《Public finance and public choice》读书笔记

这本书是由马斯格雷夫和布坎南共同完成的，全书以论文---答复---讨论的结构来呈现，这样的结构可以使读者从不同角度全面，深刻的理解与国家有关的哲学思想以及财政学理论和政策有关的重要问题。

全书从财政联邦制，税制改革，税收竞争，道德问题等多个方面叙述与讲解，每一个方面讲的都很好，我对其中的对政治行动的约束这一块提一些自己的读后感。

作者在对这个方面进行说明时，巧妙的运用了一个例子，就是两个人对只有半瓶酒这样一个状态的看法，两个人的两个看法就如同两个作者对有无约束的集体做好事或者做坏事的看法。其中布坎南对此的看法就是政府在普通的政治活动中要进行约束。在早些时候，当不是按照闵准的方式组建政府，而是由君主或者贵族进行统治的年代，就存在着应该对政府的行为施加约束的一种普遍共识。由此可见，对政府的权利进行约束这种现象不是少例或者后来才出现的，它的出现不是偶然，而是历史的必然。实际上，现在的议会就是为了限制君主的权利而进行长期斗争的产物。

这个部分说的是限制政府的权力，那对限制这个词的理解又是怎样呢？在集体行动中，人们愿意对集体的可能行为预先设立限制，这个限制不是针对自己而是他人。也就是说个人可能愿意对集体的行动施加具体的限制，但如果是私人行动的角度，就不会有自我强加的限制法则。这个其实也是很容易理解的，人都是带有自私性的，在个人利益与具体利益间，大部分人都是选择自己，这是人之常情。对限制

的具体理解还要具体的针对发出行动的主体来理解，不然就会有失偏颇，不同的主体代表不同的利益，同时各自的主体又都希望各自的利益达到最大化，所以都会希望对其他主体的行为进行限制，很大程度上的对个人利益的侵害都是建立在主体的基础上的。

对政府的权力进行的限制，很大的程度上还是这些拥有权力的主体已经不再是一个个单个的个体，他们都已经变成一个个个体联合在一块的利益集团。说到这边，我们不难发现，布坎南通过一个例子以及图标来展现具体的利益关系，利益关系到了这么一个地步，已经不再是简简单单的利益单方面的关系，利益的涉及已经是千丝万缕的了。在利益集团中，任何一方都不会轻易伤害其他人的利益，这些都是不符合各自的利益出发点的。这么多的利益集团联系在一块，要想很好的平稳的分配这些利益，又要避免大的利益剧团发生侵吞的危险，那些小的利益剧团也会因此提出政治权利的限制，这样对于那些大的利益剧团而言，虽然会损失一些，但对于长远的发展而言，他们还是可以接受的，所以针对利益集团的权力限制，已经不仅仅是个人所希望的，同时也是利益集团内部大小团块相互妥协的产物。、

说到这边，我们不得不说一下我们天朝存在的一些现象。虽然我们是在名义和事实上都是社会主义国家，社会主义讲究的是大家平等，各方面都不会相差太多，更别提利益集团会在社会主义国家出现了，但事实上却是出现了，这不过这些利益集团的出现都是披上国家的外衣，堂而皇之的出现并且理所当然的享受国家给予的各项补贴。这些利益集团在过去很大程度上为国家做了很大的贡献，如就业方

面，税收方面等，但经历了改革开放30多年的洗礼，这些企业早已不是当初的企业，他们虽然如今依然披着国家的外衣，但他们却已经不再像过去那样为国家的发展提供各方面的支持了，他们现在有的只是钱。现在谈到这些企业，普通人都是表示羡慕，都以能进入其中工作为荣，为什么呢？在过去人们进入其中都是为国家的发展贡献自己的一份力量，而如今确实进入其中意味着高福利和高薪水以及各项好处，这些企业虽然规模很大，但有时候却不思进取，对于后来居上者，它也会运用国家权力对其进行打压，这已经是司空见惯了。对于这些企业所代表的利益，我们不去过分追究与探讨，但对于它们未来的发展我们却是不得不进行改革，改革改的不仅仅是企业里面的各项规章，同时也是这些企业所拥有的权力，只有对这些企业所代表的利益集团说不，让这些回归一个正常企业所应有的模样，才能真正让这些企业为国家的发展做贡献，为人民的利益做贡献，为社会主义的发展做贡献。

政权和政体的健康发展都必须有一个健康体制的支撑，而这个健康体制的存在很大程度上需要将体制内的各项权力进行明确限制，让其在各自的轨道上井然有序的行进。

第二篇：Auditor choice and the cost of debt capital for newly public firms

Weappreciatehelpfulcommentsfrom,JoyBegley,MarkDeFond,ShaneDikolli,BillKinney,KenKlassen,CliveLennox,ThomasMatthews,GordRichardson,JayRitter,DanSimunic(thereferee),BillySoo,RossWatts(theeditor),andparticipantsatthe2003InternationalSymposiumonAuditResearch,the2003AAAAuditSectionMidyearConference,andthe2001CAAAAnnualConference.Finally,wegratefullyacknowledge?nancialsupportfromtheSocialSciencesandHumanitiesResearchCouncil.*Correspondingauthor.Tel.:+1-709-737-3100;fax:+1-709-737-7680.

E-mailaddresses:jpittman@mun.ca(J.A.Pittman),steve.fortin@mcgill.ca(S.Fortin).

0165-4101/$-seefrontmatterr2003ElsevierB.V.Allrightsreserved.

doi:10.1016/j.jacceco.2003.06.005$

114J.A.Pittman,S.Fortin/JournalofAccountingandEconomics37(2004)113–136

1.Introduction

RajanandZingales(1998)reportevidenceofalifecycleinthepatternofcorporate?nancing,with?rmsmoredependentonexternal?nancingintheirearlyyears.However,marketfrictionssuchasasymmetricinformationthatcaninhibitthegrantingofcredittodeserving?rmsmaybemoresevereforyoung?rms(StiglitzandWeiss,1981).1Takentogether,thisresearchimpliesthattheinformationproblemsthatyoung?rmsexperiencecoincidewiththeyearsthattheymostdemandexternal?nancing.

Accordingly,?rms’earlypublicyearsprovideanopportunesettingforstudyingthein?uenceofauditorchoiceontheirinterestrates.WeexaminewhetherengagingaBigSixauditor,whichhasabrandnamereputationforsupplyingahigher-qualityauditthatenhancesthecredibilityof?nancialstatements,enablesyoung?rmstoreducetheirborrowingcosts.WepredictthattheimportanceofdebtmonitoringbyaBigSixauditorgraduallysubsideswith?rmage.Asinformationinthecapitalmarketsonyoung?rmsbecomesmoreavailable,thein?uenceon?rms’interestratesofrelyingonaBigSixauditortolowermonitoringcostsshoulddecreaseovertime.Speci?cally,weestimatethetime-seriesvariationintheimpactofauditorchoiceon?rms’interestratesintheir?rst9publicyears.Consistentwithourpredictions,wereporteconomicallysigni?cantandstatisticallyrobustevidencethathiringaBigSixauditorlowers?rms’costofdebtcapitallesswithage.Wealso?ndthat?rmswithshortprivatehistoriesbene?tmorefromchoosingaBigSixauditorintheirearlypublicyears,thoughthiseffectondebtpricingsubsidesrelativelyquicklyastheseyounger?rmsbecomebetterknowntolenders.Overall,ourresearchsuggeststhattheeconomicvalueofauditorreputationdeclineswithageasborrowersshifttowardexploitingtheirownreputationstoreduceinformationasymmetry.

Althoughpriorresearchextensivelyexaminestheroleofauditorchoiceonthecostofequitycapitalforyoungpublic?rms,thereisvirtuallynoevidenceonitseffectondebtpricingforthese?rms.2Thispapercontributestoourunderstandingofthisissuebyestimatingthein?uenceofauditorreputationonmoderatingcostlyinformationproblemsindebtmarkets.Inseparatetests,wemeasure?rmageasthenumberofyearssinceitsinitialpublicofferingandsinceitsincorporation.Thesetestsprovideevidenceonthejointimpactofauditorchoiceand?rms’(publicandprivate)experienceontheirinterestrates.

Therestofthispaperisorganizedasfollows.Section2furtherdevelopsthemotivationfortheempiricaltests,whileSection3describesthesampleformationprocess.Section4speci?esandreportsdescriptivestatisticsontheregressionInfact,empiricistsfrequentlyspecifyagetomeasuretheextentofexanteuncertaintyabout?rmvalue;e.g.,Beatty(1989)andRitter(1991).

2Thereisotherevidenceonthedebtpricingimplicationsofauditquality.McKinleyetal.’s(1985)surveyofbankloanof?cers’perceptionsofprivate?rms’creditapplicationsindicatesthatauditorreputationdoesnotaffectloanpricingdecisions.Wallace(1978)?ndsthatchoosingaBigEightauditorlowersinterestcostsformunicipalities.Byexaminingpublic?rms,ratherthanprivate?rms,welargelyavoidcomplicationsfromthepersonalreputationsand?nancialaffairsofownersnotbeingeconomicallydistinctfromtheir?rms.1

J.A.Pittman,S.Fortin/JournalofAccountingandEconomics37(2004)113–136115variablesfortheteststhatestimatetheimpactofauditorreputationon?rms’interestrates.Section5presentstheestimationmethodologyandresultsforthedebtpricingtests.Finally,theconclusionsinSection6discussdirectionsforfutureresearch.

2.Extantresearchandhypothesesdevelopment

Inthisstudy,werelatetheinformationalasymmetrythatisprevalentin?rms’earlypublicyearstoauditorchoiceindebtmarkets.DeAngelo(1981)andDataretal.(1991)arguethatlarge,prestigiouspublicaccounting?rmsconcernedaboutprotectingtheirinvestmentinreputationcapitalhavemoreincentivethanotherauditorstosupplyahigh-qualityaudit.3Balversetal.(1988)andBeatty(1989)?ndthathigh-reputationauditorsenableentrepreneurstoreducetheextentofexanteuncertaintyinnewequityissues.4CopleyandDouthett(2002)reviewtheconsiderableliteratureonthelinkbetweenauditqualityandthecostofcapitalin?rms’initialpublicofferings.Incontrast,noempiricalstudiesexaminethedebtpricingimplicationsofauditorselectionbypublic?rms.

