What does tax aggressiveness signal Evidence from stock price reactions to news about - PDF document

Whatdoestaxaggressivenesssignal?EvidencefromstockpricereactionstonewsabouttaxshelterinvolvementMichelleHanlon,JoelSlemrodRossSchoolofBusinessattheUniversityofMichigan,UnitedStatesarticleinfoabstractArticlehistory:Received11September2007Receivedinrevisedform12September2008Accepted16September2008Availableonline23September2008 JournalofPublicEconomics93(2009)126141 WearegratefultoEstelleDauchy,KylePeterson,andNemitShroffforableandtirelessresearchassistance.WeappreciatecommentsonearlierdraftsfromJosephBankman,DhammikaDharmapala,EdOutslay,TerryShevlin,andworkshopparticipantsattheNationalBureauofEconomicResearchFinancialAc 0047-2727/$seefrontmatter©2008ElsevierB.V.Allrightsreserved.10.1016/j.jpubeco.2008.09.004 ContentslistsavailableatScienceDirectJournalofPublicEconomicsjournalhomepage:www.elsevier.com/locate/econbase Furthermore,ifarmisidentiedasataxshelterpurchaser,thermmaybearreputationalandpoliticalcostsofbeinglabeledacorporatecitizen.ThemissionstatementofGeneralElectric'staxdepartmentincludesapartthatstatesthattaxstrategiesshouldnotbeharmfultothecompany'sreputation.TheyincludereputationasataxriskcategoryanddescribethecriteriaforevaluatingthistypeofriskforaparticularstrategyastheWallStreetJournalTest(e.g.,woulditlooknegativeifthecompanywerediscussedonthefrontpageoftheWallStreetJournalforthestrategy?).Publicstatementsattestingtorms'taxpaymentssupportthishypothesis.Forexample,afterthereleaseofa2004CitizensforTaxJustice(CTJ)studythatidentiedSBCasalowtaxpayer,theVicePresidentandControllerofSBC,JohnJ.Stephens,assertedthatWecarryanenormoustaxburdenforthiscountry;forthatnottobemadeclearseemsveryinappropriateWeisman,2004,p.E02).Inthelettertoshareholdersthatopensthe2004Wal-MartAnnualReport,LeeScott,PresidentandCEO,tellsshareholders:IalsowillreportthatWal-Martpaid$4billioninU.S.federalincometaxesinscalyear2004.ToborrowapagefrommyfriendWarrenBuffett'sannualreportlettertotheshareholdersofBerkshireHathaway,thismeansthatif446othertaxpayerspaidthesameamountasWal-Mart,nootherbusinessorindividualintheUnitedStateswouldhavehadtopayfederaltaxeslastyear(basedon2003U.S.federaltaxreceiptsof$1.782trillion).Onemightinferfromthesestatementsthatatleastsomermsperceiveapoliticalcosttobeinglabeledastaxaggressive.Inaddition,insomeinstancesinvestorsmayinterpretnewsaboutarm'staxaggressivenessasevidencenotonlyaboutabehaviortowardtheIRS,butalsoaboutinsiders'willingnesstobeaggressivewiththeinvestorsaswell(e.g.,Desaietal.,2007).IfshareholderssuspectthatmanagementswhoareaggressivewiththeIRSarealsoaggressivein,forexample,reportingtheiraccountingearnings,thenthemarketmaygrowsuspiciousoftheaccuracyofthecompany'snancialstatements.Totheextentthatthishappens,themarketreactiontothenewsthatarmistaxaggressivecouldbenegative.Intheory,then,newsabouttaxaggressivenesscouldeitherboostordepressarm'sstockprice.ThetensioniswonderfullyillustratedbytheresponseofWhirlpool'staxcounsel,RobertKenney,tothenewsthatWhirlpoolhadthecurrenteffectivetaxrateonthelistofrmsnamedbythe1984CTJstudy:It'sadoubleedgedsword.WeoweittoourshareholderstotakelegitimateandlegalmeanstokeepdowntaxesBirnbaumandMurray,1987,p.12).Thispaperattemptstosortoutthemultipleavenuesbywhichnewsofcorporatetaxaggressivenessmightaffectthemarketvalueofcorporationsbyinvestigatingthemarketreactiontoaninitialpressmentionthatarmwasinvolvedinacorporatetaxshelter.Westudytheseeventsbothbecausetaxsheltersareunderintensescrutiny,andbecausethereisverylittlerigorousquantitativeresearchonthermsthatengageintaxsheltering.Weconstructoursamplebysearchingpressarticles,announcementsandnewswiresforthetermtaxshelteralongwithcorporatecorporation,andthenrestrictingthesampletocaseswherethenewsreferstosometypeofaccusationorannouncementinthepressthatthermengagedinataxshelter.Wedothiswhilefullyrecognizingthatthereisnouniversally-accepteddenitionofataxshelter,andthatsomeofthermsinoursamplementionedinthepressasengaginginataxshelterdidnoteverreceiveaNoticeofDeciencyfromtheIRSregardingtheirparticipationinthementionedshelter.ndthat,onaverage,acompany'sstockpricedeclineswhenthereisnewsaboutitsinvolvementintaxshelters.Wealsondsomelimitedevidenceregardingwhatexplainsthecross-sectionalvariationinthereaction.Forexample,thereactionismorenegativeforrmsintheretailsector,suggestingthatpartofthereactionmaybeaconsumer/taxpayerbacklash.Inaddition,withrelativelyhighdisclosedcasheffectivetaxrateshavealessnegativereaction,consistentwiththemarketreactingpositivelytoevidencethatthesermswerenotastax-passiveaspreviouslybelieved.Wealsoinvestigatewhetherthegovernancestructureofthermaffectsthemarketreaction,andndamorecomplexrelationthandescribedinDesaiandDharmapala(inpress).Wenotethatsomeofthesecross-sectionalresultsaresensitivetotheproxyused,forexample,themeasureacompany'sgovernancequalityoritsvulnerabilitytoapublicbacklashregarding(lackof)corporatecitizenship.Thepaperproceedsasfollows.Section2summarizesearlierresearchconcerningthemarketreactiontocorporatemisdeeds.Section3discussesthepotentialsourcesforthemarket'sresponsetotaxnoncompliancenewsanddevelopssomehypothesesaboutwhatcharacteristicsmightaffectthestockpricereaction.Section4describesourresearchdesignanddata,andSection5presentsourempiricalresults.Section6examinesthestockpricereactiontoadifferentkindofnewsabouttaxaggressiveness,andSection7concludes.2.Relatedliterature2.1.ThemarketresponsetocorporatemisdeedsEarlierliteraturehasanalyzedthemarketpricereactiontonewsregardingawiderangeofcorporatefraud,includingnon-tax-relatedfraudagainstthegovernment.Notably,though,thisliteraturegenerallyignoresthepossibilitythatshareholdersmightwantthecompanytobeoptimallyaggressiveintheirdealingswithothercompaniesandwiththegovernment. TakenfromthepresentationmadebyRickDAvino,VPandSeniorTaxCounsel,GECapitalandNBCUniversal,attheTaxCouncilPolicyInstituteConference,February2007.Bankman(2004)mentionsnegativepublicityasafactorinexplainingcorporatedemandfortaxshelters,andnotesthatitmaybegoodorbad.Throughoutthepaperweusethetermtaxaggressivenessinreferencetothepotentialbehaviorofourtaxshelterrmsample.Forconvenienceinthepaperwesometimesusethetermstaxaggressivenessandtaxshelteringinterchangeably,although,werecognizethattaxsheltersareasubsetoftaxaggressivebehaviors.Thetermtaxshelterhasnouniversallyaccepteddenition,andcanbeusedinavarietyofwaysdependingonthecontext.Ouruseofthetermreferstocomplextransactionsusedbycorporationstoobtainsignicanttaxbenetsprobablyneverintendedbythetaxcode;thesetransactionsmaynotbeillegalperseandtheiruse,ifdetected,maytriggerlengthyprocessesofIRSassessmentandjudicialappeal.Ourempiricalanalysisreferstopressusesofthisterm,andthereforerelatestotheinterpretationgivenbythepressoutletsthemselves.Weintendtoprimarilycaptureoff-the-shelftaxshelterstrategiespromotedbylaw,investmentbanking,andaccountingrms.TheAppendixcontainsadescriptionofthesheltersweincludeinoursample.WerecognizethattherearecertainlyothertaxsheltersinexistencForoneexception,seeGrahamandTucker(2006)Slemrod(2004)reviewstheeconomicissuespertainingtocorporatetaxaggressiveness.Alistofthetaxshelterrmsincludedinoursampleisavailablefromtheauthors.127M.Hanlon,J.Slemrod/JournalofPublicEconomics93(2009)126 Forexample,theideathatfuturenesandpenaltiesconstitutepartofthemarket'sresponsehasbeenstudiedpreviouslyinthecontextsofnon-taxmisdeedsbyKarpoffandLott(1993)andBoschandEckard(1991).Thattheresolutionofacourtcasecouldhaveimplicationsforfuturecase-relatedincomestreamsintermsofadeclineintherm'sexpectedafter-taxprots,assumingthatthecase-relatedactivitieswillcease,becurtailed,orwillbecontinuedatahighercostofconcealment,isconsideredbyKarpoffandLott(1993)BoschandEckard(1991)addressthepossibilitythatnewsofonecaseofaggressivenesscouldraiseconcernsthatthecompany'smanagementisengagedinbeingaggressivewitheveryone,distinguishingtwocomponents.First,customersandsuppliersmightbecomewaryofdealingwiththerm,therebyincreasingfuturetransactioncostsandperhapscausingcustomersandsupplierstodealwithothercompanies(asinKleinandLefer,1981).Second,thecasemightsignalthatthermcouldbeengaginginotherasyetunknownaggressiveactivitythatcouldleadtoprosecutionandassociatedcosts.Inaddition,itmaysignalthatthedishonestyextendstothenancialaccountingstatementsinotherwordsmanagementislyingtotheshareholders(seeDesaiandDharmapala,inpressandDesaietal.,2007BoschandEckardalsoraisethepossibilitythatnewsofaggressiveness,especiallyaggressivenessdetectedandpotentiallysanctioned,couldbeasignalofmanagementincompetenceassumingthatsmartmanagerswouldn'tgetcaughtBoschandEckard,1991,p.310)oritcouldrevealthatthemanagersmisestimatedtheprobabilityofbeingcaught.2.2.EventstudiesofcorporatemisdeedsunrelatedtotaxaggressivenessThereisaconsiderableliteraturethatusesevent-studymethodologytoexaminetheeffectsonstockmarketvalueofinformationrevealedaboutcorporatemisdeeds.Forexample,Ellert(1975)examinesthemarketpriceeffectof566rmsindictedinhorizontalpricexingconspiracycasesbetween1935and1971andndsastatisticallysignicantabnormalreturnof1.1%intheindictmentmonth.Strachanetal.(1983)lookat47pricermsinthe1970sandreportthatthesermshavestatisticallycantnegativeabnormalreturnsonthereportingdateandonthedaybeforeofabout0.8%and0.6%,respectively.etal.(1982)studydailyreturnstoexaminetheimpactofantitrustsuits,includingpricexingsuits,on34rmsbetween1937and1974,andndanaveragesample-wide6%declineinshareprice.Inaddition,BoschandEckard(1991)examinethestockpricereactiontofederalindictmentsforpricexingfor127rmsduringtheperiod1962to1980,andndanaverageabnormalreturnof1.08%;theyestimatethatonly13%ofthatcanbeattributedtovariouslegalcostsandarguethattheremainderisduetothepresentvalueoflostmonopolyproKarpoffandLott(1993)studythestockpricereactiontoannouncementsduring1978to1987concerningcorporatefraudcasesinwhichthedamagedpartydoesbusinesswiththeaccusedndthatreportsofregulationviolationsdonottriggersignicantabnormalreturns,althoughreportsofotherfraudsdo.Forexample,theyreportthatforinitialpressreportsofcorporatefraudagainstprivateparties,stockpricedeclinedbyanaverageof1.22%;forfraudsagainstgovernmentagencies(non-tax),thelossinvalueisonaverage1.67%;andfornancialreportingfrauds,thenegativereactionwas4.66%.Finally,wheninvestigatingallegationsofenvironmentalviolations,Karpoffetal.