Professional Documents
Culture Documents
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp
.
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
Think about Richard Scarry's Cars and pathologies that may develop underthese more
Trucksand Things That Go.1 Think about what realistic conditions.
that book would have looked like in sequential For me, the study of asymmetricinformation
decades of the last century had RichardScarry was a very first step towardthe realizationof a
been alive in each of them to delight and amuse dream. That dream was the development of a
children and parents. Each subsequent decade behavioral macroeconomics in the original
has seen the developmentof ever more special- spiritof JohnMaynardKeynes' General Theory
ized vehicles. We started with the Model T (1936). Macroeconomicswould then no longer
Ford. We now have more models of backhoe suffer from the "adhockery"of the neoclassical
loaders than even the most precocious four- synthesis, which had overriddenthe emphasisin
year-old can identify. The General Theoryon the role of psychologi-
What relevance does this have for econom- cal and sociological factors, such as cognitive
ics? In the late 1960's there was a shift in the bias, reciprocity, fairness, herding, and social
job description of economic theorists. Prior to status. My dream was to strengthenmacroeco-
that time microeconomic theory was mainly nomic theory by incorporating assumptions
concerned with analyzing the purely competi- honed to the observation of such behavior. A
tive, general-equilibrium model based upon team of people has participatedin the realiza-
profit maximization by firms and utility maxi- tion of this dream. Kurt Vonnegut would call
mization by consumers. The macroeconomics this team a kerass, "a group of people who are
of the day, the so-called neoclassical synthesis, unknowingly working together toward some
appendeda fixed money wage to such a general- common goal fostered by a largercosmic influ-
equilibriumsystem. "Sticky money wages" ex- ence."2 In this lecture I shall describe some of
plained departuresfrom full employment and the behavioralmodels developed by this kerass
business-cycle fluctuations. Since that time, to provide plausible explanationsfor macroeco-
both micro- and macroeconomics have devel- nomic phenomenawhich are centralto Keynes-
oped a Scarry-ful book of models designed to ian economics.
incorporateinto economic theory a whole vari- For the sake of background,let me take you
ety of realistic behaviors. For example, "The back a bit in time to review some history of
Market for 'Lemons'" explored how markets macroeconomicthought. In the late 1960's the
with asymmetric information operate. Buyers New Classical economists saw the same weak-
and sellers commonly possess different, not nesses in the microfoundationsof macroeco-
identical, information.My paper examined the nomics that have motivated me. They hated its
lack of rigor.And they sackedit. They then held
a celebratory bonfire, with an article entitled
t This articleis a revised version of the lectureGeorge A. "AfterKeynesian Macroeconomics."3The new
Akerlof delivered in Stockholm, Sweden, on December 8, version of macroeconomicsthat they produced
2001, when he received the Bank of Sweden Prize in Eco- became standardin the 1970's. Following its
nomic Sciences in Memory of Alfred Nobel. The article is
copyright ? The Nobel Foundation2001 and is published
neoclassical synthesis predecessor,New Classi-
here with the permission of the Nobel Foundation. cal macroeconomicswas based on the compet-
* Department of Economics, University of California, itive, general-equilibriummodel. But it differed
Berkeley, CA 94720-3880. I thankJanetYellen for extraor- in being much more zealous in insisting that all
dinarily helpful discussions and editorial assistance. I also decisions-consumption and labor supply by
thank Henry Aaron, William Dickens, Ernst Fehr, William
Gale, and Robert Shiller for invaluable comments and the
Canadian Institute for Advanced Research for generous
financial support. 2 See (http://www.gibbsonline.com/gibbsbooks.html).
1 See Scarry (1974). 3 See Robert E. Lucas, Jr. and Thomas
Sargent (1979).
411
In what follows I shall describe how behav- The failure of credit marketsis one of the major
ioral macroeconomists, incorporatingrealistic reasons for underdevelopment. Even where
assumptionsgroundedin psychological and so- mechanisms such as reputationand repeat sales
ciological observation, have produced models arise to overcome the problem of asymmetric
that comfortablyaccount for each of these mac- information, such institutions become a major
roeconomic phenomena.In the spiritof Keynes' determinantof marketstructure.
General Theory, behavioral macroeconomists To understandthe origins of the economics of
are rebuilding the microfoundationsthat were asymmetricinformationin markets,it is useful
sacked by the New Classical economics. I shall to reflect on the more general intellectualrevo-
begin my review by describing one of my ear- lution thatwas occurringat the time. Priorto the
liest attempts in this field, which led to the early 1960's, economic theorists rarely con-
discovery of the role of asymmetricinformation structed models customized to capture unique
in markets. institutions or specific market characteristics.
Edward Chamberlin's monopolistic competi-
I. Asymmetric Information tion and Joan Robinson's equivalent8 were
taught in graduateand even a few undergradu-
I first came upon the problemsresultingfrom ate courses. However, such "specific" models
asymmetric information in an early investiga- were the rareexception;they were presentednot
tion of a leading cause for fluctuationsin output as central sights, but instead as excursions into
and employment-large variations in the sales the countryside, for the adventurousor those
of new cars.6 I thought that illiquidity, due to with an extra day to spare.9 During the early
the fact that sellers of used cars know more than 1960's, however, "special" models began to
the buyers of used cars, might explain the high proliferateas growth theorists,working slightly
volatility of automobilepurchases.7In trying to outside the norms of standardprice-theoretic
make such a macroeconomic model, I got di- economics, began to constructmodels with spe-
verted.I discoveredthatthe informationalprob- cialized technological features:putty-clay, vin-
lems that exist in the used car market were tage capital, and learning by doing. The
potentially present to some degree in all mar- incorporationinto models of such specialized
kets. In some markets,asymmetricinformation technologies violated no established price-
is fairly easily soluble by repeat sale and by theoretic norm, but it sowed the seed for the
reputation.In other markets, such as insurance revolutionthatwas to come. Duringthe summer
markets,credit markets, and the marketfor la- of 1969, I first heard the word model used as a
bor, asymmetric information between buyers verb, and not just as a noun.10It is no coinci-
and sellers is not easily soluble and results in dence thatjust a few months earlier "The Mar-
serious market breakdowns. For example, the ket for 'Lemons'" had been accepted for
elderly have a hard time getting health insur- publication.1 The "modeling" of asymmetric
ance; small businesses are likely to be credit- information in markets was to price theory
rationed;and minoritiesare likely to experience what the "modeling"of putty-clay,vintage cap-
statistical discrimination in the labor market ital, and learning by doing had been to growth
because people are lumped together into cate-
gories of those with similar observable traits.
8 See Robinson
(1942) and Chamberlin(1962).
