Você está na página 1de 18

Markets with Asymmetric Information: The Contributions of George Akerlof, Michael

Spence and Joseph Stiglitz


Author(s): Karl-Gustaf Lofgren, Torsten Persson, Jorgen W. Weibull
Source: The Scandinavian Journal of Economics, Vol. 104, No. 2 (Jun., 2002), pp. 195-211
Published by: Blackwell Publishing on behalf of The Scandinavian Journal of Economics
Stable URL: http://www.jstor.org/stable/3441066
Accessed: 11/11/2009 12:32
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=black.
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.

Blackwell Publishing and The Scandinavian Journal of Economics are collaborating with JSTOR to digitize,
preserve and extend access to The Scandinavian Journal of Economics.

http://www.jstor.org

Scand. J of Economics 104(2), 195-211, 2002

Marketswith AsymmetricInformation:
The Contributionsof GeorgeAkerlof,
Michael Spence and Joseph Stiglitz*
Karl-GustafLofgren
Universityof Umea,SE-90187 Umea,Sweden
karl-gustaf.lofgren@econ.umu.se

TorstenPersson
StockholmUniversity,SE-10691 Stockholm,Sweden
torsten.persson@iies.su.se

Jorgen W Weibull
StockholmSchoolof Economics,SE-11383 Stockholm,Sweden
jorgen.weibull@hhs.se

adverseselection;signaling;screening
Keywords:Asymmetricinformation;
JEL classification: D8

I. Introduction
Formorethantwo decades,researchon incentivesandmarketequilibriumin
situationswith asymmetricinformationhas been a prolificpartof economic
theory.In 1996,the Bankof SwedenPrizein EconomicSciencesin Memory
of Alfred Nobel recognizedJamesMirrleesand WilliamVickrey'sfundamentalcontributionsto the theoryof incentivesunderasymmetricinformation, in particulartheirapplicationsto the designof optimalincometaxation
and resourceallocationthroughdifferenttypes of auctions. The modem
theoryof marketswith asymmetricinformationrests firmlyon the workof
this year'sprizewinners:George Akerlof (Universityof California,Berkeley), Michael Spence (StanfordUniversity)and Joseph Stiglitz (Columbia
University).This work has indeed transformedthe way economiststhink
aboutthe functioningof markets.Analyticalmethodssuggestedby Akerlof,

*Wewouldlike to thankToreEllingsen,BertilHolmlund,BertilNaslund,BernardSalani6and
LarsE. 0. Svenssonforhelpfulcomments.
? Royal Swedish Academy of Sciences. Published by Blackwell Publishers, 108 Cowley Road, Oxford OX4 IJF, UK
and 350 Main Street, Malden, MA 02148, USA.

196 K.-G. Lofgren,T Persson and J W Weibull

Spence and Stiglitz have been applied in explaining many social and
economicinstitutions,especiallydifferenttypesof contracts.1
Asymmetricinformationis a commonfeatureof marketinteractions.The
seller of a good often knows more about its qualitythan the prospective
buyer. The job applicanttypically knows more about his ability than his
potentialemployer.The buyer of an insurancepolicy usually knows more
about her individualrisk than the insurancecompany.Such asymmetries
immediatelygive rise to a numberof questions.What happensto prices,
tradedquantitiesandthe qualityof tradedgoods, if agentson one side of the
market are better informed than those on the other? What can betterinformedagentsdo to improvetheir individualmarketoutcome?Whatcan
less-informedagents do? In their work, this year's laureatesaddressed
preciselythese questions.
Theirpioneeringcontributionshave given economiststools for analyzing
a broad spectrumof issues. Abundantapplicationsrange from traditional
agriculturalmarketsto modem financialmarkets.Why are interestratesso
high in many local creditmarketsin developingcountries?Why do people
looking for a good used car typicallyturnto a dealerratherthan a private
seller? Why do some firmspay dividendseven when these are taxed more
heavilythan capitalgains?Why is it in the interestof insurancecompanies
to offer a menu of policies, which combine premiums, coverage and
deductiblesin differentways? Why do wealthy landownersnot bear the
entireharvestrisk in contractswith poor tenants?These questionsdeal with
familiar-but seeminglyvery different-phenomena.In all cases, however,
a key to the answerrelies on one and the same observation:one side of the
marketis betterinformedthanthe other.The borrowerknowsmorethanthe
lender about his creditworthiness;the seller knows more than the buyer
aboutthe qualityof his car;the CEOandboardof a firmknowmorethanthe
shareholdersaboutthe profitabilityof the firm;insuranceclientsknowmore
thanthe insurancecompanyabouttheiraccidentrisk;andtenantsknowmore
thanthe landowneraboutharvestingconditionsandtheirown workeffort.
More specifically,the contributionsof the laureatesmay be summarized
as follows. Akerlof showed how informationalasymmetriescan produce
adverse selection in markets.When lenders or car buyers have imperfect
information,borrowerswith weak repaymentprospectsor sellers of lowqualitycarsmay thuscrowdout everyoneelse fromtheirside of the market,
stifling mutuallyadvantageoustransactions.Spence demonstratedthat informed economic agents in such markets may have incentives to take
'Riley (2001) surveys the developmentsin the economicsof informationover the last 25
years,emphasizingmanyof the issueswe raisehere.
Dates in bracketsreferto publicationsby Akerlof,SpenceandStiglitz(see Bibliographies);
datesin parenthesesreferto worksby otherauthorslistedat the end of this article.
? Royal Swedish Academy of Sciences.