(1999)reportthatthereisanaveragenegativeabnormalstockpricereactionof1.58%whentheallegedviolationisrstannouncedandanegative1.92%returnwhenthermischargedwithorsuedforaviolation.Morerecently,therehavebeenstudiesthatinvestigatethemarketreactiontonewsrelatedtospecinancialaccountingmisdeedssuchasbeingaccusedofaviolationofGenerallyAcceptedAccountingPrinciplesbytheSecuritiesandExchangeCommission(SEC)andofannouncingarestatementofthenancialstatementsoftheForexample,Dechowetal.(1996)provideevidenceforasampleof92rmsaccusedbytheSECofwrongdoingthat,upontheannouncementthatthemanipulatedearnings,stockpricesdeclinedbyanaverageof8.8%.Consistentwiththeseestimates,Palmroseetal.(2004)negativeaverageabnormalreturnsofabout9%overa2-dayannouncementwindow;inaddition,theabnormalreturnismorenegativewithindicationsoffraud,thelargerthesizeoftheadjustment,andforrestatementsattributedtotheauditors.Theyalsondthatthereactionisgreater,ceterisparibus,ifnoamountisgivenintheannouncement.Theauthorshypothesizethatthenegativesignalassociatedwithfraudandauditor-initiatedrestatementsisassociatedwithanincreaseininvestors'expectedmonitoringcosts,whilehighermaterialityisassociatedwiththegreaterrevisionsoffutureperformanceexpectations.Insum,earlierliteraturehasgenerallyfound(oftenlarge)negativestockmarketresponsestocorporatemisdeeds,buthasnotinvestigatedthemarketresponsetonewsaboutcorporatetaxaggressiveness. Inanattempttoseewhetherinstitutionalinvestorsthinkpoorlyofrmsengagedintaxshelters(ortax-aggressivebehavior),wesearchedtheWebforarticlesthatmentionedtheCaliforniaPublicEmployeesRetirementSystem(CalPERS)andtaxshelters(oraggressiveness).WefoundseveralarticlesmentioningthatCalPERSthreatenedtodivestandblockstatepurchasesofstocksandbondsofU.S.companiesthatmakeuseofforeigntaxhavens.AlthoughonearticlequotedPhilipAngelides,CaliforniaTreasurer,assayingforeigntaxsheltersarejustanotherwaytogamethesystem(SFGate.com,July25,2002),otherarticlespresentCalPERSsconcernasbeingaboutshareholderrightsratherthantheavoidanceoftax.Forexample,ontheCalSTRS(CaliforniaStateTeachersRetirementSystem)WebsiteisanarticleaboutinstitutionalinvestorstakingoutafullpageadintheUSATodaycallingforIngersollRand(whichhadmoveditsheadquarterstoBermudafortaxreasons)tomovebacktotheU.S.citingasubstantiallossofshareholderrightsthatgoeswithoff-shoreincorporationhttp://www.calstrs.com/Newsroom.Whats%20New/callforvote.aspxFurther,aTaxBriefsarticlefromLevinandWeiser,LLC(March3,2003)statesthatCalPERSbelievesBermudaincorporationmakesitmoredifcultforTycoshareholderstoholdthecompany,itsofcersanddirectorslegallyaccountableintheeventofwrongdoing(page6).KarpoffandLott(1993)distinguishfourkindsoffraud,includingfraudsofgovernments,buttheserefertocasesinwhichtheaccusedrmcheatedorwasaccusedofcheatingonimplicitorexplicitcontractswithagovernmentagency,suchasoverchargingonmilitarycontracts.Taxevasioncasesarenotbecausethefocusoftheanalysisarecasesinwhichsomeorallofthecostsofthefraudcanbeinternalizedbythermthroughitsrepeatedinteractionswithcustomers,suppliers,employees,andinvestors.Oneoftheotherthreesubcategoriesisnancialreportingfrauds,whichincludescaseswhereagentsofthemisrepresentedorareaccusedofmisrepresentingthenancialcondition.Anotherisregulatoryviolations,suchasfailuretoreportcurrencytransactions,inwhichitisnotclearthatthermviolatedanimplicitorexplicitcontractwithaninvestororstakeholder.RestatementsdonotnecessarilyinvolveanaccusationofwrongdoingbytheSEC.However,thermrestatesbecausethepriornancialstatementswereincorrectandthustherestatementconstitutesanexplicitacknowledgementthatexistingnancialstatementsdonotconformtoGAAPPalmroseetal.,2004128M.Hanlon,J.Slemrod/JournalofPublicEconomics93(2009)126 2.3.RelevantempiricaltaxresearchForreasonsweelaborateonbelow,eventsrelatedtotaxaggressivenesssharesome,butnotall,characteristicswiththenon-taxeventsdiscussedabove.Whiletherearenopreviousanalysesofthestockmarketreactiontonewsoftaxaggressiveness,thereareseveralrelatedstudies.DesaiandDharmapala(inpress)investigatehowinvestorsvaluemanagerialactionsdesignedsolelytominimizecorporatetaxobligations.Theyproceedbyregressing,overacross-sectionofcompanies,Tobin's(marketvaluedividedbythereplacementcostofassets)onaproxyfortaxavoidance,measuredasanestimateofthebook-taxdifferencesofthermlessanestimateoftheportionofthebook-taxdifferencesduetoearningsmanagement(i.e.,thetotalaccrualsoftherm).Theyndthattheirproxyfortaxavoidanceispositivelyrelatedtormvalueforwell-governedrms,butinsignicantlyrelatedtormvalueforpoorlygovernedrms.Theauthorsinterprettheirevidenceasconsistentwithagencycostsmitigatingthebenetstoshareholdersofcorporatetaxavoidance.Inotherwords,themanagers'taxshelteringdecisionsarerelatedtotheirabilitytodivertvalue,sothatinpoorlygovernedrmsthetaxshelteringsignalsahigherlikelihoodofmanagerialwealthdiversionandthusonnetaddsnovalue.Note,though,thatDesaiandDharmapala(inpress)isnotaneventstudy.Incontrast,theanalysiswereportbelowexaminesadditionalcross-sectionaldeterminantsbeyondgovernanceanddoesnotrelyonproxiesoftaxaggressivenesssuchasbook-taxdifferences(whichcanbeduetomanyfactors).Instead,itusesthepressmentionofarminataxsheltertoexaminethemarket'sperceptionofthebehavior.Thereisalsoresearchexaminingthewealtheffectsofchangesinthetaxtreatmentofspecictransactionsorcompaniesundertakingspecictax-favoredtransactions.DhaliwalandErickson(1998)investigatethemarketresponsestocourtrulingsabouttheamortizationofintangibleassets,alowercourt'sdisallowanceofthedepreciationofcertainintangibleassets,followedbytheSupremeCourt'sreversalofthisdecision.Theyndnegativeandpositivepricereactions,respectively,forrmsnotnamedinthecasebutwithsimilaracquiredintangibles.Althoughitsresearchdesignissimilartooursinthatitinvestigatesmarketreactionstoatypeoftaxnews,itsaimistogaugewhetherthemarketimpoundsfuturecashowsfromtaxdeductions(or,onecouldgeneralize,othertaxlawchanges)intosharepriceanddoesnotaddresstheconceptuallydistinctreputationeffectsofbeingcalledtaxaggressive.Inaddition,severalstudiesexaminethemarketresponsetoacompanyannouncingitsintenttoundergoacorporateinversionDesaiandHines,2002;Cloydetal.,2002;andSeidaandWempe,2004).Thistypeofstudyisquitedifferentthanours,asitconcernsthenewsofacompanydecision,ratherthananexogenouseventrelatedtocompanydecisions.Themarketresponsetoa(e.g.,tax-related)companydecisionwouldbenegativeonlyifthemarketdisagreedwiththedecisionmakers'assessmentofitsvalue.Ourstudyoftaxshelternewstsmorecloselywiththestudiesofthestockpricereactiontonon-tax-relatedenforcementactions.Thenewsgenerallyconcernsagovernmentenforcementactionorasimplepressmention(withoutenforcementaction)pertainingto,althoughpossiblyalsorelatedongoing,actionsofthecompany.Becausetheeventtestedisnotarmaction,onecannotpresumethatitisvalue-enhancing.Nopreviousstudyhasaddressedthissetofissues.3.AsimplemodelofthemarketreactiontonewsoftaxaggressivenessManyoftheseconsiderationscanbeillustratedinasimplemodelofthemarketvaluationoftheafter-taxstreamofincome.Supposethatthesharepricedependsontheexpectedafter-taxincomegoingtotheshareholders,whichequalsthetrueincomeofthe,minusincomedivertedfromtheshareholdersbythemanagers,,minusexpectedtaxpayments.Weallowtheamountdivertedtobenegativelyrelatedtothecompany'sgovernance,representedbyExpectedtaxpaymentsareproportionaltotaxableincome,whichisminustheamountofshelteredincome,,andtheprobabilitythattheshelteringwillbedetectedandsubjecttoaproportionalpenaltyimposedatrateonimproperlyshelteredtax.Theprobabilityofdetectionandpenaltymayincreasewiththeamountofshelteringdone,so).Withtheseassumptions,wehave:ðÞ¼ðÞðTheshareholderspreferthatbechosentomaximize,whichwouldrequirethat=0and=[1(1+)]/['(1+where0.However,theyareconcernedthatthiswillnotbedone,perhapsbecausethemanagersmakedecisionsnottobutrathertomaximizeamanagerialutilityfunction)).Variousmeasures,suchasincentive-basedcompensation,areinstitutedtoinducethemanagerstomaximizeratherthan(.),butthesemeasuresareimperfect,andweabstractherefromtheircostandeffectiveness.Inthismodel,newsabouttaxshelteringmayhaveseveraldistincteffects.First,itmaychangeshareholders'viewsabouttheextenttowhicharebeingchosentomaximize.Itmaychangeshareholders'viewsabouttheeffectivenessofthetaxauthorityinpolicingtaxshelters,sotheirperceptionof(.)changes.Finally,forsomeamountofsheltering,callit,thatthenewsrevealshasbeendetectedandpenalized,theprospectivetaxsavingsfromthatshelterarelostandtheprobabilityofpenaltygoestoone;becausethisisnotanongoingevent,weattachadiscountfactor,,toit.Ignoringsecond-orderterms,wecanapproximatethechangeinexpectedafter-taxprotsasfollows:vetermsofexpression(2)eachhaveastraightforwardinterpretation.Therstterm,(1),concernsanychangeinincome,aftertax,thatmaycomefromaconsumerbacklashagainstwhatmightbeseenasun-civicbehavior,representedbyorrelatedtochangedperceptionsbybusinesscustomersofthereliabilityofthebusiness,whichmightdependonthegovernance Notethatinthismodel,unlikeinShackelfordetal.