GEORGE
L. PERRY
BrookingsInstitution
WhatHave We Learned
about Disinflation?
U.S. MONETARYPOLICYsince 1979has been gearedto stoppinginflation.
It has had considerablesuccess in doing so, but at a great cost in lost
output and high unemployment, not only in the United States, but
throughoutthe world. In Europeunemploymentrose steadilyfrom 1979
to 1982 and is expected to rise furtherin 1983. In the United States,
unemploymentrose to 10.7percentat the troughof the recession in the
fourthquarterof 1982,morethan I1/2points above the previouspostwar
recordreachedat the worst point of the severe 1975recession.
The durationof economic weakness in the presentdisinflationperiod
distinguishesit from previous postwar recessions even more than the
amountby which unemploymentrose and the level that unemployment
reached. If the briefbounce-backin economic activity aftermid-1980is
ignored,the recent U.S. recession lasted twelve quarters.The previous
postwar record was the five-quarterrecession of 1974-75. All other
postwar recessions had lasted less than a year. By the end of 1983the
cumulativeexcess of unemploymentover its 1979level will be about 11
percentagepoint-years, correspondingto an estimated $700 billion to
$900billionof forgoneGNP in today's prices.
This severe recession has been accompaniedby a dramaticslowing
in the rateof inflation.Table 1 summarizesseveral measuresof inflation
that reflect that slowdown, includingwidely used measures of actual
inflationrates, which are affected by many special developments not
closely associated with the underlying inflation problem, and some
alternativemeasuresof the underlyinginflationrate. Duringthe 1978I thankJudithD. KleinmanandPatriciaJ. Reganfor researchassistance.
587
(O
o
0
TT'o
N
0
C1
1 oo1
(-
(-o
ooooo
~~0*.~0~0*.
:
00 o.,
0
~~
00
(N0;;
ON
~
o-o t--
0 tr 06V
00
~ol
--oo
cl
m
ONOr?
CN 10 _O
.0~'~0
O
o
ON
E
0 ON
-
oo
--
0S
o
O
000r
cr
ON
0tNo
ON
O.
NNoo
O
--
oCo6o
FooooO O
~0
-)-00
-
.00
C
0
0
3-~~~~~~~~~~~~~7
~~ ~ ~ ~ ~
C
ci
ci
CN
rl
00
00
0
730
~ ~~ ~ ~ ~ ~ ~ ~ ~
3C
~~~~~~~~~~~~~Z>
Cu~
-
0o0
73
CZ
0
.0
0
~ ~
~
~ ~
z
Zou .Z
o
~~~~~
c
>
w5
George L. Perry
589
80 perioda combinationof priceincreasesin food andenergy(andrising
interest rates as they affected the shelter component of the consumer
price index) pushedthe actualinflationrate to more than 10 percent, as
measuredby any of the principalprice indexes shown in the top half of
the table. In partbecause of the cessation or reversalof these developments, these inflationrates slowed dramaticallyby 1982,and even more
by the firsthalf of 1983.
Both the accelerationof inflationand its subsequent slowdown are
greatly attenuated in the measures of the underlying inflation rate
displayedin the bottom half of the table. Nonetheless, the slowdown is
still substantial,amountingto 4 to 5'/2 percentagepoints between the
most inflationaryyear, 1980,andthe firsthalfof 1983,usingall measures
except the volatile producerprice index, which has had a much larger
swing.
One importantquestion raised by this experience is whether the
disinflationthat has been accomplishedis more than might have been
predicted on the basis of historical experience or whether it is about
what would have been expected along with the huge recession that has
occurred.The answer to this question, in turn, should shed some light
on theoretical issues concerninghow the economy works and on the
options confrontingpolicymakers. In particular,a central issue to be
consideredin lightof the experienceof the pastfew years is whetherthe
relationbetween output and inflationwill remainthe same when there
has been a significantchange in the policy regime, such as that which
occurredin the fall of 1979.