Priorresearchsuggeststhatinformationproblemssteadilysubsidewithageas?rms’accumulateahistoryinthecapitalmarkets.Forexample,Lang(1991)providestheoryandevidencethatthemagnitudeofstockpricereactionstoearningsannouncementsdiminishwithage,whichheattributestothegradualrevelationof?rm-speci?cinformationovertime.Similarly,Diamond(1989)arguesthatyoung?rmssuffermoresevereassetsubstitutionandmoralhazardproblems.Hemodelsthedynamicsofborrowers’incentiveswithlenderslearningovertimefromobserving?rms’creditrecords.

Dattaetal.(1999)reportevidenceconsistentwithDiamond’s(1989)predictionthat?rmslowertheirinterestratesbydevelopingtheirreputations.Lendersmaypreferthatyoung?rms,whicharejustformingtheirreputationsfordebtservicing,havehigher-qualityaudits.Thisimpliesthatthemarginalbene?tthatahigher-qualityauditprovidesthroughlowering?rms’borrowingcostswillbedecreasinginage.Essentially,auditorreputationsubstitutesforthe?edglingreputationsof?rmswithshortcredithistories.

Engagingahigh-reputationauditormaycontributetotheef?cientresolutionofcontractingproblemsbyproducingvaluableinformationaboutborrowers(seeJensenandMeckling,1976;WattsandZimmerman,1986).SincethereisonlyPriorevidencethatBigSix?rmsreceivebrandnamepricepremiaincompetitivemarketsalsoimpliesthatqualitydifferentiationexists;e.g.,Craswelletal.(1995).Craswelletal.explainthatalthoughallpublicaccounting?rmsmustcomplywithminimumprofessionalstandards,theBigSix?rmsvoluntarilyinvestinhigherlevelsofexpertiseandhaveincentivestoprovidehigher-qualityauditstoprotecttheirreputations.

4Aseparateresearchstreamexamineswhetherauditorselectionprovidesanexantesignalofinsurancecoveragetoinvestorsinequityinitialpublicofferings;e.g.,MenonandWilliams(1994).Inadditiontoresearchonnewshareissues,otherstudiesarguethatthelargerpublicaccounting?rmsarebrandnamesuppliersofhigher-qualityaudits;e.g.,TeohandWong(1993)andBeckeretal.(1998).3

116J.A.Pittman,S.Fortin/JournalofAccountingandEconomics37(2004)113–136

limitedinformationavailableonyoung?rms,lendersdependlargelyon?rms’disclosurestoevaluatetheirperformanceandfutureprospects.Lenderswouldhavetoconductcostlyinformationproductionandmonitoringusingalternativesourcesintheabsenceofcredible?nancialstatementsfortheserelativelysmall,obscure?rms.Thiscredibility,whichaBigSixauditorcanenhanceby,forexample,improvingtheprecisionin?rms’earningsreports(DeAngelo,1981;Balversetal.,1988),lowersthemonitoringcostsofdebtholders.

ExtantresearchsuggeststhatBigSixauditorsprovidesuperiormonitoring,especiallyfor?rmssufferingfromsevereinformationproblems.Francisetal.(1999)?ndthat?rmswithotherwiserelativelyhighuncertaintyaboutreportedearningsareinducedtohireaBigSixauditortobolsterthecredibilityoftheir?nancialstatements.Theyreportevidencethatthisexternalmonitoringconstrainsaggressiveandpotentiallyopportunisticreportingofaccruals-basedearnings.Inaddition,Beckeretal.(1998)documentthat?rmswithBigSixauditorshavelowerdiscretionaryaccruals,whichtheyargueisindicativeofhigh-qualityauditorspermitting?rmslessaccounting?exibility.TeohandWong(1993)reportthatclientsofBigSixauditorshavehigherearningsresponsecoef?cients.

Overall,thisevidenceimpliesthatthehigherqualityofBigSixauditsreducesearningsmanagement,whichisimportantsincedebtcontractsengenderincentivestoadjustearningsthroughaccountingchanges(Sweeney,1994)anddiscretionaryaccruals(DeFondandJiambalvo,1994)toavoidaccounting-basedcovenantviolations.Further,Teohetal.(1998)?ndthatearningsmanagementpriortoinitialequityissuesaffectsaccrualsin?rms’earlypublicyears.Thereductioninuncertaintyaboutthequalityof?rms’earningswouldjustifylendersconsideringauditorchoicewhenpricingdebtcontracts.Sinceuncertaintyinthecapitalmarketsabout?rms’earningsisdecreasingintheirage(Lang,1991),young?rmsmayparticularlybene?tfromchoosingaBigSixauditor.Intheabsenceofsuchmonitoring,lenderswillpriceprotectagainstexcessiveearningsmanagementbyrequiring?rmsnotretainingaBigSixauditortopayhigherinterestrates.

Iflendersgraduallyrelymoreon?rmreputationandlessonauditorchoicewhendetermininginterestrates,thenthein?uenceofselectingaBigSixauditorshouldbemoredetectableinasampleofyoung?rmsthathaveonlyshorttrackrecords.Thisstudybeginsbyexamining?rmsovertheir?rst9publicyearstoestimatetheevolutionintheimpactofauditorreputationonborrowingcosts.Weexpectthatthehigh-qualitymonitoringofaBigSixauditorwillenableyoung?rmstobettermoderateinformationproblemsintheirearlypublicyears,whichleadstothefollowinghypothesis(statedinalternateform):

H1:Thein?uenceofauditorreputationonlowering?rms’interestrateswill

subsidewithage.

Ourinitialtestsdeliberatelyignorethat?rmstypicallyhaveprivateoperatingexperiencewhentheygopublic.FirmsthatalreadyhavelengthycredithistoriesattheirIPOdatesmaynotbeabletofurthermateriallyreduceinformationasymmetrybyhiringaBigSixauditor.These?rmscandependontheirowngoodreputations

J.A.Pittman,S.Fortin/JournalofAccountingandEconomics37(2004)113–136117withlenderstolowertheirinterestrates,therebylargelyeliminatingthedebt-monitoringbene?tofhavingahigh-qualityaudit.Infact,Ritter(1991)?ndsthatstockperformanceimprovesthelongerthedurationbetweenincorporationandinitialpublicoffering,whichwouldvalidatelendersperceivingcreditrisktobehigherfor?rmswithlessprivateexperience.5Incomparisonto?rmswithlongerprivatehistories,lenderslearnmoreaboutyounger?rmsintheirearlypublicyears.Althoughauditorchoicewillaffectyounger?rms’borrowingcostsmore,itsimpactwilldissipatemorequicklyovertimeastheybecomebetterknown.Thisargumentmotivatesthefollowinghypotheses(alsostatedinalternateform):

H2A:Thein?uenceofauditorreputationonloweringinterestrateswillbestronger

in?rmswithshortprivatehistories.

H2B:Thein?uenceofauditorreputationonloweringinterestrateswillsubsidemore

quicklywithagein?rmswithshortprivatehistories.

3.Sampleselection

Thesampleweusetoexaminethedeterminantsofdebtpricingformaturing?rmsisextractedfromthosethatwentpublicfrom1977to1988.Alistofthe3458SEC-registeredinitialpublicofferingsconductedbetweenJanuary1,1977andDecember31,1988wasobtainedfromSecurityDataCorporation.WecomparethislisttotheAnnualIndustrialandResearchCompustat?lestoidentifytheIPOsforwhichanydataisavailable.Table1reportsthatthisscreeningreducesthesampleto2180?rms.Weexcludesomeindustriesfromthesamplebecausetheircapitalmarketdecisionsarefundamentallydifferentfromthoseofother?rmsduetoregulationandthe?nancialnatureoftheiroperations(seediscussioninPittmanandKlassen,2001).Wedeleteutilities(SICcodesfrom4911to4941)since?rmsinregulatedindustriesmighthaveincentivestohaverelativelyhighleverage.6Utilities’stablecash?owsmayincreasetheirdebtcapacity,whiledelegatingdiscretionoverinvestmentpolicytoregulatoryauthoritiesmayreducetheirassetsubstitutionandunder-investmentproblems.

Inaddition,weremove?rmsin?nancial(SICcodesfrom6022to6200),insurance(SICcodesfrom6312to6400)andrealestate(SICcodesfrom6500to6799)industries.Explicit(orimplicit)investorinsuranceschemessuchasdepositinsurancemaystronglyin?uencethecreditdecisionsof?rmssuchasbanksandinsurancecompanies.Infact,theirdebt-likeliabilitiesarenotstrictlycomparabletothedebtissuedbynon-?nancial?rms.Further,regulationssuchasminimumcapitalrequirementsmayaffecttheircapitalstructures.Theindustryscreeningeliminates249?rmsfromthesample.

Otherempirical(e.g.,LjungqvistandWilhelm,2003)andanalytical(e.g.,ChemmanurandFulghieri,1999)researchsuggeststhatthereismoreuncertaintyabout?rmsgoingpublicearlierintheirlives.6Otherindustrieswereregulatedatdifferentpointsduringthe1977–1997periodofthisstudy.Allofourresultswerevirtuallyidenticalwhen?rmsoperatingintherailroad(SIC4011),trucking(SIC4213),airline(SIC4512),andtelecommunications(SIC4812and4813)industrieswereremovedfromthesample.5

Table1

Sampleselectionsummary

Calendaryear197719781979198019811982198319841985198619871988TotalNumberofSECregisteredIPOs323862xxxxxxxxxxxx573557284151673458Numberof?rmsnotfollowedbyCompustatsince(11)(20)(22)(67)(125)(40)(246)(129)(131)(290)(150)(47)(1278)initialpublicoffering

Numberof?rmsfromutilities,?nancial,insurance,(2)(3)(4)(11)(26)(7)(51)(24)(29)(53)(27)(12)(249)andreal-estateindustries

Numberof?rmsthatdidnotsurvivethroughtheir(14)(2)(13)(33)(99)(47)(210)(123)(110)(195)(123)(65)(1034)?rst9yearsofpublicoperation

Numberof?rmswithmissingCompustat(2)(2)(4)(8)(16)(6)(36)(21)(22)(49)(29)(9)(204)observations

Numberof?rmswithextremeobservations(3)(8)(9)(14)(42)(13)(72)(25)(30)(51)(41)(14)(322)Numberof?rmsinthesample031016409703533904520371118J.A.Pittman,S.Fortin/JournalofAccountingandEconomics

37

(2004)

113–136

J.A.Pittman,S.Fortin/JournalofAccountingandEconomics37(2004)113–136

119

Thisstudyinitiallyfollows?rmsfortheir?rst9publicyears,whichresultsinremovingthosethatweredelistedfromtheCompustatdatabasebecauseofmergers,acquisitions,bankruptcy,orliquidationduringthisperiod.Restrictingthesampletothe?rmsthatsurvivefor9consecutiveyearsensuresthatanyevidencesupportingourpredictionsisnotdrivenbychangestothecompositionofthesampleovertime.