Marketswith asymmetricinformation 197

observableand costly actionsto crediblysignal theirprivateinformationto


uninformedagents,so as to improvetheirmarketoutcome.Cardealersmay
thus offer guaranteeson used cars to signal high quality,and the management of a firmmay incurthe additionaltax cost of dividendsso as to signal
high profitability.Stiglitz showedhow poorly informedagents can extract
informationfrom those who are better informedby offering a menu of
alternativecontractsfor a specific transaction,so-called screeningthrough
self-selection.Insurancecompaniesmay thus divide their clients into risk
classes by offering differentpolicies where, say, lower premiumscan be
exchangedfor higherdeductibles.Stiglitz also analyzeda rangeof similar
mechanismsin othermarkets,includingthe commonsharecropping
contract
betweenlandlordsandtheirbetter-informed
tenants.
Sections II throughIV of this paper give a brief account of the most
fundamentalcontributionsby each of the laureates.Section V describes
some applicationsand empirical tests of their models. Suggestions for
furtherreadingconcludethe paper.

II. George Akerlof


Akerlof'sarticle, "The Marketfor 'Lemons':QualityUncertaintyand the
MarketMechanism"[1970a], is probablythe single most importantcontributionto the literatureon economicsof information.Thispaperhas all the
typicalfeaturesof a trulyseminalpiece. It introducesa simplebutprofound
and universalidea, offers numerousinterestingimplicationsand points to
broadapplications.Nowadays,Akerlof'sinsightsregardingadverseselection
are routinelytaughtin microeconomicscoursesat the undergraduate
level.2
His essay analyzesa marketfor a productwheresellers are betterinformed
than buyersabout the qualityof the good; one example is the marketfor
used cars. Since then, "lemons" (a colloquialismfor defective cars) has
becomea well-knownmetaphorin everyeconomist'svocabulary.
Akerlof's idea may be illustratedby a simple example.Assume that a
good is sold in indivisibleunits and is availablein two qualities,low and
high, in fixed sharesA and 1 - A. Each buyer is potentiallyinterestedin
purchasingone unit, but cannot observe the differencebetween the two
qualitiesat the time of the purchase.All buyershave the same valuationof
the two qualities:one unit of low qualityis worthwL dollarsto the buyer,
while one high-qualityunit is worth wH > wL dollars.Each seller knows
the qualityof the units he sells, and values low-qualityunits at vL < wL
dollarsandhigh-qualityunitsat vH < wH dollars.
2More recently, the term "private information" or "hidden information" has become increasingly common in describing such situations. Those terms say more about the causes of the

phenomenonwhereas"adverseselection"emphasizesits consequences.
? Royal Swedish Academy of Sciences.

198 K.-G. Lofgren,T Persson and

W Weibull

If there were separate markets for low and high quality, every price
between vL and wL would support beneficial transactions for both parties in
the market for low quality, as would every price between vH and wH in the
market for high quality. Those transactions would constitute a socially
efficient outcome: all gains from trade would be realized. But if the markets
are not regulated and buyers cannot observe product quality, unscrupulous
sellers of low-quality products would choose to trade on the market for high
quality. In practice, the markets would merge into a single market with one
and the same price for all units. Suppose that this occurs and that the sellers'
valuation of high quality exceeds the consumers' average valuation. Algebraically, this case is represented by the inequality vH > w, where the
average valuation w is given by w =- AwL+ (1 - A)wH. If both qualities are
sold, consumers are willing to pay at most w. But this maximum price falls
short of vH, the minimum price at which sellers of high-quality goods are
willing to part from their units. High-quality sellers therefore leave the
market until only low-quality goods, the lemons, remain for sale.3 In other
words, Adam Smith's invisible hand is not always as effective as traditional
economics had us believe.4
In his paper, Akerlof not only explains how private information may lead
to the malfunctioning of markets. He also points to the frequency with which
such informational asymmetries occur and their far-reaching consequences.
Among his examples are social segregation in labor markets and difficulties
for elderly people in buying individual medical insurance. Akerlof emphasizes applications to developing countries. One of his examples of adverse
selection is drawn from credit markets in India in the 1960s, where local
moneylenders charged interest rates that were twice as high as the rates in
large cities. However, a middleman trying to arbitrage between these
markets, without knowing the local borrowers' creditworthiness, risks
attracting those with poor repayment prospects and becomes liable to heavy
losses.
Another fundamental insight is that economic agents' attempts to protect
themselves from the adverse consequences of informational asymmetries
may explain existing institutions. Guarantees offered by professional dealers
in the used-car market is but one of many examples. In fact, Akerlof
concludes his essay by suggesting that "this (adverse selection) may indeed
3A very earlyprototypeof Akerlof'sresultis usuallyreferredto as Gresham'slaw:"badmoney
drives out good". (ThomasGresham,1519-1579, was an adviserto Queen ElizabethI on
currencymatters.)But as Akerlof[1970a,p. 490] himselfpointsout, the analogyis somewhat
lame; in Gresham'slaw both sellers and buyerscan presumablydistinguishbetween"good"
and"bad"money.
4Classicaleconomicanalysisdisregarding
asymmetricinformationwouldmisleadinglypredict
thatgoods of both qualitieswouldbe sold on the market,at a price close to the consumers'
averagevaluation.
? Royal Swedish Academy of Sciences.

