cohen and zarowin 2010

内容需要下载文档才能查看 内容需要下载文档才能查看

JournalofAccountingandEconomics50(2010)2–19

ContentslistsavailableatScienceDirect

JournalofAccountingandEconomics

journalhomepage:http://wendang.chazidian.com/locate/jae

Accrual-basedandrealearningsmanagementactivitiesaroundseasonedequityofferings

DanielA.Cohen1,PaulZarowinÃ

NewYorkUniversity,SternSchoolofBusiness,NY10012-1118,USA

articleinfo

Articlehistory:

Received15January2008

Receivedinrevisedform

22December2009

Accepted18January2010

Availableonline2February2010

JELClassi?cation:

G14

G32

M4

M41

Keywords:

Seasonedequityofferings

Earningsmanagement

Accountingchoices

AccountingAccruals

RealactivitiesabstractWeshowthatSEO?rmsengageinrealactivitiesmanipulation,andthedeclineinpost-SEOperformanceduetotherealactivitiesmanagementismoreseverethanthatduetoaccrualmanagement.Ourevidenceisimportant,becauseitshowsthatpost-SEOoperatingunderperformanceisdrivennotjustbyaccrualreversals,butalsore?ectstherealconsequencesofoperationaldecisionsmadetomanageearnings.Wealsoshowhow?rms’choicesofrealversusaccrual-basedearningsmanagementactivitiesaroundSEOsvarypredictablyasafunctionofthe?rm’sabilitytouseaccrualmanagementandthecostsofdoingso.&2010ElsevierB.V.Allrightsreserved.

1.Introduction

Inthispaper,weexaminebothrealandaccrual-basedearningsmanagementactivitiesaroundseasonedequityofferings(SEOs).PriorresearchhasdocumentedthatSEOsareassociatedwithearningsmanagementandsubsequentdeclinesinoperatingperformance(e.g.,Rangan,1998;Teohetal.,1998;Shivakumar,2000;DuCharmeetal.,2004).Thesestudiesfocusedexclusivelyonaccrual-basedmanipulation,andtheyattributedthepost-SEOperformancedeclineprimarilytoaccrualreversals.Inadditiontoaccrualmanipulation,however,?rmscanmanageearningsbyalteringrealactivities(e.g.,Gunny,2005;Roychowdhury,2006;Zang,2006).Thedistinctionisimportant,becausewhileaccrual-basedearningsmanagementactivitieshavenodirectcash?owconsequences,realactivitiesmanipulationsaffectcash?ows.Werefertorealactivitiesmanipulationasactionsmanagerstakethatdeviatefromnormalbusinesspractices.2

WecontributetotheliteraturebyshowingthatSEO?rmsengageinrealactivitiesmanipulationintheyearoftheSEO,andthedeclineinpost-SEOperformanceduetotherealactivitiesmanagementismoreseverethanthatduetoaccrualÃCorrespondingauthor.Tel.:þ2129980015.

E-mailaddresses:dcohen@stern.nyu.edu(D.A.Cohen),pzarowin@stern.nyu.edu(P.Zarowin).

Tel.:þ2129980267.

2Thisde?nitionisconsistentwithRoychowdhury(2006)whode?nesrealactivitiesmanipulationsas‘‘ymanagementactionsthatdeviatefromnormalbusinesspractices,undertakenwiththeprimaryobjectiveofmeetingcertainearningsthresholds.’’1

0165-4101/$-seefrontmatter&2010ElsevierB.V.Allrightsreserved.doi:10.1016/j.jacceco.2010.01.002

D.A.Cohen,P.Zarowin/JournalofAccountingandEconomics50(2010)2–193

management.Ourevidenceisimportant,becauseitshowsthatpreviouslydocumentedpost-SEOoperatingunder-performanceisdrivennotjustbyaccrualreversals,butalsore?ectstherealconsequencesofoperationaldecisionsmadetomanageearningsatthetimeoftheSEO.

