Professional Documents
Culture Documents
Levy Prospect Theory RC and IR
Levy Prospect Theory RC and IR
Blackwell Publishing and The International Studies Association are collaborating with JSTOR to digitize,
preserve and extend access to International Studies Quarterly.
http://www.jstor.org
S.
LEVY
RutgersUniversity
A half-decadeafterthe firstsystematic
applicationsof prospecttheoiyto
international
relations,scholarscontinueto debate itspotentialutilityas a
oftheexperimental
framewvork.
Keyquestionsincludethevalidity
theoretical
international
behaviorthat
theirrelevanceforreal-world
themselves,
findings
and
decisionsbycollectiveactorsin interactive
settings,
involveshigh-stakes
withrespecttorationalchoice.In this
theconceptualstatusofprospecttheory
debatesovertheseissues.I review
and methodological
essayI assesstheoretical
workin social psychologyand experimentaleconomicsand conclude that
challengesto the externalvalidityof prospecttheoiy-basedhypothesesfor
behavioraremuchmoreseriousthanchallengestotheirinternal
international
betweenprospecttheoryand expectedvalidity.I emphasizethesimilarities
regardinglossaversionand thereflection
theory,
arguethathypotheses
utility
effectare easilysubsumedwithinthe latter,and thatevidenceof fi-aming
are moreproblematicforthe
effects
and nonlinearresponsesto probabilities
forfutureresearchincludetheconstruction
theory.I concludethatpriorities
ofhypotheses
on theframing
offoreignpolicydecisionsand researchdesigns
fortestingthem;theincorporation
lossaversion,and thereflection
offraming,
decisionmaking;and experieffect
intotheoriesofcollectiveand interactive
tothepoliticaland strategic
contextofforeign
mentalresearchthatissensitive
policydecisionmaking.
It is ironic thatjust as rational choice has become the most influential paradigm in
international relations and political science over the last decade, expected-utility
theoiy has come under increasing attack by experimental and empirical evidence
of systematicviolations of the expected-utilityprinciple in individual-choice behavior. Experimental evidence suggests that people tend to evaluate choices with
respect to a reference point, overweight losses relative to comparable gains, engage
in risk-aversebehavior in choices among gains but risk-acceptantbehavior in choices
among losses, and respond to probabilities in a nonlinear manner. Many of these
patterns have been confirmed by field studies of consumer, investment, and insurance behavior. This challenge to expected-utility theory has been reinforced by the
development of prospect theory (Kahneman and Tversky, 1979), which integrates
these descriptive patterns into an alternative theory of riskychoice.
Ration-aland Cognitive
Autihor'snote:Earlierversionsof thisarticlewere presentedat a conferenceon-"Con-trasting
(1994), at thePeace ScienceSociety(1995), and at colloquiaat Duke Univelrsity,
TheoriesofWar"atTexas A & M University
ofCaliforniaat San Diego, and at Columbia
theUniversity
ofCalifomiaat Berkeley,the University
Ohio StateUniiversity,
I would also like to thankJeffrey
Berejikian,Bill Boettcher,Bill Clark,PatrickJames,Rose McDermott,Cliff
University.
fortheirhelpfu-l
comments.
Mor-gan,
MichaelSinatra,NancySpaulding,JaniceStein,and HarrisonWagner? 1997 Initertnational
StudiesAssociation.
350 MaisiStreet,Malden,MA 02148,USA, and 108 CowleyRoad,OxfordOX4 lJF,UK.
PuLblishecd
byBlackwellPublisliers,
88
ProspectTheory,Rational Choice
JACK S. LEVY
89
90
Thzeory,
RationalChoice
Prospect
Furtherevidence in supportof the hypothesesof loss aversionand the endowment effectcan be found in the observed tendencytoward statusquo choices.
Because people treatthecostsofmovingawayfromthe statusquo as losses and the
benefitsof moving away fromthe statusquo as gains, and then overweightthe
formerrelativeto the latter,people stayat the statusquo more frequentlythan
expected-utility
theorywould predict. This statusquo bias is demonstratedin a
number of experimentaland field studies of consumerand investmentbehavior
(Knetsch and Sinden, 1984; Samuelson and Zeckhauser, 1988; Knetsch, 1989;
Hartman,Doane, and Woo, 1991).