(2008),earningsreportedonnancialstatementsdonotaffectthesharepriceforagivenafter-taxincome.129M.Hanlon,J.Slemrod/JournalofPublicEconomics93(2009)126 ofthe.Thesecondterm,),isthedeclineinexpectedafter-taxincomeduetoanincreasedperceptionofinsiderdiversion.Thethirdterm,)(1+),isthechangeinexpectedincomethatresultsfromanincreaseintheperceivedvalueof.Thistermisnegativeif(i.e.,ifshelteringwasperceivedtobegreaterthanoptimal),butcanbepositiveif(ifthemarkethadthoughtthecorporationwasinsufcientlyaggressive);itiszeroifshelteringwaspreviouslythoughttobeoptimal(Thefourthtermofexpression(2),(1+),resultsfromanincreasedperceptionoftheeffectivenessofthetaxauthorityinpenalizingfuturetaxshelteractivity;thisisalwaysnegative,andislargerwhentheamountofshelteringthedoesislarger.Finally,thelastterm,)(1+,isthelossfromanynewsthatsomediscreteamountofsheltering,,willbepenalizedforsurewhen,beforethenews,therewasaprobabilityofthishappening.Dependingonthecharacterofthenewsandonthermvaluesof,theeffectofthenewsontherelativevaluesofwillvary.Althoughwecannotseparatelyidentifyeachofthese,weattempttoshedlightonthenatureofthestockmarketreactionbyinvestigatingcross-sectionaldifferencesinmarketreactions.First,weexaminethecasheffectivetaxrate(CASHETR)fromthenancialstatementsforthetwoyearspriortothenewsthatthermwasinvolvedinataxshelterasaproxyforthemarket'sexanteperceptionsregardingthetaxaggressivenessoftherm();thisiscalculatedasthecashtaxespaidsummedoverthetwoyearsdividedbypre-taxincomesummedoverthetwoyears.Ourhypothesisisthat,thehighertheCASHETR,thelesslikelythatthemarketwouldexpectthermtobesufcientlytaxaggressive,andthemorelikelyitistotakethetaxshelternewsasapositivesignalofoptimalaggressiveness.Asaresult,thehighertherm'sCASHETR,themorepositive,orlessnegative,thereactionuponthenewsannouncement.Second,wepredictthatrmsintheretailindustrythatdealdirectlywithconsumerswillhaveamorenegativereactionthanotherrms.Thesermsmaybemoresusceptibletobeingpubliclyperceived,andpenalized,forbeingunconscionableorunpatriotic.Finally,wehypothesizethatpoorlygovernedrmswillhaveamorenegativemarketreactionrelativetowell-governedforatleasttworeasons.First,itcouldbethatthemarketperceivesthatthewell-governedrmsaremorelikelytogetawaywiththeshelteringormorelikelytobeengagedinsmallerorlessriskyshelters.Second,suspicionscouldbearousedthatthepoorlygovernedrmsarenotsimplybeingaggressivewiththeIRSbuttheyareaggressivetowardseveryone,includingtheshareholders,sothat)islargerformorepoorlygoverned4.Sampleandresearchdesign4.1.SampleselectionToobtainoursampleofrms,werstdoabroadsearchofallsourcesintheFactivadatabase.Weusethesearchtermssheltercorporatecorporationtocaptureallarticlesorpressreleasesthatcontainthoseterms.WeconductthesearchoverthedaterangebeginningJanuary1,1990andendingSeptember1,2004(thedatewestartedthesearchprocess).Fromthissearchweobtainabeginningsampleof6293articles,newswirereleases,orothertypesofpressmentions.Wethenhadtworesearchassistantsreadthroughthearticles,retainingonlythosethatcontainthenameofacompanyassociatedinsomewaywithhavingataxshelter.Thetypesofarticlesdiscardedbythisprocesspertaintothetopicoftaxsheltersgenerally,withoutmentioninganyparticularcompany.Theremainingsampleofpotentialobservationscontain824articles.Weexcludearticlesaboutcorporateinversions,whichhavebeenstudiedseparately,becauseweviewtheseasafundamentalrestructuringofamultinationalcorporationratherthanasitengaginginasetoftaxreducingtransactions.Wealsoexcludearticlesabouttransferpricing,becausethesecasesarenotgenerallyconsideredtaxsheltersalthoughtheyaresometimescalledsuchbythepress.Weexcludearticlesaboutspecictaxprovisionsandthecompaniesthattakeadvantageofthem(legally),whichthepressreferredtoasataxshelteractivity,suchasForeignSalesCorporationsandtheAgriculturalLandTaxprovision.Finally,weexcludearticlesabouttaxesinothercountries(e.g.,thePetroleumReserveTax),articlesaboutprivatecompanies,andnon-income-taxshelterissues.Aftertheseexclusions,weretainasampleof601articles.Table1presentsasummaryofoursampleselectionprocess.Wethenfurtherreducethesamplebyexcludingmultipleobservations(408articles)ofthesamerm-shelter,retainingonlyrstarticleabouttherm-shelter.WeexcludeobservationswithmissingdataintheCRSPdatabase(45articles)andforvariousotherreasonsspelledoutinTable1.Thisleavesuswithasampleof108articles,pertainingto97rms,someofwhichareengagedinmultipleshelters.Wecreatefoursomewhatoverlappingsamplesfromtheseremaining108articles.Therstsample,whichistheonethatweutilizeforourmainanalyses,containstherstarticleforeveryshelterthatarmisin.Thearticlesinthissampledonothavetobeinamajorpresssource,butcanbeinanypressarticleorrelease.Inaddition,ifarmisintwodifferentsheltersthatrmwillhavetwodifferentarticles(dates)includedinthesample.Thissamplecontainsall108rm-sheltercombinations.Thesecond Weusethecasheffectivetaxrateratherthanthecurrenteffectivetaxratetoavoidinterpretationissuesresultingfromtherecordingofataxcontireserve(whichoverstatesnancialaccountingexpenserelativetothecashtaxesactuallypaid).Whenarmtakesanaggressivetaxpositionthattheythinkmightbedisallowedinthefuturethermistorecordtherelated(potentiallyowedinthefuture)taxexpensewhendeterminingitsnancialaccountingearnings.Thereserveforthepossiblefuturetaxesiscalledthetaxcontingencyreserve(andwasalsoknownasthetaxcushionpriortorecentregulatchanges).Therewaslittledisclosureofthisamount,however,priortotheenactmentofFinancialAccountingStandardsBoardInterpretationNumber48,whichoccurredafteroursampleperiodends.SeeDyrengetal.(2008)fordistributionalstatisticsanddataoncasheffectivetaxrates.Weuseataxratecalculatedoverthetwoyearspriortotheoftheshelterinvolvementratherthanforatimeperiodpriortoactualshelterinvolvement(whichisunknowntous).Inprinciple,thelattercouldbepositivelyassociatedwithtaxaggressivenessifhigh-tax-ratermshaveahigherincentivetoengageintaxshelters.Thetaxrateimmediatelyprecedingthenewswouldreectanyeffectonthetaxratefromtheshelteritself.Allmultinationalcorporationshavetosettransferprices.Thedisputesarise,ofcourse,becauseofhowaggressivethermsareintheirdecisiontosettheprice.130M.Hanlon,J.Slemrod/JournalofPublicEconomics93(2009)126 sampleincludesonlythemajorpress(denedasoneofthemajornewswires(DowJones,AP,orReuters),theWallStreetJournal,ortheWashingtonPostarticlesoftherm-sheltercombinationsintherstsample.Thus,thisgroupisasubsetoftherstsampleandcontains72rm-shelterobservations.Thethirdsamplecoversbothmajorpresssourcesandnon-majorpresssources,butexcludesrmsthatdiscussedtheshelterintheirnancialstatementspriortothedateoftherstpressmentionbecausethiswouldlessentheamountofnewsinthepressarticle.ToobtainthermsthatdiscussedtheshelterintheirSEClings,wesearchedthe10-Klingsofeachoftheshelterrmsfortwoyearspriortothedateofthepressarticle.Wefound15rmswhose10-Ksmadesuchamentionandthusareexcludedfromthisthirdsubsample(=93).Finally,wecreateasubsamplethatstartswiththefullsampleandthenexcludesobservationswherethermhadotherearningsnewsduringoureventwindow.Tocollectthesedata,wegatherearningsannouncementdatesand10-K,8-K,and10-Qlingdatesforoursamplerms.Ifanyoftheseannouncementsorlingsoccurredduringthe3-dayeventwindowofthepressarticleweexcludethatobservationfromthissubsample.Thisisanattempttoisolatemorepreciselythemarketreactiontobeonlyinresponsetothenewsarticle(=76).4.2.EventdateandunivariatestatisticaltestsWeuseanevent-studymethodologytotestthemarketreactiontonewsthatarmhasengagedintaxaggressivebehavior.Weexaminethe3-daywindowcenteredonthedayofthepressmention.Weincludethedaypriortothereleasetocaptureanyeffectofnewsavailabletothemarketbeforethestoryandthedayaftertoprovidetimeforthemarkettoreact.Weuseamarketadjustedmodelbasedonavalue-weightedindex(withdividends)toestimateabnormalreturns;thismodelsubtractstheCRSPmarketindexreturnfromacompany'sdailyreturntoobtainthemarketadjustedabnormalreturn(AR)foreachdayandcompany.Thedailyabnormalreturnsarethensummedtocalculatethecumulativeabnormalreturn(CAR)foragiventimeperiod.Wereportfourmeasuresofstatisticalsignicance.Therstistheunadjusted-statisticusingthestandarddeviationofthesampleCARs.ThesecondisthestatisticdescribedinPatell(1976),whichestimatesaseparatestandarderrorforeachsecurity-eventandassumescross-sectionalindependence.Third,wereportvar,whichcompensatesforthepossiblevarianceincreaseduringtheeventperiod;totheextenttheeventperiodisassociatedwithincreaseduncertainty(i.e.,greaterreturnvariability),theuseofhistoricalorpost-eventtime-seriesvariabilitymightunderstatethetruevariabilityoftheeventperiodabnormalperformance(KothariandWarner,2004).Finally,wereporttheresultsfromanon-parametricgeneralizedsigntest.The Table1SampleselectionNumberofarticlesfromasearchofFactiva1/1/19909/1/2004 PotentiallyvalidarticlessummarizedbyRAsArticlesaboutinversionsArticlesabouttransferpricingArticlesaboutforeignsalescorporations(FSCs)ArticlesaboutPetroleumReserveTaxArticlesaboutAgriculturalLandTaxBreakArticlesaboutprivatecompaniesornon-incometax 117601Subsequentarticlesaboutthesamerm-shelterObservationswithmissingdataArticlesaboutsalesleasebacksArticleswithavaguementionofshelteractivity(e.g.,Enronesq)ArticleswithothernewsinthesamearticleMiscellaneousincometaxbutnon-observations Remainingobservationsfornalsample108Non-incometaxissuesincludearticlesaboutissuessuchasforeigntaxacts(e.g.,StampDutyReserveTax),propertytax,trackingstocktaxissues,reinsurance,andHongKongaircraftleasing.Inaddition,articlesaboutthesellersofthetaxshelters(e.g.,MerrillLynch)areexcluded.Thesearticlesincludediscussionsofincometaxissuesbutarenotincludedinoursample.ExamplesofitemsnotincludedarearticlesaboutEnronandWorldCom.Weexcludetheseobservationsbecausetherewassomuchothernewsinthemarketaboutthosecompaniesatthesametime.WeexcludetwoarticlesaboutU.divisionsofforeigncompaniesbecausetheremaybeadifferentmarketreactionwithrespecttoforeigncompanies.Weexcludeoneobservationaboutainvolvementinasyntheticfuelinvestmentbecausethermhadveryclearlydescribeditsactivitiesintheseinvestmentsinpriorconferencecallsandearningsreports,onearticleaboutlow-incomehousingcredits,onearticleaboutTeresaHeinzKerryinvestinginmunicipalbonds,oneaboutacompany'sinvestmentinaircraftleasebonds,twoarticlesaboutcompanieslobbyingontransferpricingissues,threearticlesaboutinvestmentsinMonthlyIncomePreferredShares(MIPS)andFelinePrideinstruments,andthreearticlesaboutataxissueatNationalCityBankthatwaslatercorrectedtobeaboutCityNationalBank. WenotethatFactivadoesnotincludefulltextoftheNewYorkTimes.Whilewehaveonearticlefromaabstract(whichisincludedinFactiva),thissamestorywascoveredbytheWallStreetJournalWeestimateourregressionsoversubsamplesofthemajor-press-onlysampleaswell(e.g.,asubsamplethatexcludesrmswithotherearningsnewsinthesameeventwindow.)131M.Hanlon,J.Slemrod/JournalofPublicEconomics93(2009)126 