The incidence of earnings management on information asymmetry in an uncertain environment some canadian evidence

2010 emphasize that accruals models can be estimated at the firm level, which allows variation across firms in the determinants of normal accruals.. 2 The balance sheet method of accruals

evidence

ESG UQÀM, School of Management, University of Quebec at Montreal, P.O Box 8888, Montréal, Québec, Canada H3C 3P8

a r t i c l e i n f o

Keywords:

Corporate governance

Earnings management

Environmental uncertainty

Information asymmetry

a b s t r a c t

© 2013 Published by Elsevier Inc

1 Introduction

Inthispaper,weinvestigatetheassociationbetweenearningsmanagementandinformationasymmetryconsidering environmentaluncertainty.Thetheoryofthefirm(e.g.Child,1972;Williamson,1975)recognizesthatenvironmental uncer-taintyplacessignificantconstraintsonfirms,affectingstrategyanddecision-making.Althoughfirmsareconstrainedbythe natureoftheirenvironment,managersdohaveopportunitiestorespondstrategicallytouncertainty(Ghosh&Olsen,2009) Oneoftheseopportunitiesisearningsmanagement.Theextentofopportunisticearningsmanagementislikelytobehigher wheninformationasymmetryishigh(Dye,1988;Trueman&Titman,1988).Inturn,earningsmanagementcouldincrease theuncertaintyforinvestorsaboutthedistributionofafirm’sfuturecashflows,whichwouldcreateinformationasymmetry betweeninformedandlessinformedinvestors(Bhattacharya,Desai,&Venkataraman,2012)

Twodimensionsgenerallycharacterizeenvironmentaluncertainty:complexityanddynamism(Child,1972;Thompson,

1967).AccordingtoThompson(1967)andTerreberry(1968),acomplexenvironmentisoneinwhichinteractiverelationships relevantfordecisionmakingrequireahighdegreeofabstractioninordertoproducemanageablemappings.Adynamic environmentisoneinwhichrelevantfactorsfordecisionmakingareinaconstantstateofchange

Priorresearchsuggeststhatcomplexityoftheenvironmentincreasesthedifficultyforinvestorstoassessearnings man-agement(Lim,Ding,&Thong,2008).Financialreportingisexpectedtobemorecomplexforfirmswithdiversifiedbusiness andgeographicaloperations.Hence,weexpectearningsmanagementtoincreasewiththelevelofdiversificationandtobe moredifficulttodetectbystockmarketparticipants

∗ Corresponding author Tel.: +1 514 987 3000; fax: +1 514 987 6629.

E-mail address: cormier.denis@uqam.ca (D Cormier).

1061-9518/$ – see front matter © 2013 Published by Elsevier Inc.

&Lev,2000).Insuchanasymmetricalcontext,wecanarguethatearningsmanagementismorelikelytooccur(e.g.Francis, LaFond,Olsson,&Schipper,2005)andlesslikelytobedetectedbymarketparticipants.Dynamicfirmsarealsocharacterized

bysalesvolatility,whichaffectsmanagerialdecisions(e.g.Child,1972;Cyert&March,1963;Williamson,1975).Itisassumed thatitisinmanagers’intereststoreducethevariabilityofreportedearnings(Ghosh&Olsen,2009;Gul,Chen,&Tsui,2003) butinavolatileenvironment,earningsmanagementisexpectedtobemoredifficulttodetectbecauseofalackofstabilityin accountingfigures.Therefore,wecanexpectearningsmanagementoccurringinadynamicenvironmenttoaffectinformation asymmetrytoalesserextentthaninastableenvironment

TheCanadianinstitutionalenvironmentmayprovidemanagersuniqueearningsmanagementopportunities.Anotable differencedistinguishestheCanadianfromtheU.S.equitymarket.IntheU.S.,ownershipinpubliclytradedfirmsishighly dispersedwhileinCanada,ownershipishighlyconcentrated.Infirmswithconcentratedownership,thereisareal possi-bilitythatdominantshareholdersmaymistreatorexpropriateoutsideshareholdersbyearningsmanipulation.Therefore, differencesinearningsmanagementopportunitiesmayexistbetweenCanadaandtheU.S.However,bylistingtheirshares

intheU.S.,Canadianfirmsbondthemselvestobetterdisclosureandgovernancepractices(Stulz,1999).Weexpectearnings managementpracticestodifferbetweenCanadianfirmsandU.S.firmsaswellastheirimpactoninformationasymmetry.We considerthatCanadaprovidesauniquecontexttoassesstheincidenceofearningsmanagementoninformationasymmetry, especiallyinanuncertainenvironment

Ourmainfindingsarethefollowing.First,weobservethatenvironmentalcomplexityanddynamismleadtoearnings man-agement,especiallyforfirmsnotcross-listedonaU.S.stockmarket.Second,insuchcontexts,earningsmanagementaffects informationasymmetrytoalesserextent.Third,ourfindingssuggestthatinthepresenceofcomplexityanddynamism, discretionaryaccrualsaremorelikelytobedetectedbyinvestorsforfirmslistedonaU.S.stockexchange.Thisresultis consis-tentwiththeviewthattheU.S.stockmarketismoreliquidandtransparentthantheCanadianmarketinthewayinformation

iscollectedandanalyzed,andthusisinabetterpositiontodetectearningsmanagementinacontextofuncertainty.Finally, thequalityofcorporategovernanceisassociatedwithlessearningsmanagement.Overall,ourfindingssupporttheview thataccrualanomalyispartlydrivenbyinvestors’failuretocorrectlyassessfutureearningsimplicationsofaccruals(Gong,

Li,&Xie,2009)

Comparedwithpriorresearch,thepresentstudyinnovatesbyinvestigatinghowthelevelofcomplexityanddynamism affectsthestockmarket’sassessmentofearningsmanagement.Bycombiningananalysisofspecificenvironmentalincentives

toengageinearningsmanagementwithanassessmentoftheincidenceoninformationasymmetry,thisstudycontributes

totheunderstandingofearningsmanagementimplications.Resultsalsosuggesttheimportanceofcontrollingforthe endogenousnatureofearningsmanagement,namelycorporategovernancemechanisms

