Compensation and Tournament Theory Empirical Tes

NORC at the University of ChicagoThe University of ChicagoExecutive Compensation and Tournament Theory: Empirical Tests on Danish DataAuthor(s): Tor ErikssonSource:

Journal of Labor Economics,

Vol. 17, No. 2 (April 1999), pp. 262-280Published by: The University of Chicago Press on behalf of the Society of Labor Economists and theNORC at the University of ChicagoStable URL: /stable/10.1086/209920 .Accessed: 26/06/2013 18:03Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at ./page/info/about/policies/OR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range ofcontent in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new formsof scholarship. For more information about JSTOR, please contact support@ University of Chicago Press, Society of Labor Economists, NORC at the University of Chicago, TheUniversity of Chicago are collaborating with JSTOR to digitize, preserve and extend access to Journal ofLabor ://

This content downloaded from 137.222.114.240 on Wed, 26 Jun 2013 18:03:40 PMAll use subject to JSTOR Terms and Conditions

ExecutiveCompensationandTournamentTheory:EmpiricalTestsonDanishDataTorEriksson,AarhusSchoolofBusinessandCentreforLabourMarketandSocialResearchThisarticleaddstotheempirieveralpropositionsoftournamentmodelsonarichdatasetcontaininginformationabout2,600executivesin210Danishfi,Arepaydifferentialsbetweenjoblevelsconsistentwithrelativecom-pensation?Ispaydispersionbetweenlevelshigherinnoisyenviron-ments?Isthedispersionaffectedbythenumberoftournamentparti-cipants?Isaveragepaylowerinfirmswithmorecompressedpaystructures?Doeswiderpaydispersionenhancefirmperformance?uctionMyintentioninthisarticleistotestsomepredictionsthathaveemergedinthetheoryoftournamentsasatheoryofpaystructuresoffiabasecomesfromamajorDanishconsultingfirmandcontainsfairlydetailedEarlierversionsofthisarticlewerepresentedatmeetingsin1996:theSocietyofLaborEconomistsmeetingsinChicagoinMay;theinternationalconferenceonComparativeAnalysisofEnterpriseDatainHelsinkiinJune;aNordicWork-˚rhusinSeptember;andtheEuropeanshoponManagerialCompensationinAAssociationofLabourEconomistsconferenceinChania,Greece,lcommentsbyGeorgeBaker,MartinConyon,PeterJensen,goalsotoMetteLaustenandAnjaBaastrupNielsenwhohelpedmewithportionsofthedataset.[JournalofLaborEconomics,1999,vol.17,no.2]᭧htsreserved.0734-306X/99/1702-0005$02.50262This content downloaded from 137.222.114.240 on Wed, 26 Jun 2013 18:03:40 PMAll use subject to JSTOR Terms and Conditions

TournamentTheory263informationaboutmanagers,theirjobs,theircompensation,andthefi,itusesactualdataonexecutives;second,thebasisofouranalysisismoregeneralasthedataarefromover260firms(andnotfromacasestudyofasinglefirm);andthird,itisoneofthefirstinvestigy,itattemptstotestforseveralaspectsoftournamenttheoryonthesamedataset,whereasmostpreviousstudies,saveMain,O’Reilly,andWade(1993),haveexamdifferentialsbetweenjoblevels,controllingforindividualandfirmcharacteristics,consistentwithrelativecompensation?Istheprizeinthetournamentaffectedbythenumberofparticipants?Isthepaydispersionbetweenjoblevelsgreaterinnoisybusinessenvironments?Doesawiderspreadinpayenhancefirmperformance?Aretheredifferencesbetweenfirmsinthisrespect?Istheaveragepaylowerinfirmswithmorecompressedpaystructures?Commontothesequestionsisthefocusonthepaystructurewithinfistinguishestournamentmodelsfromsomeothertheoriesofpaylikehumancapital,agency,andlearningmodels,whicharessedbytworecentsurveysonpayandpromotiondynamicswithinfirms,Gibbons(1996)andPrendergast(1996),theliteratureischaracterizedbythetheoreticalmodelsbeingsetuptoexplainafewstylizedfactsandbytheempiricalworkbeithorsconcludewithpleasformoretheoreticalworkaimingatexplanationsofbroaderpatternsofevidenceandforempirimpiricalverificationoftheories,analternativeapproachistoconsidermghoneortwooftheindividualpredictionsmayalsobegivenanalternative,possiblyequallyplausible,explanation,thestrategyoftestingseveralpredictionsonthesamedatasetprovidesuswithasharpertestofthetournamenttheoryagainstthealtehevirtuesoftournamenttheoryisthatitencompassesseveralpredictionsthateachmaybeexplainedbyadifferent,alternativetheory,ionIIsomebasictheoreticalnotionsandearlierworkarebriefltionsoftestsThis content downloaded from 137.222.114.240 on Wed, 26 Jun 2013 18:03:40 PMAll use subject to JSTOR Terms and Conditions