DisinflationPolicy and Inflation Models
One model of the inflation process, the wage norm hypothesis,
explicitlydistinguishesbetween the cyclical variationin inflationand an
ongoingnormrate of wage increasearoundwhich the cyclical variation
takes place.' Cyclical variationsin employmentand inflationarejointly
determinedby the optimalresponseof firmsto variationsin the demand
for theiroutput. Althoughexpectationscan, in principle,alterthe wage
1. George L. Perry, "Inflationin Theoryand Practice,"BPEA, 1:1980,pp. 207-41;
and Arthur M. Okun, Prices and Quantities: A Macroeconomic Analysis (Brookings
Institution,1981).
590
Brookings Papers on Economic Activity, 2:1983
norm, in practicethis hypothesisrelates historicalchanges in the norm
to experienced,ratherthananticipated,changesin the economy.
In a wage normequation,shiftsin the norm,such as those thatwould
arise from sustainedepisodes of inflationor economic slack, show up
empiricallyas shiftsin the constantterm.Thecoefficientsrelatingoutput
and inflationin the shortrunare determinedby institutionaland behavioral characteristicsthat are largelyunaffectedby changes in macroeconomicpolicies or by changesin the wage normitself. In empiricalwork,
I have modeledhistoricalnormsas stable except for occasional discontinuous shifts.2 Because wage norms do not respond in a linear or
systematic way to any one measure of economic performance-for
instance, they would respond less to a large spurt in the CPI coming
from supply shocks (1978-80) thanto an equivalentincrease in the CPI
associated with a sustained period of very low unemployment(196669)-I preferredthis procedureto moreelaborateways one mightmodel
normshifts. It has providedan adequate,simplecharacterizationof the
past.
For the present episode, this model predictedthat the policy aimed
persistentlyat disinflationwould eventually shift the wage normdownward, but only afterimposingunusuallygreator sustainedweakness in
the real economy. Under these conditions, the wage norm equation
would begin to exhibit forecast errorsindicatingthat wage normswere
shiftingdownward.
Most empirical studies of wage or price inflation have used past
inflationrates, ratherthan the concept of wage norms, to help explain
current inflation. The effect of lagged inflation in these models has
generallybeen interpretedas representingthe influenceof inflationary
expectationson currentinflation,where these expectations are formed
by an adaptiveprocess. In suchmodels, developmentsaffectinginflation
continue to have an influencelong afterthey occur. And the effect of a
sustained shock affecting inflation grows with the passage of time.
Because the laggedinflationtermsdirectlyembodylong-runand expectational effects, a sustained macroeconomicpolicy change is not expected to altereitherthe constanttermor the coefficientsof the equation
relatingoutputand inflation.
2. Perry, "Inflationin Theoryand Practice."CharlesSchultzefoundthis characterizationof normchangesexplainedinflationthroughoutthe peacetimeyearsof this century.
See CharlesL. Schultze, "Some MacroFoundationsfor MicroTheory,"BPEA,2:1981,
pp. 521-76.
George L. Perry
591
New classical models differin importantrespects from either of the
modelsjustdescribed.Theyrejectadaptiveexpectationsas a description
of how futureprice expectationsareformed.Generallythey relatethese
expectations to the policies that are anticipated,but that idea has never
been implementedin a widely accepted way in a predictive equation.
The problemis in determiningwhat economic agents expect policy to
be and what amount of inflationthey think will accompany it. These
models also hypothesizethat marketsalways clear and that in doing so
they incorporateexpectedfutureprices in an importantway. In the pure
form of such models, levels of outputand employmentdeviate fromfull
employmentlevels only because economic agents are mistakenabout
policy now and in the future. They interpretthe very flat short-run
Phillipscurve estimatedfrom historicaldata as a reflectionof mistakes
by economic agents resultingfromunpredictablepolicies.
The centralpolicy implicationof this theory is that policy should be
steady and predictable,a prescriptionusuallyinterpretedas supporting
monetarism.Accordingto the theory, the mistakesof economic agents
would be minimized under such a policy and the short-runrelation
between output and unemploymentwould become very steep. If the
monetarypolicy changeof 1979was understoodas a changeto a steadier
and more disinflationarypolicy regime, this theory predictedthat inflation would slow more promptly and output and employment would
decline less thanhistoricalequationswouldforecast.