Recentresearchshowsincreasedappreciationfortheimportanceofunderstandinghow?rmsmanageearningsthroughrealactivitiesmanipulationinadditiontoaccrual-basedactivities(e.g.,Gunny,2005;Roychowdhury,2006;Zang,2006).Forexample,Roychowdhury(2006)?ndsevidencethat?rmsusemultiplerealearningsmanagementmethodsinordertomeetcertain?nancialreportingbenchmarkstoavoidreportingannuallosses.Inparticular,hisevidencesuggeststhatmanagersareprovidingpricediscountstotemporarilyboostsales,reducingdiscretionaryexpendituresinordertoimprovereportedmargins,andoverproducingtolowerthecostofgoodssold.

Inarecentsurveyoftopexecutives,Grahametal.(2005)provideevidencesuggestingthatmanagerspreferrealearningsmanagementactivitiescomparedtoaccrual-basedearningsmanagement.Thisisthecasesincerealmanagementactivitiesarelesslikelytobescrutinizedbyauditorsandregulators,andthuspotentiallyhaveagreaterprobabilityofnotbeingdetected,althoughtheconsequencesofsuchactivitiescanbeeconomicallysigni?canttothe?rm.Moreover,consistentwiththeconjecturesmadebyGrahametal.(2005),Cohenetal.(2008)?ndthatmanagershaveshiftedawayfromaccrualtorealearningsmanagementinthepostSarbanes–OxleyAct(SOX)period.Thisevidenceimpliesthatinthepost-SOXperiodfollowinghighlypublicizedaccountingscandals,theneedtoavoiddetectionofaccrual-basedearningsmanagementisgreaterthaninpreviousperiods,inducingmanagerstoshiftfromaccrual-basedtorealearningsmanagementactivities.

Despitetheincreasinginterestinandimportanceofrealearningsmanagementactivities,nostudytodatehasexaminedwhetherandhow?rmsengageinrealearningsmanagementaroundSEOs,andhowrealandaccrual-basedearningsmanagementinteractaroundtheseimportantcorporateevents.We?llthisgapintheliterature.

Tocaptureaccrual-basedearningsmanagementwefollowpriorstudiesthatusethecross-sectionalJones(1991)model(e.g.,DeFondandJiambalvo,1994;Subramanyam,1996).Tocapturerealearningsmanagement,wefollowRoychowdhury(2006)andestimateabnormallevelsofcash?owsfromoperations,discretionaryexpenses(advertising,R&D,andSG&A),andproductioncosts.Inaddition,wecombinethesethreemeasuresintotwocomprehensiveaggregatemetricsofrealearningsmanagement.

Consistentwithpriorstudies(e.g.,Teohetal.,1998,Rangan,1998,amongothers)we?ndthat?rmsuseaccrual-basedearningsmanagementaroundSEOs.Inaddition,alsoconsistentwiththesestudies,we?ndthatSEO?rmstendtobothoutperformtheirindustrypeersintheperiodprecedingtheSEOandunderperformtheirpeersfollowingtheSEO,asevidencedbytheirreturnsonassets(ROA).

Wemakethreecontributionstotheliterature.First,consistentwiththeevidenceinZang(2006)that?rmsusemultipleearningsmanagementstrategies,wedocumentthat?rmsusereal,aswellasaccrual-based,earningsmanagementactivitiesaroundSEOs.Second,weshowhowthetendencyfor?rmstotradeoffrealversusaccrualearningsmanagementaroundSEOsvariescross-sectionally.Weestimateatwo-stagemodelusingtheHeckman(1979)methodtocontrolfor?rms’self-selectiontomanagereportedearnings.Inthe?rststage,weestimateaparsimoniousselectionmodeltoexplain?rms’overalldecisionstoengageinearningsmanagementornot.Next,conditionalonthis?rststageanalysis,inthesecondstageweanalyzethefactorsdeterminingthepreferenceforrealearningsmanagementstrategiesascomparedtoaccrualsbasedones.

