Full text of New Neoclassical Synthesis and the Role of Monetary Policy

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

M a r v in G o o d frien d a n d R o b e r t G . K in g
FEDERAL RESERVE BANK OF RICHMOND AND UNIVERSITY OF
A N D UNIVERSITY OF VIRGINIA, NBER, AND FEDERAL
RESERVE BANK OF RICHMOND

The N e w

N e o c l a s s i c a l S y n t h e s i s a n d th e

R o l e o f M o n e t a r y Policy

2. In trod u ction
It is com m on for macroeconomics to be portrayed as a field in intellectual
disarray, with major and persistent disagreements about methodology
and substance betw een competing camps of researchers. One frequently
discussed m easure of disarray is the distance between the flexible price
m odels o f the n ew classical macroeconomics and real-business-cyde
(RBC) analysis, in which monetary policy is essentially unimportant for
real activity, and the sticky-price models of the N ew Keynesian econom­
ics, in w hich m onetary policy is viewed as central to the evolution of real
activity. For policymakers and the economists that advise them, this
perceived intellectual disarray makes it difficult to em ploy recent and
ongoing developm ents in macroeconomics.
The intellectual currents of the last ten years are, however, subject to a
very different interpretation: macroeconomics is m oving toward a New
Neoclassical Synthesis. In the 1960s, the original synthesis involved a com­
m itm ent to three— som etim es conflicting— principles: a desire to pro­
vide practical macroeconomic policy advice, a belief that short-run price
stickiness w as at the root of economic fluctuations, and a commitment to
m odeling m acroeconomic behavior using the same optimization ap­
proach com m only em ployed in microeconomics.

ThispaperbenefitedfrompresentationsattheBankofEnglandandtheworkshopon
"MonetaryPolicy,PriceStability,andtheStructureofGoodsandLaborMarkets"spon­
soredbytheBankofItaly,CentroPaoloBaffi,andIGIER.Theauthorsacknowledgehelpful
commentsfromB.Bemanke, O.Blanchard,C.Goodhart,M. Dotsey,B.Hetzel,B.McCallum,E.McGrattan,E.Nelson,J.Rotemberg,K.West,andA.Wolman.Theopinions
aresolelythoseoftheauthorsand do notnecessarilyrepresentthoseoftheFederal
ReserveSystem.




232 • GOODFRIEND & KING
The N e w Neoclassical Synthesis inherits the spirit o f the old, in that it
com bines Keynesian and classical elem ents. Methodologically, the new
synthesis involves the systematic application o f intertemporal optimiza­
tion and rational expectations as stressed by Robert Lucas. In the synthe­
sis, these ideas are applied to the pricing and output decisions at the
heart of K eynesian m odels, n ew and old, as w ell as to the consumption,
investm ent, and factor supply decisions that are at the heart of classical
and RBC m odels. M oreover, the n ew synthesis also em bodies the in­
sights of monetarists, such as M ilton Friedman and Karl Brunner, regard­
ing the theory and practice of m onetary policy.
Thus, there are n ew dynam ic m icroeconom ic foundations for macro­
econom ics. These com m on m ethodological ideas are implemented in
m odels that range from the flexible, small m odels o f academic research
to the n ew rational-expectations policy m odel o f the Federal Reserve
Board. The N e w Neoclassical Synthesis (N N S ) suggests a set of major
conclusions about the role of m onetary policy. First, N N S models sug­
gest that m onetary policy actions can have an im portant effect on real
econom ic activity, persisting over several years, du e to gradual adjust­
m ent of individual prices and the general price level. Second, even in
settings w ith costly price adjustm ent, the m odels su ggest little long-run
trade-off betw een inflation and real activity. Third, the m odels suggest
significant gains from eliminating inflation, which stem from increased
transactions efficiency and reduced relative price distortions. Fourth, the
m odels im ply that credibility plays an important role in understanding
the effects o f m onetary policy. These four ideas are consistent with the
public statem ents o f central bankers from a w ide range of countries.
In addition to the general points, N N S m odels allow the analysis of
alternative m onetary policy rules within a rational-expectations setting.
It is in this role that they can inform — rather than confirm — the priors of
central bankers. The credibility o f m onetary policy appears intuitively to
require a sim ple and transparent rule. But w hich one? W e use the NNS
approach to develop a set o f principles and practical guidelines for neu­
tral m onetary policy, defined as that w hich supports output at its poten­
tial level in an environm ent of stable prices. The n ew synthesis suggests
that such a m onetary policy involves stabilizing the average markup of
price over marginal cost. In turn, this im plies a m onetary policy regime
o f inflation targets, w hich vary relatively little th r o u g h time. A lth o u gh
price stability has been lon g suggested as a prim ary objective for mone­
tary policy, a num ber o f major questions have arisen about its desirabil­
ity in practice. W e confront a range of im plem entation issues, including
the response to com m odity price shocks, the lon g and variable lags
betw een m onetary policy and the price level, the potential policy trade­




The New Neoclassical Synthesis •233
off betw een price and output variability, and the use of a short-term
interest rate as the policy instrument.
The organization of our discussion is as follows. In Section 2, we
describe the general approach of the original neoclassical synthesis as it
was articulated by Paul Samuelson. In Section 3, we review why the
original neoclassical synthesis was never fully accepted by monetarists,
even at the height o f its influence in the 1960s, and then was more
fundam entally challenged by the rational-expectations revolution. We
then turn to m ore recent work in macroeconomics covering RBC models
in Section 4 , and N e w Keynesian economics in Section 5.
The N N S is introduced and described in Section

6.

We analyze the

effect o f m onetary policy within the new synthesis using two comple­
mentary approaches. First, w e em ploy the standard Keynesian method
that view s m onetary policy as affecting real aggregate demand. Second,
we use an RBC-style alternative which views variations in the average
markup as a source of variations in aggregate supply; these markup
variations are analogous to the effects of tax shocks in RBC models. We
use the insights of the previous sections to develop principles for mone­
tary policy in Section
Section

8.

Section

9 is

7 and

practical guidelines for monetary policy in

a summ ary and conclusion.

2 . The N eoclassical Synthesis
A s popularized by Paul Samuelson ,1 the neoclassical synthesis was ad­
vertised as an engine o f analysis which offered a Keynesian view of the
determination o f national incom e— business cycles arising from changes
in aggregate dem an d because of wage and price stickiness— and neoclas­
sical principles to guide microeconomic analysis. In our discussion of the
neoclassical synthesis, w e consider three major issues: the nature of the
monetary transm ission mechanism, the interaction of inflation and real
activity, and the role o f monetary policy.

2.1 THE M ON ETARY TRANSMISSION MECHANISM
The basic m acroeconom ic framework of the neoclassical synthesis was
the IS-LM m odel. The neoclassical synthesis generated a number of
advances in the 1950s and 1960s to make this framework more consistent
with individual choice and to incorporate the dynamic elements that
were so evidently necessary for econometric modeling of macroeco­
nom ic time series. Theoretical work rationalized the demand for m oney

1.Anearlydescriptionoftheneoclassicalsynthesisisfoundinthe1955editionofSamuelson'sEconomics,andthematuresynthesisisdiscussedinthe1967edition(Samuelson,
1967)
.




234 ■ GOODFRIEND & KING
as arising from individual choice at the margin, leading to a microeconom ic explanation of the interest rate and scale variables in the mone­
tary sector. The synthesis stimulated advances in the theory of consump­
tion and investm ent based on individual choice over time. Econometric
w ork on m o n ey dem and and investm ent developed dynamic partial
adjustm ent specifications.
These n ew elem ents were introduced into large-scale m odels of the
m acroeconom y. O u r discussion focuses on the Federal Reserve System's
M PS m o d el, w hich w as developed because "n o existing m odel has as its
m ajor purpose the quantification o f m onetary policy and its effects on
the ec o n o m y ," as de Leeuw and Gramlich (1968, p. 11) reported. The
M PS m od el initially included the core elem ents o f the IS-LM framework:
a financial block, an investm ent block, and a consum ption-inventory
block. The structure o f production possibilities and the nature of w ageprice dynam ics were view ed as important, but secondary in the early
stage o f m odel developm ent. Relative to other then-existing models, the
MPS m odel suggested larger effects o f m onetary policy because it incor­
porated a significant effect o f long-term interest rates on investment and
its estimated lags in the dem and for m o n ey suggested m uch faster adjust­
m ent than in earlier m odels.
In its fully developed form , circa 1972, the M P S m odel incorporated
several structural features that are worth stressing. It w as designed to
have long-run properties like that o f the consensus growth model of
Robert Solow, including the specification o f an aggregate production func­
tion im plying a constant labor share o f national incom e in the face of
trend productivity grow th. A s explained in A n d o (1974), however, the
M PS m o d el had a short-run production function w hich linked output to
labor input roughly one for o ne, as a result o f variations in the utilization
o f capital. The empirical m otivation for this feature is displayed in Figure

1 : over the course o f business

cycles, total m an -h ou rs and output display

similar am plitude, with m easured capacity utilization strongly pro­
cyclical. For the m ost part, these cyclical variations in total hours arise
largely from variations in em ploym en t rather than average hours per
w orker .2

2.2 INFLATION A N D REAL ACTIVITY
In the early years of the neoclassical synthesis, macroeconometric models
w ere constructed and practical policy analysis w as undertaken assuming
that nom inal w ages and prices evolved independently from real activity

2.IneachpanelofFigure1,outputisthelightersolidline.Thedataarefilteredtoisolate
periodiccomponentsbetween6and24quartersinduration.




The New Neoclassical Synthesis * 235
Figure 1 HOURS, EMPLOYMENT, AN D UTILIZATION




Aggregate Hours

236 • GOODFRIEND & KING
and its determinants. In fact, in the 1950s, there was relatively little vari­
ability in inflation. By the m id-1960s this prem ise could no longer be
m aintained— inflation becam e a serious policy concern and it was plain
to see that inflation w as related to developm ents in the econom y .3
The Phillips curve thus becam e a central part of macroeconomic model­
ing and policy analysis. M acroeconom ic m odels were closed with wage
and price sectors that indicated major trade-offs betw een the rate of
inflation and the level o f real activity. The M P S m odel specified that the
price level w as determ ined by a markup o f price over marginal cost, with
the nom inal w age rate being a central determinant of cost. In addition,
the M PS m odel m ade the markup depen d on the extent o f utilization and
allow ed the price level to gradually adjust toward marginal cost (Ando,
1974, pp. 544, 552). The M PS version o f the Phillips curve also specified
that the rate o f w age inflation depen ded on the u n em ploym en t rate and
the lagged rate o f change of nom inal prices. W ith these three assum p­
tions taken together, as in de M enil and Enzler (1972), the MPS model
suggested that the effect o f reducing the long-run rate o f inflation from
5 % to 0% w as an increase in the u n em ploym en t rate from 3 .5 % to 7% .
The nature o f the trade-off betw een inflation and unem ploym ent be­
came central to m acroeconom ic policy, as w ell as to macroeconomic
m odeling. Policy advisers worried about a w a g e-p ric e spiral and were
concerned that inflation could develop a m o m en tu m o f its ow n, as
appeared to be the case in the recession o f 1 9 5 7 -5 8 (O ku n et aL, 1969,
p. 96; O k u n , 1970, p.

8).

By the standards o f later years, the outcomes

for inflation and u n em ploym en t w ere favorable in the 1950s and 1960s.
The Phillips correlation held up remarkably w ell throughout the 1960s.4
Yet econom ic advisors operating within the synthesis tradition were
pessimistic about the prospects for tam ing inflation.

2.3 THE ROLE OF M ONETARY PO LICY
The practitioners o f the neoclassical synthesis saw a n eed for activist
aggregate d em an d m anagem ent. Given the degree of short-run pricelevel stickiness built into the neoclassical synthesis, m onetary policy was

3.deMenilandEnzler(
1972)report"thefirstlargeeconometricmodelsofthe1940sand
1950shadrelativelylittletodowithwagesandprices.AslateasI960,oneofthemajor
U.S.modelsdidnothavewage orpriceequations.Inthelate1950s,theauthorsof

anothermodelreportedthatforallpracticalpurposespriceandwagemovementswere
independentofrealvariablesintheirmodel.However,postwarexperiencehasfocused
attentionmore and more on theproblemofinflationand hasshown thatthereare
cruciallinksbetweenrealvariablesandpricesandwagesthatimplyatradeoffbetween
realoutputandemploymentontheonehandandinflationontheother."
4. SeeTobin's(1972,p.48)discussionofthecrueldilemma.




The New Neoclassical Synthesis •237
recognized to have potentially powerful effects. Yet, in practice, policy
advisors w orking within the synthesis viewed monetary policy as play­
ing a perm issive role in supporting fiscal policy initiatives. Moreover,
economists regarded the effect of market rates on interest-sensitive com­
ponents o f aggregate dem and as less important than direct credit effects
(Okun et al., 1969, pp. 8 5 -9 2 ). They thought monetary policy worked
primarily b y affecting the availability of financial intermediary credit,
with particular importance attached to the effect on spreads between
market rates and then-regulated deposit rates. Accordingly, there was a
reluctance to let the burden of stabilization policy fall on monetary pol­
icy, since it w orked by a distortion of sorts .5
In spite o f a reluctance to use it, practitioners of the neoclassical synthe­
sis recognized that monetary policy could control inflation. Okun's (1970,
p.

8) view

w as representative: "th e basic cure for inflation is to remove or

offset its cause: cut aggregate demand by fiscal or monetary policy suffi­
ciently so that m on ey spending will no longer exceed the value of g o o d s/'
James Tobin could say of the 1966 tightening of monetary policy to fight
sharply rising inflation that "th e burden of restraint fell almost wholly on
the Fed w hich acted vigorously and courageously ."6
Thus, m onetary policy in the neoclassical synthesis was regarded as a
powerful instrument, but one ill suited to controlling inflation or to
undertaking stabilization policy. While monetary policy could control
inflation in theory, the practical view was that inflation was mainly gov­
erned b y psychological factors and m omentum, so that monetary policy
could have only a very gradual effect. Since monetary policy created dis­
tortions across sectors, fiscal policy was better suited for controlling the
business cycle.

3. M o n eta rism and Rational Expectations
W h en it em erged in the 1960s, monetarism seemed to threaten the neo­
classical synthesis. Partly, this was because monetarists portrayed them­
selves as intellectual descendants of the pre-Keynesian quantity theory
of m oney, as articulated by Irving Fisher and others. Partly, it was be­
cause m onetarists questioned so much of synthesis doctrine, e .g ., the
effectiveness o f fiscal policy and the structural stability of the Phillips
curve. In the 1970s and 1980s, m any monetarist insights were to be

5. One particularconcernwas thatchangingcreditavailabilitywould createinstability

inthose sectorsmost dependent on financialintermediaries:smallbusinessesand
individuals.
6.Tobin(1974,p.35)
.