nIVisconcernedwithshapeofthepayandjob-levelrelationship,fiandExistingWorkToillustratesomepoints,IprovideasimplemodelthatfollowstheonesetoutinLazearandRosen(1981).Therearetwoidenticalplayers,ehasafixedprizetothewinner,W1,andtheloserreceivesanotherfixedprize,yers’output(performance),qiÅmi/ei,iÅj,k,(1)dependsontheplayer’seffortlevel(action),m,andarandomcomponent(,luck)utilityofeffortisdescribedbythecostofeffortfunctionCÅC(m),anditisassumedthatbothC؅andCЉbabilitythatjwinsW1dependspositivelyonhowmucheffortheputsforth(m1)andnegativelyontheactionsoftheotherplayer(mk).Inaddition,ectedutilityofthejthplayerisP(W10C(mj))/(10P)(W20C(mj))ÅP(W10W2)/W20C(mj),babilitythatjwinsisthenprob(qjúqk)Åprob((ek0ej)õ(mj0mk))Åprob((mj0mk)úz))ÅG(mj0mk),wherezÅek0ej;zÇg(z),GÅthecumulativedensityfunctionofz,andE(z)Åtheplayersmaximizes(3)ditionsforoptimum(assuminganinteriorsolution)are(W10W2)dP/dmi0dC/dmiÅ0and22(W10W2)d2P/dm2i0dC/dmiõ0,(2)(3)iÅj,k.(4)This content downloaded from 137.222.114.240 on Wed, 26 Jun 2013 18:03:40 PMAll use subject to JSTOR Terms and Conditions

TournamentTheory265Ifbothplayersaremaximizing(3)takingtheotherplayer’sactionasgiven,thenusing(3),dP/dmjÅdG(mj0mk)/dmjÅg(mj0mk),whichaftersubstitutioninto(4)givesplayerj’sbestreactionfunction;(W10W2)g(mj0mk)ÅdC/se,giventheassumptionofidenticalplayers,bothchoosesthesameeffortlevel,andplayerk’equilibrium,mjÅmk,,PÅ0.5).Thus,(W10W2)g(0)0dCi/dmiÅ0.(5)Giventhelaborsupplycharacterizedby(5),firmsmaximizeprofitsperworker(whichhereisequaltothenumberofjobslots).Itcaneasilybeseen(byLazearandRosen1981)thattheaveragewagenecessarytoattractemployeestothefirmandtheoptimalwagespreadare(W1/W2)/2ÅC(m)andW10W2Å1/g(0),(7)(6)on(5),theequilibriumlevelofelsoftheprizasewherethereareseveralpositionswithinthefirm,tournamenttheorypredicts(seeRosen1986)thattherewillbeaninbecausethevalueofwinningnotonlyisthewinner’sprizeatthatlevelbutalsoincludesequence,,however,thatatthefinallevelthereisnofurtherprizetobewon,andCEOsshould,therefore,,tournamenttheorypredictsanextraordi-narilylar content downloaded from 137.222.114.240 on Wed, 26 Jun 2013 18:03:40 PMAll use subject to JSTOR Terms and Conditions

266ErikssonThesecondimplicationof(5)—andalsoanimplicationof(7)—isthat,,thesmallerg(0)becomes),,inproductionenvironmentsinwhichcontributionsofluckorotherrandomfactorstooutputareimportant,firmsusealartedoutbyLazear(1995),thisincentiveroleofsalariesmaybeimporodelabove,r,iftheydo,itisnolongerneceersknowtheirownabilityaswellasthatoftheotherplayers,sableknowtheyarelesslikelytowin,andtheablewinbytheirinnateability(seeKnoeberandThurman[1994]foradiscussion).Anothers,replayers,ssobvious,however,howtheeffectofeffortonexaminedbyMcLaughlin(1988),whoderivesthefollowingexpressionsfortheoptimalprizespreadandeffort,respectively,inatournamentwithncontestants:DW*Å(dP(n)/dm)/((dP(n)/dm)2/SCЉ/4)andC؅(m*)Å1/(1/SCЉ/4(dP(n)/dm)2),(9)(8)whereP(n)istheprobabilityofwinninginan-contestanttournament,giventhattheothern01contestantssupplyeffortm*,Sisthedegreeofriskaversion,andCЉis,asbefore,eethat,underriskneutrality(SÅ0),theprizespreadisincreasingintournamentsize,ectofamarginauently,toinduceeffort,hecasewithrisk-aversecontestants,itingresultfortheoptimalprizespreadisfragile:r,thespreadisincreasinginthetournamentsizerange,elsketchedabovealsoabstractsfromthefactthatseniormanagementofafirmoftenactsasateamperforminghighlyinterdepen-This content downloaded from 137.222.114.240 on Wed, 26 Jun 2013 18:03:40 PMAll use subject to JSTOR Terms and Conditions

