The Effects of External Inflationary Shocks (Brookings Papers On

JAMES L. PIERCE Staff, Board of Governors of the Federal Reserve System and

JARED J. ENZLER Staff, Board of Governors of the Federal Reserve System

The Effects of External Inflationary Shocks

THE ECONOMYIS ALWAYSVULNERABLE to a variety of external influences or shocksthat have important impacts on income,employment, and prices. Whilethese externalshocks are unforeseeableand unavoidable,economic policy must somehowdeal with theirconsequences. Lately an alarmingnumber of upwardjolts to priceshave come from sourcesbeyond the normalinteraction of production,wages, and prices. Onewas the relativedecline in the valueof the dollarfollowing the abandon- ment of the systemof fixedexchange rates, which raisedthe pricesof importedgoods and contributedto the risein farmprices as exportscompeted with domesticconsumption. A numberof otherevents shook the economy at about the same time. Crop failuresin the Soviet Union resultedin a giganticsale of Americangrain. The Peruviananchovy catch mysteriously

Note: The views expressedin this paper are our own and do not necessarilyreflect those of the Federal Reserve Board or its staff. We want to thank members of the Brookings panel for their many constructivecomments on an earlier version of this paper.

13 14 Brookings Papers on Economic Activity, 1:1974 dropped,contributing to a worldwideprotein shortage and higherprices for feedstocks.Then floods in the Midwestdestroyed part of the soybean crop, drivingfeedstock prices still higher.Late in 1973,a cartelof oil pro- ducersgot the upperhand over the buyercartel and forcedfuel pricesdras- tically upward.Finally, the dismantlingof domesticprice controls may undo whateverdownward pressures the programhad on the price level. Becausethese inflationarysurges have substantialeffects on real income and on relativeprices, their implications for economicstabilization policy deserveexamination. The consequencesof the recentupward shocks to priceswill be analyzed underthe assumptionthat they are permanent,even though they may be overshadowedby downwarddisturbances from other sources, or evenpartly reversethemselves (anchovies may reappearoff the coast of Peru and soybean fieldsmay yield a bumpercrop, for example).This assumptionper- mits analysisof the longer-runconsequences of externaldisturbances that are not quicklyreversed. Table 1 presentsrecent growth rates of selectedprice deflators from the nationalincome accounts. Farm and importprices have risenat a tremen- dous rate; many of these increaseshave found their way into consumer prices,which have risensubstantially relative to the nonfarmbusiness deflator.The extentto whichmonetary policy has contributedto the current high ratesof inflationis a subjectof considerablecontroversy. We will not enterinto this debate,but ratherwill deal with the impactof externalinfla-

Table1. Changesfrom Preceding Quarter in SelectedPrice Deflators, NationalIncome Accounts Basis, 1972 and 1973 Seasonallyadjusted annual rate of change in percent Gross Consumer Nonfarin Yearand national nondurables business, Farm quarter product and services adjusteda business Imports 1972:1 5.7 3.0 5.0 20.6 3.8 2 1.6 2.9 0.8 19.2 15.4 3 2.7 3.2 1.8 28.7 6.7 4 3.3 4.3 2.8 23.2 7.9 1973:1 6.0 6.9 4.3 50.5 12.1 2 7.2 8.7 5.4 75.8 40.9 3 7.0 7.2 5.0 107.4 20.0 4 8.7 10.0 7.1 4.2 35.6 Source: Official data from the U.S. Bureau of Economic Analysis. a. Nonfarm business price deflator adjusted for federal indirect business taxes. James L. Pierce and Jared J. Enzler 15 tionaryshocks to the economy and with the specialproblems that these disturbancespose for monetarypolicy.' An importantissue for monetarypolicy is the type of responsethat is appropriateto externallyinduced surges in inflation.Some observershave arguedthat monetary policy should respond vigorously to theseprice surges, whileothers believe that policy should "accommodate" them, at leastin the shortrun. This paperargues that the conceptof accommodationis a slip- peryone in this context.Furthermore, the effectsof pricedisturbances may dependin part on the source of the shock, and so, therefore,may the appropriateresponse of monetarypolicy. To examinethese matters, we have used a modifiedversion of the MPS (MIT-Penn-SSRC)quarterly econometric model to obtain quantitative estimatesof the effects on individualsectors and on the economy as a whole of imposingdomestic price controls and of a rise in foreign oil prices.The modelhas been adjustedand modifiedin waysdescribed below to make it a more suitabletool for analyzingthe specialproblems under study. The followingsection attempts to assessthe initialimpact of externalin- flationaryshocks on individualsectors of the economyconsidered in isolation. The sectorsmost relevantfor the analysisinclude the moneymarket, consumptionexpenditures, and the interactionof wages and prices. The next section reportsa simulationof the full model to study the effectsof these shockson the economyas a whole, and concludeswith a discussion of the possibleimplications for monetarypolicy.

InitialImpacts of Shocks

In this section, we start with a general,qualitative, discussion of the initial impactsof the shocks on each sector, and then turn to a more de- taileddiscussion of theseimpacts on the relevantsectors of the MPSmodel.

A BRIEF OVERVIEW First we considerthe impact that a sharp increasein the price of an importedcommodity-in this case, oil-has on the demandfor money.

1. While fiscal policy can, and probably should, play an important role in com- bating the effects of externalshocks, this paper will deal only with monetarypolicy. 16 BrookingsPapers on EconomicActivity, 1:1974 In keepingwith accepted doctrine, we shallassume that the public'sdemand for money varies directlywith the volume of transactionsand inversely with marketinterest rates.2 But this discussioncalls attentionto the well- known problemsof an appropriatemeasure of transactions. Gross nationalproduct, which is often used as a proxyfor transactions, ignoressuch potential complications as the level of integrationof the economy and the absolutelevels of importsand exports.Thus, for example,if import prices rise relative to other prices, the GNP deflatorwill remain unchanged.The rise in the prices of consumptionand investment goods containingimports will be offsetby the correspondingrise in prices of imports themselves.Thus, both nominal and real GNP remain unchanged. However,importers and purchasersof final consumptionand investmentgoods will wish to hold largermoney balances at giveninterest rates and nominalGNP to supportthe largervalue of their transactions. Thus, whenimport prices rise, the correspondingrise in transactions-and hence moneydemand-are missedif GNP is used as the measureof transactions. If the pricesof exportsrise relative to otherprices, the GNP deflatorand nominalGNP willboth rise.The demandfor moneyshould rise with GNP, but probablyless than when domesticprices rise; domesticpurchasers of finalinvestment and consumptiongoods feel no need to hold largermoney balancesbecause the valueof theirtransactions has not risen. In analyzingthe increasein oil prices,we will use GNP plus importsas our measureof transactionsdemand, and will provide evidence on the importanceof thisarbitrary adjustment. While this measureprobably over- states the impacton transactionsof an increasein pricesof importedoil, it sets an upperlimit to thateffect. Because we do not considerexport prices, no adjustmentsneed be made to the transactionsvariable for this factor. The initialeffects on the monetarysector of an increasein the pricesof importedoil follow from this discussionof transactionsdemand. Assume that the volumeof real transactionsis unchanged,at least initially,by a rise in importprices (or, more realistically,that any initial drop in real transactionsis small relativeto the price increase).The nominalvalue of transactions(GNP plus imports)increases, as just described,augmenting

2. For an excellentdiscussion of money demandfunctions, see StephenM. Goldfeld, "The Demand for Money Revisited," Brookings Papers on Economic Activity (3:1973), pp. 577-638. Hereafterthis documentwill be referredto as BPEA, followed by the date. JamesL. Pierce and JaredJ. Enzler 17 the demandfor moneyat giveninterest rates and GNP. Thus,other things equal, interestrates will rise if the money stock is fixed. The analysisof the initialimpact of the increasein the priceof imported oil on the real volume of consumptionexpenditures is straightforward. Althoughhigher import prices do not raisethe GNP deflatorthey do raise the pricesof consumptiongoods. This rise reduces the realvalue of disposable incomeand of householdnet worth,since the nominalvalues of these variablesare unaffected.Thus, initialiythe priceincrease will reducereal consumptionexpenditures. Nominal consumptionexpenditures are likely to rise at firstbecause consumers wili adjustreal purchasesslowly to their reducedreal incomesand net worth.Thus, nominal saving drops initially. The direction,if not the size, of the initial impacton wages and prices also can be anticipated.The surgein importprices wili make the goods that workersbuy moreexpensive, putting pressure on wages,and in turn tend- ing to raisedomestic prices. The pricesof import-competinggoods shouldalso respond.An increase in the worldprice of an internationallytraded commodity like oil would, in the absenceof regulation,probably spur the sameincrease in the domestic price.Prices of energysubstitutes, such as coal, shouldrise as demand shiftsfrom oil. Domesticenergy producers may find it profitableto export theirproduct, thus furtherbidding up the oil prices. These increasesin domestic prices wili reinforcethe impacts of new prices of imports.In particular,the transactionsdemand for money will expand(this time becausenominal GNP rises),putting additional upward pressureon interestrates; real consumptionexpenditures will drop further due to the additionaldecline in real disposableincome and wealth; finally, wagesand prices wili tend to risein responseto theincrease in domesticprices. A risein domesticoil priceswill not be as deflationaryas a risein import prices of equal size because corporate profits and investmentwill be greater,and the stock marketwill not experienceas seriousa declineas results when profitsgo to foreigners.Because of these differences,the effects of a rise in domesticoil pricesare qualitativelymore like those of decon- trollingdomestic prices when they are underpressure to rise-the second economicevent consideredin this paper.

RELEVANT SECTORS OF THE MPS MODEL Theimpacts described above can be givensomewhat greater precision by analyzingthem in terms of the sectors of the MPS model. The demand 18 BrookingsPapers on EconomicActivity, 1:1974 equationfor the public'sholdings of demanddeposits, as modifiedby the additionof imports,is shownin equation(1):

(1) In DD _- 0.28 In DD-1 (PjY) + (Pimim) (PyY)+ (Pimim)

- 0.06 In RTB - 0.12 In RSD

+ 0.08 n RDIS -034 In RDI&1_ NN-005. The demandfor nominal demanddeposit balances(DD) varies directly withreal GNP (y), realimports (im), the GNP and importdeflators (P, and Pim),changes in the FederalReserve discount rate (RDIS), and population (N). The demandvaries negatively with the Treasurybill rate (RTB)and the savingsdeposit rate (RSD). The basic notion in this formulationis that firmsand householdsdesire demand deposit balances for transactionspurposes. In equation (1), changesin the pricesof both domesticaliyproduced and importedgoods have a powerfulimpact on moneydemand.3 A 1 percentrise in a weighted averageof domesticand importprices produces a 0.72 percentrise in deposit demandin that quarterand a 1 percentrise in the long run. Thus, duringrecent periods when import prices have risenby as much as 10 percent in a quarter,the demand for money has increasedsignificantly.4 For a givenmoney stock thesesurges in moneydemand exert upward pressure on interestrates. The consumptionfunction in the model is

= (2) N b ai + Edp N + 0.054 N i-=O =Oi __ -cnVi PCOnlN' where bo = 0.10, E = 0.66; and do = 0.016, di = 0.054. i i Real per capitaconsumption (con/N) dependson a distributedlag on real

3. A similar adjustmentfor imports is not necessaryfor the currencyequation because it uses the value of consumer expenditures,which include import prices, as the transactionsvariable. The currency equation is not shown because, except for this difference,it is basicallyof the same form as the demand deposits equation. 4. A 10 percentincrease in the price of imports would raise the value of GNP plus imports by about 1 percent. Thus, given interest rates, deposit demand will rise by 1 percentin the long run. Currencydemand also has a unitary elasticity with respect to prices in the long run, but it is based on personal consumption expenditures,which constitute about two-thirds of GNP. Thus, currencydemand will rise by 1.5 percent in this case. JamesL. Pierce andJared J. Enzler 19 per capitadisposable income (Yd/N), a distributedlag on the common stockportion (VCW) of realper capitaconsumer net worth,and the value of the remainingportion of net worth(VOW) at the beginningof the quarter. The coefficientsindicate that a rise in the price of consumptiongoods (PC,,)will have a sizableeffect on real consumptionthrough the reduction in real disposableincome and real net worth.The veryshort lag for VOW is troublesomeand its implicationswill be analyzedbelow. Wage-priceinteractions are a key vehiclethrough which externalinfla- tionaryshocks are transmittedto the economy.We shall discussthe wage- price sector in considerabledetail becausewe have reestimatedthe wage equation,and its propertiesare ratherdifferent from those in the conven- tionalMPS model.5We shall also attemptto isolatethose variablesin the wage-pricesector that will be most importantfor the outcomeof the full- economysimulations that follow.