Wealsoremove?rmsifanyoftheobservationsnecessarytoconstructtheregressionvariablesaremissing,whichlowersthesampleto693?rms.Finally,discardingextremeobservationsinthedata,whichweexplaininSection4,reducesthe?nalsampleto371?rms(seeTable1forasummary).Asimposingthesescreeningcriteriaresultsinattritionof89percentoftheoriginalsample,the?nalsamplemaydiffersystematicallyfromthepopulationofsurvivors.Althoughinspectionofthedatasuggeststhattheindustry(Table2)andcalendaryearclusteringinthesampleresemblestheclusteringintheinitialpopulation,wemoreformallyconsidertheeconometricimplicationsof?rmattritiononthesampleselectionprocessinSection5.

Table2

IndustrydistributionofsampleTwo-digitSICcode13202728353637383942484950565859738087

Number9127172944631691171087122179

Industrydescription

OilandgasextractionFood

PrintingandpublishingChemicals

IndustrialandcommercialmachineryElectronicandelectricalequipmentTransportationequipment

Measuring,analyzing,andcontrollingequipment

MiscellaneousmanufacturingindustriesFreighttransportationandwarehousingCommunications

Electric,gas,andsanitaryservicesWholesaletrade—durablegoodsApparelandaccessorystoresRestaurants

MiscellaneousretailBusinessservicesHealthservices

Engineering,accounting,researchmanagement,andrelatedservices

Percentage2.43.21.94.67.811.81.68.41.62.43.01.92.72.21.93.25.71.92.4

Sub-total

262109371

Industrieswithfewerthansix?rms

70.629.4100

Total

120J.A.Pittman,S.Fortin/JournalofAccountingandEconomics37(2004)113–136

4.Regressionvariablesanddescriptivestatistics

Thissectiondescribesthedependentandexplanatoryvariablesthatareusedintheempiricaltestsofthehypotheses.Weemphasizethespeci?cationofthethreevariablesofprimaryinterestinthisstudy,?rms’interestrates,ages,andauditorchoices.Wealsomotivateandspecifythecontrolvariablesthatrepresentotherpotentialdeterminantsofdebtpricing;i.e.,theunderlyingcostofcapitaland?rm-speci?ccharacteristics.Thissectionconcludeswithdescriptivestatisticsforthesevariables.

4.1.Interestrate

Thedependentvariableistheinterestrateonthe?rm’sdebt,whichiscalculatedasitsinterestexpensefortheyeardividedbyitsaverageshort-andlong-termdebtduringtheyear.However,descriptivestatisticsindicatethatthisvariableisanoisyproxyforthe?rm’sinterestrateunlesswetrimthedatatoaddressextremeobservations.

Ifa?rm-yearobservationisoutsidethe5thor95thpercentilesofthepooleddistribution,thenwediscardthe?rmunlesswecanreplacetheseextremeobservationswithinterestratescalculatedusingquarterlydata.7Droppingobservationsisoftenappliedtohandlenoiseresultingfromaccountingratioshavingverysmalldenominators(Dechow,1994).Thisrestricts?rms’interestratestorangetoamaximumof20.7percent,whichisabout180basispointshigherthanthehighestprimerateinthe1977–1997studyperiod(Table3).Thesamplecontains?rmswithnodebtoutstandingduringtheyearthathave0percentinterestrates.8Althoughtrimmingexplanatoryvariablesdoesnotaffecttheexpectedvalueoftheregressionslopes,excessivetrimmingcanin?atetheirstandarderrors.Thetrimmingprocedurethatweapplygeneratesstandarderrorsthataresmallerthanfortheregressioncontainingallobservationsorforthosethatusealternatetrimmingmethods.9

Forthequarterlydata,wecomputeyearlyinterestratesastheannualizedaverageofthefourquarterlyinterestrates.Whileitmayhavebeenpreferabletoestimatealldatapointsusingquarterlydata,thequarterlyCompustattapecoversfewer?rmsthanitsannualtape,particularlyintheearlysampleyears.AllresultsonhypothesesH1,H2A,andH2Barerobusttousingstrictlyannualorstrictlyquarterlydata.8AllevidenceonH1,H2A,andH2Bremainsafterexcluding?rmswith0percentinterestratesandafterrunningtobitregressionstoaddresspotentialcensorshipbiasarisingfromtreatinginterestrateasacontinuousvariable.

9WeexaminethepossibilitythatthetrimmingprocedurechosenspuriouslyinducestheevidenceonthehypothesesbyapplyingthefollowingalternateproceduresforhandlingextremeobservationsanddatacodingerrorsmadebyCompustat.Inseparatetests,wediscardobservationsoutsidethe1stand99thpercentilesandwinsorizeatthe5thand95thandatthe1stand99thpercentiles.Inaddition,anotherspeci?cationconstrainsthe?rms’interestratestorangefromthemeanprimerateforthatyearandtheprimerateplus0.075;i.e.,observationsbelowprimearesettoprimeandthoseaboveprimeplus0.075aresettoprimeplus0.075.Wealsore-specifytoconstraintheirinterestratestorangefromprimetoprimeplus0.05andprimetoprimeplus0.10.Allresultsreportedareessentiallyunaffectedbyanyofthesere-speci?cations.7

J.A.Pittman,S.Fortin/JournalofAccountingandEconomics37(2004)113–136121

4.2.Firmage

ThetheoryofreputationformationindebtmarketsbyDiamond(1989)predictsthatinterestrateswilldeclineovertimeas?rmscompilegoodcredithistories.Wemeasure?rmageasthenumberofyearssincetheywentpublicwhentestingH1.ForH2AandH2B,weexaminewhethercross-sectionaldifferencesin?rms’privateages,whichisthenumberofyearsbetweentheirincorporationandinitialpublicoffering,affectthein?uenceofauditorchoiceon?rms’borrowingcostsintheir?rst9publicyears.

4.3.Auditorchoice

Oursamplehasobservationsthatcovertheperiod1977–1997duringwhichmergersinvolvingthelargepublicaccounting?rmsoccurred.Forthe1989–1997portionofoursampleperiod,wefollowpreviousresearchbyspecifyingauditorchoicewithavariableindicatingwhetherthe?rmretainsaBigSixauditor;e.g.,Francisetal.(1999).Forthe1977–1988portionofoursampleperiod,thisindicatorvariableidenti?espublicaccounting?rmsthatwereamongtheBigEight;e.g.,Beatty(1989).10

4.4.Underlyingcostofcapital

Welargelyrelyontheaverageprimeratefortheyeartocontrolforchangesintheunderlyingcostofcapital.Therisk-freerateandadefaultpremiumforthelender’sbestcustomerscomprisetheprimerate.Firmsthatarenotamongthelender’sbestcustomersincuranadditionaldefaultpremium.TheannualdifferencebetweentheyieldonBAA-ratedcorporatebondsandtheyieldon10-yeargovernmentbondscontrolsforaggregatevariationinthisdefaultpremium(PetersenandRajan,1994).

4.5.Firmcharacteristics

Thefollowingvariablescontrolforvariationindebtpricingattributableto?rm-speci?ccharacteristicsotherthanitsreputationandauditorchoice.Thesecontrolsproxyforcreditrisk,withobservablyriskierborrowerspredictedtopayhigherinterestrates.

Wepredictaninverserelationbetweeninterestratesand?rmsize,whichismeasuredwiththenaturallogarithmofoneplustotalassets,sincecreditorsperceivelarger?rmsaslessriskyandthereareeconomiesofscaleindebtproductioncosts(Careyetal.,1993).Thelogarithmicspeci?cationthatprovidesfortheexpecteddecreasingmarginalimpactofsizefollowsextantresearch;e.g.,Blackwelletal.(1998).

Forexpositionalconvenience,wealwaysuse‘‘BigSix’’todescribethelargepublicaccounting?rms,despitethatthesearereally‘‘BigEight’’?rmsbeforethe1989mergersthatresultedintheformationofDeloitte&ToucheandErnst&Young.10

Table3

Summarystatistics

Panel(A)Continuousvariables

VariableNMeanStd.dev.MedianMinMax

Interestrate33390.0930.0400.09600.207

Age333952.582519

Primerate33390.0910.0210.0880.0600.189

Defaultpremium33390.0200.0030.0200.0130.031

Leverage33390.2740.2390.22100.992

Cash?ow33390.0640.1650.085à4.5351.898

Size33393.6051.4203.4770.32310.954

Assetstructure33390.5610.8780.466029.017

Panel(B)Discretevariables

VariableNPercent

BigSixauditor298289.31

Negativebook912.73

equityindicator

Panel(C)Correlations

VariableInterestrateAgePrimerateDefaultLeverageCash?owSizeAssetBigSixNegative

premiumstructureauditorequity

Interestrateà0.1810.3110.0510.210à0.0870.0290.065à0.0530.089

(0.001)(0.001)(0.003)(0.001)(0.001)(0.094)(0.001)(0.002)(0.001)

Ageà0.145à0.415à0.4360.070à0.0500.1120.1540.0690.068

(0.001)(0.001)(0.001)(0.001)(0.004)(0.001)(0.001)(0.001)(0.001)122J.A.Pittman,S.Fortin/JournalofAccountingandEconomics37(2004)113–136

