财务会计 外文

Earningsforecastdisclosureregulationandearningsmanagement:evidencefromTaiwanIPOfirms

BikkiJaggi*·Chen-lungChin·

Hsiou-weiWilliamLin·PichengLee

󰁢C

SpringerScience+BusinessMedia,Inc.2006

AbstractThisstudyexamineswhethertheTaiwaneseregulationrequiringdisclosureofearningsforecastsintheIPOsresultedindisclosureofmoreoptimisticearningsforecastsandwhethertheforecasterrorwasreducedmorebymanipulatingthereportedearningsratherthanrevisingtheearningsforecaststomeettheforecasterrorthreshold.Thestudyisbasedon759forecastsissuedbytheTaiwaneseIPOfirmsfrom1994to2001,i.e.8-yearperiodaftertheregulationwasmodifiedtoincreasetheforecasterrorthresholdto20%.

Thefindingsshowthatthedisclosureregulationresultedinmoreoptimisticforecaststhanconservativeforecasts,especiallyforfirmsexpectingbetterperformanceintheforecastyearcomparedtothepreviousyear.Firmsdisclosingoptimisticearningsforecastsengagedmoreinmanipulationofreportedearningsthanrevisionofforecaststomeettheforecasterrorthreshold.Thesefindingsthussuggestthatthedisclosureregulationresultedinearningsmanipulation,whichreducedthequalityofreportedearnings.

B.Jaggi(󰀋)

FacultyofManagementSchoolofBusiness,RutgersUniversity,NewBrunswickPiscataway,NJ08854Tel:(732)445-3539

e-mail:jaggi@rbsmail.rutgers.edu

C.-L.Chin

DepartmentofAccounting,NationalChengChiUniversity,Taipei,Taiwan

H.-W.W.Lin

DepartmentandGraduateInstituteofInternationalBusiness,NationalTaiwanUniversity,Taipei,Taiwain

P.Lee

DepartmentofAccounting,LubinSchoolofBusiness,PaceUniversity,NewYork,NY10038*Wereceivedvaluablecommentsatthe2003AmericanAccountingAssociationand2004EuorpeanAccountingAssociationannualmeetings.WealsothanktheparticipantsattheresearchseminarsatRutgersUniversity,CityUniversityofHongKong,andPaceUniversity,fortheirinsightfulcomments.PichengLeeespeciallythanksPaceUniversityfor2003summerresearchgrant.

Springer

276RevQuantFinanAcc(2006)26:275–299

KeywordsEarningsmanagement.Disclosureregulation.Discretionaryaccruals.Mandatoryearningsforecasts.Forecasterrorthreshold

1.Introduction

In1991,theTaiwanSecuritiesandFuturesExchangeCommission(TSFEC)issuedaregu-lationrequiringtheIPOfirmstoincludeannualearningsforecastsintheIPOprospectusesandalsodiscloseforecastsfortwoyearssubsequenttotheissuanceofIPO.1Theunderlyingobjectiveofthisregulationhasbeentoreduceasymmetryofinformationamonginsidersandinvestorsatlargebydisseminatingforecastinformationtoallpotentialinvestorsonafairandequitablebasis,2sothattheusefulnessofearningsinformationforinvestmentdecisionscouldbeenhanced.Theregulationalsosetsaforecasterrorthresholdandincludesaprovisionthatallowsthefirmstorevisetheforecasts,apparentlytodiscouragethemtomanipulatethereportedearningsformeetingtheforecasterrorthreshold.Theregulation’sintentionthusappearstoreducetheasymmetryofinformationwithoutloweringtheinformationqualityofreportedearningsthroughmanipulation.

ThisstudyempiricallytestswhetherthedisclosureregulationencouragedTaiwanfirmstoissuemoreoptimisticforecastsandwhethertheforecasterrorofoptimisticforecastswasreducedbymanipulatingthereportingearningsusingpositivediscretionaryaccru-alsratherthanrevisingtheearningsforecastsdownward,asallowedbytheregulations.WeargueinthispaperthatthedisclosurerequirementismorelikelytoencourageIPOfirmstoissuemoreoptimisticforecastsforsendingpositivesignalstothemarketsothatIPOproceedscouldbemaximized.Theoptimisticforecastswillespeciallybeis-suedwhenthereisanexpectationofanimprovementintheoperatingperformancedur-ingtheforecastyearcomparedtothepreviousyear.Wefurtherargueinthispaperthatmanagerswillchoosetheoptionofadjustingthereportedearningsupwardusingposi-tivediscretionaryaccrualstoreducetheforecasterrorofoptimisticforecastsformeetingtheforecasterrorthreshold.Theywouldavoidtheoptionofrevisingtheoptimisticfore-castsdownwardbecausedownwardrevisionscouldbeinterpretedasnegativesignalsbythemarket.

Thestudyisbasedon759earningsforecaststhatweredisclosedby253TaiwanindustrialIPOfirmsfrom1994through2001.3Discretionaryaccrualsareusedasaproxyforearningsmanagement,andchangeinthereturnonassets(ROA)onanex-postbasisisusedasaproxyforexpectationofbetterfutureoperatingperformance.WeconductOLSregressionaswellasLogitregressionanalysestoevaluatetheassociationbetweenoptimisticforecasts

TheTSFECissued“GuidelinesforDisclosureofFinancialForecastsbyPublicCompanies”,whichrequiredthecompaniesdesiringtolistwiththeTaiwanStockExchange(TSE)todisclosefinancialforecastspursuanttothelisting.ThepreparationoffinancialforecastsshallbesubjecttotheStatementsofFinancialAccountingStandardsNo.16“PreparationofFinancialForecasts”publishedbytheAccountingResearchandDevelopmentFoundationofRepublicofChinainTaiwan.

TheTSFEC’sadministrativerule(No.82-Taiwan-Finance-Securities-(VI)-02581issuedonOctober30,1993)mentionsthatdisclosureofforwardlookinginformationwillreduceinformationgapbetweeninformedanduninformedtraders,andthiswillfinallypreventtheuseofinsideinformationformakingarbitraryprofits.Theregulationwasissuedin1991andtheforecasterrorthresholdwasoriginallysetat10%offorecasts.In1994,theforecasterrorthresholdwasrevisedto20%Inordertohaveuniformsampleforthestudy,wefocusonIPOsissuedaftertheforecasterrorthresholdwasrevised,i.e.1994.Springer

32

1

RevQuantFinanAcc(2006)26:275–299277

andexpectationofbetteroperatingperformanceduringtheforecastyear.T-testsandnon-parametricWilcoxontestsareconductedtoevaluatewhetherreductionintheforecasterrorisachievedthroughearningsmanagementorforecastrevisions.

TheresultsofthisstudyshowthatTaiwanIPOfirmsissuedmoreoptimisticforecaststhanconservativeforecastsafterthedisclosureregulationwasimposed.TheOLSandLogitregressionresultsshowthatthereisapositiveassociationbetweenoptimisticforecastsandchangeinROAintheforecastyearcomparedtothepre-forecastyear,suggestingthattheIPOfirmsespeciallyissuedmoreoptimisticforecastswhenoperatingperformanceduringtheforecastyearwasexpectedtobebetterthanthepreviousyear.Theresultswithregardtoreductionofforecasterrorshowthatinthecaseofoptimisticforecaststheforecasterrorthresholdismetmostlybyadjustingthereportedearningsupwardusingdiscretionaryaccru-alsthanrevisingearningsforecasts.Theseresultsthusindicatethattheregulationencouragedmanagerstodisclosemoreoptimisticforecastsandthereportedearningsaremanipulatedtomeettheforecasterrorthreshold,whichreducesthequalityofearningsinformation.Thesefindingsthussuggestthattheregulationhasnotachievedtheobjectiveofenhancingtheuse-fulnessofreportedearningsforinvestmentdecisionsbecausethequalityofreportedearningshasbeencompromised.

Thestudymakesthefollowingimportantcontributionstotheaccountingliterature.First,thefindingsshowthatmandatorydisclosureofearningsforecastsislikelytoresultinmoreoptimisticforecastsbyIPOfirms.Second,managersarelikelytomanipulatethereportedearningstomeettheforecasterrorthreshold,especiallyofoptimisticforecasts,andthiswillreducethequalityofreportedinformation.Thus,amandatoryregulationfordisclo-sureofearningsforecasts,especiallyintheSouthEastAsiancountries,whicharegen-erallyassociatedwithlowlegalinvestorprotection,islikelytoreducethequalityofre-portedearnings,whichwillenhancetheusefulnessofearningsinformationforinvestmentdecisions.

Theremainderofthepaperisorganizedasfollows.InthepartII,webrieflydescribetheTSFECregulationandprovidealiteraturereviewrelatedtothestudy.TheresearchdesignisdiscussedinpartIII,whichincludesdiscussiononhypotheses,researchmethodologyusedinthestudy,andsampleselectionprocedures.TheresultsarepresentedinpartIVandconclusionisprovidedinpartV.

2.TSFECregulationandliteraturereview2.1.TSFECregulation

Disclosureofearningsforecastsonavoluntarybasisisnotcommonincountrieswithde-velopingeconomies.4Consequently,investors,whohavenoaccesstoinsideinformation,dependonsourcesotherthandisclosuresbymanagersforforward-lookinginformation.Thus,non-availabilityofreliableinformationonthefirms’futureoperatingperformancetoallpotentialinvestorsonanequitablebasisresultsinsignificantinformationasymmetrybe-tweeninformedanduninformedinvestors.InordertoreduceinformationasymmetryforIPO

4

OurinvestigationindicatedthattheTaiwanIPOfirmsrarelyissuedearningsforecastsbeforetheTSFECregulation.Moreover,thedevelopmentofearningsforecastsbysecurityanalystsisnotacommonphenomenoninTaiwan.

Springer

278RevQuantFinanAcc(2006)26:275–299

firms,5theTSFECissuedaregulationin1991thatmadeitmandatoryforTaiwanIPOfirmstoincludeearningsforecastsintheirIPOprospectusesandalsodiscloseearningsforecastsfortwoyearsafterissuanceofIPOs.

