sign in sign up
<<
Home , Economic expansion, Recession, Robert Hall (economist)

ROBERT J. GORDON NorthwesternUniversity

Alternative Responses of Policy to External Supply Shocks

DURING 1973 and 1974 reductionsin suppliesof food (throughnatural causes)and of oil (throughunnatural causes) simultaneously lowered the real incomeof U.S. nonfarmworkers and raisedthe rate of .An inflation-cum-recessioninduced by lowersupplies of rawmaterials may call for a policyresponse different from the traditionaltonic of demandrestric- tion calledfor by a "garden-variety"inflation generated by excessdemand. In lightof the noveltyof the 1974situation, the sharpdivergence of policy recommendationsamong is not surprising.Some analyzedthe episodewithin the context of standardmacroeconomic analysis, treatingthe 1973-74acceleration of inflationas a delayedconsequence of the accelerationin monetarygrowth during 1972, and the 1974-75recession as a delayedconsequence of the sharpdeceleration in monetarygrowth that beganin June 1974.The policy advice of this group, consistinglargely of economistsgenerally identified as "monetarists,"was to maintaina con- stantor even slightlyreduced rate of growthof the moneysupply.' Arthur Note: This paper was supportedby National Science Foundation Grant GS-39701. It was inspired,as was a previouspaper in another area, as an attempt to reconcilethe views of and Arthur Okun. I am grateful to Michael Parkin and participantsin the Brookingspanel for helpful suggestions. 1. See Allan Meltzer,"A Plan for SubduingInflation" (a dialoguebetween Allan H. Meltzerand two editorialstaff membersof Fortune),Fortune, vol. 90 (September1974), pp. 112ff.In the same month, when the supply (MI) had risen 5.8 percent over the preceding twelve months, Milton Friedman wrote: ". . . until a few months ago at 183 184 Brookings Papers on Economic Activity, 1:1975 Okunput forth the contrastingview that an attemptby policymakersto maintainfixed growth in nominalincome ignored the "macroeconomicex- ternalities"of commodityshortages: total real output falls by more than the declinein farmoutput, through an extrainduced loss of nonfarmout- put.2An implicationof Okun'sargument is that, whilestabilization policy cannotre-create the lost farmoutput, it can minimizeor eliminatethe in- duced loss of nonfarm output by promoting a higher growth rate of nominal income. The inflationin 1973and 1974can be regardedas a combinationof an underlying"hard-core" inflation, inherited from the 1960s and perhaps aggravatedby the rapid pace of economicexpansion between 1971 and 1973,with a set of four temporary"bubbles": (1) the 1972-74shortfall of farmsupplies to U.S. consumers,caused in the firsttwo yearsby buoyant foreigndemand and in the thirdby domesticsupply ; (2) the re- strictionof oil productionenforced by the cartel of the Organizationof PetroleumExporting Countries (OPEC); (3) the end of priceand wagecon- trols in 1974; and (4) the devaluationsof the dollar in 1971 and 1973. Althoughthese events may have permanentlyraised the pricelevel, such a one-shotrise generatesonly a temporaryincrease in the rate of inflation.3 This paperdeals with the issues raisedby an inflationinitiated not by excessdemand but by commodityshortages. Although its formalanalysis treatsan externalshock that takesthe form of a declinein farmoutput, its basicconclusions apply with only minorchanges to the casesof oil and de- valuation.What policies are availableto minimizethe indirecteffects on output?What are the conditionsunder which expansivepolicy actions best, these high interestrates have been accompaniedby extremelyhigh rates of mone- tary growth.... Recent rates of monetarygrowth are not too low. If anythingthey are still too high to bring inflation to an end in a reasonableperiod of time." See Milton Friedman,"Is MoneyToo Tight?"Newsweek, vol. 84 (September23, 1974),p. 82. Fried- man'sstand on monetarypolicy was takendespite his recognitionthat specialfactors had contributedto the 1974 inflation.He attributedroughly half of it to increasesin oil and food , to the lifting of controls, and to precautionaryincreases against re- newed price controls. See Milton Friedman,"Inflation Prospects," Newsweek, vol. 84 (November4, 1974), p. 84. 2. ArthurOkun, "Incomes Inflation and the Policy Alternatives,"in "The Econo- mists Conferenceon Inflation, September5, 1974, Washington,D.C.; September23, 1974,New York, New York," vol. 1, "Report"(1974; processed),pp. 365-75. A formal analysisof the externalityargument is presentedbelow. 3. The list could perhapsbe expandedby two smallerbubbles-the increasesin prices in fear of reimpositionof controls, and the overshooting of prices beyond the levelsjustified by shortagesdue to speculativeinventory hoarding. Robert J. Gordon 185 takento counteracta temporarydecline in farmoutput will causea perma- nent increasein the rateof inflation?What are the relativeadvantages and disadvantagesof income-taxreductions, food subsidies, and expansive monetarypolicy as policyresponses? Finally, how woulduniversal escala- tion (or "indexation")of contractsaffect the resultsof the analysis?

The PolarCases

To establishthe range of possibilities,the followingtwo sectionscom- pare the responsesof two hypotheticaleconomies, one with perfectflexi- bility of pricesand wagesand the other with absoluterigidity in the non- farm sector. These cases serve to illuminatethe more complicatedand relevantanalysis of a realisticeconomy in whichnonfarm prices and are neitherperfectly flexible nor absolutelyfixed.