The asymmetry
betweenlosses and gains is manifestednot onlyin loss aversion
and the endowmenteffectbut also in riskorientation.
People tend to be risk-averse
withrespect to gains and risk-acceptantwithrespect to losses.8 This means that
individualvalue functionsare usuallyconcave in the domain of gains and convex
in thedomain oflosses,witha reflection
around thereferencepoint(Kahneman
effect
and Tversky,1979:268). This patternof riskorientationsis repeatedlyfoundfora
varietyof individualsand situations,but it may break down forverysmall probabilitiesor for catastrophiclosses (Slovic, Fischhoff,Lichtenstein,Corrigan,and
Combs, 1977; Fishburn and Kochenberger, 1979; Kahneman and Tversky,
1979:282-3; Slovic and Lichtenstein,1983; Tverskyand Kahneman, 1986; Quattroneand Tversky,1988).9
Because of the asymmetry
of gains and losses and the role of the referencepoint
in definingthese distinctdomains, the identification,
orframing,of the reference
point can have a criticaleffecton choice. A change in framecan resultin a change
in preferences(preference
reversal)even ifthevalues and probabilitiesassociatedwith
outcomesremain the same. This has repeatedlybeen demonstratedin laboratory
experiments(Gretherand Plott,1979; Kahneman and Tversky,1979; Slovic and
Lichtenstein,1983; Tverskyand Kahneman, 1986; Tversky,Slovic,and Kahneman,
1990; Camerer, 1995:652-65; Roth, 1995:68-75). In choices betweenalternative
medical treatmentprocedures, for example, it makes a differencewhether a
particulartreatmenthas a 90 percentsurvivalrateor a 10 percentmortality
rate.10
In many simple choice problems the framingof the referencepoint is largely
predeterminedby the situation(or the experimentaldesign). In a staticsituation
8 Loss aver-sion-i
and r-isk
or-ientation
are analyticallydistinct.Loss aver-sion
is r-eflected
in the steepnessof thevalue
on the loss side, whereas r-isk
orientationis reflectedin the curvatureof the value function.The conceptsof
f'unction
loss aversion,the erndowment
effect,and the statusquo bias can all be applied to choice behaviorunder conditionsof
cer-tainty
wher-e
thereare no r-isky
gamblesas wellas to choicesunder-i-isk.Tverskyand Kahneman(1991), forexample,
develop a model of loss aversionin risklesschoice thathas nothingto do withlotter-ies.
9There is a potential levels-of-an-alysis
problem that requir-esfur-therexploration. Experimentaltestson r-isk
in the domains of losses an-dgains gener-ally
pr-opernsities
utilize "between-subjects"
designs,so that distinctsets of
subjectsar-etestedfor-choices in each domain. The concernis thatsin-glesubjectsin a "within-subjects"
design might
in r-epeatedchoices.
moreeasilyr-ecognizethe equlivalenceofchoice problemsand ther-efore
behave more consistently
Consequently,aggregateresultsfrombetween-subjects
designsdo not logicallyimplya reflectioneffectin individuals,
and we do not knowhow manyof the 70 per-cent
who are riskaverse forgains belon-gto the 70 percentwho are risk
stillexist
acceptantforlosses.Experimentsbased on an individual-level
analysisshowthattendenciestowar-d
reflectivity
but thattheyare weaker(He-sheyand Schoeniaker,1980; Camnerer,
1995:633).
10The preferencereversalphenomenonalso createsa potentiallyseriousmethodologicalpr-oblem
forutilitytheoi-y
because it raises the question of "elicitationbiases": differentmethods of elicitingprefer-ences
generate different
and thusdifferent
r-esponses
preferenceorders(Camerer-,1995:625-6). This is evidentfromthefollowingexperiment:
Consider twogamblesthatare mathematically
equivalentin termsofexpectedvalue: one involvesa highprobabilityof
a modestamountof money(the P lottery)and the otherinvolvesa low pr-obability
winn-ing
ofwinninga largeramount
thatif
of money(the $ lottei-y).
