Lucky CEOs and Lucky Directors

LuckyCEOsandLuckyDirectors

LUCIANA.BEBCHUK,YANIVGRINSTEIN,andURSPEYER∗

ABSTRACT

Westudytherelationbetweenopportunistictimingofoptiongrantsandcorporategovernancefailures,focusingon“lucky”grantsawardedatthelowestpriceofthegrantmonth.Optiongrantpracticesweredesignedtoprovideluckygrantsnotonlytoexecutivesbutalsotoindependentdirectors.LuckygrantstobothCEOsanddirectorsweretheproductofdeliberatechoices,notoffirms’routines,andweretimedtomakethemmoreprofitable.LuckygrantsareassociatedwithhigherCEOcompensationfromothersources,nomajorityofindependentdirectors,nooutsideblockholderonthecompensationcommittee,andalong-servingCEO.

THEOPPORTUNISTICTIMINGofexecutives’optiongrants—viabackdating,“spring-loading”basedontheuseofinsideinformation,orotherwise—hasattractedagreatdealofattention.TheSECandasmallarmyofprivatelawfirmshiredbycompanieshaveinvestigatedpastgrantpractices.Morethan200companieshavecomeunderscrutiny,anddozensofexecutivesanddirectorshavebeenforcedtoresign.

Workinfinancialeconomicshascontributedsubstantiallytoidentifyingtheexistenceofopportunistictiming.TheliteratureonthetimingofoptiongrantsbeginswiththeseminalworkbyYermack(1997),whoshowsthatstockpricesexhibitednegativeabnormalreturnspriortoagrantdateandpositiveabnor-malreturnsafterward.AboodyandKasznik(2000)andChauvinandShenoy(2001)suggestthatthesereturnpatternswerepartlyduetothemanipulationoffirms’informationdisclosures,whileLie(2005)providesevidencethatback-datingwasanimportantcauseoftheabnormalstockreturnsprecedingandfollowinggrantdates.Collins,Gong,andLi(2005),HeronandLie(2007),andNarayananandSeyhun(2006)showthatthepatternsofpre-andpostgrant

iswiththeHarvardLawSchoolandtheNationalBureauofEconomicResearch,

GrinsteiniswiththeJohnsonSchoolofManagement,CornellUniversity,andPeyeriswithIN-SEAD.Thispaperintegratestwodiscussionpaperscirculatedearlier,“LuckyCEOs”and“LuckyDirectors.”Forhelpfulcommentsandconversations,wearegratefultoananonymousrefereeandassociateeditor,JohnGraham(theEditor),NadineBaudot-Trajtenberg,JohnBizjak,AlmaCohen,CharlesHadlock,RandyHeron,IraKay,ErikLie,MPNarayanan,andworkshopparticipantsattheNBERcorporatefinancemeeting,theAmericanEconomicAssociationandAmericanLawandEconomicAssociationannualmeetings,theUniversityofDelawareSymposiumonBackdatingofStockOptions,TheUniversityofVirginiaLawandFinanceConference,Harvard,theHerzliaInterdisciplinaryCenter,LondonBusinessSchool,Tel-AvivUniversity,andYaleSchoolofManage-ment.Forfinancialsupport,wewouldliketothanktheJohnM.OlinCenterforLaw,Economics,andBusiness,theHarvardLawSchoolProgramonCorporateGovernance,andtheGuggenheimFoundation.

∗Bebchuk

2363

23R

TheJournalofFinance󰀆

returnswereinfluencedbytheadoptionoftheSarbanes-OxleyAct(SOX),whichisconsistentwiththeexistenceofbackdatinggiventhatSOXmadebackdatingmoredifficult.

Inthispaper,wecontributetounderstandingthecorporategovernancede-terminantsandimplicationsofopportunisticoptiontimingpractices.Overall,ouranalysisprovidessupportfortheviewthatopportunistictimingpracticesreflectgovernancebreakdownsandraisegovernanceconcerns.Inparticular,wefindthat:opportunistictimingwascorrelatedwithfactorsassociatedwithgreaterCEOinfluenceoncorporatedecision-making,suchasalackofama-jorityofindependentdirectorsoralong-servingCEO;grantstoindependentdirectorswerealsoopportunisticallytimed,andthistimingwasnotmerelyabyproductofsimultaneousawardstoexecutivesoroffirmsroutinelytimingalloptiongrants;andluckygrantstoindependentdirectorswereassociatedwithmoreCEOluckandCEOcompensation.Wealsofindthat,ratherthanbeingasubstituteforotherformsofcompensation,gainsfromopportunistictimingwereawardedtoCEOswithlargertotalcompensationfromothersources,andopportunistictimingwasnotdrivenbyfirmhabitbutrather,foranygivenfirm,theuseofsuchtimingwasitselftimedtoincreaseitsprofitabilityforrecipients.

Westudytheuniverseofallat-the-money,unscheduledoptiongrantsawardedtotheCEOsandindependentdirectorsofpubliccompaniesduringthedecadeof1996to2005.Ourinvestigationfocuseson“lucky”grants,whicharegrantsgivenatthelowestpriceofthemonth.Opportunistictimingcanresultinanabnormallyhighfractionofgrantsbeing“luckygrants.”Inoursample,about15%ofthegrantstoCEOsand11%ofthegrantstodirectorswereluckybeforetheadoptionoftheSOX.Wecontributetopriorandcurrentworkonopportunistictimingbyinvestigatingseveralquestionsandhypothe-sesthatarerelevanttoassessingthecorporategovernancesignificanceandthedeterminantsofopportunistictimingpractices.

Tobegin,weprovidethefirstevidencethatgrantstoindependentdirectorswereaffectedbyopportunistictiminginwaysthatlinkdirectorluckandCEOluck.Eventhoughtheexistenceofopportunisticallytimedoptiongrantstoex-ecutivesiswidelyrecognized,ithasgenerallybeenassumedthatindependentdirectors,whoplayanimportantoversightroleinthecurrentmodelofcorpo-rategovernance,didnotbenefitdirectlyfromsuchtiming.Weshow,however,thatawardstoindependentdirectorswerethemselvesopportunisticallytimed.Wefurthershowthatthetimingofdirectorgrantswasnotamerebyprod-uctofdirectorgrantsbeingawardedatthesametimeasexecutivegrants.Inparticular,foranygivenfirmandCEO,wefindthattheoddsofaCEOgrantbeingluckyweresignificantlyhigherwhentheindependentdirectorsofthefirmreceivedgrantsonthesamedate.Inaddition,directorgranteventsnotcoincidingwithawardstoexecutiveswerealsoopportunisticallytimedand,moreover,weremorelikelytobeluckywhentheCEOreceivedaluckygrantinthesameorprioryear.Wealsoshowthatthetimingofdirectorgrantswasnotabyproductoffirmsroutinelytimingalloptiongrants;allofourresultsconcerningdirectorluckcontinuetoholdwhenoneremovesfromthedata(thesmallnumberof)firmsthatprovidedluckygrantstoallgrantrecipients.By

LuckyCEOsandLuckyDirectors2365

providingevidencethattimingpracticeswerestructuredinamannerconsis-tentwithmakingindependentdirectorsbeneficiariesofthesepractices,wecontributeanimportantinputforassessingtheroleofindependentdirectorsbothinconnectionwithopportunistictimingpracticesandmoregenerally.Oursecondcontributionistoprovidethefirstanalysisoftherelationbe-tweenopportunistictimingandtotalreportedcompensation.Wetest,butdonotfindsupportfor,thehypothesisthatgainsfromopportunistictimingwereasubstituteforothermeansofcompensation.Controllingforsize,performance,tenure,andotherfirmandCEOcharacteristicsthatdeterminecompensation,CEOsbenefitingfromluckygrantsalsoreceivedsignificantlyhighertotalcom-pensationfromothersources.Also,consistentwiththehypothesisthatinde-pendentdirectorswhoreceivedluckygrantsthemselveswerelessinclinedtoprovideacheckoncompensationdecisions,wefindthattotalCEOcompensa-tionwashigher(controllingforstandardcompensationdeterminants)infirmsthatgrantedluckygrantstoindependentdirectors.

Inthecourseofanalyzingtherelationbetweentotalreportedcompensationandgainsfromopportunistictiming,wederiveestimatesofthelatter.Incon-trasttosuggestionsthatthepotentialgainstoCEOsfromopportunisticallytimedgrantsmighthavebeenratherlimited(see,forexample,Walker(2006)),our(conservative)estimatesindicatethatthesegainswererathersignificant.WeestimatethatthegaintoCEOsfromluckygrantsduetoopportunistictim-ingexceeded,onaverage,20%ofthereportedvalueofthegrant,andadded,onaverage,morethan10%totheCEO’stotalreportedcompensationfortheyear.1

OurthirdcontributionistoidentifyanassociationbetweenopportunistictimingofCEOs’andindependentdirectors’grantsandcertainaspectsoffirmgovernanceandmanagement.Inparticular,wefindthatopportunistictimingwascorrelatedwithvariablesassociatedwithCEOinfluenceovertheinternaldecision-makingprocesses:thelackofamajorityofindependentdirectorsontheboardandlongCEOtenure.Inaddition,wefindthatamajorityofinde-pendentdirectorsontheboardisonlyeffectiveatreducingCEOluckiftheindependentdirectorsthemselvesdidnotreceiveluckygrants.Wefurtherfindthat,althoughtheexistenceofanindependentcompensationcommitteewasnotitselfassociatedwithareducedlikelihoodofopportunistictiming,aninde-pendentcommitteewithatleastonelargeblockholderonitwasassociatedwithsuchareduction.Ourresultshighlightthattheeffectivenessofindependentdirectorscouldwelldependonfactorsotherthantheirformalclassificationasindependent.2

andJarrell(2009),Narayanan,Schipani,andSeyhun(2007),andWalker(2006)es-timatethegainsfromopportunistictiminginsamplesofbetween50and150firmsthathavecomeunderscrutiny.Incontrast,westudytheissuefortheentiresetofgrantstoCEOsatpubliccompanies.

2Bizjaketal.(2009)andCollins,Gong,andLi(2009)findacorrelationbetweentimingandlackofamajorityofindependentdirectorsontheboard.Unlikethecurrentpaper,however,thesestudiesdonotinvestigatehowtheassociationbetweenboardindependenceandtimingdependedonindependentdirectorsreceivingluckygrants,theydonotidentifytheassociationwefindbetweenopportunistictimingandCEOtenureaswellasthatbetweenopportunistictimingand

1Bernile

2366R

TheJournalofFinance󰀆

ThesefindingscomplementexistingresearchconcerningthelinkbetweenopportunistictimingofCEOgrantsandcertainaspectsofgovernancesuchasboardinterlocks(Bizjak,Lemmon,andWhitby(2009)),thequalityofthefirm’sauditor(HeronandLie(2009)),andthecompositionofthecompensationcom-mittee(Yermack(1997)).Ourfindingsalsocomplementandreinforceresearchshowingthatdirectorindependenceisassociatedwithimprovedcompensationanddisclosure(e.g.,ChhaochhariaandGrinstein(2009)andBeasley(2000)),thatCEOtenureiscorrelatedwithincreasedCEOinfluenceoncompensa-tiondecisions(e.g.,Core,Holthausen,andLarcker(1999)andHarfordandLi(2007)),andthatoutsideblockholdersimprovecompensationarrangements(BertrandandMullainathan(2000,2001)).

Finally,withpriorworkfocusingoncross-sectionaldifferencesamongfirms,wecontributebyprovidingatime-seriesfixed-effectpaneldataanalysisofopportunistictiming,controllingforunobservablefirmandCEOcharacteristicsthatcouldbecorrelatedwithopportunistictiming.Thisanalysisallowsustotestthehypothesisthat,amongthefirmsengagedinopportunistictiming,suchtimingwasmerelytheproductofroutine.Inconsistentwiththisview,wefindthatforanygivenCEOandfirm,grantstobothCEOsanddirectorsweremorelikelytobeluckyinmonthsinwhichthepotentialpayoffsfromsuchluckwererelativelyhigher.Thus,theperiodinwhichopportunistictimingoccurredwasitselfopportunisticallytimed.Thispatternisconsistentwiththeviewthatopportunistictimingreflectsaneconomicdecisionthatissensitivetopayoffsratherthanapracticehabituallyfollowedbysomefirms.

Althoughfirmsdidnotcommonlyengageinopportunistictimingonallpos-sibleoccasions,wedofindevidenceofsignificantpersistence.TheoddsofaCEO’sgrantbeingluckyweresignificantlyhigher,controllingforCEOandfirmcharacteristicsinourdataset,whenaprecedinggranttotheCEOwasluckyaswell.Theseresultsindicatethat,beyondthecharacteristicsweiden-tifyasassociatedwithopportunistictiming,thereareadditionalunobservabletraitsoffirmsandCEOsthatledtoahighertendencyforopportunistictiming.Theremainderofthispaperisorganizedasfollows.SectionIdescribesourdataandprovidessummarystatistics.SectionIIanalyzestherelationbetweenCEOluckandtheluckofindependentdirectors.SectionIIIstudiestherela-tionbetweengainsfromluckygrantsandreportedcompensationfromothersources.SectionIVinvestigatestherelationbetweenoptiontimingandgov-ernancearrangements,thelevelofCEOinfluence,andpayoffsfromgettingaluckygrant.SectionVconcludes.

I.DataandSummaryStatistics

A.DataSources

WeconstructadatasetofoptiongrantsawardedtoCEOsandindependentdirectorsbetween1996and2005usingThomsonFinancial’sInsiderTrading

thecompositionofthecompensationcommittee,andtheydonotstudytheassociationbetweengovernanceandthetimingofdirectorgrants.

LuckyCEOsandLuckyDirectors2367

database,whichincludesallinsiders’filingsofequitytransactionsinForms3,4,5,and144.Inthecourseofconstructingthisdataset,weuseproceduressimilartothoseusedbyHeronandLie(2007,2009)andNarayananandSeyhun(2008).OurdatasetincludesobservationswithacleanseindicatorofR(“dataverifiedthroughthecleansingprocess”),H(“cleansedwithaveryhighlevelofconfidence”),orC(“arecordaddedtononderivativetableorderivativetableinordertocorrespondwitharecordontheopposingtable”).Wecombineanyduplicategrantsthatareawardedonagivendatetoagivenindividualinagivencompany.ThepricedatacomefromtheCRSPdatabase,andwerequirestockreturnstobeavailablefortheentiremonthofthegrantdate.

Fromthissampleweeliminategrantsthatarescheduled,astheymightbelesslikelytohavebeenopportunisticallytimed.Agrantisdefinedasscheduledifanadditionalgrantwasawardedonthesamedateplus/minus1dayintheprecedingyear(HeronandLie(2007)).Wefurthereliminategrantsthatweregiveninmonthswherethefirm’sstockwentex-dividend;totheextentthatfirmsschedulegrantsafteranex-dividenddate,thegrantpricemightfallbelowthestockpricesprecedingtheex-dateevenintheabsenceofanybackdatingorspring-loading.Inaddition,becausesomefirmsschedulegrantstodirectorsonthedateoftheannualshareholdermeeting,weuseannualmeetingdatesfromtheInvestorResponsibilityResearchCenter(IRRC)database(availableforabout25%ofthesamplefirms)andeliminatealldirectorgrantsfallingwithin+/−1dayoftheannualmeeting.

