llFYI1~MAY/JUNE 1995

Allan H. Meitzer is a professor of political economy and public policy at Carnegie Mellon University. An earlier version of this article was pre-sented at the 1994 Konstanz Seminar in Insel Reichenau, Germany, and some material is included from earlier work with Karl Brunner andAlex (ukierman. The author thanks Dale Henderson and Benneff McCallum for comments on earlier drafts.

Information,Sticky Prices andMacroeconomicFoundations

Allan H. Meltzer

5izor decades. macroeconomists have us-~fntened to criticism from their professional

I colleagues about the absence of micro-foundations from most of what they say anddo. A typical comment is: “I don’t understandmuch about macroeconomics; how is it relat-ed to economics?” A possibly less-frivolouscomment is that nothing can be said aboutmacroeconomic policy until economistsdevelop a macroeconomic theory from amicroeconomic foundation. Anything less isbranded “ad hoc” and dismissed.

Much of the work in macroeconomics ofthe past two decades has responded to thesecriticisms by attempting to build macroeco-nomics on an explicit micro-foundation.That is a worthwhile goal hut it opens thequestion: Which foundation should that be?The current generation of academicresearchers are as divided about the appro-priate analytic model as their predecessors.Analytic paradigms now include multipleequilibrium, real business cycle, neo-Keynesian, monetary-rational expectations,and eclectic models. Most have a micro-foundation, hut it is not the same foundation.

I do not question the presumption that amicro-foundation is useful. At issue is whatthe foundation should be. I question therelevance for monetary and macroeconomicsof micro-foundations which feature a repre-sentative agent who trades on a complete setof Arrow-Debreu markets. Despite somelimited successes, it is time to questionwhether this now widely accepted approachis likely to be fruitful and to suggest why I

believe its success will be limited. Successis relative, of course. Real business cycletheory has developed an explicit analysis ofthe transmission of productivity and terms oftrade shocks. These shocks, though widelyrecognized earlier, had not been made thesubject of an explicit model. Neo-Keynesiansand others have produced some suggestionsabout pricing. Overlapping-generationsmodels of money, intertemporal substitutiontheories of unemployment, and productivityshock theories of the business cycle have notproved fruitful. These models have not pro-duced either an accepted foundation formacro theory or averified theory of aggregateoutput, prices and interest rates. Perhapsthey will in the future, but the results to dateare not promising.

There are two main ways to tie microand macro theories. The first is aggregation.Aggregation is an important and muchneglected issue, but it is not my main theme.Most current or recent research dispenseswith the aggregation problem by assuming arepresentative individual. Here the old andnew macroeconomics are equally deficientand open to charges of “arm waving” and “adhocery” The representative individual is auseful working assumption for some purpos-es, but the representative individual discardsan important difference between marketsand individuals.

The second problem is the specific

micro-foundation used for macro theoryMost of my discussion is directed there. I donot attempt to survey a large literature on

micro theory uncertainty and industrial orga-nization. The implications of some of thisliterature for price stickiness is ably summa-

rized in Gordon (i990). This paper is apersonal statement, reflecting joint workas noted earlier.

The problem with existing micro-foun-dations that I emphasize is the neglect ofcosts of information. In Walrasian micro-

foundations, all trades take place at marketclearing prices, and there is no cost ofacquiring information. An auctioneer calls

FEDERAL RESERVE SANE OF St.. Louts

101

llE~IF~MAY/JUN. 1995

out the prices at zero cost and does not closetransactions until all transactors are at anequilibrium. There is a numeraire, but thereis no way for money to disturb the real valueof production or purchases. Non-neutralitycannot arise. Attempts to graft a monetarydisturbance onto these micro-foundationsseem misdirected.

The hypothesis that all observed prices

are market clearing prices does not implythat all individuals, behaving rationally,know those prices and fully adjust to them.Information is costly to acquire; time andresources must be used to collect, processand interpret—the latter especially—newdata. Even in markets dominated by pricetakers, there are differences between theinformation processed and known to arbi-tragers and specialists and what is known to

non-specialists. The representative individ-ual paradigm ignores this distinction.

One way to proceed is by aggregatingheterogenous individuals who face differentcosts of acquiring infonnation, and much newwork has taken this path. As Gordon (1990)emphasizes, at any time there are many differ-ent layers of pricing and output includedin aggregate prices and output—suppliers,suppliers of suppliers, foreign producers, andso on. Instead of trying to aggregate overthese many, diverse and changing levels ofdecisions, it may prove more useful to treatthem as part of a stochastic process.

One of the problems with the hypothesisthat “sticky” prices reflect decisions at differ-ent levels of aggregation comes out clearlyin Gordon’s paper. He argues that sluggishprice adjustment reflects marginal cost pric-ing. “[Agentsl care about the relation oftheir own price to their own costs, not toaggregate nominal demand. Unless a singleagent believes that the actions of all otheragents will make its marginal costs mimicthe behavior of nominal demand with mini-mal lags, the aggregate price level cannotmimic nominal demand, and Keynesianoutput fluctuations result.”

This argument has a perverse implication.Farmers or farms that operate in commoditymarkets with many competitors should heslowest to adjust to aggregate demandshocks. Yet commodity prices typically

adjust quickly Large industrial firms—General Motors, Mitsubishi—know that theyare a relatively large part of the economy, sothey should adjust to aggregate demandmore promptly than commodity producers.Typically, they do not.

Missing from Gordon’s analysis is thedifference in information and in costs ofacquiring information. Commodities aretraded on open markets, so prices promptlyreflect changes in the factors affectingdemand and supply Prices of autos, steel,heavy industrial products and consumerdurables are not set in organized markets.Information on which to base prices ismore uncertain.

As Keynes recognized, a principal miss-ing element is uncertainty about the future.Uncertainty and its twin, costs of informa-tion, make it rational for some firms toadjust prices slowly Prices may be “sticky”as economists have observed for about aslong as there has been a discipline.

Some economists will scoff that stickyprices are irrational or equivalent to leavingfive, 50 or 500 dollar bills in the street.This is an error arising from dependence onWalrasian foundations and neglect of infor-mation costs. Yet, there is nothing novelabout invoking costs of acquiring informationto explain sluggish price adjusunent orsticky prices. The positive slope of Lucas’(1972) aggregate supply curve arises fromconfusion between relative and absoluteprice changes; prices in that model do notimmediately adjust to new information. Thecost of learning whether a change in demandis an aggregate or relative change is infinitefor one period and zero thereafter. In theneo-Keynesian models discussed in Ball andMankiw (1994), costs of price adjustment—so-called menu costs—and “real rigidities”are assumed to be present. The authorsaccept that one of the principal costs of priceadjustment is the cost of acquiring informa-tion relevant for a decision about how muchto change price.

Given this widespread acceptance ofinformation or transaction costs, whatremains to be done? Lucas’ (1972) modelimplies more rapid adjustment of prices andoutput to new information than we observe

FSflSAL RSSSSVI BANK OF ST. LOUIS

102

DFYIF~MAY/SUNS 1995

in practice. Could it be rational to adjustmore slowly than in one week or month?Is the compelhng fact that data on the pricelevel are released monthly? Or is this infor-mation subject to error so that it takes longerto disseminate, interpret and act on thesedata? Neo-Keynesian models build on themodel of an imperfectly competitive firm,with strong implications for profits andexcess capacity that do not find empiricalsupport at the micro level. If monopolisticelements and menu costs are the source ofsticky prices, there are testable implicationsfor the profits of different types of firms; thereis a significant problem of reconciling contin-uous excess capacity with rational behaviorFurther, as Gordon (1990) has noted, costsof adjusting output are neglected in thesemodels. Yet these costs may be larger formany firms than costs of price adjustment.

