determinants of capital structure choice-titman

8/12/2019 Determinants of Capital Structure Choice-Titman

1/20

American Finance Association

The Determinants of Capital Structure ChoiceAuthor(s): Sheridan Titman and Roberto WesselsSource: The Journal of Finance, Vol. 43, No. 1 (Mar., 1988), pp. 1-19Published by: Blackwell Publishingfor the American Finance AssociationStable URL: http://www.jstor.org/stable/2328319.

Accessed: 19/01/2011 02:07

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at.http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless

you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you

may use content in the JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at.http://www.jstor.org/action/showPublisher?publisherCode=black..

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed

page of such transmission.

JSTOR 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 forms

of scholarship. For more information about JSTOR, please contact [email protected].

Blackwell PublishingandAmerican Finance Associationare collaborating with JSTOR to digitize, preserve

and extend access to The Journal of Finance.

http://www.jstor.org
http://www.jstor.org/action/showPublisher?publisherCode=blackhttp://www.jstor.org/action/showPublisher?publisherCode=afinahttp://www.jstor.org/stable/2328319?origin=JSTOR-pdfhttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/action/showPublisher?publisherCode=blackhttp://www.jstor.org/action/showPublisher?publisherCode=blackhttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/stable/2328319?origin=JSTOR-pdfhttp://www.jstor.org/action/showPublisher?publisherCode=afinahttp://www.jstor.org/action/showPublisher?publisherCode=black


2/20

THE JOURNALOF FINANCE * VOL.XLIII, NO. 1 * MARCH1988

The Determinants of Capital Structure ChoiceSHERIDANTITMANand ROBERTOWESSELS*

ABSTRACTThis paper analyzes the explanatorypower of some of the recent theoriesof optimalcapital structure.The study extends empiricalworkon capital structure heory in threeways. First, it examines a much broader et of capital structure heories,manyof whichhave not previouslybeenanalyzedempirically.Second,since the theorieshave differentempirical implications in regardto different types of debt instruments,the authorsanalyzemeasuresof short-term, ong-term,and convertibledebt ratherthan an aggre-gate measure of total debt. Third, the study uses a factor-analytictechnique thatmitigates the measurementproblemsencounteredwhen workingwith proxy variables.

IN RECENTYEARS,A number of theories have been proposed to explain thevariation n debt ratios across firms. The theoriessuggestthat firms select capitalstructuresdependingon attributes that determinethe various costs and benefitsassociated with debtand equity financing.Empiricalwork in this area has laggedbehind the theoreticalresearch,perhapsbecause the relevantfirm attributesareexpressedin terms of fairlyabstractconceptsthat are not directlyobservable.The basic approach taken in previous empiricalwork has been to estimateregression equations with proxies for the unobservabletheoretical attributes.This approachhas a numberof problems.First, there may be no unique repre-sentation of the attributes we wish to measure. There are often many possibleproxiesfor a particularattribute,and researchers, acking theoretical guidelines,may be tempted to select those variablesthat work best in terms of statisticalgoodness-of-fit criteria, thereby biasing their interpretationof the significancelevels of their tests. Second, it is often difficult to find measuresof particularattributesthat are unrelatedto otherattributesthat are of interest.Thus, selectedproxy variables may be measuring the effects of several different attributes.Third,since the observedvariablesareimperfectrepresentationsof the attributesthey are supposed to measure, their use in regression analysis introduces anerrors-in-variableproblem. Finally, measurementerrorsin the proxy variablesmaybe correlatedwith measurementerrorsin the dependent variables, creatingspurious correlations even when the unobserved attribute being measured isunrelatedto the dependentvariable.

*Both authors are from the University of California,Los Angeles, and Wessels is also fromErasmusUniversity,Rotterdam.We gratefullyacknowledgehe researchassistanceprovidedby JimBrandon,Won Lee, and Erik Sirri and helpful comments from our UCLA colleagues, especiallyJulianFranks,DavidMayers,Ron Masulis,and WalterTorous.We also receivedhelpfulcommentsfrom seminarparticipantsat UCLA and the University of Rochester.Titman received financialsupport rom the Batterymarch ellowshipprogramand fromthe UCLAFoundation orResearch nFinancialMarketsand Institutions.Wesselsreceivedfinancialsupportfromthe NetherlandsOrga-nization forthe Advancementof Pure Research(Z.W. 0.).1


3/20

2 The Journal of FinanceThis study extends empiricalworkon capital structuretheory in three ways.'First, it extends the range of theoretical determinants of capital structurebyexaminingsome recently developedtheoriesthat have not, as yet, been analyzedempirically.Second, since some of these theories have differentempirical mpli-

cations with regardto differenttypes of debt instruments,we analyze separatemeasuresof short-term, ong-term,and convertibledebtratherthan an aggregatemeasure of total debt. Third, a technique is used that explicitly recognizes andmitigates the measurementproblems discussed above.This technique, which is an extension of the factor-analytic approach tomeasuringunobservedorlatent variables, s known as linear structuralmodeling.2Very briefly,this method assumes that, althoughthe relevantattributes are notdirectly observable,we can observe a number of indicator variables that arelinear functions of one or more attributesand a random errorterm. There is, inthis specification,a direct analogywith the return-generatingprocess assumedto hold in the Arbitrage Pricing Theory. While the identifying restrictionsimposedon ourmodel are different,the techniquefor estimatingit is verysimilarto the procedureused by Roll and Ross [29] to test the APT.Ourresults suggest that firms with unique or specialized productshave rela-tively low debt ratios. Uniqueness is categorized by the firms' expendituresonresearch and development, selling expenses, and the rate at which employeesvoluntarilyleave their jobs. We also find that smallerfirms tend to use signifi-cantlymore short-termdebtthan largerfirms.Ourmodelexplains virtuallynoneof the variation in convertible debt ratios across firms and finds no evidence tosupporttheoretical work that predicts that debt ratios are related to a firm'sexpected growth, non-debt tax shields, volatility, or the collateral value of itsassets. We do, however, find some support for the propositionthat profitablefirmshave relatively less debt relative to the market value of their equity.

I. Determinants of Capital StructureIn this section, we present a brief discussion of the attributes that differenttheories of capital structure suggest may affect the firm's debt-equity choice.These attributes are denoted asset structure, non-debt tax shields, growth,uniqueness, ndustryclassification, size, earningsvolatility,andprofitability.Theattributes, their relation to the optimal capital structurechoice, and their ob-servable indicators are discussed below.

