经济在实验室里弗农l.Smith(1)

EconomicsintheLaboratoryVernonL.SmithVernonL.SmithisRegents’ProfessorofEconomics,andResearchDirectoroftheEconomicScienceLaboratory,UniversityofArizona,Tucson,Arizona.In1992,hewasnamedaDistinguishedFellowoftheAmericanEconomicAssociation.Whydoeconomistsconductexperiments?Toanswerthatquestion,itisfirstnecessarybrieflytospecifytheingredientsofanexperiment.Everylaboratoryexperimentisdefinedbyanenvironment,specifyingtheinitialendowments,preferencesandcoststhatmotivateexchange.Thisenvironmentiscontrolledusingmonetaryrewardstoinducethedesiredspecificvalue/costconfiguration(Smith,1991,6).1Anexperimentalsousesaninstitutiondefiningthelanguage(messages)ofmarketcommunication(bids,offers,acceptances),therulesthatgoverntheexchangeofinformation,andtherulesunderwhichmessagesbecomebindingcontracts.Thisinstitutionisdefinedbytheexperimentalinstructionswhichdescribethemessagesandproceduresofthemarket,whicharemostoftencomputercontrolled.Finally,thereistheobservedbehavioroftheparticipantsintheexperimentsasafunctionoftheenvironmentandinstitutionthatconstitutethecontrolledvariables.Usingthisframeworkofenvironment,institution,andbehavior,Icanthinkofatleastsevenprominentreasonsintheliteratureastowhyeconomistsconductexperiments.Undoubtedly,therearemore(DavisandHolt,1992,chapter1andpassim).1.Testatheory,ordiscriminatebetweentheories.Thismotivationcomesfromtheeconomicandgametheoryliterature.Wetestatheorybycomparingitsmessageoritsoutcomeimplicationswiththeexperimentalobservations.Thegreaterthefrequencywithwhichtheobservationshitthesepredictions,inthecontextofadesigninwhichhitsareunlikelytooccurbychance,thebetterthetheory.2Examplescanbefoundintheauctionliterature(Smith,1991,25-29),whereriskaversemodelsofbiddinginDutchandfirstpricesealedbidauctionsarefavoredbythedataoverriskneutralmodels,while1Whereappropriate,referencestoworkbymeandmycoauthorswillbetothepapernumbersinSmith(1991).2Selten(1989)offersameasureofpredictivesuccess.Iusethetermspredictionandimplicationofatheoreticalmodelinterchangeably.Consistencywithapredictiondoesnotrequirethatthetheorybedoneinadvanceofanobservation.dominatestrategyauctionssuchastheEnglish,whoseoutcomesarepredictedtobeindependentofriskattitude,performwellinthelaboratory.Ofcourse,theoriessubjectedtosufficientlyrigoroustestsarenearlyalwaysfoundtoneedimprovement;thisleadstothesecondreasonfordoingexperiments.2.Explorethecausesofatheory’sfailure.Whentheobservationsofanexperimentfailtoconformtotheimplicationsofthetheory,thefirstthingtobedoneistoreexaminethedesign,andtobesurethatthepredictivefailureisthefaultofthetheory.Well-articulatedtheoriesformallymodeltheenvironmentandthetradingrules,andtheexperimentalistseekstoreproducetheseconditionsofthetheory.Inthecourseoftestingwhentheexperimentaldesigncontinuestoseemappropriateandthetheorystillfails,thistendstoencourageanexperimentalexaminationdesignedtodiscoverthecause.Establishingtheanatomyoffailureisessentialtoanyresearchprogramconcernedwithmodifyingthetheory.Examplesaretobefoundinthebargainingliterature(Roth,1987;HoffmanandSpitzer,1985;Hoffman,etal,1992;Bolton,1991)andincommonvalueauctions(KagelandLevin,1986;CoxandSmith,1992).Oftentheoriesthatinitiallyperformpoorlyshowimprovementifsubjectsaregivenmoreexperience(CoxandSmith,1992),orthepayoffsareincreased(SmithandWalker,1993),butsometimesthesemeasuresfailtoyieldresultsthatimprovethetheory’sperformance(SmithandWalker,1993).3.Establishempiricalregularitiesasabasisfornewtheory.Well-formulatedtheoriesinmostsciencestendtobeprecededbymuchobservation,whichinturnstimulatescuriosityastowhataccountsforthedocumentedregularities.Microeconomictheorytendstobuilduponsimplifyingassumptions,andtoeschewattemptstomodelmanyofthecomplextradingandcontractinginstitutionsthatweobserve.Butinthelaboratory,especiallywithcomputerization,institutionswithcomplextradingrulesareaseasytostudyasaresimplesingleunitauctions.Thismakesitpossibletorangebeyondtheconfinesofcurrenttheorytoestablishempiricalregularitieswhichcanenabletheoriststoseeinadvancewhatarethedifficultproblemsonwhichitisworththeirwhiletowork.Thecontinuousdoubleauction,usedtheworldover,isafineexample.Inthisinstitutionbuyersannouncebidprices,sellersannounceoffers,oraskingprices.Anynewbid(offer)mustbeatapricewhichislower(higher)thanthestandingbid(offer);i.e.thebid-askedspreadmustnarrow.Abindingcontractoccurswhenabuyeracceptsaseller’sask,oraselleracceptsabuyer’sbid.Contractsoccurinsequenceasnewbids,asksandacceptancesoccur.Becauseofitsrobustequilibratingpropertieswithsmallnumbersoftraderspossessingonlyprivateinformationthisinstitution(Smith,1991,1,2,6)wasstudiedextensivelyinthelaboratorylongbeforetheattemptsbyR.Wilson,D.Friedmanandotherstomodelit(seeFriedmanandRust,1992,forreferences).4.Compareenvironments.Comparingenvironmentsusingthesameinstitutionpermitsaninvestigationoftherobustnessofthatinstitution.Theobjectiveistostressthetheorywithextremeenvironmentalconditionsunderwhichaninstitution’sestablishedpropertiesmaybegintobreakdown.Thus,incommonvalueauctions(wheretheitemhasthesamevaluetoallbidder