My principal criticism of both neo-classical and neo-Keynesian approaches isthat they seek to explain why firms delayusing available information. By putting theissue in that framework, they neglect theuncertainty that surrounds much of theaggregate and disaggregated data. This paperuses several strands of earlier work to explainrational price setting and gradual adjustmentas a response to uncertainty about what cur-rent information implies. Information iscostly to acquire and to interpret as inBrunner and Meltzer (1971) or Alchian(1977). As in Bomhoff (1983), a principaldifficulty in interpreting information isuncertainty about how long changes willpersist. This is the central idea developedin Brunner, Cukierman and Meltzer (1983),but we took the idea from Muth’s (1961)seminal paper on rational expectations. InMeltzer (1982), 1 used these ideas to discussprice setting.

There are three separable aspects of pric-ing to consider. First, (some) prices are setby firms. Second, firms choose nominalvalues; they do not index or set a relativeprice. Third, prices that are set change lessfrequently than prices in auction markets.

Each aspect is important for macroeco-nomics. If some prices were not fixed innominal value for at least one period, relativeprices would be invariant to a monetary

change. Delays in recognizing permanentchanges in money or its growth rate wouldaffect only real balances. At best, we wouldbe forced to fall back on the real balanceeffect in consumption to explain short-termreal effects of changes in money A rationalreason for setting some, but not all, nominalprices permits a more direct effect throughinventory adjustment. Firms that holdinventories can both anticipate future pricechanges and buffer current transitory pricechanges by varying inventories. Becausesome prices are determined in auction mar-kets, price setting introduces different speedsof adjustment and relative price changes.My extensive work with Brunner emphasizedthe relative prices of assets and output.This introduces the difference between thereplacement cost, or the cost of currentproduction, and the market price of existingassets. Asset prices adjust more rapidly thanoutput prices particularly for assets thattrade in organized markets. Hence, informa-tion costs (and transaction costs) are implicitin that framework as an explanation ofchanges in relative prices. My emphasishere is on information as in Brunner,Cukierman and Meltzer (1983). Relativeprice changes in response to nominal shocksare part of that story, but they remain inthe background.

kJ ~ At!’ •PRICES?

Postwar recessions typically last aboutnine months on average. Costs of informa-tion or other explanations of sluggish adjust-ment must be able to explain this timing,and estimates of price stickiness should beconsistent with data on cyclical fluctuations.I digress therefore to consider a recentattempt to measure stickiness. Blinder(1991) reports preliminary findings from asurvey of pricing decisions by managers of arandom sample of corporations with annualsales of $10 million located in the northeast-ern United States. Blinder’s survey producedtwo very different sets of estimates.

First, Blinder reports the answer to aquestion about how frequently a finnchanges its price. He finds that the medianfirm changes price about once a year. Nearly

FSBUAL RESIRVI SANK Of ST. LOUIS

103

llF~IF~MAY/SUNS 7995

Beooiee nod Money l19941estimate the effects on pricedispersion oi factors such osodvertisieg, tronsportcosts, nodtoncentrotion using cross-sectionInto for specrlrc commodities.They find effects of sonic of theseetemerts on price dispersior.

40 percent of the firms were at the median.More than one-sixth of the firms changedprices less frequently than once a year, so 55percent of the prices change no more thanonce a year.

Second, Blinder also reports the respon-dents’ estimate of the mean lag of actualprices behind increases and decreases indemand and cost. There is considerableuniformity in these responses. The mean lagfor increases or decreases in demand andcost is three to lour months.

The second set of data suggest consider-able uniformity of response to the four shocks.There is little evidence of the asymmetricadjustment associated with Keynesian down-ward price inflexibility The data also sug-gest an inconsistency that Blinder does notmention. How do we reconcile a mean delayof three to four months in response to changesin demand and cost with the report that 76percent of the sample changes price no morethan twice a year and, as noted, 55 percentchanges price no more than once a year?Are demand and cost changes infrequent?It is difficult to reconcile the assumption thatcosts change infrequently with evidenceshowing that commodity prices and otheropen market prices change daily and typicallyfall in recessions and rise in expansions.

Bhnder did not ask whether firms adjusttheir prices fully in response to changes indemand and cost. They may adjust partially,as implied by rational behavior under uncer-tainty about the persistence or permanenceof announced changes, or they may antici-pate future changes and adjust prices morethan current changes in cost or demand.One way to reconcile the different responsesis to assume that respondents answer thetwo questions in different ways. Supposethey treat price as a scalar when asked aboutthe frequency of price changes but include inprice adjustment more than just price setting.On this interpretation, quoted prices are onecomponent of a vector of terms and conditionsrelevant to sellers and buyers. The theoreticalterm price used in economics typically sub-suanes delivery time, discounts, advertisingallowances, volume rebates, payment termsand other conditions used to adjust the buyer’scost and the seller’s net receipts. Blinder’s

early results confirm that delivery lagsand service are moderately important for65 percent of the firms in his sample; hereports delivery lags and service are the mostcommonly cited reason for price stickiness.’Delivery lags and service are one way ofadjusting the theoretical term “price” whileleaving the quoted price unchanged.

Assume that firms initially respond tochanges in cost and demand by adjustingdeliveries, advertising allowances, discounts,and so on while leaving quoted pricesunchanged for several months or longer. Ifmanagers are uncertain about the duration ofchanges in demand or cost, they can changeother components of the price vector to testthe market’s response. By changing dehveiytenns, or offering or removing discounts,firms can change their revenues or thebuyers cost without changing the quotedprice. This pricing model can be used torationalize the familiar Keynesian supplycurve—a reverse L—when quoted prices aredistinguished from other terms in the pricevector. Equally the model can explain thedifference in response to the questions inBlinder’s survey

Figure 1 shows the initial response to anincrease in demand. On the left, the quotedprice (p) does not respond to a perceivedchange in demand when output is belowcapacity; the firm (or industry) increasesoutput (q) with little or no change in quotedprice. On the right, the price vector (p),includes other dimensions ol the firm’s price;supply is drawn as a positively sloped linearfunction. An increase in perceived demandfrom d© to d, induces the firm to reduceadvertising allowances, retnove discounts,or change some other component of theprice vector, so (p) increases.

If the firm’s initial response reflectsuncertainty about the persistence ofincreased demand, the response changes asinformation accrues. An increase in thequoted price may substitute for or supple-ment other components of the price vector.As perception of the magnitude of the per-manent increase becomes clearer, the firmmay recognize that the new demand curve isat, or to the right of, d,. Or the firm mayraise p, with p unchanged.

FIDERAL RISSSVR SANK OF ST. LOUIS

104

II [~I1WMAY/SUNS 1995

We need not explore the many possibili-ties. The main point is that uncertainty aboutthe degree of persistence and the use of aprice vector permit this hypothesis to accountfor very different responses within a rationalexpectations framework. In particular, thefirm’s revenues may respond instantly toincreases in demand, but quoted prices mayadjust with a long lag, as Blinder found.The reasoning is symmetric. Perceived reduc-tions in demand may induce firms to offerdiscounts and allowances with p unchanged.As information accnoes and perceptionschange, the actual price, p, is reduced.

The Keynesian supply function does notwork for changes in cost if the demand curveis not kinked. Changes in cost shift thesupply curve, so prices change instantlywhether the supply has a reverse L-shape oris monotonically increasing. Since Blinder’ssurvey finds that price responds about aspromptly to changes in cost as to changes indemand, this evidence rejects the Keynesiansupply curve.1

The distinction between price as a vectorand a scalar does not reconcile the differencesin timing reported in the survey Although Ibelieve that prices, terms and conditionschange at different rates when there is uncer-tainty about the persistence of the marketconditions inducing firms to change prices,those differences are neglected hereafter.Price stickiness will mean that it takes aboutthree or four months on average for prices torespond to demand and cost changes. Thethree-to-foter month delay that Blinder reportsincludes at least one quarterly reporting periodat which managements announce earningsand sales and, most importantly observereported earnings, inventories and sales ofcompeting firms.3 Knowledge of competitors’results helps the firm to supplement tradegossip and other informal data sources todecide whether a persistent change in marketdemand (or industry costs) has occurred.