A version of this theory associated primarilywith William Fellner
stresses the credibilityof policy in a moregeneralway.3In this version,
what matters is that policy convincingly promises it will not tolerate
inflation,quiteapartfrompreciselyhow policy is conductedin the short
run. The credibilityhypothesisalso anticipatedthatthe changein policy
regimein 1979,once widely believed, would bringdown inflationmore
promptlyand with a smallerloss of outputand employmentthanwould
be predictedby empiricalmodels fit to historicalperiods. And furthermore, it predictsthat the coefficientsrelatingoutputand inflationin the
short run will change under a consistent policy, makingthe apparent
Phillipscurve steeper.
As these formulationssuggest, some of the important,distinctive
predictionsof these alternativehypotheses could be tested only after a
complete cycle of disinflationand subsequentrecovery, when it would
3. William Fellner, Towards a Reconstruction of Macroeconomics: Problems of
Theory and Policy (American Enterprise Institute, 1976), pp. 116-18.
592
Brookings Papers on Economic Activity, 2:1983
be possible to observe whetherthe short-runrelationsbetween output
and inflationhad changedand whether a new inflationnorm had been
established.One implicationof those theories that stress expectations
of policy should be testable already:they promisedthat a clear policy
commitmentto disinflationwould yield better results-measured as
more promptdisinflationand more disinflationrelative to the increase
inunemployment-than modelsbasedon pastexperiencewouldpredict.
In this paper,I review how the actualpathof inflationduringthe past
few years compares with the path predicted by some equations fit to
historicaldata. I also look at some disaggregatedevidence aboutwhere
wage disinflationhas been most pronouncedand how that experience
compareswith the past. These comparisonsshouldindicatewhetherthe
economic performanceof this periodsupportsthe claimthatthe change
in monetary policy initiated in 1979 has speeded up the disinflation
process.
It should be noted that, if expectations are simply formed by actual
experience, there is no useful distinctionbetween the wage norm and
credibility hypotheses in their predictions for the present period of
disinflation.Both would then requirean extended period of low actual
inflation,broughtabout by an extended depressionin economic conditions, in order to shift wage inflationdownwardby more thanjust the
predictedcyclical response. There is no quantitativepredictionabout
how large such an eventual shift down mightbe. If equations that rely
on lagged inflationto capturelong-runor expectationaleffects should
overpredictin such a period, it would be evidence of a more favorable
outcome thanaveragepast behaviorwould predict.
Wage Inflation
The present disinflationis first examinedby looking at the behavior
of wages. Because wage costs representabouttwo-thirdsof value added
in the economy, wage behavioris the maindeterminantof business costs
andhence of the underlyingtrendin priceinflationin most sectors of the
economy.
ECONOMY-WIDE
WAGES
Table 2 comparesthe actualincreasesin the averagehourlyearnings
index with the increasepredictedby two closely relatedwage equations
593
George L. Perry
Table 2. Actual and Predicted Wage Changes, 1979:1 to 1983:2a
Percent
Equation2-2
Equation2-1
Error
Predicted
change
Error
8.1
8.4
8.3
8.4
-0.2
- 1.8
-0.2
-0.4
8.0
8.5
8.2
8.4
- 0.0
- 1.8
-0.1
-0.5
9.0
8.3
0.7
8.3
0.7
1980:2
3
4
9.3
8.6
9.8
7.8
8.0
7.9
1.5
0.6
1.9
8.0
8.3
8.3
1.4
0.3
1.5
1981:1
2
3
4
9.4
7.6
8.1
7.0
7.9
7.7
7.4
7.2
1.5
-0.1
0.7
-- 0.2
8.2
7.9
7.5
7.1
1.2
- 0.2
0.6
- 0.2
1982:1
2
3
4
6.8
5.8
5.8
4.7
6.9
6.4
6.3
6.0
-0.1
- 0.6
- 0.5
- 1.3
6.9
6.4
6.3
5.9
-
1983:1
2
5.2
3.4
5.8
5.7
-0.6
- 2.4
5.5
5.5
-0.3
- 2.1
Actual
change
Predicted
change
1979:1
2
3
4
7.9
6.6
8.1
8.0
1980:1
Period
In sample
Out of sample
0.2
0.5
0.4
1.2
Source: Estimations by the author based on George L. Perry, 'Inflation in Theory and Practice," BPEA, 1:1980,
pp. 207-41.