Theremainderofthepaperisorganizedasfollows.Section2presentsthetheoreticalbackgroundanddevelops hypothe-ses.Thestudy’smethodologyisdescribedinSection3.ResultsarepresentedinSection4.Finally,Section5providesa discussionofthepotentialimplicationoftheresults

2 Background and hypotheses

2.1 Stockmarketassessmentofearningsmanagement

Priorresearchshowsarelationshipbetweenearningsmanagementandinformationasymmetryinthemarketplace Measuringaccrualsmanagementasthestandarddeviationofresidualsfromregressionsrelatingcurrentaccrualstocash flows,Francis,LaFond,Olsson,andSchipper(2004)andFrancisetal.(2005)findthatearningsmanagementisassociated withhigherinformationasymmetry,leadingtohighercostsofdebtandequity.Theyalsoshowthatinvestorsputmore importance(inthedeterminationofthecostofcapital)onaccrualsthatreflectintrinsicfeaturesofthefirm’sbusiness model,relativetoaccrualsthatreflectacombinationofpurenoiseandopportunisticchoicesandmanagement’sattemptsto makeearningsmoreinformative.Inthesamevein,LiuandWysocki(2007)arguethatthedocumentedrelationshipbetween accrualmanagementandthecostofcapitalisdrivenprimarilybythevolatilityoffirms’operatingactivitiesthatarenot relatedtoaccountingchoicesandlesssubjecttomanagerialmanipulation

However,priorevidencealsosuggeststhatinvestors’abilitytoassessearningsmanagement maybeimperfect.The market’sinabilitytofullydetectearningsmanagementisassociatedwithanincreaseintheheterogeneityofmarketbeliefs (Ronen&Yaari,2008).Theaccrualanomalyischaracterizedbystockmarketsoverweightingtheaccrualpersistence.Pincus, Rajgopal,andVenkatachalam(2007)findthatstockpricestendtooverweighttheroleofaccrualspersistence,especially discretionaryaccruals.Theauthorsobservenegativeabnormalreturnsinyeart+1forfirmswithpositivediscretionary accrualsinyeart.ThisisparticularlythecaseforcountrieshavingcommonlawtraditionssuchasAustralia,Canada,theUK andtheU.S.SoaresandStark(2009)reachthesameconclusionforaBritishsample(1989–2004)sincetheyfindthataverage annualabnormalreturnsgenerallydeclineaspriorperiodaccrualsmovefromlowtohigh.Thisoutcomeisconsistentwith theaccrualsanomalysinceinvestorsoverweightthepersistenceofaccrualsandunderweightthepersistenceofcashflows

inpredictingthenextperiod’searnings.Anexplanationoftheaccrualanomalyisthat,asdocumentedbyLevandNissim (2006),investorsavoidextreme-accrualsfirmsbecauseoftheirattributessuchassmallsize,lowprofitability,andhigh risk.LevandNissimalsoobservethatthehighinformationandtransactioncostsassociatedwiththeimplementationofa

profitableaccrualstrategyreduceprofitsforinvestorswhotradeonaccrualsinformation.Therefore,weexpectearnings managementtoincreaseinformationasymmetryinthestockmarket

2.2 Environmentaluncertaintyandstockmarketassessmentofearningsmanagement

Thevarioustermsusedtodepictenvironmentaluncertaintyfallgenerallyintotwocategories(Child,1972;Terreberry, 1968;Thompson,1967):complexityanddynamism(orinstability).Mintzberg(1979)describesthreedimensionsofthe environmentsimilartothoseproposedbyChild(1972)butaddsnewfeaturesforeach.Heintroducesthetermmarket diversificationtoreflectwhatThompson(1967)meansbyheterogeneityandChild(1972)bycomplexity.Mintzberg(1979)

reservesthetermcomplexityforthedegreeofsophisticatedknowledgenecessarytooperateinagivenenvironmentofa technicalorscientificnature

Environmentaldynamismreferstotherateofchangeandinstabilityoftheenvironment(Dess&Beard,1984).Dynamic environmentsarecharacterizedbychangesinvariousmarketelements,suchascustomerpreferences,technology,and competitorstructure.AccordingtoDessandBeard(1984),uncertaintyistheoutcomeofcomplexityanddynamism.Earnings managementinacontextofuncertaintyislikelytoreinforcetheaccrualanomalygiventhedifficultyforinvestorstodetect earningsmanagementinsuchacontext.Ifsuchisthecase,theimpactofearningsmanagementoninformationasymmetry willbelowerinanuncertainenvironmentthaninastableenvironment

Inthisstudy,weassessenvironmentaluncertaintybyreferringtocomplexityanddynamism.Weassertthatacomplex anddynamicenvironmentweakenstherelationshipbetweenearningsmanagementandinformationasymmetry

2.2.1 Complexity

Thereismixedevidenceconcerningtherelationshipbetweencomplexityandinformationasymmetry.Accountingis morecomplexforfirmswithdiversifiedbusinessandgeographicaloperations.Agrawal,Jaffe,andMandelker(1992)find thatacquisitionsthatarefocusdecreasing(negativefocus),suchasthosemotivatedbydiversificationorfinancialmotives likeasset-stripping,experiencesuperiorlong-termperformancethanacquisitionswhicharefocuspreservingorincreasing (positivefocus).Incontrast,Megginson,Morgan,andNail(2004)findevidenceofpositivefocusoutperformance,andhence

apositiverelationshipbetweencorporatefocusandlong-termacquisitionperformance.ErwinandPerry(2000)examine theeffectofforeignacquisitionsbyU.S.firmsonanalystforecasterrors.Forasampleoffocus-preservingforeignacquisitions andfocus-decreasingacquisitions,theyfindthatpost-mergeranalystforecasterrorsaresignificantlyhigherforU.S.firms thatchoosetoexpandinternationallyoutsidetheircorebusinesssegment,relativetothosethatundertakeglobalexpansion withintheircorebusiness.However,usingearningsforecasterrorsandearningsforecastdispersiontomeasureinformation asymmetry,Jiraporn,Miller,Yoon,andYoungSang(2008)showthatdiversifiedfirmsdonotsuffermoresevereinformation asymmetrycomparedwithnon-diversifiedfirms