Finally,wecheckwhetherthestrikepriceofthegrantis“closeenough”totheclosingpriceofthegrantdate,ortotheclosingprice1daybeforeorafterthegrant,wherea“closeenough”priceisdefinedasapricethatiswithin1%ofthestrikeprice.Thedatewiththenearestclosingpricetothestrikepriceisthendefinedastheeffectivegrantdate.3

Inouranalysis,wefocusontwoimportantgroupsofindividualsthatreceiveoptiongrants.ThefirstgroupincludestheCEO.FollowingHeronandLie(2007),wedefineanindividualasaCEOifheorsheisidentifiedintheThom-sondatabaseeitherasaCEOorthepresidentofthecompany(rolecodeCEOorP).Thesecondgroupcomprisesindependentdirectors,whoplayakeyoversightroleincurrentcorporategovernance.Anindividualisconsideredanindepen-dentdirectorifheorsheisdefinedasadirectorintheThomsondatabase(rolecodeD)andisnotdefinedashavinganyotherroleinthecompany.4

Wedefineourunitofobservationasagrantevent.IntheCEOsample,agranteventisdefinedasadayinwhichtheCEOreceivesoneormoreoptiongrants.Inthedirectorsample,agranteventisdefinedasadayinwhichoneormoredirectorsreceiveoneormoreoptiongrants.

withHeronandLie(2007),wealsofindalargefractionofgrantswhosestrike

pricedoesnotcoincidewiththegrant-dateprice.HeronandLiediscussthepossiblereasonsfordeviationofthegrant-datepricefromthestrikeprice.

4ForthesetofcompaniesforwhichwealsohaveinformationfromIRRC,thedatasetofdirectorsthatweconstructispracticallyidenticaltothesetthatwouldobtainusingtheindependentclassificationprovidedbytheIRRCdatabase.Inparticular,morethan99%ofthedirectorsweclassifyasindependentdirectorsusingtheabovealgorithmarealsoclassifiedasindependentbytheIRRC.

3Consistent

2368R

TheJournalofFinance󰀆

OurCEOgranteventssamplecontains19,036CEOgranteventsin5,819firms.InourCEOsample,4,510CEOshaveonegrantevent,1,874havetwo,1,050havethreegrantevents,and1,386havefourormoregrantevents.Ourdirectorgranteventssamplecontains25,888granteventsin6,441differentfirms.Theaveragenumberofdirectorsgettingagrantinagranteventis3.07.B.LuckyGrantEvents

Theliteratureontheopportunistictimingofoptiongrants(startingwithYer-mack(1997))andthemorerecentliteratureonbackdating(Lie(2005),HeronandLie(2007,2009),andNarayananandSeyhun(2006,2008))hasfocusedontheexistenceofnegativepregrantstockreturnsand/orpositivepostgrantstockreturnsasthebasisfordetectingandinvestigatingabnormalgrantdatepatterns.5Incontrast,ourstrategyistoexamineat-the-moneygrantsondaysinwhichthestockpriceswereatthebottomofthepricedistribution.6

Inparticular,inthispaperweusethegrantmonthastheexamined“look-back”period-–thatis,wefocusongrantsthatweregivenatthelowestpriceofthemonth.Wecallthesegrants“lucky”grants.7Wecomparetheincidenceofluckygrantsrelativethebenchmarkcaseinwhichthegrantdateischosenwithoutregardtothepricedistribution.8

WhileourchoiceofperiodallowsustofocusonmanipulationinstancesthatwerelikelyofgreatesteconomicsignificanceforCEOsandshareholders,ouranalysisdoesnotfullycaptureinstancesofopportunistictimingbasedonshortlook-backperiods.NarayananandSeyhun(2008)demonstratethat,especiallyduringthepost-SOXperiod,therehavelikelybeennumerousinstancesinwhichgrantsweremisdatedbyafewdays,oftenbyjust1or2days.Thus,our

tendencyofgrantdatestoranklowratherthanhighinthedistributionofpricesisnoted

byHeronandLie(2007)andNarayananandSeyhun(2008),butthesestudiesincludepre-andpostgrantreturns,notpriceranks,asthemaintoolofanalysis.

6Bebchuk,Grinstein,andPeyer(2006a)discussindetailthepotentialadvantagesanddisad-vantagesoffocusingonpricerankscomparedwithpre-andpostgrantreturns.

7Bebchuk,Grinstein,andPeyer(2006a,2006b)extendtheanalysisinthispaperbyinvestigatingtheentirerankdistributionofgrantsduringthemonth,thecalendarquarter,andthecalendaryear.

8Althoughwerefertothebenchmarkasoneof“randomselection”ofgranteventdates,thisisnotmeanttoinvolveastrictlyrandomassignmentbutratheroneinwhichgrantdatesareselectedonthebasisoffactorsthatareindependentofprice-rankconsiderations.Itmightbesuggestedthat,totheextentthat(i)therearesomeconsiderationsthatleadfirmstoconcentrateawardsearlyinthemonthand(ii)stockpricestrendupwardovertime,thebenchmarkprobabilityshouldbeadjustedupward.However,theactualdistributionofoptiongranteventdatesissymmetricaroundthemiddleofthemonth.Granteventsoccurredonaverageonday15.94ofthemonthintheCEOsample,andonday15.96ofthemonthinthedirectorsample,androughlythesamenumberofgranteventsoccurredbeforeandafterthemediantradingdayofthemonthinbothsamples.Theupwardtrendofstockpricesis,onaverage,slowrelativetothevolatility,sopricesatthebeginningandendofamonthdonotsignificantlydifferintheiroddsofbeingthelowestpriceofthemonth.Tobecautious,however,weconductrobustnesschecksthatverifythatourresultsholdwhencontrollingforthelocationofthegrantdatewithinthecalendarmonth(seefootnote21forfurtherdetails).

5The

LuckyCEOsandLuckyDirectors

14%12%10%8%6%4%2%0%Lowest (lucky)Highest2369

CEOsDirectorsFigure1.FrequencyofGrantsatLowestandHighestStockPriceoftheMonth.

analysisinvestigatesanimportantsubsetofopportunistictimingpractices,butnotallofthem.

Thefigureaboveshowsthepercentageofgranteventsinoursamplethatwereatthelowestpriceofthemonthandatthehighestpriceofthemonth.About12%oftheCEOgranteventswerereportedtobegivenatthelowestpriceofthemonth,whereasonly4%ofthegranteventswerereportedtobegivenatthehighestpriceofthemonth.Similarly,about9.5%ofdirectorgranteventsweregivenatthelowestpriceofthemonthcomparedto6%atthehighestpriceofthemonth.9Thus,thereisaclearasymmetrybetweentheincidenceofgranteventdatesatthelowestandhighestpricesofthemonth.10

Togetanestimateofthenumberofgranteventsthatwereopportunisticallytimedinbothsamples,wecomparetheactualnumberofluckygranteventsto

that,onaverage,oursamplefirmshavelessthan20tradingdayseachmonthwherethe

pricesaredifferentinalldays.Thenumberofdaysduringthemonthinwhichthereisactualtradingislessthan20inmanycases,andinmanycasesthehighest(orlowest)priceofthemonthobtainson2ormoretradingdays.Thus,underrandomassignment,theexpectedfractionofgrantsonthehighestpricedayofthemonthisnotsimply5%but,onaverage,higher.

10Thisasymmetryexistsnotonlybetweentheincidenceofgranteventsatthelowestandthehighestpriceofthemonthbutalsobetweenotherpricelevelsatthebottomandtopofthepricedistribution.Overall,weobserveaclearmonotonicrelationbetweentherankofthepriceinamonthandthepercentageofgrantsgivenatthatrank.Toillustrate,intheCEOgranteventssample,thefrequencyofgranteventsisthehighestatthelowestpriceofthemonth(12%),secondhighestatthesecondlowestpriceofthemonth(9%),andthirdhighestatthethirdlowestpricelevel(8%).Conversely,thefrequencyofgranteventsislowestatthehighestpricelevel(4%),secondlowestatthesecondhighestlevel(5%),andsoforth.SeeBebchuketal.(2006a,2006b)forafullsetofdescriptivestatistics.Thesepapersalsoofferaregressionanalysisidentifyingtheextenttowhichdayswiththelowestpriceofthemontharemorelikelybechosenasgrantdates.

9Note

2370R

TheJournalofFinance󰀆

theexpectednumberofgranteventsthatwouldhavebeenluckyifgranteventswereallocatedrandomlyduringthemonth.Foreverymonth,wecalculatetheexpectednumberofgranteventsthatwouldhavebeenluckyifgranteventswererandomlyassignedoverthetradingdaysduringthemonth.11Thisestimationisdonebycalculatingforeachindividualgrantevent,assumingrandomassignment,theprobabilityofbeinggrantedatthelowestpriceofthemonth,andthenaggregatingtheseprobabilitiesacrossallgrantevents.Duetothelargenumberofgranteventsinvolved,arandomassignmentishighlyunlikelytodeviatesignificantlyfromtheexpectednumberwecalculate.WeconductthisanalysisbothforthesampleofdirectorgranteventsandforthesampleofCEOgrantevents.

TableIshowsourestimationresults.Weestimatethatoverthefullsampleperiodof1996to2005,1,163luckyCEOgranteventsand804luckydirectorgrantevents(about50%and33%ofallluckygrantevents,respectively)wereduetoopportunistictiming.12

ThepercentageofCEOluckygranteventsthatwereluckyduetooppor-tunistictimingwasabout55%beforeSOXand35%afterward.Thepercentageofdirectorgranteventsthatwereluckyduetoopportunistictimingwasabout36%beforeSOXand25%afterSOX.Sincealargefractionofluckygranteventsowesitsstatustoopportunistictiming,luckygranteventscanprovideausefulbasisforstudyingthefactorslikelytobeassociatedwithsuchtiming.

TableIalsogivesusasenseofthemagnitudeofthediscountinexercisepricethatopportunistictimingcouldproduce.Forluckydirectorgrantevents(luckyCEOgrantevents),thegrantpricewas11%(12%)lower,onaverage,thanthemedianpriceofthemonth.

TableIIprovidesourestimatesofthenumberoffirmsassociatedwithoppor-tunistictimingofCEOanddirectorgrantevents.Ourestimationmethodologyincludescalculatingthedifferencebetweentheactualandexpectednumberofgranteventsgivenonthelowestpricedayofthemonth.13

11The

scenarioofrandomassignmentalsoassumesthat,afterthedayisrandomlyselected,the

distributionofpricesamongthemonth’sdifferentdaysisnotaffectedbytheactionsthatinsidersareexpectedtotake.Theprobabilityofadaybeingthelowestpricedayiscomputedusingtheratioofthenumberofdaysinthegrantmonththathavethelowestpricetothetotalnumberoftradingdaysinthatfirm’sstockduringthegrantmonth.

12OurestimateforopportunisticallytimedgranteventsatthelowestpriceofthemonthismoreconservativethanthefigureestimatedbyHeronandLie(2009)forthetotalnumberofopportunisticallytimedgrants.Aswediscussabove,wedonotattempttocapture“small-scale”backdatinginwhichgrantsweremisdatedwithinashortperiodoftime.Incontrast,theHeron-Liemethodology,whichisbasedonthecomparisonofpre-andpostgrantreturns,attemptstocapturesuchinstancesofmanipulationaswell.SeealsoBebchuketal.(2006a,2006b),whoestimatedeviationfromrandomassignmentusingrankdistributionoverlongerperiods.

13Asbefore,webasethecalculationoftheexpectednumberofluckygranteventsonthechancethataparticulargranteventoccursonthelowestpricedayofthemonth(i.e.,numberoflowestpricedayspermonthdividedbynumberoftradingdayspermonth).Forfirmsthathavetwograntevents,theexpectednumberisbasedonthechancethatatleastoneofthetwogranteventsisgivenonthelowestpricedayofthemonth;similarlogicappliesformoregrantseventsandfortheanalysisoffirmswithmultiplegrantevents.

LuckyCEOsandLuckyDirectors

TableI

2371

EstimatingtheIncidenceofOpportunisticTiming

ThetableshowstheactualnumberofgranteventsthatfellonthelowestdayofthemonthinthedirectorgrantsampleandtheCEOgrantsample.Italsoshowsthenumberofgranteventsexpectedtofallonthelowestdayofthemonthifthegrantdatewasrandomlyselected.Weestimatetheprobabilityofobservingagranteventonthelowestpriceofthemonthbycountingthenumberofdaysinthemonthwherethepriceisthelowestanddividingitbythetotalnumberoftradingdaysofthestockinthatmonth.Thetablecomparestheestimatetotheactualnumberofgranteventsthatwerethelowest.Wealsoshowtheaverageratiooftheexercisepricetothemedianstockpriceinthemonth.GranteventsbeforetheSarbanes-OxleyAct(SOX)areoneswhosegranteventdatewasbeforeSeptember1,2002,andgrantsafterSOXareoneswhosegranteventdatewasonorafterSeptember1,2002.Thesampleconsistsof26,209optiongranteventstodirectorsand19,036granteventstoCEOsbetween1996and2005.

IncidencesofLuckyGrantEvents

CEOGrantEvents

Overall

ActualnumberofgranteventsActual/total

ExpectednumberofgranteventsActual−expected

(Actual−expected)/actual(Actual−expected)/total

Exerciseprice/medianstockpriceObservationsBeforeSOX

ActualnumberofgranteventsActual/total

ExpectednumberofgranteventsActual−expected

(Actual−expected)/actual(Actual−expected)/total

Exerciseprice/medianstockpriceObservationsAfterSOX

ActualnumberofgranteventsActual/total

ExpectednumberofgranteventsActual−expected

(Actual−expected)/actual(Actual−expected)/total

Exerciseprice/medianstockpriceObservations

2,32912.2%1,1661,16349.9%6.1%0.8819,0361,74114.5%78595654.9%8.0%0.8711,9985888.4%38120735.3%2.9%0.917,038

DirectorGrantEvents

2,4739.4%1,66980432.5%2.8%0.26,2091,70710.9%1,09860935.7%3.5%0.8815,7097667.3%57119525.4%1.7%0.9110,500

PanelAindicatesthatthenumberoffirmswithoneormoreluckyCEOgranteventexceedstheestimatednumberofsuchfirmsunderrandomassignmentby722.Thisfigureimpliesthatabout12%ofallfirmsinoursamplehadluckyCEOgranteventsduetoopportunistictiming.

PanelBindicatesthatthenumberoffirmswithoneormoreluckydirec-torgranteventexceedstheestimatednumberofsuchfirmsunderrandom

2372R

TheJournalofFinance󰀆

TableII

EstimatingtheIncidenceofFirmsAssociatedwithOpportunistic

Timing

ThetableshowstheactualnumberoffirmsthathadatleastonegranteventthatfellonthelowestdayofthemonthinthedirectorgrantsampleandtheCEOgrantsample(luckygrant).Italsoshowsthenumberoffirmsexpectedtohaveatleastoneluckygranteventfallingonthelowestdayofthemonthifthegrantdatewasrandomlyselected.Weestimatetheprobabilityofobservingagranteventonthelowestpriceofthemonthbycountingthenumberofdaysinthemonthwherethepriceisthelowestanddividingitbythetotalnumberoftradingdaysofthestockinthatmonth.Thetablecomparestheestimatetotheactualnumberofgranteventsthatwerethelowest.Thesampleconsistsof26,209optiongranteventstodirectorsand19,036granteventstoCEOsbetween1996and2005afterexcludingeventsthatweregivenontheannualmeetingdate(+/−1day).