Primerate0.262à0.476à0.2030.0120.010à0.097à0.048à0.044à0.025

(0.001)(0.001)(0.001)(0.481)(0.544)(0.001)(0.005)(0.012)(0.153)

Def.premium0.084à0.4100.094à0.077à0.009à0.126à0.044à0.041à0.041

(0.001)(0.001)(0.001)(0.001)(0.598)(0.001)(0.011)(0.018)(0.018)

Leverage0.2250.081à0.019à0.094à0.3600.2260.2430.0060.180

(0.001)(0.001)(0.265)(0.001)(0.001)(0.001)(0.001)(0.709)(0.001)

Cash?owà0.047à0.039à0.0080.000à0.1490.1190.1630.025à0.173

(0.007)(0.024)(0.642)(0.982)(0.001)(0.001)(0.001)(0.142)(0.001)

Size0.0800.107à0.117à0.01350.2240.1600.0780.1880.036

(0.001)(0.001)(0.001)(0.001)(0.001)(0.001)(0.001)(0.001)(0.036)

Assetstructure0.0550.099à0.048à0.0210.0800.218à0.0440.0310.040

(0.001)(0.001)(0.006)(0.220)(0.001)(0.001)(0.011)(0.074)(0.020)

BigSixauditorà0.0550.069à0.062à0.0530.0180.0150.1810.0230.034

(0.001)(0.001)(0.001)(0.002)(0.300)(0.396)(0.001)(0.189)(0.049)

Neg.bookequity0.0760.068à0.040à0.0420.221à0.2630.0470.0050.034

(0.001)(0.001)(0.021)(0.016)(0.001)(0.001)(0.007)(0.765)(0.049)

Notes:Thistablepresentssummarystatisticsforthe3339?rm-yearobservationsovertheperiod1977–1997usedinthehypothesestestsonthefullbalancedpanel.(A)presentsthedistributionalstatisticsforthecontinuousvariablesand(B)presentsthedistributionforthediscretevariables.Tables1and2,respectively,providedescriptivestatisticsonthecalendaryearandindustrydistributions.(C)reportscorrelationsfortheregressionvariables,withPearsoncorrelationspresentedbelowthediagonalandSpearmancorrelationspresentedabovethediagonal;relatedprobabilityvaluesarepresentedinparentheses.Continuousvariablesareasfollows(alldollar-denominatedvariablesarestatedinmillions).Interestrateisinterestexpensedividedbytheaverageshort-andlong-termdebtfortheyear.Firmageisthenumberofyearsthathaveelapsedsincethe?rm’sinitialpublicoffering.Primerateistheaverageprimeratefortheyear.ThedefaultpremiumisthedifferencebetweentheyieldonBAA-ratedcorporatebondsandtheyieldon10-yeargovernmentbondsfortheyear.Leverageisthebookvalueofthesumofshort-andlong-termde?atedby?rmmarketvalue(thesumofmarketvalueofequityandbookvalueoftotaldebt).Cash?owiscash?owfromoperationsscaledbytotalassets.Firmsizeisthenaturallogarithmofoneplustotalassets.Assetstructureistotalproperty,plantandequipmentscaledbytotalassets.

Indicatorvariablesareasfollows.Theyeardummyvariablesindicatethecalendaryearinwhichthe?rmunderwentitsinitialpublicoffering.Theindustrydummyidenti?esthe?rm’sone-digitSICcode.TheBigSixauditorvariableindicatesifthe?rmretainsaBigSix(BigEightbefore1989)auditor.Thenegativebookequitydummyindicatesifthebookvalueofcommonequityisnegative.J.A.Pittman,S.Fortin/JournalofAccountingandEconomics37(2004)113–136

123

124J.A.Pittman,S.Fortin/JournalofAccountingandEconomics37(2004)113–136

Wealsopredictthatinterestrateswillbeincreasinginleverage,whichisde?nedastotalshort-andlong-termdebtscaledby?rmmarketvalue(PetersenandRajan,1994).Thecontrolforpro?tabilityiscash?owfromoperations(PetersenandRajan,1994),whichispredictedtohaveanegativecoef?cientsince?rmsthatcangeneratemorecashinternallyareinabetterpositiontoservicetheirdebts.Wespecifyadummyvariabletoidentifywhenthebookvalueofcommonequityisnegativetore?ectthat?rmsexperiencing?nancialdistressmayincurhigherborrowingcosts(Grahametal.,1998).

Priorevidencesuggeststhatinterestratesareincreasingincollateral(Johnetal.,2003,amongothers),whichisconsistentwiththeperceptioninthebankingindustrythatriskierborrowersmustprovidesecurityfortheirloans(Morsman,1986).Thisimpliesthatthecoef?cientonthecontrolforassetstructure,whichisthefractionoftotalassetsinproperty,plant,andequipment,willbepositive.

Weincludeanindicatorforissueyeartocapturelingeringvariationincreditrisk.11Table1reportsdramaticcyclicalswingsinthenumberofIPOsconductedbetween1977and1988.Thereisconsiderabledebateabouttheaveragequalityof?rmsgoingpublicinhotandcoldmarketsfornewissues(LoughranandRitter,1995).Finally,werelyonone-digitSICcodestocontrolforindustryin?uencesondebtpricing(Stoughtonetal.,2001).12

Table3reportsdescriptivestatisticsthatindicatethatthereissubstantialvariationacrossallcontinuousvariables.Thereisevidencethatsomecontrolvariables,especiallytheprimerate,defaultpremium,andsizeproxies,arecorrelatedwiththeBigSixindicatoror?rmage.Thisraisestheconcernthatotherdeterminantsofdebtpricingmightexplaintheobservedseparateandjointin?uencesof?rms’auditorchoiceandage.Toaddresspotentialmulti-collinearityissues,weremovefromtheequationsoneatatimeexplanatoryvariableswithabsolutecross-correlationswithauditorchoiceor?rmageexceeding0.15.Noneofthesere-speci?cationsqualitativelyaffecttheregressionresultspresentedbelow.Laterinthispaper,wefurtherexplorepotentialconfoundingfrom?rmsize.

5.Estimationmethodologyandresults

WebeginthissectionbydevelopingourresearchdesignforexaminingthepredictioninH1thatthedebt-monitoringbene?tthataBigSixauditorprovidesthroughlowering?rms’interestratesgraduallysubsideswithage.FortheempiricaltestsofthepredictionsinH2AandH2Bthattheimpactofauditorchoiceondebtpricingwillbestronger,butwilldissipatemorequicklyovertime,in?rmswithshortOurresearchdesignpartiallysuppressesthein?uenceofchangingmacroeconomicconditionsondebtpricingbyaligningthedataineventtimestartingwiththe?rms’?rstpublicyear,ratherthanincalendartime.Thisshouldreduceconfoundingadmittedbyshiftsinprevailingconditionsotherthanchangesintheprimerateanddefaultpremium,forwhichwespecifyseparatecontrols.

12Althoughthisrequiresspecifying50industrydummies,re-runningthetestsusingthemoreconventionaltwo-digitSICcodedummiesdoesnotqualitativelyaffecttheresultsreportedinthispaper.11

J.A.Pittman,S.Fortin/JournalofAccountingandEconomics37(2004)113–136125privatehistories,wesimplydividethesampleusedtotestH1accordingto?rms’medianprivateagewhentheywentpublic.

5.1.PrimarytestsofH1:Firms’publicexperience

Weinitiallyestimatethefollowingmodeltoexaminetheevolutionintheimpactofauditorchoiceon?rms’borrowingcosts:

INTERESTRATEit?atbXitteit;e1TwhereINTERESTRATEitistheinterestrateincurredby?rmiattimet,aistheinterceptthatiscommontoall?rms,Xitisavectorofexplanatoryvariables,andeitistheerrorterm.Theregressionvariables,whichSection4describesinmoredetail,are:

INTERESTRATEitInterestrateisinterestexpensedividedbytheaverageoftotal

short-andlong-termdebtduringtheyear.

AGEitAgeisthenumberofyearsthathaveelapsedsincethe?rm’s

IPO.

BIGSIXitThisindicatorvariablehasavalueofonewhenthe?rm

retainsaBigSixauditor;zerootherwise.?AGEitBIGSIXitThisinteractionistheproductofageandtheBigSixauditor

indicatorvariable.

PRIMEtPrimeistheaverageprimeratefortheyear.DEFAULTtThedefaultspreadisthedifferencebetweentheyieldon

BAA-ratedcorporatebondsandtheyieldon10-year

governmentbondsfortheyear.

LEVERAGEitLeverageisthebookvalueoftotalshort-andlong-termdebt

de?atedby?rmmarketvalue(sumofthemarketvalueof

equityandthebookvalueoftotaldebt).

CASH-FLOWitCash?owiscash?owfromoperationsscaledbytotalassets.SIZEitFirmsizeisthenaturallogarithmofoneplustotalassets.ASSETSitAssetstructureistotalproperty,plantandequipmentscaled

bytotalassets.

NEG.EQUITYitThenegativebookequitydummyindicatesifthebookvalue

ofequityisnegative.

YEARiThisisthecalendaryearinwhichthe?rmhaditsinitial

publicoffering.

INDUSTRYiThisindicatesthe?rm’sone-digitSICcode.

WestartbyestimatingthismodelusingordinaryleastsquaresundertheassumptionthattheerrorterminEq.(1)isindependentoftheexplanatoryvariables.Thisgeneratesresultsthataredirectlycomparabletopriorresearchonthedeterminantsof?rms’interestrates.Thistestalsoprovidesthe?rstcross-sectionalevidenceonwhetherauditorchoiceaffectsdebtpricing.However,thepaneldatatests,whichcontrolforunobserved?rm-speci?ceffectstoavoidomittedvariable

126J.A.Pittman,S.Fortin/JournalofAccountingandEconomics37(2004)113–136

biasandtore?netheobservationofwithin-?rmdynamics,thatfollowareourprimaryevidenceonH1andH2B.

Thesigni?canceofthepooledOLSresultsreportedincolumn(1)ofTable4arecalculatedwithstandarderrorsobtainedusingWhite’s(1980)heteroscedasticitycorrection.Thenegativeandstatisticallysigni?cantcoef?cientontheagevariablesupportsDiamond’s(1989)theorythat?rms’interestrateswilldeclineovertime.