Thomas(2002)reachesthesameconclusionasJirapornetal.(2008)examiningtherelationshipbetweencorporate diversificationandasymmetricinformationproxiesderivedfromanalystforecasterrorsandforecastdispersion,aswellas abnormalreturnsassociatedwithearningsannouncements.Thomas(2002)arguesthat,eveniftheerrorsoutsidersmake

inforecastingsegmentcashflowsarelargerthantheerrorstheymakeinforecastingfocusedfirmcashflows,iftheseerrors arenotperfectlypositivelycorrelated,theconsolidatedforecastmaybemoreaccuratethanaforecastforafocusedfirm Concerningtherelationshipbetweendiversificationandearningsmanagement,Limetal.(2008),basedonasampleof seasonedequityoffering,findthatcurrentdiscretionaryaccrualsarehigheramongdiversifiedfirmsthanfornon-diversified ones.Theirevidenceisconsistentwiththeviewthatearningsmanagementisrelatedtotheextentofafirm’sdiversification Hence,giventheincreaseinthedifficultyforinvestorstoassessearningsmanagementfordiversifiedfirms,weexpectthe positiverelationshipbetweenearningsmanagementandinformationasymmetrytobeweakenedforfirmswithdiversified operations,i.e.,businessoperationsandgeographicaloperations

H1. Acomplexenvironmentweakenstherelationshipbetweenearningsmanagementandinformationasymmetry

2.2.2 Dynamism

Dynamicfirms,especiallythosehighlyinvolvedinintangibleassets,attractmarketparticipantsandespeciallyfinancial analysts(Barth,Kasznik,&McNichols,2001).Thosefirmsofferagrowthpotentialandarethenmorescrutinizedbyinvestors AccordingtoMilliken(1987),afirm’stopmanagementismostlikelytoexperiencetheso-called“responseuncertainty” eitherinthecourseofchoosingfroma numberofpossiblestrategiesorinthecourseofformulatingaresponsetoan immediatethreatintheenvironment.InvestingintensivelyinR&Dmaybearesponsetosuchuncertainty.AboodyandLev (2000)arguethattherelativeuniquenessofR&Dinvestmentsmakesitdifficultforoutsiderstolearnabouttheproductivity andvalueofagivenfirm’sR&Dfromtheperformanceandproductsofotherfirmsintheindustry.Furthermore,while mostphysicalandfinancialassetsaretradedinorganizedmarkets,R&Disnot.NoassetpricesderivedirectlyfromR&D information.Thus,thelevelofR&Dcontributestoinformationasymmetry.Giventhescarcityofpublicinformationabout R&Dactivities,AboodyandLev(2000)hypothesizeanddocumentthatR&Dcontributestoinformationasymmetrybetween firmsandinvestors.Insuchacontext,wecanarguethatearningsmanagementismorelikelytooccur(e.g.Francisetal.,

2005)andlesslikelytobedetectedbymarketparticipants

Milliken(1987)definestheso-called“effectuncertainty”asaninabilitytopredictwhatthenatureoftheimpactofafuture stateoftheenvironmentorenvironmentalchangewillbeontheorganization.Inacontextofvolatility,firmsarefacingthe effectuncertainty.Volatilityisapowerfulcontextualfactoraffectingmanagerialdecisionsfordynamicfirms(Child,1972; Cyert&March,1963;Williamson,1975).Priorresearchsuggeststhatmanagershaveincentivestoreducethevariabilityof reportedearnings(e.g.Guletal.,2003;Wang&Williams,1994).LeblebiciandSalancik(1981)findthatbankloanofficers searchformoreinformationwhenmakingloandecisionsinavolatileenvironment.Accountingstandardsprovidesadegree

offlexibility,givingopportunisticdiscretiontomanagersinreportingearningsinanattempttoreducethevariabilityin reportedearningsviaaccrualmanagement(e.g.Bannister&Newman,1996)

Salesorearningsvariabilityincreasesthelevelofdifficultyforinvestorstoassessearningsmanagement.Inthecontext

ofsalesorearningsvolatility,managershaveincentivestoreducethevariabilityofreportedearnings(Guletal.,2003) Forfirmswithunstablesalesorearnings,earningsmanagementmightbedifficulttodetectbymarketparticipants.Prior research(e.g.Ghosh&Olsen,2009)showsthatmanagersusediscretionaryaccrualstoreducethevariabilityinreported earningswhenfacinghighsalesvolatility.Inourview,inacontextofhighsalesvariation,itismoredifficultforinvestorsto assessearningsmanagement.Therefore,weexpectthepositiverelationshipbetweendiscretionaryaccrualsandinformation asymmetrytobeweakenedforfirmsinvolvedinresearchanddevelopmentactivitiesandthosefacingunstablesales Thisgivesrisetothesecondhypothesis:

H2. Adynamicenvironmentweakenstherelationshipbetweenearningsmanagementandinformationasymmetry

3 Method

3.1 Sample

Wefixedthesamplebasedonaninitialdatasetinwhichwecollecteddataaboutcorporategovernance.Ourstarting pointis189non-financialfirmsrepresentingtheTorontoStockExchangeS&P/TSXIndexidentifiedin2002.Mergersand acquisitions,bankruptciesanddelistingsreducedoursamplefrom189to155in2005.Foraninitialsampleof155firms, therearemissingdataforgovernancevariablesandsharepricevolatility(18firms).Thesamplecomprises137firmslisted

ontheTorontostockexchange(TSE)in2005(133firmsforthebid-askspreadestimationmodel).Theresulting137firms represent80%ofthemarketcapitalizationofnon-financialfirmslistedontheTSE

Governancevariablesarecollectedfromproxystatements.Sharepricevolatilityandbid-askspreadaremeasuredfor theyear2005whilefinancialandgovernancevariableswerecollectedbasedontheinformationavailableinStockGuide(a Canadiandatabase)inthespringof2005,i.e.2004financialstatements.Bid-askspreadiscollectedfromCanadianMarket ResearchCenterdatabase(CFRMTSX).Samplefirmsoperateinthefollowingindustries(one-digitSICcodes):materials; energy;consumerproducts&services;industrials;Informationtechnology

