The Effect of the 2003 Dividend Tax Cut on Corpora

ASSA2006SESSION:NewEvidenceAbouttheImpactofTaxingCorporate-SourceIncome(H2)Presiding:JOELSLEMROD,UniversityofMichiganTheEffectofthe2003DividendTaxCutonCorporateBehavior:InterpretingtheEvidenceByRajChettyandEmmanuelSaez∗∗RajChetty,[e-mail],EmmanuelSaez,[e-mail],UniversityofCalifornia,DepartmentofEconomics,549EvansHall#3880,Berkeley,CA94720.WethankAlanAuerbach,MartinFeldstein,JohnFriedman,GustavoGrullon,JoelSlemrod,andconferenceparticipantsforveryhelpfulcommentsanddiscussions.JosephRosenbergandJamesSlyprovidedoutstandingresearchassistance.FinancialsupportfromNSFgrantsSES-0134946andSES-0452605isgratefullyacknowledged.The2003dividendtaxreformhasgeneratedrenewedinterestinunderstandingtheeco-nomiceffectsofdividendtaxation.Thereformintroducedfavoredtaxtreatmentofindividualdividendincome,wherebydividendsaretaxedatarateof15percentinsteadoffacingtheregularprogressiveindividualincometaxschedulewithatoprateof35percent.Severalrecentstudieshaveusedthe2003taxcutasa“naturalexperiment”tolearnabouttheeffectsofdividendtaxationoncorporatebehavior.Thesestudieshaveobtaineddivergentempiricalresultsdespiteusingthesameunderlyingdata.Thegoalofthispaperistore-examinetheevidenceusingnewlyavailabledataandreconcilesomeofthecontradictoryresultsinthisrecentliterature.Wefocusonthreequestions:(1)Didthetaxcutcausethesurgeindividendsorwereotherfactorsresponsible?(2)Didthetaxcutinducesubstitutionofrepurchasesfordividends,ordidtotalpayoutrise?(3)Didthetaxcutinducemoreefficientdistributionofinvestmentfundsacrossfirms?1DataandBasicAnalysisofDividendPayoutsInthissection,weextendtheseriesreportedinourearlierstudy(RajChettyandEmmanuelSaez,2005)toincludenewlyavailabledatathrough2005Q2.Asinallpreviousstudies,weusedatafromtheCenterforResearchinSecurityPrices(CRSP)forourprimarilyanalysisofdividendpayments,supplementedwithdatafromCOMPUSTATforothervariablessuchassharerepurchases.1WefocusonthesampleofallCRSPfirmsexcludingforeignfirms,financialfirms,andutilityfirms.Wecallthissamplethe“core”sample,whichcontainsonaverage5,000firmsineachquarter.Weclassifydividendsintoregulardividendsandspecialdividends.Regulardividendsareperiodicandrecurrent(ingeneralquarterly),andtendtobeverysmooth.Firmsveryrarelycutregulardividendpayments(assuchacutsendsabadsignaltothemarket),andhenceinitiateorincreasedividendsonlywhentheyareconfidenttheycansustainsuchpaymentsinthefuture.Specialdividendsareone-time,non-recurringevents.Threeeventsareparticularlyrelevantforouranalysis.First,thereformwasinitiallyproposedonJanuary7,2003.Second,thetaxcutwasofficiallysignedintolawonMay28,2003,andwasmaderetroactivetothebeginningof2003andsettoexpireattheendof2008.1CRSPcontainsfinancialdataonallcompanieslistedontheNYSE,AMEX,andNASDAQ,whicharethethreelargeststockexchangesintheUnitedStates.1Finally,afterthere-electionofPresidentBushinNovember2004,itbecamealmostcertainthatthereformwouldindeedlastatleastuntil2008.Thesethreekeydatesareshownonthegraphsbelowusingverticallines.Figure1plotsaggregateregulardividendsforthecoresampleataquarterlyfrequencybetween1981-Q3and2005-Q2inreal2004dollars.Totalregulardividendsstagnatedaround$25billionfrom1998to2002andthenrosetoabout$33billionby2005.Alargefractionoftheincreasetookplaceinthelasttwoquartersof2003afterthetaxcutwassignedintolawinlateMay,2003.Dividendssurgedagainbyapproximately$2billioninthefirst2quartersof2005,justafterBushwasre-electedandthetaxcutsecureduntilatleast2008.Hence,thedividendtaxcutappearstohaveinducedafairlylong-termshiftinpolicyasopposedtoaone-timesurgeindistributions.Ifthepost-2003increasesinaggregatedividendsaredueexclusivelytothetaxcut,thereformappearstohaveraisedaggregateregulardividendsbyroughly30percentrelativetothelevelin2002-Q4.However,theseincreasesmaynotbedueentirelytothetaxcutbecausedividendsalsoincreasegraduallyovertime(thoughneverasfast)inearlierperiods.Figure1alsoplotsthetotalamountofspecialdividends,whichalsoincreasedsubstantiallyafterthe2003taxcut.Thespikein2004-Q4isdueentirelytoa$32billionspecialpaymentbyMicrosoft.[INSERTFIGURE1HERE]AsdescribedinChettyandSaez(2005),itisdifficulttomakepreciseinferencesabouttheeffectofthedividendtaxcutonaggregateamountsbecauseofentryandexiteffectsandtheconcentrationofthedividendpaymentsdistribution.Clearerevidencecomesfromthetimeseriesofregulardividendinitiationsandterminations,whichareunaffectedbytheseeconometricissues.Figure2showsthatinitiationssurgedafterthelawwasenacted.Thenumberofinitiationsinthethreequartersimmediatelyfollowingenactment(2003-Q3,2003-Q4,and2004-Q1)arethethreehighestamongthe80quartersweconsider.Thespikeininitiationsdiesdownrapidlyafterthetaxcut,asonemightexpect,untiltheBushelection.[INSERTFIGURE2HERE]22DisappearingDividendsandtheCausalityoftheTaxCutInaninfluentialstudy,EugeneFamaandKennethFrench(2001)observedthatthefractionofpubliclytradedfirmspayingdividendshasdeclinedsteadilyfrom1980to2000,atrendtheyterm“disappearingdividends.”BrandonJulioandDavidIkenberry(2004)extendtheFamaandFrenchanalysisanddocument“reappearingdividends”startinginlate2000inthecoreCRSPsample.Thisresult