TheTSFECregulationalsostipulatespenaltyforfirmsthatdonotmeettheforecasterrorthreshold.Thefirmsviolatingtheforecasterrorthresholdarenotallowedtoraiseadditionalcapitalinthemarketwithoutspecialpermission.Thoughtheoriginalthresholdwassetat10%offorecasterror,itwasrevisedto20%effective1994.Accordingtotherevisedforecasterrorthreshold,theactualreportedearningsshallnotdeviatemorethan20%fromthepredictedearningsineitherdirection.Thefirmsare,however,allowedtorevisetheirforecastswithoutanyrestrictiononthenumberofrevisionsbeforeactualearningsarereported.6Theintendedobjectiveofallowingforecastrevisionsisprobablytoencouragemanagerstoissuemorerealisticearningsforecasts.TheTSFECmay,however,alsoaskthefirmstorevisetheearningsforecastsifitconsideredthattheforecastswereunrealistic.7Ifviolatingfirmsintendtoraiseadditionalcapitalinthemarket,theyarerequiredtorequestforareviewprocess.8

2.2.Literaturereview

MostcountriesallowdisclosureofearningsforecastsbytheIPOfirmsonavoluntarybasis.WebrieflyreviewtheliteraturedealingwithearningsforecastsissuedbyIPOfirms(includedintheIPOprospectuses)intheUK,NewZealand,Australia,CanadaandothercountriesintheSouthEastAsianregion.ThemainfocusofthisreviewistoexaminewhethertheIPOfirmsissuedmoreoptimisticorconservativeforecastsandhowtheforecasterrorisreduced.Additionally,wediscussrelevantstudiesontheUSIPOfirms,especiallywithregardtoearningsmanagementbythesefirmsduringthepre-IPOperiod.

Onanoverallbasis,thereviewindicatesthatdisclosureofearningsforecasts,especiallybytheUK,AustralianandCanadianfirmsonavoluntarybasisdidnotresultindisclosureofmoreoptimisticforecasts,probablybecauseofwell-developedfinancialmarketsinthesecountries.ThefindingsofstudiesontheforecastsissuedbytheUKIPOfirmsindicatethattheseforecastshavegenerallybeenmoreconservativethanoptimistic,i.e.thereportedearningsaregenerallyhigherthanearningsforecasts.Thus,therehasbeennoneedtoadjustthereportedearningsupwardtoreducetheforecasterrors.

Wangetal.(2005)explorethedifferencesinIPOanomaliesbeforeandafterthenewregulationwasissuedinJune,1991.Theempiricalresultssupporttheirpresuppositionsthattheinitialreturns,honeymoonperiodsandaftermarketreturnsofmandatorydisclosersaftertheimpositionofthe1991financialforecastsregulationweresignificantlylowerthanthoseofnon-mandatorydisclosers.

TheTSFECregulationhasplacednolimitonthenumberofrevisionssolongastherevisionsareissuedbeforetheactualearningsarereported.

Examplesofthedirectivestorevisetheforecastsarefoundinthefollowingarticle,whichisexcerptedfromChinaTimespressdatedOctober14,1998:“AfterthetwocasesofundulybiasedmanagementforecastsdonebyTung-HoSteel...andActionElectronics,TSFEChasrequestedeightmorefirmstorevisetheirfinancialforecasts.Sincetherearemanyotherfirmscomingoutwithdownwardforecastrevisions,itisexpectedthatmorefirmswillshowuponTSFEC’ssanctioninglist.”

TSFEC’sattitudeplaysanimportantpartintheprocessofIPOapplication.AsstatedinArticle12oftheCriteriaGoverningtheOfferingandIssuanceofSecuritiesbySecuritiesIssuersinTaiwan,newsharesofalistedcompanycanbeissueduponapprovalofTSFEC.IfafirmisrequiredbytheTSFECtocorrectthepubliclydisclosedfinancialforecastfortwotimesintheyearofapplicationorduringthetwopreviousyearsoreventhefinancialforecastsaremodifiedmorethantwotimesinanyyear,theTSFECmaynotallowafirmtoraiseequityordebtcapitalinthemarket.Springer

8765

RevQuantFinanAcc(2006)26:275–299279

Priorto1991,theAustralianfirmsrarelydisclosedearningsforecastsintheIPOprospec-tuses,andiftheforecastsweredisclosed,theywerefoundtobehighlyinaccurate(e.g.Brown,etal.,2000).Afterthe1991CorporationsLaw,therehasbeenanincreaseinthefrequencyofforecastsincludedintheIPOprospectuses,andtheiraccuracyalsoimproved(HowandYeo,2001).Thesestudies,however,donotevaluatehowtheforecasterrorisreduced.

ThoughtheOntarioSecuritiesCommissionallowedtheCanadianfirmstoincludeearningsforecastsintheirIPOprospectusesonavoluntarybasisfrom1982,thefirmsrarelyincludedtheearningsforecastsintheirIPOprospectuses.AccordingtoClarkson,etal.(1992),theCanadianIPOfirmsdiscloseearningsforecastsonlywhentheybelievethattheforecastswouldimprovemarketexpectations.

UnlikemanyotherBritishCommonwealthcountries,theNewZealandsecuritiesregula-tionsrequiredisclosureofprofitforecastsintheIPOprospectuses(FirthandSmith,1992).TheforecastsincludedintheIPOprospectusesofBritsishfirmsare,however,foundtobeconservative(e.g.DevandWebb,1972).Mak’s(1989)findingsreportthatonanoverallbasistherehasbeenanabsoluteearningsforecasterrorintherangeof100percentfortheNewZealandfirms,andthatthereisapositiveassociationbetweenforecasterrorandforecasthorizon.Inarecentstudybasedontheirsamplefromtheperiod1987through1994,Hsu,etal.(2000)findthattheforecastsissuedbytheNewZealandfirmsinfacthavebeenmorepessimistic.

TheIPOfirmsinsomeSouthEastAsiancountriesalsoissueearningsforecastsonavoluntarybasisandseveralstudieshaveexaminedaccuracyoftheirforecasts.WithregardtotheHongKongIPOfirms,Chanetal.(1996)findthatthat60%offorecasterrorsoftheirsamplefirmshaveameanabsoluteforecasterrorof18%.Jaggi(1997)reportedthattheearningsforecastsofHongKongfirmsarecomparativelymoreaccuratethantheIPOforecastsfromothercountries,suchasAustralia,Canada,andNewZealand.ChengandFirth(2000)alsofindthattheforecastsbyIPOfirmsaremoreaccuratethantheIPOforecastsfromothercountries.

SomeothercountriesintheSouthEastAsianregion,suchasMalaysiaandSingapore,requiredisclosureofearningsforecastsintheIPOprospectuses.Jelic,etal.(1998)haveexaminedtheaccuracyofearningsforecastsincludedintheIPOprospectusesofMalaysianfirmsandtheyfindthat,onaverage,themeanforecasterroris+33%,whiletheabsoluteforecasterroris55%.Firthetal.(1995),whoexaminedforecastaccuracyof114SingaporeIPOprospectusesfortheperiodof1980to1993,reportameanabsoluteforecasterrorof10%andameansignedforecasterrorof20%forthesefirms.TheirfindingsthusindicatethatearningsforecastsprovidedinSingaporeanIPOprospectusesaremoreaccuratethanthosemadeinAustraliaandUK.

TheChinaSecuritiesRegulatoryCommission(CSRC)alsorequirestheChineseIPOfirmstodiscloseprofitforecastoftheIPOyear.ChenandFirth(1999)reportthatonaveragetheforecastsincludedintheprospectusesoftheChineseIPOfirmshavebeenconservative.ThemagnitudeofabsoluteforecasterrorsisfoundtobehigherthanthatoftheHongKongandSingaporefirms,butlowerthanthatoftheAustralian,Canadian,andNewZealandfirms.TheUSSecuritiesandExchangeCommission(SEC)doesnotrequireearningsforecaststobeincludedintheIPOprospectuses,andtheUSfirmsrarelyincludeearningsforecastsintheIPOprospectusesonavoluntarybasis.SeveralstudieshavebeenconductedwhethertheIPOfirmsmanipulatetheirearningsduringthepre-IPOperiod.ThefindingsofthesestudiesshowthattheUSfirmsgenerallyengageinupwardmanipulationofreportedearningsduringthepre-IPOyearstosendpositivesignalstothemarket(e.g.Friedlan,1994;Teoh,WelchandWong,1998b;Teoh,WelchandRao,1998;DuCharme,MalatestaandSefcia,2001).Asaresultofupwardmanipulationofreportedearningsinthepre-IPOyears,aportionofthe

Springer

280RevQuantFinanAcc(2006)26:275–299

earningsisshiftedfromthepost-IPOperiodtothepre-IPOperiod.Consequently,theIPOfirmsarefoundtobeassociatedwithunderperformanceduringthepost-IPOperiod(e.g.Teoh,WelchandWong,1998b).

ThoughtheUSIPOfirmsgenerallydonotissueearningsforecasts,someofthem,how-ever,arefoundtoissueearningsforecastsonavoluntarybasisaftertheIPOperiod.StudieshavebeenconductedtoexaminewhethertheissuanceofearningsforecastsbytheUSonavoluntarybasisduringthenon-IPOperiodwouldresultinmanipulationofreportedearnings.Ithasbeenreportedthattheforecast-issuingfirmsgenerallyengageinearningsmanipulationtoreducetheforecasterror.Kasznik(1999)findsthatthefirmsgenerallyusepositivediscre-tionaryaccrualstoreducetheforecasterror,whereasJaggiandSannella(1995)reportthatthemanagersmaychoosetheaccountingmethodthatwouldreducetheforecasterror.Themotivationforreducingtheforecasterrorisgenerallytoavoidcostsassociatedwithpoten-tiallegalactionsbyshareholdersifthereportedearningswoulddeviatefromtheforecastedearningsconsiderably(e.g.Skinner,1994;Frankel,JohnsonandNelson,1995;Teoh,WelchandRao,1998b;Baginski,Hassell,2002).

3.Researchdesign

Inthissection,wefirstdevelophypothesesforthestudyandthenexplainhowthefore-casterrorismeasuredanddiscretionaryaccruals,whichareusedasaproxyforearningsmanagement,arecalculated.Wealsodiscusstheproceduresforsampleselection.3.1.Hypotheses

3.1.1.Disclosureregulationandearningsforecasts

TheregulatorybodiesintheUSAandotherindustrializedcommonlawcountriesgenerallydonotrequireinclusionofearningsforecastsintheIPOprospectuses.Thefirmsmay,however,includeearningsforecastsintheIPOprospectusesonavoluntarybasisinsomecountries.AnimportantreasonfornotincludingearningsforecastsintheIPOprospectusesonavoluntarybasisinthecommonlawcountriesisprobablytoavoidpotentiallegalactionbyinvestorsiftheforecastsarenotrealized.Intheabsenceofearningsforecasts,theUSIPOfirmsaregenerallyfoundtosendpositivesignalstothemarketontheirIPOsbymanipulatingthereportedearningsduringthepre-IPOyears(e.gTeoh,WelchandRao,1998b;Clarkson,etal.,1992).