a. Wage changes are measured by the change in the log of the average hourly earnings index. Predicted wage
changes, w, come from the following equations (t = statistics in parentheses):
(2-1)
w = 0.2 + 20.0U'
- 8.0U-1 + 0.2PL + dummies
(-1.7)
(2.7)
(4.5)
standard error of estimate, 0.8,
(2-2)
w = 0.6 + 25.3U-1 - 15.U-1
(5.9)
(- 3.7)
+ 0.3PL + dumnmies
(4.3)
standard error of estimate, 0.8,
where U is the weighted unemployment rate, and PL is the change in the log of the lagged CPI. The equations differ
in how historical norm shifts were modeled. See Perry, "Inflation in Theory and Practice," for further details.
Equations 2-1 and 2-2 correspond, respectively, to equations 4-7 and 4-5 in that article. The sample period for both
is 1954:1 to 1980:1.
I presentedin my papercited above.4The estimatedequationsexplicitly
allowedfor a normshiftafterthe extendeddemandinflationof the 1960s,
and differ only in how they treat the transitionalquartersof that shift
statistically.Allowingfor this shift sharplyreducedthe effect of lagged
inflationon currentwage changes comparedto the effect estimatedby
4. Perry,"Inflationin TheoryandPractice,"p. 228, equations4-5 and4-7.
594
Brookings Papers on Economic Activity, 2:1983
most conventionalmodels. In the model that underliesthe equations,
substantialand sustainedpredictionerrorsare thus takenas evidence of
a normshift.5
Under the new policy regime, which began in 1979:4,wages at first
rose somewhat faster than predicted, althoughthe average underpredictionin these firstsix quartersis only slightlylargerthanthe standard
errorof the equationsof 0.8 percentwithinthe sampleperiod.Still, there
is no sign of exceptionallypromptdisinflationin these results.
Before 1982:4neitherequationin table 2 shows evidence that wages
were slowingmorethanpredicted.Even the overpredictionsfrom 1982:4
and 1983:1average only 0.75 and 0.95 percent in the equations, again
little differentfrom their standarderrors of 0.8 percent. Only the 3.4
percent rate of increase in average hourly earningsin 1983:2is significantly slower than predictedby either equation. One such observation
is not important.But if overpredictionsof the size in that quarterare
maintained,they would indicatethe wage normis shiftingdownward.
UNION WAGES
These recent developments in average wages have occurred in a
climate of unprecedentedconcessions in importantunion wage contracts. Table 3 compares first-yearwage adjustmentsin major union
wage situations in the present disinflationwith those adjustmentsthat
occurred around earlier recessions. Because of long-term contracts,
these first-year adjustments, which include changes under contract
reopenings, are more sensitive indicators of developments in union
wages thanare total effective adjustmentsor other averagesof all union
wages. The three groups of years in the table chronicle developments
aroundthe past three business cycles, startingwith the last year before
the downturnin the economy.
It is hard to draw inferences about the inflationprocess from the
episode of the early 1970sbecause of the price controls in effect in that
period.The President'sPay Boarddid hold down androllback wages in
1972 and 1973 and the Dunlop committee was intensely involved in
slowing wage settlementsin the constructionindustryduringthe same
5. Ibid., pp. 207-41; and Arthur M. Okun, Prices and Quantities.
595
George L. Perry
Table 3. First-Year Wage Adjustments and Average Hourly Earnings
around Recessions, Selected Periods, 1969 through First Half of 1983
Percent
First-yearwage adjustmentsa
Year
Manufac- Construction
turing
Other
1969
7.9
13.1
9.6
1970
1971
1972b
1973b
8.1
10.9
6.6
5.9
17.6
12.6
6.9
5.0
14.2
12.2
8.2
6.0
1973b
5.9
5.0
1974
1975
1976
1977
8.7
9.8
8.9
8.4
11.0
8.0
6.1
6.3
1979
1980
6.9
7.4
8.8
13.6
1981
1982
1983,first
7.2
2.8
half
-1.9
Percent of adjustments
with no change or
decline
Average
Manufac- Construction
turing
hourly
earnings
3
6.7
1
2
1
n.a.
n.a.