9 For example, I could well imagine a graduatestudent
being unawareof Harold Hotelling's (1929) model of spa-
also looked at the implicationsof S-s policy for the relation tial competition. I cannot rememberit in the graduatecur-
between the shifts in the ideal price and the actual price riculum and remember finding it tucked away as an
being charged.See Akerlof (1973, 1979) for analysis of the appendix to Chamberlin'sMonopolistic Competition.
effects of target-thresholdmonitoring on the short-runin- 10Conversationwith Michael Rothschild in Cambridge,
come and interest elasticity of the demand for money. Massachusetts,summerof 1969. I rememberthe usage just
6 See Akerlof (1970). as many people today may remember the first time they
7 Frederic S. Mishkin
(1976) later developed the ideas heard someone say they would "grow the economy."
that set me on this course initially. He showed why the 1 I do not have the exact date of the acceptanceof this
demand for automobiles is more volatile because cars are article,but I rememberthat it took slightly more than a year
illiquid due to asymmetricinformation. between acceptance and publication.
theory.12It was the first application of a new may be due to negative supply shocks, which
economic orientationin which models are con- cause workersto withdrawfrom the labor force
structed with careful attention to realistic mi- and eschew the jobs which are available. Any
croeconomic detail. This development has account of the business cycle based on volun-
broughteconomic theorymuch closer to the fine tary variationsin job-taking faces a significant
grain of economic reality. Almost inevitably, empirical difficulty-to explain why quits de-
the analysisof informationasymmetrieswas the cline in cyclical downturns.If higherunemploy-
first fruit of this new modeling orientation. It ment resultsfrom workers'rejectionof the poor
was the ripest fruitfor picking. In the remainder returnsfrom work, quits should rise along with
of this essay I shall discuss the payoffs of this unemployment.But there are fewer quits, not
new orientationfor the new field of behavioral more, when unemploymentrises. The procyclic
macroeconomics. behavior of quits is indisputable.'4
Instead of denying the very existence of
II. Involuntary Unemployment involuntaryunemployment, behavioral macro-
economists have provided coherent explana-
I once had an economistfriendwho saidthathe tions. Efficiency wage theories, which first
could not sell his house, a complaintthatI reiter- appeared in the 1970's and 1980's, make the
atedsympatheticallyto one of his colleagues.The concept of involuntary unemployment mean-
colleaguerespondedthattherewas only one prob- ingful.15 These models posit that, for reasons
lem: the house was unreasonablypriced. At a such as morale, fairness, insider power, or
lowerpricethe housewouldsell, perhapsinstantly. asymmetricinformation,employershave strong
New Classical economics views involuntary motives to pay workersmore thanthe minimum
unemployment as a logical impossibility, like necessary to attract them.'6 Such "efficiency
my friend's inabilityto sell his house. Could not wages" are above marketclearing, so that jobs
an unemployed worker obtain a job if only she are rationed and some workers cannot obtain
were willing to reduce her reservation wage? them. These workers are involuntarily unem-
The New Classical answer is yes: unemployed ployed. In the next section I will extend this
workers are those searching for work (hence reasoning to explain why involuntary unem-
unemployed,ratherthan out of the labor force) ployment varies cyclically.
but rejectingjobs thatare availablebecause they The pervasive empirical finding of a wide
had expected better pay. The unemployed may spreadof earnings for seemingly similar work-
be unhappythat they cannot sell their labor at ers is stronglysuggestive of the near ubiquityof
the wage or salary that they would ideally like, efficiency wages. Long before the efficiency
but except for those affected by the minimum wage was a gleam in the eye of macroecono-
wage or union bargaining,they are voluntarily, mists, labor economists had documented wide
not involuntarily,unemployed.Everyonecan get a dispersionin earnings across seemingly similar
job at the market-clearing wage. In New Classical jobs and among workers with apparentlyiden-
of
theory,periods decliningemployment-business-
cycle downturns-may be caused by an unex- 14
This question was raised by James Tobin (1972). For
pected decline in aggregate demand, which some data on the countercyclical behavior of quits, see
leaves workersmistakenlyholding out for nom- Akerlof et al. (1988). Kenneth J. McLaughlin (1991) has
inal wages that exceed the new market-clearing attemptedto reconcile the procyclicality of quits with New
level.13 Alternatively, declining employment Classicaleconomicsas follows: He definesquitsas employee-
initiated separations, and layoffs as firm-induced separa-
tions. In McLaughlin'smodel a positive productivityshock
causes more workersto ask for wage increases. Since some
12
See Robert M. Solow (1959, 1962) and Kenneth J. requests are rejected, quits rise as unemploymentdeclines.
Arrow (1962). But why should firms' wage offers lag behind worker de-
13
This theory suffers from a further theoretical diffi- mands in the face of a positive productivityshock?
culty. Since aggregateunemploymentis readily observable 15 An excellent concise summary of this literature is
with a shortlag, workersshouldconditiontheirexpectations given by Janet L. Yellen (1984).
of prevailingwage distributionson the aggregateunemploy- 16 The inclusion here of insider-outsidermodels is taking
ment rate. Such conditioning would eliminate serial corre- an especially broad interpretationof the concept of effi-
lation in unemployment. ciency wages.
tical characteristics.17 Analysis of panel data ory from psychology), and adherenceto group
indicates that workers of the same quality re- norms (referencegroup theory in sociology and
ceive different wages depending upon their theory of group formation in psychology). In
place of work. Moreover, data show that work- the earliest "sociological" version of efficiency
ers who switch industriesreceive wage changes wage theory based on gift exchange, firms give
that are correlated with the respective wage workers above market-clearing wages and
differentialsbetween the industries.18Industries workersreciprocatein their commitmentto the
with higher pay (conditionalon characteristics) firm.21The payment of above-market-clearing
also have lower quit rates, suggesting that pay wages may also be motivatedby considerations
differences are not simply compensatingdiffer- of fairness: in accordancewith the psychologi-
entials due to different working conditions or cal theory of equity, workers may exert less
benefits.19It thus appears that there are "good effort insofaras their wage falls shortof what is
jobs" and "badjobs." considered fair.22Groupnorms typically deter-
The existence of good jobs and bad jobs mine the conceptions workers form about how
makes the concept of involuntary unemploy- gifts should be reciprocatedand what consti-
ment meaningful: unemployed workers are tutes a fair wage. In the laboratory,Ernst Fehr
willing to accept, but cannot obtain,jobs iden- and his coauthors have established the impor-
tical to those currently held by workers with tance both of reciprocal behavior and social
identical ability. At the same time, involuntarily norms for worker effort in experimental set-
unemployed workers may eschew the lower- tings.23My favorite version of efficiency wages
paying or lower-skilled jobs that are available. is the insider-outsidermodel, whereby insider
The definition of involuntary unemployment workers prevent the firm from hiring outsiders
implicit in efficiency wage theory accords with at a market-clearingwage lower than what the
the facts and agrees with commonly held per- insiders are currently receiving.24This theory
ceptions. A meaningful concept of involuntary implicitly assumes that insiders have the ability
unemployment constitutes an important first to sabotage the inclusion of new workersinto a
step forward in rebuilding the foundations of firm. A detailed study by Donald Roy of an
Keynesian economics. Illinois machine shop reveals the dynamics by
But why do firms pay wages above rock which this may occur: In Roy's machine shop,
bottom? In my view, psychological and socio- insiders establishedgroupnormsconcerningef-
logical explanations for efficiency wages are fort and colluded to prevent the hiring of rate-
empiricallymost convincing.20Three important bustingoutside workers.Workerswho produced
considerations are: reciprocity (gift exchange more than the level of output considered"fair"
theory from anthropology),fairness (equity the- were ostracizedby others.2 Collusion by insid-
ers against outsidersis a compelling motive for
many firms to pay wages that are above market
17 See John T. Dunlop (1957). clearing.
18
See William T. Dickens and LawrenceF. Katz (1987) An alternativeversion of efficiency wage the-
and Alan B. Krueger and Lawrence H. Summers (1988). ory, grounded in asymmetric information,
Note that these studies are for the United States in a period views above-market-clearingwages as a disci-
when unionizationwas quite weak; it is thus unlikely to be
the major factor in such wage differentials. In contrast, plinary device. In the Shapiro-Stiglitz model,
Dunlop's wage differentials may have been mainly the firms pay "high"wages to reduce the incentive
result of differentialsin union power. of workers to shirk. The attemptof all firms to
19See Kruegerand Summers (1988).