Unliketheregulatorybodiesinthecommonlawindustrializedcountries,theTaiwanregulatorybody,i.e.theTaiwanSecuritiesandFuturesExchangeCommission(TSFEC),issuedaregulationrequiringthefirmstoincludeearningsforecastsintheIPOprospectusesandtodiscloseearningsforecastinformationfortwoyearsafterissuingtheIPO.ThemainpurposeofthisregulationhasbeentoreduceasymmetryofinformationwhenIPOsareissued.Itwasrealizedthatnotenoughfuture-orientedinformationwasavailabletoinvestorsontheIPOfirmsformakingtheirinvestmentdecisionspriortoearningsdisclosureregulation.9Themandatorydisclosureregulationwasimposedexpectingthatreliablefuture-orientedearningsinformationwouldbecomeavailabletoallinvestorsonafairandequitablebasis.

Only5outof89IPOfirmsmadeearningsforecastsonavoluntarybasisfortheperiodof1983to1991Junebeforeearningsdisclosureregulationwasimposed.Springer

9

RevQuantFinanAcc(2006)26:275–299281

Inthispaper,wearguethattherequirementofissuingearningsforecastswouldencouragemanagerstoissuemoreoptimisticforecaststhanrealisticorconservativeearningsforecasts.ManagerswouldbemotivatedtomaximizetheirIPOproceedsbysendingpositivesignalstothemarketonthefirm’sfutureperformancethroughoptimisticforecasts.10TheexpectationofoptimisticforecastsisalsosupportedonthegroundthatinvestorprotectionrightsundertheTaiwaneselegalsystem,similartothelegalsystemofothercountriesintheAsianPacificBasinregion,areweak(e.g.LaPortaetal.,1998;FanandWong,2002).Inordertodetermanagerstoissuebiasedearningsforecastsandprotectinvestors’interests,theTSFECregulationincludesapenaltyclauseifthefirmsfailtomeettheforecasterrorthreshold.Thispenaltyisintheformofspecialpermissionforraisingadditionalcapitalinthemarket.Thethrustofthispenaltyis,however,weakenedbyanotherprovisionintheregulationthatenablesthefirmstoobtainawaiverinspecialcaseswhentheforecastthresholdisviolated.Inviewofweakinvestorlegalprotectionrightsandnot-so-stringentpenaltyundertheregulation,wearguethatmanagersofIPOfirmswouldhaveastrongmotivationtoissueopti-misticforecastsforsendingpositivesignalstothemarket,andthiswillreducethereliabilityofearningsforecasts.Wetestthisexpectationonthefollowinghypothesis:

H1:TheIPOfirmsissuemoreoptimisticforecaststhanrealisticorconservativeforecastsaftertheimpositionofTSFECregulationrequiringtheIPOfirmstodiscloseearningsforecasts.

3.1.2.Optimisticforecastsandfutureexpectedoperatingperformance

EventhoughtheTSFEC’sdisclosurerequirementwouldprovideastrongmotivationtoman-agersofIPOfirmstoissueoptimisticforecaststomaximizetheirIPOproceedsbecauseoflowlegalpotentialcostsforissuingsuchforecasts,weexpectthemanagerstobeconcernedabouttheirreputationandcredibilityinthemarketplaceforissuingoptimisticforecastswith-outconsideringthefirm’sfutureoperatingperformance.Inordertomaintaintheirreputationandcredibility,weexpectthemanagerstoissueoptimisticforecastsespeciallywhentheyexpectanimprovementinthefirm’sfutureoperatingperformance.Ifthefutureoperatingperformanceisnotexpectedtoimprove,themanagerswouldbecautiousinissuingoptimisticforecasttoavoidtarnishingtheirreputation,whichcouldresultinlosingcredibilityfortheminthemarketplace.

Weusethefirm’sfutureoperatingperformanceonanex-postbasisasaproxyformanagers’expectation,andmeasuretheoperatingperformancebythereturnonassets(ROA),basedonthefirm’soperatingnetincome.WeusethechangeinROAfortheforecastperiodoverthepre-forecastperiodtoevaluatewhetherthemanagerexpectedanimprovementintheoperatingperformanceduringtheforecastperiod.ApositivechangeintheROAwouldindicateanimprovement,whereasnochangeoranegativechangewouldreflectanexpectationofnoimprovementintheoperatingperformance.Wedevelopthefollowinghypothesistotestthisexpectation:11

H2:ThereisapositiveassociationbetweenoptimisticearningsforecastsandchangeinROA,aproxyfortheexpectationonoperatingperformance.

FindingsoftheexistingstudiesbasedonUSfirmsshowthattheIPOfirmssendpositivesignalstothemarkettooptimizetheIPOproceeds(Teoh,WelchandRao,1998,199b).

11

10

WealsotestthishypothesisontheROAchangebasedonthepre-managedearnings.

Springer

282RevQuantFinanAcc(2006)26:275–299

3.1.3.Earningsmanagementandforecasterror

TheTSFECregulationsetsaforecasterrorthresholdwhichthefirmsarerequiredtomeet.Iftheyviolatethisthreshold,theyareheldinviolationoftheregulationandpenaltyforthisviolationisthattheywouldnotbeallowedtoraisefundsinthemarketinfuturewithoutspecialpermission.Thus,iftheforecasterrorisoutsidethethreshold,managerswillhavetomakeastrategicdecisionwhethertointerveneandbringtheforecasterrorwithinthethresholdornot.Eventhoughthepenaltyfornotmeetingtheforecasterrorthresholdisnotsosevere,wearguethatmanagerswouldattempttostaywithintheforecasterrorthresholdbecauseitwouldenhancetheircredibility.

Reductionintheforecasterrorwouldbeachievedeitherbyrevisingtheearningsfore-casts,asallowedundertheTSFECregulation,orbymanipulatingthereportedearningsusingdiscretionaryaccruals.Wearguethatmanagerswouldchosethealternativeofreducingtheforecasterror,especiallyofoptimisticforecasts,byadjustingthereportedearningsusingdis-cretionaryaccrualsratherthanrevisingtheinitialearningsforecasts.Inthecaseofoptimisticforecast,wearguethatdownwardrevisionwouldbeavoidedbecauseitcouldsendnegativesignalstothemarket,whichwouldhaveanegativeimpactonthemanagers’credibility(e.g.Kasznik,1999).Inthecaseofconservativeforecasts,managerscouldadjusttheearningsforecastsupwardoradjustthereportedearningsdownwardandcreate“cookiejar”reservesforuseinthefutureperiods.Thechoicebetweenforecastrevisionsandearningsmanagementwouldbemadedependingonthesituationandtoachievethesetobjectiveunderthatsituation.Inthisstudy,wetestwhethertheforecasterrorofoptimisticforecastswillbereducedbymanipulatingthereportedearningsusingdiscretionaryaccrualsorrevisingearningsforecasts.Thefollowinghypothesisisdevelopedtotesttheassociationbetweenearningsmanagementandforecasterrorreduction:

H3:Theforecasterrors,especiallyofoptimisticforecasts,arereducedmorebyadjustingthereportedearningsusingdiscretionaryaccrualsthanbyrevisingtheearningsforecasts.3.2.Calculationofforecasterrors

Theforecasterroriscalculatedasthedifferencebetweenthepredictedearnings(PE)andreportedearnings(RE)deflatedbytheabsolutevalueofPE.TheabsolutevalueofPEisusedasascalertoeliminatetheeffectofminusvaluesofPEinthedenominator.12FE=(PE−RE)/|PE|Where,

FE=ForecastError

PE=Predicted(Forecasted)EarningsRE=ReportedEarnings

?PE?=AbsolutePredictedEarnings

(1)

12

ThisformulaisalsousedbyTSFECtoevaluatetheforecasterrorofIPOs.Springer

RevQuantFinanAcc(2006)26:275–299283

Weidentifyinitialforecasts(PEIN)andlastrevisedforecasts(PELR),13andsimilarlyweidentifyactualreportedearnings(RE),andactualpre-managedearnings(PME).ThePMEisobtainedbydeductingtotaldiscretionaryaccruals(TDA)fromthereportedearnings(RE):PME=(RE−TDA)

(2)

Wecalculatethreetypesofforecasterrorstoevaluatethefrequencyandmagnitudeoferrorsatdifferentstagesofthemanagerialdecisionmakingprocess.First,wecalculatetheforecasterrorsbasedonthedeviationsbetweeninitialpredictedearnings(PEIN)(withoutanyrevisions)andactualpre-managedearnings(PME),andthiserroristermedasFE1forecasterror.Thisforecasterrorisusedtoevaluatetherelativemagnitudeofdifferencebetweeninitialforecastsandpre-managedearnings,whichwillsignifytheamountofforecasterrorthatneedstobereducedformeetingtheforecasterrorthreshold.FE1=(PEIN−PME)/|PEIN|

(3)

ThepositivesignofFE1(FE1>0)indicatesthatthepredictedearningsarehigherthanactualearnings,andinthiscasetheforecastsarereferredtoasoptimisticforecasts.Iftheforecasterrorislessthan0(FE1<0),itmeansthatthereportedearningsarehigherthanthepredictedearnings,andtheseforecastsarereferredtoasconservativeforecasts.

