外汇市场实验研究：Experimental Foreign Exchange Markets

1ExperimentalForeignExchangeMarketsEricO’N.Fisher*andFrankS.KellyRunningHead:ForeignExchangeExperimentsAbstract:Weanalyzeexperimentalmarketswheresubjectsbuyandselltwodifferentassets.Theassets’propertiesvaryacrosstreatments,andtheirrelativepriceistheexchangerate.Althoughbothassetsuniformlyexhibitbubbles,theexchangeratesatisfiescross-currencyarbitrage.Thereisnoevidenceofapositiveriskpremiuminthesemarkets,andalmostallsubjects’forecastsoftheexchangeratearerational.(JELClassificationCodes:F3,C9)*Fisher:DepartmentofEconomics,TheOhioStateUniversity,410ArpsHall,1945NorthHighStreet,Columbus,OH43210,USA.Kelly:DepartmentofEconomics,IUPUI,516CavanaughHall,425UniversityBoulevard,Indianapolis,IN46202,USA.FisherthanksTheOhioStateUniversityfora1993SeedGrant.ThispaperwaswrittenwhileFisherwasavisitingfellowattheAustralianNationalUniversity.WethankJamesNieberding,MelaneyFullerton,ZusunRhee,ColinFeng,andRaphaelSolomonforcapableresearchassistance.Wearegratefulforthecommentsofthreereferees,JeffBergstrand,IkeBrannon,andmanyseminarparticipants.Allthedataandtestsdescribedinthispaperandtheinstructionsforallthetreatmentsforthisexperimentareavailableat(1982)isaclassicexample,andPlottandSunder(1982,1988)areotherimportantpapers.2Thesecondstrandfocusesondynamicenvironments,andtheseminalpaperisbySmith,Suchanek,andWilliams(1988).Theseexperimentalmarketsalmostalwaysexhibitpricebubbles.3Theasset’spriceisinitiallybelowitsfundamental,risesaboveitinthemiddleperiods,andcrashesdowntoitinthelastfewperiods.41DavisandHolt(1993)surveyofmuchoftheworkinexperimentaleconomicsingeneral,andSunder(1995)surveystheliteratureonassetmarketsessionsinparticular.2O’BrienandSrivastava(1991)replicatestaticbutinformationallycomplexenvironmentswithseveralassets,andtheirworkiscloselyrelatedtoourown.Theyfindlesssupportforthepredictionsofrationalexpectationstheorythanisthenorm,buttheyalsoshowthatassetpricessatisfycross-equationrestrictions.3Abubbleisanysystematicdeviationofanasset’spricefromitsfundamental,andarationalbubblefurthersatisfiessomeintertemporalarbitragecondition,typicallyanEulerequation.FloodandHodrick(1990)surveythenascentliteratureontestingforbubblesinforeignexchangemarketsinthefield.4Smith,Suchanek,andWilliams(1988)argueimplicitlythatbubblesarisebecauseofexpectedcapitalgains,butSunder(1995)suggeststhattheymaybeartifactsoflearning.King,Smith,Williams,andvan3Almostalltheliteratureusingfielddatafromthemodernperiodoffloatingexchangeratesemphasizesthatthestochasticprocessesdrivingexchangeratesappearnottobestationary.Forexample,Mark(1990)andFisherandPark(1991)showthatthemanybilateralexchangerateshaveunitroots,andmostmoderntheoriesemphasizethatassetpricesfollowmartingales.Hence,itisappropriatetodesignexperimentswherethemarketisnotstationary.ThisconsiderationledustoadopttheframeworkofSmith,Suchanek,andWilliams.Weaddonebasicelementtotheirdesign:simultaneoustradingintwoassets.Insometreatments,thedifferencesbetweentheseassetsisspurious;inothers,oneassetisriskierthantheother,oronepaystwicetheexpecteddividendoftheother.Ofcourse,assetmarketsbytheirverynatureallowforinter-temporalarbitrage.Inforeignexchangemarkets,asinourexperiment,tradersalsofacethepossibilityofcross-assetarbitrage.Ourmainconclusionisthatsubjectsgetthecrossexchangerateright.Ourmarketsareoraldouble-auctionswheretraderssellassetsfordollarsanddollarsforassets.Henceatradertakingadvantageofaperceivedarbitrageopportunitymust,forexample,sellblueassetsfordollarsandthensubsequentlybuyredassetswithdollars.Thismayseemstrangeatfirstblush,butmodernforeignexchangemarketsworkexactlyinthismanner.5Foreignexchangeistradedinthefieldinover-the-countermarketsmadebylargecommercialbanks.ButtheinformationalefficienciesinherentinanBoening(1993)findbubblesevenwithshortsellingandfuturesmarkets,andvanBoening,Williams,andLaMaster(1993)havedemonstrateditalsooccursincallmarkets.5SeeChorafas(1992,p.57)orGrabbe(1986,p.63).4oraldouble-auctioncreatelaboratorymarketswhosepropertiesarequitesimilartothoseofthethickforeignexchangemarketsinthefield.Itisimportanttoemphasizethatthereareonlytwoindependentexchangeratesinourexperiments:thedollarpricesoftheblueandredassets.Thecruxofourresultsisthattheimpliedblue-redcross-rateiswellbehaved.Thisfindingisallthemorestrikingbecausethedollarpricesofbothassetsaresignificantlydifferentfromtheirfundamentals.Thusbubblesariseineachmarket,buttheyarenotindependent!Whyarethesedividend-bearingassetsforeignexchange?Theseassetscapturetwoofthetraditionalrolesofmoney:transactionsandstoreofvalue.First,anasset’sdiv