238 • GOODFRIEND & KING
incorporated into the broad-based synthesis, and for good reason: mone­
tarism w as a set o f principles for practical policy advice, it was commit­
ted to neoclassical reasoning, and it too identified the source of business
cycles in short-run price-level stickiness .7 H ow ever, at the same time,
Lucas's critique o f macroeconometric policy and the subsequent intro­
duction of rational expectations into m acroeconom ics led to a broader
questioning o f the neoclassical synthesis.
The quantity theory— the heart o f m onetarism — suggested organizing
m onetary analysis in terms of the su pply of nom inal m oney and the
dem an d for real m on ey balances. This focus had implications for the
m onetary transmission m echanism , for the linkage betw een inflation
and real activity, and for the role o f m onetary policy.

3.1 THE M ONETARY TRANSM ISSION M ECHANISM
The basic monetarist fram ew ork w as the quantity equation, which we
introduce using notation that w e carry throughout the paper. According
to the quantity theory, nom inal incom e (Y f) is the result of the stock of
m o n ey (M f) and its velocity (v t):

log Y t = log M t + log v r

(3.1)

M onetarists m ade the quantity theory operational b y taking money as
a u to n om ou s .8 M onetarists also constructed an econometric model on
the basis of their analytical fram ew ork. The St. Louis m odel of Anderson
and Jordan (1968) w as sim ply the quantity equation in a distributed-lag
context, with a flexible specification introduced to capture the dynamic
adjustm ent o f m o n ey dem an d and m o n ey supply.
The monetarist view o f the transm ission m echanism w as sharply at
o d d s with the neoclassical synthesis, w hich tended to view the main
channels o f transm ission as w orking through credit availability and sec­
ondly through the effect o f long-term interest rates on investment*
Monetarists regarded both of those channels as secondary. They focused
on m o n ey rather than credit channels.
Follow ing Irving Fisher, monetarists recognized that nom inal interest
rates contained a real com pon en t and a prem ium for expected inflation.
Like other lags, those in expectation form ation w ere taken to be long and
variable. A s a practical matter, th ou gh, monetarists regarded m ost of the
7. See, for instance, Friedman (1970).
S.

Fullyoperationalmonetaristanalysisalsorequiredassumptionsaboutvelodty.Insome
contextsvelocitywas assumed constant,inothers,autonomous. More sophisticated
analysesmadevelodtyafunctionofasmallsetofmacrovariables.




The New Neoclassical Synthesis •239
variation in long-term rates as reflecting inflation premia, giving long
rates a relatively minor role in the transmission of monetary policy to
real activity.

3.2 INFLATION A N D REAL ACTIVITY
Monetarists also differed in their view of the linkage between inflation
and real activity. For the m ost part, monetarists acknowledged that they
had no reliable theory to predict the short-run division of nominal in­
com e grow th betw een the price level (Pt) and real output (y,)— they had
no short-run price equation. In various ways, they interpreted the apparent
short-run nonneutrality of m oney as the result of price-level stickiness.
But they observed that the effect of monetary policy actions on the
econ om y w as long and variable. They tended to attribute that variability
to differences in the degree to which policy actions were expected, be­
cause expectations determined the degree to which prices and wages
w ould adjust to neutralize an injection of money.
These expectational considerations were made explicit by Friedman
(1968), w h o described h ow incomplete adjustment of expectations could
lead w ages and prices to respond sluggishly to changes in money. A t the
same tim e, Friedman suggested that sustained inflation should not affect
real activity in the long run, defined as a situation in which expectations
were correct, since output w ould then be determined by real forces .9
Friedman's suggestions were well timed. A s shown in Figure 2, inflation
increased sharply in the 1970s with little accompanying expansion of real
activity.10

3.3 THE ROLE OF MONETARY POLICY
M onetarists saw a dramatically different role for monetary policy as well.
Distrustful o f discretionary and activist monetary policy, they sought to
formulate sim ple fixed rules for policy. With Friedman and Schwartz's
(1963) interpretation of the Great Depression in mind, they believed that

9. ThebuildersoftheSt.Louismodelsoughttodevelopapriceequationalongtheselines
(seeAndersonandCarlson,1972),whichincorporatedthesimultaneousdetermination

ofpriceandoutputandalong-terminterestrateasameasureofexpectedinflation.
NBER turningpointsplottedasverticaldashedlines.Unemploymentandinflation
moved inverselyduringallmajorpostwarrecessions.Business-cyclecomponentsof
inflationand unemploymentarenegativelycorrelatedinastablemanneroverthe
postwarperiod.However,low-frequencytrend componentsofinflationandunemploy­
ment(cycleswithperiodicitygreaterthanthreeyears)bearrelativelylittlerelationship
toeachotherand virtuallynonetoNBER business-cycleepisodes.Notshown, the
high-frequencyirregular componentsofinflationandunemploymentarealsoessen­
tiallyunrelated,withinflationhavingmuch morevolatilityathighfrequenciesthan
unemployment.

20.Figure2 displaysU.S.inflation(thedarkline)andunemployment(thelightline),with




240 ■ GOODFR1END & KING

Figure 2 INFLATION A N D UNEM PLOYM ENT

the m onetary authority sh ou ld avoid major m onetary shocks to the
m acroeconom y, suggesting a rule in w hich the quantity o f m oney grew
at a constant rate sufficient to accom m odate trend productivity growth
(Friedman 1960). A fter arguing that sustained inflation has little effect on
real activity, Friedman (1969) described a long-run m onetary regime that
involved sustained deflation, m aking the nom inal interest rate zero and
thereby providing for an optim al quantity o f m oney.
In practice, there w ere also im portant differences in the suggested role
o f m onetary policy over the business cycle. W h ile the policy advisors of
the neoclassical synthesis sough t to have the Federal Reserve maintain
un ch an ged interest rates as fiscal policy w as varied, monetarists thought




The New Neoclassical Synthesis •241
interest-rate sm oothing contributed to fluctuations in real economic activ­
ity by m aking the m on ey stock vary procyclically.

3.4 RATION AL EXPECTATIONS
A s w as the case with monetarism, the introduction of rational expecta­
tions into macroeconomics in the early 1970s at first seemed incompati­
ble with the neoclassical synthesis. This was particularly ironic in that
John M u th motivated his rational-expectations hypothesis by suggesting
that individuals form expectations optimally, which is a natural exten­
sion o f the neoclassical principle that the economy is inhabited by ra­
tional, m axim izing agents.
The early n ew classical m odels, such as that of Sargent and Wallace
(1975), incorporated Friedman's view that perceived variations in money
led sim ply to changes in prices, with only misperceived monetary
changes having real effects .11 Coupled with rational expectations, this
strong neutrality mechanism led to very specific and controversial state­
m ents about the role of monetary policy. First, as in the monetarist
analysis, the central bank should avoid creating monetary shocks. Sec­
ond, a w ide class o f monetary rules led to the same fluctuations in real
activity, since real effects of perceived variations in money would be
neutralized b y price-level m ovem ents.

3.5 CREDIBILITY
Even though the policy-ineffectiveness result was fragile, other farreaching implications carry over to most m o d em macroeconomic models,
including the sticky-price framework that w e discuss below. Rationalexpectations reasoning teaches that the effect of a given shock cannot be
calculated w ithout understanding its persistence or the extent to which it
w as expected and prepared for in advance. This point, delivered force­
fully in Lucas (1976), revolutionized policy analysis, implying that one
cannot predict the effect o f a policy action at a point in time without
taking account of the nature of the policy regime from which it comes.
Sargent (1986) tied these ideas explicitly to the nature of the inflation
process: "inflation only seems to have a m om entum of its own. It is actu­
ally the long-term governm ent policy of persistently running large defi­
cits and creating m on ey at high rates that imparts the mom entum to the
inflation rate ."12Reviewing a series of historical episodes in which coun­
tries tried to reduce high inflation rates, he argued that the costs of
disinflation— forgone output— were much smaller if the government s

11. McCallum (1980) discusses the robustness of the policy neutrality proposition.
22. Sargent (1986, p. 41).




242 • GOODFRIEND & KING
com m itm ent to disinflation w as credible than if it w as not. Yet, ironically,
the n ew classical m acroeconom ic m o d el assigned little importance to
credibility. In that m od el, the future intentions o f the central bank are
very important for the evolution o f the price level, because they affect ex­
pected inflation, but they are o f limited relevance for real activity so long
as they are accurately perceived. Consequently, while m any central
banks view ed credibility as important, they were reluctant to use the new
classical m acroeconom ic m odel for analysis o f m onetary policy issues.

4. Real B usiness C ycles
A lth o u gh rational expectations were introduced into macroeconomics to
study the links betw een real and nom inal variables, its implications were
m ore systematically w orked out within the real-business-cycle research
program . The strong m onetary neutrality built into RBC models has
precluded their w idespread use in m acroeconom ic policy analysis to
date. But w e see RBC logic as a central part o f the N e w Neoclassical
Synthesis. O n e reason is that the RBC program constructs models in
which the alternative policies can be com pared on the basis of measures
of the utility benefits or costs, rather than on the basis o f ad hoc objec­
tives. A n oth er is that the RBC fram ew ork allows for the analysis of
policy and other shocks in the d yn am ic-stoch astic context of a fully
specified system , as called for b y rational-expectations reasoning. The
RBC program integrates and clarifies the intertemporal substitution that
is at the heart o f m acroeconom ics— involving consum ption, investment,
and labor-supply behavior— and in so doin g it clarifies the determinants
of the real rate o f interest. Finally, RBC m od els provide insights into the
nature o f cyclical nonneutralities in N N S m odels and also describe
m acroeconom ic outcom es under neutral m onetary policy.

4.1 THE CORE ELEMENTS OF RBC M ODELS
The RBC approach em p loys real general equilibrium m odels to study
m acroeconom ic ph en om en a. O n e key elem ent is the intertemporal opti­
m ization approach to consum ption and labor supply. A nother is the
similar intertem poral analysis o f investm ent and labor dem and, arising
from the profit-m axim izing decisions o f firms. Plans of households and
firms are then com bined into a general equilibrium, in which quantities
and prices are sim ultaneously determ ined.

4.2 PRODUCTIVITY SH O CKS
The RBC program focused m acroeconim ists on the procylicality of the
m easured productivity of factor inputs. In the h an ds of Prescott (1986)




The New Neoclassical Synthesis * 243
and Plosser (1989), the basic RBC model was seen to be capable of
generating business cycles that resembled those of the U.S. and other
econom ies w h en it was driven by Solow residuals. For the purpose of
defining these residuals and for discussing other issues below, we write
the production function as constant returns to scale in labor (n) and
capital (k), shifting through time as a result of productivity shocks (a):
yt =

kt).

(4pl)

In the RBC m odel, productivity shocks have two sets of effects on
output. O n e is that they mechanically raise or lower output, as stressed
by Solow in his fam ous decomposition,

(4.2)

where sk and sn are the factor shares of labor and capital. However,
productivity shocks also exert effects on macroeconomic activity, be­
cause they affect marginal product (factor demand) schedules. These
marginal (substitution) influences interact with the smoothing motiva­
tion built into households' preferences to govern the dynamic response
of the econom y. A temporary rise in current productivity, for example,
makes it m ore valuable for households to work (to cut back on leisure)
and to invest (to postpone current consumption). Within the RBC
m odel, these m echanism s explain, for example, the procylicality of labor
input and the high-am plitude response of investment. The RBC ap­
proach forces a researcher to explain the response of the macroeconomy
in terms o f substitution and wealth effects on households.
A m ajor question about the RBC approach has been the measurement
of productivity shocks, particularly whether the Solow method mismeasures factor inputs. Subsequent research has focused on variable capital
utilization as one source of mismeasurement: recent work by Burnside,
Eichenbaum , and Rebelo (1995) cuts dow n the variability of the Solow
residual so substantially that an adherent of the RBC approach may
worry that there is little left in the way of productivity shocks.

4.3 RATIONALIZING HIGH SUPPLY ELASTICITIES
By focusing attention on the supply side, RBC modelers provoked many
questions, one o f the m ost basic being: are the high-amplitude labor sup­
ply variations assum ed in RBC m odels counterfactual? Early RBC models
assum ed that aggregate labor supply varied solely by an individual




244 • GOODFRIEND & KING
worker (the representative agent) changing the num ber of hours worked.
This m echanism is arguably inconsistent w ith microeconom ic evidence
on labor supply.
Yet, over the course of the business cycle there are large changes in
w ork effort. A s illustrated b y com parison o f panels (a) and (c) of Figure 1,
these m ostly arise from changes in the num ber of em ployed individuals,
rather than in the num ber o f hours w orked by each individual. Important
modifications o f the basic RBC fram ew ork have m odeled such move­
m ents into and out of the w ork force, yielding extremely high aggregate
labor su pply elasticities w hile maintaining small micro elasticities. Other
recent studies feature variable capital utilization, with a supply of capital
services that is highly sensitive to changes in factor prices so that utiliza­
tion is strongly procyclical .13 Overall, the m odern RBC approach de­
scribes a m acroeconom y that is highly sensitive to real shocks. Hall (1991)
points out that m an y approaches to business cycles require a "highsubstitution" econ om y like that constructed by RBC researchers.

4.4 M O N EY IN RBC MODELS
Early in the RBC research program , a m onetary sector was added to
explore the types o f business-cycle correlations betw een m oney and
output that could em erge if productivity shocks were the main driving
factor (King and Plosser, 1984). A t a later stage of research, the effects of
the inflation tax were explored (C ooley and H an sen , 1989). From this
research and other w ork over the last decade, a num ber o f conclusions
have em erged that are broadly shared b y m acroeconom ists. First, en­
dogen ou s variations in m o n ey su pply arising from the joint actions of
private banks and the m onetary authority at least partly explain the
business-cycle correlation o f m o n ey and output. Second, while versions
o f RBC m odels supplem en ted with a m onetary sector can in principle
explain the correlation o f m o n ey and output, they do less w ell at explain­
ing the cyclical variation in real and nom inal interest rates (Sims, 1992),
suggesting that there is m ore to the cycle than real productivity shocks
that cause sym pathetic variations in m oney. Third, the predicted conse­
quences o f cyclical variations in expected inflation are quantitatively
small w ithin flexible-price m o d els, if m o n ey dem an d is m odeled via cash
in advance or with an explicit transactions technology. That is, for
business-cycle pu rposes, an RBC m odel with an explicit monetary mech­
anism w orks a lot like an RBC m odel with a m o n ey dem an d function just
tacked on after a real general equilibrium analysis.

23.ChoandCooley(1994)showhowheterogeneityoffixedcostsofgoingtoworkcanlead

tolargework-forceadjustmentsandsmallhoursadjustments.Theselaborsupplyand
capacityutilizationdevelopmentsarereviewedinKingandRebelo(1997).