20 See Katz (1986) and Alan S. Blinder and Don H. Choi
(or significant) relative to the size of the Rule-of-thumbpricing behavior takes many
shock.30 We dubbed the rule-of-thumb strate- forms.For example,staggeredprice (wage) mod-
gies employed by firms with inertial price set- els, in which firms keep nominalprices (wages)
ting "near-rational"since the losses they suffer fixed for a period of time, correspondclosely to
from their departurefrom complete optimiza- descriptionsof price- (wage-) settingprocesses.32
tion are second-order (or small). In the Taylor staggeredcontractmodel, during
The logic of the key result-that near-rational each period,half of all firms set a nominalprice
price stickinessis sufficientto impartsignificant which they maintain for the succeeding two-
power to monetarypolicy-is simple. With mo- periodinterval.33A variantof the staggeredcon-
nopolistic competition, each firm's profit func- tractmodel, due to GuillermoA. Calvo, assumes
tion is second-differentiablein its own price so insteadthat a fixed nominalprice is reset at ran-
that the profit function is flat in the neighbor- domly varyingintervals.34New Classicalecono-
hood of the optimumown-price.In consequence, mistsobjectto bothrenditionsof the model,on the
any deviation from the profit-maximizingprice groundsthat such price setting is not maximiz-
causesa loss in profitsthatis small-second-order ing.35Of course,they areright:insteadof keeping
with respectto the size of those deviations.But if nominalpricesunchangedduringa fixed interval,
the deviationsfrom the optimumof a large num- Taylor's and Calvo's firms would do better by
ber of firmsare similar-for example,if they are establishingpricesthatvary withinthe intervalin
all slow to adjusttheirprices following a change accordancewith the firm's expectationsof the
in the money supply-then real balances (the money supply (aggregatedemand).Such profit-
money supply deflated by the price level) maximizingbehaviorwould again rendermoney
change by a first-order amount relative to a supply changes neutral. However, price-setting
situationwith fully optimizing price-settingbe- (wage-setting) strategies of the Taylor/Calvo
havior. This first-orderchange in real balances, type are near-rational:the small amountof nom-
in turn, causes first-orderchanges in aggregate inal rigidity that characterizesthese models is
demand,output,and employment.For example, sufficient to allow monetarypolicy to be stabi-
suppose that the money supply increases by a lizing, yet the losses relative to a strategy that
fraction E and a fraction of firms keep their varies prices within the pricing interval are
prices unchanged.Each firm's losses, relative to second-order.36There are many other forms of
fully optimizing behavior, are approximately
proportionalto the squareof E. If E is 0.05, for
example, its square is quite a small number, but shocks to the money supply change real variables by a
0.0025, so the losses from price stickiness are first-orderamount. In Mankiw's formulationsmall "menu
apt to be small. However, assuming money de- costs," which are fixed costs for making a price change,
mand is proportionalto income, the change in inhibitprice changes with effects on equilibriumoutputthat
real outputis first-order-proportional are an order larger than the menu cost.
to E.(With 32
Especially see Carlton(1986).
fully maximizing behavior by all firms,the change 33See Akerlof (1969), Stanley Fischer (1977), and John
in the money supply leaves output unchanged.) Taylor (1979).
Thus small deviationsfrom completerationality- 34 See Calvo
(1983).
35
indeed small and reasonable deviations from See Barro (1977) for this complaint about staggered
contractmodels.
completerationality-reverse the conclusionthat 36 See Akerlof and Yellen (1991). Technically, it turns
expected changes in the money supply have no out that the amplitudeof the business cycle, as measuredby
effect on real income and output.31 the standarddeviation of (log) income rises due to Taylor's
staggeredcontractsby an amountthat is proportionalto the
standarddeviation of the pricing "error"made by Taylor's
firms. Monetarypolicy can offset this price stickiness and
30
See Akerlof and Yellen (1985a, b), N. Gregory Man- reduce business-cycle volatility. But the losses realized by
kiw (1985), Michael Parkin(1986), and Olivier Blanchard firms from the use of Taylor-type staggered contracts are
and Nobihiro Kiyotaki (1987). second-order,proportionalto the variance of shocks to the
31 The same results hold in a number of alternative system. In this sense, staggered pricing has a first-order
frameworks. For example, if firms set profit-maximizing effect on both the size of the business cycle and the stabi-
efficiency wages, nominal wage stickiness is a form of lizing propertiesof monetarypolicy. But the nonmaximiz-
rule-of-thumb behavior with similar consequences: the ing behavior which allows monetarypolicy to stabilize the
losses to the firm holding wages constant are second-order, economy results in losses that are second-order.
tions consistent with equality between actual different from unity. The inference was drawn
and expected inflation-is vertical because that the lagged inflationterms in such estimates
there is one and only one unemploymentrate: correspond to expected inflation, which is an
the "natural rate"-at which actual and ex- autoregressiveweighted average of past infla-
pected inflation match. tion, and that the coefficient on expected infla-
To see why the long-run Phillips curve must tion in determining current inflation is one.43
be vertical, imagine that a centralbank attempts Finally, there is a bias for economists to accept
via monetarypolicy to hold unemploymentbe- rationally based null hypotheses, even though
low the naturalrate. With labor marketsabnor- accepted only by tests with relatively low
mally tight, workers demand nominal wage power.44
increases in excess of expected inflation (plus Economists should not have acceptedthe nat-
normal real wage cum productivity gains). ural rate hypothesis so readily. There are both
Firms,in turn,pass the associatedcost increases theoretical and empirical reasons to be highly
into prices, so thatinflationexceeds what work- suspicious. Theoretically, the natural rate hy-
ers initially anticipated when they bargained. pothesis reminds me of a common diet book
With unemploymentbelow the naturalrate, ac- rule of thumb. According to that rule of thumb
tual inflation therefore exceeds expected infla- for every 3,200 calories extra that we eat, we
tion. Ex post, workers have been fooled. So, gain a pound. For every 3,200 calories less, we
over time, inflationaryexpectations, and infla- lose a pound. This always makes me imagine
tion in turn, accelerates. With unemployment twin brothers. One of these twin brotherseats
held below the naturalrate, the consequence is just enough to keep his weight even. The other
ever accelerating inflation. Similarly, the twin eats one more 100-calorie cookie per day.
Friedman-Phelpsmodel predicts that a central If the rule of thumb is right, after one year the
bank attemptingto hold unemployment above cookie eater is 11 pounds heavier than his
the natural rate indefinitely eventually causes brother. After a decade he is 110 pounds
accelerating deflation. Only the naturalrate of heavier.Fifty years later, should he live so long,
unemploymentyields steady inflation. he would be 550 pounds heavier. Just as ex-
Economists accepted the naturalrate hypoth- pected, the rule of thumb does break down
esis remarkablyquickly after it was first pro- when extrapolatedover long time periods:more
posed by Friedman and Phelps in the late accuraterenditions of the relationshipbetween
1960's. Three things conspired in its favor. weight and calories show that the maintenance
First, it seemed to explain remarkablywell the of higher weight requires extra caloric intake.
inflation-unemployment experience of the Happily the twins' weights will not diverge
1960's and 1970's. At the low unemployment forever. Similarly my guess is that for at least
rates of the late 1960's, inflation rose, which some band of unemployment rates, inflation
apparentlydrove up inflationaryexpectations, would asymptoteto a constantvalue ratherthan
shifting the short-rununemployment inflation accelerateor decelerateindefinitely.Such a pri-
trade-offoutward.Thus the 1970's began with a ori reasoningcould be wrong, but the errorfrom
much less favorable unemployment inflation overextrapolationof the diet book rule of thumb
trade-offthan the 1960's. (Analysts ignored the warns us that the natural rate hypothesis is
equally plausible explanation that as inflation rather odd. At very low unemployment rates,
increased, as it did in the late 1960's, wage the Friedman/Phelpsprediction of accelerating
bargains and price setting began to take infla- inflation seems quite possibly reasonable and
tionaryexpectations,which had previouslybeen
ignored, into account.)42Second, empirical es-
timates of the Phillips curve yielded coefficients 43We should here note Thomas J. Sargent's (1971)
on past inflationwhose sum was not statistically criticism that the coefficient on lagged inflation will not
equal one in an accelerationistmodel if the process gener-
ating inflation is stable, without a unit root.