3.2 Empiricalmodel

Thisstudyattemptstoprovideanintegratedanalysisofearningsmanagementandinformationasymmetry.Wereferto thefollowingsimultaneousequations:

ABSDACCi,t=ˇ0+ˇ1SIZEi,t+ˇ2INDDIRi,t+ˇ3CEOCHAIRi,t+ˇ4BODSIZEi,t

+ˇ5BODSIZESQRi,t+ˇ6ACSIZEi,t+ˇ6ACSIZEi,t (1)

SPV/BASi,t=ˇ0+ˇ1SYSRISKi,t+ˇ2FFLOi,t+ˇ3ANFOLLi,t+ˇ4ABSDACCi,t+ˇ5ABSDACC∗USi,t

+ˇ6ABSDACC∗ENVUNCi,t+ˇ7ABSDACC∗ENVUNC∗USi,t+ˇ8ENVUNCi,t+ˇ9ENVUNC∗USi,t+ˇ10USi,t (2) whereABSDACCistheabsolutevalueofestimateddiscretionaryaccrualsscaledbylaggedtotalassets;SPVistheshareprice volatility;BASisthebid-askspread;SIZEisthenaturallogarithmoftotalassets;INDDIRisthepercentageofindependent directorsontheboard;CEOCHAIRistheindicatorvariabletakingthevalueof1iftheCEOalsoistheChairmanoftheboard,

0otherwise;BODSIZEisthenumberofdirectorsontheboard;BODSIZESQRisthesquareofthenumberofdirectorsonthe board;ACSIZEisthenumberofmembersontheauditcommittee;SYSRISKisthesystematicrisk(beta);FFLOisthepercentage

offreefloat;ANFOLListhenumberofanalystsfollowingafirm;ENVUNCistheenvironmentaluncertainty(SEGMENTS,RD, SCV);USistheU.S.listing

3.2.1 Estimationofearningsmanagement

AccordingtoSchipper’s(1989)definition,earningsmanagementis“apurposefulinterventionintheexternalfinancial reportingprocess,withtheintentofobtainingsomeprivategain(asopposedto,say,merelyfacilitatingtheneutraloperation

oftheprocess)”.ConsistentwithHanandWang(1998)andEricksonandWang(1999),earningsmanagementisestimated

byacross-sectionalregressionforthe2001to2003period(155firms× 3years=465firm-yearobservations).The estima-tionofnormalaccrualsiscross-sectionalbasedonindustryspecificobservations(5industriesfor3years).Industry-level

estimationsremovethevariationinthenormalaccrualsthatiscommonacrossfirmsinthesameindustry(Burgstahler, Hail,&Leuz,2006;Dechow,Sloan,&Sweeney,1995;Dechow,Sloan,&Sweeney,1996;Francisetal.,2005;Leuz,Nanda,& Wysocki,2003).1Observationsvaryamongindustries:from43forindustrialsto174forconsumerproductsandservices

Weintroducediscretionaryaccrualsinabsolutevaluetocaptureearningsmanagement

HribarandCollins(2002)arguethatthedifferencebetweennetincomeandcashflowfromoperationsisthecorrect measureoftotalaccrualsandthattheuseofabalancesheetapproachmayleadtoasystematicbiasindiscretionaryaccruals estimation.Theyshowthatbalancesheetaccrualsestimatesarepredictablybiasedinstudieswherethepartitioningevent

iscorrelatedwitheithermergersoracquisitions,ordiscontinuedoperations.Theauthorsdemonstratethattestsofmarket mispricingofaccrualsareunderstatedduetoerroneousclassificationof“extreme”accrualsfirms.2 Whileafirm’stotal accrualsareeasilyaccessiblefromitsfinancialstatements,normalanddiscretionaryaccrualsarenotdirectlyobservable andmustbeinferredthroughanestimationmodel

Jones’s(1991)modelperformspoorlyforfirmswith“extreme”performance(Dechowetal.,1995;Kothari,Leone,& Wasley,2005).Sincewefocusonenvironmentaluncertaintythatislikelytoinvolvefirmswithextremeperformance(e.g highsalesvolatility),wewillrelyonDechowandDichev(2002)(DD)whomodelaccrualsasafunctionofcurrent,past, andfuturecashflows.Accrualsanticipatefuturecashcollections/paymentsandreversewhencashpreviouslyrecognized

inaccrualsiscollected/paid.DDfocusonshort-termworkingcapitalaccruals.However,accordingtoMcNichols(2002),DD modeldoesnotcontrolforfundamentalfactorsinfluencingaccruals.WebaseaccrualsestimationonDDmodelasmodified

byMcNicholscombiningDDandJonesmodels.BasedonJonesmodel,fixedassetsandyear-to-yearchangeinsales(a firm’sunderlyingperformance)areaddedtotheregressionestimations.Wefocusonworkingcapitalaccruals,whichare accrualsnetofdepreciation,amortizationandunusualitems.Unusualitems(field70inStockGuidedatabase)includesuch elementsasnon-recurringgainsandlosses,restructuringprovisions,write-offs,andothernon-operatinggainsorlosses Thisisconsistentwiththeviewthatshort-termaccrualsarehardertodetectbymarketparticipantsthanarelong-term accruals(e.g.impairmentofassets)

Foragivenindustry,workingcapitalaccrualsaremodeledinthefollowingmanner:

WCaccrualsit =˛1+˛2Operatingcashflowit−1+˛3Operatingcashflowit+˛4Operatingcashflowit+1

+˛4Changeinsalesit+˛4Fixedassetsit+εit (3)

Thecoefficientsfromtheaboveregressions(variablesscaledbylaggedtotalassets)arethenusedtocomputenormal accruals.Meanindustrycoefficientsareasfollows(averageadjustedR-squareof51.3%)3:

−0.065Operatingcashflowit−1−0.568Operatingcashflowit+0.271Operatingcashflowit+1

+0.092Changeinsalesit−0.029Fixedassetsit

3.2.2 Determinantsofearningsmanagement

Firmsize(SIZE).Largefirmstakeintoaccountthereputationcostswhenengaginginearningsmanagement.Thesefirms mayhaveabetterknowledgeofthemarketenvironment,bettercontrolovertheiroperationsandbetterunderstandingof theirbusinessesrelativetosmallfirms(Kimetal.,2003).Largefirmsmayhavemoresophisticatedinternalcontrolsystems andhave morequalifiedinternalauditorsascomparedtosmallfirms.Moreover,priorresearchdocumentsa negative associationbetweenfirmsizeandextremeaccruals(Lev&Nissim,2006).Hence,weexpectfirmsizetobenegativelyrelated

toearningsmanagement

Corporategovernance.Weexpectthequalityofcorporategovernancetobenegativelyrelatedtoearningsmanagement Corporategovernancemitigatesthedegreeofearningsmanagementandimprovesthequalityoffinancialreporting(Beasley, Carcello,Hermanson,&Lapides,2000;Warfield,Wild,&Wild,1995).Governancevariablesareintroducedtocapturehow corporategovernanceactingasamonitoringfactoraffectsearningsmanagement.Theboard’smonitoringinfluences man-agerialdiscretionandinducesfirmstomoretransparencyinorganizationalperformancemeasurementandreporting(Eng

&Mak,2003;Fama,1980).Fivevariablesareusedtocapturetheboardeffectiveness:Independentdirectors(INDDIR);board

1 Dechow et al (2010) emphasize that accruals models can be estimated at the firm level, which allows variation across firms in the determinants

of normal accruals However, firm-level estimation assumes time-invariant parameter estimates and imposes sample survivorship biases Therefore, the models are most frequently estimated at the industry level This specification assumes constant coefficient estimates within the industry Hence, the authors argue that some firms may have large residuals because of variation induced by industry membership rather than because of earnings management.

2 The balance sheet method of accruals may create bias in the estimation of normal accruals when the firm is involved in mergers and acquisitions Hence, the change in working capital accounts can be affected by the operation of mergers and acquisitions without any earnings management intention The Canadian stock market is very active in mergers and acquisitions.

3 Coefficients do not vary substantially when we estimate accruals based on a cross-sectional regression for the whole sample (Adjusted R-square; 51.7%).

WC accrualsit = −0.125 Operating cash flowit−1

(0.000)

− 0.465 Operating cash flowit

(0.000)

+ 0.260 Operating cash flowit+1

(0.000)

+ 0.072 Change in salesit (0.000)

− 0.014 Fixed assetsit (0.257)

+ εit

chairduality(CEOCHAIR);boardsize(BODSIZE);boardsizesquared(BODSIZESQR);andauditcommitteesize(ACSIZE).4Ghosh, Marra,andDoocheol(2010)findthatearningsmanagementdoesnotvarywithboardcompositionandstructure,orwith theauditcommitteecomposition,expertise,andownership.However,theyfindthatfirmswithsmallerboardsandaudit committeeshavelargerdiscretionaryaccruals.Somepriorstudiesassumetherelationshipbetweenboardsizeandboard performancetobecurvilinear(e.g.Eisenberg,Sundgren,&Wells,1998;Golden&Zajac,2001;Vafeas,1999;Yermack,1996)

Tocontrolforthepossiblecurvilinearityintherelationshipbetweenboardsizeandearningsmanagement,weincludethe variableboardsizesquaredasaninstrument.WetreatABSDACCasanendogenousvariableandgovernancevariablesas instruments.5

3.2.3 Determinantsofinformationasymmetry

Severalproxiesexisttocaptureinformationasymmetry.Welker(1995),Healy,Hutton,andPalepu(1999),Leuzand Verrecchia(2000),andFrancisetal.(2005)showthattheextentofinformationasymmetry– proxiedbybid-askspread, sharepricevolatilityorstockliquidity(tradingvolume)–isnegativelyassociatedwithdisclosurequality.Inthecurrent study,weusesharepricevolatilityandbid-askspreadtoassessinformationasymmetry.Sharepricevolatilityismeasured

asthestandarddeviationofpercentagechangesindailystockpricesfortheyear2005

Priorstudiessuggestthatvariousfirmattributesarerelatedtoinformationasymmetry(Leuz&Verrecchia,2000).Based

onthatliterature,weuseanalystfollowing,systematicrisk,andfreefloataskeydeterminantsofsharepricevolatilityand bid-askspread.Thesefourvariablesareintroducedascontrolvariablesforinformationasymmetry

Systematicrisk(SYSRISK).Apositiverelationshipisexpectedbetweensystematicriskandinformationasymmetry.The higherafirm’ssystematicrisk(beta),themoredifficultitisforinvestorstoaccuratelyassessafirm’svalueandthemore likelytheyareexpectedtoincurinformationcoststoassessitsriskdrivers.Priorresearchdocumentsanassociationbetween systematicriskandthecostofcapital(e.g.Botosan&Plumlee,2005;Botosan,1997;Hail&Leuz,2006;Leuz&Verrecchia, 2000;Mikhail,Walther,&Willis,2004).BetaisextractedfromStockGuidedatabaseandiscomputedbasedonpercentage stockpricechangeweekoverweekforaperiodof260weeksendingattheendof2005fiscalyear

Freefloat(FFLO).Weexpectanegativeassociationbetweenfreefloatandinformationasymmetry.Controlblocksusually havegreateraccesstoprivateinformationthanadiffuseownership(Leuz&Verrecchia,2000).Freefloatisusedasan inverseproxyforinsidercontrol.Thevariableismeasuredasoneminusthepercentageofvotingsharesthatarecloselyheld (percentageofvotesattachedtothesharesofafirmheldbydirectors,andindividualsorcompaniesthatownmorethan 10%ofsharesoutstanding)

Analystfollowing(ANFOLL).Weexpectanegativerelationshipbetweenthenumberoffinancialanalyststhatfollowafirm

onanannualbasisandinformationasymmetry.Afirm’sanalystfollowingmayproxyfortheextentofitscommunication withfinancialanalysts(Leuz,2003).Priorresearchsuggeststhatanalystcoveragereducesasymmetryinthestockmarket (Alford&Berger,1999)