'Recent cross-sectionalstudies includeToy et al. [40], Ferri and Jones [14],Flath and Knoeber[15],Marsh[24], Chaplinsky 11],Titman [38],Castanias[8],Bradley,Jarrell,andKim [6], Auerbach[2], and Long and Malitz [23]. The papersby Titman [38],Bradley,Jarrell,and Kim [6], Auerbach[2], and Long and Malitz [23] examinevariables that are similar to some of those examined here.The studies find a negative relationbetweenboth research and developmentand advertising andleverage but have mixed findings relating to the different measures of non-debt tax shields andleverageandvolatility and leverage.Also see Schwartzand Aronson [30], Scott [31], and Scott andMartin [32] for evidence of industryeffects in capitalstructurechoice.

2 References o linear structuralmodelingcan also be found in the literatureunder the headingsof analysis-of-covariancetructures,path analysis, causal models,and content-variablesmodels.Anontechnical ntroduction o the subjectprovidingmany references s BentlerandBonett [4].


4/20

CapitalStructure 3A. CollateralValue of Assets

Most capital structure theories arguethat the type of assets ownedby a firmin some way affects its capital structure choice. Scott [33] suggests that, byselling secured debt, firms increase the value of their equity by expropriatingwealth from their existing unsecured creditors.3Arguments put forth by Myersand Majluf [28] also suggest that firms may find it advantageous o sell secureddebt. Their model demonstratesthat there may be costs associated with issuingsecuritiesaboutwhich the firm'smanagershave better informationthan outsideshareholders.Issuing debt securedby propertywith known values avoids thesecosts. For this reason, firms with assets that can be used as collateral may beexpected to issue more debt to take advantageof this opportunity.Work by Galai and Masulis [16], Jensen and Meckling [20], and Myers [26]suggeststhat stockholdersof leveragedfirms have an incentive to invest subop-timally to expropriate wealth from the firm's bondholders.This incentive mayalso induce a positive relation between debt iatios and the capacityof firms tocollateralize heir debt.If the debtcan be collateralized, he borrower s restrictedto use the funds for a specified project.Since no such guaranteecan be used forprojects hat cannot be collateralized,creditorsmay requiremore favorable erms,which in turn may lead such firmsto use equity ratherthan debt financing.The tendencyof managersto consumemore than the optimallevel of perquis-ites may produce the opposite relation between collateralizablecapital and debtlevels. Grossman and Hart [18] suggest that higher debt levels diminish thistendency because of the increased threat of bankruptcy. Managers of highlylevered firms will also be less able to consume excessive perquisites sincebondholders(or bankers) are inclined to closely monitor such firms. The costsassociated with this agencyrelationmay be higherfor firmswith assets that areless collateralizable ince monitoringthe capital outlaysof such firmsis probablymore difficult.For this reason,firmswith less collateralizableassets may choosehigherdebt levels to limit their managers'consumptionof perquisites.The estimated model incorporates two indicators for the collateral valueattribute. They include the ratio of intangible assets to total assets (INT/TA)and the ratio of inventory plus gross plant and equipment to total assets(IGP/TA). The first indicator is negativelyrelatedto the collateralvalue attrib-ute, while the second is positively related to collateral value.B. Non-Debt TaxShields

DeAngelo and Masulis [12] present a model of optimal capital structurethatincorporates he impact of corporate axes, personal taxes, and non-debt-relatedcorporate tax shields. They argue that tax deductions for depreciation andinvestment tax credits are substitutes for the tax benefits of debt financing.Asa result,firms with largenon-debt tax shieldsrelative to their expectedcash flowinclude less debt in their capital structures.Indicatorsof non-debt tax shields include the ratios of investment tax creditsover total assets (ITC/TA), depreciationover total assets (DITA), and a direct

'See Smith and Warner[35] for a commenton Scott's model.


5/20

4 The Journalof Financeestimateof non-debt tax shields over total assets (NDT/TA). The lattermeasureis calculatedfrom observedfederal income tax payments (T), operatingincome(OI), interest payments (i), and the corporate ax rate duringour sampleperiod(48%),using the followingequation:

NDT = OI-i-- 0.48'which follows from the equality

T= 0.48(0I- i-NDT).These indicators measurethe current tax deductions associated with capitalequipment and, hence, only partially capture the non-debt tax shield variable

suggestedby DeAngeloand Masulis.First, this attribute excludes tax deductionsthat arenot associatedwith capitalequipment,suchas researchanddevelopmentand selling expenses. (These variables,used as indicatorsof another attribute,arediscussedlater.) Moreimportant,ournon-debttax shield attributerepresentstax deductions rather than tax deductionsnet of true economicdepreciationandexpenses, which is the economic attribute suggestedby theory. Unfortunately,this preferableattribute wouldbe very difficultto measure.C. Growth

As we mentionedpreviously,equity-controlled irmshave a tendencyto investsuboptimallyto expropriatewealth from the firm'sbondholders.The cost asso-ciated with this agency relationship is likely to be higher for firms in growingindustries, which have more flexibility in their choice of future investments.Expectedfuturegrowthshould thus be negativelyrelatedto long-termdebt levels.Myers, however,noted that this agency problemis mitigated if the firm issuesshort-term rather than long-term debt. This suggeststhat short-termdebt ratiosmight actually be positively related to growthrates if growingfirms substituteshort-termfinancing for long-termfinancing.Jensen and Meckling [20], Smithand Warner[36], and Green [17] arguedthat the agencycosts will be reduced ffirms issue convertibledebt. This suggests that convertibledebt ratios may bepositively related to growthopportunities.It should also be noted that growthopportunitiesare capital assets that addvalue to a firmbut cannot be collateralizedand do not generatecurrenttaxableincome.Forthis reason, the argumentsput forth in the previoussubsections alsosuggesta negativerelation between debt and growthopportunities.Indicatorsof growth include capital expendituresover total assets (CE/TA)andthe growthof total assets measuredby the percentagechangein total assets(GTA). Since firms generally engage in researchand developmentto generatefuture investments, researchand developmentover sales (RD/S) also serves asan indicator of the growthattribute.4

4We also consideredusingprice/earningsratiosas an indicatorof growth.However, his variableis determined n part by the firm's everageratioand, hence,is subject o biasdueto reversecausality.