TournamentTheory267dentworkand,so,compensationbasedonindividualperformancemaybeinappropriatebecausssedbyLazear(1989),paycompressionmaydominatetournamentaspectsinso-calledhawkishfihis,weallowforemployeestobehavestrategicallyagainsttheirrivalsbyaugmenting(1)witha‘‘sabotage’’parameterfi,whichshowstheharmj(k)caninflictonk(j):qjÅmj0fk/ejandqkÅmk0fj/ek.(10)Assabotagerequireseffort,thecostofeffortfunctionnowbecomes:C(mi,fi).DenotingthepartialswithC1andC2(bothassumedtobepositive),respectively,thefirst-orderconditionstotheemployees’max-imizationproblemare(W10W2)dP/dmiÅC1(mi,fi)and(W10W2)dP/dfiÅC2(mi,fi).Fromthesewecanseethat,foragivenwagespread,theeffortislowerforplayerswhocansabotageothers(providedC12ú0).Thefirst-orderconditionsforthefirm’smaximizationproblemincaseofabsenceandpresenceofstrategicalbehavior,respectively,are(10C1)dm/dW1Å0,(10C1)dm/dW2Å0,(10C1)dm/dW10(1/C2)df/dW1Å0,and(10C1)dm/dW20(1/C2)df/dW2Åy,asC2ispositive,equilibriumeffortishigherinevidenceoftournamentnotionshasaboveallbeenobtainedfromstudiesofsports(EhrenbergandBognanno1990;BeckerandHuselid1992)andincontrolledexperiments(Bull,Schotter,andWeigelt1987).Studiesbasedondataonactualexecutivesarethinontheground,simplybecausedatasetscontaininginformationaboutseveralmanagersperfirmarehardtofithestudieshavefocusedontheconvexityofthepaystructure.O’Reilly,Main,andCrystal(1988),Leonard(1990),andMainetal.(13)(12)(11)This content downloaded from 137.222.114.240 on Wed, 26 Jun 2013 18:03:40 PMAll use subject to JSTOR Terms and Conditions

268Eriksson(1993),allusingthesamedataset,haveshownthatdifferencesincompen-srresultsareobtainedbyLambert,Larcker,andWeigelt(1993)andintwodetailedstudiesofthepersonnelrecordsofasinglefirm,Lazear(1992)andBaker,Gibbs,andHolmstro¨m(1994).Inarecentstudy,Conyon(1995),usingalargesampleofBritishfionalevidenceissomewhatmoremixed,however.O’Reillyetal.(1988)findanegativerelationship,andMainetal.(1993)apositiverelationship,al.(1993)alsoconsidertheeffectsofthepaystruc-tureonfirmperformance,findGarvey(1998)examinetheeffectsofpfindstrongevidenceofatournamentstructure;strongpromotionincentiveswerearandThurman(1994)studytheperfestsofpredictionsconcerningtheeffectsofprizelevelandprizedifferentials,theeffectsofability,andtheexistenceofhandicapsystems,edintheintroduction,thisarticlediffersfrompreviousworkinthscriptionThebulkofthedatausedinthisarticlecomesfromanunbalancedpanelcontaininginformationaboutapproximately2,600managersinabout210Danishfirms(peryear)duringthe4-yearperiod1992–ahavebeenobtainedfromconfidentialfilesofamajorDanishconsultingfirmandprovide,inadditiontoannualcompensationdata,fairlydetailedinformationabouttheindividualcharacteristicsofmanagers,theirjobs,andthefipensationvariableincludessalaryandbonuscomponentsaswellastheemployers’contributionstopensionfunds(allthreeareavail-ableasseparatevariables).Arelativelysmallproportion—20%–25%ofallmanagersandathirdoftheCEOs—arepaidbonusesand/orcommissions,andtheiraverageshareoftotalcompensationvariesbe-tween10%and12%ptions,deferredcompensation(exceptcontributionstopensions),issionisnotlikelytoaffectourresultsmuch,umerationdatasethasbeenaugmentedwithfurtherinforma-tiononthefirmsregardingtheirperformance(accountingprofits,sales)inthe8-yearperiod1987–formationhasbeenderivedfromThis content downloaded from 137.222.114.240 on Wed, 26 Jun 2013 18:03:40 PMAll use subject to JSTOR Terms and Conditions

TournamentTheory269anannualhandbookofallDanishfirmswithanannualturnoverexceeding40millionDanishkronerin1994pricesormorethan50employeescalledGreens—Børsensha˚istinguishingfeaturesofDanishfi,mostDanishfiextent,thatissimpr,theaveragesizeoffirmsinDenmarkissmallalsorelativetotheotherNordiccountries,and,inparticular,thelargeDanishfirmsaremuchsmallerthanthoseinFinland,Norway,,thecorporategovernancesystemisgenerallyoftheNordic-Germantype,inwhichmangersaremonitoredbyrepresentativesofbanks,largeshareholders,andcloselyrelatedfi,itdiffersinparticularfromtheNorthernAmericansystem,whichischaracterizedbyboardsmadeupmainlyof‘‘outsiders’’,severalDanishfiuently,thereisonlyarelativelysmallnumberoffipliesalsotothefirmsinmydataset.1Themajorityofthefirmsinthisdatasetaremedium-sizedorlargefirms(intheDanishsense),andthedataare,therefore,notrepresentativeofallDanishfir,thesampleatmydisposalisfairlyrepresen-tativeofthemedium-sizedandlargefirmswJobLevelsThefirstofthetestsIcarrorwhetherdifferentialsinpaybetweenlevels(definedinalternativeways)inestimatesofthepaydiffer-encesbetweenadjacentorganizationallevels,Iestimatecompensationequations,bothfromsingleyearsandfromashortpanel(whichallowsustoaccountforfixedeffects),ofthefollowingform:WijtÅai/bXijt/gLijt/eijt,(14)whereWisthelogarithmofannualcompensation,aiareindividualfixedeffects,Xisavectorofindividualandfirmcharacteristics,torXincludesage,tenureincurrentposition,educationallevel,industry,numberofemployees,numberofsubordinates,and(logof)sales(andyeardummiesinpanelestimations).ThevectorLwillbedefi,theItshouldbenoted,however,thatthefirmsinthesamplearenotheadedbyowner-managers.1This content downloaded from 137.222.114.240 on Wed, 26 Jun 2013 18:03:40 PMAll use subject to JSTOR Terms and Conditions