44We shall see an example of such bias below when we
42This alternative explanation was given by Otto review Summers'criticism of the acceptanceof the random
Eckstein and Roger Brinner(1972), but did not make it into walk hypothesisbased on failureto reject by tests with very
the mainstream. low power against alternativehypotheses.
empirically relevant.45 But I am suspicious even with allowance for such shifts, estimates
about the theory's applicability when unem- of the naturalrate possess high standarderrors.
ployment is high. Douglas Staiger et al. (1997) compute a 95-
My suspicions regardingthe naturalrate hy- percent confidence interval for the U.S. natural
pothesis are supported by an empirical fact, rate which exceeds 5 percentagepoints; this is
which reveals that its applicability is not uni- more than three times the standarddeviation of
versal. Unemploymentin the United States for the U.S. monthly unemploymentrate over the
the whole of the 1930's was indisputably in last 50 years.
excess-surely greatly in excess-of any plau- In recent papers, William Dickens, George
sible naturalrate. According to the naturalrate Perry, and I have explored two behavioralhy-
hypothesis, price deflationshould have acceler- potheses that,contraryto the naturalratemodel,
ated for the whole decade. That did not happen. produce a stable trade-off between unemploy-
Prices fell for a time, but deflationstopped after ment and inflation at sufficiently high unem-
1932; there was no significantdeflation for the ployment and low inflation rates. The first
next ten years, despite extremely high unem- hypothesis is "pure Keynes": workers resist,
ployment. This evidence suggests that, at least and firms rarely impose, cuts in nominal pay.
aftersome time, at high levels of unemployment The second hypothesis concerns the role of in-
and low inflationrates, the naturalrate hypoth- flationary expectations in wage bargains: we
esis breaks down. Such a failure would not be argue that, at very low inflation, a significant
terribly serious for a theory derived from em- number of workers do not consider inflation
pirical observation, but it constitutes a serious sufficiently salient to be factored into their de-
flaw for a relationship derived from a priori cisions. However, as inflation increases, the
principles, principles that are accepted because losses from ignoring it also rise, and therefore
they are supposedto be always and everywhere an increasingnumberof firms and workerstake
true. it into account in bargaining.
The evidence of the 1930's is not unique. Keynes' assumptionthat workersresist nom-
Modem economies display similar characteris- inal wage cuts was consistent with his intuitive
tics. For example, Pierre Fortin estimates that understandingof psychology. The assumption
from 1992 to 2000, the Canadianeconomy ex- also coincides with psychological theory and
periencedalmost 12 points of unemploymentin evidence. Prospecttheoryposits thatindividuals
excess of a very conservative, 8-percent esti- evaluate changes in their circumstancesaccord-
mate of NAIRU.46 During that same period, ing to the gains or losses they entail relative to
inflation averaged a very low 11/2percent per some reference point. The evidence suggests
year. According to naturalrate theory, core in- that individuals place much greater weight on
flation should have declined by roughly 6 per- avoiding losses than on incurringgains. Daniel
centage points, since a typical estimate of the Kahneman and Amos Tversky (1979) have
Phillips curve slope is 1/2.Instead, inflation de- demonstrated that many experimental results
clined over that period by only 0.1 percent. which are inconsistent with expected utility
Econometric evidence further suggests that maximization can be rationalized by prospect
the naturalratetheoryrests on shifty sandrather theory. Downward wage rigidity is a natural
than bedrock. Time-varying estimates of the implication of prospect theory if the current
naturalrate show that it changes over time; but, money wage is taken as a reference point by
workers in measuringgains and losses. In sup-
port of this view, Eldar Shafir et al. (1997)
found in a questionnairestudy that individuals'
45 The occurrenceof hyperinflationwith low unemploy-
ment maintainedsufficiently long is one prediction of the mental frames are defined not just in the real
theory. The frequentoccurrenceof hyperinflationseems to terms hypothesizedby classical economists but
supportthe theory. But these hyperinflationshave occurred also exhibit some money illusion.
when governments have lost fiscal credibility (and could Numerous empirical studies document that
only pay their deficits by seigniorage).It may be the loss of
fiscal credibility, not the maintenance of low unemploy- money wages are, in fact, downward sticky.
ment, which is the cause of the hyperinflation. Using panel data, David Cardand Dean Hyslop
46 Observationdue to Fortin in Fortin et al. (2001). (1997) and Shulamit Kahn (1997) found that
distributions of nominal wage changes are impose nominal wage cuts results in higher
asymmetricaroundzero. Fortinfound a remark- permanentrates of unemployment.Because the
able pileup of wage changes at zero in Canadian real wages at which labor is supplied are higher
data. From 1992 to 1994, when Canadianinfla- at every level of employment when inflation is
tion was 1.2 percentand the unemploymentrate low, the unemploymentrate consistent with sta-
averaged 11.0 percent, only 5.7 percent of non- ble inflationrises as inflationfalls to low levels.
COLA union agreements had first-year wage Spillovers produce an aggregate employment
cuts, whereas47 percenthad wage freezes.47In impact which exceeds the employmentchanges
detailed interviews in Connecticut, Bewley in those firms that are constrainedby their in-
found that managersare willing to cut nominal ability to cut wages. Thus, a benefit of a little
wages only as a last resort.4 To investigate inflation is that it "greases the wheels of the
whether firms cut total compensation through labor market."
benefit cuts as opposed to money wage cuts, Simulations of a model with intersectoral
David E. Lebow et al. examined the individual shocks and aversion on the part of firms to
industriescovered by the EmploymentCost In- nominal wage cuts suggests that, with realisti-
dex: they found that benefit cuts are only a cally chosen parameters,the trade-off between
minor substitutefor nominalwage cuts.49Using inflation and unemployment is severe at very
Swiss data, Fehr and Lorenz Goette found that low rates of inflation,when productivitygrowth
even a seven-year period of low inflation and is low. For example, a permanentreductionin
low productivity growth did not increase the inflationfrom 2 percentper year to zero results
frequency of money wage cuts.50 in a permanent increase in unemployment of
At low inflation there is a long-run trade-off approximately 2 percentage points.52 Estima-
between outputand inflationif there is aversion tion of a Phillips curve for the United States
to nominal pay cuts. Unlike the Friedman- after World War II, correspondingto the simu-
Phelps model, in which such a trade-offis tran- lation model just described, gives similar re-
sitory, long-term increases in inflation (if it is sults. When the Phillips curve thus estimatedis
close to zero) resultin significantlyless employ- used to simulate the inflationexperience of the
ment and more output.51The logic goes as 1930's, the fit is shockingly close to actualU.S.
follows. In both good times and bad, some firms inflation experience during the depression.53A
and industriesdo betterthanothers.Wages need comparable simulation of the standardnatural
to adjust to accommodate these differences in rate model, in contrast,counterfactually,shows
economic fortunes. In times of moderateinfla- acceleratingdeflation throughoutthe 1930's.
tion and productivitygrowth,relativewages can An alternativebehavioraltheory also gener-
easily adjust.Unlucky firmscan raise the wages ates a permanenttrade-offbetween inflationand
they pay by less than the average, while the unemploymentat low inflation. This theory is
lucky firms can give above-average increases. based on the idea that because inflation is not
However, if productivity growth is low (as it salient when it is low, anticipated future
was from the early 1970's through the mid- changes in the price level are ignored in wage
1990's in the United States) and there is no bargaining.54With monopolistic competition
inflation, firms that need to cut their real wages and efficiency wages such ignoranceof inflation
can do so only by cutting the money wages of when it is low is near-rational.55The psychol-
their employees. Under realistic assumptions ogy of just noticeable differences and cogni-
about the variability and serial correlation of tive psychology both suggest that people tend
demand shocks across firms, the needed fre-
quency of nominal cuts rises rapidlyas inflation
declines. An aversion on the part of firms to 52
See Akerlof et al. (1996).