6/20

CapitalStructure 5D. Uniqueness

Titman [39] presents a modelin whicha firm's liquidationdecision is causallylinked to its bankruptcystatus. As a result, the costs that firms can potentiallyimpose on their customers, suppliers,and workers by liquidatingare relevant totheir capital structuredecisions.Customers,workers,and suppliersof firms thatproduce unique or specialized products probably suffer relatively high costs inthe event that they liquidate. Their workers and suppliers probably have job-specific skills and capital, and their customers may find it difficult to findalternative servicing for their relatively unique products. For these reasons,uniquenessis expectedto be negativelyrelated to debt ratios.Indiators of uniqueness include expenditures on research and developmentover sales (RD/S), selling expenses over sales (SEIS), and quit rates (QR), thepercentage of the industry'stotal work force that voluntarilyleft their jobs inthe sample years. It is postulated that RD/S measuresuniqueness because firmsthat sell products with close substitutes ar'e likely to do less research anddevelopment since their innovations can be moreeasily duplicated.In addition,successful research and development projects lead to new productsthat differfrom those existing in the market. Firms with relatively unique products areexpected to advertise more and, in general, spend morein promotingand sellingtheir products. Hence, SE/S is expected to be positively related to uniqueness.However, t is expected that firms in industrieswith high quit rates areprobablyrelatively less uniquesince firms that produce relatively unique productstend toemploy workerswith high levels of job-specifichumancapitalwho will thus findit costly to leave theirjobs.It is apparent from two of the indicatorsof uniqueness, RD/S and SEIS, thatthis attribute may also be related to non-debt tax shields and collateral value.Researchand developmentand some selling expenses (such as advertising)canbe considered capital goods that are immediately expensed and cannot be usedas collateral.Given that our estimation technique can only imperfectlycontrolfor these other attributes,the uniquenessattributemay be negativelyrelated tothe observed debt ratio because of its positive correlationwith non-debt taxshields and its negativecorrelationwith collateral value.E. Industry Classification

Titman [39] suggests that firms that makeproducts requiring he availabilityof specialized servicing and spare parts will find liquidation especially costly.This indicates that firms manufacturingmachines and equipment should befinanced with relatively less debt. To measurethis, we include a dummyvariableequal to one for firms with SIC codes between 3400 and 4000 (firms producingmachines and equipment) and zero otherwise as a separate attribute affectingthe debt ratios.F. Size

A numberof authors have suggestedthat leverageratiosmaybe relatedto firmsize. Warner [41] and Ang, Chua, and McConnell [1] provide evidence that


7/20

6 The Journal of Financesuggests that direct bankruptcycosts appear to constitute a largerproportionofa firm's value as that value decreases.It is also the case that relatively large firmstend to be more diversified and less prone to bankruptcy. These argumentssuggest that large firms should be more highly leveraged.

The cost of issuing debt and equity securities is also related to firm size. Inparticular,small firms pay much more than largefirmsto issue new equity (seeSmith [34]) and also somewhat more to issue long-term debt. This suggests thatsmall firms may be more leveraged than large firms and may prefer to borrowshort term (throughbank loans) rather than issue long-term debt because of thelower fixed costs associated with this alternative.We use the natural logarithmof sales (LnS) and quit rates (QR) as indicatorsof size.5 The logarithmic transformation of sales reflects our view that a sizeeffect, if it exists, affects mainlythe very small firms. The inclusion of quit rates,as an indicatorof size, reflects the phenomenon that large firms, which oftenoffer wider career opportunitiesto their employees, have lower quit rates.G. Volatility

Many authors have also suggested that a firm's optimal debt level is adecreasingfunction of the volatility of earnings.6We were only able to includeone indicator of volatility that cannot be directly affected by the firm's debtlevel.7It is the standard deviation of the percentage changein operating ncome(SIGOI). Since it is the only indicator of volatility, we must assume that itmeasuresthis attributewithout error.H. Profitability

Myers [27]cites evidencefrom Donaldson [13]andBrealeyandMyers [7] thatsuggests that firms prefer raising capital, first from retained earnings, secondfromdebt,and third fromissuingnew equity.He suggeststhat this behaviormaybe due to the costs of issuing new equity. These can be the costs discussed inMyersand Majluf [28]that arise becauseof asymmetric nformation,orthey canbe transaction costs. In either case, the past profitabilityof a firm, and hencethe amount of earningsavailable to be retained,should be an importantdeter-minant of its current capital structure.We use the ratios of operatingincomeover sales (OI/S) and operatingincome over total assets (OI/TA) as indicatorsof profitability.'An unreportedmodelwas estimatedthat included he natural logarithmof total assets (LnTA)as a third indicatorof size. The very high correlationbetween LnTAand LnS (about 0.98) createdanear singularityin the covariance matrix, causing problems in estimating the model. However,

parameter stimatesof the structuralmodel are not sensitive to the choicebetweenLnS or LnTAasan indicator or size.6 Counter-exampleso this basichypothesishave beendemonstrated. See, forexample,Castaniasand DeAngelo [9], Jaffe and Westerfield[19],and Bradley,Jarrell,and Kim [6].)'Other possible indicators,such as a firm's stock beta or total volatility, are of coursepartiallydetermined y the firm'sdebt ratio.Calculating unlevered etas and unlevered olatilities equiresaccurate-measurementsf the marketvalueof the firm'sdebtratioas it evolves over time and of thetax gain to leverage.The potential for spuriouscorrelationarises if the impact of leverageand taxesis not completelypurged rom these volatilityestimates.


8/20

Capital Structure 7II. Measures of Capital Structure

Six measures of financial leverageare used in this study. They are long-term,short-term,and convertibledebt dividedby marketand by book values of equity.8Althoughthese variablescould have been combined to extract a common debtratio attribute, which could in turn be regressed against the independentattributes,there is goodreason for not doingthis. Some of the theories of capitalstructurehave different implications for the differenttypes of debt, and, for thereasons discussed below, the predicted coefficients in the structural model maydifferaccording o whether debt ratios aremeasured n terms of book or marketvalues. Moreover,measurementerrorsin the dependent variablesare subsumedin the disturbance erm and do not bias the regressioncoefficients.Data limitations forceus to measure debt in terms of book values rather thanmarket values. It would, perhaps, have been better if market value data wereavailablefor debt. However,Bowman [5] demonstrated hat the cross-sectionalcorrelationbetweenthe book value and market value of debt is very large,so themisspecification due to using book value measures is probably fairly small.Furthermore,we have no reason to suspect that the cross-sectionaldifferencesbetween market values and book values of debt should be correlatedwith any ofthe determinantsof capital structuresuggestedby theory,so no obvious bias willresult because of this misspecification.There are, however,some other importantsourcesof spuriouscorrelation.Thedependent variables used in this study can potentially be correlatedwith theexplanatoryvariables even if debt levels areset randomly.Consider irst the casewhere managersset their debt levels according o some randomlyselected targetratio measured at book value.9 This would not be irrationalif capital structurewere in fact irrelevant. If managersset debt levels in terms of book value ratherthan market value ratios, then differences in market values across firms thatarise for reasons other than differences in their book values (such as differentgrowth opportunities) will not necessarily affect the total amount of debt theyissue. Since these differencesdo,of course,affect the marketvalue of their equity,this will have the effect of causingfirms with higher market/bookvalue ratios tohave lowerdebt/marketvalue ratios. Since firmswith growth opportunities andrelatively low amounts of collateralizableassets tend to have relatively highmarketvalue/book value ratios, a spurious relation might exist between debt/market value and these variables, creating statistically significant coefficientestimates even if the book value debt ratios are selectedrandomly.108We also examinedthese debt levels dividedby total assets and marketvalue of equity plus bookvalue of debt and preferred tock. The resultsusing these dependentvariableswerevery similar tothose reportedhere.'There is evidencethat managersdo think in termsof book values.See, for example,the survey

evidencepresented n Stonehillet al. [37].10It may be easier to understandhow this spuriouscorrelationarises in the case where all firmshave the same book debt ratios. In this case, the cross-sectionalvariation n debt/marketvalue willbe determinedentirely by the variation in the differencesbetween book and market values acrossfirms. Variables that are related to this difference will, therefore, also be related to debt/marketvalue.One shouldnote that the previouslycited empiricalworksharesthis potential forspuriousresults.Clearly,the different measuresof firm size and industry classifications used in past studies arecorrelatedwith their marketvalue/book value ratios.


