ESSAYSINEXECUTIVEINCENTIVESAND LABORMARKETS By DamienAlexanderBrooks hi hi ADISSERTATION Submittedto MichiganStateUniversity inpartialentoftherequirements forthedegreeof BusinessAdministration-Finance-DoctorofPhilosophy 2015 ABSTRACT ESSAYSINEXECUTIVEINCENTIVESANDLABORMARKETS By DamienAlexanderBrooks Thisdissertationiscomposedofasetofstudiesthatexaminetheincentivesof seniorexecutivesandthematchingbetweenexecutivesandInoneessay,I examinetheincentivesoftmanagerstostrategicallytimetheirdisclosureof insidertrades.Ievidenceconsistentwiththehypothesisthatinsiderschoosethe timingoftheirdisclosurestominimizeanynegativeinformationsignalsthatmaybe conveyedbytheirtradingdecisions.Inasecondessay,Iconsiderthematchingpro- cessbetweenarealassetsanditsmanagerialhumancapital.Ithatcertain typesofCEOeducationalarehighlycorrelatedwithandindustrychar- acteristics.Thisevidenceindicatesthateducationalbackgroundscaptureimportant managerialcharacteristicsthathaveatimpactontheoptimalmatching processbetweenCEOsandInaessay,Iexaminetheroleofaown- ershipstructureonthetypesofseniormanagersthattheisabletoattract.I thatthepresenceofcertaintypesofblockownersleadstoasubstantiveconstraint inthesetofcandidatesthatarewillingtoconsiderleadingathussuggestinga substantialhumancapitalcosttocertainownershipstructures. DedicatedtomywifeAli,andtomyparentsPhilipandKathiBrooks iii ACKNOWLEDGEMENTS IamgratefultoBooth(CommitteeMember),WilliamGrieser,Charles Hadlock(Chair),JoergPicard,AndreiSimonov(CommitteeMember),andMiriam Schwartz-Ziv(CommitteeMember)forusefulcommentsandguidance. iv TABLEOFCONTENTS LISTOFTABLES :::::::::::::::::::::::::::::::: vii CHAPTERIStrategicRevelationofInsiderTrades ........ 1 1.1Introduction........................... 1 1.2LiteratureReviewandMotivation............... 3 1.3TheSample............................ 8 1.4StylizedFacts........................... 10 1.5EmpiricalAnalysis........................ 12 1.5.1Hypotheses....................... 12 1.5.2ReportingSpread................... 13 1.5.3OpportunityforMultipleTrades........... 14 1.5.4yofInsiderTrading............ 16 1.5.5Insidery.................. 20 1.5.6InformativenessofInsiderTrading.......... 25 1.6Conclusions............................ 28 CHAPTERIIMatchingCEOEducationstoFirmCharacteristics 29 2.1Introduction........................... 29 2.2LiteratureReviewandMotivation............... 30 2.2.1HiringMarketObservations............. 30 2.2.2ofBackgroundonPerformance........ 31 2.2.3ManagerialStyle.................... 33 2.3StudyMotivation........................ 35 2.4TheSample............................ 36 2.4.1Compustat....................... 36 2.4.2BoardexCompanyRecords.............. 36 2.4.3BoardexPersonRecords............... 37 2.4.4BoardexEducationRecords............. 38 2.4.5CombiningtheData.................. 39 2.4.6EducationQuality................... 40 2.5EmpiricalAnalysis........................ 42 2.5.1RegressionFramework................ 43 2.5.2DegreeType...................... 43 2.5.3SchoolCharacteristics................. 52 2.5.4SchoolQuality..................... 58 2.6Conclusions............................ 65 CHAPTERIIITheRoleofOwnershipinExecutiveLaborMarkets 68 3.1Introduction........................... 68 3.2LiteratureReviewandMotivation............... 69 3.2.1CEOBackgrounds................... 69 v 3.2.2ExecutiveLaborMarket............... 71 3.3TheSample............................ 73 3.3.1FactsetActiveCompanies.............. 73 3.3.2FactsetInactiveCompanyIndividuals........ 74 3.3.3FactsetInactiveCompanyInstitutions........ 74 3.3.4FactsetInactiveCompanyFunds........... 75 3.3.5CRSP/CompustatMerged.............. 75 3.3.6BoardexCompanyRecords.............. 76 3.3.7BoardexPersonRecords............... 76 3.3.8BoardexEducationRecords............. 76 3.3.9CombiningtheData.................. 77 3.3.10Variablenitions.................. 78 3.4EmpiricalAnalysis........................ 81 3.4.1RegressionFramework................ 83 3.4.2DegreeType...................... 84 3.4.3SchoolCharacteristics................. 91 3.4.4SchoolQuality..................... 97 3.5Conclusions............................ 103 BIBLIOGRAPHY :::::::::::::::::::::::::::::::: 104 vi LISTOFTABLES Table1.1SummaryStatistics......................... 10 Table1.2SummaryStatisticsforReporting................. 12 Table1.3SummaryStatisticsforReportingbyNetTransactionDirection. 13 Table1.4MultipleTradesasaDeterminantofReportingDelay...... 15 Table1.5SummaryStatisticsforCARAfterTransactionDate....... 17 Table1.6AbsoluteReportingDelayasDeterminedbyPost-TradingCAR. 18 Table1.7NetReportingDelayasDeterminedbyPost-TradingCAR.... 19 Table1.8HoldingPeriodCARbyJob.................... 21 Table1.9AbsoluteReportingDelayasDeterminedbyHoldingPeriodCAR 23 Table1.10NetReportingDelayasDeterminedbyHoldingPeriodCAR.. 24 Table1.11SummaryStatisticsforCARAfterReportDate.......... 25 Table1.12AbsoluteReportingDelayasDeterminedbyPost-DisclosureCAR 26 Table1.13NetReportingDelayasDeterminedbyPost-DisclosureCAR.. 27 Table2.1SummaryStatistics......................... 44 Table2.2DeterminantsoftheCEOHoldingaLawDegree......... 45 Table2.3DeterminantsoftheCEOHoldingaDoctoralDegree....... 47 Table2.4DeterminantsoftheCEOHoldinganMBADegree........ 48 Table2.5DeterminantsoftheCEOHoldingaMaster'sDegree....... 50 Table2.6DeterminantsoftheCEOHoldingaGraduateDegree...... 51 Table2.7DeterminantsoftheCEOGraduatingfromaPrivateSchool... 53 Table2.8DeterminantsoftheCEOGraduatingfromaRichSchool.... 55 Table2.9DeterminantsoftheCEOGraduatingfromaHighClassSchool. 56 Table2.10DeterminantsoftheCEOGraduatingfromaReligiousSchool.. 57 Table2.11DeterminantsoftheCEOGraduatingfromaForeignSchool.. 59 Table2.12DeterminantsoftheCEOGraduatingfromaPrestigiousSchool. 60 vii Table2.13DeterminantsoftheCEOGraduatingfromaTopSchool.... 61 Table2.14DeterminantsoftheCEOHoldinganMBAfromanEliteProgram 63 Table2.15DeterminantsoftheCEOHoldinganMBAfromaTopProgram 64 Table2.16DeterminantsoftheCEOGraduatingfromanEliteSchool... 66 Table3.1SummaryStatistics......................... 82 Table3.2DeterminantsoftheCEOHoldingaLawDegree......... 85 Table3.3DeterminantsoftheCEOHoldinganMBADegree........ 86 Table3.4DeterminantsoftheCEOHoldingaMaster'sDegree....... 87 Table3.5DeterminantsoftheCEOHoldingaDoctoralDegree....... 88 Table3.6DeterminantsoftheCEOHoldingaGraduateDegree...... 90 Table3.7DeterminantsoftheCEOGraduatingfromaPrivateSchool... 91 Table3.8DeterminantsoftheCEOGraduatingfromaRichSchool.... 92 Table3.9DeterminantsoftheCEOGraduatingfromaHighClassSchool. 93 Table3.10DeterminantsoftheCEOGraduatingfromaReligiousSchool.. 94 Table3.11DeterminantsoftheCEOGraduatingfromaForeignSchool.. 96 Table3.12DeterminantsoftheCEOGraduatingfromaPrestigiousSchool. 98 Table3.13DeterminantsoftheCEOGraduatingfromaTopSchool.... 99 Table3.14DeterminantsoftheCEOHoldinganMBAfromanEliteProgram 100 Table3.15DeterminantsoftheCEOHoldinganMBAfromaTopProgram 101 Table3.16DeterminantsoftheCEOGraduatingfromanEliteSchool... 102 viii CHAPTERI StrategicRevelationofInsiderTrades 1.1 Introduction UnderSection16oftheSecuritiesExchangeActof1934(henceforthknownas \theAct"),insiders(whichincludedirectors,andbownersof 10%ormoreofaclassofequity)mustreporttheirtradingactivitiesinthes stocktotheSecuritiesandExchangeCommission(SEC).Inthecaseofrestricted shares,insidersmustaForm144forauthorizationpriortotrading.This makespublictheintendedtradingactivity,astheinsidermustreporttheexpected directionofthetrade(buyorsell)andthenumberofshares,whiletheassociated authorizationexpiresafter90days.However,thisistheexception. Allinsidertrades(includingtheactualtransactioninvolvingtheauthorizedre- strictedshares)arereportedafterthefact.Insidersmustthreeforms:theForm 3,whichisduewithin10daysoftheinsiderbecomingareportingpersonandre- portsanypre-existingtransactions;aForm4,whichisduebytheendofthesecond businessdayafteralmostalltypesoftransactions;andaForm5,whichisanannual reportoftheperson'sholdingsintheTogether,theseformscanrecreatean insider'scompletetradinghistoryforthestock. Therehasbeennoscholarshiporinquirytomyknowledgethatexploresthe cacyoftheseregulations.Section21oftheActenumeratespenaltiesforviolationsof spsectionsandlaws.Whileinsidertradingisamongthese,Section16compli- anceisnotlistedormentioned.Thus,itfallsunderthepurviewofthecivilremedies inSection21(a)(3).TheSECisabletopursueviolationsbothoftheActitself 1 andproceduraloradministrativeviolations(e.g.,failuretocomplywithdocument requestsandnoshowsforSEChearings)incourt.Thepenaltiesfora\naturalper- son"beginat$5,000foratierrisingto$50,000forthesecondtierand $100,000forthethirdtier.Inallcases,thepenaltyisthegreaterofthegivenamount orthegrossamountofpecuniarygainaccruingfromtheviolation. TheextenttowhichtheSECchoosestoorcanenforcethisstatuteinthecontextof Section16isunclear.Infact,theonlyobviouspenaltyforSection16non-compliance isthatthemustdisclosethedelinquencyinits10-K.Priorto2003,a10-K acknowledgingareportingperson'sdelinquentreportingwasgivenaspecialdesigna- tion:10-K405.However,thiswasdropped,asappliedthelabelinconsistently. Therelativeobscurityofthispunishment,combinedwithitsrelianceonthe (overwhichthereportingpersonlikelyholdstasthemonitorof resortsuggeststhatitwouldlikelybeantdeterrent.Certainly,one mightarguethatrepeatedviolationsofSection16mightprovideprobablecausefor aninsidertradinginvestigation,butthelevelatwhichthisbecomesaconcernis unclear. Thisisespeciallyaconcerninthelastdecade.Section403oftheSarbanes- OxleyActof2002requiredthattheSECmovetoelectronicandpostingof insidertradingdocuments.InRelease33-8230,theSECissuedarulethatall Forms3,4,and5mustbeviaEDGAR.Noothermeansweretobeaccepted, andnohardshipexemptionwastobeavailable.ThisruletookonJune30, 2003.Amajorbofthetransitionwasthatthetimebetweentheand thepublicdisseminationofareportwasallbuteliminated.EDGARsubmissions aregenerallyavailabletoinvestorslessthananhouraftertheirsubmission.One canimmediatelyseethegain:submissionofmagnetictapeswouldrequire timeforshipping,aswellasrequiretimeforSECtoreviewandprocessthe data,whileelectronicsubmissionstoEDGARaretransmittedmuchfasterandcan 2 bequicklyreviewedalgorithmically.Thisimpliesthatanydelaysinreportingare verylikelytobeintentional,asmuchofthenoiseintheprocesshasbeeneliminated. Section16alsoprohibitstwolessinnocuousactivities.First,itestablishesthe regulatorybasisfordisgorgementofshort-runonthepresumptionofinside information.Shareholders,andtheitself,areeligibletosueformadebya reportingpersonderivingfromabuyandsalecombinationwithinasixmonthperiod. Thisimpliesthattherelevantholdingperiodforaninsiderissixmonths.Second,it prohibitsshortingbyinsiders.Thismeansthatwhileinsiderscanavoidlosses,they cannotcapturefromdownturnsviatradingactivity.Theseaspectswillhelp informmyanalysis. 1.2 LiteratureReviewandMotivation Tomyknowledge,noresearchhasexaminedthelatenessoftradingdisclosures. Abodyofresearchhasbuiltontheimplicationsofthesethough.Huddart, Hughes,andLevine(2001)incorporatetheForm4requirementintotheequilibrium fromKyle(1985).Theinformedtraderissubstitutedforaninsiderwhonowmust reporttheirtradesimmediatelyafterthemarketclears.Thisresultsinaccelerated pricediscoveryandlowerfortheinsider.Thisframeworkprovidesalens intotwoofmandatoryreporting:itactstosimultaneouslyincreasemarket anddecreaseincentivesforinsidertrading.However,intheequilibrium, insidersareabletoprotectsomeoftheirbyadoptingadissimulationstrategy andaddinganoisy,randomcomponenttotheirtradingstrategy.Thishelpstoensure thatthepricediscoveryprocessremainsacceleratedbypreservingtheinsider'sability toreap Ofcourse,theinsiderisassumedtobeHowever,thisisnar- rowlytobestrictlyinthecontextofhisorherinsidertradingactivity.Sup- posethatonestartswiththeHuddart,Hughes,andLevine(2001)model.Suppose 3 thatthisinsiderisnotjustareportingperson,butacorporateForthesake ofexposition,IwillassumetheinsiderisaCEO,butthediscussionshouldbegener- alizabletoalowerrankedexecutive(albeitnotasstrongly).Now,endowthisCEO withawage, w ,thatispaideachperiodafterthemarketclears.Further,suppose thatthereexistssomeprobabilitythatheorshewillbeterminated, ˝ t ,whichisa functionofthestockprice: ˝ t ( p t )= ˝ t 1 + p t + t Here, representstheportionoftheirperformanceevaluationthatisorthogonalto thestockprice(e.g.,boardpreferencesformanagerialstyle).Onecanseethatthis wouldresultinatobjectivefunction.Themanager'ssingleperiod functionwouldnowbe: ˇ t = x t ( v p t ) w ( ˝ t ˝ t 1 ) = x t ( v p t ) wp t w ( t t 1 ) Noticethat wp t istheexpectedlossinsalaryfromtheCEO'stradingactivity, while w ( t t 1 )istheexpectedlossinsalaryfromothersources(whichwould beirrelevanttotheCEO'stradingstrategy).ThiswouldlendtheCEOt incentivesthantheyhadpreviously.Thatis,allelseequal,theCEOhasanincentive tobuyandplaceupwardpressureontheprice.So,onecanseehowpuretrading mightnotbetheonlyobjectiveforacorporateintheHuddart,Hughes,and Levine(2001)framework.Rather,tradingcouldalsoserveasaformofemployment insurancefortheexecutive.ConsiderthatiftheCEOisabletoacquiresomeform ofinsurancethroughtrading,itwouldlikelyaltertheirallocation. Whiletherehasbeennotheoreticalworkinthisarea,empiricalstudieshavees- tablishedthestockpriceasadriverforCEOturnover.Weisbach(1988)that 4 poorstockperformancecanbeusedasapredictorofCEOresignations,especially inthecaseofanoutsider-dominatedboard.Likewise,positivereturnscorrespondto resignationannouncements.Dikolli,Mayew,andNanda(2012)alsoanassoci- ationbetweenperformanceandCEOsurvival.Whilethisrelationshipweakens overthecourseoftheCEO'stenure,itcanbequitestrongearlyintheexecutive's timewiththeAnalternativechannelisdiscussedbyMartinandMcConnell (1991),whosthatCEOsoftendertakeovertargetsaresubjecttoahigh turnoverrateaftercompletionofthetakeover,andthatthesetargetswerecon- sistentunderperformers.So,poorstockperformancecanbothleadtheboardtooust theCEOorinduceatakeoverwhichremovestheCEO.Inbothcases,theCEOhas anincentivetoraisethestockpriceinordertoprotecthisorherjob. Ofcourse,thereisnoroominthemodelforaconsistentinanyversion oftheKylemodel.Atsomepoint,theinsidermustEvenwithmyproposed extension,itseemsunlikelythatdrivingthepricetlyabovetruevaluewould beanidealstrategy.Scholarshiphasshown,however,thatsuchacanactuallybe value-increasing.KhannaandSonti(2004)showthat,inasituationinwhichmultiple informedinvestorsareabletosequentiallyplacetheirorders,itisoptimalforthethird insidertoherdundercertainconditions.Sp,thesestraders'signalsneedto benon-worthlessandtheymustpossessinventorylevelsaboveaderivedminimum inordertosupportthisherdingequilibrium.Mostimportantly,theyarguethatthis equilibriumcanleadtovaluecreation,asthehigherstockpricescanalleviate constraintsontheandallowfornewinvestments. Giventhatreportingpersonstendtobeexactlythosetraderswithrelativelylarge inventoriesofsharesinthe(bownersbyandtypically throughcompensationpackages),itiseasytoapplythismodeltomycontext.In- corporatingthejobherewouldbeevenmoredirect:herdingwouldnotonly produceamorevaluablecompany,butitwouldalsodecreasetheirlikelihoodofter- 5 mination.Hence,evenifthetradersarenotparticularlytheinsidermight stillontheirportfolio(consistingofboththetradingandtheexpected wageInordertotakeadvantageoftheherding,though,insiderswouldneed somemannerinwhichtokeepconstantpositivepressureonthepricesothatthe bubbleforms.Onewaymightbetoincorporatetheabilityoftheinsidertolate ConsidertheofallowingalateintheHuddart,Hughes,andLevine (2001)context.Thatis,ratherthantheinsider'stradesbeingguaranteedtobe reported,theCEOhastheoptiontostrategicallydelaybyoneperiod(sothey wouldbeincorporatedafterthenextauction,ratherthantheoneinwhichthetrades tookplace).Forexample,anytradesfromtime t =1wouldnotbeincorporateduntil aftertradingconcludesat t =2.Onerealizesthatthismakesthetwoperiodmodel problematic,asitwouldalmostsurelybeoptimalforinsiderstodelayreportingall trades.Inthissituation,themodelcollapsestothatofKyle(1985).Amodelwithat leastthreeperiodswouldbeneededinordertomodelthisappropriately.However,I canmakeconjecturesagainbasedonthesingleperiodmodel.RecallthattheCEO isoptimizingoverreapingtradingandminimizingthelossinexpectedwage. Withasingularfocusontradingthedecisionistrivial:obviously,theCEO shoulddelayallreporting.However,ifthelossinexpectedwageistlyhigh, onecouldenvisiontheCEOdisclosingontimeinordertoapplysomepositiveprice pressure. However,theHuddart,Hughes,andLevine(2001)modelisbasedonacrucial assumption:theinsidermustimmediatelyandconsistentlyreporttheirtradesprior toanyfurthertrading.WhilethisassumptioniscertainlyinthespiritofSection 16,androbustatlargetimescales,itcanbedistortionaryinthefaceofsomeshort- termobservations.Properregulationofinsidertradingseemstobeimportantfor capitalmarket.BhattacharyaandDaouk(2002)thatinsidertrading lawshavelittleonface.However,thecostofequityfallstly 6 afteranation'sprosecutionforinsidertrading.Thissuggeststhatthemarket isreluctanttoinvestinequitywithoutsomesignthatthegovernmentiswilling enforceitslawsrestrictinginsiders'activities.Thisaddssigntomystudy,as indicationsthattheSECisabdicatingitsenforcementrolemightreversethis OnechannelwherethishasbeenobservedisthroughcorporateAsignif- icantbodyofresearchhasdocumentedthepropensityoftotheirForms 10-Kand10-Qlate.Alford,Jones,andZmijewski(1994)werethetopresent suchevidence,astheyfoundthat20%of10-Kswereaftertheduedate.Inthis context,theSECprovidesforanautomatic15dayextensionviaaForm12b-25 However,theauthorsfoundthatlessthanonethirdoftheselateshadac- tuallyfortheextension.Theauthorsevidencethatmostmissingthe deadlinearesimplyunabletomeetthedeadlineforvariousreasons.This ist,astheauthorspointoutthat,atleasttothatpoint,theSEC wasincapableoflevyinganydirectpenaltiesonRather,penalties rangedfromasuspensionofashelfregistrationtoasuspensionoftradingin thestock. ThesecanbedirecostsforaastheSECcouldpotentiallyallbuteliminate theabilitytoraiseexternalfunds.However,donotseemtogooutoftheir waytoavoidthesepenalties.Cao,Calderon,Chandra,andWang(2010)revisitthis resultbyexaminingthereasonsgivenforthedelayontheForm12b-25(extension) Theycreatemeasuresofaccountingandinformationsystemcontrolquality andusethesetopredictthemarket'sreactiontotheextension.Thesearefoundto bestrongdrivers,whiledistressfallsaway.Thissuggeststhatperhaps couldhaveavoidedthesebutduetoinstitutionalchoicesfailedtodoso. Ratherthanagroupofinsiders,though,Iamtypicallyconsideringthecaseofa singleinsidersignalingtheentireexternalmarket.Inthismanner,Imustconsider thepsychologicalofthedisclosuresonmarketparticipants.ShefrinandStat- 7 man(1985)extendKahnemanandTversky's(1979)prospecttheorytoelaboratea frameworkwhichisconsistentwiththebehaviorofinvestorsinsellingwinnerstoo soonandholdingloserstoolong.Atpartofthisextensioncomesinthe formofregretavoidanceandself-control(orlackthereof):investorslacktheself- controltoforcethemselvestorealizealoss,insteadpreferringtohangonandhopeto eventuallyturnaByanalogy,onecanapplythistothereportingdecision:I couldexpecttoseeinsidersbeingquicktoreportagoodtradingdecision(\winner") andslowtoreportabadtradingdecision(\loser"). Onecouldsupposethatinsidersmightevenfromamorepublicversion ofthesameshameavoidanceastheinvestorsShefrinandStatman(1985)discuss. Whereasthosetradersneededonlyconsiderpotentialrecriminationsfromtheirown ego,reportingpersonsfacepublicscrutiny.Now,allegosinthemarketarebrought tobearontheirdecisions.Onecouldenvisionthattheseothermarketparticipants employsomeformofmentalaccountinginwhichpoortradingreturnsfromsophisti- catedtradersgeneratesapseudo-wealthThatis,theuninformedfeelwealthier whentheinformedtraders'matchorfallbelowtheirown.Withthisinmind, whethertheinsiderdecidestobuyorsell,theymightbehesitanttorevealthatchoice foraslongasitcanpossiblybedelayedwithoutprovokingregulatoryaction. 