ActualNo.ofFirmsatLowest

ExpectedNo.ofFirmsatLowest

(Actual−Expected)/Actual

(Actual−Expected)/Total

No.ofGrants

Firms

Actual−Expected

PanelA:CEOGrantEvents:DistributionbyFirm

12345>All

1,8801,1068605691,4045,819

2962542622127291,753

1381491521254671,031

15810511087262722

53.4%41.3%42.0%41.0%35.9%41.2%

8.4%9.5%12.8%15.3%18.7%12.4%

PanelB:DirectorGrantEvents:DistributionbyFirm

12345>All

1,5971,0247615502,1196,051

1952042151881,0091,811

1331391481348421,395

62656754167416

32.0%31.9%31.3%28.6%16.6%22.9%

3.9%6.3%8.8%9.8%7.9%6.9%

assignmentby416.Thisfigureimpliesthatluckydirectorgranteventsthatwereduetoopportunistictimingtookplaceinabout7%ofallfirmsinoursample.14

Foreachofthe12Fama–Frenchindustries,weproduceanestimateoftheincidenceofluckygrantsduetoopportunistictiming.Thisanalysisshows15thattherewasasignificantincidenceoftimingofCEOgranteventsineachofthe12Fama–Frenchindustries,andthattherewasasignificantincidenceoftimingofdirectorgranteventsineachofthe12industriesotherthanutilities.

opportunistictimingthatweestimateabovemightbeduenotonlytobackdatingbut

alsotospring-loading,basedontheuseofprivateinformation.Bebchuketal.(2006a,2006b)show,however,thatatleastsomeoftheidentifiedinstancesofopportunistictimingareduetobackdatingratherthanspring-loading.SimilarfindingshavebeenobtainedbyHeronandLie(2007)andNarayananandSeyhun(2006).

15Forthefulldetailsofthisindustry-levelanalysis,seeBebchuketal.(2006a,2006b).

14The

LuckyCEOsandLuckyDirectors2373

Finally,totheextentthatboardmeetingsinwhichoptiongrantsarede-terminedarealsotheonesthatprecedeearningsannouncements,ahigher-than-expectedincidenceofluckygrantsmightbeduetoapositivestockreturndriftfollowingpositiveearningssurprises.Wethereforeexaminewhetherourresultsaredrivenbygrantsthatweregiveninmonthsinwhichquarterlyearn-ingswereannounced.WereestimatethenumbersinTableIexcludingallgranteventstakingplaceinmonthswithaquarterlyearningsannouncement.WefindthatexcludingtheseeventsdoesnotchangesignificantlyanyofourresultsinTableI.16Werepeatthisrobustnesscheckthroughout,andalltheresultsdisplayedinthispaperarerobusttotheremovalofgranteventsawardedduringmonthswithquarterlyearningannouncements(seetheInternetAppendix17).

II.CEOLuckandtheLuckofIndependentDirectors

InexaminingtheopportunistictimingofCEOs’grantevents,onenaturalquestiontoaskiswhatrole,ifany,independentdirectorsplayed.Totheextentthatopportunistictimingiscausedbyagencyproblems,onemaywonderwhydirectorsfailedtopreventit.Onepossibleexplanationisthatdirectorsdidnotknowaboutopportunistictiming.Anotherpossibleexplanationisthatdirectorshadincentivestoallowsuchpracticestocontinue,oratleastnottolearnaboutthem.InhisopeningstatementattheSenateFinanceCommitteehearingonbackdating,then-ChairmanGrassleyexpressedconcernsthat“boardsofdirectorswereeitherasleepattheswitch,orinsomecases,willingaccomplicesthemselves.”18

Inthissection,weexplorethequestionofdirectors’incentives.Weshowthatgrantdateswereselectedinwaysthatmadeindependentdirectorsbeneficia-riesofopportunistictimingpractices.Wealsoshowthatthetimingofdirectorgrantswasnotmerelyabyproductofdirectorshappeningtoreceivegrantsonthesamedateasexecutiveawardsoroffirmsroutinelytiminggrantstoallrecipients.Rather,corporatedecision-makerschosetoalsoprovideluckygrantstoindependentdirectorsandnotonlyexecutives.Finally,ouranalysisinthisandsubsequentsectionsshowsthatdirectors’luckwasassociatedwithimprovedoutcomesfortheCEO,bothintermsofincreasedoddsofreceivingluckygrantsandintermsofreceivinghighercompensationfromothersources.

granteventsinmonthswherethefirmmakesaquarterlyearningsannouncement

reducesthenumberofCEOgrantsfrom19,036to11,997,andthenumberofluckygranteventsfrom2,329to1,535.However,thefractionofluckygrantevents(andthefractionofunexpectedluckygrants)doesnotchangesignificantlyandgoesfrom12.2%to12.8%(6.1%to6.5%).Similarly,fordirectorgrantevents,thetotalnumberofgranteventschangesfrom26,209to16,634,andthenumberofluckygrantsfrom2,473to1,628.Thefractionofluckygrantevents(andthefractionofunexpectedluckygrants)changesfrom9.4%to9.8%(2.8%to3.4%).

17TheInternetAppendixisavailableathttp://www.afajof.org/supplements.asp.

18OpeningStatementofSen.ChuckGrassley,Chairman,FinanceCommitteeHearing,“Ex-ecutiveCompensation:BackdatingtotheFuture”,September6,2006,availableathttp://www.senate.gov/∼finance/.

16Excluding

2374R

TheJournalofFinance󰀆

A.CEOLuckandSimultaneousAwardstoOutsideDirectors

Thefactthatsomedirectorgranteventswereopportunisticallytimeddoesnotindicatethatthosemakingthetimingdecisionssoughttomakeindepen-dentdirectorsthebeneficiariesofopportunistictiming.Somedirectorgranteventscouldhavecoincided,forvariousreasons,withgrantsawardedtoexec-utives.Infact,inoursample,29%ofalldirectorgranteventscoincidedwithawardstooneormoreexecutives.19Giventhatsomedirectorgranteventsco-incidedwithexecutivegrantevents,itmightbepossiblethatdecision-makersintendingtoopportunisticallytimetheoptionsawardedtoexecutiveswouldhaveproducedopportunistictimingofdirectorgrants,evenwithouthavinganydesiretoproducethelatteroutcome.

Iftheopportunistictimingofdirectorgrantswasamerebyproductofthetimingofexecutivegrants,thelikelihoodofaCEOgranteventbeingluckywouldnotbeexpectedtobecorrelatedwithwhetherindependentdirectorsreceivedanoptiongrantonthesameday.Totestthishypothesis,werunthreeregressions—apooledregression,aregressionusingfirmfixedeffects,andaregressionusingCEOfixedeffects—onoursampleofCEOgranteventsanddisplaytheresultsinTableIV.Inallregressions,thedependentvariableisadummyvariableindicatingwhetheraCEOgranteventwaslucky.Theindependentvariableofinterestisadummyvariableindicatingwhetheradirector’sgranteventcoincidedwiththeCEOgrantevent.

Weincludeseveralcontrolvariablesintheregressions(seeTableIIIfordefinitions).Thefirstisfirmsize,aspriorworkshowsthatopportunistictimingwasmorelikelytooccurinsmallerfirms(HeronandLie(2009)).Ourvariableforsizeisthenaturallogofrelativemarketcapitalization,definedastheratioofthemarketcapitalizationofthefirmatthegrantdatedividedbythemedianmarketcapitalizationofallthefirmsinoursampleforthatyear.20SOXimposedstricterreportingrequirements,makingbackdatingmoredifficult(NarayananandSeyhun(2006)andHeronandLie(2007)),sooursecondcontrolvariableisadummyvariableequaltooneifthegrantwasgivenpost-SOX.Inthepooledregression,wealsoincludeaneweconomydummytocontrolforthepossibilitythatopportunistictimingwasmoreprevalentamongneweconomy(hi-tech)firms(HeronandLie(2009)).OurdefinitionoftheneweconomyfollowsMurphy(2003),whodefinedneweconomyfirmsasfirmsthatbelongtothefollowingfour-digitSICcodes:3570,3571,3572,3576,3577,3661,3674,4812,4813,5045,5961,7370,7371,7372,and7373.

Finally,evenunderrandomselectionofdates,agranteventwouldbemorelikelytobeluckywhenmoretradingdatesinthemonthhadapriceequaltothelowestpricelevelofthemonth.Also,whentherewasonly1daywiththis

executives’granteventsareidentifiedfromThompsonusingthesamesample

selectionmechanismasforCEOsbutrequiringanexecutivetitleastherolecode.

20Wechoosethisbenchmark,ratherthanabenchmarkbasedonthedistributionofallfirmsinCompustat,toensureanevensizedistributionacrossfirmsinthesample.OurresultsremainessentiallythesamewhenwedividemarketcapitalizationbythemedianmarketcapitalizationofthefirmsinCompustatinthatyear.

19Non-CEO

LuckyCEOsandLuckyDirectors

TableIII

2375

VariableDefinitions

VariableLuckygrant

Simultaneousgranttodirectors

LuckyCEOcurrentorprioryear

LuckyCEOcurrentyear

Luckydirectorcurrentyear

Relativegainfromluck(CEO)

Definition

DummyvariableequaltooneifthegrantwasgivenonthedatewherethelowestpriceofthemonthprevailedandzerootherwiseDummyvariableequaltooneifadirectorgrantoccurredonthesamedayasagranttotheCEOandzerootherwise

DummyvariableequaltooneiftheCEOreceivedaluckygrantinthecurrentorprioryearandzerootherwise

DummyvariableequaltooneifatleastonegranteventtotheCEOduringthefiscalyearwasonthelowestdayofthemonthDummyvariableequaltooneifatleastonegranteventtothe

directorsduringthefiscalyearwasonthelowestdayofthemonthRatiooftheimpliedunderreportedoptionvaluetototal

compensation,wheretheimpliedunderreportingistheratioofthebenchmarktograntvalueminusone,timestheBlack-ScholesvalueoftheoptionsreportedbyExecuComp.ThebenchmarkvalueiscomputedastheBlack-Scholesvalueofanoptionwiththestrikepriceofthegrant,butwherethegrantdatepriceisthemedianpriceofthemonth.Allotherparametersoftheoptiongrantareheldconstant.Thestandarddeviationofthedailystockreturnsiscalculatedovertheyearpriortothefiscalyearinwhichthegrantwasgiven.ForCEOswithmultipleluckygrantsperfiscalyeartheimpliedunderreportingistheaveragerelativegainfromluckacrossallgrants.

Dummyvariableequaltooneifthepreviousgranteventwasatthelowestpriceofthemonthandzerootherwise.Ifthereisnopreviousgrantevent,thenthedummyvariableisequaltozeroDummyvariableequaltooneifthepreviousgranteventwasatanypriceotherthanthelowestpriceofthemonthandzerootherwise.Ifthereisnopreviousgrantevent,thenthedummyvariableisequaltozero

Naturallogofthebookvalueoftheassetsattheendofthefiscalyear

Operatingincomedividedbybookvalueofassets

Bookvalueofliabilitiesplusthemarketvalueofequityalldividedbythebookvalueofassets,wheretheindustryadjustmentismadeatthetwo-digitSIClevel

Ratioofthebookvalueoflong-termdebtdividedbythebookvalueofassets

Cumulativestockreturnintheyearofthegrant(t)Cumulativestockreturnintheyearofthegrant(t−1)

Naturallogoftheratiobetweenthemarketcapofthefirmattheendoftheyearandthemedianmarketcapofthefirmsinthesampleforthatyear

DummyvariableequaltooneforfirmswithSICcodesasdefinedinMurphy(2003)

DummyvariableequaltooneforgrantsafterSeptember1,2002andzerootherwise

Fractionoftradingdayspermonthwithlowestprice

(continued)

PreviousgranteventluckyPreviousgranteventnotlucky

LogbookvalueReturnonassets(ROA)

Industry-adjustedTobin’sQLeverageStockreturntStockreturnt-1Relativesize

NeweconomySOX

Daysinmonthlowest

2376R

TheJournalofFinance󰀆

TableIII—Continued

Variable

ln(totalcompensation)Differencebetweenthemedianandlowestprice

Marketcomponentofthemedian

price—lowestpricedifferenceFirm-specific

componentofthemedian

price—lowestpricedifference

CEOoutsiderdummyIndependentcompensation

committeedummyIndependentboarddummyTenure

5%blockholderoncompensation

committeedummy

Definition

Naturallogofthevariabletdc1fromExecuComp

Naturallogofoneplusthereturntoshareholdersfromthelowestpriceofthemonthinwhichtheoptionsweregrantedtothemedianpriceofthatmonth

Naturallogofoneplusmarketreturnfromtheminimumpricedaytothemedianpriceday

Totalreturnminusthemarketreturnfromtheminimumpricedaytothemedianpriceday

DummyvariableequaltooneiftheCEOwasnotemployedinthefirmbeforebecomingtheCEO,andzerootherwise

DummyvariableequaltooneifthecompensationcommitteeconsistedonlyofindependentdirectorsandzerootherwiseDummyvariableequaltooneiftheboardhadamajorityofindependentdirectorsandzerootherwise

NaturallogofoneplusthenumberofyearsthattheCEOservedinthecompany

Dummyvariableequaltooneiftherewasatleastonedirectorwhoheld5%ormoreofthesharesinthecompanyandzerootherwise

pricelevel,theprobabilitythatitwouldbeselectedislowerwhenthemonthhadmoretradingdays.Therefore,weaddafourthcontrolvariableequaltotheratioofthenumberofdaysinthemonthofthegranteventwithclosingpricesequaltothelowestpriceofthemonthtothenumberoftradingdaysinthefirm’sstockinthegrantmonth.21

TableIVpresentstheresults.TheresultsindicatethatthelikelihoodthataCEOgranteventwasluckywashigher(significantatthe5%levelorbetter)whentheCEOgranteventwasonthesamedayasthedirector’sgrantevent.Forexample,intheCEOfixedeffectsregression,theoddsofaCEOgrantbeingluckyincreasedbyexp(0.301)=1.35(or35%)whentheCEOgranteventcoincidedwithagranteventtoindependentdirectors.

theInternetAppendix,wealsoadd,hereandinallsubsequentregressionswiththedepen-dentvariablebeingwhetheragranteventwaslucky,additionalcontrolstoverifythatourresultsarerobusttothepossibilityofanupwarddriftinstockpricesduringthegrantmonth(seefootnote8).Inonetypeofregression,weadddummiesforthelocationofthegrantdayinthesequenceofthecalendardaysofthemonth(firstdayofthemonth,seconddayofthemonth,etc.).Inasecondtypeofregression,weadddummiesforthelocationofthedayinthesequenceoftradingdays(firsttradingdayofthemonth,secondtradingdayofthemonth,etc.).Addingthesecontrolsdoesnotchangeanyoftheresultsreportedinthispaper.

21In

LuckyCEOsandLuckyDirectors

TableIV

2377

CEOLuckandSimultaneousAwardstoDirectors

ThetableshowslogitregressionresultsforthesampleofCEOgrantevents.Duetodataavail-ability,thesampleisreducedto18,543observations.Thedependentvariableisadummyvariableforwhetherthegrantwasgivenonthedatewherethelowestpriceofthemonthprevailedandzerootherwise.TheindependentvariablesaredefinedinTableIII.Thenumbersinparenthesesaretheestimatedstandarderrorsofthecoefficients,adjustedforclusteringattheexecutivelevel.∗∗and∗∗∗indicatesignificanceatthe5%and1%level,respectively.

