一个测量金融危机的新方法

ANewApproachtoMeasuringFinancialContagionKee-HongBaeKoreaUniversityG.AndrewKarolyiOhioStateUniversityReneÂM.StulzOhioStateUniversityandNBERThisarticleproposesanewapproachtoevaluatecontagioninfinancialmarkets.Ourmeasureofcontagioncapturesthecoincidenceofextremereturnshocksacrosscountrieswithinaregionandacrossregions.Wecharacterizetheextentofcontagion,itseconomicsignificance,anditsdeterminantsusingamultinomiallogisticregressionmodel.Applyingourapproachtodailyreturnsofemergingmarketsduringthe1990s,wefindthatcontagionispredictableanddependsonregionalinterestrates,exchangeratechanges,andconditionalstockreturnvolatility.Evidencethatcontagionisstrongerforextremenegativereturnsthanforextremepositivereturnsismixed.Since1997,economists,policymakers,andjournalistshavetalkedaboutthe``Asianflu.''IthasgenerallybeenperceivedthattheadversecurrencyandstockmarketshockthatfirstaffectedThailandinJuly1997propa-gatedacrosstheworldwithlittleregardforeconomicfundamentalsintheaffectedcountries.BeforetheAsianflu,therewasthe1994Mexican``Tequilacrisis,''andsincethen,the1998``Russianvirus.''Emergingmarketseconomiccrises,ingeneral,havebeencharacterizedascontagious.AccordingtoWebster'sdictionary,contagionisdefinedas``adiseasethatcanbecommunicatedrapidlythroughdirectorindirectcontact.''Emergingmarketseconomiccriseshaveledtomassivebailoutstoquellcontagionandhavereducedsupportforfreecapitalmobility.BaeandKarolyigratefultotheDiceCenterforResearchonFinancialEconomicsforsupport;BaealsoappreciatesthefinancialsupportoftheSKResearchAward.WethankTomSantner,MarkBerliner,BobLeone,andStanLemeshowforusefuldiscussionsonmethodology,SteveCecchetti,PeterChristoffersen,CraigDoidge,BarryEichengreen,VihangErrunza,DavidHirshleifer,MatthewPritsker,RobertoRigobon,RichardRoll,KarenWruck,and,especially,ananonymousreferee,andtheeditor,CamHarvey,forcomments.CommentsfromseminarparticipantsatHongKongUniversityofScienceandTechnology,KoreaUniversity,McGillUniversity,YaleUniversity,MichiganStateUniversity,UniversiteitMaastricht,OhioStateUniversity,RiceUniversity,MonteVeritaRiskManagementCon-ference,FederalReserveBankofChicagoAnnualConferenceonBankStructureandCompetition,GlobalInvestmentConferenceonInternationalInvesting,andtheNYSEConferenceonGlobalEquityMarketsinTransitionimprovedthearticle.AddresscorrespondencetoReneM.Stulz,OhioStateUniversity,FisherCollegeofBusiness,2100NeilAve.806AFisherHall,Columbus,OH43210,ore-mail:[e-mail]iesFall2003Vol.16,No.3,pp.717±763,DOI:10.1093/rfs/hhg012ã2003TheSocietyforFinancialStudiesTheInternationalMonetaryFund(IMF)deputymanagingdirectoratthetime,StanleyFischer,rationalizedthe1994Mexicanbailoutinthisway:``Ofcourse,therewasanotherjustification:contagioneffects.Theywerethereandtheyweresubstantial.''1ContagionhasledBhagwati(1998)toarguethat``capitalflowsarecharacterized...bypanicsandmanias.''Ifmarketsworkthisway,itisnotsurprisingthatStiglitz(1998)calledforgreaterregulationofcapitalflows,arguingthat``...developingcountriesaremorevulnerabletovacillationsininternationalflowsthaneverbefore.''Eventhoughthiscontagionconnotespowerfulimagesofeconomicandfinancialplagues,itisdifficulttostudyscientifically.Evidenceofthisdifficultyisthatthereislittleagreementonevendefiningwhatfinancialcontagionmeans.2Sinceequitymarketvaluationsreflectfutureeconomicactivity,muchofrecentresearchattemptstolearnaboutcontagionbyinvestigatingwhetherequitymarketsmovemorecloselytogetherintur-bulentperiods.Thereareconsiderablestatisticaldifficultiesinvolvedintestinghypothesesofchangesincorrelationsacrossquietandturbulentperiodsandrecentinvestigationsofthisissuefindatbestmixedresults[seeBaigandGoldfajn(1999)andForbesandRigobon(2001)].Nevertheless,theredoesnotseemtobestrongevidencethatstockreturnsinonecountryaremorehighlycorrelatedwithreturnsinothercountriesduringcrisisperiodsonceonetakesintoaccountthefactthatcorrelationestimatesarelikelybiased.Arelatedliteraturedemonstratesthat,eventhoughcorrela-tionschangeovertime,itisdifficulttoexplainchangesincorrelations.3Perhapsthemostimportantlimitationoftheseinvestigationsoffinan-cialcontagionisthattheyfocusonassetreturncorrelationsinthefirstplace.Noneoftheconcernsexpressedaboutcontagionseemtobebasedonlinearmeasuresofassociationformacroeconomicorfinancialmarketevents.Infact,theseconcernsaregenerallyfoundedonthepresumptionthatthereissomethingdifferentaboutextremelybadeventsthatleadstoirrationaloutcomes,excessvolatility,andevenpanics.Inthecontextofstockreturns,thismeansthatifpanicgripsinvestorsasstockreturnsfallandleadsthemtoignoreeconomicfundamentals,onewouldexpectlarge1SeehisstatementinCalvo(1996,p.323).2Forareviewofthedifficultiesindefiningcontagion,seeDornbusch,Park,andClaessens(2001).ArecentbookbyClaessensandForbes(2001)publishedthisandmorethan20otherarticlesfromaFebruary2000WorldBank/IMFconference,InternationalFinancialContagion:HowitSpreadsandHowitCanbeStopped.Thesearticlesincludetheoreticalmodels,aconceptualcontribution,countrycasestudies,andbroad-basedempiricalstudies.Otherimportantrecentcontribu