Marketswith asymmetricinformation 199

explain many economic institutions".This prophecyhas come true; his


approachhas generatedan entireliteratureanalyzinghow economicinstitutionsmaymitigatethe consequencesof asymmetricinformation.
In a laterarticle,"TheEconomicsof Casteandof the Rat-RaceandOther
WoefulTales" [1976c], Akerlofentersinto a more thoroughdiscussionof
the significanceof informationalasymmetriesin widely differingcontexts,
such as the caste system,factoryworkingconditionsand sharecropping.
He
uses illustrativeexamplesto showhow certainvariables,called "indicators",
not only provideimportantefficiency-enhancing
economicinformation,but
also
cause
the
to
become
may
economy
trappedin an undesirableequilibrium.In the case of sharecropping,
wheretenancyis repaidby a fixedshare
of the harvest,a tenant'svolume of productionacts as an indicatorof his
workefforton the farm.On the assemblyline in a factory,the speed of the
conveyorbelt acts as an indicatorof the workers'ability,and can therefore
be used as an instrumentto distinguishbetween workers of different
abilities.
Apart from his work on asymmetricinformation,Akerlof has been
innovativein enrichingeconomic theorywith insights from sociology and
social anthropology.Severalof his paperson the labormarkethave examined how emotions such as "reciprocity"towardsan employerand "fairness" towards colleagues can contributeto higher wages and thereby
see Akerlof[ 1980b, 1982c]andAkerlofandYellen[ 1990a].
unemployment;
This kind of emotionallymotivatedbehaviorhas recentlybeen confirmed
experimentally,see e.g. Fehr and Schmidt (1999, 2000), and has also
receivedempiricalsupportfrominterviewsurveys;see e.g. Bewley (1999).

III. Michael Spence


Spence'smost importantworkdemonstrateshow agentsin a marketcan use
signalingto counteractthe effects of adverseselection.In the asymmetricinformationcontext,signalingrefersto observableactionstakenby economic agents to convince the oppositepartyof the value or qualityof their
products.Spence'smain contributionswere to develop and formalizethis
idea andto demonstrateandanalyzeits implications.5A fundamentalinsight
is that signalingcan succeed only if the signalingcost differs sufficiently
among the "senders". Subsequentresearch contains many applications
whichextendthe theoryof signalingandconfirmits importancein different
markets.
Spence'sseminalpaper"JobMarketSignaling"[1973a]andbookMarket
Signaling[ 1974a]bothdealwith educationas a signalin the labormarket.If
an employercannot distinguishbetween high- and low-productivitylabor
5Informal versionsof this ideacan be tracedto the sociologicalliterature;
see Berg(1970).
? Royal Swedish Academy of Sciences.

200 K.-G. Lofgren,T Persson and J W Weibull

when hiring new workers,the labor marketmight collapse into a market


where only those with low productivityare hired at a low wage-this is
analogousto the adverse-selectionoutcomein Akerlof'smarketwhereonly
lemonsremain.
Spence'sanalysisof how signalingmayprovidea way out of this situation
can be illustratedby slightly extendingAkerlof's simple example above.
Assumefirstthatjob applicants(the "sellers")can acquireeducationbefore
workers,wL,
enteringthe labormarket.The productivityof low-productivity
is below thatof high-productivity
wH,
and
the
workers,
populationsharesof
the two groups are A and 1 - A, respectively.Although employers (the
"buyers")cannot directlyobserve the workers'productivity,they can observethe workers'educationallevel. Educationis measuredon a continuous
scale s > 0, andthe necessarycost-in termsof effort,expensesor time-to
reacheach level is lower for high-productivity
individuals.To focus on the
signalingaspect, Spence assumesthat educationdoes not affect a worker's
productivity,andthateducationhas no consumptionvaluefor the individual.
Otherthingsbeing equal,thejob applicantthuschoosesas little educationas
possible.6 Under perfect information,perfect competition and constant
returnsto scale, all applicantswould thus choose the minimaleducational
level, s = 0, and each applicantwould be paid accordingto his or her
productivity.Underasymmetricinformation,by contrast,high-productivity
workersmay acquireeducationas a signalof theirability.
To see this, assume that all employersexpect all job applicantswith at
least a certaineducationallevel s H > 0 to have high productivity,but all
othersto have low productivity.Can these expectationsbe self-fulfillingin
equilibrium?Under perfect competitionand constantreturnsto scale, all
applicantswith educationallevel s H or higherare offereda wage equal to
their expected productivity,wH, whereas those with a lower educational
level are offeredthe wage wL. Such wage settingis illustratedby the stepwise schedulein Figure1. Giventhis wage schedule,eachjob applicantwill
choose eitherthe lowestpossibleeducationsL = 0 and obtainthe low wage
wL, or the higher educationallevel s H and the higher wage w H. An
educationbetweenthese levels does not yield a wage higherthan wL, but
costs more; similarly,an educationabove sH does not yield a wage higher
thanwH, butcosts more.
In Figure1job applicants'preferencesarerepresentedby two indifference
curves, which are drawnto capturethe assumptionthat educationis less
costly for high-productivityindividuals.The flattercurve throughpoint A
thusrepresentsthose education-wagecombinations(s, w) thathigh-productivity individualsfind equallygood as their expectededucation-wagepair
6Obviously, job applicants' incentives to acquire education will be strengthened under the
more realistic assumption that education enhances productivity.
? Royal Swedish Academy of Sciences.

Marketswith asymmetricinformation 201

sH

Fig. 1. Indifference curve for low-productivityjob applicants (steep) and indifference


curve for high-productivityjob applicants(flat)

(sH, wH). All points northwest of this curve are regarded as better than this
alternative, while all points to the southeast are regardedas worse. Likewise,
the steeper curve through B indicates education-wage combinations that
low-productivity individuals find equally good as the minimum education
L = 0 and
wage wL.7
With these preferences, high-productivity individuals choose educational
level sH, neither more nor less, and receive the higher wage, as this
alternative A gives them a better outcome than alternative B. Conversely,
low-productivity individuals optimally choose the minimum educational
level at B; they are worse off with alternative A, as the higher wage does not
compensate for their high cost of education. Employers' expectations that
workers with different productivity choose different educational levels are
indeed self-fulfilling in this signaling equilibrium. Instead of a market
failure, where high-productivity individuals remain outside of the market
(e.g., by moving away or setting up their own business), these workers
participate in the labor market and acquire a costly education solely to
distinguish themselves from low-productivityjob applicants.
7Thecrucialassumptionthatmoreproductiveapplicantsfindit less costly (by any margin)to
increasetheireducationlevel-the flatterindifferencecurvein Figure1 is closely relatedto
Mirrlees's(1971) so-calledsingle-crossingcondition.A similarconditionis foundin numerous
contextsin modem microeconomictheoryand is often referredto as the Mirrlees-Spence
condition.
,

Royal Swedish Academy of Sciences.