TheforecasterrorFE2iscalculatedonthebasisoflastrevisedforecast(PELR)andpre-managementearnings(PME),andthisforecasterrorsignifiesthemagnitudeofforecasterroraftertheforecastsarerevised.Iftheforecasterrorthresholdismetbyrevisingtheearningsforecasts,therewillbenoneedforadjustingactualearnings.FE2=(PELR−PME)/|PELR|

(4)

Third,wecalculatetheforecasterrorbasedonlastrevisedforecastandreportedearnings,anditistermedasFE3forecasterror.Thismeasureindicateswhethertheforecasterroriswithintheforecasterrorthresholdanditwouldalsoprovideinformationonthemagnitudeofforecasterrorreducedbyrevisingtheforecastsand/oradjustingactualearnings.FE3=(PELR−RE)/|PELR|

3.3.Calculationofdiscretionaryaccruals

Moststudiesintheliteraturehaveusedtotaldiscretionaryaccruals(TDA)toevaluateearningsmanagement.CalculationofTDAarebasedonthetotalassets,changesinrevenuesandaccountsreceivables,andfixedassetsofthefirms(e.g.Jones,1991;DeFondandJiambalvo,1994;JaggiandLee,2002).14Recently,ithasbeenpointedoutthatestimationofdiscretionary

Thestatisticsonrevisionofforecastsshowsthatmostforecastswererevisedonlyonce,whereassomeofthemwererevisedtwiceandonlythreeforecastswererevisedthreetimes.Wecalculateforecasterrorbasedoninitialforecastsforallforecasts.Theforecasterrorsonlastrevisionwillonlybecalculatedforthoseforecaststhatwererevised.Iftheforecastwasnotrevised,theforecasterrorforinitialforecastsandlastrevisedforecastwillbethesame.Incasetheforecastwasrevisedonce,thentheforecasterrorforlastrevisionwillbebasedonthisfirstrevision,andthesimilarprocedureisfollowedforsecondandthirdrevisionofforecasts.Theeffectivenessofdiversifiedcross-sectionalandtime-seriesdiscretionaryaccrualmodelshasbeenex-aminedbyDechow,SloanandSweeney(1995),Guay,Kothari,Watts(1996)andSubramanyam(1996).They

Springer

1413

(5)

284RevQuantFinanAcc(2006)26:275–299

accrualisalsoinfluencedbythefirm’sperformance.Therefore,thefirm’sperformance,proxiedbythefirm’sROA,isincludedinestimatingdiscretionaryaccruals(e.g.Kothari,LeoneandWasley,2005;Ashbaugh,LafondandMayhew,2003).Lately,severalstudiesalsohaveemphasizedtheimportanceofcurrentdiscretionaryaccruals(CDA)inevaluatingearningsmanagement(e.g.Ashbaugh,LafondandMayhew,2003;Becker,etal.,1998;Frankel,JohnsonandNelson,2002).Therefore,wedecidedtouseboththegrowth-adjustedTDAandCDAtoobtainrobustresults.

TDAiscalculatedinthefollowingsteps:Inthefirststep,totalaccruals(TA)arecalculatedasthedifferencebetweennetincomebeforeextraordinaryitemsandoperatingcashflows:TAit=NIBEit−OCFitwhere,

TAit=totalaccrualsforfirmiinyeart,

NIBEit=netincomebeforeextraordinaryitemsforfirmiinyeart,OCFit=operatingcashflows.

Inthesecondstep,parametercoefficientsareestimated:

TAit/Ait−1=α1[1/Ait−1]+β1[󰀑REVit/Ait−1−󰀑ARit/Ait−1]+β2[PPEit/Ait−1]

+β3[ROAit−1]+εit

Where,

TAit=totalaccrualsforfirmiinyeart,Ait−1=totalassetsforfirmiinyeart−1,

󰀑REVit=changeinnetrevenuesforfirmiinyeart,

󰀑ARit=changeinaccountsreceivableforfirmiinyeart,PPEit=grosspropertyplantandequipmentforfirmiinyeart,ROAit−1=operatingincomebytotalassetsforfirmiinyeart−1,εit=errortermforfirmiinyeart.

Asalaststep,thetotaldiscretionaryaccruals(TDA)arecalculatedusingtheestimatedparametersfromequation(7):

(7)(6)

TDA=TAit/Ait−1−{a1[1/Ait−1]+b1[󰀑REVit/Ait−1−󰀑ARit/Ait−1]

+b2[PPEit/Ait−1]+b3[ROAit−1]}

whereuitrepresentstotaldiscretionaryaccruals(TDA)forfirmiintheeventyeart.InordertocalculateCDA,wefirstcalculatetotalcurrentaccruals(TCA),asfollows:TCAit=(󰀑CAit−󰀑CASHit)−(󰀑CLit−󰀑STDit)

(9)(8)

havealsopointedouttheweaknessesofeachmodel.Thefindingsofastudy(Dechow,SloanandSweeney,1995)indicatethatmodifiedcross-sectionalJonesmodelperformsbetterthananyothermodels.Subramanyam(1996)suggeststhecross-sectionalversionoftheJonesmodelhasstatisticalpropertiesthatmakeitbetter,exante,thanitstime-seriescounterpart.Springer

RevQuantFinanAcc(2006)26:275–299285

Where,

TCAit=totalcurrentaccrualsforfirmiinyeart,CAit=currentassetsforfirmiinyeart,CASHit=cashforfirmiinyeart,

󰀑CAit−󰀑CASHit=changeinthedifferenceofcurrentassetsandcashforfirmiinyeart,

󰀑CLit=changeincurrentliabilitiesforfirmiinyeart,

󰀑STDit=changeintheportionoflong-termliabilitiesduewithinayearofFirmiforperiodt.

TheparametersforcalculatingCDAareestimatedbyusingthefollowingequation:TCAit/Ait−1=α1[1/Ait−1]+β1[󰀑REVit/Ait−1−󰀑ARit/Ait−1]

+β2[ROAit−1]+εit

Where,

TCAit=totalcurrentaccrualsforfirmiinyeart,Ait−1=totalassetsforfirmiinyeart−1,

󰀑REVit=changeinnetrevenuesforfirmiinyeart,󰀑ARit=changeinaccountsreceivableforfirmiinyeart

ROAit−1=operatingincomebytotalassetsforfirmiinyeart−1,εit=errortermforfirmiinyeart.

Thecurrentdiscretionaryaccruals(CDA)arecalculatedbyusingtheestimatedparametersasfollows:

CDAit=CAit/Ait−1−{a1[1/Ait−1]+b1[󰀑REVit/Ait−1−󰀑ARit/Ait−1]

+b2[ROAit−1]}

whereCDAitrepresentsthediscretionarycurrentaccrualsforfirmiintheeventyeart.3.4.Sampleselection

ThestudyisbasedonearningsforecastsofTaiwanindustrialIPOfirmsfrom1994through2001withfiscalyearendingDecember31.15ThesefirmsareidentifiedfromtheTaiwanEconomicJournaldatabase(TEJ).TheIPOfirmsbelongingtoutilities,financialinstitutionsandotherregulatedindustriesareexcludedfromthesamplebecauseoftheirspecialregulatorynature,whichcouldconfoundevaluationofmandatoryregulationondisclosureofearningsforecasts.Weidentified253firmsthatissued75916earningsforecastsduringtheIPOyear(t-periodforecasts)andtwoyearssubsequenttotheIPOyear(t+1andt+2periodsforecasts).Outof759forecasts,328havebeenrevisedand431havenotbeenrevised(seeTable5forrevisiondetails).Thoughsomeinitialforecastsarerevisedmorethanonceduring

Weexcludetheperiodfrom1991through1993fromtheanalysestofocusonasingleforecasterrorthresholdthroughoutthestudyperiod.Theoriginalforecasterrorthresholdwas10%,whichwasrevisedto20%in1994.Thetotalnumberofobservationsforthesamplefirmsshould759(3×253).Becausedatawerenotavailableforsomeobservationsfort+2periodforecasts,thenumberofobservationshasbeenreducedto677.

Springer

1615

(10)

(11)

286

Table1DistributionofsamplebyyearandindustryPanelA:Distributionofforecastsbyyear

t-periodforecasts

t+1-periodforecasts

RevQuantFinanAcc(2006)26:275–299

T+2-periodforecasts

#ofTotalNon-revisedRevisedNon-revisedRevisedNon-revisedRevisedYearfirmsforecastTotalforecastforecastTotalforecastforecastTotalforecastforecast199422199539199628199719199820199928200055200142Total2536611784576084165126759223928192028554225312179810122613107102219111016292914622392819202855422538161481010172010314231411101838221502239281920285542253111412881129251181125161112172617135PanelB:NumberofforecastsbyindustryIndustryCode111213141516171820212223and2425262729

IndustriesCementFoodsPlasticsTextiles

Electric&machineryElectricAppliance&CableChemicalsGlass&CeramicsSteel&IronRubberElectronicsConstructionTransportationTourism

Wholesale&RetailOthers

Total

Numberoffirms154152041121011321971318253Numberofforecasts3151245601233630339657213954759thefiscalyearinwhichtheywereissued,weuselastrevisioninouranalysis.Ananalysisofrevisionsindicatesthatmostrevisionsareclusteredaroundthetimewhenthesemi-annualreportsareissued.DistributionofsampleobservationsbyyearandindustryisprovidedinTable1.

Distributionofforecastsoveryearsindicatesthat,withtheexceptionofyear2000and2001,thereisnosignificantdifferenceinthenumberofforecastsissuedoverdifferentyears.Thenumberofforecastsissuedinyear2000and2001ishighprobablybecausemoreIPOswereissuedbyelectronicfirmsduringthisperiod.

Springer

RevQuantFinanAcc(2006)26:275–299287

Distributionofsampleforecastsbyindustryindicatesthatalargenumberofindustrygroupsarecoveredinthestudy,butthemainconcentrationappearstobeontheelectronicgroup,followedbyelectric/machineryandconstruction.

4.Results

4.1.Optimisticversusconservativeforecasts

WecalculatetheforecasterrorFE1usingequation(3),whichreflectsthedifferencebetweeninitialearningsforecastsandpre-managedactualearnings(earningsbeforeadjustments)andthisforecasterroralsoindicateswhethertheinitialforecastisoptimisticorconservative.Apositiveforecastwillsignifythattheinitialforecastisoptimistic,i.e.thepredictedearningsarehigherthanthepre-managedearnings,whereasanegativeforecasterrorwillsignifythattheearningsforecastisconservative,i.e.thepredictedearningsarelowerthanthepre-managedearnings.ThefrequencyandmagnitudeofforecasterrorFE1areprovidedinTable2.

Theresultsonthefrequencyofoptimisticandconservativeforecasts(PanelA,Table2)showthat66.93%ofthetotalforecastsareoptimistic(FE1>0),whereas33.07%areconser-vative(FE1<0).Wetestwhetherthedifferenceinthefrequencydistributionofoptimisticandconservativeforecastsisstatisticallysignificantusingbinomialtestsforthetotalsam-pleaswellasindividualyears.Weconductbinomialtestonthehypothesisthatoptimisticandconservativeforecastsareequallydistributed.Theresultsshowthatthefrequencyofoptimisticforecastsissignificantlyhigherthantheprobabilityof50%forthetotalsampleaswellasforindividualyears.TheseresultsthusprovidesupporttoourhypothesisH1thatthenumberofoptimisticforecastsishigherthanthatofconservativeforecasts,suggestingthatmorefirmsissuedoptimisticforecaststhanconservativeforecastsaftertheregulationondisclosureofearningsforecastswasimposed.