6
2
6.6
7.2
6.2
6.2
6.0
1
2
6.2
10.2
11.9
8.6
8.0
1
3
3
1
8
11
4
8.0
8.4
7.2
7.6
7.6
9.5
*
*
*
*
8.0
9.0
13.5
6.5
9.8
4.3
13
52
1
15
9.1
6.8
1.6
5.4
72
60
4.9c
*
*
*
Source: BLS, Current Wage Developments, various issues.
n.a. Not available.
* Less than 0.5 percent.
a. Settlements covering one thousand or more workers.
b. Year of wage controls.
c. Change from one year earlier.
period.6 That strategy was an alternative to the present attempt at
disinflationthrough monetary restraint alone. It has been criticized
becauseit distortedrelativepricesin the economyandbecause aggregate
restraint did not accompany the direct controls. Ultimately it was
abandoned under the inflationarypressures of the supply shocks of
1973-74.
A cleaner comparisonis availablebetween the last two recessions.
First-yearunionwage adjustmentsjumpedup in 1974-75when controls
ended in an environmentof double-digitinflationrates stemmingfrom
the food and energy price explosions. Thereafter, they slowed only
6. The committeechairedby JohnT. Dunlopwas formallyknownas the Construction
IndustryStabilizationCommittee.
Brookings Papers on Economic Activity, 2:1983
596
slightly in manufacturingand substantiallyoutside manufacturing.By
comparison,first-yearwage adjustmentshave slowed dramaticallyin
recent quarters,to the point of negative average adjustmentsin manufacturingin the firsthalfof 1983.
Table 4. Employment Changes of Production Workers from Prerecession Year,
Three Cycles, 1970 through First Half of 1983
Percent
Manufacturing
Cycle and
year
1970-73
1970
1971
1972
1973
Con-
Total
Total
Durable
goods
struction
private
sector
-4.9
- 8.3
-6.9
- 11.4
-0.7
2.0
-0.1
-0.1
-4.9
-6.9
0.5
0.9
8.1
3.6
13.0
8.3
1974-77
1974
- 1.3
-0.8
- 3.3
1.2
1975
1976
1977
- 12.1
- 8.1
-4.7
- 13.4
-9.3
-4.8
- 17.5
- 17.4
- 11.3
-2.3
1.3
5.7
1980-83
1980
1981
1982
-5.7
-6.9
- 15.1
-7.3
-8.9
- 19.3
-4.0
-8.8
- 14.6
0.0
0.9
- 1.3
- 18.0
- 22.9
- 22.1
-2.7
1983, first half
Source: BLS, Employment and Earnings, various issues.
Thedifferencebetweenthese adjustments,whichcover majorunions,
and broadermeasuresof economy-widewage movements, such as the
averagehourlyearningsindex shown in the table, shows that the wage
slowdown has been unusuallygreatamongmajorunionsthis time, with
union wages in manufacturingand constructionunderespecially great
pressure. Indeed, the concentrationof wage pressures is even greater
thanthe wage adjustmentsrevealbecause of the compositionof changes
withinall first-yearadjustments.In 1982,52 percent, and in the firsthalf
of 1983, 72 percent, of the wage adjustmentsin manufacturingrepresented either no change or a decline in wage rates. What had been a
negligiblephenomenonin manufacturingindustriesin the past became
a dominantone in thisrecession. Inthe firsthalfof 1983anunprecedented
60 percentof wage adjustmentsin the constructionindustryshowed no
changeor decline.
George L. Perry
597
These unusualwage developmentshave correspondedwith the exceptional and prolonged employment declines shown in table 4. In
durablegoods industries,where the most publicizedwage concessions
have occurred,productionworkeremploymentfell 24 percentbetween
1979and the firstquarterof 1983.Over a comparableperiod startingin
1973, the decline was only 8 percent. Even at its low point in 1975,
employmentin durablegoods industrieswas only 13 percent below its
1973level. Thiscorrespondenceof extremewageweaknesswithextreme
employment weakness does not support the idea that the present
slowdownin economy-widewage averagesis comingfrom any unusual
expectations about macroeconomicpolicy. Such policy expectations
should,presumably,influencewage decisions everywhere.
