits
in failed banks.
Apart from the severity o f agricultural
distress, the preponderance o f failures among
very small banks suggests that bank size itself,
or som e other characteristic o f sm all banks,
might explain their relatively high failure rate.
T he failure rate in 1 9 3 0 -3 1 o f national and
state-chartered banks was inversely correlated
w ith bank size, declining from 25 percent of
active banks on Ju n e 3 0 , 1930, for banks with

F ig u r e 3

Deposits in Suspended B anks Per D ollar of
Deposits in Active B an ks, 1 9 2 9 - 3 2


o f 0 .8 5 ) and Nevada again had the highest
rate o f deposits in failed banks at 16 percent.
Still, com parison o f Figures 2 and 3 reveals
that n ot all states w ith high bank failure rates
also had high rates of deposits in failed banks,
and that some states with relatively low failure
rates had high rates of deposits in failed banks.
C onnecticut, for example, had a relatively low
bank failure rate (3 .6 percent), but a relatively
high rate o f deposits in failed banks (3 .9 per­
cent). O n the other hand, Georgia had a high
failure rate (8 .0 percent), but a com paratively
low rate o f deposits in failed banks (1 .6 per­
cent) because m ost o f the banks that failed in
Georgia were quite small. Besides Nevada,
other states with high rates o f deposits in failed
banks included South Carolina, Florida, North
C arolina, Iowa, M ississippi and Arkansas, all
w ith rates above 7 percent. States w ith low
rates o f deposits in failed banks include those
in the N ortheast, California and scattered
others. T he m ean rate across all states was
3 .4 percent, while the median was 2.1 percent.
W hat explains interstate differences in
bank failure rates and in the rate of deposits in
failed banks? One hypothesis is that banking
distress was more severe in regions suffering
the largest declines in econom ic activity
because banks in those regions likely experi­
enced the largest losses on their loans and
other assets. The extent to w hich per capita
income fell during 1929-32 ranged from 3 2 per­
cent in M assachusetts to 5 6 percent in
M ississippi (b oth the m ean and median
declines were 4 4 percent). It seem s reason­



in California, w ith som e 3 0 0 belonging to
the Bank o f Italy (the forerunner o f Bank of
A m erica). California had nearly twice as
many branch offices as it had banks. Rhode
Island was the only other state having more
branches than banks.2 If the opportunity to
branch afforded banks greater diversification,
or perm itted them to operate at a m ore effi­
cient scale, states that allowed branching
m ight have had lower bank failure rates.3
A second policy that could have affected
bank failure rates is deposit insurance. Eight
states — Kansas, Mississippi, Nebraska, North
D akota, O klahom a, South D akota, Texas and
W ashington — enacted insurance systems
for their state-chartered banks follow ing the
“Panic of 1 9 0 7 .” In each system , insurance
prem ium s were low and unrelated to failure
risk, thereby creating a subsidy that appears
to have caused more bank entry and greater
risk-taking than would have otherwise occurred
(see Calomiris, 1989, 1992; and W heelock,
1992a, 1 9 9 3 ).
Banks proliferated throughout the United
States in the two decades before 1920. In
1 9 0 0 , the U nited States had 1 2 ,4 2 7 banks.
By 1 9 2 0 , the num ber had reached 3 0 ,2 9 1 ,
thanks in part to rapid growth in agricultural
states during the com m odity price boom of
W orld W ar I (Board o f G overnors, 19 5 9 ).
T he num ber o f banks increased particularly
fast in states w ith deposit insurance systems,
such as N orth Dakota, w hich by 1 9 2 0 had
one bank for every 7 2 0 persons, the m ost
o f any state.
The wartime boom cam e to an end in
1920. Commodity prices collapsed, triggering
widespread bank failures in rural areas. Sub­
sequently, states w ith the highest num bers of
banks per capita in 1 9 2 0 suffered the highest
failure rates, and members o f state insurance
systems had higher failure rates than uninsured
banks. By 1929, each of the state insurance
system s was either insolvent or closed by
state authorities. Because none o f the systems
carried a state guaranty, depositors, rather
than taxpayers, suffered losses if insurance
premiums were inadequate. An exception was
Mississippi, where the state assumed the oblig­
ations o f its insurance system and issued bonds
to reimburse depositors of failed banks. Further
detail about the state insurance system s can

fewer than $15 0 ,0 0 0 o f loans and investments,
to 2 percent for banks with at least $ 5 0 m illion
o f loans and investm ents (Federal Reserve
Board, non-dated publication, p. 6 7 ). If small
banks were less diversified than large banks,
either geographically or along product lines,
they m ight have been more vulnerable to a
dow nturn in a given m arket. F or exam ple,
W hite (1 9 8 6 ) argues that their greater involve­
ment in the securities business m ight have left
large banks better diversified and, hence, less
likely to fail than small banks. Accordingly, a
predom inance o f sm all, undiversified unit
banks might explain the generally higher bank
failure rates o f the rural Midwest and South.
A lack o f diversification might not explain
entirely why the failure rate o f small banks
exceeded that o f large banks. Typically, small
banks had state charters and the failure rate
of state-chartered banks during the Depression
exceeded that o f national banks. In 1929,
the failure rates o f national and state banks
were 0.8 and 3.4 percent, respectively; in 1930,
they were 2.2 and 7.1 percent; in 1931, 6.0 and
12.1 percent; and in 1932, 4 .5 and 8 .7 percent
(Bremer, 1935, p. 4 6 ). D ifferences in regula­
tion or supervision might explain the relatively
high failure rate of state-chartered banks and,
hence, of small banks. F or exam ple, in most
states, national banks had higher m inim um
capital requirem ents and were su b ject to
greater restrictions on real estate lending
than state-chartered institutions.
Apart from differences in the regulation
or supervision o f national and state banks,
other state banking policies might have affected
state banking markets or failure rates. Branch
banking restrictions, for example, can hamper
diversification and, to the extent that the timing
or magnitude o f a decline in econom ic activity
varies geographically, a bank with m ultiple
offices m ight be able to offset losses in one
region with profits in another. Although unit
banking predominated in the United States in
the 1930s, several states perm itted at least
limited branching w ithin their borders. In
1930, nine states, including Arizona, California
and North Carolina, permitted state-wide
branching, and 12 others perm itted limited
branching. Banks in 18 states had no branches
at all. As of Ju n e 1930, U.S. commercial banks
operated 3 ,6 1 8 branches. O f these, 8 5 3 were




2 Aggregate data on branch bonking
ore from the Federal Reserve Board
(December, 1930, p. 8 12 ). The
data for the Bank of Italy are from
Tippetts (1929, p. 335) and are
for 192/.
3 Although not common at Hie time,
multiple-bank holding companies
olso could have provided some geo­
graphic diversification. This article,
however, does not investigate
whether holding componies affect­
ed state failure rates.



be found in Federal D eposit Insurance
Corporation (1 9 5 6 ) or Calom iris (1 9 8 9 ).
Although states w ith deposit insurance
systems had high num bers o f bank failures
during the 1920s, they still had significantly
m ore banks per capita in 1929 than other
states. Generally, the m ore banks per capita
a state had in 1929, the higher its bank failure
rate during 1 9 2 9 -3 2 . (T h e correlation coeffi­
cient is 0 .3 3 , w hich is significant at the 0 .0 5
level). Thus, by affecting the number o f banks
per capita or other aspects of market structure,
or if banks that had been m em bers o f state
deposit insurance system s continued to hold
riskier portfolios, deposit insurance could
have contributed to bank failures during the
Great Depression.


caused more entry and encouraged greater
risk-taking than would have otherw ise
occurred and, hence, the banking system s of
states w ith insurance m ight have been m ore
vulnerable to a decline in econom ic activity.
In other words, failure rates m ight have been
higher because deposit insurance generated
m ore banks than were econom ically viable
once insurance had ended, or because banks
that had been insured continued to hold
especially risky portfolios.
Apart from its im pact on the num ber of
banks per capita, the collapse o f state deposit
insurance systems in the 1920s caused declines
in the num ber o f state-chartered banks rela­
tive to the number o f national banks. In 1908,
the Com ptroller o f the Currency ruled that
national banks could not jo in state deposit
insurance systems. This led to a relative
increase in the num ber and deposit shares
o f state-chartered banks in the states enacting
insurance system s. The decade-long shake­
out o f rural banks that followed the collapse
o f com m odity prices in 1 9 2 0 reduced the
number of state banks. More than 5,700 banks
failed in the ’20s, and Alston, Grove and
W heelock (1 9 9 4 ) show that rural failure rates
were higher in states w ith deposit insurance
system s, after controlling for the extent of
agricultural distress. Moreover, W h eelock
(1 9 9 3 ) finds that the demise o f deposit insur­
ance caused especially large declines in the
relative num ber of state-chartered banks, both
because the rate o f failure am ong insured
state banks was high and because many state
banks sw itched to national charters to escape
state insurance system s. These effects were
especially large in states where the insurance
systems collapsed (o r were closed by state
authorities) early in the decade. F or this
reason, the im pact o f deposit insurance on
market structure, and hence on failures during
the 19 3 0 s, m ight differ in states where insur­
ance ended early in the ’2 0 s from its im pact
in other insurance states. Therefore, I include
one dumm y variable, set equal to 1 in states
in w hich insurance lasted to either 1 9 2 8
or 1 9 2 9 (M ississippi, N orth D akota and
N ebraska), and to zero otherw ise. I set a
second dumm y equal to 1 in states in w hich
insurance ended by the m id -1 9 2 0 s (Kansas,
O klahom a, South Dakota and Texas), and to

The m ain objective o f this article is to
discern w hether differences in state banking
policies contributed to interstate variation in
failure rates during the Great Depression.
Accordingly, in m odeling the determ inants of
b ank failure rates during 1 9 2 9 - 3 2 ,1 control
for cross-state differences in the level o f eco­
nom ic distress by including the percentage
change in per capita incom e, and the average
annual farm and business failure rates as inde­
pendent variables. I expect that the more per
capita incom e fell and the higher the rates of
farm and business failures, the higher were
state bank failure rates and rates o f deposits
in failed banks.
To test w hether w ithin-state branching
helped to lim it failures, perhaps by enabling
greater diversification or scale, I include the
ratio o f branches to operating banks in 1930
as another independent variable. I expect
that failure rates and the rate o f deposits in
failed banks were low er where branching
was m ore prevalent.
N ext, I include dummy variables
reflecting w hether a state had a deposit
insurance system during the 1920s. By
affecting a state’s banking m arket structure,
deposit insurance could have had an im pact
on failure rates during the 1930s even though
insurance no longer existed. Deposit insurance


m a r c h




Table 1

D eterm inants of In terstate V a ria tio n in B an k F a ilu re R ates
Dependent Variable: average failure rate, 1929-32, models 1-4;
Dependent Variable: log of the ratio of deposits to failed banks, 1929-32, models 5-8

(1 .8 3 )*
A Per

capita income

Farm failure rate

Business failure rate


- 8 - 9 8. 8 2
( 1 .7 3 ) *



- 7 .9 1
( 1 .7 7 ) *

- 5 .7 1
( 0 .7 2 )


0 .1 6
(2 .6 3 ) * *

- 0 .7 3 - 0 .1 5
(0 .5 2 )
(0 .1 1 )

Branching ratio
(1 .3 3 )
Dl (MS, ND, NE)
(1 .3 5 )
Dl (KS, OK, SD, TX)
( 2 .8 5 ) * * *

0 .1 6
(1 .8 7 )*


(1 .6 7 )
- 0 .0 3
(0 .0 3 )

- 0 .4 2
(0 .2 6 )


- 4 .4 0
(1 .1 8 )

- 1 .9 1
(0 .5 3 )

- 1 .6 6
(0 .5 0 )

0 .0 3
(1 .1 1 )

0 .0 3
(1 .2 7 )

(0 .3 3 )

-0 .0 0 4
(0 .1 6 )

0 .2 8
(0 .7 1 )

0 .3 4
(0 .8 7 )

0 .3 4
(0 .9 5 )

(1 .0 9 )

- 1 .-31 3. 7 3
(1 .0 2 )

-1 .0 0

(2 .2 1 ) * *

2.31 2 .3 0
(1 .3 7 )

- 0 .7 8
( 1 .8 2 )*

0 .1 6
(0 .2 4 )

In Banks per capita

0 .1 3
(0 .2 1 )

-0 .3 8
(0 .6 5 )

- 2- 39. 6 1
. 9
(2 .2 4 ) * *

0 .0 6
(0 .1 0 )

- 1 .5 2
(1 .0 2 )

In National bank ratio

- 2 .4 0 - 1 .5 0
(2 .1 1 ) * *

(1 .3 8 )
In Average bank size

- 0 .7 6



.8 6


- 0 .0 9
(0 .2 3 )

- 1 .0 3
- 1 .0 0
( 2 .7 6 ) * * * ( 2 .7 3 ) * * *

(0 .8 5 )
Adjusted R2


- 3 .4 7
- 7 .2 2
- 6 .5 4
- 6 .3 8
(4 .9 5 ) * * * ( 4 .1 4 ) * * * ( 4 .3 5 ) * * * ( 1 .2 3 )

- 2 3 .5 7 - 2 6 .9 3
- 2 4 .1 6
- 1 9 .9 2
- 5 .7 9
( 2 .7 9 ) * * * (2 .4 7 ) * *
( 2 .0 3 ) * * (1 .7 3 )*
(2 .4 1 ) * *
( 3 .2 8 ) * * *


.1 5

- 0 .4 1
(1 .3 2 )

Note: The coefficients of models 1-4 are multiplied by 100; absolute values of t-statistks are in parentheses; ***, * * and * indicate statistical signifi­
cance at the .01, .05 and .10 levels. The adjusted R! is presented for use in comparing alternative specifications but, because of the heteroscedasticity
correction, does not indicate the proportion of the variation in the dependent variable explained by models 1-4.