PERFECT PRICE FLEXIBILITY

Theeconomy encounters no problemsin adjustingto an externalshock- say, a crop failure-if both farm and nonfarmprices and wages are per- fectlyflexible. In this case the marketfor nonfarmgoods and labor always clears,and no involuntaryunemployment can arise.A briefexamination of this case servesas a point of comparisonwith the diametricallyopposite case of fixed prices. The treatmentof all cases incorporatesseveral common assumptions. The economyis closed, with all outputof both sectorsproduced and con- sumedin the domesticeconomy. Farm output is exogenous,produced by a factorthat is not mobilebetween the two sectorsand consumedentirely in the nonfarmsector. The exogenoussupply of farm output, QFI is equated to the demand:

= (1) QF AN (PN') whereA is a constant,ao is the nonfarmincome of demandfor farmproducts, a, is the absolutevalue of the priceelasticity (which through- out the paperis assumedto be less than unity),and PF andPN are, respec- tively, price indexes for farm and nonfarmoutput. A rearrangementof 186 BrookingsPapers on EconomicActivity, 1:1975 (1) relates the -clearingrelative price, PF/PN, to the exogenous supplyof farmoutput and the level of nonfarmoutput, QN:

(2) PF= [AQN } Ni QF For any given supply of farm products,an increasein nonfarmoutput raisesthe demandfor farm products,and hence the relativeprice, by an amountthat dependspositively on the incomeelasticity, ao, and negatively on the priceelasticity, a,. The relativeprice depends, in part,on the level of nonfarmoutput, except in the specialcase of a zero incomeelasticity. Nonfarmoutput is assumedto be producedwith labor and some other fixedfactor, like .Knowledge and technologyis assumedfixed, so that labor inputdetermines nonfarm output. Given the population,if the supplyof labor does not respondto changesin the real wage, both labor inputand nonfarmoutput are fixed. In this case, a crop failurechanges the relativeprice of farmproducts but not the level of nonfarmoutput. Since the wagerate that nonfarmfirms can affordto pay to a given numberof workersis limitedby nonfarmprices, any increasein the relativeprice of farmproducts reduces the real wage of workers,when the latteris defined in terms of a consumerprice includingboth farm and nonfarm products. If, however,a lowerreal wage causes workers to reducetheir labor input, eitherby withdrawingfrom the labor force or by workingfewer hours per week, a crop failuremust reducenonfarm output.4 This responsein the nonfarmlabor marketthus providesa second relationshipbetween non- farmoutput and the relativeprice of farmproducts, in additionto equation (2) above, allowing the simultaneousdetermination of both variables.5 Hence,output and relativeprices in each sectorare beyondthe controlof

4. A third case, not discussedhere, is a negativelysloped labor supply curve. Most cross-sectionevidence for the appearsto supporta verticalcurve for adult male workers,a positively sloped response of women and teenagersto an increase in their own real wage, and a negativeresponse of wives to an increasein their husbands' real wage. See the evidence cited in Robert J. Gordon, "The Cost of Higher ,"BPEA (1:1973), table 2, p. 159. 5. The exact form of the second relationshipis zkPF~~~~_ 0-(-1) /(b+e)

V PN / whereD is a constant, k is the share of farm productsin consumerexpenditures, and b and e are, respectively,the elasticitiesof the nonfarmlabor demandand supply curves. RobertJ. Gordon 187 policymakers.If the choices of individualsbetween leisure and labor are sociallyaccepted, any reductionin employmentcaused by the voluntary withdrawalof laborinput in responseto a lowerreal wage is of no concern for stabilizationpolicy, sincethat reductionis purelyvoluntary. What, if anything,can stabilizationpolicy accomplishwhen nonfarm pricesare perfectly flexible? Aggregate-demand policy controlsthe level of nominalincome (that is, grossnational product in currentdollars), which is sufficientto set the nominalnonfarm sincethe valuesof all real variableshave been determined.If policymakersfollow a rulethat calls for constantnominal income, then a crop failuremust cause nominal nonfarm pricesto fall, but the overallaverage price level must rise.6If, on the other hand,policymakers achieve constant overall prices by reducingnominal in- come, they would preventa redistributionof income from creditorsand pensionersto debtors.Even if the expectedrate of inflationand the level of the interestrate are unaffected,the higherthe price level, the smallerthe fractionof incomea debtorwill requireto servicehis . Whetheror not the labor supply shrinksin the flexible-pricecase, the welfareof nonfarmworkers is reduced.7Not only does a cropfailure reduce totalreal output, but also, as as the demandfor farmproducts is price inelastic,it transfersincome from workers to farmers,who enjoya windfall. Whilethe problemis not one of stabilization,society might wish to reduce or eliminatethe transferby a redistributivetax policy that, for example, levies a windfall-profitstax on farmersto financea subsidyon nonfarm productspurchased by nonfarmworkers. However, the case for redistribu- tivetax-subsidy schemes is not obvious,nor is therean obviousline between temporaryevents justifying redistribution and those that do not.

COMPLETE WAGE AND PRICE RIGIDITY In the case of perfectprice flexibility, nonfarm output is eitherfixed or determinedby workers'decisions about labor supply,leaving the nonfarm pricelevel to be determinedby stabilizationpolicy. If, on the otherhand, 6. The nonfarmprice level falls if the price elasticity of demandfor farm productsis (approximately)less than unity; the overallprice level must rise, becausereal output has fallen and nominalincome is assumedconstant. 7. Althoughworkers who reducelabor input obtain leisureworth the real wage at the margin,they lose partof theirproducers' surplus earned on inframarginalunits of work. In parallelfashion, farmersgain a producers'surplus from the increasein the relative price of their output. 188 Brookings Papers on Economic Activity, 1:1975 the nonfarm wage rate is rigid and nonfarm prices are "marked up" over the wage rate by a constant fraction, then nonfarm prices are fixed and non- farm real output is determined by . Nominal income, Y, is the sum of total nominal spending in each sector:

(3) Y= PFQF + PNQN; equation (2) can be substituted into (3) to obtain

(4) YIPN = [A QNaOQF (1 a1)](la1) + QN* When nominal income is held fixed by a policy rule, the wage rate and nonfarm prices are rigid, and the demand for farm products is income and price inelastic (a0 < 1 and a, < 1), then nonfarm output varies in the same direction as farm output, even if the supply of nonfarm labor is completely unresponsiveto chan.gesin the real wage. Since the of farm output rises and nominal income is fixed, the value of nonfarm output must fall. With nonfarm prices rigid, nonfarm output must drop, causing involuntary un- . The crop failure thus carries with it a real ""effect. Just as stabilization policy can alter nominal nonfarm spending and the price level of the nonfarm sector in the flexible-pricecase, so it can alter that sector's nominal spending, real output, and employment in the rigid-wage case. In this extreme case, the multiplier can be derived when the market- clearing condition for farm output, (2) above, is written in the form of percentage changes:

