Академический Документы
Профессиональный Документы
Культура Документы
Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless
you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you
may use content in the JSTOR archive only for your personal, non-commercial use.
Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at .
http://www.jstor.org/action/showPublisher?publisherCode=aea. .
Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed
page of such transmission.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.
American Economic Association is collaborating with JSTOR to digitize, preserve and extend access to The
American Economic Review.
http://www.jstor.org
Does Money IllusionMatter?
Until recently, the notion of money illusion Bewley, 1998; Alan S. Blinder et al., 1998). It
seemed to be thoroughlydiscreditedin modem has puzzled economists for decadesbecause it is
economics. James Tobin (1972) described the quite difficult to explain in an equilibrium
negative attitude of most economic theorists model with maximizing individuals. Instead of
towards money illusion as follows: "An eco- money illusion other factors like informational
nomic theoristcan, of course, commit no greater frictions (RobertE. Lucas, Jr., 1972), staggering
crime thanto assume money illusion" (p. 3). As of contracts (e.g., Stanley Fischer, 1977; John
a consequence, money illusion has been anath- B. Taylor, 1979), costs of price adjustment(N.
ema to the profession for several decades. The Gregory Mankiw, 1985), and near-rationality
index of the Handbookof MonetaryEconomics (Akerlof and Janet L. Yellen, 1985) have been
(Benjamin M. Friedman and Frank M. Hahn, invoked to explain nominal inertia.
1990), for example, does not even mention the In this paper we do not contest the potential
term "money illusion." In principle, money il- relevance of these explanations. We do, how-
lusion could provide an explanationfor the in- ever, arguethat money illusion has prematurely
ertiaof nominalprices and wages and, thus, for been dismissed as a potential candidatefor the
the nonneutralityof money. The stickiness of explanation of sluggish nominal price adjust-
nominal prices and wages seems to be an im- ment. Our argumentis based on rigorousexper-
portant phenomenon (see, e.g., George A. imental evidence from a price-settinggame that
Akerlof et al., 1996; Ben S. Bernanke and isolates money illusion from other potentialde-
Kevin Carey, 1996; David Card and Dean terminantsof nominal inertia. In particular,we
Hyslop, 1997; ShulamitKahn, 1997; TrumanF. show thataftera fully anticipatednegative nom-
inal shock, long-lasting nominal inertia pre-
vails, even if informationalfrictions, costs of
* Fehr: Institute for Empirical Research in Economics, price adjustmentand staggeringare absent. Our
University of Zurich,CH-8006 Zurich, Switzerland;Tyran: results indicate that the direct and indirect ef-
Departmentof Economics, University of St. Gallen, CH- fects of money illusion are the major determi-
9000 St. Gallen, Switzerland.We are particularlygrateful
for two excellent referee reports and for comments by
nants of this long-lasting nominal inertia. We
George Akerlof, LindaBabcock, Jim Cox, Urs Fischbacher, show, in addition, that money illusion causes
Simon Gachter, Ed Glaeser, Lorenz Goette, Charles much less nominal inertia after a fully antici-
Goodhart,ReinhardSelten, Dick Thaler,Fransvan Winden, pated positive nominal shock. This result is
and Michael Waldman.In addition, we acknowledge help- reminiscent of the Keynesian proposition that
ful comments by the participantsof many seminars around
the globe. ErnstFehr gratefully acknowledges the hospital- downwardwage rigidity causes asymmetricre-
ity of the Center for Economic Studies (CESIFO) in Mu- sponses to monetary shocks. Yet, since we ob-
nich. Valuable research assistance has been provided by tain our result in a price-setting game, the
MartinBrown, BeatriceZanella, and Tobias Schneider.We asymmetricresponse cannot be directly related
are grateful for financial support by the Swiss National
Science Foundationunder Project No. 1214-051000.97/1, to downwardwage rigidity. Our results suggest
and by the EU-TMR Research Network (Project No. that the asymmetry is caused by a particular
FMRX-CT98-0238). form of money illusion arising from people
1239
1240 THEAMERICANECONOMICREVIEW DECEMBER2001
taking nominal payoffs as a proxy for real pay- librium. Although there are no direct costs of
offs. After a negative money shock, nominal forming expectations in our experiments, it is
payoffs decline because prices tend to decline, quite likely thatthe task of formingexpectations
while after a positive shock nominal payoffs involves cognitive costs. Taken together, the
increase because prices tend to rise. If these results of our experiments suggest that money
changes in nominalpayoffs are taken as a proxy illusion matters, i.e., money illusion should be
for changes in real payoffs there will be more considered as a serious candidatein the expla-
reluctance to adjust prices to the new equilib- nation of nominal inertia.
rium after a negative shock. The rest of the paperis organizedas follows:
Our experiments also allow us to judge the In Section I we discuss the notion of money
relative importance of the direct and indirect illusion and its potential aggregateimplications
effects of money illusion on nominal inertia. in more detail. In Section II we argue that
The direct effects of money illusion are defined experimentalmethodsare appropriatefor study-
as those effects that are the direct result of ing whether money illusion matters and we
individual optimization mistakes. The indirect present our experimentaldesign. In Section III
effects of money illusion are defined as those the experimentalresults of the design with the
effects that arise because some agents expect negative nominal shock are presented. Section
that others are prone to money illusion and, as a IV argues that the nature of money illusion in
consequence, they behave differently. The dis- our experiment suggests that after a positive
tinction between the direct and the indirect ef- nominal shock there should be less nominal
fects of money illusion is importantbecause inertia.This conjectureis tested in a design with
many economists seem to believe that money a positive nominal shock. In the final section we
illusion is not a widespreadphenomenonat the summarizeand interpretour main results.
individual level, i.e., that the direct effects of
money illusion are small. The textbookexample I. Money Illusion at the Individual
where all nominal prices and nominal incomes and the Aggregate Level
are doubled nicely illustrates this view. It is
hardto believe that many people make an indi- A. Money Illusion at the IndividualLevel
vidual optimizationmistake by choosing a dif-
ferentbundleof goods when prices and incomes Wassily Leontief (1936) defined money illu-
are doubled. Ourresults clearly show, however, sion as a violation of the "homogeneitypostu-
that it would be misleading to conclude that late." This postulate stipulatesthat demand and
money illusion is largely irrelevantbecause the supply functions are homogeneous of degree
directeffects of money illusion are small. In our zero in all nominalprices which means thatthey
experimentsthe directeffects of money illusion depend only on relative and not on absolute
on nominal inertia after the negative shock are prices. Although other authors have used
also rathersmall but the total effects neverthe- slightly different definitions, the intuition be-
less are very large. The reason for this findingis hind their definitionsseems to be rathersimilar.