270Erikssongestimatesarederivedfromamodelthatcontrolsforindividualtraits(andindividual-specificfixedeffects)aswellassomefiybeimportantassomepartoftheinter-levelpaydifferencesmayrefltself-evidenthowtodefiaatmydisposalcontaininformationaboutjobsaccordingtotheirfunction(production,sales,logistics,personnel,etc.),formalpositionasreportedbythefirm(CEO,vice-president,higher-levelmanager[fagdirektør],andlower-levelmanager[fagchef]),membershipintheboardorthetop-managementgroup,andresponsibilitylevel(seebelow).Innoneofthesedescriptionsarejoblevelsidentifirtocheckthesensitivityoftheresultstothejob-leveldefinitionadopted,Ihaveusedthreealternativesetsoflevels(orpositions)tdetaileddescription,which,however,isavailableonlyforaportionofthewholesample,isaclassifi-cationofpositionsintoninelevelsaccordingtoajobauthoritiesevalua-tionsystemcreatedbytheconsultingfissificationisbasedongrades(1–6)giventosixfactors:complexityoftheproblemstobesolved,independenceindecisionmaking,reporting,responsibility,experience,ondclassificationisacruderversionofthefirstone2andclassi-fiesthepositionsheldintofivedifferentresponsibilitylevels.3Alljobsinthesamplearecoveredbythisclassifirdsetofleveldummieshasbeenconstructedfromtwopiecesofinformation:thetitlesofpositionsasreportedbythefivesussixlevels:CEO,vice-president,aboardmem-berhigher-levelmanager,anon–boardmemberhigher-levelmanager,aboardortop-leassifispace,belowIreportonlyresultsfromusingthethirdclassifiimatesfromusingtheothertwoclassificationswerequitesimilartothosereportedhereThetwolowestlevelsinthecruderclassificationcorrespondlargelytolevels1–4inthemoredetailedclassification,levels3and4to5and6,andlevel5tolevels7to9.3Thejobsareclassifiedintothreemainresponsibilitylevels:thetacticallevel,thestrategicallevel,andthepolicylevel,twolowestresponsibilitylevels,,apositionatthelowerstrategicallevelinvolvesmakingpropositionsregardingprincipalstrate-giesandplansforthefirm,whereasapersoninthepositionatthehigherstrategicallevelhastheauthoritytomakethosedecisions.2This content downloaded from 137.222.114.240 on Wed, 26 Jun 2013 18:03:40 PMAll use subject to JSTOR Terms and Conditions

TournamentTheoryTable1MeanPay(BaseSalaryplusBonus)byLevelsinCorporateHierarchyLevel/PositionLower-levelmanager,nomembershipLower-levelmanager,boardortop-levelgroupHigher-levelmanager,nomembershipHigher-levelmanager,top-levelgroupHigher-levelmanager,boardmemberVice-presidentCEON7297149TotalPay483.1524.9844.8728.6835.1903.11,250.0/12.4/60.9013.7/14.6/8.1/38.4271Difference(%)(andaresetoutinalongerversionofthisarticle,whichisavailablefrommeonrequest.)ragepay(basesalaryplusbonus)foreachlevelandtheproportionalpaydifferencesbetweenadjacentlevels,startingfromthelowest,arepresentedintable1.4Thefiguresarefrom1994,butthecorrespondingfiectiveofwhichoneofthelevelclassificationsareused,ovesupthecorporatehierarchy,thepaydifferencesincrease.5Clearly,thes,letusturntoconsiderwhetherthisalsoremainstobethecasewhenwecontrolforindividualandfies2and3,theestimatesfromequation(7)ffi-cientsarethoseofleveldummiesinestimationsinwhichthedependentvariable,totalpay,isinlogsandthe(omitted)mepreviousstudies(,Leonard1990;Lazear1992;andBakeretal.1994),,inregressionswithjustonesetofexplanatoryvariablesincludedatatime,thehumancapitalvariablesexplainabout20%ofthevariance,whereasjobleveldummiesexplainabout60%.Addingthejob-leveldummiestoaspecificationwithstandardhumancapitalvariables,industrydummies,andfirmcharacteristicssignificantlyTheresultspresentedareallbasedonttureremainslargelyunalteredwhenwelookatbasesalaryonly.5However,ereason,therathersmallgroupofhigher-levelmanagerswhoareneitherboardmembersnormembersoftopmanagementgroupsinthefirmsearnmorethantheircolleagueswhoaremembersofthesebodies.(Thisunexpectedresultisfoundforeachyearinthesample.)4This content downloaded from 137.222.114.240 on Wed, 26 Jun 2013 18:03:40 PMAll use subject to JSTOR Terms and Conditions