1.3 TheSample IusedataoninsiderstocktransactionsfromtheThomsonReutersInsiderFiling DataFeed.Sp,Ifocusmyattentiononthenon-derivativetransactiontable (Table1).ThiscontainsdataonForm3,4,and5entriesrangingfromJanuary1, 1986,toMarch25,2014. Thedataexistsattheperson-company-tradelevel.Eachtradeisassignedacode (reportedbytheinsideronthesummarizingitsnature.Iselectthosecodesthat thatareinherentlydiscretionary(P,S,V,I,andJ).So,thingssuchasgiftsorrequired 8 distributionsareexcluded,whileopenmarketpurchasesorsalesareincluded.My assumptionisthatwhilethesetransactionsmightimpacttheinsider'sdiscretionary tradingstrategy,theyshouldnotthereportingoftheinsider'sdiscretionary trades. Sincereportingrequirementsareatthedailylevel,thereislittlethatcanbedone belowthatfrequency(e.g.,onecanonlyattempttoguessthetimeatwhichthetrade tookplaceduringthereportedday).So,Iaggregatethedataattheperson-company- daylevel.Iretainonlythoseobservationsinwhichalltradeswereabletobev orcleansedbyThomson(codedasR,H,andL).Next,Idropanyperson-company- daysinwhichaSECreceiptdateisreportedaspriortothetransactiondate.These twostepsshouldremovethemajorityofanyerroneousdatafromthesample. Next,Iexcludeanyobservationswithtradesthatwerereportedonaformbesides Form4(e.g.,Forms3and5).Thisallowsforthemostcomparabilityofthedata: thedeadlineforaForm5wouldbecantlytfromthatofaForm4, anditisnotnecessarilyclearthatalltransactionsonaForm5shouldhavebeen reportedthere(e.g.,selecttransactionsareexemptfromForm4requirements,but delinquentForm4transactionsarealsosupposedtobereportedonaForm5).Given theEDGARstandardizationIdiscardalltradingpriortotheJune30,2003, edate.Thisallowsmetostrictlyfocusontheinsider'spropensitytodelay whatshouldbeafairlyinstantaneousprocess. Finally,ImergetheinsiderdatawithdailypricingdatafromtheCenterfor ResearchinSecurityPrices(CRSP)inordertoinsomemissingpricedata.This implicitlyassumesthattheinsiderboughtorsoldattheclosingprice,whichshould biasagainstanyresults.Finally,Idropanyobservationswhichwerewitha negativepricefromCRSP.Itisnotclearthattheestimatedpricefromtheclosing bid-askspreadisagoodestimateforthepriceatwhichtheinsidertraded. Finally,Iremoveanyinsider-company-daycombinationsthatgeneratemorethan 9 oneThatis,Ikeeponlythosereportingallactivityonthesameform.This isdonetoensurethataccidentalmisreportingdoesnotmyresults.This shouldbiasagainstany Thisprocessyieldsasampleof928,302insider-company-daycombinations.These rangefromJuly1,2003,toMarch25,2014.Further,thesampleincludesobservations from100,005distinctinsidersand11,829uniqueSummarystatisticsforseveral measuresarepresentedinTable1. Table1.1SummaryStatistics VariableMeanStd.Dev.NObserved Sizein: Shares-36337.1842777063.122928,302 Value-1073949.82959156797.159928,302 Buyby: Shares0.3580.479332,533 Value0.3540.478328,915 Jobs: CEO0.1290.335119,634 CFO0.0630.24358.474 CIO0.0020.0411,543 COO0.0250.15723,520 CTO0.0130.11311,935 President0.0950.29388,121 EVP0.0660.24861,248 GC0.0260.1624,483 Controller0.0110.10610,583 1.4 StylizedFacts Inmysample,Iseethatthetwobusinessdaydeadlinecanresultinarangeof anywherefromtwotoecalendardays.Forexample,ifaninsiderplacesatrade ontheFridayprecedingMartinLutherKing,Jr.Day(whichisalwaysrecognizedas thethirdMondayinJanuary),theirdeadlinewouldbethatWednesday(no areallowedonSECholidays),whileatradeplacedontheTuesdayimmediatelyafter MartinLutherKing,Jr.DaywouldhaveareportingdeadlineonthatThursday. 10 Whiletheamountoftradingtimeislargelyunchanged,onecouldenvisionthatsome short-livedinformationmightneverbedisseminatedtothemarket.Alternatively, theadditionalcalendartimecouldallowmoretimeforotherinsiderstodiscover andsubsequentlytradeontheinformationbeforeitisintheprice.My assumptionisthatthiscouldleadtosomeoverheating,asthetradingpressuresthe pricetomove,andthenthesignalfromtheadds\new"pressureinthesame direction. Moreimportantly,theHuddart,Hughes,andLevine(2001)frameworkimplicitly assumesperfectenforcement.Thatis,theformsarealwayscorrectlyandon- time,allowingthemarketmakertoincorporatetheirinformationcontentintoprices beforethenextauction.Mysampleshowsthatthisisnotnecessarilythecase:ap- proximately8%ofthesampleisatleastonedaylate,withhalfofthoseatleast 14dayslate,and2%ofthesamplebeingatleast67dayslate.Amongtheexclu- sionswere2,496,679Form4sand5sthatreportedholdingswithnocorresponding tradereported.Thisrepresentedapproximately20.11%ofthetotaldatasetpriorto cleaning.Thomson1,395,418recordsforirregularities,totalingapproximately 10.71%ofthetotaldata.Thisimpliesthatapproximately29.41%oftheseSEC ingsaremissinginformationorincorrecttosomedegree.Onecouldargue,ofcourse, thatthisisadataaggregationproblem.Thatis,thestudy'sahistoricalperspective removesanysoftinformationitmighthavecarriedthatwouldhavesubstitutedfor thehardinformationitshouldprovide.ThisisnotanexplanationIcanruleout. Myfocus,then,isonthetimelinessofandconsistencyinTable 2presentssummarystatisticsforthenumberofanddatesassociated witheachinsider-company-daypriortothecullingofmultipleinsider-company- days.Mysamplecontains41,107insider-company-daycombinationswithmorethan oneassociatedThiscorrespondstoapproximately4.24%ofthesample.I furtherobserve23,547insider-company-daycombinations,approximately2.71%of 11 thesample,withmorethanonedate.Giventhenatureofthetransactionsin thisstudy,andthenatureoftheinvestorsinthesample,eventheserelativelymodest numbersseemtoohightobeexplainedbyhumanerror. Table1.2SummaryStatisticsforReporting VariableMeanStd.Dev.Min.Max.N Sizeofsalesby: Shares-34767.6662813310.394-1523960704519750112969,409 Value-1096116.99559423898.711-2598352896026511536128969,409 Filings1.0550.409185969,409 FilingDates1.0250.1616969,409 1.5 EmpiricalAnalysis 1.5.1 Hypotheses Asdiscussedabove,thereistheoreticalntobelievethatinvestorsvary theirreportingdependingonwhetherthetransactionisabuyorsell.Inaccordance withthisanalysis,Iclaimthatinsidersshouldnotbeoverlyintheirtrading. RecalltheresultfromBhattacharyaandDaouk(2002):costsofequitydecreasewhen insidertradinglawsareenforced.Thissuggeststhatinvestorshaveanaversionto insidersontheirprivateinformation.Primarily,Iwouldexpectthis tomanifestitselfinthesellcase,astherecouldbesomepresumptionofinsiders intentionallyharmingthecompany(andshareholdersbyextension)inorderto Alongsimilarlines,Iwouldexpecttoseeinsidersdisclosesoonerwhentheyexpect thenewstobegoodfortheInsidersshouldknowwhatthemarketreactionwill betotheirdisclosure.Recallthattheyaretradingonsomeprivateinformation.They aloneknowthisinformation,sotheyshouldbeabletopredictthereactiontotheir trading. 12 1.5.2 ReportingSpread Iconsiderthenumberofdaysittakestodiscloseabuyversusthenumberofdays todiscloseasellattheperson-company-daylevel.Thisdistributionistly skewed,soIusealogtransformationonthenumberofdays.Inordertopreserve thedatathroughthetransformation,Iaddone(six)tothenumberofdaysinthe absolute(net)case,sothattheminimumineachisequaltoone.Summarystatistics forthereportingspreadgiventhiscalculationarereportedinPanelAofTable3.I performasimilaroperationbywinsorizingthetailsofthedistributionofthenumber ofdaysatthe1%level.Thesummarystatisticsforthiscalculationarereportedin PanelBofTable3. Table1.3SummaryStatisticsforReportingbyNetTransactionDirection PanelA:LogTransform VariableMeanMedianStd.Dev.Min.Max.N Overall: Absolute1.1771.0991.04108.055928,302 Net1.7671.6090.82708.055928,302 BuyDays: Absolute1.3851.0991.29308.055332,533 Net1.9281.7921.04708.055332,533 SellDays: Absolute1.0611.0990.84807.997595,769 Net1.6771.6090.65707.998595,769 PanelB:Winsorized VariableMeanMedianStd.Dev.Min.Max.N Overall: Absolute7.487225.70174928,302 Net4.605-125.565-4170928,302 BuyDays: Absolute12.089233.8270174332,533 Net9.129033.661-4170332,533 SellDays: Absolute4.919219.2880174595,769 Net2.081-119.188-4170595,769 Iseethatthereappearstobeatendencytowardreportingbuyslateversussells. Atthemean,theseemstobenearsevendays,oraboutaweek.Obviously, 13 thiscanbeexplainedtosomedegreebyafatuppertail,buteventhemedianshows someinthenetreportingtimecase.Thisrelationshipappearsconsistent acrossthetwoapproaches. 1.5.3 OpportunityforMultipleTrades Ifonebeginswiththeassumptionthatinsiderspossessactionableprivateinfor- mationandtradeonthatinformationwiththeintentionofmakingtrading themostobviousreasononemightconsiderforexplainingdelaysinreportingisthat insidersintendtomakeanothertradeontheinformationwhileitremains Hence,Iwouldexpecttoseelongerdelaysincaseswhereasecondtradeoccurs. However,onealsowonderswhetherinsidersmightactuallyengageinthisstrategy. ConsiderthattheSECmightbemorelikelytoinvestigatepersonswhoappeartobe manipulatingthedelayperiodforWhilethereisevidencethattheSECislax inenforcingreportingrequirements,oneshouldnotcarelesslyoverreach.Iestimate thefollowingregressioninordertotestforthisbehavior: Delay t = 0 + 1 ST + 2 SD + 3 ST SD + v + u Here,STisanindicatorforasecondtradebeforetheandSDisanindicator fortheperson'snexttradebeinginthesamedirection(buyorsell). v isavectorof Table4reportstheresultsofthisestimation. AfewthingsareapparentfromtheresultsinTable4.First,theredoesseemto beadelayinreportingiftheinsiderplansonmakingasecondtrade.However,the estimateddelayisreducedbyapproximatelytwo-thirdsinthecasethattheinsider makesanothertradeinthesamedirection(e.g.,abuyfollowingabuy).Thissuggests thatinsidersmightrealizethatcontinuingtotradeinthesamedirectionafteradelay couldraiseSECsuspicions. 14 Table1.4MultipleTradesasaDeterminantofReportingDelay PanelA:LogTransform (1)(2)(3)(4)(5)(6) NetdelayNetdelayNetdelayNetdelayNetdelayNetdelay 2nd2.235***2.103***2.419***1.741***1.481***1.876*** (106.95)(83.63)(64.76)(85.04)(57.81)(51.97) Direction-0.142***-0.147***-0.118***-0.0758***-0.0785***-0.0484*** (-76.51)(-45.02)(-55.48)(-39.24)(-21.98)(-22.55) 2nd&Dir-1.244***-0.633***-1.715***-0.967***-0.381***-1.346*** (-58.71)(-24.33)(-45.73)(-46.93)(-14.69)(-37.22) Intercept1.715***1.780***1.661***1.742***1.921***1.649*** (980.87)(597.27)(818.67)(457.15)(216.23)(431.13) N 928302332533595769926438331515594923 R 2 0.2220.2920.1840.4230.5370.405 adj. R 2 0.2220.2920.1840.4150.5210.395 JobNoNoNoYesYesYes FirmNoNoNoYesYesYes YearNoNoNoYesYesYes PanelB:Winsorized (1)(2)(3)(4)(5)(6) NetdelayNetdelayNetdelayNetdelayNetdelayNetdelay 2nd67.84***62.34***76.77***52.28***43.81***58.81*** (72.67)(55.53)(46.26)(59.20)(40.29)(37.34) Direction-3.096***-3.631***-2.317***-1.405***-1.433***-0.704*** (-62.11)(-41.02)(-41.39)(-25.86)(-13.49)(-12.30) 2nd&Dir-43.42***-20.79***-62.53***-33.97***-12.98***-49.38*** (-46.04)(-18.01)(-37.56)(-38.23)(-11.75)(-31.32) Intercept2.913***4.522***1.611***3.906***8.401***1.732*** (60.05)(53.70)(29.32)(33.22)(28.47)(16.30) N 928302332533595769926438331515594923 R 2 0.1540.2290.1170.3720.4780.396 adj. R 2 0.1540.2290.1170.3640.4610.386 JobNoNoNoYesYesYes FirmNoNoNoYesYesYes YearNoNoNoYesYesYes t statisticsinparentheses * p< 0 : 10,** p< 0 : 05,*** p< 0 : 01 15 Itisimportanttonote,however,thatonecannotrelytooheavilyonthemag- nitudesoftheestimatespresentedinthetable.Itseemsalmostcertainthatthese estimatesarebytheextremeuppertailofthedistribution(atsomepoint, itseemslikelythattheinsiderwillmakeanothertrade,evenifitisbasedoninfor- mationunrelatedtothepriortrade). 1.5.4 yofInsiderTrading Thatinsidersdelayinordertotradeagaineasilyreconcileswiththenotionthat theyaretradingtomaximizetradingSincetheirreportingisduesoquickly aftertradeexecution,onemightsurmisethattheychoosetoengageintradeswitha focusonlong-term,durableinformation.Thiswouldimplylowerimmediates thatwouldinevitablyseemlesssuspicioustoSECinvestigators.Iusethisconjecture toformatestofHypothesis1,asIwishtoidentifytheshortrunabnormalproy oftrades.Todoso,IsupplementthedatawithreturndatafromCRSPandFama FrenchfactordatafromtheKennethFrenchDataLibrary.Iestimatethethreeday cumulativeabnormalreturn(CAR)foratradeonday t asthesumofthefollowing threedays'abnormalreturnsasestimatedviathethreefactormodelestablishedin FamaandFrench(1993): CAR t = 3 X i =1 ( \ abnormalreturn t + i ) TheseCARsarealwayscalculatedinlongterms.Recallthatinsiderscannotshort theirequity,soreferringtotheirpersonalCARafteraselliscounterfactual. Rather,IspeakintermsoftheCAR.ThesummarystatisticsfortheseCARs arepresentedinTable5.IalsoreportCARswinsorizedonbothtailsatthe1%level. Recallthattheseestimatesareforathreedayholdingperiod.Thisisobviously nottheholdingperiodthatinsidersfocuson,asdisgorgementrulesprohibitreaping 16 Table1.5SummaryStatisticsforCARAfterTransactionDate VariableMeanMedianStd.Dev.Min.Max.N Overall: CAR.000729-0.0004330.06123-0.860235.6647765,781 WinsorizedCAR0.000358-0.0004330.0519-0.1542630.1765765,781 BuyDays: CAR0.000405-0.0010760.06786-0.860235.6647243,631 WinsorizedCAR-0.000195-0.001080.05443-0.1542630.176487243,631 SellDays: CAR0.00088-0.0001620.0579-0.824432.4237522,150 WinsorizedCAR0.000617-0.0001620.05067-0.1542630.1765522,150 suchshortterm(withrent-seekingshareholdersasanemeansofen- forcement).However,thesewouldbeindicativeoftheabnormalreturnsthatinsiders haveearnedpriortoorduringtheirreportingdecisionperiod.Inparticular,removing outliersleadstonegativemeanCARsondaysthatinsidersbuy,whilepositivemean CARsremainondaysthatinsiderssell.Now,Iwishtodeterminetheexplanatory powertheseCARsholdforthereportingspread.Thisismodeledbytheregression model: ReportingSpread i;j;t = 0 + 1 CAR j;t + v + u Here, i idenaparticularinsider, j representsaparticularcompany,and t aparticulartradingdate.Notethat CAR j;t isacompany-datelevelvariable,asit isinvarianttothechoiceofinsider. v representsavectorofcontrolvariables.I employalongallthreevectors:ajobthatproxiesfortheperson's characteristics(e.g.,CEO,CFO,COO,etc.),acompanyandayearThe resultsofthisestimationarepresentedinTable6fortheabsolutereportingspread measureandTable7forthenetreportingspreadmeasures. Table6showslittletonorelationshipinthethreespIaneg- ativerelationshipinthenetsellcaseofPanelB,butthisrelationshipisnon-existent inthelogspHowever,theinclusionofyieldsat positiverelationship.Iseethisespeciallyinthenetbuycase(Equation(5)),whileit 17 Table1.6AbsoluteReportingDelayasDeterminedbyPost-TradingCAR PanelA:LogTransform (1)(2)(3)(4)(5)(6) AbsoluteAbsoluteAbsoluteAbsoluteAbsoluteAbsolute CAR-0.001300.0340-0.008790.0429***0.101***-0.00523 (-0.07)(0.93)(-0.46)(2.64)(3.28)(-0.31) Intercept1.115***1.345***1.008***1.161***1.481***1.025*** (1020.24)(529.56)(962.44)(245.22)(114.06)(226.60) N 765781243631522150765781243631522150 R 2 0.0000.0000.0000.2380.3890.195 adj. R 2 -0.000-0.000-0.0000.2300.3710.183 JobNoNoNoYesYesYes FirmNoNoNoYesYesYes YearNoNoNoYesYesYes PanelB:Winsorized (1)(2)(3)(4)(5)(6) AbsoluteAbsoluteAbsoluteAbsoluteAbsoluteAbsolute CAR-0.6060.483-0.974**0.815**2.014***-0.376 (-1.35)(0.51)(-2.35)(2.11)(2.62)(-1.07) Intercept6.226***11.24***3.885***7.243***14.39***4.368*** (238.92)(171.26)(175.62)(65.08)(41.27)(48.16) N 765806243642522164765806243642522164 R 2 0.0000.0000.0000.2280.3540.220 adj. R 2 0.000-0.0000.0000.2200.3350.209 FirmNoNoNoYesYesYes YearNoNoNoYesYesYes t statisticsinparentheses * p< 0 : 10,** p< 0 : 05,*** p< 0 : 01 18 fadesawayinthenetsellcase(Equation(6)).Thissuggeststhereissometendency todelaybuyreportingastheyofthenetbuytradeincreases.However, onemustacknowledgethatthethreedayCARsbeingusedherearesmall,whichwill leadtoaneconomicallyminiscule(equivalenttoafewminutes). Table1.7NetReportingDelayasDeterminedbyPost-TradingCAR PanelA:LogTransform (1)(2)(3)(4)(5)(6) NetNetNetNetNetNet CAR-0.006410.0319-0.01930.0282**0.0828***-0.0176 (-0.44)(1.09)(-1.33)(2.21)(3.43)(-1.37) Intercept1.719***1.895***1.637***1.751***1.998***1.647*** (2000.94)(926.00)(2053.99)(466.93)(188.44)(473.64) N 765781243631522150765781243631522150 R 2 0.0000.0000.0000.2370.3880.194 adj. R 2 -0.0000.0000.0000.2290.3690.183 JobNoNoNoYesYesYes FirmNoNoNoYesYesYes YearNoNoNoYesYesYes PanelB:Winsorized (1)(2)(3)(4)(5)(6) NetNetNetNetNetNet CAR-0.5920.491-0.962**0.798**2.008***-0.403 (-1.33)(0.53)(-2.33)(2.08)(2.63)(-1.15) Intercept3.352***8.276***1.054***4.340***11.37***1.514*** (129.34)(126.71)(47.93)(39.18)(32.75)(16.77) N 765806243642522164765806243642522164 R 2 0.0000.0000.0000.2280.3540.220 adj. R 2 0.000-0.0000.0000.2200.3340.209 JobNoNoNoYesYesYes FirmNoNoNoYesYesYes YearNoNoNoYesYesYes t statisticsinparentheses * p< 0 : 10,** p< 0 : 05,*** p< 0 : 01 Table7strengthensthisnotionbyfocusingonthenetreportingspread.Recall thatthismeasureremovesthenoiseintroducedbyholidaysandweekendsinthe absolutemeasure.Hence,itappearsthattherelationshipfoundinTable6isnot simplyaspuriousartifactofcalendarnoise.Theoveralltrendrepresentedbythe 19 interceptshowsastrongtendencytodelayreportingbuysversussells,andthe oftheCARseemstobuildonthis.Notice,however,thattheCARwillrepresenta smallportionoftheoverallestimate.Thus,itappearsthattheapparentreturnis notaprimarydriverofthereportingspread. 1.5.5 Insidery Ifinsidersarenotincorporatingtheirreportinginatradingstrategy,one mightwonderwhethertheyareengaginginat-makingstrategyatall.Thatis, doinsidersmakeprontheirinvestments?Insidershavebeendocumentedtoearn excessreturnsontheirtradingstrategies.Givenmyfocus,however,Iwishtorestrict attentiontothoseinsiderswhoarecersintheRecallthattheminimum holdingperiodtobeabletotissixmonths(givendisgorgementrules).Ifinsiders aretradingonreliableprivateinformationandseekingatradingtheyshould haveapositiveholdingperiodCARas: HPCAR i;j;t = 126 X i =1 ( \ abnormalreturn j;t ) Thisisespeciallytrueinthenetbuycase.Considerthataredispropor- tionatelyinvestedinthe.Thismeanstheirtendencyshouldbetodivesttheir portfolioanddiversifytheirrisk.Weseethatinthenumberofbuysversusthe numberofsells.However,thismeansthatthebuysshouldbeespeciallyp Table8showstheresultofasetofregressionsofholdingperiodCARonasetofjob indicators. Equations(1)and(3)representthenetbuycase,whileEquations(2)and(4) displaythenetsellresults.Equations(2)and(4)includeandyear IseethatmostinsidersareabletoreapapositiveHPCARontheirbuysbefore includingThissquareswiththetypicalintheliterature.Oddly, 20 Table1.8HoldingPeriodCARbyJob (1)(2)(3)(4) HPCARHPCARHPCARHPCAR CEO0.0270***0.0147***0.00503-0.00136 (5.35)(5.94)(1.09)(-0.63) CFO0.0238***0.0313***0.00208-0.000388 (4.64)(11.05)(0.46)(-0.17) CIO0.244***0.0292-0.00298-0.0120 (11.43)(1.62)(-0.17)(-1.05) COO0.0217***0.0110**0.00307-0.00947*** (2.64)(2.56)(0.43)(-2.66) CTO-0.006260.0802***-0.00182-0.00687 (-0.37)(13.47)(-0.14)(-1.42) President0.002810.00725**-0.005050.00242 (0.50)(2.48)(-1.02)(0.96) EVP-0.0168***-0.0172***0.000902-0.00121 (-3.95)(-7.17)(0.23)(-0.56) GC0.0208***0.0332***0.00614-0.000474 (2.71)(9.28)(0.92)(-0.16) Controller-0.0000266-0.0253***0.002600.000742 (-0.00)(-5.46)(0.33)(0.18) Intercept-0.00878***0.0225***-0.0167***0.0174*** (-7.12)(26.69)(-3.62)(7.19) N 215078462532215078462532 R 2 0.0010.0010.3420.371 adj. R 2 0.0010.0010.3210.362 FirmNoNoYesYes YearNoNoYesYes t statisticsinparentheses * p< 0 : 10,** p< 0 : 05,*** p< 0 : 01 21 CIOs'tradingseemsespeciallyinthisspHowever,Equation(2) suggeststhattheyarelesssuccessfulattimingtheirsells.Onlycontrollersmanageto sellaheadofanegativeHPCAR(recallthatHPCARisfromtheperspective, soapositiveHPCARfollowingasaleisa\bad"sale).Allotherssellpriortopositive HPCARs.IncludingessentiallywipesoutjobIamleft onlywithatnegativeinterceptinthenetbuycase.COOstimetheir selldecisionsslightlybetterthantheirpeers,buttheystillsellaheadofapositive HPCAR.Thissuggeststhatatleastoneofthethreeessentialassumptionsiswrong: eitherinsidersdonothavereliableprivateinformation,theydonottradeonprivate informationtheypossess(perhapsforfearofprosecution),ortheyhavenon-trading motivations. Onewonderswhetherinsidersvarytheirreportingbasedontheirholdingperiod returns.Duetothelengthofthetimeperiod,thisseemsunlikely.Inanycase,Table 9reportstheresultsofregressingtheabsolutereportingspreadontheHPCAR.Table 10substitutesthenetreportingspreadasthedependentvariable. Table9suggeststhatthereislittlewhenincludingthevectorofef- fects.Beforetheseareincluded,Ihighlytrelationships,withthedelay decreasinginHPCARoverall.So,thebetterthedoesoverthesixmonthholding period,thefasteronewouldexpecttheinsidertoreport.Thisisespeciallytrueinthe netbuycase,butitalsoholdsinthenetsellcase.Again,duetohowHPCARsare calculated,thisimpliesanopposingrelationship.Thatis,\good"buysarereported faster,but\bad"salesarereportedfaster.However,thisrelationshipfailswhenin- corporatingtheasonlyEquation(4)ofPanelBremainst.In thatcase,weseeasimilarbutsmallertothatofEquation(1). Table10producesresultshighlysimilartothatofTable9inthe spHowever,nowIactuallydosomeintheoverall spsuggestingaslightlymorerobustrelationship. 