Wealsoevaluateforecasterrorsofoptimisticandconservativeforecastsbyconductingt-testonthemeanvaluesofforecastsforthetwogroupsandWilcoxontestonthemedianvaluesofforecastsofthetwogroups.TheresultsarecontainedinPanelBoftheTable2.Theresultsshowthatthemeansaswellasmediansofforecasterrorsofoptimisticforecastsaresignificantlyhighercomparedtothoseofconservativeforecastsforthetotalsampleaswellasforthesub-samplesforallperiods.Thedifferenceinthemeansaswellasmediansisstatisticallysignificantforthetotalsampleaswellasfortandt+1forecasts,whereasthedifferenceinthemeanvaluesoft+2forecastsoftwogroupsisstatisticallyinsignificant.ThesefindingsindicatethattheforecasterrorsforoptimisticforecastsarehigherthantheforecasterrorofconservativeforecastsandprovideadditionalsupporttohypothesisH1.4.2.Operatingperformanceandoptimisticforecasts

Weusethechangeinreturnofassets(ROA)intheforecastyearfromtheprecedingyearasaproxyfortheexpectationontheoperatingperformanceforevaluatingtheassociationbetweenissuanceofoptimisticforecastsandexpectationofbettereconomicperformanceduringtheforecastyear.First,wecomparethefrequencydistributionofoptimisticforecastsinthegroupsofobservationswithhighandlowROAduringtheforecastyearcomparedtothepreviousyear,i.e.positiveandnegativechangeinROAintheforecastyear.TheresultsshowthatthefrequencyofoptimisticforecastsissignificantlyhigherinthegroupofobservationswithpositivechangeinROAintheforecastyear.ThefrequencyofoptimisticforecastsinthegroupofpositivechangeinROAintheforecastyearis57.14%comparedto42.86%for

Springer

Table2Frequencydistributionandforecasterrorsofoptimisticandconservativeforecasts288

Springer

Conservativeforecasts(FE1a<0)Percentage67.98b***81789225169.17b***63.64b***66.93b***NPercentage32.0230.8336.3633.07Conservativeforecasts(FE1a<0)Median1.17720.77801.04930.9428251−1.150392−1.357278−0.912681−1.1444NMeanMedian−0.4355−0.6113−0.8007−0.6302t-valuefordifferenceinabsolutemeanvaluesc2.41∗∗1.32∗1.60∗2.84∗∗∗Z-statisticsfordifferenceinabsolutemedianvaluesd4.1061∗∗∗1.8275∗∗1.02603.9651∗∗∗PanelA:FrequencyofoptimisticandconservativeforecastsOptimisticforecasts(FE1a>0)PeriodNt(IPOyear)172t+1175t+2161Total508PanelB:MeanandmedianforecasterrorsofoptimisticandconservativeforecastOptimisticforecasts(FE1a>0)PeriodNMeant(IPOyear)1721.7594t+11751.0048t+21611.7374Total5081.4922RevQuantFinanAcc(2006)26:275–299

∗∗∗1%significancelevel,∗∗5%significancelevel,∗10%significancelevel,onbasisofone-tailedtests.aForecasterrorFE1iscalculatedasthedifferencebetweeninitialforecastedearnings(PE)andpre-managedearnings(PME)deflatedbytheabsolutevalueofinitialforecastedINearnings(|PEIN|),(i.e.(PEIN-PME)/|PEIN|).Pre-managedearnings(PME)arecalculatedbydeductingtotaldiscretionaryaccruals(TDA)fromthereportedearnings(RE).Totaldiscretionaryaccruals(TDA)arederivedusingJones’modifiedcross-sectionalmodelcontrolledforperformance(seeequation(6),(7)and(8)fordetails).bThe50%asabenchmarkisusedtotestthetendencyofoptimisticandconservativeearningsforecast.cThet-testiscomputedonthedifferenceintheabsolutevalue(magnitude)ofmeanoptimisticandconservativeforecasts.dWilcoxonZ-statisticsiscomputedbasedonthedifferenceintheabsolutevalue(magnitude)ofmedianoptimisticandconservativeforecasts.RevQuantFinanAcc(2006)26:275–299289

thegroupwithnegativechangeintheROA.Thedifferenceisstatisticallysignificantatthe1%level.

Second,weevaluatetheassociationbetweenoptimistic/conservativeforecastsandchangeinROAbyconductingLogitregressionandcontrollingfortheeffectofothervariablesthatmayinfluencethechoiceofoptimisticandconservativeforecasts.TheLogittestisconductedbycodingtheforecastswithpositiveforecasterrors(optimisticforecasts)as1,andtheforecastswithnegativeforecasterrors(conservativeforecasts)as0.ThefollowingLogitmodelisused:

Prob(OptimisticForecast=1)=F(β0+β1CHGROAit+β2SIZEit

+β3HORIZONit+β4BETAit+β5MBit−1+εit)

where:

F(β’X)=eβ’X/(1+eβ’X),

CHGROAit=Changeinreturnsonassets(ROA)inyeartandt−1.SIZEit=logarithmoftotalassetsinyeart.

HORIZONit=Numberofmonthsbetweeninitialmanagementearningsforecastsan-nouncementdateandearningsannouncementdate.BETAit=Systematicriskforfiscalyear-end.

MBit−1=Market-to-bookratioatthebeginningofthefiscalyear.

TheROAiscalculatedbydividingtheoperatingincomebytotalassets.TheCHGROAitisachangeinreturnsonassets(ROA)inyeartandt−1.17Theuseofcontrolvariablesisconsistentwithearlierstudies.Someearlierstudiesindicatethatforecastoptimismmaybeinfluencedbythefirmsize(e.g.Atiase,1985;Becker,etal.,1998).ChoiandZiebart(2000)showthatoptimisticbiasincreasesastheforecasthorizon(HORIZON)increases.Thesystematicrisk(BETA)isincludedtocapturetheeffectofriskonforecasterrors(e.g.Imhoff,1978),andmarket-to-bookratioisaddedtocontrolforthepotentialgrowtheffect(ChoiandZiebart,2000).TheresultsarecontainedinTable3(PanelA).

TheLogitregressionresults(PanelA)showthatthecoefficientforthechangeinROAispositiveandisstatisticallysignificantatthe.001level.TheresultsbasedonthechangeinROAmeasuredbythepre-managedearnings(PME),i.e.PMEtoperiodtminusPMEforperiodt−1,alsoshowsignificantlypositivelyassociation.TheseresultsthussupportourhypothesisH2thatoptimisticforecastsareespeciallyissuedwhenthefutureoperatingperformance,proxiedbyROA,isexpectedtobehigh.

Inordertoevaluatewhetherthemagnitudeofoptimisticandconservativeforecastsisassociatedwiththemagnitudeofexpectedimprovementintheperformance,weconductanOLSregressiontestonthecontinuousvariableofforecasterrors,wherethepositiveforecasterrorsreflectoptimisticforecastsandnegativeforecasterrorsreflectconservativeforecasts.Thefollowingregressionmodelisused.

FE1it=β0+β1CHGROAit+β2SIZEit+β3HORIZONit

+β4BETAit+β5MBit−1+εit

17

(12)

(13)

WealsoconductatestonthechangeinROAbasedonthepre-managedearnings,whicharecalculatedasreportedearningsadjustedbydiscretionaryaccruals.

Springer

290RevQuantFinanAcc(2006)26:275–299

Table3Associationbetweenoptimistic/conservativeforecastsandfirmperformanceforthetotalsamplePanelA:LogitRegressionAnalysis

Prob(OptimisticForecast=1)=F(β0+β1CHGROAit+β2SIZEit+β3HORIZONit

+β4BETAit+β5MBit−1+εit)Variables

InterceptCHGROAitbSIZEitc

HORIZONitdBETAiteMBit−1fChi-squarePseudo-R2hCorrectrate

No.ofObservationsOptimisticforecastsConservativeforecastsN

Coefficient−5.270340.57000.29700.19180.4874−0.0035

Chi-square7.620399.8251∗∗∗5.4789∗∗57.2664∗∗∗4.4660∗∗2.0406

Marginaleffectg

266.63∗∗∗0.41780.9%508251759

7.99020.05840.03770.0960−0.0007

PanelB:OLSRegressionAnalysis

FE1ita=β0+β1CHGROAit+β2SIZEit+β3HORIZONit+β4BETAit+β5MBit−1+εitVariablesInterceptCHGROAitbSIZEitc

HORIZONitdBETAiteMBit−1fAdj.RF-statisticsp-valueN

∗∗∗1%

Coefficient−3.464123.17860.18120.1032−0.3590−0.0007

t-value−1.216.37∗∗∗∗0.962.79∗∗∗−1.07−0.61

0.063710.77<.0001759

significancelevel,∗∗5%significancelevel,∗10%significancelevel,onbasisofone-tailedtests.aFE1=Earningsforecasterrors;ForecasterrorFE1iscalculatedasthedifferencebetweeninitialforecastedearnings(PEIN)andpre-managedearnings(PME)deflatedbytheabsolutevalueofinitialforecastedearnings(|PEIN|),(i.e.(PEIN-PME)/|PEIN|).Pre-managedearnings(PME)arecalculatedbydeductingtotaldiscretionaryaccruals(TDA)fromthereportedearnings(RE).Totaldiscretionaryaccruals(TDA)arederivedusingJones’modifiedcross-sectionalmodelcontrolledforperformance(seeequation(6),(7)and(8)fordetails).

bCHGROA=Changeinreturnsonassetsinyeartandt−1;returnsonassets=operatingincome

it

dividedbytotalassets.

cSIZE=Logarithmtotalassetsinyeart

it

dHORIZON=Numberofmonthsbetweeninitialmanagementearningsforecastsannouncementdate

it

andearningsannouncementdate.eBetait=Systematicriskforfiscalyear-end.fMB

it−1=fiscalyear’sbeginningmarket-to-bookratio.

gThemarginaleffectarecomputeaseβ’X/(1+eβ’X)2,whereβ’XiscomputedatthemeanvaluesofX.hPseudo-R2isMcFadden’smeasureofgoodnessoffit.Springer

RevQuantFinanAcc(2006)26:275–299291

where,

FE1it=ForecastErrorbasedonunadjusted(pre-managed)earningsforfirmiandperiodt.