WholeSample

FirmFixedEffect

CEOFixedEffect

LuckyCEOGrantDummy

DependentVariable

SimultaneousgrantstodirectorsRelativesizeNeweconomySOX

DaysinmonthlowestConstantObservations

(1)0.188∗∗∗(0.061)−0.066∗∗∗(0.012)0.267∗∗∗(0.062)−0.563∗∗∗(0.052)5.484∗∗∗(0.404)−2.222∗∗∗(0.042)18,543

(2)0.231∗∗(0.091)0.156∗∗∗(0.046)−0.552∗∗∗(0.071)5.907∗∗∗(0.715)

(3)0.301∗∗∗(0.104)0.165∗∗∗(0.053)−0.572∗∗∗(0.084)6.446∗∗∗(0.885)

18,54318,543

Onepossibleexplanationforthisfindingisthat,whenselectingwhichCEOgrantstoopportunisticallytime,decision-makerspreferredtotimeCEOgrantsthatcoincidedwithawardstoindependentdirectors.Analternativeexplana-tionisthat,whenselectingwhentoprovidegrantssimultaneouslytotheCEOandindependentdirectors,decision-makerschosetodosowhengrantswereopportunisticallytimed.Undereitherexplanation,choicesweremadewithanaimtobenefitindependentdirectorsinconnectionwiththeprovisionofbene-fitstotheCEO.Thus,ourresultsdonotsupportthehypothesisthatdecision-makerssoughttoopportunisticallytimeCEOgrantswhilehavingnointerestin(andbeingindifferenttotheprospectof)providingindependentdirectorswithluckygrants.

B.DirectorLuckandSimultaneousAwardstoCEOs

HavingexaminedtherelationshipbetweenthelikelihoodofaCEOgranteventbeingluckyandtheexistenceofasimultaneousawardtodirectors,wealsoexaminehowthelikelihoodofaluckydirectorgranteventdependedontheexistenceofasimultaneousawardtoexecutives.Werunaregressionofwhetheradirectorgranteventwasluckyon(i)whethertheCEObutnotnon-CEOexecutivesreceivedagrantonthatdate;(ii)whethertheCEOandoneor

2378R

TheJournalofFinance󰀆

TableV

DirectorLuckandSimultaneousAwardtotheCEO

Thetableshowslogitregressionresultsforthesampleofdirectorgrantevents.Duetodataavailability,thesampleisreducedto25,888grantevents.Thedependentvariableisadummyequaltooneifagranteventisatthelowestpriceofthemonth.“CEObutnototherexecsgetgrant”isadummyequaltooneiftheCEObutnootherexecutivealsoreceivedagrantonthesameday.“CEOandotherexecsgetgrant,”and“OtherexecsbutnotCEOgetgrant”aredefinedaccordingly.∗and∗∗∗indicatesignificanceatthe10%and1%levels,respectively.Standarderrorsareshowninparentheses.Theregressionincludesalsofirmfixedeffects.DependentVariable

CEObutnototherexecsgetgrantCEOandotherexecsgetgrantOtherexecsbutnotCEOgetgrantRelativesizeNeweconomySOX

DaysinmonthlowestObservations

LuckyDirectorGrantDummy

0.660∗∗∗(0.215)0.650∗∗∗(0.084)0.348∗∗∗(0.079)0.067∗(0.039)0.446(0.302)−0.313∗∗∗(0.0)7.561∗∗∗(0.547)25,888

morenon-CEOexecutivesreceivedagrantonthatdate;and(iii)whetheroneormorenon-CEOexecutives,butnottheCEO,receivedagrantonthatdate.Theregressionincludesourstandardcontrolsandfirmfixedeffects,andwepresenttheresultsinTableV.

Thecoefficientsonallthreedummyvariablesarepositiveandsignificantatthe1%level,indicatingthattheoddsthatadirectorgranteventwasluckyincreasedwhenanexecutivealsoreceivedagrantonthesameday.At-testshowsthatdirectorswhoreceivedgrantsonthesamedayastheCEO(whetherornotanon-CEOexecutivealsoreceivedoptions)improvedtheoddsofbeingluckybymorethanifdirectorsreceivedgrantstogetherwithoneormorenon-CEOexecutive.Thus,theresultsareagainconsistentwiththepossibilitythattheopportunistictimingofdirectorgrantswastheproductofchoicesthattiedtheinterestsofindependentdirectorswiththoseofexecutivesingeneralandtheCEOinparticular.

C.DirectorGrantEventsNotCoincidingwithExecutiveAwards

Ourfindingsthusfarraisethequestionofwhetherthehigher-than-normalincidenceofdirectorgranteventsthatwereluckywasfullydrivenbyeventscoincidingwithexecutiveawards.Toexplorethisquestion,werunaregressionsimilartothosedisplayedinTableVbutrestrictthesampletograntevents

LuckyCEOsandLuckyDirectors

TableVI

2379

DirectorLucknotSimultaneouswithExecutives

Thesampleconsistsof18,376granteventswhereoutsidedirectorsreceivedoptiongrantsandnoexecutivereceivedagrantonthesameday.Theregressionsarelogitregressionswherethestandarderrorsareclusteredbycompanyandreportedunderneaththecoefficients.Thedependentvariableisadummyequaltooneonthedayinthecalendarmonthofthegranteventandzerootherwise.Theindependentvariablesaredummyvariablesequaltooneifthepriceonagivendayisthelowest(secondlowest,thirdlowestprice)ofthemonth.Allregressionsincludeadditionalpriceranks(seetheInternetAppendixforthefullspecification).Regression1includesallgrantevents,regression2onlythosepre-SOX,andregression3onlythosepost-SOX.∗∗and∗∗∗indicatesignificanceatthe5%and1%levels,respectively.

GrantDateDummy

DependentVariableLuckySecondlowestThirdlowestConstantObservations

(1)All0.450∗∗∗(0.027)0.201∗∗∗(0.027)0.122∗∗∗(0.026)−3.034∗∗∗(0.004)367,620

(2)Pre-SOX0.513∗∗∗(0.033)0.221∗∗∗(0.033)0.130∗∗∗(0.032)−3.045∗∗∗(0.006)212,654

(3)Post-SOX0.329∗∗∗(0.044)0.168∗∗∗(0.047)0.115∗∗(0.048)−3.019∗∗∗(0.006)154,966

notcoincidingwithawardstoexecutives.Specifically,weregresswhetheradaywasselectedforadirectorgranteventonwhetherthedayhadthelowestpriceofthemonth.

TheresultsdisplayedinTableVIindicatetheexistenceofopportunistictim-ingevenintheuniverseofdirectorgranteventsnotcoincidingwithexecutiveawards.Thisisthecasenotonlyforthe1996to2005periodasawhole,butalsoforthesubperiodsseparatedbytheadoptionofSOX.Asexpected,daysatthebottomofthepricedistributionweremorelikelytobechosenfordirectorgranteventsbeforeSOXthanafterSOX.Overall,TableVIshowsthattheincidenceofluckamongdirectorgranteventsnotcoincidingwithexecutiveawardswassignificantlyhigherthanrandom.

Focusingonthesetofdirectorgranteventsnotcoincidingwithexecutiveawards,weexamineinTableVIIwhethertheopportunistictimingidentifiedwithinthissetinvolvedanychoiceslinkingdirectorluckandCEOluck.Inparticular,weareinterestedinwhetheradirectorgranteventnotcoincidingwithexecutiveawardswasmorelikelytobeluckyiftheCEOhadaluckygranteventonadifferentoccasionduringthecurrentortheprioryear.

IntheregressionsofTableVII,PanelA,thedependentvariableisadummyvariableforwhetheragivendaywasselectedforadirectorgrantevent.ThesampleconsistsofalltradingdaysinmonthsthathaddirectorgranteventsnotcoincidingwithCEOgrantevents.Thevariableofinterestisaninteractiontermbetweenadummyindicatingwhethertheday’sstockpricewasatthe

2380R

TheJournalofFinance󰀆

TableVII

DirectorLuckandCEOLuck

Thesampleconsistsofalltradingdaysinmonthswhereoutsidedirectorshadatleastoneoptiongranteventandnootherexecutivereceivedagrantonthesameday(atotalof18,376grantevents).InPanelA,theregressionsarelogitregressionswheretheerrorsareclusteredbycompanyandreportedunderneaththecoefficients.Thedependentvariableisadummyequaltooneifthetradingdaywasagranteventdayandzerootherwise.TheindependentvariablesaredescribedinTableIII.Regression1includesallgrantevents,regression2onlythosepre-SOX,andregression3onlythosepost-SOX.Allregressionsincludeadditionalpriceranksfromthesecondlowesttothefifthlowestandthefivehighestpriceranks(seetheInternetAppendix).PanelBcontainsapooledpanelregressionwheretheerrorsareclusteredbyfirm(regression1)andafirmfixedeffectregression(regression2).Thedependentvariableisadummyequaltooneifthegranteventwasluckyandzerootherwise.∗∗∗indicatessignificanceatthe1%level.

PanelA:DependentVariableIsGrantDateDummy

SampleLuckyDirector

LuckyDirector×

LuckyCEOcurrentorprioryearLuckyCEOcurrentorprioryearConstantObservations

(1)All0.365∗∗∗(0.028)0.525∗∗∗(0.081)−0.057∗∗∗(0.010)−3.003∗∗∗(0.003)367,620

(2)Pre-SOX0.417∗∗∗(0.034)0.571∗∗∗(0.100)−0.071∗∗∗(0.014)−3.006∗∗∗(0.004)212,654

(3)Post-SOX0.262∗∗∗(0.047)0.438∗∗∗(0.137)−0.036∗∗∗(0.013)−2.998∗∗∗(0.004)154,966

PanelB:DependentVariableIsLuckyDirectorGrantDummy

Regression

LuckyCEOcurrentorprioryearRelativesizeNeweconomySOX

DaysinmonthlowestConstantObservations

(1)

Pooledpanel0.502∗∗∗(0.075)−0.042∗∗∗(0.014)0.013(0.068)−0.205∗∗∗(0.052)6.937∗∗∗(0.404)−1.610∗∗∗(0.235)18,376

(2)

Firmfixedeffects

0.500∗∗∗(0.111)0.025(0.043)−0.252∗∗∗(0.072)8.221∗∗∗(0.1)

18,376

lowestlevelofthemonthandadummyvariableequaltooneiftheCEOhadaluckygranteventintheyearofthegranteventortheprecedingyear.Werunthreeregressions—oneforthewholeperiod,oneforthepre-SOXperiod,andoneforthepost-SOXperiod.Inalloftheregressions,thecoefficientontheinteractiontermispositiveandsignificantatthe1%level.Theseresults

LuckyCEOsandLuckyDirectors2381

indicatethatthroughouttheperiod,aswellasduringthetwosubperiods,adirectorgranteventnotcoincidingwithexecutiveawardswasmorelikelytobeluckyiftheCEOreceivedaluckygrantonadifferentoccasionduringthecurrentortheprioryear.

InPanelBofTableVII,thesampleconsistsofalldirectorgranteventsnotcoincidingwithCEOgranteventsandthedependentvariableisadummyequaltooneifagivendirectorgranteventthatdidnotcoincidewithexecutiveawardswaslucky.TheindependentvariablesarewhethertheCEOreceivedaluckygrantduringthecurrentorprioryear,whichisourvariableofinterest,andstandardcontrols.Column1reportstheresultsofapooledregression,whilecolumn2reportstheresultsofafirmfixedeffectsregressionfocusingonwithin-firmvariationovertimebetweenperiodsinwhichtheCEOwasandwasnotlucky.Inbothcolumns,thecoefficientonthedummyindicatingwhethertheCEOhadaluckygranteventduringthecurrentorprioryearispositiveandsignificantatthe1%level.Economically,thecoefficientinregression1indicatesthattheoddsofadirectorgranteventbeingluckyincreasedby1.65=exp(0.502)or65%iftheCEOalsohadaluckygranteventinthecurrentorprioryear.

D.IsDirectorLuckDuetoRoutineTiming?

Finally,weexplorewhetherthehigher-than-normalincidenceofdirectorluckcouldhavebeenamerebyproductoffirms“routinely”timingallgrantstoeveryone.Wefindthatourresultsconcerningdirectorluckwerenotduetosuchroutinetiming.

Tobegin,weobservethattheincidenceoffirmsthatprovidedluckygrantstoallrecipientswasverylow.Firmsthatcommonlyawardedluckygrantsdidnotuniformlyprovidesuchgrantstoallrecipients,andalsoprovidedgrantsthatwerenotluckytosomerecipients.22Inparticular,amongthefirmswiththreeormoregranteventsduringtheperiodofourstudy,only1.2%hadgranteventsthatwerealwayslucky(80firms);amongthefirmswiththreeormoregranteventsduringthepre-SOXperiod,only1.7%providedluckygranteventstoeveryoneduringthisperiod(94firms).Onayear-by-yearbasis,amongfirm-yearobservationswiththreeormoregranteventsduringagivenyear,only4.9%hadgrantsthatwereallluckyinagivenyear(292firm-yearobservations).Furthermore,theincidenceofuniformluckisonly1%amongthefirmswithfourormoregranteventsduringtheperiodofourstudy,1.3%amongthefirmswithfourormoregranteventsduringthepre-SOXperiod,and3.7%amongthefirm-yearobservationsoffirmswithfourormoregranteventsduringagivenyear.

Moreimportantly,ifweremovealltheobservationsfromthedatasetofdi-rectorgranteventsthatcouldhavebeenluckyduetoaroutinetimingpractice,

startwiththetotalsampleofoptiongrants(248,084)toeveryonewherewehavethe

necessaryinformationonstockprices.Werestrictoursampletothosefirmsthatprovidedatleastonegranttoadirectorduringoursampleperiod.

22We

2382R

TheJournalofFinance󰀆

allofourresultsfromthefullsamplehold.Inparticular,weperformthreere-movalstrategies.First,weremovealltheluckydirectorgranteventsoffirmswithtwoormoregranteventsthatprovidedluckygrantstoeveryoneduringthesampleperiodoratleastduringthepre-SOXperiod.Next,toallowforthepossibilitythataroutineluckpracticewasintroducedjustbeforealuckydirec-torgrantevent,weremoveallluckydirectorgranteventsthatwerefollowedbygrantevent(s)thatwereallluckyduringthesampleperiodoratleastuntiltheadoptionofSOX.Third,toallowforthepossibilitythatafirmhadaroutineluckpracticejustinagivenyear,weremovealltheluckydirectorgranteventsoffirmsthathadtwoormoregranteventsinthatyearandwhosegranteventsinthatyearwerealllucky.Excludingtheseluckydirectorgranteventsfromourdata,wefindthatallourprecedingresultsconcerningdirectorluck,aswellasalltheresultspresentedlaterinthispaper,continuetohold.23

Overall,theevidencewepresentinthissectionisconsistentwiththehy-pothesisthatfirmschosetoprovideopportunisticallytimedoptiongrantstodirectors.Thesubsequentsectionsexaminewaysinwhichfirmsmakingsuchchoicesdifferedfromothers.Amongotherthings,weshallseethatCEOcom-pensation(relativetopeercompanies)washigher,andthelikelihoodofluckyCEOgrantsassociatedwithamajorityofindependentdirectorsontheboardwaslower,infirmsprovidingluckygrantstodirectors.

III.TotalReportedCEOCompensationandLuckyGrants

Inthissection,weinvestigatetherelationbetweentheCEO’stotalre-portedcompensationandluck.Inparticular,wetestthehypothesisthatfirmsusedgainsfromluckygrantstotheCEO,whichwerenotreflectedinpub-liclyreportedcompensationfigures,asa“substitute”forformsofCEOcom-pensationthathadtobepubliclyreported.Wedonotfindsupportforthishypothesisbut,rather,findapositivecorrelationbetweenCEOreportedcom-pensationandluckyCEOgrants.WealsoinvestigatetherelationbetweentotalreportedCEOcompensationanddirectorluckandfindevidencethatluckygrantstoindependentdirectorswereassociatedwithhighertotalCEOcompensation.