zero otherw ise. The insurance system s of
these states had all ceased to function by 1926,
though, in som e cases, did not officially close
u n til a later date. A lthough W ashington had

however, many o f W ashington’s state banks
sw itched to federal charters.
If deposit insurance left states w ith m ore
banks than were econom ically viable, or with

an insurance system , it collapsed after the
failure o f the first, and largest, insured bank
in 1921. Because of its short life, 1 treat
W ashington as not having had insurance.
Like other states where insurance ended early,

banks having especially risky portfolios, the
coefficients on one or b o th o f these dumm ies
should be positive in the failure rate regres­
sions. On the other hand, if insurance caused
the relative num ber o f banks w ith federal





from such m odels appear to be norm ally
distributed, w hile those from m odels o f the
D eterm inants of In terstate V ariatio n in
level o f the rate o f deposits in failed banks
B an k M a rk e t Structure
do not. Because the value o f this variable for
Dependent Variables: log of banks per capita (equation 1);
Rhode Island is zero, I omitted this observation
log of the ratio of national banks to all banks (equation 2);
w hen estim ating the reported regressions.
log of deposits per bank (equation 3)
Assigning an arbitrarily sm all value to this
observation and re-estimating the models does
not, 8however, substantially alter the results.
- 1 .2 0
- 0 .2 4
9 .0
(2 .2 9 ) ”
(0 .4 6 )
( 2 6 .2 2 ) * * * states with few banks, each bank failure
has a larger im pact on the failure rate than in
states w ith m any banks, w h ich could cause
In Population density
- 0 .1 2
- 0 . 00 .0 5
the errors o f the m odel to be larger in states
(2 .5 3 ) * *
(1 .5 4 ) (1 .6 1 )
w ith fewer banks. H ence, to correct for heteroscedasticity, each variable in the failure
In Farm population
0 .0 6
- 0 .2 9
- 0 .6 7
( 0 .5 5 )
( 2 .6 1 ) * * *
( 9 rate* *models has been m ultiplied by the square
.0 0 ) *
root o f the average annual number o f operating
Branching ratio
- 0 . 7 7 0 .0 6
0 .8 2 banks. W eighting gives m ore im portance to
( 4 .5 8 ) * * *
(0 .3 8 )
( 7 .3 3 ) ” *states, located m ainly in the M idwest
and South, that had large num bers o f banks,
and less to 1states w ith fewer banks (like those
Dl (MS, ND, NE)
0 .4 5
- 0 .2 2
- 0 .3
in the W est and N ortheast).
(0 .9 4 )
(1 .8 8 )*
( 1 .9 5 )*
M odels 1 and 5 include only m easures o f
econom ic activity as independent variables.
Dl (KS, OK, SD, IX )
0 .4 0
0 .3 3
- 0 .3 5
Th ) *
(1 .9 3 )*
(1 .6 2 )
(2 .5 4e *m ore per capita incom e fell during
1 9 2 9 -3 2 , the higher were both the failure
rate and the rate o f deposits in failed banks.
Minimum capital
- 0 .0 0 3
- 0 .00.0003
Typically, states w ith large declines in per
(0 .5 7 )
(1 .5 3 )
(0 .7 8 )
capita incom e also had relatively high farm
Adjusted R2
.91 failure rates, and the correlation coefficient
betw een these two variables is -0 .5 7 , w hich
is significant at the 0 .0 1 level. Nevertheless,
Note: Absolute values of t-statistics are in parentheses; * * * , * * and * indicate statisti­
in the bank failure rate m odel (m odel 1), the
cal significance at the .0 1 , .0 5 and .1 0 levels.
coefficient on the farm failure rate is positive
and statistically significant, indicating th at a
higher farm failure rate caused a higher b an k
charters to be unusually high in these states,
failure rate. T h e coefficients on the business
and if differences in regulation or supervision
failure rate, on the other hand, are never
caused national banks to have low er failure
statistically different from zero.
rates than state-chartered banks, deposit insur­
Models 2 and 6 include the branch
ance m ight have caused fewer Depression-era
banking ratio and deposit insurance dummy
bank failures. Conceivably, the coefficients
variables as additional regressors. In both
on the two insurance dum m ies could differ if
m odels, the coefficient on the branching
the effects o f insurance differed between states
ratio indicates that bank failure rates were
whose systems ended early and those in which
low er where branching was m ore prevalent,
4 The branching ratio is positively cor­
insurance lasted until decade’s end.
though only in m odel 6 is the coefficient
related with both the percentage
Table 1 reports regression estim ates in
statistically significant.4 This suggests that,
change in per capita income and
which the average annual bank failure rate and
w here perm itted, branching lowered bank
business failure rote. If the latter is
the log o f the average annual rate o f deposits
failure rates. Perhaps this occurred because
omitted, the coefficient on the
in failed banks are the dependent variables.
branching banks were b etter diversified, or
branching ratio is larger ond statisti­
I estim ate m odels o f the log o f the rate of
possibly because branching enabled banks to
cally significant in the bank failure
deposits in failed banks because the residuals
rate regression.
achieve econom ies o f scale. States w ith more
Table 2




collinearity m ight explain the absence o f a
significant relationship betw een bank size,
or the num ber o f banks per capita, and failure
rates. The correlation coefficient betw een
the logs o f the national bank ratio and average
deposits per bank is 0 .4 9 , w hich is significant
at the 0 .0 1 level, while that betw een the logs
o f the national bank ratio and num ber of
banks per capita is -0 .3 4 , w hich is significant
at the 0 .0 2 level. T he correlation betw een
the national bank ratio and bank size m akes
it impossible to determine whether differences
in regulation or supervision o f state and
national banks had an im pact on failures,
except as they m ight have influenced bank
size. The absence of a significant relationship
betw een bank size and failure rates, however,
suggests that any influence size had on failures
is reflected in the ratio o f national banks to
all banks.
The inclu sion o f banks per capita, the
national bank ratio and average bank size in
the bank failure rate m odel reduces the sta­
tistical significance o f the percentage change
in per capita incom e and the farm failure
rate. T he correlation coefficients betw een
the m arket structure variables and the two
m easures o f econom ic activity are all statisti­
cally significant at the 1 percent level. The
states w ith the largest declines in per capita
incom e and the highest farm failure rates
also had the highest num bers o f banks per
capita, the lowest national bank ratios and
sm allest average bank sizes. Although the
D epression affected the entire nation, rural
farm ing regions were h it especially hard.
Unfortunately, these states also tended to
have banking m arkets consisting o f many
sm all, undiversified banks. Thu s, it is diffi­
cu lt, if n ot im possible, to apportion the
com paratively high bank failure rates of
these states betw een changes in the level of
econom ic activity and the vulnerability of
their banking system s.' T h e evidence pre­
sented here, however, suggests that banking
m arket structure affected the perform ance of
state banking system s, and adds weight to
other research associating banking distress
and declining econom ic activity in the 1930s
w ith banking system fragility (see Bernanke
and Jam es, 1 9 9 1 ; Calom iris, 1 9 9 3 ; and
Grossm an, 1 9 9 4 ).

branching tended to have larger banks. The
correlation coefficient betw een the branching
ratio and the log o f deposits per bank is 0 .7 2 ,
w hich is significant at the 0 .0 1 level.
The coefficients on the deposit insurance
dummy variables in m odel 6 are n ot statisti­
cally different from zero. In model 2, however,
the dumm y for states where insurance ended
by the m id -1920s (Kansas, O klahom a, South
Dakota and Texas) has a negative and signifi­
cant coefficient. In these states, the average
annual bank failure rate was some 3 percentage
points low er because o f deposit insurance.
Apparently, deposit insurance affected market
structure in a way that reduced failure rates.
This m ight be explained by the comparatively
large increase in the relative number of national
banks, w hich had lower failure rates than state
banks, in these states. M odels 3 and 7 test
for this possibility by including the log o f the
ratio o f national banks to total banks as an
additional independent variable. D oing so
reduces som ew hat the absolute size and sta­
tistical significance o f the deposit insurance
coefficient. Insurance appears to have had
no effect on the national bank ratio in states
where insurance lasted at least until 1928.
The coefficient on deposit insurance for these
states is positive and fairly large (though not
statistically significant), suggesting that insur­
ance affected bank failure rates in these states
by causing excessive num bers of banks or
M odels 4 and 8 further indicate how
banking market structure affected bank failure
rates during the Depression. I exclude
branching and deposit insurance from these
specifications because their effects on bank
failures appear to have worked through their
influence on market structure. Further analysis
of the determ inants of m arket structure is pre­
sented in the n ext section.
The m arket structure measures included
in m odels 4 and 8 are the logs o f banks per
capita in 1929, the ratio o f national to all banks
in 1929, and the average volum e o f deposits
per bank in 1929. O nly the coefficients on
the ratio o f national to all banks is statistically
significant. Its negative coefficients indicate
that bank failure rates and rates o f deposits
in failed banks were sm aller where national
banks were relatively more prevalent. M ulti-



s Conceivably, income fell more and
farm failure rates were higher in
these states because their banking
market structures were more vul­
nerable to bonk failures. As noted
previously, the measures of eco­
nomic activity ideally would be
treated os dependent variables in o
simultoneous-equations framework
to capture any impact of bank fail­
ures on economic activity.





T he same variables are included in a
model o f the log o f the ratio o f the num ber
of national banks to total banks. T he m ost
im portant determ inant o f this ratio is the
ratio o f farm to total state population: the
greater the fraction o f the population in agri­
culture, the lower the relative num ber o f
national banks. N ational banks were more
prevalent in N ortheastern m anufacturing
states and other states where agriculture was
relatively unim portant. O n the one hand,
this reflects the low er population density o f
agricultural states, and that such states often
set low m inim um capital requirem ents for
their state-chartered banks to ensure the
presence o f banking facilities in rural areas.
Note that the coefficient on the m inim um
capital ratio is positive and, for a one-tail
test, statistically significant at the 0 .1 0 level.
State-chartered banks also typically enjoyed
fewer lending restrictions than national banks,
especially on real estate loans. Consequently,
state-chartered banks were able to serve more
of the banking needs o f agricultural borrowers.
In model 2, the coefficients on the two
deposit insurance dum m y variables differ
significantly from one another. The coefficient
on the variable for states where deposit
insurance ended early in the 1920s (Kansas,
Oklahoma, South Dakota and Texas) is positive
and, for a one-tail test, statistically significant
at the 0 .1 0 level. The failure o f large num bers
o f state banks, and the decision o f others
to sw itch to national charters, explain why
deposit insurance had a positive influence
on the national bank ratio in these states.
Insurance lasted longer, and generally per­
formed better, in M ississippi, N orth Dakota
and N ebraska and, h ence, there was no effect
o f insurance on the national bank ratio.
Finally, the coefficient on the branching
ratio is n ot statistically different from zero.
U ntil the M cFadden A ct o f 1 9 2 7 enabled
national banks to open branches, virtually
all branching was done by state-chartered
institutions. T h e ability to branch m ight
have increased the dem and for state charters
and, hence, all else equal, had a negative
influence on the national bank ratio. On
the other hand, it could have also held down
state chartering because branch offices sub­
stituted for independent state banks.

The evidence presented in the preceding
section shows that government policies affected
bank failure rates during the Depression, at
least in part, by causing differences in banking
market structure across states. Further insight
into the effects o f governm ent policies on
m arket outcom es can thus be gleaned from
studying interstate variations in banking
m arket structure.
W h eelock (1 9 9 3 ) investigates the impact
o f governm ent policies on banking m arket
structure during the 1920s. There, I show
that the num ber o f banks per capita was lower
where bran ch banking was m ore prevalent,
in states that im posed high m inim um capital
requirem ents on state-chartered banks, and
in states w ith deposit insurance system s. In
addition, the num ber of banks per capita was
lower in m ore densely populated states. The
costs o f transportation and com m unication
m ake the finding o f an inverse relationship
betw een population density and the num ber
o f banks per capita unsurprising. An inverse
relationship between the prevalence o f branch­
ing and the number o f banks per capita is also
n ot surprising. W here perm itted, branch
offices can serve markets that otherwise would
require independent banks. To the extent that
branches substitute for unit banks, the number
o f banks per capita will be lower. Finally,
because deposit insurance subsidized entry,
and was instituted only in unit banking states,
it caused the number of banks per capita to be
higher than it would otherw ise have been.
M odel 1 o f Table 2 reports a regression
o f the log o f banks per capita in 1 9 2 9 on the
log of population density, the branching ratio,
the deposit insurance dumm y variables, the
log o f the ratio o f farm to total state popula­
tion, and the m inim um capital requirem ent
imposed on state banks. Only the coefficients
on the latter two variables are insignificant.
As expected, the less densely populated a state
was, the higher were banks per capita. In
addition, in states where there were deposit
insurance system s, or where branch banking
was less prevalent, banks per capita were
again higher.



distribution o f bank failures during the
D epression was in part a function o f m arket
structure and governm ent banking policies.

Model 3 o f Table 2 is a regression of
the log o f average deposits per bank in 1929.
This variable is negatively correlated w ith
the num ber o f banks per capita and positively
correlated with the national bank ratio. Hence,
the estim ates o f this model are unsurprising.
Banks were larger in m ore densely populated
states and where agriculture was less im por­
tant. Average bank size was also larger where
branching was m ore prevalent. Apparently,
branching enabled banks to achieve larger
scale than they otherw ise would. Finally,
states w hich had deposit insurance systems
tended to have, on average, sm aller banks.
These were uniform ly rural states that prohib­
ited branching. D eposit insurance provided a
subsidy that, because of branching restrictions,
led to the entry o f many small un it banks.
The demise o f deposit insurance removed
this subsidy and, at least in four states, con ­
tributed to a shift toward m ore banks with
federal charters. Despite this, the negative
im pact o f insurance on average bank size
apparently remained in 1929.

Alston, Lee J ., Grove, W ayne A., and David C. Wheelock. "W hy Do
Banks Fa il? Evidence from the 1 9 2 0 Explorations in Economic
s ,"
History (October 1 9 9 4 ) , pp. 4 0 9 -3 1 .
Bankers Encyclopedia Company.Polk's Bonkers Encyclopedia
(March 1 9 2 9 ).
Bernanke, Ben S . "Nonmonetary Effects of the Financial Crises in the
Propagation of the Great Depression,"
American Economic Deview
(June 1 9 8 3 ) , pp. 2 5 7 -7 6 .
_ _ _ _ _ _ _ _ ,_ and Jam es, Harold. "The Gold Standard, Deflation, and
Financial Crises in the Great Depression: An International
Comparison," in Hubbard, R. Glenn, erf..
Financial Markets and
Financial Crises. University of Chicago Press, 1 9 9 1 , pp. 3 3 -6 8 .
Board of Governors of the Federal Reserve System .
Banking and
Monetary Statistics, 1 91 4-41 ( 1 9 4 3 ) .
_________. All Bank Statistics, 1 8 9 6 -1 9 5 5 (1 9 5 9 ) .

Bremer, C. D. American Bank Failures. Columbia University Press, 1 9 3 5 .
Calomiris, Charles W. "Financial Factors in the Great Depression,"
Journal of Economic Perspectives (spring 1 9 9 3 ) , pp. 61-8 6 .

In response to the bank failures o f the
Great D epression, Congress enacted federal
deposit insurance, im posed new restrictions
on the activities o f com m ercial banks, and
m aintained a strict prohibition o f interstate
branching. A lthough these policies appeared
to work well for many years, their weaknesses
were exposed in the 1980s, prompting reforms.
Looking back, econom ic historians have
demonstrated the destabilizing effects of deposit
insurance and branch banking restrictions in
the 1920s. This article illum inates how these
policies affected banking m arket structure
and, ultimately, state-level bank failure rates
during the Depression. Even though state
deposit insurance had ended by 1 9 2 9 , its
effects lingered into the 1930s, causing both
higher numbers o f banks per capita and higher
ratios of national banks to total banks in states
that earlier had insurance system s. At the
sam e time, branching restrictions, where
enforced, contributed to the sm all average
size o f un it banks and to their higher rate of
failure during the Depression. T hus, as others
have show n for the 1980s, the geographic


_ _ _ _ _ _ _ _"Do 'Vulnerable' Economies Need Deposit Insurance?
Lessons from U .S . Agriculture in the 1 9 2 0 s ," in Philip L. Brock, editor,
If Texas Were Chile: A Primer on Banking Reform. Institute for
Contemporary Studies, 1 9 9 2 , pp. 2 3 7 -3 1 4 .
_ _ _ _ _ _ _ _ . _ "Deposit Insurance: Lessons from the Record," Federal
Reserve Bank of ChicagoEconomic Perspectives (M ay/Ju n e 1 9 8 9 ) ,
pp. 10-30 .
_ _ _ _ _ _ _ _ ,_ and Joseph R. M ason. “Contagion and Bonk Failures
During the Great Depression: The June 1 9 3 2 Chicago Banking Panic,"
working paper, 1 9 9 4 .
Department of Agriculture. "The Form Real Estate Situation, 1 9 3 0 - 3 1 ,”
Bureau of Agricultural EconomicsCircular No. 209 ( 1 9 3 1 ) .
_ _ _ _ _ _ _ _ . _ "The Form Real Estate Situation, 1 9 3 2 -3 3 ," Bureau of
Agricultural EconomicsCircular No. 309 ( 1 9 3 3 ) .
Department of Commerce, Bureau of the Census.
Fifteenth Census of
the United States, Population, vol. I ( 1 9 3 1 ) .
_________. Fifteenth Census of the United States, Agriculture,

volume IV ( 1 9 3 2 ) .
_________. Statistical Abstract of the United States ( 1 9 3 0 , 1 9 3 1 ,

1 9 3 2 ,1 9 3 3 ).
_________. Personal Income By States Since 1929, A Supplement to
the Survey of Current Business ( 1 9 5 6 ) .