Weempiricallymodel?rms’choicestouserealand/oraccrualearningsmanagementstrategiesaroundSEOsasafunctionoftheirabilitytouseaccrualmanagementandthecostsofdoingso.Werepresenta?rm’sabilitytomanageearningsthroughaccrualsbyitsnetoperatingassetsposition(followingBartonandSimko,2002),andthecostsofsuchbehaviorproxiedbyitsauditorcharacteristicsandlitigationprobability.We?ndthatthepresenceofaBig8auditor,longerauditortenure,beinginahigh-litigationindustry,andthelevelofnetoperatingassetsareallpositivelyassociatedwiththetendencytouserealearningsmanagementaroundthetimeoftheSEO.AlthoughweuseourmodeltostudySEOs,weemphasizethatthemodelcouldbeusedmoregenerallytostudyhow?rmstradeoffbetweenthetwomethodsofearningsmanagement,animportanttopicforfutureresearch.

Third,andmostimportant,wecomparethepost-SEOoperatingperformanceassociatedwithaccrualversusrealearningsmanagementactivities.Despitethewidespreadbeliefthatrealearningsmanagementhasgreaternegativeconsequencesthanaccrual-basedearningsmanagement(Grahametal.,2005;Gunny,2005;Zang,2006),thereexistsnoempiricalevidencetodateontheirrelativeeffects.Weprovidethe?rstevidenceonthisimportantissuebyexaminingtheeffectofeachtypeofearningsmanagementactivityona?rm’sfutureperformance.

FollowingRangan(1998),weassesstheeconomicsigni?canceofeachearningsmanagementmethod(i.e.,discretionaryaccruals,abnormalcash?owsfromoperations,abnormalproductioncosts,andabnormaldiscretionaryexpenses)bycalculatingtheimpactofaone-standarddeviationchangeinthecorrespondingmethodonpost-SEOROA.We?ndthatrealearningsmanagementismorelikelythandiscretionaryaccrualstobeassociatedwithearningsdeclines.Thus,ourevidenceshowsthattheeffectsofrealearningsmanagementactivitiesonsubsequentoperatingperformancearelikelygreaterthantheeffectsofaccrualearningsmanagement,atleastintheSEOcontext.

Therestofthepaperisorganizedasfollows.Section2reviewstheliteratureonearningsmanagementaroundSEOsandrealearningsmanagementandstatesourhypotheses.Section3discussesourempiricalmethodology,includingoursampleconstructionandestimationequations.Section4discussesourempiricalevidenceon?rms’useofrealandaccrualearningsmanagementaroundSEOs.Section5discussescross-sectionaldeterminantsofearningsmanagementstrategies

4D.A.Cohen,P.Zarowin/JournalofAccountingandEconomics50(2010)2–19

forSEO?rms.Section6providesevidenceonaccrualandrealearningsmanagementactivitiesaroundthetimeoftheSEOandthepost-SEOoperatingperformanceassociatedwithbothearningsmanagementmethods.Section7concludes.

2.Relatedliteratureandhypothesesdevelopment

Ourpaperunitestwostreamsofresearchthathavepreviouslybeendisparate:oneonrealearningsmanagementandtheotheronearningsmanagementaroundSEOs.3We?rstdiscussrelatedresearch,andthenwebuildontheexistingbodyofevidencetostateourhypotheses.

2.1.Accrual-basedearningsmanagementaroundSEOs

BeginningwithRangan(1998)andTeohetal.(1998),researchershavebeenconcernedwithwhether?rmsmanagereportedearningsduringSEOs,andtheaccountingandstockmarketconsequencesassociatedwithsuchactivities.Thisliteraturehasbeenmotivatedbytheempirical?ndingsthatSEOsarefollowedbybothpoorstockreturnsandpoorearningsperformance,leadingresearcherstosuspectthatearningshavebeenmanagedupwardinanticipationoftheSEO,andsubsequentlyreverse(e.g.,Teohetal.,1998;Rangan,1998;Shivakumar,2000).

