您的当前位置:首页正文

Ogneva, Subramanyam and Raghunandan AR 2007

2022-07-22 来源:帮我找美食网
THEACCOUNTINGREVIEWVol.82,No.52007

pp.1255–1297

InternalControlWeaknessandCostofEquity:EvidencefromSOXSection

404Disclosures

MariaOgneva

UniversityofSouthernCaliforniaK.R.Subramanyam

UniversityofSouthernCaliforniaK.Raghunandan

FloridaInternationalUniversity

ABSTRACT:Weexaminetheassociationbetweencostofequityandinternalcontrolweakness(ICW)forfirmsthatfiledfirst-timeSection404reportswiththeSEC.Usingseveralproxies,wefindhigherimpliedcostofequityassociatedwithICWfirmsthanforacontrolsampleoffirmsthatdisclosednoICW.However,thehighercostofequityassociatedwithICWdisappearsaftercontrollingforprimitivefirmcharacteristicsandforanalystforecastbias.Overall,wefindthat,onaverage,ICWsarenotdirectlyas-sociatedwithhighercostofequity.

S

I.INTRODUCTION

ection404oftheSarbanes-OxleyActof2002(SOX,U.S.HouseofRepresentatives2002)requiresthatmanagersmaintainadequateinternalcontrolsoverfinancialre-portingandprovideanauditor-attestedassessmentoftheireffectivenessintheannualreport(10-Kfiling),highlightingdiscoveredmaterialweaknesses.Section404hasbeenextremelycontroversial.CompaniesmaintainthatthehighcostsofcomplyingwithSection404arenotcommensuratewithitsperceivedbenefitsand,hence,callformodificationorrepealofthesection(e.g.,AmericanBankersAssociation[ABA]2005;FinancialExecu-tivesInstitute[FEI]2005;Microsoft2005;Powell2005).RegulatorsinsistthattheinternalcontrolrequirementsofSection404shouldresultinhigherqualityfinancialreporting,whichinturnshouldlowercostofcapital(U.S.HouseofRepresentatives2005;U.S.Senate2004;Donaldson2005b).Financialadvisorsmaintainthatinvestorsshoulddemandahigherriskpremiumforfirmswithinternalcontrolweaknessesandcreditratingagenciesnote

Theauthorsgratefullyacknowledgehelpfulcommentsandsuggestionsfromtwoanonymousreferees,DanDhaliwal(editor),SurajSrinivasan,SarahMcVay,andtheparticipantsofthe2005HaskellandWhiteCorporateGovernanceConferenceatCaliforniaStateUniversity,Fullerton,the2006MidyearFinancialAccountingandReportingSection(FARS)meetinginAtlanta,andthe2006AmericanAccountingAssociationAnnualMeetinginWashington,D.C.MariaOgnevaalsothankstheDeloitteFoundationforfinancialsupport.Editor’snote:ThispaperwasacceptedbyDanDhaliwal.

SubmittedOctober2005

AcceptedApril2007

1255

1256Ogneva,Subramanyam,andRaghunandan

thatinternalcontrolweakness(hereafter,ICW)isanimportantconsiderationintheratingprocess(Wilfert2005;Moody’sInvestorService2004;FitchRatings2005).Inlightofthisdebate,itisimportanttoinvestigatewhetherICWisindeedassociatedwithhighercostofcapital.Accordingly,ourpaperusesalargesampleofSection404disclosurestoexaminewhetherreportedICWsareassociatedwithhigherimpliedcostofequity.

ICWcanpertaintospecificaccountingissues(suchasrevenuerecognitionorinventoryaccounting)orbroadercontrolissues(suchasadequacyandtrainingofpersonnelor‘‘toneatthetop’’).Accordingly,thereareatleasttworeasonstoexpecthighercostofequityforICWfirms.First,ICWscanresultinpooraccountingquality,therebyincreasinginformationrisk.Informationriskhasbeenboththeoretically(e.g.,EasleyandO’Hara2004;Lambertetal.2007)andempirically(Francisetal.2004,2005)linkedtohighercostofequity.Second,ICWmaybesymptomaticofpoormanagementcontrolsingeneral,whichcouldincreasebusinessriskandthecostofequitytotheextentthisriskissystematic.

Westudy2,515firmsthatfiledfirst-timeSection404reportswiththeSECduringtheperiodfromNovember2004toJanuary2006.Ofthe2,515firms,346filedadverseopinions(i.e.,acknowledgedoneormoreICWs)and2,169filedcleanreports.Ourprimarytestsuseimpliedcostofequity,whichweestimatebyinvertingtherelationbetweenstockpricesandanalysts’earningsforecasts.Becauseoferrorinmeasuringimpliedcostofequity,weestimateseveralalternativeproxiesthatusedifferentunderlyingmodelsandassumptions,aswellasdifferentdata.Tocontrolformacroeconomiceffects,weemployapurecross-sectionaldesignbyestimatingcostofequityasofJune2004foralloursamplefirms.WealsoreplicateouranalysesasofJune2006toincorporateannouncementeffects,ifany,ofoursamplefirms’Section404disclosuresonourcost-of-equityestimates.Inunivariateanalysis,wefindthatmostofourimpliedcost-of-equityproxiesare,onaverage,signifi-cantlyhigherforICWfirmsthanforcontrolfirms(i.e.,thosethatfilecleanreports),duringbothJune2004andJune2006.

GeandMcVay(2005),Ashbaugh-Skaifeetal.(2007),andDoyleetal.(2007a)findthatfirmsreportingICW(bothunderSection404andundertherelatedSection302)arerelativelymoredistressedandhavelowermarketvalues,greatercomplexityofoperations,higherincidenceoforganizationalchange(suchasM&Aandrestructuringactivity),andgreateraccountingmeasurementrisk.Therefore,itispossiblethatthehighercostofequityassociatedwithICWmaybeattributabletotheseprimitivefirmcharacteristicsratherthantoICWperse.Additionally,wefindthatanalysts’forecastsaremoreoptimisticforICWfirmsthanforcontrolfirms,whichcouldgeneratearelativelystrongerupwardbiasinthecost-of-equityestimatesforICWfirms.Accordingly,weexaminedifferencesacrossICWandcontrolfirmsaftercontrollingforthesefactors.ForbothJune2004andJune2006,wefindnosignificantdifferencesbetweenICWandcontrolfirmsformostofourcost-of-equityproxiesaftercontrollingforprimitivefirmcharacteristicsandanalystforecastbias(withtheexceptionoftheValueLineexpectedreturnproxy).TheseresultsleadustoconcludethatICWsdonotappeartobedirectlyassociatedwithcostofequity.

Animportantalternativeexplanationforourresultsisthatextrememeasurementerrorinourimpliedcost-of-equityproxies(EastonandMonahan2005;Guayetal.2005)sig-nificantlyreducesthepowerofourteststodetectadifferencebetweenICWandcontrolfirms.1Accordingly,weconductthefollowingadditionalanalyses.First,wereplicateEaston

1Therearetwoargumentsagainstthelowpowerexplanation.First,wedofindadifferencebetweenICWandcontrolfirmsinimpliedcostofequity.Thesedifferencesvanishonlyaftercontrollingforfirmcharacteristicsandanalystforecastbias.Therefore,itismorelikelythatour‘‘non-result’’arisesfromcorrelatedomittedvariablebiasratherthanlowpower.Second,weconductadditionalanalysesthatsuggestthatourtestsarepowerfulenoughtodetectevenreasonablysmalldifferencesincostofequity,suchasadifferenceof50basispoints.

TheAccountingReview,October2007

InternalControlWeaknessandCostofEquity:EvidencefromSOXSection404Disclosures1257

andMonahan’s(2005)constructvaliditytestsforourcost-of-equityproxiesduringoursampleperiod.ConsistentwithEastonandMonahan(2005),wefindthatmostofourproxiesfailtheseconstructvaliditytests.However,theproxyproposedbyClausandThomas(2001)—whichalsosuggestsnodifferencebetweenICWandcontrolfirms—performswellinthesetests;specifically,itsassociationwithfutureexpectedreturnsisnotstatisticallydifferentfromthatpredictedbytheoryanditsmeasurementerrorisinsignifi-cant.Second,weexaminepatternsinvariousunderlyingdriversofcostofequity(riskattributes)thattheorysuggestsmaybeassociatedwithICW,includingearningsqualityproxies,informationasymmetryproxies,andmarketriskattributes.Ingeneral,wefindnodifferencebetweenICWandcontrolfirmsfortheseriskattributesaftercontrollingforprimitivefirmcharacteristics.Overall,theseadditionaltestsincreasetheconfidencethatourprimaryinferencesarenotdrivenbymeasurementerrorintheimpliedcost-of-equityproxies.

OurstudycontributestotheintensedebateregardingcostsandbenefitsoftheSection404provisionsofSOX.Ourevidencesuggeststhatcostofequity,onaverage,isunlikelydirectlyassociatedwithICWreportedunderSection404.Oneinferencethatcouldbedrawnfromthisfindingisthattherearepotentiallynosignificantcost-of-equitybenefitsassociatedwithmaintainingandreportingoneffectiveinternalcontrolsforfinancialreporting.SincereducedcostofcapitalwasclaimedasamajorbenefitoftheSection404requirements,ourresultshaveimportantimplicationsforthisregulatorydebatebyquestioningthecost-benefittrade-offsofthisregulation.

Suchastronginferencemustbetemperedinlightofconflictingevidencefromotherconcurrentstudies.UsingValueLine’sexpectedratesofreturnasaproxyforcostofequity,Ashbaugh-Skaifeetal.(2006)findthatICWs(reportedunderbothSection302andSection404)areassociatedwithhighercostofequity,evenaftercontrollingforfirmcharacteristics,andthatsubsequentremediationoftheseweaknessesresultsinloweredcostofequity.OurValueLineresultsareconsistentwiththoseofAshbaugh-Skaifeetal.(2006);however,wefindthattheseresultsaredrivenbysamplingbiasesarisingfromValueLine’srestrictedcoverage.2Also,usingashort-windowdesignBeneishetal.(2006)documentanegativestockpricereactiontoICWdisclosures(underbothSection302andSection404),whichtheyattributetoloweredearnings’expectationsandincreasedcostofequity.3However,consistentwithourresults,Beneishetal.(2006)documentnosignificantstockpriceorcost-of-equityeffectsassociatedwithSection404disclosures.ThelackofasignificantassociationbetweencostofequityandSection404disclosuresmaybeattributedto(1)thelowermaterialitythresholdsemployedbyauditorswhendisclosingSection404materialweaknesses(Doyleetal.2007b);and/or(2)thesuperiorinformationenvironmentsofthelargerfirmsthataresubjecttotheacceleratedfilingrequirementofthissection.Therefore,itispossiblethatICWcouldhavecost-of-equityimplicationsthatarenotdetectableinourstudy,whichfocusesonSection404disclosuresmadebylargerfirms.

Therestofthepaperproceedsasfollows.SectionIImotivatesourstudyanddiscussestheoreticalconsiderationsrelatingICWandcostofequity.SectionIIIdescribesthesampleandreportsdescriptiveinformation.SectionIVpresentsourprimaryresults.SectionVpresentsadditionalanalysesandSectionVIconcludes.

23TheValueLineexpectedreturnproxyalsoperformspoorlyinEastonandMonahan’s(2004)testsofconstructvalidity.ThisfurtherreducesthereliabilityoftheValueLineresults.

Hammersleyetal.(2005)alsostudymarketreactionstoSection302reports.

TheAccountingReview,October2007

1258Ogneva,Subramanyam,andRaghunandan

II.MOTIVATION

TheRegulatoryDebate

Whileevaluationofinternalcontrolsoverfinancialreportinghaslongbeenanimportantelementoftheauditingprocess(Kinneyetal.1990;Hermanson2000),therewerenorequirementsformakingpublicdisclosuresregardinginternalcontrolsinthepre-SOXera,exceptwhenan8-Kwasfiledfollowinganauditorchange.Sections302and404ofSOXarethefirstbroad-basedstatutoryrequirementsrelatingtopublicdisclosureofinternalcontroleffectiveness.Inparticular,Section404ofSOX(whichbecameeffectiveforac-celeratedfilersforfiscalyearsendingonorafterNovember15,2004)requiresthatcom-paniesincludeanauditor-attestedreportontheeffectivenessoftheirinternalcontrolsoverfinancialreportingintheirannual10-Kfilingsthatspecificallyhighlightsanydiscoveredmaterialweaknesses.Inaddition,AuditingStandardNo.2(ASNo.2)issuedbythePublicCompanyAccountingOversightBoard(PCAOB)inMarch2004requirestheauditortorenderanopinionregardingtheclient’sinternalcontrolsoverfinancialreporting.

Section404’srequirementswithrespecttointernalcontrolshavearguablybeenthemostcontroversialprovisionsofSOX.Muchofthecontroversysurroundingthissectionrelatestothehighcostsofcompliance,whichmanyperceivearenotcommensuratewiththecorrespondingbenefits.Forexample,94percentofrespondentstoaMarch2005surveybytheFinancialExecutivesInstituteindicatedthatthecostsofcompliancefaroutweighedthebenefits(FEI2005),andmanySECregistrantsandbusinessassociationshavecom-plainedaboutthecostsassociatedwiththeinternalcontrolreportingrequirements,callingfortherevision(andevenrepeal)oftherequirements(e.g.,AmericanElectronicsAssoci-ation[AEA]2005;FEI2005).Partlyinresponsetosuchconcerns,boththeU.S.SenateandtheU.S.HouseofRepresentativeshaveheldhearingsonthecostsassociatedwiththeinternalcontrolreportingrequirementsofSOX(U.S.Senate2004;U.S.HouseofRepre-sentatives2005).

RegulatorsandauditorshighlighttheperceivedbenefitsofSection404,whichlargelyrelatetoitseffectoncostofcapital.Forexample,SECChairmanWilliamDonaldsonnotesthattherequirementsofSOX,andinparticularSection404,are‘‘importantbecausetheyhaveproduced,andwillproduce,improvementsthathelptorestoreandreinforceinvestorconfidenceinourmarkets,andlowerthecostofcapitaltoissuers’’(Donaldson2005a).Consistentwiththeaboveargument,Ernst&YoungnotesthateffectiveimplementationoftheinternalcontrolreportingrequirementsofSOXshouldleadto‘‘potentialbenefitstoinvestors...[that]aresignificantintermsofimprovementsinthereliabilityofperiodicfinancialreports’’andthatasaresult‘‘issuersshouldhavealowercostofcapital’’(Ernst&Young2005).4UsergroupssuggestthattheinternalcontrolassessmentsunderSection404willinflu-encetheirdecisionsinwaysthataffectcostofcapital.Financialadvisorssuggestthatinvestorsshoulddemandahigherriskpremiumforcompanieswithineffectiveordeficientinternalcontrols,whichareassociatedwithahigherprobabilitythataccountingproblemsexist(Wilfert2005).Also,credit-ratingagenciesindicatethatinternalcontrolweaknesses,especiallyarisingfromcompany-wideproblems,areanimportantfactoraffectingcredit-ratingdecisions(Moody’sInvestorService2004;FitchRatings2005).

4Inasimilarvein,Deloitte&TouchenotesthatcomplyingwithSection404wouldleadtomultiplebenefitsincluding‘‘potentialforreducedcostofcapitalduetocapitalmarketvaluationofahighqualitycontrolenvi-ronmentwhichimprovesthepredictabilityandreliabilityofoperatingresults’’(Deloitte&Touche2005).Also,KPMG’sbrochurelists‘‘increasedconfidenceofthebondmarketsandreducedcostofcapital’’as‘‘compellingrewards’’ofinternalcontrolcompliance(KPMG2005).

TheAccountingReview,October2007

InternalControlWeaknessandCostofEquity:EvidencefromSOXSection404Disclosures1259

Also,manycompaniesvoluntarilyreportedinformationregardingtheadequacyoftheirinternalcontrolsinannualreportsinthepre-SOXera(seeBronsonetal.2006;McMullenetal.1996).Ingeneral,voluntarydisclosuresregardinginternalcontrolsweremoreprev-alentamonglargercompaniesandthesedisclosures,almostwithoutexception,happenedtobe‘‘clean,’’i.e.,therewerenoreportsofanyinternalcontroldeficiencies.Thissuggeststhatcompaniesdidbelievethattherewerebenefitsfromsignalinggoodinternalcontrolstothemarkets.

Insummary,thereisanintensedebateaboutthecostsversusbenefitsofSection404.Ontheonehand,preparersobjecttotheexcessivecostsofcompliance.Ontheotherhand,regulatorsandauditorspointtothepotentialbenefits,whichlargelyarisethroughreducedcostofcapitalduetoimprovedfinancialtransparency.

InternalControlsandCostofEquityCapital

ImplicitintheargumentsoftheproponentsofSection404isthefollowingtwo-steplink.First,effectiveinternalcontrolsareexpectedtoreduceaccountingrisk,therebypo-tentiallyimprovingaccountingquality.Second,improvedaccountingquality(orreducedaccountingrisk)ishypothesizedtoreducecost-of-equitycapital.Inthissection,wetheo-reticallyexplorethistwo-steplinkbetweeninternalcontrolsandcostofequity.WefirstexaminethenatureofreportedICWsandhowtheycouldincreaseaccountingrisk.Wethenexplorethetheoreticalrelationbetweenaccounting(information)qualityandcost-of-equitycapital.Finally,wehypothesizeanalternativemechanismbywhichICWcouldbeindirectlyassociatedwithcostofequity.