INFORMATION AND PRICESMITING

Anyone familiar with hterature on thebehavior of firms knows that there are manyrational reasons for firms to set prices. This

p p

section discusses one reason that is rathergeneral and can be incorporated readilyinto macroeconomics. Price setting isconsidered as a response to uncertainty orcosts of acquiring information about themarket clearing price by at least one (large)group of market participants. At times,prices convey information known to theseller hut not to the buyer. At times,neither the buyer nor the seller is certainabout the market clearing price. Amongthe possible reasons on the cost and demandsides, this section emphasizes rationalreasons why buyers and sellers do notinstantly know the permanence of changesin demand. They develop contracts andmarket arrangements to deal withthis uncertainty.

This approach differs from the literatureon so-called menu costs, The menu cost lit-erature emphasizes costs of changing prices.5

These costs are recognized as relatively small(Ball and Mankiw, 1994). Moreover, the menucost model does not explain why sellers facedifferent costs and adjust at different speeds.I do not challenge the existence of menu costs,but the emphasis here is on uncertainty andinformation costs—the cost of learning aboutcurrent and prospective market conditions.Information costs differ by firm and industry.They depend on the way in which marketsare organized and the types of contractsthat emerge to reduce costs of uncertaintyinformation and moral hazard.

5”

FEDERAL RESSRVS BANK OF ST. LOUIS

S©

q q

‘loll and Mookiw 11994) use

the ability to generate a reversekhope (Keynesian) supply curveOs evidence for a model of menucosts and imperfect competition.

This shonld not soffest thot thetiming is constoot. As in Moth11961), the timing ofthe responsedepends or the relative vamioncesof permorent ond trnnsitory compo’nents. See below.

Mod of this section is token fromMeltzer (1912), ord especiallyBrorner mod Meltzer (1993).Additinnal exomples of mlionnlprice setting when information iscostly to ocqeimn con be found inthose references. A recent paperby Hentnn and tocos (forthcoming)onolyzes the effects of tmrnsactinncosts, such as bid’osk sproods,investor heterogeneity and persis’terce of shocks. They find thommrry pozzles ir osset pricing conbe reduced nr removed by usingmodels with tronsoction costs ondpersistence. Bribers ond Cnsimnnn(19911 model price odiosmrnentwhen firms most learn oboot thepersistence of chrnges in demand.

5See McCalfum 11919) for, citiceldiscossion of menu costs no a bosisfor rational price’level stickiness.

105

II 1~I1WMAY/JUNE 1995

Market Orr-zanizatkinFrank Knight was an early expositor of

the economics of information. Knight (1933),argued that decisions about how muchand when to produce require a pooling ofinformation about individual decisions topurchase. When aggregated, the individualdecisions constitute a demand curve.

Firms reduce uncertainty for consumersby pooling information. In principle, indi-viduals can contract in advance for the goodsand services that they want. Organizingretail firms that pool information is an efficientalternative to futures contracts if individualsare less certain about the magnitude and tim-ing of their purchases than firms are aboutmarket demand. Knight appeals to the lawof large numbers to explain finns’ advantage.He compares pooling by firms to insuranceand concludes that the two differ in animportant way; a firm’s pricing and outputdecisions are non-insurable because theyrequire more subjective judgment, and errorsare less likely to cancel across firms.

In Knight’s view, firms produce for inven-tory using pooled information about expecteddemand. This arrangement shifts uncertaintyfrom individuals to firms and, by pooling,reduces the social cost of bearing uncertainty

Knight’s argument is one of several thatlinks information costs and uncertainty toprice setting. A non-uniform distribution ofinformation is critical for these arguments.In an auction market, all market participantsmust have information about the qualitiesof the goods traded and their prices. In thestandard Walrasian model, this is accom-plished by: (I) assuming the presence ofan auctioneer who calls out the prices;(2) allowing recontracting; and (3) letting alltrades be made simultaneously These assump-tions are necessary for equilibrium. They leaveno role for monetary disturbances.

The necessary conditions are frequentlyviolated in practice. Some people have acomparative advantage in acquiring informa-tion. Some receive information about marketconditions as a by-product of other activityFor example, in securities markets, there arebrokers, dealers and market makers whoacquire specialized information in the courseof trading. Assembling all or a sufficiently

large numnber of market participants alsoimposes severe constraints. For many reasons,including the law of large numbers, peoplechoose different times for their marketactivities. The commitment to assemble atpre-specified thmes for all marketing purposesis costly, if not impossible. The use of anagent introduces costs of monitoringand supervision.

In practice, markets operate in manydifferent ways. One alternative commonlyfound in oriental countries is called a bazaar.Prices are not posted, and there is no auctioneermechanism. Buyers and sellers negotiate(haggle) until a transaction is completed or thenegotiation terminates. Considering a bazaarbrings out the importance of information.

The bazaar requires an investment oftime by transactors. Where the auctionmarket requires simultaneous arrival of allparticipants or their agents, the bazaardepends on a trickle of arrivals. It cannotcope with large groups arriving simultane-ously because each negotiation is separate.It is not surprising that the bazaar is foundin comparatively simple, low-incomeeconomies. In these circumstances, theallocation of time to the bargaining processmnay he partly a constemption good. Withrising opportunity costs of time, the disad-vantages of the bargaining process exceedthe benefits.

The working of a bazaar restricts itsapplication. There are severe limits on thenumber of transactions per period. Sellerscannot serve several customers simultane-ously Delegating bargaining to employeesposes both an incentive and a moral hazardproblem. Buyers incur costs to learn aboutreservation prices, since information isrevealed only by commitments to transactfirm offers to buy or sell. An accepted offerto purchase may be above the seller’s reserva-tion price; a refusal to sell maybe based onan incorrect inference that the buyer is willingto pay more. Learning the reservation priceand negotiating the market price often requiresa series of offers. There is no certainty thatthe reservation price is revealed. Subsequenttransactors on one side of the market do notknow the history of past transactions. Theymust invest their own time.

FEDERAL RESERVE SANK OF ST. LOUIS

106

H Pd F WMAY/JUNE 1995

° targeJapanese ins that offerlifetime omplaymeat portially indexby tying bonuses ta some measureal the irw’s proltobility. Withlifetime emplaymert, it mayberational to assowe ohot errors will

lappronimately) cancel oner time.tifetimeemployment coatroctsimpose costs or irms (owaars) ifoutput falls helow some eapecmedvoloe arrd impose costs on workersif ootpot is above the eapectednelue. The iotsor cor he comper-sated more readily by adiustegbameses, but nogetive bonuses orenot observed, An unexpectedlylong period of slow growth eliminatal same of these contracts.

relocate or search for a particular input at aninconvenient time.

Labor Mo.rtrersThe conditions leading to price setting

also apply to the labor market. The termsnegotiated and the time horizon built into anagreement depend on the assessments of theparties. There are no organized futures mar-kets for labor. Both parties use all availableinformation to form their uncertain assess-ments. An assessment of the market is a(subjective) probability distribution. Themore diffuse the distribution, the shorter thetime covered by the arrangement.

Realizations often deviate from theexpectations implied by the subjective prob-ability distributions. Both parties have toinfer from realizations whether the unexpectedchanges are transitory or permanent.

A transitory change does not changethe expected value, so the unanticipatedgains and losses do not change the informa-tion on which the agreement was based.Either party may believe that a costlessrevision of the bargain to adjust to a transitorychange would be beneficial, but attempts atrevision for each such change raise the costof transacting and eliminate the benefits of alonger-term agreement.

A more permanent change in conditionsposes a different problem. The initial assess-ment of at least one party must he revised inthe light of the new information. If the stakesare sufficiently high, the permanent changemay justify the cost of renegotiation.