WebeginbyestimatingthemagnitudeofgainstoCEOsfromopportunistictimingofoptiongrants.Althoughwedonotknowforcertainwhichluckygrantswereproducedbybackdating,weidentifyapoolofgrants—thoseawardedatthelowestpriceofthemonth—inwhichalargefractionwaslikelyproducedbyopportunistictiming.Therefore,itisworthestimatingthepo-tentialgainthataCEOwouldhavederivedfromhavingagrantplacedinthispoolviaopportunistictiming,assumingthatthegrantwasindeedsoplaced.24

alsoconfirmedinasimilarwaythatallofourresultsconcerningCEOluckinthispaper

arerobusttoimplementingeachofthethreestrategiesforremovingluckyCEOgranteventsthatcouldbeduetoroutinetimingbythefirm.

24Forthefullanalysis,seeBebchuk,Grinstein,andPeyer(2006a).

23We

LuckyCEOsandLuckyDirectors2383

Tothisend,wefirstcalculatethevalueofeachgrantintheconsideredpool,assumingthatitwasgrantedonthedatereportedusingtheparametersgivenintheThomsondatabaseforthegrantdate,maturitydate,strikeprice,andnumberofoptionsgranted.25Assumingthegrantwasbackdated,wethencomputetheaverageratiosofthreebenchmarkestimatesofthegrant’sactualvalueundertruthfulreporting.Onecomparisonbenchmarkisthevaluetheoptionhad,assumingthatitwasgrantednotonthereporteddatebutonadateinthegrantmonthinwhichthepricewasequaltothemonth’smedianprice.Thesecondcomparisonbenchmarkistheexpectedvalueofthegrant,assumingitwasgrantednotonthereporteddatebutonarandomlyselecteddayduringthegrantmonth(thatis,assumingitwasgivenonanyofthesedayswiththesameprobability).26ThethirdcomparisonbenchmarkisthevaluethattheCEO’soptionhadattheendofthegrantmonth.27

Wefindthatthethreemethodsaboveyieldverysimilarresults.Assumingluckygrantsowedtheirstatustoopportunistictiming,theyhadavaluethatwas,onaverage,20%to21%higherthanthevalueofthegrantintheabsenceofsuchtiming.28Theluckygrantshadavaluethatwas20%higherthanthevalueofagrantwiththemedianpriceofthemonthortheexpectedvalueofagrantwhosedatewasrandomlyselectedamongthedaysofthemonth.Bytheendofthegrantmonth,luckygrantshadavaluethatwas,onaverage,21%higherthantheirvaluebasedonthereportedgrantdate.OurestimateofthedollargaintotheCEOranges(dependingonthemethod)from1.4to1.7milliondollars.

WealsoestimatetheratioofaCEO’sgainfromanopportunisticallytimedgrantinoneofthreepricesatthebottomofthedistributiontothetotalcom-pensationoftheCEO.DataontotalcompensationcomefromExecuComp(thetdc1variable),reducingthesampletothosecompaniesforwhichwehavedatafromExecuComp.Toderivethisestimate,wetaketheBlack-ScholesvalueoftheoptionsreportedbyExecuComp,anduseourmethodsforestimatingthepercentageofthisvaluethattheCEOgainedassumingthegrantwasback-dated.Wethenestimatetheaverageratioofsuchunreportedgainstototalreportedcompensationandgetestimatesof9%to10%.Becauseourprocedure

ordertocalculateBlack-Scholesvalues,weusethe3-monthT-billastherisk-freerate.As

aproxyforvolatility,weusethestandarddeviationofdailyreturnsintheyearpriortothegrant.Grantswithfewerthan30daysofstockreturnsinthepreviousyearareexcluded.

26ThisvalueiscomputedastheaverageoverBlack-Scholesoptionvaluesinthegrantmonth,wherethedailyoptionvaluesarebasedonthestrikepriceoftheactualgrantbutthestockpriceisthepriceoftheparticulardayofthemonth.Allotherparametersareheldconstant.

27Thisvalueiscomputedusingthestrikepriceoftheactualgrantandthestockpriceatthelasttradingdayofthemonth.

28Thereisanaspectofourfindingsthatmakesourestimatesofthepercentageofunderreportingconservative.Notknowingwhichgrantsintheluckygrantspoolwereproducedbybackdating,weassumethatthemanipulatedgrantsinthispoolweresimilarincharacteristicstotheother(nonmanipulated)grantsinthepool.However,ourresultsbelowsuggestthatmanipulationmighthavebeenmorelikelytooccurwhenthedifferencebetweenthelowestandmedianpriceofthemonthwashigh,whichincreasedthepercentageappreciationingrantvalueduetobackdatingtothemonth’slowestprice.

25In

2384R

TheJournalofFinance󰀆

forderivingthepercentagebywhichmanipulatedgrantswereunderreportedisconservative,webelievethattheseestimatesareconservativeandlikelysignif-icantlyunderstated.Thus,ourestimates,basedonthefullsampleofgrantingfirms,suggestthatthegainsfromgrantmanipulationwereaneconomicallysignificantpartofCEOs’regularcompensation.

WhilewehavethusfarexaminedtheprofitstoCEOsusingthestandardgrantdatevaluationmethod,wealsoverifythatopportunistictimingproducedsignificantexpostprofitstoCEOsexercisingopportunisticallytimedgrants.CEOsandotherinsidersarerequiredtoreportnotonlyoptiongrantsbutalsoexercisesofsuchgrantsandtheThompsondatabaseincludesrecordsofoptionexercises.Toidentifyeventsinwhichoptionsawardedinaluckygranteventweresubsequentlyexercised,wematchexerciseswithawardsusingthepersonID,thefirmID,theexerciseprice,andthegrantexpirationdate.WethencomparetheprofitsthatCEOsmadefromexercisingluckyoptionswiththeprofitsthattheywouldhavemadehadtheoptionsbeenawardednotatthelowestpriceofthegrantmonthbutinsteadatthemedianpriceofthatmonth.Weareabletoidentifyrealizationeventswithrespectto42%oftheluckygrants.Itshouldbenotedthatsomeoftheluckygrantswerestilloutstandingattheendofoursampleperiod(December2005),andthattheseluckygrantsmighthaveproducedsomerealizationeventsafter2005.Wefindthat,onaverage,theprofitsinrealizationeventsofoptionsawardedaspartofaluckygrantwerehigherby13%comparedwiththescenarioinwhichthegrantwasawardedatthemedianpriceofthegrantmonth.TheseresultsconfirmthatopportunistictimingsignificantlyincreasedCEOs’gainsfromoptionawardsnotonlyexantebutalsoexpost.

GiventhatthegainstoCEOsfromtheopportunistictimingofoptiongrantsweresignificant,weturntoexaminethehypothesisthatthesegainswerepro-videdasasubstituteforotherformsofcompensation.Afindingthatfirmsusingthisformofsubstitutecompensationpaidlowercompensationthroughothersources(relativetopeercompanies)wouldbeconsistentwiththisview.There-fore,wetestwhether,usingstandardcontrols,totalreportedcompensationwaslowerforCEOswhoweretherecipientsofluckygrants.

TableVIIIpresentsregressionresultsforthesubsampleofCEOsforwhomdataareavailableinExecuComp.Thedependentvariableisthenaturalloga-rithmoftotalcompensation(tdc1)fromExecuComp.Thisvariableisthesumofthesalary,bonuses,othercompensation,valueofrestrictedstock,andBlack-ScholesvalueoftheoptiongrantstotheCEO.WenotethattheBlack-ScholesvalueoftheCEO’soptiongrantsinExecuCompdoesnotincludethegainfrommanipulatingthegrantdate.Theoptionvalueiscalculatedassumingthattheoptionsareat-the-money,withastrikepricebasedonthereportedgrantdate.Totheextentthattheoptiongrantswerebackdated,theirtruevalueexceededthevaluereportedbycompaniesandincludedintdc1.Theexcessofthistruevalueoverthereportedvalueisthegainfromluck.29Thus,thereportedvalue

thereportedBlack-Scholesvalueof(backdated)optionsfromtdc1wouldprevent

usfromaccountingforthepossibilitythatcompaniesthatprovidedCEOswithbackdatedoptions

29Omitting

LuckyCEOsandLuckyDirectors

TableVIII

2385

ReportedCompensationandLuckyGrants

ThetableshowsregressionresultswherethedependentvariableisthenaturallogoftotalCEOcompensation,definedasthenaturallogofthevariabletdc1fromExecuComp.Thesampleinre-gressions1–4consistsofallfirm-yearsinExecuCompthatappearalsointheCEOgrantdatabase.Regressions5and6excludefirm-yearobservationswheretheCEOandtheindependentdirectorshavereceivedatleastonegrantonthesamedayduringtheyear.TheindependentvariablesaredescribedinTableIII.Wereportcoefficientsandstandarderrors(inparentheses)ofOLSregres-sions(regressions2,4,and6arefirmfixedeffectsregressions)withyearandindustrydummies(atthetwo-digitSIClevel).Errorsareclusteredatthefirmlevel.∗,∗∗,and∗∗∗indicatesignificanceatthe10%,5%,and1%levels,respectively.

ln(TotalCompensation)

DependentVariableLuckyCEOcurrentyearRelativegainfromluck(CEO)Luckydirectorcurrentyear

StandarddeviationofreturnsLogbookvalueROA

Industry-adjustedTobin’sQLeverageStockreturntStockreturnt−1NeweconomyTenureTenure2CEOage<50CEOage>65Constant

ObservationsR2

YeardummiesIndustrydummiesFirmfixedeffects

(1)0.076∗∗(0.035)

(2)0.056∗(0.032)

0.009(0.009)

6.963∗∗∗(1.586)0.485∗∗∗(0.013)0.173(0.174)0.099∗∗∗(0.011)−0.274∗∗∗(0.095)0.043(0.033)0.172∗∗∗(0.029)0.227∗∗∗(0.062)0.011∗∗(0.005)−0.000(0.000)−0.003(0.038)−0.237∗∗∗(0.086)3.470∗∗∗(0.141)4,3250.55YesYesNo

2.431∗(1.426)0.534∗∗∗(0.035)0.230(0.149)0.051∗∗∗(0.008)−0.421∗∗∗(0.113)0.030(0.025)0.146∗∗∗(0.023)

7.847∗∗∗(1.692)0.4∗∗∗(0.013)0.131(0.175)0.108∗∗∗(0.012)−0.339∗∗∗(0.093)0.022(0.033)0.169∗∗∗(0.030)0.233∗∗∗(0.065)0.012∗∗(0.005)−0.000∗(0.000)0.017(0.040)−0.179∗∗(0.087)3.582∗∗∗(0.143)4,3250.54YesYesNo(3)

(4)

(5)0.071∗(0.041)0.110∗∗(0.049)6.426∗∗∗(1.770)0.484∗∗∗(0.014)0.220(0.198)0.109∗∗∗(0.014)−0.231∗∗(0.108)0.042(0.040)0.186∗∗∗(0.035)0.222∗∗∗(0.073)0.012∗∗(0.005)−0.000(0.000)−0.021(0.044)−0.222∗∗(0.101)3.401∗∗∗(0.156)3,1790.55YesYesNo

(6)0.056∗(0.031)0.088∗(0.045)2.070(1.771)0.533∗∗∗(0.041)0.448∗∗(0.186)0.058∗∗∗(0.011)−0.362∗∗(0.143)0.042(0.031)0.118∗∗∗(0.028)

0.019∗∗∗(0.006)

2.361(1.494)0.546∗∗∗(0.036)0.136(0.152)0.055∗∗∗(0.009)−0.579∗∗∗(0.123)0.028(0.026)0.138∗∗∗(0.023)

0.004(0.006)−0.000(0.000)0.066(0.044)−0.274∗∗∗(0.090)0.004(0.006)−0.000∗(0.000)0.066(0.046)−0.216∗∗(0.091)0.004(0.006)−0.000(0.000)0.085(0.052)−0.291∗∗∗(0.104)

4,3250.21YesYesYes4,3250.21YesYesYes3,1790.21YesYesYes

2386R

TheJournalofFinance󰀆

ofthegrantdoesnotincludethegainfromopportunistictimingandtherefore,absentanylinkbetweendirectorluck,CEOluck,anddirectors’incentivestobenefittheCEO,shouldbeindependentofthedecisiontogranttheCEOorthedirectorsluckygrants.

Incolumns1and2,theindependentvariableofinterestisadummycalled“luckyCEO,”whichisequaltooneifthegrantwasgivenatthelowestpriceofthemonth.Incolumns3and4,theindependentvariableofinterestis“relativegainfromluck(CEO),”andisdefinedasthegainfromluckintheeventthatthegrantislucky(whichisthuszerowhenthegrantisnotlucky)dividedbytotalreportedcompensation.30

Wecontrolforotherknowndeterminantsofthelevelofcompensation,namely:thestandarddeviationofthedailystockreturnsintheyearpriortothefiscalyearwherethegrantwasgiven;thelogofthebookvalueofassets;thefirm’sreturnonassets;industry-adjustedTobin’sQ;thefirm’sleverage;thefirm’sstockreturnsintheyearofcompensationand(separately)theprioryear;adummyforwhetherthefirmisaneweconomyfirm;CEOage;indus-trydummies;andCEOtenure.AllcontrolvariablesarefromExecuCompandCompustat.Regressions1,3,and5arepooledOLSregressions.OnepotentialconcernaboutourpanelregressionsisthattheymightnotcontrolforsomerelevantbutunobservablecharacteristicsoftheCEO.Wethereforereportfirmfixedeffectsregressionsinregressions2,4,and6.WeclustertheerrorsbyCEOstocorrectforpotentialcorrelationsacrossthelevelsofcompensationforthesameCEO.31

ThecoefficientsonluckyCEOinregressions1and2arepositiveandsig-nificant(atthe5%and10%levels),allowingustorejectthehypothesisthatfirmsgrantingoptionsatthelowestpriceofthemonthpaidlowercompensa-tionrelativetopeersthroughother,reportedsources.ThemagnitudeofthecoefficientsuggeststhatwhentheCEOreceivedaluckygrant,thecompensa-tiontotheCEOfromothersourceswashigherbyabout7%(sincethisisalogregression,givingaluckygranttotheCEOwasassociatedwithanincreaseofexp(0.07)incompensation,whichisroughly1.07).Asimilarconclusionarisesfromregressions3and4inwhichthecoefficienton“relativegainfromluck”ispositiveandinregression4issignificantatthe1%level.Thus,thehigherwasaCEO’sgain(ifany)fromreceivingaluckygrant,thehigherwastheCEO’stotalcompensationfromother,reportedsources.32

Notethatourresultsinthissectionnotonlyfailtofindanegativecorrelationbetweenluckandtotalreportedcompensationfromothersources,aspredicted

reducedthenumberofoptions(andthusthereportedBlack-Scholesvalueoftheoptions)inordertooffsetthegainfromluckproducedbybackdating.

30Theresultsdisplayedinthetableuseourfirstmethodofestimatinggainsfromluck,whichassumesthatopportunisticallytimedluckygrantswere,infact,givenonadaywithapriceequaltothemonth’smedian.Usingtheothermethods,alloftheregressionsinthetableyieldsimilarresultstothosedisplayed(seetheInternetAppendix).

31Weincludeyeardummiestoaccountforunobservabletimetrends(beyondtheSOXeffect).32Infurtherexplorationofthesubstitutionhypothesis,wealsotestedthehypothesisthatluckygrantswereprovidedasatax-efficientsubstitutefornonperformancecompensationwhosedeductibilitywaslimitedbysection162(m)ofthetaxcodeto$1million.Underthishypothesis,firmsreachingthe162(m)limitationofthe$1millionnonperformance-basedcompensationwould

LuckyCEOsandLuckyDirectors2387

bythesubstitutionhypothesis,buttheyidentifyapositivecorrelation.OnepossibleexplanationforthispositivecorrelationisthatCEOs’influenceoverthepay-settingprocess,orsomeothergovernanceproblemconcerningthepay-settingprocess,resultedbothinluckygrantstotheCEOandintheCEOsreceivinghighercompensationfromothersources.