Internalcontrolsoverfinancialreportingaredesignedtoassurereliabilityofaccountinginformation,therebyprovidingexternaluserswithfinancialstatementsofpotentiallyhigherquality.5Reportedmaterialweaknessesininternalcontrolsencompassawidevarietyofissues.GeandMcVay(2005)findthatamajorityofICWsdisclosedpursuanttoSection302ofSOXpertaintoproblemswithspecificaccounts(e.g.,inventory,accountsreceivables/payables).OtherICWsarebroaderinscopeandcanindirectlyaffectorgani-zationalcontrolprocessesinadditiontoimpairingfinancialaccountingquality.Suchdefi-cienciescouldpertaintosystemsororganizationaldesignflawssuchasimpropersegre-gationofdutiesorpoorinformationsystemssecurity.ICWmayalsoarisefrominadequateorpoorlytrainedpersonnel—GeandMcVay(2005)notethatpersonnelproblemsareattherootofmanyofthereportedweaknesses.Finally,ICWcanarisefromthehighestlevelsoftheorganization,includingpoorcorporategovernancestructuresandthetopmanage-ment’sattitudetowardinternalcontrols(‘‘toneatthetop’’).

ThePCAOB(2004)definesamaterialweaknessininternalcontrolsoverfinancialreportingasoneormoredeficienciesthat‘‘resultsinmorethanaremotelikelihoodthatamaterialmisstatementoftheannualorinterimfinancialstatementswillnotbepreventedordetected.’’ICWcansignificantlyincreasethelikelihoodofmisstatementinfinancialstatementsforatleastthreereasons.First,ICWcanincreasethelikelihoodofaccountingerror.Forexample,inadequatelytrainedstaffincreasesthelikelihoodoferrorsinthein-terpretationandapplicationofGAAP.Second,asGeandMcVay(2005)observe,systemicinternalcontrolproblemscanincreasethescopeformanagingearningsatvariouslevelsintheorganization.Forexample,internalcontrolproblemsrelatedtocontractualagreementswithcustomersandthemonitoringofsuchcontractscanmakeiteasierformanagersto

5Forexample,accordingtoPCAOBAuditingStandardNo.2(PCAOB2004),internalcontroloverfinancialreportingis‘‘aprocessdesigned...toprovidereasonableassuranceregardingthereliabilityoffinancialreportingandthepreparationoffinancialstatementsforexternalpurposesinaccordancewithGAAP.’’

TheAccountingReview,October2007

1260Ogneva,Subramanyam,andRaghunandan

manipulaterecognitionofrevenue.Finally,materialICWscanalsoincreasethepropensityforfinancialfraud.Forexample,inadequateorimpropersegregationofdutiescancreateopportunitiesforemployeefraud.Accordingly,ICWincreaseaccountingriskandthereforehavethepotentialtoimpairaccountingquality.

TheoreticalworkbyEasleyandO’Hara(2004)andLambertetal.(2007)suggeststhatinformationrisk(riskarisingfrompoorinformationquality)mayhaveanondiversifiablecomponentthatispricedbythemarket.EasleyandO’Haraarguethatinformationqualityrepresentsauniqueriskfactorarisingfrominformationasymmetry.Intheirmodel,unin-formedinvestorsdemandhigherexpectedreturnforsecuritieswithgreaterinformationasymmetrytoprotectthemselvesfromtheirinformationdisadvantagevis-a`-visinformedinvestors.Poor(public)informationqualityincreasescostofcapitalbecauseitexacerbatesinformationasymmetry.Lambertetal.(2007)disputeEasleyandO’Hara’s(2004)claimthatinformationqualityrepresentsauniqueriskfactor,butsuggestthatinformationqualitycanaffectcost-of-equitycapitalbothdirectlyandindirectly.Thedirecteffectarisesbecausesuperiorinformationqualityreducesafirm’sbetabydampeningconditionalcovarianceofthefirm’scashflowswiththemarket.Theindirecteffectoccursbecausepoorinformationqualitycanaffectthefirm’srealdecisionsbyincreasingagencyproblems.Recentempiricalevidence(Francisetal.2004)documentsaninverserelationshipbetweenearningsqualityandimpliedcostofequity,therebyprovidingsupportforthetheoreticalpredictionsthatinformationqualitycanaffectcost-of-equitycapital.

Thusfar,wesuggestthatICWcanleadtohighercostofcapitalthroughitspotentialeffectonqualityofaccountinginformation.Alternatively,poorinternalcontrolsoverfi-nancialreportingmayalsobeassociatedwithgeneralmanagementcontroldeficiencies,includinginadequatecontrolsoverbusinessprocessesandpoorcorporategovernancestruc-tures.Suchassociationisexpectedtobeespeciallystrongwheninternalcontrolweaknessesoverfinancialreportingarecausedbysystemicproblemssuchastopmanagement’sphi-losophytowardinternalcontrols(‘‘toneatthetop’’).Poormanagementcontrolsmayleadtoexcessiverisktakingbymanagers(e.g.,investinginhighlyriskyprojects),especiallyduringperiodsofpoorperformance.6Increasedrisktakingcanincreasecashflowvolatility,whichwillbepricedtotheextentitissystematic.Excessiverisktakingcanalsoincreasetheprobabilityofbusinessfailure,whichmaybepricedbythestockmarket(FamaandFrench1995).7WethereforehypothesizethatICWcanbeindirectlyassociatedwithhighercostofequitybecauseitissymptomaticofgenerallypoorcorporatecontrolsandincreasedbusinessrisk.

III.SAMPLEANDDESCRIPTIVEANALYSIS

SampleSelection

Weobtainourinitialsampleof3,802firmsthatfiledfirst-timeSection404reportswiththeSECbetweenNovember2004andJanuary2006fromtheAuditAnalyticsdatabase.Thissamplecomprises573adverse(i.e.,withmaterialinternalcontrolweakness)and3,229

67Managers’propensitytowardexcessiverisktakinghasbeendocumentedinthestrategicmanagementandcor-poratefinanceliteratures.Severalpriorstudiesshowthatmanagersexhibitapropensitytoengageinexcessiverisktakingwhenthefirm’sperformanceislaggingcomparedtosomebenchmark.Forexample,Bowman(1982)andFiegenbaumandThomas(1988)provideevidenceofthehigheroverallriskinessofthosefirmsthatperformworsethantheirindustrypeers.Further,studiesonmutualfundmanagers’behaviordocumentthatmanagerstendto‘‘gamble’’byincreasingthevolatilityofthemanagedportfoliosintheendofthefiscalyearifthefundhasbeenunderperformingintheinterimquarters(e.g.,Brownetal.1996;ChevalierandEllison1997).

Anecdotalevidence(e.g.,EnronandWorldCom)isconsistentwithpoorinternalcontrolsincreasingbusinessriskandleadingtobankruptcy.

TheAccountingReview,October2007

InternalControlWeaknessandCostofEquity:EvidencefromSOXSection404Disclosures1261

cleanreports.Outofthisinitialsample,2,821firms(approximately74percent)meetourcommondatarequirementsforimplied-cost-of-equityestimationasofJune2004:(1)con-sensus(median)one-andtwo-year-aheadforecastsofEPSintheI/B/E/SdatabaseasofJune2004;(2)sharepriceandnumberofsharesoutstandingasofJune2004intheI/B/E/Sdatabase;and(3)bookvalueanddividendpayoutfor2003intheCompustatannualdatabase.Weloseanother232firmsbecauseofnon-availabilityofdataforcontrolvariablesinourmultipleregressionmodels.Finally,weeliminate74firmswithcleanSection404reportsthathadpreviouslyreportedmaterialweaknessesunderSection302.Accordingly,ourfinalsamplefortheJune2004analysescomprises2,515firmsofwhich346firms(14percent)haveadverseSection404reports(ICWfirms)and2,169firms(86percent)havecleanSection404andSection302reports(controlfirms).Theexactnumberofobservationsinspecifictests,however,willbeloweranddependon(1)theuniquedatarequirementsoftheimplied-cost-of-equitymeasureorothervariablesusedinthattestand(2)deletionofoutliers.

FactorsAssociatedwithInternalControlWeakness

GeandMcVay(2005),Ashbaugh-Skaifeetal.(2007),andDoyleetal.(2007a)showthatICWdisclosuresunderSections302and404areassociatedwithanumberoffirmcharacteristics.Inthissection,weexaminewhetherthefactorsidentifiedbytheseauthorsexplaintheincidenceofICWinourSection404-onlysample.Accordingly,weestimatethefollowingcross-sectionallogitmodel(wheresubscriptidenotesfirm):

Prob(WEAKi)ϭf(␤ϩ␤1*LOGSEGMENTSiϩ␤2*FOREIGNiϩ␤3*M&Aiϩ␤4*RESTRUCTUREiϩ␤5*SALEGRWiϩ␤6*INVENTORYiϩ␤7*LOGMKTViϩ␤8*LOSSiϩ␤9*RZSCOREiϩ␤10*LOGAGEi)

(1)

where:

WEAKϭindicatorvariablethattakesavalueof1ifafirmhasanICW,and

0otherwise;

LOGSEGMENTSϭlogarithmofnumberofbusinesssegmentsreportedin2004

Compustatsegmentfileforfiscalyear2003;

FOREIGNϭindicatorvariableequalto1ifthefirmhasanon-zeroforeign

currencytranslation(Compustatdataitem#150)infiscalyear2003,and0otherwise;

M&Aϭindicatorvariableequalto1ifacompanywasinvolvedinmergers

oracquisitionsoverthethreeyearsprecedingJune2004,and0otherwise(theM&AdatacomefromtheSDCPlatinumdatabase);

RESTRUCTUREϭindicatorvariableequalto1ifatleastoneofCompustatannual

dataitems#376,#377,#378,or#379isnotequalto0foranyfiscalyearin2001–2003,and0otherwise;

SALEGRWϭindicatorvariablethatisequalto1ifindustry-adjustedgrowthin

sales(Compustatdataitem#12)fallsintothetopquintileinfiscalyear2003,and0otherwise;

INVENTORYϭinventory(Compustatannualdataitem#3)overtotalassets

(Compustatannualdataitem#6);

TheAccountingReview,October2007

1262Ogneva,Subramanyam,andRaghunandan

LOGMKTVϭlogarithmofmarketvalueofequity,wheremarketvalueofequity

isequaltopricepersharemultipliedbynumberofsharesoutstandinginI/B/E/SasoftheendofJune2004;

LOSSϭindicatorvariableequalto1ifearningsbeforeextraordinaryitems

(Compustatdataitem#18)in2002and2003sumtolessthan0;

RZSCOREϭdecreasingdecilerankofAltman’s(1968)Z-Score;weuse

decreasingrankssothatlargervaluesindicatehigherbankruptcyprobabilities;and

LOGAGEϭlogarithmofnumberofyearsthefirmexistsintheCRSPdatabase.WeincludevariablesfromGeandMcVay(2005),Ashbaugh-Skaifeetal.(2007),andDoyleetal.(2007a)thatarehypothesizedtobeassociatedwiththediscoveryofICW.SinceSection404disclosuresaremandatory,wedonotincludevariablesthatareassociatedwithcompanies’propensitytoreportICWoncesuchaweaknessisdiscovered.Thefun-damentalfactorsassociatedwiththeexistenceofinternalcontroldeficienciesthatweex-plorepertainto:(1)complexityofoperations;(2)organizationalchange;(3)accountingapplicationmeasurementrisk;and(4)resourceconstraintindicators.Weincludetwocom-plexitymeasures:numberofbusinesssegments(SEGMENTS)andforeignoperations(FOREIGN).Toidentifyorganizationalchange,weincludeindicatorvariablestorepresentrecentmergeroracquisitionactivity(M&A)orrestructuringactivity(RESTRUCTURE).Wecaptureaccountingapplicationandmeasurementriskthroughsalesgrowth(SALEGRW)andlevelofinventory(INVENTORY).Finally,wemeasureresourceconstraintsthroughthemarketvalueofequity(MKTV),existenceofaloss(LOSS),andbankruptcyprobability(RZSCORE).Weincludeageofthefirm(AGE)sinceDoyleetal.(2007a)suggestthatolderfirmshavebetterinternalcontrols.

Table1,PanelAreportsmeanandmedianvaluesoftheseeconomiccharacteristicsseparatelyfortheICWandcontrolsamples,andthedifferencesacrossthesegroups(alongwithstatisticalsignificance).WefindthatICWfirms(1)havemuchhigheraccountingapplicationriskasevidencedbyhighersalesgrowth(SALEGRW)andinventorylevels(INVENTORY);(2)havehigherresourceconstraintsasevidencedbylowermarketvalue(MKTV),greaterincidenceofaloss(LOSS),andhigherbankruptcyprobability(RZSCORE);and(3)aresomewhatyounger(i.e.,lowerAGE)thanthecontrolfirms.However,wefindmixedevidenceon(1)complexityofoperations—lowernumberofsegments(SEGMENTS)butweaklyhigherincidenceofforeignoperations(FOREIGN)forICWfirms;and(2)organizationalchange—lowerincidenceofmergers(M&A)buthigherincidenceofrestruc-turingactivity(RESTRUCTURE)forICWfirms.

Table1,PanelBreportsresultsofthelogitestimationofEquation(1).ThelogitresultsarelargelyconsistentwiththedescriptiveanalysisinPanelA.Specifically,thelikelihoodofhavinganinternalcontrolweaknessincreaseswiththeincidenceofforeignoperations(FOREIGN);increaseswithaccountapplicationriskfactorsSALEGRWandINVENTORY;increaseswhenacompanyexperiencesanorganizationalchangesuchasarestructuringactivity(RESTRUCTURE);andincreaseswithresourceconstraints,thatis,lowermarketvalue(MKTV),greaterlosses(LOSS),andhigherbankruptcyprobability(RZSCORE).

Insummary,wefindthatICWsareassociatedwithincidenceofforeignoperations,restructuringactivity,accountapplicationrisk(higherinventorylevelsandsalesgrowth),andresourceconstraints(lowermarketvalue,greaterlosses,andbankruptcyrisk).OurresultsaregenerallyconsistentwiththoseinAshbaugh-Skaifeetal.(2007),GeandMcVay(2005),andDoyleetal.(2007a).

TheAccountingReview,October2007

InternalControlTABLE1

FirmCharacteristicsAssociatedwithInternalControlWeaknesses:June2004

PanelA:MeanandMedianDifferencesbetweenWeaknessandControlFirms

MKTV($MM)RZSCORESEGMENTSFOREIGNSALEGRWINVENTORYM&ATheRESTRUCTUREAccountingLOSSAGE

Review,PREDFERR1OctoberPREDFERR2VLPREDFERR1

2007#Obs.FullUsed

Sample2,4653,2862,5155.2212,5152.1172,5150.1902,5150.1732,4900.0792,5150.6422,5150.4152,5150.2532,51517.583

2,465Ϫ0.0272,465Ϫ0.0161,085

Ϫ0.017

Means

ICWControl1,2893,6025.7375.1391.9882.1380.2170.1860.2110.1660.0940.0760.5900.6500.4620.4080.3840.23214.538

18.069

Ϫ0.028Ϫ0.027Ϫ0.013Ϫ0.016Ϫ0.021

Ϫ0.017

ICWϪFullControl

Sample

Ϫ2,313

850(0.00)0.5985.000(0.00)Ϫ0.1491.000(0.10)0.031—(0.18)0.045—(0.04)0.0170.028(0.01)Ϫ0.060—(0.03)0.055—(0.05)0.152—(0.00)Ϫ3.53111.000

(0.00)Ϫ0.001Ϫ0.026(0.26)0.003Ϫ0.017(0.00)Ϫ0.004Ϫ0.016

(0.00)

Medians

ICWϪICWControl

Control

478973Ϫ495(0.00)6.0005.0001.000(0.00)1.0001.0000.000(0.06)————————0.0340.0280.006(0.06)————————————9.000

12.000

Ϫ3.000(0.00)Ϫ0.027Ϫ0.026Ϫ0.001(0.23)Ϫ0.016Ϫ0.0180.002(0.01)Ϫ0.018

Ϫ0.016

Ϫ0.002(0.00)

(continuedonnextpage)

WeaknessandCostofEquity:EvidencefromSOXSection404Disclosures1263TheAccountingTABLE1(continued)

Review,PanelB:LogitAnalysis

IndependentVariable؍WEAKOctoberDependentVariablesParameterEstimatePrϾ␹2InterceptϪ6.620

(0.00)RZSCORE0.097

(0.00)2007LOGMKTVϪ0.528(0.00)LOGAGEϪ0.120(0.12)LOSS

0.357(0.01)LOGSEGMENTSϪ0.024(0.83)FOREIGN0.430(0.01)M&A

Ϫ0.049(0.71)SALEGRW

0.441(0.01)RESTRUCTURE0.466(0.00)INVENTORY

2.115(0.00)LikelihoodRatioX2(PrϾX2)206.962(Ͻ.0001)

ObservationsUsed

2,465

PanelAcomparesmeanandmediancharacteristicsfor2,515firmsthatfiledfirst-timeSection404opinionsduringtheNovember2004toJanuary2006period.Ofthe2,515firms,346filedadversereports(ICWfirms)and2,169filedcleanreports(controlfirms).Observationswithvaluesfallingintothetopandbottomextreme

percentilesoftherespectivecontinuousvariablearedeleted.Two-sidedp-valuesforcomparisonswiththecontrolsamplearereportedinparentheses:p-valuesformeancomparisonsarebasedont-statistics,andp-valuesformediancomparisonsarebasedonaWilcoxonsignedranktest.

PanelBreportsresultsofestimatingthelogitmodel:Prob(WEAK*SALEGRWi)ϭf(␤ϩ␤␤1*LOGSEGMENTSiϩ␤2*FOREIGNiϩ␤3*M&Aiϩ␤4*RESTRUCTUREϩ␤i5iϩ␤6*INVENTORYiϩ␤7*LOGMKTViϩ␤8*LOSSiϩ9*RZSCOREiϩ␤10*LOGAGEothervariablesi).p-valuesarebasedonChi-squaredtests.WEAKisanindicatorvariablethattakesonavalueof1ifthefirmdisclosesanICW,and0otherwise.Allareasdefinedbelow.Allobservationswithstudentizedresiduals(ofanidenticalOLSregression)fallingintothetopandbottomextremepercentilesaredeleted,resultinginasampleof2,465observations.