Experimentalstudies(by economistsas well as psychologists)repeatedlydemonstr-ate
asked to choose betweenthe two,mostpeople preferthe P lottery.However,iftheyare asked to give the lowestp-ice
at whichtheywould sell each lotteryif theyowned both, most people put a higherprice or-reser-vationlevel on the
$ lotter-y
(Gr-etherand Plott,1979; Slovic and Lichtenstein,1983; Tver-sky,
Slovic,and Kahneman, 1990; Tverskyan-d
Thaler, 1990; Caimler-er,
1995:657-65). Apparentlychoices are correlatedwithprobabilitieswhilepricesare correlated
JACK S. LFVY
91
92
RationalChoice
Prospect
Theory,
expected-utility
theorypositsthatutilitiesand probabilitiesare givenequal weight.
Althoughsmall probabilitiesare overweighted,extremelysmall probabilitiesgenerateunpredictablebehavior.I I Extremelyunlikelyeventssometimesare treatedas
if theywere impossible and at other times are overweightedin actors' decision
calculations (Kahneman and Tversky,1979:282-3), as illustratedby the unpredictabilityof insurancebehaviorwithrespectto rare catastrophes.
Kahneman and Tversky(1979) developed prospect theoryto integratethese
observedpatternsinto an alternativetheoryof riskychoice. They distinguishtwo
phases in the choice process. In the editing
phasethe actor identifiesthe reference
point,the available options,the possible outcomes,and the value and probability
ofeach oftheseoutcomes.In theevaluation
phaseshe combinesthevalues ofpossible
outcomes (as reflectedin an S-shaped valuefunction)with theirweightedprobabilities(as reflectedin theprobability
weighting
function)and then maximizesover
'2
the product(the "prospectiveutility",).
Note that attitudestoward risk are determined by the combination of the
S-shaped value functionand theprobability
weightingfunctionand notbythevalue
functionalone. Althoughthiscombinationusuallygeneratesriskaversionforgains
of smallprobabilitiescan trigger
and riskacceptance forlosses,the overweighting
a reversalof riskpropensitiesunder certainconditions,depending on the precise
shapes of the two functions(Kahneman and Tversky,1979; Levy, 1992a: 183-4).
Thus, the existenceof smallprobabilitiesis a necessarybut not sufficient
condition
forriskaversionin the domain oflosses and riskacceptancein thedomain ofgains.
In termsof debates betweenprospecttheoryand expected-utility
theory,reference dependence and framingeffectsare critical.If people alwaysframedaround
the statusquo itwould be possibleto subsumeloss aversionand thereflectioneffect
withinexpected-utility
theorybypositingan S-shapedutilityfunctionwitha steeper
slope on theloss side. But preferencereversalsinducedbychangesin framesrather
than by changes in subjectiveutilitiesor probabilitiesare much more difficult
to
reconcilewithexpected-utility
theoryor withrationalchoice theoriesmore generally. Evidence that behavior varies depending on whetherthe glass is seen as
or half-full
does not easilylend itselfto a rationalchoice explanation.
half-empty
Referencedependence is inconsistentwiththe Coase theorem,which suggests
that if transactioncosts are negligiblethe valuation of a propertyrightis independent of who owns that right,so that initial entitlementsdo not affectfinal
allocations (Kahneman, Knetsch, and Thaler, 1990; Camerer, 1995:670). But
evidence suggeststhatinitialentitlementsdo make a differenceand thatrates of
exchange are affectedby whetherone is givingup a good or acquiring it. An
a partofher endowment
individualmayprefergood x to good ywhenx is currently
but prefergood y to good x when y is part of her endowment,so thatpreferences
are reference-dependent
(Tverskyand Kahneman, 1991:1039; Tverskyand Simonson, 1993).13More generally,thereis evidencethatone's preferencebetweenx and
y may depend on the presence or absence of a third option z (Tverskyand
Kahneman, 1991; Tverskyand Simonson, 1993).
1I1Althoughth-er-e
shiftsto unldeiweighting
is no conclusiveevidenceas to the specificpoin-tat which oveiweightiing
oi whetherthispointvariessignifican-tly
acr-ossindividualsor conditions,preliminar-y
eviden-cesuggeststhatit fallsin
the .10-.15 range (Her-shey
and Schoemaker,1980).
12 It should be clear thatprospecttheoi-y
belongs to the generalclass ofutilitymaximizationtheories,whichinvolve
theol-y
thecombin-ationofprobabilitiesand utilitiesand some formofmaximizationofvaluLe.