202 K.-G. Lofgren,T Persson and J. W Weibull

Absent furtherconditions,there is a whole continuumof educational


levels sH with correspondingsignalingequilibria.In each of these,however,
incentivecompatibilityrequiresthat the expectedlevel of educationnot be
so high that high-productivityindividualspreferto refrainfrom education,
or so low thatlow-productivity
applicantspreferto educatethemselvesup to
thatlevel. Geometrically,these conditionsimplythatpointslike B lie below
the indifferencecurve of high-productivityindividualsthroughany equilibriumpointcorrespondingto A, andpointslike A lie below the indifference
individualsthroughpointB.
curveof low-productivity
indicates
that a certain signaling equilibriumis the
Spence [1973a]
individualsopt
In
this
most
efficient.
equilibrium,high-productivity
socially
for (and are expected to do so by employers)the minimumeducationto
distinguishthemselves from those with low productivity.In other words,
high-productivityworkers choose the combination given by point C
in Figure 1. Low-productivityworkers are then indifferentbetween the
education-wage combination(?, wH) at point C and the combination
individuals,conversely,
(0, wL) at their chosen point B. High-productivity
prefer point C to B. Riley (1975) showed that this is the only signaling
by employers.Spence's
equilibriumwhich is robustto wage experimentation
signalingmodel also spurreda flurryof game-theoreticresearch.In particular, variousrefinementsof the Nash equilibriumconcepthave been developed to discriminatebetween the many signaling equilibriain Spence's
model.Manyof these refinementsselect the sociallymost efficientsignaling
equilibrium.An influentialpaperin this genreis Cho andKreps(1987).
Spence[1973a, 1974a]also demonstratesthe existenceof otherequilibria,
e.g. one where no applicantacquireseducation.Assume that employersdo
not expect educationto be a productivitysignal, i.e., they expect all job
applicants,regardlessof education,to have the averageproductivityon the
market: w = twL + (1 - A)wH. Employers then offer this wage to all job

applicants,and their expectationsare self-fulfilling,as it is optimalfor all


applicantsto choose the minimumlevel of educationsL = 0. Spence also
notes the possibilityof equilibriawhere differentgroupsof applicantshave
differenteducationalincentiveseven in the absence of innate differences
between the groups.For example, high-productivitymen (blacks)may be
expected to acquire another level of education than equally productive
women (whites).In such equilibria,the returnsto educationdifferbetween
men andwomen,or blacksandwhites,as do theirinvestmentsin education.
Apart from his work on signaling, Spence has made distinguished
contributionsto the field of industrialorganization.Duringthe period19751985, he was one of the pioneersin the wave of game-theoryinspiredwork
withinthe so-callednew industrialorganizationtheory.His most important
studiesin this areadeal with monopolisticcompetition[1976d] and market
entry [1977c]. Spence'smodels of marketequilibriumundermonopolistic
? Royal Swedish Academy of Sciences.

Marketswith asymmetricinformation 203


competition have also been influential in other fields, such as growth theory
and internationaltrade.
IV. Joseph Stiglitz
Stiglitz's classical article with Rothschild on adverse selection, "Equilibrium
in Competitive Insurance Markets:An Essay on the Economics of Imperfect
Information" [1976d], is a natural complement to the analyses in Akerlof
[1970a] and Spence [1973a, 1974a].8 Rothschild and Stiglitz ask what
uninformed agents can do to improve their outcome in a market with
asymmetric information. More specifically, they consider an insurance
market where companies do not have information on individual customers'
risk situation. The (uninformed) companies offer their (informed) customers
different combinations of premiums and deductibles and, under certain
conditions, customers choose the policy preferred by the companies. Such
screening through self-selection is closely related to Vickrey (1945) and
Mirrlees's (1971) analyses of optimal income taxation, where a tax authority
(unaware of private productivities and preferences) gives wage earners
incentives to choose the "right" amount of work effort.9
Rothschild and Stiglitz's model may be illustrated by means of a simple
example. Assume that all individuals in an insurance market are identical,
except for the probability of damage. Initially, all individuals have the same
income y. A high-risk individual incurs a loss of income d < y with
probability pH and a low-risk individual suffers the same loss of income
with the lower probabilitypL, with 0 < pL < pH < 1. In analogy with
Akerlof's buyer and Spence's employer, who do not know the sellers' quality
or the job applicants' productivity, the insurance companies cannot observe
the individual policy-holders' risk. From the perspective of an insurance
company, policy-holders with a high probability pH of injury are of "low
quality", while policy-holders with a low probability pL are of "high
quality". In analogy with the previous examples, there is perfect competition
in the insurance market, and insurance companies are risk neutral (cf. the
earlier assumption of constant returns to scale).10 An insurance contract
(a, b) specifies a premium a and an amount of compensation b in the case of
income loss d-the deductible is thus the difference d - b. Under perfect
information, high-risk individuals pay the actuarially fair premium
8Salopand Salop (1976) similarlyanalyzehow firmscan use self-selectionwhen employing
workerswithprivateinformationabouttheirpropensityto quit.
9Stiglitz[1975e] actuallyused the word "screening",but addressedwhat is todayknownas
signaling.Stiglitzrefersto Arrow(1973) and Spence[1973a],while discussingandextending
theirideas.
10Stiglitz[1977d]providesan analysisof the monopolycase.
? Royal Swedish Academy of Sciences.