(continuedonnextpage)

1264Ogneva,Subramanyam,andRaghunandanTABLE1(continued)

VariableDefinitions:

(LOG)SEGMENTSϭ(logarithmof)numberofbusinesssegmentsreportedin2004CompustatSegmentfileforfiscal2003;

FOREIGNϭanindicatorvariableequalto1ifthefirmhasanon-zeroforeigncurrencytranslation(Compustatdataitem#150)infiscal2003;

M&Aϭanindicatorvariableequalto1ifacompanywasinvolvedinmergersoracquisitionsoverthethreeyearsprecedingJune2004,whereM&A

datacomefromtheSDCPlatinumdatabase;

RESTRUCTUREϭanindicatorvariableequalto1ifatleastoneofCompustatannualdataitems#376,#377,#378,or#379isnotequalto0foranyfiscal

yearin2001–2003;

SALEGRWϭanindicatorvariablethatisequalto1ifindustry-adjustedgrowthinsales(Compustatdataitem#12)fallsintothetopquintileinfiscal

2003;

INVENTORYϭinventory(Compustatannualdataitem#3)overtotalassets(Compustatannualdataitem#6);

(LOG)MKTVϭ(logarithmof)marketvalueofequity,wheremarketvalueofequityisequaltopricepersharemultipliedbynumberofsharesoutstanding

inI/B/E/SasoftheendofJune2004;

TheLOSSϭanindicatorvariableequalto1ifearningsbeforeextraordinaryitems(Compustatdataitem#18)in2002and2003sumtolessthan0;RZSCOREϭdecreasingdecilerankofAltman’s(1968)Z-Score(computedasdecilerankofZ-scoretimes–1andplus10);Accounting(LOG)AGEϭ(logarithmof)numberofyearsthefirmexistsinCRSPdatabase;and

(VL)PREDFERR1,2ϭ(ValueLine)I/B/E/Sone-ortwo-year-aheadpredictedforecasterrors,computedusingtheLiuandSu(2005)model(seeAppendixBfor

details).

Review,October2007InternalControlWeaknessandCostofEquity:EvidencefromSOXSection404Disclosures12651266Ogneva,Subramanyam,andRaghunandan

IV.IMPLIEDCOSTOFEQUITYANDINTERNALCONTROLWEAKNESSMeasuringImpliedCostofEquity

Asinpriorresearch(e.g.,Botosan1997),weuseimpliedcostofequitytoproxyforthefirms’cost-of-equitycapital.Implied-cost-of-equityestimatesareinternalratesofreturn(IRR)thatequatecurrentstockpricestoexpectedfuturepayoffsandareestimatedbyinvertingtherelationbetweenanalysts’forecastsoffutureearningsandcurrentstockprices.Priorresearchgeneratesseveralalternativeestimatesofimpliedcostofequity.Ingeneral,theestimationmethodsdifferintermsofthevaluationmodelusedandassumptionsre-gardingterminalvaluecomputation.Priorresearchevaluatesalternativeempiricalmeasuresofimpliedcostofequityandreachesdifferentconclusionsontheirrelativemeritsanddemerits(e.g.,Guayetal.2005;EastonandMonahan2005;BotosanandPlumlee2005).Wereportresultsusingseveralalternativeestimatestoensurethatourresultsarerobusttomethodchoice.

Thefirsttwomethodsthatweusearebook-valuegrowthspecificationsbasedonOhlson(1995):theindustryROEmethoddevelopedbyGebhardtetal.(2001),labeledGLS;andtheeconomy-widegrowthmethodofClausandThomas(2001),labeledCT.ThenextthreemethodsareearningsgrowthspecificationsbasedonOhlsonandJuettner-Nauroth(2005):theunrestrictedabnormalearningsgrowthmethodofGodeandMohanram(2003),labeledGM;therestrictedabnormalearningsgrowthmethoddevelopedbyEaston(2004),labeledPEG;andtheunrestrictedabnormalgrowthmethodincorporatinglong-termgrowthforecasts,whichwederiveandlabelOJ.

Allofthefivemethodsabovemakeassumptionsregardinglong-termearningsgrowthratesusedtoderiveterminalvalues.Theseassumptionsintroducesignificantmeasurementerrorintheestimationofcostofequity(EastonandMonahan2005).Accordingly,weestimatetwoothermeasuresofcostofequitythatarearguablylesssusceptibletoerrorsinlong-termgrowthassumptions.First,weusetheexpectedratesofreturnprovidedbyValueLine(VL),whichhavebeenusedbyseveralauthors(e.g.,Francisetal.2004;Ashbaugh-Skaifeetal.2006).Second,weusethemethodologydevelopedbyEaston(2004)tosimultaneouslyestimatecostofequityandlong-termgrowthrates(EAST).SincetheEaston(2004)methodcanestimatecostofequityonlyforaportfolioofstocks,weestimateby-firmmeasuresofcostofequitythroughabootstrappingprocedureusingrandomlyse-lectedfirms.Weprovidedetailsonthederivationandestimationofallourcost-of-equitymeasuresinAppendixA.

Thedatanecessaryforcost-of-equityestimationareobtainedfromtheI/B/E/SandCompustatdatabases.Animportantdesignfeatureinourstudyisthatwemeasurecostofequityforourentiresampleoffirmsatanidenticalpointintime,whichensuresthatourresultsarenotcontaminatedbyintertemporalchangesincostofequityattributabletomacroeconomicfactors.Inourprimarytests,weestimateimpliedcostofequityasoftheendofJune2004,whichisatleastfivemonthspriortoanyoftheSection404disclosures.8BychoosingapointintimebeforethedisclosureofSection404reports,weinvestigatewhetherICWfirmsareinherentlyriskierthancontrolfirms.However,suchadesigndoesnotincorporateICWdisclosureeffects.Accordingly,wealsoconductsensitivityanalysesbyreplicatingallourtestsasofJune2006.

8TheValueLineInvestmentSurveyhasathree-monthcycleforcoveringallfollowedstocks.OurValueLineestimatesusedinourprimaryanalysespertaintothesecondcalendarquarterof2004andarenotderivedatthesamepointintimeforallsamplefirms.Similarly,oursupplementaryanalysesuseValueLineestimatespertainingtothesecondcalendarquarterof2006.

TheAccountingReview,October2007

InternalControlWeaknessandCostofEquity:EvidencefromSOXSection404Disclosures1267

PrimaryAnalyses:June2004UnivariateResults

Table2,PanelAreportsthesamplesizesforeachofthecost-of-equitymeasures(estimatedattheendofJune2004).ExceptforVL,themeasuresgeneratesamplesofatleast2,107firms(i.e.,atleast84percentofallavailablefirms),withPEGgeneratingthelargestsample(2,369firms,or94percentofallavailablefirms).TheVLmeasureisavailableforonly1,107firms,whichislessthan50percentofallavailablefirms.ThesmallersampleislargelyduetoVL’slimitedcoverage.TheproportionofICWfirmsrangesfrom12percentto14percentformostofthesamples;theVLsampleisagainexceptional,withfewerthan9percentICWfirms.Unreportedresultssuggestthatthedistributionpa-rametersforourcost-of-equityestimatesaregenerallyconsistentwiththosereportedinpriorresearch(e.g.,Guayetal.2005;EastonandMonahan2005;BotosanandPlumlee2002,2005).

Table2,PanelAexaminesdifferencesinmeanandmediancostofequitybetweenICWandcontrolfirmsattheendofJune2004foreachofoursevenmeasures.Theactualnumberofobservationsusedinourtestsissmallerthantheavailablesamplebecausewedeleteoutliers.MeancostofequityishigherforICWfirmscomparedtocontrolfirmsforallmeasureswithexceptionofEAST(thedifferencesinCTandEASTarenotstatisticallysignificantatconventionallevels).Thedifferencesrangefromalowof–41basispointsforEASTto289basispointsforVL;theaveragedifferenceacrossallsevenmeasuresis80basispoints.Mediandifferencesshowqualitativelysimilarpatterns,althoughthedif-ferencesaresomewhatweakerbotheconomicallyandstatistically(theaveragedifferenceinmediansacrossallsevenmeasuresis73basispoints).Overall,univariatecomparisonsindicatethatfirmsreportingICWaregenerallyassociatedwithhighercostofequityattheendofJune2004.

ControllingforFirmCharacteristicsandForecastBias

WenextexaminewhethertheICWfirmscontinuetoexhibithighercostofequityaftercontrollingforcorrelatedomittedvariables.Wecontrolfortwosetsoffactors.First,wecontrolforvariousprimitivefirmcharacteristicsthatareassociatedwithICW.InSectionIIIweshowthatseveralcharacteristics,suchassize,complexityofoperations,anddistress,areassociatedwithICWfirms.Someofthesecharacteristics(e.g.,sizeanddistress)havebeenconjecturedtoproxyforsystematicriskinthefinanceliterature(FamaandFrench1995,1996).Inaddition,factorssuchascomplexityofoperationscouldincreaseafirm’ssystematicrisk.Therefore,itisimportanttoexaminewhetherthehighercostofequityassociatedwithICWfirmsisindeedattributabletoICWperseorarisesmerelyfromcorrelationwithmoreprimitivefirmcharacteristics.Second,wecontrolforbiasinanalysts’forecasts.Sinceimplied-cost-of-equityestimatesuseanalysts’forecastsasinputs,forecastbiashasbeenshowntocausebiasintheseestimates(EastonandMonahan2005;Guayetal.2005).Table1revealsdifferencesbetweenICWandcontrolfirmsinone-year-aheadandtwo-year-aheadanalysts’predictedforecasterrors.Accordingly,asinDhaliwaletal.(2005),wecontrolforanalystforecastbias.

WeexaminedifferencesinimpliedcostsofequityacrossICWandcontrolfirmsthroughthefollowingcross-sectionalOLSregressionthatincludesallfirmcharacteristicsfromEquation(1)andaproxyforpredictedforecasterrors(wheresubscriptidenotesfirm):

TheAccountingReview,October2007

TheAccounting1268TABLE2

DifferencesacrossWeaknessandControlFirmsforVariousImpliedCostofEquityMeasures:June2004

OJ

GLSCTPEGGMVLEASTPanelA:MeanandMedianDifferencesbetweenWeaknessandControlFirms

Review,SampleSize

2,107OctoberObservationsUsed

2,065ICW

0.1073MeansControlSample

0.10272007ICW–Control(0.00)0.0046ICW

0.1048MediansControlSample

0.1011ICW–Control

(0.00)

0.0038PanelB:MultipleRegressionAnalysisIntercept0.0909(0.00)WEAK0.0008(0.56)LOGMKTVϪ0.0021(0.00)RZSCORE(0.95)0.0000LOGSEGMENTSϪ0.0010(0.16)FOREIGNϪ0.0007(0.56)SALEGRW

0.0113(0.00)

2,2482,1862,2042,1440.08110.08100.07920.0808(0.11)0.0018(0.89)0.00020.07910.08090.07840.0802(0.18)

0.0007(0.80)

0.00070.04270.0576(0.00)(0.00)Ϫ0.0016Ϫ0.0016(0.07)(0.17)Ϫ0.0022Ϫ0.0011(0.00)(0.00)(0.00)0.0004(0.00)0.00130.00160.0000(0.00)(0.98)0.0016Ϫ0.0014(0.03)(0.17)0.00150.0069(0.05)

(0.00)

2,3692,2062,3232,1620.12370.12220.10850.1130(0.00)0.0152(0.00)0.00920.10680.11160.09640.1047(0.00)

0.0103(0.00)

0.00700.03480.0433(0.00)(0.00)0.00130.0004(0.58)(0.83)0.0076Ϫ0.0060(0.00)(0.00)(0.00)0.0023(0.00)0.00120.00010.0008(0.92)(0.45)0.00780.0069(0.00)(0.00)0.00300.0046(0.17)

(0.01)

1,1072,2061,0882,1620.13860.10080.10970.1048Ϫ0.0041(0.00)0.0289(0.49)0.13500.10490.10500.1061Ϫ0.0012(0.00)

0.0300(0.99)

0.15930.1649(0.00)(0.00)0.02080.0019(0.00)(0.75)0.00660.0024(0.00)(0.11)Ϫ0.0013Ϫ0.0019(0.07)(0.03)Ϫ0.00580.0034(0.02)(0.29)Ϫ0.00650.0077(0.09)(0.14)0.00660.0096(0.18)

(0.08)

(continuedonnextpage)

Ogneva,Subramanyam,andRaghunandanϪϪTABLE2(continued)

INVENTORY0.02540.03410.04510.04050.02260.0239Ϫ0.0438

(0.00)(0.00)(0.00)(0.00)(0.00)(0.07)(0.01)M&A

Ϫ0.0039Ϫ0.0005Ϫ0.0034Ϫ0.0041Ϫ0.00320.0016Ϫ0.0062(0.00)(0.46)(0.00)(0.02)(0.02)(0.67)(0.16)RESTRUCTURE0.00070.00290.00000.00880.00630.00290.0061(0.53)(0.00)(0.98)(0.00)(0.00)(0.40)(0.17)LOSSϪ0.00200.0042Ϫ0.00580.04280.03080.0070Ϫ0.0167(0.12)(0.00)(0.00)(0.00)(0.00)(0.17)(0.00)LOGAGEϪ0.0013Ϫ0.00140.00050.00070.0022Ϫ0.0109Ϫ0.0047(0.02)(0.00)(0.28)(0.47)(0.00)(0.00)(0.05)PREDFERRϪ0.0490Ϫ0.7824Ϫ0.32260.3303Ϫ0.3375Ϫ1.49200.8883(0.17)(0.00)(0.00)(0.00)(0.00)(0.00)(0.00)ObservationsUsed2,0652,2042,1442,3232,1621,0852,162AdjustedR211.67%

48.71%

20.11%

36.09%

28.49%

13.86%

4.23%

ThetablecomparesimpliedcostofcapitalestimatesasofJune2004forasampleof2,515firmsthatfiledfirst-timeSection404opinionsduringtheNovember2004ThetoJanuary2006period.Ofthe2,515firms,346filedadversereports(ICWfirms)and2,169filedcleanreports(controlfirms).Samplesizevariesdependingonthecost-of-equityproxyused.Cost-of-equityproxiesare:OJ:ameasurebasedontheOhlsonandJuettner-Nauroth(2005)modelincorporatinganalysts’long-termearningsAccountinggrowthforecasts;GLS:theindustryROEmethoddevelopedbyGebhardtetal.(2001);CT:theeconomy-widegrowthmethodofClausandThomas(2001);PEG:therestrictedabnormalearningsgrowth(orPEG)methoddevelopedbyEaston(2004);GM:theunrestrictedabnormalearningsgrowthmethodofGodeandMohanram(2003);VL:expectedrateofreturnaccordingtotheValueLineInvestmentSurvey;andEAST:costofequityderivedfromourmodifiedversionofsimultaneouscost-of-equityandgrowthcomputation.Detailsontheestimationofallcost-of-equitymeasuresareprovidedinAppendixA.

PanelAcontainsmeanandmediancost-of-equitycomparisons.ObservationswithvaluesfallingintothetopandbottomextremepercentilesoftherespectivevariableReview,aredeleted.‘‘Samplesize’’indicatesthesizeofthesamplebeforetruncationand‘‘ObservationsUsed’’aftertruncation.Two-sidedp-valuesforcomparisonswithcontrolsamplearereportedinparentheses:p-valuesformeancomparisonsarebasedont-statisticsandp-valuesformediancomparisonsarebasedonaWilcoxonsignedranktest.

October(continuedonnextpage)

2007InternalControlWeaknessandCostofEquity:EvidencefromSOXSection404Disclosures1269TheAccountingReview,OctoberTABLE2(continued)

2007PanelBreportsresultsforthefollowingregression:

COECiϭ␦ϩ␦0*WEAKiϩ␦1*LOGSEGMENTSiϩ␦2*FOREIGNiϩ␦3*M&Aiϩ␦4*RESTRUCTUREiϩ␦5*SALEGRWiϩ␦6*INVENTORYiwhere:

ϩ␦7*LOGMKTVibϩ␦8*LOSSiϩ␦9*RZSCOREiϩ␦10*LOGAGEiϩ␦11*PREDFERRiϩ␧

COECϭoneofthesevencost-of-equityestimatesdefinedabove;

WEAKϭindicatorvariablethattakesonavalueof1ifthefirmdisclosesanICW,and0otherwise;and

PREDFERRϭaverageofone-year-aheadandtwo-year-aheadpredictedforecasterrors,computedusingtheLiuandSu(2005)model.

FortheValueLinecost-of-equityanalysis,PREDFERRϭone-year-aheadpredictedforecasterrorestimatedusingValueLinedata(seeAppendixBfordetails).AllothercontrolvariablesareasdefinedinTable1.

Wedelete1percentoftheobservationswiththesmallestandthelargestvaluesofstudentizedresidualstoremovetheeffectofoutliers.Two-sidedp-valuesarereportedinparentheses.

1270Ogneva,Subramanyam,andRaghunandanInternalControlWeaknessandCostofEquity:EvidencefromSOXSection404Disclosures1271

COECiϭ␦ϩ␦0*WEAKiϩ␦1*LOGSEGMENTSiϩ␦2*FOREIGNiϩ␦3*M&Aiϩ␦4*RESTRUCTUREiϩ␦5*SALEGRWiϩ␦6*INVENTORYiϩ␦7*LOGMKTViϩ␦8*LOSSiϩ␦9*RZSCOREiϩ␦10*LOGAGEiϩ␦11*PREDFERRiϩ␧i,

(2)

where:

COECϭoneofthesevenproxiesforimpliedcost-of-equitycapital;and

PREDFERRϭone-year-ahead(averageofone-andtwo-year-ahead)predictedforecast

errorfortheVLregression(allotherregressions),computedusingtheLiuandSu(2005)model;seeAppendixBfordetails.Allothervariablesareasdefinedearlier.