(5) PF PN aqF+ aoN where lower-case ps and qs denote percentage changes between the initial situation and the new situation after the crops have failed:

PF = (PFI - PFO)/PFO. If policymakers hold nominal income constant, the change of nominal in- come-that is, a weighted average of spending in the two sectors as defined in (3) above-must be zero:

(6) sY = = k(PF + qF) + toak)peN + qS), where k is the share of farm spending in total spending. Substituting (5) RobertJ. Gordon 189 into(6) yields,after some rearrangement, the percentagechange in nonfarm outputrelative to the exogenouschange in farm output:

- a,) (7) q.N_ = k(l qF kao + (1-k)ala To takea simpleexample, assume that the initialshare of expenditurein the farmsector, k, is 10 percent,and that the incomeand priceelasticities are,respectively, zero and 20 percent(ao = 0 and a, = 0.2). In this casethe elasticityof nonfarmoutput to a changein farmoutput is 4/9. Withinitial levelsof expenditureof $100billion and $900billion in the two sectors,a 10 percentloss in farm output($10 billion) causes a 4.44 percentdecline in nonfarmoutput ($40 billion). Thus the , C, of the $10 billion crop failureis

(8) C =-[kq, + (1- k)q,]Y = ak _ $10 billion =50 bilion. - 0.2 =$0blin Sincethe nonfarmprice level is rigid,policymakers can fully offset the multipliereffect of the crop failureon nonfarmoutput with no deleterious sideeffects. Nominal income must simplyincrease sufficiently to leavenon- farmoutput unchanged by the crop failure.This "fullyaccommodating" policyresponse can be calculatedfrom (6) whenqN (as well as PN) is equal to zero: k(1 - a,) (9) y = k(p + q,) =-q al With the parametersof the previousexample, nominal income should be raisedby 4 percent-$40 billion-to counteractthe $40 billionloss of non- farmoutput that would have occurredhad nominalincome been allowed to remainfixed. Theconsumer price index, an averageof the fixednonfarm price and the higherfarm price, must rise, and policymakerscannot avoid acceptingthis higheroverall price level, just as they cannotre-create the lost crops.8But 8. A positivevalue for the income elasticityof demandfor farm productsreduces the multiplier,since lower nonfarm output moderates the iilcrease in the relative price neededto clearthe farm output market,and this in turn releasesmore of the fixed level of nominalincome for the supportof nonfarmoutput. When ao = 0.2, the elasticityof nonfarmoutput is reducedfrom 4.44 to 4.0 percent,the social cost from $50 billion to $46 billion, and the necessarynominal income offset from $40 billion to $36 billion. 190 BrookingsPapers on EconomicActivity, 1:1975 stabilizationpolicy can eliminate the wasteful"multiplier" loss in nonfarm outputand associatedinvoluntary nonfarm unemployment by providing enoughextra nominal income to makeroom for both the originallevel of nonfarmspending (fixed price and initialreal output)and the higherlevel of spendingon farmproducts.9

PartialPrice Adjustment

NO COST-OF-LIVING EFFECT ON WAGES At thispoint the policymakeris tornbetween the conflictingadvice of the flexible-pricemodel, whichrecommends a reductionin nominalincome to stabilizethe price level, and that of therigid-price model, which recommends an increasein nominalincome to avoid involuntaryunemployment. The simplestintermediate model allows the rate of change of nonfarmprices (PN,where small letters now denotepercentage changes per unit of time)to adjustby a fraction,X, of the differencebetween the market-clearingvalue of the flexible price, N, and the current price, PN:

(10) PN = X(PN -PN) Whennominal income is held constant,&N during the periodof the crop failurelies belowthe initialnonfarm price level (PNo)and the rateof change of nonfarmprices is negativeuntil they arebrought into line withPN. Since PNOlies above the market-clearingvalue, N, the initial consequence of the crop failureis a declinein nonfarmoutput and the creationof ,as in the rigid-priceanalysis of the previous section. Throughtime, however,downward adjustment of the nonfarmprice level makes more of nominal income availablefor nonfarmoutput, and the severityof the recessionis graduallymitigated. Finally, PN ends its decline whenit reachesits market-clearinglevel, N, at whichpoint involuntary un- employmentis eliminated.The processis reversedwhen the cropsreturn to normal;at the low nonfarmprice level, N, the constantlevel of nominal incomeallows nonfarm output to rise aboveits initialvalue, and an output and employment"boom" continues until PN has returnedto PNO The temporaryrecession, as well as the subsequenttemporary boom in

9. Nonfarmoutput might have fallenas in the flexible-pricecase if the supplyof labor were voluntarilyreduced in responseto the lower real wage. Robert J. Gordon 191 output,can be eliminated,as describedin the previoussection, by a policy of accommodatingnominal income. If nominal income is raised by the amountcalculated in equation(9), the market-clearingvalue of PN during the periodof the cropfailure is by definitionequal to the initialprice level, PNOand no downwardadjustment in nonfarmprices takes place. Now a policyof accommodatingnominal income imposes on societythe cost of a higherprice level than one that aims at constantnominal income, and a moresubstantial (albeit temporary) redistribution from creditors and pen- sionersto debtors.The choicebetween the policieshas no long-runconse- quencesfor the level of pricesor output,or for the rate of inflation.10

SOME COST-OF-LIVING EFFECTS ON WAGES The previoussection assumesthat higher farm prices have no direct effecton nonfarmwages and prices,and thus ignoresthe possibilitythat a policy of accommodatingnominal income may permanentlyincrease the rate of inflation.As a point of departurefor developinga more realistic mechanismfor adjustingnonfarm prices, which allows for the possibilityof an equilibriumnonzero inflation rate, (10) may be reformulatedas