Morerelevanttothisstudy,thenegativeandstatisticallysigni?cantcoef?cientontheBigSixauditorvariableimpliesthatrelyingonahigh-qualityauditorreduces?rms’interestrates,onaverage,by143basispoints.13Toourknowledge,thisisthe?rstestimateofthedebtpricingimplicationsofauditorselectionbypublic?rms.Inthisregression,weevaluatethetime-seriesvariationinthein?uenceofauditorreputationonborrowingcostsbyinteracting?rmagewiththeBigSixauditorindicator.Thepositiveandstronglystatisticallysigni?cantcoef?cientonthisinteractionisconsistentwiththepredictioninH1thatretainingaBigSixauditoraffects?rms’interestrateslesswithage.

Exceptfortheestimatesfor?rmsizeandthedefaultpremium,thepooledOLSresultsforthecontrolvariablesarestatisticallysigni?cantinthepredicteddirections.14Infact,theresultsreportedincolumn(1)ofTable4aresimilartothoseprovidedbypriorcross-sectionalstudies;e.g.,PetersenandRajan(1994)andBlackwelletal.(1998).

However,theseresults,particularlytheevidencesupportingH1,maybedrivenbyomittedvariablebiaswiththeexplanatoryvariablesrepresentingunobservable?rm-speci?ceffects.Weaddresstheconcernthat?rm-speci?ccorrelatedomittedvariablesarepresentinthedatabyspecifyinganinterceptforeach?rm,ai:Thiseliminatesthetime-invariantportionoftheerrorterminEq.(1),therebyyieldingconsistentcoef?cientestimates:

INTERESTRATEit?aitbXitteit:e2TTheresultsreportedincolumn(2)ofTable4areforatwo-way?xedeffectsmodelwithcorrectionforunspeci?edheteroscedasticity.AnF-teststronglyrejectsthenullhypothesisthattheconstantterms(the?xed?rmeffects)areallidentical.Inthepresenceofthe?rmeffects,anotherF-testsupportsincludingtime-speci?cparameterscorrespondingtotheprimerateandthedefaultpremiumfortheyear.

Thoughstillstatisticallysigni?cantatthe1percentlevel,thiscoef?cientestimatedropstoà0.0058whentheinteractionbetweenauditorchoiceand?rmageisremovedfromthiscross-sectionalregression.14AlthoughthepaneldatateststhatfollowprovideourmainempiricalevidenceonH1,thestatisticalsigni?canceofthesecross-sectionalresultsmaybeexaggeratedbytheserialcorrelationoftheerrorterms.However,theevidence(p-valueunder0.01)thatthein?uenceofauditorqualityoninterestratesisdecreasingin?rmageremainswhenthepooledequation(1)isre-estimatedusingGeneralizedMethodofMoments(seeHansen,1982),whichproducesstandarderrorsthatarenotaffectedbyserialcorrelationofsuccessiveobservationsandcross-sectionalheteroscedasticity.Inaddition,exceptfortheproxyforcash?owfromoperations,theothervariablescontinuetobesigni?cantinthepredicteddirections.13

Table4

Debtpricingresults—OLSand?xedeffectsestimation

OLSregression:INTERESTRATEit=a+b1AGEit+b2BIGSIXit+b3AGE?BIGSIXit+b4PRIMEt+b5DEFAULTt+b6LEVERAGEit+b

+bINDUSTRY7CASH-

FLOWit+b8SIZEit9ASSETSit+b10NEG.EQUITYit+fi+dYEARi+eit

Fixedeffectsregression:INTERESTRATEit=ai+b

SIZE1AGEit+b

e2BIGSIXit+b3AGE?BIGSIXit+b4PRIMEt+b5DEFAULTt+b6LEVERAGEit+b7CASH-

FLOWit+b8it+b9ASSETSit+b10NEG.EQUITYit+it

VariablePredictionBalancedpanelsUnbalancedpanels

FullSize-truncatedSize-matchedFullRestricted

PooledOLSFixedeffectsPooledOLSFixedeffectsPooledOLSFixedeffectsPooledOLSFixedeffectsPooledOLSFixedeffects

12345678910

Intercept?0.0703???0.1023???0.0677???0.1001???0.1045???0.0634???0.1073???Ageàà0.0031???à0.0024???à0.0031???à0.0024???0.0606???0.0858???0.0336???

à0.0032???à0.0032???à0.0017???à0.0022???à0.0022???à0.0025???

BigSixauditoràà0.0143???à0.0014à0.0143???à0.0007à0.0116??à0.0023à0.0102???à0.0017à0.0095???à0.0037Age?BigSixauditor+0.0019???0.0016???0.0018??0.0016??0.0023??0.0019??0.0007?0.0009??0.0008?0.0012???Primerate+0.3297???0.3338???0.3326???0.3373???0.3454???0.3454???0.4035???

Defaultpremium+0.17400.05130.17920.04830.22590.11360.5964???0.2726???0.3654???0.3147???

à0.15570.4849???0.0227

Leverage+0.0326???0.0101???0.0371???0.0120???0.0260???0.0118?0.0460???0.0193???0.0210???0.0088???Cash?owàà0.0069??à0.0046à0.0080??à0.0040à0.00370.00090.0002à0.0003à0.0025?à0.0012Sizeà0.0031???à0.0019?0.0036???à0.00150.0026???0.0062???0.0022???0.00030.0012???à0.0026???Assetstructure+0.0038???0.00060.0039???0.00070.0134???0.0140??0.0034???0.0013?0.0018???0.0004Negativebookequity+0.0102???0.0089??0.0011??0.0092??0.0210???0.0134??0.0083???0.0034?0.0096???0.0058??AdjustedR20.1580.5530.1690.5540.2100.5280.1320.5590.0970.492F-statistic27.17??106.69??27.85??103.46??10.86??92.22??49.48??63.95??27.98??84.61??Numberofobservations3339333931xxxxxxxxxxxx87633763360386038

Notes:Thistablepresentsregressionresultsforthedebtpricingmodelsusingpooledcross-sectional,time-seriesordinaryleastsquares(OLS)andtwo-way?xedeffectsestimationforthe?rms’?rstthroughninthyearsofpublicoperation.Thefullbalancedsampleiscompiledbydiscardingtheentiretime-seriesof?rmsifanymissingobservationsareencounteredinthe9years.Thesize-truncatedsub-sampleconsistsofall?rmsfromthefullbalancedpanelwithtotalassetsrangingfrom$0.6million(theminimumtotalassetsinthesetof?rmsusingaBigSixauditor)and$291.9million(themaximumtotalassetsinthesetof?rmsusinganon-BigSixauditor).Thesize-matchedbalancedsub-sampleconsistsof56?rmsusingaBigSixauditorand56?rmsusinganon-BigSixauditorthatwerematchedascloselyaspossibleontotalassets(themaximumpermissibledifferenceintotalassetswas10percent).Thefullunbalancedpaneliscompiledbydiscardingonlythe?rmyearwhenmissingobservationsareencountered.Therestrictedunbalancedpanelallows?rmstoleaveanytimeduringtheir?rst9publicyears,butnotreturnlatertothesample.Thedependentvariableandtheexplanatoryvariablesarespeci?edinTable3.TheOLSregressionsincludeunreporteddummyvariablesidentifyingthe?rm’sone-digitSICcodeandthecalendaryearofitsinitialpublicoffering.RegressionequationF-statisticssigni?cantatlessthan0.001areidenti?edbya??superscript.Inthistable,thesubscriptsiandtidentify?rmsandtime,respectively.Thesuperscriptasterisksindicateexplanatoryvariablecoef?cientsigni?canceatp-valueslessthan0.10(?),0.05(??),and0.01(???)inone-tailedtestswheredirectionalpredictionsaremadeandtwo-tailedtestsotherwise.J.A.Pittman,S.Fortin/JournalofAccountingandEconomics37(2004)113–136127

128J.A.Pittman,S.Fortin/JournalofAccountingandEconomics37(2004)113–136

The?xedeffectsresults,whichpreservethetime-seriesvariationwhileaccountingforindividual?rmheterogeneity,alsoprovideevidenceconsistentwithourpredictionsabouttheevolutionin?rms’interestratesandtheimpactofauditorchoiceduringtheir?rst9publicyears.Speci?cally,Diamond’s(1989)theorythat?rms’debt?nancingcostsaredecreasingin?rmageisstronglysupported.Similarly,consistentwithH1,thepositiveandstronglystatisticallysigni?cantinteractionbetween?rmageandtheBigSixauditorindicatorimpliesthatthein?uenceofauditorchoiceon?rms’interestratesgraduallysubsidesovertime.15

Infact,calibratingthistime-seriesevidencesuggeststhatborrowerandauditorreputationsareeconomicallyimportantdeterminantsofdebtpricing.Thecoef?cientestimateofà0.0024on?rmageimpliesthat,onaverage,?rms’interestratesdecreaseby216basispointsduringtheir?rst9publicyears.Thecoef?cientestimateof0.0016ontheage-BigSixauditorinteractionimpliesthat,onaverage,thebene?tofexternalmonitoringbyahigh-qualityauditordropsby144basispointsduringtheseyears.Tohelpputtheseestimatesinperspective,Table3reportsthat?rms’mean(median)interestrateoverthese9yearsis9.3percent(9.6percent).

Collectively,thesewithin-?rmresultsprovideevidenceonhow?rmandauditorreputationsaffectthedebtpricingprocessovertime.Young?rmscanlowertheirinterestratesbyacquiringatrackrecordforrepayingtheirdebts.However,intheirveryearlypublicyearswhen?rmsonlyhaveshortcredithistoriesforlenderstoobserve,these?rmsmostbene?tfromrelyingonthesuperiordebtmonitoringofaBigSixauditortomoderateinformationproblems.Insummary,theevidencereportedincolumns(1)and(2)ofTable4suggeststhat?rmageandauditorchoiceseparatelyandjointlyaffectyoungpublic?rms’borrowingcosts.