9/20

8 The Journal of FinanceSimilar spurious relations will be induced between debt ratios measured atbook valueandthe explanatoryvariablesif firms select debt levels in accordancewith marketvalue target ratios. If some firmsuse book value targets while othersuse marketvalue targets, both dependent variableswill be spuriously correlated

with the independent variables. Fortunately, the book and market value debtratios induce spurious correlation in opposite directions. Using dependent vari-ables scaled by both book values and marketvalues may then make it possibleto separate the effects of capital structure suggestedby theory, which predictscoefficient estimates of the same sign for both dependent variable groups, fromthese spurious effects.

III. DataThe variables discussed in the previous sections were analyzed over the 1974through1982 time period.The source of all the data except for the quit rates isthe Annual Compustat Industrial Files. The quit-rate data are from the U.S.Departmentof Labor,Bureauof LaborStatistics, Employmentand Earningspublication.These data are availableonly at the four-digit (SIC code) industrylevel for manufacturing irms.From the total sample, we deleted all the observations that did not have acomplete record on the variables included in our analysis. Furthermore,sincemany of the indicator variablesare scaled by total assets or average operatingincome, we were forcedto delete a small number of observations that includednegative values for one of these variables. These requirements may bias oursampletowardrelatively largefirms. In total, 469 firms were available.The sampling period was divided into three subperiodsof three years each,over which sample averagesof the variableswere calculated.Averagingover threeyears reducesthe measurementerror due to randomyear-to-yearfluctuationsinthe variables.The dependentvariables were measured duringthe 1977 through1979subperiod.Two of the indicatorsof expectedfuture growth,the growth rateof total assets (GTA) and capital expendituresover total assets (CE/TA), weremeasured over the period 1980 through 1982. By doing this, we are using therealized values as (imperfect) proxies of the values expected when the capitalstructure decision was made. The variables used to measureuniqueness, non-debt tax shields, asset structure,and the industryclassification were measuredcontemporaneouslywith the dependent variables, i.e., during the period 1977through 1979. The variables used as indicators of size and profitability weretaken from the 1974 through 1976 period. Measuringthe profitabilityattributeduringthe earlierperiod allows us to determinewhetherprofitabilityhas morethan just a short-term effect on observedleverageratios. Measuringsize in theearlierperiods avoids creating a spuriousrelation between size and debt ratiosthat arisesbecause of the relationbetween size and past profitability (profitablefirms become larger) and the short-term relation between profitability andleverage (profitablefirms increase their net worth). Finally, the standard devia-tion of the changein operatingincome was measuredusing all nine years in thesamplein order to obtain as efficient a measureas possible.


10/20

Capital Structure 9In comparing our results with results from previous research, it should benoted that our sample is somewhat more restricted than others. Our samplereflects the fact that most of the theories were developed with the knowledgethat regulatedandunregulated irmshavevery different capitalstructures.Given

this and other well-knowncapital structure differences between broad industrygroups,we think that it is moreappropriate o test the capital structuretheorieson a sample restricted only to those firms in the manufacturingsector of theeconomy. Of course, the drawbackto limiting the sample is the loss in powerassociatedwith reducingthe variationin the independentvariables.IV. The Model Specification

Section II discussed a number of attributes and their indicators that may intheory affect a firm'scapital structurechoice.Unfortunately,the theories do notspecifythe functional formsdescribinghow the attributes relate to the indicatorsandthe debt ratios. The statisticalproceduresusedto estimatethe modelrequirethat these relations be linear.The model we estimate is an applicationof the LISREL system developedbyK. Joreskog and D. Sorbom.11 t can be conveniently thought of as a factor-analytic model consisting of two parts: a measurementmodel and a structuralmodelthat are estimated simultaneously.In the measurementmodel,unobserv-able firm-specificattributes are measuredby relating them to observable vari-ables, e.g., accounting data. In the structuralmodel, measured debt ratios arespecifiedas functions of the attributesdefinedin the measurementmodel.The measurementmodelcan be specifiedas follows:

x = A+6,(1)where x is a q x 1 vector of observable indicators, t is an m x 1 vector ofunobservableattributes,and A is a q x m matrix of regressioncoefficients of xon t. Errorsof measurement are representedby the vector 3. In our model,wehave fifteen indicator variables foreight attributes-thus, x is 15 x 1 and A is 15x 8.The structuralmodel can be specifiedas the following system of equations:

y = rt + c, (2)wherey is a p x 1 vector of debt ratios, r is a p x m matrix of factor loadings,and e is a p x 1 vector of disturbance terms. The model is estimated for twoseparate3 x 1 vectors of debt:short-term, ong-term,and convertibledebt scaledby book value and marketvalue of equity.Equation (1) simply states that, although the firm-specific attributes thatsupposedlydeterminecapital structures cannot be observed,a number of othervariables denoted as indicators,that are imperfect measuresof the attributes,areobservable.These indicator variables can be expressedas linear functions of oneor moreof the unobservableattributes and a random measurementerror.

Fora lucidintroduction o the systemand its many applications, ee Joreskogand Sorbom[23];for a critical reviewsee Bentler [3];a detailed descriptionof the technical proceduresof LISREL isgiven in Joreskog[21].