Rangan(1998)?ndsthat,onaverage,SEO?rmshavepositiveabnormalaccruals(i.e.,upwardlymanagedreportedearnings)duringtheyeararoundtheSEO,andthattheseaccrualspredictbothearningsreversalsandpoorstockperformanceinthefollowingyear.Heinterpretshis?ndingstomeanthat?rmsmanageearningsupwardaroundSEOs,andthatthestockmarketismisledbytheupwardlymanagedearnings,temporarilyovervaluingissuing?rmsandthenbeingdisappointedbytheirpredictableearningsdeclines.Teohetal.(1998)reportsimilarevidencetoRangan,withtheadditional?ndingthatSEOissuerswhoupwardlymanageearningsmore(greaterpositiveabnormalaccruals)havelowerpost-eventstockreturnsandsubsequentearnings.TheirinterpretationisthesameastheoneofferedbyRangan(1998).Bothofthesestudiesattributethedeclineinpost-SEOoperatingperformanceprimarilytoaccrualreversals.

Shivakumar(2000)also?ndsevidenceconsistentwithaccrualsearningsmanagementaroundSEOs,butincontrasttoRanganandTeohetal.,heshowsthatthestockmarketdoesnotreactinef?cientlytotheupwardlymanagedearnings,butthatinvestorsrationallyundotheseeffects.HeattributesRangan’sandTeohetal.’s?ndingsofabnormalstockreturnstotestmisspeci?cation,arguingthattheearningsmanagementisnotdesignedtofoolormisleadinvestors,butisitselfarationalresponsetothemarket’santicipationthat?rmswillupwardlymanageearningsaroundtheSEO.

DuCharmeetal.(2004)addalegaldimensiontotheresearchonSEOs.TheyshowthatabnormalaccrualsarehighestforSEOsthataresubsequentlysued,andsettlementamountsarealsopositivelyrelatedtothelevelsofabnormalaccruals.Likewise,post-SEOearningsreversalsaremorepronouncedandpost-SEOstockreturnsarelowerforlitigatedSEOs.Theyinterprettheirresultsasindicatingthattheearningsmanipulationdrivesthepost-SEOlitigation.Collectively,thesepapers?ndstrongevidenceofaccrualsearningsmanagementaroundSEOs,andlinktheearningsmanagementactivitiestopost-eventlitigation.Thereisdisagreement,however,astowhetherthestockmarketisindeedmisledbytheearningsmanagementactivitiesaroundSEOs.

2.2.Realearningsmanagement

Althoughrealearningsmanagementhasnotbeenaswidelystudiedasaccrual-basedearningsmanagement,Grahametal.(2005)survey?ndsthatmanagerspreferrealactivitiesmanipulation(e.g.,reducingdiscretionaryexpendituresorcapitalinvestments)overaccrualsmanipulationasawaytomanagereportedearnings.Theserealearningsmanagementactivitiesaresigni?cantlydifferentthanaccrual-basedonesastheyhavedirecteffectsoncash?ows.Grahametal.(2005,p.32)?nd

ystrongevidencethatmanagerstakerealeconomicactionstomaintainaccountingappearances.Inparticular,80%ofsurveyparticipantsreportthattheywoulddecreasediscretionaryspendingonR&D,advertising,andmaintenancetomeetanearningstarget.Morethanhalf(55.3%)statethattheywoulddelaystartinganewprojecttomeetanearningstarget,evenifsuchadelayentailedasmallsacri?ceinvalue.

Thereareatleasttworeasonsforexecutives’greaterwillingnesstomanageearningsthroughrealactivitiesthanthroughaccruals.First,accrual-basedearningsmanagementismorelikelytodrawauditororregulatoryscrutinythanrealdecisions,suchasthoserelatedtoproductpricing,production,andexpendituresonR&Doradvertising.Second,relyingonaccrualmanipulationaloneisrisky.Therealizedshortfallbetweenunmanagedearningsandthedesiredthresholdcanexceedtheamountbywhichitispossibletomanipulateaccrualsaftertheendofthe?scalperiod.Ifreportedincomefallsbelowthethresholdandallaccrual-basedstrategiestomeetitareexhausted,managersareleftwithnooptionsbecauserealactivitiescannotbeadjustedatoraftertheendofthe?scalreportingperiod.