Tippetts, Charles S . State Banks and the Federal Reserve System.
D. Van Nostrand Co., 1 9 2 9 .

Federal Deposit Insurance Corporation.Annual Report ( 1 9 5 6 ) .
Federal Reserve Board. Bulletin (December 1 9 3 0 ).

Wheelock, David C. "Government Policy and Banking M arket Structure
in the 1 9 2 0 s , “
Journal of Economic History (Decem ber 1 9 9 3 ) ,
pp. 8 5 7 -7 9 .

_ _ _ _ _ _ _ _. _ "Bank Suspensions in the United States, 1 8 9 2 -1 9 3 1 ,"
(not dated).
Friedm an, Milton, and Anna 1. Schw artz. Monetary History of the
United States, 1867-1960. Princeton University Press, 1 9 6 3 .

_ _ _ _ _ _ _ _ . _ "Deposit Insurance and Bank Failures: New Evidence from
the 1 9 2 0 s ,"Economic Inquiry (July 1 9 9 2 ) , pp. 5 3 0 -4 3 .

Grossman, Richard S . "The Shoe that Didn't Drop: Explaining Banking
Stability During the Great Depression,"
Journal of Economic History
(Septem ber 1 9 9 4 ) , pp. 6 5 4 -8 2 .

_ _ _ _ _ _ _ _ . _ "Monetary Policy in the Great Depression: W hat the Fed
Did, and W hy," thisReview (M arch/April 1 9 9 2 ) , pp. 3 -28.
White, Eugene N. "Before the Glass-Steagall Act: An Analysis of the
Investment Banking Activities of National Banks,"
Explorations in
Economic History (January 1 9 8 6 ) , pp. 3 3 -5 5 .

Keeley, Michael C. "Deposit Insurance, Risk, and M arket Power in
Banking," American Economic Review (December 1 9 9 0 ) ,
pp. 1 1 8 3 -2 0 0 .

_ _ _ _ _ _ _ _. _ "A Reinterpretation of the Banking Crisis of 1 9 3 0 ,"
Journal of Economic History (March 1 9 8 4 ) , pp. 1 1 9 -3 8 .

Kryznowski, Lawrence, and Gordon S . Roberts. ''Canadian Banking
Solvency, 1 9 2 2 -1 9 4 0 ,"
Journal of Money, Credit and Bonking
(August 1 9 9 3 ) , pp. 3 6 1 -7 6 .

Wicker, Elmus. "A Reconsideration of the Causes of the Banking
Panic of 1 9 3 0 ,"Journal of Economic History (Septem ber 1 9 8 0 ) ,
pp. 5 7 1 -8 4 .

Stauffer, Robert F. "The Bank Failures of 1 9 3 0 -3Journal of Money,
1 ,"
Credit and Banking (February 1 9 8 1 ) , pp. 1 0 9 -1 3 .
Temin, Peter. Did Monetary Forces Cause the Great Depression? W. W.
Norton, 1 9 7 6 .

A p p e n d ix

Bank failures and deposits in failed banks: Board of Governors
( 1 9 4 3 , pp. 2 8 4 -8 5 ).

Farm population: Department of Commerce ( 1 9 3 2 , p. 4 0 ) .
Minimum capital requirement for state banks: Bankers Encyclopedia
Company (March 1 9 2 9 ) .

Branch bank offices: Federal Reserve Board (December 1 9 3 0 ,
pp. 8 1 1 -1 2 ).

Number of banks and bank deposits: Board of Governors ( 1 9 5 9 ) .

Business failure rate: Department of Commerce,
Statistical Abstract
( 1 9 3 0 , 1 9 3 1 , 1 9 3 2 , 1 9 3 3 ).

Per capita personal income: Department of Commerce ( 1 9 5 6 , p. 1 4 2 ) .
Population density: Department of Commerce ( 1 9 3 1 , p. 1 3 ) .

Farm failure rate: Department of Agriculture ( 1 9 3 1 , 1 9 3 3 ) .




Michael Ulan is an economic analyst at the U.S. Department of State. William G. Dewald is director of research at the Federal Reserve Bank of
St. Louis. Jam es B. Bullard is a senior economist at the Federal Reserve Bank of St. Louis. Jerram Betts provided research assistance. Any
view s expressed in this article are those of the authors and not necessarily those of the U.S. Department of State, the Federal Reserve Bank of
St. Louis or the Federal Reserve System.


U .S . Official
Forecasts of
G -7 Economies,
1 9 7 6 -9 0

(W E O ). The A dm inistration forecasts and
their accuracy are reported along w ith a
number o f alternative forecasts. The primary
com parison is to projections made for the
G -7 by the O rganization for E con o m ic Coop­
eration and Developm ent (O E C D ) and by
D ata Resources Incorporated (D R I). F o r the
U nited States only, we also com pare the
Adm inistration forecasts to those made by
the Blue Chip consensus and the U .S. Federal
Reserve “G reenbook.”1 F o r each country
and for the G -7 nations taken as a w hole, the
outlooks are evaluated on the basis o f the
differences between predictions and outcomes.
T he predictions and outcom es are expressed
in terms of year-over-year percentage changes.
T he statistics cited are the sum o f squared
errors, the m ean squared errors, the root mean
squared errors (RM SE) and the bias (sum of
prediction m inus outcom e). W e th in k these
measures provide simple but effective summary
statistics useful in evaluating forecast accuracy.

M ichael U lan,
W illiam G . Dewald
and Jam es B. Bullard
orecasts are regularly used in making fiscal
and m onetary policy decisions. F or many
decisionmakers, the likely short-term effect
o f a proposed action is a m ajor concern in
deciding whether to im plem ent a particular
policy. Such decisions are typically made in
the context o f considerable uncertainty, not
only about what the likely effects o f a partic­
ular action m ight be, but also about the
m om entum and direction o f aggregate eco­
nom ic variables in them selves. T hus, an
im portant concern from a policy point o f
view is the exten t to w hich forecasts are reli­
able representations o f econom ic outcom es
at relatively short horizons, such as a year.
The purpose o f this article is to report
facts concerning the accuracy o f the U.S.
official forecasts o f real output growth and
inflation from 1976 to 1990 for the Group of
Seven (G -7 ) econom ies: Canada, France,
Italy, Jap an , the U nited Kingdom , the U nited
States and W est Germany. Though widely
distributed w ithin the governm ent, the
A dm inistration forecasts had been classified
and n ot available to the public. W e obtained
the forecasts for years through 1 9 9 0 under
a Freedom o f Inform ation Act request with
the helpful cooperation o f the Treasury
Departm ent.
T h e accuracy o f these forecasts is m ea­
sured against the standard of actual real output
growth and inflation as subsequently published
in the Treasury’s W orld Econ om ic O utlook


Administration Forecasts of Real
Output Growth
The errors in the Administration forecasts
o f real gross national product (G N P) (gross
domestic product, GDP, in some cases) growth
in the G -7 nations are shown in Figure 1. The
sum m ary statistics relating to the errors in
these forecasts appear in Table 1. The sum
o f squared errors o f the A dm inistration’s real
output growth forecasts is largest for Canada,
the U nited States and W est Germany. Ju st
under h alf o f the forecast errors were o f a dif­
ferent sign from the errors o f the preceding
year. The num ber o f sign reversals o f forecast
error, n o t counting a zero error as a sign
change, ranged from four for Ja p a n to eight
for the U nited States.
Like the other forecasters, the Adm inis­
tration sim ply missed the deep recessions in
1 9 8 2 in the United States and Canada. The
A dm inistration forecasted 3 .4 percent real


1 The Blue Chip Economic Indicators
newsletter is published monthly
and surveys major forecasters of
the U.S. economy. The Federal
Reserve "Greenbook" is a document
distributed to top-level staff and
Federal Open Market Committee
(FOMC) members shortly before
each FOMC meeting. The FOMC is
the primary policymaking arm of the
Federal Reserve. Greenbook infor­
mation is classified for five years
following each FOMC meeting.

Fig u re 1

A d m in istratio n Fo recast E rro rs, R e a l Output G row th
Percentage Points



4 -



M jy u lg








H U.K.


lsr v r J Japan

KSSSS1 W.Gernt




I France



I Italy

output growth for the United States in 1982
and 3 .2 percent for Canada. The outcom e
was a 1.9 percent decrease in output in the
U nited States and a 4 .4 percent decrease in
Canada, one o f the deepest recessions in
either country sin ce the end o f W orld W ar II.
In absolute term s, the 1 9 8 2 forecast errors
for U.S. and Canadian econom ic growth
were tw o-to-three tim es as large as any for a
non-N orth Am erican G -7 econom y over the
15 years covered here, as all o f the largest
absolute forecast errors were betw een 2 per­
centage points and 3 percentage points for
the rem aining countries. W h ile output fell
in som e other G -7 econom ies in 1 9 8 2 , other
nations did not experience a comparable
reversal o f fortunes.
There were, however, som e large declines
in real output growth in other countries in
other years. Italy, for instance, experienced a
4 .2 percentage p oint fall in its output growth
rate (from 4 .0 percent to -0 .2 percent) between

Table 1

Descriptive Statistics, Errors in
A dm inistration Forecasts of R eal
GNP/GDP G row th, 1 9 7 6 - 9 0

C o u n try
United States
West G erm any
United Kingdom
G-7 total

Sum o f
M ean
S q u ared S q u a red
E rro rs
E rro r
4 7 .6 3

3 .1
1.78 8

B ia s


6 .1 5
2 .4 8
7 .7
9 .2
1 .4 0
- 1 .2
1 5 .7 9
1.05- 0 . 7
1 .03
3 9 .2 5
2 .6 2
1 .6 2
- 2 .7
- 2 .3
2 1 .6 7
1 .2
2 7 3 .7 0

growth in Italy and the U nited Kingdom was
larger in non -tu rnin g-point years than during
these turning-point episodes. In the case of
Italy, the largest error was for 1 9 7 6 , w hen the
nation’s econom y experienced a substantial
upturn. In that year, the change in direction

1 9 8 0 and 1 9 8 1 , and the U nited Kingdom
witnessed a 4 .3 percentage point decline
(from 1.5 percent to -2 .8 percent) betw een
1 9 7 9 and 1 9 8 0 . A dm inistration forecasts in
these instances, however, were n ot so wide
off the m ark as for the U.S. and Canadian
forecasts for 1 9 8 2 . Moreover, the error in
the A dm inistration’s forecasts o f real output


of the Italian economy (a total of 9.3 percentage
points— from a decline o f 3 .7 percent in 1975
to output growth o f 5 .6 percent in 1 9 7 6 ) was
actually greater than the percentage point
changes in the direction o f real output growth
in the U .S. and Canadian econom ies in 1 9 8 2 ,




Figu re 2

A d m inistration Fo recast Erro rs, Inflation
Percentage Points


F 7 s \ Japan



I France



I U.K.

I Italy

than the consum er price index (C P I).
H ence, for this year the forecast error is cal­
Descriptive Statistics, Errors in
culated w ith respect to the change in that
Adm inistration Forecasts of
m easure rather than the CPI. The sum of
Inflation, 1 9 7 6 - 9 0
squared errors is largest for Italy and the
U nited Kingdom. T he large error in the fore­
Sum of
cast o f United Kingdom inflation in 1 9 7 8 is
Squared Squared
attributable prim arily to a decline in inflation
in that year; inflation fell from 1 5 .8 percent
United States
4 0 .1 9 .6 8
- 2 .1
in 1977 to 8.3 percent in 1978. It subsequendy
3 2 .2 7
3 .9
- 2 .1
rebounded to 1 3 .4 percent in 1 9 7 9 . During
1.23 1 .5 2 1 0.9
1 9 7 8 , there were price controls in force on
5 5 .4 6
3 .7 0
0 .0
som e com ponents o f the CPI m arket basket
West Germ any
18.7 0
and, at governm ent urging, unions moderat­
1 74 .65
11.6-4 2 0 . 3
3.41 ed their wage demands. In 1979, with the
United Kingdom
103 .25
6 .8 8
2 .6 2
- 9 .5
election o f a Conservative governm ent, the
G-7 total
4 4 9 .0 3
4 .2 8
2 .0 7 unions 3returned to no-holds-barred wage
-2 0 .
bargaining, and the governm ent n o t only
removed price controls b u t also increased the
but the error in the Administration forecast
rate o f value-added tax applicable to several
o f Italian real GDP growth in 1976 was only
item s in the CPI m arket basket, boosting
-2.7 percentage points.
inflation during that year.
As Table 2 reveals, the A dm inistration
Administration Inflation Forecasts
tended to underpredict inflation in Italy and
The Adm inistration forecast errors for
the U nited Kingdom , countries w ith high
inflation from 1976 to 1990 in the G -7 nations
average inflation rates, and to overpredict
are show n in Figure 2, while Table 2 presents
inflation in W est Germ any and Japan , cou n­
the associated sum mary statistics. The
tries w ith comparatively low inflation. Errors
Adm inistration forecasts o f U.S. inflation in
in one direction were followed by errors in
1 9 8 0 pertained to the GNP deflator rather
the other direction about a third o f the time—



Table 3

less than was the case for real output growth.
The number of reversals of sign of the forecast
error ranged from three for Jap an and W est
Germany to six for both France and the United
Kingdom , again, not including a zero error
as a change in sign.

Descriptive Statistics, Errors in
Forecasts of R eal Output Grow th
by Country, 1 9 7 6 * 9 0



United States

Administration Forecasts and OECD
Projections, Real Output Growth,

2 Since the OECD uses the personol
consumption deflator rather than
the CPI as its measure of inflation,
its inflation projections ore not con­
sidered here.