Environmentaluncertainty(ENVUNC).Environmentaluncertaintyreferstofirmcomplexityanddynamismandiscaptured

bythenumberofbusinessandgeographicalsegments(SEGMENTS),R&Dintensity(RD),andsalescoefficientofvariation (SCV).Weestimatethreeseparateregressionsadding inturninteractiontermsABSDACC*SEGMENTS,ABSDACC*RD, and ABSDACC*SCVtotheregressions.Wealsomeasureoveralluncertaintybythecombinationofthethreevariables

WeexpectcoefficientsoninteractiontermsABSDACC*SEGMENTS,ABSDACC*RD,andABSDACC*SCVtobenegatively associ-atedwithinformationasymmetry.ThevariableSEGMENTSproxiesforcomplexityandismeasuredasoneifthetotalnumber

ofsegmentsisgreaterthanthesamplemedian.DynamismiscapturedbythevariablesRDandSCV.RDisabinaryvariable measuredasR&Dscaledbysalesgreaterthanthesamplemedian.SCVisabinaryvariablethattakesthevalueofoneifthe salescoefficientofvariationisgreaterthanthesamplemedian.SCViscalculatedforeachfirmbasedonannualobservations from2000to2004.Priorresearch(e.g.Ghosh&Olsen,2009)showsthatmanagersusediscretionaryaccrualstoreduce variabilityinreportedearningsforfirmswithhighsalesvolatility.Inacontextofhighsalesvolatility,itismoredifficultfor investorstoassessearningsmanagement

U.S.listing(US).Managerialincentivesforearningsmanagementareaffectedbytheintensitywithwhichafirmis mon-itoredbyinvestorsortheiragentssuchassecuritiesregulatorsandfinancialanalysts(Healyetal.,1999).TheSEChas traditionallybeendiligentinitspursuitoffirmswithinappropriatedisclosureandreportingpractices(e.g.,Feroz,Park,

&Pastena,1991).Moreover,marketefficiencyisenhancedifthenumberofanalystsfollowingthefirmincreases(Lang& Lundholm,1996).Thenumbersinvolved,aswellasthegeographicaldispersionofanalystsintheUnitedStates,ensurethat firmsfaceacriticalaudienceandhighscrutinywhenreportingtoinvestors.Inthisvein,basedonasampleofCanadianoil andgasfirms,CormierandMagnan(2002)showthatmanagerialincentivesforearningsmanagement,aswellasinvestors’ appreciationofvariousperformancemetrics,canbeaffectedbythemagnitudeofexternalmonitoringthataccompaniesa U.S.stockexchangelisting

4 In Canada, audit committees must comprise at least three independent members We consider that adding a few more members could enhance the monitoring role of the audit committee.

5 Prior studies document that the presence of an independent and competent board of directors should limit managers’ ability to manage earnings ( Klein, 2002; Peasnell et al., 2003 ) More recently, Chang and Sun (2009) show that the passage of the Sarbanes-Oxley Act (SOX) improves the effectiveness of an

Table 1

Descriptive statistics.

N = 137 Minimum Maximum Mean Standard deviation Mean Non-U.S listing (N = 69) Mean U.S listing (N = 68)

NACC is estimated normal accruals scaled by lagged total assets; DACC is estimated discretionary accruals scaled by lagged total assets; ABSDACC is the absolute value of estimated discretionary accruals scaled by lagged total assets; BAS is the bid-ask spread; SPV is share price volatility; ANFOLL is the number

of analysts following a firm; SYSRISK is Systematic risk (beta); FFLO is the percentage free float; RD is the amount of R&D in % of sales; SEGMENTS is the number of business and geographical segments; SCV is the sales coefficient of variation calculated for each firm based on annual observations from 2000

to 2004; INDDIR is the percentage of independent directors on the board; CEOCHAIR is an indicator variable taking on the value of 1 if the CEO also is the Chairman of the board; 0 otherwise; BODSIZE is the number of directors on the board; ACSIZE is the number of members on the audit committee; US is an indicator variable for U.S listing.

ABSDACC*ENVUNC*UStoassesstheeffectofU.S.cross-listingoninformationasymmetryandexpectU.S.marketstobeina betterpositiontodetectearningsmanagementincontextsofcomplexityanddynamism.Weexpectanegativerelationship betweentheinteractiontermABSDACC*ENVUNC*USandinformationasymmetry

4 Results

4.1 Descriptivestatistics

Table1providessomedescriptivestatistics.Samplefirmsarequitelarge(totalassetsaveraging$5billion).About78%

ofsamplefirmsarefreefloat.Firmsarefollowedonaveragebysevenfinancialanalysts.Discretionaryaccrualsareon average1.3%oftotalassetswhiletheabsolutevalueoftotaldiscretionaryaccrualsaverages5.2%oftotalassets.6Onaverage, firmsinvestinR&D inaproportionof4.0%oftotalassets,operateinclosetofivebusinessdivisionsandgeographical segments,andexhibithighsalesvolatility(meancoefficientofsalesvariationof29%).Moreover,36%ofsamplefirmshave independentdirectorswhile20%arepresentingCEOandboardchairduality.Halfofsamplefirmsarecross-listedonaU.S stockexchange.Wealsoobservemoreenvironmentaluncertainty(segments,R&Dintensity,andsalesvolatility,)for cross-listedfirms.Finally,thelevelofdiscretionaryaccruals(inabsolutevalue)ismuchlessimportantforcross-listedfirms(0.048 versus0.056)comparedwithnon-cross-listedfirms.Therearenosignificantdifferencesingovernancevariablesamong cross-listedandnon-cross-listedfirms

InTable2,wepresentdiscretionaryaccrualsinabsolutevalue(ABSDACC)for highandlowlevelsofenvironmental uncertainty,namely,salescoefficientofvariation,numberofsegments,andresearchanddevelopmentintensity.Results confirmourexpectations.Wealsopresentdifferencesfornon-U.S.versusU.S.listing.WeobservemoreABSDACCfornon-U.S listingforfirmsthatarefacinghighuncertainty.U.S.listingfirmsdonotshowsignificantdifferencesinthelevelofABSDACC betweenhighandlowuncertainty