11/20

10 TheJournalof FinanceTable I

The Structureof the MeasurementModelNDT/TA 0 0 X1,3 0 0 0 0 0 6 61ITC/TA 0 0 X2,3 0 0 0 0 0 6 62D/TA 0 0 X3,3 0 0 0 0 0 6 63RD/S X4,1 X4,2 0 0 0 0 0 0 44 64SE/S 0 X5,2 0 0 0 0 0 0 6 65CE/TA X6,1 0 0 0 0 0 0 0 46 66INT/TA 0 0 0 X7,4 0 0 0 0 47 67IGP/TA = 0 0 0 X8,4 0 0 0 0 X L8 + 68LnS 0 0 0 0 X9,5 0 0 0 69GTA X10,1 0 0 0 0 0 0 0 610QR 0 X11,2 0 0 X11,5 0 0 0 allOI/TA 0 0 0 0 0 X12,6 0 ? 612OI/S 0 0 0 0 0 X12,7 0 0 613SIGOI 0 0 0 0 0 0 1 0 0IDUM o 0 0 0 0 0 0 1 0

The principaladvantageof this estimationprocedureover standardregressionmodels is that it explicitly specifies the relation between the unobservableattributes and the observablevariables. However, in order to identify the esti-mated equations, additional structure must be added. In most factor-analysismodels,the commonfactorsare constrainedto be orthogonaland scaledto haveunit variances,and the residualsare assumedto be uncorrelated.However,sincethe commonfactorsin this studyaregivendefinite interpretationsby identifyingthem with specific attributes, the assumption that the common factors areuncorrelatedis untenable since many firm-specific attributes are likely to berelated (e.g., size and growth). For this reason, the correlations among theunobservableattributes (thematrixT ) are estimatedwithin the model.Ofcourse,in orderto achieve identification, additional restrictions on the parametersofthe modelmust be, mposed.In total, we have imposed 105 restrictionson the matrixA of factor loadings.These are shown in Table I as the factorloadingsthat are exogenouslyspecifiedto equal either one or zero. For example, since RD/S is not assumedto be anindicatorof size, its factor loadingon the size attributeis set to zero and is notestimatedwithin the model. In addition,we have also constrainedthe measure-ment errorin the equationof indicatorvariablesSIGOIand IDUM to be zero,implyingthat the factorloadingsof these variableson their respectiveattributesare constrained o equalone.Also,we have assumedthat the measurementerrors,6, are uncorrelatedwith each other, with the attributes, and with the errorsinthe structuralequations.Since the restrictions may not all be appropriate, nterpretationsof the esti-mates shouldbe made with caution. It is quite likely, for example,that some ofthe measurementerrors may in fact be correlated.It is unfortunatethat there isan arbitraryelement in the choice of identifying restrictions;however,similarrestrictionsmust be made implicitly in orderto interpreta standardregressionmodelthat uses proxyvariables.


12/20

Capital Structure 11In contrast to the measurement model, the structural model is totally unre-stricted. The model estimates the impactof each of the attributes on each of thedifferent debt ratios. In otherwords,none of the factorloadingsin the structuralequationsis fixed exogenously.In addition,the correlationsbetweenthe residual

errorsin the structural equations are estimated within the model. This allowsfor the possibility that there exist additional attributes, not considered in themodel,that are determinantsof each of the debt ratios.V. Estimates of the Parameters

The parametersof our model can be estimated by fitting the covariance matrixof observablevariables implied by the specification of the model (i) to thecovariance matrix (S) of these variables observed from the sample. In theLISREL system, this is done by minimizingthe function,

F = log(det ) + tr(SX1) - log(det S) - (p + q), (3)with respect to the vector of parametersof the matrices referred o above.Thisfitting function is derived from maximum-likelihoodproceduresand assumesthat the observedvariables are conditionally multinormallydistributed.Our estimates of the parametersof the measurement model are presentedinTables II and III. The estimates are generallyin accord with our a priori ideasabouthow well the indicatorvariablesmeasure the unobservedattributes.Boththe direction and the magnitude,as well as the statistical significance,of theestimates suggestthat these indicatorscapturethe conceptswe wish to consideras determinantsof capital structure choice.12The estimates of the structural coefficients are presentedin Table IV. Thesecoefficients specify the estimated impact of the unobservedattributes on theobserved debt ratios. For the most part, the coefficient estimates for the long-term and short-term debt ratios were of the predicted sign. However, many ofthe estimated coefficients are fairly small in magnitude and are statisticallyinsignificant.Iniparticular, he attributes representingnon-debt tax shields, assetstructure,and volatility do not appearto be related to the various measuresofleverage. Moreover,the estimated models explain virtually none of the cross-sectional variationin the convertibledebt ratios.Some of the coefficient estimates are both largein magnitudeand statisticallysignificant. The large negative coefficient estimate for the uniquenessattribute(Q2) indicates that firms characterized as having relatively large researchanddevelopment expendituresand high selling expenses, and that have employeeswith relatively low quit rates, tend to have low debt ratios. The coefficient

12 The goodnessof fit of the measurementmodel was evaluatedby testing the modelagainsttwoalternatives.(See Bentler and Bonett [4].) The first alternativemodelis one in whichthe observedvariablesare assumedto be mutuallyuncorrelated-the x2-statisticequals 1893,with 42 degreesoffreedom,which for this test is highly significant.The second alternativemodel specifies that thecovariancematrix s totallyunrestricted.The x2-statisticequals 378,with 63 degreesof freedom;hisresult is also highly significant, ndicating hat, althoughourmodel capturesa significantpartof theinformation n the samplecovariancematrix, relaxingone or more of the imposedrestrictionscouldimprove he fit of the model.


13/20

12 The Journal of Finance

TableII

Measure

Model:

Factor

Loadingsfor

Independ

Variables

Attributes

Variable~1

~

4

~

6

~

8

2

(Growth)

(Uniqueness)

(Non-Debt)

(Collateral

(Size)

(Profitability)

(Volatility)

(IDUM)

Tax

Shields

Value)

NDT/TA

0.779

0.393

(26.7)

ITC/TA

0.606

0.744

(19.2)

D/TA

0.848

0.280

(30.1)

RD/S

0.246

0.781

0.401

(6.6)

(21.6)

SE/S

0.681

0.536

(19.7)

CE/TA

0.951

0.095

(26.4)

INT/TA

-0.331

0.891

(-8.7)

IGP/TA

1.180

-0.392

(15.7)

LnS

0.938

0.120

(7.9)

GTA

0.471

0.778

(13.9)

QR

-0.228

-0.273

0.896

(-5.6)

(-5.5)

OI/TA

0.641

0.589

(18.8)

OI/S

0.998

0.005

(27.8)

SIGOI

1.000

0.000

IDUM

1.000

0.000

a

Reported

t-statisticsarein

parentheses.


14/20

Capital Structure 13

TableIII

Estimated

Correlation

between

Attributes

Attribute

41

t2

43

t4

t5

46

47

48

4t

(Growth)

1.00

42

(Uniqueness)

-0.18

1.00

43

(Non-DebtTax

Shields)

0.72

-0.04

1.00

4

(Asset

Structure)

0.27

-0.39

0.47

1.00

45(Size)

0.15

-0.18

0.19

0.28

1.00

46

(Profitability)

0.53

0.12

0.46

0.12

-0.02

1.00

47

(Volatility)

-0.08

-0.01

-0.02

0.03

-0.11

-0.04

1.00

t8

(Industry

Dummy)

-0.14

0.38

-0.13

-0.22

-0.24

-0.10

-0.05

1.00

TableIV

Estimatesof

Structural

CoefficientsaAttributes

Debt

Measures

4,

42

43

44

48

46

47

48

(Growth)

(Uniqueness)

(Non-Debt

(Asset

(Size)

(Profitability)

(Volatility)

(Industry

TaxShields)

Structure)

Dummy)

1.

