The (Un)secured Debt Puzzle: Evidence from U.S. Public ......5The literature is extense in terms of...

33
The (Un)secured Debt Puzzle: Evidence from U.S. Public Manufacturing Firms, 1994-2010 * Kizkitza Biguri Universitat Autonoma de Barcelona First Version: June 2013 June 11, 2014 Abstract The benefits of pledging (inside) collateral have been widely discussed in the literature because collateral helps solving market imperfections that are caused by asymmetric information and limited contract enforceabil- ity problems. However, the implicit assumption made is that the only financial contract available is secured debt, ruling out the role of unse- cured debt, which represents 64% of total debt outstanding of U.S. public manufacturing firms from 1994 to 2010. My paper has three interesting results. First, I show that debt structure is not solely determined by col- lateral, but by the interaction between collateral and financial strength, which determines unsecured debt. In addition to this, I show that col- lateral only plays a role for those firms that are financially constrained. Second, I prove that higher collateral does not increase borrowing capac- ity by lowering the financial frictions faced, but only secured debt bor- rowing capacity. This result allows for a complementary channel to the so-called collateral channel, the unsecured channel. Moreover, I test the pecking-order hypothesis and conclude that firms have a clear preference for unsecured debt because it allows to minimize total costs of financing. * I want to thank Filippo Ippolito (UPF), Gregory Udell (Kelley School of Business, Uni- versity of Indiana), Ronald Mann (Columbia Law School), Lynn LoPucki (UCLA School of Law), the participants of the UPF Finance Seminar and the UPF Student Seminar for use- ful comments. Finally, I would also want to thank my advisor Hugo Rodriguez Mendizabal (IAE-CSIC) for amazing supervising work and unconditional support. 1

Transcript of The (Un)secured Debt Puzzle: Evidence from U.S. Public ......5The literature is extense in terms of...

Page 1: The (Un)secured Debt Puzzle: Evidence from U.S. Public ......5The literature is extense in terms of consumers’ unsecured credit, including papers as Chatterjee, Corbae and Nakajima

The (Un)secured Debt Puzzle: Evidence from

U.S. Public Manufacturing Firms, 1994-2010∗

Kizkitza BiguriUniversitat Autonoma de Barcelona

First Version: June 2013

June 11, 2014

Abstract

The benefits of pledging (inside) collateral have been widely discussedin the literature because collateral helps solving market imperfections thatare caused by asymmetric information and limited contract enforceabil-ity problems. However, the implicit assumption made is that the onlyfinancial contract available is secured debt, ruling out the role of unse-cured debt, which represents 64% of total debt outstanding of U.S. publicmanufacturing firms from 1994 to 2010. My paper has three interestingresults. First, I show that debt structure is not solely determined by col-lateral, but by the interaction between collateral and financial strength,which determines unsecured debt. In addition to this, I show that col-lateral only plays a role for those firms that are financially constrained.Second, I prove that higher collateral does not increase borrowing capac-ity by lowering the financial frictions faced, but only secured debt bor-rowing capacity. This result allows for a complementary channel to theso-called collateral channel, the unsecured channel. Moreover, I test thepecking-order hypothesis and conclude that firms have a clear preferencefor unsecured debt because it allows to minimize total costs of financing.

∗I want to thank Filippo Ippolito (UPF), Gregory Udell (Kelley School of Business, Uni-versity of Indiana), Ronald Mann (Columbia Law School), Lynn LoPucki (UCLA School ofLaw), the participants of the UPF Finance Seminar and the UPF Student Seminar for use-ful comments. Finally, I would also want to thank my advisor Hugo Rodriguez Mendizabal(IAE-CSIC) for amazing supervising work and unconditional support.

1

Page 2: The (Un)secured Debt Puzzle: Evidence from U.S. Public ......5The literature is extense in terms of consumers’ unsecured credit, including papers as Chatterjee, Corbae and Nakajima

1 Introduction

The benefits of pledging (inside) collateral have been widely discussed in theliterature. Collateral helps solving market imperfections that are caused byasymmetric information problems, such as moral hazard, adverse selection orcostly state verification. Myers (1977), claimed that granting collateral reducesagency costs by alleviating the underinvestment problem. Similarly, Smith andWarner (1979b) and Stulz and Johnson (1985) show that pledging collateralmay lower a firm’s total cost of debt by preventing asset substitution, reducingforeclosure costs, limiting claim dilution and mitigating underinvestment. Fi-nally, Smith and Warner (1979a) argue that collateral reduces adverse selectionproblems under asymmetric information because creditors cannot distinguishbetween good and bad borrowers1.

