Are Investors Reluctant to Realize their Losses

AreInvestorsReluctanttoRealizeTheirLosses?TERRANCEODEAN*ABSTRACTItestthedispositioneffect,thetendencyofinvestorstoholdlosinginvestmentstoolongandsellwinninginvestmentstoosoon,byanalyzingtradingrecordsfor10,000accountsatalargediscountbrokeragehouse.Theseinvestorsdemonstrateastrongpreferenceforrealizingwinnersratherthanlosers.Theirbehaviordoesnotappeartobemotivatedbyadesiretorebalanceportfolios,ortoavoidthehighertradingcostsoflowpricedstocks.Norisitjustifiedbysubsequentportfolioperformance.Fortaxableinvestments,itissuboptimalandleadstolowerafter-taxreturns.Tax-motivatedsellingismostevidentinDecember.THETENDENCYTOHOLDLOSERStoolongandsellwinnerstoosoonhasbeenlabeledthedispositioneffectbyShefrinandStatman~1985!.Fortaxableinvestmentsthedispositioneffectpredictsthatpeoplewillbehavequitedif-ferentlythantheywouldiftheypaidattentiontotaxconsequences.Totestthedispositioneffect,Iobtainedthetradingrecordsfrom1987through1993for10,000accountsatalargediscountbrokeragehouse.Ananalysisoftheserecordsshowsthat,overall,investorsrealizetheirgainsmorereadilythantheirlosses.Theanalysisalsoindicatesthatmanyinvestorsengageintax-motivatedselling,especiallyinDecember.Alternativeexplanationshavebeenproposedforwhyinvestorsmightrealizetheirprofitableinvestmentswhileretainingtheirlosinginvestments.Investorsmayrationally,orirrationally,believethattheircurrentloserswillinthefutureoutperformtheircurrent*UniversityofCalifornia,Davis.ThispaperisbasedonmydissertationattheUniversityofCalifornia,Berkeley.Iwouldliketothankananonymousreferee,BradBarber,PeterKlein,HayneLeland,RichardLyons,DavidModest,JohnNofsinger,JamesPoterba,MarkRubinstein,PaulRuud,RichardSansing,RichardThaler,BrettTrueman,andparticipantsattheBerkeleyPrograminFinance,theNBERbehavioralfinancemeeting,theFinancialManagementAsso-ciationConference,theAmericanFinanceAssociationmeetings,andseminarparticipantsatUCBerkeley,theYaleSchoolofManagement,theUniversityofCalifornia,Davis,theUniver-sityofSouthernCalifornia,theUniversityofNorthCarolina,DukeUniversity,theWhartonSchool,StanfordUniversity,theUniversityofOregon,HarvardUniversity,theMassachusettsInstituteofTechnology,theAmosTuckSchool,theUniversityofChicago,theUniversityofBritishColumbia,NorthwesternUniversity,theUniversityofTexas,UCLA,theUniversityofMichigan,andColumbiaUniversityforhelpfulcomments.IwouldalsoliketothankJeremyEvnineandespeciallythediscountbrokeragehousethatprovidedthedatanecessaryforthisstudy.FinancialsupportfromtheNasdaqFoundationisgratefullyacknowledged.THEJOURNALOFFINANCE•VOL.LIII,NO.5•OCTOBER19981775winners.Theymaysellwinnerstorebalancetheirportfolios.Ortheymayrefrainfromsellinglosersduetothehighertransactionscostsoftradingatlowerprices.Ifind,however,thatwhenthedataarecontrolledforrebal-ancingandforshareprice,thedispositioneffectisstillobserved.Andthewinninginvestmentsthatinvestorschoosetosellcontinueinsubsequentmonthstooutperformtheloserstheykeep.Thenextsectionofthepaperdiscussesthedispositioneffectandlitera-turerelatedtoit.SectionIIdescribesthedatasetandSectionIIIdescribestheempiricalstudyanditsfindings.SectionIVdiscussesthesefindingsandSectionVconcludes.I.TheDispositionEffectA.ProspectTheoryThedispositioneffectisoneimplicationofextendingKahnemanandTver-sky’s~1979!prospecttheorytoinvestments.Underprospecttheory,whenfacedwithchoicesinvolvingsimpletwoandthreeoutcomelotteries,peoplebehaveasifmaximizingan“S”-shapedvaluefunction~seeFigure1!.Thisvaluefunctionissimilartoastandardutilityfunctionexceptthatitisde-finedongainsandlossesratherthanonlevelsofwealth.Thefunctionisconcaveinthedomainofgainsandconvexinthedomainoflosses.Itisalsosteeperforlossesthanforgains,whichimpliesthatpeoplearegenerallyrisk-averse.Criticaltothisvaluefunctionisthereferencepointfromwhichgainsandlossesaremeasured.Usuallythestatusquoistakenastheref-erencepoint;however,“therearesituationsinwhichgainsandlossesareFigure1.Prospecttheoryvaluefunction.1776TheJournalofFinancecodedrelativetoanexpectationoraspirationlevelthatdiffersfromthestatusquo....Apersonwhohasnotmadepeacewithhislossesislikelytoacceptgamblesthatwouldbeunacceptabletohimotherwise”~KahnemanandTversky~1979!!.Forexample,supposeaninvestorpurchasesastockthatshebelievestohaveanexpectedreturnhighenoughtojustifyitsrisk.Ifthestockappre-ciatesandtheinvestorcontinuestousethepurchasepriceasareferencepoint,thestockpricewillthenbeinamoreconcave,morerisk-averse,partoftheinvestor’svaluefunction.Itmaybethatthestock’sexpectedreturncontinuestojustifyitsrisk.However,iftheinvestorsomewhatlowersherexpectationofthestock’sreturn,shewillbelikelytosellthestock.Whatif,insteadofappreciating,thestockdeclines?Thenitspriceisintheconvex,risk-seeking,partofthevaluefunction.Heretheinvestorwillcontinuetoholdthestockevenifitsexpectedreturnfallslowerthanwouldhavebeennecessaryforhertojustifyitsoriginalpurchase.Thustheinvestor’sbeliefaboutexpectedreturnmustfallfurthertomotivatethesaleofastockthathasalreadydeclinedthanonethathasappreciated.Similarly,consideraninvestorwhoholdstwostocks.Oneisup;theotherisdown.Iftheinvestorisfacedwitha