THE WAGE EQUATION The wage equationis a modifiedPhillips curve in which the percentage changesof wages (PL), measuredby compensationper manhourin nonfarm privatedomestic business, varies inverselywith the unemployment rate (UL) and directlywith the percentagechange in consumptiongoods prices(Pc,0), the level of social securityand unemploymenttaxes (SS), and the percentage change of the minimum wage (Pmin). The estimated equation is 1 UL (3) -0PL 0031 0017 +0004SS+0.009 /pnin PL_ = .31J UL1 Pmin-1

0.9912 1 O.87 APconl 00 + 0 j99 3 O. 0.87i P 1 0.007D - 0.00005, 5', 0.87 con_i-10 wherewage increasesare measuredas percentageannual rates and D is a dummyvariable to capturethe wage freeze in 1971:4 and wage release in 1972:1. The initial impact of the increasein the price of consumptiongoods (Pcon)on the growthof wagesis smallbut in the long run consumerprices will rise almost proportionatelyto wages, given the unemploymentrate.

5. George de Menil and JaredJ. Enzler, "Pricesand Wages in the FRB-MIT-Penn Econometric Model," in Otto Eckstein (ed.), The Econometrics of Price Determination (Board of Governorsof the Federal ReserveSystem, 1970). 20 Brookings Papers on Economic Activity, 1:1974 Thus, the wage equationused in this study does not allow a significant tradeoffin the long run betweeninflation and the unemploymentrate. Be- cause the long-runcoefficient on the rate of changeof pricesis essentially unity, attemptsto reduce unemploymentbelow its "natural"rate ulti- matelywill fuel an explosiveincrease in the rateof inflation.This property, and some othersthat will be discussedbelow, calis for a justificationof this wageequation. In the light of recentwork by such researchersas Robert Gordon and George Perry,6our equationmay seem to lag the currentstate of the art. Workingwithin the contextof a largeeconometric model, however,poses problemsfor theconstruction of a singlesector. Thus, explanatory variables endogenousto the economicsystem cannot be employedunless equations for thosevariables can be produced.For this reasonthe fixed-weightunem- ploymentrate by age and sex that Perryuses as the labor marketvariable is difficultto incorporate.We encountersimilar difficulties with Gordon's formulationof the dependentvariable in the wage equationas a serieson hourly earningscorrected for changesin overtimeand interindustryem- ploymentshifts. Equallyimportant is the needto strikea compromisebetween the ability to explainan individualsector and the abilityto track the entiresystem. Althoughmost equationsin the MPS model are estimatedby single-equation methods,the best-fittingform of an estimatedequation clearly is not always the ultimatebest choice. The effectsthat variousalternative estimatesof a particularequation have on the trackingproperties of the model and on its responsesto various kinds of shocks are importantcriteria. Theseproperties can be tested only by simulation. One of the most vexingquestions in this regardis the size of the long-run coefficienton the priceterm in the wage equation.Until recentlythe best- fittingequations of the type presentedhere had estimatedlong-run coeffi- cients substantiallyless than unity. Accordingto these equations,a con- siderabletradeoff between unemployment and inflationwas possibleeven in the long run.Once the 1969-71experience is addedto the sampleperiod, however, the estimatedprice coefficientusually increasesmarkedly to about unity. Furthermore,the averagelag for the price term in most of these equationsis remarkablyshort. Gordon has demonstratedthat the wage equationsusing the concepts

6. George L. Perry, "ChangingLabor Marketsand Inflation,"BPEA (3:1970), pp. 411-41; Robert J. Gordon, "Wage-PriceControls and the Shifting Phillips Curve," BPEA (2:1972), pp. 385-421. James L. Pierce and Jared J. Enzler 21 developedby Perry,noted above, do not yield a high long-runcoefficient for pricechange even whenestimated through 1970.7 In the courseof estimation,we triedto approximatethe effectof Perry'shypothesis concerning compositionof the unemploymentrate by addingto the wage equationa numberof variablesdesigned to capturethis effect.The proportionof teen- agersin the adultpopulation was testedand consistentlyyielded the wrong sign. The proportionof personsaged 20-24 yearshad the correctsign but was statisticallyinsignificant, and the coefficientwas so smallthat this variable could accountfor only negligibleshifts in the Phillipscurve. We tried enteringchildren under 5 yearsas a percentof the adultpopulation, on the theorythat when the birthrate is high, femaleparticipation rates are low and a givenunemployment rate shouldput greaterdownward pressure on wage changes.This variableconsistently had the right sign but was very weak. Most important,none of these variables,either alone or in combination,had a noticeableeffect on the pricecoefficient. We wereunable to estimatea wage equationwith a long-runcoefficeint below 0.8. We also tried Gordon's suggestionof a flexible price coefficientthat dependson the rate of inflation.8We attemptedto estimateequations with two pricevariables, a short distributedlag on the total rate of increasein consumerprices and a longer distributedlag on the excess of inflation abovesome threshold.This formulationintroduced a kink in the long-run Phillipscurve. When the thresholdwas high, the sum of the price coeffi- cientsalways exceeded 0.8 and the thresholdvariable itself was not significant. Whenthe thresholdwas lowered,the thresholdvariable became sig- nificantand the sum of the coefficientsrose to approximatelyunity. Al- thoughthis specificationworked well in estimation,the full model proved to be highlyunstable when simulatedwith this sort of equation.When the long-runcoefficient on pricesis near unity and when the lag on pricesis short, errorsin otherequations quickly cumulate to largeerrors in wages and prices.9We then experimentedwith forcinglonger price lags into the wageequation and finallysettled on equation(3), whichuses a singleexpo- nentiallag based on 0.87; that is, roughly16 percent- ( 0.87 100 12 1~0.87i) 7. Ibid. 8. Ibid. 9. This unstableprocess will occurbefore the higherprices can force up interestrates and unemployment sufficiently to correct the original disturbance. 22 BrookingsPapers on EconomicActivity, 1:1974 -of an increasein P60nin the previousquarter is translatedinto a wagein- creasein the currentquarter and the weightsdecline exponentially there- afteras powersof 0.87. It was possibleto introducethe longerlag with a modestloss in fit. The best-fittingequation we estimatedhas a standard errorof about 1.3 percentper quarterat annualrates comparedto 1.4 in equation(3). Theseerrors are only slightlylarger than those foundin other studiesdespite the fact that the equationuses the relativelyerratic series on compensationper manhour. The average lag implied by the 0.87 coefficientis 4.9 quarters. This modificationproduced a wage equationthat seemedmore in ac- cordancewith our observationson the dynamicsof wage determination. We werealso comfortedby the improvementin the full-modelsimulations when this equationwas used, in the sense that the resultsaccorded better with our a prioriexpectations concerning the reactionof the economyto shocks.The longerlag on pricesgives the systemtime to reactand to cor- rectany errors in simulatedprices. Not surprisingly,as thelength of the price lag is extended,the size of the coefficienton the inverseof unemploymentis increased.In the estimatedequations using short price lags, the unemployment coefficientis about one-halfof that reportedin equation(3). Thus, the size of the coefficientand the length of the price lag have offsetting effectsand the reactionof the wage-pricesystem to changesin the unemploymentrate is not very differentwhen the length of the lag on the price term is altered.Lengthening the lag strengthensthe initial impact of a givenchange of the unemploymentrate on the rate of wageincreases, but wageswill reactmore slowly to the priceincrease. Lengthening the lag does moderatethe volatilityof the wage-pricesystem in responseto other influences,such as an increasein unfilledorders or productivity.Thus, the longerlag form makesthe full model more stable. This discussionemphasizes the uncertaintiesinherent in modelingthe wage process.The structureof the wage equationis obviouslyimportant to the questionwe are addressing.Yet a considerablenumber of hypoth- eses aboutthe structureof the wage-pricesector fit the facts about equally well, but carrydifferent implications for the impactof changesin the full model of the economy.This is obviouslya sourceof concern. Figure 1 showsdynamic simulations of the percentagechanges of wages using actual values of the explanatoryvariables, including prices, in the wageequation. A four-quarterannual percentage change, 100(PL,-PLX4)/ (PL,-4), was constructedin orderto smooth out fluctuationsin the actual 0-

p 4 ~ ~ ~ ~ CC.~ p ~ ~ ~ ~ ~ ~ C

prz~~~~~~~

C~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~C

op.4~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~4 0Pk 24 BrookingsPapers on EconomicActivity, 1:1974 data that would make the chart difficultto read. The wage equationper- formsquite well overthe period,but, surprisingly,shows little or no effect of the wage-pricecontrol program once the slowdownin pricesis allowed for, as it is in this simulation.This findingsuggests that the only influence controlshad on wageswas the indirectif powerfulone assertedthrough the downwardpressure on prices.10This findingis not simply the result of estimatingthe wage equationover the periodthat includedthe pricecontrol episode.In the simulationsof dozensof differentwage equations, either using dummyvariables for the controlperiod or endingthe sampleperiod in 1971:2, we have neverfound any evidenceof a directeffect on wages otherthan in the freezequarter 1971:4.

THE PRICE EQUATION The priceequation in the model is

(4) ln Pf = 0.2901 ln PL + 0.073 EPD -0.040 UOPD-1

- 0.001 TIME - 0.0075 E ln i=0 Mul_

-0.013 In pPr + 0.710 ln Pn,1

The equationcan be viewedas expressingprices (private nonfarm deflator net of excisetaxes, Pnf) as a markupover labor costs (PL).11The markup variespositively with the ratio of unfilledorders of producers'durables to shipments(UOPD/EPD), whichserves as a measureof capacityutilization; negativelywith a productivitytime trend;and negatively(but weakly)with cyclicalvariation in outputper manhour(ynf/MH) aroundits trend. The markupalso variesnegatively with the changein raw materialprices (Prm) as an impactvariable. The nonfarmbusiness sector now accountsfor 85 to 90 percentof real GNP. Governmentis excludedbecause the "price"of governmentis reallya wage,and farmoutput is excludedbecause it is sub- ject to forces,such as weather,that we cannot explainin economicterms.