The New Neoclassical Synthesis * 245
4.5 AN ALYSIS OF SUSTAINED INFLATION
Studies of the costs of steady inflation conducted under the RBC rubric
have led to a revised understanding of the benefits that may be obtained
from low ering inflation. A basic reference in this area is Lucas (1993),
w ho calculates that the welfare cost of a 7% inflation may be about 1% of
output using a variant of the shopping-time model of money demand.
Since Lucas's transactions technology has no satiation level of cash bal­
ances, m ost o f his estimated gains from lowering inflation to the Fried*
man (1969) level arise as a result of deflation. However, estimating the
parameters o f a shopping-tim e model with annual U.S. data over 19151992, W olm an (1996) concludes that the U.S. experience appears more
consistent with a transactions technology with a satiation level of cash
balances. This alternative m oney demand model provides roughly the
same total gain from lowering inflation, but locates most of it between
7% and zero inflation.

4.6 FISCAL POLICY A N D FISCAL SHOCKS IN AN RBC SETTING
Another important topic of RBC analysis has been the study of fiscal
policy and fiscal disturbances in real general equilibrium. In the RBC
m odel, changes in tax rates have a powerful effect on real activity. In
particular, variations in a comprehensive income or sales tax affect the
after-tax real factor returns to labor arid capital, inducing substitutions
betw een g o o d s and across time that influence the quantities of work
effort and investm ent chosen by a representative agent. For example,
the after-tax real w age is

dnt
where rf is the tax rate at date t and w t is the real wage rate at t. Thus,
from the standpoint o f the marginal return to work, the tax works just
like a productivity shock. Accordingly, changes in comprehensive in­
com e taxes exert a high-octane influence on the RBC model.
RBC studies o f actual U .S. fiscal shocks, like that of McGrattan (1994),
com e to an ironic conclusion. Changes in tax rates have powerful effects
on m acroeconom ic activity, but since the variation in measured U.S.
capital and incom e tax rates at business-cycle frequencies is small, these
shocks d o not contribute much to overall business-cycle variability. H o w ­
ever, w e see below that changes in markups can be interpreted as taxes
o f a potentially cyclically volatile form.




246 * GOODFRIEND & KING

5. N e w K eyn esia n E conom ics
The N e w Keynesian approach to macroeconom ics evolved in response
to the monetarist controversy and to fundam ental questions raised by
Lucas's critique, and in order to provide an alternative to the competitive
flexible-price fram ew ork of RBC analysis. O ur discussion of this wideranging research program will be divided into three parts. W e first re­
view early w ork by G ordon (1982) and Taylor (1980). W e then discuss
m ore recent n ew K eynesian m icroeconom ic foundations, which high­
light m onopolistic com petition and costly price adjustm ent. Finally, we
focus on optim izing price adjustm ent in a dynam ic setting.

5.1 FIRST-GENERATION NEW KEYN ESIAN MODELS
In first generation N e w Keynesian m o d els, G ordon (1982) and Taylor
(1980) m odernized the specification o f the w age-price block to incorpo­
rate monetarist and rational-expectations insights.
5 .1 .1 G ordon's Price Equation

O n the empirical side, G ordon (1982) esti­

m ated price dynam ics u sin g a m onetarist proxim ate exogen eity of nominal
aggregate dem an d. A bstaining from separate consideration of nominal
w ages because he view ed their dynam ics as essentially identical to those
o f prices, G ordon estim ated price equations o f the form
tTt = A(L)tt,_1 + G(log Y t -

log Y f_j) + ps, +

(5.1)

w here 1rt = log P t — log P t-1 is the rate o f inflation, A(L) is a polynomial in
the lag operator, log Y t - log Y t^ is nom inal incom e grow th, ps, captures
the effects o f observable price shocks, and % is an error term.
G ordon interpreted the A(L) coefficients as indicating h ow the price
level gradually adjusts tow ard a lon g-run level required by nominal
incom e and a "natural rate" level o f real activity. There were three main
findings o f G o rd o n 's investigation: First, there w a s a numerically small
value o f G in the price equation. Estimating quarterly price equations
over nearly a century of data and several subsam ples, Gordon found
slope coefficients in the range o f G =

0. 10,

indicating a small impact

effect o f output on prices equal to G /(l + G) = 0 .0 9 .14 Second, lags were
estimated to be very important in the price equation: the mean lag be­
tween output and prices w as m ore than a year. G ordon interpreted this
as evidence for gradual adjustm ent o f the price level to changes in nomi­
nal expenditure.
14. The impact effect is interpreted using the identity log Yt - log

(logy, - logyIM).




= log P, - log pt-i +

The New Neoclassical Synthesis * 247
H ow ever, G ordon also found remarkable changes in his estimates
when the ninety years of data was split into three or more subperiods.
W ithin the early subsample running from 1892 to 1929, there were major
shifts in the effects of nominal income during the war period 1915-1922.
In particular, the estimated coefficient on nominal income rose substan­
tially, w ith a big difference arising between expected nominal income
growth (G = 0.47) and unexpected nominal income growth (G = 0.25).
M easures o f supply shocks— notably energy and commodity prices—
became increasingly important in the post-World War II sample period.
Finally, the sum of coefficients on lagged inflation, A(l), rose substan­
tially from 0 .4 during 1892-1929 to more than 1 during 1954-1980.

5 .1 .2 Taylor's Rational-Expectations Approach to Wage Setting

The most

hardy o f the first generation of N ew Keynesian rational-expectations
m acroeconomic m odels is that of Taylor (1980). In m odem terminology,
Taylor's vision w as that the firm and its workers set a fixed wage over
the life o f a J-period contract. W age bargains were assumed to be stag­
gered through time with 1// of the contracts set each period. The sim­
plest mathematical representation of Taylor's wage-setting mechanism is
as fo llow s .15 The nom inal wage rate set at date t, log YJ*t, depends on the
average price level expected over the contract, (1//) Sjlg Et log PHj; on the
average labor-market tightness [incorporated as (hlj)

Etet+j, where e, is

the labor-market tightness at date t and h governs the wage response to
this tightness]; and on a wage shock (yt):

log wr = - ^

e, log pl+j+

JM

- § E‘eM +

(5-2)

j 7=0

Taylor (1980) adopted a very simple macroeconomic model to focus on
the consequences of this wage-setting behavior. First, Taylor specified
that the price level was a simple average of wages, motivated by refer­
ence to a m onopolist with constant marginal cost selecting a fixed
m arkup,

35.Taylor(1980)assumedthatcurrentwagesdependedonpastandfuturewages:
M
log Wf - 2

/-l
bi loS WT-y + 2

/=o

h /"*
¥ * IoS ^

j=o

+ 7E

£^

+

>M

With the contract weights being b. = (1/(1 - /)) (1 - W

obtainedbysubstituting(5.3)into(5.2).




This is a reduced form

248 • GOODFRIEND & KING

io g P f = 7 § i o g w ^

(5.3)

Jm
Second, like G ordon (1982), Taylor m ade the monetarist assumption that
nom inal expenditure w as determ ined b y a quantity equation. Third,
Taylor assum ed that labor-market tightness related to output: et = g x log
y t. Fourth, Taylor assum ed an activist m o n ey stock rule for monetary
policy, specifically that

log M t = g 2 log Pf.

(5-4)

A s with the earlier N e w Classical rational-expectations m odels, Taylor/s
rational-expectations m odel required specification o f the monetary au­
thority's behavior. Rather than taking the m onetary authority to be a
source o f business-cycle im pulses, he view ed it as adjusting the money
stock to the price level with a response coefficient g2.

5 .2 .3 Business-C ycle and Policy Implications o f Taylor's Framework

There are

four implications o f the Taylor fram ew ork. First, Taylor produced a
"h u m p e d -sh a p e d " pattern o f cyclical output (unem ploym ent) dynamics
in response to w age shocks vt, w hich Taylor suggested w as a measure of
success, because a num ber o f empirical researchers had estimated timeseries m odels w hich im plied such profiles. Second, Taylor dem onstrated
that the policy rule mattered for the evolution of real activity. Third, Taylor
highlighted a n ew m onetary policy trade-off betw een the variability of
output and the variability o f inflation w ithin his m o d el, even with the
maintained assum ption that there w as n o long-run trade-off between the
rate of inflation and the level of output. If velocity shocks were small, for
exam ple, then a central bank could largely eliminate real variability by
accom m odating price-level m ovem en ts (g2 close to one), but this would
require greater variability in the price level. Fourth, he sh ow ed that ra­
tional expectations mattered a great deal— for the response of the econ­
om y to shocks and for the design o f m onetary policy rules— by contrast­
ing his results with those based on extrapolative expectations.
Importantly, sticky-w age and sticky-price rational-expectations mod­
els like Taylor's also explained the m ain findings of G ordon , at least in
broad form . Lags o f nom inal w ages and prices w ere important state
variables in these m odels, reflecting gradual adjustm ent to real and nomi­
nal shocks.

Moreover the effects of proximately exogenous variations in

nominal incom e depen ded in a central m anner o n h o w persistent these




The New Neoclassical Synthesis * 249
were expected to be, since (5.2) indicated that price expectations played
a m ajor role in w age setting.
Taken together with his subsequent work on larger macroeconomet­
ric m odels incorporating gradual price adjustment, Taylor's theoretical
m odel had a major intellectual impact. Yet, at the same time, there was
an uneasiness about the staggered wage models of Taylor. In the United
States, in particular, only a small portion of the labor force was subject
to explicit multiperiod contracts. Further, the microeconomic underpin­
nings o f the wage-setting process were sketchy.16

5.2 SECOND-GENERATION NEW KEYNESIAN MODELS
In the next stage of research, N ew Keynesian economists shifted the
location o f nom inal stickiness from wages to prices.17 In this new work,
price-setting firms were explicitly m odeled as monopolistic competitors.
The imperfect-competition framework was used to explain the real out­
put effect o f m on ey w hen prices were subject to costs of adjustment, to
develop various amplification mechanisms, and to highlight the poten­
tial social costs o f business cycles.

5 .2 .1 E xplicit M onopolistic Competition M odels

During the 1980s, implica­

tions o f m onopolistic competition were explored in a wide range of fields,
including econom ic growth, international trade and finance, and macro­
econom ics. In each case, imperfect competition held the promise of under­
standing issues that were puzzling from the perspective of competitive
theory. In macroeconomics, monopolistic competition was important for
analyzing h o w firms set prices. In the standard competitive setting, firms
take market prices as given and adjust quantity in response to variations in
prices and costs. By contrast, in Blanchard and Kiyotaki (1987) and
Rotem berg (1987), firms are monopolistic competitors and set prices in or­
der to m axim ize profit. These studies take consumption to be an aggregate
o f a continuum o f differentia ted products, ct = L^1cf(z)1_ 1/*dz]u 1)/€. A nindividual firm producing the product z faces the constant elasticity demand

(5.5)

17. These NewKeynesiandevelopmentsareencapsulatedinMankiwandRomer(1991)and
surveyedinMankiw(1990),Romer(1993),andRotemberg(1987).Inpart,New Key-

nesianeconomistssoughttoavoidtheoreticalcriticismsofwagecontractingmo es.n
part,theythoughtthatpricestickinessseemedpervasiveandsticky-pncemodi
consistentwiththesomewhatprocyclicalrealwagesfoundinthedata(Mankiw,1990).




250 • GOODFRIEND & KING
which is shifted by the aggregate price level and the level o f aggregate
consumption dem and. Investment and governm ent purchases could be
view ed similarly as aggregates of differentiated products, leading to a
version o f (5.5) that replaced c, with an aggregate dem an d measure. The
implied form o f the (perfect) price index associated w ith aggregate expen­
diture is

1

a

\l/(l-e)
(5.6)

P t( z ) ^ ed z j

Accordingly, with a nominal marginal cost o f % , an optimizing firm
w ould set its price at a constant m arkup over marginal cost, Pt(z) =
[e/(e — 1)]

with the markup being given by the conventional formula.

Thus, monopolistic competition rationalizes a firm setting a price and
setting it at a level greater than marginal cost. Imperfect competition
does not, by itself, rationalize nominal stickiness.
5.2.2 Incorporation o f Nominal Stickiness

A t the m icroeconom ic level,

stickiness of nominal prices is a feature o f our everyday life. Thus, if we
are developing "m icro foundations for m a cro econ o m ics/' it is important
to have m odels that can explain these observed pricing practices. The
most direct explanation is that small real costs o f changing nominal
prices— m enu costs— account for sticky prices. It is an open question as
to whether small m enu costs can lead to sustained stickiness of the
prices of individual goods, particularly in a situation o f positive inflation.
For the m ost part, in the N ew Keynesian m odeling approach, the dis­
crete and occasional adjustment of individual prices is sim ply a feature
of the environment, rationalized in m ore or less elaborate w ays. In this
paper, as in that literature, w e focus less on w h y individual prices might
be set in advance and more on the implications that discrete and occa­
sional individual price adjustment has for the behavior of the aggregate
price level and real economic activity.
5 .2.3 The Causes and Consequences o f M on eta ry Business Cycles

New

Keynesian economists also have stressed that im perfect competition is
important for the effect of m oney on output if there is nom inal price
stickiness. To see the pow er of this argum ent, think about the perfectcompetition case. If dem and rises, but price rem ains the sam e, the firm
will not respond, routing its potential custom ers elsew here. By contrast,
if its price is fixed at a level that exceeds marginal cost, then it is desirable
for an individual firm to expand its output if its dem an d rises. The easy
case is if marginal cost is unrelated to the firm 's output, for then it will
absorb all o f the dem and variation w ithout suffering a decline in its




The NewNeoclassical Synthesis * 251
markup. Even if marginal cost rises with output, either at the level of the
firm or in general equilibrium, then it will continue to be profitable to
satisfy dem an d so long as price exceeds marginal cost. In response to a
general econom ic expansion— a rise in ct in (5. 5) above— it is plausible
that m arginal cost increases because firms must pay higher real wages to
secure the labor input to produce additional output. Accordingly, N ew
Keynesians highlight the importance of procyclical movements in real
wages and m arginal cost.
A s a related matter, N ew Keynesian analysis also suggests a new set of
conclusions for welfare analysis of the business cycle. With monopolistic
competition, market power of firms means that there is too low a level of
em ploym ent and output on average. The New Keynesian analysis thus
provides a coherent account of the temptation to expand the economy
present in the literature on time-inconsistent monetary policies (Barro
and G o rd o n , 1983 ).18 Further, monetary policymakers should not be
indifferent about short-run changes in employment that arise from
changes in m o n ey w h en prices are sticky. Notably, a decrease in employ*
ment and outpu t that results from a contractionary monetary policy
lowers the w elfare o f the representative individual by increasing m o­
nopoly distortions.
5 .2.4 The O rigins and Implications o f M onopolistic Competition

There are a

range o f econom ic mechanisms, of course, that are consistent with mo*
nopolistic com petition. To us, the most plausible is that firms face im­
portant fixed costs, including general overhead costs. These suggest
m odifying the production function to
y* =

kt) -

tf>],

(S*7)

where <P is a m easure o f fixed costs, which plausibly are assumed to
display the sam e factor intensity requirements and technical shifts which
govern final output. W ith such a production function, the representative
firm has constant marginal cost (at given factor prices) and diminishing
average cost.
Hall (1988) dem onstrates that the modified Solow decomposition is

y>

-(1

dn,

dk,\

+ 4 > ) [ s n — + sk — \ +
V
nt
k, J

da,
a,

(5-8)
v

,

18. Ireland(1996b)providesafullyarticulatedmodelof

h ° w

im p e r fe c t

stickypriceslead toexcessiveinflationwhen themonetary
commititsfutureactions.




c^petitionand
ty

252 • GOODFRIEND & KING
w here sn and sk are total cost shares and

0

is the ratio of overhead to

variable cost. This decom position highlights the consequences of over*
head costs. First, the standard Solow residual varies with the business
cycle even if there are no productivity shocks. Second, there is an amplifi­
cation m echanism , so that a one-percent change in labor changes output
by (1 4- <[>) sn percent.
The N e w K eynesian approach allows for a w ide range of assumptions
about the nature and extent o f imperfect com petition. If there are no
pure m o n op o ly profits, then the m arkup o f price over marginal cost
m ust sim ply cover overhead costs, i.e ., w e m u st have fi =

1

+ <j> on

average, w hich w e assum e throughout our discussion. In various quanti­
tative exercises below, it will also be necessary for us to take a stand on
the value o f the steady-state markup. C om pared to som e other recent
studies, w e take a small value, fi =

1 . 1,

w hich corresponds to a

"n e t " m arkup and a dem an d elasticity o f about l l

.19 W e

10%

do this for two

reasons. First, it is broadly consistent w ith observed markups in the
construction and autom obile service industries, i .e ., m arkups in the
range o f 7% to 15% in contracts and bills o f sale. Second, it is consistent
with the detailed empirical studies o f Basu and F em ald (1997).