(I11) PN P + jZ5 wherepN* is the rate of changeof the expectednonfarm price level, Z is the excessdemand for labor,and] is an adjustmentcoefficient. Assume that the expectedlevel of nonfarmprices remainsconstant (p, = 0) after a crop failure;then, so long as the pricelevel is aboveits market-clearingvalue- PN > P.T in (10)-the resultinginvoluntary nonfarm unemployment means thatZ <0 in(1). Equation(11) is simplyan "expectationalPhillips curve," the of whichhave receivedextensive analysis and empiricaltesting in recent years.A slightlymore complicatedbut substantiallymore realisticversion can be developedif (ignoringproductivity change) it is assumedthat the rate of growthof the wage rate, w, is equal to that of the expectedprice level plus a fraction,j, of the excessdemand for labor,Z: (12) w=p*+jZ. Theexpected price level relevant for wagedecisions is a weightedaverage of

10. Such consequencesmight ensue to the extent that the -inducingpolicy cuts real investmentand thus endows future generationswith a lower capital stock. 192 Brookings Papers on Economic Activity, 1:1975 theexpected nonfarm price, PN, which defines the valueof labor'smarginal product,and the expectedconsumer price index, P*, adjustedfor the -taxfactor, T*, usedby workersto calculatetheir real after- wage rate.Thus (12) becomes11

(13) w - g(PZ + t) + (I When the coefficientg is greaterthan zero, the wage rate dependsnot only on the nonfarmproduct price, but also on farmprices and the pay- roll tax rate.In the extremecase, wheng has a value of unity,all of the in- creasein consumerprices relative to nonfarmproduct prices resulting from a cropfailure is passedthrough to the wagerate, and real wages do not fall. NVhenthe wage equationis interpretedas the adjustmentpath in a neo- classicalmodel of the labormarket, the parameterg is the ratio of the elas- ticityof the laborsupply curve to the sum of thatelasticity and the elasticity of the demandcurve, and is zerowhen the supplyof labordoes not respond to changesin the real wage.12But in alternativelabor marketsettings the value of g might be nonzeroeven if labor were suppliedinelastically. In unionizedindustries, for instance,the strikeweapon might be used to pass throughsome or all of an increasein farm pricesin higherwages. Quite apartfrom unions, competitivefirms might offer risk-averseemployees a wagecontract indexed to the consumerprice index, trading this real-wage insurancefor a reductionin the averagereal wage.13 The followinganalysis

11. Equation(13) has been estimatedin Robert J. Gordon, "Inflationin Recession and Recovery,"BPEA (1: 1971),table 1, equation(11). The equationhas also been used in empiricalwork for the United Kingdomby MichaelParkin and his collaboratorsand has been derivedexplicitly in Michael Parkin, Michael T. Sumner,and R. Ward, "The Effectsof ExcessDemand, Generalized Expectations, and Wage-PriceControls on Wage Inflationin the U.K.," in Karl Brunner(ed.), a conferencevolume on controls (Amster- dam: North-Holland,1975), forthcoming. 12. A more complexversion with severalvarieties of , cyclicalvariations in pro- ductivitygrowth, and other complications,is analyzed in Robert J. Gordon, "Inter- relations between Domestic and InternationalTheories of Inflation,"in R. Z. Aliber (ed.), ThePolitical Economy of MonetaryReform, forthcoming. 13. The idea of "wage insurance"'as an explanation of rigid wages was developed simultaneouslyand independentlyby C. Azariadis,"Implicit Contracts and Underem- ployment,"Journ2al of ,vol. 83 (1975), forthcoming;Martin N. Baily, "Wagesand Employmentunder UncertainDemand," Reviewof EconomicStudies, vol. 41 (January1974), pp. 37-50; and Donald F. Gordon, "A Neo-Classical Theory of KeynesianUnemployment," in Karl Brunner and Allan Meltzer (eds.), The Phillips Curveand Public Policy, Carnegie-RochesterConference Series, vol. 1 (Amsterdam: North-Holland,1975). RobertJ. Gordon 193 will discussthe consequencesof differentvalues of g as thoughthey result from an expectationalmechanism in wage bargaining,but the interpreta- tion couldreadily be adaptedto cover othercases. An equationfor the price of outputin the nonfarmsector is now required. In line with considerableevidence, the nonfarmprice level is set as a "'markupfraction" multiplied by "standard"unit labor cost-that is, the wage rate dividedby productivityat some "standard"level of -withthe size of the markupfraction dependent on the demand for .14Assuming a constant level of standardproductivity (equalto 1.0), the priceequation becomes

(14) PN= wxc, whereX is an indexof excesscommodity demand and c is the percentage responseof the inflationrate to the rate of growthof output. Whenthe wageand priceequations are combinedwith the definitionof consumerprices,

(15) PC =kp-k) a relationshipbetween changes in nonfarmand farmprices is obtained:

(16) PN = (1 - gk)p* + g(kp* + t*) + jZ + cx. As in equation(11), the basicforce that allowsinvoluntary unemployment to persistis the partialdownward adjustment of pricesin the face of excess labor (and commodity)supply. What differenceis made by a value of g greaterthan zero? The analysisis identicalto that of (11), of course,if the expectedfarm price is unaffectedby a temporaryincrease in the actuallevel. Onthe otherhand, a cropfailure may lead individualsto reviseupward the levelof farmprices that they expect during their wage contracts (in 1972-74, U.S. domesticfood consumershad "threelean years").In this casea "wage push"is exertedby farmprices, which raises the nonfarmprice level above the adjustmentpath describedby (10) and (11), in turn "usingup" more of the fixedlevel of nominalincome, raising the multiplier,and aggravating the recession. If g is positiveand if expectedfarm prices respond to the higheractual level,the resultsdepend on how expectationsadjust to pricechanges in the nonfarmsector. One possibility is thatexpectations adapt to past changesin