10. For a similarresult, see Robert J. Gordon, "The Response of Wages and Prices to the First Two Years of Controls,"BPEA (3:1973), pp. 765-78. 11. Since it determinesthe relationshipbetween prices and wages and since labor requirementsare determinedelsewhere in the model, the equation also determinesthe profit share. JamesL. Pierce andJared J. Enzler 25 The nonfarmdeflator is a value-addeddeflator; that is, it is calculatedas a weightedaverage of the pricesof businessoutput less a weightedaverage of the prices of the raw materialsbusinesses buy. For the businesssector as a whole, the raw materialinputs consist mainly of imports and farm products.If prices of these raw materialsincreases, and the cost is not passed along to final consumers,the nonfarmbusiness deflator actually falls. Becauseof this phenomenon,changes in raw materialsprices (Pm,) enterthe deflatorequation with a negativesign. This termhas only a tran- sitory effect,however, as these changesare eventuallypassed on to consumers.Finally, the equationhas a laggeddependent variable with a co- efficientindicating that if any of the determinantsof the pricelevel change, about 30 percentof the completeprice adjustment is madein one quarter. Figure2 showsa dynamicsimulation of the priceequation using actual valuesof the right-handvariables, except for the laggeddependent variable whichtakes on the predictedvalue from the previousquarter. The equation performsquite well until 1971:3, whenit beginsto simulateprices too high, whichsuggests that the pricecontrol program has had a dampeningeffect on the pricelevel. It shouldbe stressedthat the simulationdoes not capture the full effectof the pricecontrols on the pricelevel. We arguebelow that if pricecontrols had not been imposed,the higherprices that would have ensuedwould have led to higherwages and thento evenhigher prices. This interactionof wagesand pricesconsiderably enlarges the effectof controls that appearsin Figure2. Closingthe wage-pricesector calls for threeidentities:

(5) Pcnd UcndPnf

(6) Pcd = UcdPnf (7) Pp _cndPCnd + 0.OlPCd(WCD + O.O1RCB.KCd) con - Cnd + (WCD + 0.0379Kcd) Equation(5) relatesthe deflatorfor consumernondurables and services (Pcnd)to the nonfarmbusiness deflator by the ratio U6nd. Equation(6) does the samefor the consumerdurable deflator (PCd) using the ratio UCd. Actual historicalvalues for these ratios were used in each quarter.We do not attemptto explainthese ratios, since they dependin large part on three items for whichbehavioral equations are difficultto construct:the priceof importedgoods, the priceof farmproducts, and the federalexcise tax rate. The consumptionprice deflator (P0on) is definedas a weightedaverage of the price of nondurablesand the price of the servicesof durables.It is 00

/~~~~~~~~~~~~~~~~~~~~~~~~~

*U * x

CU - , t t} q - o o Uz t James L. Pierce and Jared J. Enzler 27 determinedin equation(7) as the ratio of the sum of (1) current-dollar purchasesof consumernondurables and services(Cnd Pcnd) and (2) current- dollarconsumption of the servicesof the stock of durablegoods, to the sum of Cndand the consumption(in 1958dollars) of the servicesof the stock of durable goods. The term PCd(WCD+ 0.01RCB KCd)measures the imputed rentalservices from durablegoods in currentdollars, where WCD is the depreciationon consumerdurables and K,d is the stock of consumer durables,both in 1958dollars, and RCB is the currentinterest rate on corporate bonds. The expression(WCD + 0.0379KCd)measures the imputed servicesfrom durablesin 1958dollars, where 0.0379 is the averagevalue of RCB in 1958.

INTERACTION OF WAGES AND PRICES How does the wage-pricesector, considered in isolationfrom the rest of the full model, respondto an increasein consumptionprices? Wages will rise in responseto this shock, and spur a furtherincrease in prices.But if unemploymentis unchangedand if the long-runprice coefficientin the wageequation is less thanunity, these feedbacks will not accumulateindefi- nitely.In this case, the rate of priceincrease will eventuallyreturn to what it wouldhave been in the absenceof the shock, althoughthe level of prices will be higher.If, however,the long-runcoefficient is unity, this feedback mechanismwill continue without end, and the rate of inflationwill be permanentlyhigher, but by only a limitedamount even in this case. For inflationto increasewithout limit requiresa permanentreduction in the unemploymentrate. In orderto studythe adjustmentof the wage-pricesector in responseto an externalprice shock, we did two simulationsof the sectorfor the seven- year period 1967through 1973. In the first,the sectorwas simulatedas is. In the second, the ratio of the consumernondurable deflator to the nonfarm businessdeflator (U,fnd of equation5) was increasedby 1 percentin the first quarterof the simulation(raising PC0n by an annualrate of 3.44 percentthat quarter)and held at the new higherlevel to capturethe effect of a permanentrise in the price level of a commoditylike importedoil. The effects of this permanentincrease in Ucndon the rate of change of wages, the nonfarmdeflator, and PC0nare measuredby the differencebetween the two simulations,and are reportedin Table 2. The deflatorfor consumernondurables and services(P,0n) increases at a 28 Brookings Papers on Economic Activity, 1:1974 Table 2. SimulatedEffect on the Wage-PriceSector of a Rise in ConsumptionPrices, by SelectedQuarters Deviation from control simulationin annual percentrate of change Privatenonfarm deflatornet of Consumptionprice Quarterof Compensation excise taxes deflator simulation per manhour Pf Pcon 1 0.00 0.00 3.44 2 0.54 0.16 0.15 3 0.50 0.26 0.24 4 0.47 0.32 0.30 5 0.45 0.36 0.36 6 0.46 0.38 0.38 7 0.46 0.40 0.41 8 0.46 0.42 0.41 12 0.49 0.46 0.46 16 0.42 0.44 0.43 20 0.43 0.43 0.44 24 0.42 0.42 0.43 28 0.41 0.41 0.42 Source: Simulations in which the ratio of the consumer nondurable deflator to the nonfarm business deflator is held 1 percent above historical values in each quarter of the simulations. See text for detailed discussion.

3.44 percentannual rate in the firstquarter of the simulationwithout affecting compensationper manhour or domestic nonfarm prices, since Peon entersthe wageequation only with a lag. In the second quarterthe rate of wage changerises by about one-halfpercent (at annualrates) and prices by about one-sixthpercent. The feedbacksbetween these equationscon- tinue and eventuallythe rate of increaseof both prices and wages settles at a level about0.4 percentabove that in the simulationwithout the price disturbance. These results obtain, of course, only in a wage-price sector in isolation. Table 2 can be regarded as an approximation to what would happen in the full systemonly if monetaryand fiscalpolicy weremanaged in such a way as to offset the effects of the price disturbance on unemployment, the ratio of unfilled orders to shipments, and productivity. If policy were not so accommodating, the rise in prices would drive interest rates up, production down, and the unemployment rate up, thus reducing the upward pressure JamesL. Pierce andJared J. Enzler 29 on prices. In fact, if the money stock were not expanded,output would haveto remainat a reducedlevel until both the level and the rate of change of pricesreturned to the valuesthey wouldhave attainedin the absenceof the originaldisturbance in prices. Thereis reasonto suspectthat wagesmay be less responsiveto external priceshocks than this modelindicates. We do not know exactlywhat role the rateof changeof pricesplays in the wageequation. If it captureschanges in the bargainingposition of labor,a rise in importprices will have weaker effectsthan one in domesticprices because the valueof the marginalproduct of labor will not have risen.Furthermore, to the extentthat the coefficient representswage earners'expectations of future inflation,one might questionwhether an increasein pricesdue to somethinglike the formation of a cartel by the oil-producingcountries would be extrapolatedin the mannerthe priceterm indicates. After all, the industrycannot be cartelized more than once. To the extentthat wageearners are awareof the sourceof the initialrise in price,they may not reactas stronglyas the wageequation implies. In a similarvein, it has been arguedthat expectationsare based not only on past values of the variablein question,but also on knowledgeof the structureof the economic system and of the reactionsof policy makers. Supposepeople expectthe authoritiesto increasethe unemploymentrate to, say, 6 percentand hold it therewhenever inflation exceeded some maxi- mum acceptablevalue. When this point is reached,individuals would reason that the authoritieswould not allow this situationto continueindefi- nitely,but would act to reducethe inflationrate slowly. In this case, price expectationswould not dependupon previousprices alone, as they do in the wageequation used here.

The Effectsof ExternalShocks on the Economyas a Whole

In this and the next sectionwe reportsimulations of the full MPS model to analyzethe effectsof two priceshocks to theeconomy: cartelization in the foreignoil-producing countries and the applicationof pricecontrols. The emphasiswill be on the interactionamong the sectorsof the economy,and on determiningthe pressuresthat would develop, under the assumptionthat monetarypolicy, as measuredby growthin M1 (currencyplus demandde- 30 BrookingsPapers on EconomicActivity, 1:1974 posits), does not respondto the shock. Fiscal policy, too, is assumedto remainunchanged.12 Thedegree to whichpolicy should adjust to each of theseshocks depends uponone's preferences concerning the tradeoffbetween unemployment and inflation.We have nothingto add to this debate.In a later section, however, we do discussthe impact of an "accommodating"monetary policy on the initialeffects of a rise in oil prices. In carryingout the experiments,we first simulatedeach of the model equationsseparately using actual values of the right-handvariables, and recordedthe errorsfor each equationfor each quarter.Next, we simulated the equationssimultaneously in the full modelusing values of the variables that werecalculated by the model on the right-handside, addingin each quarterthe errormade by that equationwhen it was simulatedseparately. Becauseeach equationsimulated in this mannertracks the actual values exactly,the systemtracks exactly. In the thirdsimulation we enteredsome disturbance(for example,a change in the price of oil) and measuredits effectby noting the differencebetween the second and third simulations. This procedurehas two advantages.First, because the model is non- linear,its responsesvary with the conditionsunder which the responseis tested.Feeding the errorsinto each equationpermits a test nearhistorical values of the endogenousvariables even in the face of serious equation errors.Second, and equallyimportant, it allowsus to modifyvarious equations arbitrarilyand still remainnear historicalvalues of the relevanten- dogenousvariables. Otherwise it wouldbe necessaryto reestimatethe equations involved. An exampleshould demonstratethe importanceof this procedurefor modifyingthe model's structure.Suppose that we have measuredthe response of the system to a change in the money stock. We now wish to knowhow the systemwould have reacted to the samechange if the interest

12. We definedan unchangedfiscal policy as consisting of (1) unchangedtax rates (tax collectionsare allowed to vary with income), (2) unchangedconstant-dollar federal purchases (current-dollarpurchases of goods move proportionally to the nonfarm business deflator, and employee compensation moves proportionally to the private compensationrate), and (3) unchangedcurrent-dollar transfer payments. The last of these assumptionsis debatablein this context and so we repeatedsome of the experiments using an assumption of unchanged constant-dollar transfers, with negligible differencesin the results. James L. Pierce and Jared J. Enzler 31 elasticityof demandfor moneywere half the value specifiedin the money demandequation. With the interestelasticity coefficient cut in half, we sim- ulate the alteredmoney demandequation using actual variableson the right-handside, and recordthe errorfor each quarterin placeof the errors arisingfrom the unalteredequation. Simulating the new system simulta- neously,we add the new set of errorsto each equationin each quarter. As in the test of the unalteredsystem, this second simulationwill track historyexactly. In the third simulationwe again changethe money stock and note the differencebetween the second and third simulations.We shouldfind a greatereffect on interestrates than we foundin the unaltered systemsince now interestrates will needto changemore in orderto accommodatethe changedmoney stock. This techniquecan also be used to add a newvariable arbitrarily to an equation.We shallmake use of the capacity to modifythe modelin theseways later in this sectionin orderto do sensitivity analyseswith certainimportant parameters in the model.