5.3 D YN A M IC PRICE-SETTING MODELS
M odels o f price dynam ics based on fixed real costs o f changing nominal
prices were first developed in the early 1970s. In these m odels, firms
choose the tim ing and m agnitude o f their price adjustm ents in response
to the state o f the econom y, including the average rate o f inflation and
the stage o f the business cycle. This state-dependent approach to pricing is
attractive from a m icroeconom ic perspective because ( 1) individual firms
are observed to discretely adjust their prices at infrequent intervals of
apparently stochastic length, and (2) firms are m ore likely to adjust price
w h en there are large shocks to their markets or sustained inflation.
H ow ever, it has proved difficult to introduce this form o f price adjust­
m ent into com plete m acroeconom ic m odels. Caplin and Leahy (1991)
indicated that the consequences could be major, but also that many
simplifications were necessary to characterize the imperfectly competi­
tive equilibrium w ith costly price adjustm ent, including extreme restric­
tions on the rules of the central bank, on the behavior o f consum ers, and
on the nature o f m o n ey dem and. Th u s, w hile state-dependent pricing is
natural, existing m od els have been ill suited for empirical analysis or
exam ination o f alternative m onetary policy rules. For this reason, the

19. Usingn

)asinthetextabove,,/i= 1.1correspondstoe= 11.

= e/(e — 1




The NewNeoclassical Synthesis •253
em phasis in N e w Keynesian literature has been on time-dependent price
adjustm ent rules which specify that firms have exogenous opportunities
for price adjustm ent.

5 .3 .1 A n Intertem poral Approach to Price Setting

Following Calvo (1983),

we consider h o w a rational firm would select its price today given that it
will have to keep it fixed for an interval of stochastic length. To opera­
tionalize this idea, w e use notation and structure from a recent study of
time and state-dependent pricing .20 A s in the imperfect-competition
m odel above, w e can posit a large number of firms— technically a contin­
uum of firm s— and suppose that a fraction
periods ago, for j = 0, 1,

last adjusted their price j

1 , Accordingly,

the date-f conditional

probability o f the next adjustment at date t + j is o)j t +/a>0,. When the
dem and elasticity is assum ed constant, so thaty,(z) = [Pt(z)IPlY *d t with Pt
being the perfect price index and dt an aggregate demand construct, then
the optimal price is restricted by

'

e -1

Ef 2l~4 m

+/A ,)

dt+j) '

where ^ t+j is nom inal marginal cost at t + j and p jA nj/At is the discount
factor for date-f + ; contingent cash flow s .21 The general price adjustment
rule (5.9) derives from an equating o f marginal revenue and marginal cost
in a dynam ic setting and has a convenient approximate form that we use
below .22In particular, w hen the inflation rate is close to zero, then log P f is
approxim ately log (e/(l - e)) + [l/'Z}hZ lP h<*>h]

% +j. That is, the

price is a discounted distributed lead of expected nominal marginal cost,
with the w eights related to the frequency distribution of price adjustment
dates. Equivalently, denoting real marginal cost as tftt and using three

20.Dotsey,King,andWolman (
1996)
.TheapproachthereisageneralizationofCalvo's
(1983)approachtopricesetting.
22.Althoughwe willfocusontime-dependentpricinginourdiscussionbelow,therei
s

somerecentworkthathassoughttomakethetimingofpriceadjustmentendogenous
withinaframeworklikethatjustdiscussed(Dotsey,King,andWolman,1996).There
arethreegeneralimplicationsofthislineofresearch.First,theadjustmentprobabilities
*51 t varythroughtimewiththestatevariablesofthemodel,butwestillobtain
(5.9).Moreover,theapproximation(5.10)isrobusttostatedependence,solongasthe
inflationrateisclosetozero.Second,themodelmustallowfortimevariationsinthe
resourcesusedinpriceadjustment.However,sincethelevelsoftheseresourcesare
assumed tobesmallinmostNew Keynesianmodels,thedirectresourceeffectsof
thesearelikelytobeminor.Third,therearetime-varyingfractionsofthefirmswhich
lastadjustedtheirpricesj = 1, 2, . . . ,J periodsago.
22.TheseapproximationsarederivedinDotsey,King,andWolman(1996)
.




254 • GOODFRIEND & KING
identities (log % = log P t + log ijjt, log i(/t — - l o g

and log jx = log (e/(l -

e)), w e can express the optimal price as

1
-----r vt,1ru -

(5.10)

i.e ., as dep en d in g on the future path o f the price level and on the
deviation o f real marginal cost from its steady-state level.
5 .3 .2 The Price Level

To com plete the dynam ic pricing m odel, we need

an equation that aggregates prices across firms into the general price
level. W ith all firms that adjust at date t choosing P f, the perfect price
aggregator is

(5.11)

so that the price level depen ds on pricing decisions and adjustment pat­
terns. If variation in the adjustm ent patterns is small over the business
cycle— as in tim e-depen den t m od els or som e state-dependent models—
and the inflation rate is low, then there is a comparable approximation,

1-1
(5.12)

M
w hich w e can pair with (5.10). These tw o equations (5.10) and (5.12) are
a convenient representation o f the central "price b lo ck " o f the N N S
m odels that w e describe in the next section.
5 .3 .3 Com parison w ith Taylor's D yn am ic S ystem

Based on intertemporal

optim ization and three simplifications (low inflation, constant elasticity
o f d em an d , and small variations in adjustm ent patterns), w e have ob­
tained a pair o f loglinear equations (5. 10) and (5. 12) describing price
dynam ics. These broadly resem ble the forw ard-looking wage-setting
and backw ard-looking price-level equations used by Taylor, but with
additional flexibility in the distributed lead and lag m echanism s because
of the use o f a stochastic adjustm ent m odel.
There is, how ever, one notable om ission: there are n o price shocks in
our pair o f behavioral expressions. This is a co m m o n outcom e in eco­
nom ic m odelin g: optim ization theory leads one to view shocks as arising




The NewNeoclassical Synthesis •255
from m ore primitive events which affect economic decision makers. A s
w e shall see, our optimization approach allows for many types of events
that are typically described as price shocks [as, for example, the comm od­
ity price variations included by Gordon (1982) in his empirical specifica­
tion], H ow ever, these exert an influence on prices through marginal
cost, rather than directly, according to the theory developed in the next
section.

6 . The N e w Synthesis: Description and Mechanics
The N e w Neoclassical Synthesis is defined by two central elements.
Building o n n ew classical macroeconomics and RBC analysis, it incorpo­
rates intertemporal optimization and rational expectations into dynamic
m acroeconom ic m odels. Building on N ew Keynesian economics, it in­
corporates imperfect competition and costly price adjustment. Like the
RBC program , it seeks to develop quantitative models of economic
fluctuations.
The N N S is currently displayed in three distinct modelling scales.
First, there are small analytical models that can be used to study a range
of theoretical and empirical issues while retaining sufficient tractability
that they can be solved by hand. Second, there are medium-scale
macroeconom ic m odels analogous to those developed by RBC research­
ers that are being used to address a wide range of positive and norma­
tive issu es .23 Third, there is the new FRB/US large-scale model of the
A m erican econ om y developed over the last few years, which is now the
principal m odel em ployed for policy evaluation by the Federal Reserve
Board .24
W e call the n ew style of macroeconomics research the New Neoclassi­
cal Synthesis because it inherits the spirit of the old synthesis discussed
in Section

2. N N S

m odels offer policy advice based on the idea that price

stickiness implies that aggregate demand is a key determinant of real
econom ic activity in the short run. N N S models imply that monetary
policy exerts a pow erful influence on real activity. This has both positive
and norm ative implications. From a positive point of view, the central

23.A recentpartialsurveyiscontainedinNelson(1997)
.
24.Braytonet al. (1996)provideadescriptionofthenewFRB-USmodel,whichincorpo­

ratesrationalexpectationsanddynamicspecificationsintoconsumption,investment,
prices,andwages. Thenewmodeldisplaysnolong-runtrade-offbetweeninflation
andrealactivity.Expectationsarecentraltothedynamicconsequencesofmonetary
and fiscalactions. While the FRB-US model does not relyascompletely on in­
tertemporaloptimizationassomesmalleracademicmodelsandcontainsadifferent
processofwage determination,itneverthelesssharesmanyothercentralstructural
featuresoftheNNSapproach.




256 •GOODFRIEND & KING
conclusion is that economic fluctuations cannot be interpreted or under­
stood independently o f monetary policy. This is true notwithstanding
the fact that the RBC m odel at the core of the N N S assigns a potentially
large role to productivity, fiscal policy, or relative price shocks. From a
norm ative perspective the N N S says that aggregate dem and must be
m anaged by m onetary policy in order to deliver efficient macroeconomic
outcom es. In other w ords, the N N S creates an urgent dem and for mone­
tary policy advice.
The N e w Neoclassical Synthesis also supplies that advice. The combi­
nation o f rational forward-looking price setting, m onopolistic competi­
tion, and RBC com ponents in the N N S provides guidance for monetaiy
policy based on the follow ing reasoning. First of all, stationary monetary
policy m u st respect the RBC determinants of real econom ic activity on
average over time. That is, even though output m ay be demanddetermined on a period-by-period basis in the N N S , output must be
supply-determ ined on average. Second, the N N S locates the transmis­
sion of m onetary policy to real activity in its influence on the ratio of the
average firm 's price to marginal cost o f production, w hich w e call the
average m arkup. A m onetary policy action which raises aggregate de­
mand raises marginal cost and lowers the average m arkup. This lower
average markup sustains the increase in output and em ploym ent, be­
cause it w orks like a tax reduction in an RBC setting. Third, there is little
long-run trade-off betw een inflation and real activity at low inflation
rates. Illustrating this point, w e sh ow w ithin a Tayloresque version of
optimal pricing— one in w hich the typical firm adjusts its price once per
year— that the steady-state markup tax is m inim ized by monetary policy
that pursues near-zero inflation. T h u s, the recom m endation is that
monetary policy should stabilize the path o f the price level in order to
keep output at its potential. This policy is "a ctiv ist" in that the authority
m ust m anage aggregate dem an d to accom m odate any supply-side distur­
bances to output.
The p o w er of the n ew synthesis lies in the com plem entarity of its New
Keynesian and RBC com p on en ts, w hich are compatible because of their
shared reliance on m icroeconom ics. The N e w Synthesis allows knowl­
edge gained from N e w Keynesian and RBC studies to be brought to bear
on business-qrcle and m onetary policy questions in a single coherent
m odel. In doing so, the n ew synthesis strengthens our understanding of
econom ic fluctuations. This and subsequent sections elaborate on the
key features and implications o f N N S m o d els. The balance of this section
covers so m e preliminaries— the basic m echanics o f m arkups, the average m arkup as a tax on econom ic activity, relative prices as productivity
shocks, and the pow er and limitations o f m onetary policy. N N S princi-




The NewNeoclassical Synthesis * 257
pies and practical guidelines for monetary policy are developed in Sec­
tions 7 and 8 respectively.

6.1 H O W MONETARY POLICY AFFECTS THE REAL ECONOMY
In the n ew synthesis, monetary policy has effects which resemble those
of productivity and fiscal shocks, producing substitution and wealth
effects o n the econom y as in RBC models. Variations in the average
markup charged by firms affect marginal returns to factors in a way that
is similar to productivity shocks or changes in comprehensive taxes;
changes in relative prices across firms work like the level effects of pro­
ductivity shocks or changes in government purchases.
6.1.1 M arginal and A verage Markups

Two measures of the markup play a

major role in m odels of the N N S .25 A s suggested above, the average
markup o f price over marginal cost plays a prominent role in the transmis­
sion o f m onetary policy. A t any point in time, though, only a subset of
firms are adjusting prices and setting a new markup level, which we call
the marginal m arkup.26
Formally, the marginal markup is the ratio of price to marginal cost for
firms that are adjusting their price in period f, i.e.,

W e k n o w from Section

5 that P* depends on the expectations that adjust­

ing firms have about future economic conditions, including the price
level and m arginal cost. The average markup is the ratio of price to
marginal cost for the average firm in the economy (the ratio of the price
level to m arginal cost ),27

m

= Zl

(6.2)

For analyzing th e determination of real economic activity within period
1 is

t, it is the average markup that is central. From this stan poin ,

25. Rotemberg and Woodford (
1991)provide a surveyofaltenative
determination and some suggestive empirical evidence
a simpler model with
26. The terminology of average and marginal markup » used m a simpler model w,m
Calvo-style price setting by King and Wolman (19?
• i mst is the same
27. Capital is assumed to b l perfectly mobile among firms, so the marginal cost is
for all firms in equilibrium.