Negotiations proceed more smoothlywhen both parties share the reassessment ofmarket conditions. Differences in assessmentprovide evidence on the extent of uncertaintyStrikes and lockouts increase with differencesin assessment, for example, at the start of aperiod of inflation or disinflation, if bothparthes could agree on the actual shocks thatoccurred in the past and on how long theywill persist, they could contract in advanceto compensate for unanticipated changesafter they occur. Permanent changes innominal values would not be allowed to affectreal wages. Permanent changes in produc-tivity would he paid to workers if positive

and by workers if negative. That we do notobserve contracts of this kind suggests thatassessments differ even after the event andthat reaching a common assessment of thepast is costly Of course, setting nominalwages is not costless either. Bargaining ornegotiating to correct for unforeseen eventscan be costly, privately and socially if thereare strikes or layoffs.°

Differences in information can explainprice setting, but they do not fully explainwhy firms and employees often set nominalwages or prices. There must be some addi-tional cost of setting relative prices or benefitfrom setting nominal wages (or prices). Oneexplanation is that the parties do not agreeon the interpretation of real wages and, par-ticularly at low rates of inflation, have diffi-culty agreeing on an appropriate index.

There are two different meanings of realwages. One meaning expresses real wages interms of the product of the firm at which theworker is employed. The second refers to thebasket of goods and services that the workercan purchase. The problems of setting con-tract terms differ in the two cases.

Contracts that set wages in relation toproductivity require a satisfactory solutionto the measurement of productivity Whereprecise measurement is difficult, as in serviceindustries or managerial tasks, real wagecontracts are difficult to write. Evenwhen productivity is measured reliably asin piece-work systems, the measure doesnot translate directly into a real wagerate. Valuation is required; often some(more or less) arbitrary system must be usedto impute the price of the final product tothe various inputs. Imputation and valuationbring tcvo additional problems. One is moralhazard; the employer has some incentive toadopt a cost accounting systeen that benefitshim. The other is the difference betweenrelative and absolute prices. The employer ismainly concerned with the relation of wagesto product prices. The employee is concernedalso, and perhaps most, about the relation ofwages to the price of consumables—the sec-ond meaning of real wages. A frequentcompromise in periods of inflation is partialindexation to the price of consumables.

Real wage setting with full indexation

FEDERAL RESERVE BANK OF ST. LOUIS

108

D1~I1WMAY/JUNE 199S

A seller who posts prices reduces thebuyers’ costs of acquiring information aboutprices. Buyers’ costs of comparing prices bydifferent sellers are reduced. The socialadvantage of price setting is greatest whereone party to the transaction has more infor-mation about market conditions. Someexamples illustrate this argument.

information about prices conveys morethan just purchase cost. Posted prices canalso reduce costs of acquiring informationabout quality For example, a restaurantowner must decide on the market he wantsto serve. This decision influences the kindand quality of food served, the services offered,and the prices charged. By posting prices, theowner informs the buyer about his choices.Although the buyer must sample to judgequality the correlation between quality andprice helps the buyer to decide whether tosample. A policy of frequently changingprices reduces information and places therestaurant at a competitive disadvantage.

The organization of the diamond marketprovides additional evidence on the role ofinformation in market organization. Thewholesale diamond market is an auctionmarket dominated by buyers and sellers whoare specialists. Traders rely on their ownskill in judging quahty knowledge of pricesand other attributes. The retail market isvery different. ‘[he sellers are mainly special-ists; the buyers typically have much lessinformation than the sellers. By postingprices, sellers exploit the correlation betweenquality and price to inform bteyers. Buyersfind it less costly to invest in informationabout the seller than to invest in informationabout the quality of diamonds, so sellers useresources to build reputation. If costs ofacquiring information about the quality ofdiamonds were to fall to a minimal value,these arrangements would change. Diamondsmight he sold in supermarkets or in retailaucthon markets.

Price setting is valued by transactors evenin some auction markets. In well-organizedauction markets, we find people willing topay for the right to purchase or sell at fixednominal prices. The contracts expressingthese rights, known as “put” and “call”options, give the owner the right to buy or

sell at a fixed price within a fixed time period.The prices of the puts and calls are determinedin auction markets. The price of these optionsis the cost that people pay for the right to tradein the future at prices fixed today Similarlyin commodity markets, hedgers pay tochange uncertain future prices into knownvalues. They pay a fee for the right to buy orsell at fixed nominal prices. In such markets,information about current and currentlyanticipated future prices is available. The feepermits transactors to avoid uncertain futureprice changes.

Costs of information and transactionsare not uniform across goods, so no singleform of market organization dominates allothers. The specific reasons transactors arewilling to pay for puts and calls differ fromthe reason for price setting in the diamondmarket, just as the diamond market differsfrom the restaurant. Each is related, however,to costs of information. The organizationof the diamond market reduces the costs ofbearing uncertainty about quality The marketfor puts and calls permits asset owners orspeculators to limit risk of wealth changes.And there are other market arrangementswhere price setting is useful. Cataloguesmust post prices. Cotitracts fix prices for ayear or more on housing rentals, magazinesubscriptions, and automobile leases.These many different examples suggestthat there are advantages in different typesof contracts.

The examples suggest a way to modelprice setting formally. The seller has infor-mation that is costly for the buyer to acquire.The seller internalizes the cost of acquiringthe information; it is part of his specializedknowledge and he revises his information inthe process of buying inputs. By posting aprice, he exploits the correlation between priceand quality In goods markets, the seller mayoffer a particular type of put—an option forthe buyer to return the merchandise if thequality is not as represented or, perhaps, if thebuyer finds the same merchandise at a lowerprice. The buyer pays for the good and forthe put, but the purchase cost is lower thanunder alternative forms of organization. Inservice markets, the buyer may purchaseincreased certainty that he will not have to

FEDERAL RESERVE BANK OF ST. LOUIS

107

H F~11WMAY/JUNE I99S

is rare. This suggests that at low levels ofinflation, buyers and sefiers prefer cx postadjustment through negotiation to relianceon an imprecise or imperfect index. As infla-tion increases, costs of non-indexation riserelative to costs of indexation. More partieschoose a mixed strategy of partial indexation.Experience in Israel and Brazil suggests thatat relatively high rates of inflation, indexationis nearly, but never fully complete. There isalways some lag in adjustment. Data forcountries with high inflation also suggestthat workers willingly forego indexation ifoffered relatively stable prices.

The choice of an index is a problem forboth parties. Some of the problem wouldbe removed if shocks could be identifiedunambiguously, if one-time price changes(transitory changes in the rate of inflation)could be separated from permanent changesin inflation, or if all shocks were of onekind—for example, permanent, nominalor real aggregative, or real allocathve. Theabsence of reliable information prevents settle-ment on an optimal indexation formula.

Nominal wage contracting is also notideal. Different types of contracts are used toadjust nominal wages for inflation. In periodsof low or moderate inflation, we observecontracts that differ in duration, in the extentof formal indexation, and in the use of clausespermitting reopening of the wage agreementduring the life of the contract. We observealso that the types of contracts change withthe rate of inflation and that employers canbe induced to compensate for (some) pastprice changes when (non-indexed) contractsare renewed. At high rates of inflation, firmsand other market participants monitor therate of inflation. Costs that were previouslymarginal costs of information became fixedor quasi-fixed costs of information. Nominalprices adjust more frequently

Retail store leases differ from wagecontracts. Leases are often indexed to thevolume of sales. Sales are more easilymonitored and therefore less subject to moralhazard. Valuation is based on receipts, someasurement is not as much of a problem asfor profits or productivity Both parties havean interest in maintaining the property

Bond contracts provide another example

of the problem of choosing an index. Privateparties do not issue price-level hnked bonds.Under the gold standard, however, firmsoffered to pay in gold. Buyers and sellerscould agree on this index of long-term value.Once this common measure, related to thevalue of money became less relevant, indexedbonds were rare. Inability to agree on anindex left no agreement. In Britain, Israel,Brazil and a few other countries, the govern-ment resolved this problem by issuing anindexed bond.

Comparison of the choices made inmarkets for labor, rental property and bondssuggests that agreement on an index is mostdifficult when prices can change because ofreal and nominal shocks, and changes can bepermanent or transitory. This should notsuggest that non-indexation is optimal.Contracting parties find many different solu-tions but, as experience in labor marketsshows, full indexation is rare.