nullhypothesisforthistestisthatthefractionofpositivereturnsisthesameintheeventperiodasintheestimationperiod.Welabelthisstatistic5.Results5.1.DescriptivestatisticsTable2presentssimpledescriptivestatisticsofthevariablesweuseinourtests.PanelApresentsthedataforthefullsampleandPanelBpresentsthedataforthearticlesinmajorpresssourcesonly.Forthemainsampleofallrm-shelterstheaverageCARis0.0053,andforthesampleofrmswherethepressmentionwasinamajornewssourcetheaverageCARis0.0120.Asmentioned,wemeasuretheCASHETRastherm'saveragecasheffectivetaxrate(cashtaxespaid(Compustatdata#317)/totalpre-taxbookincome(data#170))overthetwoyearspriortothepressmention.EventhoughthesermsaretaxshelterthecasheffectivetaxratesarenotstrikinglylowrelativetotheU.S.statutorytaxrateof35%.Becausethismeasureisnot Table2DescriptivestatisticsPanelA:AllMeanS.D.MinimumMedianMaximumCAR1080.0050.0380.1300.0040.203CASHETR1080.2810.1370.00.3040.5MISSINGCASHETR1080.2960.459001MISSINGCASHETRETR1080.0970.153000.5LOSS1080.0650.247001RETAIL1080.1570.366001ADVERTISING1080.0150.027000.123HIGHADVERTISING1080.1850.390001GOVSCORE1089.2223.65401015ENTRENCHSCORE(normalized)1080.3410.23200.3330.667NON-ENTRENCHSCORE(normalized)1080.3770.16700.3890.611MISSINGENTRENCHSCORE1080.1110.316001PanelB:Allrm-shelters,majornewssourceonlyMeanS.D.MinimumMedianMaximumCAR720.0120.0350.1300.0070.060CASHETR720.2690.13900.2870.5MISSINGCASHETR720.2640.444001MISSINGCASHETRETR720.8300.144000.5LOSS720.0830.278001RETAIL720.1670.375001ADVERTISING720.0150.027000.123HIGHADVERTISING720.1940.398001GOVSCORE728.9863.85109.515ENTRENCHSCORE(normalized)720.3130.23100.3330.667NON-ENTRENCHSCORE(normalized)720.3720.17300.3890.611MISSINGENTRENCHSCORE720.1250.333001NotestoTable2:CARsarethecumulativeabnormalreturnscalculatedovera3-daywindowcenteredonthedateofthenewsandcalculatedusingmarketadjustedreturns(i.e.,takingtherawreturnforthermlesstheCRSPvalue-weightedindexreturnforthesameday).CASHETRiscomputedbytakingthesumofcashtaxespaidovertwoyearspriortothenewsoftheshelteranddividingbythesumofworldwidepre-taxbookincomeoverthetwoyearspriortothenewsoftheshelter.IfthedenominatorofthisratioisalossweresettheCASHETRvariabletozero.Ifthedataarenotavailablefortwoyearspriortonewsoftheshelter,weuseonlytheprioryeardatatocomputetheCASHETR.Weusetheratioofthesumandnottheaverageoftheratiosinordertominimizethenumberoflossobservationsandtheeffectofextreme1-yearCASHETRs.Ifthermdoesnotdisclosecashtaxespaidwereplacethenumeratorwithcurrenttaxexpenseandwhenthatisnotavailablewereplacethenumeratorwithtotaltaxexpense.Wethencreateaseparateindicatorvariable(MISSINGCASHETR)andsetthatindicatorequaltooneforwherethissubstitutionhasbeenmadeandzerootherwise.WeincludethisindicatorvariableandtheindicatorinteractedwithETR,whichisthereplacedvalueofCASHETRdescribedabove(i.e.,theuseofcurrenttaxexpenseand/ortaxexpenseasthenumeratorbecausecashtaxespaidisnotavailable).CASHETR(andETR)isresettozeroforaminimum,and.5foramaximumtoeliminatetheeffectofoutliers.LOSSisanindicatorsettooneforrmswithnegativepre-taxbookincomeinthecalculationofCASHETR,andzerootherwise.RETAILisanindicatorvariablesetequaltooneforrmsintheretailindustryandzerootherwise.ADVERTISINGisacontinuousvariableofadvertisingexpensescaledbycurrentyearsales.HIGHADVERTISINGisanindicatorvariablesettooneforrmsthathaveagreaterthanmedian(forsamplermsthatdiscloseadvertisingexpense)ratioofadvertisingexpensetosalesandzerootherwise.GOVSCOREisthecontinuousgovernancescorefromGompersetal.(2003)asavailableonWRDS.ENTRENCHSCOREisacontinuousvariableofthesixprovisionsBebchuketal.(2005)determinearemostcloselyrelatedtomanagemententrenchmentandgovernanceproblems(thismeasureisnormalized).NON-ENTRENCHSCORErepresentstheremaining18provisionsoftheGompers'scorenotincludedintheENTRENCHSCORE(thismeasureisnormalized).MISSINGENTRENCHSCOREisanindicatorvariablesettoonermswheretheGOVSCOREand/ortheENTRENCHSCOREarenotavailable.Inaddition,whenGOVSCOREand/orENTRENCHSCOREaremissingwereplacetheirmissingvaluewithzero.Thesampleofallrm-shelters(the108observations)includesthenewsofshelterinvolvementinallnewssourcesavailableonFactivaandthesampleofthemajornewssourceonly(the72observations)includesonlynewsinTheWallStreetJournalTheWashingtonPost,orthemajornewswires(APOnline,DowJonesNewsWire,orReuters). Onepossibleexplanationisthathigh-ratermsarethosethatshelter.132M.Hanlon,J.Slemrod/JournalofPublicEconomics93(2009)126 computableforallrms,wereplacethismeasurewithasubstituteundertwoconditions.First,ifthismeasureisnotcomputablebecausethermhasanegativepre-taxbookincome,weincludeanindicatorvariablelabeledLOSSsettoonefortheserms,andzerootherwise.Fortheselossrms,wesettheCASHETRcontinuousvariabletozero.Second,ifthermdoesnotdisclosecashtaxespaidwereplacethenumeratoroftheCASHETRwithcurrenttaxexpense(data#16data#50)orifcurrenttaxexpenseisnotavailable,wereplacethenumeratorwithtotaltaxes(data#16)(whichthenmakesthisvariableatraditionalnancialaccountingETR).Inthesecaseswethenincludeaseparateindicator(MISSINGCASHETR)setequaltooneforrmswherecashtaxespaidarenotavailable,andzerootherwise.WetheninteractthisindicatorvariablewiththeCASHETRvariable(whichwerelabelintheinteractiontobeETRbecausethesermsdonothaveatrueCASHETR),tobesuretherelationwiththeCARisnotdifferentfortheETR(i.e.,thereplacedCASHETR)variable.Finally,inordertolimittheinuenceofmeasurementerrorduetooutliers,weresetanyvaluesofCASHETR(andETR)greaterthan0.5to0.5,andanyvalueslessthanzerotozero.ToinvestigatewhetherthereisadifferentialCARforrmsintheretailindustrywesetanindicatorvariable(RETAIL)equaltooneforrmsintheretailindustry(NAICScodes4445)andzerootherwise.Approximately16%ofthefullsampleand17%ofthesampleofmajorpressarticlesarermsintheretailindustry.Wealsoexaminetherms'advertisingexpense(data#45)asafractionofsales(data#12)asanotherproxyforbeingintheconsumerspotlight.Manyrmsdonotdiscloseadvertisingexpenseandwheremissing,weresetthemissingtozero.Wethencreateanindicatorvariablewhichissettoonewhentheratiooftherm'sadvertisingexpensetosalesisgreaterthanthesamplemedian(ofrmswithnonmissingadvertisingexpense)andzerootherwise.Finally,wemeasuregovernance(GOVSCORE)usingtheGompersetal.(2003)indexofshareholderrightsthatrangesfrom0to24,whereahighvalueindicateslowqualitygovernance.Inaddition,weparsethismeasureintotwoseparateindicators.TherstiswhatwecalltheENTRENCHSCORE.ThisindexwasdevisedbyBebchuketal.(2005)tofocusonthesubsetofprovisionsthataremostimportantforrmgovernancethoseindicatingmanagerentrenchment.Bebchuketal.(2005)documentevidenceconsistentwiththeprovisionstheyidentifyasbeingthemostrelated(negatively)tormvalueandreturns.Theentrenchmentprovisionsbelongtotwocategories:constitutionallimitationsonshareholders'votingpowerandtakeoverreadinessprovisions.cally,thefollowingsixcomponentsoftheGompersetal.(2003)indexareincludedintheENTRENCHSCORE:(1)staggeredboards(oneifstaggeredboardispresent),(2)limitstoamendby-laws,(3)and(4)supermajorityrequirementsformergersandcharteramendments,(5)poisonpillsand(6)goldenparachutes.Weincludeavariablemeasuringthepresenceoftheother18provisionsandrefertoitasNON-ENTRENCHSCORE.Theseprovisionsincludefairpriceprovisionsandbusinesscombinationstatutes,blankcheckpreferredstock,limitsonpowertocallaspecialmeetingandtoactbywrittenconsent,andliabilityandcationprovisions.AsinBebchuketal.(2005)boththeENTRENCHSCORE(06)andtheNON-ENTRENCHSCORE(018)countallprovisionsequally,givingonepointforeachoftheprovisionsthermhas.Weincludeeachmeasureasacontinuousvariable,normalizedtorangefromzerotoone.Table2revealsthattheaveragegovernancescoreforoursampleofrmsisaround9whichissimilartotheaveragegovernancescoreinDesaiandDharmapala(inpress)(averagegovernancescoreof9.265).Theyexaminealargersampleofrmsinwhichonlyaverysmallportionareaccusedofparticipatingintaxshelters.Thus,itseemsthat,onaverage,rmsaccusedofparticipatingintaxsheltersarenotmorepoorlygovernedthanotherrms.TheaverageENTRENCHSCOREforoursampleisaroundtwo(notnormalized),whichissimilartothatinBebchuketal.(2005)wherethemeanvariedfrom2.30to2.82.TherearesomermsforwhichwecannotobtainGOVSCOREand/oranENTRENCHSCORE.Inthiscasewesetthegovernancevariablesequaltozero;wecreateanindicatorvariable(MISSINGENTRENCHSCORE)whichissettooneforthesermsandzerootherwise.5.2.Overallandby-categoryaveragesTable3reportstheresultsofthetestsconductedtoaddressthehypothesesposedabove.ThetoprowofPanelArevealsthat,forthesampleofobservationswhichincludestherstshelternewsforeachrm-shelterfromallpresssources(=108),themeanCARis0.53%,butthisestimatefallsjustshortofbeingsignicantlydifferentfromzeroatstandardlevelsofcondence.If,however,wemeasuretheabnormalreturnsforthesampleincludingonlymajorpressmentions(andnotmentionsinsmallerpapersornewssources),theaverageCARis1.20%,whichissignicantlydifferentfromzeroforeachofthe-testsweutilize.Thus,thesimplestandmoststraightforwardevent-studymethodologysuggeststhat,onaverage,themarketreactsnegativelytonewsthatarmisinataxshelter.TheremainingpanelsofTable3reportontheCARbysubcategoriesofrms.PanelsB1andB2dividetheobservationsintothosewithaCASHETRatorabovethesamplemedianandthosewithobservationsbelowthesamplemedian.PanelB1presentstheresultsfortheentiresampleof108observations.ThermswithaCASHETRbelowthemedianhaveanaverageCARof1.32%,whichissignicantlydifferentfromzeroatconventionallevels.ThermswithaCASHETRatorabovethemedianvalueis0.06%,whichisnotsignicantlydifferentfromzero;notably,thedifferencebetweenthetworeturnsissignicant(-valueof0.0525,one-tailed).Thus,thereissomesupportforthehypothesisthatrmswithahigherCASHETRwillhavearelativelylessnegative(i.e.,morepositive)reactiontonewsabouttaxaggressiveness.InPanelB2wepresenttheresultsforthesubsampleof72observations Bebchuketal.