The literature on limited contract enforceability also highlights the role of col-lateral in solving financial frictions. Papers like Bernanke, Gertler and Gilchrist(1996), Kiyotaki and Moore (1997) or Livdan, Sapriza and Zhang (2009) con-tain a borrowing constraint for firms which allows them to borrow up to theexpected liquidation value of the tangible assets of the firm. That is, the collat-eral they hold. The intuition in this type of financial frictions models is clear:the higher the collateral the firm can pledge, the higher the debt financing itcan achieve and therefore, the lower the probability of not undertaking someprofitable investment project due to credit constraints. Moreover, these type ofmodels give a predominant role to collateral in the propagation and amplifica-tion of exogenous shock to the real economy through the balance sheet channeland the bank-lending channel, from a demand of credit perspective.

However, all these papers, implicitly focus on secured debt financing and deter-mine that below some collateral threshold, firms will be financially constrained.Secured debt is by definition debt that is backed with collateral of the firm(asset-based lending). A specific set of assets are encumbered in order to signthe debt contract, such that in the event of default, the creditor has the rightto liquidate the assets attached to the debt contract in order to satisfy debtrepayment. On the other hand, unsecured debt is not backed with collateralof the firm (cash flow-based lending). Creditors extend credit on an unsecuredbasis according to the financial strength of the firm and the cash flows that thefirm is expected to generate in the future through investment decisions2.

1I will base the analysis on inside collateral, the pledgeable assets of the firm, as opposedto external collateral, which considers the personal assets of shareholders. The majority of thepapers highlighting the benefits of collateral, actually consider external collateral. However,external collateral does not play any role in public firms that are atomistically owned, whichis the set-up that I will consider in the present work.

2This definition of debt structure heterogeneity, the combination of secured and unsecureddebt, becomes relevant in terms of the priority structure upon default. Secured claims aresenior to unsecured and the incentives that both types of creditors face are very different.

2

Page 3: The (Un)secured Debt Puzzle: Evidence from U.S. Public ......5The literature is extense in terms of consumers’ unsecured credit, including papers as Chatterjee, Corbae and Nakajima

So far, the literature has not focused on the role of unsecured debt3. Thisis surprising as unsecured debt is as important in the financing structure offirms as secured debt at least quantitatively. In particular, 64% of debt out-sanding of U.S. public firms is unsecured, using Standard & Poor’s4 Compustatdatabase, for a period covered from 1994 to 2010.

As opposed to the economics and finance literature5, back since the late 70’sthe law literature has emphasized about the relevant role that unsecured debtplays in the context of firm’s financial and investment policies and, on creditor’sbargaining process upon default. They introduced a very popular concept, the“secured debt puzzle”6: the fact that regardless of the benefits that pledgingcollateral might offer, firms that are large and financially strong rather want torely on unsecured debt in order to finance their investment projects. Accordingto this concept, debt choice is determined by firm characteristics other thancollateral availability.

According to the results derived in the present paper, the role of unsecureddebt is relevant. If we allow for a second dimension to play a role, besides col-lateral availability of the firm, I show that debt structure is determined by theinteraction between collateral and financial strength. Moreover, collateral onlyplays a role for firms that are financially constrained. Second, I show that interms of the balance sheet channel, more collateral only means more secureddebt borrowing capacity, which allows for a complementary channel: the unse-cured channel. Furthermore, when collateral availability increases, firms thatare financially constrained, surprisingly, also show a preference for unsecureddebt. Finally, I test the pecking-order hypothesis and conclude that the strongpreference for unsecured relies on the lower financing costs attached.