that money illusion renders price expectations This intuition says that if the real incentive
very sticky after the negative shock, which- structure,that is, the objective situation, an in-
under conditions of strategic complementar- dividual faces remains unchanged,the real de-
ity-induces agents to choose sticky prices. cisions of an illusion-free individual do not
This result lends supportto theories that stress change either.Two crucial assumptionsunderly
that small amounts of individual-level irratio- this intuition:First, the objective function of the
nality can have large aggregateeffects (Akerlof individualdoes not depend on nominalbut only
and Yellen, 1985; John Haltiwanger and on real magnitudes. Second, people perceive
Michael Waldman,1985, 1989; ThomasRussell that purely nominal changes do not affect their
and RichardThaler, 1985). It also lends support opportunity set. For example, people have to
to the view of George W. Evans and Garey understandthat an equi-proportionatechange in
Ramey (1992, 1998) thatcostly expectationfor- all nominal magnitudes leaves the real con-
mation causes expectationsand prices to adjust straintsunaffected.Whetherpeople are, in fact,
only graduallyto the rationalexpectationsequi- able to pierce the veil of money, i.e., whether
VOL.91 NO. 5 FEHR AND TYRAN:DOES MONEYILLUSIONMATTER? 1241
they understandthat purely nominal changes with a given aggregate price level. However,
leave their objective circumstancesunchanged, this rule is inappropriatein situationswhere the
is an empirical question. Irving Fisher (1928), aggregateprice level is changing.Therefore,the
for example, was convinced that ordinarypeo- applicationof this rule in an environmentwith
ple are, in general, prone to money illusion. changing aggregateprices constitutesa form of
More recently Eldar Shafir et al. (1997) pro- money illusion.
vided interesting questionnaire evidence indi-
cating that frequentlyone or both preconditions B. Money Illusion at the Aggregate Level
for the absence of money illusion are violated.
Their results suggest that the preferences of In the past, economists frequently invoked
many people as well as their perceptionsof the the assumption of money illusion to account
constraints are affected by nominal values. for the short-runnonneutralityof money (e.g.,
Moreover, the answers of many people do not Fisher, 1928). However, since the success of
only indicate that they themselves are prone to the rational expectations revolution, econo-
money illusion but that they also expect other mists have been extremely reluctant to invoke
people's behavior to be affected by money money illusion to explain the short-run non-
illusion. neutrality of money. A common feature of the
Since the absence of money illusion means models of New Classical and New Keynesian
that an individual's preferences, perceptions macroeconomists is that they exclusively fo-
and, hence, choices of real magnitudes are not cus on the equilibrium states of their econo-
affected by purelynominalchanges, it is natural mies. In general, they remain silent on how
to view money illusion as a framing or repre- economic agents move from one equilibrium
sentation effect. From this viewpoint, an indi- to the other. In models that exclusively focus
vidual exhibits money illusion if his or her on equilibrium, the assumption of the absence
decisions depend on whetherthe same environ- of money illusion is very intuitive because it
ment is representedin nominal or real terms. is difficult to imagine that an illusion could
There is a large body of experimentalresearch persist in equilibrium. However, there is a
that shows thatalternativerepresentationsof the strong a priori argument that money illusion
same situation may well lead to systematically is likely to affect the adjustmentprocess of an
different responses (ReinhardSelten and Claus economy after a fully anticipated monetary
C. Berg, 1970; Amos Tversky and Daniel shock. This argument is based on the simple
Kahneman,1981). Representationeffects seem fact that in an interactive situation the failure
to arise because people tend to adopt the par- of some agents to fully adjust to the nominal
ticular frame that is presentedand evaluate the shock will, in general, provide incentives for
options within this frame. Because some op- other agents to not fully adjust to the shock,
tions loom larger in one representationthan in either. Thus, there may be a snowball effect
another, alternative framings of the same op- that causes less than full adjustment for a
tions may provoke different choices. prolonged period of time.
It is importantto note that the nominal rep- This can be illustrated in the context of a
resentationof an economic situationis probably monopolistically competitive economy as ana-
the naturalrepresentationfor most people. This lyzed in, for example, Akerlof and Yellen
is so because most economic transactions in (1985) or Olivier Jean Blanchardand Nobuhiro
people's lives involve the use of money and, Kiyotaki (1987). To keep the argumentsimple
hence, are framed in nominal terms. Therefore, we focus solely on the firms' behavior. The
it is likely that people often perceive and think reduced-formreal profit function for firms in
about economic problems in nominal terms these models can be written as
which may induce money illusion. A rather
basic form of money illusion occurs when peo- (1) -rrj= -fi(Pi/P, M/P), i = 1, ... n
ple take nominal values or changes in nominal
values as a proxy for real values or changes in where rriis firm i's real profit,Pi is the nominal
real values, respectively. Note that this rule of price set by firm i, P is the aggregate price
thumb makes perfect sense in an environment level, M denotes the supply of money, and n the
1242 THEAMERICANECONOMICREVIEW DECEMBER2001
numberof firms.1In these models MiP is pro- incentive to partly compensate the behavior of
portional to real aggregate demand. For sim- the nonrationalones so that the latter have a
plicity, we assume identical firms, the absence disproportionatelysmall impact on the aggre-
of menu costs and informationalfrictions, and a gate outcome. The results of Haltiwanger and
uniqueand symmetricequilibriumP* = PJ, for Waldman (1989) thus suggest that, given stra-
all i, j. In this equilibriumeach firm maximizes tegic complementarity,the existence of a small
real profits by setting P* = P*. Since (1) is group of subjects that suffer from money illu-
homogeneousof degree zero in Pi, P, and M, a sion may generate substantialnominal inertia.
change in M to AM (A 0 1) leads to post-shock However, while this is a plausible theoretical
equilibriumvalues of AP* and AP*. argument,there is, to our knowledge, no empir-
Supposenow thatthereis one groupof agents ical evidence for the claim that a small amount
who suffers from money illusion and does, of money illusion may generate substantial
therefore,not fully adjust their nominal prices nominal inertia.2
to AP*. Suppose furtherthat there is a second
group of agents that anticipatesthe behavior of II. An ExperimentalApproach
the first group. The second group, therefore, to MoneyIllusion
anticipates a change in real aggregate demand
M/P such that their members, in general, have One way to rigorously examine whether
an incentive to choose a price that differs from money illusion matters,is to look for a natural
AP*, too. Whetherthe interactionbetween these experiment in which an exogenous and fully
groups causes aggregate nominal inertia de- anticipated monetaryshock occurs. In order to
pends in an importantway on the strategic en- unambiguously identify whether the shock is
vironment,that is, whether agents' actions are fully anticipated,the researcherneeds to know
strategic complements or strategic substitutes. individualinformationsets before the shock. To
Haltiwangerand Waldman (1989) have shown judge whether the anticipated shock causes a
that in the presence of strategic complementar- disequilibrium and nominal inertia, the re-
ity between agents' decisions, the existence of a searcherhas to know the equilibriumvalues of
small group of nonrational subjects can have nominal prices before and after the shock.
large effects on the process of adjustmentto Moreover, to examine whether money illusion
equilibrium.In the above-mentionedmodel of causes nominal inertia, the researcher should
monopolistic competition, strategic comple- identify two similarnaturalexperiments.In one
mentaritymeans that firm i's profitmaximizing experiment the "world" should be framed in
nominal price P' is positively related to the nominal terms while in the other experimentit
aggregate price level P. This means that firms should be framed in real terms. In our view, it
which believe that, because of money illusion, seems extremely difficult, if not impossible, to
the prices of other agents are kept close to the meet the above requirementswith field data. In
pre-shock equilibriumhave a rationalreason to fact, the exogeneity of monetarypolicy and the
choose a nominal price that is also close to the causality between money and outputis a matter
pre-shock equilibrium. of considerable debate (e.g., Christina D.