Forecasting Prices
An additionallook at the presentdisinflationis providedby examining
price ratherthan wage forecasts. Prices are more volatile than wages
and, partly because of this, may reveal changes in the economy's
responseto recentpolicy thatare not yet apparentin wage statistics. To
examine this possibility the disinflationperiod is next examined using
theequationforthefixed-weightGNPpricedeflatorpresentedpreviously
by Gordon and King.7This equation utilizes a distributedlag of past
inflationrates, ratherthan explicit normshifts, to capturelong-runand
expectationaleffects.
The main source of instabilityin the Gordon-Kingequationappears
to come from three variables capturingthe effects of certain relative
prices, includingthe prices of foreigntradedgoods: the relative prices
of food and energy, the relativeprice of total imports,and the effective
exchangerate. In view of this instability,threeformsof the equationare
used, with each estimatedfor the originaldata period, 1954:2to 1980:4
(shown in footnote a of table 5). In additionto the independentvariables
shown, each equationincludesvariablesfor the minimumwage rate, the
effective social securitytax rate, and dummyvariablesfor the periodof
Nixon price controls. These had little or no effect in the forecast period
and are omittedfor simplicity.
Equation 5-1 is the previously published Gordon-Kingequation,
which is reestimatedhere with the recently revised nationalaccounts
7. Robert J. Gordon and Stephen R. King, "The Output Cost of Disinflationin
TraditionalandVectorAutoregressiveModels,"BPEA, 1:1982,pp. 205-42.
598
Brookings Papers on Economic Activity, 2:1983
Table5. Predictedand ActualInflationRates, Fixed WeightGNP Deflator,
SelectedPeriods,1980:4through1983:2
Percent
1982:4-1983:2
1980:4-1981:4
1981:4-1982.:4
Actual change
8.8
5.1
3.9
Forecast
Equation5-1
Equation5-2
Equation5-3
6.2
7.4
8.4
4.2
4.3
5.6
3.1
2.8
4.5
Errors(actualminus
forecast)
Equation5-1
Equation5-2
Equation5-3
2.6
1.4
0.4
0.9
0.8
-0.5
0.8
1.1
-0.6
Source: Estimations provided by Robert J. Gordon, as described in the text, based on Robert J. Gordon and
Stephen R. King, "The Output Cost of Disinflation in Traditional and Vector Autoregressive Models," BPEA,
1:1982, pp. 205-42.
a. The coefficients on the main independent variables in the price equations are shown below; all are logs or
changes in logs and all are significant at the 5 percent level. For simplicity, some additional variables of little or no
importance in the forecast period are not shown. See the first equation in table 2 of Gordon and King, p. 218, for
the detailed specifications. The dependent variable in each case is the change in the log of the fixed weight GNP
price deflator.
Independentvariables
Lagged inflation
Equation
1954:21966:4
1967:11980:4
Output
ratio
Productivity
growth
5-1
5-2
5-3
0.92
0.97
0.94
1.04
1.08
1.10
0.36
0.35
0.36
-0.15
- 0.23
-0.21
Food
and
energy
prices
0.44
0.10
0.21
Inport
prices
Exchange
rate
Standard
error
of estimation
0.06
0.10
0.08
-0.13
- 0.08
...
0.803
0.862
0.857
data. Equation5-2uses the exchangerateover the 1961-80periodrather
thanjustthe 1975-80periodused in 5-1. Equation5-3dropsthe exchange
rate variablealtogetherand relies on food and energy prices and total
importprices to captureforeigntradeprice effects.