204 K.-G. Lofgren,T Persson and . W Weibull


aH = pHd, low-risk individuals pay the actuarially fair premium aL

pLd,

and both groups are fully compensated in case of damage, b = d (competition drives down premiums and deductibles to their minimal levels).
In the case of asymmetric information, Rothschild and Stiglitz establish
that equilibria may be divided into two main types: pooling and separating.
In a pooling equilibrium, all individuals buy the same insurance, while in a
separating equilibrium they purchase different contracts. Rothschild and
Stiglitz show that no (pure-strategy) pooling equilibrium exists in their
model. The reason is that in such an equilibrium an insurance company
could profitably cream-skim the market by instead offering a contract that is
better for low-risk individuals but worse for high-risk individuals. Whereas
in Akerlof's model the price became too low for high-quality sellers, here
the equilibrium premium would be too high for low-risk individuals. The
only potential equilibrium in the Rothschild-Stiglitz model is a unique
separating equilibrium, where two distinct insurance contracts are sold in the
market. One contract (aH, bH) is purchased by all high-risk individuals, the
other (aL, bL) by all low-risk individuals. The first contract provides full
coverage at a relatively high premium: aH > aL and bH = d, while the
second contract combines the lower premium with only partial coverage:
bL < d. Consequently, each customer chooses between one contract without
any deductible, and another contract with a lower premium and a deductible.
In equilibrium, the deductible barely scares away the high-risk individuals,
who are tempted by the lower premium but choose the higher premium in
order to avoid the deductible.
This unique separating equilibrium corresponds to the socially most
efficient signaling equilibrium, point C of Figure 1 in the simple illustration
of Spence's model.11 In both models, "good types" incur a cost in equilibrium; high-productivity workers have to educate themselves beyond those
who are less productive, and low-risk insurance clients do not obtain full
coverage, respectively. As noted above, Rothschild and Stiglitz also identify
conditions under which no (pure-strategy)equilibrium exists-non-existence
is a feature not shared with Spence's model.12 The fact that at most one
equilibrium exists is typical of screening models, as is the correspondence

llRiley's (1975) robustnesstest, with respectto experimentingemployers,led to the same


equilibriumas in Spence'smodel. In fact, Riley'sidea is not wholly unlikethatof Rothschild
andStiglitz[1976d].However,accordingto Riley (2001, p. 438), RothschildandStiglitzmade
"... a more radicaldeparturefrom Spence'sanalysisby proposingthatthe model shouldbe
viewedas a non-cooperative
gamebetweenthe consumers".
12Thenon-existenceproblem has spurredsome theoreticalresearch.Wilson (1977), for
example,suggestsa less stringentdefinitionof equilibrium,basedon the ideathatunprofitable
contractscan be withdrawn.This renderscertainotherwiseprofitabledeviationsunprofitable
andmakesexistencemorelikely.
? Royal Swedish Academy of Sciences.

Marketswith asymmetricinformation 205

between the unique screeningequilibriumand the socially most efficient


signalingequilibriumin the associatedsignalingmodel.
Rothschildand Stiglitz'sarticle has been very influential.In particular,
their classification of equilibria has become a paradigm;pooling and
separatingequilibriaare now standardconceptsin microeconomictheoryin
generalandin informationeconomicsin particular.
But this article is only one of Stiglitz's many contributionsregarding
markets with asymmetric information.He is probably the most cited
researcherwithinthe informationeconomicsliterature-perhapsalso within
a wider domainof microeconomics.In his largeproduction,often with coauthors,Stiglitzhas time and againpointedout that economicmodels may
be quite misleading if they disregardinformationalasymmetries.The
message has been that many marketstake on a different guise in the
perspectiveof asymmetricinformation,as do the conclusionsregardingthe
appropriateforms of public-sectorregulation.Several of his essays have
becomeimportantsteppingstonesfor furtherresearch.
Two papers co-authoredby Stiglitz and Weiss [1981c, 1983d] analyze
credit markets.13Stiglitz and Weiss show that to reduce losses from bad
loans,it maybe optimalfor imperfectlyinformedbanksto rationthe volume
of loansinsteadof raisingthe lendingrate,as wouldbe predictedby classical
economicanalysis.Since creditrationingis so common,these insightswere
importantstepstowardsa morerealistictheoryof creditmarkets.Theyhave
had a substantialimpactin the fields of corporatefinance,monetarytheory
andmacroeconomics.
Stiglitz's work with Grossman[1980c] investigatesthe hypothesis of
efficiency on financial markets. It introducesthe so-called GrossmanStiglitzparadox:if a marketwere informationally
efficient-i.e., all relevant
informationis reflectedin marketprices-no agentwouldhave an incentive
to acquirethe informationon which prices are based. But if everyoneis
uninformed,thenit pays some agentto becomeinformed.Thus,an informationallyefficientequilibriumdoes not exist. This workhas exertedconsiderable influencein financialeconomics.
Stiglitz has also proposedan information-based
explanationof involuntary unemployment.A widely cited articleby Shapiroand Stiglitz [1984c]
modelwith so-calledefficiencywages.14By definidevelopsa labor-market
tion, an efficiencywage exceedsa worker'sreservationwage (thewage level
which makeshim indifferentbetweenremainingon the job or quitting)and
thusgives workersincentivesto performwell (moreefficiently)to keeptheir
13StiglitzandWeissalso studymoralhazard,a conceptalreadyusedby Arrow(1963) to refer
to situationswherean economicagentcannotobservesome relevantactionof anotheragent
aftera contracthas been signed.
researchwith similarideasis reportedin BowlesandBoyer(1988).
14Concurrent
(

Royal Swedish Academy of Sciences.