272Table2EstimatedPayDifferences(logDifferentials)Position*Lower-levelmanager,boardortop-levelgroupHigher-levelmanager,nomembershipHigher-levelmanager,top-levelgroupHigher-levelmanager,memberofboardVicepresidentCEOAdjustedR2Numberofobservations1992.121(.012).399(.023).324(.014).489(.021).628(.021).989(.022).6992,2891993.103(.011).413(.024).319(.015).464(.019).629(.020).951(.022).7242,2481994.110(.011).420(.024).340(.014).480(.020).620(.020)1.031(.025).7192,5021995.125(.010).437(.025).337(.016).475(.018).695(.021)1.011(.026).7012,111ErikssonNOTE.—Standarderrorsareinparentheses.*Omittedcategory:egressorsincludedwereage,tenureincurrentposition,levelofeducation,industry,numberofemployeesinthefirm,thenumberofsubordinates,eseenfromthetable,imatescanbeusedtocalculatetheincreaseinrewardfrommovingfromoneleveltothenextholdingindividualtraitsandfi,fromtheresultsfor1995intable2,6weTable3Fixed-EffectsEstimationResultsPositionsLower-levelmanager,board/topmemberHigher-levelmanager,nomembershipHigher-levelmanager,top-levelgroupHigher-levelmanager,boardmemberVicepresidentCEOAdjustedR2Hausman’stest(x2(2))NumberofobservationsNOTE.—tionResults.101.375.327.472.607.923(.010)(.025)(.015)(.020)(.030)(.020).7747.259,150Again,sothatthedecreaseinrewardsbetweenthethirdandthefourthlevelmaybeduetorelativelyfewobservationsforthegroupofhigher-levelmanagerswhoareneitherboardnortop-managementgroupmembers.6This content downloaded from 137.222.114.240 on Wed, 26 Jun 2013 18:03:40 PMAll use subject to JSTOR Terms and Conditions

TournamentTheory273findthatincreasesinrewards,beginningfromthelowestlevel,are13.3,36.6,09.5,14.8,24.6and37.1%.Thecorrespondingfiguresfromtable3are10.6,31.6,04.7,15.6,14.5,and37.2%.Thelastfigureisanestirisonofthesefigureswiththoseintable1showsthat,ingeneral,controllingforindividualandfirmcharacteristicerde-creaseoccursasIcaterforheterogeneities(likedifferencesinabilitydiscussedabove)byanindividualfixed-effectsspecifiultssetoutintable3confingesarerathersmall,however,resultoftheeconometricexecreaseinsprear,Ihavenotbeenabletoisolateanextraordinarilylargeincreaseintherewardatarize,theconvexityoftherelationshipbetweenpayandlevelsinhierarchydocumentedaboveisconfitswellinwiththeresultsofanearlierstudyusingthesamedata,ErikssonandLausten(1996),ellbethatexecutivepayhaslittletodowiththeabsoluteperformanceoftheCEOorotherseniormanagersandthatinsteadthghawideningpaygapthroughthecorporatehierarchyisakeypredictionoftournamentmodels,thepatternobserveddoesnotimplytournamenttheoryasothereconomictheories(seeRosen1992)aswellassociologicaltheories(seeO’Reillyetal.1988),forexample,providedsuperiors’decisionsaffectdirectlytheproductivityoflower-levelemployees,sortingofmoreablepersonsintohigher-levelpositxpaystructuremayalsoariseintaskassignmentmodels;see,forexample,Waldman(1984).Theassignmentofanemployeetonewtaskssignalsinformationabouthisprodus,however,someevidenceintheanalysisabovesuggestingthatpayisnotlikelytoequalmarginalproduct,namely,thefactthatpayislargelyattachedtojoblevelsandconsiderablylesstoindividualcharacteristicsreflspectsofTournamentTheoryTheaiminthissectionistotrytotestsomeotherpredizetwoThis content downloaded from 137.222.114.240 on Wed, 26 Jun 2013 18:03:40 PMAll use subject to JSTOR Terms and Conditions