Thus, under strategic complementarityratio- Romer and David H. Romer, 1989, 1994; Kevin
nal firms have an incentive to partly imitate the D. Hoover and Steven J. Perez, 1994; Wilbur
behavior of the nonrationalfirms which gives John Coleman, 1996). In addition, full knowl-
the latter a disproportionatelylarge impact on edge of the pre- and post-shock equilibrium
the aggregate price level. In contrast, in the
presenceof strategicsubstitutability,i.e., if P' is
negatively related to P, rational firms have an 2
Since strategic complementarityis importantfor our
argument in favor of the aggregate relevance of (beliefs
about) money illusion, one would like to know to what
1 Equation (1) already incorporates(i) the maximizing extent it does prevail in naturally occurring economies.
behavior of all households, (ii) the cost-minimizingbehav- SeonghwanOh and Waldman(1990, 1994), Russell Cooper
ior of all firms for given output and wages levels, (iii) the and Haltiwanger(1996), and Blinder et al. (1998) provide
equilibrium real wage, and (iv) the equilibrium relation evidence in favor of the relevance of strategiccomplemen-
between real aggregate demand and real money balances. tarity in real economies.
VOL 91 NO. S FEHR AND TYRAN.DOES MONEYILLUSIONMA7TER? 1243
values of nominal prices is clearly beyond the They were free to change their nominal prices
informationcontent of presently available field in each period at no cost. The playersinteracted
data. In reality, almost all business transactions anonymously via computer terminals. Each
are shroudedin nominal money, i.e., it is very treatmentcondition had 2T periods. During the
difficult to find real-world examples of a real first T periods of a session the money supply
frame. was given by Mo. Then we implementeda fully
In an appropriatelaboratorysetting,however, anticipated monetary shock by reducing the
the above-mentioneddata requirementscan be money supply to M1. This shock and the fact
met. The techniquesof experimentaleconomics that the post-shockphase again lasted T periods
allow the implementation of exogenous and was common knowledge.
fully anticipatednominal shocks and the exper- Our majorinterestconcerns subjects' pricing
imenter can exert full control over pre- and behaviorin the post-shockphase. The pre-shock
post-shock equilibrium values of nominal phase serves the purpose to make subjects ac-
prices. In addition, the experimenter controls quainted with the computer terminal and the
the framing of the situation, e.g., whether sub- decision environment. In addition, and more
jects receive the payoff informationin nominal importantly,the pre-shock phase allows us to
or in real terms. These enhancedcontrol oppor- see whether subjects reach equilibrium in the
tunities suggest that laboratory experiments pre-shock phase. After all, one can only argue
provide valuable informationregardingthe im- that money illusion is a disequilibratingforce if
pact of money illusion on nominal inertia, equilibriumhas in fact been reachedbefore the
which may complementand help to interpretthe shock.
results of studies based on field data (for field The real payoff of subject i, 7ni, is given by
evidence see, e.g., Michael Abbott and Orley
Ashenfelter, 1976; Beth T. Niemi and Cynthia (2) n i= 7Ti(Pi, P_i, M) i = 1, ... , n
B. Lloyd, 1981). The use of experimentalmeth-
ods also distinguishesour examinationfrom the where Pi denotes i's nominal price, P-i repre-
study of Shafiret al. (1997). While these authors sents the averageprice of the other n - 1 group
asked subjects hypothetical questions, we di- members while M denotes a nominal shock
rectly observe the evolution of individual and variable(money supply).The nominalpayoff of
aggregate behavior after a nominal shock. subject i is given by P_iri. In total, we have
four treatmentconditions and the payoff func-
A. General Description of the tions (2) are the same in all conditions.The four
ExperimentalDesign conditions differ along two dimensions (see
Table 1). Thefirst dimensionconcernsthe fram-
To study the impact of money illusion, we ing of the situation, i.e., whether payoffs are
designed an n-player pricing game with stra- representedin real or in nominal terms. In the
tegic complementarity and a unique equilib- real treatments,denotedby RC and RH, subjects
rium. The pricing game was divided into a received the payoff informationin real terms
pre-shock and a post-shock phase. All n while in the nominal treatments,denotedby NC
players simultaneously had to determine their and NH, payoffs were representedin nominal
nominal prices in each period of the game. terms.Thus, to computetheirreal payoffs in the
1244 THEAMERICANECONOMICREVIEW DECEMBER2001
nominal treatmentssubjects-hadto divide their RC we can test the hypothesis whetherthere are
nominal payoffs P_i7i by P_. individual-levelirrationalitiesotherthan money
The second dimension concerns the fact illusion.
whether our experimentalsubjects face n - 1 In the NC, in contrast, money illusion can
preprogrammed computerized players or affect the behavior of individuals because as a
whetherthey face n - 1 other human subjects. part of the individual optimizationproblemhu-
The crucial point here is that in the computer- man subjects have to correctly deflate nominal
ized condition where one human subject faces payoffs at the various (Pi, PJ) combinations.
n - 1 preprogrammedcomputers,the subjectis Hence, by comparing the post-shock prices of
informed about the aggregate response rule of human subjects in the RC and the NC we can
the computersin advance. The response rule of observe whether there exists money illusion at
the computersis given by the best replies of the the individual level.
computers [based on the computers' payoff In the RH, as in the RC, individual-level
functions (2)]. Therefore, there is no strategic irrationalityother than money illusion can play
uncertaintyand, hence, no need to form expec- a role. However, in the RH the adjustmentto the
tations aboutthe behaviorof the otherplayersin new post-shock equilibriumis not just an indi-
this condition. Moreover, since the computers vidual optimizationproblemfor the humansub-
play best replies, their behavior rules out any jects. In the RH, adjustment to the new
money illusion or any other form of irrational- equilibriumalso involves the solution of a com-
ity. In contrast, in the condition with human plex coordinationproblem.3It cannot be taken
opponents each subject faces the task of form- for granted that subjects instantaneously suc-
ing expectations about the other players' price ceed to act according to the new post-shock
choices. This necessarily also involves a guess equilibrium.A plausible reason for this is that
about the extent to which other players are the complexity of subjects' task is greaterin the
affected by money illusion. RH comparedto the RC.
The conditionswith computerizedplayers es- The RH and the RC are used to examine
sentially boil down to individual decision- to what extent individual-level irrationalities,
making experiments in which human subjects other than money illusion, together with the
can maximize their money earningsby playing coordinationproblem contributeto nominal in-
optimally against the known aggregate best re- ertia. The difference in price adjustment be-
ply of the n - 1 computerizedplayers. Note tween the RH and the RC measures the impact
that in the computerizedconditions the indirect of the coordinationproblem plus the impact of
effects of money illusion, which operatevia the the interactionbetween the coordinationprob-
expectations that other players are affected by lem and the individualirrationalitiesthat are not
money illusion, can play no role because the related to money illusion. Interaction effects
computers play best reply. These conditions, occur when these individualirrationalitiescause
therefore, allow us to examine to what extent slow adjustment by some subjects after the
money illusion has direct effects on nominal shock which- due to strategic complementar-
inertia, i.e., to what extent it simply causes ity-induces the other subjectsto adjustslowly,
individual optimizationmistakes. In the condi- too. A particularlyinterestingcase arises if we
tions with humanopponentsthe indirecteffects find no nominal inertia in the RC while in the
of money illusion can, in addition, also play a RH nominal inertia prevails. In this case all of
role. the nominal inertia in the RH can be attributed
The experimentaldesign in Table 1 allows to to the coordinationproblem because individual
isolate various potentially important determi-
nants of nominal inertia. In the RC, money
illusion is ruledout at the individualand, hence, 3There is an importantliteratureon coordinationprob-
also the aggregate level. Therefore, if we ob- lems in macroeconomic models (see, e.g., Cooper, 1999)
serve in the RC a slow adjustmentof the nom- that is based on the existence of multipleequilibria.We use
the term "coordinationproblem"in a differentway because,
inal price chosen by the humansubject afterthe even in the case of a unique equilibrium, subjects face a
shock, money illusion cannot be the source of coordinationproblem: Nash-equilibriumplay presupposes
this nominal inertia. Thus, with the help of the that subjects have coordinated(equilibrium)expectations.