206 K.-G. Lofgren,T Persson and J W Weibull

jobs. In Shapiroand Stiglitz'smodel, an employeris assumedto carryout


randomsurveysamonghis employeesto observetheirworkeffort.A worker
caughtshirkingis firedand ends up with his reservationwage (by looking
for anotherjob or settingup his own business),a level lowerthanif he had
refrainedfrom shirkingand instead kept his job at the prevailingwage.
Optimalbehaviorof both employersand employeesresults in equilibrium
unemployment.Shapiroand Stiglitz'smodel is an importantingredientin
modemlaborandmacroeconomics.
Stiglitz's researchhas also been crucial for the emergenceof modem
developmenteconomics. He has shown that economic incentives under
asymmetricinformationare not merely academicabstractions,but highly
concretephenomenawith far-reachingexplanatoryvalue in the analysisof
institutionsand marketconditionsin developingeconomies.One of his first
studiesof informationalasymmetries[1974g] deals with sharecropping,an
ancient but still common form of contracting.As the term implies, the
contractregulateshow the harvestshouldbe dividedbetweena landowner
and his tenants. The size of a harvest generally depends on external
circumstancessuch as weatherand on the tenants'work effort. Underthe
conventionalassumptionthat absoluterisk aversionis decreasingin wealth,
the optimaloutcomewould be to let the richerparty(here,the landowner)
bearthe entirerisk. In practice,however,the harvestis dividedup between
the partiesaccordingto fixed shares,usuallyhalf each. Stiglitz [1974g] and
Akerlof [1976c] both attemptedto explainthis relationin terms of asymmetric informationbetween the two parties. Since the landownerusually
cannot observe tenants'work effort, an optimalcontractstrikesa balance
betweenrisk sharingand incentives,lettingthe tenantsassumesome share
of the risk.
In additionto his workon the economicsof information,Stiglitzhas made
significant contributionsto public economics, especially the theory of
optimaltaxation,as in AtkinsonandStiglitz[1976b],industrialorganization,
as in Dixit andStiglitz[1977c],andthe economicsof naturalresources,as in
Stiglitz[1974e] andDasguptaandStiglitz[1980g].

V. Applications and Evidence


Akerlof,SpenceandStiglitz'sworkon marketsandinformationasymmetries
is fundamentalto modem microeconomictheory.Theirresearchhas helped
us understandmarketphenomenawhich could not be fully capturedby
traditionalneoclassicaltheory.Moreover,their models have been used to
explainthe emergenceof many social institutionsthat counteractthe negative effects of informationalasymmetries.The range of applicationis
remarkable:from financialmarkets,throughindustrialorganization,all the
way to issues in economic development.This section offers a selection of
? Royal Swedish Academy of Sciences.

Marketswith asymmetricinformation 207

such applicationsfrom recent research and a brief discussion of some


empiricaltests of the models.
In financialeconomics, e.g. Myers and Majluf (1984) have shown how
shareholderscan becomevictimsof adverseselectionamongfirms.In a new
sector(suchas today'sIT) most firmsmay appearidenticalin the eyes of an
uninformedinvestor,while some insidersmay havebetterinformationabout
the futureprofitabilityof such firms. Firmswith less than averageprofitability will thereforebe overvaluedby the stock marketwhere, of course,
uninitiatedinvestorsalso trade.Such firmswill thereforepreferto finance
new projectsby issuingnew shares(as opposedto debt).Firmswith higher
thanaverageprofitability,on the otherhand,will be undervaluedand find it
costly to expandby shareissue. Underasymmetricinformation,the "lowquality"firms(with low futureprofitability)thustend to growmorerapidly
than "high-quality"firms, implying that the market will graduallybe
dominatedby "lemons".Whenuninitiatedinvestorsultimatelydiscoverthis,
sharepricesfall (the IT bubblebursts).
Another puzzle in financial economics is why some firms choose to
distributedividendsto theirshareholders,even if dividendsaremoreheavily
taxed (due to double taxation)than capitalgains, as has been the case in
some countries.A cheaperalternativewould be to retainthe profitswithin
the firm and favor shareholdersby way of capital gains througha higher
shareprice.JohnandWilliams(1985) showthat,underasymmetricinformafirmon
tion, dividendscan act as a crediblesignalfor a "high-profitability"
the stock market.Firmswith positive insiderinformationpay dividendsto
theirshareholders,but this signal is too costly for firmswith inferiorinsider
information.The stock marketthus interpretscostly dividendsas a credible
signal for favorableprospectsand thereforepays a high price for the stock.
Undercertainconditions,the shareprice rises enoughto compensateshareholders for the extra tax they have to pay on dividends-a separating
equilibriumis achieved.
In the sphere of industrialorganization,numerousapplicationshave
shownhow consumersmay interpretprice settingand advertisingas signals
for good quality.As in Spence's[1973a, 1974a] model of a labor market
with high- and low-productivityworkers,equilibriacan arise when it is
profitablefor firmswith high-qualityproductsto engage in costly advertising, whereasfirmswhichproducelow-qualitygoods refrain;see e.g. Nelson
(1974) andMilgromandRoberts(1986). Tirole(1988) providesan extensive
overview of other applicationsof informationeconomics in the field of
industrialorganization.
In labor economics, Waldman(1984) examinesa situationwhere firms
competingfor laboruse the job assignmentof a competitor'semployeeas a
signal of his ability.Because an employerdoes not want to signal the true
capacityof a good employeeto potentialcompetitors,employeesmightnot
? Royal Swedish Academy of Sciences.