TheOLSregressionresultsarecontainedinPanelBofTable3.TheresultsshowthatthecoefficientforthechangeinROAissignificantlypositive,18suggestingthatthereisapositiveassociationbetweentheexpectationonthemagnitudeofimprovementintheperformanceandthemagnitudeofoptimisticforecasts.TheseresultsprovideadditionalsupporttoourhypothesisH2.Theresultsalsoshowthatthereisapositiveassociationbetweenoptimisticforecastsandforecasthorizon(HORIZON),indicatingthatforecastswithlongerforecasthorizonaremoreoptimistic.Thisresultisconsistentwiththeresultsofpriorstudies(e.g.ChoiandZiebart,2000;Baginski,1997)

Weconductedanadditionaltestontheassociationbetweenoptimistic/conservativefore-castsandchangeinROAontheIPO-yearforecastsonly(t-periodforecasts).TheresultsfortheIPOyeararesimilartothoseforthetotalsample.

RecognizingthattheassociationbetweenthechangeinROAandchoiceofoptimisticforecastsmaybesubjecttotwodifferentinterpretations,wewouldliketoaddcautioustoourresults.Asarguedinthispaper,apositivechangeinROAfortheforecastyear,whichreflectsmanagement’sexpectationofbetteroperatingperformanceintheforecastyear,isinterpretedtosuggestthattheexpectationofbetterperformanceresultsinissuanceofoptimisticforecasts.Ontheotherhand,itcouldbearguedthatahigherROAduringtheforecastmaybetheresultofmanipulationofreportedearnings,whichwouldbethecaseofself-fulfillingprophecyonachievingahigheroperatingperformanceduringtheforecastyear.Therefore,wecautiouslyinterpretourfindingstosupportourhypothesisH2thattheoptimisticforecastsareassociatedwiththeexpectationofbetteroperatingperformanceduringtheforecastyear.4.3.Reductionofforecasterrors

4.3.1.Discretionaryaccrualsandforecasterror

Becauseitisdifficulttomakedirectevaluationofmanipulationofreportedearningsbymanagementtoreducetheforecasterror,weevaluateearningsmanipulationbyevaluatingdiscretionaryaccruals,aproxyforearningsmanagement.Theuseofdiscretionaryaccrualsasaproxyforearningsmanagementisconsistentwithseveralearlierstudiesonearningsmanagement.WecalculateTDAandCDAusingequations(8)and(11)respectively,butwetabulatetheresultsbasedonCDAbecausecurrentdiscretionaryaccrualsareconsideredtobemoresusceptibletoearningsmanagementintheshortrunthantotaldiscretionaryaccruals(e.g.Ashbaugh,LafondandMayhew,2003;Becker,etal.,1998;Frankel,JohnsonandNelson,2002.)TheTDAresults,whicharenottabulated,are,however,similartotheCDAresults.Thefrequencyoffirmsusingpositiveandnegativecurrentdiscretionaryaccruals(CDA)separatelyforoptimisticandconservativeforecastsisprovidedinTable4.

White’s(1980)methodisusedtoevaluatethesignificanceofcoefficients.Allvarianceinflationfactors(VIF)arelessthan10,indicatingthattheregressionresultsarenotaffectedbymulticollinearity(Kennedy,1992).DWstatisticsshowthatautocorrelationisinsignificant.Theresultsremainunchangedwhenthecoefficientsarere-estimatedusingBelsley,Kuh,andWelsch(1980)model.Weeliminateobservationsfromtheanalysesif(1)theabsolutevalueofstudentizedresidualsisgreaterthan2,(2)studentizedresidualwithahatmatrixvalueisgreaterthan2*P/N(pisthenumberofparametersandnisnumberofobservations),or(3)aDFFITSstatisticisgreaterthan2(P/N)1/2.

Springer

18

292RevQuantFinanAcc(2006)26:275–299

Table4Currentdiscretionaryaccrualsbyoptimisticandconservativeforecasts

PanelA:Frequencyofpositiveandnegativetotaldiscretionaryaccrualsforoptimisticforecasts

Optimisticforecasts(FE1a>0)Positivecurrent

Discretionaryaccrualsb

Periodt(IPOyear)T+1T+2Total

TotalN172175161508

N148148133429

%86.0584.5782.6184.45

Negativecurrent

DiscretionaryaccrualsbN24272879

%13.9515.4317.3915.55

Z-statisticsfor%difference9.455∗∗∗9.146∗∗∗8.275∗∗∗15.529∗∗∗

PanelB:Frequencyofpositiveandnegativetotaldiscretionaryaccrualsforconservativeforecasts

Conservativeforecasts(FE1a<0)Positivecurrent

Discretionaryaccrualsb

Periodt(IPOyear)T+1T+2Total

aForecast

Negativecurrent

DiscretionaryaccrualsbN

TotalN817892251

N1110526

13.5812.825.4310.36

Z-statisticsfor%difference−6.555∗∗∗−6.567∗∗∗−8.550∗∗∗−12.560∗∗∗

706887225

86.4287.1894.5789.34

∗∗∗1%significancelevel,∗∗5%significancelevel,∗10%significancelevel,onbasisofone-tailedtests.

errorasFE1iscalculatedasthedifferencebetweeninitialforecastedearnings(PEIN)and

pre-managedearnings(PME)deflatedbytheabsolutevalueofinitialforecastedearnings(|PEIN|).Pre-managedearnings(PME)arecalculatedbydeductingtotaldiscretionaryaccruals(TDA)fromthereportedearnings(RE).

bInordertoobtaincurrentdiscretionaryaccruals(CDA),wefirstcalculatetotalcurrentaccruals(TCA)asfollowing:TCAit=(󰀑CAit−󰀑CASHit)−(󰀑CLit−󰀑STDit).TCAit=totalcurrentaccrualsforfirmiinyeart,CAit=currentassetsforfirmiinyeart,CASHit=cashforfirmiinyeart,(󰀑CAit−󰀑CASHit)=changeindifferenceofcurrentsandcashforfirmiinyeart,and󰀑CLit=changeincurrentliabilitiesforfirmiinyeart.󰀑STDit=changeintheportionoflong-termliabilitiesduewithinayear.Thenon-discretionarycurrentdiscretionaryaccrualsarederivedusingJones’modifiedcross-sectionalmodelcontrolledforperformance(seeequation(10)fordetails).Finally,currentdiscretionaryaccruals(CDA)arecalculatedbyusingtheestimatedparametersa1andb1intheequation(10)asfollow-ing:CDAit=TCAit/Ait−1−{a1[1/Ait−1]+b1[󰀑REVit/Ait−1−󰀑ARit/Ait−1]+b2[ROAit−1],whereCDAitrepresentsthediscretionarycurrentaccrualsforfirmiintheeventyeart.

TheresultscontainedinPanelAoftheTable4showthatonanoverallbasis,84.45%ofoptimisticforecastsareassociatedwithpositivecurrentdiscretionaryandonly15.55%withnegativecurrentdiscretionaryaccruals.TheresultsfortheIPOyear,t+1,andt+2yearsarealmostsimilar.TheresultscontainedinPanelBofthetableshowthat89.34%oftheconservativeforecastsareassociatedwiththeuseofnegativediscretionaryaccrualswhereas17.8%withpositivediscretionaryaccruals.Theseresultsthusindicatethatpositivediscretionaryaccrualsareextensivelyusedforreducingtheforecasterrorofoptimisticfore-casts.Negativediscretionaryaccrualsareprimarilyusedforreducingtheforecasterrorofconservativeforecaststomeettheforecasterrorthreshold.

Springer

RevQuantFinanAcc(2006)26:275–299Table5FrequencyofforecastrevisionsPanelA:Frequencyofforecastrevisions

Totalinitialforecasts253253253759

Numberofinitialforecastsnotrevised107103118328

Numberofinitialforecastsrevised146150135431

NumberofrevisionsofinitialforecastsRevisedonce112115111338

Revisedtwice30302383

Revisedthree3418

Revisedforecast1102

293

Periodt(IPOyear)t+1t+2Total

PanelB:Upwardanddownwardrevisions

Numberofinitialforecastsrevised146150135431

Numberofearningsforecasts,includingrevisions185191160536

UpwardrevisionsN765052178

Percentage14.189.339.7033.21

DownwardrevisionsN109141108358

Percentage20.3326.3120.1566.79

Periodt(IPOyear)t+1t+2Total

PanelC:Upwardanddownwardrevisionsbyoptimisticandconservativeforecasts

Totalrevisedforecasts

Optimistic(FE1a>0)Conservative(FE1a<0)

∗∗∗1%aForecast

UpwardrevisionsN52103

Percentage18.64%67.76

DownwardrevisionsN22749

Percentage81.36%32.24%

Z-statistcsfordifferenceinupwardand

downwardrevisions−10.476∗∗∗4.379∗∗∗

279152

significancelevel,∗∗5%significancelevel,∗10%significancelevel,onbasisofone-tailedtests.errorasFE1iscalculatedasthedifferencebetweeninitialforecastedearnings(PEIN)andpre-managedearnings(PME)deflatedbytheabsolutevalueofinitialforecastedearnings(|PEIN|).Pre-managedearnings(PME)arecalculatedbydeductingtotaldiscretionaryaccruals(TDA)fromthereportedearnings(RE).Totaldiscretionaryaccruals(TDA)arederivedusingJones’modifiedcross-sectionalmodel(seeequations(6),(7)and(8)fordetails).