(2005)forthereasoningforincludingthesixprovisionsintheENTRENCHSCOREandtheother18intheNON-ENTRENCHSCORE.ThegovernanceprovisionsareavailableonWRDSfromtheRiskMetrics(previouslyIIRC)database.TheENTRENCHSCOREisavailableonaWebpageprovidedintheBebchuketal.(2005)paper.NotethatweincludethegovernancescoresfortheyearnearesttotheyearofthenewsarticleandthatthereisnotnecessarilyanENTRENCHSCOREavailableforeveryrmthathasaGOVSCORE.Weconducttestsexcludingthesermsfromthesampleintheregressionanalysesaswell.TheARday-by-dayforthefullsampleisareturnof0.41%onday1,areturnof0.02%onday0,andareturnof0.09%onday+1.133M.Hanlon,J.Slemrod/JournalofPublicEconomics93(2009)126 wherethepressmentionisinamajorpresssource.Thereturnsare1.53%forthelowCASHETRgroupandstatisticallycantlynegative,and1.24%forthehighCASHETRgroupbutnotstatisticallysignicant;thedifferencebetweentheseisnotstatisticallysignicant(-valueof0.3730,one-tailed).PanelsC1andC2,showthesampledividedintoretailandnon-retailrms.Wendthattheabnormalreturnfortheretailaverages2.60%,comparedto0.14%forthenon-retailrms(thedifferenceissignicantlydifferentfromzero).Forthesubsampleofmajorpressmentions,theresultsareevenstarker:theabnormalreturnfortheretailrmsaverages4.13%,comparedto Table3CumulativeabnormalreturnsPanelA:Cumulativeabnormalreturnspresentedbysubsample#obsMeanCAR(%)Pos/NegttSample1:Allrm-shelters,rstpressmention1080.5347/611.4221.5781.5781.222Sample2:Allrm-shelters,majorpressmentiononly721.2029/432.8721.583Sample3:Firstrm-shelter,excludingrmswithpriorstatementmention0.6040/531.4741.7041.6971.174Sample4:Firstrm-shelter,excludingrmswithearningsannouncementsorSEClingsineventwindow0.9430/462.2711.689PanelB1:CumulativeabnormalreturnsfortherstpressmentionsubsampleabovemedianCASHETRandbelowmedianCASHETR#obsMeanCAR(%)Pos/NegttCASHETRabovemedian540.0628/260.0980.2200.2020.237CASHETRbelowmedian471.3216/312.718NoCASHETR(lossindenominatorLOSS 71.173/41.1550.6091.1580.068Totalsample108PanelB2:CumulativeabnormalreturnsfortherstpressmentionmajornewssourcesubsampleabovemedianCASHETRandbelowmedianCASHETR#obsMeanCAR(%)Pos/NegttCASHETRabovemedian331.2415/181.9471.3551.153CASHETRbelowmedian331.5312/212.9761.476NoCASHETR(lossindenominatorLOSSrms)0.852/40.7460.5630.978Totalsample72PanelC1:Cumulativeabnormalreturnsfortherstpressmentionsubsampleretailindustrycomparedtonon-retailindustries#obsMeanCAR(%)Pos/NegttRETAI2.607/101.9560.647NonRETAIL0.1440/510.7371.052Totalsample108PanelC2:Cumulativeabnormalreturnsfortherstpressmentionmajornewssourcesubsampleretailindustrycomparedtonon-retailindustries#obsMeanCAR(%)Pos/NegttRETAIL124.134/82.7291.136NonRETAIL0.6225/351.6931.4751.5271.227Totalsample72PanelD1:CumulativeabnormalreturnsfortherstpressmentionsubsamplebygovernancegroupwheregovernanceisdenedusingtheentrenchmentportionoftheGompersetal.(2003)scorefromBebchuketal.(2005)#obsMeanCAR(%)Pos/NegttWell-governedlowentrenchmentscore490.3023/260.477Poorlygovernedhighentrenchmentscore480.5419/291.1481.0781.395Nogovernanceorentrenchmentindex1.485/61.3831.6351.4570.191Totalsample108PanelD2:Cumulativeabnormalreturnsfortherstpressmentionsubsamplebygovernancegroupwheregovernanceisdenedusingtheportionoftheetal.(2003)scorethatisnotpartoftheentrenchmentscorefromBebchuketal.(2005)#obsMeanCAR(%)Pos/NegttWell-governedlownon-entrenchmentgovernancescore450.3623/220.5410.4690.4490.257Poorlygovernedhighnon-entrenchmentgovernancescore511.0619/321.9082.0771.799Nogovernanceorentrenchmentindex1.575/71.6041.7081.597Totalsample108NotestoTable3:CARsarethecumulativeabnormalreturnscalculatedovera3-daywindowcenteredonthedateofthenewsandcalculatedusingmarketadjustedreturns(i.e.,takingtherawreturnforthermlesstheCRSPvalue-weightedindexreturnforthesameday).WeclassifyrmsashavinganabovemedianCASHETRiftheiraverageCASHETR(data#317/data#170)overthetwoyearspriortothenewsisabovethesamplemedianofthismeasure.Ifthermhasnegativetotalpre-taxbookincome(data#170)overthetwoyearswecodethermasaLOSSrm.WeclassifyrmsasRETAILiftheyareintheretailsector(NAICScodes4445).Wermsaswell-governediftheirENTRENCHSCOREorNON-ENTRENCHSCOREisbelowthesamplemedianandpoorlygovernedifthosescoresareabovethesamplemedian.Thesampleofallrm-shelters(the108observations)includesthenewsofshelterinvolvementinallnewssourcesavailableonFactivaandthesampleofthemajornewssource(the72observations)includesonlynewsinTheWallStreetJournalTheWashingtonPost,orthemajornewswires(APOnline,DowJonesNewsWire,orReuters).Signicancelevelsareasfollows:indicatessignicanceat.01,at.05,andat.10,two-tailed.134M.Hanlon,J.Slemrod/JournalofPublicEconomics93(2009)126 forthenon-retailrms.Thus,thepointestimatesandstatisticallevelsareconsistentwiththehypothesisthatthepossibilityofanegativeconsumerreactiontoanindicationofcorporatecitizenshipmakesretailrmsrelativelymorevulnerabletonewsoftheirtaxaggressiveness.However,werecognizethattheinterpretationoftheseresultsissubjecttotheconcernthatengagementintaxsheltersisendogenous.Thus,forexample,itispossiblethatretailrmsarelesslikelytopursuetaxsheltersbut,iftheydoso,theexpectedbenetwouldbehigherthanotherwiseinordertooffsetthehigherexpectedcost.Itisalsopossiblethatthetypeofsheltersthatretailrmsengageinandthatbecomenewsworthyissystematicallydifferentthanthetypeofsheltersthatmakethenewsinothersectors.Theseresultsshouldbeinterpretedwiththesecaveatsinmind.PanelsD1andD2concerncorporategovernance.UsingtheENTRENCHSCORE(PanelD1),themeanCARsforthewell-governedandpoorlygovernedrms(splitatsamplemediansforthistable)are0.30%and0.54%,respectively.Neitherreturnisstatisticallycantfromzero(andthedifferencebetweenthereturnsisinsignicant).WhenweusetheNON-ENTRENCHSCORE(PanelD2),themeanCARsforthewell-governedandpoorlygovernedrms(againsplitatsamplemedians)are0.36%and1.06%,respectively.Theabnormalreturnsfortheobservationsofrelativelypoorlygovernedcompaniesareatleastmarginallysigniforeachofthethreeteststatisticsandalsousingthenon-parametricsigntest;thereturnsforthewell-governedrmsarenotcantusinganytest(thedifferencebetweenthetworeturnsissignicantat0.075,two-tailed).Theseresultsarebroadlyconsistentwiththehypothesisthatrelativelypoorlygovernedrmswillhaveamorenegativeresponsetothenewsabouttaxaggressiveness;however,thisonlyholdsusingtheNON-ENTRENCHSCORE.5.3.Cross-sectionalanalysisofexcessreturnsTofurtherexamineourhypotheses,andinparticulartoinvestigatethepartialrelationshipbetweenrmcharacteristicsandthestockpricereactiontotaxshelternews,wenextinvestigatethecross-sectionalrelationbetweenrmcharacteristicsandtheeventwindowreturns.TheregressionresultsarepresentedinTable4,fortheall-rm-sheltersampleinPanelAandforthesamplelimitedtoonlymajorpresssourcesinPanelB.Wepresentseveralspecications,butfocusourattentionrstonColumn(1)ofPanelA.TheestimatedcoefcientonCASHETRis,asourreasoningsuggests,positiveandsignicant.Theestimateof0.055suggeststhataonestandarddeviation(equalto0.137)lowervalueofCASHETRisassociatedwitha0.75%lowervalueoftheCARsurroundingataxshelternewsrelease.WehypothesizedthatrmsintheretailsectormightbemorenegativelyimpactedbynewsofataxshelterinvolvementbecauseofthepossibilityofaconsumerreactiontoperceptionsofnotbeingagoodcorporateThishypothesisisalsosupportedbythedata.Theestimatedcoefcientontheindicatorvariableforbeingintheretailsectoris0.028andcantlydifferentfromzero,suggestingthatthemarketreactionis2.8%lowerforaretailrmcomparedtonon-retailWenextturntoanalyzingthecorporategovernancevariables.Thepredictionisthatwell-governedrmswillhavealessnegativereaction,consistentwithDesaiandDharmapala(inpress).Becausethegovernancescoresareconstructedsuchthatahighernumbermeansmorepoorlygoverned,onewouldexpectanegativerelationifourhypothesisiscorrect.ThedatainTable4suggestthatthereisinfactasignicantlynegativerelationbetweentheNON-ENTRENCHSCOREandtheabnormalreturns;but,surprisingly,thereisasignipositiverelationbetweentheENTRENCHSCOREandtheabnormalreturns.Botharehighlycant.IftheENTRENCHSCOREisthescoremorerepresentativeofgovernancethenonewouldexpectthisvariabletobecantlynegativeandtheNON-ENTRENCHSCOREtobeinsignicant.Wecanonlyofferanex-postrationalizationontheseresults.Bebchuketal.(2005)documentthatincreasesintheENTRENCHSCORE,butnottheNON-ENTRENCHSCORE,aremonotonicallyassociatedwitheconomicallysignicantreductionsinrmvaluationThus,themarketseemstohavealreadyimpoundedtheformer,butnotthelatter,setofindicatorsofpoorgovernanceintoshareprices.ItisplausiblethatwhenthemarketlearnsoftaxshelteractivityforthermswithahighNON-ENTRENCHSCOREitreactsnegativelybecausethenewscon