258 • GOODFRIEND & KING
important to stress that the average m arkup is just the reciprocal of real
marginal cost,
W

1

pt

ft

(6.3)

Th u s, procyclical variation o f real marginal cost— w hich m any econo­
mists find realistic— directly im plies a countercyclical average markup.
The average and marginal m arkups can m ove very differently from
each other in response to shocks. In response to sustained increases in
nom inal aggregate dem an d, for exam ple, the m arkup falls for firms not
adjusting price, but the higher inflation motivates adjusting firms to
choose a higher m arkup. Thus the short-run effect o f a sustained in­
crease in dem an d is that the marginal m arkup rises and the average
markup falls. W e will return to this point in Section 7.
6 .2 .2 The A verage M arkup as a D istortin g Tax

Firms produce output with

capital and labor services. Since they are m onopolistically competitive,
their factor dem an ds are based on cost m inim ization at a demanddeterm ined output level. A necessary condition for cost minimization is
that the value marginal product of every factor is equated to its rental
price. U sin g Wt to denote nom inal m arginal cost as above and letting
W t be the nom inal w age rate, the efficiency condition for labor is W, =
tyflt dF(nt, kt)/dnt, and there is a com parable condition for capital services.
D ividing each side o f this expression by the price level, the real wage is
equated to real marginal cost tim es the m arginal product of labor.
dF(nt, kt)

=

dnt

1
fj,t

dF(nt, kt)

„

------------,

(6-4)

dnt

w here the last equality follow s directly from the fact that the average
markup and real marginal cost are reciprocals. A gain , a similar equality
of real factor prices and real value marginal products holds for capital
services.
Thus, variations in the average markup w ork just like a comprehen­
sive tax w hich a firm m ust pay on factor inputs. In the case of labor
d em an d, for exam ple, the average markup drives a w edge between the
real w age and the marginal product of labor, just as the tax w edge did in
(4.3). A higher m arkup raises the implicit tax on labor and capital.
6 .2 .3 R elative Price D ispersion as a P rodu ctivity Shift

In addition to the

average m arkup, there is a second important source o f distortions inher­




The NewNeoclassical Synthesis •259
ent in N N S m odels. Since som e individual prices are sticky, changes in
the general price level bring about changes in relative prices. This disper­
sion o f relative prices results in a misallocation of aggregate output
across alternative uses of final goods. To exposit this misallocation, we
define aggregate output as the simple sum 28

y<" [

J

y fe )d z .
0

Suppose further, as in (5.5) above, that demand is given by the constantelasticity specification, y t{z) = [Pt(z)/PtY % , with dt being the level of bene­
fit derived in final (consumption or investment) use. Then the distribu­
tion o f relative prices influences the extent of end-use benefit from final
output:

dt = ----------- ^ -----------

fo [P t(z)/Pt] -'d z ’
The norm ative consequences of variations in this composite measure of
relative prices are analogous to those of a total-factor-augmenting pro­
ductivity shock.

6.2 THE TRANSMISSION MECHANISM
The N e w Neoclassical Synthesis provides two complementary ways of
thinking about the transmission of monetary policy actions to real eco­
nom ic activity, which w e view as the aggregate-demand and markup-tax
approaches.
6.2 .1

A gg rega te Dem and

From a traditional perspective, changes in the

quantity o f m on ey alter aggregate dem and, which calls forth changes in
aggregate supply. W h en N N S models are interpreted in this manner—
taking real aggregate dem and as determined by monetary policy— the
results are sensible at each point in time. Yet, this interpretation is incom­
plete for tw o reasons: the price level m ay respond to monetary policy
within the period, and the focus is shifted away from real marginal cost,
which is an important element of N N S m odels .29

28, Thisdefinitionisconsistentwithourdiscussionaboveand drawson Yun's(1996)

work,whichshowsthatitisconsistentwithcompetitivefactormarketsanddemanddeterminedoutput.
29. The evolutionofrealmarginalcostovertimeiscentraltodynamicpricingmodels.
Generally,changesinrealmarginalcostare
=(s„dwjwt +skdzjz) -dajat, where
z i
stherentalpriceofcapital.Thus,smallresponsesofwagesandrentalpricesto




260 • GOODFRIEND & KING
6 .2 .2 The M arkup Tax

A n alternative view o f the m onetary transmission

m echanism is suggested by the idea that the m arkup can be interpreted
as a tax and, in particular, as a change in a generalized output (sales) tax
that affects the rewards to capital and labor. From an RBC perspective the
influence o f m onetary policy on econom ic activity can be analyzed using
the relatively w ell-understood effects o f com prehensive tax changes on
m acroeconom ic activity, w hich w e review ed in Section 4 above. This
view places m ovem en ts in the average m arkup and real marginal cost at
the center o f the m echanism by w hich m onetary policy influences real
econom ic activity. It is similarly incom plete, how ever, in that it does not
incorporate the influence o f the price level on the average markup, nor
does it recognize the role o f real marginal cost in the evolution of prices.
Yet, the average m arkup remains a useful sum m ary statistic for monetary
transmission.

6.3 THE POWER A N D LIMITATIONS OF M ONETARY PO LICY
Like its nam esake predecessor, the N e w Neoclassical Synthesis views
m onetary policy as having the potential to exert a m ajor influence on
econom ic activity, though w ithin clearly defined limits. Moreover, that
influence can likewise be u n derstood to operate via distortions, albeit
different ones than identified in the original synthesis o f the 1960s.

6.3.1 W hat M on eta ry Policy Can D o

To illustrate the pow er o f monetary

policy, it is useful to study the sim plest possible price-setting m odel, one
with tw o-period staggered price-setting. In this setting, it m ight be sup­
posed that m onetary policy has limited p ow er for influencing real activ­
ity because pricing decisions are m ade just one period in advance, but
w e will see that m onetary policy is still very pow erful. In the two-period
setting w ith a>0 = ^

=

the approxim ate equation for the price level

(5.12) is log P t = ^(log Pf + log P*_!). The forw ard-looking price-setting
equation (5.10) is

log p* = r r ^ i o g p, + log^A/o + p e ,p i+, + p log Et(<pl+m

changesinoutput,assuggestedbytheU.S.aggregatedataandbuiltintoRBCmodels,
implysmallresponsesofmarginalcost.Morespecifically,itisnecessarytolookbehind
theprecedingcostdecompositiontofactor-marketequilibriuminordertodetermine
theresponsivenessofmarginalcostandtogainamorecompleteunderstandingofthe
evolutionofrealactivityandthepricelevelovertime.




The NewNeoclassical Synthesis •261
C om bining the equations in this two-period price block, we can express
the price level as

log Pt = logPr_1 + log (</vty)

+ 22

p

1 log Et (<w *w •

(6.5)

This rational-expectations solution for the price level displays two
important features that carry over to longer-horizon pricing models.
First, the price level is partly predetermined by prices set in the past.
Second, prices set by currently adjusting firms depend on current and
future real marginal cost. In fact, the price level depends on an infinite
distributed lead of expected real marginal cost even though each firm
m ust keep its price fixed for only two periods. Expectations of future
real marginal cost matter for current pricing because each firm knows
that it will keep its price fixed for some period of time. Moreover, each
firm cares about what prices will be next period in setting its price
today, and so it cares what prices firms will set next period, and so on
into the future.
In order to think about the evolution of the price level and output in this
sim ple N N S m odel w e need to understand the behavior of real marginal
cost. To do so, recall once more that real marginal cost is just the reciprocal
of the average markup, so w e can write log (fx jp ) = - l o g

The RBC

analysis above indicated that variations in the markup tax can exert a
pow erful inverse effect on em ploym ent and output. Such effects can be
com plex in a fully dynamic RBC setting, but for heuristic purposes con­
sider the sim ple inverse relationship

fit
log — = -< p {log y t -

__
log y t),

(6.6)

where y t is the flexible price level of output, i.e., that obtained in a
noncom petitive RBC m odel with a constant markup /x. Since real mar­
ginal cost is, in turn, given by l o g ( t /^ ) - <p(log 1h “ log f t ) ' the Parame*
ter (p is the elasticity o f real marginal cost with respect to an "output
g a p ."
N o w su ppose that monetary equilibrium is given by a quantity equa­
tion such as (3.1): log M t = log Pt + log y t - log vt, where vt is the velocity
process. Substituting for the price level and output in the quantity equa-




262 • GOODFRIEND & KING
tion w ith the price level and markup expressions, w e arrive at an expres­
sion relating the m o n ey stock to current and expected future markups:
log (M ,) = log (P*_,) + log y, + log (v,)
----------- -- log OVm) - 2 J ) j8> log Et (n ,+j/n) .

(6.7)

Thus as lon g as the m onetary authority follow s a policy that supports a
determinate distributed lead o f expected markups (a relatively weak
condition), the preceding expression indicates that it can choose the
m o n ey stock to produce an arbitrary pattern of small variations in the
average m arkup over time. The m onetary authority w ou ld have similar
leverage over the path o f the m arkup, real marginal cost, and output in
m ore general N N S m odels as w ell.
O n e w a y to sum m arize this p ow er is that the m onetary authority can
choose an arbitrary stationary stochastic process for the markup tax
relative to a m ean /x.30 H ow ever, the m onetary authority can produce
variations in the average m arkup only by accepting the implications for
prices and m oney. In particular, m arkup stabilization and price-level
stabilization are intimately related in N N S m o d els, a point w e shall
return to w h en w e discuss the role o f m onetary policy.
6 .3 .2 W hat M on eta ry Polio/ C annot D o

H ow ever, the analogy to taxation

is incom plete. A lth o u g h the m onetary authority can choose h ow the
m arkup tax m oves through tim e, there is little that it can do to affect the
steady-state level o f the m arkup, because the N N S incorporates forwardlooking price setting. A s w e discuss in the next section, at low inflation
rates the level o f the steady-state m arkup is nearly invariant to the infla­
tion rate and so is essentially determ ined by the extent of monopoly
pow er in the private sector. In addition, there are som e restrictions across
the short run and lon g run, as in any rational-expectations m odel. The
m ore persistent the m onetary authority's planned m ovem en ts in the
m arkup tax, the larger are their inflationary consequences.

7. G u id in g P rin cip les fo r M o n e ta r y P olicy
The N e w Neoclassical Synthesis m akes the strong recom m endation that a
central bank sh ou ld target near-zero inflation. In this section w e spell out

30. Whenwestatethepowerofmonetarypolicythisway,i
tisimportanttorememberthat
weareconsideringthesortsofsmallvariationsimplicitintheloglinearizations(5.10)
and (5.12),respectingtherequirementthatallfirmshavepriceatleastasgreatas

marginalcost.




The New Neoclassical Synthesis •263
the principles underlying this prescription. For concreteness and simplic­
ity, w e work within the time-dependent price-setting model developed
above. The principles are sufficiently general, however, that they will
guide m onetary policy in other N N S models as well. The role of monetary
policy in the n ew synthesis derives from two sources. First, the underly­
ing microeconom ic structure suggests that it is desirable to stabilize the
average markup, avoiding a source of time-varying distortions to the
macroeconomy. Second, forward-looking price-setting behavior makes it
feasible to design simple policies that will accomplish this stabilization.

7.1 THE OPTIMAL RATE OF INFLATION
W h a t are the implications of the new synthesis for the optimal rate of
inflation? W h ile a complete analysis of this topic is beyond the scope of
the present stu d y w e can identify several key features that are impor­
tant. First, the rate o f inflation affects the distribution of relative prices in
any m od el w ith price stickiness, which in turn has effects on the end-use
value of output that w e described above. These are minimized when
there is zero inflation. Second, the average markup depends on the rate
o f inflation: in the example that w e study further below, the average
m arkup is m inim ized at a rate of inflation that is near zero. Third, if
resources are expended adjusting prices, then these are minimized at
zero inflation. H ence, on these three grounds, the incorporation of im­
perfect competition and price stickiness leads to the suggestion that a
rate o f inflation close to zero is desirable.
H ow ever, Friedman (1969) earlier argued that it was desirable to have
expected deflation, so that the short-term nominal interest rate was zero.
Th u s, a complete analysis o f the optimal rate of inflation must balance
the m onetary benefits from disinflation with the distortion costs associ­
ated with deflation.
7 .1.2

Effects on M arkups 31

Early Keynesian analyses recognized that

steady inflation w ould erode the market power of firms, suggesting bene­
fits to sustained inflation. How ever, dynamic models of price setting
su ggest at best a small positive inflation rate on these grounds. Moreover,
these m odels of price setting also suggest that larger rates of inflation will
raise, rather than lower, average markups because of expected inflation
effects. W e use Figure 3 to display the main ingredients of this conclusion.
First, in any m odel with sticky prices, positive inflation does mechanically
erode the relative prices of firms which are not adjusting, or, equivalently,

32.ThediscussioninthissectiondrawsheavilyonKingandWolman(1996)
,whoanalyze

thelinkbetweeninflationand theaveragemarkup inaCalvo-stylemodelofprice
setting.




264 ■ GOODFRIEND & KING

Figure 3 THE EFFECTS OF STEADY-STATE INFLATION
A : R a tio o f P to P *

B : M a rg in a l M a rk u p

1.010

1.000
0.990
0.980
0.970

—

^

0.960
0.950
0.940
0.930
2

3

4

5

6

7

8

10

1.150
1.140
1.130
1 1.120
1 Mio
1 1.100
S’ 1.090
2
1.080
1.070
1.060

%9

10

Infla tion

In f la t io n

D : R e la tiv e P r ic e D isto rtio n

Average Maikup

C : A v era g e M ark u p

In fla tio n

there will be higher relative prices for those firms that are adjusting. To
provide an idea of the quantitative im portance of this channel, panel A
sh ow s the effect o f inflation on PIP* w ith a dem an d elasticity o f 11 and the
4-quarter staggering o f price adjustm ent suggested by som e of Taylor's
work (a)j = 0.25 for j =

0, 1 , 2, 3). A 10%

annual inflation rate lowers P/P* by

about 4 % . Second, confronted w ith a situation o f higher steady-state
inflation, a rational price-setting firm has an incentive to raise its marginal
m arkup. U sin g the sam e parameter values as above, panel B show s that a

10%

inflation rate causes

/a *

= P * /^ to increase to 1.15 from the zero-

inflation level e/(e — 1) = 1.10. T h u s, firms raise the marginal markup
substantially in response to ancitipated inflation.
The average m arkup em bodies both the inflation-erosion and expectedinflation effects, since it is sim ply the product

fL — (P*/^0 (P/P*)-32

32. Itispossibletoshowanalyticallythatinflationhasanegativeeffectonaverage mark­

upsnearzeroinflation:

= n I ---------------------- I < o.




The NewNeoclassical Synthesis •265
Accordingly, panel C displays the combined effect of inflation on the
average m arkup, yielding three results of interest. First, the smallest
value o f the average markup occurs at a positive inflation rate, but
this rate is not very different from zero. Second, the effect of inflation
on the markup is positive at higher inflation rates. Third, the overall
effect o f inflation on the average markup is very small quantitatively
near zero inflation. How ever, larger inflations actually raise rather
than low er the average markup: increasing the inflation rate to
per year from zero produces an increase in /i from

10%

e ! ( e - 1) = 1.10 to

1.1044. T h u s, with small inflations or deflations, the monetary author­
ity cannot influence the average markup by very much in the N N S
m odel.