Table3presentscorrelations.Consistentwithourexpectation,ABSDACCispositivelyassociatedwithsharepricevolatility (SPV)(0.24).WeobserveapositiveassociationbetweenSPVandR&Dintensity(RD)(0.17),whichisinlinewithpriorevidence (e.g.Aboody&Lev,2000).WealsoobserveanassociationbetweenABSDACCandboardsize(BODSIZE)(−0.12)andaudit committeesize(ACSIZE)(−0.17).ThisresultshowstheimportancetotreatABSDACCendogenously.Moreover,weobservea positiverelationshipbetweenABSDACCandsalesvolatility(SCV)(0.22).Finally,U.S.listing(US)isassociatedwithsystematic

6 We split the sample based on the median of a measure of an overall uncertainty (combining SEGMENTS, RD and SCV: 1 if two or three uncertainty measures are larger than the sample median, 0 otherwise) Results (not tabulated) show that normal accruals are equals among the two groups (2.9%

Table 2

Descriptive statistics Mean discretionary accruals in absolute value based on dynamism, diversity and volatility.

p value

SEGMENTS is the number of business and geographical segments; RD is the amount of R&D in % of sales; SCV is the sales coefficient of variation calculated for each firm based on annual observations from 2000 to 2004.

Table 3

Correlations.

1 SPV * 0.48 * 0.24 * 0.26 0.13 * −0.44 * −0.14 0.04 0.05 * −0.38 * −0.33 * 0.17 −0.04 * 0.13 0.02

3 ABSDACC 1 * 0.13 0.06 * −0.26 * −0.18 0.08 −0.05 * −0.12 * −0.17 * 0.10 −0.03 * 0.22 −0.08

a SPV is share price volatility; BAS is the bid-ask spread; ABSDACC is the absolute value of estimated discretionary accruals scaled by lagged total assets; SYSRISK is Systematic risk (beta); FFLO is the percentage of free float; SIZE is the natural logarithm of total assets; ANFOLL is the number of analysts following

a firm; INDDIR is the percentage of independent directors on the board; CEOCHAIR is an indicator variable taking on the value of 1 if the CEO also is the Chairman of the board; 0 otherwise; BODSIZE is the number of directors on the board; ACSIZE is the number of members on the audit committee; RD is the amount of R&D in % of sales; SEGMENTS is the number of business and geographical segments; SCV is the sales coefficient of variation calculated for each firm based on annual observations from 2000 to 2004; US is an indicator variable for U.S listing.

* Significant at 0.10.

risk(SYSRISK)(0.29),freefloat(FFLO)(0.17),firmsize(SIZE)(0.14),analystfollowing(ANFOLL)(0.14),RD(0.20)andcomplexity (SEGMENTS)(0.16).Bid-askspread(BAS)ispositivelyassociatedwithABSDACC(0.28)andnegativelyassociatedwithSIZE (−0.31),ANFOLL(−0.19),andBODSIZE(−0.14)

4.2 Multivariateanalyses

Sincewepositthatcorporategovernanceaffectsearningsmanagementandinformationasymmetrysimultaneously,

we firstdetermine whetheran interactionexists betweenthese variables using Hausman test The Studentt-test of thecoefficientforthevariableresidualsconstitutestheHausmantest.Usingthisprocedure,fortheoveralluncertainty regression,we rejectthenull hypothesisof noendogeneity withrespecttoSPV andABSDACC (t-test=−4.57;p<0.00) andwithrespecttoBASandABSDACC(t-test=−3.65;p<0.00).Therefore,wetreatABSDACCasanendogenousvariable

Inlightofthisdiagnostic,werelyonathree-stageestimationprocedureforasystemofsimultaneousequations.3SLS (whichcombines2SLSandSeeminglyUnrelatedLeastSquare–SURE)mayimprovetheefficiencyofparameterestimates whenthere iscontemporaneous correlationoferrors acrossequations.Moreover,thegreatertheintra-equation mul-ticollinearity,themorelikely 3SLSis tohave aconsiderablegain inefficiencyfortheentiresystemofSURE (Binkley,

1982) In practice, thecontemporaneous correlation matrix is estimatedusing OLS residuals For overall uncertainty regressions,weobserveasignificantcorrelationoferrorsacrossequations(−0.14betweenABSDACCandSPVequations and −0.10between ABSDACC and BASequations) Concerningintra-equation multicollinearity, we observethat some interactionterms arehighly correlated ABSDACCis correlated at 0.86with ABSDACC*SCVand 0.87 withABSDACC*RD Sincemulticollinearitycouldbeanissue,SUREislikelytoimprovetheefficiencyoftheentiresystem(Binkley,1982) STATAsoftwareisbeingused.Finally,weexcludefromregressionsobservationswithstandardizedresidualsexceeding two

Table 4

3SLS estimation of the relationship between discretionary accruals and share price volatility in interaction with environmental uncertainty.

Environmental uncertainty Complexity segments Dynamism R&D Dynamism sales volatility Overall uncertainty

ABSDACC regression

SPV regression

Coefficient difference ABSDACC and

ABSDACC*ENVUNC

Coefficient difference ABSDACC and

ABSDACC*US

Coefficient difference

ABSDACC*ENVUNC and

ABSDACC*ENVUNC*US

ABSDACC is the absolute value of estimated discretionary accruals scaled by lagged total assets; SPV is share price volatility; SIZE is the natural logarithm

of total assets; INDDIR is the percentage of independent directors on the board; CEOCHAIR is an indicator variable taking on the value of 1 if the CEO also

is the Chairman of the board, 0 otherwise; BODSIZE is the number of directors on the board; ACSIZE is the number of members on the audit committee; SYSRISK is the systematic risk (beta); FFLO is the percentage of free float; ANFOLL is the number of analysts following a firm; ENVUNC is the environmental uncertainty variable; US is an indicator variable for U.S listing.

* p < 0.10.

** p < 0.05.

*** p < 0.01.

One-tailed if directional prediction, two-tailed otherwise.

Endogenous variables: ABSDACC, SPV.