VOL.91 NO. 5 FEHR AND TYRAN.DOES MONEYILLUSIONMATTER? 1245
irrationalities(other than money illusion) are ference in the deviations of the post-shock av-
absent. erage prices from equilibriumbetween the NC
In the NH, subjects face the same coordina- and the RC, ApNC - APRC, measures the
tion problem as in the RH. We are, however, aggregate impact of individual-level money il-
particularlyinterested in the impact of adding lusion on nominal inertia. Note that ApNC _
the nominal frameto this coordinationproblem, ApRC = pNC _ pRC because the equilibrium
i.e., in a comparisonof the NH and the RH. This price is identical across conditions. The differ-
comparisonallows us to isolate the total effects ence between the NH and the RH, APNH -
of money illusion in an environmentwhere sub- ApRH = pNH _ pRH, measures the total ef-
jects face a coordination problem. The total fects of money illusion which consist of the
effects of money illusion in this environment individual-leveleffects plus the indirecteffects.
consist of the direct effects of individual-level Thus, if there are no indirect effects of money
money illusion as exhibited in the NC plus the illusion the total effects must be equal to the
indirect "multiplier"effects of individual-level individual-leveleffects: pNH - pRH = pNC _
illusion. These "multiplier"effects may arise pRC. If there are, however, indirect effects we
because in our setting with human opponents, should observe that pNH - pRH > pNC -
Note thatthe realpayoff widoes not dependon ticular strategy is (weakly) dominated. In the
the averageprice P of all groupmembersbut on real frame a (weakly) dominatedstrategyPi has
P_i. This feature makes it particularlyeasy to (weakly)_smaller real payoff numbers at any
play a best reply for a given expectation about level of P-i. In the nominal frame a (weakly)
the other players' average price. If we made wi dominated strategy Pi has (weakly) smaller
dependenton P, so that Pi affects P, it would nominal payoff numbers at any level of P,i.
have been much more difficult for i to compute Thus, to eliminate (weakly) dominated strate-
the best reply (see also below). It is also worth- gies in either frame, subjects only need to elim-
while to point out that the nominal payoff for inate those strategiesthat have (weakly) smaller
each subject i is given by P_iri and not by (real or nominal) payoff numbersat any given
P7i. This makes the computation of the real level of P- _. Since, in the conditionwith human
payoffs from a given nominalpayoff much eas- opponents, the best-reply function and, hence,
ier because the deflatoris independentof one's the numberof (weakly) dominatedstrategiesis
own price choice. exactly the same underthe real and the nominal
Properties (i) and (iii) above were imple- frame, there is, in the absence of money illu-
mented because our analysis focuses on the sion, no reason why adjustmentshould differ
impact of money illusion on the adjustment across the RH and the NH.
process of an economy with a unique money-
neutral equilibriumP*, i = 1,. n. To see C. ExperimentalProcedures and Parameters
thatproperties(i) and (iii) imply neutrality,note
thata changein M fromMoto AMoleavesrealpay- All major experimentalparametersare sum-
offs unaffected if prices change to APi and marized in Table 2. The experiment was con-
AP_i. Moreover, if Pi, i = 1, n, is a best ducted in a computerized laboratory with a
reply to P_i at MO, AP' also is a best reply group size of n = 4. The groupcompositiondid
to AP-i at AMo. Thus, AP* for all i is the not change throughoutthe whole experiment,
post-shock equilibrium. i.e., for 2T periods. In each group there were
Property(ii) capturesstrategiccomplementa- two types of subjects: Subjects of type x and
rity and was implementedfor the reasons given subjects of type y. Payoff functions differed
in Section I, subsection B. In our pilot experi- among the types. This difference implied that
ments we initially implemented a price-setting x-types had to choose a relatively low price in
game with monopolistic competition.However, equilibriumwhile y-types had to choose a rel-
it turnedout that subjects quickly realized that atively high price (see Table 2 for details).
under monopolistic competition cooperative There is no particularreason for our choice of
gains can be achieved by out-of-equilibrium the group size because there are no strong con-
behavior.Therefore,both in the nominalas well jectures about the net effects of a different
as in the real frame, subjects systematically group size. On the one hand, a largergroup size
tried to achieve real payoff gains through out- enhances the chances that there are individuals
of-equilibriumbehavior. Only towards the end with money illusion in a group. On the other
of each phase these attemptsvanished. Thus, in hand, the relative impact of an individual on
the pre- as well as in the post-shockphase of our averageprices becomes smaller.With regardto
pilot experiments, adjustmenttowards equilib- the heterogeneity of the players' payoff func-
rium was strongly retardedby attempts to co- tions, the case of four differentpayoff functions
operate.To remove this confoundwith the other would be the most realistic but also the most
sources of nominal inertia we chose payoff complicated case. Therefore, we went for an
functions that ensured that the equilibriumwas intermediate solution with only two types of
the unique Pareto-efficientpoint in the whole players.6
payoff space [property(iv)]. In the pre-shock phase of each treatmentthe
Finally, property (v) means that there is a
method for finding the equilibriumthat works 6 The payoff functionsof the two types were the same up
exactly in the same way in the real as well as in to a parallelshift. Except for P * andP k all parametersof the
the nominal frame. Note that the framing of payoff function specified in the Appendix are the same for
payoffs has no impact at all on whether a par- both types.
VOL.91 NO. S FEHR AND TYRAN:DOES MONEYILLUSIONMATTER? 1247
TABLE2-EXPERIMENTALDESIGN
feasible integer combination of (Pi, P-i). To formation sheet that informed them about the
inform subjects about the payoffs of the other P-i response of the three computers to each
type, each subject also received the payoff ma- price choice Pi E { 1, 2, ..., 30 }. Fifty percent
trix of the other type. This informationcondi- of the human subjects in these conditions were
tion was common knowledge. The presentation endowed with the payoff function of an
of payoffs in the form of a matrix made it x-player, the other 50 percent had the payoff
particularlyeasy to find the best reply for any function of a y-player.
given P_i: The subjectjust had to look for the At the end of the final pre-shock period the
highest real or nominal payoff in the column nominal shock was implementedin the follow-
associated with P-i.9 In fact, one of the first ing way: Subjects were publicly informed that
things most subjectsdid afterwe distributedthe x- and y-types received new payoff tables.