West G erm any

T he O ECD ’s projections o f econom ic
growth for G-7 nations between 1976 and 1990
are readily available for com parison w ith the
A dm inistration predictions.2 T he O ECD staff
issues its projections in the Economic Outlook
twice each year— around mid-year and in
December. W e compared the December OECD
projections (prepared in mid-November) with
A dm inistration forecasts, although the latter
were generally made earlier. Sum m ary statis­
tics covering the A dm inistration’s predictions
and O ECD projections over 1 9 7 6 -9 0 appear
in Table 3.
T he O ECD m akes several assum ptions
about m em bers’ econom ies in projecting each
nation’s econom ic growth. The organization
assumes that the exchange rate of the nation’s
currency during a year rem ains at the level of
November in the previous year (the m onth the
projections were prepared), that fiscal policy
will rem ain unchanged and that the price of
oil relative to that o f O ECD exports o f m anu­
factures will rem ain constant. T he reasoning
behind these assum ptions is that the O ECD
is “advising” its m em ber governm ents where
they are headed econom ically if they continue
to pursue current policies, n ot taking into
account prospective changes in policies. Hence,
the OECD considers its product a projection
rather than a forecast.
Table 3 shows that, for each G -7 nation
except Italy, the sum o f squared errors o f the
O ECD real growth p rojection is sm aller than
that for the A dm inistration’s forecast errors.
For both the OECD and the Administration, the
smallest sum of squared errors was achieved in
the case of France, while the country evidently
posing the m ost difficulty over this period was

United Kingdom
G-7 total

Sum of
Squared Squared

4 7 .6 3


3 .1 8
2 2 .5 3

1.231 .5 0

5 7 .4 0

6 .1 5
3 .8 3

2 9 .2 4
1 7.7 6


3 9 .2 5

2 7 3 .7 0
1 99 .72


- 5 .0



2 .6 9

1 .4 5

0 .7

- 2 .7

- 2 .3
- 1 0 .9


Canada. F o r the A dm inistration, the second
w orst case was the U nited States.
To what might one attribute the generally
greater accuracy o f the O ECD projections
compared w ith the Adm inistration’s forecasts?
One factor m ight be that O ECD projections
of real output growth in the G -7 nations were
made closer to the beginning o f the forecast
year. T he O ECD m ight also be in a better
position to closely follow the econom ic per­
formance of many nations by evaluating world­
wide influences than is the A dm inistration,
w hose forecasts are largely dependent on
inputs from individual countries. O n the
other hand, the O ECD procedure sim ply
assumes unchanged fiscal policies, exchange
rates and real oil prices. These might be fac­
tors that would lead to less accuracy in their
p rojections to the extent that such factors
have a predictable effect on real output growth.


- 6 .2

1.51 .2 0

1 .90

7 .7
2 .0

- 1 . 2 .4 0

2 .6 2

2 1 .6 7
18.9 9


2 .4 8

1.01 1.03 - 3 . 7

2 7 .2 9
4 0 .3 7

- 4 .0



15.7 9 1.05
15.3 9


- 3 5 .2

- 7 .4


W ith one observation deleted, the sum of
squared errors o f both the A dm inistration
and O ECD forecasts tend to be m uch sm aller
and m uch the same.
O ur data set contained com plete DRI
forecasts o f econom ic growth and inflation for
all the G -7 countries for the period o f 1983
to 1 9 9 0 . T h e sum m ary statistics pertaining
to these DRI forecasts are com pared w ith
those o f the A dm inistration and the O ECD
over the same period in Tables 5 and 6 . Save
for the real growth forecasts for Italy, Japan
and the U nited States, the DRI forecasts over
this evaluation period were m ore accurate
than either the Adm inistration forecasts or the
O ECD projections. DRI was also m ore accu ­
rate than the A dm inistration in forecasting
inflation for every country except Jap an
during this period.

Descriptive Statistics, Errors in
Forecasts of R eal Output Growth
by Country, 1 9 7 6 - 9 0
(largest erro r omitted)


Sum of
Squared Squared

United States


19.5 4
15.7 7



- 6 .6

3 4 .4 5
2 8 .2 4

2 .4 6
2 .0 8


- 3 .5

2 1 .9 5
1 2 .9 2

0 .9 2

1 .25
0 .9 6

1 .5
- 4 .0

West G erm any

11.7 9
10.5 5

0 .8 4
0 .7 5

0 .9 2
0 .8 8

- 1 .5

3 3 .0 0
2 1 .7 7

2 .3 6


- 0 .2


2 0 .6 2
2 3 .5 6



0 .4
- 6 .8

17.2 6
1 3 .7 0

0 .9 8

0 .9 9

- 0 .6
-5 .1

1 2 6 .5 0



8 .3

United Kingdom
G-7 total

1 9 7 7 -9 0
Real Output Growth
T he Blue Chip E conom ic Indicators
consensus forecast o f year-over-year real
econom ic growth in the U nited States has
been published monthly since 1976 (first fore­
casting 1 9 7 7 ). A consensus forecast o f the
year-over-year change in the CPI has been
published since 1 9 7 9 (forecasting 1 9 8 0 ).
B oth the num ber and the identities o f partic­
ipating private-sector forecasters have changed
over time. Though DRI, O ECD and Federal
Reserve G reenbook forecasts for the United
States are available for the full period for which
we have A dm inistration forecasts, we only
com pared the five forecasting records for the
period for w hich the Blue Chip consensus has
been available. As show n in Table 7, the Blue
Chip, DRI, Federal Reserve and O ECD projec­
tions o f U.S. econom ic growth were each more
accurate than the Administration forecasts
over this period, with the OECD achieving the

* Largest error is either -2.35 in 1976, implying a bios of -2.65, or
2.35 in 1982, implying a bias of -7.35.
** Bias is -30 if 1976 error is omitted, or -34.7 if 1982 error is omitted.

Both the A dm inistration forecasts and
the O ECD projections o f real growth display
bias, according to our measure— bu t in
opposite directions. T he A dm inistration’s
real output growth forecasts in total are
biased upward, w ith the m ain contributors
to the total being the errors associated w ith
the United States and Canada. In contrast,
except for Canadian econ om ic growth, the
O ECD projections are biased downward—
substantially for the cases o f Italy and the
U nited Kingdom.
Since one large error can m ar an other­
wise good perform ance, we also exam ined
the data w ith the largest error om itted; the
descriptive statistics are presented in Table 4.

1 9 95

greatest overall accuracy. T he Adm inistration
forecasts for Canadian real output growth



1 9

Table 5

Table 6

Descriptive Statistics, Errors in
Forecasts of R eal Output Growth
by Country, 1 9 8 3 - 9 0

D escriptive Statistics, Errors in
Forecasts of Inflation by
Country, 1 9 8 3 - 9 0


United States

Sum of
Squared Squared

1 5.4 8

United Kingdom
G-7 total

2 .2 7
1 2 .9 6
1 1 .0 7

10.4 7

8 .8 2 1 .1 0
7 .2 5

10.4 4
0 6.31
.7 9
11.2 0


Japan 5 . 6
- 5 . 0 France
DRI4 . 5

0 .7 3

- 3 .9
0 .8 6
- 3 .3

1 .1 2
- 4 .5
1 .86 - 5 . 5 1.36
1 .0 5


0 .7
0 .9 5

West G erm any
DRI- 3 . 0
- 4 .DRI
United Kingdom
G-7 total
0 .2


1 .6 3
1 .1 5
9 .4 8

1 .2 8

0 .9 0
1 2 .7 8

0 .7 6

2 .2 7
18 .7 7
.1 0


5 .3

1 .1 7
8 .8 9


3 .7 2
.4 6
3 .3 9
2 .2


7 3 .9 9

2 .0 4
1.15 1 .3 2


- 4 .4
- 2 .0
2 2.9

m onths into a recession that would not bottom
- 1 6 .4
1.423 0 . 7
1.19 out for 14 months. The Greenbook and OECD
1 .6 0
1.27 forecasts, both o f which, it is im portant to add,
- 2 4 .3


were made later in the year, were considerably
better, predicting -0 .6 and -0 .5 , respectively,
versus an actual outcom e o f -1.9.
Table 8 show s the effects o f om itting
the largest error in com puting the accuracy
o f these forecasts. T h e errors in the A dm inis­
tration, Blue Chip, DRI, Federal Reserve and
O ECD projections o f U .S. real output growth
for 1 9 8 2 were 5 .3 , 4 .5 , 4 .3 , 1.3 and 1.4 per­
centage points, respectively. T h e largest
Federal Reserve error was -2 .5 , recorded in
1 9 8 1 , while the largest O ECD error was 2.6,
recorded in 1990. The largest errors in the
A dm inistration, O ECD and DRI forecasts o f
Canadian real output growth were 7 .6 ,5 .4 and

vation for 1 9 8 2 greatly improves the accuracy
o f the A dm inistration forecasts. In fact, the
same observation accounted for the greatest
error in the forecasts o f the Adm inistration,
the Blue Chip consensus and DRI. In
Septem ber 1 9 8 1 , m any forecasters predicted


5 .3
6 .8

3 .6

in 1 9 8 2 even though it was already several

1 01 .14 1.81
8 9 .5 7

5 .9

1 .2 0


- 0 .2

5 .51.51
1 .0 5

9 .3 3

2 .7

2 .50 .9 5

1 8 .1 5

2 7 .1 8

3 .6 1 .0 9
0 .5 8


2 9 .7 6

3 .0
2 .8

- 4 .8
0 .8 9- 5 . 1
1.40 - 5 . 1
1.83 positive econom ic growth for the U.S. economy

were less accurate than those o f either DRI
or the OECD.
F o r U.S. econom ic growth, there was a
“rosy scen ario” positive bias o f the Adm inis­
tration forecasts w hich was approached in
m agnitude only by the negative bias o f the
Federal Reserve forecasts. As noted in the
previous section, the exclusion o f the obser­

13.0 0
9 .1 6

- 3 .1
2 .0 7
- 4 .5

1 .62- 7 . 8

2 .8 2

United States
DRI 5 . 4
Canada . 7

3 .4

8 .9 9 1 .12
5 .8 6
8 .9 2
1 .12
2 2 .5 4
10.0 7
1 .26
14.8 6


2 .1 4
2 .1 9

16.5 6
2 .3 4

1 8 .7 4
1 8 .1 8

1 .39

2 1.44
.0 9

16.6 9

West Germ any

1 .94
1 7.0 8
1 7 .5 3


Sum of
Squared Squared



Table 7

Table 8

Descriptive Statistics, Errors in
Forecasts of U .S. and Canadian
R eal Output G row th, 1 9 7 7 - 9 0

Descriptive Statistics, Errors in
Forecasts of U .S. and C anadian
R eal Output G row th, 1 9 7 7 - 9 0
(larg est erro r omitted)


Sum of
Squared Squared

United States
Blue Chip


3 .4 0 1.84
2 .3 0
2 .6 2
2 9 .2 9
2 .0 9
2 2 .4 7

4 7 .5 9
3 2 .1 3
3 6 .7 2

9 1 .4 0
6 8 .6 4
5 6 .9 8


4 .9 0
4 .0 7

8 .6
2 .6
2 .0 2

6.5 percentage points, respectively. W ith the
largest error omitted, the Blue Chip consensus
ranks first in accuracy for the United States.


1 9.5 0
0 .9 4 8
- 2 .0
0 .8
1.18- 4 . 9
1 8.2 3
2 3 .0 4
1 .7 7
- 2 .8
3 3 .6 4 2 .5 9
2 6 .3 9
1 .43
2 .0 3
2 7 .8 2
2 .1 4

- 2 .3

forecast the output decline in 1982. Deleting
that observation substantially enhances mea­
sured forecast accuracy, reducing the RMSE
from 1.78 to 1.18 over the 1976 to 1990 period.
U.S. official forecasts were better w ith respect
to inflation, as the A dm inistration was one o f
the best among those compared in forecasting
U.S. CPI inflation betw een 1 9 8 0 and 1990.
The Administration’s forecasts o f economic
growth for alm ost all G -7 countries were less
accurate than the O ECD projections for the
period 1 9 7 6 to 1990. T he biases in the
Adm inistration’s forecasts tend to be positive;
those in the forecasts o f U.S. and Canadian
real output growth are particularly large.
The biases in the O ECD projections tend to
be negative; those associated with projections
o f Italian and U.K. real output growth are
large. F or the G -7 as a whole, the projections
o f the O ECD are m uch m ore accurate than
those o f the Adm inistration. Over the 1983
to 1 9 9 0 period, DRI was m ore accurate than
either the A dm inistration or the O ECD for
four o f the G -7 countries.
The differences betw een the forecast
errors o f the A dm inistration and the forecast
(o r p rojection) errors o f the other forecasters
may arise from differences in the tim es at
w hich the forecasts or projections were pre­
pared, a situation that may have influenced
the quality o f the historical baseline available
to forecasters and the values o f exogenous

As show n in Table 9, in contrast to the
situation w ith respect to real output growth,
the A dm inistration was a m arginally more
accurate forecaster o f U.S. inflation over the
period 1 9 8 0 -9 0 than the G reenbook and also
more accurate than the Blue Chip survey. DRI
was the m ost accurate overall for the United
States, and DRI also predicted Canadian infla­
tion more accurately than the Administration.
Sum mary statistics w ith the largest forecast
error om itted are presented in Table 10. In
this case, the Adm inistration forecasts hold
up very well against those o f the other fore­
casters for the United States, as do D RI’s
inflation forecasts for Canada.

Com paring Adm inistration forecasts to
Blue Chip consensus, DRI, Federal Reserve
G reenbook and O ECD predictions o f real
output growth in the U.S. economy, we find
that the A dm inistration tended to see the
future more optim istically and less accurately
than the other forecasters. M uch, though
not all, o f that rosy perspective was connected
w ith the failure o f the A dm inistration to


United States
Blue Chip

- 5 . 3 1.45
- 3 .7

62.5 36

Sum of
Squared Squared



- 2 .8


- 6 .3

m a r c h




Table 9

Table 1 0

Descriptive Statistics, Errors in
Forecasts of U .S. and Canadian
Inflation, 1 9 8 0 - 9 0

Descriptive Statistics, Errors in
Forecasts of U .S. and C anad ian
Inflation, 1 9 8 0 - 9 0
(larg est erro r omitted)


United States
Blue Chip

Sum of
Squared Squared


14.3 7 1.31
17.3 8
13.1 3
15.9 4
15.0 2
11.0 6

1 .1 4
3 .3
1 .0 9 1 .1 9 5 .3
1 .20


1 .21 .17
1 .0 0

United States
6 .4
Blue Chip
DRI 5 .2
- 1 . 4 Administration

variables assumed in predicting the future
paths o f the econom ies. N onetheless, so far
as we can ascertain, every forecast we have
evaluated was a genuine prognostication of

Sum of
Squared Squared

11.6 2

9 .0 80.91
1 .16
9 .1 3
1 1 .5 3




0 .9 5

1 .0
4 .0

A p p en d ix


1 The Treasury thought the Council's
forecast of U.S. economic growth in
1983 wos too high and substituted
the Blue Chip consensus forecost.
(As it turned out, the Treasury—
that is, Blue Chip— forecost wos
also too high, but not so high as
the Council's.)
2 The dates of the Administration
forecosts for the next year range
from September through December
of the previous yeor.

judgm ents about the national economies. The
Blue Chip consensus forecasts are the mean
values o f the forecasts o f the firm s covered in
the Blue Chip surveys. T he DRI forecasts are
based on the outputs o f the DRI m odel and
the ju d gm ents o f that firm ’s staff. T h e O ECD
p rojections are prepared by m em bers o f that
organization’s staff. The Federal Reserve fore­
casts are prepared by the staff o f the Federal
Reserve Board.
T he A dm inistration, Blue Chip, DRI
and Federal Reserve forecasts, and the O ECD
outlooks have appeared several times each
year and are frequently revised. T h e WEO
forecasts evaluated in this article are the last
predictions o f b oth econom ic growth and
inflation for the n ext year m ade during the
previous year.2 T he Blue Chip and DRI fore­
casts for the U .S. and Canadian econom ies
selected for com parison to the Adm inistration
forecasts were those published during the
same m onths as the Adm inistration forecasts.
The DRI forecasts begin w ith those for 1976
and run through those for 1 9 9 0 . A com plete
set o f DRI forecasts for all o f the G -7 countries




econom ic growth and inflation m ade in the
closing m onths o f a year w ith respect to the
n ext year.