LT/MVE

-0.068

-0.263

-0.058

0.041

-0.033

-0.213

-0.031

-0.106

(-0.7)

(-3.7)

(-0.6)

(0.8)

(-0.6)

(-3.7)

(-0.7)

(-2.1)

ST/MVE

-0.112

-0.260

-0.041

-0.046

-0.183

-0.179

-0.017

-0.063

(-1.2)

(-3.7)

(-0.4)

(-0.9)

(-3.2)

(-3.1)

(-0.4)

(-1.2)

C/MVE

-0.067

-0.076

-0.050

0.004

0.055

-0.108

-0.027

0.026

(-0.7)

(-1.0)

(-0.5)

(0.1)

(1.0)

(-1.8)

(-0.6)

(0.5)

2.

LT/BVE

0.230

-0.281

-0.113

-0.076

-0.132

-0.052

-0.043

-0.066

(2.4)

(-3.6)

(-1.1)

(-1.4)

(-2.3)

(-0.9)

(-0.9)

(-1.2)

STIBVE

0.140

-0.185

-0.079

-0.096

-0.284

-0.044

-0.038

-0.051

(1.5)

(-2.4)

(-0.8)

(-1.7)

(-4.1)

(-0.7)

(-0.8)

(-0.9)

C/BVE

0.028

-0.065

-0.156

-0.019

0.050

0.026

-0.016

0.074

(0.3)

(-0.8)

(-1.5)

(-0.3)

(0.9)

(0.4)

(0.3)

(1.3)

'The

coefficient

estimatesarescaledto

representthe

estimated

changeinthe

dependent

variable,

relativetoits

variance,

with

respecttoachangeinan

attribute,

relativetoits

variance.

Reported

t-statisticsarein

parentheses.


15/20

14 The Journal of Financeestimate of -0.263 in the equation with LT/MVE indicates that firms that differin uniqueness by one variance are expectedto have long-term debt ratios thatdifferby 0.263 variances. This evidence,alongwith the estimated coefficientsofthe industry dummy variable, are consistent with the implications of Titman[39]. As mentioned previously, a negative relation between uniquenessand thedebt ratios could also be due to the relationbetweenthis attribute and non-debttax shields and collateral values.Although the reportedt-statistics for the coefficientsof the uniqueness attrib-ute arequite high,we feel that their statistical significanceshould be interpretedcautiously.The reported -statistics are basedonthe assumptionsof independent,identical, and normally distributed error terms-assumptions that are surelyviolatedby our data. To providefurtherevidence about the statisticalsignificanceof the relation between uniqueness, the industry dummy,and the measureddebtratios, we compared the estimated likelihood function for the reported modelwith the likelihoodfunctionof the modelwith the coefficients of uniquenessandthe industrydummyconstrainedto equalzero.The difference n these likelihoodfunctions has a x2 distributionwith six degreesof freedom.The estimatedx2 forthe debt ratios scaled by market values equals forty, which is statisticallysignificant at well beyond the 0.005 level. With the debt ratios scaled by bookvalue of equity, this statistic equals seventeen, which is significant at the 0.01level. Given these results, we feel comfortable in asserting that the evidencesupportsthe implicationof Titman [39] that firms that can potentially imposehigh costs on their customers, workers,and suppliersin the event of liquidationtend to choose lower debt ratios.The evidence also indicates that small firms tend to use significantly moreshort-term financing than large firms. This difference in financing practiceprobablyreflects the hightransaction costs that small firmsface whenthey issuelong-term debt or equity. Our finding that small firms use more short-termfinancing may also provide some insights aboutpossible risk factors underlyingthe small-firmeffect. By borrowingmore short term, these firms are particu-larly sensitive to temporary economic downturns that have less of an effect onlargerfirms that are less leveragedand use longer term financing. (See Chan,Chen,and Hsieh [10] for a similarargumentand evidencerelatingto this.)The results also suggestthat size is relatedto LTD/BVE but not LTD/MVE.This finding may be due to the positive relation between our size attribute andthe total market value of the firm. Firms with high market values relative totheir book values have higher borrowingcapacities and hence have higher debtlevels relative to their book values.Thus, rather than indicatinga size effect, wethink that this evidencesuggeststhat manyfirmsareguidedby the market valueof their equitywhen selectingtheir long-termdebt levels.

Coefficient estimates for the profitability ttribute are largeand have hight-statistics in the equations with debt over market value of equity-dependentvariables,but they are not statistically significantin the equationswith the debtmeasures scaled by book value of equity. This suggests that increases in themarketvalueof equity,dueto an increasein operating ncome,arenot completelyoffset by an increase in the firm'sborrowing.This providesadditionalevidencesupporting he importanceof transaction costs and is consistent with the obser-


16/20

CapitalStructure 15vation of Myers [27] regardingwhat he calls the pecking order theory thatfirmsprefer internal to external financing. However, the evidencesuggeststhatborrowing s increased to the extent that the higher operatingincome leads toan increase in the book value of equity (through increasesin retainedearnings).This suggests that many firms do in fact use book value target debt-to-equityratios.It should be emphasized that the significant coefficient estimates for profita-bility and size are not necessarily inconsistent with the hypothesis of capitalstructure irrelevance. As we mentioned in Section III, significant coefficientestimates for either (but not both) the market value or book value equationsareconsistent with debt ratiosbeingchosen randomly.Similarly,we should not viewthe positive coefficient estimate of the growth attribute in the long-term debtover book value of the equityequationas necessarilybeing inconsistent with theagency-and tax-based theories that predicta negativecoefficient for this attrib-ute. The observedpositive coefficient simply implies that, since growthoppor-tunities add value to a firm, they increasethe firm's debt capacity and, hence,the ratio of debt to book value, since this additional value is not reflectedin thefirm's book value.