Several papers argue that debt structure is relevant for many aspects of firmstrategic decisions such as investment and dividend policies. For example, Rauhand Sufi (2010) show that corporate debt is different in terms of types, sourcesand maturities using a sample of U.S. public firms. Additionally, they show that

3A recent empirical paper by Giambona and Golec (2012) shows a positive correlationbetween the firms’ investment opportunities, measured by Tobin’s q, and the usage of unse-cured debt. They claim for the existence of a growth opportunity channel of debt structureof firms, giving a predominant role to unsecured debt. These results are consistent with anearlier study by Barclay and Smith (1995), which show that firms trying to exploit investmentopportunities do not rely on secured debt.

4S&P’s hereafter.5The literature is extense in terms of consumers’ unsecured credit, including papers as

Chatterjee, Corbae and Nakajima (2007) or Chatterjee, Corbae and Rios-Rull (2008) and interms of trade credit (unsecured), with papers like Petersen and Rajan (1997), Casamatta(2003) and Cunat (2007). Trade credit is not financial debt, but becomes relevant upondefault.

6The strand of this literature contains Jackson and Kroman (1979), Schwartz (1981), Lev-more (1982), Picker (1982), White (1984), Buckley (1986), Scott (1986), Shupack (1989),Bowers (1991), Triantis (1992), Adler (1993), Barnes (1993), LoPucki (1994) and Mann (1995,1996), among others.

3

Page 4: The (Un)secured Debt Puzzle: Evidence from U.S. Public ......5The literature is extense in terms of consumers’ unsecured credit, including papers as Chatterjee, Corbae and Nakajima

firms do not significantly vary amount of debt from year to year, but instead,adjust its composition. Colla, Ippolito and Li (2012) analyze corporate debtstructure of U.S. public firms and relate usage to demand/supply factors. Theyconclude, as in Graham and Leary (2011), that credit constraints prevent firmsfrom having their preferred debt structure and that optimal capital structurehas to be designed in order to minimize firm’s costs of financing7.

On the other hand, the literature has also emphasized about the empirical rela-tionship between collateral and borrower’s risk. Berger and Udell (1990) provethat not only secured loans are riskier (implying that unsecured loans are safer),but also secured creditors are riskier too8. Therefore, high-risk firms are will-ing to pledge higher collateral and ultimately, borrow more on a secured basis.They called this empirical fact the “sorting-by-observed-risk paradigm”9.Theirstudy is consistent with later results by Carey, Prowse, Rea and Udell (1993) inwhich they analyze the private placement market and conclude that riskier firmsborrow secured debt and stronger firms unsecured debt (with looser covenants).This can be supporting evidence for the “(un)secured debt puzzle”, as largerfirms that are financially strong can be considered as low-risk firms and there-fore, are expected to rely more on unsecured debt.

These studies also reflect that the role of collateral can be secondary for certaindebt contracts, but they do not deepen on the relevant firm characteristics thatmake collateral not as important as the rest of the literature is trying to em-phasize. The cited deficit in the literature is one of the additional goals of thispaper, which is related to both, the “sorting-by-private-information paradigm”by Berger and Udell (1990) and the “(un)secured debt puzzle”.

The present paper is closely related to two papers: i) Berger and Udell (1990)and Giambona and Golec (2012). The papers differ in many aspects. First,Berger and Udell (1990) only consider bank loans in their analysis, while I willbe considering all types of available debt instruments. Second, they considerall commercial and industrial loans in the Federal Reserve’s Survey of Terms ofBank Lending and thus; they are considering both, private and public compa-nies, while I only focus on U.S. public manufacturing companies. Finally, theyconsider the spreads on debt types as a function of loan characteristics, whilethis paper wants to deepen on borrower characteristics.

Similarly, this paper also differs from the Giambona and Golec (2012) paper.

7Previous papers in the literature that recognize debt structure heterogeneity and seek tounderstand the reasons for it are Diamond (1991a, 1993), Park (2000), Bolton and Freixas(2000) and DeMarzo and Fishman (2007) among many others.

8The authors use data from commercial and industrial loans from the Federal Reserve’sSurvey of Terms of Bank Lending, containing information on over one million business loansfor years 1977, 1981, 1983 and 1987.

9As opposed to the “sorting-by-private-information paradigm”, which establishes a nega-tive relationship between collateral and borrower’s risk. Literature validating this view includeBesanko and Thankor (1987a,b), Chan and Kanatas (1985) and Bester (1985).