Whereasexpected-utility
positsa lin-earcombinationofutilitiesand probabilities,prospecttheol-y
positsthatutilitiesare weightedbya fuLnction
thatis nonlinearin probabilities.
13 Onletheoreticalimplicationis thatitis not alwayspossibleto construct
cur-ves
thatare reversibleaiid
indifference
nonin-tersecting
(Knetsch,1989).
JACK S. LEVY
93
These reference-dependent,
or framing,effects
violatetheassumptionofconsistentand transitivepreferencesthatis essentialnot onlyto expected-utility
theory
but also to nearlyall rationalchoice theoriesas well.14 Framingeffectscannot be
reconciledwiththeprincipleofthe"independenceofirrelevantalternatives"(Luce
and Raiffa,1957:27; Tverskyand Simonson, 1993:1179) or the related invariance
assumption,which requires that logicallyidenticalchoice problems should yield
identicalresults(Tverskyand Kahneman, 1986:S252-7). Althoughinvarianceis not
a formalaxiom ofexpected-utility
theory,itis an "invisiblebackgroundassumption"
of the theory(Camerer, 1995:652). Arrow(1982:6) refersto invarianceas "extensionality"and describesit as a "fundamentalelementof rationality."Ferejohnand
Satz (1995:80) concede thatcases in which"choices are based on preferencesbut
preferencescan depend on the situation,"do not fit "classical" conceptions of
rationality,but suggest "weaker forms of rationality"that do not satisfythe
independence hypothesis, such as regret models (Bell, 1982; Loomes and
Sugden, 1982).'5
In sumrnary,
framingeffectsare not consistentwithexpected-utility
theoryand
probablynot consistentwithmostconventionalconceptionsof rationalchoice,but
theoristshave constructedweakerformsof rationalchoice theorythatare compatible withframingeffects.
Implications
forInternational
Relations
94
ProspectThzeory,
Rational Chzoice
standai-dpr-ocedui-e
forelicitingpi-efer-ences
is the Becker,DeGi-oot,and Mar-schak(1964) procedure,in
whiclhr-ational
subjectshave strongincentivesto statetheirtrueresel-vationpricesforsellinglotterieswithwhichthey
have been enldowed.For a bh-ief
descriptionsee Roth (1995:19-20).
18Econ-omists
gen-erally
use monetaryamouLn-ts
thatare comparable (when-cum--ulated)
to wage ratesof emllployed
membersof the subjectpools-typically$8-$20 per hour.
19Plott (1987:120) suggeststhat there is more evidence for incentiveeffectsin grouLpas opposed to individual
17 The
behavior.
JACK S. LEVY
95
96
Prospect
RationalChzoice
Theory,
23 Findingsof systematic
of
anomalies in expected-utility
theorygain some fur-ther
supportfromlaboratorystuLdies
animalbehavior,wheresimilarpatternsofchoiceunderconditionsofriskare obser-ved
(Battalio,Kagel, and McDonald,
1985; Kagel, 1987). Needless to say,questionsof cross-speciesgeneralizability
is highlycontestedterritory.
24 This is consistentwithexperimentalstudies of the endowmenteffect,
which generallyfind that selling prices
exceed buyingpricesby twoto one.
JACK S. LEVY
97
It is fairto concludethatfindingsofreferencedependence,theendowmenteffect,
the reflectioneffect,framingeffectsand preferencereversals,and nonlinear responses to probabilitiesin individualchoice have held up quite well in the face of
substantialcriticalscrutiny.Numerous and varied attemptsto demonstratethat
theseanomaliescan be explained byabsence offinancialincentivesor opportunities
forlearning,bytransactioncosts,or othereconomicvariableshave failed.Camerer
(1995:674), for example, concludes that "not a single major recent (post-1970)
anomaly has been 'destroyed'by hostile replication."It is true that the proper
controlsreduce the magnitudeof observed anomalies to a certainextent,which
suggeststhe need foradditionalresearchto determinethe conditionsunderwhich
thisis mostlikelyto occur,but the bottomline is thatthe anomalies persist.
The theoreticalimplicationsofthesedescriptivefindingsare less clear. Ifwe were
to adopt a Popperian criterionand evaluate theoriesby direct testsagainst the
evidence (Popper, 1959), thenwe mighthave to rejectexpected-utility
theoryas a
positivetheoryof individualchoice. As the economistPlott(1987:117) concludes,
"Iftheonlyquestionposed is Rationalchoice,trueor false?thentheansweris clearly
false.... The weakestformsofthe classicalpreferencehypothesisare systematically
at odds withthe facts."