TofurtherexploretherelationbetweenCEOtotalcompensationandtimingpractices,wetestthehypothesisthatdirectorswhoreceivedluckygrantswerelesslikelytoresisthigherCEOpay.Underthishypothesis,luckygrantstodirectorswereassociatedwithhigherCEOcompensation(relativetopeers).Totestthishypothesis,weaddtoregressions1and2theluckydirectordummyindicatingwhetherthefirmgrantedluckygrantstoindependentdirectorsinthatyear.Althoughtheresultsareverysimilarusingthefullsample(seetheInternetAppendix),werestrictthesampletofirm-yearobservationswheretheCEOandthedirectorsdidnotgetagrantonthesamedayduringtheyeartobeabletoassesstheimpactofluckydirectorgrantsthatwerenotmerelyabyproductofCEOluck.Theresults,whicharedisplayedincolumn5(pooledregression)andcolumn6(firmfixedeffectsregression)ofTableVIIIindicatethatthecoefficientsontheluckydirectordummiesarepositiveandsignificant.Thecoefficientsareofsimilarmagnitudeinbothregressions,andtheresultsindicatethatwhendirectorsreceivedaluckygrant,theCEO’scompensationfortheyearwashigherbyabout11%(exp(0.110))usingtheestimatesofregression5.Thus,theevidenceisconsistentwiththehypothesisthatdirectorswhoreceivedluckygrantswerelesslikelytoconstraintheCEO’stotalreportedcompensation.

Inthenextsection,wefurtherinvestigatetheassociationbetweenCEOinfluence,governancearrangements,andluckygrants.

IV.FactorsAssociatedwithCEOandDirectorLuck

Thissectionexamineswhichfirm-,CEO-,andgrant-levelfactors(otherthantheexistenceofsimultaneousawardstoindependentdirectors)areasso-ciatedwithopportunistictiming.Inparticular,weaskwhetherthepoten-tialbenefittotheCEOfromopportunisticallytimedgrantsandtheleveloftheCEO’sinfluenceoninternaldecision-makingarecorrelatedwiththe

benefitfromgrantingnonperformance-basedcompensationviaopportunisticallytimedoptionsbecausesuchoptionsaretaxdeductible.Thepossibilitythatbackdatinghadbeenpartlymotivatedbysection162(m)wasstressedinaWallStreetJournaleditorial(“BackdatingtotheFuture,”October12,2006)andwasoneofthefactorsleadingtohearingsonthetaxtreatmentofexecutivepayheldbytheSenateFinanceCommittee.Totestthishypothesis,weexaminedwhetherluckygrantswererelatedtotheleveloftheCEO’ssalary.Inparticular,weranlogitregressionswherethesampleconsistsofalltradingdaysinmonthsinwhichaCEOreceivedagrantandthedependentvariableisequaltooneifthetradingdayhasagrantdayandzerootherwise.Thevariablesofinterestwereaninteractiontermbetweenadummyvariableindicatingwhetherthetradingdatehadtheloweststockpriceofthemonthandthreesalary-leveldummiesindicatingwhethertheCEO’ssalarywaslowerthan,equalto,orhigherthan$1million.Wefoundthatalltheinteractionvariableswereinsignificant,thusprovidingnoevidencefortheviewthatopportunistictimingcouldbeexplainedasanattempttoprovidenonperformancecompensationinatax-efficientwaythatavoidssection162(m)penalties(seeBebchuk,Grinstein,andPeyer(2006a)).

2388R

TheJournalofFinance󰀆

occurrenceofluckygranteventsbothintheCEOsampleandthedirectorsample.

Sincemanyluckygranteventsowetheirstatustoopportunistictiming,luckygranteventsprovideausefulbasisforstudyingthefactorslikelytobeassociatedwithsuchtiming.Naturally,someluckygranteventsweretheproductof“pureluck.”However,granteventsthatwereluckyasaresultofpureluckarenotexpectedtobesystematicallyrelatedtofirm,CEO,orgrantcharacteristics(atleastaftercontrollingforfactorsaffectingtheprobabilityofluckunderrandomselection,suchasthefractionofdaysofthegranteventmonththathadstockpricesatthelowestlevelofthemonth).Thus,totheextentthatluckygranteventsarecorrelatedwithsuchcharacteristics,thiscorrelationcanbeattributedtocorrelationbetweenopportunisticallytimedluckygrantsandthesecharacteristics.

WeruntheregressionsbelowoncewhenthedependentvariableisadummyforaluckydirectorgranteventandoncewhenthedependentvariableisadummyforaluckyCEOgrantevent.Inalloftheregressions,wecontrolforrelativefirmsize,SOX,andthefractionoftradingdaysofthemonththathadastockpriceatthelowestlevelofthemonth.Insomeofthespecifications,weincludefirmorCEOfixedeffects.Whenfixedeffectsarenotincluded,wealsocontrolforwhetherthefirmwasinaneweconomyindustry.AcrossallregressionsweclustertheerrorsbyCEOtocorrectforpotentialcorrelationsacrossthelikelihoodofluckygrantevents,eithertotheCEOortothedirectors,amongthesameCEOs.

A.TheTimingofOpportunisticTiming

WefirstexaminetheextenttowhichgranteventstoCEOsanddirectorsinagivenfirmweremorelikelytobeluckywhenthepayoffsfromsuchluckwerehigher.Ifopportunistictimingofoptiongrantswastheproductofeconomicdecisionsaimedatmaximizinginsidepayoffs,thenweshouldexpecttoseeahigherpropensityofgranteventsbeingluckyinmonthswithhigherpotentialgainsfromsuchtiming.Alternatively,ifopportunistictimingwastheproductofhabitualfollowingoffirms’practices,thenweshouldnotexpecttofindsuchasignificantassociation.

Toexplorethisissue,weemployfirmandCEOfixedeffectsregressiontech-niques.Thisapproachallowsustoinvestigatewhether,duringthetermofanygivenCEOinagivencompany,opportunistictimingwasmorelikelyonocca-sionsinwhichtheassociatedpayoffswerelarger.Ourfixedeffectanalysisaddstothecross-sectionalanalysisofHeronandLie(2009),whoreportapositivecorrelationbetweenopportunistictimingandafirm’sstockreturnvolatility.Whiletheircross-sectionalfindingsmaybeduetoopportunistictimingbeingmorecommoninfirmswithhighvolatility(suchashigh-techfirms),ouranal-ysisfocusesondifferencesintheoddsofluckygranteventsovertimewithinthesamefirm.

Thedependentvariableinourfixedeffectregressionsisadummyvariableindicatingwhetheragranteventwaslucky.Werunourregressionsseparately

LuckyCEOsandLuckyDirectors

TableIX

23

TimingofOpportunisticTiming

Thetableshowsthefixedeffectlogitregressionresultswherethedependentvariableisadummyvariableforwhetherthegranteventwasatthelowestpriceofthemonthandzerootherwise.TheindependentvariablesaredescribedinTableIII.Duetodataavailability,thesampleisreducedto18,543observationsforCEOgranteventsand25,888fordirectorgrantevents.Thenumbersinparenthesesaretheestimatedstandarderrorsofthecoefficients,adjustedforclusteringattheexecutivelevel.∗,∗∗,∗∗∗indicatesignificanceatthe10%,5%,and1%levels,respectively.

Sample:DirectorGrant

EventsFirmFixedEffect1.678∗∗∗(0.299)

CEOFixedEffect

Sample:CEOGrantEvents

DependentVariable:LuckyGrantDummyDifferencebetweenthemedianandlowestpriceMarketcomponentofthemedianprice−lowestpricedifference

Firm-specificcomponentofthemedianprice−lowestpricedifferenceRelativesizeSOX

DaysinmonthlowestObservations

FirmFixedEffect0.0∗∗∗(0.250)

CEOFixedEffect

1.832∗(1.107)0.771∗∗∗(0.267)

3.111∗∗∗(0.937)1.597∗∗∗(0.304)

0.156∗∗∗(0.046)−0.548∗∗∗(0.071)6.285∗∗∗(0.729)18,543

0.160∗∗∗(0.053)−0.565∗∗∗(0.084)6.782∗∗∗(0.902)18,543

0.086∗∗(0.039)−0.265∗∗∗(0.065)8.493∗∗∗(0.773)25,888

0.084∗∗(0.039)−0.263∗∗∗(0.065)8.547∗∗∗(0.776)25,888

forCEOgranteventsanddirectorgrantevents.Theindependentvariableofinterestisthepercentagedifferencebetweenthelowestandmedianpriceofthegrantmonth(inlog)asanindependentvariable.Thisvariableisusedasaproxyforthepotentialpayoffsfromturningagrantactuallygivenonanotherdayintoaluckygrant.

TheresultsofthesefixedeffectregressionsaredisplayedinTableIX.Wefindthat,controllingforCEOfixedeffects,thecoefficientonthelowest-mediandifferenceispositiveandsignificantbothintheCEOsampleregression(column1)andinthedirectorsampleregression(column3).TheseresultsindicatethatthecorrelationbetweenopportunistictimingandstockpricevolatilitydocumentedinHeronandLie(2009)wasnotdueentirelytocross-sectionaldifferences,thatis,differencesbetweenhigh-andlow-volatilityfirms.DuringthetenureofanygivenCEOinanygivenfirm,CEOgranteventsanddirectorgranteventsweremorelikelytobeluckyinmonthsinwhichthedifferencebetweenthelowestandmedianpricewasrelativelylarge,thatis,potentialgainsfromopportunistictimingwerelarger.

Theregressionsdisplayedincolumns2and4ofTableIXdecomposethepercentagedifferencebetweenthelowestandmedianpriceofthegrantmonth

2390R

TheJournalofFinance󰀆

intotwocomponents:thecomponentdrivenbymarketpricemovements,andthecomponentdrivenbyfirm-specificpricemovements.33Wefindthatboththemarketcomponentandthefirm-specificcomponentarepositiveandsignificantinbothregressions.Overall,theresultsofTableIXindicatethatthetimingofwhentoengageinopportunistictimingwasitselfopportunisticallytimed—thatis,doneinawaythatwasbetter-than-randomfromtheperspectiveofCEOsanddirectors.Theresultsarethusconsistentwiththeviewthattheuseofopportunistictimingwasnotaresultofathoughtless,habitualfollowingofapractice,butratheraconsequenceofaneconomicdecisionthatwassensitivetopayoffs.

B.LuckandCEOInfluence

Inthissection,wetestthehypothesisthattheoddsofagranteventbe-ingluckyarehigherinthepresenceoffactorsassociatedwithgreaterCEOinfluenceandpoweroverinternalpay-settingprocesses.Totheextentthatop-portunistictimingwasmerelyaproductofarationalbusinessdecisionbythefirmtoprovidenonperformancecompensation,theincidenceofluckygrantsshouldnotcorrelatewithsuchfactors.However,ifopportunistictimingwasproducedbyagencyproblems,thenluckygranteventsareexpectedtocorrelatewithsuchfactors.

B.1.CEOTenureandBackground

Asintheprecedingsection,werunregressions,forboththeCEOgrantssam-pleandthedirectorgrantssample,inwhichthedependentvariableiswhetheragranteventwaslucky.Inadditiontotheexplanatoryvariablesincludedinourearlierbenchmarkregression(seeTableIX),weaddCEOownership(anditssquare)asadditionalcontrolvariables.Ourvariableofinterest,obtainedfromtheExecuCompdatabase,isCEOtenure(inlogs).34ThelongertheCEOtenure,themoreinfluencetheCEOcanbeexpectedtohavehadondirectorsandinternalpaypractices(Coreetal.(1999),Cyert,Kang,andKumar(2002),andHarfordandLi(2007)).

WereporttheresultsinTableX.Columns1and5displaytheresultsoftheregressionforthesampleofCEOgranteventsanddirectorgrantevents,respectively.Inbothregressions,consistentwiththehypothesisthatoppor-tunistictimingwascorrelatedwithCEOpower,thecoefficientonCEOtenureispositiveandsignificant.

Wefurtherexploretheeffectoftenureonthelikelihoodofluckbysepa-ratingthetenureeffectbetweenCEOswhosepreviouspositionwaswiththefirm(insiderCEOs)andCEOswhosepreviouspositionwaswithadifferent

33ThisdecompositionisinthespiritofLie(2005),whousesittoshowthatthepatternofreturns

accompanyinggranteventsreflectsbackdatingandnotmerelyspring-loading.

34TheneedforinformationaboutthefirmtobeavailableonExecuCompreducesthesizeofthesampleto6,001(CEOgranteventssample)and7,990(directorgranteventssample).

TableXLuckandCorporateGovernanceThetableshowslogitregressionresultswherethedependentvariableisadummyvariableforwhetherthegrantwasgivenatthedateatwhichthelowestpriceofthemonthprevailedandzerootherwise.Thesampleconsistsof6,001(7,990)CEO(director)observationswithatleastonegranteventandwithgovernanceinformationatthefirmlevelfromtheIRRCandtheExecuCompdatabases.Regression4excludesCEOgrantswheredirectorsdidreceiveagrantonthesameday.TheindependentvariablesaredescribedinTableIII.Thenumbersinparenthesesaretheestimatedstandarderrorsofthecoefficients,adjustedforclusteringattheexecutivelevel.∗,∗∗,and∗∗∗indicatesignificanceatthe10%,5%,and1%levels,respectively.CEOLuck(1)0.1395∗∗(0.052)0.179∗∗∗(0.0)(2)(3)(4)(5)0.017∗∗(0.009)DirectorLuck(6)(7)DependentVariable:LuckyGrantDummyCEOtenureCEOtenure×CEOoutsiderdummyCEOtenure×CEOinsiderdummyCEOoutsiderdummy0.359∗∗∗(0.117)0.096∗(0.057)−0.614∗∗(0.293)0.023∗∗(0.012)−0.001(0.010)−0.297∗(0.166)LuckyCEOsandLuckyDirectors

Independentboarddummy0.374∗∗∗(0.139)0.158∗∗(0.071)−0.354(0.349)−0.440∗∗∗(0.146)0.037∗∗(0.015)0.015(0.011)−0.169(0.203)−0.255∗(0.145)Independentboarddummy×LuckydirectorcurrentyearLuckydirectorcurrentyear0.100(0.327)Independentcompensationcommitteedummy−0.790∗∗∗(0.152)0.719∗∗∗(0.199)0.247∗∗(0.116)0.112(0.334)0.087(0.167)(continued)2391

2392

TableX—ContinuedCEOLuck(1)0.595(0.594)−1.231∗(0.713)0.530(0.594)−1.176∗(0.721)(2)(3)(4)(5)(6)DirectorLuck(7)−0.018(0.600)−0.753∗∗(0.325)DependentVariable:LuckyGrantDummy5%blockholderoncompensationcommitteedummy5%blockholderoncompensationcommittee×independentcompensationcommitteedummyControlvariables:CEOownership>5%and<25%dummyCEOownership>25%dummyRelativesizeNeweconomyR

TheJournalofFinance󰀆

SOXDifferencebetweenthemedianandlowestpriceDaysinmonthlowestConstant6,0010.376∗∗(0.143)0.456(0.330)−0.05∗(0.029)0.093(0.132)−0.281∗∗∗(0.102)1.710∗∗∗(0.388)7.396∗∗∗(1.756)−2.918∗∗∗(0.165)0.127(0.193)−0.519(0.546)0.007(0.036)0.125(0.1)−0.212∗(0.120)2.609∗∗∗(0.473)13.866∗∗∗(2.5)−3.231∗∗∗(0.385)6,0014,1993,8110.128(0.185)−0.296(0.519)0.003(0.037)−0.262∗∗∗(0.122)−0.221∗∗(0.117)2.562∗∗∗(0.483)13.830∗∗∗(2.608)−3.277∗∗∗(0.381)0.348∗∗(0.145)0.393(0.333)−0.052∗(0.029)0.097(0.134)−0.277∗∗∗(0.102)1.726∗∗∗(0.388)7.594∗∗∗(1.769)−2.841∗∗∗(0.169)0.470∗∗(0.194)−0.386(0.509)−0.032(0.036)0.133(0.175)−0.110(0.110)2.508∗∗∗(0.582)7.862∗∗∗(2.036)−3.394∗∗∗(0.197)7,9900.426∗∗(0.167)−0.344(0.433)−0.044(0.032)0.063(0.159)−0.082(0.105)2.071∗∗∗(0.529)7.610∗∗∗(1.620)−3.098∗∗∗(0.170)7,9900.371∗(0.203)−0.788(0.550)−0.005(0.040)0.211(0.200)0.011(0.126)3.267∗∗∗(0.662)26.75∗(15.00)−3.339∗∗∗(0.296)5,181ObservationsLuckyCEOsandLuckyDirectors2393

firm(outsideCEOs).Weexpecttenuretohaveagreatereffectonthelike-lihoodofluckwhentheCEOwasanoutsiderthanwhentheCEOwasaninsider,sinceinsiderCEOsarelikelytohaverelationshipswithdirectorsevenintheirfirstyearsasCEOs.WethereforeincludeadummyvariableforwhethertheCEOwasanoutsider,andwealsointeractthetenurevariablewithtwodummyvariables:oneforwhethertheCEOwasaninsiderandtheotherforwhethertheCEOwasanoutsider.Wepresenttheresultsincolumns2and6.ThecoefficientsontheCEOoutsiderdummyarenegativeandsignificantinboththeCEOregressionandthedirectorregression.Furthermore,thecoeffi-cientsonthetenure–outsiderinteractionvariablearelargerthanthoseonthetenure–insiderinteractionvariable.