CARTELIZATION OF AN IMPORT INDUSTRY We shall attemptto estimatethe economic impact of the increasein crudeoil pricesfrom about $3.50to about $10 per barrelthat occurredin 1973after the formationof the oil productioncartel. We have ignoredthe effectsof the temporaryembargo on oil shipmentsto the United States, which,while sharp,were probablynot abiding. To some extent the resultsof the increasein oil prices will dependon price and income elasticitiesof demandthat at this point are uncertain, on cross elasticitiesof demandbetween energy sources, and on the exis- tence of backup energy technologiesabout which very little is known. Somesweeping assumptions about these matters are necessary: (1) that the cartelizationwill be permanent;(2) that the cartel will adjustthe dollar price of oil in responseto changes in the U.S. price level to maintain the relativeprices of U.S. goods and oil; (3) that the price elasticityof demandfor oil is low; (4) that the suppliersof importedoil are willingto take financialclaims in paymentand not spend them for U.S. exports. We also assumea world of floatingexchange rates in which both import and export prices move proportionatelywith the domestic nonfarm businessdeflator. In this case a generalinflation has no price effects on the quantitiesof internationallytraded goods consumed. The effect of 32 BrookingsPapers on EconomicActivity, 1:1974 real incomeon importdemand remains in force. Finally,we assumeaway all third-countryeffects. Becauseof technicaldifficulties in simulatingevents into the relatively distantfuture, and becausewe wishedto use the errorcoding procedure describedearlier when alteringthe structureof the model, we moved cartelizationand the oil crisis back to the beginning of 1967.13This procedureallows us to simulatethe effectsof the oil case for seven years, which is probablymuch longer than several of our assumptionscould be maintained.14 A necessaryfirst step in the analysisis to determinethe initial impact of the increasein prices of importedoil on U.S. prices. In principlewe could calculatehow much this increaseadds to the currentbills for imports and for final goods and raise the various price deflatorsfor 1967 accordingly.However, this procedurecannot be used for differentcom- ponents of final output because we do not have complete information abouthow the increaseis distributedamong the pricedeflators for various products.Clearly, consumer nondurables and servicesbear the bruntof it. In additionto the substantialportion of petroleumused by privateauto- mobiles,much of the productof industriesthat rely heavilyon petroleum, such as air transportand electricutilities, falls into these categories.For simplicitywe assumethat all of the increasein import prices is initially reflectedin these sectors.On this assumption,the deflatorfor consumer nondurablesand serviceswould rise by about 2 percent.Since the value- added deflator(Praf) is not affecteddirectly, we altered equation (5) by increasingUCf,d by 2 percentabove its historicalvalue.15 In fact, oil pricesdid not riseto theirultimate level all at once. Moreover, since the simultaneousembargo on shipmentsto the United States by some producercountries reduced the quantityof importedoil, the initial directeffect of the higherworld prices for crudeoil on the domesticprice level was less severethan the above calculationsimply. To allow for this we phasedin the upwardadjustment in equal incrementsover threequar-

13. Economic conditions obviously were not the same in 1967 as they are in 1974. Some evidence on the importance of initial conditions for the simulation results is discussedbelow. 14. Severalof the equationsuse nominalinterest rates prior to 1967 and real interest rates thereafter,so an earlierstarting date was not feasible. 15. Then, since we wantedthe deflatorfor domesticallyproduced nondurable goods (Pcndcnd - PimiM) to be unaffectedby the oil price increase,we calculatedthe value of Pimthat gave this result. James L. Pierce and JaredJ. Enzler 33 ters, with the ratio Und maintained2 percenthigher in everyquarter thereafter. These proceduresstill leave the problem of pricing domestic oil and substitutesand of distributingthis price increaseby final product. By mid-1974,the averageprice of domesticoil had risen about $3 a barrel in the wakeof the risein the priceof importedoil. Domesticprice controls, of course, containedthe rise, and once the furor over petroleumprices subsides,oil producerswill probablybe able to exactprices that are closer to world levels. The rise in domesticoil prices causes an increasein the nonfarmbusiness deflator relative to wages,and in the profitshare of total output.The pricesof oil substituteshave also risen.The price of coal has risensubstantially (as wouldthe priceof naturalgas in the absenceof regulation). These effects should also increasethe price level and the profit sharein these industries. But even if the pricesof oil and otherenergy sources go up and remain high, the shareof profitsin gross nationalproduct might not rise, except in the short run, by as much as the crudefigures on oil revenuessuggest. Expandeddomestic production of oil, coal, and naturalgas, along with the developmentof other energysources, would have a depressingeffect on both foreign and domestic oil prices. Not only will there be a sub- stitution away from oil productsto other energy productsand to other expendituresbut in the long run resourceswill be attractedby high profits until the marginal,if not the average,productivity of capital in the oil industryrecedes to approximatelythat of other industries.Thus, in the long run the profitshare in GNP will tend to returnto the value it would have had in the absenceof cartelization.However, we believethat oil producerswill be morelikely to be able to maintainthe higherprofitability of their industrythan will otherimport-competing goods industries.Our be- lief restson the fact that largerprofits are reallya rent of oil wells and that the rentalprice can remainhigh even though inducedinvestment reduces the marginalproduct of capitalin this industry. Clearly,many of these issues are more politicaland technologicalthan economic, and we are forced to fall back on assumption.We assume(1) that the worldprice of oil remainsat its new level of $10, and (2) that the increasein the profitshare brought about by a rise in the averageprice of domestic crude oil from about $3.50 to $6.30 per barrel is permanent. Further expansionin the profit share, arising from the higher price of domesticpetroleum substitutes and deregulationof both domesticcrude 34 Brookings Papers on Economic Activity, 1:1974 oil and its substitutes,is assumedto be offset by heavierinvestment in energyindustries and the developmentof new technologies.In the present state of the world,these assumptionscould be outdatedwithin weeks. The next step is to approximatethe impactof the increasesin domestic oil priceson the relevantprice deflators. The increasein the dollarvalue of domesticcrude oil productionis the equivalentof just over 1 percentof the current-dollarvalue of nonfarmdomestic business output in 1973:4.16 To accountfor the assumedresponse of domesticoil prices,we adjusted the nonfarm business deflatorequation so that the deflatorwas 1 percent higher relative to factor costs than was the case in the control simulations. Likethe pricesof importedoil, those of domesticoil will not reachtheir new level all at once; furthermore,some time will elapse between the rise in the pricesof importedoil and the increasein the pricesto the final consumersof productsmade from it. To approximatethis combination of factors,we adjustedthe price equationupward by only 0.5 percentin the initialquarter, then graduallyincreased the adjustmentto 1 percent.'7 Raisingthe prices of both domesticand importedoil increasesthe deflatorsfor consumernondurables and services3 percentrelative to wages whenthe phased-inadjustments are complete. The priceof generalgovern- ment holds relativeto wages, and the product prices of the remaining sectorsof the economyrise 1 percentrelative to wages. Implicitin theseadjustments to the deflatorsare assumptions concerning various elasticitiesof demandfor petroleumproducts and of supply of domesticoil. Are these elasticitiesreasonable? In the eight years ending in 1973, annual consumptionof petroleumproducts rose just under 50 percent,from 3,749 million to 5,617million barrels. Assume that the same rateof increasewould have continued had it not been for the increasein oil

16. Projectedproduction of domestic crude oil for early 1974 was about 11 million barrelsper day, which at $3.50 per barrelis valued at $14.1 billion per year. At $6.30 per barrel,the value would be $25.3 billion per year. In the fourthquarter of 1973private nonfarm domestic output was $1,073 billion. Increasing this total by $11.2 billion (25.3-14.1) while leaving real output unchangedwould increase the domestic nonfarm price deflatorby slightly over 1 percent. 17. The adjustmentwas made by adding to equation (4) a constant 0.005 in the initial quarterof the high-oil-pricesimulations and 0.003 in each quarter thereafter. Since the coefficientof the lagged dependentvariable is 0.7, the adjustmentapproaches 0.003/(1.O-0.7) = 0.01, or 1 percentsince the equation is in logarithms. JamesL. Pierce andJared J. Enzler 35 prices.Further assume that consumernondurables and servicesand nonfarm business output grow on averageat the long-runnatural rate of growth of the economy(around 3.5 percentper year)-that is, about 32 percentin the same period. Our importprice adjustmentof 2 percentin UC,,is correct only if the proportionof importedoil in total real consumption of nondurablesand servicesremains constant; our domestic price adjustmentof 1 percentin Pnf is correct only if domesticoil production is a constant proportionof nonfarm business output. For all theseassumptions to be met, both importedoil consumptionand domestic oil productionmust also increaseby 32 percentover the period. Thus, we have implicitlyassumed that the quantity of oil consumed will be reduced by 12.0 percent over the next eight years-that is, from 150 percent of its 1973 level to 132 percent.We estimatethe averagefinal sale price of petroleumproducts in 1973 to be roughly30 cents per gallon; the price increaseswe are assuming will raise the average retail value by 10 cents per gallon, or 33 percent.The assumedlong-run (eight- year) price elasticityof demandis then - 12/33, or about -0.36. This figureis perhapsa bit on the low side, but not improbable. Next, is the assumptionof a rate of growth of domestic petroleum output of about 3.5 percent reasonable?Domestic productionactually declinedslightly between 1970 and 1973and if the worldprice of oil had not risen,the declinewould probably have accelerated.Thus our assumptions imply a sizable price effect on domestic production. Imported petroleum,on the otherhand, grew at about a 12 percentannual rate in the eight years ended in 1973. Our assumptionsimply that this rate of growth drops to about 3.5 percent,which again is a substantialeffect. The oil price increasewould also be expectedto stimulateinvestment. For a time capitalexpenditures will be necessaryto expanddomestic oil output, to develop new energyresources, to expand production of pe- troleumsubstitutes, and to adopt energy-savingtechnology. We do not know the magnitudesof these effectsand have not attemptedto change the investmentfunction to account for them. The function does allow investmentto respondto changes in output and the cost of capital in responseto the rise in oil prices. The assumptionswe have had to make are more likely to hold in the short run than over a longer period of years. Small errorsin the assumed demand and supply elasticitieswill loom large with the passage 36 Brookings Papers on Economic Activity, 1:1974 Table 3. Simulated Economic Impact of the 1973 Increase in Oil Prices, by Selected Quarters Deviation from control simulation

Conswnp- Unem- Grossnational product tion ex- Disposable ploy- Quarter penditures income ment of Billions Billions (billions (billions rate Treasury simu- of current of 1958 of 1958 of 1958 (per- bill rate lation dollars dollars dollars) dollars) cent) (percent) 1 0.1 -2.5 -2.7 -5.5 0.1 0.3 2 -3.1 -6.0 -5.3 -10.2 0.4 0.3 3 -7.2 -10.0 -8.3 -14.7 0.6 0.3 4 -10.4 -13.4 -10.4 -16.4 0.9 0.2 5 -13.4 -16.4 -12.4 -18.2 1.1 0.1 6 -16.4 -19.1 -14.3 -20.0 1.3 0.0 7 -18.6 -20.8 -15.8 -21.4 1.5 -0.1 8 -20.0 -21.5 -16.9 -22.7 1.6 -0.2 12 -16.2 -15.2 -17.2 -23.9 1.5 -0.2 16 -21.6 -14.4 -18.2 -24.7 1.7 -0.3 20 -33.7 -15.1 -17.4 -25.4 1.8 -0.7 24 -57.3 -19.5 -18.6 -28.3 2.1 -1.2 28 -75.5 -12.9 -17.1 -28.1 1.9 -2.4 Source: Simulations covering the 1967-73 period; the price of oil rises from $3.50 to $10.00 per barrelfor imports and to $6.30 per barrel for domestic production, as described in the text, while the money supply and fiscal policy are unchanged from the control simulation. of time.The successof the foreignoil cartelis harderto predictthe further ahead one looks. And investmentresponse is likely to become more significantafter a period of time. For all these reasons, the analysisis probablyconsiderably more reliablefor the near term than for several yearsdown the road.