But "True or false?,"in termsof directtestsof a theoryor itshypothesesagainst
the empiricalevidence,is not the onlyquestion,and social scientistshave increasinglyshiftedto a Lakatosian conceptionof scientificprogressand a demand that
theoriesbe testedagainst alternativetheories(Lakatos, 1970). Prospecttheoryis
one of a number of alternative theories of individual choice that behavioral
decision theoristshave constructedin an attemptto integrateobservedanomalies
in expected-utility
theoryinto a more descriptively
accuratetheoryof choice.25
Experimentalsocial psychologistsand economistshave conducted endless tests
of these theoriesagainsteach otherand againstexpected-utility
theory(Camerer,
1989, 1995; Harless and Camerer, 1994). The resultsare mixed. The general
conclusionis that"choice under uncertainty
is a fieldin flux"(Machina, 1987:121),
thatthereis "no consensusas to whetherany of the manyalternativetheoriesare
better than expected utilitytheoryin explaining behavior under uncertainty"
(Sopher and Gigliotti,1993:75), and that"none ofthealternativesto expectedutility
theory... consistently
organizethe data, so we have a long wayto go beforehaving
a complete descriptivemodel of choice under uncertainty"(Battalio,Kagel, and
Jiranyakul,1990:46).
There is anothersense in whichthe question "True or false?"is misleading.As
social scientistsour aim is not to testexpected-utility
theoryper se (whichtaken
alone is empiricallyemptyand nontestable)but modelsof social behaviorthatare
builton themicrofoundations
ofexpected-utility
theory(or prospecttheoryor some
alternative).We must also be carefulto specifyexactlywhat it is thatwe want to
explain. Althoughexpected-utility
theorymodels of individualchoice behaviorare
descriptivelyproblematic,there is a fairamount of evidence fromexperimental
studiesofassetmarketsand auctionsthatutilitymaximizationmodels do fairlywell
in accountingformanyaspectsofaggregatemarketbehavior(Friedman,Harrison,
and Salmon, 1984; Plott,1987; Smith,1991; Wittman,1991; Sunder, 1995; Roth,
informational
1995:49-60; Kagel, 1995). Experience,learning,incentive,
efficiencies,
98
ProspectTheory,Rational Choice
JACK S. LEVY
99
100
ProspectTheory,Rational Choice
JACK S. LEVY
101
36 In truth,
"as if' assumption
thereis a debate betweenthoserationalchoice theor-ists
who adopt theinstrumentalist
and focuLson the pr-edictive
powerof rationalchoice theories,and "realists"who insistthatthe rationalityassumptio
is a descr-iptively
accur-atelaw of behavior. Sin-ceM. Friedman (1953), most formaldecision and game theor-ists,
particularly
in intern-ational
r-elations,
adopt the instrumentalist
view,and I shall assume thatperspectivehere; but for
a br-ief
summalyof the debate see J. Friedman(1996).
102
ProspectTheory,Rational Choice
JACKS. LEVY
103
104
ProspectTheory,Rational Choice
also be spurious,ifanothervariablesimultaneously
influencesan actor'spreferences
over outcomesand her selectionof a referencepoint (Jervis,1992).38
This discussionmakesitclearthatone critically
importanttaskforfutureresearch
involvesthe theoreticaldevelopmentand empiricalvalidationof linkagesbetween
loss aversion,the reflectioneffect,and framingeffectsat the individuallevel and
decisionmakingat the collectivelevel. Most usefulwould be the developmentof a
formaltheoryofcollectivechoicebased on themicrofoundations
ofprospecttheory.
This is an ambitioustheoreticaltask,but thereare otherthingsthatcan be done
short of this that would add to the utilityof prospect theoryfor international
relations.
One approach would be to incorporateframing,loss aversion,the reflection
effect,and the certaintyeffectintostandardmodels of bureaucraticpolitics.These
hypothesescould help to explain thesourcesofthepolicypreferencesofindividual
actors in the decision-makingprocess,while conventionalhypothesesabout the
nature of the bureaucraticprocess explain how these preferencesof different
individualsin different
roles and different
degrees of politicalinfluenceget aggregated to produce collectivepolicy choices. Heads of organizationsmay be more
concerned about maintainingtheir organizational power and budgets than in
increasingthem,forexample, and theymay treattemporarybudgetaryincreases
as a permanentpart of the organization'sendowment.