The data used in this article com e
prim arily from the World Economic Outlook
(W EO ) prepared by the U nited States Depart­
m ent o f the Treasury, Blue Chip Economic
Indicators, D RI’s various Reviews, the OECD
Economic Outlook and the Federal Reserve’s
G reenbook. T he A dm inistration forecasts o f
G -7 nations’ econom ic growth and inflation
have been made since 1 9 7 5 (for 1 9 7 6 ). The
forecasts evaluated in this article cover 1 9 7 6
to 1990, the last year for w hich forecasts have
been cleared by the Treasury for release to the
public. This is also the last year for w hich
the G reenbook forecasts are cleared for public
release. W ith one m ajor exception, the
Administration forecasts for the U.S. econom y
are those o f the Council o f Econom ic Advisers.1
Forecasts for the other G -7 econom ies are
produced by Treasury financial attaches at
U.S. embassies in the capitals o f these nations.
T he attaches review the host-governm ent
and host-cou ntry private-sector forecasts for
the econom ies o f the nations to w hich they
are posted and base their own forecasts on
such inform ation, together w ith their own



7 .0 6

0 .9 6

1 .1 5

- 0 .9
0 .8 4


is available for each year since 1 9 8 3 . The
O ECD projections are those published in
D ecem ber for the n ext year, beginning w ith
the outlook for 1976. T h e Federal Reserve
forecasts are those associated w ith the last
Greenbook issued in a given calendar year
(usually D ecem ber).
GNP and GDP data are frequently revised.
It was necessary to choose a fixed target to
w hich to com pare the forecasts. W e used the
Treasury D epartm ent’s historical data, w hich
it provided along w ith its forecasts in each
issue o f the WEO. Generally, historical data
on GNP or GNP changes for a particular year
continue to appear in the WEO for about

18 m onths follow ing the end o f that year.
T h e last historical citation o f the annual
change in national GNP or GDP appearing in
the WEO is the outcom e to w hich the forecasts
are com pared.3 A lthough CPI data tend n ot
to be revised after they are issued, a sim ilar
procedure has been followed in selecting the
inflation data w ith w hich to com pare the
forecasts. Because the Treasury presents no
historical data for growth or inflation in 1978,
we have com pared its forecasts for 1 9 7 8 w ith
outcom es taken from the 1981 International
Financial Statistics (IF S ) yearbook.1

3 In 1986, Canada changed the
emphasis in its National Income
and Product Accounts (NIPA) from
GNP to GDP and stopped explicitly
reporting historical real GNP data in
its official bulletin. National Income
and Expenditure Accounts (NIFA).
When the Canadian NIPA focus
shifted, the Administration began to
forecast GDP instead of GNP for
Canada and reported historical GDP
data in the WFO. Since the 1985
and 1986 growth forecasts for
Canada prepared by the
Administration pertained to GNP, we
obtained real GNP growth data for
1985 and 1986 with which to
compare the forecasts.
4 The Administration's 1980 inflation
forecast for the United States,
which appeared in the September
1979 WFO, pertained to the GNP
deflator rather than the CPI. The
deflator calculated on the basis of
data appearing in the 1981 IFS
yearbook was used to test the
accuracy of this forecast. Given the
Administration data, the September
1979 Blue Chip, DRI and Federal
Reserve forecasts of the increase in
the U.S. GNP deflator— rather
than CPI inflation— ore employed
in this comparison.



r e v i e w


Joseph A. Ritter is a senior economist at the Federal Reserve Bonk of St. Louis. Heidi L. Beyer provided research assistance.


An Outsider's
G uide to Real
Business Cycle
M o d e lin g

(1 ) D ecisions o f firms and consum ers should
be derived from fully specified intertem poral
optim ization problem s w ith rational expecta­
tions. (2 ) The general equilibrium o f the
m odel m ust be fully specified. (3 ) B oth the
qualitative and quantitative properties o f the
model should be studied. Lucas argued in
1 9 8 0 , before w ork began on RBC models,
that theoretical developm ents beginning with
H icks, Arrow and Debreu allowed m odern
econom ists to begin w ork w hich m et the
first two criteria. T he dram atic fall in the
price o f capital (com puters) has made it
possible to m eet the third criterion as well,
allow ing m acroeconom ists to explore their
models in m uch greater depth (although this
potential is n o t always realized). This article
is concerned m ostly w ith giving an outsider
a feel for how the third requirem ent is met.
It proceeds by describing the theory underlying
a standard RBC m odel, explaining what con­
stitutes an equilibrium , and then delving into
the m echanics o f solving a specific model
(H ansen’s landm ark indivisible labor m odel)
using a specific technique. I conclude w ith
two illustrations o f how the basic m ethodol­
ogy can be extended to study fiscal and
m onetary policy.

Joseph A . Ritter
One exhibits understanding o f business cycles
by constructing a model in the most literal
sense: a fu lly articulated artificial economy
which behaves through time so as to imitate
closely the time series behavior o f actual
Robert E. Lucas (1 9 7 7 )

uring the last decade, guided by Lucas’
principle, the real business cycle (RBC)
model has become a standard tool for a
large share o f macroeconomists. T he tool has
found such widespread applicability that pro­
ponents o f this approach to m acroeconom ic
modeling (and those w ith proper sensitivity
training) now prefer a m ore generic label:
com putable dynamic general equilibrium
model. O ther demographic groups often
regard the customs and rituals o f RBC propo­
nents w ith som e degree o f bafflem ent. The
goal o f this article is to dispel som e o f the
aura o f mystery that surrounds— from an out­
sider’s point of view— the specification, cali­
bration, solution and evaluation of RBC mod­
els.' It is thus concerned more with the “how’ ”
of RBC modeling than with the “why’s ” (or, for
that matter, the “why not’s ”).2 Broader intro­
ductions to real business cycle modeling can
be found in Blanchard and F isch er (1 9 8 9 ,


The typical RBC model is an ArrowDebreu type economy, specifically a onesector stochastic growth m odel. M any iden­
tical consum ers who live forever m axim ize
expected utility (derived from goods and
leisure) su b ject to an intertem poral budget
constraint. Competitive firms purchase factors
in com petitive m arkets. U ncertainty com es
from a stochastic shock to the econom y’s
production technology.
For simplicity, suppose that consum ers
own capital directly and rent it to firms. Firm s
buy capital and labor services from consumers
and use them to produce a single output which
can be used as either consum ption or invest­
m ent. O utput is the num eraire. T h e firms’
technology is described by an aggregate con-

chapter 7 ), M cCallum (1 9 8 9 ), Plosser (1 9 89 )
and Stadler (1 9 9 4 ). The pioneering papers are
Kydland and Prescott (1 9 8 2 ) and Long and
Plosser (1983).
Three criteria have guided the m odelbuilding process in the RBC literature.


1 Outsiders would include, omong
others, those who (like the outfior)
were educated where these models
were not in favor and those who
(like the author) finished school
before these models held a large
market share.
2 In this spirit, the programs used in
the article are available on the
FRED electronic bulletin board. For
more information, see the bock
cover of this issue.

stant-returns-to-scale production function
w hich includes an exogenous aggregate tech ­
nology shock, A,: AtF(K „L,), where K, and L,
are the aggregate levels o f capital and labor.3
The A, are usually taken to be serially corre­
lated, b u t the exact specification can be post­
poned. These are sim ple com petitive firms
w hich will purchase labor and capital services
at wage
and rental price Rt up to the point
w here their marginal products equal Wc and
R„ respectively:
(1 )

determ ined by the m odel; once equilibrium
quantities are know n, they can sim ply be
substituted into the E uler equation for each
asset to determ ine its price.
Since consum ers and firms are identical,
this artificial econom y is m athem atically
identical to a representative agent econom y
in w hich one price-taking consumer sells labor
and capital services to a single price-taking
firm. O n th e surface, finding an equilibrium
appears to be a very daunting task. Even
though we have reduced the num ber o f co n ­
sum ers and firms to one each, we still have
an infinite num ber of goods: consum ption
and leisure in various states o f the world at
dates from 0 to infinity. However, a great deal
is know n about the theory underlying this
type o f econom y (Stokey, Lucas and Prescott,
1 9 8 9 ), and this theory provides im portant
tools that allow sim ulations to be constructed.

Wt = AtFL(K t,L t), Rt = A c k(K„ L,).

Let I be a consum er’s tim e endowm ent,
1, the am ount o f labor she supplies, c, her
consum ption, k, her holdings o f capital and i,
her rate o f gross investm ent. (Upper case will
be reserved for aggregate variables.) T he con­
sum er takes prices W, and R, as given. Given
a starting value k0, she chooses paths, that is,
contingency plans, for i, and I, to m axim ize


E0j;f> u (c t, i —

F o r the representative consum er, the
state o f this econom y at the beginning o f t is
sum m arized by the individual’s capital stock
kh the aggregate capital stock K, and the state
o f technology A,. Thus, the m axim um lifetime
utility attainable by the consum er will be a
function V o f k„ K, and At. V (kt, K t, A,) is the
value function for the consum er’s utility m ax­
im ization problem .
T he core o f the problem is to find V. To
start, substitute the budget constraint ( 2 ) into
the consumer’s utility function, then substitute
marginal products for W, and Rt as described by
equation 1. T he latter substitution im plicitly
defines the consum er’s rational expectations
o f factor prices in terms o f present and future
values o f aggregate labor and capital. In other
words, the consum er does n ot care about Kt
and L, per se; they merely contain the same
inform ation as W, and R,.' W e now have

su b ject to a budget flow constraint
(2 )

c t + i , < R ,k t + W tlt

and a description o f how capital accum ulates
and depreciates:
(3 )
3 For o long time, the technology
shock 4 , wos the driving force in
most RBC models (hence, the
"real"). Both proponents and
opponents recognized this os the
Achilles heel of this line of research.
One response has been the devel­
opment of models in which technol­
ogy is not the only source of uncer­
tainty (the last section contains two
examples), though the criticism
goes deeper than simply claiming
that there are other kinds of shocks
(see Stadler, 1994, section IV A ).
4 The exact sequence of substitutions
here is designed to hammer the
model into the mold required later
for a specific numerical solution
method. For example, i, could easi­
ly be eliminated from the problem
using 3 , but that would be inconve­
nient loter.

k, = ( l - 5 ) f e 1
_1+ i [ ,


For present purposes, it is more useful to
fram e the solution in term s o f decision rules
which prescribe it and I, as functions o f current
state variables, kh K, and At:

i, = i(k t, K„ A,),

I, = l(k t, K„ At).

These decision rules depend only on the state
variables w hich fully describe the position of
the econom y at the beginning o f t and which,
therefore, contain all of the information needed
to decide optimal levels o f i, and 1,. A great
deal o f inform ation about how the econom y
works— about the structure o f the model,
in other words— will be embedded in these
fu nctions w hen we find them.
In addition to capital, there may be many
financial assets with a net supply o f zero, but,
since consumers are identical and the economy
is closed, these assets would be redundant.
Nevertheless, the prices o f these assets are

E 0 X j 8 'u ( a , F k(K„ L ,) k t
[t =0
+ A ,F l (K„ Lt ) I - i „ i - l ()

In period 0, the consum er is choosing i0 and
1 Rewrite utility as


m a r c h

(4 )

/ A pril

u(A0Fk (K 0,L 0)k0
+A0Fl (K 0.



V(fe„K„A ,)

= u(i(fe„ K„ A , ) , l ( k r K„ A , ) , k t,


K „ L ( K „ A ,) ,A ,)

+/JE1| X ^ u ( A , F K(K„L,)fe,

+0 E{v(fe,+,K,+,A,+
1 1 I)|A,}

+ A lFK( K „ L , ) ( - i 1J - l , )

(7 )

Exam ination o f the second term in 4 reveals
an im portant feature o f the optim ization
problem ; apart from the values o f state
variables, the consum er will solve exactly the
same problem in period 1 as in period 0. For
an optim al plan, this recursion is sum m a­
rized in the Bellm an equation for the con ­
sum er’s problem at t:
(5 )

and I(K „ A t ) = i(K „ K r, A ,).
Condition 6 says that, given expectations L (-),
decision rules i(-) and 1(0 are optim al for
consumers. The equations in 7 say that
expectations of aggregate labor supply and
investment are consistent with individual
decisions. Only a sm all num ber o f exam ples
can be solved analytically (for exam ple, Long
and Plosser, 1 9 8 3 ), so we m ust now turn to
the computer.

VCkt,K „ A t)
= m ax ju ( i t,I,,k t,K (, Lt,A t)
i„l, I


+ 0 E v { ( f e I+1, K 1
+1,A 1
+1)|A(} }

To com pare the tim e-series behavior of
the m odel’s equilibrium w ith the tim e-series
behavior o f actual econom ies— to evaluate
its quantitative im plications— requires that we
sim ulate the model using specific functional
form s and param eter values. T h e process o f
choosing param eter values, calibration, is
deferred to the n ext section. T he n ext few
sections illustrate solution procedures by
fully specifying, calibrating and solving Gary
H ansen’s (1 9 8 5 ) indivisible labor m odel, an
early landmark in the RBC literature. Hansen’s
relatively sim ple model provides a clear illus­
tration, but should not be taken as the state of
the art. The solution follows one of the popular
linear-quadratic approxim ation methods/

The m axim ization on the right is subject to
the constraint (3 ) that connects investm ent
and capital. Em bedded in 5 is Richard
Bellm an’s deep insight that if you know the
value o f your problem n ext period for the
various values o f state variables, it is a rela­
tively sim ple m atter— a static m axim iza­
tion— to figure out the optim al action now.
There are four kinds o f variables in 5:
individual decision variables (i„ I,); an indi­
vidual state variable ( k ,); economy-wide state
variables (K„ A,); and an economy-wide variable
determ ined in t (L,).5 The state variables are
determ ined at the start o f t or inherited from
t — T h e contem poraneously determ ined
econom y-w ide variable L, appears because
(for m athem atical reasons evident below )
we have substituted out W, and Rt. These
m arket-clearing prices would otherw ise sum ­
marize the inform ation contained in K, and
L, that is relevant to the consum er’s decision.
O ur task is to find a recursive competitive
equilibrium, that is, decision rules i(k„ K„ A,)
and l(k„ K„ At) for the household, functions
I(K„ A,) and L(K„ A,) determ ining aggregate
investm ent and labor, and a value function
V(k„ K„ At) such that

L (K l, A t ) = l(K l,K „ A l )

Functional Forms
Subsequently, the production function is
assumed to be Cobb-Douglas:

AtF (K t,L t) = e*' K°L\-e.
T he technology shock eA is driven by an
A R (1) process:
(8 )

A, =

+ e ,,


where the £, are independent and identically


s Hansen and Prescott (1995) would
call Z,an "aggregate decision vari­
able." Other solution methods
would solve for pricing rules that
specify IV, and It,as functions of
stote variables rather than for on
"aggregate decision rule" for Lt .
In either case, the functions capture
the substance of the assumption of
rational expectations.
6 Taylor and Uhlig (1990) compare a
number of approaches to solving
stochastic growth models. Other
papers in the same issue of the
Journal of Business and Economic
Statistics provide short descriptions
of the various solution methods.