SIMULATIONRESULTS The problemhas now been definedand we can turnto a seriesof simulation experimentsdesigned to give quantitativeestimates of the impacts of the rise in oil prices.Table 3 shows the differencesbetween the high- oil-priceand controlsimulations. Initially, the higheroil pricespowerfully depressincome and employment,for severalreasons. First, the increasein import prices alone, when fully passed on to James L. Pierce and Jared J. Enzler 37

consumers,reduces real disposableincome by about 1.8 percent'8-equivalent in effect to a 12 percent income tax surchargeat 1973:4 levels of the relevantmagnitudes. Second, the increasein prices of domesticoil further depressesreal disposable income by about 0.9 percent. Some of this loss will be returnedto individualsin the form of dividendsand, therefore,does not have quite the depressingeffect of the increase in

18. The 1.8 percentis the proportionby which current-dollardisposable income in 1973:4 ($917.8 billion) would have had to rise to cover the increasedprice of imported oil ($16.6 billion). This, like other direct effectsdiscussed here, cannot be read from the tables. The figures in the tables are the result of a simultaneous solution of the full system. The changein disposableincome reportedthere is a result not only of the direct effect of the increasein oil prices on disposableincome, but of indirect effects as well, since changed prices and real incomes lead first to changed production and then to furtherchanges in disposableincome. 38 BrookingsPapers on EconomicActivity, 1:1974 import prices. Moreover,the higher domestic prices will also enhance shareprices and reduce the costs of capital,thus inducing more investment. Third,the increasedprices of importsalso diminishthe value of house- hold net worthmeasured in terms of consumergoods. In the early quarters of the simulation,this accountsfor about $2 billion (1958 prices)of the reducedconsumption in the simulation.Fourth, the increasedprices for domesticoil furthercut the real value of consumernet worth; again the effectsare not as severeas those imposedby higherprices for importsbe- causethe richerprofits of oil companiessupport the valueof the corporate- share portion of consumernet worth. Finally, once the phase-in of oil pricesis complete,the increasein the value of transactionsraises the demand for money by enoughto increasethe Treasurybill rate by 60 basis points. Thesefive effectscombine to exerta powerfuldownward force on real outputand consumption. Unlikemost downwardshifts in real demand,that resultingfrom the increasein oil pricesdoes not lead quicklyto lower interestrates for a given growthin M1. In fact, the bill rate is not significantlylower until the fifth yearof the simulation.There are two reasonsfor this slow reduction.First, higherprices raise the value of transactions,and interestrates cannot fall until real outputand priceshave droppedsufficiently to bringthe value of transactionsbelow the levelsin the controlsimulation. Second, the increase in pricessets up an interactionthat keepscompensation per manhour up to around its control levels for the first six quarters,and holds consumer prices above their control levels for five years. These higher prices also supportinterest rates, depriving the systemfor some time of the stabilizing effects the financialsector ordinarilyoffers when real activity declines. Thus,the multiplier-acceleratorprocess is allowedto go fartherthan usual. Givenour assumptions of unchangedmoney growth, the rateof inflation eventuallymust returnto its originallevel. This processtakes a very long time to work itself out. The rise in oil pricesraises the overallprice level abovea point that can be supportedby the existingmoney stock. The cur- tailedoutput and unemploymenteventually act to lower the rate of inflation as the economytries to squeezeout the excessin prices.By the time it succeeds,however, the annualrate of inflationhas been reducedby almost 2 percentand prices overshootthe mark. At the end of the simulation period,output is recovering.19 19. Real GNP is not down by nearly as much as the 1.9 percent higher unemployment rate would imply, given Okun'slaw. Among the reasonsfor this developmenttwo are especiallyimportant: (1) By the end of the simulation,the productof state and local James L. Pierce and Jared J. Enzler 39 This overshootof the pricelevel is surprisingat firstglance. It is, however, an almost inescapableresult of the standardeconomic model. Sup- pose we have an economicsystem growing along an equilibriumpath and for some reasonthe pricelevel is displacedupward. If the money stock is not adjustedto accommodatethis development,interest rates will rise, depressingthe level of real output.In turn, unemploymentwill rise, and begin to restrainthe rate of changeof wagesand prices.As long as prices are too high (and thereforeoutput and employmenttoo low) to be con- sistentwith the equilibriumgrowth path, the rate of changeof priceswill decline.When pricesfinally return to the level they would have occupied in the absenceof the disturbance,their rate of changeis too low and they overshoot.The processthen reverses.If the systemis stable,both the level and the rate of changeof priceswill eventuallyachieve their equilibrium values. Thereis no guaranteethat this cyclicalmechanism is stable, and in the MPS model,it is barelyso. The model may not pinpointthe time at which this price overshootoccurs, but simulationsover a period long enough for the overshoot to occur serve to show how the adjustment works. Eventuallythe inflationrate, unemployment,real output, and the bill rate must returnto the neighborhoodof their control levels. Becausethe value of transactionshas risen, nominal GNP, prices, and wages must settle somewhatbelow control simulationlevels. The simulationis not long enoughfor all these eventsto work themselvesout. The deflationarypower of the oil crisis turns out to be substantial.In particular,the reductionin real outputis sufficientto restrainwages in the early quartersof the simulation.This sharp reaction could be exces- sive. If prices of consumer goods increase, is it reasonableto believe that householdsrapidly adjust their consumptionto the fall in the real value of theirassets, as the model predicts?

TESTS OF SENSITIVITY To test the importanceof this factorin the simulationresult, we altered the reactionof the consumptionfunction to changesin real net worth.In general governmentis lower and privatebusiness output is higherin the case involving higher oil prices. Real output per employeeis much higherin the privatesector than in the governmentsector. (2) Labor productivityin this model depends positively on the level of capacityutilization; when utilizationis high, overheademployees are used more effectively.The lower real output in the early quartersof the oil simulationsreduces the ending level of capacityand raises utilizationrates. 40 Brookings Papers on Economic Activity, 1:1974 the first three years of the simulation, current-dollarnet worth excluding common stocks (VOW) was virtually unchanged by the increase in oil prices. Thereafter it dropped slightly, with the reduction not reaching $50 billion until the fifth year. Thus, nearly all the reduction in real net worth in the early part of the simulations is caused by higher prices of consumer goods. To test the contribution that the rapid response of real consumption to the reduction in real net worth made to the simulation results, we merely suspended the term on VOW in the consumption function by setting its coefficient to zero. This is equivalent to assuming that households never react to the fall in the real value of VOW. The results of this test are shown in Table 4. As expected, the initial downward effect on consumption is mitigated, and the initial upward effects on prices and wages are magnified. The differences are not great, however. After a few quarters, Table 4 looks much like Table 3. Also,

Table4. SimulatedEconomic Impact of the Increasein Oil Prices witha ModifiedConsumption Function, by SelectedQuarters Deviation from control simulation

Consump- Unem- Grossnational product tion ex- Disposable ploy- Quarter penditures income ment of Billions Billions (billions (billions rate Treasury simu- of current of 1958 of 1958 of 1958 (per- bill rate lation dollars dollars dollars) dollars) cent) (percent) 1 1.3 -1.5 -1.6 -5.1 0.1 0.3 2 0.2 -3.4 -3.1 -9.1 0.2 0.4 3 -1.6 -6.1 -5.1 -12.9 0.4 0.6 4 -3.4 -8.7 -6.7 -14.2 0.5 0.5 5 -5.4 -11.5 -8.5 -15.7 0.7 0.5 6 -7.7 -14.2 -10.2 -17.5 0.9 0.4 7 -9.6 -16.2 -11.5 -18.9 1.1 0.4 8 -10.8 -17.5 -12.6 -20.1 1.2 0.4 12 -10.7 -16.7 -14.6 -23.0 1.4 0.5 16 -14.9 -17.6 -16.4 -24.8 1.7 0.1 20 -24.6 -19.8 -17.5 -27.0 2.0 -0.2 24 -48.8 -27.5 -20.9 -32.7 2.5 -1.0 28 -63.1 -20.5 -20.7 -34.0 2.5 -2.0 Source: Simulations covering the 1967-73 period. Changes in real consumer net worth other than common stocks are assumed to have no effect on consumption. Changes in oil prices and assumptions about policy are the same as those for Table 3. James L. Pierce and Jared J. Enzler 41 while the early effects are moderated by this change, the later effects are intensified. These shifts result because prices in this simulation have risen even farther above the level that can be supported by the existing money stock. Certainly the correct specification of the consumption function is closer to the one used in Table 3 than to that in Table 4. This test demonstrates that the results reported in Table 3 would not be significantly altered by any reasonable slowing of the reaction of the consumption function to changes in net worth. We also tested the sensitivity of the results to our selection of imports plus GNP as the appropriatemeasure of transactions in the money demand function. The test consisted of dropping the modified money demand equation and repeating the experiment with the original money demand function, in which transactions are represented by nominal GNP. The re-

Compensationper Consumptionprice Valueof manhour deflator Corporatecorporate Exports Imports profits shares Annual Annual (billions (billions (billions (billions rate of Index rate of of of of of Amount change (1958= change current current current current (cents) (percent) 100) (percent) dollars) dollars) dollars) dollars) -0.1 -0.2 1.2 4.3 0.2 2.8 1.7 19.6 0.1 0.3 2.2 3.3 0.3 5.6 1.1 17.9 0.5 0.5 3.0 2.8 0.3 8.0 -0.1 5.8 1.1 0.7 3.3 0.8 0.4 7.6 -1.2 2.8 1.5 0.4 3.5 0.5 0.5 7.4 -2.0 7.3 1.6 0.1 3.6 0.4 0.6 7.2 -2.4 16.7 1.7 0.0 3.8 0.3 0.7 7.1 -2.4 25.8 1.5 -0.2 3.9 0.2 0.7 7.1 -1.9 44.8 0.4 -0.4 4.2 -0.1 0.8 7.9 3.0 64.9 -0.6 -0.3 4.0 -0.3 0.7 8.2 5.1 55.6 -3.8 -0.9 3.2 -0.9 0.3 8.1 7.9 49.0 -10.5 -1.7 1.5 -1.5 -0.6 5.8 9.2 32.9 -21.4 -2.3 -1.4 -2.1 -2.8 5.0 27.8 94.9 42 Brookings Papers on Economic Activity, 1:1974 sults are reported in Table 5. A comparison with Table 3 indicates that the effects of this adjustment are not large. The modification works in the expected direction. The unemployment rate is not driven as high and the price level ends up somewhat higher. It has sometimes been argued that the interaction between prices and wages depends upon whether the increase in prices originates within the nonfarm business sector or outside it. Prices can affect wages through two avenues: through demands by workers for higher wages to cover a rise in consumption prices, and through the willingness of firms to pay higher wages when the value of the marginal product of labor has risen. Price increases whose source is outside the nonfarm business sector do not raise the value of labor's marginal product and should, therefore, lead to smaller wage gains. Thus a price increase resulting from an external shock that raises P,,,, and not Pnfmay not have as large an impact on wages, at least initially, as an ordinary rise in all prices. Because the correlation

