The determinants of capital structure choice

AmericanFinanceAssociationTheDeterminantsofCapitalStructureChoiceAuthor(s):SheridanTitmanandRobertoWesselsSource:TheJournalofFinance,Vol.43,No.1(Mar.,1988),pp.1-19Publishedby:BlackwellPublishingfortheAmericanFinanceAssociationStableURL::17/01/200907:06YouruseoftheJSTORarchiveindicatesyouracceptanceofJSTOR'sTermsandConditionsofUse,availableat://=black.EachcopyofanypartofaJSTORtransmissionmustcontainthesamecopyrightnoticethatappearsonthescreenorprintedpageofsuchtransmission.JSTORisanot-for-profitorganizationfoundedin1995tobuildtrusteddigitalarchivesforscholarship.Weworkwiththescholarlycommunitytopreservetheirworkandthematerialstheyrelyupon,andtobuildacommonresearchplatformthatpromotesthediscoveryanduseoftheseresources.FormoreinformationaboutJSTOR,[e-mail]ericanFinanceAssociationarecollaboratingwithJSTORtodigitize,preserveandextendaccesstoTheJournalofFinance.*VOL.XLIII,NO.1*MARCH1988TheDeterminantsofCapitalStructureChoiceSHERIDANTITMANandROBERTOWESSELS*ABSTRACTThispaperanalyzestheexplanatorypowerofsomeoftherecenttheoriesofoptimalcapitalstructure.Thestudyextendsempiricalworkoncapitalstructuretheoryinthreeways.First,itexaminesamuchbroadersetofcapitalstructuretheories,manyofwhichhavenotpreviouslybeenanalyzedempirically.Second,sincethetheorieshavedifferentempiricalimplicationsinregardtodifferenttypesofdebtinstruments,theauthorsanalyzemeasuresofshort-term,long-term,andconvertibledebtratherthananaggre-gatemeasureoftotaldebt.Third,thestudyusesafactor-analytictechniquethatmitigatesthemeasurementproblemsencounteredwhenworkingwithproxyvariables.INRECENTYEARS,Anumberoftheorieshavebeenproposedtoexplainthevariationindebtratiosacrossfirms.Thetheoriessuggestthatfirmsselectcapitalstructuresdependingonattributesthatdeterminethevariouscostsandbenefitsassociatedwithdebtandequityfinancing.Empiricalworkinthisareahaslaggedbehindthetheoreticalresearch,perhapsbecausetherelevantfirmattributesareexpressedintermsoffairlyabstractconceptsthatarenotdirectlyobservable.Thebasicapproachtakeninpreviousempiricalworkhasbeentoestimateregressionequationswithproxiesfortheunobservabletheoreticalattributes.Thisapproachhasanumberofproblems.First,theremaybenouniquerepre-sentationoftheattributeswewishtomeasure.Thereareoftenmanypossibleproxiesforaparticularattribute,andresearchers,lackingtheoreticalguidelines,maybetemptedtoselectthosevariablesthatworkbestintermsofstatisticalgoodness-of-fitcriteria,therebybiasingtheirinterpretationofthesignificancelevelsoftheirtests.Second,itisoftendifficulttofindmeasuresofparticularattributesthatareunrelatedtootherattributesthatareofinterest.Thus,selectedproxyvariablesmaybemeasuringtheeffectsofseveraldifferentattributes.Third,sincetheobservedvariablesareimperfectrepresentationsoftheattributestheyaresupposedtomeasure,theiruseinregressionanalysisintroducesanerrors-in-variableproblem.Finally,measurementerrorsintheproxyvariablesmaybecorrelatedwithmeasurementerrorsinthedependentvariables,creatingspuriouscorrelationsevenwhentheunobservedattributebeingmeasuredisunrelatedtothedependentvariable.*BothauthorsarefromtheUniversityofCalifornia,LosAngeles,andWesselsisalsofromErasmusUniversity,Rotterdam.WegratefullyacknowledgetheresearchassistanceprovidedbyJimBrandon,WonLee,andErikSirriandhelpfulcommentsfromourUCLAcolleagues,especiallyJulianFranks,DavidMayers,RonMasulis,andWalterTorous.WealsoreceivedhelpfulcommentsfromseminarparticipantsatUCLAandtheUniversityofRochester.TitmanreceivedfinancialsupportfromtheBatterymarchfellowshipprogramandfromtheUCLAFoundationforResearchinFinancialMarketsandInstitutions.WesselsreceivedfinancialsupportfromtheNetherlandsOrga-nizationfortheAdvancementofPureResearch(Z.W.0.).12TheJournalofFinanceThisstudyextendsempiricalworkoncapitalstructuretheoryinthreeways.'First,itextendstherangeoftheoreticaldeterminantsofcapitalstructurebyexaminingsomerecentlydevelopedtheoriesthathavenot,asyet,beenanalyzedempirically.Second,sincesomeofthesetheorieshavedifferentempiricalimpli-cationswithregardtodifferenttypesofdebtinstruments,weanalyzeseparatemeasuresofshort-term,long-term,andconvertibledebtratherthananaggregatemeasureoftotaldebt.Third,atechniqueisusedthatexplicitlyrecognizesandmitigatesthemeasurementproblemsdiscussedabove.Thistechnique,whichisanextensionofthefactor-analyticapproachtomeasuringunobservedorlatentvariables,isknownaslinearstructuralmodeling.2Verybriefly,thismethodassumesthat,althoughtherelevantattributesarenotdirectlyobservable,wecanobserveanumberofindica