A Stylized Model

In several papers, I have used a model ofpermanent and transitory changes based onBomhoff (1983) to study the frequency ofshocks and their interaction. A main conclu-sion of this work is that there is no reason toexpect constancy or even repetition in thefrequency distribution of shocks. The publiccannot use data from the past to anticipatethe future relative frequency of permanentand transitory shocks or real and nominalshocks. These frequencies change with publicpolicy decisions, policy rules at home andabroad, weather, inventions or innovations,changes in market structure, and other factorsaffecting an economy’s structural relations.

The model has three types of shocks:transitory shocks to the level; permanentshocks to the level (which are transitoryshocks to the growth rate); and permanentshocks to the growth rate. Let xr be the currentvalue of real output and Pr the current valueof the price level. Both prices and output canbe affected by each of the three shocks so that

Xr~~Xr+E~

Cr= ~~o~r+Yr

Xr ~-a +

FEDERAl. RESERVE BANK OF ST. LOUIS

109

HIYIIWMAY/JUNE 1995

and

Pa P~+ ~

P~ Pr~r+Pr~r

Pr = + Zr

There are stochastic elements in the growthpaths of output and inflation in addition totransitory and permanent changes in the levelsof output and prices. Much confusion in thediscussion of inflation has been caused bythe use of “inflation” to refer both to levelchanges such as oil shocks (distributed overtime) and persistent changes in the maintainedrate of change.

Suppose we now introduce a commontype of Phillips relation between p and x inthe neighborhood of Pc = 1.0.

xa- kr = a(pr - Pa.)

The way in which prices will change over timedepends on the permanence of the shocks.It takes time for agents to learn whetherthe shocks change a,, v, or Z, and, from thesimultaneity of x and p, shocks to ~, y~and

Pa’ The path by which prices adjust—or thedegree to which they are sticky—depends onthe nature of the shocks. It is entirely ratio-nal in this framework for prices to be stickyand for the speed of adjustment to differfrom one episode to another,

The model in Brunner, Cukierman andMeltzer (1983) conveys some of the ideasjust discussed. The main idea of the modelcan be written in a general way Considerthe following system of simultaneous equa-tions for a macroeconomic system. In theunderlying micro model, firms set price atthe start of the period and hold inventories.Shocks are revealed after production decisionsare made.

(1) .IIY,. P,~Ma, ~ai, ~r’ r,, -‘c, ma] 0,

where y = output, p = price level, h = stockof inventories, i = nominal interest rate,r = real interest rate, x = exogenous realshock, and m = nominal money stock. Underfull information about the structure of theshocks and in the absence of transaction costs,

the price level reflects current informationabout money Money is neutral. Incompleteinformation of some sort is a necessarycondition for significant monetary effects,but it is not sufficient.

Lucas’ (1972) hypothesis about incom-plete information restricts uncertainty tomisperception of current shocks; people donot know whether a shock to demand is spe-cific to their product or is a general increasein demand. It is now generally recognizedthat Lucas’ hypothesis produces a responseof real output to a nominal shock but doesnot generate as much persistence as is foundin cyclical fluctuations of output and prices.Persistence must be introduced. The problemis to introduce persistence without introduc-ing an implausible reason for neglectinginformation in current output, prices, interestrates and other variables.

In Brunner, Cukierman and Meltzer(1983), a representative producer sets priceand output at the start of each period usingall available information at the time. Thisincludes, in particular, the variables in equa-tion 1—inventory levels and changes, interestrates and current policy When making priceand output decisions, producers are uncertacnabout the permanence of observed shocks.As in the earlier discussion, they do notrespond to changes that are perceived tobe transitory.

Let x~’and m* denote the perceived per-manent components of x and cm. Knowingthese values, producers set output and theprice level; y~’and p’°denote the producers’decisions. They are determined from a subsetof the system of simultaneous equations:

(2) ujYa*, Pr*, Ah,*, ~rr, 1a*~r*, x’°,ma*] = 0.

The actual values are x and mm, but x - x’°andmm - are ignored in adjusting prices andoutput. Transitory changes are not innocuous.With y~and p’°adjusted to x~and rn*, nomi-nal and real interest rates and the change ininventories adjust to the perceived transitorychanges, shown as equation 3:

(3) -RYr*, p~Ah0

, hj~dr, ~a, ~~r’ mm,] = 0.

Output, inventories and other real values

FEDERAL RESERVE BANK OF ST. LOUIS

110

H 1~I1WMAY/JUNE 199$

respond to both nominal and real shocks inthis model by amounts that depend on thesize of the misperception. As new informationarrives, producers revise their beliefs aboutthe permanence of shocks; p’°.y~’and all othervariables adjust to the changed perception.

An econometrician examining the datagenerated by this model would at times findserially correlated changes in output andserially correlated errors. Serial correlationarises following a large permanent real ornominal shock if the shock is believed for atime to be transitory. As time passes, anderrors are repeated, perceptions adjust. Evenif the unconditional error in the populationis serially uncorrelated, misperception of thepermanence of shocks can lead producers tomake errors that, cx post, are serially correlat-ed. A model of this kind may explain whysome researchers have found cx post realeffects of anticipated changes in money See,inter alios, Mishkin (1983).

Ahhough prices do not fully adjust toshocks when they occur, decisions are entirelyrational. Producers use all available informa-tion, but they misinterpret the nature of theshock. Once they perceive that the shock ispermanent, prices and output fully reflectthe information.

The length of the recognition lag dependson the relative variance of permanent andtransitory shocks. The larger the variance ofpermanent shocks, relative to transitoryshocks, the shorter the recognition lag. If allshocks were permanent, prices and outputwould adjust top* andy* values as soon as theshocks occurred. The lag in our model wouldbe one period, as in the Lucas model. If allshocks are transitory y and p change, but y~and p~’never adjust.

In Keynesian models, inflexible prices(or wages) and gradual adjustment are taken asevidence of disequilibrium. The informationstructure of the model here implies that thisinference is invalid. Buyers and sellers useall available information and adjust to amarket equilibrium.

There are three types of equilibrium.At any moment, there is a permanent stockequilibrium characterized by the state variablesx* and m*. This equilibrium occurs when

= 0. The values of all variables are adjusted

to the perceived permanent shocks andthe condition of unchanging inventories.Each firm uses resources at the profit maxi-mizing rate. No firm seeks to expand orcontract output or change its price andinventory position.

A permanent equilibrium is less encom-passing. The state variables in this case aref, m~and ~o’ All other variables adjust tothese conditions. If x” and rn* remainunchanged, the permanent equilibriumconverges over time to the permanentstock equilibrium.

A transitory equilibriuan imposes anadjustment of the system to given values,mm nn*, x x*, y* p* and ~ao Inventoryadjustment and interest rate changes producethe transitory short-run equilibrium.

In Brunner, Cukierman and Meltzer(1983) we show how real variables respondto monetary shocks in a model of this kind.All expectations are rational. No informationis wasted once it is correctly perceived.Misperceptions occur, so the system adjustssluggishly to information that, cx post, turnsout to be permanent. The discussion here,following the original, uses inventories as arepresentative real variable but the responseof real output is similar.

Figure 2 shows the adjustment path.Asterisks denotes permanent values, and asuperscript a denotes actual values.

“Up to period t, the economy is at anequilibrium: H = H~.During period t,there is an unanticipated increasein money growth. Interest ratesfall; with prices fixed for the period,aggregate demand increases, andinventories are reduced. Inability toidentify permanent shocks means thatthe perceived value of cm changes asinformation about the rate of monetaryexpansion becomes available. Forecasterrors remain on one side of zero forseveral periods. Forecast errors rein-force the cyclical deviation of inventories.A large permanent increase in moneygrowth, that is not immediately recog-nized as permanent, increases twa,

and adds to the cyclical deviationof inventories.

FEDERAL RESERVE BANK OF ST. LOUIS

111

H1~I1WMAY/JUNE 1995

I I I I I It+l HZ t+3 to5 time

An asterisk in the table indicates the expected path of inventories.The outer envelopeshows the actual path.