instructionswas to mark the best replies in the These tables were based on M1 = M0/3. Again
payoff tables. each type received the payoff table for his own
After subjects had read the instructions,but and the other type. Subjects kept the pre-shock
before the startof the experiment,each subject tables and were encouragedto comparethe pre-
had to solve a series of exercises(see the Appen- and post-shock tables. They were told that, ex-
dix in Fehr and Tyran, 2000). These exercises cept for payoff tables, everything else would
involvedthe computationof theirown payoff and remain unchanged. They were given enough
the payoff of theiropponentsfor given hypothet- time to study the new payoff tables and to
ical strategyprofiles.In the nominaltreatments,in choose Pi for the firstpost-shock period.10 This
particular,subjectshad to computethe real pay- procedure ensured that in the first post-shock
offs from their nominal payoff tables for given period subjects faced an exogenous and fully
hypotheticalstrategyprofiles.The subjectsknew anticipatednegative nominal shock. At the be-
that we did not startthe experimentuntil every ginning of this period it was also common
participantin a session had solved all exercises knowledge that the experiment would last for
successfully.All subjectswerein fact ableto solve anotherT periods.
the exercises. By this training procedure we Before we proceed to the experimental re-
wantedto rule out thatsubjectsdo not know how sults, it needs to be emphasizedthat in a given
to properly deflate nominal values. It is quite phase the numberof dominatedprice choices is
likely that this procedurediminishedthe amount identical across all treatments.It is, however,
of individual-levelmoney illusion in our experi- not identical between the pre- and the post-
ment.It was motivatedby the questionwhethera shock phase. Since the money supply is lower in
small amountof individual-levelmoney illusion the post-shock phase the numberof dominated
will cause long-lasting nominal inertia in the NH strategiesis also lower in this phase. Note that
becauseof the indirecteffects of money illusion. the smallernumberof dominatedstrategiesin the
Obviously,the case for the relevanceof money post-shock phase is an inevitable result of the
illusion is strongerif we observe large indirect fact that the money supply is reducedwhile the
effects. nominal strategy space and the nominal ac-
In the treatmentswith computerized oppo- counting unit is kept constant." Due to the
nents, subjects received the same instructions
and payoff tables as in the treatmentswith hu-
man opponents. In addition, subjects were in- 10
Subjects were told that they had ten minutes to study
formed that the decisions of the other three the new payoff tables and, in addition, three minutes to
players in the group would be made by prepro- make a decision for the first post-shock period. Yet, almost
all subjects made their decision well before the 13 minutes
grammed computers. Subjects received an in- had elapsed. In the subsequentperiods subjects also rarely
exhausted their time limits.
l l A change in the nominal price in the post-shock phase
9 If a subjectis uncertainaboutthe truevalue of P_ i, the (i.e., at M013) by one unit has the same real effects as a
calculationof the best reply requires,of course, to take into change in the nominal price by three units in the pre-shock
account the subjective distribution of P_i and not only phase (i.e., at MO). This means that if a nominal price is
the expectation of P-,. However, for simplicity, in the strictly dominated in the post-shock phase there will, in
following we will use the term "best reply"in the sense of general, be three nominal prices that are strictly dominated
a best reply to the expectation of P_i. in the pre-shock phase.
VOL.91 NO. S FEHR AND TYRAN:DOES MONEYILLUSIONMA1TER? 1249
TABLE3-EvoLUrIoN OFPRICES
ANDEFFCIENCY
LOSSES
OVER
TIME
the RC there are no or only extremelysmall real the equilibriumis efficient it is also a measure
incomes losses in the post-shockphase. To ver- of the efficiency loss. Columns 5 and 6 of Table
ify this claim we calculate by how much actual 3 present the evolution of the average value of
real income of player i, 7ri, falls short of real s t over all playersin the RC and in the NC. The
income in equilibrium7*. For this purpose we two columns indicate that after the shock the
have computed sit, (= * - 7rit)/7r* for all average efficiency loss is most of the time zero
players in each period t. si, is a measure of the in the RC and always lower than in the NC.
income loss relativeto the equilibriumpayoff as Taken together, the results of the treatments
a percentage of the equilibrium payoff. Since with computerizedopponentsindicate thatthere
VOL.91 NO. S FEHR AND TYRAN:DOES MONEYILLUSIONMATTER? 1251
20
Nominalwithhumanopponents
- - -Nominal withcomputerized
opponents
- [/ with
-Real/\\g, humanopponents
18
--Real withcomputerizedopponents
16
14
10
4
-20 -18 -16 -14 -12 -10 -8 -6 -4 -2 1 3 5 7 9 11 13 15 17 19
period
FIGURE1. EVOLUTION
OF AVERAGEPRICES
is a small amount of money illusion at the shows thatin the firstpost-shockperiod average
individual level but beyond that there is no prices in the RH are 3.1 units above the average
individual irrationality.The small amount of equilibrium price of P1 = 6. This deviation
individual-level money illusion is suggested by quickly decreases to 1.4 units in period 3 and
the small price differencesbetween the NC and after period 4 the deviation is never largerthan
the RC after the shock. The absence of other one unit. This patternof averagebehavioris not
forms of individualirrationalityis suggested by an artifactof aggregationbut is also revealed at
the perfect adjustment to the shock and the the level of individual choices. In the final pre-
generally high incidence of equilibriumplay in shock period 93 percent of the subjects in the
the RC. RH play exactly their equilibriumstrategies.In
the firstpost-shockperiodonly 35 percentof the
B. Nominal Price Adjustmentas a subjects play the new equilibriumand 23 per-
CoordinationProblem cent of the subjects are only one or two price
units above the equilibrium.The other 42 per-
The fact that in the RC the adjustmentto the cent are more than two units above the equilib-
post-shock equilibriumis perfect makes the in- rium. Yet, after only three periods the
terpretationof the deviation of prices from the distribution of individual price choices has
post-shock equilibrium in the RH particularly moved much closer to the equilibrium.In period
easy. It means thatthe whole deviationis due to 4, 45 percent of all subjects play exactly the
the fact that subjectsin the RH face a relatively equilibrium, 48 percent are one or two units
complex coordinationproblem.The majorfacts above and only 7 percentaremore thantwo units
about price adjustmentin the RH are displayed above the equilibrium.This post-shock evolu-
in Table 3 and Figure 1. Column 3 of Table 3 tion of prices indicates that the coordination
1252 THEAMERICANECONOMICREVIEW DECEMBER2001
problem initially causes considerable nominal TABLE 4-DEVIATION FROM POST-SHOCK EQUILIBRIUM IN
TREATMENTS WITH HUMAN OPPONENTS
inertiabut that after a few periods this effect is
rather small because prices are again close to Real Treatmentwith Nominal Treatment
the equilibrium. Post-Shock Human Opponents with Human Opponents
Our description of the pattern of nominal Period (RH) Coefficient a, (NH) Coefficient 3
inertia in the RH is also supportedby formal 1 3.10*** 7.14"**
statistical tests. To check how long average 2 1.68*" 6.86***
group prices in the RH and the NH deviate 3 1.43 5.43***
significantly from the equilibrium we ran the 4 0.90 4.41***
following regression for the post-shock phase: 5 1.00 3.86***
6 0.55 4.18***
7 0.25 3.77***
19 20 8 0.35 3.05***
9 0.25 2.70**
(3) Pit - P*= atdt + E ot(l- dt) 10 0.83 2.59***
t=1 t=l
11 0.13 2.05***
12 0.23 1.61**
where Pit denotes the average price of group i 13 0.18 1.18
in period t. dt 1 if the price observationin 14 0.18 0.89
period t comes from the RH. The coefficients at 15 0.10 0.70
measure the deviation from equilibriumin the 16 0.13 1.25
17 0.03 0.80
RH while the coefficients Pt measure the devi- 18 0.13 1.20
ation in the NH.13The results of regression (3) 19 0.05 1.45
are summarized in Table 4. The table shows 20 0.95
that, at the 5-percentlevel, averageprices in the
Notes: Pit- = a d?tdt+ E- -d=). ,(1
RH deviate significantly from the equilibrium
dt = I if price observation in period t is from RH.