The same can be done for alternativemodels of the foreignpolicy process.
Domesticpublics,forexample, maybe more influencedbypolicies thathurttheir
standardoflivingthanthosethatimproveit,morebypoliciesthatharmthenational
interestthan those thatenhance it, and consequentlytheymay be more likelyto
punish political leaders for failuresthan to reward them for successes. In this
"two-levelgame" (Putnam,1988) withloss-aversepublicseven politicalleaderswho
were not themselvesinfluencedby loss aversionor framingwould findit in their
interestto behave as iftheywere.
In thesewaysthe loss aversionhypothesisand the reflectioneffectcan make an
importantcontributionto rationaltheoriesof politicsbyspecifying
both the shape
oftheutilityfunctionand theotherparametersthatare exogenous in thosemodels.
An alternativeapproach to theaggregationproblemwould focuson group dynamics. One hypothesismightbe that group settingslead to "compromisedframes,"
including"mixedframes"thatinvolveboth gainsand losses,thatare notas extreme
as individualframes(Shafir,1992:314).39One implicationis thatframechanges-and
consequentlyframing-induced
preferencereversals-willbe lowerin frequencyand
magnitude in group settingsas opposed to individualdecision making,but this
needs to be developed theoretically
and explored empirically.
Howeverit is done, some formof integrationofprospecttheoryhypothesesinto
broader models of the foreignpolicyprocess is a criticaltaskforfutureresearch,
forprospecttheoryas a theoryof individualchoice does not in itselfconstitutea
general theoryof foreignpolicy.
Problem
TheStrategic
Interaction
JACKS.
LEVY
105
106
ProspectTheory,Rational Choice
expected-utility
theoryin experimentaland fieldresearch.This leads naturallyto
the viewof a paradigmaticdebate betweenprospecttheoryand rationalchoice.
In the sectionon externalvalidityI have emphasized the methodologicalproblems involvedin empiricallydifferentiating
a prospecttheoryexplanationfroma
rationalchoice explanationbased on expected-valuemaximization,and thisraises
the question of the criteriabywhichwe evaluate competingtheories.ElsewhereI
have argued thatexpected-utility
theoryis moreparsimoniousthanprospecttheory
because it is axiomaticallybased and because it requiresfewerparametersforfull
specification.Expected-utility
theoryis also based on a normativelyappealing set
of axioms, whereas prospect theorymakes no normative claims (Tverskyand
Kahneman,1986:S272).40On thesegroundsI argued thatin case studyapplications
of prospect theorythe burden of proof should be on the proponent of prospect
theoryto testher theoryor explanationagainsta rivalrationalchoice explanation
and to show not only that the evidence is consistentwith a prospect theory
explanation,but also thatthe theoryprovidesa betterexplanationofthatbehavior
than does a rationalchoice explanationbased on a straightforward
expected-value
calculation(Levy,1992b:296-7).
AlthoughI continueto believethatin the end empiricalapplicationsofprospect
theoryin internationalrelationsand politicalscience more generallywill have to
demonstratethatthe theoryprovidesa betterexplanationof a set of phenomena
than does an alternativerationalchoice theory,I believe thatmanyinternational
relationsanalystshave gone too farin definingtheirtaskin termsofa paradigmatic
debate betweenprospecttheoryand rationalchoice. I also believe thatgiven the
earlystagesin the developmentand applicationofprospecttheoryin international
relations,it is prematureto focus primarilyon empirical testsbetween the two
paradigms.