53 This is done by sequentially feeding in the simulated
inflationof the previous periodto derive adaptivelythe next
47 See Fortin (1995, 1996). period's inflationaryexpectations.The fit is so excellent that
48
See Bewley (1999). there must be a component of luck.
49 See Lebow et al. 54
(1999). Past inflation is incorporatedindirectlybecause wage
50 See Fehr and Goette take into account the wages paid by competitors.
(2000). bargains
51 See Tobin (1972). 55See Akerlof et al.
(2000).
to ignore variables that are unimportant to deflation. Centralbankerswho accept the text-
their decisions.56Econometricestimates of the book version of the natural rate hypothesis
Phillips curve which allow for the possibility should follow the advice of Oliver Cromwell to
that past inflation has a different impact on the General Assembly of the Church of Scot-
current inflation when inflation is high than land: "I beseech you in the bowels of Christ,
when it is low are consistent with this hypoth- think it possible you may be mistaken."It is no
esis: at high inflation,the sum of coefficients on coincidence that the leading survey of cognitive
past inflationis close to one.57At low inflation, psychology uses this citation to demonstratea
this sum of coefficients is much closer to zero. common perceptualerror:overconfidence.60
Similarly,regressionsusing survey measuresof
expected inflation as an independent variable V. Undersaving
yield much higher coefficients on the expected
inflation term at high inflation than at low in- It is common wisdom that people save too
flation.58Not surprisinglythen, when periods of little. To compensate for this failure, most de-
low and high inflationare combined to estimate veloped country governments heavily support
a nonlinearmodel of the influence of inflation- the elderly in retirement. In addition, a very
ary expectations we find that their impact de- large number of employers require and subsi-
pends on the recent history of inflation. dize pension contributionsof their employees.
The demonstrationby behavioralmacroeco- Many forms of saving receive tax advantage.
nomics that very low inflation has the cost of Even with these legs up, the common wisdom is
permanentlyhigh unemploymentand low out- that financialassets of most households still fall
put, has important implications for monetary considerably short of what they need to main-
policy. Most of us think of central bankers as tain their consumptionin retirement.6'
cautious, conservative, and safe. But I consider For New Classical economics, saving too lit-
many to be dangerous drivers: to avoid the tle or too much, like involuntary unemploy-
oncoming traffic of inflation, they drive on the ment, is an impossibility, a straightforward
far edge of the road, keeping inflation too low contradictionof the assumptionsof the model.
and unemploymenttoo high. Duringthe 1990's, Since saving is the result of individual utility
Canadahad very low inflation and an unprece- maximization, it must, absent externalities, be
dentedunemploymentgap-close to 4 percent-
age points-with the United States.59Europe 60
See Nisbett and Ross (1980). This book is one of the
has also had high unemploymentand very low
leading primersfor the psychology of behavioralmacroeco-
inflation.Japanhas gone much further,allowing nomics. Curiously, cognitive psychologists have a much
more empiricalbasis for their theories than economists.
61 Eric M. Engen et al. (1999, p. 97) reach the opposite
conclusion. They compare the actual wealth with that de-
56This formulation is also influenced by the public's rived in a calibrated optimization model. Their preferred
mental frame regarding inflation. Robert J. Shiller calibrationhas a rate of time preferenceof 3 percent. With
(1997a, b) has elicited the differences in mental frame be- data from the U.S. Health and Retirement Survey with a
tween the public and economists by questionnaire re- broad definition of wealth to include all home equity, 60.5
sponses. percent of households have more than the median optimal
57 One is not, however, necessarilythe magic numberfor wealth in the calibrated model. But I would focus on an
the reasons noted earlier by Sargent (1971). alternativeresultfrom theirsimulations.If we exclude home
58Such regressions address the problem suggested by equity investmentin spendablefinancialcapital,and assume
Sargent that the naturalrate model should produce coeffi- a zero rate of intertemporaltime discount,only 29.9 percent
cients on expected inflation that correspondto the money of households reach the preretirementage of 60 or 61 with
supply rule, and those coefficients need not be equal to more than the optimal median wealth for someone of their
unity. If expectations are observed without error,the coef- age (p. 99, Table 5). Like the discussants,both for empirical
ficient on expected inflationwith naturalrate theory should and a priorireasons, I view a zero rate of discount as more
be unity. Error in the expectations data should bias its correct. This conforms to people's stated preference for
coefficient downward,but it should not, as observed, result nondeclining consumption at a zero rate of interest (see
in changes in the coefficient, unless there are also changes below) and it weights utility at different ages on a one-for-
in the errorof observationbetween periods of high and low one basis. My choice to exclude home equity capital as-
inflation. sumes that retireesshould not have to leave their homes for
593.8 percent from 1990 to 1999, accordingto the Eco- financialreasons, or to reverse-mortgagethem, as they get
nomic Report of the President (2000, Table B-107). older.
just right. Behavioral macroeconomics,in con- contrast to the constant discount rates that are
trast,has developed theoreticaltools and empir- standardin neoclassical theory, the hyperbolic
ical strategiesto advanceunderstandingof such function assumes that the discountrates used to
time-inconsistentbehavior. evaluate trade-offs between adjacent periods
A key theoretical innovation permittingsys- decline as the time horizon lengthens: individ-
tematic analysis of time-inconsistentbehavioris uals use high discount rates to evaluate options
the recognitionthatindividualsmay maximize a that requirean immediate sacrifice for a future
utility function that is divorced from that repre- rewardand lower discount rates when the same
senting "truewelfare." Once this distinction is sacrifice is deferredinto the future. Thus, they
accepted, "saving too little" becomes a mean- are patient in making choices requiringgratifi-
ingful concept. The idea can be illustratedby cation delays when those sacrificesare deferred;
the ancient myth of the lemmings, who every but impatient in delaying gratification in the
few years are said to converge in a deathmarch, shortrun. Because presentconsumptionis more
which ends with theirfinal plunge into the sea.62 salient than future consumption, individuals
The alleged behaviorof those lemmings reveals procrastinate about saving. The hyperbolic
a distinctioncommon among psychologists, but function accords closely with experimental
rarefor economists. Unless the lemmings expe- findings:Humanand animalsubjectsare far less
rience an unusualepiphanyin that final plunge, willing to delay gratificationimmediately than
their utility or welfare is given by one function; to commit to such delays in the future.66
yet they maximize another. Two forms of procrastination may result
Think about it: the popular view of saving, from hyperbolic discounting. "Naive procrasti-
that people undersave, is similarly described. nation"occurs when an individual assumes in-
Determining whether people save too much or correctly that her utility function will be
too little involves asking whether people, like different in the future. She mistakenly projects
the lemmings, have one (intertemporal)utility that, althoughtoday is salient, tomorrowwill be
function which describes their welfare, but different. She fails to see that tomorrow's self
maximize another.63Such evidence as there is will be different from today's self, so that to-
suggests potentially large difference between morrow will be just as salient as today once it
the two concepts. High negative rates of time has moved one step closer. The naive procras-
discount are necessary to explain actual wealth- tinator mistakenly believes that she will save
earnings ratios.64Yet, questionnaireresponses (diet, exercise, quit smoking, etc.) tomorrow,
on the consumption-savingtrade-offs that peo- although she has not done so today, and is
ple think they ought to make reveals an inter- surprisedthat the sacrifices deferred today are
temporal discount rate that is on average also deferred again tomorrow. More sophisti-
slightly positive.5 cated procrastinationtakes the form of preprop-
The hyperbolic discount function, which has eration, according to the terminology of Ted
been used to study intertemporal savings O'Donoghueand Rabin (1999). The prepropera-
choices, can be used to formalizethe distinction tor has fully rationalexpectationsaboutwho her
between the utility function that describes ac- futureself will be. She says to herself:thereis no
tual saving behaviorand the utility functionthat reason to save today if tomorrowis going to be
measuresthe welfare resultingfrom thatbehav- especiallysalient.If tomorrowis especiallysalient
ior. The hyperbolic function capturesthe diffi-
culty people have in exercising self-control. In 66
See Robert H. Strotz (1956), Phelps and Robert A.