14. See the evidencepresented in Gordon, "Inflationin Recession and Recovery," p. 129. 194 Brookings Papers on Economic Activity, 1:1975 nonfarmprices. The expectedlevel of nonfarmprices for the next period would then be set equal to the currentlevel extrapolatedby an expected rate of nonfarminflation estimated from its past rate. Just after a crop failure,such adaptivenonfarm expectations would worseninflation, since nonfarmprice expectations would be raisedin responseto the highercur- rentprice level caused by the feedthroughof farmprices to wages.15And, if nominalincome is heldconstant, the higherlevel of nonfarmprices worsens the initialstages of the recession.But soon the adjustmentof nonfarmprice expectationswould begin to operatein the oppositedirection, reducing in- flationand the magnitude of the recession,since it wouldamplify the down- wardadjustment of nonfarmwages and pricesin responseto excesslabor supply. In ,adaptive nonfarm price expectations amplify the fluctuationsin nonfarmoutput and pricesin responseto a crop failureas long as nominal incomeis held constant.If, on the otherhand, policymakers pursue a fully accommodatingpolicy for nominalincome, which prevents the emergence of excesslabor supply, raise expected nonfarm prices- the "base"around which the adjustmentof pricestakes place-and endow the economywith a permanentlyhigher price level. So long as the crop failureis temporary,the rate of inflationis not permanentlyaffected, since the declinein farmprices at the end of the failurefeeds through to expecta- tions and ends the upwardadjustment of expectednonfarm prices. But an accommodatingpolicy for nominal income wouldpermanently raise the rateof inflationin the case of a permanentsupply reduction, brought about, for example,by an eternaloil cartel.

The Potentialfor Tax Policy

In any realisticcase, a policy accommodatingnominal income (such as an increasein the money supplysufficient to eliminatethe nonfarmmulti- pliereffect of a cropfailure) has the disadvantageof raisingthe pricelevel

15. Correspondingto (7) above is a multiplierformula that takes into account the feedthroughof farm prices to wages (but not the effect on prices of excess labor or commoditysupply): fN= k[(l -ai)(1 -gk) +g] l qF aogk + [kao + (1 - k)al](1 - gk)f Comparedto the case a, = 0.2, ao = 0.2, and g = 0, which yields an elasticity of 0.4, the 0.2 value for g (assumed in the simulation below) increasesthe elasticity to 0.492. Robert J. Gordon 195 relativeto an alternativepolicy aimed at constantnominal income. Changes in tax ratesand subsidies,on the otherhand, not only operateon income but also can directlyalter the price level. A reductionin the payrolltax rate, for instance,narrows the "wedge"between market prices and after- tax factorcost, and henceallows firms to chargea lowerprice while paying workersthe sameafter-tax wage rate. Along with a reductionin tax rates, policymakersmust take steps (cutting governmentexpenditures or the moneystock, for example)to maintain,as I shall assume,an unchanged path of nominalincome.16 A reductionin taxeswill lowerconsumer prices most if appliedto those taxeswhose burden is borneby consumersrather than factorsof produc- tion. At one extreme,changes in state salestaxes are probablyshifted for- wardto consumerprices by nearly100 percent, while at the otherextreme, changesin the corporationincome tax affectmainly capital income and do not appearto be substantiallyshifted forward.17 The personalincome tax is an intermediatecase and appearsto be shiftedforward to consumersby roughly20 percent.18In the absenceof a universalfederal sales tax, the policyoption that would the greatestreduction in pricesfor a given loss of revenuewould be a federalgovernment bribe to inducereductions in stateand local sales taxes. If this mechanismwere rejectedas administra- tivelyclumsy or politicallyinfeasible, the federalgovernment could subsi- dize nonfarmoutput to offset the impactof the higherfarm prices on the consumerprice index.19 A constantnominal income would thereby be suffi- cient for both the higherfarm-price level neededto clear that marketand the originallevel of nonfarmoutput, since the after-subsidynonfarm price wouldbe pusheddown to the market-clearinglevel, PN. The size of the re- quiredsubsidy relative to GNP is givenby equation(9)-for instance,$40 billionin the simpleexample spelled out above. Possibly,such a subsidycould be financedby a windfall-profitstax on

16. In principle,if no offsettingaction is taken, the price level may be eitherraised or lowered.See Alan S. Blinder,"Can IncreasesBe Inflationary?An Expos- itory Note," NationalTax Journal,vol. 26 (June 1973), pp. 295-301. 17. RobertJ. Gordon,"The Incidenceof the CorporationIncome Tax in U.S. Manu- facturing,1925-62," American Econiomic Review, vol. 57 (September1967), pp. 731-58. 18. See Gordon, "Inflationin Recession and Recovery," table 1, where the tax co- efficientrefers to the personalincome tax plus the social securitytax paid by employees. 19. A subsidyfor farm productswould raise demandabove the reducedsupply and hencewould be infeasiblewithout a commodityinventory or bufferstock. A subsidyfor nonfarmproducts would not requirehigher nonfarm output than initially, but would simply offset the multipliereffect and allow the original full employmentlevel of non- farm output to be maintained. 196 Brookings Papers on Economic Activity, 1:1975 farmersif societyfelt this temporaryevent justified income redistribution. Anotheralternative would be bond finance,which would redistributein- comefrom future generations to presentones. Stillanother solution would be the establishmentof a "pricestabilization fund" that would pay nonfarm subsidiesin yearsof low farmproduction, financed by a nonfarmsales tax in yearsof bumpercrops and low farmprices.20 Symmetric supply fluctua- tionswould allow this remedy, but asymmetricevents like thoseengendered by the OPECoil cartelwould not.