5.2.PrimarytestsofH2aandH2b:Firms’privateexperience

OurresearchdesignfortestingH1intheprevioussectionmeasuresa?rm’sageasthenumberofyearsthathaveelapsedsinceitsIPO,although?rmsordinarilyhaveaprivateoperatinghistorywhentheygopublic.Itwouldbeinterestingtoexaminethedeterminantsofdebtpricingfortheprivateoperatingyearsof?rmsinoursample.However,sincewedonothaveaccesstotheir?nancialstatementsfortheseyears,weresorttocoarselytestingthepredictionsinH2AandH2Bthatthein?uenceofauditorreputationoninterestrateswillbestronger,andwilldiminishmorequicklyovertime,in?rmswithshortprivatehistories.

Thecoef?cientsonmanyexplanatoryvariables,includingtheBigSixauditorindicator,tendtoapproachzeroorareestimatedimpreciselyinthe?xedeffectsregression.However,thetheoriesmotivatingthesepotentialdeterminantsofdebtpricingintendtoexplaindifferencesacross?rms.Enteringthe?rm?xedeffectsthatremovecross-sectionalvariationinthedatameansthatthesehypothesesmaynolongerbetestable.OnlytheOLSresultsthatarereportedincolumn(1)ofTable4arevalidforthesecross-sectionalpredictions.15

J.A.Pittman,S.Fortin/JournalofAccountingandEconomics37(2004)113–136129Thisinvolvesre-examiningtheevidencereportedinTable4afterbisectingthesampleby?rms’medianprivateage,de?nedasthenumberofyearsbetweentheirincorporationandtheirIPOs.16Priorresearchsorts?rmsbysuchwithin-samplecharacteristics;e.g.,RajanandZingales(1998)andPittmanandKlassen(2001).

Table5presentsevidencefortheolderandyounger?rmsub-samplesthatsuggeststhatthedeterminantsofdebtpricingin?rms’earlypublicyearsdependontheirprivateages.First,theimpactofpublicexperienceonthetime-seriesvariationintheirrespectiveinterestratessupportsDiamond’s(1989)theoryofreputationformation.Inthe?xedeffectsresultsincolumns(2)and(4),thecoef?cientestimateon?rmageisonlynegativeandstatisticallysigni?cantfortheyounger?rms.Thisimpliesthatlenderscontinuetolearnaboutthissetofborrowersbyobservingtheircredithistoriesduringtheirearlypublicyears.Incontrast,itappearsthattheolder?rmshavelargelycementedtheirreputationsduringtheirprivateyears.Theseresultscontributetoextensivepriorresearchthatsuggeststhatinformationasymmetryisworsefor?rmsgoingpublicearlierintheirlives;e.g.,ChemmanurandFulghieri,1999;LjungqvistandWilhelm,2003.

Second,theevidenceinTable5isconsistentwithH2A.ThepooledOLSresultsincolumns(1)and(3)indicatethatchoosingaBigSixauditorlowerstheinterestratesoftheyounger?rmsmorethantheolder?rms.Inbothregressions,thecoef?cientestimateonauditorchoiceisnegativeandsigni?cantlydifferentfromzero.17However,anF-testrevealsthatthiscoef?cientissigni?cantlygreaterforthesub-sampleofyounger?rms.Theyounger?rms,whicharelesserknowntolenders,canreducetheirinterestratesmoreintheirearlypublicyearsbyrelyingontheirauditors’reputationsaswellastheirown.

Finally,consistentwithH2B,thedecliningimportanceovertimeofauditorchoiceisstrongerinthesub-sampleofyounger?rms.18Infact,accordingtotheresultsWecollecteddataon?rms’incorporationdatesfromWard’sBusinessDirectoryofU.S.PrivateandPublicCompanies,Moody’sInternationalManual,Moody’sIndustrialManual,Moody’sOTCIndustrialManual,DirectoryofCorporateAf?liations,andawebsitemaintainedbyProfessorJayRitteroftheUniversityofFlorida.Themedianprivateageis10years,measuredfromyearofincorporationtoyearofinitialpublicoffering.Thedistributionishighlyskewed;e.g.,some?rmswerefoundedmorethan100yearsbeforegoingpublic,suchthatmeanprivateageisanunreliableestimateofcentraltendencyinthissample.Wediscardobservationsfor39?rmsforwhichwecouldnotdetermineincorporationdates.17InpooledOLStestswithoutthe?rmage-BigSixauditorinteraction,theauditorchoicevariableisonlysigni?cantforthesub-sampleofyounger?rms.

18AlthoughseparateHausman(1978)testscon?rmthatthe?xedeffectsmodelistheproperdesignchoiceforexaminingH1andH2B,externalvalidityisunderminedbecausetheresultsonlyapplytothe?rmsinthesesamples,whichonlycontainabout11percentofthepopulationofinitialequityissuesoccurringbetween1977and1988(seeTable1).Therearealmostcertainlymechanicalexplanationsforsomesampleattritionsuchas?rmsnotbeingimmediatelyfollowedbyCompustat;i.e.,oursampleselectionprocessinadvertentlyexcludes?rmsthatbelonginthesampleofthosesurvivingfor9publicyears.Sinceweareeagertoprovideevidencetojustifyinferencesabouttheentirepopulation,were-estimatetheequationswitharandomeffectsmodeland?ndevidencestronglysupportingthehypotheses.Further,GrilichesandHausman(1986)explainthatobservingconsistentestimatesacrossalternativepaneldataestimationtechniquessuggeststheabsenceofseriouserrorsinvariablesproblems.16

130

J.A.Pittman,S.Fortin/JournalofAccountingandEconomics37(2004)113–136

Table5

DebtpricingOLSand?xedeffectsregressionresults—balancedpanelpartitionedbyprivateageatIPOdate

OLSregression

INTERESTRATEit=a+b1AGEit+b2BIGSIXit+b3AGE?BIGSIXit+b4PRIMEt+b5DEFAULTt+b6LEVERAGEit+b7CASH-FLOWit+b8SIZEit+b9ASSETSit+b10NEG.EQUITYit+fINDUSTRYi+dYEARi+eit

Fixedeffectsregression

INTERESTRATEit=ai+b1AGEit+b2BIGSIXit+b3AGE?BIGSIXit+b4PRIMEt+b5DEFAULTt+b6LEVERAGEit+b7CASH-FLOWit+b8SIZEit+b9ASSETSit+b10NEG.EQUITYit+eitVariable

Prediction

Older?rmsPooledOLS

InterceptAge

BigSixauditor

Age?BigSixauditorPrimerate

DefaultpremiumLeverageCash?owSize

Assetstructure

NegativebookequityAdjustedR2F-statistic

Numberofobservations

?àà++++àà++

1

0.0819???à0.0024??à0.0081?0.00110.3304???à0.25970.0203???à0.0139??0.0024???0.0070???0.0079?0.13710.30??1413

Fixedeffects2

0.1042???à0.00110.0086?0.00020.3420???à0.28740.00460.0018à0.0036?0.00560.00600.488115.48??1413

Younger?rmsPooledOLS3

0.0644???à0.0047???à0.0202???0.0033???0.3235???0.40210.0423???à0.00460.0036???0.0030???0.0107??0.18415.81??1575

Fixedeffects4

0.0921???à0.0035???à0.00190.0026???0.3342???0.28490.0105??à0.0056à0.00160.00040.0155???0.57789.80??1575

Notes:Thistablepresentsregressionresultsforthedebtpricingmodelsforolderandyounger?rmsusingordinaryleastsquaresandtwo-way?xedeffectsestimationforthe?rms’?rstthroughninthyearsofpublicoperation.Thebalancedsamplediscardstheentiretime-seriesof?rmsifanymissingobservationsareencounteredinthe9years.Older?rmsarede?nedasthoseforwhichthedurationoftheirprivatehistory,whichismeasuredasthenumberofyearsthathaveelapsedbetweentheirincorporationandtheirinitialpublicoffering,isabovethemedianforthesample.Younger?rmsarede?nedasthosewithprivatehistoriesthatareshorterthanthemedianforthesample.Thedependentvariableandtheexplanatoryvariablesarespeci?edinTable3.TheOLSregressionsincludeinitialpublicofferingyearandone-digitSICcodeindustrydummyvariables;coef?cientestimatesforthesearenotreported.RegressionequationF-testssigni?cantatlessthan0.001areidenti?edbya??superscript.Inthistable,thesubscriptsiandtidentify?rmsandtime,respectively.Thesuperscriptasterisksindicateexplanatoryvariablecoef?cientsigni?canceatp-valueslessthan0.10(?),0.05(??),and0.01(???)inone-tailedtestswheredirectionalpredictionsaremadeandtwo-tailedtestsotherwise.

reportedincolumns(2)and(4)ofTable5,thecoef?cientontheage-BigSixauditorinteractionisonlypositiveandstatisticallysigni?cantforthissetof?rms;thecorrespondingcoef?cientforthesub-sampleofolder?rmsisnotstatisticallydifferentthanzero.AnF-testindicatesthatthecoef?cientontheinteractionissigni?cantlygreaterforthesub-sampleofyounger?rms.Together,theevidencein

J.A.Pittman,S.Fortin/JournalofAccountingandEconomics37(2004)113–136131thissectionsuggeststhattheseparateandcombinedimpactofborrowerandauditorreputationsoninterestratesisstrongerin?rmswithshortprivatehistorieswhentheygopublic.

5.3.Sensitivityanalysis

ThissectionexaminesthesensitivityofthereportedempiricalresultsthatareconsistentwithH1,H2A,andH2Bbyexploringwhetherthisevidencepersistsforaseriesofvariableandsamplere-speci?cationsandalternateestimationtechniques.

5.3.1.Firmsize

Extantresearchsuggeststhatlarger?rmstendtochooseBigSixauditors(Francisetal.,1999)andpaylowerinterestrates(PetersenandRajan,1994).Sincethisimpliesthatourmainevidencecouldbedrivenby?rmsize,wefollowBlackwelletal.’s(1998)methodologyforhandlingthispotentialconfounding.19

Weinitiallyre-estimatetheequationsonasize-truncatedsub-samplethatincludesall?rmswithtotalassetsrangingfrom$0.6million(theminimumtotalassetsof?rmswithBigSixauditors)and$291.9million(themaximumtotalassetsof?rmswithnon-BigSixauditors).Thissampleisintendedtoprovideevidenceonwhetherextremelysmallorlarge?rmsareresponsibleforourmainresults.20However,similartoBlackwelletal.(1998),descriptivestatisticsindicatethat,withinthissub-sample,theclientsofBigSixauditorsarestillsigni?cantlylargerthantheclientsofnon-BigSixauditors.