22 Table1.9AbsoluteReportingDelayasDeterminedbyHoldingPeriodCAR PanelA:LogTransform (1)(2)(3)(4)(5)(6) AbsoluteAbsoluteAbsoluteAbsoluteAbsoluteAbsolute HPCAR-0.0302***-0.0486***-0.00417*-0.00416-0.006750.000325 (-12.14)(-9.14)(-1.67)(-1.53)(-1.28)(0.12) Intercept1.118***1.353***1.008***1.157***1.478***1.022*** (957.15)(498.57)(902.41)(235.50)(109.59)(216.74) N 677610215078462532677610215078462532 R 2 0.0000.0000.0000.2370.3880.191 adj. R 2 0.0000.0000.0000.2280.3680.179 JobNoNoNoYesYesYes FirmNoNoNoYesYesYes YearNoNoNoYesYesYes PanelB:Winsorized (1)(2)(3)(4)(5)(6) AbsoluteAbsoluteAbsoluteAbsoluteAbsoluteAbsolute HPCAR-0.859***-1.192***-0.328***-0.127*-0.143-0.0400 (-14.34)(-8.69)(-6.12)(-1.94)(-1.03)(-0.68) Intercept6.265***11.36***3.882***7.026***14.06***4.254*** (224.84)(162.02)(164.39)(61.60)(39.43)(45.14) N 677634215093462541677634215093462541 R 2 0.0000.0000.0000.2240.3490.216 adj. R 2 0.0000.0000.0000.2160.3280.204 JobNoNoNoYesYesYes FirmNoNoNoYesYesYes YearNoNoNoYesYesYes t statisticsinparentheses * p< 0 : 10,** p< 0 : 05,*** p< 0 : 01 23 Table1.10NetReportingDelayasDeterminedbyHoldingPeriodCAR PanelA:LogTransform (1)(2)(3)(4)(5)(6) NetNetNetNetNetNet HPCAR-0.0254***-0.0367***-0.00693***-0.00599***-0.00356-0.00491** (-12.96)(-8.56)(-3.65)(-2.80)(-0.83)(-2.32) Intercept1.720***1.900***1.635***1.746***1.996***1.643*** (1872.66)(868.81)(1922.98)(450.33)(182.13)(453.90) N 677610215078462532677610215078462532 R 2 0.0000.0000.0000.2350.3850.190 adj. R 2 0.0000.0000.0000.2270.3660.178 JobNoNoNoYesYesYes FirmNoNoNoYesYesYes YearNoNoNoYesYesYes PanelB:Winsorized (1)(2)(3)(4)(5)(6) NetNetNetNetNetNet HPCAR-0.851***-1.178***-0.329***-0.137**-0.131-0.0628 (-14.28)(-8.63)(-6.18)(-2.09)(-0.95)(-1.07) Intercept3.388***8.387***1.048***4.123***11.05***1.397*** (122.25)(120.23)(44.64)(36.33)(31.11)(14.89) N 677634215093462541677634215093462541 R 2 0.0000.0000.0000.2240.3480.216 adj. R 2 0.0000.0000.0000.2150.3280.204 JobNoNoNoYesYesYes FirmNoNoNoYesYesYes YearNoNoNoYesYesYes t statisticsinparentheses * p< 0 : 10,** p< 0 : 05,*** p< 0 : 01 24 1.5.6 InformativenessofInsiderTrading Akeyempiricalquestionisalsothatoftheinformativenessofinsiderdisclosures. Thatis,doesthestockpricereacttothenewsofaninsidertransaction?Ipostulate inHypothesis2thattheinsidershouldanticipateanyreactionandalterhisorher disclosurebehavioraccordingly.Now,basedonthestylizedfactsenumeratedin Section5,Ihypothesizethatthereisareaction.Iadoptaverysimilarapproachto theyanalysisabove.IcalculateCARsforthethreedaysfollowingtheSEC receiptdate.Inordertosimplifytheanalysis,Irestrictattentiontoinsider-company- datecombinationsthatcorrespondtoasingleSECdate.Thisrepresentsthe vastmajorityofthesample(approximately95.67%).Thesummarystatisticsfor theseCARsarepresentedinTable11. Table1.11SummaryStatisticsforCARAfterReportDate VariableMeanMedianStd.Dev.Min.Max.N Overall: CAR0.000707-0.00040.06138-0.848923.09614764,085 WinsorizedCAR0.000308-0.00040.05196-0.155450.17609764,085 BuyDays: CAR-0.000237-0.0014410.0659-0.82013.09614243,943 WinsorizedCAR-0.000824-0.0014410.05422-0.155450.17609243,943 SellDays: CAR0.001150.0000590.05914-0.848922.42366520,142 WinsorizedCAR0.0008390.0000590.05086-0.155450.17609520,142 Again,Iseethatwinsorizingreducesthemeanreturn,especiallyinthenetbuy state.Again,theexpectedreturnmakesitseemunlikelythatinsidersaretrading underpuretradingmaximization. Table12presentstheresultsfortheabsolutereportingspread,whileTable13 presentstheresultsforthenetreportingspread. Table12showsaerentialarisingbetweenbuysandsells.Inthiscase,Isee that\bad"buysandsalesarereportedfaster.However,includingthevectorof seemstoreversetherelationship.InPanelA,almostallfallsaway, 25 Table1.12AbsoluteReportingDelayasDeterminedbyPost-DisclosureCAR PanelA:LogTransform (1)(2)(3)(4)(5)(6) AbsoluteAbsoluteAbsoluteAbsoluteAbsoluteAbsolute CAR-0.01800.0954**-0.0418**-0.01480.00513-0.0298* (-0.89)(2.22)(-2.18)(-0.89)(0.16)(-1.73) Intercept1.112***1.338***1.005***1.168***1.485***1.028*** (1023.64)(532.09)(964.43)(247.65)(114.92)(227.61) N 764085243943520142764085243943520142 R 2 0.0000.0000.0000.2350.3840.188 adj. R 2 0.0000.0000.0000.2270.3660.177 JobNoNoNoYesYesYes FirmNoNoNoYesYesYes YearNoNoNoYesYesYes PanelB:Winsorized (1)(2)(3)(4)(5)(6) AbsoluteAbsoluteAbsoluteAbsoluteAbsoluteAbsolute CAR-0.5243.917***-2.192***-0.988**-1.673*-1.293*** (-0.95)(3.04)(-5.25)(-2.47)(-1.88)(-3.65) Intercept6.134***11.05***3.829***7.398***14.46***4.412*** (238.22)(170.61)(175.33)(66.89)(41.68)(48.96) N 764107243952520155764107243952520155 R 2 0.0000.0000.0000.2220.3450.208 adj. R 2 0.0000.0000.0000.2140.3250.197 JobNoNoNoYesYesYes FirmNoNoNoYesYesYes YearNoNoNoYesYesYes t statisticsinparentheses * p< 0 : 10,** p< 0 : 05,*** p< 0 : 01 26 butinthewinsorizedmeanspIseeanegativerelationshipbetweenthe CARandthedelayinbothcases.Thissuggeststhattheinsiderreportsa\good" buyfasterwhencontrollingforthevariouswhiletheystillreport\bad" salesfaster.IftheCARisindeeddeterminedbythequalityoftheinformationbeing disseminated,itwouldappearthatsalesaredisclosedsoonerthelessvalue-negative theinformationtheinsiderisprivyto.Evidenceonbuysseemsabittoomixed. Table1.13NetReportingDelayasDeterminedbyPost-DisclosureCAR PanelA:LogTransform (1)(2)(3)(4)(5)(6) NetNetNetNetNetNet CAR-0.02450.0727**-0.0488***-0.0266**-0.0273-0.0322** (-1.52)(2.06)(-3.34)(-2.05)(-1.04)(-2.46) Intercept1.716***1.889***1.635***1.756***2.001***1.649*** (2015.22)(934.04)(2066.49)(470.87)(189.69)(475.66) N 764085243943520142764085243943520142 R 2 0.0000.0000.0000.2320.3810.185 adj. R 2 0.0000.0000.0000.2240.3630.174 JobNoNoNoYesYesYes FirmNoNoNoYesYesYes YearNoNoNoYesYesYes PanelB:Winsorized (1)(2)(3)(4)(5)(6) NetNetNetNetNetNet CAR-0.5293.821***-2.160***-1.016**-1.773**-1.281*** (-0.96)(2.98)(-5.20)(-2.55)(-2.00)(-3.64) Intercept3.262***8.090***1.000***4.496***11.45***1.556*** (127.38)(125.50)(46.04)(40.85)(33.14)(17.35) N 764107243952520155764107243952520155 R 2 0.0000.0000.0000.2220.3450.208 adj. R 2 0.0000.0000.0000.2140.3250.197 JobNoNoNoYesYesYes FirmNoNoNoYesYesYes YearNoNoNoYesYesYes t statisticsinparentheses * p< 0 : 10,** p< 0 : 05,*** p< 0 : 01 Table13producesessentiallythesameoutcomeasinTable12(withtheobvious adjustmenttotheintercepts).Thus,thedecreaseinbuyreportingtimeinthenet 27 buycaseappearsrobusttotheexclusionofthedeadlineperiod.Thesameholds trueforthesellreportingaccelerationinthenetsellcase.Takentogether,thisadds credencetotheclaimthatinsidersreport\good"buysand\bad"sellsfaster. 1.6 Conclusions Inthispaper,Iestablishacontextinwhichtradersareabletostrategicallymanip- ulatethereportingoftheirtradesinordertoachievenon-tradingmaximizing objectives.Inatheoreticalcontext,Isuggestthatanexecutive'scareerconcernscan alterhisorherexpectedfunctionientlyenoughtoresultinttrad- ingbehavior.Applyingtheseaspectstoexistingmodelsappearsitwouldresultina convergenceofpredictions,inthatinformedherdingfunctionsnear-equivalentlytoa competitivemarketmaker.Ileavetheproofofthesesuppositionsforfutureresearch. Empirically,Ishowthatinsidersinfactdovarytheirtradereportingbasedon theirnetdailytrading.IcalculateCARsinanattempttoseparatetheinsider's tradingmotiveversusotherincentives.Ithatthenetreportingdelayfor buysisincreasinginCARforthethreedaysfollowingthetrade.Thissuggeststhat buysarereportedmoreslowlyastheinsider'sexpectationofincreases.I thatthenetreportingdelayisdecreasinginCARforthethreedaysfollowingthe disclosure.Iarguethatthissuggestsinsidersexpeditereportingofbuyswhenthey expectthemarketreactiontobemorepositive,whichsupportsacareerincentive intradingbehavior.Ibelievethisisthepapertoprovideevidenceofother motivationsforinsidertrading.Ileavetheimplicationsofthistofutureresearch. 28 CHAPTERII MatchingCEOEducationstoFirmCharacteristics 2.1 Introduction CEOscomefromavarietyoftbackgrounds.Manyareclassicallytrained, withundergraduatebusinessdegreesandMBAs.However,thisishardlytheonly path:CEOscanhavetrainingasengineers,doctors,lawyers,ormyriadotherpro- fessions.Mattersarefurthercomplicatedbythevariationsamongtheeducationsof eventhosehavingthesametraining.Bothpublicandprivateschoolsarerepresented atthehighestechelons. The\obvious"path,whichhasbeenimplicitlysuggestedbytheexistingliterature, wouldbetohiretheCEOwiththe\best"education.Thiscouldserveoneoftwo purposes:eithertheCEO'seducationisinherentlybetter,makingherasuperior manager,ortheCEO'sabilitytofollowan\elite"educationalpathwayservesasa signaloftheirinherentquality.Ineithercase,oneshouldseethatwhichare abletoattracttheseidealexecutivesshouldbgreatlybydoingso.Theproblem withthisthinkingisthatitnecessarilyimpliesthatavaryingfromthisidealpath mustnecessarilybechoosingsomesortofmanagerialdiscount.Further,itoverlooks recentevidenceregardingCEOhiringandperformance.Weattempttoreconcile theseissues. 29 2.2 LiteratureReviewandMotivation 2.2.1 HiringMarketObservations Thereisavastliteratureontheeducationofexecutives.Baruch(2009)that morethanhalfoftheCEOsoflargeinternationalholdanMBA.Theynote thatoftenattempttorecruitMBAsformanagementpositionsorencourage theirlower-rankingemployeestopursueMBAsasagatingmechanismforpromo- tion.However,itappearsthatthemarketmightbemovingawayfromtheMBA, asDataretal.(2010)documenttwocountervailingtrends:notonlyareconsult- ingincreasinglyrecruitingnon-MBAs,butenrollmentsinMBAprogramshave precipitouslyfallen,especiallyforlower-rankedprograms. Thiscallsintoquestionthesimplerelationshipofbackgrounddrivingperfor- manceforseveralreasons.First,onemustwonderwhy{ifanMBAreallyistheideal manager{theremainderoflargeinternationalchoosenottoimitatethischoice. ThelargeinternationalBaruch(2009)focusesonshouldnotbeconstrainedin hiringorrecruiting,whichsuggeststhattheymustbemakinganintentionalchoice. Second,iftheMBAisideal,whyisthisseculartrendawayfrommanagerspossessing oneoccurring?Onemightsuggestamorelikelyalternative:themarketisadjusting toanewoptimaleducationcredential.Whileitresolvestheissueofsub-optimality, itraisesthequestionofwhyoptimalmanagerialeducationistime-varying.That is,whywastheMBAoptimal,andwhyisitnolongeroptimal?Theseare questions,andtheyraisedoubtsaboutexistinginterpretations. Perhapsevenmoremotivatingforthisstudyisthenotionthatthemarketseems toactinalargelymonolithicmannerinCEOhiring.Thatis,themarketwasheavily investedinMBAsasCEOs,buthavechosenbyandlargetobeginmoving awayfromMBAs.Thissuggeststhattheremustbesomedrivingforcesbehindthe decisions.Thatis,perhapsratherthanattemptingtohireuniquelytalented 30 CEOstoincreasetheirvalue,perhapstheareattemptingtohiretheCEOswith aneducationalbackgroundtailoredtotheircircumstances.Alternatively, perhapsthesearechoosingwhattheybelievetobethe\best"educational regardlessoftheperformanceimplications.Thatis,boardmembersmay choosetorelyonmoretangiblefactorswhenchoosinganewCEO.Eitherexplanation isconsistentwiththeoveralltrendspreviousstudieshavehighlighted. 2.2.2 ofBackgroundonPerformance Muchacademicresearchhasfocusedonpredictingeithersalaryorperformance basedontheCEO'sbackground.OnewayinwhichtheCEO'seducationalback- groundcanmanifestitselfisthroughtheCEO'sexpertise.Earlystudiesfoundthat R&DisbythetypeofeducationtheCEOreceived.Someofthesestudies includedTylerandSteensma(1998),FinkelsteinandHambrick(1996),andBarker andMueller(2002).Allsimilarresults,showingthatCEOswithtechnicaldegrees investedmoreheavilyonR&Dthantheirpeerswithlaworbusinessbackgrounds. Likewise,otherstudies,suchasGrahamandHarvey(2001,2002)andGraham,Har- vey,andRajgopal(2005),havearguedthatexecutiveswithabusinesseducationlikely holdanadvantageovertheirnon-businessdegreepeersbasedontheirknowledgeof decision-makingtechniquessuchasnetpresentvalueandthecapitalasset pricingmodel. Mian(2001)pointsoutthatthesespecializationctsmightnotholdtruefor CEOsasmuchasCFOs,astheCEOpositionisuniqueintheextenttowhichits responsibilitiesarebroadandwide-ranging,whileotherexecutivescanfocusontheir sparea.Iqbal(2015)reinforcesthisnotion,thatCEOswithbusiness degreesactuallyhedgedlessthanthosewithindustryspdegreesintheoilin- dustry.However,hethattheCFOsalmostunanimouslyholdbusiness degrees,withnorenceineducationalfocusorqualitybetweenthoseinthat 31 hedgeandthosethatdonot. Otherstudiesfocusedonaneducationalqualityhypothesis:perhapsCEOsfrom betterschoolsmakebetterCEOsthanthosefrommorepoorly-regardedschools.In fact,Burt(1992)andBelliveau,O'Reilly,andWade(1996)foundthatgraduates frommoreselectiveschoolsbfromthenetworkingtheseschoolsprovide.The authorsthenfoundsomeevidencethatthesetiescouldimproveperformance. Deary(2004)andFreyandDetterman(2004)posethequestionasoneofinnate ability.TheauthorsarguethatmoreintelligentCEOs,basedonattendingschools withhigheraverageentranceexamscores,performbetterthantheirpeers.Regardless ofthemechanism,subsequentstudieshaveprovidedfurtherevidenceforeducation qualitydrivingperformance,asPerez-Gonzalez(2006)CEOswithIvyLeague degreeshaveimprovedperformance,whileMaxametal.(2006)thathedgefund managerswithdegreesfromhighlyrankedschoolsoutperformtheirpeers. Theevidenceisnotunanimous,though.GottesmanandMorey(2010)useTobin's Qtomeasureperformance,andtheynorelationshipbetweenperformance andtheCEO'seducation.NotonlyistherenobtoholdinganMBAversus anon-businessdegree,buttheauthorsnobtoholdingagraduatedegree versusonlyanundergraduatedegree.Likewise,Bhagatetal.(2011)showsno oftheCEO'sschoolonaslong-termperformance.Itisnotsimplyaquestionof whetherapositiverelationshipexists,though.Infact,Jalbertetal.(2002)establishes anegativecorrelationbetweentheCEO'seducationandareturnonassets. Further,BarkerandMueller(2002)arguethatMBAstendtobemoreriskaverse, preventingthemfromtakingthecorrect{albeitrisky{actions. Otherstudieshavefoundsimilarlymixedresults.Jalbertetal.(2011)attempt tocreatearankingsystemforuniversitiesbasedonthenumberofCEOstheyhave placedintheForbes500(nowconvertedtotheForbesGlobal2000).Indoingso, theyseethataselectgroupofeliteuniversitiesdominatetheCEOhiringmarket. 32 Unexpectedly,theauthorsthattheserankingshavelittlecorrelationwiththe appointedCEOs'salaries.Thisseemstobeanoddresultiftheyareactuallyadding value,astheearlierstudiessuggested.GottesmanandMorey(2010)focusonthe performance.UsingasampleofallNYSEwhoseCEOhasatleasta bachelor'sdegree,theynorelationshipbetweentheCEO'seducationalcharac- teristicsandtheperformance.Giventhesystematicnaturewithwhich seemtohire,thisraisesquestions. andJonson(2013)approachtheissueofeducationandresults moredirectly.TheyattempttoisolatetherelationshipbetweenaCEO's educationandtheperformance.Theyessentiallynorelationshipbetween thetwo.Thatis,thebusinesseducationofamanagerneitherhelpednorharmed thesperformance.Thiswastrueregardlessofwhetherthemanagerheldan MBAorantbusinessdegree.Astheauthorspointout,thiscouldillustrate thesubstitutabilityofbusinesseducation.Thatis,ratherthanpushingemployees toattendatraditionalMBAprogramorhiringmanagerswhoalreadyhaveanMBA degree,anemployermightarrangeforexecutiveeducationcoursesorothermeans ofleadershipdevelopment.O'Leonard(2014)illustratesthis,asthepublicationdoc- umentsthatapproximatelyone-quarteroftrainingdollarsspentbyU.S.were spentonleadershipdevelopment. 2.2.3 ManagerialStyle IftheevidenceismixedregardingtheCEO'seducation,perhapswecouldbuild onthenotionfromBarkerandMueller(2002):aCEO'seducationmightinstead instillinthemaparticularstyleofmanagementthatvalue.Thereisan extensiveliteratureonmanagerialstylethatcanbeexploitedforthispurpose. Earlystudiesinthisareadiscussedthewaysinwhichtmanagerialin- centivesandattitudescouldmanifestinCEOdecisions.ShleiferandVishny(1989), 33 aswellasMorck,Shleifer,andVishny(1990)wereamongthetosuggestthat inmanagerialabilitycoulddictatetheinvestmentbehaviorofmanagers. Laterstudiesexpandedthisanalysistothemanager'sutilityfunction.Thesestud- iesincludedRotembergandSaloner(1993,1994,2000)andAggarwalandSamwick (2003,2006),whofocusedonthemanager'sriskaversioncocient. MurphyandZimmerman(1993)focusoninvestmentbehavior,someev- idencebutconcludingthatitismostlyduetounderperformancepriortothe change.ThisisreinforcedbyDenisandDenis(1995),whothatdismissals precedeincreasedy.They,alongwithWeisbach(1995)andBennedsen, Perez-Gonzalez,andWolfenzon(2007),thatgenerallyreduceinvestment followingcertaintypesofCEOturnovers.Thesearealsoevidentinvarious otherdecisionsthatCEOswouldbeatleastnominallyinchargeof,asillustrated inAdams,Almeida,andFerreira(2005),Bamber,Jiang,andWang(2010),Dyreng, Hanlon,andMaydew(2010),FrankandGoyal(2007),andGraham,Li,andQiu (2009). BertrandandSchoar(2003)thatmanagerialtransferbetween companies,implyingthatmanagerscarrydecision-makingbiaseswiththembetween employers.Thisist,asitsuggeststhatmanagerialbiasesmustpre-date theircurrentemployment.Oneplacewherethismightpresentitselfisintheman- ager'seducation. Onemanagerialcharacteristicthathasbeenwell-documentedisovence. MalmendierandTate(2005,2005,2008),aswellasGoelandThakor(2008)and others,havedocumentedthethatmanagerialovcanhaveona Thehaverangedacrossessentiallyallmanagerialdecisions,frominvestments to,anditevenextendstoacquisitions.Itisnotunrealistictobelievethat amanager'seducationmighteithercreateorsignaltheirov Fee,Hadlock,andPierce(2013)mixedevidenceofCEOstyle.WhenCEOs 34 leaveexogenously,thereappearstobenotchangeinpolicy.However, whenCEOsleaveforendogenousreasons,theredoesseemtobeachangeinpolicy. Moreover,themanager'sstyledoesnotseemtocarryoverbetweenemployers.This suggeststhattheboardhiresamanagertothestyletheywish,ratherthanman- agersimposingtheirstyleontheTranslatingthisresulttotheCEOeducation space,wewouldexpectthataCEO'seducationmatters,butthataresp callychoosingtheparticulareducationalcharacteristicstheyarelookingfor.Thisis especiallyrelevantwithrespecttobehavioralimplicationsintheliterature. 2.3 StudyMotivation Asmentioned,muchacademicinquiryhasbeenfocusedontheofanewlyap- pointeddirectororbackgroundonthesubsequentperformance.These studiestakebackgroundasameasureofqualityorexpertise.Thisisproblematic,as thereismixedevidenceonwhetherthe\better"manageractuallybthe performanceinanymeaningfulway.Likewise,itisunclearthatthereisan\ideal" educationalpathforthemarketasawhole. Evenifoneweretoassumethatthisisthecase,weareonlyabletoobserve itthroughmultipleFirst,someparticularbackgroundneedstoexhibitout- performance.BoardsneedtoobservethisandaccuratelyattributeittotheCEO's educationalbackground.Then,theyneedtohireaccordingly.ProspectiveCEOsalso needtoobservetheoutperformanceandaccuratelyassesstheirownvalueentering negotiations.Then,thetwoneedtobematchedtogether.Wewouldonlybeable toseetheappropriatewhenalloftheseoccur.Alternatively,weproposea slightlytmechanism,wherethereisnosingleoutperformer,butratherpartic- ularskillsparticularsituationsmorethanothers.Ifweassumethateducational backgroundscanproxyforthepresenceoftheseskillsorattitudes,thisisconsistent withthepriorliterature. 35 2.4 TheSample ThisstudyreliesonanoveldatasetintheCEObackgroundliterature.There areseveralcomponentsofthisdatasetthatwillbedescribedseparatelybelow. 2.4.1 Compustat WeidentifyexecutivesusingtheCompustatResearchInsightCDsfrom1990 through2007.Thesediscslistthefourhighestrankingexecutives,asderivedfrom thewiththeUnitedStatesSecuritiesandExchangeCommission.The namesinthesearethentrimmedofallextraneousspaces,andalllettersare convertedtouppercase.Fromhere,wemergethedatawithCompustatbyGVKEY. WeeventuallywishtomergewithBoardex,sowewillneedidenthattranslate reliably.So,weconverttheCUSIPtothesixdigitform.Thiseliminatesthe threedigits,whichconsistofatwodigitidenforaparticularsecurityissue andacheckdigit,leavingonlythesixdigits,whichareuniquetotheitself. Thisshouldbeuniformacrossdataproviders. 2.4.2 BoardexCompanyRecords WeimporttheCompanyCharacteristicsfromBoardex.Thishasbasic identifyinginformationregardingathatcanbeusedtomergewithexternal datasources.TheseincludethetickerandtheISIN.Thetickerisdirectlycompa- rable.However,CompustatlacksanISINobservation,sowemustconverttheISIN toaCUSIP.TypicalISINsare12digits,wherethetwodigitsareacountry identhenextninearetheCUSIP,andtheldigitisacheckdigit.Aswith theCompustatdata,wewishtoeliminateanymistakesfrommismatchingsecurities, sogeneratethesixdigitCUSIP(thethirdthroughtheeighthdigitoftheISIN). 