7.1 .2

R elative-Price D istortions from Inflation

If there is no inflation in the

steady state, then all relative prices will be one and the end-use value of
output will be m axim ized. Further, small changes in relative prices near
this initial point will have no effect on the ratio d jy tf so that there will be
no productivity effect of small business cycles or small rates of inflation
or deflation. H ow ever, using a demand elasticity of e =

11 and 4-quarter

10%

annual inflation

staggering o f prices as above, we calculate that a

rate will low er the end-use value of output by 0.4% and more generally
display the relationship between inflation and relative-price distortions
in panel D o f Figure S .33Thus, the N N S framework indicates a quantita­
tively important direct social cost of sustained inflation arising from
relative-price distortions.
Taking these findings concerning the average markup and the size of
relative-price distortions together with the observation from Section 4
that there are relatively small gains from reducing inflation from zero to
the Friedman rule, the N N S model recommends that the monetary au­
thority target a near-zero rate of inflation. Since the productivity effects
o f relative-price distortions are minor near zero inflation, in what follows
w e focus solely on m ovem ents in the average markup in considering the
response o f the macroeconomy to various shocks.

Thus, acasecanbemade forreducingmonopolisticcompetitiondistortionsviaa
positiveinflationratewithintheNNS approach.Thisderivationisrelatedtothoseof
BenabouandKonieczny(1994)inaverydifferentsetup.Goodfnend(1997)makesa
similarcaseforpositiveinflationinamodelinwhichthereisa zoneofindeterminacy
fortheaveragemarkup.
.
33.Thatis,i
twilllowertheratiodfyt fromthezero-inflationlevelofunityto0.996.




266 * GOODFRIEND & KING

7.2 M ON ETARY PO LICY A N D THE BUSINESS CYCLE
H o w sh ou ld m onetary policy vary over the course o f the business cycle?
W e argue the objective o f the m onetary authority should be to produce a
constant path for the average m arkup or, equivalently, for real marginal
cost. W h ile m arkup constancy is an ad h oc objective, it is attractive to us
for three related reasons. First, it brings about the sam e response of the
real econ om y to various shocks as w o u ld arise if all prices were perfectly
flexible. Second, it corresponds to tax sm oothin g as recom m ended in the
public-finance literature .34 Finally, it is consistent w ith the traditional
su ggested focus o f m onetary policy, w hich is to eliminate gaps between
actual output and a tim e-varying level of potential (capacity) output.
O u r recom m endation am ounts to a neutral m onetary policy in the
sense that it keeps the average m arkup at its steady-state level and
m akes the N N S m o d el behave like a noncom petitive RBC economy.
Neutral m onetary policy accom m odates shocks that w o u ld alter the equi­
librium levels o f output and em p lo y m en t with flexible prices, such as
changes in productivity, fiscal policy, and international relative price
changes, and som e that w ou ld not, such as m o n ey dem an d shifts.
7.2.2 N eutral M on eta ry Policy and Price D ynam ics

N N S price dynamics

involve forw ard-looking and backw ard-looking com ponents, as dis­
cussed in the previous section. To a first approxim ation [as in (5.10)], an
adjusting firm sets its price at

log P ? =

(E, log P t+j + log ( * l+im
^h=oP

j*= o

To a first approxim ation [as in (5.12)], the price level is log P t =

<of log

P l­
u n d e r the neutral-m onetary-policy requirem ent, real marginal cost is
constant n o w and at all future dates, so that price setting depends only
on the expected future path o f the price level. Accordingly, the two
dynam ic equations im ply an expectational difference equation that can
be solved to determ ine the price level and inflation implications of neu­
tral m onetary policy. It is possible to produce a general mathematical
solution to this difference equation, but instead, w e look at several spe-

34. Existinganalysesofdynamicallyoptimaltaxationinastochasticgeneralequilibrium

settingaresupportiveofthisassumption.Notably,inaneconomywithelasticsupply
oflaborandcapitalservices,Zhu(1995)showsthatthereislittlevariationintaxrates
oneitherfactor.Ireland(1996a)isanimportantstartonstudyingoptimalmonetary
policyinenvironmentswithimperfectcompetitionandstickypricesthatdrawsonthe
optimal-taxationapproach.




The NewNeoclassical Synthesis ■267
cial cases o f this solution to provide an intuitive understanding of the
implications o f neutral monetary policy.
7.2.2 The D esirability o f a Constant Price Level

The benchmark result is

that a constant priceJevel is a neutral monetary policy. That is, if we set
log P* = log P t = log P at all dates in the price equations (5.10) and (5.12),
then a present value of real marginal cost must be expected to be zero at
all dates,

0=

Et I jlg p }a)/(X}hZl0

<oh) log(^(+;/^ ), which can only be satisfied

by a constant level of real marginal cost .35There are two equivalent ways
of stating this conclusion. Directly, a monetary authority committed to
targeting a growth path for the price level must do so by maintaining
constant real marginal cost or equivalently a constant average markup.
Alternatively, one can say that a monetary authority committed to neu­
tral policy m ust target a constant inflation rate.
7.2.3 The M on ey-S u p p ly Process Supporting Neutral Outcomes

Under neu­

tral m onetary policy, output behaves according to a monopolistically
com petitive real business cycle with a constant markup /x in the face of
shocks to technology, fiscal policy, and international relative prices. Neu­
tral policy eliminates output gaps, making y t = y t at all dates.
U nder a neutral policy, the monetary authority accommodates varia­
tions in m o n ey dem and to insure that excesses or shortages of money do
not create aggregate dem and disturbances. To work out the implications
for m o n ey supply, suppose that the price-level path under neutral policy
is given by log P t = log P
and

7t is

+ tt, where log P t is the log of the price level

the trend rate of inflation. Since inflation is constant, variations

in the real (r) and nominal (jR) interest rate are identical (Rt - F, + tt).
Then, if the m o n ey dem and is log M t = log Pt - my log y t - mRRt ~ vt, the
m o n ey stock m ust be
log M t = log P t - my log y t + mR( r t +

tt)

- log vr

(7.1)

That is: the monetary authority should accommodate movements in
output and interest rates obtaining in the RBC model, and velocity
shocks, too.
7.2 .4 Initial Conditions and Inflation Transitions

The optimal pricing equa­

tions readily allow for a characterization of neutral monetary policy un­
der m ore general conditions. Two decades ago, Edmund Phelps and

35.More generally,anyprice-levelpathwithaconstantinflationrateatal
ldatesalso

stabilizesthemarkup. This conclusionisobtainedby similarreasoningand more
algebra,togetherwithsettingp = 1.




268 * GOODFRIEND & KING
Guillerm o C alvo studied the disinflation problem in a basic fixed-wage
m odel with a mathematical structure similar to (5.10) and (5 . 12).36 Two
key features o f neutral m onetary policy carry over to the economics of
disinflation. First, the average m arkup m u st be constant through time,
w hich am ounts to requiring that the price adjustm ent decision depend
only on the expected future path o f the price level: log P? = Et S jlj (Pj ^
p h a)h) log Pf+;. Second, the path o f the price level is just a function of
the price adjustm ent decisions m ade at various dates: log P t ~

log

P*_;-. W h e n w e solve the resulting expectational difference equation as­
su m ing that the steady-state inflation rate is zero, the "sta b le " solution is
of the form

log P t -

log P U = §

°}(log P t f - log P l h l ),

(7.2)

M
w here the coefficients

are functions o f the parameters

and /3. That

is, there is a unique path o f price adjustm ents w hich m u st occur if there
is to be a constant average m arkup.
There are a n um ber o f implications o f this P h elp s-C a lv o neutraldisinflation form ula. First, neutral m onetary policy could equivalently be
stated as a rule for the growth rate o f n ew ly se t prices,

77?

— log Pf - log

PJLj. Second, given that w e have determ ined the grow th rate 77? necessary
for a neutral m onetary policy, w e can u se the price-level equation (5. 12) to
determ ine the neutral transition path for the m easured rate of inflation,

7Tt =

log P t ~ log P,_j = Zj~j (Oj7lf_;,

In Section 7.1 above w e u sed a 4-quarter Taylor m o d el to get an idea of
the quantitative sensitivity o f the average m arkup to inflation in a steady
state. In that m o d el, it turns out that a neutral transition to 7? = 0 takes
the form

<

-

-0.437Tf_, -

0 .1 2 tt?_2.

(7.3)

That is, w h en w e begin in an inflationary steady state with a quarterly
rate o f inflation of, say, 2 .5 % (so that the annual inflation rate is initially

10% ),

then there m u st be a price decrease on the part of adjusting firms

36.TheresultsarereportedinPhelps(1978)
,whichcontainsanappendixcoauthoredwith

Calvo. The appendix toourworking paper containsour derivationoftheneutral
monetarypolicyunderthemoregeneralconditionsnecessaryforthevariousscenarios
discussedinthetext.WethankOlivierBlanchardandJulioRotembergforalertingusto
thisreference.




The NewNeoclassical Synthesis •269
equal to - ( 0 . 4 2 + 0.12) x 0.025 = -0 .0 1 3 7 5 in the impact period of a
neutral disinflation. This price decline is necessary to stabilize the aver­
age m arkup given the past price rises built into the system, i.e., the
initial conditions

=

7j?_2 =

0.025. With this aggressive policy action,

the actual inflation rate rrt = £ (tt* + tt* j + tt?_2+

7rf_3) drops from 2.5%

to

about 1 .5 % in the impact period of the policy and subsequently declines
quickly to zero over the course of one year.37

7 .2 .5 Im perfect Control o f the Price Level

We can also operationalize neu­

tral m onetary policy w hen the monetary authority has imperfect control
o f the price level. In such a setting, the monetary authority cannot
achieve perfect control of the markup tax, but can keep it from varying in
expected value. That is, its policy rule can make (l/SJlJ, p ho>h) £ f_,

a>;-

log(a^+/a0 = 0. The preceding results then apply to the expected compo­
nent o f m onetary policy, with an additional price adjustment shock intro­
duced into the analysis. That is, with imperfect control of the price level,
neutral m onetary policy takes the form

= S a/77? - /+ &

(7,4)

7=1

where

= log Pf -

E(_j log P f. Thus, the central bank accommodates

som e portion of price-level targeting errors, as in Taylor's analysis.

7 .2 .6 Com parison o f Inflation Targets and Price-Level Rules

Many central

banks pursue inflation targets which allow for base drift in the price level.
In our setting, a return to a fixed-price-level path is undesirable, since it
requires variations in the average markup.
W e can use the preceding analysis to quantify how much base drift is
desirable in the setup with 4-quarter staggered price setting given in (7.3).
Su ppose that incomplete information leads to a targeting error,

>

0,

which the m onetary authority learns of at the end of the current period.
H o w m u ch o f the forecasting error in the price level should be reversed

37.Thepreciseformofpricestickinessisimportantforthedetailsofneutraldisinflation.
Withtwo-periodstaggeredpricesettingasinSection6.3,neutralmonetarypolicywith
azeroinflationtargetimplies< =0forallperiodsafterthedisinflationbegins sotha

thepathofthepricelevelislogP = logP* = logP? Theinflationratem thefirst
periodofthepolicyisaccordinglytt=J(0.025),asone-halfoftheagentscatchuptothe
othersatlogPf_r




270 • GOODFRIEND & KING
eventually ?38 In the current setup (7.3), the desirable long-run effect on
the price level is sim ply

------^----------------------- &------------ = 0.6
1 - 2 $

oj

1 -

( - 0 . 4 3 - 0.12)

Thus, the m onetary authority allows about six-tenths o f a price forecast­
ing error to feed through into the general price level in the long run.

8 . The Practice o f M o n eta ry P olicy
W hile price stability has long been suggested as a prim ary objective for
monetary policy, a num ber of questions have arisen about its practical
desirability and feasibility. This section takes up four major concerns
using the approach o f the N e w Neoclassical Synthesis. First, the effects
of oil and other com m odity price shocks have been lon g discussed by
Keynesian econom ists as a reason for not stabilizing the price level.
Second, M ilton Friedman and other monetarists have questioned the
desirability of inflation targeting on the basis of their reading of mone­
tary history and. the long and variable lags in the link betw een money
and prices. Third, N e w Keynesians such as John Taylor have suggested
the existence of important trade-offs betw een output and inflation vari­
ability. Fourth, central bankers routinely worry about the tactics of using
their preferred policy instrument, a short-term interest rate. In address­
ing these issues below, w e illustrate h o w the n ew synthesis can guide
the practice o f m onetary policy.
8.1 A N OIL SH O CK IN THE NEW SYNTHESIS MODEL
Oil shocks pose a difficult problem for m onetary policy because they can
create inflation and un em ploym en t at the sam e time. This problem,
however, m akes oil shocks particularly instructive for illustrating the
mechanics of the N N S framework and its prescriptive p ow er for mone­
tary policy. The analysis also highlights the com plem entarity of RBC and
Keynesian reasoning that is inherited by N N S m odels.
It is natural to think of an oil shock as a restriction in the supply of oil
available for use in the production o f final good s. Firms produce output
by com bining (after overhead) capital and labor services w ith oil. Since
firms are monopolistically com petitive, output is dem and-determ ined.
38. We calculate the effect of such a forecasting error on the long-run price level under the
rule 7jf = 2;=1
+ £t using the same approach employed in the literature on
stochastic trends, since 77? follows an autoregressive process under the optimal policy-




The NewNeoclassical Synthesis •271
For any level of final dem and, optimal factor demands require the mar­
ginal cost o f producing output for a firm to be the same for an increase in
any o f the three factors of production. By analogy to (6.4), measured in
units o f the final-good aggregate, optimal use of energy requires that

w here cft is the quantity of energy (oil) input and jft is the real price of oil.
This gives us two independent marginal conditions for the three factors,
plus the production function itself that relates the three factor uses to the
dem and-determ ined level of output. A firm chooses optimal factor uses
taking factor prices as given. In general equilibrium, factor prices adjust
to clear the factor markets, and, by influencing the markup, factor-price
adjustm ents also help clear the final-goods market.
Since the price level is sticky, output is governed by aggregate demand
in the short run. Thus, w e need to take a stand on how aggregate
dem an d will behave in order to say how the system responds to the oil
shock. For illustrative purposes our strategy is to ask what aggregate
dem an d policy should do, and to assume that monetary policy supports
that level o f aggregate dem and.
W e benchm ark the optimal policy response with RBC reasoning. By con­
struction, the standard competitive RBC model would respond efficiently
to the oil shock. For our purposes, the key feature of the competitive RBC
m o d el is that firms price output at the marginal cost of production. The
gross m arkup is always 1 in the standard RBC model. A necessary condi­
tion for the N N S m odel to respond efficiently is that it also maintains a
constant m arkup. Thus, the N N S recommends that monetary policy
should aim to stabilize the markup against the oil shock, not accommodat­
ing any o f the oil price rise in higher inflation.
W ith neutral monetary policy in place, w e can ask how the N N S
m o d el w o u ld respond to the oil shock. A t the initial levels of factor
inputs, output, and price, the rise in the price of oil raises the nominal
marginal cost and hence cuts the markup. In order for monetary policy
to restore the markup to its initial level, policy must depress aggregate
dem an d and cut em ploym ent. From the Keynesian perspective, such a
recom m endation sounds like adding insult to injury— causing em ploy­
m ent to fall just w h en materials costs are high. Yet, RBC reasoning says
that the econ om y should produce less when the marginal cost of produc­
tion is temporarily high. That reasoning also suggests that the extent of
the proper cut in dem and depends on the expected persistence of the oil