Consequently,were-estimatetheregressionsonasize-matchedsub-sample.Thisinvolvesmatchingeach?rmusinganon-BigSixauditorintheir?rstyeartothe?rmusingaBigSixauditorthatisclosestontotalassets.21Wealsorequirethantheabsolutevalueofthedifferenceintotalassetsbetweeneachpairislessthan10percenttoavoidpoormatches.Asmatchingonacontinuousvariablemeansthatwehavetoacceptimperfectmatches,wecontinuetoincludetheproxyfor?rmsizeintheseregressions.At-testrevealsthatthe56?rmsusingBigSixauditorsandthe56?rmsusingnon-BigSixauditorsinthissub-samplearestatisticallyindistinguishableonsize.Similarly,aWilcoxonsigned-ranktestthatmakeslessstrongdistributionalassumptionsalsoindicatesnodiscernabledifferencein?rmsize.

WethankanDanSimunicforinsightfullyidentifyingthisissueandsuggestingthetestsinthissection.Otherstudiesonauditqualityandthecostofcapitalmatch?rmsonsize;e.g.,Felthametal.(1991).20Wealsore-runthemaintestsafterbisectingeachbalancedandunbalancedpanelfromTables4and5accordingtomedian?rmsizeand?ndthatevidenceconsistentwithourpredictionspersistsforbothhalvesofthesedistributions,implyingthattheBigSixauditorvariabledoesnotjustproxyfor?rmsize.21Additionaltestsindicatethatmatchingusingthelast(ninth)publicyearratherthanthe?rstpublicyeardoesnotqualitativelymattertoanyoftheresultsreportedinthissection.Sinceidentifyingonlysinglematchesdecimatesthesample;i.e.,wekeeponly112oftheoriginal371?rms,wecompileanothersamplethatcontainsupto?vesuitable(totalassetswithin10percent)?rmswithBigSixauditorsmatchedtothe?rmswithnon-BigSixauditors(Klassenetal.,2003).Thisincreasesthesampleto231?rms.EstimationusingweightedleastsquaresforthesemultiplematchesalsoprovidessimilarevidenceonH1,H2A,andH2B.19

132J.A.Pittman,S.Fortin/JournalofAccountingandEconomics37(2004)113–136

Table4reportsregressionresultsincolumns(3)–(6)forthesesize-truncatedandsize-matchedsub-samples.Theseresultsaresimilartothosereportedincolumns(1)and(2)forthefullsample,implyingthattheevidencesupportingH1isnotanartifactof?rmsize.Inparticular,themagnitudeandstatisticalsigni?canceoftheOLSand?xedeffectscoef?cientestimatesfor?rmage,theBigSixauditorindicator,andtheinteractionofageandtheBigSixauditorindicatorareconsistentacrossallthreesamples.

WeprovideevidenceonH2AandH2BinTable5aftersplittingthefullsampleintoyoungerandolder?rmsaccordingtotheirprivateages.Results(nottabulated)fromre-estimatingtheseregressionsonsize-truncatedandsize-matchedsub-samplesoftheseyoungerandolder?rmsremainconsistentwiththesepredictions.

5.3.2.Sampleattrition

Wecompilethebalancedpaneldatabyremoving?rmsthatbecamebankrupt,wereliquidated,wereacquired,ormergedwithanother?rmduringtheir?rst9publicyears.Imposingthesescreenspreventsbiasfromthecompositionofthesamplechangingovertime.However,thereareatleasttworeasonstoexaminecertainunbalancedpanels—permitting?rmstoleaveand,inonespeci?cation,re-enterthesampleduringthe9years—whentestingthepredictions.

First,considerableattritionforreasonsotherthannon-survivaloccurredwhenthepanelwasconstructedbydiscardingobservationsthatdonotappearinallyears(seeTable1).Forexample,a?rm’sentire9-yeartime-seriesislostifevenasingleinterestrateobservationwasconsideredextremebecauseitwasoutsidethe5thor95thpercentiles.Boththefullandsplit-sampletestswereseparatelyre-estimatedforanunbalancedpanelinwhich?rmswerepermittedtoenterandleavethesampleduringthe9-yearperiod.Inthesesamples,welosefewerthan10percentofthe?rm-yearobservationsbyreplacing,ifpossible,extremevalueswithinterestratescalculatedfromquarterlydata.TheseresultsprovideevidenceconsistentwithH1(seecolumns

(7)and(8)ofTable4)aswellasH2AandH2B(nottabulated),implyingthatthescreeningusedtocompilethebalancedpaneldatadoesnotseriouslyundermineexternalvalidity.22

Second,severalcorporate?nancetheoriesarguethatchangestosamplecompositionaffectsdebtpricing.Forexample,onemechanisminDiamond(1989)thatgeneratesapatternofdiminishinginterestratesovertimeishavingyoung?rmsthatdefaultontheirdebtsleavethecreditmarket.Thebalancedpanelpartiallysuppressestheimpactofthismechanismontheevolutioninthedeterminantsof?rms’interestrates.Priorresearchimpliesthatlendersre?netheirinformationabout?rmsthatsurvivebyobservingthosethatdefault.Accordingly,were-runtheThisevidenceforH1,H2A,andH2Bremainsaftermoreformallytreatingthispotentialsampleselectionbias.WeapplyHeckman’s(1979)two-stepestimationprocedurethatgeneratesconsistentestimatesinthepresenceofattritionbyeliminatingthebiasthatcouldarisefromomittedvariableswhenthesampleisnotrandom.The?xedeffectsregressionthatmodelswithin-?rmvariationindebtpricingremovesanytime-invariantsampleselectionbiassincethecorrectiontermforselectivityisabsorbedinthe?rm-speci?ceffect.Weexaminepotentialattritionbiasbecause?rmsmaybeexcludedfromthesampleforreasonsbesidesnon-survival.22

J.A.Pittman,S.Fortin/JournalofAccountingandEconomics37(2004)113–136133hypothesestestsonunbalancedpanelscomprisedof?rmsthatwereallowedtoleave,butnotlaterreturnto,thesampleundertheassumptionthatborrowersthatdefaultareforeverexcludedfromthecreditmarket.Theseresultsprovideevidenceatthe1-percentlevelconsistentwithH1(seecolumns(9)and(10)ofTable4)andH2AandH2B(nottabulated).

5.3.3.Year-to-yeartests

Weemploythefollowingmethodologytoprovideyear-to-yearevidenceonthepredictioninH1thathiringaBigSixauditoraffectsdebtpricinglessovertime(Lang,1991;PittmanandKlassen,2001).Afterremovingtheage-BigSixauditorinteraction,were-estimateEq.(1)inseparatecross-sectionalregressionsforeachofthe?rms’?rst9publicyearswitharandomcoef?cientsmodelthatattributesparameterheterogeneitytostochasticvariation.Thisestimationtechnique,whichhasbeenmodi?edtocorrectforheteroscedasticityandavoidsserialcorrelationcomplications,permitscross-sectionalvariationintheauditorchoicecoef?cients.

Plottingtheseninecoef?cientsrevealsanalmostmonotonicallydecliningtrendinthein?uenceofauditorchoiceon?rms’interestratesovertime.ConsistentwithH1,thePearsoncorrelationoftheninecoef?cientswith?rmageis0.79,whichissigni?cantatthe1-percentlevelinaone-tailedtestassumingindependence.However,sincethiscorrelationignorestheprecisionofthecoef?cientestimates,werunaweightedlinearregressionwiththeleastweightassignedtotheobservationsthataremeasuredwiththemosterror.Theweightedleastsquaresregressionoftheauditorchoiceestimatesfromtheninecross-sectionalrandomcoef?cientsmodelson?rmagealsoproducesevidenceatthe1-percentlevelthatapositivepatternexists.

Similarly,applyingthisproceduretotheyoungerandolder?rmsub-samplesprovidestime-seriesevidenceatthe1-percentlevelsupportingthepredictioninH2Bthatthesubsidingin?uenceofauditorreputationoninterestrateswithagewillbestrongerin?rmswithshortprivatehistories.

5.3.4.Interestratespeci?cation

ThedependentvariableforthetestsreportedinTables4and5isinterestexpensefortheyeardividedbyaverageshort-andlong-termdebtduringtheyear.Replacingthisdenominatorwithlong-termdebtorthesumofshort-,long-term,andconvertibledebtproducesqualitativelysimilarevidenceonthehypotheses.23

5.3.5.Publicagespeci?cation

Sincetheimpactondebtpricingofanadditionalyearofpublicexistencemaydeclineovertime,wespecifyvariousnon-lineartransformationsof?rmage(theStein(1992)andothersarguethatconvertibledebtprovidescost-effective?nancingwheninformationalasymmetriesarerelativelyseveresuchasfortheyoung,small?rmsinoursample.Specifyingonlylong-termdebtinthedenominatoravoidsconfoundingfromseasonal?uctuationsin?rms’short-termdebt.23

134J.A.Pittman,S.Fortin/JournalofAccountingandEconomics37(2004)113–136

naturallogarithmofoneplusage,second-orderlogs,squareroots,andreciprocals)toprovidethedatamore?exibility.Evidencesupportingourpredictionsisfoundforeachspeci?cation.Further,R2wasslightlyhigherinboththefull(Table4)andsplit(Table5)sampleswiththelinearspeci?cationof?rmage,validatingourdesignchoice.