Table5. SimulatedEconomic Impact of the Increasein Oil Prices witha ModifiedMoney Demand Equation, by SelectedQuarters Deviation from control simulation

Consump- Unem- Grossnational product tion ex- Disposable ploy- Quarter penditures income ment of Billions Billions (billions (billions rate Treasury simu- of current of 1958 of 1958 of 1958 (per- bill rate lation dollars dollars dollars) dollars) cent) (percent) 1 0.2 -2.4 -2.6 -5.5 0.1 0.1 2 -2.7 -5.6 -5.1 -10.0 0.3 0.0 3 -6.1 -9.1 -7.9 -14.3 0.6 -0. 1 4 -8.4 -11.9 -9.7 -15.8 0.8 -0.1 5 -10.5 -14.3 -11.5 -17.4 1.0 -0.2 6 -12.6 -16.5 -13.2 -19.0 1.1 -0.3 7 -14.0 -17.8 -14.5 -20.2 1.3 -0.3 8 -14.5 -18.1 -15.3 -21.2 1.4 -0.3 12 -5.5 -9.7 -14.6 -21.4 1.2 0.0 16 -9.7 -10.2 -15.8 -22.2 1.3 -0.2 20 -17.5 -10.6 -14.8 -22.9 1.4 -0.5 24 -34.0 -14.2 -16.3 -25.5 1.6 -0.8 28 -52.0 -12.6 -17.0 -26.8 1.6 -1.9 Source: Simulations covering the 1967-73 period. Money demand is assumed to be unaffected by higher import prices. Assumptions about the changes in oil prices and about policy are the same as those for Table 3. James L. Pierce and Jared J. Enzler 43 between consumption prices and nonfarm business prices is very close, their differential effects are difficult to estimate; we were in fact unable to do so. The question is critical to our discussion, however, so we decided to test how much difference it might make if the price effect estimated in the model (which used consumer prices only) were in fact distributed between consumer and producer prices. We again arbitrarilyaltered the original form of the model by breaking the long-run price coefficient of 0.989 in the wage equation into two parts. We entered two price terms with the same distributed lag into the equation. The first term was on the rate of change of the nonfarm business deflator, and we gave this term a coefficient of 0.5. The other distributed lag was on the rate of change of P,,nnand we gave this term a coefficient of 0.489. Simulation results with this modification to the wage equation are reported in Table 6. As expected, this change helps repress the rate of wage inflation in the

Consump- Unem- Grossnational product tion ex- Disposable ploy- Quarter penditures income ment of Billions Billions (billions (billions rate Treasury simu- of current of 1958 of 1958 of 1958 (per- bill rate lation dollars dollars dollars) dollars) cent) (percent) 1 0.1 -2.5 -2.7 -5.5 0.1 0.3 2 -3.3 -6.0 -5.3 -10.3 0.4 0.3 3 -7.8 -10.1 -8.4 -15.0 0.6 0.3 4 -11.6 -13.5 -10.5 -16.9 0.9 0.1 5 -15.3 -16.4 -12.6 -18.8 1.1 0.0 6 -18.9 -18.9 -14.4 -20.7 1.3 -0.2 7 -21.8 -20.5 -15.8 -22.0 1.4 -0.3 8 -23.8 -20.9 -16.7 -23.0 1.5 -0.4 12 -21.0 -12.3 -15.8 -23.1 1.4 -0.5 16 -27.2 -9.7 -15.9 -22.9 1.4 -0.6 20 -40.8 -9.1 -13.8 -22.1 1.4 -0.9 24 -62.7 -11.0 -13.7 -22.8 1.5 -1.3 28 -78.1 -3.4 -11.3 -20.6 1.1 -2.4 Source: Simulations covering the 1967-73 period. It is assumed that half the price effect on wages comes from consumer prices and half from producer prices. Assumptions about changes in oil prices and about policy are the same as those for Table 3. early part of the simulation.The adjustmentto the increasein oil prices is less painful.The earlyrates of inflationare somewhatreduced and the maximumrise in unemploymentin the later part of the simulationsis considerablyreduced. The most implausiblefeature of the simulationis that wage inflationis immediatelyreduced by the increasein oil prices. This may be possible,but seems unlikely.In any event, the modification of the wage equationdoes not producea majorchange in the outcome. We next modifiedthe assumptionsof the basic model to allow for a rise in exportsin responseto the dollaraccumulations of oil-producingnations. In this variation,real exportsslowly rose until the nominalvalue of the incrementmatched the increasein the value of importedoil. In the second quarterof the simulation,real exports were increasedenough to push nominal exports up by 5 percent of the amount of the extra payments for oil importsin the firstquarter of the simulation.In successivequarters, James L. Pierce and Jared J. Enzler 45

this processwas repeated,in 5 percentincrements, until in the twenty-first quarter,the extra value of exportsmatched the increasedoil payments. The resultsof this experimentare reportedin Table7. Againthey reflect no majordeparture from the resultsof Table3. Increasingexports length- ens the period of adjustmentbecause aggregate demand does not fall as rapidly.Despite this upwardshift in aggregatedemand, the economymust enduresufficient unemployment to squeezethe excessinflation from prices. The adjustment,while reduced in magnitude,is extendedin time.20 Initialconditions in the economymay be crucialto the issue of policy responsesto the increasein oil prices.Conditions in 1967obviously were not the same as those in 1974.In particular,the unemploymentrate was 20. It appearsthat any adjustmentin the investmentfunction to allow for expanded oil exploration or coal mining, or for developmentof alternativeenergy technologies, would have similar effects. 46 Brookings Papers on Economic Activity, 1:1974 Table 7. Simulated Economic Impact of the Increase in Oil Prices with Rising U.S. Exports, by Selected Quarters Deviation from control simulation

Consump- Unem- Grossnational product tion ex- Disposable ploy- Quarter penditures income ment of Billions Billions (billions (billions rate Treasury simu- of current of 1958 of 1958 of 1958 (per- bill rate lation dollars dollars dollars) dollars) cent) (percent) 1 0.1 -2.5 -2.7 -5.5 0.1 0.3 2 -2.8 -5.7 -5.3 -10.1 0.3 0.3 3 -6.1 -9.1 -8.2 -14.4 0.6 0.4 4 -8.5 -12.0 -10.1 -15.9 0.8 0.3 5 -10.6 -14.3 -12.1 -17.4 1.0 0.2 6 -12.6 -16.4 -13.8 -18.9 1.1 0.1 7 -13.8 -17.6 -15.1 -20.0 1.3 0.1 8 -14.2 -18.0 -16.0 -21.0 1.4 0.1 12 -8.0 -11.3 -16.5 -21.8 1.3 0.3 16 -9.3 -9.6 -17.3 -22.3 1.3 0.1 20 -13.0 -7.4 -16.1 -22.0 1.3 -0. 1 24 -27.0 -9.7 -16.6 -23.8 1.4 -0.6 28 -35.3 -3.6 -15.7 -23.7 1.2 -1.3 Source: Simulations covering the 1967-73 period. The value of exports is allowed to rise to match the higher value of oil imports. Assumptions about the changes in oil prices and about policy are the same as those in Table 3. lower in 1967-69 than it has been recently.21To examine this possibility, we conducted the experiment presented in Table 3 starting in the first quarter of 1970 rather than 1967. This shift allows us to analyze the effects of an oil price increase under unemployment rates significantly higher than those in 1967. Of course, only four years could be simulated in this experiment. The results are reported in Table 8; again they reveal no major changes in the results. The most notable difference is that the price adjustment has been slowed since at higher unemployment rates the Phillips curve is flatter-that is, a given rise in the unemployment rate

21 If our model werelinear, the resultswould not depend on initial conditions. This model, however,has a numberof nonlinearities;the prime exampleis the Phillips curve since its usual hyperbolicformulation means much greaterchanges in inflationfor given changesin the unemploymentrate at low levels of unemploymentthan at high levels. James L. Pierce and Jared J. Enzler 47

Compensationper Conisumptionprice Valueof manihour deflator Corporatecorporate Exports Imports profits shares Annual Annual (billions (billions (billions (billions rate of Index rate of of of of of Amount change (1958= change current current current current (cents) (percent) 100) (percent) dollars) dollars) dollars) dollars) -0.3 -0.4 1.2 4.2 0.2 2.7 1.2 18.6 -0.3 0.0 2.1 3.1 0.5 5.3 0.3 17.1 -0.1 0.2 2.9 2.6 1.1 7.4 -0.9 6.8 0.3 0.5 3.1 0.5 1.6 6.9 -1.9 5.1 0.4 0.2 3.2 0.3 2.1 6.7 -2.2 11.1 0.4 -0.1 3.3 0.2 2.6 6.5 -2.2 23.4 0.2 -0.2 3.4 0.1 3.1 6.5 -1.6 34.8 -0.2 -0.4 3.4 0.0 3.6 6.5 -0.3 59.1 -1.5 -0.4 3.5 -0.2 5.8 7.9 8.3 89.9 -2.6 -0.4 3.2 -0.5 8.2 8.6 11.8 82.7 -5.6 -0.7 2.4 -0.7 11.0 9.1 15.2 91.5 -10.7 -1.2 1.1 -1.1 12.6 8.3 16.0 77.3 -18.4 -1.4 -0.9 -1.4 14.0 8.6 29.8 91.7

yields a smallerreduction in the inflationrate. This pattern makes the adjustmentmore difficultsince a larger rise in unemploymentmust be enduredto squeeze out the excess from prices. In this respect, 1974 is probablymore like 1970than like 1967.