36 2.4.3 BoardexPersonRecords WebeginbyimportingtheCharacteristicsfromBoardex.Thiscon- tainsseveralpointsofinformationabouteachcorporatesuchasherposition intheandthedateshebeganorendedtherole.Allinformationispresented asreportedbytheInordertoensurecompatibility,thedataiscleanedtobe consistent.Thestringbeginningandendingdatesareparsedforyears,months,and days.Thosevaluesthatarepresentareretained.Otherwise,weassumethebroadest possibledaterangeinordertoensurethedataisavailableformerginglater.Hence, missingbeginningyearsareassumedtobe1990,missingbeginningmonthsareas- sumedtobeJanuary,andmissingbeginningdaysareassumedtobetheofthe month.Conversely,missingendingyearsareassumedtobe2013,missingending monthsareassumedtobeDecember,andmissingendingdaysareassumedtobe thelesserof31ortheappropriatenumberofdaysinamonth.Thismeansthatan unknownmonthisassignedanendingdayvalueof31,whileAprilwouldbeassigned anendingdayvalueof30.Oncethemissingvaluesarein,thedatesarerecon- structedtoformproperbeginningandendingreferencepoints.Next,wegenerate annualobservationsforeveryyearbetweenthereportedbeginningandendingdate. OurBoardexdatabeginsin1989,sothisresultsinupto25observationsforeachof- spanningbetweenthelaterof1989andtheirreportedstartdateandtheearlier of2013andtheirreportedenddate. Atthesametime,weimporttheDirectorCharacteristicsfromBoardex.This datacontainsbasicinformationondirectors,asoftheannualreportdate.As before,wecleanthesereporteddateobservationswhennecessary.Whenvaluesof \Current"arereported,wereplacethemwithJanuary1,2011,tocorrespondwiththe endofthedataset.Next,anyobservationsthataremissingdayvaluesareassigned adayvalueof1.Thedatesarethenreconstructed.Finally,wecreateobservations forthetwoyearspriortotheannualreportdate.So,adirectorasofAugust1,2010, 37 wouldhaveobservationsfor2008,2009,and2010.Thisisdoneinordertoensure thatanobservationdoesnotfallthroughthecracksduetomistimingthedirector's actualpresenceatthe Fromhere,weappendtheCharacteristicsdatatotheDirectorCharacter- isticsNow,weattempttocleanthenamesofthecersanddirectors.Namesin Boardexcommonlyhavesuperlativesorothertitlesinthemthatleadto inmatchingwithotherdatasources.Weattempttoparsethemajorityoftheseout. First,weconvertallnamestouppercaseforconsistency.Next,weaddaspaceto thebeginningandtheendingofthestring.Thisseeminglyextraneousstepallowsus toensurethataccidentlymatchingiskepttoaminimum.Forexample,oneofthe titlesbeingcleanedis\DOCTOR"(otherforms,suchas\DR"or\DR.",arealso cleaned).Ifwesimplylookfortheword\DOCTOR"inthename,itispossiblethat wecouldeliminateatleastpartoftheperson'sname.ConsiderifaCEOhappens tohavethename\JUANDOCTORE"{afterremovingthesimplestring,wewould beleftwith\JUANE"(anincorrectparsing).Addingthespacesallowsustosearch for\DOCTOR"in\JUANDOCTORE"andpreventsucherrorsfromoccurring. Oncethiscleaningprocedureiscomplete,anyextraneousspacingisremoved,and thedataisreadytobemergedintothemaindataset. 2.4.4 BoardexEducationRecords WeimporttheDirectorEducationsfromBoardex.Thiscontainseduca- tionalbackgroundinformationonatportionoftheBoardexuniverse.For eachdegreeorothereducationalevent(professionalorganizationsarealsoincluded) foraperson,thereisanobservation.Hence,eachmanagermighthavemultipleen- tries.Infact,somehavenumerousreported,withthemaximumbeing 27(mostlyhonorarydegrees).Inordertokeepthedataatamanageablesizewhen merging,wecompresseachperson'sobservationsintoasingleobservation.So,we 38 generate27ofeachschoolvariable(auniqueidenfortheschool,theschool's name,thedegreethepersonreceived,theyearthepersonreceivedthedegree,etc.). Then,wekeeptheobservationforeachperson. 2.4.5 CombiningtheData WeresumeusingtheCompustatdata.Now,wewishtomergeitwiththeBoardex persondata.However,evenhavingcleanedtheBoardexdata,thisistodo directly,soitwillinvolveseveralsteps.First,wejointheCompustatdatatothe Boardexpersondatausingthename.Thiscreatesallpossiblematches onnames.Fromhere,wemergeintheBoardexcompanydatausingtheBoardex companyidenNow,wechecktoseethatthecompaniesarecorrect:wekeep theobservationifthetickersandCUSIPsmatch.Otherwise,wediscardthem.Next, wedropanyduplicateerson-yearobservations.Theremainingobservationsare labeledasdirectmatches. Next,weresumewiththeCompustatdata.Wemergeontheearlevelwith thedirectmatches,keepingonlythosethatdonotmerge.Wetaketheseleftover observationsandmergethembyCUSIPwiththeBoardexcompanydata.Wesave allthesuccessfullymergedobservationsasapass.Then,weretainalltheobser- vationsthatfailedtomerge.WeattempttomergethesebytickerusingtheBoardex companydata.Wedropthoseobservationsthatdonotmergebeforeappendingthe passobservations.Finally,wemergethisdatasetbyearwiththeBoardex persondata,usingtheBoardexidenNow,wehaveproperlymatched butnotnecessarilyproperlymatchedpeople.Fromhere,wekeeponlythose observationsthataremanagerobservationsbetweenthestartandenddateordirector observationswheretheCompustatdatadateismorethan300daysawayfromthe Boardexannualreportdate.Next,wetrimanyexcessspacesoutofthedata.Atthis point,wenotehowmanywordsareineachname.Then,weaddspacestothebegin- 39 ningandendoftheBoardexname.WesearchforeachwordintheResearchInsight nameintheBoardexnamestring.Ifawordisfound,wecountasuccess,remove itfromtheBoardexname,andcontinueuntilwehavesearchedforallthewordsin theResearchInsightname.Finally,wekeepthoseobservationswithasuccesstotal greaterthan50 Finally,weappendthedirectandindirectmatchestogether.Next,wemerge themwiththeBoardexeducationdatausingBoardexpersonidenFromhere, wecreate27copiesofeachobservation.Doingsoallowsustoreplacethe27versions ofeachvariablewithasingleeducationalitemperobservation.Finally,wesortby yearontheersonlevel,keepingonlytheobservation.Wepresumethat theearliestobservationrepresentstheyearinwhichthepersonservedasCEO oftheThisallowsustointerpretthedataasacrosssection. 2.4.6 EducationQuality Inordertoincreasethecomparabilityoftheeducationdata,wecompilealist ofthevarioustypesofdegreeslistedinthedata.Doingsoallowsustoclassify eachdegreetypeintoanappropriatecategory.Forexample,aBAorBSdegree isabachelor'sdegree,whileanAAorASisanassociate'sdegree.Ultimately,we classify330degreetypesintoeightcategories.Forourpurposesgoingforward,the mostimportantareUndergraduate(acombinationofBachelor'sandAssociate's, withpreferencetotheBachelor's),Master's,MBA,Law,andDoctor.Wecarefully parsethereporteddegrees,givingprecedencetoearlierreporteddegrees(inorderto dealwiththepossibilityofhonorarydegrees,alongwithotherdatacontaminations) andpositively-idendegreetitles.WithinMBAs,wereporttheBusinessWeek rankingsinordertolaterassessquality.Next,sincesomeschoolnamesaremissing, weassumethatapersonattendedthesameschoolforundergraduateandgraduate study,byreplacingtheemptyschoolnamefortheundergraduateinstitutionwiththe 40 person'sgraduateinstitution. Additionally,werecognizethattheBoardexdataisinconsistentwithschool names.Forexample,weobservethat\UofM,"\UniversityofMichigan,"\Michi- gan,"and\StephenRossCollegeofBusiness"areallgivenuniqueidendespite ostensiblyreferringtothesameuniversity.Thisposesaproblem,soweassigna codetoasmanyoftheschoolnamesascanbepositivelyiden(meaningallof theUniversityofMichiganmonikersabovewouldhavethesameidenIndoing so,wedropobservationsfromprofessionalorganizations,governmentalorganizations, highschools,andothernon-highereducationinstitutions. Next,wemaketheassumptionthattheaverageCEOisapproximately50years old(roughlyequaltothemeaninourdata).Thus,weincorporatehand-collected datafromHawes(1978)oncollegesanduniversities.Thisdatagiveswide-ranging informationonvariouscolleges,yieldinginformationonschoolquality,enrollment, selectivity,cost,andmore.Webelievethatthisdataismoreappropriate thancurrentcollegerankings,asitshouldbemoreaccurateregardingthecollege's particularsituationatthetimetheCEOattended.Weassignthecorrespondingcode fromourpreviousofBoardexentriestothecorrespondingobservation fromthebook'sdata.Foranyschoolsmissingselectivitydata,weassumetheyare simplynotveryselective.Weperformsimilaroperationsforallvariableswheredoing somakessense. Finally,weassignqualitydummiestothedata.Wecreatetwomeasuresofprestige basedonHawes'sreportedadmissionyandselectivity.Ifadmissiony isa1(thehighestlevel),andselectivityisratedgreaterthan85,weprestige1to be1.Likewise,ifadmissionyisa2andselectivityisratedgreaterthan90,we prestige1tobea1.Otherwise,itiszero.Weprestige2inasimilar,but slightlystricter,manner,increasingtheselectivityratingby5foreachconditional. Wesaythatthepersonattendedgraduateschooliftheyhaveanidenmaster's, 41 MBA,law,ordoctoraldegree.WeaprestigiousMBAtobeoneinthetop 350inthecountry,whileaneliteMBAisconsideredtobeinthetop10usingthe BusinessWeek rankings. Inaddition,wearichschoolasthosewitha75thpercentileorhigher predictedtotalcost,whenregressingcostonfacultysalaries,totalenrollment,a dummyfortheschoolbeingprivate,anditsadmissiony.Next,wea highclassschoolasonewitha75thpercentileorhigherpredictedratioofalumni onthe SocialRegister tototalenrollmentwhenregressingonitscost,itsadmission y,andadummyfortheschoolbeingprivate.Last,aschooliseliteifitsvalues forrichandprestige2areboth1. 2.5 EmpiricalAnalysis Havingconstructedthesample,weturntoananalysisofthefactorssurrounding achoiceofCEO.Wefocusonsixkeyvariables,inbothaunivariateandmul- tivariatecontext,usingbothsimpleregressionandlogittechniques.Thesevariables focusonthreeprimaryaspectsoftheFirst,thelog-transformoftotalassets servesasaproxyforthesize.Second,wecontrolforthecharacter- isticsoftheInordertodoso,weincorporateintangibleassetsasashareof totalassets,liabilitiesasashareoftotalassets,andR&Dexpenseasapercentage ofrevenue.Finally,weconsidertheperformance.Ourproxiesfor thisaspectarethelaggedmargin,thelaggedreturnonassets,andthelagged Tobin'sQ.Theseindependentvariablesareusedtopredictavarietyofeducational variables.Thesedependentvariablesincludedummiesforthetypeofdegree,the prestigeoftheschool,andthequalityoftheschool.Allt-statisticsarebasedon robuststandarderrors.SummarystatisticsforourvariablesarepresentedinTable1. PanelAsummarizesthedependentvariables,whilePanelBsummarizestheindepen- dentvariables.Wecanseethatmostofthedependentvariablevaluesarerelatively 42 closetozero,duetotherelativelylownumberofyesses. 2.5.1 RegressionFramework Weuseamultivariateregressionanalysistotestthehypothesesinourstudy.We performtheseanalysesusingthreebasicmodels.Themodelstrictlyincludes thedependentvariablesofinterest.Thesecondincorporatesindustryandyear Thethirdreplacesthevariablesandindustrywiththerespective industrymeansandtheeldeviationsfromthemean. \ DependentVariable = b 0 + c 1 IndependentVariables \ DependentVariable = b 0 + c 1 IndependentVariable + b 1 IndustryFE + b 2 YearFE \ DependentVariable = b 0 + c 1 Deviation + b 1 IndustryMeans + b 2 YearFE Eachmodelisrepeatedusingalogitfunctionalform.Thelogitmodelsarerun usingtheclosestequivalentfunctionalform.Theindustryandmean calculationsareaggregatedatthe2-digitSIClevel,andthemeancalculationsare performedonanannualbasis.Thisapproachallowsustoparseouttheraw ofthevariablesandanynon-linearitythatexistsinthevariousrelationships.While wecannotfullyresolvetheendogeneitypresentinthemodel,wehopethatthis tlyisolatestheparticularrelationshipsinquestion. 2.5.2 DegreeType Weconsiderthedegreeheldbythemanager.Sp,weconsiderwhether themanagermightholdalawdegree,anMBA,amaster'sdegreeofanykind,a doctoraldegree,oragraduatedegreeofanykind. Thespecializeddegreethatonemightsurmiseprovidesspecializedandnec- essaryskillsisalawdegree.Table2presentstheresultsofregressionsoflawdegree 43 Table2.1SummaryStatistics PanelA:DependentVariables VariableMeanStd.Dev.Min.Max.N law0.0750.263016314 doctor0.0960.295016314 mba0.3520.478016314 master0.1930.394016314 gradschool0.6030.489016314 private0.4360.496016314 religious0.1420.349016314 foreign0.0650.247016314 prestige10.2810.45016314 prestige20.2370.425016314 prestigiousmba0.1860.389016314 elitemba0.1340.341016314 rich0.220.414016314 highclass0.2370.426016314 elite0.1750.38016314 PanelB:IndependentVariables VariableMeanStd.Dev.Min.Max.N FirmLevel logassets5.7621.953-1.26213.426313 pctintan0.1490.18400.9415674 pctliab0.520.414011.356296 pctrndrev4.09757.601-2.2631927.1824010 lagpm-4.67130.707-9468111.9586164 lagroa-0.0930.575-26.9316.8216279 lagtobinq2.2424.6050.002169.0713590 IndustryMeans indlogassets5.7621.0072.17712.0686313 indpctintan0.1490.09600.7095674 indpctliab0.520.1650.0322.7216296 indpctrndrev4.09711.339062.6594010 indlagpm-4.6716.725-33218.4386164 indlagroa-0.0930.218-5.3152.2256279 indlagtobinq2.2422.0070.05360.4523590 DeviationfromIndustryMean devlogassets01.674-6.4136.8466313 devpctintan00.157-0.4830.7515674 devpctliab00.379-1.8089.2846296 devpctrndrev056.474-62.6591864.5234010 devlagpm0129.632-9340.201128.6826164 devlagroa00.532-26.214.5966279 devlagtobinq04.144-59.538164.5773590 44 Table2.2DeterminantsoftheCEOHoldingaLawDegree (1)(2)(3)(4)(5)(6) OLS1Logit1OLS2Logit2OLS3Logit3 logassets0.005110.08240.004630.0732 (1.64)(1.72) (1.46)(1.42) pctintan0.01310.2260.02770.524 (0.42)(0.42)(0.87)(0.92) pctliab0.02740.4300.01970.348 (1.50)(2.23) (1.08)(1.49) pctrndrev-0.0000536-0.0126-0.0000521-0.0111 (-2.01) (-1.26)(-1.97) (-1.15) lagpm-0.0000991-0.00133-0.0000981-0.000980 (-18.15) (-0.70)(-14.75) (-3.35) lagroa0.005690.1020.004140.0550 (0.86)(0.62)(0.63)(0.31) lagtobinq-0.000806-0.01240.0005360.0107 (-0.29)(-0.20)(0.18)(0.16) devlogassets0.004980.0812 (1.54)(1.62) devpctintan0.02640.492 (0.79)(0.85) devpctliab0.01920.324 (1.03)(1.48) devpctrndrev-0.0000523-0.0116 (-1.92) (-1.05) devlagpm-0.000102-0.00156 (-18.16) (-0.48) devlagroa0.002120.0240 (0.29)(0.14) devlagtobinq-0.00003550.00295 (-0.01)(0.05) Intercept0.0193-3.424-0.142-2.988-0.0791-3.306 (0.98)(-9.38) (-3.13) (-2.61) (-1.37)(-3.31) IndustryNoneNoneFEFEMeanMean YearNoneNoneFEFEFEFE N 217521752175205921752174 R 2 0.01190.06320.0198 t statisticsinparentheses p<: 10, p<: 05, p<: 01 45 dummyonoursetofindependentvariables.Weseelittleevidenceofasystematic tendencytowardhiringaCEOwithalawdegreebasedonthesize.However, weastrongnegativerelationshiponthemarginintheyearpriortothe CEO'shiring.Onepossibleinterpretationisthattendtohire lawyersasCEOinordertousetheirexpertisetollystrengthenthe incomestatement,asitislikelythatareorganizationofthebookswouldlikely resultinamuchfasterturnaroundthanamorestrategicfocus. WealsoseeamoderatenegativerelationshipbetweentheR&Dexpense andtheCEOhavingalawdegree.Whiletherelationshipdisappearsinthelogit spthisstillsuggestsaweaktendencytowardhiringalawyerasCEO. Themostlikelyexplanationwouldbetousetheirlegaltrainingtobetterprotectthe intellectualproperty. Doctoraldegreespresentanotheropportunityfortoharnessspecialized skills.Table3presentstheresultsofourmodelsusingadoctoraldegreedummy asthedependentvariable.Thereisnofromsize,butweseet alongboththecharacteristicsandyvectors.Weseentlynegative forbothintangibilityandliabilities. TheysweseeonlaggedROAandlaggedTobin'sQareweaker, butstilllargelyt.WeseethattheofROAisnegative,suggestingthat tendtohavesuppressedyintheyearpriortotheCEOturnoverevent. However,Tobin'sQexhibitsapositive{albeitweaker{relationshiptothepresenceofa doctoraldegree.Sincethemarketdoesnotappeartohaveanegativeoutlookonthe correspondingtoitslowerROA,wecanconcludethatthehassomefuture earningpotential.Giventheiranalyticaltraining,adoctoraldegree-holdermightbe thekeytounlockingthatpotentialintheeyesofthe AnMBA-holder,ontheotherhand,isamuchmorecommonchoice.Table4 presentstheresultsofourregressionsusinganMBAdummyasthedependentvari- 46 Table2.3DeterminantsoftheCEOHoldingaDoctoralDegree (1)(2)(3)(4)(5)(6) OLS1Logit1OLS2Logit2OLS3Logit3 logassets-0.00323-0.03590.001270.0272 (-0.87)(-0.82)(0.33)(0.63) pctintan-0.118-1.254-0.0847-0.932 (-3.31) (-2.98) (-2.20) (-2.11) pctliab-0.0664-0.880-0.0593-0.828 (-5.03) (-3.29) (-4.40) (-3.26) pctrndrev0.0001610.0006580.0000468-0.0000248 (0.79)(0.76)(0.27)(-0.04) lagpm-0.00000436-0.0000240-0.0000120-0.000114 (-0.17)(-0.17)(-0.52)(-0.82) lagroa-0.0538-0.370-0.0407-0.310 (-2.53) (-2.23) (-2.11) (-1.96) lagtobinq0.006850.04630.003270.0257 (2.29) (2.60) (1.04)(1.21) devlogassets0.0006520.0123 (0.16)(0.27) devpctintan-0.110-1.148 (-2.74) (-2.49) devpctliab-0.0573-0.791 (-4.36) (-3.03) devpctrndrev0.0000399-0.0000666 (0.22)(-0.09) devlagpm-0.0000147-0.000112 (-0.58)(-0.75) devlagroa-0.0410-0.289 (-1.90) (-1.65) devlagtobinq0.002800.0165 (0.82)(0.71) Intercept0.164-1.4290.0163-0.9650.155-1.643 (6.26) (-5.84) (0.34)(-1.85) (3.13) (-2.25) IndustryNoneNoneFEFEMeanMean YearNoneNoneFEFEFEFE N 217521752175195021752174 R 2 0.02850.09260.0630 t statisticsinparentheses p<: 10, p<: 05, p<: 01 47 Table2.4DeterminantsoftheCEOHoldinganMBADegree (1)(2)(3)(4)(5)(6) OLS1Logit1OLS2Logit2OLS3Logit3 logassets0.01050.04280.01290.0544 (1.85) (1.76) (2.11) (2.05) pctintan-0.0557-0.242-0.0943-0.420 (-0.97)(-0.96)(-1.52)(-1.53) pctliab0.01790.08150.02430.110 (0.68)(0.72)(0.92)(0.95) pctrndrev0.00007040.0003240.00005210.000252 (0.37)(0.41)(0.25)(0.29) lagpm0.00002740.0001930.00002540.000182 (2.24) (1.14)(1.89) (1.07) lagroa0.03330.1710.03770.197 (2.17) (1.79) (2.43) (1.99) lagtobinq-0.00137-0.00575-0.00237-0.0105 (-0.41)(-0.38)(-0.70)(-0.67) devlogassets0.01340.0550 (2.12) (2.00) devpctintan-0.0799-0.346 (-1.26)(-1.24) devpctliab0.02930.133 (1.09)(1.15) devpctrndrev0.00003950.000202 (0.19)(0.24) devlagpm0.00003280.000234 (2.47) (1.28) devlagroa0.03750.203 (2.27) (1.87) devlagtobinq-0.00183-0.00784 (-0.50)(-0.47) Intercept0.320-0.734-0.261-1.468-0.0599-0.677 (8.78) (-4.61) (-1.12)(-1.24)(-0.63)(-1.41) IndustryNoneNoneFEFEMeanMean YearNoneNoneFEFEFEFE N 217521752175215621752174 R 2 0.00470.03360.0111 t statisticsinparentheses p<: 10, p<: 05, p<: 01 48 able.Weseeapositiveforthesize.Thiswouldsuggestthat choosemorebusiness-orientedCEOsastheybecomelargerandmoremature.This couldbetheresultofincreasingspecializationofexecutivefunctionsinanexpanding bureaucracyoraneedforbusinessexpertiseinamature Thesecondexplanationbecomesmoretoembracewhenoneconsiders thepositiverelationshipsexhibitedbythelaggedproyvariables,though.The tcoientsonlaggedROAandmildlytcotsonlagged marginsuggestthatthesarealreadyingoodhealthwhentheturnover eventtakesplace.Thissuggeststhatthebusinessexpertisemightnotbestrictly necessary,lendingcredencetothebureaucraticexplanation. Suppose,however,thatwebroadenournettothosewithanykindofmaster's degree.Table5presentstheregressioncotsinthiscontext.Weseethatthe sizegoaway.Further,weseethattheyhaveweakened, withlaggedmarginremainingtonlyintheOLSsps.This remainsconsistentwiththeMBAstory,asmsingoodhealthtargetnewCEOs withmoreadvancedorspecializedskills.Sincewearenolongerfocusingonany particularspecialization,itmakessensethattheofspecializationisweakened. Moreinteresting,however,istheweaklynegativeobservedonliabilitiesas apercentageofassets.Thus,withlowerleveragetendtobemorelikelytohave aCEOwithamaster'sdegree.Onemightsurmisethatadvancedtrainingleadsto lessrisk-taking.Alternatively,CEOswithmaster'sdegreesmightbeasignofmore rigidpromotionstructure,suggestingamoreconservativecorporateenvironment. Ofcourse,thesesameargumentscouldbeusedinthemoregeneralcontextof havinganygraduatedegree.Table6illustratestheoutcomeofourregressionsusing agenericgraduateschooldummyasthedependentvariable.Weseeapositivesize againsuggestingthatlargerprefertheirCEOspossesssomeformof advancedtraining.Onealsoseesanegativerelationshipbetweenintangibilityand 49 Table2.5DeterminantsoftheCEOHoldingaMaster'sDegree (1)(2)(3)(4)(5)(6) OLS1Logit1OLS2Logit2OLS3Logit3 logassets-0.00368-0.02150.0000755-0.000914 (-0.79)(-0.75)(0.02)(-0.03) pctintan-0.0147-0.0924-0.0270-0.160 (-0.30)(-0.31)(-0.50)(-0.50) pctliab-0.0445-0.301-0.0350-0.218 (-2.18) (-1.74) (-1.65) (-1.39) pctrndrev0.00009870.0005050.0001460.000769 (0.51)(0.61)(0.72)(0.87) lagpm0.00003070.0009100.00002910.000750 (3.71) (0.45)(3.01) (0.42) lagroa-0.0178-0.105-0.00778-0.0444 (-0.97)(-1.09)(-0.42)(-0.48) lagtobinq0.002320.01290.002770.0165 (0.71)(0.76)(0.84)(1.00) devlogassets0.001410.00868 (0.27)(0.28) devpctintan-0.0453-0.275 (-0.83)(-0.85) devpctliab-0.0290-0.191 (-1.38)(-1.16) devpctrndrev0.0001380.000727 (0.72)(0.89) devlagpm0.00003340.000793 (3.08) (0.45) devlagroa-0.00671-0.0433 (-0.36)(-0.45) devlagtobinq0.002170.0122 (0.58)(0.62) Intercept0.252-1.0580.0655-1.0800.154-0.947 (8.01) (-5.73) (0.27)(-0.85)(1.97) (-1.68) IndustryNoneNoneFEFEMeanMean YearNoneNoneFEFEFEFE N 217521752175209421752174 R 2 0.00400.04520.0139 t statisticsinparentheses p<: 10, p<: 05, p<: 01 50 Table2.6DeterminantsoftheCEOHoldingaGraduateDegree (1)(2)(3)(4)(5)(6) OLS1Logit1OLS2Logit2OLS3Logit3 logassets0.01020.04610.01850.0861 (1.79) (1.85) (3.05) (3.09) pctintan-0.141-0.575-0.151-0.652 (-2.42) (-2.34) (-2.43) (-2.40) pctliab-0.0279-0.139-0.0174-0.0926 (-1.10)(-1.23)(-0.67)(-0.77) pctrndrev0.0002860.004810.0002360.00324 (3.57) (0.93)(3.75) (0.87) lagpm-0.0000492-0.00249-0.0000569-0.00212 (-3.56) (-1.24)(-4.95) (-1.10) lagroa-0.00896-0.009120.007640.0594 (-0.37)(-0.09)(0.33)(0.62) lagtobinq0.002740.009650.000300-0.000610 (0.75)(0.57)(0.08)(-0.04) devlogassets0.01880.0846 (2.96) (2.98) devpctintan-0.159-0.673 (-2.49) (-2.45) devpctliab-0.00719-0.0478 (-0.28)(-0.40) devpctrndrev0.0002240.00295 (3.23) (0.78) devlagpm-0.0000525-0.00270 (-3.59) (-1.45) devlagroa0.005040.0541 (0.21)(0.53) devlagtobinq-0.000496-0.00445 (-0.12)(-0.24) Intercept0.6020.404-0.493-1.8530.09400.617 (16.30) (2.51) (-2.08) (-1.50)(0.99)(1.24) IndustryNoneNoneFEFEMeanMean YearNoneNoneFEFEFEFE N 217521752175215621752174 R 2 0.00590.05780.0256 t statisticsinparentheses p<: 10, p<: 05, p<: 01 51 thepresenceofagraduatedegree.Conversely,thereisapositiverelationshipwith therm'sR&Dexpense.Onesurmisesthatgraduatedegreesareblfors withhighlycapitalintensiveresearch.Thatis,whicharefocusedonresearch butrequirealargeamountofequipmentinordertoperformthisresearch.Thiscould beameansthroughwhichattempttomitigateagencyissuesandprotecttheir capitalinvestments. Theinconsistentlytnegativeoflaggedmarginsuggeststhat inrelativelypoorhealtharemorelikelytochooseaCEOwithanadvanced degree.Thisislikelyduetoabeliefthattheadvancedtrainingcouldaidthein improvingitsy.Giventhatthedebtratioseemstohavenoect, itappearstherecouldbesometruthinthisbelief. Takentogether,weseeevidencethatsystematicallychooseCEOswithpar- ticulartrainingorskills.Itappearsthatgraduatedegreesareonewayinwhich areabletoselectaCEOpossessingtheirdesiredskillset.Moreover,itappearsthat thisisnotstrictlylimitedtothe\obvious"MBA,butitstretchestoallgraduate degrees. 