,Iinvestigatewhetherinterfirmdifferencesinortwopredictionsspecifictotournamentmodels,namely,thatrewarddifferencesareaffectedbythenumberoftournamentparticipantsandthatpaxaminewhetheraveragepayislowerinfi,Icarryoutasimpletestofakeypredictionconcern-ingtheefficiencyoftournamentpaystructures,namely,thatawiderpaydispersionenhancestheeconomicperformanceoffistigatingtheseaspectsoftournamentmodels,theunitsofobser-vationsarefirmsandtheirmanagerialpaystructures.7Thesampleana-lyzedbelowconsistsofthosefirmsforwhichIhaveobservationsonaminimumoffiveemployees(oneofwhichistheCEO)andcompleterecordsonfirmperformancefortheperiod1987–94andonmanagerialcompensationforall4years1992–estrictionsreducesthesampleto111fihlin’s(1988)analysissuggestsasatestofthepresenceoftour-namentstestingfortheexistenceofapositiverelationshipbetweenCEOpayand(giventheaveragepayofthetournamentparticipants)se,inordertocarryoutsuchatest,theparticipantsinthetournamenthavetobeidentifir,asmanyofthecompaniesinthedatasetdonothaveformalvice-presidentpositions,Ihave,followingO’Reillyetal.(1988),decidedtousethemanagersthatarereportedbythefirmstohavesignificantresponsibilities—thatis,themanagerswhosejobsareclassifiedasbeingatthepolicylevel—endentvariableisthelogdifferencebultsfromestimationsondatafor1992and1994arepresentedintable4(seealsotable5below).Theestimatesdosuggestthat,control-lingforfirmsize,agreaternumberofcontestantsincreasesthewinningprize,nitudeoftheeffectisrathermodest,howninSectionII,apredictionemanatingfromtournamentmodelsisalargerspreadinpayinfirmsoperatinginnoisyorriskyenvironmenuently,weexpectfirmsinindustrieswheredethertestcouldbewhethertournamentstructuresareobservedinthefirmsthataccordingtothetheoryaremorelikelytoadoptthem,namely,firmsinwhichmeasurementcosts,thefirm-specifichumancapitalelement,r,becauseofthegreatdemandsonthedata,suchtestsareextremelydifficulttoperform.7This content downloaded from 137.222.114.240 on Wed, 26 Jun 2013 18:03:40 PMAll use subject to JSTOR Terms and Conditions

TournamentTheoryTable4TestoftheEffectoftheNumberofContestantsDependentVariable:logCEOPay0logAverageManagerialPay1992ConstantNumberofcontestantsFirmsize(logsales)R2.175(.039).018(.006).0002(.0001).1021994.187(.052).017(.005).0002(.0001).097275NOTE.—nproblemwithattemptingtotestthishypothesisis,ofcourse,tofindavariablethataccuratelycapturesdifferencesinfirms’(industries’)asetprovidesinformationaboutthesalesofthefirmsintheperiod1987–eseseriesIhaveforeachfirmcalculatedthecoefficientofvariationof(deflated)sales.8Thesecoefficientsofvariationisthefiermeasureisderivedinasimilarfashion,ficientsofvariationwerecalculatedfromtheIndustrialStatis-tics,producedbyStatisticsDenmark,forthe1987–93periodforvolumeofproductionforeachofthetwo-digitlevelindustriesthefirmsinthesampleareoperatingin.9Itriedtwoalternativedependentvariables:thecoefficientofvariationoftotaackofthecoefficientofvariationvariableisthatitmaybeaffectedbydifferencesinthenumberofmanage-rialpositionsinfiingtotheestimationresultssetoutintable5,thereisindeedapositiveandstatisticallysignificantrelationshipbetweenthevariabilityofthesales(production)ofthefirm(industry)andtheintrafieaseinthe‘‘noise’’asmeasuredbyanincreaseinthecoefficientofvariationoffirmsalesby1SEincreasesthecoeffiarelativelymodesteffect,ally,ininterpretingtheresults,oneshouldkeepinmindthatthecoefficientofvariancemeasurescanatbestonlybecrudeproxiesThefirmsthathavehadchangesinsalesowingtoacquisitionsofotherfirmsorsalesofthepartsofthefirmhadtobediscarded.9Somefissignedthemtotheindustriesreportedbythemselvesastheirmainindustry.8This content downloaded from 137.222.114.240 on Wed, 26 Jun 2013 18:03:40 PMAll use subject to JSTOR Terms and Conditions

276ErikssonTable5TestoftheEffectofaNoisyBusinessEnvironmentonSpreadofPay,1994DependentVariableCVofPay(1)ConstantFirmsize(logsales)CVoffirmsalesCVofindustryoutputNumberofcontestantsAdjustedR2.064.311(.054).011(.005)(2).295(.059)CEO-ContestantsDifference(3).122(.034).0002(.0001).014(.008).015(.008).121(4).111(.033).0002(.0001).016(.005).012(.004).134.015(.003).072NOTE.—Åcoeffierfirmdifferencesinnoiseduetoproductionuncertainty,luck,measurementerror,orotherfactorsbeyondthemanagers’ultsofthetwotestscarriedoutsofarclearlyarddifferencesarelargerthemoreimportantarerandomfactorsforthedevelopmentoftheperformanceofthefiveanalysishas,however,beenconcernedwithwhatthingslooklike,,letusnowturntoconsidertheconsequencesofthepaystruc-tureonfiainfromtheequilibriumin(5),thewiderthepaydispersion,r,asdiscussedbyLazear(1989;1995),theremayalsobeincentivemotivesforfirmstoadoptamorecompressedpaystructure.10Inordertoattract(theright)peopletoparticipateinatournament,thespreadcannotbe‘‘toobig.’’Moreover,ifthecooperationofthemanagersisessentialforthesuccessofthefirm,rewarfirmsbenefitfromtheirtopmanagersactingasateam,however.11Forthosefirmsforwhichcooperationislessimportant—SeealsoMilgromandRoberts(1990)forarelatedanalysisofemployeesengagingincostlyrent-seekingbehaviorwhenthepotentialfromredistributionofcompensationislarge.11AspointedoutbyLazear(1995),analternativetopaycompressionasameansofreducinganticooperativebehaviorofmanagersistosetupthestructure10This content downloaded from 137.222.114.240 on Wed, 26 Jun 2013 18:03:40 PMAll use subject to JSTOR Terms and Conditions