ACCOMMODATIVEMONETARY POLICY One obvious policy responseto an event like the surge in oil prices would be to accommodatethe originalincrease in priceswith an expansion of the money stock. We could derivea monetarypolicy that maxi- mizes an objectivefunction based on, say, the unemploymentrate and the rate of inflation.We havechosen not to attemptthis, partlybecause we do not want to enterinto a debateabout the parametersof this objective function.Instead we have chosento analyzethe simplecase in which the 48 Brookings Papers on Economic Activity, 1:1974 Table 8. Simulated Economic Impact of the Increase in Oil Prices with Differing Initial Economic Conditions, by Selected Quarters Deviation from control simulation

Consump- Unem- Grossnational product tion ex- Disposable ploy- Quarter penditures income ment of Billions Billions (billionis (billions rate Treasury simu- of current of 1958 of 1958 of 1958 (per- bill rate lation dollars dollars dollars) dollars) cent) (percent) 1 -0.2 -2.9 -3.1 -6.2 0.1 0.4 2 -4.3 -6.9 -6.0 -11.1 0.3 0.5 3 -8.9 -11.4 -9.3 -15.4 0.6 0.5 4 -11.8 -14.7 -10.9 -15.8 0.8 0.4 5 -14.1 -17.8 -13.1 -17.7 1.0 0.2 6 -16.6 -20.9 -15.2 -19.1 1.2 0.2 7 -17.8 -23.1 -17.0 -20.7 1.4 0.4 8 -18.0 -24.7 -18.6 -22.3 1.5 0.3 12 -6.1 -22.0 -21.8 -28.1 1.8 0.9 16 -12.1 -27.6 -25.9 -32.9 2.3 1.1 Source: Simulations starting with the economy of 1970 rather than 1967. The assumptions about the changes in oil prices and about policy are the same as those in Table 3. monetaryauthority responds by increasingthe money stock in proportion to the increasein transactionsdemand-the sum of current-dollarGNP plus imports-resulting from the initial rise in prices for both imported and domestic oil. After this initial accommodation,the money stock is assumed to grow at its old constant rate. The results of this exercise are shown in Table 9. As expected, real output declines much less in this case. A declinestill occurs, of course,because even thoughthe monetaryau- thorityhas offsetthe impactof the originalprice increases on transactions, the system still reacts to the reduction in real income and wealth. The expansionin the moneystock also allowsthe wage-pricespiral to continue to a much greaterextent. The averageinflation rate over the seven-year period is about 0.7 percenthigher than when the authoritiesdo not accommodate the increase in transactions. In the long run, the inflation rate must returnto its value in the control simulation.Achieving this result requires that the average unemployment rate be sufficiently high in the transitionperiod. Whether or not the outcome of this "accommodating" monetarypolicy is regardedas superiorto the outcomewith unchanged policy, shown in Table 3, dependsupon the weightsassigned to the unemploymentand inflationobjectives. James L. Pierce and Jared J. Enzler 49

Compensationper Consumptionprice Valueof manhour deflator Corporatecorporate Exports Imports profits shares Annual Ainnual (billions (billionis (billionis (billions rate of Index rate of of of of of Amount change (1958= change current current current current (cents) (percent) 100) (percent) dollars) dollars) dollars) dollars) -0.3 -0.3 1.4 4.3 0.3 3.4 1.3 16.8 -0.2 0.2 2.5 3.3 0.4 6.6 -0.8 6.3 0.4 0.5 3.4 2.7 0.4 9.2 -3.7 -2.3 1.4 1.0 3.7 0.8 0.6 8.9 -7.1 -3.8 2.2 0.8 4.0 0.6 0.7 8.6 -8.0 -8.6 3.0 0.7 4.2 0.7 0.8 8.4 -10.0 -5.5 3.8 0.6 4.6 0.8 1.0 8.3 -10.4 5.4 4.4 0.6 4.8 0.8 1.0 8.4 -9.8 17.1 6.9 0.3 6.1 0.7 1.8 9.1 3.0 123.3 6.8 -0.5 7.0 -0.2 2.8 8.8 4.7 83.2

Accommodativemonetary policy has no single definition.The definition that we choseis one that sometimeshas been advancedas a reasonable responseto an exogenousprice shock. Other definitionswould lead to differentresults involving, among other things, differentamounts of unemploymentand inflation.However, since the increasein oil prices has severalkinds of impact on the economy, no monetarypolicy is capable of offsettingall its effects.

Wage and Price Controls

We conclude with an examinationof the recent wage-pricecontrols and some observationson what might be expectedfrom their removal.22 The impact of the controlswas estimatedby comparingtwo simulations of the full systemholding fiscal and monetarypolicy unchangedin the

22. Robert Gordon has recently provided an analysis of the controls using only a wage-pricesector. His analysisis approximatelyequivalent to assumingthat policy was so adjusted that the controls had no effect on the unemploymentrate. See Gordon, "Response of Wages and Prices." 50 BrookingsPapers on EconomicActivity, 1:1974 Table 9. Simulated Economic Impact of the Increase in Oil Prices with AccommodatingMonetary Policy, by Selected Quarters Deviation from control simulation

Consump- Unem- Grossnationial product tionzex- Disposable ploy- Quarter penditures income ment of Billions Billions (billions (billions rate Treasury simu- of current of 1958 of 1958 of 1958 (per- bill rate lation dollars dollars dollars) dollars) cent) (percent) 1 0.4 -2.1 -2.6 -5.4 0.1 -0.2 2 -1.9 -5.0 -4.8 -9.7 0.3 -0.4 3 -4.2 -7.7 -7.3 -13.7 0.5 -0.6 4 -5.3 -9.6 -8.7 -14.9 0.6 -0.5 5 -6.1 -11.2 -10.2 -16.2 0.8 -0.5 6 -7.0 -12.7 -11.6 -17.4 0.9 -0.6 7 -7.4 -13.6 -12.6 -18.3 1.0 -0.5 8 -6.7 -13.5 -13.2 -19.1 1.1 -0.5 12 10.4 -1.9 -10.8 -17.6 0.6 0.1 16 9.3 -3.8 -12.4 -18.6 0.8 0.1 20 8.9 -3.7 -10.9 -19.0 0.8 0.0 24 -2.1 -10.3 -13.7 -22.1 1.0 -0.2 28 -26.8 -20.2 -19.1 -27.7 1.6 -1.4 Source: Simulations covering the 1967-73 period. It is assumed that the money stock is increased to accommodate the transactions demand of the original price increases. The increase in oil prices is the same as that assumed in Table 3.

samemanner as in the experimentreported in Table 3. In the firstsimula- tion, the residualsof the price equationfor the period of price controls wereadded to the priceequation; in the secondthey were not. Therationale forthis procedure is thenegligible character of theresiduals of theprice equation priorto the introductionof the controls,revealed in Figure2, so thatwe assumethat the errorsthereafter are due entirelyto controls.Under this assumption,the differencebetween the two simulationsmeasures the effects of the controlsin the full system.These are summarizedin Table 10. Over the whole intervalthrough the end of 1973, the price controls raised real output by an estimatedcumulative total of $17 billion over what it would have been in their absence. The price of consumption goods (P,0n)ends up 3.8 points (2.5 percent)higher without the controls, afterbeing as much as 5.1 points (3.7 percent)higher at a time coinciding with the end of PhaseII. By the end of the period,prices are risingsome- James L. Pierce and Jared J. Enzler 51