The InflationaryConsequences of an AccommodatingPolicy

A nonfarmsubsidy appears to be almostideal in principle,eliminating involuntarynonfarm unemployment and avertingmost (but not all) of the increasein consumerprices.21 But its rapidimplementation may pose ad- ministrativeor politicalissues, and its financingraises difficult problems. An alternativeis an accommodatingpolicy for nominal income, which couldeliminate involuntary nonfarm unemployment at the cost of a higher pricelevel. A rough numericalestimate of these inflationaryconsequences is pre- sentedin figure1. A simplemodel has been simulatedto illustratethe con- sequencesof a hypothetical10 percentdecline in farmoutput lasting twelve quarters.The model consists of the farm market-clearingequation (2) combinedwith the nonfarmprice adjustment equation (16). The simulated responseof the rate of wage increaseto excess labor supplyis relatively slight,as U.S. evidencesuggests, but excesscommodity demand is assumed to have a substantialimpact on nonfarmprices relative to wages. Other parametersare identicalto those used in the multiplierexamples in the previoussection (details are spelledout in the appendix). The "basic"simulation, A, illustratedby the solid line in figure1, shows that a crop failureaccompanied by a policy of constantnominal income createsa recession,the severityof whichgradually eases as nonfarmprices adjustdownward in responseto excesssupply. The "optimisticaccommo- dation"simulation, B, assumesthat policymakers raise nominal income to maintainthe originallevel of nonfarmoutput and that the expectedlevel of

20. Inventoriesof farm products are ruled out by the assumption that the supply is sufficientlysevere to exhaust them. 21. See note 6 above, 0