VI. RobustnessAn examination of the correlation matrixof the sampledata (Table V) providessome insights aboutthe robustnessof our results.Particularlynoteworthyis thehigh negative simple correlationbetween OI/TA and the various debt ratios.This relation can potentially create a problem in interpretingthe correlationbetween variables scaledby either OIor TA and the debt ratio measures.The best examples of this are the indicators of non-debt tax shields. Forinstance, the simplecorrelationbetween NDT/TA and the different measures ofleverage s strongly negative. While this correlation s predictedby the DeAngeloand Masulis [12] model, it should be noted that the large negative correlationmaybe dueto the largepositive correlationbetweenOI/TAand NDT/TA causedby their common denominators.In the estimated structural model, where wecontrol for the profitabilityattribute that is measuredby OI/TA and OI/S, thecoefficient estimate for the non-debt tax shield attribute is not statisticallysignificant. Moreover, f we replace the denominatorsof the non-debt tax shieldindicatorswith OI, the simplecorrelationsare still just as strong but are reversed.Forexample, NDT/OI is strongly negativelycorrelatedwith OI/TA and stronglypositively correlated with the measures of leverage.Using indicators scaled byOIfor the non-debt tax shield attribute leads to positive coefficient estimatesthat aresometimesmarginallystatistically significantin the structuralequations.While this resultis inconsistent with the DeAngeloand Masulis model, it is mostlikely caused by the way the variables used as indicators are scaled.We expect that similar changes in coefficient estimates caused by scalingindicatorvariablesby OIrather than TA would be found for other attributes.For example,IGP/OI is very highly correlatedwith the inverse of OI/TA and istherefore probably strongly positively correlated with leverage. This positivecorrelationcould be put forth as evidencein supportof the hypothesisthat firms


17/20

16 TheJournalof Finance

TableVCorrelation

Matrix

LT/ST/C/

LT/ST/C/

NDT/ITC/D/

RD/SEICEI

INTIIGP/

GTAQR

/O

SIGOI

IDUM

MVE

MVEMVEBVEBVEBVETATATASS

TA

TATA

LnS

TAS

LT/MVE

1.

ST/MVE

0.661.

C/MVE

0.29

0.191.

LT/BVE

0.73

0.47

0.151.

ST/BVE

0.43

0.75

0.10

0.661.

C/BVE

0.15

0.10

0.89

0.14

0.111.

NDT/TA

-0.25

-0.32

-0.15

-0.11

-0.17

0.141.

ITC/TA

-0.06

-0.14

-0.06

0.09

-0.02

-0.07

0.461.

D/TA

-0.08

-0.12

-0.10

0.02

-0.02

-0.09

0.66

0.521.

RD/S

-0.27

-0.24

-0.07

-0.19

-0.12

-0.03

0.30

-0.04

0.111.

SE/S

-0.25

-0.14

-0.08

-0.20

-0.06

-0.04

0.06

-0.24

-0.10

0.501.

CE/TA

-0.14

-0.20

-0.13

0.14

0.04

-0.06

0.51

0.47

0.58

0.09

-0.131.

INT/TA

0.03

0.02

0.13

0.00

-0.01

0.12

-0.13

-0.11

-0.17

-0.03

0.23

-0.091.

IGP/TA

0.09

0.06

0.03

0.01

0.11

0.08

0.37

0.39

0.51

-0.22

-0.43

0.31

-0.39L.

LnS

0.04

-0.14

0.05

-0.07

-0.24

0.01

0.18

-0.01

0.17

-0.01

-0.25

0.14

-0.09

0.311.

GTA

-0.20

-0.22

-0.17

0.04

-0.01

-0.10

0.24

0.26

0.27

0.18

0.07

0.45

-0.03

-0.01

-0.181.

QR

0.11

0.19

0.06

0.10

0.14

0.02

-0.13

0.03

-0.01

-0.23

-0.01

-0.04

0.07

-0.13

-0.22

0.071.

OI/TA

-0.38

-0.34

-0.23-0.24

-0.19

-0.17

0.31

0.18

0.29

0.03

0.07

0.25

-0.04

0.09

-0.01

0.20

-0.061.

OI/S

-0.29

-0.28

-0.18

-0.02

-0.03

-0.05

0.41

0.20

0.39

0.19

0.12

0.50

-0.06

0.14

-0.02

0.29

-0.13

0.641.

SIGOI

0.00

0.03

-0.02

-0.04

-0.01

-0.02

-0.02

-0.02

-0.01

0.02

-0.05

-0.07

0.04

-0.04

-0.10

-0.09

-0.04

-0.02

-0.04

1.

IDUM

-0.17

-0.07

0.01

-0.13

-0.04

0.06

-0.08

-0.16

-0.11

0.32

0.16

-0.14

-0.04

-0.25

0.23

-0.03

-0.18

-0.11

-0.10

-0.06L


18/20

CapitalStructure 17with assets that have high collateral value choose high debt levels. However,wefeel that such a variable wouldactuallybe measuringprofitabilityand thereforethought it more appropriate o scale the variableby TA.The above discussion suggests that one shouldbe cautiouswhen interpretingvariablesscaled by operating ncome (OI)that arepositivelycorrelatedwith debtratios and to a lesser extent variables scaled by total assets (TA) that arenegativelyrelated to the debt ratios.Fortunately,the indicators of uniqueness ,the attribute that appears o do the best job explainingdebtratios,are not scaledby either TA or OI. Research and developmentexpenditurescould conceivablyhave been scaled by OI or TA; however,the correlationbetween debt and thisvariable is not nearlyas sensitive to its scalingas, for example,NDT. The otherindicatorsof uniqueness ,SE/S and QR,suggestno alternativescaling variable;hence, the robustness of the correlation between these variables and the debtratios is not a seriousissue.

VII. Summary and ConclusionThis paper introduceda factor-analytictechnique for estimating the impact ofunobservableattributeson the choice of corporatedebt ratios. While our resultsare not conclusive, they serve to documentempirical regularitiesthat are con-sistent with existing theory.In particular,we find that debt levels are negativelyrelatedto the uniqueness f a firm's ine of business. This evidenceis consistentwith the implicationsof Titman [39] that firmsthat can potentiallyimpose highcosts on their customers,workers,and suppliersin the event of liquidationhavelowerdebt ratios.The results also indicate that transactioncosts may be an important determi-nant of capital structure choice. Short-term debt ratios were shown to benegativelyrelatedto firm size, possibly reflectingthe relatively high transactioncosts small firms face when issuing long-term financial instruments. Sincetransactioncosts aregenerallyassumedto be small relativeto other determinantsof capital structure, their importance in this study suggests that the variousleverage-relatedcosts and benefits may not be particularly significant. In thissense, althoughthe results suggestthat capital structures are chosen systemati-cally, they are in line with Miller's [25] argumentthat the costs and benefitsassociated with this decision are small. Additional evidence relating to theimportance of transaction costs is providedby the negative relation betweenmeasures of past profitabilityand current debt levels scaledby the market valueof equity. This evidence also supports some of the implications of Myers andMajluf[28] and Myers [27].Our results do not provide support for an effect on debt ratios arising fromnon-debt tax shields, volatility, collateral value, or future growth.However, itremains an open question whetherour measurementmodel does indeedcapturethe relevantaspectsof the attributessuggestedby these theories.Onecould arguethat the predictedeffects were not uncoveredbecause the indicatorsused in thisstudy do not adequatelyreflect the nature of the attributessuggestedby theory.If stronger linkages between observable indicator variables and the relevantattributes can be developed,then the methods suggested in this paper can beused to test morepreciselythe extant theories of optimal capital structure.