Table2,PanelBreportsresultsofestimatingtheabovemodel(asofJune2004).ThenumberofobservationsusedintheregressionsissmallerthantheavailablesamplereportedinPanelAbecausewedeleteinfluentialobservations.Afterincludingcontrolsforunder-lyingfirmcharacteristicsandforecastbias,wefindlittleevidencesuggestingthattheim-pliedcostofequityishigherforICWfirms.Specifically,thecoefficientsonWEAKareeithernegativeorinsignificantatconventionallevelsofsignificance,suggestingthatimpliedcostofequityisnothigherforICWfirmsthanforcontrolfirms.TheoneexceptionisVL,wheretheICWfirmsdisplaybotheconomically(208basispoints)andstatistically(pϽ0.01)highercostofequity.

WemakenopredictionsaboutthedirectionoftherelationbetweencharacteristicsassociatedwithICWandimpliedcostofequity,becausewithfewexceptionsitisdifficulttopredictthedirection.Accordingly,weonlyreportsomeoftheconsistentpatternsthatweobserveacrossourregressions.Smallerfirmsize(lowerMKTV)andhigherbankruptcyprobability(RZSCORE)arepositivelyassociatedwithimpliedcostofequitywiththeex-ceptionofVLandEAST.9Applicationriskmeasures(SALEGRWandINVENTORY)aregenerallypositivelyassociatedwithcostofequity.Theassociationbetweencostofequityandorganizationalchangeismixed:mergers(M&A)appeartoreducecostofequitywhilerestructuringactivity(RESTRUCTURE)appearstoincreasecostofequity.Finally,predictedforecastbias(PREDFERR)isstronglynegativelyassociatedwithimpliedcostofequityformostmeasures,suggestingthatanalysts’forecastoptimismtendstoincreasecost-of-equityestimates(notethatnegativevaluesofPREDFERRindicateforecastoptimism).10Overall,weareunabletofindevidencesuggestingthatICWisdirectlyassociatedwithimpliedcostofequityasofJune2004.Thehighercostofequityreportedintheunivariate

910Sincemarketvalueofequityisariskproxy(e.g.,FamaandFrench1996),includingmarketvalueofequityintheregressionmaycapturevariationinriskattributabletoSection404weakness.Accordingly,inasensitivitytestwereplacelog(marketvalue)withlog(totalassets)andfindqualitativelysimilarresults.

Someofthesignsonthecoefficientsintheregressionsareoppositefromourexpectations(e.g.,positive(negative)coefficientsonLOGMKTV(RZSCORE)fortheVLandEASTregressions).Weruleoutmulticol-linearityasareasonfortheseresults—themaximumvarianceinflationfactor(nottabulated)doesnotexceed1.4inanyofourmodels.Also,wefindthatinmostcases,theanomalousrelationisalsopresentattheunivariatelevel,suggestingthattheanomalousrelationsareapeculiarityoftheSection404sample.Forexample,theunivariatecorrelationsbetweenVLandLOGMKTV(RZSCORE)arenegative(positive)whenweusetheentireCompustat-ValueLineuniversetocomputecorrelationcoefficients.

TheAccountingReview,October2007

1272Ogneva,Subramanyam,andRaghunandan

analyses,withtheexceptionofVL,appearstobeattributabletoprimitivefirmcharacter-isticsthatareassociatedwithICW.11WeanalyzethedifferentialVLresultsindetailinalatersection.

AlternativeEstimationPeriod

Ourprimaryanalyses,whichareconductedpriortotheSection404disclosures(June2004),ignorethedisclosureeffectsofICW,ifany,onstockprices.Toensurethatourresultsarenotsensitivetothechoiceofperiodunderanalysis,wealsoreplicateourprimarytestsasofJune2006.SinceoursamplecomprisesSection404disclosuresuptoJanuary2006,examiningcostofequityasofJune2006ensuresthatallpotentialdisclosureeffectsareincorporatedinourcost-of-equityestimates.

Table3reportsresultsoftheJune2006analyses.OursamplefortheJune2006analysescomprises2,357firms(311ICWand2,046controlfirms),althoughtheactualnumberofobservationsusedissmallerbasedonavailabilityofcostofequitydataforouralternativemeasuresaswellasdeletionofoutliers.Wereportbothunivariate(PanelA)andmultivar-iateregression(PanelB)analyses.Forourcross-sectionalmultivariateregressions,weem-ploythefollowingmodel(wheresubscriptidenotesfirm):

COECiϭ␦ϩ␦0*WEAKiϩ␦1*REMEDiϩ␦2*LOGSEGMENTSiϩ␦3*FOREIGNiϩ␦4*M&Aiϩ␦5*RESTRUCTUREiϩ␦6*SALEGRWiϩ␦7*INVENTORYiϩ␦8*LOGMKTViϩ␦9*LOSSiϩ␦10*RZSCOREiϩ␦11*LOGAGEiϩ␦12*PREDFERRiϩ␧i,

(3)

where:

REMEDϭindicatorvariablethatequals1ifafirmreportsanadverseSection404report

initsfirstyearbutreportsacleanreportsubsequently,and0otherwise.Allothervariablesareasdefinedearlier,althoughtheestimationperiodsaredifferenttocorrespondtothe2006timeperiod.Equation(3)isidenticaltoEquation(2)butfortheinclusionofREMED,whichisanindicatorvariablethatswitchesonforICWfirmsthatsubsequentlyremediatetheirinternalcontrolweaknesses.Weincludeacontrolforreme-diationbecausesomeofoursamplefirmshadreportedtheirsecondyearofSection404disclosuresbyJune2006andAshbaugh-Skaifeetal.(2006)arguethatcostofequitydeclinesforfirmsthatremediatetheirinternalcontroldeficiencies.

TheresultsfromourJune2006analysesarequalitativelysimilartothosefromourJune2004analyses.Thatis,whilemeanandmediancostofequityarehigherforICWfirmsforsomeofourcost-of-equityproxies,theWEAKcoefficient(whichexaminesdif-ferencesaftercontrollingforfactorsassociatedwithICW)isinsignificantatconventionallevels.Interestingly,thereislittleevidencethatfirmsthatremediatedtheirinternalcontrol

11Toseewhetherour‘‘non-result’’inthemultipleregressionsisduetolowpower,weexaminethepowerofourregressions.Wefindthattheprobabilityofrejectingthenullof‘‘nodifference’’whenthe‘‘true’’differenceincostofequityis50(100)basispointsis0.93(1.00),1.00(1.00),0.99(1.00),0.59(0.99),0.74(1.00),0.14(0.42),and0.13(0.39)fortheregressionswithOJ,GLS,CT,PEG,GM,VL,andEAST,respectively,asthedependentvariables.Thissuggeststhat,withtheexceptionoftheVLandEASTproxies,ourregressionsrejectthenullwithalmostcertaintyfora100basis-pointdifferenceandwithhighprobabilityevenfora50basis-pointsdifference.Therefore,lowpowerisunlikelythecauseofour‘‘non-result.’’

TheAccountingReview,October2007

InternalControlTABLE3

DifferencesacrossWeaknessandControlFirmsforVariousImpliedCostofEquityMeasures:June2006

OJ

GLSCTPEGGMVLEASTPanelA:MeanandMedianDifferencesbetweenWeaknessandControlFirms

SampleSize

1,910ObservationsUsed

1,872ICW

0.1111MeansControlSample

0.1071ICW–Control(0.02)0.0040ICW

0.1077MediansControlSample

0.1051ICW–Control

(0.02)

0.0026PanelB:MultipleRegressionAnalysisIntercept0.1052

(0.00)WEAK0.0014(0.47)TheREMED0.0037Accounting(0.29);LOGMKTVϪ0.0006(0.11)RZSCOREϪ0.0007(0.00)Review,LOGSEGMENTS0.0009(0.28)OctoberFOREIGN0.0007(0.58)SALEGRW0.0083(0.00)

20072,0641,9432,0241,9050.08100.08620.08090.0866Ϫ0.0004(0.95)0.0001(0.76)0.08110.08580.08010.0855(0.97)

0.0010(0.77)

0.00040.04790.0768(0.00)(0.00)Ϫ0.00170.0018(0.17)(0.21)0.0017Ϫ0.0037(0.43)(0.14)Ϫ0.00170.0006(0.00)(0.05)0.00060.0003(0.00)(0.05)0.00180.0006(0.00)(0.35)0.00270.0004(0.00)(0.67)Ϫ0.00030.0028(0.78)

(0.01)

2,1972,0722,1552,0320.12170.12410.10590.01590.11280.0113(0.00)(0.00)0.10630.11610.09550.1053(0.00)

0.0108(0.00)

0.01080.05140.0610(0.00)(0.00)0.00190.0018(0.51)(0.45)0.01310.0063(0.01)(0.16)0.0060Ϫ0.0051(0.00)(0.00)0.00050.0001(0.15)(0.86)0.00200.0009(0.14)(0.45)0.00430.0013(0.04)(0.46)0.00830.0083(0.00)

(0.00)

1,1182,0751,0972,0350.11550.10860.10930.00620.1120Ϫ0.0034(0.34)(0.68)0.11500.10790.11000.1128Ϫ0.0048(0.32)

0.0050(0.56)

0.17160.1465(0.00)(0.00)0.0184Ϫ0.0045(0.02)(0.64)Ϫ0.03360.0182(0.01)(0.31)0.0063Ϫ0.0005(0.00)(0.83)Ϫ0.0032Ϫ0.0047(0.00)(0.00)Ϫ0.00590.0055(0.02)(0.22)Ϫ0.0077Ϫ0.0029(0.05)(0.68)0.00020.0139(0.98)

(0.09)

(continuedonnextpage)

WeaknessandCostofEquity:EvidencefromSOXSection404Disclosures1273ϪϪϪTheAccountingTABLE3(continued)

Review,OJ

GLSCTPEGGMVL

EAST

INVENTORY0.01820.03190.02240.06570.0287Ϫ0.0094

Ϫ0.0456

October(0.00)(0.00)(0.00)(0.00)(0.00)(0.51)(0.10)M&A

Ϫ0.0013Ϫ0.0011Ϫ0.00240.0018Ϫ0.00060.0056Ϫ0.0013(0.26)(0.14)(0.01)(0.34)(0.71)(0.18)(0.84)2007RESTRUCTUREϪ0.00230.0027Ϫ0.00170.00480.00500.00490.0091(0.04)(0.00)(0.05)(0.01)(0.00)(0.17)(0.14)LOSS0.00910.00440.00510.05810.02600.0247Ϫ0.0129(0.00)(0.00)(0.00)(0.00)(0.00)(0.00)(0.24)LOGAGEϪ0.0027Ϫ0.0005Ϫ0.0005Ϫ0.00010.0011Ϫ0.0120Ϫ0.0094(0.00)(0.31)(0.38)(0.93)(0.26)(0.00)(0.02)PREDFERRϪ0.4665Ϫ1.0952Ϫ0.89840.1743Ϫ0.5548Ϫ1.4692Ϫ0.6082(0.00)(0.00)(0.00)(0.01)(0.00)(0.00)(0.11)AdjustedR29.18%

40.48%

24.26%

30.23%

15.46%

8.89%1.02%

ThetablecomparesimpliedcostofcapitalestimatesasofJune2006forasampleof2,357firmsthatfiledfirst-timeSection404opinionsduringtheNovember2004toJanuary2006period.Ofthe2,357firms,311filedadversereports(ICWfirms)and2,046filedcleanreports(controlfirms).Samplesizevariesdependingonthecost-of-equityproxyused.Cost-of-equityproxiesare:OJ:ameasurebasedontheOhlsonandJuettner-Nauroth(2005)modelincorporatinganalysts’long-termearningsgrowthforecasts;GLS:theindustryROEmethoddevelopedbyGebhardtetal.(2001);CT:theeconomy-widegrowthmethodofClausandThomas(2001);GM:theunrestrictedabnormalearningsgrowthmethodofGodeandMohanram(2003);PEG:therestrictedabnormalearningsgrowth(orPEG)methoddevelopedbyEaston(2004);EAST:costofequityderivedfromourmodifiedversionofsimultaneouscost-of-equityandgrowthcomputation;andVL:expectedrateofreturnaccordingtotheValueLineInvestmentSurvey.Detailsontheestimationofallcost-of-equitymeasuresareprovidedinAppendixA.

PanelAcontainsmeanandmediancost-of-equitycomparisons.Observationswithvaluesfallingintothetopandbottomextremepercentilesoftherespectivevariablearedeleted.‘‘Samplesize’’indicatesthesizeofthesamplebeforetruncationand‘‘Observationsused’’aftertruncation.Two-sidedp-valuesforcomparisonswithcontrolsamplearereportedinparentheses:p-valuesformeancomparisonsarebasedont-statisticsandp-valuesformediancomparisonsarebasedonaWilcoxonsignedranktest.

(continuedonnextpage)

1274Ogneva,Subramanyam,andRaghunandanTABLE3(continued)

PanelBreportsresultsforthefollowingregression:

COECiϭ␦ϩ␦0*WEAKiϩ␦1*REMEDiϩ␦2*LOGSEGMENTSiϩ␦3*FOREIGNiϩ␦4*M&Aiϩ␦5*RESTRUCTUREiϩ␦6*SALEGRWiϩ␦7*INVENTORYiϩ␦8*LOGMKTViϩ␦9*LOSSiϩ␦10*RZSCOREiϩ␦11*LOGAGEiϩ␦12*PREDFERRiϩ␧

where:COECϭoneofthesevencost-of-equityestimatesdefinedabove;

WEAKϭanindicatorvariablethattakesonavalueof1ifthefirmdisclosesanICW,and0otherwise;

REMEDϭanindicatorvariablethattakesonavalueof1ifthefirmdisclosedanICWin2004andreportednoICWin2006,and0otherwise;andPREDFERR

ϭ

averageofone-year-aheadandtwo-year-aheadpredictedforecasterrors,computedusingtheLiuandSu(2005)model.

TheFortheValueLinecost-of-equityanalysis,PREDFERRϭone-year-aheadpredictedforecasterrorestimatedusingValueLinedata(seeAppendixBfordetails).AllAccountingothercontrolvariablesareasdefinedinTable1.

Wedelete1percentoftheobservationswiththesmallestandthelargestvaluesofstudentizedresidualstoremovetheeffectofoutliers.Two-sidedp-valuesarereportedinparentheses.

Review,October2007InternalControlWeaknessandCostofEquity:EvidencefromSOXSection404Disclosures12751276Ogneva,Subramanyam,andRaghunandan

weaknesseshavelowercostofequity.TheVLresultscontinuetobedifferent:thereisevidencethattheVLcostofequityishigherforICWfirmsthanforcontrolfirms,butthisdifferenceismorethancompensatedforthosefirmsthatsubsequentlyremediatetheweak-nesses.OurVLresultsareconsistentwiththoseofAshbaugh-Skaifeetal(2006).12AnalysisbyTypeofWeakness

Finally,weexaminewhethertherearedifferencesinourresultswhenwepartitionoursamplebytypeofweakness.Doyleetal.(2007a)suggestthatcertaintypesofICWsaremoreegregiousthanothers.Accordingly,weusetheclassificationsprovidedbyAuditAnalyticsandexaminewhetherdifferenttypesofICWshavehighercostsofequitythanourcontrolfirms.WegroupICWsintoseveral(overlapping)subclassificationsincludingpresenceofmultipleweaknesses;presenceofpersonnelproblems;presenceofgovernanceissues;presenceofaccounting/informationsystems-relatedproblems;presenceofsystemicproblems;presenceofaccountingrestatements;andlatefilers,i.e.,acceleratedfilersthatdelayedfilingtheirSection404reportsforatleastsixmonthsfollowingthefiscalyear-end.Whilewefindthatsomeofthesepartitionshavehighercostsofequitythanourcontrolfirmsinunivariatetests,noneofthepartitionshavesignificantlydifferentcostsofequityinmultipleregressions,bothasofJune2004andJune2006.13Forbrevity,wedonottabulatetheseresults.

V.ADDITIONALANALYSES

Inthissectionwereportresultsofadditionalanalysesandsensitivitytestsaimedatincreasingtheconfidenceinourmainempiricalresults.Wefirstevaluatetheconstructvalidityofourcost-of-equityproxiesthroughfuturereturnsassociationandmeasurementerroranalyses.Next,weexaminewhetherseveralintermediateriskattributes(proxiesforearningsquality,informationasymmetry,andsystematicrisk)areassociatedwithICW.CorroboratingevidenceusingriskattributesthataretheoreticallyexpectedtobeassociatedwithICWshouldprovideassurancethatourprimaryresultsarenotdrivenbymeasurementerrorinourcost-of-equityproxies.Finally,weexplorewhyourVLresultsaredifferentfromthoseusingotherproxies.

TestsofConstructValidity

Theultimatetestofconstructvalidityofacost-of-equityproxyisitsassociationwithfutureexpectedreturnsonthefirm’sstock.EastonandMonahan(2005)(hereafter,E&M)proposeamethodthatevaluatestheassociationofacost-of-equityproxywithfutureex-pectedreturnsthrougharegressionoffuturerealizedreturnsonthecost-of-equityproxyaftercontrollingforfuturecashflowanddiscountrateshocks.Aftercontrollingforcashflowanddiscountratenews,E&Mfindthatmostcost-of-equityproxiesarenegativelycorrelatedwithrealizedstockreturnsandhavesignificantmeasurementerror.Accordingly,theyconcludethatmostcost-of-equityproxies(withtheexceptionofCTundercertainconditions)lackconstructvalidity.

Toevaluateconstructvalidityofoursevencost-of-equityproxies,wereplicatetheE&Manalysesforoursample(asofJune2004).TheE&Manalysisprimarilyinvolvesregressing

1213WealsoexaminechangesincostofequitybetweenJune2004andJune2006anddonotfindanyevidencethatcostofequityincreasedoverthisperiodfortheICWfirms(withtheexceptionoftheVLmeasure).