208 K.-G. Lofgren,T.Persson and J W Weibull

necessarilybe assignedtaskswhichmaximizetheircontributionto the firm's


profits. Such allocation of labor within firms might be optimal for an
individualfirm in a labor-competitivesituation,but resultsin social inefficiency.

Bernhardt(1995) develops these argumentsinto an analysis of promotions, explainingwhy low-educationemployeespromotedto high positions
are usually extraordinarilycapable. An employerwho wants to hide his
private informationabout employees from a competingemployerhas an
incentivenot to promotecompetentworkers.For a promotionto be profitable, a low-educatedworker thereforehas to be sufficientlycapable to
compensatefor the higherwage the firmis forcedto pay to retaina worker
whose competenceis revealedto potentialcompetitors.Similarmechanisms
can also explainwage discrimination.Milgromand Oster(1987) point out
that such discriminationleads to social inefficiency when workers are
assignedto the wrongjobs or are not given sufficientincentivesto become
bettereducated.
Riley (1979) makesan earlyattemptto empiricallytest Spence'ssignaling
model. Riley's idea is that signaling should be most importantin those
sectorsof the economywhereworkerproductivityis difficultto measure.In
suchsectors,wages andeducationarethusexpectedto be stronglycorrelated
at the outset of a worker'scareer,whereasthe correlationshouldbe weaker
in sectorswhere productivityis more easily observed.Over time, as firms
learn more about the productivityof their employees,the correlationbetween wages and educationshould become weaker,particularlyin sectors
where productivityis hard to measure.Riley was able to confirmthese
effects empirically.More recent tests of Spence's signaling model were
carriedout by LangandKropp(1986) andBedard(2001). Both studiesshow
thatdataon high-schoolenrollmentsand dropoutratesare consistentwith a
model.
signalingmodelandinconsistentwith a purehuman-capital
In their empiricalanalysis of firing on a labor marketwith asymmetric
information,GibbonsandKatz(1991) test the relevanceof adverseselection
and signaling.If firms can freely decide which employeesshouldbe fired,
other agents on the labor marketwill conclude that the ability of fired
workersis below average(they are "lemons").Workerswho are alike in all
other(measurable)respects,but who hadto leave theirjobs becausethe firm
closed down,shouldthus find it easierto get a newjob andreceive a higher
wage. Basedon a largesampleof redundantworkers,GibbonsandKatzfind
empiricalsupportfor thesepredictions.
Farberand Gibbons(1996) developedSpence'ssignalingmodelby allowing employersto obtain informationon workerproductivityby observing
their careers. The model predicts that the wage effect of education is
independentof the length of time a workerhas been on the labormarket,
which are
whereasthe wage effect of constant,unobservablecharacteristics,
? Royal Swedish Academy of Sciences.

Marketswith asymmetricinformation 209


positively correlated with worker ability, increases with the time a worker
has been employed. Both predictions are consistent with data regarding
young people on the US labor market.
Acemoglu and Pischke (1998) show that asymmetric information about
worker ability can explain on-the-job training in firms. The mechanism
resembles that in Waldman (1984) and Gibbons and Katz (1991). Informational asymmetries concerning a trained worker's productivity generate a
monopsony (a buyer monopoly) on the local labor market, implying that the
firm can successively pay for training by a wage which falls short of the
competitive wage. The predictions are empirically supported when confronted with data from the German apprentice system.
Other attempts to test for the predicted effects of asymmetric information
have produced ambiguous results. One difficulty with such tests is to
distinguish, in practice, between adverse selection and moral hazard;another
is that screening and signaling partially eliminate the effects of informational
asymmetries.15
In recent years, many insights from the economics of information have
been incorporated into development economics. It is perhaps not so surprising that models suggested by Akerlof and Stiglitz have had a large influence
in this field, as their early studies were largely inspired by issues in development economics. Prime examples are Akerlof's lemons model and Stiglitz's
sharecropping model. Extensions of the latter have been used to explain
institutional relationships between landowners and tenants, such as why
landowners often grant credit to tenants (it has positive incentive effects on
work effort). Arguments based on asymmetric information have also been
used to clarify the dichotomy between modem and traditional sectors in
developing economies. Basu (1997) is an example of a modem advanced
textbook in development economics that builds heavily on the economics of
information.

VI. Suggested Reading


The laureates' own original works remain highly recommended reading: see
e.g. Akerlof [1970a, 1976c], Spence [1973a, 1974a], Rothschild and Stiglitz
[1976d], Stiglitz and Weiss [1981c] and Shapiro and Stiglitz [1984c]. Riley
15Adirecttest carriedout by Bond(1982) on datafroma marketfor second-handsmalltrucks
does not lend supportto the asymmetricinformationhypothesis.Dahlby(1983, 1992) finds
some supportfor adverseselectionusingaggregatedataon Canadiancarinsurance.In a study
of data froma car insurancecompany,Puelz and Snow (1994) find supportfor both adverse
selectionand signaling.Chiapporiand Salani6(2000a)examinewhetherindividualswho buy
car insurancewith bettercoveragehave more accidents.They are unableto find statistical
supportfor such a correlation.For a surveyof tests of asymmetricinformationmodels, see
ChiapporiandSalanie(2000b).
? Royal Swedish Academy of Sciences.

210

K.-G. L6fgren, T Persson and J W. Weibull

(2001) gives a detailed survey of economic analyses of markets with


asymmetric information. Salanie (1997) provides an introduction to asymmetric information modeling in the context of contract theory. Gibbons
(1992) offers an accessible introduction to game-theoretic modeling of
asymmetric information. A more advanced introductionto adverse selection,
signaling and screening can be found in Chapter 13 of Mas-Colell, Whinston
and Green (1995).