EarningsmanagementaroundSEOsisreallypartofalargerliteraturestudyingearningsmanagementaroundcorporateeventssuchasIPOs,managementbuyouts,stockrepurchases,andstockforstockacquisitions.3

D.A.Cohen,P.Zarowin/JournalofAccountingandEconomics50(2010)2–195

Consistentwiththesepredictions,researchershavedocumentedvariationsinR&Dexpendituresandassetsaleslinkedto?rmsmeetingand/orbeatingearningsbenchmarks.Forexample,Bartov(1993)?ndsthat?rmswithnegativeearningschangesreporthigherpro?tsfromassetsales,suggestingthatthepro?tsareusedtobluntthebadearningsnews.DechowandSloan(1991)?ndthatexecutivesneartheendoftheirtenurereduceR&Dexpenditurestoincreaseshort-termearnings.Inrelatedstudies,Baberetal.(1991)andBushee(1998)reportevidenceconsistentwith?rmsreducingR&Dexpenditurestomeetearningsbenchmarkssuchaspositiveearningsorpreviousyear’searnings.

Threerecentrelatedstudiesexaminerealearningsmanagementactivitiesandtheircapitalmarketconsequences.Roychowdhury(2006)focusesonrealactivitiesmanipulations,whichhede?nesasmanagementactionsthatdeviatefromnormalbusinesspractices,undertakenwiththeprimaryobjectivetomisleadcertainstakeholdersintobelievingthatearningsbenchmarkshavebeenmetinthenormalcourseofoperations.Focusingonthezeroearningsthresholdandexaminingannualdata,he?ndsevidenceconsistentwith?rmstryingtoavoidreportinglossesinthreeways:(1)boostingsalesthroughacceleratingtheirtimingand/orgeneratingadditionalunsustainablesalesthroughincreasedpricediscountsormorelenientcreditterms;(2)overproducingandtherebyallocatingmoreoverheadtoinventoryandlesstocostofgoodssold,whichleadstolowercostofgoodssoldandincreasedoperatingmargins;or(3)aggressivelyreducingaggregatediscretionaryexpenses(de?nedasthesumofR&D,advertising,andSG&Aexpenses)toimprovemargins.Thisreductionismostlikelytooccurwhensuchdiscretionaryexpensesdonotgenerateimmediaterevenuesandincome.Zang(2006)analyzesthetradeoffsbetweenaccrualmanipulationsandrealearningsmanagement.Shesuggeststhatdecisionstomanageearningsthrough‘‘real’’actionsprecededecisionstomanageearningsthroughaccruals.Herresultsshowthatrealmanipulationispositivelycorrelatedwiththecostsofaccrualmanipulation,andthataccrualandrealmanipulationsarenegativelycorrelated.These?ndingsleadhertoconcludethatmanagerstreatthetwostrategiesassubstitutes.

Inrelatedresearch,Gunny(2005)examinestheconsequencesofrealearningsmanagementand?ndsthatrealearningsmanagementhasasigni?cantnegativeimpactonfutureoperatingperformance.Additionally,itappearsthatcapitalmarketsparticipantsmostlyrecognizethefutureearningsimplicationsofmanagers’myopicbehaviors.

Insummary,thereisastrongevidenceofrealearningsmanagementactivities,achievedviamultiplemeansanditislikelylinkedtomeetingcertainearningsbenchmarks.Buildingontheabove,we?rstseektoexaminewhetherrealearningsmanagementactivitiesareusedasearningsmanagementmethodsaroundSEOs.Therefore,our?rsthypothesis(inalternativeform)is

H1:.FirmsthatissueSEOsexhibitevidenceofrealactivitiesmanipulation.

Oursecondobjectiveistoexaminewhetherthereexistsanysubstitutionorcomplementaryrelationbetweenrealandaccrual-basedstrategies.Inparticular,weinvestigatethecross-sectionaldeterminantsofthechoices?rmsmakebetweendifferentearningsmanagementactivitiesaroundSEOs,tobetterunderstandwhycertain?rmswillchoosedifferentearningsmanagementmethodsaroundtheyearoftheSEO.