5.3.6.Othertests

TheevidencethattheimpactofauditorchoiceondebtpricingisconsistentwithH1,H2A,andH2Bremainsafterremoving?rmsthatchangedtheir?scalyear-endstosynchronizethetestsat1-yearintervals.Thereportedresultsarealsonotcalendaryear,industry,orBigSixauditorspeci?candrepresentapervasiveeconomictrend,ratherthanin?uentialobservationsdominatingthedata.Inaddition,serialcorrelationthroughrepeatedmeasurementsofa?rmisnotresponsiblefortheevidence.24

6.Conclusions

Weexaminethelinkbetweenauditorchoiceanddebtpricingfornewlypublic?rms.These?rmsparticularlydependonobtainingexternal?nancingdespiteexperiencingseriousinformationproblemsintheirearlypublicyears.Ourcross-sectionalevidencesuggeststhatchoosingaBigSixauditor,whichcanreducedebt-monitoringcostsbyenhancingthecredibilityof?nancialstatements,enablesyoung?rmstolowertheirinterestrates.

Extanttheoryandevidenceimpliesthatinformationasymmetrybetweenborrowersandlendersisdecreasingin?rmagesince,forexample,young?rmshavenothadtheopportunitytoaccumulatealengthycreditrecord.Consistentwithourtime-seriespredictions,we?ndthattheimpactofauditorchoiceon?rms’interestratesdiminisheswithage.Ourevidencesuggeststhattheeconomicvalueofauditorreputationtothecostofcreditdeclinesovertimeasborrowersgraduallyshifttowardrelyingontheirownreputationstomoderateinformationasymmetry.Informationproblemsareevenworsefor?rmsthathaveshortprivatehistorieswhentheygopublicaccordingtopriorresearch.Wereportresultsfromadditionalteststhatimplythatprivateageaffectstheimportanceofauditorchoiceto?rms’borrowingcosts.Speci?cally,againconsistentwithourpredictions,we?ndthatalthoughyounger?rmsattheirIPOdatesespeciallybene?tfromretainingaBigSixauditor,thisstrongerin?uencesubsidesmoreswiftlyovertheir?rst9publicyearsastheybecomebetterknown.

Thisresearchcouldbeextendedbyexaminingthepredictionsinothersettingssuchas?rmsemergingfrombankruptcyprotection.Thesenewlyre-organized?rms,whicharejustbeginningtheprocessofrehabilitatingtheirreputationswithlenders,maybeeagertoexploitthesuperiordebtmonitoringofaBigSixauditor.ThismayAlthoughastatisticallysigni?cantAR(1)parameterindicatesthatserialcorrelationisimportant,thet-statisticsforthetestvariablesaresimilarwhenrobustestimatesofthestandarderrorsareused.24

J.A.Pittman,S.Fortin/JournalofAccountingandEconomics37(2004)113–136135beacleanertestinggroundthanthenewlypublic?rmssincebankrupt?rmscanselectabrandnewcapitalstructurebeforeleavingChapter11(AldersonandBetker,1995).Incomparison,evidenceinthisstudysuggeststhatthedurationof?rms’pre-IPOhistoriesaffectsinterestratesintheirpost-IPOyears.Finally,StiglitzandWeiss(1981),amongothers,arguethatinformationproblemsareresponsiblefortheseverecreditrationingthat?rms’encounterintheirearlyyears.Empiricalresearchonwhetherretainingahigh-qualityauditorreducescreditrationinginyoung?rmswouldcomplementourevidenceondebtpricing.

References

Alderson,M.,Betker,B.,1995.Liquidationcostsandcapitalstructure.JournalofFinancialEconomics39,45–69.

Balvers,R.J.,McDonald,B.,Miller,R.E.,1988.Underpricingofnewissuesandthechoiceofauditorasasignalofinvestmentbankerreputation.TheAccountingReview63,605–622.

Beatty,R.P.,1989.Auditorreputationandthepricingofinitialpublicofferings.TheAccountingReview64,693–709.

Becker,C.,DeFond,M.L.,Jiambalvo,J.,Subramanyam,K.R.,1998.Theeffectofauditqualityonearningsmanagement.ContemporaryAccountingResearch15,1–24.

Blackwell,D.W.,Noland,T.R.,Winters,D.B.,1998.Thevalueofauditorassurance:evidencefromloanpricing.JournalofAccountingResearch36,57–70.

Carey,M.,Prowse,S.,Rea,J.,Udell,G.,1993.Theeconomicsofprivateplacements:anewlook.FinancialMarketsInstitutionsandInstruments2,1–66.

Chemmanur,T.J.,Fulghieri,P.,1999.Atheoryofthegoingconcerndecision.ReviewofFinancialStudies12,249–279.

Copley,P.A.,DouthettJr,E.B.,2002.Theassociationbetweenauditorchoice,ownershipretained,andearningsdisclosureby?rmsmakinginitialpublicofferings.ContemporaryAccountingResearch19,49–75.

Craswell,A.,Francis,J.,Taylor,S.,1995.Auditorbrandnamereputationsandindustryspecializations.JournalofAccountingandEconomics20,297–322.

Datar,S.M.,Feltham,G.A.,Hughes,J.S.,1991.Theroleofauditsandauditqualityinvaluingnewissues.JournalofAccountingandEconomics14,3–49.

Datta,S.,Iskandardatta,M.,Patel,A.,1999.Bankmonitoringandthepricingofcorporatepublicdebt.JournalofFinancialEconomics51,435–449.

DeAngelo,L.E.,1981.Auditorsizeandauditquality.JournalofAccountingandEconomics3,183–199.Dechow,P.,1994.Accountingearningsandcash?owsasmeasuresof?rmperformance:theroleofaccountingaccruals.JournalofAccountingandEconomics18,3–42.

DeFond,M.,Jiambalvo,J.,1994.Debtcovenantviolationandmanipulationofaccruals.JournalofAccountingandEconomics17,145–176.

Diamond,D.W.,1989.Reputationacquisitionindebtmarkets.JournalofPoliticalEconomy97,828–862.Feltham,G.A.,Hughes,J.S.,Simunic,D.A.,1991.Empiricalassessmentoftheimpactofauditorqualityonthevaluationofnewissues.JournalofAccountingandEconomics14,375–399.

Francis,J.R.,Maydew,E.L.,Sparks,H.C.,1999.TheroleofBig6auditorsinthecrediblereportingofaccruals.Auditing:AJournalofPracticeandTheory18,17–34.

Graham,J.R.,Lemmon,M.L.,Schallheim,J.S.,1998.Debt,leases,taxes,andtheendogeneityofcorporatetaxstatus.JournalofFinance53,131–161.

Griliches,Z.,Hausman,J.A.,1986.Errorsinvariablesinpaneldata.JournalofEconometrics31,93–118.Hansen,L.,1982.Largesamplepropertiesofgeneralizedmethodofmomentsestimators.Econometrica50,1029–1054.

Hausman,J.A.,1978.Speci?cationtestsineconometrics.Econometrica46,1251–1271.

136J.A.Pittman,S.Fortin/JournalofAccountingandEconomics37(2004)113–136Heckman,J.J.,1979.Sampleselectionbiasasspeci?cationerror.Econometrica47,153–162.

Jensen,M.C.,Meckling,M.H.,1976.Theoryofthe?rm:managerialbehavior,agencycosts,andownershipstructure.TheAmericanEconomicReview3,305–360.

John,K.,Lynch,A.W.,Puri,M.,2003.Creditratings,collateral,andloancharacteristics:implicationsforyield.JournalofBusiness36,371–409.

Klassen,K.J.,Pittman,J.A.,Reed,M.,2003.Across-nationalcomparisonofR&Dexpendituredecisions:?nancialconstraintsandtaxincentives.ContemporyAccountingResearch,forthcoming.

Lang,M.,1991.Time-varyingstockpriceresponsetoearningsinducedbyuncertaintyaboutthetime-seriesprocessofearnings.JournalofAccountingResearch29,229–257.

Ljungqvist,A.,WilhelmJr,W.J.,2003.IPOpricinginthedot-combubble.JournalofFinance58,723–752.

Loughran,T.,Ritter,J.,1995.Thenewissuespuzzle.JournalofFinance50,23–51.

McKinley,S.,Pany,K.,Reckers,P.M.J.,1985.Anexaminationofthein?uenceofCPA?rmtype,size,andMASprovisiononloanof?cerdecisionsandperceptions.JournalofAccountingResearch23,887–896.

Menon,K.,Williams,D.D.,1994.Theinsurancehypothesisandmarketprices.TheAccountingReview69,327–342.

MorsmanJr,E.,1986.Commercialloanstructuring.JournalofCommercialBankLending58,2–20.Petersen,M.A.,Rajan,R.G.,1994.Thebene?tsoflendingrelationships:evidencefromsmallbusinessdata.JournalofFinance49,3–37.

Pittman,J.A.,Klassen,K.J.,2001.Thein?uenceof?rmmaturationon?rms’rateofadjustmenttotheiroptimalcapitalstructures.TheJournaloftheAmericanTaxAssociation23,70–94.

Rajan,R.,Zingales,L.,1998.Financialdependenceandgrowth.TheAmericanEconomicReview88,559–586.

Ritter,J.R.,1991.Thelong-runperformanceofinitialpublicofferings.JournalofFinance46,3–27.Stein,J.C.,1992.Convertiblebondsasbackdoorequity?nancing.JournalofFinancialEconomics32,3–22.

Stiglitz,J.E.,Weiss,A.,1981.Creditrationinginmarketswithimperfectinformation.TheAmericanEconomicReview71,393–410.

Stoughton,N.M.,Wong,K.P.,Zechner,J.,2001.IPOsandproductquality.JournalofBusiness74,375–408.

Sweeney,A.,1994.Debtcovenantviolationsandmanagers’responses.JournalofAccountingandEconomics17,281–308.

Teoh,S.H.,Wong,T.J.,1993.Perceivedauditorqualityandtheearningsresponsecoef?cient.TheAccountingReview68,346–366.

Teoh,S.H.,Wong,T.J.,Rao,G.,1998.Areaccrualsduringinitialpublicofferingsopportunistic?ReviewofAccountingStudies3,175–208.

Wallace,W.A.,1978.Theimpactofselected?nancialreportingpracticesandthenatureoftheauditopinionuponmunicipalinterestcostandbondrating.Ph.D.dissertation,UniversityofFlorida.Watts,R.L.,Zimmerman,J.L.,1986.PositiveAccountingTheory.PrenticeHall,EnglewoodCliffs,NJ.White,H.,1980.Aheteroscedasticity-consistentcovariancematrixandadirecttestheteroscedasticity.Econometrica48,817–838.