for two periods.Yet, from period 3 onwards,the ***Denotes significance at the 1-percentlevel.
hypothesis that average prices are in equilib- **Denotes significance at the 5-percent level.
rium can no longer be rejected.14
To what extent is nominal inertia in the RH
associated with real income losses? Column 7 perceive coordinationas a difficult problemthis
of Table 3 indicates that in the first post-shock should be reflected in subjects' confidence in
period the real income loss resulting from dis- pe . In the first few pre-shock periods, sub-
adjustmentdifferencebetween the NC and the put differently,to what extent did subjectsplay
RC, AP'c - ARC = pNc pRc. It is, for a best reply to their expectations.The vast ma-
example,easy to infer from Table 3 that, in the jority of subjects in both treatments indeed
second post-shockperiod, the adjustmentdiffer- played best replies to pei During the first ten
ence betweenthe NH andthe RH is 12.9 - 7.7 = post-shock periods 84 percent of the subjectsin
5.2 price units, while the differencebetween the the RH choose exactly the payoff-maximizing
NC and the RC is only 7.4 - 7.0 = 0.4 units. price in response to pe i and the rest of the
Hence, in this period the impact of the nominal subjects chooses prices that were close to the
frameis 13 times largerin the settingwith human best reply. In the NH there are slightly fewer
playerscomparedto the settingwithcomputerized subjects (80 percent) who chose exact best re-
players.In the tenthpost-shockperiodthe adjust- plies during the first ten post-shock periods.
mentdifferenceis still 1.8 unitsin the settingwith Yet, as in the RH, the deviations from the exact
humanplayers and only 0.6 units in the setting best reply were in general very small. The fact
with computerizedplayers. To substantiatethe that most subjects responded to pe i with a
indirecteffects of money illusion we also con- payoff-maximizing price choice suggests that
ductedperiod-wiset-testsfor the null hypothesis the greaterstickiness of the expectations in the
that the adjustmentdifferencebetween the NH NH also caused a greater stickiness of actual
andthe RH is biggerthanbetweenthe NC andthe prices in the NH.
RC. In five of the firstten post-shockperiodsthe
differencebetweenthe NH and the RH is signif- IV. Nominal Inertia after
icantlylargerat the 5-percentlevel. In view of the a Positive Money Shock
considerablevarianceacross the four conditions
this is quiteremarkable.15 Thus, the implementa- A. The Relevance of a Positive Money Shock
tion of the nominalframehas a much largerim-
pact in the settingwhere money illusion can also Our results so far indicate that the direct
have indirecteffects. effects of individual-level money illusion are
If money illusion has indirect effects we relatively small. The introductionof the nomi-
should also observe that expectations are stick- nal framein the setting with computerizedplay-
ier in the NH compared to the RH. Figure ers leads only to a small increase in nominal
2 shows the evolution of the average price ex- inertia. Nominal inertia is much more pro-
pectations over time in both treatments. The nounced, however, when money illusion can
figure shows that in the last few pre-shock pe- also affect players' expectations and can, thus,
riods, expectations are in equilibrium in both also have indirect effects. In the NH, subjects'
treatments.In the post-shock phase there are, expectations are much stickier and, as a conse-
however, striking differences. While expecta- quence, prices are much stickier.This raises the
tions are very sticky in the NH they are far less question of why expectations are so sticky in
sticky in the RH. To provide statisticalevidence the NH compared to the RH. We believe that
for this, we ran the same regression as in Table the answer to this question can be found in the
4 with expectationsdata. It turnsout that, in the existence of subjects who take nominal payoffs
NH, expectations differ significantly from the as a proxy for real payoffs. Subjects who apply
equilibrium(at p < 0.05) for 13 periods while this rule of thumb mistakenlybelieve that if all
in the RH they differ only for three periods. players choose relatively high prices, all will
Thus, there can be little doubt that the nominal reaphigh real payoffs because they all reaphigh
framecauses a large increasein the stickiness of nominal payoffs. They mistakenly believe that
price expectations.The next question then is, to there are real gains from jointly setting high
what extent this difference in expectations prices. Such subjects will, therefore, be reluc-
causes differencesin subjects' price choices. Or tantto cut theirnominalprices afterthe negative
money shock in the NH. Moreover, if the pres-
15
ence of subjects who are reluctantto cut prices
There are two subjectsin the NC condition who could
well be classified as outliers. If we run the tests without is anticipatedby other subjects, others will be
these two subjects the indirecteffects are highly significant induced to cut their price insufficiently also.
in the first nine post-shock periods. It is importantto note that the above rule of
VOL.91 NO. 5 FEHR AND TYRAN:DOES MONEYILLUSIONMATTER? 1255
20 _
- Nominal with human opponents (NH)
8-
Real with human opponents (RH)
16-
14-
10
4
-20 -18 -16 -14 -12 -10 -8 -6 -4 -2 1 3 5 7 9 11 13 15 17 19
period
FIGURE2. EVOLUTION
OF AVERAGEEXPECTATIONS
thumb cannot become effective in the RH. In only means that money illusion has a different
the RH the numbersin the payoff tables repre- impact after a positive shock compared to a
sent real payoffs which makes it completely negative shock. After a negative shock the rule
transparentthat at high nominalprices real pay- of thumb mentioned above causes a reluctance
offs are not generally higher. This means that to adjustprices (downwards)while after a pos-
the presence of subjectswho take nominal pay- itive shock it does not cause a reluctance to
offs as a proxy for real payoffs causes no reluc- adjustprices (upwards).Note furtherthat while
tance to cut nominal prices after the negative the rule of thumb implies a quicker adjustment
shock in the RH. These differencesbetween the to equilibriumafter a positive shock in the NH,
NH and the RH in the reluctanceto cut nominal the adjustmentspeed in the RH should not differ
prices also provide a rationale for the much across positive and negative shocks. The reason
stickier price expectationsin the NH. is that the rule of thumb cannot become opera-
Yet, if the above explanationfor the stickier tive in the RH.
expectations in the NH is correct, we should To test these implications of our explanation
also observe that after a positive money shock, for the much stickier expectationsin the NH we
prices and expectations adjust more quickly to conducted additional experiments with a posi-
the equilibriumthanaftera negative shock. This tive money shock. Forty-eight subjects (12
is so because after a positive shock, adjustment groups) participatedin the RH and another48
towards equilibriummeans adjustmenttowards subjects(12 groups)participatedin the NH with
higher prices and, hence, higher nominal pay- the positive money shock. The easiest way to
offs. A quicker adjustment after a positive implement a positive shock would be a reversal
shock, however, does not mean that money il- in the sequence of the money supply in our
lusion is absentwhen a positive shock occurs. It previous design. Unfortunately,this approachis
1256 THEAMERICANECONOMICREVIEW DECEMBER2001
- - Differenceof pricedeviationsNH
5- v
- - -Differenceof expectationsdeviationsNH
a
&<.8 " ----Difference of pricedeviations RH
Differenceof expectationsdeviationsRH
1
3
4 -'
-1 .. . .