The actual inflationrates in the postsampleperiod and the dynamic
forecasts fromthe three equationsare comparedin table 5. As shown in
the wage equation of table 2, prices rise faster than predicted in the
earliestperiodshown, in this case 1981.Both 5-1 and 5-2 reveal that the
inflationslowdownhas been, if anything,disappointinglysmallthroughout. Only equation5-3, which omits the exchange rate and relies on the
variablesfor relativeimportpricesandfood andenergypricesto capture
the effects of foreignprices, predictsmore inflationthan occurred, and
its overpredictionsare very small and do not occur until 1982.Thus in
the Gordon-Kingprice equation there is no evidence supportingthe
George L. Perry
599
Table 6. Predicted Effect of Individual Variables on Inflation Rate, Selected Periods,
1980:4 through 1983:2
Percentage points
Equation
1980:41981:4
1981:41982:4
1982:41983:2
Outputratio
5-1
5-2
5-3
- 1.4
- 1.5
- 1.5
- 3.1
- 3.2
-3.3
- 3.8
-4.0
-4.1
Productivitygrowth
5-1
5-2
5-3
0.1
0.1
0.1
-0.2
-0.3
-0.3
-0.3
-0.3
-0.3
Food and energy prices
5-1
5-2
5-3
-0.3
-0.1
-0.1
-0.2
0.0
-0.1
-0.9
-0.3
-0.5
Importprices
5-1
5-2
5-3
0.0
-0.1
0.0
-0.6
- 1.0
-0.9
-0.8
- 1.2
- 1.2
Exchangerate
5-1
5-2
-1.6
- 0.7
-1.3
- 1.0
-0.6
- 1.3
5-3
...
...
...
5-1
5-2
5-3
- 1.8
-0.8
-0.2
- 2.2
-2.1
- 1.0
- 2.3
-2.9
- 1.6
Variable
Addendum
All relative price
variables
Source:Same as table5.
credibilityhypothesisin connectionwiththe presentpolicyof disinflation
or pointingto a downwardshift in inflationnormsgreaterthanhistorical
equationswould predict.
Table 6 identifies the contributionof the individual independent
variablesto the slowdownin inflationpredictedby the Gordonequations.
The table shows how much inflationshouldhave changedas a result of
the independent variables taking on their actual values since 1979:4
rather than the unemployment rate staying constant at 6 percent,
productivityrising at its trend rate, and all relative prices remaining
unchanged. The estimates shown allow for both the direct effect on
inflationof the changein each variable,and its effect actingthroughthe
lagged inflationrate which is endogenous in the projectionperiod. By
the first half of 1983the shortfallof outputbelow trendaccountedfor a
slowdown in inflation of about 4 percentage points in each of the
600
Brookings Papers on Economic Activity, 2:1983
equations.This is the minimumeffect that can be attributeddirectlyto
the disinflationarypolicy.
The effects of the three relative price variables account for an
additional2.9 to 1.6 percentage points of disinflation,depending on
which equationis used. If these effects too are creditedto the policy of
the period, thatpolicy has reducedthe inflationrateby an estimated6.9
to 5.6 percentagepoints, producingthe forecasterrorsshown in table 5.
Food and energy prices had alreadyslowed during1980. So almost all
these relativepriceeffects come fromimportprices and, when it is used,
the exchange rate variable. Thus part of these anti-inflationbenefits
come at the expense of correspondingpro-inflationarypressureson the
countrieswhose dollarexchangeratesdeclined.And they are reversible
if the exchangerate moves back towardits level in earlieryears.
The Inflation Process and Policy
Monetarypolicymakerssince 1979have shown historicallyunusual
determinationto fight inflationand willingnessto raise unemployment
in the process. Althoughmany issues remain,this experience has shed
light on some questions about the inflationaryprocess and the costs of
disinflation. Developments have certainly not supported the central
promiseof someclassicalmodelsthatonce steadydisinflationarypolicies
are in place, disinflationwill take place with little output loss. The
disinflationthatoccurredwas not exceptionallyprompt,andwhenwages
finally did slow by unusually large amounts in some manufacturing
industries and construction, that development was accompanied by
unusuallylargedeclines in employment.
CREDIBILITY
VERSUS
AUSTERITY
No measure exists of what private decisionmakersthought about
policy aims in this period. Because of this, one could argue that the
promisedbenefits of credibledisinflationarypolicy have not been realized because the credibility of anti-inflationpolicy has never been
established. However, the FederalReserve has certainlyworked hard
to establish its credibility.It persisted in its restrictivepolicies, implementedthroughmoney supplytargets,long afterits predecessorswould
601
George L. Perry
have turnedto fightingrecession. And in the springof 1983it already
moved to restrain the expansion only months after it began. The
commitmentto fightinginflationhas been more clear and crediblethan
ever beforein the postwarera.