WeempiricallymodelanSEO?rm’schoicetouserealoraccrualearningsmanagementstrategiesasafunctionofthecostsofusingaccrual-basedearningsmanagementandtheabilitytodoso.Thecostsofusingaccrual-basedearningsmanagementincludethescrutinyprovidedbyauditorsandregulatorsandthepotentiallitigationpenalties(Grahametal.,2005;Zang,2006).Theabilitytouseaccrual-basedearningsmanagementisbasedonthe?rm’saccrualsmanagement?exibility,ascapturedbythecurrentlevelofnetoperatingassets(BartonandSimko,2002),sincethebalancesheetaccumulatestheeffectsofprioraccountingchoices.Therefore,oursecondhypothesisis

H2:.ThedecisionbySEO?rmstomanageearningsbyrealactivitiesmanipulationversusaccruals-basedmanipulationisassociatedwithauditorcharacteristics,litigationprobability,andaccrualsmanagement?exibility(ascapturedbynetoperatingassets).

Finally,andmostimportantly,weanalyzetheeconomicimpactofaccrualversusrealearningsmanagementbyexaminingtheeffectofeachmethodonfutureperformance.InlinewithpriorresearchonSEOs,weinvestigatethesubsequentoperatingperformanceasafunctionoftheearningsmanagementmethodusedaroundtheyearoftheSEO.Therefore,ourthirdhypothesisis

H3:.FutureoperatingperformanceofSEOsisassociatedwiththeextentofaccruals-basedandreal-activities-basedearningsmanagement.

Inthefollowingsection,wediscusstheempiricalmethodologyweemploytotestthesehypotheses.

3.Empiricalmethodology

3.1.Dataandsampledescription

OurSEOsampleconsistsof1511completedUSoffersoverthe1987to2006period,andisobtainedfromtheSecuritiesDataCompany(SDC)NewIssuedatabase.ThesamplecriteriarequiresSEOstobeissuesofcommonstocksbyUSissuersthatarelistedonNYSE,NASDAQ,orAMEXandexcludethefollowing:(1)SEOslackingCOMPUSTATannual?nancial

6D.A.Cohen,P.Zarowin/JournalofAccountingandEconomics50(2010)2–19

statementdataforthe4yearspriortotheSEO?lingdate,theofferyearandthesubsequentyear,(2)completedSEOswithofferpriceslessthan$5andwithdrawnSEOswith?lingrangemidpointlessthan$5,(3)spin-offs,(4)reverseLBOs,(5)closed-endfunds,unitinvestmenttrusts,REITSandlimitedpartnerships,(6)rightsandstandbyissues,(7)simultaneousorcombinedoffersofseveralclassesofsecuritiessuchasunitoffersofstocksandwarrantsand,(8)nondomesticandsimultaneousdomestic–internationaloffers.

Wecollectour?nancialdatafromtheCOMPUSTATannualindustrialandresearch?lesforthesampleperiod.Werestrictoursampletoallnon?nancial?rmswithavailabledata,andrequireatleast8observationsineach2-digitSICgroupingperyear.Further,werequirethateach?rm-yearobservationhasthedatanecessarytocalculatethediscretionaryaccrualsmetricsandrealearningsmanagementproxiesweemployinouranalysis.Thisrestriction(usedinotherpapersonSEOs)likelyintroducesasurvivorshipbiasintothesampleresultingintheinclusionoflargerandmoresuccessful?rms.Weexpectthatthiswillreducethevariationinourearningsmanagementmetricsresultinginamoreconservativetestofourresearchquestions.

FollowingCollinsandHribar(2002),weusecash?owsfromoperationsobtainedfromtheStatementofCashFlowsreportedundertheStatementofFinancialAccountingStandardsno.95(SFASno.95,FASB1987).4Thesampleperiodof1987–2006permitsustouseSFASno.95statementofcash?owdatatoestimateaccruals,ratherthanusingabalancesheetapproach.