Firstof all, too much scholarshipin the internationalrelationsfieldhas focused
primarilyon interparadigmaticdebates at the highestlevel rather than on the
constructionand testingof falsifiablemodels of behavior. The rational choice
theoriesofinternationalpolitics
paradigm4lsubsumesa numberofratherdifferent
(derivingin part fromthe incorporationof different
auxiliaryassumptionsabout
preferences),and a focuson the paradigmaticdebate betweenprospecttheoryand
rationalchoice obscuresthese significanttheoreticaldifferences.In the same way,
theoretical
paradigmaticdebatesbetweenrealismand liberalismobscuresignificant
variationswithineach (Levy,1994), and the furtherconceptualdevelopmentofthe
theoriesof interprospecttheoryresearchprogramis likelyto generatedifferent
nationalpolitics.Interparadigmaticdebates deflectattentionawayfromtheimportant tasks of empiricallytestingbetween alternativetheories within the same
theoreticalperspectives
paradigm and of consideringhow insightsfromdifferent
mightbe combinedto generatemore powerfultheoriesof behavior.
Anotherreason whyinternationalrelationsscholarsshould not focusprimarily
on the paradigmaticquestionofwhetherprospecttheoryprovidesa betterexplanationofinternationalpoliticalbehaviorthandoes rationalchoice/expected-utility
theoryis methodological:itis extremely
unlikelythatempiricaldata in international
politicswill provide a convincingtestof these competingtheoriesof choice. It is
to operationalizeand measure the keyvariablesand to control
simplytoo difficult
40 That is, actorswho wereaware thattheywerebehavingaccordingto the dictatesofexpected-utility
theorywould
have no incentivesto altertheirbehavior,whereasactorswho wereaware thattheirbehaviormatchedprospecttheoi-y
mightwanitto altertheirbehavior.
41 It is more usefulto conceptualizerationalchoice as a paradigm or a researchprogram-or perhaps as a "family
oftheories"or a "fieldofendeavor"(Green and Shapiro, 1994:28-30; Ferejohnand Satz, 1995:81)-than as a "theory"
ofbehavior.
JACKS.
LEVY
107
42 One example might be the Morgan and Wilson (1989) experimentaltest of a spatial bargaining model of
interinational
conflict.
43' This parallelsthecriticisms
on thegroundsthat
ofGreen and Shapiro's (1994) Pathologies
ofRationalChoiceThleory
it gives too much emphasis to the criterionof empiricalconfirmationor explanatoiysuccess and not enough to the
specificationofcausal mechanismsthatexplain behavior-(Ferejohn and Satz, 1995; Chong, 1996).
108
Prospect
Thieory,
RationalChoice
44 Itwould
ofreference
anidframing
alsobeusefultoexplorewhether
theintractable
phenomenon
depenidence
canl
To whatextent,
forexample,can thetheoiyofrationalexpectations
be reconciled
withrationalapproachtheories.
effects?
explaini
framing
JACK S. LEVY
109
BUENO DE MESQUITA,B. (1981) The WarTrap. New Haven, CT: Yale University
Press.
BUENO DE MESQUiTA,B., AND D. LALMAN(1992) Warand Reason.New Haven, CT: Yale University
Press.
CAMERER,C. F. (1989) An ExperimentalTest ofSeveralGeneralized UtilityTheories.JournalofRiskand
Uncertainty
2:61-104.
CAMERER,
C. F. (1990) "BehavioralGame Theory."In Insights
inDecisionMaking,editedbyR. M. Hogarth,
pp. 311-336. Chicago: Universityof Chicago Press.
C. F. (1995) "IndividualDecision Making." In TheHandbookofExperimental
CAMERER,
edited
Economics,
byJ. H. Kagel and A. E. Roth,pp. 587-703. Princeton,NJ: PrincetonUniversityPress.
CHEW,S. H. (1983) A Generalizationof the Quasilinear Mean withApplicationsto the Measurementof
Income Inequalityand Decision TheoryResolvingtheAllais Paradox. Econometrica
51:1065-1092.
CHONG,D. (1996) "RationalChoice Theory'sMysteriousRivals."In TheRationalChoiceControversy,
edited
byJ. Friedman,pp. 37-57. New Haven, CT: Yale UniversityPress.
COURSEY, D. L., J. L. HOVIS,ANDW. D. SCHULZE (1987) The Disparity Between Willingness to Accept and
9:71-84.
P. C. (1982) NontransitiveMeasurable Utility.
JournalofMathematical
FISHBURN,
Psychology
26:31-67.
P. C. (1984) SSB UtilityTheoryand Decision MakingUnder Uncertainty.
FISHBURN,
Mathematical
Social
Sciences8:253-285.
FISHBURN, P. C., AND G. A. KOCHENBERGER(1979)
Utility
Functions.DecisionSciences10:503-518.