Theresultsaregenerallyconsistentwiththenotionthatopportunistictim-ing,bothofCEOgrantsandofdirectorgrants,wasmorelikelywhenCEOshadmoreinfluenceoncorporatedecision-making.

B.2.BoardandCommitteeCompositionandIndependence

OurnextstepistoaddadditionalexplanatoryvariablesfromtheIRRCdatabasetoexplorehowCEOanddirectorluckwereassociatedwiththecompo-sitionoftheboardandcompensationcommittee.35Thecompensationcommit-teeisinchargeofnegotiatingacompensationcontractwiththemanager,andthereforeitisimportanttoexplorewhethercompensationcommitteestructureandincentivesofcommitteemembersplayedaroleinthedecisiontoallocateluckygrants.Theboardappointsandoverseesthecompensationcommittee.Inparticular,weareinterestedintheextenttowhichtheboardandthecompensationcommitteewereindependent.Directorindependencehasbeenviewedasaninstrumenttoimproveboardoversightingeneral,andover-sightoverexecutivecompensationinparticular.Statecorporatelawshavelongencouraged,andstockexchangeshaverecentlyrequired,amajorityofindependentdirectorsontheboardandacompensationcommitteeconsistingofindependentdirectors.Whileformalindependencerequirementscouldbein-sufficienttoeliminateCEOinfluenceondirectors(BebchukandFried(2004)),theyarelikelytoreduceit.

WeusetheIRRCdefinitionofdirectorindependencetodeterminewhetherdirectorswereindependent.Usingthesedata,weaddtwoexplanatoryvari-ables:adummyvariableforwhethertheboardhadamajorityofindependentdirectors,andadummyvariableforwhetherthecompensationcommitteecon-sistedofindependentdirectors.

Wealsoaddadummyvariableforwhetherthecompensationcommitteehadatleastoneblockholder,definedasadirectorholdingatleast5%ofthesharesofthecompanyandnotemployedbyit(informationonthenumberofsharesthateachdirectorheldisavailablefromIRRC).Anoutsideblockholderisexpected

needfortherelevantinformationaboutthefirmtobeavailableontheIRRCdatabase

furtherreducesthesizeofoursampleto4,199grantevents(CEOsample)and5,181(directorsample).

35The

2394R

TheJournalofFinance󰀆

tohaverelativelystrongincentivestomaximizesharevalueandthereforetoactindependentofmanagementandseekcompensationarrangementsthatserveshareholderinterests.Sincetheextenttowhichablockholderonthecompensationcommitteecanbeexpectedtoserveshareholdervaluemightbeinfluencedbywhethertheblockholderandothermembersofthecommitteeareindependent,wealsoincludeaninteractiondummyforwhetherthecom-pensationcommitteewasbothindependentandhadablockholderonit.

OurresultsaredisplayedinTableX,column3(CEOsample)andcolumn7(directorsample).Thecoefficientontheindependentboarddummyisnegativeandsignificantatthe1%levelinbothregressions,indicatingthatopportunistictimingisassociatedwithboardsthatlackedamajorityofindependentdirec-tors.ThesizeofthecoefficientimpliesthathavingamajorityofindependentdirectorsontheboardreducedtheoddsthataCEOgranteventwouldbeluckyby36%(=exp(−0.44)−1),andreducedtheoddsthatadirectorgranteventwouldbeluckyby23%(=exp(−0.255)−1).Thisresultisconsistentwiththeviewthatopportunistictimingreflectsgovernance/agencyproblems.

GivenourfindinginSectionIIthatsomeindependentdirectorsreceivedop-portunisticallytimedluckygrantsthemselves,wealsoexaminetherelationbetweenthetimingofdirectorgrantsandtheeffectivenessofindependentdi-rectorsinlimitingtimingpractices.Inparticular,wetestthehypothesisthattheassociationbetweenboardindependenceandloweroddsofCEOluckweak-enedwhenindependentdirectorsreceivedluckygrants.Totestthishypothe-sis,weaddtotheCEOluckregressionofcolumn3aluckydirectordummy,indicatingwhetherthedirectorsreceivedluckygrantsinthatyear,andaninteractionvariablebetweentheindependentboarddummyandtheluckydi-rectordummy.ToavoidamechanicallinkproducedbyluckydirectorgrantsthatcoincidedwithluckyCEOgrants,werestrictthesampletoobservationswheretheCEOdidnotreceiveagrantonthesamedayastheluckydirectors.Ourresultsaredisplayedincolumn4ofTableX.Wefindthatthecoefficientontheinteractionvariableispositiveandhighlysignificant.Themagnitudeofthecoefficientontheinteractionvariable(0.719)isveryclosetothecoefficientontheindependentboarddummy(−0.790)andthesumofthetwoisnotsignif-icantlydifferentfromzero.Thus,thedataareconsistentwiththehypothesisthatthecorrelationbetweenboardindependenceandreducedoddsofCEOluckwasweakened,andindeedthereisnoevidenceforitsexistence,whenindependentdirectorsreceivedluckygrantsthemselvesduringtheyear.Thecoefficientontheluckydirectordummyisalsopositiveandhighlysignificant,indicatingthattheCEOwasmorelikelytobeluckywhenindependentdi-rectorsreceivedluckygrants—evenafterexcludingdirectorgrantscoincidingwithCEOgrants.Thisresultisalsoconsistentwiththehypothesisthatluckydirectorswerelesslikelytoconstraintimingpractices.Overall,theresultsofcolumn4highlightthattheformalclassificationofdirectorsasindependentdoesnotfullydeterminetheirtruelevelofindependence.

AlltheregressionsinTableXindicatethathavingacompensationcommitteethatisindependentandthatincludesablockholderreducestheoddsofagranteventbeinglucky.Thecoefficientontheindependentcompensationcommittee

LuckyCEOsandLuckyDirectors2395

dummyandthecoefficientonthecompensationcommitteeblockholderdummyarenotsignificantbythemselvesbuttheirinteractionissignificantaswellaseconomicallymeaningful.Usingthecoefficientestimatesinregressions3and7indicatesthathavingacompensationcommitteethatisindependentandincludesablockholderreducedtheoddsofaCEOgranteventbeingluckyby71%(=exp(−1.231)−1)andreducedtheoddsofadirectorgranteventbeingluckyby53%(=exp(−0.753)−1).

Itisworthnotingthatthesignificanceoftenuredoesnotgoaway(norevensubstantiallychangeinmagnitude)whenboardandcommitteeindependencevariablesareadded.ThisresultisconsistentwiththeviewthatthepresenceorabsenceofamajorityofformallyindependentdirectorsontheboarddidnotfullydeterminetheextenttowhichtheboardwasinfluencedbytheCEOinmakingcompensationandoversightdecisions.36

Ourfindingsconcerningboardindependenceandcommitteeindependencecontributetotheliteratureonthepotentialbenefitsofindependentdirectors.Whilepreviousresearchhasnotbeenabletoestablishalinkbetweenboardindependenceandbettercorporateperformanceingeneral(e.g.,BhagatandBlack(1999,2002)andChhaochhariaandGrinstein(2007)),somespecifictypesofdecisionsforwhichsuchindependencemattershavebeenidentified(e.g.,Weisbach(1987),ByrdandHickman(1992),Shivdasani(1993),Brickley,Coles,andTerry(1994),Cotter,Shivdasani,andZenner(1997),Dann,Guercio,andPartch(2003),andGillette,Noe,andRebello(2003)).Inparticular,ithasbeenshownthatdirectorindependencehasanimpactonexecutivecompensationdecisions(e.g.,Coreetal.(1999)andChhaochhariaandGrinstein(2009))andontheincidenceoffraud(e.g.,Beasley(1996,2000)andDechow,Sloan,andSweeny(1996)).Thus,opportunistictimingisoneofthecontextsinwhichdirectorindependenceappearstomakeadifference.

Ourfindingsalsocontributetotheworkonthepotentialgovernancebenefitsofblockholders(BertrandandMullainathan(2000,2001)).Thisworkshowshowtheexistenceofalargeoutsideblockholdermakesadifferenceforcertainaspectsoffirmbehavior.Weshowthat,withrespecttoopportunistictiming,anoutsideblockholderneedstobeonthecompensationcommitteetomakeadifference.

Finally,wealsoreplacetheneweconomydummywithindustrydummiesbasedontheFama–Frenchclassificationinto12industrysectors.ConsistentwiththeunivariateresultsnotedinSectionI,wefindthatnoneoftheindustrydummiesisstatisticallysignificantlydifferentfromtheothers.Thus,wedonotfindsupportforthehypothesisthatindustrynormsandculturewereimpor-tantdriversoftimingpractices(Fleischer(2007)andWalker(2006)).Oncethepayoffsfromtimingandgovernanceprovisionsarecontrolledfor,thereisnostatisticallysignificantdifferenceacrossindustries.

additiontodirectorindependenceanddirectorownership,theremightwellbeotherchar-acteristicsofservingdirectorsthatarerelevanttotheoddsofluckygrantsthatouranalysisdoesnotidentify.Inparticular,inacontemporaneousstudythatcomplementsourwork,Bizjaketal.(2009)showalinkbetweenthespreadofoptionbackdatingandinterlockingdirectors.

36In

2396

C.SerialLuck

R

TheJournalofFinance󰀆

TheprecedingsubsectionsidentifyanumberofvariablesthatarecorrelatedwithluckygrantstoCEOsanddirectors.Undoubtedly,thereareCEOandfirmtraitsthatcouldaffecttheincidenceofluckygranteventsthatarenotincluded.CharacteristicssuchasaspectsoftheCEO’spersonalityandthefirm’scompensationstaffmightbedifficultorimpossibleforresearcherstoobserve.However,totheextentthatsuchtraitsexist,onewouldexpectlucktobe“serial”or“persistent.”Thatis,controllingforallofthevariablesusedthusfar,onewouldstillexpectagranteventtobemorelikelytobeluckyifaprecedinggrantwaslucky.Suchpersistencewouldnotbeexpected,ofcourse,underrandomselection.

Toexaminetheexistenceandmagnitudeofsuchpersistence,were-runtheregressionsreportedinTableXbutthistimeaddingtwonewdummyvariables.Thefirstdummyvariableisequaltooneiftheprecedinggranteventinourdatasetwaslucky.Theotherdummyvariableisequaltooneifaprecedinggranteventexistsanditwasnotlucky.Ourdefaultisthusgrantsthatwerenotprecededinourdatasetbyanothergrant.WerunoneregressionwherethesampleincludesallCEOgrantevents,andthedependentvariableisadummyforwhethertheCEOgranteventwaslucky,andanotherregressionwherethesampleincludesalldirectorgrantevents,andthedependentvariableisadummyforwhetherthedirectorgranteventwaslucky.

TableXIdisplaystheresults.Inbothregressions,thecoefficientonthepre-viousluckydummyispositiveandsignificant.IntheCEOregression,thecoefficientonaprecedinggranteventbeingluckyis0.366,whichimpliesthat,comparedtoCEOgranteventsforwhichwehavenoinformationaboutpreced-inggrants,aCEOgranteventprecededbyaluckyCEOgranteventwas44%morelikelytobelucky.Thecoefficientonthedummyforhavingaprecedinggrantthatwasnotlucky(whichlumpstogetherallotherpriceranks,includingprecedinggrantsatthesecondlowestpriceofthemonth)isnegativeinbothregressionsbutsignificantonlyinthedirectorregression.

Regressions3and4seektoaddressthepossibilitythatourpreviousgrantdummiessufferfromabiasbecausewecannotclassifyafirm’sfirstgrantasbeingprecededbyaluckyornot-luckygrant.Intheseregressions,weeliminatethefirstgrantofeachfirmfromthesampleandthusforeachgrantremaininginthesampleweknowwhetheritsprecedinggrantwaslucky.Thecoefficientsonthe“previousgranteventlucky”dummiesinregressions3and4arepos-itiveandsignificant(atthe1%confidence).Furthermore,thecoefficientonregression3(4)isverysimilarinmagnitudetothedifferenceinthecoefficientsbetween“previousgranteventlucky”and“previousgranteventunlucky”forregression1(2).Thus,weconcludethatthereislittlereasontobelievethecoefficientestimatesinregressions1and2weresignificantlybiasedduetothefactthatwedonotknowforsomegrantsthepreviousgrant’sluckstatus.