4.3.2.ForecasterrorreductionthroughdiscretionaryaccrualsversusforecastrevisionsTheforecasterrorcanbereducedeitherbyrevisingearningsforecastsorbyadjustingthereportedearnings.Weanalyzeastohowmanyforecastsarebroughtwithintheforecasterrorthresholdthroughforecastrevisionsandbyearningsmanagement.Beforethiscomparativeanalysisisconducted,weexaminethefrequencyofforecastrevisionsofoptimisticandconservativeforecasts.TheresultsarecontainedinTable5:

Outof759totalinitialforecasts,only431arerevised,andtheremaining328forecastsarenotrevised.Outof431revisedforecasts,338forecastsarerevisedonlyonceinthefiscalyear,83arerevisedtwice,whereas8and2arerevisedthreeandfourtimesrespec-tively.Thefrequencydistributionforupwardanddownwardrevisions(PanelB)showsthat66.79%ofrevisionsaredownward.Theassociationofupward/downwardrevisionswith

Springer

294RevQuantFinanAcc(2006)26:275–299

Table6Forecasterrorreductionthroughforecastrevisionsandadjustmentofreportedearningswithdiscretionaryaccruals

Numberofforecastsbrought

intothresholdusing

Numberofdiscretionaryforecastsaccrualsbroughtinto(i.e.FE3c=

(PELR−threshold

RE)/using

|PELR|)revisions

(5)=(2)−(3)(6)=(3)−(4)612283

308178486

Upperand

lowerboundsofforecasterrorsthreshold(1)Upperbound(FEd>20%)Lowerbound(FEd<−20%)Total

∗∗∗1%aForecast

Forecasterrorsbasedoninitialforecastandpre-managedearnings(i.e.FE1a=(PEIN−

PME)/|PEIN|)(2)428220648

Forecasterrorsbasedonlastrevisionsofforecastandpre-managedearnings(i.e.FE2b=(PELR−

PME)/|PEIN|)(3)e367198565

Numberoffirms

outsidethethresholdattheendoffiscalyear(i.e.FE3c=(PELR−RE)/|PELR|)(4)592079

significancelevel,∗∗5%significancelevel,∗10%significancelevel,one-tailedtests.

errorasFE1iscalculatedasthedifferencebetweeninitialforecastedearnings(PEIN)andpre-managedearnings(PME)deflatedbytheabsolutevalueofinitialforecastedearnings(|PEIN|).Pre-managedearnings(PME)arecalculatedbydeductingtotaldiscretionaryaccruals(TDA)fromthereportedearnings(RE).Totaldiscretionaryaccruals(TDA)arederivedusingJones’modifiedcross-sectionalmodel(seeequations(6),(7)and(8)fordetails).

bForecasterrorasFE2iscalculatedasthedifferencebetweenlastforecastedearnings(PE)andpre-LR

managedearnings(PME)deflatedbytheabsolutevalueoflastforecastedearnings(|PEIN|).

cForecasterrorasFE3iscalculatedasthedifferencebetweenlastforecastedearnings(PE)andreported

LR

earnings(RE)deflatedbytheabsolutevalueoflastforecastedearnings(|PELR|).

dFEisaproxyofforecastserrorasFE1,FE2,andFE3inthePanelA’scolumnof(2),(3)and(4)respectively.eThecolumn(3)ofPanelA,andthecolumn(2)ofPanelBareinsertedtomeasuretheextentofearningsrevisionsandearningsmanagement.

optimistic/conservativeforecasts(PanelC)showsthattheoptimisticforecastsareprimarilyreviseddownward(81.36%).Theresultsontheconservativeforecastsshowthat67.76%arerevisedupwardand32.24%downward.Therevisionanalysisthusindicatesthatopti-misticforecastsaremostlyreviseddownwardtoreducetheforecasterror.Ontheotherhand,thoughasignificantnumberofconservativeforecastsarerevisedupward,someforecastsarealsoreviseddownward.Downwardadjustmentofconservativeforecastssuggeststhatmanagersmayusedownwardrevisionstoreducemarketexpectationsandatthesametimeusediscretionaryaccrualstoreducetheforecasterrorformeetingtheforecasterrorthreshold.Weconductthecomparativeanalysisofforecasterrorreductionthroughearningsmanage-mentandforecastrevisionsbycalculatingtheforecasterrorsFE2andFE3usingequations(4)&(5)respectively.TheforecasterrorFE2enablesustoidentifythenumberofforecastsoutsidetheforecasterrorthresholdaftertheforecastsarerevised,andFE3enablesustoeval-uatethenumberofforecastsoutsidetheforecasterrorthresholdaftertheforecastsarerevisedandearningsareadjustedwithdiscretionaryaccruals.TheforecasterrorFE2iscalculatedonthebasisoflastrevisedforecastandpre-managedearnings,whereastheforecasterrorFE3iscalculatedonthebasisoflastrevisedforecastandreportedearnings,i.e.earningsafterearningsmanagement.TheresultsarepresentedinTable6.

Springer

RevQuantFinanAcc(2006)26:275–299295

Fig.1DistributionofforecaseerrorsofIPOs

TheresultsbasedonFE1(column2)ofPanelAshowthat648initialforecastsareoutsidetheforecasterrorthreshold,428forecastsareoutsidetheupperboundand220outsidethelowerbound.TheresultsbasedonFE2(column3)indicatethat565forecastsareoutsidetheforecasterrorthresholdaftertheforecastsarerevised(367forecastsareoutsidetheupperboundand198outsidethelowerbound).Theseresultsthussuggestthatonly83forecastsarebroughtwithintheforecastthresholdbyonlyrevisingtheforecasts,61withintheupperboundoftheforecasterrorand22withinthelowerbound.

Theresultscontainedincolumn4arebasedonFE3,i.e.afterforecastsarerevisedandearningsareadjustedwithdiscretionaryaccruals.Thenumberofforecastsoutsidetheforecasterrorthresholdis79,indicatingthatasignificantnumberoffirmsreducedtheirforecasterrorthroughearningsmanagementandforecastrevisions.

ReductionofforecasterrorsthroughearningsmanagementandforecastrevisionsisalsopresentedgraphicallyinFigure1.

Springer

296RevQuantFinanAcc(2006)26:275–299

PanelAbasedonforecasterrorFE1(i.e.beforetheforecastsarerevisedandactualearningsareadjusted)showsthattheforecasterrorsarewidelydispersed,andalargenumberofforecasterrorsareoutsidetheforecasterrorthresholdof20%.PanelBcontainsthefrequencydistributionofforecasterrorsaftertheforecastsarerevised.Theresultsshowthatthereisaslightimprovementinthedispersionofforecasterrors,buttheimprovementisinsignificant.Thispanelthussuggeststhatthenumberofforecastsbroughtwithintheforecastthresholdthroughforecastrevisionsisinsignificant.InPanelC,wepresentthefrequencydistributionofforecasterrorsafteractualearningsareadjustedwithdiscretionaryaccruals.Theresultsshowthatasignificantnumberofforecasterrorshavebeenreducedandtheforecastsarebroughtwithinthethresholdsof20%.PanelDpresentsthefrequencyofforecasterrorsbasedontherevisedforecastsandreportedearnings.Theresultsshowthattheforecasterrorsarenowconcentratedwithintheforecasterrorthreshold.

WeconductatestsuggestedbyBurgstahlerandDichev(1997)toevaluateinequalitiesatdifferentintervals.Theresults(untabulated)showthattheinequalitiesarehigheratthe20%thresholdcomparedtotheadjacentbands,andthatthereisdiscontinuityindistributionatthethresholdpoint.Theactualfrequencyattheuppertwodecilesband,i.e.from0.18to.20isrelativelyhigherthanitsexpectedfrequency,andtheresultsarestatisticallysignifi-cant(t-value=4.899,prob.>.01).Theseresultsthusindicatethattheforecastserrorsareconcentratedattheupperboundof20%threshold.

TheresultscontainedinTable6andFigure1thusshowthattheforecasterrorisprimarilyreducedthroughtheuseofdiscretionaryaccrualsandpositivediscretionaryaccrualsareusedtoadjustthereportedearningsupwardforoptimisticforecasts.Inordertohaveabetterinsightintocomparativeanalysisofearningsmanagementbyoptimisticandconservativeforecasts,wecomparetheuseofdiscretionaryaccrualsbetweenthetwogroupstomeettheforecasterrorthreshold.Theresultsindicatethatthemeansofcurrent(total)discretionaryaccrualsfortheoptimisticsub-sampleare0.096(0.089)comparedtothemeansof−0.073(−0.078)fortheconservativesub-sample.Thet-testresultsbasedontheabsolutevalueofdiscretionaryaccrualsindicatethatthedifferencebetweenthemeansofthesub-samplesisstatisticallysignificantatthe1%(10%)level.

TheaboveresultsthusprovidesupporttoourhypothesisH3thattheforecasterrorismorereducedbyadjustingthereportedearningswithdiscretionaryaccrualsthanbyrevisingtheearningsforecasts.Theuseofdiscretionaryaccrualsissignificantlyhigherforoptimisticforecaststhanconservativeforecasts.

4.4.AdditionalTest

Weconducttwoadditionalteststoevaluatetheimpactofregulationonearningsmanagement.First,wecomparediscretionaryaccrualsoftheIPOsissuedaftertheregulationwiththatoftheIPOsissuedbeforetheregulationwasimposed.Theresultsshowthatdiscretionaryaccrualsfortheperiodpriortotheregulatoryperiodarenotsignificantlyhigherthan0,whereasdiscretionaryaccrualsofIPOsfortheregulatoryperiodaresignificantlyhigherthan0.TheseresultsthusprovideadditionalsupporttotheexpectationthatearningsmanagementbyIPOfirmsincreasedconsiderablyaftertheTSFECregulationondisclosureofearningsmanagementwasimposed.

Second,wecomparediscretionaryaccrualsofthesamplefirmsfortheregulationperiodwiththoseoftheperiodaftertwoyearsofissuingIPOs(i.e.post-regulationperiod).Theresultsshowthatasignificantnumberofthesesamplefirmsareassociatedwithnegativediscretionaryaccrualsduringthepost-regulationperiod.Theseresultsthusindicatethat

Springer

RevQuantFinanAcc(2006)26:275–299297

positivediscretionaryaccrualsthatwereusedduringtheregulationperiodforreducingtheforecasterrorsrequiredreversalaftertheregulationperiod.

5.Conclusion

ThestudyhasevaluatedwhethertheTSFECregulationondisclosureofearningsforecastsresultedindisclosureofmoreoptimisticearningsbyIPOfirms,andwhethertherewasahigheruseofdiscretionaryaccruals,especiallybyoptimisticforecasts,toreducetheforecasterrorthanbyrevisingtheearningsforecasts.Thestudyalsospecificallyevaluatedwhetherthedisclosureofoptimisticforecastswasassociatedwiththemanagers’expectationonthefirms’operatingperformanceduringtheforecastyear.