InTable4,wepresentresultsofthe3SLSregressiononthedeterminantsofearningsmanagement(ABSDACC)and informa-tionasymmetry(SPV,BAS).Concerningtheimpactofcorporategovernanceonearningsmanagement,resultsfirstshowthat thenumberofmembersoftheauditcommittee(ACSIZE)isnegativelyrelatedtoABSDACCinallfourregressions.Coefficients

onthevariableBODSIZESQRarepositiveandsignificant,whichisconsistentwithanon-linearrelationshipbetweenboard sizeandearningsmanagement.Afteracertainbreakpoint,thesizeoftheboardisassociatedwithmoreearnings manage-ment.OurresultsaresomewhatconsistentwithGhoshetal.(2010)whofindthatearningsmanagementvarieswithboard sizeandauditcommitteesize.Finally,consistentwithpriorresearch,firmsize(SIZE)isnegativelyrelatedtoABSDACC Concerningthedeterminantsofsharepricevolatility(SPV),resultsareconsistentwithourhypotheses.Asexpected, systematicrisk(SYSRISK)ispositivelyrelatedtoSPV.Coefficientsonthemaineffectforenvironmentaluncertainty,i.e SEGMENTS,RD,SCVandoveralluncertaintyaresignificantandpositivelyrelatedtoSPV.ConcerningtheeffectofU.S.listing, coefficientsontheinteractiontermABSDACC*USarenegativeandsignificantinallfourregressions,whichsuggeststhatin thepresenceoflowuncertainty,earningsmanagementcreateslessasymmetryforcross-listedfirms.Forallregressions, TheStudentt-testsshowthatthesumofcoefficientsonABSDACCandABSDACC*USisstatisticallydifferentfromzero.Inthe presenceoflowenvironmentaluncertainty,thereisstillasymmetrycreatedbyearningsmanagementforU.S.listedfirms Thelowlevelofearningsmanagementunderlowuncertaintyforcross-listedfirmscouldalsoexplainitssmallimpacton informationasymmetry(seeTable2)

Complexity.ConsistentwithH1,therelationshipbetweenABSDACCandSPVissmallerforfirmswithoperationsand geographicaldiversificationsincethecoefficientontheinteractiontermABSDACC*SEGMENTSisnegativeandsignificant (−37.937;p<0.05)whilethecoefficientonABSDACCispositiveandsignificant(44.001;p<0.01).TheStudentt-testfor

Table 5

3SLS estimation of the relationship between discretionary accruals and bid-ask spread in interaction with environmental uncertainty.

Environmental uncertainty Complexity segments Dynamism R&D Dynamism sales volatility Overall uncertainty

ABSDACC regression

BAS regression

Coefficient difference ABSDACC and

ABSDACC*ENVUNC

52.43 (0.000) 29.40 (0.000) 66.41 (0.000) 11.64 (0.000) Coefficient difference ABSDACC and

ABSDACC*US

Coefficient difference

ABSDACC*ENVUNC and

ABSDACC*ENVUNC*US

ABSDACC is the absolute value of estimated discretionary accruals scaled by lagged total assets; BAS is the bid-ask spread; SIZE is the natural logarithm of total assets; INDDIR is the percentage of independent directors on the board; CEOCHAIR is an indicator variable taking on the value of 1 if the CEO also

is the Chairman of the board, 0 otherwise; BODSIZE is the number of directors on the board; ACSIZE is the number of members on the audit committee; SYSRISK is the systematic risk (beta); FFLO is the percentage of free float; ANFOLL is the number of analysts following a firm; ENVUNC is the environmental uncertainty variable; US is an indicator variable for U.S listing.

* p < 0.10.

** p < 0.05.

*** p < 0.01.

One-tailed if directional prediction, two-tailed otherwise.

Endogenous variables: ABSDACC, BAS.

coefficientdifferenceonABSDACCandABSDACC*SEGMENTSis statisticallydifferentfromzero(t=37.78;p<0.000).This resultsuggeststhatuncertaintydoesnoteliminatetheeffectofearningsmanagementonsharepricevolatility

ThecoefficientontheinteractiontermABSDACC*SEGMENTS*USispositiveandsignificant(33.451;p<0.05).TheStudent t-testshowsthatthesumofcoefficientsonABSDACC*SEGMENTSandABSDACC*SEGMENTS*USisstatisticallyclosetozero (t=2.22;p<0.136).ForU.S.listedfirms,theimpactofearningsmanagementonasymmetrywouldnotdifferinthepresence

orabsenceofcomplexity.Asasensibilityanalysis,weusethelogarithmofSEGMENTSinsteadofabinaryvariable.Essentially, results(nottabulated)remainsimilartothosepresentedinTable4.Ourfindingsareconsistentwiththeargumentthatthe U.S.stockmarketismoreliquidandtransparentthantheCanadianmarketinthewayinformationiscollectedandanalyzed, andthus,isinabetterpositiontodetectearningsmanagementinacontextofcomplexity

Dynamism.ConsistentwithH2,therelationshipbetweenABSDACCandSPVissmallerforfirmsinvolvedinR&Dactivities sincethecoefficientontheinteractiontermABSDACC*RDisnegativeandsignificant(−13.774;p<0.01).TheStudentt-test forcoefficientdifferenceonABSDACCandABSDACC*RDisstatisticallydifferentfromzero(t=24.21;p<0.000).Thisresult suggeststhatuncertaintydoesnoteliminatetheeffectofearningsmanagementonsharepricevolatility.Thecoefficient

ontheinteractiontermABSDACC*RD*USispositiveandsignificant(13.590;p<0.01).TheStudentt-testshowsthatthesum

ofcoefficientsonABSDACC*RDandABSDACC*RD*USisstatisticallyclosetozero(t=0.00;p<0.964).ForU.S.listedfirms,the impactofearningsmanagementonasymmetrywouldnotdifferinthepresenceorabsenceofdynamismasexpressedby R&Dintensity

AlsoconsistentwithH2,therelationshipbetweenABSDACCandSPVissmallerforfirmswithhighsalesvolatilitysince thecoefficientontheinteractiontermABSDACC*SCVisnegativeandsignificant(−46.773;p<0.05).TheStudentt-testfor