2.5.3 SchoolCharacteristics Ofcourse,thedegreeitselfisnottheonlythroughwhichselecttheir CEO.CompaniesalsodiscriminatebasedonthecharacteristicsoftheCEO'seduca- tion.Thereareseveralsuchcharacteristics.Whilethesehavenodirect,quantitative linktotheCEO'sknowledgeorskills,theyarenonethelesshighlyvisibleaspectsof theeducationtheCEOreceived.Thismakesthemripeforbyprospective employers. Themostimmediatelyobviouscharacteristicofaschooliswhetheritisprivate.In Table7,wepresenttheresultsofourregressionmodelsusingadummyfortheschool beingprivate.Surprisingly,weseelittleonanyvectorinthesespIn 52 Table2.7DeterminantsoftheCEOGraduatingfromaPrivateSchool (1)(2)(3)(4)(5)(6) OLS1Logit1OLS2Logit2OLS3Logit3 logassets-0.00284-0.0113-0.00153-0.00579 (-0.49)(-0.47)(-0.24)(-0.22) pctintan0.08900.3820.06710.290 (1.49)(1.56)(1.05)(1.11) pctliab-0.000994-0.0220-0.00546-0.0415 (-0.04)(-0.20)(-0.20)(-0.37) pctrndrev0.00008680.0001040.0000527-0.0000107 (0.43)(0.13)(0.25)(-0.01) lagpm-0.0000708-0.00211-0.0000744-0.00196 (-4.77) (-1.49)(-5.05) (-1.41) lagroa-0.00757-0.00910-0.0111-0.0264 (-0.36)(-0.10)(-0.51)(-0.29) lagtobinq-0.000893-0.00532-0.000547-0.00377 (-0.25)(-0.36)(-0.15)(-0.24) devlogassets0.0009920.00506 (0.15)(0.19) devpctintan0.06550.287 (1.01)(1.08) devpctliab-0.00998-0.0618 (-0.36)(-0.54) devpctrndrev0.0000524-0.0000353 (0.24)(-0.04) devlagpm-0.0000754-0.00217 (-4.47) (-1.51) devlagroa-0.0174-0.0497 (-0.79)(-0.53) devlagtobinq-0.0000318-0.00175 (-0.01)(-0.11) Intercept0.411-0.358-0.149-0.9610.1350.413 (11.02) (-2.31) (-0.63)(-0.81)(1.44)(0.87) IndustryNoneNoneFEFEMeanMean YearNoneNoneFEFEFEFE N 217521752175216121752174 R 2 0.00230.03050.0078 t statisticsinparentheses p<: 10, p<: 05, p<: 01 53 theOLSspthereisastronglytbutsmallnegativerelationship onthelaggedmargin.Asinthedegreetyperegressions,weseeunderperforming havesomepreferenceforprivatelyeducatedCEOs.Thiscouldbeinterpreted asattoquality,althoughtheprivatecategoryisfartoobroadforthistobe convincing. Anothervisibleaspectofaschool'seducationalenvironmentisthewealthofits studentbody.Presumably,these\richer"schoolsshouldprovidestudentswitha widerarrayofopportunitiesandsuperiorthanlesswealthyschoolscould. Table8isinspiredbythispresumption.However,muchliketheprevioustable,wesee littlerealThereisamoderatelytpositivesizewhenincluding theindustryandyearItislikelythatcorporationsbelievethattheseb mightbereal,especiallyasthebecomeslarger. Table9takesthisintuitionastepfurther.Aschoolthatisprimarilyattendedby upperclassstudentswouldlikelythebestnetworkingopportunities.We weakevidenceofthis,asthereisapositivesizeLargerarelikelytohave morealumniconnectionstotheseschools,makingthemmorelikelytorecruittheir graduates.Wecansomepotentialcorroborationforthisideainthemoderately positivecotonthelaggedpromargin.Betterperformingarehiring thesecandidates,allowingsomepresumptionthatthehiringisnotmadepurelyfor needbutthroughcontacts. Table10illustratesaparticularlyuniquecaseofaschool'scharacteristics:ifit hasareligious.Thisuniquetraitappearstodemonstratesomeofthe strongest{yetleastconsistent{relationshipsthatweThelaggedmargin hasastronglytnegativerelationshipintheOLSspbutitis nottinthelogitmodels.Thisinconsistencyislikelyrelatedtothelow percentageofesinthevaluesofthe religious dummyvariable,asitisa strictsubsetofthesetofprivateschools.Thereisalsoaweaklytpositive 54 Table2.8DeterminantsoftheCEOGraduatingfromaRichSchool (1)(2)(3)(4)(5)(6) OLS1Logit1OLS2Logit2OLS3Logit3 logassets0.005910.03580.01040.0654 (1.24)(1.24)(2.01) (2.07) pctintan0.003980.0227-0.0343-0.220 (0.08)(0.07)(-0.64)(-0.66) pctliab-0.00334-0.02030.000299-0.00183 (-0.13)(-0.13)(0.01)(-0.01) pctrndrev0.00003990.0002270.00005140.000295 (0.31)(0.34)(0.37)(0.43) lagpm0.000007540.00005390.00001040.0000684 (0.48)(0.39)(0.62)(0.48) lagroa-0.0135-0.0776-0.00813-0.0434 (-0.80)(-0.86)(-0.45)(-0.46) lagtobinq0.002900.01640.001070.00577 (0.96)(1.03)(0.34)(0.34) devlogassets0.01110.0687 (2.09) (2.12) devpctintan-0.0244-0.157 (-0.44)(-0.46) devpctliab-0.00208-0.0147 (-0.08)(-0.08) devpctrndrev0.00004050.000227 (0.30)(0.33) devlagpm0.00001280.0000787 (0.71)(0.56) devlagroa-0.0114-0.0644 (-0.63)(-0.67) devlagtobinq0.003670.0217 (1.10)(1.19) Intercept0.161-1.619-0.277-2.3980.135-0.300 (5.28) (-8.65) (-3.81) (-2.35) (1.68) (-0.50) IndustryNoneNoneFEFEMeanMean YearNoneNoneFEFEFEFE N 217521752175212221752174 R 2 0.00130.03500.0096 t statisticsinparentheses p<: 10, p<: 05, p<: 01 55 Table2.9DeterminantsoftheCEOGraduatingfromaHighClassSchool (1)(2)(3)(4)(5)(6) OLS1Logit1OLS2Logit2OLS3Logit3 logassets0.005200.02900.009920.0573 (1.03)(1.02)(1.81) (1.86) pctintan-0.0256-0.150-0.0692-0.399 (-0.50)(-0.50)(-1.26)(-1.26) pctliab-0.00701-0.0396-0.00243-0.0188 (-0.31)(-0.29)(-0.10)(-0.13) pctrndrev-0.0000185-0.0000849-0.000002640.00000232 (-0.17)(-0.13)(-0.02)(0.00) lagpm0.00001970.0002110.00002020.000195 (2.97) (1.19)(2.51) (1.23) lagroa-0.0135-0.0729-0.00650-0.0343 (-0.75)(-0.81)(-0.35)(-0.37) lagtobinq0.003740.01960.002800.0146 (1.16)(1.25)(0.83)(0.88) devlogassets0.01110.0631 (1.96) (1.98) devpctintan-0.0614-0.357 (-1.09)(-1.10) devpctliab-0.00100-0.00545 (-0.04)(-0.04) devpctrndrev0.000003540.0000403 (0.03)(0.06) devlagpm0.00002160.000214 (2.31) (1.23) devlagroa-0.0135-0.0724 (-0.70)(-0.75) devlagtobinq0.004150.0220 (1.16)(1.23) Intercept0.195-1.402-0.311-1.2860.125-0.313 (6.12) (-7.82) (-4.35) (-1.54)(1.54)(-0.55) IndustryNoneNoneFEFEMeanMean YearNoneNoneFEFEFEFE N 217521752175213321752174 R 2 0.00140.02630.0067 t statisticsinparentheses p<: 10, p<: 05, p<: 01 56 Table2.10DeterminantsoftheCEOGraduatingfromaReligiousSchool (1)(2)(3)(4)(5)(6) OLS1Logit1OLS2Logit2OLS3Logit3 logassets-0.00651-0.0643-0.00597-0.0635 (-1.61)(-1.78) (-1.36)(-1.58) pctintan0.05360.4300.06010.496 (1.28)(1.27)(1.36)(1.36) pctliab0.01630.1700.01480.172 (0.98)(1.28)(0.88)(1.25) pctrndrev-0.000124-0.0161-0.000138-0.0205 (-2.73) (-1.66) (-2.80) (-1.65) lagpm-0.0000907-0.00275-0.0000931-0.00374 (-13.77) (-1.04)(-12.90) (-1.31) lagroa0.02430.2780.02260.295 (2.12) (1.48)(1.88) (1.35) lagtobinq-0.00203-0.0162-0.000916-0.00801 (-0.91)(-0.65)(-0.39)(-0.28) devlogassets-0.00397-0.0432 (-0.88)(-1.08) devpctintan0.05210.422 (1.12)(1.10) devpctliab0.01830.195 (1.07)(1.45) devpctrndrev-0.000138-0.0181 (-2.77) (-1.67) devlagpm-0.0000937-0.00302 (-11.74) (-1.14) devlagroa0.02350.288 (1.91) (1.36) devlagtobinq-0.00157-0.0139 (-0.63)(-0.47) Intercept0.164-1.5670.143-0.6900.0566-1.049 (6.39) (-6.89) (0.62)(-0.55)(0.90)(-1.49) IndustryNoneNoneFEFEMeanMean YearNoneNoneFEFEFEFE N 217521752175210821752174 R 2 0.00610.03820.0093 t statisticsinparentheses p<: 10, p<: 05, p<: 01 57 onlaggedROA,however,whichmakestheyimplicationsto interpret.Finally,thereisamoderatelytnegativecotonthe R&Dexpense.ThisrelationshiptoreducedR&Dislikelyduetoapresumptionthat religiouspersonsareanti-scienleadinghighlyresearch-orientedtoshyaway. AnotheruniquetraitthataCEO'seducationcouldpossessisthatitdidnotoccur intheUnitedStates.AswesawinTable1,thereareveryfewCEOsinthesample whopursuedtheireducationataforeigninstitution.Table11showstheresultsof ourregressionmodelsbeingappliedtothisdistinction.Weseeweakin severalvariables.Sizeisweaklypositive,suggestingperhapsaninevitabilitythat largewhichareoftenmultinational,willhaveaforeign-trainedCEO.Wesee thatthecocientontheR&Dactivityispositiveinthelogitmodel.Acyni- calinterpretationmightbethatforeigneducationalinstitutionsaresuperiortotheir Americancounterpartsinanalyticalbutitwouldmorelikelybethesimplere- sultofforeign-bornemployeesworkingtheirwayupthehierarchy.Theevidence ontabilityismixed,though,asthereisapositiveandtcot onlaggedmarginintheOLSspwhilethelogitspionsshowa negativeandtrelationshipwithlaggedROA.Thisistodisentangle, anditislikelytheresultofthesmallnumberofesinoursample. WeseethatthecharacteristicsofaCEO'seducation{despitebeinghighlyvisible andquane{arenoteasilypredictablebasedonthesize,char- acteristics,ory.Whileweareabletomakesomebroadinterpretations, thereislittleprecisionwecanbringtobear. 2.5.4 SchoolQuality Theschool'scharacteristicsareeasilyquanable,butwecannotaccuratelyassess theirontheCEO'straining.Theschool'squality,however,ismoreto quantify,butitsismucheasiertodiscern.Weapproachthisaspectfromtwo 58 Table2.11DeterminantsoftheCEOGraduatingfromaForeignSchool (1)(2)(3)(4)(5)(6) OLS1Logit1OLS2Logit2OLS3Logit3 logassets0.003720.03830.005900.0705 (1.10)(0.92)(1.65) (1.64) pctintan-0.000597-0.0151-0.00362-0.0486 (-0.02)(-0.04)(-0.10)(-0.11) pctliab-0.00773-0.07320.001170.0378 (-0.47)(-0.34)(0.07)(0.22) pctrndrev0.0002470.001430.0002340.00138 (1.54)(2.15) (1.41)(2.04) lagpm0.00001390.0002520.00001690.000515 (2.40) (0.74)(2.43) (0.47) lagroa-0.0288-0.244-0.0210-0.188 (-1.82) (-2.08) (-1.35)(-1.66) lagtobinq0.001160.0143-0.000322-0.00224 (0.49)(0.63)(-0.13)(-0.08) devlogassets0.006400.0780 (1.72) (1.71) devpctintan-0.0255-0.334 (-0.69)(-0.71) devpctliab0.005720.0870 (0.34)(0.44) devpctrndrev0.0002520.00149 (1.51)(2.09) devlagpm0.00002240.000412 (2.65) (0.63) devlagroa-0.0166-0.142 (-1.02)(-1.22) devlagtobinq-0.000476-0.00456 (-0.19)(-0.17) Intercept0.0615-2.6220.392-0.583-0.0486-3.933 (2.80) (-9.73) (1.53)(-0.54)(-0.87)(-4.77) IndustryNoneNoneFEFEMeanMean YearNoneNoneFEFEFEFE N 217521752175198721752174 R 2 0.00580.03470.0180 t statisticsinparentheses p<: 10, p<: 05, p<: 01 59 directions.First,weexaminetheselectivityandadmissionsyoftheschool. Then,wemoveontothequalityoftheschool'sprogramsasassessedby Business Week .Thesemeasuresareworkable,allowingustointeractwiththeschool'squality withoutdwellingtoolongonourabilitytoestimatequality. Table2.12DeterminantsoftheCEOGraduatingfromaPrestigiousSchool (1)(2)(3)(4)(5)(6) OLS1Logit1OLS2Logit2OLS3Logit3 logassets0.003060.01520.009700.0506 (0.59)(0.58)(1.73) (1.77) pctintan0.01020.0484-0.0548-0.283 (0.19)(0.18)(-0.96)(-0.97) pctliab0.01160.05620.01940.0971 (0.45)(0.46)(0.73)(0.76) pctrndrev-0.0000229-0.0001000.000009250.0000674 (-0.19)(-0.16)(0.07)(0.10) lagpm0.00001920.0001390.00002220.000147 (1.62)(0.98)(1.65) (1.03) lagroa-0.00825-0.04050.0008320.00400 (-0.44)(-0.46)(0.04)(0.05) lagtobinq0.008710.03990.006900.0310 (2.35) (2.36) (1.73) (1.72) devlogassets0.01090.0561 (1.86) (1.89) devpctintan-0.0515-0.273 (-0.89)(-0.91) devpctliab0.01430.0694 (0.52)(0.53) devpctrndrev0.0000001930.0000124 (0.00)(0.02) devlagpm0.00002400.000152 (1.63)(1.07) devlagroa-0.00223-0.0115 (-0.11)(-0.13) devlagtobinq0.01030.0487 (2.58) (2.56) Intercept0.223-1.222-0.377-1.4840.1780.163 (6.57) (-7.09) (-4.83) (-1.80) (2.10) (0.30) IndustryNoneNoneFEFEMeanMean YearNoneNoneFEFEFEFE N 217521752175213321752174 R 2 0.00350.03820.0177 t statisticsinparentheses p<: 10, p<: 05, p<: 01 Tables12and13presenttheresultsofourregressionmodelswhentheyareapplied toourtwomeasuresofschoolprestige.Inboth,weseeapositivesizewhen 60 Table2.13DeterminantsoftheCEOGraduatingfromaTopSchool (1)(2)(3)(4)(5)(6) OLS1Logit1OLS2Logit2OLS3Logit3 logassets0.004040.02220.01040.0606 (0.82)(0.79)(1.97) (2.01) pctintan-0.0431-0.256-0.0854-0.500 (-0.86)(-0.86)(-1.57)(-1.55) pctliab0.0006840.003290.008850.0504 (0.03)(0.03)(0.36)(0.37) pctrndrev0.00001370.00008570.00004210.000243 (0.11)(0.13)(0.31)(0.37) lagpm0.00001440.0001110.00001750.000119 (1.17)(0.77)(1.25)(0.83) lagroa-0.0182-0.0938-0.00781-0.0382 (-0.97)(-1.04)(-0.41)(-0.42) lagtobinq0.006560.03290.005410.0267 (1.83) (1.95) (1.45)(1.53) devlogassets0.01160.0673 (2.12) (2.15) devpctintan-0.0775-0.466 (-1.40)(-1.41) devpctliab0.001970.0103 (0.08)(0.07) devpctrndrev0.00004420.000254 (0.32)(0.38) devlagpm0.00001740.000117 (1.17)(0.82) devlagroa-0.0121-0.0615 (-0.61)(-0.64) devlagtobinq0.007780.0398 (1.96) (2.05) Intercept0.191-1.416-0.347-1.4750.1780.193 (5.92) (-7.78) (-4.69) (-1.78) (2.25) (0.34) IndustryNoneNoneFEFEMeanMean YearNoneNoneFEFEFEFE N 217521752175212821752174 R 2 0.00310.03580.0150 t statisticsinparentheses p<: 10, p<: 05, p<: 01 61 includingindustryandyearectsalongwithapositiverelationshiptothelagged Tobin'sQwhenexcludingtheseThesizestrengthenswhenweapplythe stricterofprestige,whiletheyweakenswhenapplyingthe strictertion.OnecouldsuggestthataCEOfromaprestigiousschoolislikely aluxuryitemforaboard.Asthegrowslarger,theboardisabletoacquiresuch luxuriesmoreeasily.Likewise,theCEO'sschoolbeingsupremelyprestigiousisless wellpredictedbythelaggedTobin'sQthanifitissimplyprestigious.Atminimum, thissuggeststhattheCEO'scollegeisintendedtoengenderrespectability. AsthemostcommondegreeheldbyCEOs,wemightwanttoseethedirectimpact oftheprestigeoftheCEO'sMBAgrantingschool.WeexplorethisinTable14,where weusethedummyindicatingtheCEOholdsaprestigiousMBAasthedependent variableinourmodels.Weseeasignitandpositiveofsize,suggesting thatlargearemorelikelytohireaCEOthatholdsaprestigiousMBAthan theirsmallerpeers.Thiscanbeseenasamutuallybrelationship,asthe hasgreaterresourceswithwhichtohireaCEOandaprospectiveCEOwishes tomaximizeherearningpotential.IntheOLSspweseeapositiveand tcotonthelaggedmargin,suggestingthat aremorelikelytohireaCEOwithaprestigiousMBA.Thiscorroboratesthe notionthatthechoiceofCEOismoreamatterofopportunityorabilityratherthan need. Table15extendsthisnotionbyfocusingfurtherontothoseCEOswhoholddegrees fromeliteMBAinstitutions.Weseethattheofthesizeweakens. Muchlikethegeneralcaseofourprestigevariables,weseethatthemore subsetisnotsoughtafterasurgentlyasthefullset.Onecouldinterpretthisas themarket'srecognitionthatthereislittleadditionalbetohiringtheseelite candidates,especiallywhenconsideringthelikelylaborcost Wealsoseethatthelaggedmarginismoreconsistent,asitoccursin 62 Table2.14DeterminantsoftheCEOHoldinganMBAfromanEliteProgram (1)(2)(3)(4)(5)(6) OLS1Logit1OLS2Logit2OLS3Logit3 logassets0.01200.07610.01560.101 (2.62) (2.65) (3.09) (3.16) pctintan-0.0262-0.186-0.0702-0.482 (-0.56)(-0.59)(-1.37)(-1.38) pctliab0.02080.1510.02530.179 (0.92)(1.10)(1.11)(1.26) pctrndrev0.0001550.001010.0001550.00102 (0.87)(1.34)(0.82)(1.30) lagpm0.00002500.002490.00002510.00260 (2.37) (0.88)(2.20) (0.90) lagroa0.00113-0.01400.007930.0307 (0.08)(-0.14)(0.60)(0.31) lagtobinq0.0006830.00643-0.0001860.000437 (0.27)(0.39)(-0.07)(0.03) devlogassets0.01570.100 (3.01) (3.01) devpctintan-0.0603-0.420 (-1.15)(-1.19) devpctliab0.02530.184 (1.05)(1.28) devpctrndrev0.0001420.000963 (0.78)(1.29) devlagpm0.00003810.00282 (2.92) (1.01) devlagroa0.01020.0490 (0.74)(0.43) devlagtobinq0.002410.0183 (0.81)(0.96) Intercept0.112-1.958-0.355-2.488-0.0302-1.301 (3.86) (-10.17) (-5.13) (-2.61) (-0.40)(-2.27) IndustryNoneNoneFEFEMeanMean YearNoneNoneFEFEFEFE N 217521752175213221752174 R 2 0.00480.03260.0161 t statisticsinparentheses p<: 10, p<: 05, p<: 01 63 Table2.15DeterminantsoftheCEOHoldinganMBAfromaTopProgram (1)(2)(3)(4)(5)(6) OLS1Logit1OLS2Logit2OLS3Logit3 logassets0.007620.06340.008510.0712 (1.94) (1.91) (1.94) (1.94) pctintan-0.00445-0.0890-0.0478-0.462 (-0.11)(-0.24)(-1.07)(-1.14) pctliab0.01180.1170.01270.115 (0.64)(0.75)(0.68)(0.70) pctrndrev0.0001870.001590.0001900.00159 (1.07)(2.11) (1.06)(2.06) lagpm0.00002410.01180.00002690.0122 (2.14) (2.05) (2.26) (2.08) lagroa0.00151-0.06070.00798-0.00419 (0.13)(-0.54)(0.69)(-0.04) lagtobinq0.001790.01920.001610.0172 (0.75)(1.08)(0.65)(0.92) devlogassets0.01020.0876 (2.24) (2.23) devpctintan-0.0442-0.443 (-0.96)(-1.07) devpctliab0.009610.0984 (0.50)(0.58) devpctrndrev0.0001880.00157 (1.07)(2.10) devlagpm0.00003460.0118 (2.62) (2.04) devlagroa0.009910.0128 (0.84)(0.10) devlagtobinq0.004040.0388 (1.53)(1.89) Intercept0.0762-2.356-0.225-2.247-0.0115-1.710 (3.01) (-10.40) (-3.84) (-2.24) (-0.18)(-2.54) IndustryNoneNoneFEFEMeanMean YearNoneNoneFEFEFEFE N 217521752175211421752174 R 2 0.00340.02590.0114 t statisticsinparentheses p<: 10, p<: 05, p<: 01 64 boththeOLSandlogitspThisaddstothenotionthattheseeliteMBA holdersareviewedasluxuries,aswenowhaveamoresolidrelationshipbetween positiveresultspriortotheturnovereventandthenewCEO'sdegree.Also ofinterestisthepositiveandtoftheR&Dactivityinthelogit spThisimpliesthatperhapsdobelievetosomedegreethatthese eliteMBAholdersarebetterequippedtohandlemoresophisticatedissuesthatmight arisewithheavyresearchinvestment. Finally,Table16presentstheresultsofourmodelsusingadummyrepresenting whethertheschoolisconsideredan\elite"educationalinstitution.Recallthatthis meansthattheschoolbothour rich dummyaswellasour prestige2 dummy. Here,weseeessentiallynotsexceptforsize,whichexhibitsapositive relationship.TherearetwopossibleinterpretationsofthisFirst,onemight assumethatlargersimplyhavetherequisiteresourcestoattractthestudentsof theseinstitutions.Inanopenmarket,theyaresimplyabletobidhigherthantheir smallercounterparts.However,onemightalsosuggestthatthesearesimply chasingtheprestigethathiringsuchcandidatescanbring.Theboardmight believethatthecompany,basedonitssize,mustaim\higher"whenhiringexecutives. Itistoseparatethetwoscenarios. Hence,weseethattherearesomeconsistentrelationshipsintherealmofCEO schoolquality.Largely,itappearsthatthereisadesireforsomelevelof qualityintheCEO'straining.Beyondthispoint,though,thistrainingseemstobe viewedasanicetyratherthananecessity. 