TournamentTheory277‘‘hawkish’’firms,inLazear’sterminology—widerpaygapsmayenhanceperformance,whereasthisisnotthecasein‘‘dovish’’fiy,themaindifficultyintestingthehypothesisoftheperformance-enhancingeffectsofpaydispersionisfindingavariableorindicatorthatenablesustodistinguishbetween‘‘hawkish’’and‘‘dovish’’fiwMainetal.(1993)inusinganexecutiveteaminterdependencyindicator,con-structedastheproportionofprofitcenterheadsofthetotalnumberofmanagers,ithalow(high)proportionofmanagerswithprofitcenterheadtitlesareconsideredteamswithahigh(low)ustrialpoliticshypothesispredictsthatthelowerthedegreeofmana-gerialteaminterdependence,,theinteraointedoutearlier,mostofthefi,onecannotrelyonstockmarketindicatorsasmea-suresoffirmperformance,andIuseaccountingprofiformanceofthefirmsismeasuredasa3-yearaverageofprofitsdividedbysales.12Onceagain,Iusetwoaonalexplanatoryvariablesincludedwerefirmsize(measuredbythe4-yearaverageofthenumberofemployees),industrydummies,andtheaveragepayofthemanagersinthefiIfind?Firstofall,Ifindaweakpositiverelationshipbetweenfihepaydispersionvariables,thesealsocarrypositivecoefficients;asignificantonefortheCEO-con-testantsdifferenceandanalmostsignificantoneforthecoeffiiricaleffectofa1SEincreaseinthepaydispersionvariablesisofthemagnitude4%–5%.Theteaminterdepen-dencyvariableaswellastheinteractiontermsneverdifferedsignifi,theindustrialpoliticsargumesionsoftheaveragelogofpayonpaydispersioncontrollingforfirmsizeandindustryshow,consistentwithtournamenttheory,alowerofthefirminsuchawaythattheconsequencesofcompetitivebehaviortothefifirstwastherateofreturnonthefirm’unately,thishasonemajordrawback,namely,thatthefirm’imatesobtainedweresimilarasthoseintable6,ermeasureusedwasacrudeindexbasedontheperformanceofprofitsduringthe8-yearperiod1987–, content downloaded from 137.222.114.240 on Wed, 26 Jun 2013 18:03:40 PMAll use subject to JSTOR Terms and Conditions

278ErikssonTable6TestsofEffectsofPaySpreadonFirmPerformanceandAveragePayDependentVariablelogProfits/Sales(1)ConstantNumberofemployeesLogaveragepayCVofpayCEO-contestantsdifferenceInterdependencyindicatorInteraction(interdependencyandpayspread)AdjustedR23.092(.54).031(.011).241(.145).051(.060)0.005(.015).313(2)3.151(.57).032(.010).151(.060).039(.064).010(.020).345logofAveragePay(3)6.420(.391).0004(.0002).392(.200).111(.407).113(.222).648(4)6.398(.362).0004(.0003).204(.096).125(.750).179(.451).662NOTE.—Åcoeffiepayinfirmswithlesspaydispersion(seetable6).But,again,IfailedtofindasignificantcoeffildbenotedthatonepossiblereasonIhavedrawntwoblanksonthesevariablesisthattheestimatesaredown-warrework,Ihopetoimprovetheanalysisbyexplicitlyac-countingfordifferencesintheorganizationalstructureofthefirms,thatis,distinguishingbetweenmultidivisionalfirms,firmswithaheadquarter-subsidiariesstructure,firmsthatarepartofaconcern,potentiallyimportant,asfirmswithdifferentorganizationalstructuresmaydifferwithrespecdingRemarksInthisarticleIhaveinvestigatedsomeaspectsoftournamenttheoryusingadatasetonDanishexecutives.Ifindthatthereisastableconvexrelationbetweenpayandjoblevelsandthatthisisrelativelyrobustwithrespecttodifferencesinhowjoblevelsaredefigerthenumberofmanagersconsideredtohavesignificantresponsibilitiesinthefirm,,thepredictionoftournamentmodelsthatthereisapositiverelationshipbetwrpredictiongainingThis content downloaded from 137.222.114.240 on Wed, 26 Jun 2013 18:03:40 PMAll use subject to JSTOR Terms and Conditions