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Post-shock period
FIGURE 3. DIFFERENCES IN DEVIATIONS FROM EQUILIBRIUM ACROSS THE NEGATIVE AND THE POSITIVE SHOCK
jority of individuals plays exactly the equilib- than after the positive shock. It shows that the
rium in the final pre-shock period. Then, in deviation from equilibrium is substantially
the first post-shock period prices make a big larger after the negative shock. Between pe-
jump upwards to 20.5 and already in period 4 riod 2 and 7, e.g., the adjustment gap is four
after the shock, average prices are almost or more units bigger after the negative shock.
exactly at the new equilibrium of P* = 25. Even in period 10 the adjustment gap is still
From that period onwards prices remain very almost three units bigger.
close to the equilibrium. This contrasts The impressionconveyed by Figure 3 is con-
sharply with the adjustment process after the firmed by a more formal statistical analysis. If
negative shock where, throughout the whole we performregression(3) with the dataafterthe
post-shock period, average prices never came positive shock, it turns out that in the NH the
so close to the equilibrium. This difference in
NH adjustment paths after the negative and 16
the positive shock is depicted in Figure 3. The Let P*+ be the post-shock equilibriumprice in the
positive-shock design and let P+ be the actual post-shock
heavy line in Figure 3 shows the difference in averageprice. Analogously, let P_ be the actualpost-shock
the post-shock deviations of average prices average price in the negative-shock design and denote the
from the equilibrium between the positive and post-shock equilibriumprice in this design by P*-. Then
the negative shock.'6 The graph reveals to the heavy line in Figure 3 is given by (P - P_) -
(P + - P +), which measuresthe differencein the average
what extent in the NH the adjustment gap, deviations from equilibrium across the positive and the
i.e., the deviation of average prices from the negative shock for the first 15 periods of the post-shock
equilibrium, is larger after the negative shock phase in the NH.
1258 THEAMERICANECONOMICREVIEW DECEMBER2001
hypothesis of equilibriumplay can only be re- average price is again very close to the equi-
jected for the firstthreeperiods(at the 5-percent librium. This indicates that price adjustment
level). Rememberthat after the negative shock, after the positive shock is rather quick in the
group prices were significantly above the equi- RH-similar to the pattern after the negative
librium for 12 periods. Thus, the evidence un- shock. This similarity is also displayed in
ambiguously indicates that adjustment in the Figure 3 and by formal statistical analysis.
NH is much quicker after the positive shock, The thin line in Figure 3 is constructed anal-
which is consistent with our hypothesis that ogously to the heavy line except that we use
there is less reluctanceagainst adjustmentafter the price data from the RH. It shows that price
the positive shock. adjustment in the RH is only slightly faster
If there is indeed less reluctance against ad- after the positive shock. If we perform regres-
justment after the positive shock, at least some sion (3) with the post-shock data from the
subjects should anticipate this. Therefore we positive-shock design we get the following
should also observe that expectations are less results: The hypothesis that average prices in
sticky after the positive shock. The dashed the RH are in equilibrium can only be rejected
heavy line in Figure 3 shows the difference in for the first two periods (at the 5-percent
the post-shock deviations of the average expec- level). Note that this is exactly the same num-
tationsof P_ i from the equilibriumbetween the ber of out-of-equilibrium periods as after the
positive and the negative shock. This graph is negative shock. This suggests that the differ-
constructed analogously to the heavy line in ences in the price adjustmentacross shocks in
Figure 3 except that we used the expectations the RH are indeed negligible. The dashed thin
pe to constructit. Thus the dashed heavy line line in Figure 3, which is constructed analo-
shows to what extent the adjustmentgap in the gously to the dashed heavy line except that
expectations, i.e., the deviation of average ex- we use the expectations data from the RH,
pectations from equilibrium,is larger after the indicates that we can basically make a similar
negative shock than after the positive one. The conclusion with regard to the differences in
graph indicates that the adjustmentgap in the the adjustment of expectations across shocks.
expectations is much larger after the negative While in the NH there are large differences in
shock for many time periods. Interestingly,the the stickiness of expectations across shocks,
graph is hump-shaped,i.e., the relative sticki- in the RH the differences in expectations are
ness of expectations after the negative shock rather small.
increases in the first few periods. This is due to Thus all major regularities are consistent
the fact that between period 2 and 5 after the with our hypothesis that there are beliefs that
positive shock, expectationsrapidlyconverge to some subjects take nominal payoffs as a
equilibriumwhile they are very sticky after the proxy for real payoffs. Nonetheless, it would
negative shock. be reassuring if subjects themselves ex-
Finally, since the rule of thumb of taking pressed such a belief. To check to what extent
nominal payoffs as a proxy for real payoffs subjects indeed believed this they could indi-
cannot be operative in the RH, we should cate their degree of agreement with the fol-
observe no differences in price adjustment in lowing statement after the experiment: "I
the RH across negative and positive shocks. believed that the other subjects would inter-
Table 5 shows the evolution of average prices pret high nominal payoffs as an indicator for
in the RH after the positive shock and Figure high real payoffs." Participantscould indicate
3 illustrates the differences in average prices whether they weakly (dis)agreed, whether
and average expectations across shocks. Ta- they strongly (dis)agreed or whether they to-
ble 5 indicates that in the pre-shock phase of tally (dis)agreed with this statement. Thirty
the RH the average price is very close to the percent of the subjects in the NH agreed ei-
equilibrium PO = 12.5 already after three ther "strongly" or "totally" and further 25
periods. Immediately after the positive shock percent indicated a weak agreement. In our
there is a big upward jump in prices to 22.5, view, this can be taken as direct evidence that
only 2.5 units below the new equilibrium. a majority of the subjects believed that other
Already in the third post-shock period the subjects were affected by money illusion. In
VOL.91 NO. 5 FEHR AND TYRAN:DOES MONEYILLUSIONMA7IER? 1259
any case, these answers nicely fit with our there is field evidence indicating that positive
explanationfor the largeamountof nominaliner- and negative money shocks have asymmetric
tia observedin the NH after the negative shock. effects. While negative shocks have an output-
reducing effect, positive shocks do not seem to
V. Summary and Concluding Remarks affect output (James Peery Cover, 1992; J.
BradfordDeLong and Lawrence H. Summers,
Most economic transactionsare represented 1988). The asymmetric effects of money illu-
in nominalterms.Therefore,it seems likely that sion on price sluggishness can be consideredas
people often perceive and think abouteconomic a potential microfoundationfor this result.
problems in nominal terms which may induce Finally, anotherinterestingresult of our ex-
money illusion. However, for several decades periments is that we isolate-in addition to
money illusion has been considered as largely money illusion-a further source of nominal
irrelevantfor the nominal inertia of aggregate inertia. This source is related to the fact that in
price levels. Instead, most economists have fo- a strategicsituationsubjects do not merely face
cused on informationalfrictions, costs of price an individual optimization problem but that
adjustment,and staggeredcontracts.This paper they also have to predictother agents' behavior.