272 * GOODFRIEND & KING
shock. A shock expected to be temporary has little w ealth effect on labor
su pply and consum ption dem an d. It m ainly raises the opportunity cost
of current work relative to leisure and o f current leisure relative to future
leisure. Thus, monetary policy should act to cut aggregate dem and tem­
porarily to reflect these opportunity costs. The temporary fall in current
incom e in this case w ould cause agents to bid up real interest rates as
they attempt to borrow to sm ooth consum ption. Importantly, real inter­
est rates m ust rise to support neutral m onetary policy
A n oil shock expected to be highly persistent, on the other hand,
w ould act like a persistent negative productivity shock, creating a large
negative wealth effect that w ou ld offset the substitution effect on labor
supply. Relatively little decline in em ploym en t m ight be called for in this
case. But it w ou ld be appropriate for m onetary policy to bring about a
cut in consum ption comm ensurate with the decline in productivity due
to the lack o f availability o f oil. The willingness to cut consum ption as
incom e declines m ight produce little upw ard pressure on the real inter­
est rate. In fact, w h en one takes account of the adverse effects on invest­
ment and the capital stock that m ight accom pany w hat am ounts to a
highly persistent negative shock to productivity, there w o u ld likely be
dow nw ard pressure on real interest rates.
To sum up, one m ight reasonably ask why, in practice, oil shocks have
been inflationary. First, to the extent that oil products are produced in
competitive markets and purchased directly by consum ers, the increase
in the price of oil gets directly into the price level w ithout being interme­
diated by goods-producing firms in the sticky-price sector o f the econ­
omy. To stabilize the price level against these direct price shocks would
require pursuing aggregate dem and policy restrictive en ou gh to push
dem and and em ploym ent d o w n in the sticky-price sector, thus increas­
ing the markup there. N N S reasoning does not recom m en d increasing
the markup in the sticky-price sector to stabilize the overall price level.
Policy should be accom m odative o f such direct price shocks, especially
since they are relative-price shocks w h o se effect on inflation is tempo­
rary. Second, and equally important, central banks can be reluctant to let
real interest rates rise sharply, especially w h en a cost shock is hurting the
economy. The inflationary consequences o f oil price shocks have proba­
bly been exacerbated by central-bank attem pts to sm ooth nom inal inter­
est rates with overly expansionary m on ey grow th.

8.2 IS INFLATION TARGETING PRACTICAL?
M onetary econom ists have lon g thought that price stability has much to
recom m end it as the primary goal for m onetary policy, and recently a
num ber of central banks have adopted explicit inflation targets as a




The NewNeoclassical Synthesis •273
guide for policy .39 It has been less clear, however, that inflation targets
could play a useful role as an immediate policy objective and a criterion
for perform ance. Using the N N S , we review practical arguments that
have been advanced against inflation targeting by Friedman (1960) and
others. W e argue that these objections are unduly pessimistic when one
recognizes the role o f sticky prices and central-bank credibility in price
setting.
5.2.2

In terpreting Historical Experience

Friedman's view is based in large

part o n his w ork on the monetary history of the United States with Anna
Schwartz, in which they found lags in the effect of monetary policy to be
long and variable, ranging between half a year to over two years. Reason­
ing on the basis of the historical data, Friedman observed that "th e price
level . . . could be an effective guide only if it were possible to predict,
first, the nonm onetary effects on the price level for a considerable period
o f time in the future, and second, the length of time it will take in each
particular instance for monetary actions to have their effect. . . . " He
concluded that " . . . the link between price changes and monetary
changes over short periods is too loose and too imperfectly known to
make price level stability an objective and reasonably unambiguous
guide to policy ."40
Friedm an's inference about the advisability of inflation targeting seems
too pessim istic. In the first place, none of the data from U.S. monetary
history w ere draw n from a policy regime guided by the purposeful pur­
suit o f price stability. The gold standard prior to World War I was one in
w hich trend inflation was small by today's standards. But the United
States h ad n o central bank, and m oney growth was heavily influenced by
banking panics on a number o f occasions, and by gold flows governed by
the balance o f paym ents and the happenstance of new discoveries and
m ining techniques. A s a consequence, short-run price-level variability
w as quite significant at times during the period .41
A fter the fou n din g of Federal Reserve there was inflation during
W orld W ar I follow ed by a sharp deflation after the war; then prices
stabilized in the 1920s, and the price level fell by around one-third from
1929 to 1933. The W orld War II inflation was not reversed subsequently,
and instead the nation entered a period in which the price level more
than tripled in the three decades following the Korean War.
N N S m od els im ply that the linkages between prices and output de­
p en d sensitively on the monetary regime. Since U.S. monetary history

39.SeeHaldane(1995)andLeidermanandSvensson(
1995)
*
40. Friedman (I960,pp. 87-88)
.
.
41. FriedmanandSchwartz(1963)andMeltzerandRobinson(1989)
.




274 ■GOODFRIEND & KING
has been a succession o f very different m onetary regim es, the N N S
w o u ld predict just the kind of apparent instability in the effect o f money
fou n d by Friedman and Schwartz. Robert G o rd o n 's findings, mentioned
in Section 5 .1 , o f radically different empirical price equations across
different sam ple periods are a m anifestation o f the sam e kind o f regimedepen den t instability.
8 .2 .2 The R ole o f Credibility

If inferences from historical data can be

m isleading, w e can m ake som e conjectures about low -inflation targeting
in the N N S m odel based on the role o f central-bank credibility in the
price-setting process. A ccording to (5.10), for instance, costly price set­
ting im plies that firms care about a distributed lead of the price level and
real marginal cost in setting tod ay's price. W h e n an inflation-targeting
regim e is perfectly credible, fixed distributed leads o f both prices and
real marginal cost (the reciprical o f m arkup) anchor current price-setting
behavior .42 A d d to that som e staggering o f price setting, and the pre­
sum ption is that credibility for lo w inflation is apt to be self-enforcing to
a large extent, because in such an environm ent, firms will think less
about inflation and be less nervous about it. This confidence w ould be
reinforced further by a legislative m andate m aking low inflation a prior­
ity for m onetary policy.
The m ain question for a central bank com m itted to low inflation is
h o w "fo r g iv in g " price setters are likely to be o f policy mistakes. Mistakes
will inevitably occur due to im perfect inform ation about the economy.
But such mistakes w o u ld have little effect if caught in time, precisely
because o f the sluggishness in price setting. O f course, a central bank
that allow ed mistakes to cumulate for so m e reason, so that inflation
began to m o v e significantly higher, could turn the distributed lead in the
price equation from a stabilizing anchor into a source o f destabilizing
inflation scares .43
Inflation scares are easy to understand from the perspective of the
n ew synthesis. A central bank has an incentive to cheat on its commit­
m ent to price stability in the N N S m o d el because a m onetary policy
action can reduce the m arkup distortion and increase em ployment.
Chari, K eh oe, and Prescott (1989), for instance, m ight argue that a cen­
tral bank w ithout a precom m itm ent technology could n ot sustain a lowinflation equilibrium at all. A t a m in im u m , their argum ent suggests that

42. Ball(1995)contrastscredibleand incredibledisinflationsinsettingswith forward-

lookingpricesetting.
showshowtheycreatedproblemsformonetarypolicy.

43. Goodfriend(1993)documentsanumberofinflationscaresinthe1979-1992periodand




The NewNeoclassical Synthesis •275
the incentive to cheat makes price setters hypersensitive to policy mis­
takes in a w ay that makes a low-inflation equilibrium extremely fragile.
It seem s to us that N N S reasoning coupled with recent monetary
policy developm ents weakens considerably the force of such a point. We
think that central banks such as the Federal Reserve today largely inter­
nalize the long-run costs of cheating. A s a result of the Volcker Fed's
taking responsibility for inflation in the late 1970s and successfully bring­
ing it d o w n , the Fed is n ow widely held to be responsible for inflation.44
M oreover, the low-inflation experience since then has demonstrated
clearly the long-run benefits of price stability. Hence, we believe that the
temptation for the Fed to cheat on its low-inflation commitment is much
weaker than in the past.

8.3 INFLATION A N D OUTPUT VARIABILITY
A lth ou gh his staggered-overlapping-contract model exhibits no longrun trade-off in the level of inflation and the level of output, Taylor
(1980) sh o w ed that it does imply a trade-off between the variance of
output and the variance of inflation. O n this basis, Taylor argued that
business cycles can be reduced only by accepting increased variability of
inflation.
Since N N S m odels em body the kind of price-setting behavior assumed
by Taylor, the question arises whether such models also present policy­
makers w ith a difficult choice between inflation and output variability.
The question is o f more than academic interest, since it bears on one of
the m o st important issues in central banking today: the design of a
legislative m andate for monetary policy. Most experts agree that some
form o f clear m andate w ould improve the effectiveness of policy by tying
d o w n inflation expectations and increasing central-bank accountability.
The n ew synthesis supports such reasoning. But there is no agreement
on w hether a trade-off exists or if it does, on how to allow'for it in a
m andate.
5.3.2 Is There a Trade-off?

Recall our principle that monetary policy in

N N S m od els should aim to keep the markup constant at the low level
consistent with near-zero inflation. Thus, monetary policy should offset

44.TheFeddidnotexplicitlyassertitsresponsibilityforinflationintheinitialOctober1979

announcementsofitsdisinflationarypolicy.However,byemphasizingtheroleplayed
bymoneygrowthintheinflationprocess,andbyannouncingachangeinoperating
procedurestocontrolmoney,ineffect,theFedimplicitlyacknowledgeditsresponsibil­
ityforinflation.Today,centralbanksarewidelyunderstoodbythepublictoberespon­
sibleforinflation.




276 • GOODFRIEND & KING
shocks to aggregate dem an d. Such policy actions w o u ld not only keep
output at potential but stabilize prices as well. O n the other hand, mone­
tary policy should accom m odate productivity shocks, taking into ac­
count any associated effects on labor su pply and the capital stock. Other­
w ise, an output gap w ou ld open that w o u ld cause the markup to vary.
There is n o trade-off in either of these cases— policy should stabilize both
the m arkup and prices in response to dem an d or productivity shocks.
Even for an oil shock, society clearly faces no trade-off if oil is an interme­
diate input. W e saw above that the best outcom e is to maintain price
stability and to reduce dem an d in response to the decline in productivity.
W h a t about a N N S m od el with a flexible-price goods-producing sector
alongside the sticky-price m onopolistically competitive sector, in which
shocks could impact inflation directly? Clearly, such a modification
w ou ld n ot change the conclusion w ith respect to aggregate demand or
productivity shocks, since these sh ou ld still be offset or accommodated,
respectively.
The added price flexibility, how ever, complicates the response to an oil
shock, because the restriction in the su pply o f oil causes the oil price to
rise relative to other prices. If policy w ere to depress aggregate demand
just en ough to maintain stable prices in the sticky-price sector, oil­
intensive product prices in the flexible-price sector w o u ld still rise. The
central bank could reduce aggregate dem an d en ough to prevent the
overall price level (flexible plus sticky prices) from rising, but then it
w ou ld raise the m arkup and create an output gap in the sticky-price
sector.
Thus, policy w ou ld appear to face a trade-off betw een inflation and
output variability with respect to relative-price shocks. But even here,
N N S reasoning provides a w a y out. Practically speaking, the new synthe­
sis suggests that a central bank should aim to stabilize an index of sticky
prices alone, a core price index. This v iew accords w ell with the Keynes­
ian em phasis on a core rather than an overall cost-of-living index, and
the monetarist recom m endation to stabilize a lon g-run index and ignore
such relative price m ovem en ts as oil price shocks. W h e n w e define the
m easure o f prices that a central bank sh ou ld stabilize as a core index of
sticky prices, w e once again find that there is n o policy trade-off between
inflation and output variability.
8 .3 .2 The D esign o f a Legislative M and ate fo r M on eta ry P olicy

W hat, then,

are the implications o f the n ew synthesis for the design of a legislative
m andate for m onetary policy? First, there is n o policy trade-off between
inflation and output variability if the targeted m easure o f inflation is a
core price index o f goo d s produced b y m onopolistically competitive




The New Neoclassical Synthesis •277
firms. Second, a central bank should seek to keep output at its potential
by targeting the m inim um markup consistent with near-zero core infla­
tion. Third, according to the analysis, in Section

7. 2,

a central bank

should partially accommodate core-price-level targeting mistakes in or­
der to keep output at its potential.

8.4 TACTICAL POLICY IMPLEMENTATION
The n ew synthesis suggests that a central bank must pursue an activist
policy to target inflation. There are great difficulties in implementing an
activist policy rule, m any of them well known and long debated among
m onetary economists and central bankers, some of which were ad­
dressed above. O ur purpose in this section is to make a few additional
points suggested by the new synthesis for thinking about the practical
im plem entation of policy.
8.4 .1

Interest-R ate Policy

Central banks invariably use a short-term inter­

est rate as their monetary policy instrument. The new synthesis says that
central bankers should manage a low-inflation targeting regime by mak­
ing the short-term nominal rate mimic the real short rate that would be
ground out by a well-specified RBC model with a low, constant markup.
RBC reasoning is indispensable for thinking about how much and in
w hat direction the real rate should be moved in response to a shock. For
instance, even the direction of the appropriate real-rate response to a
productivity shock depends on the expected duration of the shock, as
w e saw above w h en w e discussed the oil shock.
A s another example of the value of RBC reasoning, consider this.
Recently, a possible pickup in productivity growth has been cited as a
reason w h y the Federal Reserve need not raise short-term real interest
rates to maintain low inflation. In fact, the standard RBC component of
the N N S m odel suggests, at a minim um , that real rates would have to
rise one for one with an increase in trend productivity growth, e.g ., a 50basis-point increase in the growth rate would be matched by a 50-basispoint increase in real interest rates .45 Importantly, rates would have to
rise even if the econom y were otherwise operating at a noninflationary
potential level of G DR
Generally speaking, central-bank management of the short-term real
interest rate is difficult for the following reason. Although the current
output gap m ay m ove relatively closely and monotonically with the

45.Thisisthecaseacrosssteadystateswhenutilityislogarithmic.Rateswouldhaveto

riseevenmoreifconsumptionwerelesssubstitutableintertemporallythanlogarithmic
utilitysuggests.Moreover,thiscalculationdoesnotallowforthetransitoryupward
pressureonrealratesduetoanaccompanyinginvestmentboom.