0

* I~~~~~~~~~~~~~~~~ 0

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

0

. 1~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~r

~0 .00 0d * .4

0 4~~~~~~~~~~~~~~~~~~~~~~~~ o 3 1 4)4)~~~~~~~~~~~~~~~~~~~~~ * U~~~~~~~~~~~~~~~~~

~~~T4 I4 .0~~~~~

C) 00~o ' * 3 0

3 4~~~~~~~~~~~~~~~~~~~~~~~0 iz~~~~~~~~~~~~~~ 198 Brookings Papers on Economic Activity, 1:1975 farmprices is adjustedupward to the higheractual farm price, but that in- dividualsmaintain their expectations about nonfarm prices. The "pessimis- tic accommodation"simulation, C, assumesthat the expectedlevel of non- farmprices is adjustedadaptively to all changesin actualnonfarm prices, whetherassociated with temporaryor permanentevents. SimulationB il- lustratesthat an accommodatingpolicy buys full employmentat the cost of a temporaryincrease in the price level and in the rate of inflation;but it permitsmore deflationafter the crops returnto normal,leaving the con- sumerprice index the sameten yearsafter the initialshock. In simulationC, the inflationrate increasesby more than it does in the optimisticcase but neverthelesstemporarily; the consumerprice index is permanentlyin- creasedby almost4 percentas the resultof the gradualupward adjustment of the expectedrate of nonfarminflation during the period of the crop failure.

The Consequencesof WageIndexing

The computersimulation program can also be used to evaluatethe con- sequencesof an externalshock for an economyin whichwages are indexed. Wageindexing can be representedby a new wageequation to replace(13): (17) w=PC+jz. The rate of changeof the realwage rate (w - Pc) now dependsonly on ex- cess labordemand. By increasingthe stabilityof the real wage,wage index- ing makeswages and prices more responsive and real output less responsive to "nominal"shocks-that is, variationsin monetarygrowth. At the same time, however,the built-inrigidity of the real wageimpedes the economy's adjustmentto "real"shocks, which require a changein the real wage.22In (17) the reductionin the realwage needed to clearthe marketfor farmout- put calls for a deeperrecession with indexingthan without. Figure 2 contraststhe path of the consumerprice index and nonfarm outputin the basicnonindexed simulation A fromfigure 1 withtwo index- ing simulations.The behavior of the wagerate under indexing is represented by (17), adjustedto makethe currentrate of wagechange equal to the rate

22. The sentencesummarizes the majorconclusion of Jo Anna Gray, "Wage Indexa- tion: A MacroeconomicApproach," working paper (University of Chicago,April 1975; processed). 0 0

I'~~~~~~~~Z

I 1 ~~~~~~~~~~~~0

0

I I~~~~~~~~~~~~~~~~~I

0 o- ~~~~~~~~~~~~~~~~~~~~~~~~0

'0 w C/l Enco co oz1~; 200 BrookingsPapers on EconomicActivity, 1:1975 of change of the CPI in the previous period, adjustedfor that period's excesslabor demand. Curve D in figure2 tracesthe effectsof wageindexing when policymakershold nominalincome constant.The increasein farm prices duringthe crop failure feeds through much more completelyto wagesand nonfarmprices when wages are indexed,using up more of the fixednominal income and requiring a muchmore substantial decline in real output(reaching a maximumof nearly15 percent)than in the basicsimula- tion. Eventually,the deep recessionbrings down the price level, freeing more of nominalincome to supportreal output. When the crop failure ends,a verylarge excess demand for labor develops.In short,wage index- ing makesboth prices and unemploymentsubstantially less stable when nominalincome is held constant in the presenceof an externalsupply shock. As before,policymakers can raisenominal income to accommodateboth higherfarm prices and the originallevel of nonfarmoutput. But this policy has very seriousinflationary consequences under wage indexing,since it preventsthe emergenceof the excesslabor supplyrequired in the lagged versionof (17) to lowerthe realwage. As illustratedby curveE in figure2, the resultis a geometricincrease in the consumerprice index (a steady6.0 percentquarterly rate of inflation)until farm output returnsto its initial level in the thirteenthquarter, by whichtime the consumerprice index has doubled.Only a bumpercrop or a policy-inducedrecession can reversethe processand bringthe consumerprice index back down.

Summaryand Conclusions

Thispaper analyzes the responseof a simpletwo-sector economy to a de- cline in outputin an externalsector wherethe price is assumedto clear markets.Its majorconclusions are, first, that no problemsarise if wages and pricesin the internalsector instantly fall to clearthe market.Any re- ductionin employmentis purelyvoluntary. The optimalpolicy is a reduc- tion in nominalincome to hold the aggregateprice index constant and avoida temporaryincrease in its level. Second,when nonfarmwage and price levels are absolutelyrigid, and whennominal income is heldfixed, the supplyreduction in the externalsec- tor has a multipliereffect, causing a recessionand involuntaryunemploy- mentin the internalsector. The "socialcost" of the supplyreduction then RobertJ. Gordon 201 exceedsthe valueof lost externaloutput by the valueof the nonfarmoutput that is squeezedout. The optimalpolicy is an increase in nominalincome designedto accommodateboth the higher externalprice level and the originallevel of internaloutput; a temporaryincrease in the aggregate pricelevel cannotbe avoidedsince the internalprice level is fixed. Third,when wages and prices are partially responsive to excesslabor and commoditydemand but wagesdo not responddirectly to higherexternal prices,the initialeffect of the externalsupply reduction is the sameas in the rigid-pricecase. If nominalincome is held constant,a recessionwill con- tinueuntil the nonfarmprice has fallen to its market-clearinglevel. As in the rigid-pricecase, the recessioncan be avoidedby an accommodating policyfor nominalincome, which temporarily (but not permanently)raises the price level comparedwith the case when nominal income is held constant. Fourth,when wages and pricesare partiallyresponsive to excesslabor and commoditydemand, and in additionexternal prices feed throughdi- rectlyto wages,the inflationand recessioncaused by the externalsupply shockare both aggravated.A policy aimedat an accommodatingnominal incomeraises the pricelevel temporarilybut not permanentlyhigher than wouldone of nonaccommodationif expectationsof the nonfarmprice level do not extrapolatethe inflationthat occursduring the periodof the supply reduction.On the otherhand, adaptive nonfarm expectations would cause the pricelevel (but not the rate of inflation)to remainpermanently higher when an accommodatingpolicy is pursued.Moreover, in the event of a permanentreduction in supply(such as one enforcedby an unbreakableoil cartel),a policy of accommodatingnominal income would raise perma- nentlythe rate of inflationof the consumerprice index. For the case of the temporarycrop failure,a superiorpolicy in principle wouldbe a subsidyto nonfarmproducts that wouldavert both the recession entailedby nonaccommodation,and the higherprice level requiredby ac- commodation.The major obstacles to a subsidyare the administrativeand politicaldifficulties of its promptimplementation, and the costs of financ- ing it. Finally, the analysis of this paper raises serious questions about the meritsof the full indexationof wagecontracts, which would shorten the lag in the adjustmentof wagesto changesin externalprices and wouldthus in- hibit the declinein the real wage requiredby an externalsupply shock. If policymakersattempt to stabilizenominal income in a wage-indexedecon- 202 Brookings Papers on Economic Activity, 1:1975 omy, any externalshock will destabilizeboth pricesand outputmore than it would in an unindexedeconomy. Any attempt to accommodatethe higherprices by raisingnominal income under indexing will imposeon the economya substantiallyhigher inflation rate for the durationof the exter- nal supplyreduction. These disadvantages of wage indexationseem to me persuasive,but do not weakenmy previouslystated supportfor fully in- dexedgovernment bonds, tax exemptions,and tax brackets.

APPENDIX

Model Usedfor Simulations

IN THIS DESCRIPTIONof the model, superscripts refer to sectors, and sub- scriptsto timeperiods. (The basicparameter assumptions and theirjustifi- cationsare listed at the end of this appendix.)Farm output, Q', dependson its base-periodlevel, adjustedby a percentagecrop failure,v: (A-1) Q' = (1 - v)QF. From (2) in the text,

(A-2) pF = pN [A(QN)ao/ QF Ia,. From (13),

(A-3) w = jZPN* + 1 + g(kpF + t) wherethe expectedfarm prices and the tax rateare set at theiractual values, andthe symbolsare as definedin the textequations. With the rateof growth of outputas a proxy for the rate of growthof excesscommodity demand, from(14):