2.6 Conclusions Weseethatachoiceofexecutiveseemstobemotivatedbyitsown cialcharacteristics.Thisquestionsthecurrentlyestablishedliterature,whichtacitly assumesanoverallidealCEOeducationforallBasedonitsanalysisofde- 65 Table2.16DeterminantsoftheCEOGraduatingfromanEliteSchool (1)(2)(3)(4)(5)(6) OLS1Logit1OLS2Logit2OLS3Logit3 logassets0.004080.02930.008930.0684 (0.93)(0.91)(1.88) (1.97) pctintan-0.0260-0.204-0.0660-0.504 (-0.57)(-0.58)(-1.33)(-1.33) pctliab-0.00441-0.03310.003790.0270 (-0.19)(-0.19)(0.16)(0.15) pctrndrev0.00006030.0003600.00008060.000505 (0.45)(0.53)(0.58)(0.73) lagpm0.000003990.00002800.000006710.0000416 (0.26)(0.20)(0.39)(0.30) lagroa-0.0108-0.0731-0.00356-0.0206 (-0.75)(-0.82)(-0.24)(-0.22) lagtobinq0.004730.02950.002770.0162 (1.59)(1.79) (0.88)(0.92) devlogassets0.009700.0728 (1.99) (2.02) devpctintan-0.0640-0.496 (-1.27)(-1.28) devpctliab0.002240.0137 (0.09)(0.08) devpctrndrev0.00006560.000397 (0.49)(0.57) devlagpm0.00001170.0000728 (0.68)(0.52) devlagroa-0.00661-0.0473 (-0.43)(-0.50) devlagtobinq0.005070.0325 (1.55)(1.70) Intercept0.130-1.863-0.233-2.4010.136-0.483 (4.62) (-9.07) (-3.44) (-2.36) (1.87) (-0.73) IndustryNoneNoneFEFEMeanMean YearNoneNoneFEFEFEFE N 217521752175212021752174 R 2 0.00210.03240.0135 t statisticsinparentheses p<: 10, p<: 05, p<: 01 66 gree,schoolcharacteristics,andschoolqualityonavectorofsize, characteristics,andtability,weareabletoteaseoutsomenovelresults. Thus,weclaimthatthisstudy'smannerofinquiryismuchmoreappropriatetothe empiricalrealityoftheexecutivehiringmarket.Thisstudyshouldserveasaninitial inquiryintothisphenomenon,allowingfuturestudiestoexpandfocustomore characteristics,aswellasaddressingtheperformanceofthatfollow(varyfrom) \normal"hiringpractices. 67 CHAPTERIII TheRoleofOwnershipinExecutiveLabor Markets 3.1 Introduction TheCEOlabormarketrepresentsaofmultipleNumer- ousstudieshaveapproachedissuesrangingfromthedeterminationofwagestojob- hopping.Ingeneral,though,thisscholarshiphaslargelyfocusedonthedemandside ofthecontractingrelationship,asitisgenerallyeasiertoconceptualizetheissues thatthefaceswhenattemptingtorecruitexecutivetalent. Thisstudyattemptstoreversethetypicalparadigm,however,byapproaching theissuefromthesupplyside.Thatis,weapproachthehiringmarketfromthe perspectiveofthepotentialCEO.Indoingso,wehopetoprovokenewlinesof inquiryinfuturescholarshipthatwillfurtherenrichourunderstandingofthebilateral executivecontractingenvironment. Toprovidethis,weassembleanoveldataset,combiningavariedsetofdata sourcesinordertodiscusschangesinthepotentialCEO'seducationalbackground. Sp,webroachtheissueofblockholders,andweexaminetheimpactthatthey mighthaveintheabilitytorecruitaCEOwithaparticularbackground.We tevidencetoarguethatpotentialexecutivesdoappeartobeimpactedby thepresenceorabsenceofablockholder,aswellastheparticulartypesofblockholders thatamighthave. 68 3.2 LiteratureReviewandMotivation Therearelargeliteraturesonbothexecutiveeducationalbackgroundsandexec- utivelabormarkets.However,itappearsthatthereislittleinteractionbetweenthe twolinesofscholarship.Wehopetoillustrateacaseinwhichthetwoactuallydo interact. 3.2.1 CEOBackgrounds Theliteratureontheeducationofexecutiveslargelyfocusesontheimpactoftheir backgroundontheirdecision-making.Thus,manystudieshavefocusedonwhatde- greestheexecutivespossess.Baruch(2009)notesthatMBAsdominatemanagement positionsforlargeinternationalasindividualswithanMBAholdoverhalfthe CEOpositionsattheseFurther,Baruchdocumentsthatthisislargelydue totheseactivelyseekingMBA-holdersformanagementpositionsorusingthe possessionofanMBAasapromotiongatingmechanism.Thisrelationshipmight bechanging,however.Dataretal.(2010)ndthatareincreasinglyrecruiting non-MBAs.NotonlyareturningawayfromMBAs,buttheauthorsdocument adramaticfallinMBAenrollments,especiallyatlesserprograms. Thequestion,ofcourse,becomeswhatpracticalethedegreeheldbytheCEO hasontheactivities.Severalstudieshaveinvestigatedthispoint.Finkelstein andHambrick(1996),TylerandSteensma(1998),andBarkerandMueller(2002) evidencethatinvestmoreheavilyinR&DwhentheirCEOhasatechnicaldegree versusabusinessorlawdegree.SubsequentstudiesdocumentbforCEOs holdingabusinessdegree,citingtheirunderstandingofmoresophisticate models,suchasthecapitalassetpricingmodelandthenetpresentvalue.These studiesincludedGrahamandHarvey(2001,2002)aswellasGraham,Harvey,and Rajgopal(2005). 69 However,thesespecializationbmightbeoverstated.Mian(2001)argues thatthesearemoreacutefortheCFOposition,astheCEOhasamuch broaderrangeofresponsibilitiesandactivities.Meanwhile,Iqbal(2015)examines theoilindustry,andhethatCEOswithbusinessdegreeshedgedlessthantheir peerswithindustry-spdegrees.Iqbaldoesthough,thatthereislittleto noineducationalqualityorfocusforCFOsbetweenthosewhohedgeand thosewhodonothedge.ThisislikelybecausealmostalltheCFOsinthestudyhold businessdegrees. AnotherpointoffocusisthequalityoftheCEO'sdegree.Burt(1992)and Belliveau,O'Reilly,andWade(1996)thatschoolselectivityprovidesnetworking btograduates,whichcanthroughtoperformance.Otherstudies changethehypothesisslightly,supposingthattheschool'squalityisasignalofthe CEO'sinnateability.ThesestudiesincludeFreyandDetterman(2004)andDeary (2004).Moregeneralstudieshavefoundbfromqualityofeducation,asPerez- Gonzalez(2006)documentsimprovedperformancerelatedtotheCEOshavingan IvyLeaguedegreeandMaxametal.(2006)thateducationalqualitycandrive hedgefundmanagerperformance. Therearecontrarianstudies,though.GottesmanandMorey(2010)andBhagat etal.(2011)norelationshipbetweenperformanceandtheCEO'seducation. andJonson(2013)takeamoresplookattheCEO'seducation, andtheyalsonoonperformance.Thiscouldbeexplainedbythe ofO'Leonard(2014)thatspentapproximatelyone-quarteroftheir trainingdollarsonleadershipdevelopment.Moreover,Jalbertetal.(2002)actually anegativeimpactoftheCEO'seducationalbackgroundandthereturn onassets.Moreconcretely,BarkerandMueller(2002)demonstratethatMBAsare sub-optimallyriskaverse. Regardlessofthethough,weseethatthereisgreatinterestintheeduca- 70 tionalbackgroundofthemanager.Itisreasonabletoassumethataconsiders theprospectiveCEO'seducationalbackgroundwhenmakinghiringdecisions.This meansthatrecruitswillbeabletomarkettheirskillsintheexecutivelabormarket, extractingapremiumfortheirThisstudyattemptstoteaseoutthe jobchoicestheseindividualsaremaking. 3.2.2 ExecutiveLaborMarket HavingestablishedtheimportanceoftheCEO'seducation,weturntothelitera- tureonCEOlabormarkets.Thecontractingliteratureoftenfocusesonthesituation inisolation,asinEdmansandGabaix(2009).However,alineofresearchhasin- corporatedthecontextoftheexecutivelabormarket.Thesestudieshaveincluded Frydman(2007)andMurphyandZabojnik(2004,2007).Thisisthough,as thereisagreatdealwedonotknowaboutthehiringprocessfortheseindividuals. Infact,OyerandSchaefer(2011)callitablackbox. However,researchhashighlightedsomeaspectsofthecontractingrelationship. AbowdandAshenfelter(1981)thatinindustrieswithmorejobriskmust paymorethanthoseinindustrieswithoutthatrisk.Kaplan(2008)furtherlinks dismissalriskwiththeexecutive'scompensation,thathigherdismissalrisk leadstohighercompensation.KaplanandMinton(2012)followthisbydocumenting adecreaseinCEOtenures,suggestingthatCEOsarebeingpaidmorewhileholding theirpositionforashorteramountoftime.Moreover,FeeandHadlock(2004) thattheseCEOsresurfaceatsmallerandearntlylessattheirnew job. Therearecomplicationsinthisrelationship,though,asEisfeldtandKuhnen (2013)andPetersandWagner(2014)thatindustryshockscanresultinaCEO's characteristicsnolongerattheircurrentemployer.Forexample,technology shockscanrenderaCEO'scompetencieslargelyobsolete.Hence,thesestudiesde- 71 velopmoresophisticatedrelationshipsbetweendismissalriskandcompensation. ThisrelationshipmightnotbelimitedtotheCEO'sskills,asitappearstoex- tendtotheirattitudesandperceptionsaswell.Sp,thereappearstobe somemoderatingintheselectionofCEOs.Campbelletal.(2009)that moderatelyoptimisticCEOsarelesslikelytobeforcedoutthanhighlyoptimisticor minimallyoptimisticCEOs.Hackbarth(2008)andGoelandThakor(2008)predict similarequilibriaforCEOovProspectiveexecutivesshouldanticipate thismeaningthattheywouldtendtowardthatprovidespacefortheir optimism.Thiscouldresultinatendencytowardlike-mindedblockholdersoraway fromblockholdersatall. Wealsoseethataindustryhasanonitspositioninthelabormarket. DengandGao(2013)andCremersandGrinstein(2013)establishthatmanagershave highermobilityinhighlypopulatedorhighlyhomogenousindustries.Accordingto Gaoetal.(2015),thismobilityispositivelyassociatedwiththepayraiseswonby \jobhopping"CEOs. Howdowedrawthetwolinesofscholarshiptogether?Occasionalstudieshave bridgedthegap.Montgomery(1991)andRees(1966)thatemployeesinthesame generallysharesocialties,includingfriendsandformercolleagues.Moreover, employeereferralscanaidinthehiringprocess.Onecouldimmediatelyconnectthis tothesocialnetworkscreatedbyacollegeenvironment. Anotherlineofinquiryinthemanageriallabormarketregardstheeofdirec- torstomanagetheirreputations.Levit(2012),Ruiz-VerduandSingh(2014),Song andThakor(2006)establishtheincentivesassociatedwithdirectors'reputational concerns,whileBouvardandLevy(2013)andBar-IsaacandDep(2014)arguethat directorsareconcernedwiththeirreputationwithtwoseparateaudiences,managers andshareholders.LevitandMalenko(2015)buildonthisbyshowinghowbuilding acertaintypeofreputationincreasesthemarketvalueofthatreputation.Itseems 72 clearthatthesedirectorsmightwanttoselectaCEOwithaparticulareducational backgroundinordertosatisfythesereputationalconcerns. 3.3 TheSample Weconstructanovelsamplefrommultipledatasources.Ataglance,wemerge ownershipdatafromFactsetwithcialdatafromtheCRSP/CompustatMerged databaseandthenmergetheresultingdatawithBoardexexecutiveeducationdata. Wewilldescribetheprocessinvolvedinmergingthesedatasourcesbelow. 3.3.1 FactsetActiveCompanies First,wecollecteddataontheownershipofactiveusingtheFactsetOwn- ership2.0interface.Thisallowsustogatherdataatanannualfrequencyregarding alltheholdersofanycurrentlyactiveThedataareobtainedalevel,sothe separatemustbecleanedandmerged.Thedatahaveseveralconsistencyissues thatneedtoberesolvedbeforethemergecantakeplace.Primarily,theheadersmust beremoved.Next,manyvariablelabelsaremissingtherelevantdate.Inordertouse thedataproperly,wemustbaclthesedatesusingthedateinthevariablelabel immediatelyprior.Basedonthedatastructure,thisinterpretationmakessense:for example,oneseesthepercentageownershipinMay2005,followedbythechangein percentageownership,followedbythepercentageownershipinMay2006,followed bythechangeinpercentageownership.Thus,thechangevariableisMay2005 andthesecondisMay2006. Onceotherminorcleaningoperationsarecompleted,thedataarereadytobe andmerged.Thisisaccomplishedbytranslatingeachvariableintoa genericvariableandmovingitsdatecomponenttoaseparatedatevariable.So,our priorexampleofpercentageownershipinMay2006becomessimplythepercentage ownershipwithadatevariablesettoMay2006.Oncethemergeiscompleted,the 73 activedataisreadytobemergedwiththeremainingFactsetdata. 3.3.2 FactsetInactiveCompanyIndividuals Next,wecollectdataontheownershipofinactiveusingtheFactsetOwn- ership3.0interface.Here,weareabletogatherlargelyequivalentinformationaswe haveonallownersoftheactivemsataquarterlylevelforindividualblockholders. Thismeansthatasimilarcleaningproceduremusttakeplace,withafew Inthiscase,thedatesareconsistentinthesensethattheyareeithergivenforan entiresetofobservationortheyaremissing.Weareable,then,to imputewhatthenextdateshouldbebasedonthepreviousdateandtheknowledge thatthedataarequarterly.Hence,weareabletoinamissingSeptember2005 dategiventhatthepriordatewasMay2005. Anotherproblemthatoccursinthisdataisduplicationofvariables.Thatis,we mighthavethesamevariableoccurtwiceinagivene.Wedeviseanalgorithm todetectanddropanyrepeatedvariablelabelspriortoassigningvariablenames. Oncethisisaccomplished,weareabletoproceedasabove.Then,theinactive individualblockholderdataisreadytobemerged. 3.3.3 FactsetInactiveCompanyInstitutions WeobtaintheFactsetCompanyInstitutionsfromWRDS.Thiscontains allinstitutionalholdings,forboththeactiveandtheinactiveSincewealready havetheactiveobservations,wemustidentifyonlythoseobservationsforthe inactiveImportantly,thisdatacontainsuniqueidenforeachin additiontothetickersandCUSIPsthatoccurinthepreviousdata.Hence,wemerge thisdatasetwiththeinactiveindividualsonthetwoseparateCUSIPvariables andtheticker.Bydoingso,weareabletogeneratealistofinactive idenWemergethislistbackintotheinstitutionaldatasothatweareleftwith 74 onlytheinactiveinstitutions.Finally,weattempttogeneratethepercentageowned foreachobservation,asthisisreportedintermsofsharesheld.Unfortunately, thereportedsharesoutstandingdataisspottyatbest.Atthispoint,wesetthisdata asideinordertoproceedon. 3.3.4 FactsetInactiveCompanyFunds WeperformasimilaroperationontheFactsetCompanyFundsfromWRDS. Thiscontainsallfundholdingsthatarenotreportedona13-F.Weperformthe samemergeoperationinordertolimitourattentiontoonlythoseholdingsininactive Next,wemergeintheFactsetFundNamesfromWRDSwhich{inaddition tocontainingthefunds'names{alsoincludesparentcompanyidenersforthefunds. Havingdoneso,weconsolidatetheobservationsbyaddingthesharesheldbyeach fundtocalculatetheparent'stotalholdings.Then,wekeeponlyasingleobservation fromeachparentbeforeattemptingtocalculatethepercentageowned.Asbefore, thisisalargelyfutileduetotheunreliablesharesoutstandingvaluesFactset contains. 3.3.5 CRSP/CompustatMerged WeusetheCRSP/CompustatMergeddatasetonWRDSasourbaseforthe variousmergeoperations.Weareabletoobtainameasureofsizefromthisdata, andwealsousethefullydilutedsharesoutstandingvaluesfromthisdatainorder tocalculatethepercentownedbytheinactiveinstitutionsandfundsfromFactset. Moreover,thisdataprovidesausefulstructuretofacilitatethemergebetweentwo fairlyunstructureddatasetsinFactsetandBoardex. 75 3.3.6 BoardexCompanyRecords ThewerequirefromBoardexistheCompanyCharacteristicsThis containstheidenthatarenecessarytomergethedatawithanoutside datasource.Sp,werelyontheISINsreportedinthisISINsare12digit companyideners.WhiletheyarenotpresentinCRSP/Compustat,theycontain theCUSIP.Sp,thethirdthroughtheeleventhdigitrepresentsthe CUSIP.So,weextractthethirdthroughtheeighthdigitinordertogenerate thesixdigitCUSIP(thesixdigitsformtheuniqueidenAt thispoint,wearereadytomergethisdata. 3.3.7 BoardexPersonRecords Next,weworkwiththeCharacteristicsfromBoardex.Here,wea tamountofdataregardingacorporatepositioninthem.This includesboththebeginningandendingtimeofthemholdingthatposition.These datesarelessthanconsistent,sowemustcleanthisdata.Indoingso,weassume thebroadestpossibledaterangeintheabsenceofinformation.So,anymissing beginningyearsareassumedtobe1990.Likewise,missingbeginningmonthsare assumedtobeJanuary,whilemissingbeginningdaysareassumedtobetheof themonth.Conversely,missingendingyearsareassumedtobe2013,whilemissing endingmonthsareassumedtobeDecember,andmissingendingdaysareassumedto betheappropriatenumberofdaysforthemonthgiven(e.g.,Decemberasanending monthwouldresultinthisvaluebeing31).Thisdone,thedataarereadytobe merged. 3.3.8 BoardexEducationRecords WealsousetheDirectorEducationsfromBoardex.Here,wedataonthe educationalbackgroundofthemajorityoftheBoardexuniverse.Itisimportantto 76 note,however,thatthisdataisnotexclusivetoearneddegrees:bothprofessional organizationsandhonorarydegreesareincluded.Thisdataisorganizedattheperson- degreelevel,meaningthattherecanbeupto27observationscorrespondingtoasingle person.Hence,wecompresseachperson'sdataintoasingleobservationbycreating 27copiesofeachvariable.Atthispoint,itisreadytobemerged. 3.3.9 CombiningtheData First,weappendthefourFactsetdatasetswiththisoneinordertocreatea masterFactsetdataset.Wealsotakethisopportunitytoconsolidatethevarious variablenamesundersingleheadingsandbacanymissingidenasbestwe can.Next,weconstruct\linktable"frombothFactsetandCRSP/Compustat, retainingonlyavariablecontainingtheobservationnumberandthevariouscompany idenWeperformaCartesianmergeofthetwolinkingonfourvariables: twoCUSIPvariables,theticker,andthetickerwithanyremoved. Next,weappendtheseintoasinglelargematchThisobviouslyhasagreat numberofduplicatesinit.Incullingtheseduplicates,weprioritizeanymatches resultingfromtheCUSIPvariables,followedbymatchesfromthefullticker.However, wedoretainanymatchesthatresultfromthe\notickersiftherewasnomatch forthatobservationonapreferredvariable. Oncethisparingdowniscompleted,wemergethemasterFactsetwiththe link,givingeachFactsetobservationaGVKEY.Now,wemovetowardcompleting themergewiththeCRSP/Compustatdata.Inordertoaccomplishthis,wegenerate threecopiesofeachFactsetobservation,onewithayearoneyearpriortothe realyear,onewithayearequaltothecalendaryear,andonewithayear oneyearaheadofthecalendaryear.So,aMay2005observationwouldbeduplicated withonebeingassigneda2004year,onebeingassigneda2005year,and onebeingassigneda2006year.Thismanipulationallowsustoreasonably 77 performamergeonayearbasiswhileeliminatingthepossibilitythat amatchmightbediscardedbasedonatimingissue. Ofcourse,ourapproachresultsinalargenumberofduplicatematches.Inor- dertoresolvethis,werelyontherelationshipoftheFactset report date withthe CRSP/Compustat datadate .Wecalculatethenumberofdaysseparatingthetwo datesforeachobservation.Then,sortingbyFactsetobservationnumbers,wekeep theobservationwiththesmallestinthetwo.Thisshouldbetheonewith themostaccurateyearaswell.ThissolvestheduplicationissuefromtheFact- setside,butthereisstillduplicationfromtheCRSP/Compustatside.Hence,we performasimilarsortontheyearlevel,wherewekeeptheobserva- tionwiththesmallestinthetwodates.Finally,wecalculatethepercentage ownedbyeachholderusingthesharesoutstandingvaluesfromCRSP/Compustat. Now,wekeeponlythosethathaveavaluegreaterthan5%. Finally,wepreparetomergetheFactset-CRSP/CompustatdatawiththeBoardex data.Thisrequiresseveralsteps.First,wemergetheFactset-CRSP/Compustatdata withtheBoardexcompanydataontheCUSIPlevel.Fromhere,wemergeinthe Boardexmanagerlistingsby companyid .Wedropanyobservationsthatareoutside theboundssetbytheCEO'sbeginningandendingdates.Last,wemergeourdata withtheBoardexeducationdataonthe directorid level.Theresultofthismerge representsoursample. 3.3.10 Variable WewishtoquantifyboththecharacteristicsoftheCEO'seducationalbackground andtheownershipstructure.Weseethattheownershipstructureisa muchsimplermatter. First,wecreateadummyvariable hasblock thatsimplycaptureswhetherthe hasaepercentblockholderatall.Next,weconstructasetofdummiesclassify- 78 ingthetypesofblockholdersthemighthave.Webasethesedonthe holder type variablesuppliedbyFactset.Themostcommonoftheseis individual , whichindicatesthatFactsetthisholderasanindividual.Likewise, invest- mentco indicatesthatthehasaninvestmentcompanyasablockholder, asFactsetclassifyingthemasaninvestmentadvisor,wealthmanagementora similarmoniker.Thedummy indicatesaFactsettionasamore broadalsuchasaninsurancecompanyorbank,while non-pr indi- catesthatFactsettheholderasaorganization,agovernment,or acharity.Wealso subordinate toshowthatFactsethaslabeledtheholderas asubsidiary,jointventure,orothernon-parententity.Toroundoutour tions, connotesaFactsettypeofpubliccompany,privatecompany,ora similarlygenericlabel. Finally,we numblockholders tobethetotalnumberofblockholdersthatthe has.Wearguethatthisisequalorsuperiortothetotalpercentageowned,aswe believethatthemerepresenceofablockholderisamajorissueforaprospectiveCEO, regardlessofhowlargetheparticularblockmightbe.Thiscompletesourownership variableconstruction. Now,wemoveontotheeducationdata.First,wewishtoclassifythetypeof degreethattheCEOhas.WetabulatealistofdegreetypesintheBoardexdata.We thatthereare330degreetypes,whichweclassifyintoeightcategories.Forthe purposesofthisstudy,wefurtherlimitattentiontoanarrowerrangeofcategories, lookingspatgraduatedegrees.Thesecategoriesinclude gradschool ,which simplyproxiesforthepresenceofagraduatedegree, masters ,whichproxiesforthe presenceofamaster'sdegree,and doctor ,whichproxiesforthepresenceofadoctoral degree.Also,withinthemaster'sdegreecategory,webreakoutthespdegreesof anMBA( mba )andalawdegree( law ).Intheeventthatagraduatedegreeisreported withoutaschoolname,weassumethatthepersonattendedtheirundergraduate 79 institutionforbothdegrees. WeassumeinthisstudythattheaverageCEOisapproximately50yearsold (whichthedataroughlyleadingustohand-collectdatafromHawes(1978) inordertotabulatevariousmeasuresofschoolqualityandschoolcharacteristics.The datafromHawesiswide-ranging,allowingustogatherinformationontheschool's selectivity,cost,quality,andmore.Moreover,webelievethatitismoreaccuratefor theCEOsinoursample,asitshouldbetterrepresentthestatusoftheschoolswhen theexecutiveswereactuallyattendingthem.Wewillgenerateseveralvariablesusing thisdata. First,wegeneratesimpledummies,suchas private ,whichisequalto1ifthe schoolisprivate.Likewise, foreign equals1whenaschoolisnotlocatedinthe UnitedStates.Wealsotabulatethevariable religious toindicateifaschoolcarries areligiousOtherwise,wea rich schoolasoneinthe75thorhigher percentileinourdatafortotalcost.Further,weuseHawes'suniquemeasureofsocial prestige(thenumberofalumniinthe SocialRegister )inordertoderiveameasure forsocialclass.Weconvertthismeasureintoaratio,thenweassignthoseinthe 75thorhigherpercentilesavalueof1for highclass . Next,wemoveontoschoolprestige.Wetwoprestigemeasuresbased onHawes'sdataonadmissionsyandselectivity.IfHawesratesadmissions yasa1(thehighestlevel)andselectivitygreaterthan85,we prestige1 tobe1.Alternatively,weallowadmissionsytodropto2ifselectivityis greaterthan90.Wealsoastrictermeasure, prestige2 ,whichtoughensthe standardsof prestige1 by5pointsinselectivityineachconditional.Wealso elite tobeadummytakingthevalueof1ifaschoolhasevaluesforboth rich and prestige2 .Finally,wemovetomoremodernratings,asweincorporate BusinessWeek 'srankingsofMBAprogramsinordertowhataprestigious MBAprogramis.WeaprestigiousMBAprogram(denotedby highmba )as 80 oneinthetop350inthecountry,whilewecategorizeaprogramaselite(denotedby supermba )ifitisinthetop10. 