Pollak (1968), George Ainslie (1992), George Loewenstein
and Drazen Prelec (1992), Laibson et al. (1998), and
62 Laibson (1999). In Akerlof (1991) I was regrettablyun-
My 1946 version of The Encyclopedia Britannica
describes as fact the march of the lemmings, which "never aware of earlier work on intertemporalinconsistency. In
ceases until they reach the sea, into which they plunge and economics this includes Strotz (1956), Phelps and Pollak
are drowned." (1968), RichardH. Thaler(1981), and Loewenstein (1987).
63 This difference is made explicit in David I. Laibson Loewenstein and Thaler (1989) give an excellent early
(1999). review of the previous literatureon dynamic inconsistency
64 See
Engen et al. (1999, pp. 157-58). including the psychological experiments and theory. See
65
See Robert S. Barsky et al. (1995, p. 34). also Ainslie (1992).
then I will spend whateversavings I have laid fraction of wage or salary increases to be set
asidetodaywhen it was also especiallysalient.So aside for savings. Consistent with hyperbolic
I shouldnot make the sacrificetoday. discounting,but not with the standardexponen-
Laibson has used hyperbolic discounting as tial model, workerschose relativelymodest sav-
the basis of a research programon saving be- ing out of currentincome but committedto save
havior and policy. With coauthors Andrea large fractions of future wage and salary in-
Repetto and Jeremy Tobacman (1998) he has creases. Within a short period of time, the av-
simulated the effects of different tax incentive erage savings rate had doubled.7
programsin a world in which consumers pre-
properate.They estimatethatlarge positive wel- VI. Asset Markets
fare effects result from small changes in
incentives to save which reduce the amount of Keynes' General Theory was the progenitor
preproperation.Because of this work the regu- of the modem behavioralfinance view of asset
lations regardingtax-advantaged401(k) savings markets.In Keynes' metaphor"professionalin-
plans have been changed. If firms so choose, vestment may be likened to those newspaper
workers may now be automatically enrolled competitionsin which competitorshave to pick
with an automatic default contribution.Adop- out the six prettiestfaces from a hundredpho-
tion of such plans significantly increases plan tographs,the prize being awardedto the com-
participationand many workers maintain their petitorwhose choice most nearlycorrespondsto
contributionsat the level of the default.67 the average preferencesof the competitorsas a
Besides the popularityof social security and whole."72Thus stock markets are too volatile
other programsthat "force"consumersto save, and also too responsiveto news. This view of the
the best evidence of undersavingis probablythe stock marketcontrastswith the efficientmarkets
observation that, upon retirement,individuals, model in which stock prices measurethe present
on average,reduceconsumptionsubstantially.68 value of futurereturnsadjustedfor risk.
In fact, consumptionat retirementdeclines dis- In the early 1980's Robert Shiller conducted
continuously.69Those with more wealth and a direct test of the Keynesian excess volatility
higher income replacement reduce their con- hypothesis. He reasoned that if stock prices
sumptionby much less. This finding is difficult really are the predictedvalue of expected future
to explain with the standardlife cycle, exponen- returns, they should vary less than the dis-
tial discounting model.70 counted returns themselves. Shiller's insight
Thalerand Shlomo Benartzi(2000) have de- was a direct application of a simple statistical
vised a savings plan to overcome workers' ten- principle: a good forecast should have lower
dency to procrastinateand have tested it on an variancethan the variablebeing forecast. If the
experimentalbasis at a mid-size manufacturing weather forecast has greater variance than the
firm: employees were invited to join a savings actualweather,the weatherforecastershould be
plan allowing them to elect, in advance, the fired.73Using 100 years of U.S. data on stock
prices and dividends, Shiller (1981) compared
the variance of detrended stock prices to the
67
variance of the detrended present discounted
See Brigitte C. Madrianand Dennis F. Shea (2001).
68
See James Banks et al. (1998) and B. Douglas
Bemheim et al. (2001).
69
Such declines might occur if retirementis associated
with negative income shocks. Bernheimet al. (2001, p. 854) 71 From 4.4 percent to 8.7 percent.This behavior is also
suggest that such an adjustmentis relatively minor. explained by prospect theory by Kahneman and Tversky
70 Retirees,of course, obtain
greaterleisure, and thus one (1979). According to prospect theory the framing of
might expect a reductionin consumptionas leisure is sub- decision-making is important and people resist taking
stituted for consumption.It is difficult, but not impossible, losses. In this context these employees do not want to take
to explain, in addition,why such substitutionvaries system- losses in their consumption.
atically both with the level of wealth and with the income 72 Keynes (1936, p. 156).
73For
replacementratio.This could occur if those with a particular example, drawingfrom a normaldistribution,the
taste for leisure in retirementhave by choice high income forecast that yields the smallest squareddeviation between
replacement ratios and have accumulated high levels of the actual draw and the forecast is the mean of the distri-
savings. bution, which is a constant with no variance at all.
values of dividends.74 He found just what more than 30 percent away from fundamen-
Keynes would have expected: the standardde- tals 35 percent of the time.77
viation of (detrended)stock prices is five times Beyond establishing the existence of excess
largerthanthe standarddeviationof (detrended) volatility, Shiller has also examined its possible
discounted dividends. These results have been causes. In IrrationalExuberance(2000), he re-
confirmedin more sophisticatedtests that prop- views the news coverage of the stock market
erly allow for the nonstationarityof both stock bubble of the 1990's and explains how the idea
prices and the present discounted values of of a "new era"both in financialmarketsand the
dividends.75 real economy was propagated.As stock prices
The results of variance-bounds tests not- rose, the "new economy" mantrawas transmit-
withstanding, belief in efficient markets was ted from person to person; individual investors
sustained by empirical results such as the acted on the opinions of the media, which ex-
finding of insignificant serial correlation in aggeratedthe effects of economic fundamentals
returns in monthly data.76 Rejection of the such as the interneton productivity.Such stock
hypothesis that returns are serially correlated marketbubblesare common;they have occurred
suggests that the stock market follows some- in many other countriesand frequentlyover the
thing close to a random walk. In response, courseof history.Indeed,Kindleberger'saccounts
Summers (1986) showed in a model of of manias and panics and Galbraith'shistory of
"fads"-with serially correlated deviations the Great Crash of 1929 are distinguishedpre-
from perfect markets-that serial correlation decessors to Irrational Exuberance.