HribarandJenkins(2004)findanincreaseincostofequityafterfirmsannouncerestatements.OurresultslikelydifferfromtheirsbecauseweexamineonlyICWfirms(i.e.,thosereportingaSection404weakness)thatalsorestatetheirfinancialstatements,ratherthanrestatementsingeneral.

TheAccountingReview,October2007

InternalControlWeaknessandCostofEquity:EvidencefromSOXSection404Disclosures1277

thelogofone-year-aheadstockreturnsonthelogoftherespectivecost-of-equitymeasure(proxyforexpectedreturn)andthelogofproxiesforcashflownewsanddiscountratenewsoverthefollowingyear.Table4reportsresultsoftheseregressions,whereeachrowcorrespondstooneofthealternativecost-of-equityproxies.Acoefficientontheexpectedreturnproxy(LOG(ER))closetoϩ1wouldindicatethatthecost-of-equityproxymeasuresexpectedreturnaccurately,i.e.,hashighconstructvalidity.WiththeexceptionoftheCTmeasure,wefindthatthecoefficientsonLOG(ER)areeithernegativeandsignificantorinsignificant.WhenCTisusedasaproxyforexpectedreturn,thecoefficientisϩ0.88,whichissignificantlypositive(pϽ0.01)andstatisticallyindistinguishablefromthetheo-reticalvalueofϩ1(pϭ0.74).

E&Malsoconductfurthertestsofmeasurementerrorbymodifyingtechniquesdevel-opedbyGarberandKlepper(1980).Theyconstructtwostatisticsfortheextentofmea-surementerror,onethatcontrolsforcorrelationinthemeasurementerrorsacrossthethreevariablesintheregressionandonethatdoesnot.InColumn8ofTable4wereportthestatisticthatcontrolsforcorrelationinmeasurementerror,asitisbetterspecified.WefindthatallproxieshavestatisticallysignificantmeasurementerrorwiththeexceptionofCT,whichalsohasthesmallestmeasurementerrorcoefficient.

Tosummarize,ourtestsofconstructvaliditysuggestthefollowing.First,withtheexceptionofCT,ourcost-of-equityproxieshavelowconstructvaliditybecauseofsignifi-cantmeasurementerror.Second,theCTproxyhashighconstructvalidity:itsregressioncoefficientonfutureexpectedreturnsisindistinguishablefromthetheoreticalvalueofϩ1anditalsohasthelowestmeasurementerror(whichisinsignificant).ResultsfromusingtheCTmethodareconsistentwithourprimaryinference,i.e.,costofequityforICWfirmsisnothigherthanthatforcontrolfirms.Therefore,theconstructvaliditytestsprovidesomeassurancethatourconclusionsarenotdrivenbymeasurementerrorinourcost-of-equityproxies.

InternalControlWeaknessandRiskAttributes

Tocorroborateourcost-of-equityresults,weexploredifferencesbetweenICWandcontrolfirmsinunderlyingriskattributesthattheorysuggestsshouldbeassociatedwithICW.InSectionII,wehypothesizeatwo-steplinkbetweenICWandcostofequity.First,werelateICWtoaccountingquality.Accordingly,wefirstexaminewhetherICWisas-sociatedwithearningsqualityproxies,suchassignedandabsolutediscretionaryaccruals.Second,wehypothesizethatinformation(accounting)qualitycanaffectcostofequitythroughinformationasymmetry(EasleyandO’Hara2004)orthroughitseffectonconven-tionalriskmeasures,suchasbeta(Lambertetal.2007).Accordingly,wenextexaminetheassociationbetweenICWandtheprobabilityofinformedtrading(PIN)ormarketriskmeasures(betaandidiosyncraticrisk).WeconductallourriskattributetestsasofJune2004.

Doyleetal.(2007b),Ashbaugh-Skaifeetal.(2007),andHoganandWilkins(2005)reportthatabsolutediscretionaryaccrualsarehigherforasampleoffirmsdisclosingma-terialweaknessesunderSections302and404.Accordingly,wefirstexaminedifferencesinbothsignedandabsoluteJones(1991)modelresiduals(discretionaryaccruals)betweenourICWandcontrolfirms.Becausediscretionaryaccrualsareaffectedbyfirmperformance,wealsoexaminedifferencesinsignedandabsoluteperformance-matcheddiscretionaryaccruals(Kotharietal.2004).

EasleyandO’Hara(2004)suggestthatinformationriskaffectssystematicriskthroughinformationasymmetry.Inacompanionpaper(Easleyetal.2002),theytesttheirtheory

TheAccountingReview,October2007

1278Ogneva,Subramanyam,andRaghunandan

TABLE4

ConstructValidityTestsofCostofEquityProxies:June2004

FutureReturnsRegression

Proxy(1)OJGLSCTPEGGMVLEAST

n(2)162917771698185416999071701

Intercept(3)0.1023(0.00)0.2403(0.00)0.0349(0.18)0.2267(0.00)0.1252(0.00)0.2234(0.00)0.1174(0.00)

LOGER(4)0.0557(0.86)Ϫ1.7012(0.00)0.8882(0.01)Ϫ1.3904(0.00)Ϫ0.1529(0.49)Ϫ0.8973(0.00)Ϫ0.0279(0.73)

LOGCN(5)1.1337(0.00)0.8805(0.00)0.9581(0.00)0.5399(0.00)0.9810(0.00)0.8086(0.00)1.0496(0.00)

LOGRN(6)0.0117(0.05)0.2010(0.00)0.0432(0.00)0.0372(0.00)0.0217(0.00)0.0333(0.00)Ϫ0.0035(0.00)

R2%(7)29.8447.5033.0931.8026.5840.5727.30

ModifiedNoiseCoefficient

(8)0.0011(0.00)0.0002(0.06)0.0001(0.27)0.0050(0.00)0.0020(0.00)0.0043(0.00)0.0093(0.00)

Thetablereportstheresultsofreliabilityevaluationforthecost-of-equitymeasuresusingtheEastonand

Monahan(2005)method.Thesampleconsistsof2,515firmsthatfiledfirst-timeSection404opinionsduringtheperiodNovember2004toJanuary2006.Theactualnumberofobservationsusedineachtest(n)islowerbecauseofadditionaldatarequiredtoimplementtheEaston-Monahananalyses.Cost-of-equityestimatesaredefinedinTable2.Columns(3)to(7)containregressioncoefficientsandR2fortheregressionsofrealized

returnsontheproxiesforexpectedreturn,cashflownews,andexpectedreturnnews:LOGRETi,tϩ1ϭInterceptϩ␣1*LOGERi,tϩ␣2*LOGCNi,tϩ1ϩ␣3*LOGRNi,tϩ1ϩ␧i,whereLOGRETi,tϩ1istherealizedreturnovertheoneyearafterthecost-of-equityestimation,LOGERiistheexpectedreturn(oneofthecost-of-equityestimates)asofJune2004,LOGCNi,tϩ1isthecashflownewsmeasuredovertheoneyearafterthecost-of-equityestimation,andLOGRNi,tϩ1isthediscountratenewsovertheoneyearafterthecost-of-equity

estimation.Allreturnmeasuresarecontinuouslycompounded.Column(8)reportsamodifiednoisecoefficientforeachcost-of-equitymeasure.

Cashflownewsismeasuredasasumoftherevisioninone-year-aheadforecastedROEandthecapitalizedrevisioninthetwo-year-aheadforecastedROE:LOGCNi,tϩ1ϭ⌬LOGFROEi,tϩ1ϩ␳/(1Ϫ␳␻)*⌬LOGFROEi,tϩ2.RevisionsrefertochangesinforecastsfromJune2004toJune2005.(AllanalysesarereplicatedusingValueLineanalystreportissuedatestomeasurerealizedreturnsandValueLineone-year-aheadearningsforecastsandlong-termearningsgrowthrateforecaststoconstructcashflownewsproxies.Theestimationperiodthereforestartsinthesecondquarterof2004andendsinthesecondquarterof2005.Two-year-aheadearningsforecastsareconstructedbyapplyinglong-termgrowthratestoone-year-aheadearningsforecasts.TheresultsaresimilartothoseusingI/B/E/Sdata.)

ForecastedROEisdefinedasEPSforecastdividedbybookvalueofequity(Compustatitem#60)dividedbynumberofsharesusedtocalculateEPS(Compustatitem#54).Weuse␳estimatesreportedinEastonand

Monahan(2005).Persistencecoefficients␻tareestimatedthroughapooledtime-seriescross-sectionalregressionforeachofthe48Fama-Frenchindustries:LOGROEi,tϪ␶ϭ␻0tϩ␻tϫLOGROEi,tϪ(␶Ϫ1),where␶isanumberbetween0and9,andROEisreturnonequity.

DiscountratenewsismeasuredasLOGRNi,tϩ1ϭ␳/(1Ϫ␳)*(LOGER1,tϩ1ϪLOGERi,t)whereLOGERi,tisthecontinuouslycompoundedcost-of-equityestimatemeasuredasofJune2004,andLOGERi,tϩ1isthecontinuouslycompoundedcost-of-equityestimatemeasuredasofJune2005.

ThedetailsofestimatingmodifiednoisecoefficientaredescribedinEastonandMonahan(2005,506–507).Allestimationsareperformedafterdeletingobservationsthatfallinthetopandbottom0.5percentforLOGRETi,tϩ1,LOGERi,LOGCNi,orLOGRNidistributions.

TheAccountingReview,October2007

InternalControlWeaknessandCostofEquity:EvidencefromSOXSection404Disclosures1279

byshowingthatPIN—ameasureofinformationasymmetry—increasesininformationrisk.Accordingly,weexaminewhethertherearedifferencesinthePINmeasurebetweenourICWandcontrolfirms.

Finally,Lambertetal.(2007)arguethatinformationriskispricedthroughconventionalmarketriskfactors,suchasbeta.Therefore,weexaminewhethertwomeasuresofmarketrisk,betaandidiosyncraticrisk(measuredasabnormalreturnvolatility),aredifferentacrossourweaknessandcontrolsamplepartitions.WhilebetaisadeterminantofexpectedreturnsunderCAPM,thereisempiricalevidencesuggestingthatidiosyncraticvolatilitymaybepricedbythestockmarket(GoyalandSanta-Clara2003).However,itmustbenotedthatmuchofidiosyncraticvolatilityisdiversifiableandthusidiosyncraticvolatilitymaybeanimperfectindicatorofsystematicrisk,andinturn,costofequity.

UnivariateResults

Table5,PanelAreportsmeanandmedianvaluesofvariousriskattributesfortheICWandcontrolfirms,alongwiththestatisticalsignificanceofthedifferences.Weexaminefourmeasuresofdiscretionaryaccruals:cross-sectionalJones-modelresiduals(DACC),absolutevalueoftheseresiduals(ADACC),performance-matchedJones-modelresiduals(PMDACC),andabsolutevalueofperformance-matchedresiduals(APMDACC).WefindthatmeanandmedianabsolutediscretionaryaccrualsarehigherforICWfirms(althoughthedifferencesaremarginallysignificantforAPMDACC)buttherearenosignificantdiffer-encesacrossICWandcontrolfirmsforanyofthesigneddiscretionaryaccruals’measures.Next,weexaminetheprobabilityofinformedtrading(PIN),beta(BETA),andabnormalreturnvolatility(RETVAR).WhileICWfirmshavehigherinformationasymmetry(PIN),thedifferencesarenotstatisticallysignificantatconventionallevels.ConsistentwithBryanandLilien(2005),averagemarketriskattributesarehigherforICWfirms(althoughthemediandifferenceinbetaisinsignificant).

Overall,wefindevidenceofloweraccrualqualityforICWfirms,suggestinghigheraccountingrisk.Thehigheraccountingrisktranslatesintomarginallyhigherinformationasymmetryandhighermarketmeasuresofsystematicrisk.

ControllingforFirmCharacteristics

ItisimportanttoverifywhetherICWcontinuestobedirectlyassociatedwithriskattributesaftercontrollingforprimitivefirmcharacteristicsthatareassociatedwithICW.Accordingly,weestimatethefollowingcross-sectionalOLSregressions(wheresubscriptidenotesfirm):

RISK-ATTRIBUTEiϭ␥ϩ␥0*WEAKiϩ␥1*LOGSEGMENTSiϩ␥2*FOREIGNiϩ␥3*M&Aiϩ␥4*RESTRUCTUREiϩ␥5*SALEGRWiϩ␥6*INVENTORYiϩ␥7*LOGMKTViϩ␥8*LOSSiϩ␥9*RZSCOREiϩ␥10*LOGAGEiϩεi,

(4)

where:

RISK-ATTRIBUTEϭoneofthesevenriskattributesexamined.

Allothervariablesareasdefinedearlier.

Table5,PanelBreportsresultsforthesesevenregressions.WefindthatWEAKisnotsignificantinexplainingeithersignedorabsolutediscretionaryaccruals,suggestingthat

TheAccountingReview,October2007

TheAccounting1280TABLE5

DifferencesacrossWeaknessandControlFirmsforVariousRiskAttributes:June2004

DACC

PMDACCADACCAPMDACCPINBETARETVARPanelA:MeanandMedianDifferencesbetweenWeaknessandControlFirms

Review,ObservationsUsed

OctoberICW

MeansControlICW–ControlSample

2007ICW

MediansControlICW–ControlSample

PanelB:MultipleRegressionInterceptWEAKLOGMKTVRZSCORELOGSEGMENTSFOREIGNSALEGRWINVENTORY

2,0632,0630.015Ϫ0.0070.013Ϫ0.0060.001Ϫ0.001(0.78)(0.84)0.0200.0050.016Ϫ0.0030.0040.008(0.41)

(0.60)

0.0088

Ϫ0.0466

(0.42)(0.01)0.0026Ϫ0.0005(0.56)(0.95)0.0029Ϫ0.0027(0.01)(0.14)0.00150.0025(0.03)(0.02)0.00050.0011(0.84)(0.77)0.0097Ϫ0.0015(0.01)(0.80)0.00120.0093(0.77)(0.14)0.00170.0222(0.90)

(0.27)

2,0632,0630.0630.0540.0860.0790.0090.007(0.01)(0.16)0.0440.0400.0590.0520.0040.007(0.03)

(0.13)

0.04090.1046(0.00)(0.00)0.00390.0066(0.27)(0.22)Ϫ0.0030Ϫ0.0009(0.00)(0.53)0.0009Ϫ0.0020(0.09)(0.02)Ϫ0.0036Ϫ0.0089(0.06)(0.00)Ϫ0.0006Ϫ0.0031(0.84)(0.49)0.02100.0254(0.00)(0.00)0.02750.0107(0.01)(0.49)

2,4802,4652,4650.1360.1311.3871.2410.0100.0070.0050.1460.003(0.28)(0.06)(0.00)0.1260.1241.1071.0820.0070.0040.0030.0250.003(0.36)

(0.20)

(0.00)

0.0416

0.3006Ϫ0.0019

(0.00)(0.11)(0.09)0.0054Ϫ0.03550.0009(0.21)(0.64)(0.04)0.0221Ϫ0.0603Ϫ0.0015(0.00)(0.00)(0.00)0.00330.0241Ϫ0.0003(0.00)(0.02)(0.00)0.0073Ϫ0.1296Ϫ0.0009(0.00)(0.00)(0.00)0.00120.2362Ϫ0.0002(0.76)(0.00)(0.65)0.01640.25950.0019(0.00)(0.00)(0.00)0.00911.06050.0051(0.47)

(0.00)

(0.00)

(continuedonnextpage)

Ogneva,Subramanyam,andRaghunandanϪϪϪϪϪϪϪϪϪϪTABLE5(continued)

M&A

0.0055Ϫ0.0061

0.0006Ϫ0.0047

Ϫ0.0049

0.1397Ϫ0.0003

(0.10)(0.24)(0.83)(0.24)(0.12)(0.01)(0.32)RESTRUCTURE0.00000.0009Ϫ0.0027Ϫ0.0032Ϫ0.00300.27400.0015(0.99)(0.86)(0.29)(0.42)(0.35)(0.00)(0.00)LOSSϪ0.0103Ϫ0.00660.01360.0231Ϫ0.03810.48510.0083(0.01)(0.29)(0.00)(0.00)(0.00)(0.00)(0.00)LOGAGE0.00280.0048Ϫ0.0070Ϫ0.00940.0084Ϫ0.0019Ϫ0.0007(0.13)(0.09)(0.00)(0.00)(0.00)(0.95)(0.00)ObservationsUsed2,0632,0632,0632,0632,4552,4652,465AdjustedR21.17%

0.16%

8.58%

5.81%

17.59%

7.51%

30.67%

ThetablecomparesimpliedcostofcapitalestimatesasofJune2004forasampleof2,515firmsthatfiledfirst-timeSection404opinionsduringtheNovember2004toJanuary2006period.Ofthe2,515firms,346filedadversereports(ICWfirms)and2,169filedcleanreports(controlfirms).Fordiscretionaryaccrualmeasures

(PIN)thenumberoffirmsinICW/controlsamplesis290/1,815(346/2,159),respectively.Theriskattributesaredefinedas:DACCisdiscretionaryaccruals,estimatedasaresidualfromacross-sectionalregressionrunforeachtwo-digitSICindustryACCiϭ␤0ϩ␤1*1/TAiϩ␤2*(⌬SALESiϪ⌬ARi)ϩ␤(Compustatitem#123)minuscashflowfromoperations(Compustatitem#308);3*PPEi⌬ϩSales␧i,whereACCistotalaccrualsestimatedasincomebeforeextraordinaryitemsischangeinsales(Compustatitem#12);⌬ARischangeinaccountsreceivable(Compustatitem#2);andPPEisproperty,plant,andequipment(Compustatitem#7).Allvariablesarescaledbybeginning-of-periodtotalassets.Aminimumoftenobservationsisrequiredforeachregressionandobservationswithanabsolutevalueofscaledaccrualsgreaterthan1areexcludedfromtheanalysis.PMDACCisperformance-matcheddiscretionaryaccruals,basedontheKotharietal.(2005)model,andiscalculatedasDACCforthesamplefirmminusDACCforthematchedcontrolfirm,wherethematchedcontrolfirmhastheclosestROAwithinthesametwo-digitSICindustry.ADACCandAPMDACCareabsolutevaluesofDACCandPMDACC,respectively.PINistheprobabilityofinformedtrading,estimatedusingtradingTheinformationfromtheTAQ2004dataset.ThemethodforestimationisdescribedinEasleyetal.(2001).BETAistheCAPMbeta,estimatedasaslopefromregressingmonthlyexcessstockreturnonmonthlyexcessmarketreturn,using12monthsofreturnspriortoJune2004.RETVARisthevarianceofresidualsfromaregressionofAccountingmonthlystockexcessreturnsonthemarketexcessreturnandtwoFamaandFrench(1996)factor-mimickingportfolioreturnsoverthe12monthspriortoJune2004.PanelAcontainsunivariatecomparisonsofmeansandmedians.Observationswithvaluesfallingintothetopandbottomextremepercentilesoftherespectivevariablearedeleted.‘‘ObservationsUsed’’indicatesthesizeofthesampleaftertruncation.Two-sidedp-valuesforcomparisonswithcontrolsamplearereportedinparentheses:p-valuesformeancomparisonsarebasedont-statistics,andp-valuesformediancomparisonsarebasedonaWilcoxonsignedranktest.PanelBreportsresultsforRegression(4):

Review,RISK-FACTORiϭ␥ϩ␥0*WEAKiϩ␥1*LOGSEGMENTSiϩ␥2*FOREIGNiϩ␥3*M&Aiϩ␥4*RESTRUCTUREiϩ␥5*SALEGRWiϩ␥6*INVENTORYiϩ␥7*LOGMKTViϩ␥8*LOSSiϩ␥9*RZSCOREiϩ␥10*LOGAGEiϩ␧i.