Ultimately determinationshould pay off. The evidence about the
slowdown in inflationthat has finally occurredin 1983is still unclear.
But in my mind,the biggest surprisewill be if inflationdoes not slow by
morethanthe cyclical wage equationspredict.Wagenormshave shifted
upwardin the past under sustainedperiods of prosperityand cyclical
inflation, and they should shift downward now under the opposite,
sustainedcondition. In the currentperiod it is hard to distinguishthis
predictionfrom those of the credibilityhypothesis if the latter require
theactualexperienceof anextendedslumpto establishpolicy credibility.
Butin this case, it has no distinctiveimplicationsfor policy. Onlyif there
is an importantdifferencebetween credibilityandausterityis the former
of any special interest.
The predictionsfrom the Gordon-Kingprice equation(5-3) must be
interpreteddifferentlyfrom either of these views. That equation does
not hypothesize norm shifts, but does continuouslyshift the short-run
Phillipscurvethroughan accelerationistlaggeddependentvariable.One
interpretationof the equation'smodest underpredictionsof inflationin
recentquarters-in specificationsthatuse the full arrayof relativeprice
variables, including the exchange rate-is that the high elasticity of
laggedinflationexaggeratesthe speed and size of the reactionof actual
inflationto its recent past. Alternatively,if that specificationis correct,
disinflationin the recent period must be interpretedas disappointingly
slow. In any case, the predictionerrorsare not largeandlend no support
to the view that crediblepolicy has speeded up the disinflationprocess
and madeit less costly.
THE
LONGER
RUN
Some importantmessagesfor theoryandpolicy maynot be clearuntil
and unless policy allows a sustained recovery in economic activity.
Presumably,the implicithope in the present disinflationpolicy is not
only that inflationmightbe sharplyreduced, if not ended, but also that
high employment might then be resumed without returning to the
underlyinginflationratesof the pastfifteenyears. All the inflationmodels
Brookings Papers on Economic Activity, 2:1983
602
are at least partially optimistic on this point, though for different
underlyingreasons. But they do offer somewhatdifferentprospectsfor
outputand inflationin the longerrun. The new classical models predict
that, underthe appropriatepolicy regime,the cyclical relationbetween
inflationandunemploymentwill changeandmaintainedfullemployment
will be compatible with price stability. The absence thus far of the
disinflationpredictedby these models casts doubt on this predictionas
well.
The inflation-outputrelation may, in fact, not be sensitive to the
policy regime.As the historicalstudiesby Gordonandby Schultzehave
shown,these coefficientshavebeen surprisinglyinvariantto the differing
policy regimes that have prevailedduringthis century.8If it turns out
that inflationcannotbe controlledsufficientlywithoutcontinuallyholding back expansion, then other programsfor dealingmore directlywith
inflationmay have to be put back on the policy agenda.
RECENT
POLICY
Whateverlessons emergefor the longerrun, the decision to turnto a
severely restrictivemonetarypolicy in 1979mustbe viewed againstthe
economic risks thatthen existed. Energyprices were risingvery rapidly
in 1979and 1980.Inflationarypsychologyhadapparentlyspreadto other
volatile markets, and wages were accelerating.Even with the turn to
restrictivepolicies andthe recession thatensued, wages rose fasterthan
predictedin 1980.Withoutthatpolicy shift, fasterinflationwouldalmost
surely have become imbeddedin the economy, and very likely a still
faster and morestubbornwage normandunderlyinginflationratewould
have resulted.
Once that accelerationwas interrupted,as it was by 1981, the costs
and benefitsof continuingthe restrictivemonetarypolicy became more
debatable.Reasonableobservers can and do differabout how long and
how much to depress the economy in order to reduce inflation. The
results presented here suggest that we have gotten about what would
have been predictedfrom past experience,just more of it because this
recession was longerand deeperthanothers.
8. Robert J. Gordon, "A Consistent Characterizationof a Near-Centuryof Price
Behavior," American Economic Review, vol. 70 (May 1980, Papers and Proceedings,
1979),pp.243-49;andCharlesL. Schultze,"SomeMacroFoundationsforMicro
Theory."