Insum,evenaftercontrollingforallofthevariablesthatwehavefoundtobeassociatedwithopportunistictimingandadditionalcontrols,agranteventtoaCEOortoindependentdirectorswasmorelikelytobeluckywhenapreceding

LuckyCEOsandLuckyDirectors

TableXI

2397

SerialLuck

Thetableshowstheresultsoflogitregressions.ThedependentvariableinthefirsttworegressionsisadummyequaltooneiftheCEOgranteventwasatthelowestpriceofthemonthandzerootherwise.Inthethirdandfourthregressions,thedependentvariableisoneifthedirectorgranteventwasatthelowestpriceofthemonthandzerootherwise.ThesamplesizeisreducedbecausedatafromExecuCompandIRRCarerequired.Thefirstandthirdregressionsincludeallgrantevents.Thesecondandfourthregressionsexcludethefirsteventinourdatabasesothatwecandeterminewhetherthepreviousgranteventwasluckyornot.TheindependentvariablesaredescribedinTableIII.Thenumbersinparenthesesaretheestimatedstandarderrorsofthecoefficients,adjustedforclusteringattheexecutivelevel.∗,∗∗,and∗∗∗indicatesignificanceatthe10%,5%,and1%levels,respectively.DependentVariable:LuckyGrantDummyPreviousgranteventlucky

PreviousgranteventnotluckyRelativesizeSOX

DifferencebetweenthemedianandlowestpriceCEOtenure×CEOoutsiderdummyCEOtenure×CEOinsiderdummy

CEOoutsiderdummyIndependentboarddummyIndependentcompensation

committeedummy5%blockholderon

compensationcommitteedummy

5%blockholderon

compensationcommittee×independentcompensationcommitteedummy

CEOownership>5%and<25%dummy

CEOownership>25%dummy

DaysinmonthlowestConstant

IndustrydummiesObservations

LuckyCEOs

0.366∗(0.192)−0.122(0.125)0.023(0.042)−0.225∗(0.122)2.376∗∗∗(0.656)0.171∗∗(0.071)0.386∗∗(0.170)−0.388(0.392)−0.420∗∗(0.165)0.155(0.322)0.575(0.555)−1.203∗(0.673)

0.559∗∗∗(0.196)

LuckyDirectors0.229∗∗(0.111)−0.424∗∗(0.192)−0.024(0.039)0.018(0.123)2.785∗∗∗(0.668)0.014∗(0.011)0.037∗∗(0.014)−0.207(0.201)−0.240∗(0.136)0.100(0.165)0.015(0.585)−0.765∗∗(0.373)0.352∗(0.203)−0.820∗(0.509)4.793∗(2.856)−3.259∗∗∗(0.453)Yes5,181

0.658∗∗∗(0.184)

0.038(0.054)−0.217∗∗∗(0.157)2.621∗∗∗(0.699)0.392(0.267)0.055(0.103)−0.634(0.634)−0.281(0.223)0.924(0.637)1.102(1.069)−1.724(1.202)

−0.029(0.042)0.012(0.126)2.470∗∗∗(0.749)0.015∗(0.011)0.035∗∗(0.014)−0.209(0.207)−0.311∗∗(0.139)0.057(0.169)−0.401(0.650)−0.965∗∗(0.435)

0.102(0.208)−0.530(0.583)14.942∗∗∗(2.628)−3.581∗∗∗(0.487)Yes4,1990.155(0.266)−0.669(1.054)17.785∗∗∗(3.518)−4.455∗∗∗(0.755)Yes2,3740.266(0.229)−0.626(0.560)4.751(3.010)−3.451∗∗∗(0.438)Yes4,794

2398R

TheJournalofFinance󰀆

granteventtothisCEOortodirectorsinthisfirm,respectively,waslucky.Thepresenceofsuchserialluckindicatesthat,beyondthefactorswehaveidentified,opportunistictimingwasalsodrivenbyadditionalCEOandfirmcharacteristicsthatsubsequentresearchwouldhopefullyidentify.37

V.Epilogue:TheEndofBackdating

TheanalysisoftheprecedingsectionsdocumentsthatthepassageofSOXin2002didnoteliminateopportunisticallytimedluckygrants;asTablesIandIIdocument,asignificantincidenceofsuchgrantsexistedduringtheperiodbetweenthepassageofSOXandtheendof2005.Inthissection,weextendtheperiodthatwehaveinvestigatedtoexaminehowbackdatingwasaffectedbythemassivemediaandregulatoryattentiondevotedtothispracticestartinginthespringof2006.

InMarch2006,theWallStreetJournal(WSJ)publishedits“perfectpayday”article,describingindetailseveralegregiousexamplesofbackdating.Thisarticlewasfollowedinsubsequentweeksbyaseriesofadditionalstoriesonthesubject.Thisseries,whichsubsequentlywonaPulitzerPrizeforpublicservicereporting,ledtonumerousprobesbyregulators,investorgroups,andplaintifflawyers.Toexaminehowtheintensepublicattentiontoopportunistictimingaffectedthepractice,weextendourdatasettotheyears2006and2007andexaminetheincidenceofluckygrantsduringthe19-monthperiodfromApril2006(themonthfollowingtheinitialWSJarticle)totheendof2007.WeuseproceduressimilartothoseemployedinSectionItoidentifyluckygrantsandtoestimatetheincidenceofluckygrantsduetoopportunistictimingduringthisperiod.DuringApril2006toDecember2007,outof2,812CEOgrantevents,158(5.6%)wereluckygrants,whichisclosetotheexpectednumberofluckygrantsunderrandomassignmentof147.Thus,duringthe19-monthperiodfollowingtheappearanceoftheWSJarticle,theactualnumberofluckyCEOgrantsexceededtheexpectednumberunderrandomallocationbyamere7%.Incontrast,repeatingthiscalculationforthe19-monthperiodendinginMarch2006(September2004toMarch2006),wefindthatthenumberofluckyCEOgrantswasalmost50%higherthanexpectedunderrandomassignment.Wefindsimilarresultswithrespecttodirectorgrantevents.Duringthe19-monthperiod(April2006toDecember2007)followingtheappearanceoftheWSJarticle,thenumberofdirectorgranteventsthatwereactuallyluckyexceededthenumberofluckygranteventsexpectedunderrandomassignmentbyabout12%.Incontrast,duringthepreceding19-monthperiod(September2004toMarch2006),thenumberofdirectgranteventsthatwereactuallyluckyexceededtheexpectednumberofluckygranteventsbyabout25%.

Overall,wefindthatoncethepracticeofbackdatingcameintothelimelightinthespringof2006,theincidenceofopportunisticallytimedluckygrants

Cremers,andPeyer(2009)identifyonesuchadditionalvariablethatcanhelp

explaintheincidenceofluckygrants—theCEOpayslice,whichtheydefineasafractionoftheaggregatecompensationawardedtothetopfiveexecutivescapturedbytheCEO.TheyshowthatahigherCEOpaysliceisassociatedwithhigheroddsoftheCEOgettingaluckygrant.

37Bebchuk,

LuckyCEOsandLuckyDirectors2399

declineddrastically.Facingclosescrutinyofthetimingofoptiongrantsbyoutsiders,decision-makersinsidefirmsrespondedbycurtailingtheiruseofopportunistictiming.Whileinsideropportunismintheparticularformofback-datingmightcontinuetobepreventedbycloseoutsidescrutiny,understandinghowsuchopportunismwasinfluencedbygovernancemechanismsandinsiders’incentivesremainsimportant.

VI.Conclusion

Inthispaper,weinvestigatetheopportunistictimingofoptionstoCEOsandindependentdirectorsduringthe1996to2005period.Opportunistictimingincreasestheincidenceofluckygrantsgivenondayswiththelowestpriceofthemonth.Weusetheoccurrenceofluckygrantsasabasisforinvestigat-ingtherelationbetweenopportunistictimingofoptiongrantsandcorporategovernance.

Weshowthatthegrantsawardedtoindependentdirectors,whoarechargedwithoverseeingthecompany’sexecutives,werethemselvesaffectedbyoppor-tunistictiming.Thetimingofdirectorgrantswasnotmerelyabyproductofthedirectorsbeingsimultaneouslyawardedgrantswithexecutivesoroffirmsroutinelytiminggrantstoallrecipients.

WealsofindthatCEOswhoreceivedluckygrantshadhigherincomefromother(reported)sourcesofcompensation,thusfindingnoevidenceforthehypothesisthatfirmsprovidingopportunisticallytimedCEOgrantsreducedCEOs’compensationfromothersources.Wealsodonotfindsupportinthedatafortheviewthatopportunistictimingwasaresultofthehabitualfollowingoffirmorindustrypractices.Firms’choiceswithrespecttowhentoengageinopportunistictimingwerethemselvestimedinwaysthatbenefitedgrantrecipients;foranygivenfirm,grantstobothCEOsandindependentdirectorsweremorelikelytobeluckywhenthepayoffsfromsuchluckwerehigher.Opportunistictiming,wefind,iscorrelatedwiththreevariablesassociatedwithgreaterCEOinfluenceonpay-setting.Inparticular,CEOgranteventsanddirectorgranteventswerebothmorelikelytobeluckywhenthecompanylackedamajorityofindependentdirectorsontheboard,whenitdidnothaveanindependentcompensationcommitteewithanoutsideblockholderonit,orwhenithadalong-servingCEO.

WefindthatCEOluckandCEOreportedcompensationwereassociatedwithdirectorluck,evenafterexcludingdirectorgrantscoincidingwithCEOgrants.Inaddition,theassociationbetweenhavingamajorityofindependentdirectorsandreducedCEOluckdisappearswhenindependentdirectorsreceivedluckygrantsthemselves.Theformalclassificationofdirectorsasindependentthusmightnotfullydeterminehowtheyperformintermsofconstrainingexecutivecompensationpracticesandlevels.

Inclosing,ouranalysisshowstheexistenceofserialluck.Thisfindingin-dicatesthat,beyondthefactorsweidentify,theremightbeothersystematicfactorsthatdriveopportunistictiming.Identifyingthesefactorswouldbeaworthwhiletaskforfutureresearch.Anexaminationoftheincidenceoflucky

2400R

TheJournalofFinance󰀆

grants,themethodologyweuseinthispaper,canbeausefultoolinsuchfutureresearchonopportunistictiming.

REFERENCES

Aboody,David,andRonKasznik,2000,CEOstockoptionawardsandthetimingofcorporate

voluntarydisclosures,JournalofAccountingandEconomics29,73–100.

Beasley,MarkS.,1996,Anempiricalanalysisoftherelationbetweentheboardofdirectorcom-positionandfinancialstatementfraud,AccountingReview71,443–455.

Beasley,MarkS.,2000,Fraudulentfinancialreporting:Considerationofindustrytraitsandcor-porategovernancemechanisms,AccountingHorizons14,441–452.

Bebchuk,LucianA.,MartinCremers,andUrsPeyer,2009,TheCEOpayslice,Workingpaper,

JournalofFinancialEconomics(forthcoming).

Bebchuk,LucianA.,andJesseFried,2004,PaywithoutPerformance:TheUnfulfilledPromiseof

ExecutiveCompensation(HarvardUniversityPress,Cambridge,Massachusetts,andLondon,England).

Bebchuk,LucianA.,YanivGrinstein,andUrsPeyer,2006a,LuckyCEOs,NBERWorkingpaper

12771,December.

Bebchuk,LucianA.,YanivGrinstein,andUrsPeyer,2006b,LuckyDirectors,NBERWorking

paper12811,December.

Bernile,Gennaro,andGreggJarrell,2009,Theimpactoftheoptionsbackdatingscandalon

shareholders’wealth,JournalofAccountingandEconomics47,2–26.

Bertrand,Marianne,andSendhilMullainathan,2000,Agentswithandwithoutprincipals,Amer-icanEconomicsReview90,203–208.

Bertrand,Marianne,andSendhilMullainathan,2001,AreCEOsrewardedforluck?Theones

withoutprincipalsare,QuarterlyJournalofEconomics116,901–932.

Bhagat,Sanjai,andBernardBlack,1999,Theuncertainrelationshipbetweenboardcomposition

andfirmperformance,TheBusinessLawyer54,921–963.

Bhagat,Sanjai,andBernardBlack,2002,Thenon-correlationbetweenboardindependenceand

long-termfirmperformance,JournalofCorporationLaw27,231–273.

Bizjak,JohnM.,MichaelL.Lemmon,andRyanJ.Whitby,2009,Optionbackdatingandboard

interlocks,TheReviewofFinancialStudies22,4821–4847.

Brickley,JamesA.,JeffreyL.Coles,andRoryL.Terry,1994,Theboardofdirectorsandthe

enactmentofpoisonpills,JournalofFinancialEconomics35,371–390.

Byrd,John,andKentHickman,1992,Dooutsidedirectorsmonitormanagers?Evidencefrom

tenderofferbids,JournalofFinancialEconomics32,195–221.

Chauvin,KeithW.,andCatherineShenoy,2001,Stockpricedecreasespriortoexecutivestock

optiongrants,JournalofCorporateFinance7,53–76.

Chhaochharia,Vidhi,andYanivGrinstein,2007,Corporategovernanceandfirmvalue:Theimpact

ofthe2002governancerules,JournalofFinance62,17–1825.

Chhaochharia,Vidhi,andYanivGrinstein,2009,Executivecompensationandboardstructure,

JournalofFinance,231–261.

Collins,DanielW.,GuojinGong,andHaidanLi,2005,TheeffectoftheSarbanes-OxleyActonthe

timingmanipulationofCEOstockoptionawards,Workingpaper,UniversityofIowa.

Collins,DanielW.,GuojinGong,andHaidanLi,2009,Corporategovernanceandbackdatingof

executivestockoptions,ContemporaryAccountingResearch26,403–445.

Core,JohnE.,RobertW.Holthausen,andDavidF.Larcker,1999,Corporategovernance,CEO

compensation,andfirmperformance,JournalofFinancialEconomics51,371–406.

Cotter,JamesF.,AnilShivdasani,andMarcZenner,1997,Doindependentdirectorsenhance

targetshareholderwealthduringtenderoffers?JournalofFinancialEconomics43,195–218.Cyert,Richard,Sok-HyonKang,andPraveenKumar,2002,Corporategovernance,takeovers,and

top-managementcompensation:Theoryandevidence,ManagementScience48,453–469.

Dann,LarryY.,DianeDelGuercio,andMeganPartch,2003,Governanceandboardsofdirectors

inclosed-endinvestmentcompanies,JournalofFinancialEconomics69,111–148.

LuckyCEOsandLuckyDirectors2401

Dechow,PatriciaM.,RichardG.Sloan,andAmyP.Sweeny,1996,Causesandconsequences

ofearningsmanipulation:AnanalysisoffirmssubjecttoenforcementactionsbytheSEC,ContemporaryAccountingResearch13,1–21.

Fleischer,Victor,2007,Optionbackdating,taxshelters,andcorporateculture,ViginiaTaxReview

26,1031–1038.

Gillette,AnnB.,ThomasH.Noe,andMichaelJ.Rebello,2003,Corporateboardcomposition,

protocols,andvotingbehavior:Experimentalevidence,JournalofFinance58,1997–2033.Harford,Jarrad,andKaiLi,2007,DecouplingCEOwealthandfirmperformance:Thecaseof

acquiringCEOs,JournalofFinance62,917–949.

Heron,RandallA.,andErikLie,2007,Doesbackdatingexplainthestockpricepatternaround

executivestockoptiongrants?JournalofFinancialEconomics83,271–295.

Heron,RandallA.,andErikLie,2009,Whatfractionofstockoptiongrantstotopexecutiveshave

beenbackdatedormanipulated?ManagementScience55,513–525.

Lie,Erik,2005,OnthetimingofCEOstockoptionawards,ManagementScience51,802–812.Murphy,KevinJ.,2003,Stock-basedpayinneweconomyfirms,JournalofAccountingandEco-nomics34,129–147.

Narayanan,M.P.,CindyA.Schipani,andH.NejatSeyhun,2007,Theeconomicimpactofback-datingofexecutivestockoptions,UniversityofMichiganLawReview105,1598–11.

Narayanan,M.P.,andH.NejatSeyhun,2006,EffectofSarbanes-Oxleyontheinfluencingof

executivecompensation.Workingpaper,UniversityofMichigan.

Narayanan,M.P.,andH.NejatSeyhun,2008,Thedatinggame:Domanagersdesignateop-tiongrantdatestoincreasetheircompensation?,ReviewofFinancialStudies21,1907–1955.

Shivdasani,Anil,1993,Boardcomposition,ownershipstructureandhostiletakeovers,Journalof

AccountingandEconomics16,167–198.

Walker,David,2006,Someobservationsonthestockoptionsbackdatingscandalof2006,Working

paperNo.06–31,BostonUniversitySchoolofLaw.

Weisbach,Michael,1987,OutsidedirectorsandCEOturnover,JournalofFinancialEconomics

20,431–460.

Yermack,David,1997,Goodtiming:CEOstockoptionawardsandcompanynewsannouncements,

JournalofFinance52,449–476.