R E V I E W
the early RBC literature, would w ork like this:
Since agents are identical in this m odel, a
Pareto optimum can be found by maximizing
utility (9 ) subject to society’s production possi­
bilities, ignoring market structure. Production
possibilities are described by the production
function, the process generating technology
shocks, and the capital accumulation equation.
T his is a m uch sim pler problem . Since the
model has no distortions, the Second W elfare
Theorem applies: The Pareto optim al alloca­
tion can be supported as a com petitive equi­
librium . Thus, the solution to the social
planner’s problem replicates the outcom es
o f a decentralized com petitive system.
Rather than taking the shortcut o f solving
the social planner’s problem , this section fol­
lows the more general m ethod described in
Hansen and Prescott (1 9 9 5 ) that also applies
to models with distortions.® Two such models
are briefly described in the section titled
“Extending the Basic M odel.”
There are two keys to finding the value
fu nction V using functional equation 5 or 5 '.
T he first is approxim ation, described shortly.
T he second is the C ontraction Mapping
Theorem , a fixed point theorem , w hich guar­
antees that for certain problem s the following
steps will converge to the value function V.
T he theorem does n o t actually apply to m any
RBC m odels, so there is no guarantee in gen­
eral, b u t this approach usually converges
anyway, finding the correct value function.

distributed norm al random variables that are
independent o f all t - 1 variables.7 It is easier
to w ork w ith A, rather than A, as a state vari­
able from this point forward.
The utility function is specified in a
som ew hat unusual way that incorporates
indivisibility o f labor inputs, the contribu­
tion o f H ansen’s paper,
(9 )

U (ct, I - l t) = lo g c , + B ( l - 1(),

where B is a constant. T he Bellm an equation
5 becom es
(5 ')

V(fct,K ,, A,)
= max|log^eA
,FK(K (,L [)l?1
+ eA
,FL( K „ L , ) l , - i 1 )
+ B ( i-I ,)

+ /jE{v(k,+1 t+
,K I,A,+1)| t} }
Flansen showed that using a representative
consum er w ith this utility fu nction produces
the same competitive allocations as individual
consum ers described in the follow ing way.
Each consum er w orks either l0 hours or not
at all, b u t gets paid in either case. T h e prob­
ability o f w orking, chosen by the consum er,
is a,. Labor supply is determ ined indirectly
by choosing a, rather than direcdy by choos­
Total labor tim e in the econom y is
thus L,= a , l0. If the utility function o f indi­
vidual consum ers is

7 This specification differs slightly from
Honsen's, which assumes that the
technology is
L,), with e,
lognormolly distributed. Choice of
an ARO) process might be consid­
ered port of calibration since a spec­
ification search is usually part of the
overall process of estimating 7 .
1 The method described here and in
the appendix differs in details from
Hansen ond Prescott's algorithm.

1. Choose an arbitrary function
V°(fc(+1,K ,tl, At+1).
2. T he problem on the right-hand side o f 5
is now a static optim ization. Solve it to
get d ecision rules it= i(k„ K„ A,) and
lt= l(k t, Kt, At). Substitute these into the
right-hand side to produce a new func­
K„ At).
3. Replace V° on the right-hand side o f 5
w ith V 'O e ^ .K ^ .A .J . Return to step two
unless V ‘ and V l+ are alm ost identical.

UCCp a , ) = logc, + A a , l o g ( l - ( 0 ),

w ith A > 0, then the representative consum er
w ill have utility 9. By m aking these m odifi­
cations, H ansen hoped to im prove on the
rather poor performance of the basic model
(with divisible labor) in m atching facts about
relationships am ong hours, em ploym ent,
output and productivity.

Unfortunately, in general, step 2 w ill n ot
produce a function that can b e w ritten down
in any com pact way, particularly given the
presence o f the expectation in the m iddle of
the right-hand side of 5. This problem is
addressed in Hansen and Prescott’s m ethod
by solving a quadratic approxim ation o f the

Iterating to Find the Value Function
H ansen’s model could be solved using
the shortcut o f finding a Pareto optim um ,
that is, solving the social planner’s problem .
That approach, w hich was used extensively in


r e v i e w

m odel, rather than the full m odel. Variations
on the linear-quadratic approxim ation have
been the m ost com m on m ethod o f solving
dynamic general equilibrium m odels, starting
with Kydland and Prescott (1 9 8 2 ).
F or sim ulation purposes, the model is
approxim ated by a Taylor series expansion
o f the utility function (as it appears in 4 after
all the substitutions) around the steady-state
equilibrium values (K , L, X) that would occur
if we set all the shocks to zero. In some models,
the zero-shock equilibrium is not a steady
state. In these cases, a sim ple change o f vari­
ables usually produces the required steady
state. If, for exam ple, the population were
assum ed to b e growing, the model would be
form ulated in per capita terms. F or H ansen’s
m odel, (5 ') becom es



one household. But the distinction between I,
and L, m ust be maintained because the
household m ust behave as if it takes prices
(W, and Rt) as given, and these would be
functions o f L„ n ot l„ if we had m ore than
one consumer.
So how should L, be handled? (Though it
does n o t appear in the m odel, we also need to
worry about I, for reasons that will become clear
momentarily.) The model assumes that house­
holds have rational expectations about L, and
I„ so they recognize that equilibrium values of
these variables will satisfy the first-order con­
ditions. In m axim izing the right-hand side of
5 " at each iteration, the first-order conditions
define a linear relationship am ong choice
variables 1 and i„ aggregates L, and lt, and
state variables:

V(fe„ K„ A,)

0 = uo, + uult + u 2lit+ u 3,Ll

= maxjz'QZ,

+ u4,Il + u 5,k l + u6lK t + un At

U 1

0 = Uol + u lll, + u 2iiI+ u 3,L l

+ pE[v(kt+ Ki+ ,A1 )|A,}}
1 +1

+ u4jI t + u5ife, +
where 2 ,= [1 A, k, K, i, 1, I, L ,]'. Including
1 as a state variable allows the quadratic
approximation to be w ritten as a single qua­
dratic form (see the appendix).
The beauty of the quadratic approximation
is threefold. First, one can guess (correcdy)
that the value fu nction is quadratic. Second,
it does n ot depend on the distribution o f e,
except for a constant that involves the covari­
ance m atrix o f £,. Third, this constan t is not
essential for our analysis because it does not
involve any o f the state variables. Because the
constant is n ot essential, we can ignore it and
the expectation along with it. F or details, see
Sargent (1987, section 1.8), but it is not difficult
to see that the iterations described above will
always produce a quadratic if V° is quadratic.

+ u7iA,.

T h e first-order conditions can be solved for
1, and i, to get household d ecision rules spec­
ified in term s o f state variables, as well as L,
and I,. However, if L, is substituted for I, and
I, for i, in the first-order conditions (thus
im posing 7 ), the equations can be solved for
L, and I, as functions o f state variables. These
solutions can be interpreted as households’
conditional expectations o f aggregate labor
supply and investment, given the current values
o f state variables. Hansen and Prescott call
these “aggregate decision rules.” The solutions
for aggregate labor supply and investm ent
replace L, and I, in the household decision
rules w hich then becom e fu nctions o f state
variables alone.’ This procedure ensures that
cond ition 7 for a recursive com petitive equi­
librium is satisfied at every iteration.
W hen the value function approxim ations
converge, we have found a value function,
decision rules for 1 and i„ and aggregate labor
supply and investm ent fu nctions that satisfy
equations 6 and 7 by construction. T he C on­
traction M apping Theorem does n ot apply to
this particular dynamic programming problem,
but the algorithm does find the value function

Imposing Equilibrium Conditions
M ost o f the pieces o f a solution m ethod
have already been described, but there is one
m issing, namely, how to handle Lt. T his is
neither a state variable nor a decision variable
of any agent. It is an aggregate outcom e o f
households’ decisions. T h e aggregation hap­
pens to be trivial here because there is only

u 61K ,


9 Though it appears here formally, I,
drops out of household decision
rules for this model, that is.

Table 1

Calibration strategies are often m uch more
com plex than this, and the ju stifications more
Percentage Standard D eviations and
elaborate, bu t they always have the same
Correlations with Output
simple purpose, to select a plausible parameter
United States1
point at w hich to study the behavior o f the
m odel. Kydland and P rescott (1 9 9 4 ) detail
calibration strategies and their own philosophy
1 .76
of calibration. Researchers often conduct
0 .8 5
0.51 inform al sensitivity analyses, varying the
0 .8 6
8 .6 0
0 .9 9
5 .75 0 .9 2
param eters w hose values are m ost uncertain.
T h e m ost com m on such exercise seem s to
00 .6 3
.0 4
0 .4 8
0 .0 6
be to vary the risk aversion param eter in
0 .7 6
0 .9 8
models in w hich utility is o f the constant
0 .4 2
0.51 relative risk0.87
aversion form.
1 As reported by Hansen, Table 1.
2 Averages across 500 simulations.

if given an appropriate V°. The results are, of
course, subject to the caveat that if shocks take
the econom y too far from the steady state, the
quadratic approxim ation may n ot be accurate.


Ibis differs slightly from Honsen's
value (0.00712) because technol­
ogy shocks ore specified in a differ­
ent way.
See Kydland and Prescott (1982,
p. 1362).

There are six unknow n param eters in
the description o f Hansen’s m odel, 6 , 8, /3, B,
7 and the variance o f £. H ansen chose values
as follows. Given Cobb-D ouglas technology,
9 is capital’s share o f output. He used an esti­
mate o f 6 = 0 .3 6 based on tim e series for the
U .S. economy. A choice o f 8 = 0 .0 2 5 (corre­
sponding to an annual depreciation rate of
10 percent) was chosen as “a good compromise
given that different types o f capital depreciate
at different rates.” A steady-state annual
riskless real interest rate o f 4 percent would
be implied by /?= 0.99. Hansen chose B = 2.85,
w hich corresponds to an apparently arbitrary
value o f A = 2 in 10, com bined w ith I0= 0 .5 3 .
T he 0 .5 3 value equated steady-state hours in
Hansen’s divisible and indivisible labor models.
The standard deviation of e was chosen so
that, for the artificial econom y with indivisible
labor, the standard deviation o f detrended
output would be about the same as that of
detrended GNP for the U.S. economy. A value
cr = 0 .0 0 7 1 7 m eets this criterio n .1 A value of
7 = 0 .9 5 was the first-order autocorrelation
coefficient o f the Solow residuals for the U.S.

O nce the value function is found and
agents’ decision rules are know n, it is rela­
tively sim ple to sim ulate the model. The
equations o f m otion for A, and Kt along with
the aggregate d ecision rule for It are a system
of three linear difference equations in three
unknow ns that can easily b e sim ulated.
(R ecall that K,= k, in equilibrium .) Starting
values are chosen for the state variables and
innovations e, are drawn randomly.
T he real and artificial data are filtered
by taking logarithm s and detrending w ith
the H odrick-Prescott filter." There are two
reasons for filtering. First, the model is
intended to explain phenom ena at business
cycle frequencies, and the H odrick-Prescott
filter highlights those frequencies. T h e
models are n o t intended to m atch long-run
growth facts, so it would be unfair to compare
low -frequency m ovem ents in the data w ith
those from the m odel. A filter that removes
low -frequency m ovem ents in the data and
model output allows the m odel to be com ­
pared to phenom ena in the data it was
designed to explain.
Second, m any m acroeconom ic time
series may n ot be stationary. If this is true,
their second m om ents do not exist. Though
it would still be possible to generate sam ple
second m om ents, there w ould b e no reason
to think that another set o f observations on
the same econom y would produce sample
second m om ents sim ilar to the first set.
Thus, there would be no point in trying to
produce models that matched a particular set

R E V I E W
of sample second m om ents. Filtering that
induces stationarity removes this problem in
the sense that samples drawn from the same
data-generating process would be expected to
produce sim ilar sample second m om ents.
There are other ways to filter the data,
by taking first differences, for example. W hile
the H odrick-Prescott filter is som ew hat con ­
troversial (Cogley and Nason, 1 9 9 5 ; King and
R ebelo, 1 9 9 3 ), proponents argue that it does
a good jo b o f em phasizing the m ovem ents in
the data that m ost m acroeconom ists would
call business cycle m ovem ents. F o r exam ple,
Kydland and Prescott (1 9 9 0 ) say “...the
im plied trend path for the logarithm o f real
GNP is close to the one that students o f busi­
ness cycles and growth would draw through
a tim e plot o f this series.” Cogley and Nason
argue, on the other hand, that if the data are
an integrated process, “the filter can generate
business cycle periodicity and co-m ovem ent
even if none are present in the original data.
In this respect, applying the HP filter to an
integrated process is sim ilar to detrending
a random w alk.”
The results in Table 1 summarize 500
simulations o f 115 periods each. The statistics
are calculated from natural logarithm s of each
series detrended using the H odrick-Prescott
filter w ith a param eter o f 1 6 00 . E ach sim ula­
tion chooses A0 from the uncondition al dis­
tribution o f
This is more difficult for K0, so
I sim ulate the m odel for 1 0 0 periods, starting
K at its steady-state value. The value after
100 periods is used as K0 for the sim ulations
reported in Table I .1 In other words, I simu­
late the model for 215 periods and throw away
the first 100 in order to get a representative
distribution o f starting capital stocks. This
procedure allows the statistics reported in the
table to be interpreted as averages across 5 0 0
identical artificial econom ies w ith indepen­
dent realizations o f the technology shocks.
Evaluation o f RBC m odels’ output has
usually been inform al. Hansen, for exam ple,
was interested in how well this indivisible labor
model performed relative to a more standard
divisible labor model (w ith utility given by

indivisible labor model and 1.4 for the U.S.
economy. He argued that the indivisible labor
model showed prom ise because standard
m odels chronically produced ratios that were
too small. Since the real world is not charac­
terized by fully indivisible labor, he argued, it
was good that the ratio for the U.S. econom y
lay betw een the two m odels’ predictions.
A few additional tools have been used
to evaluate model output. A fairly com m on
approach is to com pare im pulse response
fu nctions from the model w ith those from
a vector-autoregression on the data. W atson
(1 9 93 ) has proposed a procedure for evaluating
the fit o f a calibrated model. A variety o f new
approaches is discussed in Pagan (1 9 9 4 ),
w hich is a thought-provoking overview of
calibration exercises.

T he solution technique described above is
m ore powerful than needed to solve H ansen’s
m odel, b u t m akes it possible to outline how
more sophisticated models can be handled. I’ve
chosen two exam ples from areas in w hich
extensive contributions have been made. These
particular papers fit easily into the framework
developed above.