TheresultsshowthattheIPOforecastsaswellasforecastsfortwomandatoryyearsafterissuanceofIPOsweremoreoptimisticthanconservative.TheIPOfirmsespeciallyissuedoptimisticforecastswhentheROAintheforecasterrorwasexpectedtobehigherthantheROAoftheprecedingyear.Additionally,theresultsshowthattheforecastserrorswerereducedmorebyusingdiscretionaryaccruals,especiallyforoptimisticforecasts,thanrevisingtheearningsforecasts.AcomparisonofearningsmanagementbythesampleIPOfirmswithearningsmanagementbyIPOfirmspriortotheregulationindicatesthattheearningsmanagementincreasedconsiderablyaftertheregulation.

Thefindingsofthisstudyindicatethattheregulationondisclosureofearningsforecastshasresultedinprovidingmoreoptimisticforecastsforsendingpositivesignalstothemarket,andthereportedearningsweremanagedusingdiscretionaryaccrualsforreducingtheforecasterrorsofoptimisticforecasts.Thesefindingsthussuggestthatmandatoryearningsforecastregulationdidnotresultinissuingmorerealisticearningsforecasts.Instead,itencouragedmanipulationofreportedearnings,whichreducedthequalityofreportedearnings.Thefindingsofthisstudythussuggestthatregulationondisclosureofearningsforecastsisnotlikelytoimprovethequalityandusefulnessofreportedearnings.

BecausetheTSFEChasnotmadeanyinformationavailableonactionstakenbyitagainsttheforecasterrorviolatingfirms,ithasnotbeenpossibletoincludethepenaltyaspectintheanalysisonthechoiceofforecastrevisionsanduseofdiscretionaryaccrualstoreducetheforecasterror.Asdiscussedearlier,theTSFEChasallowedtheviolatingfirmstoappealiftheycouldjustifytheirviolation,whichmaybeduetoexchangeratechange,unexpectedimpactofnon-economicfactorssuchaslaborstrike,etc.Additionalresearch,ifinformationonpenaltyfortheviolatingfirmsbecomesavailable,willprovideabetterinsightintotheimpactofpenaltyontheassociationbetweenforecastdisclosurerequirementandearningsmanagement.

References

Ashbaugh,H.,R.LafondandB.W.Mayhew,“DoNonauditServicesCompromiseAuditorIndependence?

FurtherEvidence.”TheAccountingReview78(3),611–639(2003).

Atiase,R.,“PredisclosureInformation,FirmCapitalization,andSecurityPriceBehaviorAroundEarnings

Announcements.”JournalofAccountingResearch23(Spring),21–36(1985).

Baginski,S.P.andJ.M.Hassell,“DeterminantsofManagementForecastPrecision.”TheAccountingReview

72(2),303–312(1997).

Baginski,S.P.,J.M.HassellandM.D.Kimbrough,“TheEffectofLegalEnvironmentonVoluntaryDisclosure:

EvidencefromManagementEarningsForecastsIssuedinU.S.andCanadianMarkets.”TheAccountingReview77(1),25–50(2002).

Springer

298RevQuantFinanAcc(2006)26:275–299

Becker,C.L.,M.L.Defond,JiambalvoandK.R.Subramanyam,“TheEffectofAuditQualityonEarning

Management.”ContemporaryAccountingResearch15(Spring)1–24(1998).

Belsley,D.,E.KuhandR.Welsch,RegressionDiagnostics,Wiley,NewYork,NY,1980.

Brown,P.,A.Clark,J.HowandK.Lim,“TheAccuracyofManagementDividendForecastsinAustralia.”

Pacific-BasinFinanceJournal.(8),309–331(2000).

Burgstahler,D.andI.Dichev.,“EarningManagementtoAvoidEarningsDecreasesandLosses.”Journalof

AccountingandEconomics24(1),99–126(1997).

Chan,A.M.Y.,C.L.K.Sit,M.M.L.Tong,D.C.K.WongandR.W.Y.Chan.,“PossibleFactorsofthe

AccuracyofProspectusEarningsForecastsinHongKong.”TheInternationalJournalofAccounting.31(3),381–398(1996).

Chen,G.andM.Firth,“TheAccuracyofProfitForecastsandtheirRolesandAssociationswithIPOFirm

Valuations.”JournalofInternationalFinancialManagementandAccounting,10(3),202–226(1999).Cheng,T.Y.andM.Firth,“AnEmpiricalAnalysisoftheBiasandRationalityofProfitForecastsPublished

inNewIssueProspectuses.”JournalofBusiness,Finance,andAccounting,27(3/4)423–446(2000).Choi,J.H.andD.A.Ziebart,“AReexaminationofBiasinManagementEarningsForecasts.”Workingpaper,

HongKongUniversityofScienceandTechnology(2000).

Clarkson,P.M.,A.Dontoh,G.RichardsonandS.E.Sefcik,“TheVoluntaryInclusionofEarningsForecasts

inIPOProspectuses.”ContemporaryAccountingResearch8(2),601–626(1992).

Dechow,P.M.,R.G.SloanandA.P.Sweeney,“DetectingEarningsManagement.”TheAccountingReview

70,193–225(1995).

DeFond,M.andJ.Jiambalvo,“DebtCovenantViolationandManipulationofAccruals,AccountingChoices

inTroubledCompanies.”JournalofAccountingandEconomics17(1),145–176(1994).

Dev,S.andM.Webb,“TheAccuracyofCompanyProfitForecasts.”JournalofBusinessFinance4(3),26–39

(1972).

DuCharme,L.L.,P.H.MalatestaandS.E.Sefcia,“EarningsManagement:IPOValuationandSubsequent

Performance.”JournalofAccounting,Auditing,andFinance16(4),369–396(2001).

Fan,P.H.andT.J.Wong,“CorporateOwnershipStructureandtheInformativenessofAccountingEarnings

inEastAsia,”JournalofAccountingandEconomics33(August),401–425(2002).

Firth,M.andA.Smith,“TheAccuracyofProfitsForecastsinInitialPublicOfferingProspectuses.”Accounting

andBusinessResearch22,239–247(1992).

Firth,M.,B.C.H.Kwok,C.K.Liau-TanandG.H.Yeo,“AccuracyofProfitForecastsContainedinIPO

Prospectuses.”AccountingandBusinessReview2(1),55–83(1995).

Frankel,R.M.,M.F.JohnsonandK.K.Nelson,“TheRelationBetweenAuditor’sFeesforNonauditServices

andEarningsManagement.”TheAccountingReview77(Supplement),71–115(2002).

Friedlan,J.M.,“AccountingChoicesofIssuersofInitialPublicOfferings.”ContemporaryAccountingRe-search11(Summer),1–31(1994).

Guay,W.R.,S.P.KothariandR.L.Watts,“AMarket-BasedEvaluationofDiscretionaryAccruals.”Journal

ofAccountingandResearch34(Supplement),83–105(1996).

How,J.andJ.Yeo,“TheImpactofForecastDisclosureandAccuracyonEquityPricing:TheIPOPerspective.”

JournalofAccounting,Auditing,andFinance16(4),401–425(2001).

Hsu,W.L.,D.HayandS.Weil,“ForecastAccuracyandBiasinIPOprospectuses:RecentNewZealand

Evidence.”PacificAccountingReview12(1),27–59(2000).

Imhoff,E.,“TheRepresentativenessofManagementEarningsForecasts.”TheAccountingReview.53(3),

836–850(1978).

Jaggi,B.andP.Lee,“EarningsManagementResponsetoDebtCovenantViolationsandDebtRestructuring.”

JournalofAccounting,Auditing,andFinance.17(4),295–324(2002).

Jaggi,B.andA.Sannella,“TheAssociationBetweentheAccuaryofManagementEarningsForecastsand

DiscretionaryAccountingChanges.”JournalofAccounting,Auditing,andFinance10(1),1–21(1995).Jaggi,B.,“AccuracyofForecastInformationDisclosedintheIPOprospectusesofHongKongCompanies.”

TheInternationalJournalofAccounting,32(3),301–319(1997).

Jelic,R.,B.SaadouniandR.Briston,“TheAccuracyofEarningsForecastsinIPOProspectusontheKuala

LumpurStockExchange.”AccountingandBusinessResearch29(1),57–72(1998).

Jog,V.J.andR.J.McConomy,“VoluntaryDisclosureofManagementEarningsForecastsinIPOProspectuses.”

JournalBusinessFinance&Accounting30(1),125–167(2003).

Jones,J.J.,“EarningsManagementDuringImportReliefInvestigations.”JournalofAccountingResearch

29(Autumn),193–228(1991).

Kasznik,R.,“OntheAssociationBetweenVoluntaryDisclosureandEarningsManagement.”Journalof

AccountingResearch37(Spring),57–81(1999).

Kothari,S.,a.LeoneandC.Wasley,“PerformanceMatchedDiscretionaryAccrualMeasure.”Journalof

Accounting&Economics39,163–197(2005).Springer

RevQuantFinanAcc(2006)26:275–299299

LaPorta,R.,F.Lopez-de-Silanes,A.ShleiferandR.Vishny,“LawandFinance.”JournalofPoliticalEconomy

106,1113–1155(1998).

Mak,Y.T.,“TheDeterminantsofAccuracyofManagementEarningsForecasts:ANewZealandstudy.”

InternationalJournalofAccounting24,267–280(1989).

Skinner,D.,“WhyFirmsVoluntarilyDiscloseBadNews?”JournalofAccountingResearch32(3),38–60

(1994).

Subramanyam,K.R.,“ThePricingofDiscretionaryAccruals.”JournalofAccountingandEconomics22,

249–282(1996).

Teoh,S.,I.WelchandT.Wong,“EarningsManagementandtheUnderperformanceofSeasonedEquity

Offerings.”JournalofFinancialEconomics50(1),63–99(1998a).

Teoh,S.,I.WelchandT.Wong,“EarningsManagementandtheLong-runMarketPerformanceofInitial

PublicOfferings.”JournalofFinance53(6),1935–1974(1998b).

Teoh,S.,I.WelchandG.R.Rao,“AreEarningsDuringInitialPublicOfferingsOpportunistic?”Reviewof

AccountingStudies3(3),175–208(1998).

Wang,Y.,P.Lee,C.ChinandG.Kleinman,“TheImpactofFinancialForecastsRegulationonIPOAnomalies:

EvidencefromTaiwan.”JournalofFinancialRegulationandCompliance.13(2),146–166(2005).

White,H.,“AHeteroscedasticity-consistentCovarianceMatrixEstimatorandDirectTestforHeteroscedas-ticity.”Econometrica.48,817–838(1980).

Springer