3.4 EmpiricalAnalysis Wewishtoexaminethemeansthroughwhichaownershipstructurecan itsabilitytorecruitCEOtalent.Wehypothesizethattherearethree primarychannelsthroughwhichthesecanbeseen.First,weclaimthatthe willbelesslikelytosuccessfullyrecruitexecutiveswithadvancedorspecialized degrees.Weexaminethisthroughthepossessionofagraduatedegree.Second,we arguethatthewillhavetorecruitoutsideoftypicallypreferredcircles.Thatis, wealthyorsociallyelitecandidatesarelikelytochoosetogoelsewhere.Third,we expectthattheCEO'seducationasawholewillbeoflowerqualitythantheirpeers. Weutilizequalitymeasuresfrommajorcollegerankingsforthisanalysis. Thesethreemeasuresarelikelytobeviaseveralchannels.First,we assumethatthesizewilldriveitshiringsuccess,asitisreasonabletoassume thattheselargerareabletomoreattractivecompensationpackagesthan theirsmallercompetitors.Moretoourfocus,though,weusethreemeasuresof blockholding.First,weadummyforthepresenceofablockholder,aswe assumethatpotentiallyinterventionistownershipisanegativeaspectofapotential employer.Second,weformasetofdummyvariablesthatcontrolforthegeneral typeofblockholdersthatarepresent.Wehypothesizethatcertainblockholdersare likelytobemoreonerousthanothers,andwehopetoisolatetheseusingthese variables.Finally,welookatthenumberofblockholders,withahypothesisthat morelargeholderspresentresultsinalesspreferableenvironmentforaparticular CEO. Table1presentsthesummarystatisticsforourvariables.PanelAshowsthe statisticsforoureducationqualitymeasuresthatwillbeusedasdependentvariables, 81 Table3.1SummaryStatistics PanelA:DependentVariables VariableMeanStd.Dev.Min.Max.N law0.0730.26015908 doctor0.0970.296015908 mba0.3520.478015908 master0.1930.395015908 private0.4350.496015908 religious0.1410.348015908 foreign0.0640.245015908 prestige10.2810.45015908 prestige20.2370.425015908 gradschool0.6030.489015908 highmba0.1860.389015908 supermba0.1330.339015908 rich0.220.414015908 snooty0.2370.425015908 elite0.1760.381015908 PanelB:IndependentVariables VariableMeanStd.Dev.Min.Max.N hasblock0.3740.484015908 individual0.2330.423015908 investmentco0.0120.11015908 0.0040.064015908 0.0020.047015908 subordinate0.0360.186015908 0.0890.285015908 logassets5.571.7260.93413.4214500 82 whilePanelBtabulatesthestatisticsforourownershipmeasuresandthesize thatwewilluseasindependentvariables.Onecanseethatcertainmeasurespresent someculty,astheydonotpossessagreatdealofvariabilityinthesample.How- ever,webelievethatthesevariablesarestillusefultohighlightcertainrelationships. 3.4.1 RegressionFramework Wewishtoidentifytheofaownershipstructureonitsabilityto recruitexecutivetalent.Inordertodoso,wemaketheassumptionthatthequality ofaCEO'seducationcanbequanviathe15variableswehavened.Having doneso,wemoveontoaregressionanalysis.Wewilluseasetof3separateOLS regressionstoisolatethevariousownershipcharacteristicsoftheFirst,we attempttoisolatethethattheexistenceofablockholderhasonownership. Thisrequiresadummyindicatingthatahasashareholderofgreaterthan5% inagivenyear.Second,weattempttocharacterizetheimpactoft typesofblockholders.Todoso,weuseavectorofdummiesrepresentingcategories ofblockholders.Finally,weregressonthenumberofblockholdersasameansof teasingoutthe\size"ofablockholding.Here,weincludeavariablecounting thenumberofblockholders.Thisresultsinthefollowingsetofregressions: EducationalCharacteristic = b 0 + b 1 HasBlockholder EducationalCharacteristic = b 0 + b 1 HasBlockholder + b BlockholderType EducationalCharacteristic = b 0 + b 1 NumberofBlockholders Inallcases,wecontrolforthesizeofthembyincludingthelog-transformof theassets.Likewise,wecontrolforindustrybyincludinga atthetwo-digitSIClevel.Inaddition,wecontrolforanytimebyincluding 83 ayearInordertobetteraccountfornon-linearitythatmightarisein thesemodels,wealsoreporttheresultsofequivalentlogitmodels.Webelievethat thesemeasuresallowustoaccuratelyassesstheofownershiponexecutive recruitment. 3.4.2 DegreeType Themostdiscretemeasureofanexecutive'sisthepresenceofa degree.Bachelor'sdegreesareubiquitousenoughattheCEOleveltobelargely uninterestingforourpurposes.Instead,weconsiderwhethertheCEOhasagraduate degree.WeisolatehavingalawdegreeorMBA.Then,wemoveontomoregeneral categories,suchaspossessingamaster'sordoctoraldegree.Finally,weexaminethe generalcategoryofallgraduatedegrees. Weturnourattentiontowhethertheexecutivehasalawdegree.Wewould expectthatCEOcandidateswithlegaltrainingpossessuniqueskills,astheyhave receivedtspecializedtraining.Otherdegrees,suchasanMBA,couldbeat leastpartlysubstitutedforbyundergraduatecoursesorexecutiveeducationsessions. Thus,alawdegreeisthemostobviousplacetoseeaparticularskillsetmanifestits Table2showstheresultsofouranalysisusing law asthedependentvariable. Weseethatthesizeisthemostconsistentlytdriveroftheexecutive havingalawdegreethroughoutthesixmodels.Giventheirearningpotentialin non-executivepositions,wewouldassumethattheserecruitsarelikelydrawntothe increasedsalariesthatlargerarelikelytoHowever,theownershiptype regressiondoesshowsomeotherSpcally,weseethattheblockholderbeing anon-proorasubordinateunitofalargerorganizationresultinnegativecots intheOLSspItislikelythatanywithlargeblockholders mayhavetlimitationsplacedonthembythatrelationship.Unfortunately, 84 Table3.2DeterminantsoftheCEOHoldingaLawDegree (1)(2)(3)(4)(5)(6) OLS1Logit1OLS2Logit2OLS3Logit3 hasblock-0.00519-0.104-0.00107-0.0182 (-0.64)(-0.79)(-0.11)(-0.11) individual-0.0121-0.238 (-1.12)(-1.27) investmentco0.01460.231 (0.50)(0.54) 0.04590.583 (0.73)(0.89) -0.0601. (-3.84)***. subordinate-0.0286-0.594 (-1.99)**(-1.58) 0.02020.321 (1.60)(1.49) numblockholders-0.00177-0.0383 (-0.93)(-1.09) logassets0.003500.05390.003460.05450.003530.0541 (1.71)*(1.82)*(1.68)*(1.82)*(1.76)*(1.86)* Intercept0.102-2.3940.0987-2.4490.101-2.405 (0.71)(-1.61)(0.69)(-1.65)*(0.70)(-1.62) N 590858105908579659085810 R 2 0.03350.03460.0335 t statisticsinparentheses * p< 0 : 10,** p< 0 : 05,*** p< 0 : 01 85 wedonothavetobservationstoconthatthispersistsinthelogit spThenegativeofbeingheldbyasubordinateentitylikelyspeaks toasimilarlackofautonomy,althoughweareabletoseethatthisweakensin thelogitmodel. Table3.3DeterminantsoftheCEOHoldinganMBADegree (1)(2)(3)(4)(5)(6) OLS1Logit1OLS2Logit2OLS3Logit3 hasblock0.01110.05040.003670.0167 (0.73)(0.74)(0.19)(0.19) individual0.008300.0376 (0.38)(0.39) investmentco0.01490.0667 (0.26)(0.26) 0.09890.420 (1.02)(1.05) 0.09000.364 (0.70)(0.68) subordinate0.06940.303 (1.92)*(1.98)** -0.0265-0.119 (-1.07)(-1.07) numblockholders0.005620.0250 (1.49)(1.53) logassets0.009550.04280.009280.04160.009730.0436 (2.65)***(2.67)***(2.56)**(2.58)***(2.74)***(2.77)*** Intercept0.244-1.1240.249-1.1010.245-1.119 (1.12)(-1.17)(1.15)(-1.15)(1.13)(-1.17) N 590859065908590659085906 R 2 0.02560.02670.0259 t statisticsinparentheses * p< 0 : 10,** p< 0 : 05,*** p< 0 : 01 AnMBAdegree,ontheotherhand,isamuchmorecommontypeofdegree inabusinessenvironment.Inmuchthesamewaythatweomitanydiscussionof bachelor'sdegrees,itwouldbesurprisingtondmuchofanassomanyCEOs possessanMBAdegree.Table3essentiallythis,asweseelittle Thereissometobeingheldbyasubordinateentity,whichlikelyisa resultofworkingone'swayupthecorporateladder. Otherwise,weseeastronglytpositiverelationshipwithsize.This 86 hasatwo-sidedrelationshipwithexecutiveincentives.Primarily,onewouldbelieve thattheselargersimplyhavemoreresourceswithwhichtorecruitCEOtal- ent.Thatis,theycanhighersalariesorprovidemoreperquisitesthantheir competitors.However,onemightalsoarguethatlargerarelesslikelytohave ablockholder,suggestingthatthesizemightprovidesomeprotectionagainst futureblockholders.Giventhattheseexecutivespossessextensivebusinesstraining, itwouldstandtoreasonthattheywouldbemostlikelytodesiretorunthebusiness withoutinterferencefromlargeshareholders. Table3.4DeterminantsoftheCEOHoldingaMaster'sDegree (1)(2)(3)(4)(5)(6) OLS1Logit1OLS2Logit2OLS3Logit3 hasblock-0.00129-0.00770-0.00517-0.0353 (-0.10)(-0.09)(-0.32)(-0.34) individual0.005600.0375 (0.31)(0.32) investmentco-0.0157-0.112 (-0.33)(-0.34) -0.0314-0.311 (-0.49)(-0.49) -0.0555-0.415 (-0.58)(-0.53) subordinate-0.0200-0.134 (-0.71)(-0.70) 0.01810.122 (0.86)(0.94) numblockholders-0.00197-0.0131 (-0.66)(-0.64) logassets-0.00418-0.0280-0.00392-0.0260-0.00438-0.0293 (-1.42)(-1.42)(-1.32)(-1.31)(-1.51)(-1.51) Intercept0.0655-2.3430.0635-2.3580.0656-2.343 (0.38)(-1.65)*(0.36)(-1.66)*(0.38)(-1.65)* N 590858485908584859085848 R 2 0.04260.04290.0427 t statisticsinparentheses * p< 0 : 10,** p< 0 : 05,*** p< 0 : 01 Supposethatwewidenournet,though,toallmaster'sdegrees.Onewould assumetherewouldbesomeincentiveontheemployersidetohirecandidateswith master'sdegrees,astheyarelikelytohaveadvancedtraininginsomeareathat 87 wouldpresumablyimprovetheirhumancapital.Thisshouldprovidethepotential employeeswithsomeyinchoosingtheirposition. However,ourresultsdonotbearthisout.Instead,Table4demonstratesessen- tiallynorelationshipbetweenownershipandhavingamaster'sdegree.Weargue thatthisisduetotheheterogeneityinherentina\master's"category.Forexam- ple,whileanMBAdegreeholdermightpreferautonomy,itispossiblethatsomeone withanon-businessfocusedmaster'sdegreemightpreferamoreactivevoicefrom shareholdersinordertoprovidethemwitha\business"viewpoint. Table3.5DeterminantsoftheCEOHoldingaDoctoralDegree (1)(2)(3)(4)(5)(6) OLS1Logit1OLS2Logit2OLS3Logit3 hasblock-0.0152-0.171-0.0195-0.249 (-1.65)*(-1.55)(-1.69)*(-1.71)* individual-0.00361-0.0205 (-0.28)(-0.13) investmentco0.02490.203 (0.66)(0.57) 0.07940.928 (1.19)(1.66)* -0.0345-0.535 (-0.46)(-0.46) subordinate-0.0215-0.292 (-1.06)(-1.13) 0.03060.364 (1.97)**(2.20)** numblockholders-0.000313-0.00134 (-0.13)(-0.05) logassets-0.00614-0.0737-0.00603-0.0729-0.00536-0.0653 (-2.80)***(-2.80)***(-2.73)***(-2.74)***(-2.47)**(-2.50)** Intercept0.0166-2.3990.0145-2.3880.0137-2.527 (0.24)(-2.26)**(0.21)(-2.24)**(0.20)(-2.37)** N 590854055908540559085405 R 2 0.07970.08090.0793 t statisticsinparentheses * p< 0 : 10,** p< 0 : 05,*** p< 0 : 01 Doctoraldegreespresentamoreuniformformofdegree,though,soonewould believethatthesepotentialCEOsmighthavemorehomogenousincentives.Table 5illustratesexactlythis.Weseethatsizeisamajor,andnegative,driverof 88 potentialCEOshavingadoctoraldegree.Wewouldsupposethatthisisbecause smallerarebestabletoutilizethedoctoraldegreeholder'suniqueskillset. Onceagrowslarger,itbecomesunwieldyfortheCEOtobeinvolvedinday-to- dayfunctions,makingspecializedresearchtrainingtlylessuseful.Themost obviousreadingofthiswouldbethatdonotdesireaCEOwithadoctorate,but onemightarguethatthezealwithwhichdoctoraldegreeholdersapproachresearch makesthemwishtoavoidinsularbureaucraticenvironmentslikeamatureoften becomes. Moretoourfocus,though,weseesomeintheexistenceofablock. Sp,weseethathavingablockholderreducesthelikelihoodoftheCEOhaving adoctorate.Ontheotherhand,weseeapositiveandtrelationshiptothe blockbeingagenericcompany.Likewise,weseeabarelytpositive inthelogitspofhavingacompanyasablockholder.Onecould envisionthatthisistheresultofdoctorateholdersworkingatsmaller,research- orientedthatreceiveeitherventurecapitalinvestmentsorlargeinvestments fromend-users. Now,webroadenourfocustosimplyhavingadegreebeyondabachelor'sdegree. Thatis,theCEOpossessesanygraduatedegree.Wecouldcertainlyarguethatthese potentialCEOshavetvalueovertheirbachelor's-holdingpeers.However, Table6doesnotshowthattheyexercisetheassociatedinanyrealway.We seenotAswiththemaster'sdegreecategory,wearguethatthisis duetothelargemixofbackgroundsandtheirdincentives. Ultimately,weseeverylittletofownershipwhenlookingatbroadcategories. However,whenwespecifyandlookatparticulardegrees,suchasanMBAorlaw degree,oranarrowrangeofdegrees,likedoctoraldegrees,wedosome Thissuggeststhatpotentialexecutivesareexercisingsomeoftheirbargainingpower basedontheownershipstructure,butitalsosupportsthenotionthatdegrees 89 Table3.6DeterminantsoftheCEOHoldingaGraduateDegree (1)(2)(3)(4)(5)(6) OLS1Logit1OLS2Logit2OLS3Logit3 hasblock-0.0105-0.0461-0.0237-0.103 (-0.67)(-0.68)(-1.19)(-1.20) individual0.005240.0225 (0.23)(0.23) investmentco0.05460.254 (0.95)(0.93) 0.1280.585 (1.37)(1.28) 0.04890.239 (0.41)(0.41) subordinate0.0003410.00275 (0.01)(0.02) 0.03060.134 (1.22)(1.22) numblockholders0.001010.00440 (0.27)(0.27) logassets0.004470.01970.004630.02060.005170.0228 (1.23)(1.24)(1.27)(1.28)(1.44)(1.45) Intercept0.270-0.9760.271-0.9740.268-0.986 (1.23)(-1.07)(1.23)(-1.07)(1.21)(-1.08) N 590859065908590659085906 R 2 0.04210.04290.0421 t statisticsinparentheses * p< 0 : 10,** p< 0 : 05,*** p< 0 : 01 90 aretoowide-ranginginfocusandaptitudestodevelopanyconsistentbiases. 3.4.3 SchoolCharacteristics Anotherareainwhichthem'sownershipmightitsabilitytodrawCEO talentisbasedonthecharacteristicsoftheCEO'sschool.Thereareseveralofthese that,whilenotdirectlyrelatedtothequalityofthepotentialCEO'seducation,would undoubtedlyherattitudesandperceptionsabouttheworld.Sp,we investigatethestatusaspublicorprivate,thewealthofitsstudents,andthe socialstatusofitsstudents,aswellaswhetheritcarriesareligiousorifit isaforeignschool. Table3.7DeterminantsoftheCEOGraduatingfromaPrivateSchool (1)(2)(3)(4)(5)(6) OLS1Logit1OLS2Logit2OLS3Logit3 hasblock0.005400.02310.03080.128 (0.34)(0.35)(1.52)(1.53) individual-0.0302-0.125 (-1.34)(-1.34) investmentco-0.00558-0.0233 (-0.09)(-0.09) -0.109-0.484 (-1.19)(-1.11) -0.0538-0.229 (-0.42)(-0.40) subordinate-0.0229-0.0963 (-0.64)(-0.63) -0.0226-0.0935 (-0.88)(-0.87) numblockholders0.002720.0115 (0.70)(0.72) logassets0.005120.02140.004730.01980.005210.0218 (1.37)(1.39)(1.26)(1.28)(1.42)(1.43) Intercept0.199-1.5800.196-1.5920.199-1.577 (1.04)(-1.30)(1.03)(-1.31)(1.05)(-1.29) N 590858975908589759085897 R 2 0.02670.02760.0267 t statisticsinparentheses * p< 0 : 10,** p< 0 : 05,*** p< 0 : 01 Onewouldassumethatagraduateofaprivateschoolwouldacttlyin 91 thehiringmarketthantheirpublicschool-educatedpeers.Table7,however,shows thatthismightnotactuallybethecase.Weseenotinanyof ourmodels.Aswiththegeneraldegreecategoriesabove,wesupposethatthisis theresultofheterogeneityinprivateschools.Thatis,thereisawidespectrumof privateschools,greatlycomplicatingthemotivationsofprivateschoolgraduatesin theexecutivetalentmarket. Table3.8DeterminantsoftheCEOGraduatingfromaRichSchool (1)(2)(3)(4)(5)(6) OLS1Logit1OLS2Logit2OLS3Logit3 hasblock-0.00682-0.0431-0.00358-0.0242 (-0.53)(-0.54)(-0.22)(-0.24) individual0.0004720.00325 (0.03)(0.03) investmentco-0.0105-0.0909 (-0.23)(-0.29) -0.0434-0.308 (-0.59)(-0.55) -0.0417-0.275 (-0.43)(-0.35) subordinate-0.0508-0.347 (-1.91)*(-1.75)* 0.01070.0712 (0.52)(0.55) numblockholders0.0001790.00145 (0.06)(0.07) logassets0.008220.04760.008380.04840.008620.0500 (2.61)***(2.60)***(2.63)***(2.62)***(2.77)***(2.76)*** Intercept0.0134-1.5910.0114-1.6100.0121-1.626 (0.11)(-2.59)***(0.09)(-2.62)***(0.10)(-2.66)*** N 590858535908585359085853 R 2 0.02960.03010.0295 t statisticsinparentheses * p< 0 : 10,** p< 0 : 05,*** p< 0 : 01 Perhaps,then,weshouldfocusonaslightlyditaspect:howexpensivethe schoolis.Wegenerallyassumethatwealthierstudentsattendmoreexpensiveschools. Oneseesthatthiswouldroughlybecorrelatedwiththeprivateversuspublicresults, asprivateschoolsshouldbemoreexpensivethantheirpubliccounterparts.This wouldsuggestthatourearliersuppositionstillholdshere,leadingtheresultstowash 92 out. Infact,Table8showstintwoareas.First,weseepositive andstronglytcotsonsize.Moreover,weseeanegativeand tonhavingasubordinateentityasablockholder.Webelievethat thisillustratestheincreasedsalaryandperquisitesthatlargeareableto comparedtoothers.Perhapsmoreimportantly,thissuggeststhattheseexpensive schoolsareabletoprovidetnetworkingandtrainingfortheirgraduatesto leapfrogintolargerrightaway.Thatis,thesepotentialexecutivesareableto parlaytheireducationintomorelucrativepositions. Table3.9DeterminantsoftheCEOGraduatingfromaHighClassSchool (1)(2)(3)(4)(5)(6) OLS1Logit1OLS2Logit2OLS3Logit3 hasblock-0.0143-0.0858-0.0138-0.0815 (-1.07)(-1.11)(-0.80)(-0.82) individual0.003660.0194 (0.19)(0.17) investmentco-0.0587-0.419 (-1.36)(-1.24) -0.0511-0.349 (-0.68)(-0.61) -0.00547-0.0146 (-0.05)(-0.02) subordinate-0.000424-0.00276 (-0.01)(-0.02) 0.0004730.00330 (0.02)(0.03) numblockholders-0.000469-0.00271 (-0.14)(-0.14) logassets0.007750.04250.007800.04260.008450.0466 (2.38)**(2.37)**(2.37)**(2.37)**(2.63)***(2.62)*** Intercept0.0271-1.1360.0263-1.1470.0244-1.195 (0.21)(-1.97)**(0.21)(-1.98)**(0.19)(-2.08)** N 590858835908588359085883 R 2 0.02790.02820.0277 t statisticsinparentheses * p< 0 : 10,** p< 0 : 05,*** p< 0 : 01 Suppose,instead,thatweturnourattentiontothesocial\class"ofaschool. Table9reportstheofourusualownershipvariablesonwhethertheCEO 93 attendeda\highclass"university.Ultimately,weseethatthereislittleevidenceofa relationshipbetweenownershipanda\highclass"CEO.Wewouldarguethatthisis duetotwocompetingFirst,itisreasonabletobelievethatblockholdersare likelytocirculateinthesamecirclesastheseCEOs,givingtheiranadvantage inrecruitingthesecandidates.AcountervailingwouldbethattheseCEOsare likelytohavealternativethatwouldallowthemtoavoiddealingwithpotentially micro-managingshareholders,largelycancellingouttheinitialadvantagethat withblockholdersmighthave.Thisissomewhatsupportedbytheofthe positivesizeasthesecandidatesappeartochasetheincreasedopportunities thatthesecanprovide. Table3.10DeterminantsoftheCEOGraduatingfromaReligiousSchool (1)(2)(3)(4)(5)(6) OLS1Logit1OLS2Logit2OLS3Logit3 hasblock0.01240.1030.02990.243 (1.11)(1.11)(2.08)**(2.14)** individual-0.0230-0.186 (-1.44)(-1.47) investmentco-0.00923-0.0738 (-0.22)(-0.20) -0.0725-0.749 (-1.35)(-1.01) 0.01080.0975 (0.11)(0.12) subordinate-0.00985-0.0831 (-0.39)(-0.38) -0.0135-0.114 (-0.75)(-0.75) numblockholders0.005040.0405 (1.70)*(1.78)* logassets0.0006920.005780.0003880.003110.0007210.00592 (0.27)(0.27)(0.15)(0.15)(0.29)(0.28) Intercept0.177-1.5700.176-1.5830.179-1.555 (1.04)(-1.23)(1.02)(-1.24)(1.05)(-1.22) N 590858575908585759085857 R 2 0.01720.01800.0176 t statisticsinparentheses * p< 0 : 10,** p< 0 : 05,*** p< 0 : 01 HavingattendedareligiousinstitutionisanotherwayinwhichaCEO'seducation 94 couldimpacttheirattitudesandperceptions.Theseinstitutionsarelikelytobequite tthantheirsecularpeers.Table10showshowthisectmightmanifestitself. Inthesecondspweseethattheexistenceofablockholderresultsina beingmorelikelytorecruitaCEOfromareligiousschool.Moreover,weseethatthe numberofblockholdersispositivelyrelatedwiththeexecutiveholdingadegreefrom areligiousinstitution. WhatdrawstheserecruitstotheseWemightsuggestthattheseCEOs prefertoworkwithmorevisiblestakeholders.Thatis,someaspectoftheirexpe- riencewhileinschoolresultedinthembeingmorecomfortablewithblockholders versushavingapurelydistributedownershipstructure.Perhapsthesecandidates believethattheyareabletoworkwiththesepotentiallytroublesomeshareholders. Alternatively,perhapstheybelievetheseblockholdersholdlittlepower.Eitherway ofthinkingcouldconceivablyderivefromthecultureoftheireducationalinstitution. TurningtothenationalityoftheCEO'suniversity,weseeatTable 11illustratesasomewhatparadoxicalSplly,weseethatthepresence ofablockhasanegativeandtintheblockholdertypesp However,wehaveaslightlylargerinmagnitudepositivefromthepresenceof anindividualblockholder.Hence,foreign-trainedCEOsseemtobeaversetothe presenceofablockholderunlessthatblockholderisanindividual.Spe,wesee thatinvestmentcompanieshavealightlytnegativesuggestingthat theyaredoublyavoided. HowcanwereconciletheseOnemightsuggestthattheseforeign-trained CEOsdesiretheadvisoryvoicethatanindividualblockholdercanprovide,whereas blockholdermighttypicallybelessaggressiveinvoicingtheirconcerns.Alterna- tively,onecouldsuggestthatforeign-trainedCEOsforanynumberofreasonsmight viewindividualsaslessthreateningthanblockholders.Perhapstheymightview individualsasunlikelytoamassthevotesnecessarytothreatentheirposition. 