shows, however, that even in the absence of After any shock, the new equilibriumcan only
these factors a fully anticipatednegative nomi- be achieved if subjects have coordinated(equi-
nal shock can cause long-lastingnominalinertia librium) expectations. The comparison of ad-
that is associated with large real income losses justment paths in the real treatments with
during the adjustmentphase. Our results indi- computerizedand with humanopponentsshows
cate that a large partof this nominal inertiacan that after a fully anticipatednominal shock, it
be attributedto the direct and indirecteffects of cannot be taken for grantedthat subjectsinstan-
money illusion. The experimentsin the setting taneously succeed in solving this coordination
with computerizedopponents show that the di- problem. They will, in general, go through a
rect effects of money illusion in the form of period of disequilibriumthat is associated with
individual optimization mistakes are not very nominal inertia.Note, however, that the coordi-
frequent:The introductionof the nominalframe nation problem alone causes substantiallyless
in the setting with computerized opponents nominalinertiathanmoney illusion. It also does
causes only a small amount of nominal inertia. not cause asymmetriceffects: In the real treat-
However, the combined direct and indirect ef- ment with humanopponentsthe extent of nom-
fects of money illusion generate a very large inal inertiais very similar after the positive and
increase in nominal inertia.This is indicatedby the negative nominal shock.
the fact that the introduction of the nominal These results show that experimentscan be a
frame in the setting with human opponents useful tool for the examinationof the nature,the
causes a huge increase in the sluggishness of extent, and the impact of money illusion in
prices. Instead of two it takes 12 periods until strategiceconomic interactions.There are many
averageprices reach the post-shock equilibrium otherquestionsthatcould be usefully tackledby
in this setting. experimental methods. One open question is
The majorcause for nominal inertia after the how subjects who are familiar with money
negative shock is that subjects' expectationsare shocks in our pricing game will respondto new
very sticky. In our view this stickiness of price shocks. Future research should thus examine
expectations is related to the nature of money how the adjustmentpatternvaries when expe-
illusion in our experiment,i.e., to the belief that rienced subjects face a series of different
there are subjects who take nominal payoffs as shocks. Otherinterestingquestions concern the
a proxy for real payoffs. This conjecture is nature of nominal wage rigidity and the inter-
supportedby direct questionnaireevidence and action of price and wage policies duringadjust-
by the resultsof furtherexperimentswith a fully ment. To answer these questions it is necessary
anticipatedpositive nominal shock. It turns out to introduceworkers as separateplayers in the
that price sluggishness is much smaller after a game. The analysis of the interactionbetween
positive nominal shock than after the negative wage and price setting may prove particularly
shock. This result is also interesting insofar as useful because there are likely to be strong
1260 THEAMERICANECONOMICREVIEW DECEMBER2001
[1 + a -
V 1 +bA2
Tik 1 _( Xnf
p p)1 2-
I + c ( d/ A/r- - + e arctan(f-A)
P-ik iS the actual average price of the other the parametersa, b, c, d, e, f, and V were the
n - 1 players from the viewpoint of player i same. They were given by a = 0.5, b = 0.6,
who is of type k. P* is the equilibriumaverage c = 27, d = 1, e = 0.05,f = 20 and V = 40.
price of the other n - 1 players from the
viewpoint of a playerof type k. Pik is the actual REFERENCES
price of i who is of type k. P* is the equilib-
rium price of a player of type k. Finally, Abbott, Michael and Ashenfelter, Orley. "Labour
/ (P_ ikIM)- (P*/M). The parametersfor Supply, CommodityDemand and the Alloca-
M and all equilibriumvalues can be found in tion of Time." Review of Economic Studies,
Table 2. October 1976, 43(3), pp. 389-411.
In all periods and all experimental sessions Akerlof, George A.; Dickens, Williams T. and
Perry, George L. "The Macroeconomics of
Low Inflation."Brookings Papers on Eco-
17 In his speech on September 8, 1858, Lincoln said: nomic Activity, 1996, (1), pp. 1-59.
"You can fool all the people some of the time, and some of
the people all the time, but you cannotfool all the people all Akerlof, George A. and Yellen, Janet L. "A Near-
the time." rational Model of the Business Cycle, with
VOL.91 NO. 5 FEHR AND TYRAN:DOES MONEYILLUSIONMAITER? 1261
eds., Money. New York: Norton, 1989, pp. Romer, Christina D. and Romer, David H. "Does
244-47. MonetaryPolicy Matter?A New Test in the
Kahn, Shulamit. "Evidence of Nominal Wage Spirit of Friedmanand Schwartz,"in Olivier
Stickiness from Microdata."American Eco- Jean Blanchard and Stanley Fischer, eds.,
nomic Review, December 1997, 87(5), pp. NBER macroeconomicsannual: 1989. Cam-
993-1008. bridge, MA: MIT Press, 1989, pp. 121-70.
Leontief, Wassily. "The FundamentalAssump- . "MonetaryPolicy Matters."Journal of
tions of Mr. Keynes' Monetary Theory of Monetary Economics, August 1994, 34(1),
Unemployment."QuarterlyJournal of Eco- pp. 75-88.
nomics, November 1936, 5(4), pp. 192-97. Russell, Thomas and Thaler, Richard. "The Rel-
Lucas, Robert E., Jr. "Expectations and the Neu- evance of Quasi Rationality in Competitive
tralityof Money." Journal of Economic The- Markets."American Economic Review, De-
ory, April 1972, 4(2), pp. 103-24. cember 1985, 75(5), pp. 1071-82.
Mankiw, N. Gregory. "Small Menu Costs and Selten, Reinhard and Berg, Claus C. "Drei Experi-
Large Business Cycles: A Macroeconomic mentelle Oligopolspielserienmit Kontinuier-
Model." Quarterly Journal of Economics, lichem Zeitablauf,"in Heinz Sauermann,ed.,
May 1985, 10(2), pp. 529-38. Beitragezurexperimentellen Wirtschaftsforsch-
Niemi, Beth T. and Lloyd, CynthiaB. "Female ung, Vol. II. Tuibingen:1970, pp. 162-221.
Labor Supply in the Context of Inflation." Shafir, Eldar; Diamond, Peter and Tversky,
American Economic Review, May 1981 (Pa- Amos. "Money Illusion." QuarterlyJournal
pers and Proceedings), 71(2), pp. 70-75. of Economics, May 1997, 112(2), pp. 341-
Oh, Seonghwan and Waldman, Michael. "The 74.
MacroeconomicEffects of False Announce- Taylor, John B. "StaggeredWage Setting in a
ments." Quarterly Journal of Economics, Macro Model."AmericanEconomic Review,
November 1990, 105(4), pp. 1017-34. May 1979 (Papers and Proceedings), 69(2),
. "Strategic Complementarity Slows pp. 108-13.
Macroeconomic Adjustment to Temporary Tobin, James. "Inflationand Unemployment."
Shocks." Economic Inquiry, April 1994, American Economic Review, March 1972,
32(2), pp. 318-29. 62(1), pp. 1-18.
Patinkin,Don. "The Indeterminacyof Absolute Tversky, Amos and Kahneman, Daniel. "The
Prices in Classical Economic Theory." Framingof Decisions and the Psychology of
Econometrica, January 1949, 17(1), pp. Choice." Science, January 1981, 211, pp.
1-27. 453-58.