[14]PPT-investor-inattention-and-friday-earnings-a

InvestorInattentionandFridayEarningsAnnouncementsSTEFANODELLAVIGNAandJOSHUAM.POLLETHeYuRenQianQuestionsdoeslimitedattentionamonginvestorsaffectstockreturns?Post-earningsannouncementdriftBehavioralexplanation:•Dispositioneffect(BernardandThomas,1989,ChanJegadeeshandLakonishok,1996)•Fluctuationinoverconfidence(Frazzini,2006andGrinblattandHan,2005)•Beliefsaboutmeanreversion(Barberis,Shleifer,andVishnny,1998)•Underreactiontoinformationduetocognitivelimits(Hongandstein,1999)makesthepredictionthatdistractionsincreasethedriftExamineadecisioninwhichattentiontonewinformationplaysacrucialrole:theresponsetoearningsurprises.Immediateresponseshouldbelesspronounced;ThedelayedresponseshouldbeofgreatermagnitudeannouncementFridayotherweekdaysWeekendsdistractinvestorsandlowerthequalityofdecisionmakingyesnoAssume:ifinattentioninfluencesstockprices,weshouldobservelessimmediateresponseandmoredriftforFridayannouncement.Empiricalresults:Fridayannouncementshavea15%lowerimmediateresponseanda70%higherdelayedresponse.Othertestingmethod:aportfolioinvestingindifferentialFridaydriftearnssubstantialabnormalreturns;tradingvolumeis8%loweraroundFridayannouncementsMeaning:Thesefindingssupportexplanationsofpost-earningsannouncementdriftbasedonunderreactiontoinformationcausedbylimitedattention.contents•SectionImodel•amodeloftheresponseofstockpricestosignalsaboutearnings.•Givenlimitstoarbitrageintheformofriskaversion,thedistractedagentsaffectprices,sodistractionincreasesthepost-earningsannouncementdrift.Thecombinedresponsetotheannouncementisnotaffectedbythedistractedinvestors•SectionIIdataandsummarystatistics•A.data•B.summarystatistics•characterizesampleofearningsannouncementsfromJanuary1995untilJune2006•devisearuletoimprovetheaccuracyofannouncementdate•SectionIII•A.graphicalevidence•A1immediateresponse•A2delayedresponse•Btopandbottomquantiles•B1immediateresponse•B2delayedresponse•B3longtermresponse•B4summary•Cdelayedresponseratio•C1summary•Dregressionsincludingallannouncements•D1summary•Eportfolioreturns•E1summary•evaluatetheimmediateanddelayedstockreturnresponsetoinformationbycomparingthetopandbottomquantileoftheearningssurprise,andcombineinasummarymeasure•examinetheimmediateanddelayedresponsefortheentiresampleofearningsannouncements•constructportfoliostomeasurethedifferentialpost-earningsannouncementdriftforFridayannouncements•purchases(longposition)theFridaydriftandsells(shortposition)thenon-Fridaydriftearnsabnormalreturns•SectionIVvolumeresponse•Assume:IfinvestorsaremoredistractedonFriday,wealsoexpectlowerabnormal•Empiricalresults:volumeisindeed8%lowerforFridayannouncementsthanfornon-Fridayannouncements•meanings:provideanexplanationfortheknownfindingsofworseearnings(anddividend)announcementsonFriday(Penman(1987),Damodaran(1989),Bagnoli,Clement,andWatts(2005))I.modelGiven•allinvestorsareexposedtoasignal,butthefractionofinvestorsthatpayattentiontothissignalvaries•ManagerschoosetoannouncethesignalA.setupDt+1=δ+st+εt+1•stεtcontemporaneouslyindependentandidenticallyandindependentlyacrossperiods•1−μtofinvestorsobservesthesignalandafractionμtofinvestorsdoesnot.Assumeallwealthisconsumedattheendofeachperiod(quarter).•attentiveinvestors(whereiμt)•distractedinvestors(wherei≤μt)•B.PortfolioChoice•ahighdistractionday,denotedμt=μh•alowdistractionday,denotedμt=μl•Riskaversionriskaversionparameterγ0thedemandfortheriskyassetis•C.Equilibrium•btincreasespricestobelessresponsive•D.ResponsetoSignal•dollarexcessreturnfromt−1totasZt=Pt+Dt−(1+R)Pt−1•newinformation:thesignalstandtheunexpecteddividendDt−Et−1[Dt]•Et[LRt+1]=Et[IRt]+Et[DRt+1]•PROPOSITION1:•Et[IRt]isalinearfunctionofst,slopecoefficient(1−bt)/(1+R).•Et[DRt+1]isalinearfunctionofst,slopecoefficientbt•Et[LRt+1]isalinearfunctionofst,slopecoefficient1/(1+R).•immediateresponseproportional(1−bh)/(1−bl).•SAandSBbesignalsofdifferentquality,whereSASB•DRRt=(E[DRt|SA,μt]−E[DRt|SB,μt])/(E[LRt|SA,μt]−E[LRt|SB,μt]).•COROLLARY1:•(i)ThedelayedresponseratioDRRt=bt(1+R).•(ii)DRRtisincreasingintheshareofdistractedinvestorsμt,DRRt=0ifμt=0.•(iii)μt≥DRRt.•PROPOSITION2:•(i)Managersthatmaximizelong-termvalueareindifferent•betweenμlandμh.•(ii)Managersthatmaximizeshort-termvaluechoose•μt=μhwheneverst≤ItfollowsthatE[st|μt=μh]E[st|μt=μl]andE[IRt|μt=μh]E[IRt|μt=μl].II.DataandSummaryStatisticsA.DatabeginwithallquarterlyearningsannouncementsfromI/B/E/Sforwhichatleastoneanalystformsanearningsforecastinthe90daysbeforetheannouncement•restrictthesamplehavestockreturndatainCRSPandreportedinbothI/B/E/SandCOMPUSTATwithadifferenceofatmostfivecalendardaysbetweenthereportedannouncementdates•resultingsample:228,651announcementsfromJanuary1984toJune2006•quantifytheaccuracyofdatesrandomlyselect2,487earningsannouncementsfortheperiod1984to2002•theoptimalimputationruleforthedatediffersforthreecategoriesofannouncements①I/B/E/SandCOMPUSTATannouncementdatesdiffer:imputethedatetobetheearlierone.②BeforeJanuary1,1990:I/B/E/SandCOMPUSTATannouncementdatesagree:theWallStreetJournaldate③AfterJanuary1,1990:I/B/E/SandCOMPUSTATannouncementdatesagree:I/B/E/Sa