“The reduction in inventories sets offa process of adjustment of output andinventories. The path along whichinventories are expected to adjust atthe onset of period t+1 is shown by thepositively sloped hne from ~ tothe permanent level of inventories, Ha’.Along this path, a typical firm plans toproduce output in excess of expectedsales and build inventories. The expectedvalue of inventories by the end of periodtel is shown as H0~2. If the monetaryshock is correctly perceived as transitoryand there are no further shocks, firmsadjust along the planned path andachieve the values of inventories, 1~1

r+2’

in successive periods until the per-manent value, Ha’, is restored.

“Suppose, however, that the increasein money growth persists. In period t+2,interest rates are again pushed below thevalue expected for the period, and actualinventories, H~,,are, again, below thevalue expected, H,

2, as shown in fig. 2.

At the beginning of period 1+2, firmsand households expect the economy toadjust along the new path, from I~I~2to W. The new path reflects all theavailable information about the shocks,including behefs about the permanenceof the change in money growth andknowledge of the structural parame-ters. . If the variance of the transitorycomponent of money growth is largerelative to the variance of the permanent

component, adjustment to permanentchanges is relatively slow. Inventoriescan fall below their expected value forseveral periods and, thus, move awayfrom Ha’.

“Additional information about the per-manence of the shock that first occurredin period t is revealed each period, so thepath of adjustment toward Ha’ is notsmooth. As time passes, however, theaddition to information is small. After t+3in fig. 2, inventories adjust toward H~’unless another shock—another unantici-pated increase in money growth—lowersinventories and starts a new process oflearning and adjusting.

“Actual inventories follow the outerenvelope in fig. 2; expected inventories,H, follow the adjustment paths that startat the actual values for each period. Thefigure shows principal features of ourmodel of inventory behavior, augmentedby the effect of permanent-transitoryconfusion. Deviations from Ha’ are onone side of Ha’ for several periodsbecause of the slow adjustment of inven-tories. This feature occurs even if allshocks are white noise. In addition,information about the permanence ofshocks becomes available graduallyPeople use all information and theirbeliefs about permanent values todetermine the adjustment path, butthey make unavoidable errors becausethey learn about the permanence ofshocks gradually”

SUMMARYMy emphasis in this section is on micro-

foundations that lead to price setting andto gradual adjustment. I do not claim tohave uncovered a unique structure thatproduces sticky prices under rational behavior.There are many reasons and many validhypotheses that make both price setting andgradual price adjustment compatible withrational behavior.

Price setting is sufficient for real effectsof monetary shocks. I have discussed several

FEDEEAL RESERVE BANK OF ST. LOUIS

IBVOIItOIy

Ut

111+2

r÷t EE~

112

D1~l1WMAY/JUNE 1995

reasons for price setting—menu costs,Knight’s uncertainty argument for theexistence of firms, producers’ desire tosignal quality, purchasers’ gains fromlower cost of information and bargaining,and other differences in costs of acquir-ing information.

In principle, sellers or buyers couldset relative, not absolute, prices. To doso efficiently, buyers and sellers have toagree on an index. Price index numbersare subject to real and nominal shocks thatare sometimes permanent, sometimestransitory, and sometimes alter the rate ofprice change persistently If these shockscould be correctly identified cx post, andif their expected duration were known, theparties might agree to adjust prices aftershocks occur. Far more often, cx postadjustment is done by negotiation, orthe parties agree to partial indexation andnegotiate about the remainder. Thesearrangements are consistent with the presenceof relatively large costs of acquiring informa-tion and agreeing on what has occurred,whether the change is permanent or transitoryand how long and how large future pricechanges will he. Among these costs are thecosts associated with moral hazard if oneparty controls relevant data.

Economic contractions typically lastnine months to one year. Considerableevidence suggests that price changes maylag as much as two years behind monetaryshocks. To yield the patterns of price andoutput change observed in actual cconocnics,price setting must be joined to a rationalreason for persistence.

Permanent-transitory confusion—uncer-tainty about the duration or persistenceof shocks—provides one such condition.Under this hypothesis, lags can be long orshort and cx post errors can be seriallycorrelated, if a large permanent shock isperceived as transitory If the variance ofhe permanent shocks is high relative to thevariance of transitory shocks, the lag isrelatively short, and there is no reason forsignificant serial correlation to he observedcx post. in this case, price setters believethat most shocks are permanent, so theyadjust promptly

avmwe~awr aMRevisions to Reported Inflation(percentage points)

Horizon(quarters followIng)

Average RangeReVISIOn

SOME EUIOENCEWe cannot directly observe how people

decide on the degree of persistence in therate of inflation (or other variables). However,we can measure some of the errors thatcontribute to permanent-transitory confusionand use econometric methods to estimate thevariance of permanent and transitory errorsin the price level (or other variables). Thissection considers these sources of evidenceon the relative size of permanent andtransitory changes.

Bullard (1994) reported the size of revi-sions to quarterly reports of the rate of changeof the GNP deflator for the years 1986 through1992. He found that the mean revision forthe 28 quarters was positive in this period;early reports understated the rate of inflationand later revisions added additional amounts.Bullard also reports the range of revisions foreach of the four quarters following the periodconsidered, The reported ranges exclude themost extreme 5 percent of the revisions ineach tail, Table I reproduces Bullard’s results,

The average rate of inflation for the periodis approximately 3 percent. The range ofrevisions (excluding extreme values) is 2.3 to2.9 percentage points. The revisions are from70 percent to 130 percent of the (3 percent)average reported change. The relatively largesize of the revisions suggests that it is rationalfor the public to act as if the initial announce-ment is a very noisy indicator of the true rateof inflation,

In several papers summarized inBrunner and Meltzcr (1993), we reportforecast errors for the rate of change of outputor inflation using different models, methodsof forecast and countries. The forecasting

FEDERAL RESERVE BANK OF ST. LOUIS

2

3-1

0.2 -LOtol.5

0.3 -l.2tol.7

0.3 -0.6 to 1.7

03 -06to17

113

HIYIEWMAY/JUNE 199$

/ See Iamhoff 119831 far a divas-sioa at the tayesiaa malti’stoteKalnaaa filter.

~X is the od)asemeea log giaen by

a.&— ,.a+~..--.~ .1

and on the ratio of the connate afthe permooeat aompoaoat to theaariaace of the traasieoay campaaeae.

methods include state-of-the-art econometricmodeling, time-series analysis, judgment andcombinations of these methods. Arule-of-thumb summary is that meanabsolute errors for output growth in themajor industrial countries is 50 percent ormore of the average rate of change onequarter or one year ahead, Inflation is lessvariable than output over short periods, andits forecast error is a smaller fraction of theaverage rate of change. Still, errors in hothgrowth and inflation forecasts are large.

The data on forecast errors make apersuasive case that forecasters frequentlymisperceive future values, Data on revisionssuggest that current reports arc subjectto large errors. Errors maybe unbiased,but that is a small consolation for thosewho adjust to events that are found laternot to have occurred or those who do notadjust to changes that did occur. Thepermanent-transitory distinction implies thatpartial adjustment is opthmal. Ex post, it willseem that adjustment has been sluggish.Indeed, as noted, errors may appear to beserially correlated following large changes.

A filter can be used to separate permna-ncnt and transitory components! Let theerror term in an aggregate such as the pricelevel or output have a permanent and transi-tory’ component,

(4) =8

a’~+ E~

where p and r denote the two components.The transitory component is white noise,The permanent component has the propertyof zero mean and constant variance, ca’2,

At any time t, the expected change in thepermanent component is zero.

People observe prices and output. As inthe model of the previous section, they cannotseparate the levels or changes of the permanentor transitory components. The rate at whichthey adjust depends on the relative variancesof the transitory and permanent componentsof inflation or output growth. The larger thevariance of the permanent component relativeto the variance of the transitory component,

the quicker the economy adjusts to permanentchanges. A relatively large transitory varianceslows the response.