Wealsoinvestigateinstitutionalholdingsasanalternativemeasureofgovernancebasedontheideathatinstitutionalinvestorsmonitormanageriperformancemorethanindividualowners,thusprovidingimprovedgovernanceattherm(similartoDesaiandDharmapala(inpress)).WemeasureinstitutionalholdingsusingdatafromThomsonFinancial(Spectrum)bycomputingthefractionoftotalsharesoutstandingthatareheldbyinstitutions(banks,insurancecompanies,mutualfundparentcompanies,pensions,endowments,andprofessionalinvestmentadvisors).Inthereturnsanalysis,wendthatwithgreaterthanthemedian(inoursample)institutionalholdingshaveanabnormalreturnovertheeventwindowof0.96%(1.39%inthemajorpressmentionsample),andrmswithlessthanthemedianinstitutionalholdingshaveareturnof0.07%(0.80%inthemajorpressmentionsample).Forthefullsampleandthemajorpresssample,thedifferencebetweenthetwogroupsisinsignicantatconventionallevels.Thus,itseemsthatrmswithhighinstitutionalholdingsdonothaveasignicantlydifferentmarketreactionuponrevelationofthenewsoftaxsheltering,whichisapparentlyinconsistentwiththemonitoringstory.WenotethatthepointestimateofthecoefcientonCASHETRishigherforrmswherewereplacedthenumeratorofCASHETRwithsomethingotherthancashtaxespaid(i.e.,whenwereplacedthecashtaxespaidwithataxexpenseamountfromtheincomestatement).However,thisdifferenceisnotstatisBecausethereisanecdotalevidencethatrmsengagedintheCorporateOwnedLifeInsurance(COLI)typeofshelteraremorelikelytoberetailrms,itispossiblethatthecoefcientontheRETAILvariableisnegativebecausetheeventsrefertoCOLIshelterrms,notbecausetheyareintheretailindustry.ToinvestigatethispossibilityweincludeanindicatorvariablesetequaltooneforrmsintheCOLItaxshelterinadditiontoalltheothervariablesinthecolumn(1)cation.ThecoefcientontheRETAILvariableremainsnegativeandsignicantlydifferentfromzeroforboththefullsampleandthemajorpresssubsample.ThecoefcientontheCOLIvariableispositive,butinsigniWealsotriedsubstitutingameasureofbrandvalueforourRETAILvariableinordertocaptureabroaderproxyofthevulnerabilityofcompaniestopublperceptionsofcorporatecitizenship.Wesetanindicatorvariableequaltooneifthermwaslistedashavingoneofthetop100brandnamesasrankedinBusinessWeekmagazineinanyoftheyearsfrom2001to2005(theonlyyearsforwhichwecouldobtainthedata).BusinessWeekobtainstherankingsfromInterbrand,whichranksbrandsbasedupontheestimatedamountthebrandislikelytoearntherminthefuture.Outofour108rmsinthetaxsheltersample,only10areontheBusinessWeekbrandlist;outofour72-observationsampleofmajorpressmentions,only7areincludedontheBusinessWeeklistinanyoftheyears2001to2005.WhenweincludethisindicatorvariableintheplaceoftheRETAILindicator(intheregressionmodelinTable4column(1))thecoefcientis0.004inthefullsampleand0.009inthesampleincludingonlymajorpressmentions,however,neithercoefcientissignicantlydifferentfromzero.135M.Hanlon,J.Slemrod/JournalofPublicEconomics93(2009)126 suspicionsofpoorgovernancethatpreviouslywerenotthoughttobevalue-decreasing.Ontheotherhand,themarketpriceofrmsforwhichtheENTRENCHSCOREishighalreadyreectsareductioninvaluerelatedtothepoorgovernanceprovisions,andthusinvestorsdonotreactnegatively.Tobesure,theestimatedpositivecoefcientfortheENTRENCHSCOREvariableremainsapuzzle,andsuggeststhatthesegovernancemeasuresmaybecorrelatedwithotherunmeasuredcharacteristicsofthermthataffectthemarketvaluationsofnewsoftaxshelterinvolvement.Wediscussthispossibilityinthenextsection. Table4RegressionanalysisPanelA:All12345Constant0.0260.0150.014(0.015)(0.015)(0.016)(0.014)(0.017)CASHETR0.0550.0530.100(0.031)(0.032)(0.036)(0.028)(0.037)MISSINGCASHETR0.1050.049(0.069)(0.069)(0.070)(0.069)(0.063)MISSINGCASHETRETR0.2810.2840.3250.2480.159(0.221)(0.225)(0.224)(0.222)(0.193)LOSS0.0390.0410.043(0.016)(0.016)(0.019)(0.018)(0.018)RETAIL0.019(0.013)(0.013)(0.015)(0.014)(0.014)HIGHADVERTISING0.001(0.010)ENTRENCHSCORE0.0480.0490.0480.031(0.021)(0.022)(0.022)(0.023)(0.022)NON-ENTRENCHSCORE0.1500.1510.1500.1610.119(0.402)(0.040)(0.041)(0.044)(0.042)MISSINGENTRENCHSCORE0.0480.0480.040(0.017)(0.017)(0.018)(0.018)Observations108108969376-squared0.2910.2910.3340.3690.211PanelB:Onlymajornewssources12345Constant0.0420.0420.048(0.016)(0.016)(0.016)(0.016)(0.017)CASHETR0.0470.0480.040.098(0.037)(0.038)(0.038)(0.032)(0.042)0.0180.0150.0370.010.043(0.038)(0.035)(0.034)(0.038)(0.062)MISSINGCASHETR0.0710.0550.1230.0290.159(0.113)(0.108)(0.102)(0.111)(0.199)MISSINGCASHETRETR0.0270.0250.0360.027(0.014)(0.015)(0.016)(0.016)(0.017)RETAIL0.0430.0430.046(0.014)(0.014)(0.014)(0.014)(0.015)HIGHADVERTISING0.006(0.012)ENTRENCHSCORE0.0490.046(0.021)(0.022)(0.021)(0.021)(0.026)NON-ENTRENCHSCORE0.1890.1850.1950.1920.169(0.046)(0.047)(0.045)(0.047)(0.056)MISSINGENTRENCHSCORE0.0660.048(0.021)(0.021)(0.022)(0.023)Observations7272636648-squared0.3270.3320.4360.3890.337NotestoTable4:Standarderrorsareinparentheses.ThesampleandthevariablesareasdenedinTables2and3.Thesampleincolumn(1)isthefullsampleofallrstpressmentionsoftherm-shelter.Thesampleincolumn(2)isthesameascolumn(1),thedifferenceistheinclusionoftheadvertisingindicatorvariableasaregressor.Thesampleincolumn(3)excludesobservationsforwhichthereisnogovernancescore.Thesampleincolumn(4)excludesobservationswheretheshelterwasmentionedinthenancialstatementspriortothepressarticle.Thesampleincolumn(5)excludesobservationswheretherewassometypeofnancialaccountingnewsduringtheeventwindow(e.g.,earningsannouncement,thelingofa10-K,10-Q,or8-K).Signicancelevelsareasfollows:canceat.01,at.05,andat.10,two-tailed.Dependentvariable:cumulativeabnormalreturns. Whenweincludethemeasureofinstitutionalholdingsintheregression(analogoustocolumn(1)inTable4)asaproxyforgovernance(anindicatorvariablesetequaltooneforbelowthemedianinstitutionalholdings),thecoefcientonthatvariableispositive(0.01)inbothsamplesrmswithlessinstitutionalholdingshaveahigherreturnbutisinsignicantatconventionallevels.136M.Hanlon,J.Slemrod/JournalofPublicEconomics93(2009)126 TheremainingcolumnsinPanelAyieldverysimilarresults.Incolumn(2)weincludetheindicatorvariableHIGHADVERTISINGasanotherproxyforthepotentialforconsumerbacklash.Wendthatthecoefcientonthisvariableisessentiallyzero.Becausesorms(only40outofthe108)discloseadvertisingexpense,however,itisdifculttoinfermuchfromthisresult.Incolumn(3)weexcludermsforwhichwecouldnotobtainagovernancescorejusttobesurethattheavailabilityofthegovernancevariableisnotcorrelatedwithsomeotherfactorthatisalteringtheresults.Thisdoesnotseemtobethecase,however,astheresultsareessentiallyunchangedafterthedeletionoftheseIncolumn(4)weexcludeobservationswherewendanyevidenceofpriornancialstatementdisclosureoftheparticularsheltermentionedinthearticle.Priordiscussioninthenancialstatements(priortwoyearsof10-KsledwiththeSEC)oftheshelterwouldpre-emptthepressarticleandlikelyreducethemarketreaction.Thedatasuggestthattheinferencesareunchanged(indeed,someresultsarestronger)forthissubsampleofFinally,incolumn(5)weexcludeobservationsforwhichthereissomeotherlingnewsduringtheeventwindow.Toobtainthissubsample,wesearchtheSECWebpageandcollectthedatesofthelingsofall8-Ks,10-Qs,and10-Ksfortherms.Wealsocollecttheearningsreleasedatesforalltherms.Ifanyoftheseeventsoccurredduringoureventwindowweexcludetheobservationincolumn(5).TheresultsareunchangedexceptthatENTRENCHSCOREisnolongersignicantlypositive.InPanelBofTable4,weconductthesameanalysisusingonlythenewseventsreportedinmajorpresssources.TheresultsareverysimilartothoseinPanelA.ThemainexceptionbeingthatthesignicanceleveloftheCASHETRvariableisdiminished.However,thepointestimatesarenotthatmuchlower(0.055inPanelAversus0.047forPanelB(Column(1)))andinColumn(4)ofPanelBthecoefcientremainsverysignicant.Thus,ingeneralthemajorpresssourceresultsareconsistentwithouroverallsampleresults.Allinall,themultivariateregressionanalysesarebroadlyconsistentwiththehypothesesthispaperproposes.Newsabouttaxshelterinvolvementonaverageaffectsarm'sstockpricenegatively.Theimpact,though,variesdependingonthetypeofbeingworseforrmsintheretailsector,andwithalowcasheffectivetaxrate.Wendsomewhatmixedandsurprisingevidencewithregardtogovernance,suggestingthathowthequalityofcorporategovernanceaffectsthestockpricereactiontothiskindofnewsismorecomplicatedthanhadheretoforebeenrealized.Itisworthnotingthatthepatternofestimatedcoefcientsimpliesthat,forcertainkindsofrms,weexpectthatthereleaseofnewsofinvolvementinataxshelterwillcausethestockpricetoincrease.Thiswouldmorelikelybethecasefornon-retailcompanieswithhigheffectivetaxratesthatarewell-governedbyonemeasure(NON-ENTRENCH).Thisconclusionisconsistentwiththefactthat44%oftheabnormalreturnobservations(47outof108cases)arepositive.5.4.CaveatsandlimitationsThereareseveraladditionalcaveatsandlimitationstoourtestsandresults.Therstisthatwecannotreliablydecipherwhen,ifever,armrecordedthetaxsavingsfromshelterinvolvementinitsnancialstatements.Whenrmstakeanaggressivepositiononataxreturn,theymayrecordthetaxbenet(taxsavings)inthenancialstatementsinawaythatwouldincreaseaccountingearningsaftertax;alternatively,thermmay,iftheybelievethepositionmaynotbesustained,simultaneouslyrecordnancialaccountingreserveandthusnotreectthetaxsavingsinthecurrentperiod'searnings.Becausewecannotidentifyrmsrecordedthetaxsavings,anyobservednegativemarketreactioncouldreectthatthermbookedthetaxsavingspreviouslyandnowowesthetaxes,requiringthermtorevealthatitmadelessaftertaxthanitreportedtoshareholders.Conceivablythepricewilladjustdownwardtoreecttheloweramountofearnings.Althoughweusepressarticlesaboutthesheltersthatoftendonotmentionanyeffectonearnings,wecannotbecertainthatthepotentialearningsimpact(ratherthan,orinadditionto,theinformationthatthermistaxaggressive)isnotaffectingourresults.However,wethinkitisunlikelythatthemarkethasinformationonwhetherthermrecordedthetaxsavingsandthus,anymarketreactionduetothiswouldbebasedonitsperceptionofwhetherthermbookedthetaxsavings.Boththeactualandcertainlytheperceivedrecordingofthetaxsavingsarenotmeasurable,however,whichisapotentiallimitationofthestudy.Second,andrelatedtotheabove,rmgovernanceandhowthermsaccountforthetaxsavingscouldbecorrelated.Forexample,itcouldbethatrmswithrelativelypoorgovernancebookthetaxsavings(increasingafter-taxearnings)intheyeartheshelterisutilizedandrmswithrelativelygoodgovernancedonotbookthetaxsavings(byrecordingacontingencyreserve)untilthetaxyearhaspassedthroughIRSaudit.Whilewehavenoevidencethatthisisthecaseandcannottestitreliably(because,asmentionedabove,wedonothavedataoniforwhenthetaxsavingswererecordedforallthermsinoursample),itisacorrelationthatcouldpotentiallyaffecttheinterpretationofouranalysesofhowgovernanceaffectsthestockpricereactiontotaxshelternews.Inthepresenceofsuchacorrelation,themorenegativereactionobservedforthemorepoorlygovernedrms(NON- Asasensitivitytest,weconducttheentireanalysisforthetaxsheltersampleusingthemarketmodeltocomputecumulativeabnormalreturnsrathertthemarketadjustedmodel.Theinferencesoftheanalysisareverysimilarnomatterwhichmethodofcomputingabnormalreturnsisused.Thus,ourresults(orlackthereof)donotappeartobeduetothemethodofcomputingtheabnormalreturns.Thismechanismisdistinctfromtheimpactonafter-taxcashows.SeeShackelfordetal.