Fiscal Policy
One obvious road to follow in generalizing
the basic RBC model is to add fiscal policy to
the model. Though obvious, this was not
initially an easy road because a m inim ally
realistic m odel requires distorting taxes. T he
first generation o f solution m ethods that
relied on solving the social planner’s problem
are not appropriate for m odels in w hich the
Second Welfare Theorem is not true.
A recent contribution in this area is
M cGrattan (1 9 9 4 a , 1 9 9 4 b ), w hich developed
a model in w hich agents face stochastic tax
rates on both labor and capital incom e.
M cGrattan (1 9 9 4 b ) modified the indivisible
labor model as follows. The governm ent uses
tax revenue to fund government purchases
and lump-sum transfers. Thus, the consumer’
budget constraint (2 ) is replaced by

log ct - A log 1 He noted that the standard
deviation o f hours relative to that o f produc­
tivity in the divisible labor model was only
about 1 com pared to 1.34/0.51 = 2 .6 for the


12 The values reported in columns
three and four of Table 1 differ very
slightly from those reported by
Hansen. Sampling variation and o
slightly different process generating
X probably account for the differ­


m oney balances carried over from last period,
mM, plus the lum p-sum transfer o f seignior­
age revenue, gMt V
T h e household budget constrain t is

c, + i , < ( l - T , ) R , f e ,

+ S t ,!?, + (1 - 0,) W,I,+ £ ,,
where the capital tax allows a depreciation
deduction and
is the lum p-sum transfer.
The tax rates, Tt and <r and governm ent pur­
chases, g(, are exogenous state variables (like
the technology shock A,). T h e size o f the
transfer is determ ined by the governm ent
budget constraint w hich im poses period-byperiod budget balance:

P,c, + P,i, + m, = PtW,l, + P(R,fe,
+ " V i + gM,-!T he last term on the left represents m oney
balances carried into t + 1 .
Because positive m oney growth results
in inflation, it is necessary in this m odel to
m ake a change o f variables for the zerosh ock path to be stationary. (R ecall that we
need a steady state around w hich to form a
quadratic approxim ation.) There is a steady
state w hen the m odel is w ritten in term s of
m, = m,/M, and P, = P,/M(. The two con­
straints (transform ed by the change o f vari­
ables) are used to elim inate c, and !, from the
consum er’s utility fu nction, leaving an opti­
m ization over m, (m oney holdings carried
into £+1) and i(.
T h e m oney supply M, is an exogenous
state variable, and is added to the list o f equa­
tions o f m otion. T he endogenous state vari­
ables are fe(, K, and mt V An equation of m otion
that says that this period’s purchases o f m oney
becom e n ext period’s m oney state variable is
added to the list for endogenous state variables
(the B matrix in the notation of the appendix).
T he aggregates that must be determined
are an investm ent function I (A,, M(, K,) and an
aggregate price function P (A,, Mt, K ,). Since M,
is exogenous, there is no aggregate that co r­
responds directly w ith the d ecision variable
m,. The aggregate price level P, serves this role
instead: At each iteration, I(A ,,M ,,K () and
P(A,,M ,,K() are chosen by setting i,= I, and
m,= 1 (th at is, m,= M,) in the first-order con ­
ditions and solving for I, and P, as functions
of state variables. (F o r H ansen’s m odel, by
contrast, we set i = I t and !(=L , in the firstorder cond itions, then solved for I, and L(.)
Details on how to handle this slight variation
can be found in Cooley and H ansen (1 9 8 9 ).

£ = t ,(R t- 5 ) K , + * ,W ,L ,- G ,.
The representative agent must form expec­
tations about assum ing that her decisions
have no influence over governm ent revenue.
Thus, K, and L, rather than fe, and l( appear
in the governm ent budget constraint. In this
model, there are four exogenous state variables,
A,, T(, <t and G,. M cG rattan calibrated the
stochastic processes for these variables by
estim ating a first-order autoregression for
each. She argues that her results indicate the
addition o f these fiscal shocks to the basic
indivisible labor model brings it into “m uch
better agreem ent w ith the data.”
To solve the m odel using the procedure
outlined above, substitute the new budget
constraint into the utility function, as before.1
Then substitute the right-hand side o f the
governm ent budget constraint for
Add the
assumed stochastic processes for tax rates and
governm ent purchases to the list o f equations
o f m otion (th e m atrix A in the language o f
the appendix).

Cooley and Hansen (1 9 8 9 ) studied an
RBC model with a cash-in-advance constraint.
In the sim pler version o f their m odel, the
m oney supply M, grows at a constan t rate, g:
M , = ( l + g)M ,_1.
Households’ consum ption decisions m ust
satisfy the cash-in-advance constraint,

RBC m ethods have found application
in a wide spectrum o f questions in business
cycles, m onetary econom ics, open econom y
m acroeconom ics and finance. The linear-

P,c, <m«_j + g M ,.lt
13 McGratton octuolly used g method
described in McGrattor (1994c).

w hich says that the nom inal value o f con ­
sum ption purchases, P,c(, m ust n o t exceed



_ _ _ _ _ _ _ _ a_n d_ _ _ _ _ _ _ _ _. "Time to Build ond Aggregate
Fluctuations," Econometrica (November 1 9 8 2 ) .

quadratic approach described in this paper is
a popular tool that can be applied to a broad
cross-section o f the questions posed in this
literature. More generally, careful study o f a
specific m ethod illustrates how equilibrium
conditions m ust be tied to optim ization to
find an equilibrium numerically.

Lucas, Robert E. "Methods and Problems in Business Cycle Theory,"
Journal of Money Credit and Banking (November 1 9 8 0 , part 2 ) .
_ _ _ _ _ _ _ _ ._ "Understanding Business Cycles,"
Series on Public Policy ( 1 9 7 7 ) .
McCallum, Bennett T. "Real Business Cycle M odels" in Robert J. Barro,
e d ., Modern Business Cycle Theory. Harvard University Press, 1 9 8 9 .


M cGrattan, Ellen R. "A Progress Report on Business Cycle M odels,"
Federal Reserve Bank of Minneapolis
Quarterly Review (fall 1 9 9 4 ) .

Blanchard, Olivier Jean, and Stanley Fischer.
Lectures on
Macroeconomics. The MIT Press, 1 9 8 9 .
Cogley, Timothy, and Jam es M. Nason. "Effects of the Hodrick-Prescott
Filter on Trend and Difference Stationary Time Series: Implications for
Business Cycle Research,"
Journal of Economic Dynamics and Control
(January/February 1 9 9 5 ) .

_ _ _ _ _ _ _ _ . _ "A Note on Computing Competitive Equilibria in Linear
M odels," Journal of Economic Dynamics and Control (January 1 9 9 4 ).
Pagan, Adrian. "Calibration and Econometric Research: An Overview,"
Journal of Applied Econometrics (December 1 9 9 4 , supplem ent).

Cooley, Thom as F., ond Gary D. Hansen. "The Inflation Tax in a Real
Business Cycle M odel,"The American Economic Review (September
1 9 8 9 ).

Plosser, Charles I. "Understanding Real Business Cycles,"
Journal of
Economic Perspectives (sum m er 1 9 8 9 ) .

Hansen, Gary D. "Indivisible Labor and the Business Cycle," of
Monetary Economics (November 1 9 8 5 ).
_ _ _ _ _ _ _ _ and Edward C. Prescott. "Recursive Methods for Computing
Equilibria of Business Cycle Models," chapter 2 in Thomas F. Cooley, ed.,
Frontiers of Business Cycle Research. Princeton University Press, 1 9 9 5 .
King, Robert G ., and Sergio T. Rebelo. "Low Frequency Filtering and
Real Business Cycles,"
Journal of Economic Dynamics and Control
(January/M arch 1 9 9 3 ).

Sargent, Thomas 1. Dynamic Macroeconomic Theory. Harvard
University Press, 1 9 8 7 .
Stodler, George W. "Real Business Cycles,"
Journal of Economic
literature (Decem ber 1 9 9 4 ) .
Stokey, Nancy L., Robert E. Lucas and Edward C. Prescott.
Methods in Economic Dynamics. Harvard University Press, 1 9 8 9 .
Taylor, John B., and Harold Uhlig. "Solving Nonlinear Stochastic Growth
Models: A Comparison of Alternative Solution M ethods,"
Journal of
Business and Economic Statistics (January 1 9 9 0 ).

Kydland, Finn E ., and Edward C. Prescott. "The Computational
Experiment; An Econometric Tool," Federal Reserve Bonk of
Minneapolis StaffReport 178 (August 1 9 9 4 ) .

Watson, Mark W. "M easures of Fit for Calibrated M odels," of
Political Economy (December 1 9 9 3 ).

_ _ _ _ _ _ _ _a_n d _ _ _ _ _ _ _ _ _ _ . "Business Cycles: Real Facts and a
Monetary M yth,"Federal Reserve Bank of Minneapolis Review
(spring 1 9 9 0 ).
Long, John B ., and Charles I. Plosser. "Real Business Cycles," of
Political Economy (February 1 9 8 3 ) .

_ _ _ _ _ _ _ _. _ "The Macroeconomic Effects of Distortionary Taxation,"
Journal of Monetary Economics (June 1 9 9 4 ).



A p p e n d ix


or, com bining all three,

Let p (x ) be the num ber o f rows in the
vector x. Let z be the vector o f exogenous
state variables, s and S the vectors of individual
and aggregate endogenous state variables, d the
vector o f decision variables, and D the vector
o f “aggregate decision variables.” For reasons
o f convenience, the first elem ent o f z is
always 1. Let




X = CZ = Bj B2







S + 0

B1 0 B2+B 3 0 B4+ B 5


T h e m atrix C is p (X ) x p (Z ).
F o r H ansen’s m odel,
Z ,= [l A, k , K, i, I, I, L J ' ,





1 0
0 7

Z =
B = [0 0 ( 1 - 5 ) 0 1 0 0 0],

U sing - to denote next-period values, the
constraints are

The General Form of the Problem
U sing the notation in H ansen and
Prescott (1 9 9 5 ), the problem we wish to
solve in order to find an equilibrium is

Z = AZ

+ £ = [A 0




(A l) V (X ) = m a x r (Z ) + 0 E{V(X)|2 }


su b ject to X = C Z +

s = BZ = [Bj B2 B3 B4 B5 ]

w ith

D = D (z,S).

All nonlinear constraints have been
substituted into the return fu nction r. As
m entioned above, the D = D(z, S ) equation
sum m arizes agents’ rational expectations
about aggregate values o f their own decision
variables, for exam ple, labor supply. Since it
does not involve the choice variables d, it is not
really a constraint, but will be used to derive
decision rules that depend only on X rather
than b oth X and D. This is w hy the problem
specifies “w ith” rather than “su b ject to .”

S = [ B , B2 B3 B 4 B 5

Steady-State Solution

= [Bj 0 B2 + B 3 0 B 4 + B 5

The return function will be approximated
around a steady-state solution to the model
w hen £ = 0 . T he steady-state solution solves


the follow ing set o f p (Z ) equations in
elem ents o f Z:

quadratic problem . The value fu nction for
this kind of problem is known to be quadratic.
The first function in the sequence can be any
quadratic function, but convergence is som e­
times sensitive to the choice. Hansen and
Prescott recom m end V< 17I, where 17 is a
small negative number.
If V(n is the p (X ) x p (X ) m atrix o f the nth
quadratic value function approxim ation, the
(n+ l ) st approxim ation is given by

X = CZ (excluding the s row s)
0 = rd(Z) + Pri ( Z ) l I - p B s]~l Bd

The first vector equation contains

p ( z )+ p (S ) scalar equations.1 The second
vector equation contains p(d) scalar equations
that are the first-order conditions for d. They

ma XZ'QZ + P X 'V ^ X

take into accou nt the recursive nature o f A l,
that is, the fact that d affects every future
period by changing s. Bs is the derivative of
BZ w ith respect to s, that is, the p ( s) colum ns
o f B that correspond to the s elem ents o f Z
(the p ( z ) + 1 through p (z )+ p (s) elem ents).
Similarly, Bd consists o f the p (d ) colum ns of
B that correspond to the d positions in Z
(th e p (X ) + l through p (X )+ p(d) elem ents).

su b ject

D = GX.

The equation D = G X summarizes the “aggregate
decision rules” implied by VM (the derivation
o f these is described below ). As explained
above, this equation is not a constraint.
F irst elim inate the equations o f m otion
for state variables to get an equivalent problem
m ax Z'QZ + / 3 Z 'C V n)CZ w ith

The Quadratic Approximation as a
Quadratic Form


m ax Z 'R Z with

L et x be a k row vector and l e t b e a
(fc+ l)x (fc+ l) m atrix. Partition ’P so that 'P'1
is a scalar. Then


D = GX

D = GX.

Decision Rules


"1 "' 4*l
X . 2

to X = CZ


The first-order conditions for this problem
are the second-to-last p (d ) rows (the p (X ) +1
through p (X ) + p (d ) rows) o f

='J'11+ ('F 2 'F 12)x


0 = (R + R ')Z .

Thus, any quadratic function from R feinto R
o f a vector x can be w ritten as a single qua­
dratic form in [ 1 x'] by collecting the constant
terms in 'Pjj, the linear terms in ('P 2 + 'I'12) x,
and the squared terms in x ' ^ 22x.
The quadratic approximation to the return
function is subsequently written in this way.
D enote the quadratic approxim ation to r(Z )
by Z'QZ. The m atrix Q involves derivatives
that can be calculated either analytically or

W rite these p (d ) equations as
0 = UZ
where U is p (d ) x p (Z ).
Partition U to conform w ith the com po­
nents o f Z:

o= m

U2 U3 U4 U5 ] S

Substituting Out Constraints and
Equilibrium Conditions

Solving for d yields2

The remainder o f this appendix describes
the process o f finding a sequence o f approxi­
m ations to the value function that converges
to the actual value function for the linear-

d = -U ~ 1(U1z + U 2s + U 3S + U5D).
The aggregate decision rules are obtained


1 Numerical methods for solving
these equations usually require
dropping the equation that
describes the evolution of the state
variable 1. Most economists have
little trouble solving this equation
analytically, however.
2 In some circumstances i/4 is close
to singular, which leads to conver­
gence problems. An alternative is
to impose the equilibrium condition
0= d before solving for d, which
instead requires ((/4+ (/ s) _ l . One
caveat is needed, however. This
variotion leads to individual deci­
sion rules which have the wrong
weights on dand Dthough the
sum of the weights is correct (so
that aggregate decision rules are
correct). If, for some reason, indi­
vidual decision rules are needed,
the correct weights can be found by
choosing them so that the firstorder conditions used in finding the
steady state are satisfied with D
its equilibrium values but with



by setting d=D and s= S in the first-order
conditions. This leads to

D = - ( U ,+ U5r 1 [U, 0

T h e value fu nction is updated u n til V(n+1) and
V< ) are sufficiently close. T he (1 ,1 ) elem ent
o f V tends to converge slowly, but, because it
does n ot enter the first-order cond itions
(U Z=0), that elem ent can be ignored w hen
testing for convergence, unless the value
function itself is needed.

«/2 + U3 )]

= GX.
Substituting this into the previous equation
gives individual decision rules that are func­
tions o f state variables only:

d = - U ? (UjZ + U2s + U3S + U5GX )
= - u ; 1([u 1 U2 u 3] + u 5g ) x
= HX.