Meltzer (1986) estimated the variancesof the permanent and transitory componentsof inflation and growth for Canada, Germanythe United Kingdom, and the United Statesunder fixed and fluctuating exchange rates.The fixed exchange rate period runs from thefirst quarter of 1960 through the third quarterof 1971. Fluctuating rates begin in the fourthquarter of 1971 and end in the fourth quarterof 1984,8

Table 2 shows the ratios computed fromthese data using the adjustment equations inMuth (1960). The estimated variance ratiosfor inflation and growth changed under fluc-tuating rates, but the direction of change isnot uniform across countries. The length ofthe adjustment lag required to distinguishpermanent from transitory shocks changedmuch less. For inflation, seven of the eightvalues of the adjustment lag (A) lie between0.41 and 0.56. These values imply that, con-sistent with Blinder’s survey data, about halfof the shock to inflation is seen in the currentquarter. Between 82 percent and 96 percentof the adjustment of permanent inflationoccurs within four quarters of the initialshock to the inflation rate. This adjustmentis faster than the two-year average lag com-monly suggested. For real growth, seven ofthe eight values of A lie between 0.45 and0.67; within four quarters, 91 percent to 99percent of the adjustment occurs. The speedof output adjustment is broadly consistentwith the length of post-war recessions ifthese recessions arc interpreted as the cuanu-lative adjustment set off by a monetary orreal shock. Inflation adjusts more slowlythan growth as many studies have shown.

Clarida and Gali (1994) studiedthe response of real exchange rates to nominalshocks in four countries. For Canada and theUnited Kingdom, they were unable to findany structural effects—real or nominal—onreal exchange rates.

For Germany and Japan, the evidencesuggests that nominal shocks explain45 percent and 34 percent, respectively ofthe four-quarter-ahead forecast error variancesof the log level of bilateral real dollar exchange

FEDERAL RESEEVE BANK OF ST. LOUIS

114

llt~It~MAY/SEINE 1995

rates. Clarida and Gali note that theirestimates are consistent with the evidencefrom vector autoregressions reported inEichcnbaum and Evans (1992).

Clarida and Gali (1994) use a trivariatevector autoregression to estimate the transi-tory component of real exchange rates.°Their model includes shocks to aggregatedemand and supply The authors report theratio of the variance of transitory shocks tothe variance of actual shocks. Using theirdata, we can compute the ratio of the varianceof permanent shocks to the variance of tran-sitory shocks. The ratio covers a wide rangein these countries—from 0.42 in Germany to3.76 in Canada. These findings suggest thatmost shocks to the Canadian real exchangerate have been permanent while most shocksto the German real exchange rate were tran-sitory Hence, the Canadian real exchangerate should adjust more rapidly to shocksthan the German real rate.

Meltzer (1993) used the permanent-transitory distinction to model the U.S.multilateral real exchange rate. For both levelsand first differences under fixed and fluctuat-ing rates from 1960 to 1991, the data suggestthat there is a large permanent component inthe change of the real exchange rate and asignificant transitory component. Further,the data suggest that the multilateral realexchange rate responds to changes in thenominal stock of money The effect eventuallyvanishes, but monetary changes have realeffects until prices adjust. These findings areconsistent with short-run non-neutrality andlong-run neutrality of money if permanentchanges in money were perceived as transitoryat the time they occurred or conversely

The studies of prices, output andexchange rates support the principalarguments of the article. They are only asmall part of the evidence supporting cxpost, short-run non-neutrality They arc ofinterest because they attribute slow adjust-ment of nominal values and real effects ofnominal shocks to the difficulty of discerningthe persistence of shocks. With a non-trivialcost of acquiring information, price settingand permanent-transitory confusion implythat nominal changes have real effects thatpersist for a time.

Relative Variances of the Permanent andTransitory Components

196041Canada Germany IlK. US.

Other recent studies find evidenceof costs of acquiring information.Investors frequently pay a premium to buycountry-specific mutual funds. The premiumimplies that they could buy the individualsecurities at lower cost. If they are uncertainabout which securities to buy and when tobuy or sell, it may he rational to pay for theservices of traders who specialize in theparticular market,

Smith (1991) uses costs of acquiringinformation as one reason for the absenceof optimal portfoho diversification of worldmarket securities. The degree of diversifica-thou depends on costs of acquiring information,People know much more about values andearnings in their own market than in foreignmarkets, Prices in foreign markets may reflectfull information, but some investors eitherdo not have this information or cannot assesswhether changes are permanent. Hence,they do not respond promptly to informationabout each of these markets. They do nothold the “true” equilibrium portfolio theywould hold if information costs were zero.A principal cost in this case, as in others, isthe interpretation of available information.Permanent-transitory confusion is one partof the interpretation problem.

In Fuhrer and Moore (1993), the inflation

FEDERAL RESERVE BANK OF ST. LOUIS

Inflation 0.29 1.56 031 0.11‘•/‘ 0.41 0.69 045 0.56

Real growth 0.62 0.65 0.36 1.110.5/ 0.54 045 0.64

197 1-84(anado Germany U K. U.S.

Inflation 0.56 0.30 0.31 0.710.52 0.42 0.42 056

Real grov.ih 135 100 0.04 0.860.6/ 0.62 0.18 0.59

/ The tnivadate system has loan

logged aaloes af the change in thelog real eachoage rote, the changein tIne log ratio 0f United States~home toaotny real GD~and thedifference between IfS. and for’olga inflation. Data one qaontenlyfrom rnid’1970 tothefoonthqaorten of 1992.

115

DII\’IF~MAY/JUNE 199$

a The tnianniote system has four

lagged valaes of the change in thelog real euchonge rate, the changnin the log nofo of United States-home country real 601, and thedifference between t.S. aad for-eign inflotion, Date ore quarterly

from mid’l 970 cache barthqaactenaf 1992.

rate is sticky. Firms adjust relative prices tothe average of other sectors’ expectedrelative prices over the hfc of existingcontracts, Firms also adjust for thecurrent and expected level of output. Theautocorrelation functions based on theirmodel have very similar shapes to the auto-correlations generated by an unconstrainedvector autoregression. In particular, theyshow considerable persistence in inflationand output movements and sustainedeffects of inflation on real output. In short,Fuhrer and Moore provide evidence that isconsistent with a model in which there arccosts of learning about the permanence orpersistence of changes, and in whichpeople adopt strategies that leave roomfor misperception and real effects ofnominal changes.

Earlier work by Boschen andGrossman (1982), Gordon (1982) andMishkin (1983) also provide evidencethat supports sticky prices. Indeed, theevidence of gradual adjustment of pricesand of short-term real effects of monetarychange is common, These studies lackmicro-foundations, Price-setting in part ofthe economy and permanent-transitoryconfusion, as in Brunner, Cukierman andMeltzer (1983), reconciles this evidencewith rational behavior.

CONcLUSiONThe examples in the preceding section

are a small part of the recently accumulatedevidence showing that there is much morethan casual observation to support the mainpropositions in this paper: Nominal pricesadjust with a lag. The lag is sufficientlylong that real variables respond to nominalchanges. Costs of acquiring informationabout the persistence of observed changes—permanent-transitory confusion—is a maincomponent of the cost and a main reasonfor the lag. Even where prices reflect infor-mation fully, all individuals or firms maynot have adjusted fully to available butcostly information,

The oft-repeated comment thatmacroeconomics should he built onmicro-foundations is correct if and only

if the micro-foundations are appropriate forthe task. Standard micro theory, such asArrow-Debreu, imposes complete marketsand market clearing in each market. Thereis no role for monetary disturbances. This isnot the appropriate micro-foundation formacroeconomics. No amount of squeezing,cutting and pasting will make it so.

Rational behavior and rational expecta-tions are entirely consistent with costs ofacquiring information and the inability tofully identify permanent and transitoryshocks either when they occur or for severalquarters after. Indeed, Muth’s (1961) initialformulation of rational expectations is basedon the latter distinction,

One alternative explanation of stickyprices in recent literature relies on menucosts and imperfect competition. This expla-nation is a foundation for the so-calledL-shaped supply curve famihar from Keynesiantheories, I show that the evidence inBlinder’s (1991) survey rejects the L-shapcdsupply curve. Further, the implications ofmonopolistic competition, such as widespreadexcess capacity do not explain gradual priceadjustment in most service industries.