References
Acemoglu, D. and Pischke, S. (1998), Why Do Firms Train? Theory and Evidence,
QuarterlyJournal of Economics 113, 79-119.
Arrow,K. (1963), Uncertaintyand the Welfare Economics of Medical Care, American
Economic Review 53, 941-973.
Arrow,K. (1973), HigherEducationas a Filter,Journalof Public Economics2, 193-216.
Basu, K. (1997), AnalyticalDevelopmentEconomics,MIT Press, Cambridge,MA.
Bedard,K. (2001), HumanCapitalversus Signaling Models, UniversityAccess and High
School Dropouts,Journal of Political Economy 109, 749-775.
Berg, I. (1970), Educationand Jobs: The Great TrainingRobbery,Praeger,New York.
Bemhardt, D. (1995), Strategic Promotion and Compensation, Review of Economic
Studies 62, 315-339.
Bewley, T. (1999), WhyWagesDon't Fall During a Recession, HarvardUniversityPress,
Cambridge,MA.
Bond, E. (1982), A Direct Test of the Lemons Model: the Marketfor Used PickupTrucks,
AmericanEconomicReview 72, 836-840.
Bowles, S. and Boyer, R. (1988), Labor Discipline and Aggregate Demand, American
Economic Review 78, 395-400.
Cho, I.-K. and Kreps, D. (1987), Signaling Games and Stable Equilibria, Quarterly
Journal of Economics 102, 179-221.
Chiappori,P. and Salanie, B. (2000a), Testing for Asymmetric Informationin Insurance
Markets,Journal of Political Economy 107, 56-78.
Chiappori,P. and Salanie, B. (2000b), TestingContractTheory:a Surveyof Some Recent
Work,Invitedlecture, WorldCongressof the EconometricSociety, Seattle.
Dahlby, B. (1983), Adverse Selection and Statistical Discrimination, An Analysis of
CanadianAutomobileInsurance,Journal of Public Economics 20, 121-130.
Dahlby, B. (1992), Testing for Asymmetric Informationin CanadianAutomobile Insurance, in G. Dionne (ed.), Contributionsto InsuranceEconomics,Kluwer,Boston, MA.
Farber,H. and Gibbons, R. (1996), Learningand Wage Dynamics, QuarterlyJournal of
Economics 111, 1007-1047.
Fehr, E. and Schmidt, K. (1999), A Theory of Fairness, Competitionand Cooperation,
QuarterlyJournalof Economics 114, 817-868.
Fehr, E. and Schmidt, K. (2000), Theories of Fairness and Reciprocity: Evidence and
Economic Applications, forthcoming in M. Dewatripontet al. (eds.), Advances in
Economic Theory, Eighth WorldCongress of the Econometric Society, Cambridge
UniversityPress, Cambridge,UK.
New York.
Gibbons,R. (1992), A Primer in Game Theory, Harvester-Wheatsheaf,
Gibbons, R. and Katz, L. (1991), Layoffs and Lemons, Journal of Labor Economics 9,
351-380.
?

Royal Swedish Academy of Sciences.

Markets with asymmetric information

211

John, K. and Williams, J. (1985), Dividends, Dilution and Taxes, a Signalling Equilibrium,Journalof Finance 40, 1053-1069.
Lang, K. and Kropp,D. (1986), HumanCapitalversus Sorting:the Effects of Compulsory
AttendanceLaws, QuarterlyJournal of Economics 101, 609-624.
Mas-Colell, A., Whinston, M. and Green, J. (1995), Microeconomic Theory, Oxford
UniversityPress, Oxford.
Milgrom, P. and Roberts, J. (1986), Price and Advertising Signals of Product Quality,
Journal of Political Economy94, 795-781.
Milgrom, P. and Oster, S. (1987), Job Discrimination,MarketForces, and the Invisibility
Hypothesis,QuarterlyJournalof Economics 102, 453-476.
Mirrlees,J. (1971), An Explorationin the Theory of OptimumIncome Taxation,Review
of EconomicStudies 38, 175-208.
Myers, S. and Majluf, N. (1984), CorporateFinancing and InvestmentDecisions when
Firms Have Informationthat InvestorsDo Not Have, Journal of Financial Economics
13, 187-221.
Nelson, P. (1974), Advertising as Information,Journal of Political Economy 82, 729754.
Puelz, R. and Snow, A. (1994), Evidence on Adverse Selection: EquilibriumSignalling
and Cross-Subsidizationin the InsuranceMarket,Journal of Political Economy 102,
236-257.
Riley, J. (1975), CompetitiveSignalling,Journal of Economic Theory10, 174-186.
Riley, J. (1979), Testing the Educational Screening Hypothesis, Journal of Political
Economy87, 227-252.
Riley, J. (2001), Silver Signals, Twenty-FiveYearsof Screeningand Signaling,Journal of
EconomicLiterature39, 432-478.
Salanie,B. (1997), TheEconomicsof Contracts,a Primer,MIT Press, Cambridge,MA.
Salop, J. and Salop, S. (1976), Self-selection and Turnoverin the LaborMarket,Quarterly
Journal of Economics 90, 619-627.
Tirole,J. (1988), The Theoryof IndustrialOrganization,MIT Press, Cambridge,MA.
Vickrey,W. (1945), MeasuringMarginalUtility by Reactions to Risk, Econometrica13,
319-333.
Waldman, M. (1984), Job Assignments, Signalling, and Efficiency, Rand Journal of
Economics 15, 255-267.
Wilson, C. (1977), A Model of InsuranceMarketswith IncompleteInformation,Journal
of Economic Theory16, 167-207.

? Royal Swedish Academy of Sciences.

Você também pode gostar