95 Table3.11DeterminantsoftheCEOGraduatingfromaForeignSchool (1)(2)(3)(4)(5)(6) OLS1Logit1OLS2Logit2OLS3Logit3 hasblock-0.00738-0.119-0.0251-0.406 (-0.88)(-0.93)(-2.53)**(-2.42)** individual0.02930.435 (2.55)**(2.43)** investmentco-0.0443-0.770 (-1.79)*(-1.27) 0.01850.317 (0.35)(0.43) -0.0152-0.257 (-0.20)(-0.22) subordinate0.000780-0.0112 (0.04)(-0.04) 0.01400.195 (0.97)(1.04) numblockholders-0.00201-0.0312 (-1.01)(-1.00) logassets-0.000904-0.0142-0.000505-0.00800-0.000783-0.0115 (-0.49)(-0.45)(-0.28)(-0.25)(-0.44)(-0.38) Intercept0.0816-1.3210.0829-1.3190.0805-1.347 (0.92)(-1.67)*(0.94)(-1.69)*(0.91)(-1.72)* N 590855085908550859085508 R 2 0.02490.02680.0249 t statisticsinparentheses * p< 0 : 10,** p< 0 : 05,*** p< 0 : 01 96 Ultimately,weseethatthecharacteristicsofthepotentialCEO'seducationalin- stitutionhavemixedcts.Thecharacteristicsthataremostgenerally withqualitymeasureshavelittleifanyrelationshipwiththeownershipstruc- ture.However,wedoseesomeoftheownershiponthemoreindepen- dentcharacteristicsofreligiousaandnationality.Wefeelthatthisislargely indicativeofcountervailingonthesegraduates,though,ratherthanan absenceofimportancerelatedtoownership. 3.4.4 SchoolQuality Whiletheirdegreetypeandschoolcharacteristicscantellusagreatdealregarding theattitudesandperceptionsofapotentialCEO,thequalityoftheschoolislikelythe bestindicatoroftheirqualityasacandidate.Ofcourse,thisinformationis toquantify,butwecanstillapproximateitusingtherankingsprovidedby Business Week andotherpublications.Thisallowsustoapplyourregressionframeworkto theireducationalquality. InTables12and13,weconfrontthisissuedirectlybyapplyingourusualsp cationstoourtwomeasuresofschoolprestige.Here,weseeaconsistentlyt andpositivesizeHowever,thisectisclearlystrongerinthesetthan thelattersetofregressions.Certainly,thisislikelycausedinpartbythesmallnum- berofevaluesinourstricter prestige2 variable.However,wecouldalso supposethatthisillustratesadiscontinuityintheofhavingaprestigiousde- gree.Thatis,perhapsthosewithprestigiousdegreesunderourweakerdo chasethebofworkingforalargerbutwemightseethosewithextremely prestigiousdegreesbeingmorestronglybynon-pecuniaryaspectsoftheir potentialworkingenvironment. Moreinterestingly,weseeastronglytandnegativeofhavinga asablockholderintheOLStypespCoupledwiththe 97 Table3.12DeterminantsoftheCEOGraduatingfromaPrestigiousSchool (1)(2)(3)(4)(5)(6) OLS1Logit1OLS2Logit2OLS3Logit3 hasblock-0.00555-0.02900.003540.0187 (-0.39)(-0.40)(0.19)(0.20) individual-0.0215-0.118 (-1.07)(-1.11) investmentco-0.0572-0.351 (-1.24)(-1.17) -0.134-0.963 (-2.01)**(-1.53) -0.106-0.681 (-1.17)(-0.92) subordinate-0.0217-0.111 (-0.70)(-0.65) 0.04050.214 (1.75)*(1.81)* numblockholders0.002790.0146 (0.79)(0.80) logassets0.007400.03710.007610.03820.008090.0407 (2.19)**(2.20)**(2.23)**(2.24)**(2.42)**(2.43)** Intercept0.180-1.6320.172-1.6720.179-1.643 (0.94)(-1.26)(0.90)(-1.29)(0.93)(-1.26) N 590859005908590059085900 R 2 0.02930.03080.0294 t statisticsinparentheses * p< 0 : 10,** p< 0 : 05,*** p< 0 : 01 98 Table3.13DeterminantsoftheCEOGraduatingfromaTopSchool (1)(2)(3)(4)(5)(6) OLS1Logit1OLS2Logit2OLS3Logit3 hasblock-0.00444-0.02620.007190.0417 (-0.33)(-0.34)(0.41)(0.42) individual-0.0246-0.151 (-1.28)(-1.33) investmentco-0.0720-0.533 (-1.76)*(-1.55) -0.127-1.169 (-2.36)**(-1.60) -0.0656-0.437 (-0.73)(-0.59) subordinate-0.0207-0.120 (-0.71)(-0.66) 0.03660.219 (1.68)*(1.75)* numblockholders0.003210.0187 (0.95)(0.96) logassets0.005960.03300.006060.03360.006650.0369 (1.85)*(1.85)*(1.86)*(1.87)*(2.09)**(2.08)** Intercept0.0226-1.3850.0148-1.4180.0212-1.447 (0.19)(-2.34)**(0.13)(-2.39)**(0.18)(-2.46)** N 590858935908589359085893 R 2 0.02510.02670.0253 t statisticsinparentheses * p< 0 : 10,** p< 0 : 05,*** p< 0 : 01 99 negativeofhavinganinvestmentcompanyasablockholderinthe prestige2 spthisagainspeakstopotentialnon-pecuniarymotivations.Thatis,we wouldimaginethatthesetypesofshareholdersarethemostlikelytopromote\short- termism",atleastsuggestingthattheseexecutivesarenotchasingresults. However,wedoseeaweaklytandpositivefromhavingageneric companyasablockholder.Wewouldarguethatthesearelikelyaberrant,perhaps representingthepresenceoftheseexecutivesatsmaller,moredynamic Table3.14DeterminantsoftheCEOHoldinganMBAfromanEliteProgram (1)(2)(3)(4)(5)(6) OLS1Logit1OLS2Logit2OLS3Logit3 hasblock-0.00319-0.0225-0.000499-0.000365 (-0.26)(-0.27)(-0.03)(-0.00) individual-0.0125-0.0963 (-0.73)(-0.81) investmentco-0.0215-0.180 (-0.51)(-0.53) 0.04040.259 (0.48)(0.53) 0.08230.455 (0.68)(0.73) subordinate0.03420.244 (1.18)(1.27) 0.002440.0144 (0.13)(0.10) numblockholders0.002080.0145 (0.66)(0.69) logassets0.01190.07840.01160.07670.01230.0818 (4.04)***(4.12)***(3.93)***(4.01)***(4.25)***(4.33)*** Intercept0.226-1.3010.226-1.3000.225-1.310 (1.17)(-1.24)(1.17)(-1.23)(1.16)(-1.24) N 590859005908590059085900 R 2 0.02710.02760.0272 t statisticsinparentheses * p< 0 : 10,** p< 0 : 05,*** p< 0 : 01 Tables14and15narrowourviewtotheprestigeofanexecutive'sMBAprogram. Thiseliminatesmanyofthemotivationsthattheprestigiousdegreeholders abovelikelyhold.Instead,weseeamuchstrongerofsizehere.These tablessuggestthatMBAdegreeholdersarelikelymuchmorecareer-focusedand 100 Table3.15DeterminantsoftheCEOHoldinganMBAfromaTopProgram (1)(2)(3)(4)(5)(6) OLS1Logit1OLS2Logit2OLS3Logit3 hasblock-0.00468-0.0461-0.00263-0.0212 (-0.44)(-0.48)(-0.19)(-0.17) individual-0.0146-0.157 (-1.01)(-1.13) investmentco0.01020.0965 (0.26)(0.26) -0.0179-0.152 (-0.27)(-0.23) -0.00939-0.121 (-0.10)(-0.15) subordinate0.04120.384 (1.59)(1.78)* 0.009440.0914 (0.57)(0.56) numblockholders0.001450.0135 (0.54)(0.54) logassets0.009660.08290.009590.08250.01010.0874 (3.72)***(3.80)***(3.68)***(3.75)***(3.94)***(4.04)*** Intercept-0.0154-1.788-0.0174-1.782-0.0165-1.854 (-0.14)(-2.70)***(-0.16)(-2.68)***(-0.15)(-2.81)*** N 590858495908584959085849 R 2 0.02690.02750.0269 t statisticsinparentheses * p< 0 : 10,** p< 0 : 05,*** p< 0 : 01 101 aremorelikelytofocussolelyonthemonetarybsthatalargercan Whereasprestigiousdegreescanhavealargenumberoffocusesandmotivations, MBAgraduatesaremuchmorehomogenous.However,wedoseeaweaklyt positiveinthelogittypespationforeliteMBA-holders.Thiscouldbethe resultofaofcareerissues.Perhapsthesecandidatesaredrawntomore dynamicillustratingsomenon-pecuniarymotivations,inawaysimilartotheir larger,prestigiouspeergroup. Table3.16DeterminantsoftheCEOGraduatingfromanEliteSchool (1)(2)(3)(4)(5)(6) OLS1Logit1OLS2Logit2OLS3Logit3 hasblock-0.0106-0.0807-0.00404-0.0316 (-0.89)(-0.92)(-0.26)(-0.28) individual-0.00903-0.0727 (-0.54)(-0.57) investmentco-0.0519-0.525 (-1.47)(-1.30) -0.0703-0.748 (-1.32)(-1.02) -0.003430.0344 (-0.04)(0.04) subordinate-0.0380-0.315 (-1.57)(-1.43) 0.02000.157 (1.06)(1.11) numblockholders0.0005830.00416 (0.19)(0.19) logassets0.006620.04480.006700.04530.007270.0494 (2.26)**(2.25)**(2.27)**(2.26)**(2.52)**(2.51)** Intercept0.0544-1.5520.0507-1.5820.0523-1.621 (0.46)(-2.54)**(0.43)(-2.59)***(0.45)(-2.67)*** N 590858255908582559085825 R 2 0.02740.02830.0273 t statisticsinparentheses * p< 0 : 10,** p< 0 : 05,*** p< 0 : 01 Finally,weturntoacompositemeasure,wheretheschoolhasevaluesfor bothour rich dummyandour prestige2 dummy.Thissetshouldtrulyencompassthe graduatesofwhatareconsideredthe\best"schoolsinthecountry.Table16reports theresultsofusingthisasthedependentvariableinourregressionframework.We 102 seeatandpositivesizesuggestingthatthesecandidatesareableto capturethehighestpayingpositions,suchasthoseprovidedbylargeHowever, weseelittletonofromoursetofownershipvariables. Ultimately,thesequalitymeasuresseemtoillustratetheabilityofcandidates withhigherqualityeducationstocapturethemostlucrativeandprestigiouspositions bylargeHowever,wedoseesomeevidenceofnon-pecuniaryincentives forourbroaderrangesofprestigiousdegreeholders.Thisisilluminating,aswesee thesedisappearwhenrestrictingattentiontoprestigiousMBAdegreeholders. 3.5 Conclusions Thisstudyestablishesthattherearelegitimateofaownershipon itsabilitytorecruittalentintheexecutivelabormarket.Thesereversethe typicaltreatmentintheliterature,aswetreattheCEO'seducationasthedepen- dentvariable,whilemoststudiestakethisasaxedinput.Thisisat contribution,asthesupplysideofthelabormarketislargelyignoredintheexisting literature.Wethatthereisevidencethatthepresenceofblockholderstypically discouragemorecandidates,whileitmightencourageothers.These raiseimportantquestionsaboutthemotivationsofpotentialexecutives.Wehopeto spurfutureresearchintothisaspectofthemarket.Futurestudiesmaybeableto thisapproach,allowingfuturescholarshiptobettercontrolfortheendogeneity inherentinsuchamarketplace. 103 BIBLIOGRAPHY 104 BIBLIOGRAPHY Abowd,J.,andAshenfelter,O.1981.AnticipatedUnemployment,TemporaryLay- andCompensatingWagetials. StudiesinLaborMarkets .University ofChicagoPress,1981:141-170. Adams,R.,Almeida,H.,andFerreira,D.2005.PowerfulCEOsandTheirImpact onCorporatePerformance. ReviewofFinancialStudies ,18(4):1403-1432. Aggarwal,R.,andSamwick,A.2003.WhydoManagersDiversifytheirFirms? AgencyReconsidered. JournalofFinance ,58(1):71-118. Aggarwal,R.,andSamwick,A.2006.Empire-BuildersandShirkers:Investment, FirmPerformance,andManagerialIncentives. JournalofCorporateFinance , 12(3):489-515. Alford,A.,Jones,J.,andZmijewski,M.1994.ExtensionsandViolationsofthe StatutorySECForm10-KFilingRequirements. JournalofAccountingandEco- nomics ,17(1-2):229-254. Bamber,L.,Jiang,J.,andWang,I.2010.What'sMyStyle?Theof TopManagersonVoluntaryCorporateFinancialDisclosure. AccountingReview , 85(4):1131-1162. Bar-Isaac,H.,andDeb,J.2014.(GoodandBad)ReputationforaServantofTwo Masters. AmericanEconomicJournal:Microeconomics ,6(4):293-325. Barker,V.,andMueller,G.2002.CEOCharacteristicsandFirmR&DSpending. ManagementScience ,48(6):782-801. Baruch,Y.2009.ToMBAornottoMBA. CareerDevelopmentInternational , 14(4):388-406. Belliveau,M.,O'Reilly,C.,andWade,J.1996.SocialCapitalattheTop: ofSocialSimilarityandStatusonCEOCompensation. AcademyofManagement Journal ,39(6):1568-1593. Bennedsen,M.,Nielsen,K.,Perez-Gonzalez,F.,andWolfenzon,D.2007.Inside theFamilyFirm:TheRoleofFamiliesinSuccessionDecisionsandPerformance. QuarterlyJournalofEconomics ,122(2):647-691. Bertrand,M.,andSchoar,A.2003.ManagingwithStyle:ThetofManagers onFirmPolicies. QuarterlyJournalofEconomics ,118(4):1169-1208. 105 Bhagat,S.,Bolton,B.,andSubramanian,A.2011.CEOEducation,CEOTurnover, andFirmPerformance.No.DNV3N10,DirectorNotes,TheConferenceBoard,19. Bhattacharya,U.,andDaouk,H.2002.TheWorldPriceofInsiderTrading. The JournalofFinance ,57(1):75-108. Bouvard,M.,andLevy,R.2013.Two-SidedReputationinMarkets. McGillUniversity,WorkingPaper. Burt,R.1992. StructuralHoles:TheSocialStructureofCompetition .Cambridge, MA:HarvardUniversityPress. Campbell,T.,Gallmeyer,M.,Johnson,S.,Rutherford,J.,andStanley,B.2011.CEO OptimismandForcedTurnover. JournalofFinancialEconomics ,101(3):695-712. Cao,J.,Calderon,T.,Chandra,A.,andWang,L.2010.AnalyzingLateSECFilings forDitialImpactsofISandAccountingIssues. InternationalJournalofAc- countingInformationSystems ,11(3):189-207. Cremers,M.,andGrinstein,Y.2014.DoestheMarketforCEOTalentExplainCon- troversialCEOPayPractices? ReviewofFinance ,18(3):921-960. Datar,S.,Garvin,D.,andCullen,P.2010. RethinkingtheMBA:BusinessEducation ataCrossroads .Boston,MA:HarvardBusinessSchoolPress. Deary,I.2004. Intelligence:AVeryShortIntroduction ,Oxford,UK:OxfordUniver- sityPress. Deng,X.,andGao,H.2013.NonmonetaryQualityofLife,andExecutive Compensation. JournalofFinancialandQuantitativeAnalysis ,48(1):197-218. Denis,D.,andDenis,D.1995.PerformanceChangesFollowingTopManagement Dismissals. JournalofFinance ,50(4):1029-1057. Dikolli,S.,Mayew,W.,andNanda,D.2014.CEOTenureandthePerformance- TurnoverRelation. ReviewofAccountingStudies ,19(1):281-327. Dyreng,S.,Hanlon,M.,andMaydew,E.2010.TheofManagersonCorporate TaxAvoidance. AccountingReview ,85(4):1163-1189. Edmans,A.,andGabaix,X.2009.IsCEOPayReallyt?ASurveyofNew OptimalContractingTheories. EuropeanFinancialManagement ,15(3):486-496. Eisfeldt,A.,andKuhnen,C.2013.CEOTurnoverinaCompetitiveAssignment Framework. JournalofFinancialEconomics ,109(2):351-372. 106 Fee,C.,andHadlock,C.2004.ManagementTurnoverAcrosstheCorporateHierar- chy. JournalofAccountingandEconomics ,37(1):3-38. Fee,C.,Hadlock,C.,andPierce,J.2013.ManagerswithandwithoutStyle:Evi- denceusingExogenousVariation. ReviewofFinancialStudies ,26(3):567-601. Finkelstein,S.,andHambrick,D.1996. StrategicLeadership:TopExecutivesand theirctsonOrganizations .St.Paul,MN:WestPublishingCompany. Frank,M.,andGoyal,V.2007.CorporateLeverage:HowMuchdoManagersReally Matter?WorkingPaper,UniversityofMinnesota. Frey,M.,andDetterman,D.2004.ScholasticAssessmentorG?TheRelationship betweentheScholasticAssessmentTestandGeneralCognitiveAbility. Psycholog- icalScience ,15(6):373-378. Frydman,C.(2007).RisingthroughtheRanks.TheEvolutionoftheMarketfor CorporateExecutives,1936-2003.MassachusettsInstituteofTechnology,Working Paper. Gao,H.,Luo,J.,andTang,T.2015.ofManagerialLaborMarketonEx- ecutiveCompensation:EvidencefromJob-Hopping. JournalofAccountingand Economics ,59(2-3):203-220. Goel,A.,andThakor,A.2008.OvCEOSelection,andCorporateGov- ernance. TheJournalofFinance ,63(6):2737-2784. Gottesman,A.,andMorey,M.2010.CEOEducationalBackgroundandFirmFi- nancialPerformance. JournalofAppliedFinance ,20(2):70-82. Graham,J.,andHarvey,C.2001.TheTheoryandPracticeofCorporateFinance: EvidencefromtheField. JournalofFinancialEconomics ,60(2):187-243. Graham,J.,andHarvey,C.2002.HowdoCFOsMakeCapitalBudgetingandCap- italStructureDecisions? JournalofAppliedCorporateFinance ,15(1):8-23. Graham,J.,Harvey,C.,andRajgopal,S.2005.TheEconomicImplicationsofCor- porateFinancialReporting. JournalofAccountingandEconomics ,40(1):3-73. Graham,J.,Li,S.,andQiu,J.2012.ManagerialAttributesandExecutiveCompen- sation. ReviewofFinancialStudies ,25(1):144-186. Hackbarth,D.2008.ManagerialTraitsandCapitalStructureDecisions. Journalof FinancialandQuantitativeAnalysis ,43(4):843-881. 107 Hansen,M.,Ibarra,H.,andPeyer,U.2010.TheBestperformingCEOsintheWorld. HarvardBusinessReview ,88(1):104-113. Hawes,G.1978. HawesComprehensiveGuidetoColleges ,NewYork,NY:NewAmer- icanLibrary. Huddart,S.,Hughes,J.,andLevine,C.2001.PublicDisclosureandDissimulation ofInsiderTrades. Econometrica ,69(3):665-681. Iqbal,Z.2015.CEOAge,Education,andIntroductionofHedgingintheOiland GasIndustry. JournalofEconomicsandFinance ,39(1):189-200. Jalbert,T.,Furumo,K.,andJalbert,M.2011.DoesEducationalBackground CEOCompensationandFirmPerformance? JournalofAppliedBusinessResearch , 27(1):15-39. Jalbert,T.,Rao,R.,Jalbert,M.2002.DoesSchoolMatter?AnEmpiricalAnalysis ofCEOEducation,Compensation,andFirmPerformance. InternationalBusiness andEconomicsResearchJournal ,1(1):83-98. Kaplan,S.2008.AreU.S.CEOsOverpaid? AcademyofManagementPerspectives , 22(2):5-20. Kaplan,S.,andMinton,B.2012.HowhasCEOTurnoverChanged? International ReviewofFinance ,12(1):57-87. Khanna,N.,andSonti,R.2004.ValueCreatingStockManipulation:FeedbackEf- fectofStockPricesonFirmValue. JournalofFinancialMarkets ,7(3):237-270. Kyle,A.1985.ContinuousAuctionsandInsiderTrading. Econometrica ,53(6):1315- 1336. Levit,D.2012.Expertise,Structure,andReputationofCorporateBoards.Univer- sityofPennsylvania,WorkingPaper. Levit,D.,andMalenko,N.2015.TheLaborMarketforDirectorsandExternalities inCorporateGovernance. TheJournalofFinance ,AcceptedArticle12287. M.,andJonson,E.2013.CEOBusinessEducationandFirmFinancial Performance:ACaseforHumilityRatherthanHubris. Education+Training , 55(4):461-477. Malmendier,U.,andTate,G.2005.CEOandCorporateInvestment. JournalofFinance ,60(6):2661-2700. 108 Malmendier,U.,andTate,G.2005.DoesOvCorporateInvest- ment?CEOOvMeasuresRevisited. EuropeanFinancialManagement , 11(5):649-659. Malmendier,U.,andTate,G.2008.WhoMakesAcquisitions?CEOOv andtheMarket'sReaction. JournalofFinancialEconomics ,89(1):20-43. Martin,K.,andMcConnell,J.1991.CorporatePerformance,CorporateTakeovers, andManagementTurnover. TheJournalofFinance ,46(2):671-687. Maxam,C.,Nikbakht,E.,Petrova,M.,andSpieler,A.2006.ManagerialCharacter- isticsandHedgeFundPerformance. JournalofAppliedFinance ,16(2):57-70. Mian,S.2001.OntheChoiceandReplacementofChiefFinancial Journal ofFinancialEconomics ,60(1):143-175. Montgomery,J.1991.SocialNetworksandLabor-MarketOutcomes:Towardan EconomicAnalysis. AmericanEconomicReview ,81(5):1408-1418. Morck,R.,Shleifer,A.,andVishny,R.1990.DoManagerialObjectivesDriveBad Acquisitions? JournalofFinance ,45(1):31-48. Murphy,K.andZabojnik,J.2004.CEOPayandAppointments:AMarket-Based ExplanationforRecentTrends. AmericanEconomicReview ,94(2):192-196. Murphy,K.,andZimmerman,J.1993.FinancialPerformanceSurroundingCEO Turnover. JournalofAccountingandEconomics ,16(1):273-315. Murphy,K.andZabojnik,J.2007.ManagerialCapitalandtheMarketforCEOs. UniversityofSouthCalifornia,WorkingPaper. O'Leonard,K.2014. TheCorporateLearningFactbook:Statistics,Benchmarks,and AnalysisoftheUSCorporateTrainingMarket .BersinandAssociates.http://www. bersin.com/Practice/Detail.aspx?id=17472(accessedFebruary1,2015). Oyer,P.andSchaefer,S.2011.PersonnelEconomics:HiringandIncentives. The HandbookofLaborEconomics ,4(B):1769-1823. Perez-Gonzalez,F.2006.InheritedControlandFirmPerformance. AmericanEco- nomicReview ,96(5):1559-1588. Peters,F.,andWagner,A.2014.TheExecutiveTurnoverRiskPremium. TheJour- nalofFinance ,69(4):1529-1563. 109 Rees,A.1966.InformationNetworksinLaborMarkets. AmericanEconomicReview , 56(1-2):559-566. Rotemberg,J.,andSaloner,G.1993.LeadershipStyleandIncentives. Management Science ,39(11):1299-318. Rotemberg,J.,andSaloner,G.1994.ofNarrowBusinessStrategies. Amer- icanEconomicReview ,84(5):1331-1349. Rotemberg,J.,andSaloner,G.2000.Visionaries,Managers,andStrategicDirection. RANDJournalofEconomics ,31(4):693-716. Ruiz-Verdu,P.,andSingh,R.2014.BoardReputation,CEOPay,andCam Compensation.UniversidadCarlosIIIdeMadrid,WorkingPaper. SecuritiesandExchangeCommission.2003. FinalRule:MandatedElectronicFil- ingandWebsitePostingforForms3,4,and5 .Washington,D.C.Availableat: http://www.sec.gotm. SecuritiesandExchangeCommission.2012. SecuritiesExchangeActof1934 .Wash- ington,D.C.Availableat:https://www.sec.gov/about/laws/sea34.pdf. Shefrin,H.,andStatman,M.1985.TheDispositiontoSellWinnersTooEarlyand RideLosersTooLong:TheoryandEvidence. TheJournalofFinance ,40(3):777- 790. Shleifer,A.,andVishny,R.1989.ManagementEntrenchment:TheCaseofManager- SpInvestment. JournalofFinancialEconomics ,25(1):123-139. Song,F.,andThakor,A.2006.InformationControl,CareerConcerns,andCorpo- rateGovernance. TheJournalofFinance ,61(4):1845-1896. Tyler,B.,andSteensma,H.1998.TheofExecutives'ExperiencesandPercep- tionsontheirAssessmentofPotentialTechnicalAlliances. StrategicManagement Journal ,19(10):939-965. Weisbach,M.1988.OutsideDirectorsandCEOTurnover. JournalofFinancial Economics ,20:431-460. 110