what less in the simulationwithout price controls.This observationmay be misleading,since in the last four quartersof the simulations,the economy moves from the restrictivePhase II to the less restrictivePhases III and IV. The behaviorof compensationper manhourmay give a betteridea of the differencebetween the ending rates of inflation.By the end of the period,the rate of wage increaseis the same with or withoutthe controls. In their absence,higher prices would have drivenup interestrates for a givenmoney stock, whichin turnwould have loweredoutput and employmentand eventuallyslowed inflation. This depressingeffect is partiallyoff- set, however,by the higher level of corporate profits and stock prices in the no-controlssimulation, which stimulatesconsumption and investment demand.The differenceis greatestin the last quarterof 1972,when corporateprofits are $20 billionhigher without the controlsand the value - ~~~~~.)~ 3 ri r.'I en t a s tI en S X > b t?t B _4 B _~~~~~~V4V oB

~~~~~~~Nent- en t- as t- t- N O

;Op O 1 1 s .O

S..~~~~~~~~~~2

5 _

^ C.) I0i 00 t- -0 N 0O

E RtQXo St c; tmbo p 4 -4 q 00 0W -- '-'4- 8. .= on 0- n o Xtt r--.

;~ .E9t S 3 oo oo zo' r.oo00 00n oo00 oooo0000) {2 .J6 c; I; oC9I*_ 5t

o. .~~~~~~~~~5; , 1 B ) oot r m- oo PO c;C0 gs1 CC. oo0n 3 dics Y0, n t N ^ t N ^ t eO ._.^'w

- - 0~~~~0

*0~~~~~~~~~~~~~~~~

C)~~~~~~~~~ oe " NC ' JamesL. Pierce andJared J. Enzler 53 of common stocks is $140 billion higher.Near the end of the simulation the depressingeffects of the high level of prices, acting throughinterest rates,become the predominantfactor. Given our assumptionof no change in the money stock betweensimulations, high inflationin the earlypart of the no-controlsimulation must be offset by low inflationlater. This adjustment will requirea period of higherunemployment rates, which are observedthroughout the simulationperiod and whichreach a level that is 0.9 point higherby the end of 1973.Given the growthin the money stock, theprocess in whichunemployment rises and eventually slows inflationtakes much longer in the presentcase than it does in the case of higherprices for imported oil. The reason is that inflationarydisturbances that are purely domesticredistribute income within the economy betweenprofits and personalincomes; while a rise in prices of importedoil reducesreal national income and wealth, which immediatelyhelps set in motion the deflationaryforces that raiseunemployment. This exercisemay overstatethe price effectsof the controls.First, with controls,the pricesat whichtransactions took placemay have exceededre- corded prices because they reflectedthe eliminationof discounts and deteriorationof quality.If this phenomenonis important,the residuals in the price equation, which we interpretedas the direct effect of the controls on the price level, are smallerthan they appear.Further, it is not reasonableto expecta changein the inflationrate broughtabout by an administrativesqueeze on profitsto be repeatedor even maintained. If that is how the controlswere working, participants in wage bargaining might not have extrapolatedthe effect in the manner suggestedby the price term in the wage equation.The simulationmay thereforeoverstate the indirect downwardpressure on wage increases that derived from controlson prices. However,there is a counter-argument.We have ignoredthe possible influenceof the unusualincreases in import prices in the price-control period.These increases may have driven up the pricesof import-competing goods. Our priceequation does not capturethis effect;had therenot been controls the increasein importprices might have producedeven higher prices than those predictedin the no-controlssimulation. To the extent this is true we have understatedthe effectsof the controls. All in all, the exercisestrongly suggests that the controls considerably improvedthe inflation-unemploymenttradeoff while they were operative, particularlyduring Phase II. Whetherthe controls were a good idea is 54 BrookingsPapers on EconomicActivity, 1:1974 quite anothermatter. They have obviously caused some unwanteddis- tortions and if they had been continued without modification,the dis- tortions would have been likely to grow worse. In addition, the total removalof controlsin 1974poses new inflationaryproblems. The resultsof the varioussimulation experiments strongly suggest that the economy will respond differentlyto the decontrollingof domestic pricesthan it will to the increasedprice of domesticoil. Put most simply, the initial reduction in aggregatedemand should be smaller and the upwardpressure on pricesgreater for a price rise coming from decontrol than for a rise of the same size coming from the increasein oil prices. In 1974,policy must cope somehowwith the effectsof both these shocks at once. Commentsand Discussion

DavidI. Fand:Pierce and Enzleranalyze how externalshocks may impart inflationaryspurts to the economy. One problem is to distinguishthe effects on relativeprices-high prices-from the effects on the inflation rate-rising prices. If there were no lags, so that these externalshocks had immediateeffects on the economy, they could be rankedin terms of their potentialcontribution to inflation:a once-and-for-allchange in the relativeprice of a commoditycauses high pricesand should theoreti- cally have the least impact on the inflationrate; while an externalshock that causesprice acceleration should have the greatestinflationary impact. Becauseof the interactionsand the lags, however,a one-shotincrease in relativeprices may, in fact, affect the rate of inflationover a fairly long period.There is also a seconddifficulty. Depending on the circumstances, the shocks consideredin this paper may incorporatedifferent combina- tions of these effects:thus when a cartel raisesthe price of oil, the move may constituteeither a one-shotjump or the first in a continuingseries of rises; and the impositionof wage and price controls may be eithera one-timeinfluence or a continuingfactor affecting all prices. Pierceand Enzlerintroduce a money demandfunction that is designed to pick up any increasein money demandresulting from a rise in transactions associated with the higher import prices. Since import prices have risen by as much as 10 percentin recent quarters,the demandfor nominal money should have increasedsignificantly. While the Pierce- Enzler money demandformulation stresses the rise in transactions,the point they are making about the greaterdemand for nominal money is very similar to a view expressedrecently that 6 percentmoney growth whenthe inflationrate is 8 percentconstitutes a destructionin realbalances of approximately2 percentand is deflationary.Thus while a policy of 6 percentmoney growthis seen as liberalwith, say, a 2 percentinflation 55 56 BrookingsPapers on EconomicActivity, 1:1974 rate, it is viewedas deflationaryin an economywith an 8 percentinflation rate. Since Pierceand Enzlerframe their money demand analysis in terms of transactions,they argue that a rise in the price of imports increases transactionsand requiresan accommodatingincrease in nominalmoney. I believe that this change in money demandmay be viewed either as a sort of real-balanceeffect or as a transactionseffect. The resultsobtained by Pierce and Enzler for the cartelizationof oil importsand for wage and price controlsmay appearsurprising. In fact, consideringthe resultsof the impositionof an oil cartel shown in their Table 3, it is not clear whetherone is viewingthe effects of inflationor deflation.Inflation accelerates so sharplyin the early periods that the subsequentperiods are dominatedby deflationaryafter-effects. The actual deflationscenario in the twenty-eight-quartersimulations shown in the tables varieswith the particularassumptions introduced in each simulation. One would imaginethat if an oil cartel raises prices and generates one set of macrodynamics,the loweringof pricesthrough wage and price controls should generatethe opposite kind of effects.But the wage and price controls simulationin Table 10 is not quite the inverse of the oil cartelcase. The differencesreflect interest rate effectsrelated to the Pierce- Enzlermoney demandfunction, the differentorders of magnitudeof the two shocks, and the impact on profits and their distributionin the two cases. A comparisonof these simulationresults suggests that they are sensitiveto the assumptionsmade in introducingthe shocks arisingfrom the oil cartelizationand wage-pricecontrols. Pierceand Enzlerquestion the definitionof monetaryaccommodation, andpoint out someof the ambiguitiesin this concept.Given a recognizable and substantialexternal jolt, like the oil embargo,that could have significant effects on the economy, the monetaryauthorities are generallyex- pectedsomehow to accommodateit. But accommodationis very difficult to definein detailfor the concretecases. In one of the simulationsmone- tary accommodationis introducedin the oil cartel case by adding an incrementto the growthof money stock in the first quarterequal to the percentageincrement in current GNP plus imports arising from the higher oil prices. Obviously, accommodationcan be defined in many other ways, dependingon one's views about interest rates, prices, unemployment,and inflation. A relatedproblem is the definitionof a constant policy. Pierce and JamesL. Pierce andJared J. Enzler 57 Enzlertake the 1967-73rate of monetarygrowth to representa constant monetary policy in their simulations.But other alternativesmight be interesting.Thus one could try a simulationexperiment that embodies relativelyunchanged aggregate demand in money terms-that is, MV- or unemploymentor the rate of inflation.A comparisonof these alternative simulationexperiments might be useful. The simulationsare all conductedin terms of the growthrates in M1. Since M1 divergedfrom M2 several times in the 1967-73 period, the simulationresults with an M2 policy wouldbe interesting. Pierce and Enzler take up the oil cartel and wage-pricecontrols to analyzethe cumulativeeffects on the pricelevel and othermacroeconomic variables.They are interestedin seeingthe impactthese jolts have on the inflationrate and what kind of monetarypolicy may be appropriateto minimizetheir undesirablerepercussions. An alternativeprocedure is to turn the simulationanalysis around-to ask how much of the observed 1973inflation can be attributedto the domesticpolicy and how much to externaljolts. The simulationsreported in this papercover the 1967-73period, when there were essentiallyno supplyshortages and no bottlenecks.As Pierce and Enzlercaution, their resultsmay not apply directlyto 1973, which sufferedfrom both of these problems.Failure of supplyto respondand the impact of the restrainedsupply on the price level are factorsthat do not operate ordinarily,and that were not presentin the period of the simulation.It is thereforedifficult to apply the simulationanalysis to the period since 1973, even if the simulationresults appear reasonable for a more normalperiod. The Pierce-Enzlerpaper considers how externaljolts affect prices and the rate of inflation,and thus illustrateshow the externalenvironment can affectmovements in the pricelevel of a particularcountry. If a country is willingto take sufficientlystrong domestic measures, it may be able to overcomethese external effects, but the requiredmeasures may be extreme. The Pierce-Enzlerresults may thereforebe viewed in a largercontext as illustratingsituations in which a country, in effect, loses control over its pricelevel. The Pierce-Enzlerexperiments are for the U.S. economy- that is, a large country with a relativelysmall foreign sector; and yet even in this case these externalinfluences may have relativelylong-lasting effectson the pricelevel and on the rate of inflation.At the otherextreme, 58 BrookingsPapers on EconomicActivity, 1:1974 for a smallcountry with a largeforeign sector, it is much easierto imagine a significantimpact of the externalenvironment on developmentsin the domestic price level. Pushing this idea somewhat,it seems to suggest that a country'scontrol over its pricelevel may weakenas its dependence on the outsideworld deepens. Perhapsthe most interestingand the most importantlesson of the Pierce-Enzlersimulations is the dramatizationof the difficultiesin formu- latingpolicy to take accountof even large and recognizabledisturbances. The resultsobtained with the varioussimulations depend very much on the particularassumptions used to analyze the external shock and its impact. But even though the kind of inflation and the other macro- economicconsequences associated with a givenset of disturbancesmay be uncertain,the monetaryauthorities must neverthelessconsider alternative accommodatingmeasures and framepolicy to keep the economymoving on someacceptable track. This paper does, therefore,point up the dilemma of choosingamong policy alternatives with unknown consequences to cope with externaldisturbances whose effectsare also uncertain.

R. J. Gordon:The Pierce-Enzlerpaper is a very timely and important one. The qualitativeresults they presentare hard to quarrelwith. How- ever, I do want to question some of their specific quantitativefindings and focus on certainaspects of the model they use which I think could be improved. Table3 shows the effectsof a one-timeincrease in the price level without any accommodativeresponse from the Fed-that is, the growth of the money supplyremains unchanged. The initial effect is an upsurge of both inflation-and unemployment,which seems entirelyreasonable. And, since this is a model with a naturalunemployment rate, in the long run the economy eventuallyreturns to the unemploymentand inflation ratesthat would have prevailedhad the initialjump in the pricelevel not occurred.The only surprisingthing about this eventualoutcome is that after seven years, the economyis still not very far along in the long-run adjustmentprocess. At that time the unemploymentrate (the last entry in Table3) is just startingto recedeand is still well above its controllevel. Anothersurprise in the resultsin Table 3 is the reductionin the real wage that occurs.Wages fall 3 percentbehind consumerprices with the increasein oil pricesand they nevercatch up. This gap developsbecause James L. Pierce and Jared J. Enzler 59 the wage equationexplains too much of the rate of change of wages in terms of the Phillips curve and too little in terms of past inflation. In this model, high oil pricesbring on a recessionand high unemployment, which then holds down wages. But I think we have seen a reductionin the effectivenessof high unemploymentrates in restrainingwages. In- creases in welfare benefits and in unemploymentcompensation have diminishedthe downwardpressure that unemploymentused to exert. If I am right, the true story is even more pessimisticthan the one told in this paper. Unemploymentwill take longer to push down wages and prices, and it might not peak out for nine or ten years ratherthan the six years shown in the model. But there are conceivableadjustments to the model that would act in the opposite direction,leading to more optimisticresults. If the price elasticityfor oil is largerthan Pierceand Enzlerallow, the demandfor imported oil will be less than that in their model and the inflationary effects of the oil embargowill be smaller.If one believes the price of imported oil will fall from the $10 level they assume, one would have anotherground for optimism. A separatepoint is the relationbetween real GNP and unemployment shown in the latter quartersof the simulationsin Table 3. Okun's law is clearly violated and there is too much unemploymentfor the real shortfallin output that is shown. If one believesthe output simulation, one can be more hopefulabout unemploymentthan this model is. Another reason for optimismis brought out by Table 6. Prices may affect wages throughthe effectsthat higher productprices have on the demandfor labor as well as throughthe outwardshift that highercon- sumer prices impart to the labor supply curve. This is particularlyim- portant if adult males exhibit an inelastic labor supply so that wages depend on the nominaldemand for labor. I think this view of the wage relation helps explain why wages have been rising so moderatelyover the last year. Table 9, which detailsthe effectsof a more accommodatingmonetary policy, illustratesthe magnitudesin the inflation-unemploymenttradeoff among which societymust choose. Comparedwith the resultsin Table 3, the unemploymentrate stays about 1 percentlower for five yearsand the inflationrate is about 1 percenthigher at the end of that time. The extra inflationis probablyoverstated here because of the relativeinsensitivity 60 Brookings Papers on Economic Activity, 1:1974 of wages to more unemployment that I have noted, but the lesson still comes through: only minimal inroads can be made against inflation with even an extended recessionary situation.

GeneralDiscussion

Walter Salant and others questioned the generality of the Pierce-Enzler formulation of the transactions demand for money. Salant suggested that some fraction of exports and imports should be included in a measure of transactions demand. To the extent that the demand is associated with financing production, exports should be included; and to the extent that it is associated with financing final purchases, imports should be included. GNP already includes exports fully and the Pierce-Enzler adjustment incorporates imports fully as well. But unlike goods and services that are produced and sold in the United States, exports and imports do not embody both these aspects of transactions demand. Jared Enzler conceded that Salant had a valid conceptual point, but argued that the exact proportions in which imports and exports should enter transactions demand could not be known, and that the simple procedure of adding all imports was a substantial improvement over ignoring them at a time when import prices were rising abruptly. James Pierce noted that the Federal Reserve is now collecting data on the ownership of demand deposits, which might fa- cilitate analysis of the distribution of money balances between households and business, and hence the nature of transactions demand. William Nordhaus asserted that the oil embargo would greatly stimu- late capital investment since, in most studies, capital and energy are substitutes in production. This upsurge of investment might be a significant offset to deflationary pressures arising from higher oil prices. William Brainard replied that the response of business investment in the aggregate was not so clear; the energy shortage was curtailing expansion plans of some industries, such as utilities. Enzler explained that he and Pierce regarded the immediate response of business investment as too uncertain to permit incorporating any special effects from that source in the model. However, even if investment were stimulated, it could not offset many of the complications arising from the higher oil prices. He noted that the possibility of a substantial induced rise in exports was modeled, and pro- James L. Pierce and Jared J. Enzler 61 vided some evidence on how a surge in demand in other sectors, such as business investment, would alter the main simulation. Salant noted that a policy to offset the effects of the rise in oil prices on domestic output and employment could not also restore the levels of consumption and investment. Higher oil prices caused a deterioration in the terms of trade that required a reduction in real consumption and investment. John Kareken added that the extent of such an effect would depend on what oil exporters did with their higher revenues. In the basic case of the model, it is assumed that they never expand their demand for U.S. goods so that no need arises to aim at a lower level of domestic consumption and investment. Nicholas Kaldor offered a more pessimistic view of price developments than that embodied in the model. In view of the British experience, he believed that industrial prices and wages would rise fully to reflect the rise in oil prices, pushing the price level substantially higher than Pierce and Enzler project. Wages would not be restrained significantly by modest increases in unemployment; and, because of the worsening in unemployment that would ensue, the central bank would not try to hold to a growth path for the money supply as Pierce and Enzler had assumed.