= ]X

into the objective fu nction as follow s to
obtain the new value function:




Z'RZ =



x 'j

'J r

R 12


^21 ^22

= [x- x r ]


R 11

= [ x ' r 11+ x ' j ' r 21

X 'R 12+ X 'J 'R 22]

JX .

= X'Rn X + X 'J' R21X + X'Rn JX + X'J'R22JX
= X '[R u + J ' R 21 + R 12J + J 'R 22J]X

= x V n1 .
+ )x


States and European Com m unity for the
Integration o f High and Low Incom e
E con om ies.”

orking papers from the Federal
Reserve Bank o f St. Louis reflect pre­
lim inary results o f staff research and
are made available to encourage com m ent
and discussion, as w ell as invite suggestions
from other researchers for revision. The views
expressed in the w orking papers are those of
the authors and do n o t necessarily reflect
official positions o f the Federal Reserve Bank
o f St. Louis, the Federal Reserve Systems or
the Board o f Governors.
F o r published papers, contact the author
or the publication listed. Single copies of
w orking papers are available by w riting to:

9 4 - 0 0 1 B -Peter S. Yoo, “T he Baby Boom and
E con om ic G row th.”
9 4 - 0 0 2 A -Peter S. Yoo, “Age D istributions
and Returns o f Financial Assets.”
9 4 -0 0 3 A -P e £ e r S. Yoo, “Age D ependent
Portfolio S election.”
9 4 - 0 0 4 A -Joseph A. Ritter, “The Transition
From Barter to Fiat Money.” PUB­
LISHED: The American Economic Review
(M arch 1 9 9 5 ).

Federal Reserve Bank o f St. Louis
Research D epartm ent
PO . Box 4 4 2
St. Louis, MO 6 3 1 6 6 -0 4 4 2 .

9 4 - 0 0 5 A -Sangkyun Park, “The Bank
Capital Requirem ent and Inform ation


9 4 - 0 0 6 A - R i chard G. Anderson and Kenneth
A. Kavajecz, “An H istorical Perspective
on the Federal Reserves M onetary
Aggregates: A T im eline.” PUBLISHED:
this Review (March/April 1 9 9 4 ).

9 5 - 0 0 1 A -Patricia S. Pollard and Rowena A.
Pecchenino, “Dependent Children and
Aged Parents: Funding Education and
Social Security in an Aging Econom y.”
9 5 - 0 0 2 A -Christopher J. Neely, Amlan Roy
and C harles Whiteman, “Identification
Failure in the Intertem poral
C onsum ption Capm .”

Kenneth A. Kavajecz, “T h e Evolution o f
the Federal Reserves M onetary
Aggregates: A T im eline.”

9 4 -0 0 7 C -R ic h a rd G. Anderson and William
G. Dewald, “Replication and Scientific

9 5 -0 0 3 A -M ic h a e I J. D ueker and Patrick K.
Asea, “N on-M onotonic Long M em ory
D ynam ics in Black-M arket Prem ia.”
FO RTH CO M IN G: The Journal o f

Standards in E conom ics a Decade Later:
T he Im pact o f the JM C B P ro ject.” PUB­
LISHED: this Review
(November/December 1 9 9 4 ).

Com parative Economics.
9 5 - 0 0 4 A -Joseph A. Ritter and Joseph
Haubrich, “C om m itm ent as Irreversible
Investm ent.”

9 4 -0 0 8 B -C h r is to p h e r J. Neely, “Target
Zones and Conditional Volatility: The
R ole o f Realignm ents.”

9 5 - 0 0 5 A - M ichael J. D ueker and Andreas M.
Fischer, “Identifying Austria’s Im plicit
M onetary Target: An Alternative Test o f
the ‘Hard Currency’ Policy”

9 4 - 0 0 9 A -C hristopherJ. Neely, Dean C orbae
and Paul Weller, “T he D istribution of
Target Zone Exchange Rates Under
Alternative Realignm ent R ules.”

9 5 - 0 0 6 A -Jam es B. Bullard and John Duffy,
“O n Learning and the Stability of Cycles.”

9 4 - 0 1 0 A -C hristopherJ. Neely, “A
Reconsideration o f the Properties o f the
G eneralized M ethod o f M om ents in
Asset Pricing M odels.”

9 5 - 0 0 7 A-Cletus C. Coughlin and David C.
W heelock, “Lessons from the United




1 9 9 5

94-025A -Sangfe_yun P ark, “Banking and
D eposit Insurance As a Risk-Transfer
M echanism .”

9 4 - 0 1 1 A -Jam es B. Bullard and Joh n Keating,
“Superneutrality in Postwar Econom ies.”
9 4 - 0 1 2 A -Jam es B. Bullard and Steven H.
Russell, “M onetary Steady States in a
Low Real Interest Rate Econom y.”

9 4 - 0 2 6 B - M ic h a e l R. P akko, “C haracterizing
C ross-C ountry Consum ption
C orrelations.”

9 4 - 0 1 3 A -James B. Bullard and John Duffy,
“Learning in a Large Square Econom y.”

9 4 - 0 2 7 B - M ic h a e l J. D ueker and Richard
Startz, “M axim um -Likelihood Estim ation
o f Fractional C ointegration W ith An
A pplication to the Short End o f the
Yield Curve.”

9 4 - 0 1 AB-James B. Bullard and Joh n Duffy,
“A M odel o f Learning and Em ulation
w ith Artificial Adaptive Agents.
9 4 - 0 1 SA-M ichael J. Dueker, “M ean
Reversion in Sto ck M arket Volatility.”

9 4 - 0 2 8 A -Jam es B. Bullard and Joh n Duffy,
“U sing G enetic Algorithm s to M odel the
E volution of H etrogeneous Beliefs.”

9 4 - 0 1 6B -M ichael J. Dueker, “Com pound
Volatility Processes in EM S Exchange
R ates.”

9 4 - 0 2 9 A -Donald S. Allen, “W h y Does
Inventory Investm ent Fluctu ate So
M uch D uring C o n traction s?”

9 4 - 0 1 7A -M ichael J. D ueker and Daniel L.
Thornton, “Asym metry in the Prim e Rate
and Firm s’ Preference for Internal
F in a n ce.”

9 4 - 0 3 0 A -M ichael Ulan, W illiam G. Dewald
and Jam es B. Bullard, “U .S. O fficial
Forecasts o f Group o f Seven E con om ic
Perform ance, 1 9 7 6 -9 0 .” PUBLISHED:
this Review (March/April 1 9 9 5 ).

9 4 - 0 1 8 A -John A. Tatom and D ieter Proske,
“Are There Adverse Real Effects from
M onetary Policy Coordination? Some
Evidence from Austria, Belgium and the
N etherlands.”

9 4 - 0 3 1 A -Peter S. Yoo, “The Baby Boom and
International Capital F low s.”
9 4 - 0 3 2 A - D a n ie ! L. Thornton, “The
Inform ation C ontent o f D iscount Rate
A nnouncem ents: W h at’s Behind THe
A nnouncem ent E ffect?”

9 4 - 0 1 9A -M ichael R. P akko, “R econciling
International Risk Sharing w ith Low
Cross-Country Consum ption


9 4 - 0 2 0 B -ChristopherJ. Neely, “Realignments
o f Target Zone Exchange Rate Systems:
W hat Do We K now ?”

9 3 - 0 0 1 A -Patricia S. Pollard,
“M acroeconom ic Policy Effects in a
M onetary U n ion .”

9 4 - 0 2 1 A -David C. W heelock and Paul W.
Wilson , “Productivity Changes in US
Banking: 1 9 8 4 -9 3 .”

9 3 -0 0 2 A -D a v ic J C. W heelock and Paul W.
Wilson, “Explaining Bank Failures:
D eposit Insurance Regulation, and
Efficiency.” FO RTH CO M IN G : The

9 4 - 0 2 2 A -Alison Butler and M ichael J.
Dueker, “Product Cycles, Innovation and
Relative W ages in European Countries.

Review o f Economics and Statistics.

9 4 -0 2 3 A -S a n g k y u n Park, “M arket
D iscipline By Depositors: evidence from
reduced form equations.”

9 2 - 0 0 1 A -John A. Tatom, “T h e P-Star Model
and Austrian P rices.” PUBLISHED:
Empirica (vol. 19, 1 9 9 2 , no. 1).

9 4 -0 2 4 A -B y u n g Chan Ahn, “M onetary
Policy and the D eterm ination o f the
Interest Rate and Exchange Rate in a
Small O pen E conom y W ith Increasing
Capital M obility.”

9 2 - 0 0 2 A -David C. W heelock and Subal C.
Kum bhaker, “T h e Slack Banker Dances:
Insurance and Risk-taking in the Banking



9 1 - 0 0 5 8 -David C. W heelock and Subal C.
Kum bhaker, “W h ich Banks Choose
D eposit Insurance? Evidence o f Adverse
Selection and Moral Hazard in a Voluntary
Insurance System .” FO RTH CO M IN G:

Collapse o f the 1 9 2 0 s.” PUBLISHED:

Explorations in Economic History (vol.
3 1 , Ju ly 1994, no. 3).
9 2 - 0 0 3 A -Daniel L. Thornton, “T he M arket’s
Reaction to D iscount Changes: W h at’s
Behind the A nnouncem ent E ffect?”

Jou rnal o f Money, Credit and Banking.
9 1 - 0 0 6 A -David C. W heelock, “Regulation
and Bank Failures: New Evidence From
the Agricultural Collapse o f the 1920s.
PUBLISHED: Journal o f Economic
History (D ecem ber 19 9 2 ).

9 2 -0 0 4 A - D a n ie l L. Thornton, “W hy Do
T-Bill Rates React to D iscount Rate
Changes? FO RTH CO M IN G: Journal

o f Money, Credit and Banking.
9 2 - 0 0 5 A -John A. Tatom, Heinz Gluck and
Dieter Proske, “M onetary and Exchange
Rate Policy in Austria: An Early Example
of Policy C oordination.” PUBLISHED:

9 0 - 0 0 1 -M ichelle R. Garfmkel and Seonghwan
Oh, “Strategic D iscipline in M onetary
Policy W ith Private Inform ation: Optimal
Targeting Periods.” PUBLISHED: The
American Economic Review (M arch 1993).

Economic Policy Coordination in an Inte­
grating Europe. Bank o f Finland, 1992.
9 2 - 0 0 6 A -John A. Tatom, “Currency
Appreciation and ‘D eindustrialization’:
A European Perspective.”

9 0 - 0 0 2 -John A. Tatom, “T he L ink Betw een
M onetary Aggregates and Prices.”

9 2 - 0 0 7 A -David C. W heelock, “G overnm ent
Policy and Banking Instability:
“O verbanking” in the 1 9 2 0 s .” PUB­
LISHED: Journal o f Economic History
(D ecem ber 19 9 3 ). (Published as
“G overnm ent Policy and Banking
M arket Structure in the 1 9 2 0 s .”)

9 0 - 0 0 3 - D a v id B. N ickerson and Stanley S.
Reynolds, “O ptim al M onopoly
Investm ent and Capacity U tilization
U nder Random D em and.”

9 0 -0 0 A -M ich elle R. G arfm kel and Seonghwan
Oh, “W h en and How M uch to Talk:
Credibility and F lexib ility in M onetary
Policy W ith Private Inform ation.”

9 2 -0 0 8 A -M ic h a e ! T. Belongia, “Selecting an
Interm ediate Target Variable for and
Dallas S. Batten M onetary Policy W hen
the Goal is Price Stability.”

9 0 - 0 0 5 -David Nickerson, M arsha Courchane
and Venkatraman Sadanand,
Precom m itm ent and Random Exchange
Rates in Sym m etric Duopoly: a New
Theory o f M ulti-national P rod uction.”

9 1 - 0 0 1 D-Mark D. Flood, “M arket Structure
and Inefficiency in the Foreign
Exchange M arket.” PUBLISHED:

9 0 - 0 0 6 -Steven Russell and Marco Espinosa,
“The Inflationary Effects o f the Use of
Reserve Ratio Reductions, or Open
M arket Purchases, to Reduce M arket
Interest Rates: A Theoretical
C om parison.”

Jou rnal o f International Money and
Finance (April 19 9 4 ).
9 1 -0 0 2 B -Ja m e s B. Bullard and Alison Butler,
“N onlinearity and Chaos in Econom ic
Models: Im plications for Policy
D ecisions.” PUBLISH ED: Economic

9 0 - 0 0 7 - A lis o n Butler, “Endogenous
Innovation in a N orth-N orth M odel of
the Product C ycle.”

Journal (July 1 9 9 3 ).
9 1 - 0 0 3 A -James B. Bullard, “Collapsing
Exchange Rate Regimes: A
R einterpretation.”

9 0 - 0 0 8 -John A. Tatom, “The P-Star Approach
to the L in k Betw een M oney and P rices.”

9 1 - 0 0 4 A -Jam es B. Bullard, “Learning
Equ ilibria.” PUBLISHED: Journal o f
Economic Theory (D ecem ber 19 9 4 ).




A rticles in the upcoming M ay/June issue of R eview
(the 1994 Conference Issue: Channels of M onetary Policy)
Th eo retical Issues of
Liq u id ity Effects

Inform atio n, Stick y P rices and
R elated Problem s

Alan C. Stockman and Lee Ohanian
Discussant: Kevin D. Hoover

Allan H. M eltzer
D iscussant: Randall Wright

R eso lvin g the Evid en ce on
the Liq u id ity Effect

P an e l D iscussion:
W h at Do W e K now A bout How
M o n e tary P olicy A ffects the
Econom y?

Adrian R. Pagan and John C. Robertson
Discussant: Lawrence J. Christiano

Is Th ere a " C re d it C h a n n e l"
for M o n etary P o licy ?
R. Glenn Hubbard
D iscussant: Bruce D. Smith

D isting uishing T h eo rie s of
the M o n etary Transm ission
M echanism
Stephen G. Cecchetti
D iscussant: M ark Gertler

Panelists: Ben S. B em anke, Thom as F.
Cooley, Manfred J.M. Neumann

A ll nonproprietary and nonconfidential data and programs for the articles published
in Review are available to our readers and can be obtained from three sources:
l.F R E D (F e d e ra l R e s e rv e Econom ic

2 . R e sea rch D e p a rtm e n t, Federal

D a ta ), a n e lectro n ic b u lle tin b oard
s e rv ice . You can access FRED by

Reserve Bank of St. Louis, Post Office
Box 4 42, St. Louis, MO 6316 6 -0 4 4 2 .

dialing (3 1 4 ) 6 2 1 -1 8 2 4 through a
modem-equipped personal computer.
For a free brochure on FRED, please
call (3 1 4 ) 444-8809.

You can request data and programs on
either disk or hard copy. Please include
the author, title, issue date and page
numbers with your request.

3 . In te r-u n iv e rs ity Consortium fo r
P o litical an d So cia l R e sea rch (ICPSR).

Member institutions can request
data through the CDNet Order facility.
Nonmembers should write to: ICPSR,
Institute for Social Research, P.O. Box
1248 Ann Arbor, M ichigan 4 8 1 0 6 , or
call (3 1 3 ) 763-5010.

Federal Reserve Bank of St. Louis
Post O ffic e B ox 4 4 2
St. Louis, M is s o u ri 6 3 1 6 6 - 0 4 4 2



recycled paper