FRANCIOSI,R., P. KUJAL,R. MICHELITSCH,AND V. SMITH (1993) Experimental Tests of the Endowment
Effect.Workingpaper. UniversityofArizona.
FRIEDMAN,D., G. HARRISON,ANDJ. SALMON(1984) The Informational Efficiency of Experimental Asset
Markets.Journal
ofPoliticalEconomy
92:349-408.
FRIEDMAN,
J. (1996) "Introduction:EconomicApproachesto Politics."In TheRationalChoiceControversy,
edited byJ. Friedman,pp. 1-24. New Haven, CT: Yale UniversityPress.
edited by
M. (1953) "The MethodologyofPositiveEconomics."In Essaysin PositiveEconomics,
FRIEDMAN,
M. Friedman,pp. 3-43. Chicago: Universityof Chicago Press.
FRIEDMAN,
M., ANDL. J. SAVAGE
(1948) The UtilityAnalysisof Choices InvolvingRisk.Journal
ofPolitical
56:279-304.
Economy
GEVA,N., ANDA. MINTZ(1994) Framingthe Options forPeace in the Middle East. Paper presentedat
the Annual Conferenceof ECAAR-Israelon The PoliticalEconomyof Peace in the Middle East,
Haifa, Israel,June 20.
A Critique
inPolitical
GREEN,D. P., ANDI. SHAPIRO
(1994) Pathologies
ofRationalChoiceTheory:
ofApplications
Science.New Haven, CT: Yale UniversityPress.
GRETHER, D. M., AND C. R. PLOTT (1979) Economic Theory of Choice and the Preference Reversal
Phenomenon.American
Economic
Review69:623-638.
Modeling
Science12:378-394.
Dependence Effectson Brand Choice. Management
HARLESS,D. W. (1989) More LaboratoryEvidence on the DisparityBetweenWillingnessto Pay and
11:359-379.
CompensationDemanded.JournalofEconomicBehaviorand Organization
HARLESS, D. W., AND C. F. CAMERER (1994) The Predictive Utility of Generalized
Theories. Econometrica
62:1251-1289.
Expected Utility
ProspectTheory,Rational Choice
110
A StudyofSmallGroupsand PolicyFailure.Baltimore,MD:
HART, P. 't (1990) Groupthink
in Government:
JohnsHopkins UniversityPress.
JournalofEconomics106:141-162.
CriticalExamination.Organizational
Behaviorand HumanPerformance
25:395-418.
Science36:780-803.
13:171-186.
LEVY,J. S. (1992a) An Introductionto ProspectTheory.PoliticalPsychology
LEvY,J. S. (1992b) ProspectTheoryand InternationalRelations:TheoreticalApplicationsand Analytical
Problems.PoliticalPsychology
13:283-310.
LEvY,J. S. (1994) The TheoreticalFoundationsofPaul W. Schroeder'sInternationalSystem.International
Review16:715-744.
History
LEVY,J. S. (1996a) Loss Aversion,Framing,and Bargaining:The Implicationsof ProspectTheory for
PoliticalScienceReview17:177-193.
InternationalConflict.International
LEVY,J. S. (1996b) Hypotheseson the Framingof Decisions. Paper presentedat theAnnual Meetingof
the InternationalStudiesAssociation,San Diego, Calif.,April 17-20.
LEVY,J. S. (1997) "ProspectTheoryand the Cognitive-RationalDebate." In DecisionMakingon Warand
Peace: TheCognitive-Rational
Debate,edited byN. Geva and A. Mintz.Boulder,CO: LynneRienner.
overTime.New York: RussellSage.
LOEWENSTEIN, G., ANDJ. ELSTER, EDS. (1992) Chioice
LOOMES, G., AND R. SUGDEN (1982) RegretTheory: An AlternativeTheory of Rational Choice Under
Economic
Journal92:805-824.
Uncertainty.
LUCE, R. D., AND H. RAIFFA(1957) Gamesand Decisions.New York: Wiley.
MACHINA, M. J. (1982) "Expected Utility"AnalysisWithoutthe Independence Axiom. Econometrica
50:277-323.
MACHINA,M.J. (1987) Decision-Makingin the Presenceof Risk.Science236:537-543.
JACKS. LEVY
111
112
ProspectTheory,Rational Choice