TournamentTheory279supportisalargerpaydispersioninfiplicationoftournamenttheory,whichdiffersconsiderablyfromtheoriesofefficiencywagesandfairness,isthatmanagerialpaydiudyaddstotheverysmallliteratureontheeffectofpromotionsandpaystructuresonperformancesomeevidenceofalargermanagerialpayspreadbeingassociatedwithbetterperformanceoffionotseemtobeanydifferences,withregardtotheeffectsofpaydispersiononfirmperformance,betweenfirmswithmar,itmustbenotedthattheseresultsaretentative,astheymaybeaffectedbytheproblemsofmeasuringaccuratelytheinterdependencyofmanagersand/orfiary,Iconcludethatalmostallofmyfighsomeoftheindividualfindingsmayalsobeexplainedbyalternativetheories,thecombinationoffiimportantinviewoftheweaklinkobservedbetweenfirmperformanceandindividualmanagers’pay,asitsuggeststhatlookingatindividualwagncesBaker,George;Gibbs,Michael;andHolmstro¨m,Bengt.‘‘TheInternalEconomicsoftheFirm:EvidencefromPersonnelData.’’QuarterlyJournalofEconomics109(November1994):881–,B.,andHuselid,Mark.‘‘TheIncentiveEffectsofTournamentCompensationSystems.’’AdministrativeScienceQuarterly37(June1992):336–,Clive;Schotter,Andrew;andWeigelt,Keith.‘‘TournamentsandPieceRates:AnExperimentalStudy.’’JournalofPoliticalEconomy95(February1987):1–,Martin.‘‘AnEmpiricalTestofTournamentTheoryUsingDataonUKExecutives.’’k:UniversityofWarwick,,Robert,andGarvey,Gerald.‘‘IncentivesforHelpingontheJob:TheoryandEvidence.’’JournalofLaborEconomics16,no.1(1998):1–erg,Ronald,andBognanno,Michael.‘‘TheIncentiveEffectsofTournamentsRevisited:EvidencefromtheEuropeanPGATour.’’In-dustrialandLaborRelationsReview43(February1990):74–on,Tor,andLausten,Mette.‘‘ManagerialPayandFirmPerfor-mance—DanishEvidence.’’:AarhusSchoolofBusiness,s,Robert.‘‘IncentivesandCareersinOrganizations.’’WorkingThis content downloaded from 137.222.114.240 on Wed, 26 Jun 2013 18:03:40 PMAll use subject to JSTOR Terms and Conditions

dge,MA:NationalBureauofEconomicRe-search,—Børsensha˚agen:BørsensForlag,r,Charles,andThurman,Walther.‘‘TestingtheTheoryofTour-naments:AnEmpiricalAnalysisofBroilerProduction.’’JournalofLaborEconomics12(April1994):155–t,Richard;Larcker,David;andWeigelt,Keith.‘‘TheStructureofOrganizationalIncentives.’’AdministrativeScienceQuarterly38(Sep-tember1993):438–,Edward.‘‘PayEqualityandIndustrialPolitics.’’JournalofPoliti-calEconomy97(June1989):561–80..‘‘TheJobasaConcept.’’InPerformanceMeasurement,Evalua-tion,andIncentives,:HarvardBusi-nessSchoolPress,1992..dge,MA:MITPress,,Edward,andRosen,Sherwin.‘‘Rank-OrderTournamentsasOp-timumLaborContracts.’’JournalofPoliticalEconomy89(October1981):841–d,Jonathan.‘‘ExecutivePayandFirmPerformance.’’IndustrialandLaborRelationsReview43(February1990):13–,Brian;O’Reilly,Charles,III;andWade,James.‘‘TopExecutivePay:TournamentorTeamwork?’’JournalofLaborEconomics11(June1993):606–hlin,Kenneth.‘‘AspectsofTournamentModels:ASurvey.’’InResearchinLaborEconomics,vol.9,editedbyRonaldEhrenberg,pp.225–ich,CT:JAI,m,Paul,andRoberts,John.‘‘TheEfficiencyofEquityinOrgani-zationalDecisionProcesses.’’AmericanEconomicReview,PapersandProceedings80(May1990):154–59.O’Reilly,Charles,III;Main,Brian;andCrystal,Graef.‘‘CEOCompen-sationasTournamentandSocialComparison:ATaleofTwoTheo-ries.’’AdministrativeScienceQuarterly33(June1988):257–rgast,Canice.‘‘WhatHappenswithinFirms?ASurveyofEmpiri-calEvidenceonCompensationPolicies.’’dge,MA:NationalBureauofEconomicResearch,,Sherwin.‘‘PrizesandIncentivesinEliminationTournaments.’’AmericanEconomicReview76(September1986):701–15..‘‘ContractsandtheMarketforExecutives.’’InContractEconom-ics,:Blackwell,n,Michael.‘‘JobAssignment,Signalling,andEfficiency.’’RANDJournalofEconomics8(Summer1984):255– content downloaded from 137.222.114.240 on Wed, 26 Jun 2013 18:03:40 PMAll use subject to JSTOR Terms and Conditions