OctoberThesevenriskattributesaredefinedabove.WEAKisanindicatorvariablethattakesonavalueof1ifthefirmdisclosesanICW,and0otherwise.AllcontrolvariablesareasdefinedinTable1.

Two-sidedp-valuesarereportedinparentheses.Toavoidspuriouseffectsduetooutliers,weexcludeallobservationswithinthetopandbottom1percentof2007studentizedresiduals.

InternalControlWeaknessandCostofEquity:EvidencefromSOXSection404Disclosures12811282Ogneva,Subramanyam,andRaghunandan

thereisnodirectassociationbetweenICWandaccountingquality.ThisresultimpliesthatwhileICWsmayincreaseaccountingrisk,i.e.,thelikelihoodofsignificantaccountingqualityproblemsinthefuture,thereisnoevidenceofalready-realizedpoorerearningsquality(asmeasuredbyourdiscretionaryaccruals’proxies)associatedwithICWsreportedunderSection404.14WealsofindthatWEAKisinsignificantinregressionswithPINandBETAasdependentvariables.Theseresultssuggestthatriskdriversidentifiedbytheoryasbeingaffectedbyinformationrisk(accountingquality)arenotdirectlyassociatedwithICWreportedunderSection404.OnlyidiosyncraticvolatilityisstillpositivelyassociatedwithICWaftercontrollingforfirmcharacteristics.However,itisdebatablewhetheridiosyncraticvolatilityissystematicandhencepricedinthefirms’costofequity.

Overall,wefindlittleevidencethatvariousriskattributesthataretheoreticallyexpectedtobeassociatedwithICWaredifferentbetweenourICWandcontrolfirmsaftercontrollingforprimitivefirmcharacteristics.Thus,ourriskattributeresultslargelycorroborateourprimaryfindingsrelatingtoimpliedcostofequity.

ExploringtheDifferentValueLineResults

WegenerallyfindnodifferenceincostofequitybetweenICWandcontrolfirms,withtheexceptionofVL,whichissignificantlyhigherforICWfirmsthanforcontrolfirms(asofJune2004)evenaftercontrollingforfirmcharacteristicsandanalystforecastbias.The2004VLresultscontrastwiththoseusingallothercost-of-equityproxies.However,theseresultsareconsistentwiththosereportedbyAshbaugh-Skaifeetal.(2006),whofindhigherVLforfirmsreportinginternalcontroldeficienciesunderSections302and404.Inthissection,weexplorethereasonsforthedifferentialresultsarisingbetweentheValueLineandotherproxiesforcostofequity.

Ashbaugh-Skaifeetal.(2006)justifytheirchoiceofVLbynotingthat(1)VLisa‘‘valid’’measureofcostofequitysinceitsatisfiestherationalexpectationstestsofGuayetal.(2005),and(2)theVLsampledoesnotsufferfromthepotentiallyserioussampleselectionbiasthatsamplesusingothermeasures(suchasPEG)havebecausetheyrequirepositiveand/orincreasingearningsforecasts.Accordingly,wefocusourinquiryoncom-paringtheVLandothermeasureswithrespecttoboththeirconstructvalidityandsamplingbiases.

Forjustifyingconstructvalidity,Ashbaugh-Skaifeetal.(2006)verifythatVLmeetstherationalexpectationsrequirementsofGuayetal.(2005).However,Guayetal.reportthatanumberofothermeasures,suchasCTandGLS(andundercertainconditions,PEGandGM),alsosatisfytherationalexpectationsrequirementsintermsofpositivecorrelationwithfuturestockreturns.Itisimportanttonotethatthesemeasuresshowapositivecor-relationwithfuturestockreturnsonlyafteradjustingforanalystforecastbiasandforecastsluggishness,adjustmentsthatwe(orAshbaugh-Skaifeetal.2006)donotmakeinouranalyses.Intheabsenceoftheseadjustments,VLisnegativelycorrelatedwithfuturere-turns,asweshowinTable4.Additionally,VLfailsthemeasurementerrortestsofEastonandMonahan(2005).Therefore,wemaintainthattheVLresultsareunreliablebecausetheVLmeasurelacksconstructvalidity.

14OurresultsrelatingtodiscretionaryaccrualsareinconsistentwiththosereportedbyDoyleetal.(2007b)whodofindevidenceofgreateraccruals’managementforICWfirms.However,thisinconsistencystemsfromthedifferencebetweentheirsampleofmaterialweaknessesreportedunderSections302and404andourrestrictedsampleofICWreportedonlyunderSection404.WereplicateouranalysesusingDoyleetal.’s(2007b)expandedsampleandfindthatunsigneddiscretionaryaccrualsforICWfirmsarehigherthanforthecontrolsampleevenaftercontrollingforprimitivefirmcharacteristics.WethankSarahMcVayforprovidingdataontheirsampleofSection302materialweaknessdisclosures.

TheAccountingReview,October2007

InternalControlWeaknessandCostofEquity:EvidencefromSOXSection404Disclosures1283

Wenextexaminetheissueofsamplingbias.Mostimplied-cost-of-equitymeasuresrequirepositiveorincreasingearningsforecasts.Thisrequirementpotentiallybiasesthesampleagainstfirmsthataredistressedorexperiencingpoorearningsperformance.AsAshbaugh-Skaifeetal.(2006)note,suchsamplingbiascanbeparticularlyimportantinourcontextbecauseICWhasbeenshowntobeassociatedwithpoorperformance(Ashbaugh-Skaifeetal.2007;Doyleetal.2007a).Sucharesearcher-imposedsamplingrequirementisnotnecessaryfortheVLmeasuresincetheVLdatabaseprovidesexpectedratesofreturnsforallfirmsincludedintheirdatabase.However,aswenoteearlier,VLhasseverelyrestrictedcoveragecomparedtoI/B/E/S.Therefore,itispossiblethatVLmayalsohavepotentiallyserioussampleselectionproblemsthatarisebecauseofVL’slimitedcoverage.

TocomparepotentialbiasesintheVLandPEGsamples(wechoosePEGsinceithasthelargestcoverageamongallmeasures),weconductthefollowinganalysis.Weidentifytwosubsamples:(1)TheVLBiassample,whichcomprisesthe1,320firmsforwhichthePEGmeasurecanbeestimatedbutVLdataaremissing;and(2)thePEGBiassample,whichcomprisesthe58firmsforwhichVLdataareavailablebutthePEGcannotbeestimated.Nextwecomparemeanandmedianvaluesforvariousfirmcharacteristicsacrossthetwosubsamples,whicharereportedinTable6.

Attheoutset,wefindthattheVLsampleexcludes242ICWfirms(18.3percent)whilethePEGsampleexcludesonlyoneICWfirm(1.7percent),suggestingthatitistheVLsamplethatissignificantlymorebiasedagainstICWfirms.Inaddition,ourresultssuggestthatitistheVLsample,ratherthanthePEGsample,thatismorelikelytoexcludedis-tressedfirms.Forexample,firmsnotincludedintheVLsamplehavesignificantlyhigherbankruptcyprobabilityrankandmuchhigherincidenceoflossesthanthoseexcludedfromthePEGsample.Overall,thereisnoevidencesuggestingthatthePEGsampleisrelativelymorebiasedthantheVLsampleintermsofexcludingdistressedand/orpoorlyperformingfirms.15Table6alsoidentifiesothersignificantbiasesintheVLsample.ThefirmsthatVLexcludesvis-a`-visthosethatPEGexcludesaresignificantlysmaller,lessdiversified,younger,andmorelikelytobefromhigh-techindustries.Also,theVLsampleappearstoexcluderiskierfirms—intermsofbeta,returnvolatility,andPIN—althoughtheVLBiassamplefirmshavelowerbook-to-marketratios(probablybecauseofgreaterhigh-techconcentration).

Finally,todeterminetheeffectofsamplingbiasonourresults,weexaminedifferencesinPEGbetweenICWandcontrolsamplesforfirmsincludedintheVLsample(nottab-ulated).Weobtain1,049firmsfromtheVLsamplewithdatanecessaryforPEGestimation.WefindthatPEGissignificantlyhigherforICWfirmsthanforcontrolfirmsevenaftercontrols,althoughthemagnitudeofthedifference(53basispoints)ismuchlowerthanthatusingVL(208basispoints).ThissuggeststhatatleastpartoftheVLresultsisattributabletosamplingbias,althoughmeasurementdifferencesalsoappeartobeimportant.

Insummary,wefindthat(1)theVLmeasureperformspoorlyinconstructvaliditytestsand(2)theVLsampleappearstohavesignificantbiases.Also,theVLresultsareinconsistentacrossthetwoperiodsofanalyses(June2004andJune2006).Forthesereasons,wedonotbelievethattheJune2004VLresultssignificantlydetractfromour

15WealsoperformsimilaranalysescomparingtheCTsampletotheVLsample.WefindqualitativelyandstatisticallyidenticalresultstothosecomparingthePEGandVLsamples,i.e.,theVLsampleexcludesahigherproportionofdistressedandICWfirmsthantheCTsample.

TheAccountingReview,October2007

TheAccountingTABLE6

Review,ComparingSampleBiasesintheValueLineandPEGSamples:June2004

Means

Medians

OctoberVLBiasPEGBiasDifference:VLBiasPEGBiasDifference:

Sample

Sample؊PEGVLBiasBias

SampleSampleVLBias؊PEGBias2007RZSCORE5.9375.1720.765(0.04)6.0005.0001.000(0.03)LOSS0.3550.0690.286(0.00)————MKTV9428126Ϫ7184(0.00)4133513Ϫ3100(0.00)BMRATIO0.4660.611Ϫ0.145(0.00)0.4330.570Ϫ0.136(0.00)SEGMENTS1.7922.466Ϫ0.673(0.00)1.0002.000Ϫ1.000(0.00)FOREIGN0.1610.310Ϫ0.149(0.00)————AGE

11.22734.310Ϫ23.084(0.00)8.00032.000Ϫ24.000(0.00)HIGHTECH0.3360.0690.267(0.00)————BETA1.3690.9410.428(0.03)1.1920.8690.323(0.03)RETVAR0.0100.0030.007(0.00)0.0060.0020.004(0.00)PIN0.1410.1150.026(0.02)0.1380.1080.030(0.01)%ICW

0.183

0.017

0.166

(0.00)

Thetablecomparesasampleof1,320firmswithavailableRPEGestimatesandmissingRVLestimates(VLBiassample)toasampleof58firmswithmissingRPEGestimatesandavailableRVLestimates(PEGBiassample)fromanoriginalsampleof2,515firmsthatfiledfirst-timeSection404opinionsduringtheNovember2004toJanuary2006period.

Two-sidedp-valuesforcomparisonswiththecontrolsamplearereportedinparentheses:p-valuesformeancomparisonsarebasedont-statistics,andp-valuesformediancomparisonsarebasedonaWilcoxonsignedranktest.Univariatestatisticsforcontinuousvariablesarereportedafterobservationsfallingintotopandbottomextremepercentilesaretruncated.

(continuedonnextpage)

1284Ogneva,Subramanyam,andRaghunandanTABLE6(continued)

VariableDefinitions:

RZSCOREϭdecreasingdecilerankofAltman’s(1968)Z-Score(computedasdecilerankofZ-scoretimesϪ1andplus10);

LOSSϭanindicatorvariableequalto1ifearningsbeforeextraordinaryitems(Compustatitem#18)in2002and2003sumtolessthan0;

MKTVϭmarketvalueofequity,wheremarketvalueofequityisequaltopricepersharemultipliedbythenumberofsharesoutstandinginI/B/E/Sasofthe

endofJune2004;

BMRATIOϭtheratioofbookvalueofequitytothemarketvalueofequity,wherebookvalueisthestockholders’bookequityplusdeferredtaxesminusthebook

valueofpreferredstock,andmarketvalueofequityisdefinedaspreviously;

SEGMENTSϭnumberofbusinesssegmentsreportedin2004CompustatSegmentfileforfiscal2003;

FOREIGNϭanindicatorvariableequalto1ifthefirmhasanon-zeroforeigncurrencytranslation(Compustatitem#150)infiscal2003;

AGEϭnumberofyearsthefirmexistsintheCRSPdatabase;

HITECHϭanindicatorvariableequalto1ifafirmbelongstoahigh-techindustryaccordingtotheFrancisetal.(1994)classification;

BETAϭCAPMbetaestimatedasaslopefromregressingmonthlyexcessstockreturnonmonthlyexcessmarketreturn,usingthe12monthsofreturnspriorto

TheJune2004;

RETVARϭthevarianceofresidualsfromregressingmonthlystockexcessreturnsonthemarketexcessreturnandtwoFamaandFrench(1996)factor-mimicking

Accountingportfolioreturnsoverthe12monthspriortoJune2004;

PINϭprobabilityofinformedtrading,estimatedusingtradinginformationfromtheTAQ2004dataset;and%ICWϭpercentageofICWfirms.

Review,October2007InternalControlWeaknessandCostofEquity:EvidencefromSOXSection404Disclosures12851286Ogneva,Subramanyam,andRaghunandan

primaryinferencethattherearenocost-of-equitydifferencesbetweenICWandcontrolfirms.

VI.CONCLUSION

Weinvestigatewhetherfirmsreportingfirst-timeinternalcontrolweaknesses(ICWs)overfinancialreportingunderSection404ofSOX,onaverage,havehigherimpliedcostofequitycomparedtoacontrolsampleoffirmsthatfilecleanreports.WefindthatICWfirmshavehigherimpliedcostofequitythanfirmswithoutsuchweaknesses.However,wefindnosignificantassociationbetweenICWandcostofequityaftercontrollingforanalystforecastbiasandprimitivefirmcharacteristicsassociatedwithICW,suggestingthatthereisunlikelyanydirectrelationbetweenICWandcostofequity.Ourresultsarerobusttousingseveralalternativeproxiesforimpliedcostofequityandvariationsinsampleandtimeperiodofanalysis.Wealsofindqualitativelysimilarresultswithrespecttovariousriskattributesthataretheoreticallyrelatedtoinformationrisk.Finally,weareunabletofindsignificantlyhighercostofequity(aftercontrols)associatedevenwithegregioustypesofICW.

Ourresultssuggestthatthereareunlikelyanycost-of-equityeffectsassociatedwithfirmsreportingICWunderSection404.TheseresultsquestionanimportantbenefitofICWdisclosuresclaimedbyproponentsoftheSection404disclosures,i.e.,reducedcostofcapital.OurpaperthereforemakesanimportantcontributiontothedebateregardingthecostsandbenefitsoftheinternalcontrolstipulationsunderSOXbyquestioningtheircost-benefittrade-offs.

Weacknowledgethefollowingcaveats.First,costofequityisachallengingconcepttomeasure.Webaseourprimaryanalysesonimpliedcostofequitybecauseitdoesnotsufferfromtheproblemsinvolvedincomputingexpectedratesofreturnthroughfactorloadings(FamaandFrench1997).However,thereareseriousmeasurementproblemsas-sociatedwithestimatingimpliedcostofequityatthefirmlevel(EastonandMonahan2005).Byusingseveralalternativeproxies,controllingforforecastbias,conductingtestsofconstructvalidity,andreportingcorroboratingevidencewithriskattributesthatarethe-oreticallyassociatedwithinformationrisk,wesignificantlyminimizethelikelihoodthatourresultsareanartifactofmeasurementerrorinourcost-of-equityproxies.

Second,oursamplesuffersfromlarge-firmbiasthatarisesbothbecausesmallerfirmsareexemptfromtheacceleratedfilingrequirementsofSection404andbecauseofourdatarequirements(oursamplefirmsmustbecoveredintheI/B/E/Sdatabase).Therefore,ourinferencesmaynotapplytothesmallestfirmsintheeconomy,wheresomeofthestrongesteffectsofweakinternalcontrolsarelikelytooccur.