Gordon (1990) proposes a disaggregaredsystem to take account of the information atmany levels of the economy He argues thatprices respond to marginal cost but that mar-ginal cost for any firm depends on the pricingstrategy of its suppliers. Hence, such infor-mation enters firms’ decisions about priceand output adjustment. Information frommacrodata is much less relevant.

This argument captures some of thedynamics of pricing, but it poses unresolvedchallenges for aggregation over differentindustry structures. Moreover, Gordon’sframework imphes that commodity producersshould adjust slowly to aggregate shocks andthat large firms in durable goods industriesshould adjust promptly The stylized factssuggest that the opposite is true. One reasonis that organized commodity markets increasethe information available to commodity pro-ducers. There are no comparable markets forconsumers’ and producers’ durable goods out-put. Differences in information and the costsof acquiring information are consistent withthe stylized facts on speed of adjustment.

FEDEEAL RESERYR BANK OF ST. LOUIS

116

II FYI F~MAY/JUNE 1995

The micro-foundations suggested in thisarticle use costs of acquiring information toexplain three common observations. First,many prices are set. Second, price setterschoose nominal values. Third, the daily week-ly monthly or quarterly variances of “setprices” are small fractions of the variance ofprices in auction markets; set prices are stickier.

Households and firms do not operate ina world of full information. Incomplete infor-mation and costs of acquiring information arecentral problems of a monetary economyInformation and transaction costs explain whypeople hold and use money as a medium ofexchange (Brunner and Meltzer, 1971). Thereis considerable evidence that these costs arenot tnvtal. The article cites revisions to report-ed data, forecast errors, incomplete informa-tion about costs, profits and strategies ofcompeting firms. These examples do notexhaust the costs that firms and individualsface.

Some of these costs can be reduced byinstitutional and contractual arrangements.The arrangements that people choose maybe optimal when contracts are written but,with changes in the environment, there areunforeseen gains and losses. If the gains andlosses are transitory, their expected value iszero, It may not be worthwhile to change thecontract or the method of contracting. Oncethe change is considered persistent, gains andlosses are expected to cumulate. Adjustmentor re-negotiation becomes more appealing toat least one party

Information about permanent and tran-sitory changes in profits, prices, wages andother variables is costly to acquire. Thedistribution of shocks between real andnominal, permanent and transitory may differfrom one sample period to the next, Peoplelearn to monitor events or changes that arecostly to ignore. But learning requires acontinuous process of monitoring both whathas happened and what should he observed.This is a basic problem for firms and house-holds. As such, it is a more appropriatemicro-foundation for macroeconomics,

REFERENCESAichion, Arnroa A. “Why Money?” Joairnal of Money, Creditand

Banking (Febroory 1977, part 2), pp. 133-40.

Ball, Lautence, and N. Gregory Monkiw. “A Sticky Price Manifesto,”Carnegie Rochester Conference Series on Pub/k Policy (December1994), pp. 127-52.

Dalvees, R. i. and T. F Casimano. “Actively Learning Abott Demandand the Dynamics of lice Aditstment,” EconomicJournal (September1990) pp. 882-98.

Baumol, William J. “The Transactions Demand for Cash: An InventoryIhearetic Approach,” QuarterlyJournal ofEconomics (November1952), pp. 545-56.

Beaulieu, Joe., and Joe Motley. “The Effects of General Inflation andldiosyncmtic Cost Shocks on Within-Commodity Price Dispersion:Evidence from Microdata,” Finance and Economic Discussion SeriesNo. 94-12 (May 1994), Baard of Governors of the Fodeml ReserveSystem.

Blinder, Man S. “Why Are Pices Sticky? Preliminary Results from anInterview Study,’ The American Economic Review (May 1991),pp. 89~96.

Bamhaff, Eduard J. MonetaryUncertainly. North Holland, 1983,

Doschen, John F., and Herschel I. Grossman. “Tests of EquilibriumMacroeconomics Using Cantemporaneats Monetary Data,” Jauraal ofMonetary Economics (Navember 1982), pp. 309-33.

Brunner, Karl, and Allan H, Meltrec Money and the Economy: Issues inMonetary Analysis. Cambridge University Press, far the RaffoeleMattiali Lectures, 1993.

___________ and ___________, “The Uses af Money: Money in theTheory of an Exchange Economy,’ The American Economic Review(December 1971), pp. 784-805.

___________ and ___________, “Economies af Scale in Cash BalancesReconsidered,” Quarterly Journal of Economics (August 1967),pp. 422-36.

_________- Alex Cukierman and Allan H. Meltrer. “Money andEcanamic Activity: Inventonies and Business Cycles,” Journal ofMonetary Economics (May 1983), pp. 281-319.

Dullard, James B. “How Reliable Are lnflatian Reports?” FederalReserve Bank of St. Louis Monetary Trends (February 1994), p. 1.

Clarida, Richard, and Jardi Gali, “Sources of Real Exchange RateFluctuations: How Important Are Nominal Shacks?” CarnegieRochester Conference Series an Pub/k Pa/icy (December 1994),pp. 1-56.

Eichenbaum, Martin, and Charles Evans. “Same Empirical Evidencean the Effects ofMonetary Policy Shacks an Exchange Rates,”Federal Reserve Bank af Chicago Wading Paper No. 92~32(December 1992).

Fahrer, Jeffrey, and George Moore. “Inflation Persistence,” Finance andEconomk Discussion Series Na. 93-77 (May 1993), Board ofGovernors of the Federal Reserve System.

Gordon, Robert J. “What Is New’Keyaesian Econamics?’ Journal afEconomic Literature (September 1990), pp. 1115-71.

__________ “Price Inertia and Policy Ineffectiveness in the UnitedStates, 1890-1980,” Journal ofPolitical Econamy (December 1982),pp. 1087-1117.

FEOERAL RESERVE BANK OF ST. LOUIS

117

DFYIF~MAY/JON. T995

Heatan, John, and Debra Lucas. “The Importance of Investor Muth, John F. “Rational Expectations and the Theory of PriceHeterogeneity and Finandal Market Imperfections for the Behavior Movements,” Ecoaornetrica (July 1961), pp.315-35.ofAsset Prices,” Carnegie Rochester Conference Series an Public ,

Poli (forthcamin ) . Optimal Properties of Exponentially Weighted Forecasts,Journal a/the American Statistical Assaciallan (June 1960),

Knight, Fmnk H. Risk, Uncertainty and Profit (Na. loin repints of pp. 299-306.Scarce Classics in Economics), London School of Economics, 1933. . - ~,,

Smith, Clifford W., Jr. Globalization of Financial Markets, CarnegieLucas, Robert E., Jr. “Expectations and the Neutrality of Money,” Racltester Conference Series on Public Policy (spring 1991),

Journalof Ecanamic Theory (April 1972), pp. 1D3-24. pp. 7T96.

McCaIlum, Bennett T. “New Classical Macroeconomics: A SympatheticAccount,” Scandinavian Journalof Economics (June 1989),pp. 223-52.

Meltzer, Allan H. “Real Exchange Rates: Some Evidence from thePostwar Years,” this Review (March/ApdI 1993), pp. 103-17.

___________ “Size, Persistence and the Interrelation of Nominal andReal Shocks: Some Evidence from Four Countries,” JournalatMonetary Economics (January 1986), pp. 161-94.

_________ “Rational Expectations, Risk, Uncertainty and MarketResponses,” in Paul Wochtel, ed., Crises in the Economic and FinancialStructure. Lexington Books, 1982, pp. 3-22.

Mishkin, Fredrick S. A Rational Expectatians Approach toMacraecanametcics: Testing Policy Ineffectiveness and EfficientMarkets Models. University of Chicago Press, 1983.

FEDERAL RESERVE BASK OF ST. LOUIS

113