tests have very low power: the power of such A second major empirical finding that casts
tests is so low as to require 5,000 years worth doubt on the rationalityof the stock market is
of data before it could discriminate 50 percent the equity premium puzzle. Over the last 200
of the time between the random walk hypoth- years, the returnon equity has been significantly
esis and a fad which would drive stock prices higher than the returnon bonds. For example,
from 1802 to 1998 the real returnon a value-
weighted market equity index was 7.0 percent
per annum comparedto 2.9 percent for a rela-
74He extrapolatedfuturedividends for times beyond his tively riskless security.78Over the last 75 years,
period of observation.For a similar test also see Steven F. 1926-2000, the real returnswere 8.7 percenton
LeRoy and RichardPorter(1981). equity versus 0.7 percenton bonds, a gap of 8.0
75
See John Y. Campbell and Shiller (1987). Although
Shiller's tour de force initially seemed to clinch the case, percent. A gap of this size is huge: Jeremy J.
two technical problems cast a shadow of doubt. The first Siegel and Thaler(1997) calculatethat a $1,000
problem is that detrendingpotentially introducesa serious investment made 75 years ago would have
bias into Shiller's procedure:neither stock price series nor
dividends are stationaryand a nonstationaryseries does not
yielded $12,400 in bonds and $884,000 in
even possess a variance.The second problem relates to the
stocks. This gap is so large that rejection of
shortnessof Shiller's sample and his extrapolationof future rationality is duck soup: With intertemporal
dividends beyond the present. Allan W. Kleidon (1986) maximization of utility, the marginalutility of
showed in simulated data that the difference between the consumption today should equal the expected
variance of Shiller's detrendedstock price and of his divi- extra utility tomorrow from forgoing one unit
dend series is not large enough to confidently reject the
efficient marketnull hypothesis when returnsfollow a ran- of consumption today. With a constant rela-
dom walk. The Campbell-Shillertest allows for the nonsta- tive risk-aversion utility function, this condi-
tionarity of stock prices and dividends, provided the two tion implies that the expected equity premium
series are cointegrated.This test is also valid even if firms should equal the product of the coefficient of
smooth dividends.
The high volatility of stock prices could also be ex-
risk aversion and the covariance between the
plained by a high frequency cycle in the expected real rate growth of consumption and the return on
of returnon stocks. But such a cycle is inconsistent with
most standardclassical models of the economy, where real
returns are mainly determinedby the state of technology,
and the capital-labor ratio. In the standardclassical model 77 Kenneth D. West (1988)
similarly demonstratedthe
both technology and the capital-labor ratio change slowly. low power of Kleidon's efficient marketstest using Shiller's
76 Where not
insignificant in the statistical sense, such detrendeddata.
correlationseemed insignificantin magnitude. 78 See Rajnish Mehra (2001,
p. 1).
stock prices. For reasonable values of the VII. Poverty and Identity
coefficient of risk aversion, however, this
product is much smaller than the equity pre- If income distribution is a topic in macro-
mium, thus rejecting rational consumption be- economics, as many have professed, then be-
havior. This rejection is known as the equity havioral economics also offers insight on the
premium puzzle.79 most enduring macroeconomic problem fac-
Furtherevidence of the irrationalityof stock ing the United States: the disparity in income
prices comes from cross-sectiondata. Similarto and social condition between the majority
Shiller's time-series finding of excess volatility white population and the African-American
coupled with reversion to the mean in price/ minority. As a legacy both of slavery and the
dividends ratios, Werner F. M. De Bondt and Jim Crow discrimination that followed it,
Thaler (1985) find reversion to the mean of poverty weighs especially heavily on African-
stock returnsin a cross section: successive port- Americans. The black poverty rate of 23.6
folios formed by the previous five years' 50 percent in 2000 was roughly triple the white
most extreme winners considerably underper- rate of 7.7.83 Despite comprising only about
form the marketaverage,while portfolios of the one-eighthof the population,African-Americans
previous five years' 50 worst losers perform have almost one-fourth of all U.S. poverty.84
betterthanthe marketaverage.Otherstock mar- The reality is yet more disparate than these
ket anomalies, such as a 20-percent one-day statistics indicate because the problems of the
decline in stock prices in October 1987 in the poorest African-Americans go beyond mere
absence of any significantnews also cast doubt poverty. They include extraordinarily high
on the efficient marketshypothesis.80 rates of crime, drug and alcohol addiction,
Asset marketsare not only importantfor their out-of-wedlock births, female-headed house-
own sake, they are also importantbecause they holds, and welfare dependency. Statistics on
affect the macroeconomy,throughat least three incarceration indicate that even the worst of
channels. First, the value of assets affects these problems affect a significant fraction of
wealth and, in turn, consumption. Second, the African-Americans. Thus, for example, about
price of existing assets relative to the price of 4.5 percent of black males are either in jail or
new capital-Tobin's q ratio-affects invest- in prison.85The black male incarceration rate
ment since investment can be viewed as an exceeds the white male rate by a factor of
arbitragebetween new capital stock and claims eight to one.86 And the lifetime chances of a
to similarexisting assets.81Finally, asset values black male youth entering prison exceeds
affect the chances that firms will go bankrupt. one-fourth.87
Firms close to bankruptcyfind it difficult, if not
impossible, to borrow, and thus commonly
forgo profitableinvestment opportunities.82 83 Hispanics have a similar but less extreme history of
discrimination.
84
See (http://www.census.gov/Press-Release/www/
79 It is remarkablethat even this weak test leads to
2000/cb00-158.html).
rejection,since most theoriesof consumption,whethermax- 85 In 1996 there were 530,140 black male prisonersand
imizing or not, would suggest considerablecorrelationbe- 213,100 black non-Hispanic and 80,900 Hispanic jail in-
tween the rate of returnon stocks and the rate of growth of mates of both sexes. There were 462,500 male and 55,800
consumption. For example, such a correlation occurs if female inhabitantsof jails. Extrapolatingthe black Hispanic
consumers have a consumptionfunction which naively de- rate at 0.3 and the male/femalerate for black as the same as
pends on their wealth, or, alternatively,if the same opti- white yields 211,814 black males in jail in 1996. The black
mism that leads to high returns in the stock market also male populationwas about 1/2(30+ 0.6 x 4.7) million =
leads to consumption binges. JonathanA. Parker (2001) 32.282/2 = 16.141 million. The net result is about 4.5
suggests a possible resolutionof the equity premiumpuzzle. percent of the African-Americanmale populationin prison
80 See Romer (1993, p. 1112). or jail. Source of incarcerationrates: Correctionalpopula-
81See the literature on q theory, especially including tions of the U.S. 1996, U.S. Departmentof Justice, Table
Tobin (1969), Summers (1981), Andrew B. Abel (1982), 5.7, p. 82. Source: (http://www.census.gov/statab/www/
and Fumio Hayashi (1982). partla.html).
82 See Stewart C. Myers (1974); Michael C. Jensen and 86 See (www.hrw.org/reports/2000/usa/Table3.pdf).
William H. Meckling (1976). Owen Lamont (1995) shows 87 This is an estimate based on incarcerationrates in
how dual equilibriamay occur because of such dependence. 1993.