19/20

18 TheJournalof FinanceREFERENCES

1. J. Ang, J. Chua, and J. McConnell. The AdministrativeCosts of CorporateBankruptcy:ANote. Journal ofFinance 37 (March 1982), 219-26.2. A. Auerbach. RealDeterminantsof CorporateLeverage. n B. Friedman ed.),Corporate apitalStructures n the UnitedStates. Chicago:University of ChicagoPress, 1985.3. P. M. Bentler. SimultaneousEquation Systems as Moment StructureModels.With an Intro-duction o Latent VariableModels. ournalof Econometrics 2 (Issue of the Annals of AppliedEconometrics) March1983),13-42.4. and D. G. Bonett. SignificanceTests and Goodnessof Fit in the Analysisof CovarianceStructures. PsychologicalBulletin 88 (September1980),588-606.5. J. Bowman. The Importanceof a Market Value Measurementof Debt in Assessing Leverage.Journal of AccountingResearch18 (Spring 1980),242-54.6. M. Bradley, G. Jarrell,and E. H. Kim. On the Existence of an OptimalCapital Structure:Theory and Evidence. ournal ofFinance 39 (July 1984), 857-78.7. R. Brealeyand S. Myers. Principlesof Corporate inance. New York:McGraw-Hill,1984.

8. R. Castanias. BankruptcyRiskandOptimalCapitalStructure. ournalofFinance38 (December1983),1617-35.9. and H. DeAngelo. BusinessRisk and OptimalCapitalStructure. Unpublishedworkingpaper,Universityof Washington,1981.10. K. C. Chan, N. F. Chen, and D. Hsieh. AnExploratory nvestigationof the Firm Size Effect.JournalofFinancialEconomics14 (March 1985),451-71.11. S. Chaplinsky. The EconomicDeterminantsof Leverage:TheoriesandEvidence. UnpublishedPh.D. Dissertation,Universityof Chicago,1983.12. H. DeAngeloand R. Masulis. OptimalCapitalStructureunderCorporateand Personal Taxa-tion. JournalofFinancialEconomics8 (March 1980),3-29.13. G. Donaldson. CorporateDebt Capacity:A Study of CorporateDebt Policy and the Determi-nation of CorporateDebt Capacity. Boston: Divisionof Research,HarvardSchoolof BusinessAdministration,1961.14. M.Ferriand W. Jones. Determinants f FinancialStructure:A New MethodologicalApproach.Journalof Finance 34 (June 1979),631-44.15. D. Flath and C. Knoeber. Taxes,FailureCosts, and Optimal Industry Capital Structure:AnEmpiricalTest. Journal ofFinance35 (March 1980), 99-117.16. D. Galai and R. Masulis. TheOption Pricing Modeland the Risk Factor of Stock. Journal ofFinancialEconomics3 (January/March1976), 53-81.17. R. Green. InvestmentIncentives, Debt, and Warrants. Journal of FinancialEconomics 13(March 1984),115-35.18. S. Grossmanand 0. Hart. CorporateFinancial Structureand ManagerialIncentives. In J.McCall(ed.), The Economicsof Informationand Uncertainty.Chicago:Universityof ChicagoPress, 1982.19. J. Jaffe and R. Westerfield. Risk and OptimalDebt Level. Working Paper, University ofPennsylvania,1984.20. M. Jensen and W. Meckling. Theoryof the Firm:ManagerialBehavior,Agency Costs andOwnershipStructure. ournalofFinancialEconomics3 (October1976),305-60.21. K. G. Joreskog. StructuralEquationModels in the Social Sciences:SpecificationEstimationandTesting. In P. R. Krishnaiah ed.), 1977.22. and D. Sorbon. LISRELV, Analysisof LinearStructuralRelationshipsby the Methodof MaximumLikelihood. Chicago:National EducationalResources,1981.23. M. S. Long and E. B. Malitz. InvestmentPatterns and FinancialLeverage. n B. Friedman(ed.), CorporateCapitalStructures n the UnitedStates. Chicago:Universityof ChicagoPress,1985.24. P. Marsh. The ChoicebetweenEquity and Debt: An EmpiricalStudy. Journalof Finance 37(March1982),121-44.25. M. Miller. Debtand Taxes. JournalofFinance32 (May 1977), 261-75.26. S. Myers. Determinants f CorporateBorrowing. ournalofFinancialEconomics9 (November1977), 147-76.


20/20

Capital Structure 1927. . The Capital StructurePuzzle. Journalof Finance 39 (July 1984), 575-92.28. and N. Majluf. CorporateFinancing and Investment Decisions When Firms HaveInformation nvestorsDo Not Have. Journalof FinancialEconomics13 (June 1984),187-221.29. R. Roll and S. Ross. An EmpiricalInvestigationof the ArbitragePricing Theory. JournalofFinance 35 (December1980), 1073-1103.30. E. Schwartzand J. Aronson. Some SurrogateEvidence in Support of the Concept of OptimalFinancial Structure. ournalof Finance22 (March 1967), 10-19.31. D. Scott. Evidence n the Importance f FinancialStructure. inancialManagement (Summer1972), 45-50.32. and J. Martin. IndustryInfluenceon Financial Structure. Financial Management4(Spring 1975),67-73.33. J. Scott. Bankruptcy,SecuredDebt, and Optimal Capital Structure. Journal of Finance 32(March1977), 1-20.34. C. Smith. AlternativeMethods for Raising Capital: Rights versus UnderwrittenOfferings.

Journal of Financial Economics 5 (December1977), 273-307.35. andJ. Warner. Bankruptcy, ecuredDebt, and OptimalCapitalStructure:Comments.Journal of Finance 34 (March1979),247-52.36. . On Financial Contracting:An Analysis of Bond Covenants. Journal of FinancialEconomics7 (June 1979),117-61.37. A. Stonehill,T. Beekhuisen,R. Wright,L. Remmers,N. Toy, A. Pares, A. Shapiro,D. Egan, andT. Bates. Determinantsof CorporateFinancial Structure:A Survey of Practice in FiveCountries. Unpublishedpaper,1973.38. S. Titman. Determinants f CapitalStructure:An EmpiricalAnalysis. WorkingPaper,UCLA,1982.39. . TheEffect of CapitalStructureon a Firm'sLiquidationDecision. Journalof FinancialEconomics13 (March 1984),137-51.40. N. Toy, A. Stonehill,L. Remmers,R. Wright,and T. Beekhuisen. AComparativenternationalStudy of Growth, Profitability,and Risk as Determinantsof CorporateDebt Ratios in theManufacturing Sector. Journal of Financial and Quantitative Analysis (November 1974), 875-86.41. J. Warner. BankruptcyCosts:SomeEvidence. Journalof Finance 32 (May 1977),337-47.

determinants of capital structure choice-titman

Documents