53Risk and time__ preferences-Linking__ experiment

forthcoming,theAmericanEconomicReviewRiskandtimepreferences:LinkingexperimentalandhouseholdsurveydatafromVietnamTomomiTanaka,ColinF.CamererANDQuangNguyen*AbstractWeconductedexperimentsinVietnamesevillagestodeterminethepredictorsofriskandtimepreferences.Invillageswithhighermeanincome,peoplearelessloss-averseandmorepatient.Householdincomeiscorrelatedwithpatiencebutnotwithrisk.Weexpandmeasurementsofriskandtimepreferencesbeyondexpectedutilityandexponentialdiscounting,replacingthosemodelswithprospecttheoryandathree-parameterhyperbolicdiscountingmodel.ComparableriskparameterestimateshavebeenfoundforChinesefarmers,usingourmethod.(JELC93,D81,D9)*Tanaka:SchoolofGlobalStudies,ArizonaStateUniversity,P.O.Box875102,Tempe,AZ85287-5102(email:[e-mail]);Camerer:DivisionfortheHumanitiesandSocialSciences,CaliforniaInstituteofTechnology,MC228-77,PasadenaCA91125(email:[e-mail]);Nguyen:Grouped’AnalyseetdeThéorieEconomique(GATE),CNRS,FacultédeSciencesÉconomiquesetdeGestion,UniversitéLumière(Lyon2)(email:[e-mail]).ThisresearchwassupportedbyaBehavioralEconomicsSmallGrantfromtheRussellSageFoundation,FoundationforAdvancedStudiesonInternationalDevelopment,andinternalCaltechfundstoauthorCamerer.CommentsfromparticipantsattheESAmeeting(October2005),SEAmeeting(November2005),SJDM(November2005),ASSAmeeting(January2007),audiencesatColumbia,NYU,Bocconi,Emory,Hawaii,Caltech,UCSC,ClaremontMcKenna,Guelph,Carleton,ArizonaState,aconferenceatUT-Dallas,andfivethoughtfulanonymousrefereeswerehelpful.Thankstoourresearchcoordinators,PhanDinhKhoi,HuynhTruongHuy,NguyenAnhQuan,NguyenMauDung,andresearchassistants,BuiThanhSang,NguyenTheDu,NgoNguyenThanhTam,PhamThanhXuan,NguyenMinhDuc,TranQuangTrung,andTranTatNhat.WealsothankNguyenTheQuanoftheGeneralStatisticalOffice,forallowingustoaccessthe2002householdsurveydata.Afundamentalquestionindevelopmenteconomicsistheextenttowhicheconomicsuccessislinkedtobasicfeaturesofhumanpreferences.Ifpeopleareextremelyaversetofinancialrisk,theymaybereluctanttocreatebusinessesthatmayhaveinherentlyriskycashflows.Ifpeopleareimpatient,theymaybereluctanttoinvestandeducatetheirchildren.Takentogether,risk-aversionandimpatiencemayexplain,inpart,whysomepeopleremainpoor.WeconductedexperimentsinVietnamesevillagestodirectlymeasureriskandtimepreferencesofindividuals,andinvestigatedhowthesepreferencescorrelatewitheconomiccircumstances.Vietnamhasseveraladvantagesasafieldsite:1.Accesstoa2002livingstandardsurveyenabledustolinkdetailedsurveyresponsesfromindividualsdirectlytoexperimentalresponsesbythesameindividuals.2.MostVietnamesevillagersarepoorbutliterate.Asaresult,itisbotheasytomotivatethemwithmodestfinancialstakes,andtoensuretheycomprehendinstructions.3.Theriseofhouseholdbusinessesinthemarketeconomyhascreatedsubstantialvariationinincome.Thisincomevariationcanbecorrelatedwithpreferencemeasures.Inanycross-sectionalstudylikethis,itisdifficulttoinferthedirectionofcausalityfromcorrelation:Dopreferencescauseeconomiccircumstances(e.g.,throughbusinessformation,forexample),ordocircumstancescreatepreferences(asdescribedbySamuelBowles(1998))?Anidealstudywoulduserandomizedassignmentofindividualstoeconomiccircumstances.Asanalternative,weemployaninstrumentalvariableapproach,usingrainfallandhouseholdhead’sabilitytoworkatthetimeofsurvey,whichareunlikelytobecorrelatedwithpreferences,asinstrumentalvariablesforincome.Besidescontributingnewdata,ourpapermakesamethodologicalcontributiontoexperimentaldevelopmenteconomics.Mostpreviousexperimentsconductedinthefieldtested1modelsofriskandtimepreferencesthatcanbecharacterizedbyoneparameter.(SeeJeffreyCarpenterandJuan-CamiloCardenas(2008)forareview).ThesemodelsoftenfitexperimentaldatainWesterneducatedpopulations(ShaneFrederick,GeorgeLoewensteinandTedO'Donoghue,2002,ChrisStarmer,2000)andfielddata(ColinF.Camerer,2000)lesswellthanmodelswithmultiplecomponentsofriskandtimepreferences.Forexample,inexpectedutilitytheory(EU),riskpreferencesarecharacterizedsolelybytheconcavityofautilityfunctionformoney.Butifriskychoicesexpressprospecttheorypreferences(DanielKahnemanandAmosTversky,1979),thenutilityconcavityisnottheonlyparameterinfluencingriskpreferences—nonlinearweightingofprobabilities,andaversiontolosscomparedtogain,alsoinfluenceriskpreferences.Ourinstrumentsaredesignedtomeasurethesethreeparametersofprospecttheory,ratherthanjustoneinEU.Similarly,wemeasurethreeparametersinageneraltimediscountingmodel(JessBenhabib,AlbertoBisinandAndrewSchotter,2007),ratherthanmeasuringasingleexponentialdiscountrateasinmostotherstudies.Iftheexponentialmodelisanadequateapproximation,thenourricherinstrumentswilldeliverparametervaluesoftheextravariableswhichaffirmthevirtueofthesimplerexponential.Beforeproceedingtodesigndetailsandresults,itisusefultodiscusshowourapproachcomparestootherfieldexperiments.Fieldexperimentsindevelopmentarepowerfultoolsforpolicyevaluationbecausetheycanrandomizetreatmentsinnaturally-occurringdecisionmakingtoseehowwellaspecificpolicyworksinaspecificsettingwithapropercontrolgroup(seeEsthe