Because standard economic theory, in our crime by deterrence: raise the stakes high
view, is incapable of explaining such self- enough, as Californiadid with its "threestrikes
destructive behavior, Rachel E. Krantonand I and you're out" law, and the potential criminal
have developed models, based upon sociologi- will think twice. But the prisons are full and
cal and psychological observation, to under- crime has not stopped.An identity-basedtheory
stand the persistence of African-American suggests, in contrast,that large negative exter-
disadvantage (2000). Our theory stresses the nalities from incarcerationmay offset the short-
role of identity and the decisions that individu- run gains from deterring criminal activity
als make about who they want to be. In our throughtougherincarcerationpolicies.88Prison
theory of minority poverty, dispossessed races itself is a school for counterculturalidentity,
and classes face a Hobbesian choice. One pos- and thus the breeding ground for future crime.
sibility is to choose an identitythat adaptsto the Moreover, externalities in identity formation
dominant culture. But such an identity is argue for programs to allay crime before it
adoptedwith the knowledge thatfull acceptance has occurred.These include, for example, effec-
by membersof the dominantcultureis unlikely. tive, easily accessed drug treatmentand reha-
Such a choice is also likely to be psychologi- bilitation programs and public jobs for inner-
cally costly to oneself since it involves being city youth. Identity theory suggests that the
someone "different";family and friends, who benefits of increased expenditures for schools
are also outside the dominantculture are likely in African-Americanneighborhoodswith high
also to have negative attitudestowarda maver- povertyratesare likely to be substantial:African-
ick who has adopted it. Thus individuals are American children have been found to be par-
likely to feel that they can never fully "pass."A ticularly responsive to differences in teacher
second possibility is to adopt the historically quality and class size.89It may take the extraor-
determined alternative identity, which, for dinary teacher and close personal attention to
many minorities, is an oppositional culture. sort through student issues concerning identity
Each identity is associated with prescriptions in addition to covering the standard curricu-
for ideal behavior. In the case of the opposi- lum.90 Finally, the externalities involved in
tional identity, these prescriptions are com- identity formationargue for affirmativeaction,
monly defined in terms of what the dominant because it is a symbol of welcome for African-
culture is not. Since the prescriptions of the Americans into the white society that has re-
dominant culture endorse "self-fulfillment," jected them for so long.91
those of the oppositional culture are self-
destructive.The identityof the oppositionalcul- VIII. Conclusion
ture may be easier on the ego, but it is also
likely to be economically and physically It is now 30 years since the revolutionwhich
debilitating. began in growth theory and then swept through
This identity-basedtheory of disadvantageis microeconomics. The new microeconomics is
consistent with a considerable body of evi- standard in all graduate programs, half of a
dence. For example, it capturesthe centralfind- two-course sequence.Adoptionof the new mac-
ings of studies by authors such as Franklin roeconomicshas been slower, but the revolution
Frazier (1957), Kenneth Clark (1965), William is coming here as well. If there is any subjectin
E. B. Du Bois (1965), Ulf Hannerz(1969), Lee economics which should be behavioral, it is
Rainwater (1970), William J. Wilson (1987,
1996), and Elijah Anderson (1990). Read any
African-American biography: the uncomfort- 88 See Steven D. Levitt
(1996).
able dance between acceptance and rejection 89
See RonaldF. Ferguson(1998) on the effect of teacher
invariablytakes center stage. quality and Kruegerand Diane M. Whitmore(1999) on the
The identity theory of minority poverty has effect of class size.
90 See Lisa
social policy implicationsthatdepartfrom those Delpit (1995).
91Glenn C. Loury (1995) has suggested that affirmative
derived from standardneoclassical theory. For action may also have the opposite effect: it may exacerbate
example, the standardeconomic theory of crime blacks' sense of exclusion and make them feel that they are
and punishmentimplicitly arguesfor combating viewed as not belonging even when they do achieve.
Du Bois, William E. B. The souls of black folk. Fortin, Pierre. "Canadian Wage Settlement
Greenwich, CT: Fawcett Publications, 1965. Data."Mimeo, Universite de Quebec a Mon-
Dunlop, John T. "The Task of Contemporary treal, April 1995.
Wage Theory," in John T. Dunlop, ed., The "The Great CanadianSlump." Cana-
theory of wage determination. New York: St. dian Journal of Economics, November 1996,
Martin's Press, 1957, pp. 3-27. 29(4), pp. 761-87.
Eckstein,Ottoand Brinner,Roger."TheInflation Fortin, Pierre; Akerlof, George A.; Dickens,
Process in the United States."JointEconomic William T. and Perry, George L. "Inflation,
Committee of the Congress of the United Unemployment, and Macroeconomic Policy
States, Washington,DC, 1972. in the United States and Canada:A Common
Economic Report of the President. 2000. Framework."Mimeo, Universitede Quebec a
Engen, Eric M.; Gale, William G. and Uccello, Montreal,June 2001.
Cori E. "The Adequacy of Household Sav- Foster, James E. and Wan, Henry Y., Jr. "Invol-
ing." Brookings Papers on Economic Activ- untary Unemployment as a Principal-Agent
ity, 1999, (2), pp. 65-187. Equilibrium." American Economic Review,
Fehr, Ernst and Falk, Armin."WageRigidity in June 1984, 74(3), pp. 476-84.
a Competitive Incomplete ContractMarket." Frazier, Franklin. The black bourgeoisie: The
Journal of Political Economy, February rise of the new middle class in the United
1999, 107(1), pp. 106-34. States. New York: Free Press, 1957.
Fehr, Ernst; Gachter, Simon and Kirchsteiger, Friedman, Milton. "The Lag in Effect of Mone-
Georg. "Reciprocal Fairness and Noncom- tary Policy." Journal of Political Economy,
pensating Wage Differentials." Journal of October 1961, 69(5), pp. 447-66.
Institutional and Theoretical Economics, De- ."The Role of MonetaryPolicy."Amer-
cember 1996, 152(4), pp. 608-40. ican Economic Review, March 1968, 58(1),
Fehr, Ernst and Goette, Lorenz. "Robustness and pp. 1-17.
Real Consequencesof Nominal Wage Rigid- Gordon,RobertJ. "The Recent Acceleration in
ity." Institutefor EmpiricalResearchin Eco- Inflation and Its Lessons for the Future."
nomics (University of Zurich) Working Brookings Papers on Economic Activity,
Paper No. 44, May 2000. 1970, (1), pp. 8-41.
Fehr, Ernst; Kirchsteiger, Georg and Riedl, Arno. . "The Time-Varying NAIRU and Its
"Does FairnessPreventMarketClearing?An Implications for Economic Policy." Journal
ExperimentalInvestigation."QuarterlyJour- of Economic Perspectives, Winter 1997,
nal of Economics, May 1993, 108(2), pp. 11(1), pp. 11-32.
437-59. Hannerz, Ulf. Soulside: Inquiries into ghetto
Fehr, Ernst and Tyran, Jean-Robert. "Does culture and community. New York: Colum-
Money Illusion Matter?" American Eco- bia University Press, 1969.
nomic Review, December 2001, 91(5), pp. Hayashi, Fumio. "Tobin's Marginal q and Aver-
1239-62. age q: A Neoclassical Interpretation."
Econo-
Ferguson, Ronald F. "Can Schools Narrow the metrica, January1982, 50(1), pp. 213-24.
Test Score Gap?"in ChristopherJencks and Hicks, John R. "Mr. Keynes and the 'Classics':
Meredith Phillips, eds., The black-white test A Suggested Interpretation."Econometrica,
score gap. Washington,DC: BrookingsInsti- April 1937, 5(1), pp. 147-59.
tution Press, 1998, pp. 318-74. Hodgman,Donald R. "Credit Risk and Credit
Fischer, Stanley."Long-TermContracts,Ratio- Rationing." Quarterly Journal of Economics.
nal Expectations, and the Optimal Money May 1960, 74(2), pp. 258-78.
Supply Rule." Journal of Political Economy, Hotelling, Harold. "Stability in Competition."
February1977, 85(1), pp. 191-205. Economic Journal, March 1929, 39(153), pp.
Flanagan, Robert J.; Soskice, David W. and 41-57.
Ulman, Lloyd. Unionism, economic stabiliza- Howitt, Peter and Clower, Robert. "The Emer-
tion and incomes policies: European experi- gence of Economic Organization."Journal of
ence. Washington,DC: Brookings Institution Economic Behavior and Organization, Janu-
Press, 1983. ary 2000, 41(1), pp. 55-84.