278 * GOODFRIEND & KING
current markup in N N S m odels, the real interest rate and the markup
are not closely related. Real interest rates rise and fall in response to
various shocks in the RBC m odel, even though there is n o markup at all.
The real interest rate adjusts to equate saving and investm ent. A t any
point in time, the current real rate (and also the expected future se­
quence of real rates) n eeded to support a constant m arkup, will depend
in a com plex w a y on the nature and m agnitude o f current shocks hitting
the econom y and their expected duration.
8 .4 .2 Inflation Indicators

N N S reasoning suggests that familiar indica­

tors o f rising inflation will be less effective in a fully credible lowinflation-targeting regime. For instance, rapid inventory stockbuilding
and lengthening delivery lags warned o f inflation in the past. From the
perspective o f N N S m odels, precautionary or speculative stockbuilding
was rational precisely because m onetary policy w o u ld fail to restrain
aggregate dem and before it pressed against capacity and raised expected
real marginal cost sufficiently to cause firms to pursue inflationary price
increases. In such circumstances, rising inflation expectations would
rationally be incorporated into long-term interest rates as w ell, and bond
rates could also warn of future inflation.
In contrast, if a central bank consistently controlled inflation, firms
w ould be less likely to build up inventories or place precautionary ad­
vance orders w h en the econom y neared full em ploym en t. Expected infla­
tion w ou ld not raise bond rates. Bond rates w o u ld rise in cyclical expan­
sions only because they em bodied increases in future short-term real
interest rates expected to be brought about by the central bank. In a fully
credible low-inflation-targeting regim e, a central bank w o u ld have to
becom e more sensitive to familiar indicators than in the past, and would
likely need to develop additional indicators to guide its interest-rate
policy actions.

9. S u m m ary and C on clu sions
The m odels of the N e w Neoclassical Synthesis are com plex since they
involve intertemporal optimization, rational expectations, monopolistic
com petition, costly price adjustm ent and dynam ic price setting, and an
important role for m onetary policy. O u r main pu rposes in the paper
were threefold: to motivate the separate com p on en ts o f the n ew synthe­
sis, to present a conceptual fram ew ork for thinking about N N S m odels,
and to use that apparatus to develop recom m endations for monetary
policy.
Tw o fundam ental insights are at the core o f our fram ew ork. First,




The NewNeoclassical Synthesis •279
Keynesian and RBC mechanisms can be viewed as operating through
som ew hat different channels. Holding the average markup constant,
N N S m o d el mechanics resemble those of a pure, albeit noncompetitive,
RBC m o d el. O n the other hand, the Keynesian influence of aggregate
dem and o n em ploym ent and output works by shrinking or increasing
the m arkup, w hich acts like a distorting tax on economic activity.
Second, dynam ic costly price adjustment means that firms adjust
price according to an expected distributed lead of the price level and real
marginal cost, w here the price level is an average of current prices and
those set in the past. W e show ed that the forward-looking price-setting
equation and a price-level expression form a price block that can be
solved to express the inflation rate as a function of prices set in the past,
current real marginal cost, and a distributed lead of expected real mar­
ginal cost. Since real marginal cost is the inverse of the markup, the
evolution o f inflation in the N N S model depends importantly on current
and expected future markups.
The recom m en ded neutral monetary policy in the new synthesis fol­
low s directly from the above insights and the idea that the markup
ought to be held constant. Markup constancy is attractive because it
delivers the sam e response of the real economy to various shocks as
w ou ld arise if all prices were perfectly flexible. We showed that the
steady-state m arkup should be minimized at a near-zero inflation rate,
and argued that m ost of the benefits for monetary exchange would be
realized at near-zero inflation as well. Thus, we found that near-zero
inflation targeting was both desirable and feasible in the N N S model.
Even though the n ew synthesis inherits much of the spirit of the old, it
differs sharply in terms of the role of monetary policy. Economists work­
ing w ithin the synthesis of the 1960s were pessimistic about taming
inflation, view ing inflation as having a momentum of its own and fluctu­
ating w ith unm anageable shifts in the psychology of price setters. The
n ew synthesis also view s expectations as critical to the inflation process,
but sees expectations as amenable to management by a monetary policy
rule.
The n e w synthesis has much to say about the practical implementa­
tion o f inflation targets. Since expectations of future markups play a key
role in the inflation-generating process, successful inflation targeting
requires a credible commitment to low inflation, so that expectations of
markup constancy anchor the inflation-generating equation. In order to
maintain m arkup constancy, monetary policy must accommodate m ove­
m ents in potential G D P brought about by RBC forces such as productiv­
ity, fiscal policy, or materials cost shocks. Accom modation must be twodim ensional. First, m oney growth must satisfy induced movements in




280 • GOODFRIEND & KING
m on ey dem and. Second, the m onetary authority m u st m ove its nominal
short-term interest-rate instrum ent to support real short-term interestrate m ovem en ts called for by underlying RBC forces. Ironically, in spite
o f the fact that K eynesian effects of m onetary policy on real activity are
pow erful in N N S m od els, m onetary policy is best w h en it eliminates
Keynesian effects entirely.
Researchers have m erely scratched ti\ e surface in thinking about NNS
m odels: such m odels will surely be elaborated and im proved in the
future. Looking backward: N N S m od els should im prove our understand­
ing of m acroeconom ic outcom es during volatile inflationary periods,
such as that extending from the m id-1960s through the early 1980s,
w h en both large m onetary policy shocks and large su pply shocks were
important. M oreover, the division o f the effect of an increase in money
growth betw een inflation and output in the N N S m od el depends sensi­
tively on the extent to w hich the faster m o n ey grow th is expected to
persist. T h u s, N N S m odels should help us understand the time-varying
effect of m o n ey on prices and output that characterizes historical time
series. Looking forward: as the United States and other countries around
the w orld maintain low inflation, su pply-side forces should loom as
large as dem an d-side forces for the business cycle. W e expect N N S m od­
els to becom e increasingly important in providing m onetary policy ad­
vice in such an environm ent.

REFERENCES
Anderson, L. C ., and K. M. Carlson. (1972). A n economometric analysis of the
relation of m onetary variables to the behavior of prices and unemployment. In
The Econometrics of Price Determination , O. Eckstein (ed.). Washington DC:
Board of Governors, pp. 166-183.
— ------, and J. Jordan. (1968). M onetary and fiscal policy: A test of their relative
importance in economic stabilization. Federal Reserve Bank of St. Louis Review,
November, 11-24.
Ando, A . (1974). Some aspects of stabilization policies, the monetarist contro­
versy, and the MPS m odel. International Economic Review 15(3):541-571.
Ball, L. (1995). Disinflation w ith imperfect credibility. Journal of Monetary Econom­
ics 35(l):5-23.
Barro, R. J. (1977). Long-term contracting, sticky prices, and monetary policy.
Journal of Monetary Economics 3(3):305-316.
, and D. Gordon. (1983). Rules, discretion, and reputation in a model of
m onetary policy. Journal of Monetary Economics 12(1):101—122Basu, S., and J. G. Fem ald. (1997). Returns to scale in U.S. production: Estimates
and implications. Journal of Political Economy 105(2).*249-283.
Benabou, R „ and D. Konieczny. (1994). O n inflation and output w ith costly price
changes: A simple unifying result. American Economic Review 84(l):290-297.
Blanchard, O . J. and N. Kiyotaki. (1987). M onopolistic competition and the effects of aggregate dem and. American Economic Review 77(4):647-666.




The NewNeoclassical Synthesis •281
Brayton, F.( A . Levin, R. Tryon, and J. Williams. (1996). The evolution of macro
models at the Federal Reserve Board. Federal Reserve Board. Manuscript. Presented at Cam egie-Rochester Conference on Public Policy, November 1996
Burnside, A . C ., M. Eichenbaum, and S. T. Rebelo. (1995). Capital utilization and
returns to scale. In NBER Macroeconomics Annual 1995, B. Bemanke and 1
Rotemberg (eds.). Cambridge, MA: The MIT Press, pp. 67-110.
Calvo, G. A . (1983). Staggered prices in a utility-maximizing framework. Journal
of Monetary Economics 12(3):383-398.
Caplin, A ., and J. Leahy. (1991). State dependent pricing and the dynamics of
m oney and output. Quarterly Journal of Economics 106(3):683-708.
Chari, V. V., P. Kehoe, and E. Prescott. (1989). Time consistency and policy. In
Modern Business Cycle Theory, Robert Barro (ed.). Cambridge, MA: Harvard
University Press, pp. 265-305.
Cho, J. O ., and T. F. Cooley. (1994). Employment and hours over the business
cycle. Journal o f Economic Dynamics and Control 18(2):411-432.
Cooley, T. F., and G. D. Hansen. (1989). The inflation tax in a real business cycle
model. American Economic Review 79(4):733-748.
de Leeuw, F., and E. Gramlich. (1968). The Federal Reserve-MIT econometric
model. Board of Governors of the Federal Reserve Bank Bulletin, 11-40.
de Menil, G ., and J. Enzler. (1972). Prices and wages in the FRB-MIT-Penn
Econometric Model. In The Econometrics of Price Determination, O. Eckstein
(ed.). Washington DC: Board of Governors, pp. 277-308.
Dotsey, M ., R. G. King, and A . L. Wolman. (1996), State dependent pricing and
the general equilibrium dynamics of money and output. University of Virginia
and Federal Reserve Bank of Richmond. Manuscript.
Friedman, M. (1960). A Program for Monetary Stability. New York: Fordham Uni­
versity Press.
-------- . (1968). The role of monetary policy. American Economic Review 58(1):1—17.
---------. (1969) The optimum quantity of money. In The Optimum Quantity of
Money, and Other Essays. Chicago: Aldine Publishing Co.
— -— . (1970). The Counter-revolution in Monetary Theory. London: The Institute
for Economic Affairs.
-------- / and A . J. Schwartz (1963). A Monetary History of the United States 18671960. Princeton, NJ: Princeton University Press.
Goodfriend, M. S. (1993). Interest rate policy and the inflation scare problem:
1979-92. Federal Reserve Bank of Richmond Economic Quarterly 79(l):l-24.
— ------. (1997). A framework for the analysis of moderate inflations. Journal of
Monetary Economics 39(l):45-66.
Gordon, R. J. (1982). Price intertia and policy ineffectiveness in the United States,
1890-1980. Journal of Political Economy 90(6):1087—1117.
Haldane, A ., ed. (1995). Targeting Inflation. London: Bank of England.
Hall, R. E. (1988). The relationship between price and marginal cost in U.S.
industry. Journal of Political Economy 96(5):921-947.
■
-------- . (1991). Substitution over time in consumption and work. In Value and
Capital Fifty Years Later, L. W. McKenzie and S. Zamagni (eds.). N ew York: New
York University Press, pp. 239-267.
Ireland, P. N. (1996a). The role of countercyclical monetary policy, journal of
Political Economy 104(4):704-723.
---------. (1996b). Expectations, credibility, and time-consistent monetary policy.
Rutgers University. Manuscript.




282 * G O O D FRIEN D & KING

King, R. G ., and C. I. Plosser (1984). Money, credit and'prices in a real business
cycle. American Economic Review 74(3):363-380.
---------, and S. T. Rebelo. (1997). Real business cycles. University of Virginia.
M anuscript, Forthcoming in Handbook of Macroeconomics.
---------, and A . L. Wolman. (1996). Inflation targeting in a St. Louis model of the
21st century. Federal Reserve Bank of St. Louis Review 78(3):83-107.
Leiderman, L., and L. Svensson, eds. (1995). Inflation Targets. London: Center for
Economic Policy Research.
Lucas, R. E., Jr. (1976). Econometric policy evaluation: A critique. CarnegieRochester Conference Series on Public Polio/ 1:19-46.
---------. (1993). On the welfare cost of inflation. University of Chicago. Unpub­
lished manuscript.
Mankiw, N . G. (1990). A quick refresher course in macroeconomics. Journal of
Economic Literature 28(4):1645-1660.
---------, and D. Romer. (1991). New Keynesian Macroeconomics, 2 vols. Cambridge,
M A: The MIT Press.
M cCallum, B. T. (1980). Rational expectations and macroeconomic stabilization
policy: A n overview. Journal of Money, Credit and Banking 12:716-746.
McGrattan, E. R. (1994). The macroeconomic effects of distortionary taxation.
Journal of Monetary Economics 33(3)573-601.
Meltzer, A ., and S. Robinson. (1989). Stability under the gold standard in prac­
tice. In Monetary History and International Finance: Essays in Honor of Anna J.
Schwartz, Michael Bordo (ed.). Chicago: University of Chicago Press, pp. 163195.
Nelson, E. (1997). A fram ework for analyzing alternative m odels of nominal
rigidities. Carnegie M ellon University. Manuscript.
O kun, A . (1970). Inflation: The problems and prospects before us. In Inflation:
The Problems it Creates and the Policies It Requires, (eds.) H. Fowler and M.
Gilbert. N ew York: N ew York University Press, pp. 3-53.
---------, M. Peck, and W. Smith. (1969). The 1969 Annual Report of the Council of
Economic Advisors . Washington: U.S. Governm ent Printing Office.
Phelps, E. S. (1978). Disinflation w ithout recession: A daptive guideposts and
monetary policy. Weltwirtschaftliches Archiv 114(4):783-809.
Plosser, C. -I. (1989). Understanding real business cycles. Journal of Economic
Perspectives 3(3):51 -77 .
Prescott, E. (1986). Theory ahead of business cycle measurement. Federal Reserve
Bank of Minneapolis Quarterly Review 10(4):9-22.
Romer, D. (1993). The N ew Keynesian synthesis. Journal of Economic Perspectives
7(l):5-22.
Rotemberg, J. J. (1987). The N ew Keynesian m icrofoundations. In NBER Macro'
economics Annual 1987 , S. Fischer (ed.). Cam bridge, M A: The MIT Press, pp.
69-104.
---------, and M. Woodford. (1991). M arkups and the business cycle. NBER
Macroeconomics Annual 1991, O. J. Blanchard and S. Fischer (eds.). Cambridge,
M A.: The MIT Press, pp. 63-129.
Samuelson, P. A. (1967). Economics: An Introductory Analysis , 7 th ed. N ew York:
M cGraw-Hill Book Company.
Sargent, T. J. (1986). Rational Expectations and Inflation. N ew York: Harper and
Row.
---------, and N . Wallace. (1975). "Rational expectations," the optimal money




Comment * 283

” °n'y “ K’11'
Sh^ ' S ' A; (1992r Int®rP retin8 the macroeconomic time series facts: The effects
of m onetary policy. European Economic Review 36(5);975-1012.

aL ° L L L y 8 8 ( l ) : t - 2 3 re8ate dynamiCS and staS8ered contracts, journal ofPolitiTt l n'n J; (19/72J \ T!’ Ie c™ el f ile™
' In The Battle Against Unemployment, Arthur
O kun (ed.). N ew York: W. W. Norton and Company, pp. 44- 53,
—
. (1974). The New Economics One Decade Older. Princeton, NJ: Princeton
University Press.
Wolman, A . I. (1996). How precise are estimates of the welfare costs of infla­
tion. In Three Essays on Monetary Economics. University of Virginia. Ph D
Dissertation, pp. 55-81.
Yun, T. (1996). Nominal price rigidity, money supply endogeneity and business
cycles. Journal of Monetary Economics 37(2):345-370.
Zhu, X. (1995). Endogenous capital utilization, investor's effort and optimal
fiscal policy. Journal of Monetary Economics 36(3):655-677.