Third,ourstudyfocusesonasampleoffirmsreportingunderSection404forthefirsttime.Doyleetal.(2007b)conjecturethatICWdisclosuresunderSection404maybesubjecttolowermaterialitythresholdsbyauditors,therebyreducingthepowerofadverseSection404reportstodetectthefirmswiththemostegregiousinternalcontrolproblems.Thismayreducethepowerofourteststomoregenerallyidentifycost-of-equityeffectsofinternalcontrolweaknesses.

Finally,allofthetestsconductedinthepaperarecross-sectional.Hence,anyeconomy-wideeffectsofSOXwillgoundetected.Inparticular,SOXrequiresfirmstoprovideas-surancesabouttheirinternalcontrolsonagoing-forwardbasis.Thisisequivalenttoanexogenouspre-commitmenttodisclose,whichmaycausefirmstochangetheirbehavior.Forexample,managersmayrealizethatcapitalmarketparticipantswillbebetterinformed,whichimpliesthatmanagementwillbeheldmoreaccountablefortheiractions;hence,

TheAccountingReview,October2007

InternalControlWeaknessandCostofEquity:EvidencefromSOXSection404Disclosures1287

managersmaychoosetoimproveinternalcontrols.This,inturn,maypotentiallyreducetheexpectedrateofreturnforallfirmsifsomeofthetheorieswediscussaredescriptive.APPENDIXA

ALTERNATIVEPROXIESFORCOSTOFEQUITY

Dependingontheparticularmodelforcostofcapitalestimation,weneeduptofiveyearsofEPSforecasts.Weusemedianconsensusforecaststoproxyforthemarket’sfutureearningsexpectationsandrequirethateachobservationhavenon-missingone-andtwo-year-aheadconsensusearningsforecasts(FEPS1andFEPS2).Weusethree-,four-,andfive-year-aheadforecastsforfutureearningspershare,iftheyareavailableinI/B/E/S.Otherwise,weestimatethecorrespondingforecastedvaluesbyapplyingthelong-termgrowthratetoFEPS2.WeusestockpricepershareandforecastsofEPSandlong-termearningsgrowthfromtheI/B/E/SsummarytapeasofJune17,2004(June15,2006forsupplementaryanalysis).Bookvalueofequityandthedividendpayoutratioforthelatestfiscalyear-endpriortoJune2004(priortoJune2006forsupplementaryanalysis)areobtainedfromtheCompustatannualdatabase.Followingpriorresearch,weassumeacon-stantdividendpayoutratiothroughouttheforecastperiod.Thefollowingdatarestrictionsareimposedwhenestimatingeachcost-of-equitymeasure:positivebookvalueofequityforGebhardt,Lee,andSwaminathan(GLS)andClausandThomas(CT)methods;FEPS1Ͼ0forModifiedOhlsonandJuettner-Nauroth(OJ),GodeandMohanram(GM),andSimultaneousEstimationofCostofEquityandGrowthRates(EAST)methods;FEPS2ϾFEPS1forPrice-Earnings-Growth(PEG),GM,andEASTmethods.

Gebhardt,Lee,andSwaminathanMethod(GLS)

Gebhardtetal.(2001)usethefollowingmodeltoestimatethecostofequity:

PtϭBtϩ

FROEtϩ1ϪreFROEtϩ2ϪreBtϩBtϩ1ϩTV,

(1ϩre)(1ϩre)2wherePtϭstockprice;Btϭbookvalueofequitypershare;Btϩiϭfuturebookvalue

ofequityestimatedusingacleansurplusassumption(BtϩiϭBtϩiϪ1Ϫk*FEPStϩiϩFEPStϩi,wherekisthedividendpayoutratioandFEPSistheearningspersharefore-cast);FROEtϩiϭfuturereturnonequityforecast;TVϭterminalvalue,whichisequaltoTϪ1FROEtϩiϪreFROEtϩTϪrewithTϭ12;andreϭcost-of-equitycap-BϩBt+iϪ1(1ϩre)ire(1ϩre)TϪ1tϩTϪ1iϭ3ital.Forthefirstthreeyears,futureROEisestimatedasFROEtϩiϭFEPStϩi/BtϩiϪ1.Beyondthethirdyearuntilyear12,FROEiscomputedbylinearinterpolationtotheindustrymedianROE.16ThecostofequityiscalculatednumericallyemployingtheNewton-Raphsonmethod.Weputtheinitialvalueofthecostofequityequalto9percentinthefirstiteration;thealgorithmisconsideredtoconvergeifthepriceofastockobtainedfromtheimpliedcostofequitydeviatesfromtheactualstockpricebynomorethan$0.005.17WelabelthismeasureGLS.

͸

1617WeestimateindustrymedianROEsoverthetenyearsprecedingthecostofcapitalestimationusingallavailableCompustatannualobservationswithpositivebookvaluesofequityandnon-negativeearnings,wherethe48industriesaredefinedaccordingtotheFamaandFrench(1997)classification.

BotosanandPlumlee(2005)useasimilarconvergencecriterion.Wereplicateestimationproceduresusingalternativeinitialvaluesandobtainsimilarresults.

TheAccountingReview,October2007

1288Ogneva,Subramanyam,andRaghunandan

ClausandThomasMethod(CT)

AmethodproposedbyClausandThomas(2001)isdifferentfromGebhardtetal.(2001)intermsoftheperpetualgrowthassumptioninestimatingterminalvalue.Specifi-cally,themethodimpliesthatthevalueofafirmcanbeexpressedas:

PtϭBtϩ

ϪrB(FEPSϪrB)(1ϩg)

ϩ,͸FEPS

(1ϩr)(rϪg)(1ϩr)

5tϩiettϩ5eetϩ4eie5tϭ1whereFEPSisfutureearningspershare,Bisbookvaluepershare,aspreviouslydefined,

gisthegrowthrateofresidualearningsinperpetuityequaltotheexpectedinflationrate(risk-freerateminus3percent),andreisthecost-of-equitycapital.Theimpliedcostofequityisestimatedusingarecursivealgorithm,asdescribedearlier.WelabelthismeasureCT.

GodeandMohanramMethod(GM)

TheGodeandMohanram(2003)methodisbasedontheOhlsonandJuettner-Nauroth(2005)modelofEPSandEPSgrowth,whichequatesthevalueofafirmwiththecapitalizednext-periodEPSandthefuture‘‘abnormal’’growthinEPS.Severalpriorstudiesderiveempiricalmeasuresforcostofcapitalbyimposingvariousrestrictionsonthisbasicmodel.GodeandMohanram(2003)representthevalueofthefirmas:

Ptϭ

FEPStϩ1(FEPStϩ2ϪFEPStϩ1Ϫre(FEPStϩ1Ϫk*FEPStϩ1)

ϩ,rere(reϪg)

whereFEPSisfutureearningspershare,kisthedividendpayoutratio,asdefinedprevi-ously,gisthelong-termgrowthinabnormalearningschanges,andreisthecost-of-equity

capital.GodeandMohanram(2003)showthatthismodelspecificationprovidesanana-lyticalsolutionforthecostofequity.Thefollowingexpressiondescribesthesolution:

RGMϭAϩ

where:

Aϵg2ϭ

1k*FEPS1(␥ϩ1)ϩ;2P0(FEPS2ϪFEPS1)

.FEPS1ΊA2ϩ

FEPS1(g2Ϫ(␥Ϫ1)),P0ͩͪGodeandMohanram(2003)estimate␥(thelong-termgrowthinabnormalearningschanges)astherealrisk-freerate:␥Ϫ1ϭRfϪ3%.Weusetheten-yearT-bondrateatthedateofestimationtoproxyfortherisk-freerate.ThismeasureislabeledGM.Price-EarningsGrowthMethod(PEG)

Easton(2004)derivesarestrictedversionofGodeandMohanram(2003)model.Heshowsthatundertheassumptionofzerodividendsandzerogrowthinabnormalearningschanges,thefirm’scostofequityisobtainedfromthefollowingexpression:

TheAccountingReview,October2007

InternalControlWeaknessandCostofEquity:EvidencefromSOXSection404Disclosures1289

Ptϭ

Et[FEPStϩ2ϪFEPStϩ1]

,(re)2whereFEPSisfutureearningspershareforecast,asdefinedpreviously,andreisthecost-of-equitycapital.Fromtheaboveexpression,thecostofcapitalisproportionaltotheinverseofthePEGratio:

RPEGϭ

ΊFEPS2ϪFEPS1.P0ThecostofcapitalestimatedbytheEaston(2004)PEGspecificationislabeledPEG.ModifiedOhlsonandJuettner-NaurothMethod(OJ)

WederiveanadditionalspecificationfromtheOhlsonandJuettner-Nauroth(2005)model.WhileGodeandMonahan(2003)usethegrowthinearningsfromperiodttotϩ1toderiveshort-termearningsgrowthestimates,weusealltheinformationinanalysts’earningsforecastsincludinglong-termgrowthforecasts.

Webeginbydescribingthevalueofafirm’sequityasinOhlsonandJuettner-Nauroth(2005):

FEPStϩ1Ptϭϩ

reiϭ1͸1/r(FEPS

ϱetϩiϩ1ϩre*DIVtϩiϪR*FEPStϩi)

Ri,whereFEPSisexpectedearningspershare,DIVisexpecteddividendspershare,andRiscost-of-equitycapitalϩ1(1ϩre).

Afterexpandingthevaluationformulatoincludeexplicitforecastsforperiodsonetothreeandtheterminalvalue,weobtain:

FEPStϩ1ztϩ1ztϩ2ztϩ3Ptϭϩϩ2ϩ3ϩ

reRRR

iϭ4͸zR

ϱtϩiiztϩiϵ1/re(FEPStϩiϩ1ϩre*DIVtϩiϪR*FEPStϩi).

Next,asinOhlsonandJuettner-Nauroth(2005)andGodeandMohanram(2003)wemaketheassumptionofaconstantgrowthrateforzintheterminalperiod:

ztϩiϵ␥*zt,

where␥Ͼ1.Ifwedefine␰tϩiϵ

␰tϩiϵ

ztϩi,thenwegetthefollowingearningschangedynamics:RiR

*␰t.

Usingthesimplifiednotation,theterminalsumcanbere-writtenas:

TheAccountingReview,October2007

1290

Ogneva,Subramanyam,andRaghunandan

␰tϩ4ϩ

R

*␰tϩ4ϩ1ϩR

␥2ϭ␰tϩ4ϭ␰tϩ4ͫͫͬ␥

R2ϩ

*␰tϩ4ϩ

␥2␥3R3ϩ

*␰tϩ4ϩ

␥4␥4RR2ϩ

␥3R3R4ϩ...

ͬ

R4*␰tϩ4ϩ...

RϪ␥

.

Therefore,theentirevaluationexpressioncanbewrittenas:

Ptϭ

ϭ

FEPStϩ1ztϩ1ztϩ2ztϩ3ztϩ4R

ϩϩ2ϩ3ϩ4reRRRRRϪ␥FEPStϩ1ztϩ1ztϩ2ztϩ3z

ϩϩ2ϩ3ϩ3tϩ4.reRRRR(RϪ␥)

ͫͬWeoperationalizetheaboveexpressioninthefollowingmanner:

Ptϭ

and

ztϩiϵ1/re(FEPStϩiϩ1ϩre*kFEPStϩiϪ(1ϩre)*FEPStϩi),

whereFEPSistheforecastedfutureearningspershare,kisthedividendpayoutratio,

definedaspreviously,andreisthecost-of-equitycapital.Thelong-termgrowthrate(g)issetequaltotheinflationrate,whichweestimateastherisk-freerateminus3percent.Weusetheiterativeproceduredescribedearliertoobtainanumericestimatefortheimpliedcostofcapitalfromtheabovemodel.WelabelthismeasureOJ.

ValueLineExpectedReturn(VL)

WeuseaverageannualtotalexpectedreturnratesfromtheValueLinehistoricalesti-matesfilecomputedusingforecastsissuedinthesecondcalendarquarterof2004(secondcalendarquarterof2006foroursupplementaryanalysis).18SimultaneousEstimationofCostofEquityandGrowthRates(EAST)

Tosimultaneouslyestimatecostofequityandearningsgrowthratesforindividualfirms,weaugmentthesimultaneousestimationofcostofequityandgrowthratedevelopedinEaston(2004)bytherandomsamplingprocedureandfixedeffectsregression.

TheoriginalEaston(2004)estimationisbasedontheOhlsonandJuettner-Nauroth(2005)model:

18FEPStϩ1ztϩ1ztϩ2ztϩ3ztϩ4ϩϩϩϩre(1ϩre)(1ϩre)2(1ϩre)3(1ϩre)3(reϪg)

AccordingtotheValueLinedatamanual,thesetotalexpectedreturnratesareestimatedusingprojectedthree-tofive-year-aheadstockappreciationanddividendforecastsissuedbytheValueLineanalysts.Projectedstock

appreciationisderivedbasedonthecurrentstockpriceandtheprojectedstockpricetarget.ThelattervalueisbasedonValueLine’sestimatedearningspersharethreetofiveyearsoutmultipliedbytheexpectedprice-earningsratiointhesameperiod.Thewidthofthepricerangeisbasedonthestandarddeviationofhistoricweeklypercentpricechange.

TheAccountingReview,October2007

InternalControlWeaknessandCostofEquity:EvidencefromSOXSection404Disclosures1291

FEPStϩ1Ptϭϩ

reiϭ1͸1/r(FEPS

ϱetϩiϩ1ϩre*DIVtϩiϪR*FEPStϩi)

Ri.Assumingthatthetermunderthesumgrowsinperpetuityattherateequalto⌬agr,thevaluationformulabecomes:

Ptϭ

FEPStϩ1(FEPStϩ2ϩre*DIVtϩ1Ϫ(1ϩre)*FEPStϩ1)

ϩ.rere(reϪ⌬agr)

Afterrearrangingtheterms,thelatterequalitycanbere-writtenas:

FEPStϩ2ϩre*DIVtϩ1FEPStϩ1ϭre(reϪ⌬agr)ϩ(1ϩ⌬agr),

PtPtor,definingCEPStϩ2ϵFEPStϩ2ϩre*DIVtϩ1,␥0ϵre(reϪ⌬agr)and␥1ϵ(1ϩ⌬agr),

as:

CEPStϩ2FEPStϩ1ϭ␥0ϩ␥1.PtPtFortheportfoliooffirms,whereidenotesafirmsubscript,theaveragecostofequity

andaveragegrowthratecanthereforebeestimatedfromtheinterceptandtheslopeofalinearregression:

CEPSitϩ2FEPStiϩ1ϭ␥0ϩ␥1ϩeit.19iiPtPtThismethodprovidesaverageestimatesofcostofequityandearningsgrowthfora

portfoliooffirms.Inthecontextofcomparingcostofequityvaluesforthetwogroupsoffirms,thisposestwoproblems:(1)itisdifficulttoassessstatisticalsignificanceofthedifferenceincostofequitybetweenthetwogroupsoffirms;and(2)itisimpossibletocontrolforfirm-specificcharacteristics.Accordingly,wedevelopabootstrappingprocedurethatallowsustodeterminefirm-specificcostofequityusingtheEaston(2004)methodology.

First,werandomlyselect500firmsoutofourfullsampleoffirmswithFEPStϩ1Ͼ0andFEPStϩ2ϾFEPStϩ1andestimatetheaveragecostofequityforthisrandomlyselectedportfoliousingEaston’smethodabove.Wereplicatethisrandomsamplingprocedure5,000times(withreplacement).Theresultingaveragecostofequityforeachportfoliocanberepresentedbythecolumnvector[r1,r2,r3,r4,r5,...,r5000].Giventhateachresultingrateofreturnrjistheaveragecostofequityforfirmsrandomlydrawniniterationj,the

19Theestimationisperformedusingnumericalmethodssincethedependentvariablerequirespriorknowledgeoftherateofreturn.Theestimationprocedurestartswithassumingarateofreturnof10percent.Thentheslope(␥1)andintercept(␥0)ofthelinearregressionareestimated.Theimpliedgrowthrate(⌬agr)iscalculatedas

⌬agrϩ͙⌬agr2ϩ4␥0(␥1Ϫ1)andtherateofreturn(re)iscalculatedas.Theobtainedrateofreturnisused

2

torecalculatethedependentvariableandtheregressionisre-estimated.Theiterationscontinueuntilconvergence(nochangeinre)issatisfied.

TheAccountingReview,October2007

1292Ogneva,Subramanyam,andRaghunandan

followingrelationbetweenportfolioratesofreturnandindividualstockreturnsneedstoholdforeachportfolioj:

rjϭ

͸r*s

N1iij500

,whereriisthecostofequityforstocki,sijisanindicatorvariableequalto1ifstockiisincludediniterationj,andNisthetotalnumberofstocksinthepopulationusedforrandomsampling.

Therefore,therelationshipbetweenportfolioreturnsandindividualfirmreturnscanberepresentedintheformYϭX␤as:

΄΅΄r1r2⅐⅐

ϭ

s111s2⅐⅐

s212s2s25000⅐⅐

s313s23s5000r5000s15000⅐

4s1s424s5000⅐⅐

⅐⅐⅐⅐⅐

sN1sN2sN5000⅐⅐

΅΄΅r1r2⅐⅐

ϫ

1.500

rNTheaboverepresentationisequivalenttoafixed-effectsmodelwithnoadditionalin-dependentvariables.Thevectorofindividual-firmcostofequitycanbeestimatedbyOLSregressionwheretheaverageportfolioreturnisthedependentvariable,andNindicatorsareindependentvariables.