Table1reportsdescriptivestatisticsofourSEOsample.PanelAprovidessummaryinformationonthesizeandofferingcharacteristics.Consistentwithpreviousstudies(e.g.,Teohetal.,1998;Rangan,1998),theSEO?rmsinoursampletendtohavelowbook-to-marketratios,averagingaround0.43.Thisisexpected,because?rmstendtoissueequitywhentheirmarketvaluationsarehigh.Theaverageoffersizeisaboutofpre-offermarketvalue.

PanelBshowsthedistributionofSEOsovertime.Also,consistentwiththepreviousSEOresearch,thereissometimeclustering,asmanySEOsoccurredduringthe1990s.Forexample,during1991–1999,therewere879outofourtotalsampleof1511SEOs(58%).Thisisnotsurprising,sincethestockmarketboomedduringthisperiod,andequityvaluationswerehigh.Sincethemarketexperiencedasigni?cantdownturnbeginningin2000,therehavebeenfewerSEOs.

PanelCreportsthefrequencyofSEOsbyindustryandindicatesthat,althoughSEOsoccurinmanydifferenttypesofbusinesses,theywereespeciallyfrequentinhightechnologyareasaschemicalproducts,computerequipment,electronicequipment,andelectronicservicesduringoursampleperiod.Together,thesefourindustriescomprisemorethan40%ofthesample.This?ndingisanticipatedasthehightech?eldsledthe1990sstockmarketboom.Insummary,oursampleexhibitssimilarcharacteristicstosamplesusedinpreviousSEOresearch,indicatingthatSEOsareclusteredbothintimeandbyindustry.

3.2.Earningsmanagementmetrics

3.2.1.Accrual-basedearningsmanagement

Weuseacross-sectionalmodeltocalculatediscretionaryaccruals,whereforeachyearweestimatethemodelforeveryindustryclassi?edbyits2-digitSICcode.Thus,ourapproachpartiallycontrolsforindustry-widechangesineconomicconditionsthataffecttotalaccrualswhileallowingthecoef?cientstovaryacrosstime(Kasznik,1999;DeFondandJiambalvo,1994).5

Ourprimarymodelforestimatingdiscretionaryaccrualsisbasedonthefollowingcross-sectionalmodelestimatedforeach2digitSIC-yeargroupingasfollows:

TAit1DSALESitPPEit¼k1þk2þk3þeitAssetsi;tÀ1Assetsi;tÀ1Assetsi;tÀ1Assetsi;tÀ1ð1Þ

where,for?scalyeartand?rmi,TArepresentsthetotalaccrualsde?nedas

TAit=EBXIitÀCFOit,whereEBXIistheearningsbeforeextraordinaryitemsanddiscontinuedoperations(annualCompustatdataitem123)andCFOistheoperatingcash?ows(fromcontinuingoperations)takenfromthestatementofcash?ows(annualCompustatdataitem308–annualCompustatdataitem124).Asseti,tÀ1representstotalassets(annualCompustatdataitem6),DSALESitisthechangeinrevenues(annualCompustatdataitem12)fromtheprecedingyear,andPPEitisthegrossvalueofproperty,plantandequipment(annualCompustatdataitem7).

Thecoef?cientestimatesfromEq.(1)areusedtoestimatethe?rm-speci?cnormalaccruals(NAit)foroursample?rms

^NAit¼k11PPEit^DSALESitþk^þk23i;tÀ1i;tÀ1i;tÀ1ð2Þ

whereourmeasureofdiscretionaryaccrualsisthedifferencebetweentotalaccrualsandthe?ttednormalaccruals,de?nedasDAit=(TAit/Asseti,tÀ1)ÀNAit.

SFASno.95requires?rmstopresentastatementofcash?owsfor?scalyearsendingafterJuly15,1988.Some?rmsearly-adoptedSFASno.95,sooursamplebeginsin1987.

5Weobtainqualitativelythesameresultswhenweuseatime-seriesapproachthatassumestemporalstationarityoftheparametersforeach?rm.4