(A-4) Pt= wt + cqt_l. The consumerprice index is (A-5) pc= kPF + (1-k)Pv. (I have omittedequations that convertlevels to rates of growth,and vice versa.).,When expectationsare adaptive,the expectedlevel of nonfarm pricesis extrapolatedfrom the actuallevel of the previousperiod by an ex- pectedinflation rate that is a distributedlag of past inflationrates, with RobertJ. Gordon 203 weightsfrom my "Inflationin Recessionand Recovery,"table A-1, trun- catedto the firstten values and constrainedto add to unity: 10 /N (A-6) p = pN -+ Pt ( ij++EUiPN U~tt, wherethe ui are the weights.The expectedfarm price level, Pr, is always equalto its actualvalue, pt. In the simulationsthat hold nominalincome, Yt, constant,the level of nonfarmoutput is a residual,and labordemand fluctuates by a fraction,n, of the changein output,while labor supplyis assumedconstant:

(A-7) Qt= (Y t therefore, the excess demand for labor, Zt, is

(A-8) zt = Zt-1(l + nqN). In the simulationsthat vary nominalincome to hold real nonfarmoutput constant,(A-7) and (A-8) are replacedby (A-9) y = PFQF + PrQN and (A-10) zt = 0. In the simulationsof wage indexing,(A-3) is replacedby (A-1) wt = PCU1+jZt-.

The basicparameter assumptions and theirjustifications are as follows: a,, the priceelasticity of demandfor farmproducts, is 0.2, a value sug- gestedto the authorby Dale E. Hathaway.Hathaway also suggested0.2 as a valuefor the incomeelasticity, ao. v, the percentagereduction in farm output,is 0.10, an arbitrarychoice. j, the percentagechange in wage growthfor a change of 1 percentage pointin the excessdemand for labor,is set equalto 0.13, to correspondto the more pessimistic assumption in Tobin's recent BPEA paper.23 (Note, however,that Tobin allows for no reactionin the priceequation.) g, the responseof wagechange to changesin farmprices, is 0.2, roughly consistentwith my evidencein "Inflationin Recession and Recovery," table 1.

23. , " in 1974 and Beyond," BPEA (1:1974), pp. 229-30. 204 Brookings Papers on Economic Activity, 1:1975 c, the percentageresponse of the inflationrate to the rate of growthof output,holding wage growthconstant, is 0.15. This impliesthat a 10 per- cent reductionof outputrelative to trendreduces the price-wageratio by 1.5 percent.This is somewhatlarger than the 1.0 percentestimate implied by the coefficienton the ratio of unfilledorders to capacityin "Inflationin Recessionand Recovery,"because of my finding that the transactions pricesof producers'durable areflexible relative to the list pricesused in that earlier paper. See my Measurement of Durable Goods Prices (Na- tionalBureau of EconomicResearch, 1975), forthcoming, chapter 5. k, the shareof the farm sector in initial spending,is 0.10, an arbitrary choice. n, the shareof a changein outputtaking the form of a changein labor input,is set at 0.5, allowinghalf of the outputfluctuation to be reflectedin .

Discussion

WILLIAMPOOLE questionedthe relevanceof Gordon's shock model to the actualsituation of 1973-74.Poole pointedout that, historically,going back to the 19thcentury, serious normally have been led by pri- maryproducts. These commoditiesare tradedin highly competitivemar- kets and theirsupply elasticities are relativelylow in the short run. More- over,inventory speculation plays a majorrole in the determinationof their prices.For these reasons,particularly large increasesin the prices of pri- maryproducts are a classicphenomenon of inflation.In Poole'sjudgment, the historicalrecord at least raises questionsthat Gordon had not an- sweredbefore assuming that agriculturalshortages and the like playedan unusualrole in the 1973-74experience. In response,Gordon emphasized that the shortfallin crops and the actionsof the oil cartelwere observable phenomenaof recentyears that properlycould be regardedas shocks.The fact that these shocks impingedon a world with high levels of aggregate demandmay dilute, but does not eliminate,the relevanceof the shock model. Anothermajor reservation about the Gordon model was expressedby MartinFeldstein. He noted that wagerigidities were the basicreason that Robert J. Gordon 205 shocksproduce recession in Gordon'smodel, and arguedthat theserigidi- ties reflectthe expectationsof laborand managementwith respectto gov- ernmentpolicy. Basically, wages are rigidbecause private decisionmakers bet that the governmentwill not let unemploymentget so high as to pull down wages.Gordon's proposed policy of accommodatingshocks by al- lowingnominal income to growfaster would strengthen these expectations, intensifyingthe downwardrigidities and making it evenharder to dealwith new shocks. Severaldiscussants raised questions about how uncertaintyof forecast- ing the exogenoussectors affected Gordon's findings. John Kareken called attentionto the analyticalproblem of incorporatingdisturbances from stochasticvariables like agriculturalsupplies into a comparative-statics model which essentiallyhas no room for .William Brainard agreedthat stochasticmodels were needed to evaluatepolicy strategiesto anticipateshocks. But he viewedGordon's paper as a constructiveattempt to assessthe abilityof policy to respondto shocks after they are known. He pointed out that, once a crop is harvested,some policy actions can influencethe pricelevel duringthe next year, even thougha fully optimal responsewould depend upon the expectedcrop. Michael Wachtersug- gestedthat the recentrecord of agriculturecould be viewedas a stringof threeshortage shocks in a row. It is temptingto accommodateeach one of themindividually as specificand transitoryshortages to ease theirimpact; but a successionof such decisionsaccommodates inflation. Somecontrasting opinions were expressed about the natureof wagebe- haviorinsofar as it influencesthe value of the g coefficientin Gordon's model.Robert Hall inferred from recent U.S. wagebehavior that food and fuel priceshad not fed throughinto wages.The real price of land and raw materialshas risen in relationto real wages,just as shiftsdictated. James Tobin agreed with Hall that one did not observe,and shouldnot haveexpected, a majorescalation of money-wageincreases as a resultof risesin the pricesreceived by entrepreneurs(like farmersand oil producers)who werenot hiringlabor. In fact, the employersof American labordid not havethe wherewithalto pay majorincreases in wages.But he statedthat neither the empiricalnor analyticalevidence demonstrated that the coefficientwas zero ratherthan some smallfraction-like the 0.2 value that Gordonused illustrativelyin his paper. FrancoModigliani cautioned that the behaviorof real wages gave no evidenceof the magnitudeof the pass-throughinto wagesof cost-of-living 206 Brookings Papers on Economic Activity, 1:1975 itemsnot producedby labor.Even if thatpass-through was sizable,as long as all pricesadjusted rapidly and fully to increasesin wages, real wages wouldcontinue to be squeezedby the pricerises in the exogenoussectors. He pointedto his own empiricalresults, which showed some effectof the priceof wholesalefoodstuffs on nonfoodconsumer prices, as evidencethat some influence(although not a terriblystrong one) prevailsin reality. Feldsteinthought that the recentmoderation of wageincreases might indi- catethat labor and management had madethe samemistake as economists in underpredictinginflation. If they expectedthe food and fuel inflationto be temporary,their failure to pass those priceincreases into wageswould be understandable. Gordon'snegative verdict on wage indexationinterested a numberof participants.In Brainard'sview, wage indexationhad undesirableeffects in Gordon'smodel partiallybecause wages are linked to a cost-of-living indexthat includes imports, food, andthe like. If the priceindex comprised only itemsmade by domesticworkers, the resultscould be quite different and morefavorable. Granting that such a form of indexationwould offer less insuranceon real wagesto workers,Brainard thought that laborcould recognizethat real wagescannot be maintainedin the face of majorprice increasesin the exogenoussectors. Modigliani strongly supported Brain- ard's suggestion,contending that, for many purposesof indexation,the priceindex used for escalationshould be restrictedto value addedin the privatenonfarm sector. Hall felt that this was likely to approximatean indexingof wagesby other wages,given the preponderanceof wage costs in privatenonfarm value added. MarinaWhitman noted that Gordonattributed a dual functionto -influencingthe cost as well as the demandside of the economy. Thisdual influence raised the possibilitythat an excisetax cut, for example, could bringdown unemploymentand yet have enough cost-reducingin- fluenceto curbinflation. She wonderedwhether it could ever be clear in the real world that the cost-reducingelement of any tax cut would out- weighits demand-raisingeffects. Joseph Pechman elaborated on this point, stressingthat the weaponsin the arsenalof fiscal policy that can most readilyexert cost-reducing effects are novel and thus far have been politi- callyunsalable.