Inunreportedanalyseswecomparethemeancostofequityobtainedusingtheaboveprocedureforinternalcontrolweakness(ICW)andcontrolfirms,withthoseobtainedbyestimatingtheEaston(2004)methodseparatelyfortheICWandcontrolsubsamples.Wefindthattheaveragecostofequityusingourbootstrappingprocedure(costofequitydirectlyapplyingEaston’smethod)forthe295ICWfirmsand1,911controlfirmsare0.1105(0.1147)and0.1056(0.1044),respectively.20Thissuggeststhatourprocedurees-timatestheaveragecostofequityfortheICWandcontrolfirmswithsignificantaccuracy.APPENDIXB

PREDICTEDFORECASTERRORS

PredictedforecasterrorsareestimatedusingLiuandSu(2005).Themodelincludesvariablesknowntobeassociatedwithsystematicbiasesinanalystforecasts:variablesrelatedtounderreactiontopastinformation,namely,(1)pastearningssurprise,(2)stockreturns,and(3)recentanalystearningsforecastrevisions,andvariablesrelatedtooverre-actiontopastinformation,namely,(4)book-to-marketratio,(5)forwardearnings-to-priceratio,(6)long-termgrowthforecast,(7)pastsalesgrowth,(8)investmentsinproperty,plant,andequipment,(9)investmentinotherlong-livedassets,and(10)theaccrualcomponentofearnings.Wealsoaddavariablemeasuringthehorizonoftheforecastsinceitishasbeenshownthatanalystsareprogressivelylessoptimisticastheygetclosertotheannualearningsannouncements(Richardsonetal.2004).Theunderreactionvariablesareexpectedtohavenegativepredictivecoefficients,andtheoverreactionvariablesareexpectedtohavepositivepredictivecoefficients.

20NotethatthesenumbersdifferfromthoseinTable2,becausetheresultsreportedinTable2areaftertruncationofoutliers.

TheAccountingReview,October2007

InternalControlWeaknessandCostofEquity:EvidencefromSOXSection404Disclosures1293

Themodelisestimatedonapooledsamplewitheitherone-year-ortwo-year-aheadforecasterrorsasdependentvariables:

FERR1iϭ␤0ϩ␤1*GRWSALESiϩ␤2*ACCiϩ␤3*⌬PPEiϩ␤4*⌬OLTAiϩ␤5*LTGiϩ␤6*BMRATIOiϩ␤7*FEPiϩ␤8*REV1iϩ␤9*QFERRiϩ␤10*RETURNiϩ␤11*HORIZ1iϩ␧i;

FERR2iϭ␥0ϩ␥1*GRWSALESiϩ␥2*ACCiϩ␥3*⌬PPEiϩ␥4*⌬OLTAiϩ␥5*LTGiϩ␥6*BMRATIOiϩ␥7*FEPiϩ␥8*REV2iϩ␥9*QFERRiϩ␥10*RETURNiϩ␥11*HORIZ2iϩ␩i;

where:

FERR1,FERR2ϭanalystforecasterror,estimatedasactualI/B/E/SEPSminus

eitherone-ortwo-year-aheadconsensus(median)EPSforecastissuedinJuneofYeartϩ1(‘‘forecastdate’’),scaledbystockpriceonthe‘‘forecastdate’’;

GRWSALESϭpercentgrowthinsales(Compustatitem#12)fromYeartϪ5to

Yeart;whereYeartisthemostrecentfiscalyearendedbeforethe‘‘forecastdate’’;

ACCϭtotalaccrualsforYeartestimatedusingbalance-sheetmethod,21scaledbybeginning-of-yeartotalassets;

⌬PPEϭchangeinproperty,plant,andequipment(Compustatitem#7)over

Yeartscaledbybeginning-of-yeartotalassets;

⌬OLTAϭchangeinotherlong-termassets22overYeartscaledbybeginning-of-yeartotalassets;

LTGϭmedianlong-termgrowthinearningsforecastonthe‘‘forecast

date’’;

BMRATIOϭbook-to-marketratio;

FEPϭforwardearnings-to-priceratio(one-year-aheadmedianEPS

forecastissuedonthe‘‘forecastdate,’’dividedbystockpriceonthe‘‘forecastdate’’);

REV1,REV2ϭrevisioninone-ortwo-year-aheadconsensusforecastoverthethree

monthspriorto‘‘forecastdate,’’scaledbystockpriceonthe‘‘forecastdate’’;

QFERRϭerrorinthemostrecentquarterlyEPSforecastpriortothe

‘‘forecastdate’’;

RETURNϭstockreturnoverthe12monthspriortothe‘‘forecastdate’’;and

HORIZ1,HORIZ2ϭdaysremainingbeforetheendofthefiscalyearforwhichtheone-ortwo-year-aheadforecastismadecalculatedonthe‘‘forecastdate.’’

2122OperatingAccrualsareestimatedaschangeincurrentassets(Compustatitem#4)–changeincash(Compustatitem#11)–changeincurrentliabilities(Compustatitem#5)ϩchangeincurrentportionoflong-termdebt(Compustatitem#34)–depreciation(Compustatitem#14).

Calculatedaschangeintotalassets(Compustatitem#6)minuschangeincurrentassets(Compustatitem#4)minuschangeinPPE(Compustatitem#7).

TheAccountingReview,October2007

1294Ogneva,Subramanyam,andRaghunandan

Weestimatetheabovemodelseparatelyforthepooledsampleusingallavailableforecastsmadein1999–2003(1998–2002)fortheone-(two-)year-aheadforecasterrorprediction.Ourestimationsamplecomprises16,321(13,088)firm-yearsfortheone-(two-)year-aheadforecasterrorpredictionmodel,respectively.Wecomputepredictedone-andtwo-year-aheadforecasterrorsforforecastsmadeinJune2004(i.e.,forecastsusedinourcostofequitycalculations).Specifically,weapplycoefficientsofthemodelestimatedforthepooledholdoutsampletothevaluesofpredictivevariablesasofJune2004foreachindividualfirminoursample.Theestimatederrorshavesignificantpredictivepowerforoursamplefirms.ThecorrelationoftheestimatedpredictedforecasterrorPREDFERR1withtherealizedone-year-aheadforecasterroris0.25(0.15)foraPearson(Spearman)test,significantatthe0.01level.Thetwopredictederrorsarehighlycorrelatedwitheachother—thePearson(Spearman)correlationcoefficientisequalto0.70(0.69).Accordingly,toavoidmulticollinearityproblemswhenusingbothPREDFERR1andPREDFERR2inOLSregressions,weusethesimpleaverageofPREDFERR1andPREDFERR2inourtests.

InourValueLinecostofequityanalyses,weestimateone-year-aheadpredictedforecasterrorbysubstitutingI/B/E/SforecastswithValueLineforecasts.23Accordingly,the‘‘fore-castdates,’’realizedEPS,andpricesarealsotakenfromtheValueLinehistoricalestimatesfile.

REFERENCES

AmericanBankers’Association(ABA).2005.CommentssubmittedtotheSECRoundtableonIm-plementationofInternalControlReportingProvisions.Washington,D.C.:SEC.Availableat:http://www.sec.gov/spotlight/soxcomp/soxcomp-brunner.pdf.

AmericanElectronicsAssociation(AEA).2005.Sarbanes-OxleySection404:The‘‘Section’’ofUn-intendedConsequencesandItsImpactonSmallBusiness.Washington,D.C.:AEA.

Ashbaugh-Skaife,H.,D.W.Collins,W.R.Kinney,andR.LaFond.2006.Theeffectofinternal

controldeficienciesonfirmriskandcost-of-equitycapital.Workingpaper,UniversityofWisconsin–Madison.

———,———,and———.2007.Thediscoveryandreportingofinternalcontroldeficienciesprior

toSOX-mandatedaudits.JournalofAccountingandEconomics44(1–2):166–192.

Beneish,M.D.,M.Billings,andL.Hodder.2006.Internalcontrolweaknessesandinformation

uncertainty.Workingpaper,IndianaUniversity.

Botosan,C.A.1997.Disclosurelevelandthecost-of-equitycapital.TheAccountingReview72:323–

350.

———,andM.Plumlee.2002.Are-examinationofdisclosurelevelsandexpectedcost-of-equity

capital.JournalofAccountingResearch40:21–40.

———,and———.2005.Assessingalternativeproxiesfortheexpectedriskpremium.TheAc-countingReview80:21–53.

Bowman,E.H.1982.Riskseekingbytroubledfirms.SloanManagementReview23:33–42.

Bronson,S.N.,J.V.Carcello,andK.Raghunandan.2006.Firmcharacteristicsandvoluntaryman-agementreportsoninternalcontrol.Auditing:AJournalofPractice&Theory(November):25–39.

Brown,K.C.,W.V.Harlow,andL.T.Starks.1996.Oftournamentsandtemptations:Ananalysis

ofmanagerialincentivesinthemutualfundindustry.JournalofFinance51:85–110.

23Two-year-aheadforecastsarenotavailableintheValueLinehistoricalestimatesfile.

TheAccountingReview,October2007

InternalControlWeaknessandCostofEquity:EvidencefromSOXSection404Disclosures1295

Bryan,S.H.,andS.B.Lilien.2005.Characteristicsoffirmswithmaterialweaknessesininternal

control:anassessmentofSection404ofSarbanes-Oxley.Workingpaper,WakeForestUniversity.

Chevalier,J.,andG.Ellison.1997.Risktakingbymutualfundsasaresponsetoincentives.Journal

ofPoliticalEconomy105:1167–1200.

Claus,J.,andJ.Thomas.2001.Equitypremiaaslowasthreepercent?Evidencefromanalysts’

earningsforecastsfordomesticandinternationalstockmarkets.JournalofFinance56:1629–1666.

Deloitte&Touche.2005.HowcancomplyingwithSarbanes-Oxleyprovisionsbenefitnot-for-profithealthcareorganizations?Deloitte&ToucheLLP,NewYork,NY.

Dhaliwal,D.,L.Krull,O.Z.Li,andW.Mozer.2005.Dividendtaxesandimpliedcost-of-equity

capital.JournalofAccountingResearch43:675–708.

Donaldson,W.2005a.Testimony:ConcerningtheimpactoftheSarbanes-OxleyAct.HouseCom-mitteeonFinancialServices.Washington,D.C.:GovernmentPrintingOffice.———.2005b.We’vebeenlistening.WallStreetJournal(March29):A14.

Doyle,J.T.,W.Ge,andS.E.McVay.2007a.Determinantsofweaknessesininternalcontrolover

financialreporting.JournalofAccountingandEconomics44(1–2):193–223.

———,———,and———.2007b.Accrualsqualityandinternalcontroloverfinancialreporting.

TheAccountingReview82:1141–1170.

Easley,D.,R.F.Engle,M.O’Hara,andL.Wu.2001.Time-varyingarrivalratesofinformedand

uninformedtrades.Workingpaper,AFA2002AtlantaMeetings.

———,S.Hvidkjaer,andM.O’Hara.2002.Isinformationriskadeterminantofassetreturns?The

JournalofFinance58:2185–2221.

———,andM.O’Hara.2004.Informationandthecostofcapital.JournalofFinance59:1553–

1583.

Easton,P.2004.PEratios,PEGratios,andestimatingtheimpliedexpectedrateofreturnonequity

capital.TheAccountingReview79:73–95.

———,andS.Monahan.2005.Anevaluationofaccounting-basedmeasuresofexpectedreturns.The

AccountingReview80:501–538.

Ernst&Young.2005.SubmissionlettertotheSEC.FileNo.4-497,implementationofSarbanes-Oxleyinternalcontrolsprovisions.Washington,D.C.:SEC.Availableat:http://www.sec.gov/spotlight/soxcomp/soxcomp-turley.pdf.

Fama,E.,andK.French.1995.Sizeandbook-to-marketfactorsinearningsandreturns.Journalof

Finance50:131–155.

———,and———.1996.Multifactorexplanationsofassetpricinganomalies.JournalofFinance

51:55–84.

———,and———.1997.Industrycostsofequity.JournalofFinancialEconomics43:153–194.Fiegenbaum,A.,andH.Thomas.1988.Attitudestowardriskandtherisk-returnparadox:Prospect

theoryexplanations.AcademyofManagementJournal31:85–106.

FinancialExecutivesInstitute(FEI).2005.Commentsofthecommitteeoncorporatereporting.Sub-mittedtotheSECRoundtableonImplementationofInternalControlReportingProvisions.SEC:Washington,D.C.:FEI.Availableat:http://www.sec.gov/spotlight/soxcomp/soxcomp-brod.pdf

FitchRatings.2005.Fitch’sApproachtoEvaluatingManagementandAuditorAssessmentsofInternal

Control.NewYork,NY:Fitch.

Francis,J.,D.Philbrick,andK.Schipper.1994.Shareholderlitigationandcorporatedisclosures.

JournalofAccountingResearch32:137–164.

———,R.LaFond,P.Olsson,andK.Schipper.2004.Costsofequityandearningsattributes.The

AccountingReview79:967–1010.

———,———,———,and———.2005.Themarketpricingofaccrualsquality.JournalofAc-countingandEconomics39:295–327.

Garber,S.andS.Klepper.1980.‘‘Administrativepricing’’orcompetitioncoupledwitherrorsof

measurement.InternationalEconomicReview21:413–435.

TheAccountingReview,October2007

1296Ogneva,Subramanyam,andRaghunandan

Ge,W.,andS.McVay.2005.Thedisclosureofmaterialweaknessesininternalcontrolafterthe

Sarbanes-OxleyAct.AccountingHorizons(September):137–158.

Gebhardt,W.R.,C.M.C.Lee,andB.Swaminathan.2001.Towardanexantecost-of-capital.Journal

ofAccountingResearch39:135–176.

Gode,D.,andP.Mohanram.2003.InferringcostofcapitalusingtheOhlson-Juettnermodel.Review

ofAccountingStudies8:399–431.

Goyal,A.,andP.Santa-Clara.2003.Idiosyncraticriskmatters!JournalofFinance58:975–1007.Guay,W.,S.P.Kothari,andS.Shu.2005.Propertiesofimpliedcostofcapitalusinganalysts’

forecasts.Workingpaper,UniversityofPennsylvania,MassachusettsInstituteofTechnology,andBostonCollege.

Hammersley,J.,L.Myers,andC.Shakespeare,2005.Marketreactionstointernalcontrolweakness

disclosures.Workingpaper,UniversityofGeorgia,UniversityofIllinoisatUrbana–Champaign,andUniversityofMichigan.

Hermanson,H.2000.Ananalysisofthedemandforreportingoninternalcontrol.Accounting

Horizons14(3):325–341.

Hogan,C.,andM.Wilkins.2005.Internalcontrolweaknessesandearningsmanagement.Working

paper,SouthernMethodistUniversityandTexasA&MUniversity.

Hribar,P.,andN.Jenkins.2004.Theeffectofaccountingrestatementsonearningsrevisionsandthe

estimatedcostofcapital.ReviewofAccountingStudies9:337–356.

Jones,J.J.1991.Earningsmanagementduringimportreliefinvestigations.JournalofAccounting

Research29:193–228.

Kinney,W.,M,Maher,andD.Wright.1990.Assertions-basedstandardsforintegratedinternalcontrol.

AccountingHorizons4(4):1–8.

Kothari,S.,A.Leone,andC.Wasley.2004.Performancematcheddiscretionaryaccrualmeasures.

JournalofAccounting&Economics39:163–197.

KPMG.2005.CorporateGovernanceReforms:ImplicationsforNot-for-ProfitHealthcare.Available

at:http://www.kpmg.com/aci/docs/ACIRTHLSp04HlthcWebFNL.pdf

Lambert,R.,C.Leuz,andR.E.Verrecchia.2007.Accountinginformation,disclosure,andthecost

ofcapital.JournalofAccountingResearch45(2):385–420.

Liu,J.,andW.Su.2005.Forecastinganalystforecasterrors.Workingpaper,UniversityofCalifornia,

LosAngeles.

McMullen,D.A.,K.Raghunandan,andD.V.Rama.1996.Internalcontrolreportsandfinancial

reportingproblems.AccountingHorizons10(December):67–75.

Microsoft,Inc.2005.CommentssubmittedtotheSECRoundtableonImplementationofInternal

ControlReportingProvisions.Washington,D.C.:SEC.Availableat:http://www.sec.gov/news/press/4-497/microsoft033105.pdf.

Moody’sInvestorsService(Moody’s).2004.SpecialComment:Section404ReportsonInternalCon-trol.NewYork,NY:Moody’sInvestorsService.

Ohlson,J.1995.Earnings,bookvalues,anddividendsinequityvaluation.ContemporaryAccounting

Research11:661–687.

———,andB.Juettner-Nauroth.2005.ExpectedEPSandEPSgrowthasdeterminantsofvalue.

ReviewofAccountingStudies10:349–365.

Powell,S.S.2005.CostsofSarbanes-Oxleyareoutofcontrol.LetterstotheEditor.WallStreet

Journal(March21):A17.

PublicCompanyAccountingOversightBoard(PCAOB).2004.AnAuditofInternalControlOver

FinancialReportingPerformedinConjunctionwithAnAuditofFinancialStatements.AuditingStandardNo.2.Washington,D.C.:PCAOB.

Richardson,S.,S.H.Teoh,andP.Wysocki.2004.Thewalkdowntobeatableanalystforecasts:The

roleofequityissuanceandinsidertradingincentives.ContemporaryAccountingResearch21:885–924.

U.S.HouseofRepresentatives.2002.TheSarbanes-OxleyActof2002.PublicLaw107-204[H.R.

3763].Washington,D.C.:GovernmentPrintingOffice.

TheAccountingReview,October2007

InternalControlWeaknessandCostofEquity:EvidencefromSOXSection404Disclosures1297

———.2005.TheimpactoftheSarbanes-Oxleyact.HearingbeforetheCommitteeonFinancial

Services.April21.Washington,D.C.:GovernmentPrintingOffice.

U.S.Senate.2004.ExaminingtheimpactoftheSarbanes-Oxleyactanddevelopmentsconcerning

internationalcommerce.HearingsbeforetheCommitteeonBanking,Housing,andUrbanAf-fairs.S.108-850.September9.Washington,D.C.:GovernmentPrintingOffice.

Wilfert,G.E.2005.Thefirstwaveofsection404reporting.OrangeCountyBusinessJournal28

(Feb7–Feb13):4.

TheAccountingReview,October2007

因篇幅问题不能全部显示,请点此查看更多更全内容

热门图文

Top