(2008)Duringoursampleperiod,whenrmstookanaggressivetaxpositionthattheythoughtmightbedisallowedinthefuturethermwassupposedtorecordtherelated(saved)taxexpensewhendeterminingitsnancialaccountingearningsinordertopreventoverstatingtheaccountingearnings.Thereserveforthesavedtaxeswascalledthetaxcontingencyreserve(andwasalsoknownasthetaxcushionpriortorecentregulatoryscrutiny).SeeGleasonandMills(2002)foradiscussionofhowlittlewasdisclosedaboutthetaxcontingencyreserve.WethankJoeBankmanforhismentionofthisissue,whichhereferstoasthegivebackhypothesis.137M.Hanlon,J.Slemrod/JournalofPublicEconomics93(2009)126 ENTRENCHSCORE)inourresultscould,inpart,beattributabletothermhavingineffectoverstatedpriorearningsduetoclaimedtaxsavingsthatthermcannotrealize.Third,werecognizethatourresultsarenotnecessarilygeneralizable.Forexample,themarket'sperceptionoftaxaggressivenessmaychangeovertime,sosheltersinthefuturemayreceiveamoreorlessseverereactionthanwedocumentinthisstudy.ThechangingregulatorylandscapeinrecentyearsincludingtheenactmentandimplementationoftheSarbanesOxleyAct,theissuanceoftheFinancialAccountingStandardsBoardInterpretationNumber48(whichprovidesguidanceontheaccountingfortaxcontingencies),thenewM-3disclosurerules,andtherequireddisclosuresoflistedtransactionswilllikelychangebothwhetherrmsengageintaxsheltersandhowthemarketreactstonewsthatrmshaveengagedinthem.Finally,itisworthreiteratingthatifthemarketlearnsabouttheshelterinadvanceofthenews,orifthemarketcanestimatermsaretaxaggressiveorengagingintaxshelterspriortothereleaseofthenews,thenweshouldnotexpectareaction(otherthantonewsaboutpenaltiesandlossoftheuseoftheshelter)onthedaytheshelterinvolvementornewsoftaxaggressivenessisreleasedinthepress.Werecognizethat,asisthecasewithanyshort-windoweventstudy,wecannotbecertainweareactuallycapturingtheeventthatcontainsthenews.Wetestthesubsampleofrmsforwhichthermhadnotdiscussedtheshelterinpriornancialstatements,andindeedndstrongercross-sectionalresults.Ofcourse,wecannotbeentirelysurethatthenewshasspreadbyothermeansthatwehavenotdetected.Forexample,itispossiblethatinsomecasesthemarketknowsthermisinvolvedinthetaxshelterorperhapsisevenincourtbeforethestoryispickedupbythepress.Therealsocouldbetradingbyinsidersbeforetheinformationbecomespublicinanymeaningfulsense.Wetestthemarketreactiontotherstannouncementinthepress;but,iftheinformationabouttheshelterwasalreadyinthemarket,thenwehaveunderstated(ormoregenerallymisstated)theeffect.6.MarketresponsetoadifferentkindofnewsabouttaxaggressivenessOnechallengeofstudyingnewsaboutcompanies'involvementintaxsheltersisthatitisdifculttodisentanglethereputationeffectofthermbeingtaxaggressivefromthemarketeffectofthepotentialfuturecostsoflosingtheshelterandincurringlegalcosts.Tohelpdisentanglethesedistinctmechanismsofresponse,wealsoanalyzethemarketreactiontoadifferentkindoftaxnews:thereleaseofstudiesbyCitizensforTaxJusticein1984,2000,and2004thatfeaturedcalculationsofthecurrenteffectivetaxratesofover200rms.Eachofthesestudieshighlightsthefactthatmanylarge,protable,U.S.companiespayverylittletax.Certainlytheheadlinesinnewsstoriesafterthestudieswerereleasedstressedthisangle:128BigFirmsPaidNoFederalIncomeTaxes,LosAngelesHeraldExaminer,October6,1984)and41BigFirmsPayNoTax,StudySays,ReutersNewsService,October21,2000).Indeed,the2004CTJreportbeginsasfollows:FollowingthetragiceventsofSeptember11,2001,Americansjoinedtogetheringriefandsolidaritytosupporteachotherandourcountry.CorporateAmericaalsoralliedtoacause,butitwasn'tinsupportofournation.Onthecontrary,itwastaxavoidance(p.1).Ifevercompaniesareaccusedofbeingpoorcorporatecitizenswithregardtotheirtaxpayingbehavior,itisinthesestudies.Thesereportsdonotlinkanyofthermstoillegalactivity,though,andthusshouldnothavetheconfoundingeffectsofexpectedlegalcostsorexpectedlossfromnotbeingabletousetheirtaxavoidancemethodsinthefuture.Theresearchdesigntrade-offisthatitisalsopossiblethatbeinglabeledaspayingrelativelylowtaxesbytheCTJdoesnotinvokethesamenotionofinvolvementinataxshelter,andthusthereputationeffectisarguablylower.EachCTJstudycomputesrmcurrenteffectivetaxrates(theCTJusescurrenttaxexpenseinthenumerator)andthengroupsrmsbasedonthesecomputedcurrenteffectivetaxratesintocategoriessuchasrmsthatpaidnotaxes.Weperformsimilareventstudytestsasthosedescribedaboveforoursheltersample,centeredonthedayofthepresscoverage,whichisthedayafterthestudyisreleased.ThesampleoverwhichweestimatetheregressionisthegroupingofrmstheCTJclaimstohavepaidnotaxinayear,orhadalowertaxratethantheaverageAmericanfamily.Weconsolidatethermsfromallofthethreestudiesintooursample.Oursampleofrmsis193observations.Inunreportedresults,theaverageCARis0.10%(andnotsignicantlydifferentfromzero).Notably,theaverageCASHETRoftheCTJlow-taxrmsisquiteabitlowerthanfortheshelterrms.ThismakessensebecausetheCTJhasdeterminedthattheyhavepaidverylittleintaxesoverthetimeperiodsofthestudies.WhenweestimatearegressionsimilartothatinTable4forthisCTJ AconferencediscussantalertedustoColgatestaxsheltercase.Usingasearchsimilartoours,therstarticlethediscussantcouldndwasaMay,1996WallStreetJournalarticleabouttestimonybroughtupinthecourtcase.Thecasewaswellknowninmosttaxcircleswellbeforethearticle,however,becauseitwasalreadyincourt.WecheckedoursampleandnotedthatoursearchpickedupanarticledatedAugust17,1990,alsofromtheWallStreetJournal.Thus,weincludeanarticlepriortothecasegoingtocourtandpriortoatleastthatinformationbeingsopubliclyavailablepriortooureventdate.Nevertheless,thegeneralpointiscertainlyvalid.Webelievethepressarticlemethodologyweemploymayoftenresultinourrecordinganearlierdatethanwouldamethodologybasedonlings(becausemanyofoureventsarepriortothecasegoingtocourt)ordisclosuresinnancialstatements.However,itisstillpossiblethatthenewsisinthemarketbeforeoureventdate,weakeningthepowerofourtests.Ourresultscouldbeunderstatedifrmswhoseauditorwaspreviouslyaccusedofsellingaggressivesheltersexperiencedthemajorityofthemarketreactionuponthereleaseofthenewsabouttheauditor.WeinvestigatewhetherthereisanydifferenceinthemarketreactiontoourpressnewsforrmsauditedbyKPMG,thermwiththemostvisiblelegalissuesovertaxshelters,ascomparedtootherauditorstoseeifthereisevidencethatthisisthecase.WendthatthemarketreactionfortheKPMGrmsisnodifferentthanforthenon-KPMGrms.Thus,itdoesnotappearthatthestockpricereactionofrmsauditedbyKPMGiscompletelynegatedbyanypriornewsabouttheirauditor.MoredetailsoftheCTJanalysiscanbeobtainedfromtheauthors.138M.Hanlon,J.Slemrod/JournalofPublicEconomics93(2009)126 sample(untabulated),thedatarevealnosignicantcross-sectionalvariationforthevariablesCASHETR,RETAIL,oranyofthegovernancemetrics.Ascautionedabove,oneexplanationisthatbeingcategorizedasalow-taxpayerbyCitizensforTaxJusticearguablydoesnotraisethesamekindofsuspicionsregardingubiquitousdoubledealingasinvolvementinataxsheltercasemight,andthusthemarketresponsetolossofreputationorfearofshareholderexpropriationislikelytobemuted.AnotherexplanationisthatbecausetheCTJbasetheircomputationsatleastinpartonrmdisclosuresthatarealreadypubliclyavailable,thereleaseofthestudyisnotreallynewsabouttherms'taxbehaviors.Overall,themarketreactiontotheCTJstudiesisnotablymoremutedthanitsreactiontorminvolvementintaxshelters.7.ConclusionsInordertomaximizethevalueoftherm,shareholderswouldliketominimizecorporatetaxpaymentsnetoftheprivatecostsofdoingso;inotherwordstheywantthecompanytobeoptimallyaggressive.Therehasbeenlittlerigorousempiricalanalysisofthebenetsandcoststocorporationsofbeingtaxaggressive.Inthispaper,weattempttollthisvoid,atleastinpart,byinvestigatingthemarketreactiontoaninitialpressmentionthatarmwasinvolvedinacorporatetaxshelter.ndthat,onaverage,acompany'sstockpricedeclineswhenthereisnewsaboutitsinvolvementintaxshelters.Wealsouncoversomeevidenceofcross-sectionalvariationinthereturns.Forexample,thestockpricedeclineismorenegativeforrmsintheretailsector,suggestingthatpartofthereactionmaybeaconsumer/taxpayerbacklash.Thereactionseemstobelessnegativermswithahighercasheffectivetaxrate,consistentwiththemarketinterpretingthenewsasapositivesignaloftaxaggressiveness.Intermsofgovernance,wendthatprovisionsnotrelatedtomanagemententrenchmentarenegativelyrelatedtothemarketreactionAppendixAOursampleconsistsofdifferenttypesoftaxshelters.Webrieydescribeeachtypeofshelterbelow.rstshelterinvolvestheuseofcorporateownedlifeinsurance(COLI).Thisshelterinvolvesthepurchaseoflifeinsurancepoliciesonmanyemployeesofthecompanynotjustkey-manpoliciesbutpoliciesonevenverylowlevelemployees(e.g.,checkersingrocerystoresandWal-Mart).Thermthenimmediatelyborrowsagainstthesepolicies,deductingtheinterestontheseloans(whichareusuallyloansfromthelifeinsurancecompany).Followingthis,thermretainstheproceedsintheeventoftheemployee'sdeathevenifthatpersonnolongerworksforthecompany.Manytimestheemployeedidnotknowtheinsurancepolicyexisted.Inoursampleof108newsevents,32pertaintoCOLI.Thesecondtypeoftaxshelterinoursampleisonewherethermtransfersitsintangibleassets(e.g.,abrandname)toacorporationorganizedinastatethatdoesnothaveanincometax.Therm(morespecicallyitsvariousentitiesfortaxpurposes)paystheothercorporationfortheuseoftheintangible.Thermthendeductsthesepaymentsonitsreturnsintaxablestates,reducingincomesubjecttostatetax.Theotherentityreportstheincomebut,becauseitdoesnothaveNexusinastatewithanincometax,thereisnostateincometaxdueonthisincome,thusprovidinganincometaxshelterfromstatetaxes.ToysR"UsisthemostwidelyknownexampleofthistypeofsheltertheytransferredGeoffreytheGiraffeastheintangibleasset.Inoursampleof108observations,thereare16observationsassociatedwiththistypeofshelter.ThethirdtypeofshelterinoursampleisknownastheStep-DownPreferredStocktaxshelter(orFast-PayPreferred).Underthisshelter,preferredstockissuesweresoldtotax-exemptinvestorssuchaspensionfunds.Thisisacomplexvehiclethateffectivelyallowsacompanytoissueequityanddeductinterestandprincipalpayments.Thecompanyengaginginthesheltercreatesarealestateinvestmenttrust(REIT)inpartnershipwithapensionfund(whichistax-exempt).Thecompanyprovidestherealestateandreceivescommonstock;thepensionfundprovidescashandreceivespreferredstock.TheREITthenlendsthecompanyfundssecuredbyitsrealestateholdings.Whenthecompanyrepaystheloan,thecashgoestopaythedividendsandsomeoftheprincipalonthepensionfund'spreferredstock.Boththedividendandprincipalpaymentsaredeductibleasinterest.Inoursampleof108observations,thereare12observationsassociatedwiththistypeofshelter.ThefourthtypeofshelterinoursampleisknownastheLease-InLease-OuttaxshelterortheSale-InLease-Outtaxshelter(LILOorSILO).Althoughthereareseveralmethodsbywhichthiscanbearranged,theconceptbehindthisshelterisforatax-exemptmunicipalitytodepreciationtaxdeductionstoaU.S.corporation.Generally,aU.S.corporationwillleaseorbuyarealassetsuchasapowerplant,astadium,asubwaysystem,etc.fromatax-exemptmunicipality(U.S.orforeign)andthensubleasethepropertybacktothesametax-exemptparty.Thecorporationwillgenerallyprepayitsleaseobligation(ifaLILO).TheU.S.companyengagingintheshelterthenobtainstheaccelerateddepreciationdeductionsonitstaxreturn(butexpensesthedepreciationmoreslowlyandoveralongertimeperiodfornancialaccountingpurposesoronlyrecordsrentalexpensefornancialaccountingpurposes).Inoursampleof108observations,9observationsareassociatedwiththistypeofshelter. Wenote,though,thatthereisahigherpercentageofrmswheretheCASHETRismissinginthissample,mostlikelybecausethesampleyearsareearlierrmswerenotrequiredtodisclosecashowstatementsduringtherstofthethreestudies(priorto1988).Asbefore,wereplacecashtaxespaidinthemeasureoftheETRwithcurrenttaxexpenseortotaltaxexpenseandincludeintheregressionaninteractiontermofMISSINGCASHETRandETR(thereplacvaluesforCASHETR).Thisisdoneinordertodetermineifthereplacedvariablehasadifferentrelationthanthevariablecomputedwithcashtaxespaidinthenumerator.139M.Hanlon,J.Slemrod/JournalofPublicEconomics93(2009)126 fthtypeofshelterinoursampleisknownastheContestedLiabilityAccelerationStrategytaxshelter(alsoknownasCLAS).ThisshelterwasdevisedandmarketedbyKPMG,andtheIRShassaidthatitgeneratedatleast$1.7billionintaxsavingsformorethantwodozencompanies.Theideabehindthisshelterwastoacceleratethetimingoftaxsettlementsoflawsuitsandotherclaims.Deductionsgenerallyarenotallowedforsuchliabilitiesuntiltheclaimantsarepaid,butoneexceptionunderthetaxcodeinvolvesthetransferofmoneyand/orotherpropertytoacontestedliabilitytrustbeforetheclaimsareresolved.Thus,acompanyengaginginthisshelterstrategywouldestablishatrustwithitselfasthebeneciaryandthentransfernoncashassetstothetrust.Thevalueoftheseitemstransferred(generallystockoranintercompanynote)issupposedtorepresenttheamountthecompanyisexpectedtopaytoresolvetheclaimsitisstillcontesting.Bymovingthesenoncashitemstothetrust,thecompanycantakethedeductionsearlierthanotherwiseallowed,reducingtaxableincome.Inoursampleof108observations,8observationsareassociatedwiththistypeofshelter.AnothertypeofshelterinoursampleisknownastheContingentPaymentInstallmentSaletaxshelter.Generally,thegoalofthisshelteristoproducepapercapitallossesthatcanbeusedagainstrealcapitalgainstheU.S.corporationhasgeneratedfromsomeothersource.Therearealternativemethodstoemploythisstrategy,butatypicaldealmayworkasfollows.Threepartiesformapartnershiptoacquireaxed-incomesecuritythatdoesnottradepublicly.TheU.S.corporationseekingtoengageinthetaxshelterwillputupapproximately9%ofthefunds,atax-exempt(usuallyaforeign)entitywillfund90%,andanotherthirdpartynancierwillputupthenal1%.Thepartnershipwillthensellthesecurityinaninstallmentsale,receivingmostoftheproceedsimmediatelyandtheremainder(contingentonthesecurity'svalueatthattime)inthethirdyear.Thepartnershipwillshowalarge(paper)gainintherstyear,mostofwhichwillbeallocabletothetax-exemptparty.Then,inthesecondyearofthepartnership'slife,thetax-exemptpartywillexitthepartnershipbeingboughtoutbytheU.S.corporation.Thepartnershipwillthenshow(paper)lossesinthesecondandthirdyears,mostofwhichwillbeallocabletotheU.S.taxshelteringcorporation,thusoffsettingthecapitalgainthecompanygeneratedfromothersources.Inoursampleof108observations,8observationsareassociatedwiththistypeofshelter.TheseventhtypeoftaxshelterinoursampleisknownastheRegulatedInvestmentTrusttaxshelter.Thisstrategywasestablishedtotakeadvantageofafederallawthatgovernsmutualfunds.TheU.S.company(usuallyabank)registersasubsidiarywiththeSECasaregulatedinvestmentcompanyundertheInvestmentCompanyActof1940.Thebankthentransferssomeoftheirloanportfoliosandotherassetsintothefunds,andusestheinterestandotherincometheygeneratetopaythemselvesdividends.MostofthetaxbenetsarestatetaxbenetsinstatessuchasCalifornia,whichexemptsmoneytransferredbetweensubsidiariesandcorporateparents.Wenote,though,thattheWallStreetJournalreportedthat,onebankfund,NBTInvestmentCo.,saidinanSEClingthatitwasoptimisticthatitsstructurewillbesufcienttorelieveitfromallorsubstantiallyallfederalandstateincometaxes'Simpson,2003,italicsadded).Inoursampleof108observations,6observationsarerelatedtothistypeoftaxshelter.AnothertypeoftaxshelteristheOffshoreIntellectualPropertyHaventaxshelter.Underthisstrategy,rmstransferpatentsonintellectualpropertytoasubsidiaryestablishedinataxhaven(generallysomewherelikeBermudaoranotherlow-taxjurisdiction).ThenroyaltiesfromsalesoftheproductsmadeoutsidetheUnitedStatesowtotheBermudasubsidiaryandcanstaytheretax-freeuntiltheU.S.parentcompanydecidestorepatriatethesefundsbacktotheUnitedStatesatwhichtimetherepatriatedearningswouldbetaxed.Oursampleof108observationsincludes5observationsrelatedtothisshelter.Oursampleincludesothertaxshelters(12rmsinoursample)forwhichthereareveryfew(usuallyonly1)observations.TheseothersheltersincludeCrossBorderDividendCapture(e.g.,Compaq'staxshelterinvolvingADRs),taxsheltersinvolvingshampartnerships,MoneyMarketPrincipalStrips,subsidiarysalesthatgenerateddoublelossesorwerelabeledreorganizationsinordertoobtainintracompanydividendtreatment,andothers.ReferencesBankman,J.,2004.Anacademic'sviewofthetaxshelterbattle.In:Aaron,H.J.,Slemrod,J.(Eds.),TheCrisisinTaxAdministration.BrookingsInstitution,Washington,D.C.,pp.937.Bebchuk,L.,Cohen,A.,Ferrell,A.,2005.Whatmattersincorporategovernance?Workingpaper.HarvardLawSchool.Birnbaum,J.H.,Murray,A.S.,1987.ShowdownatGucciGulch:Lawmakers,Lobbyists,andtheUnlikelyTriumphofTaxReform.RandomHouse,Inc.,NewYork,NY.Bosch,J.-C.,EckardJr.,E.W.,1991.Theprotabilityofpricexing:evidencefromstockmarketreactiontofederalindictments.ReviewofEconomicsandStatistics73,317.Cloyd,C.B.,Mills,L.,Weaver,C.,2002.FirmvaluationeffectsoftheexpatriationofU.S.corporationstotaxhavencountries.JournaloftheAmericanTaxationAssociationSupplement87110.Dechow,P.,Sloan,R.,Sweeney,A.,1996.Causesandconsequencesofearningsmanipulations:ananalysisofrmssubjecttoenforcementactionsbytheSEC.ContemporaryAccountingResearch13,1Desai,M.,Dharmapala,D.,inpress.Corporatetaxavoidanceandrmvalue.ReviewofEconomicsandStatistics.Desai,M.,Hines,J.,2002.Expectationsandexpatriations:tracingthecausesandconsequencesofcorporateinversions.NationalTaxJournal55,4440.Desai,M.,Dyck,A.,Zingales,L.,2007.Theftandtaxation.JournalofFinancialEconomicsNo.3,591623(June).Dhaliwal,D.S.,Erickson,M.,1998.Wealtheffectsoftax-relatedcourtrulings.JournaloftheAmericanTaxAssociation20,2149.Dyreng,S.D.,Hanlon,M.,Maydew,E.L.,2008.Long-runcorporatetaxavoidance.AccountingReview83,61Ellert,J.C.,1975.AntitrustEnforcementandStockholderReturns.Unpublisheddoctoraldissertation.UniversityofChicago.Garbade,K.D.,Silber,W.L.,White,L.J.,1982.Marketreactiontothelingofantitrustsuits:anaggregateandcross-sectionanalysis.ReviewofEconomicsandStatistics64,686691.Gleason,C.,Mills,L.,2002.Materialityandcontingenttaxliabilityreporting.AccountingReview77,317Gompers,P.A.,Ishii,J.,Metrick,A.,2003.Corporategovernanceandequityprices.QuarterlyJournalofEconomics118,107155. TheWallStreetJournal,June16,2004,KPMGShelterShaved$1.7BillionOffTaxesof29LargeCompanies.140M.Hanlon,J.Slemrod/JournalofPublicEconomics93(2009)126 Graham,J.R.,Tucker,A.,2006.Taxsheltersandcorporatedebtpolicy.JournalofFinancialEconomics81,563Karpoff,J.,Lott,L.J.,1993.Thereputationalpenaltyrmsbearforcommittingfraud.JournalofLawandEconomics36,757Karpoff,J.,Lott,J.,Rankine,G.,1999.Environmentalviolations,legalpenalties,andreputationcosts.Workingpaper.UniversityofChicagoLawSchool.Klein,B.,Lefer,K.B.,1981.Theroleofmarketforcesinassuringcontractualperformance.JournalofPoliticalEconomy89,615641.Kothari,S.P.,Warner,J.,2004.Econometricsofeventstudies.In:EspenEckbo,B.(Ed.),Workingpaper,nalversiontoappearintheHandbookofCorporateFinance:EmpiricalCorporateFinance.Elsevier,North-Holland.LosAngelesHeraldExaminer,1984.128bigrmspaidnofederalincometaxes.(October6).Novack,J.,Saunders,L.,1998.ThehustlingofXratedtaxshelters:respectabletaxprofessionalsandrespectablecorporateclientsareexploitingtheexoticaofmoderncorporatenancetoindulgeinextravaganttax-dodgingschemes.Forbes198(December14).Palmrose,Z.,Richardson,V.,Scholz,S.,2004.Determinantsofmarketreactionstorestatementannouncements.JournalofAccounting&Economics37,59Patell,J.,1976.Corporateforecastsofearningspershareandstockpricebehavior:empiricaltests.JournalofAccountingResearch14,246276.ReutersNewsService,2000.41bigrmspaynotax,studysays.(October21).Seida,J.A.,Wempe,W.F.,2004.Investors'andmanagers'reactionstocorporateinversionstransactions.Workingpaper.UniversityofNotreDame.Shackelford,D.,Slemrod,J.,Sallee,J.,2008.Aunifyingmodelofhowtaxesandaccountingrulesaffecttherealdecisionsofcorporations.Workingpaper.UniversityofMichigan.Simpson,G.,2003.Banksshiftedbillionsintofundsshelteringincomefromtaxes.WallStreetJournalA1August7.Slemrod,J.,2004.Theeconomicsofcorporatetaxselshness.NationalTaxJournal57,877899(December).Strachan,J.L.,Smith,D.B.,Beedles,W.L.,1983.Thepricereactionto(alleged)corporatecrime.FinancialReview18,121132.Weisman,J.,2004.Liberalgroupdecriescorporatetaxslide.WashingtonPostE02September23.141M.Hanlon,J.Slemrod/JournalofPublicEconomics93(2009)126