4

Page 5: The (Un)secured Debt Puzzle: Evidence from U.S. Public ......5The literature is extense in terms of consumers’ unsecured credit, including papers as Chatterjee, Corbae and Nakajima

First, they uniquely link the relevance of unsecured debt to high-growth firms.Second, another important issue they omit is the importance of the success ofinvestment projects in generating cash flows high enough to maintain a high fi-nancial strength despite the increase in leverage. Therefore, high and stable cashflows are the key in order to maintain a debt structure with a higher proportionof unsecured debt. Finally, they make a crucial assumption: all short-term debtis assumed to be unsecured, when clearly this is not the case.

The goal of this paper is to analyze the empirical relationship between 5 en-dogenous firm characteristics: debt and capital structure, collateral, financialstrength and size in order to derive stylized facts about the role of unsecureddebt and how unsecured debt borrowing is determined. Concretely, first, Iwill empirically prove that debt structure is not solely determined by collateralavailability, financial strength plays a key role and I show that the (un)secureddebt puzzle holds. That is, I will shed light on the irrelevance of collateral forspecific debt contracts depending on firm characteristics, provided that 64%of debt in debt structure of U.S. public manufacturing firms is unsecured andunsecured debt does not rely on collateral, but on the financial strength of thefirm. Finally, I will provide one mechanism that allows for the stylized evidencepreviously described: interest rates for unsecured debt are lower than those ofsecured, consistent with findings in Berger and Udell (1990).

My paper is a purely descriptive paper about the relationship between unse-cured debt and the rest of relevant firm characteristics defined, which intendsto provide an intuition about what is behind the relationship between these fiveendogenous variables. Therefore, it is not the aim of the paper to provide atheory about unsecured debt or to imply causality from firm characteristics todebt or capital structure.

The structure of the paper is as follows. Section 2 will describe the sam-ple and will present descriptive summary statistics. Section 3 introduces the“(un)secured debt puzzle” and presents regression results for the determinantsof debt and capital structure and will prove the pecking-order hypothesis forunsecured debt. Section 4, provides descriptive evidence for one possible mech-anism for the strong preference for unsecured: price discrimination in unsecureddebt markets. Finally, section 5 concludes.

2 Data Overview

The key firm characteristics choice is based on the determinants of secured andunsecured debt: secured debt depends on collateral, while unsecured debt de-pends on financial strength. However, the “(un)secured debt puzzle” involvesanother important firm characteristic: size. According to this theory, size alsoplays a predominant role in the determination of debt choice: the larger the size

5

Page 6: The (Un)secured Debt Puzzle: Evidence from U.S. Public ......5The literature is extense in terms of consumers’ unsecured credit, including papers as Chatterjee, Corbae and Nakajima

of the firm, the higher the usage of unsecured debt10.

Appendix A contains the specific definitions and data items in Compustat. Idefine unsecured in debt structure as unsecured debt over total debt11, i) unse-cured and ii) secured in capital structure as i) unsecured and ii) secured debtover total assets respectively. Financial strength is defined as the book-value ofequity over book-value of equity plus total debt12. Following the usual definitionfor tangibility, I define available collateral as property, plant and equipment, netof depreciation, over total assets13. Size is defined as the book-value of totalassets14.

I am also interested on controlling for other firm-level characteristics: market-to-book or Tobin’s q, defined as the ratio between the market-value of assetsover the book-value of assets, which is a proxy for investment opportunities andprofitability, defined as operating profits over total assets of the firm. Both vari-ables have been used in previous studies as determinants of the capital structureof firms15.

2.1 Sample Description

I will concentrate on U.S. firms traded on the AMEX, NASDAQ, and NYSEfrom the manufacturing sector16 covered by S&P’s database Compustat from1994 to 2010. Appendix A contains specific definitions and data items in Com-pustat. Appendix B contains a detailed description of the sample correctionsperformed. My final sample for the manufacturing sector comprises 25,096 firm-year observations.

Table 1 presents summary statistics, including mean, median and standard de-viation for firm-year observations in the sample. Results show that the average(median) firm holds 64% (79%) of unsecured debt in debt structure, while hold-ing 15% (11%) of unsecured debt in capital structure, as opposed to the 8%

10Many papers in the literature have established a relationship between size and the proba-bility of pledging collateral. While Leeth and Scott (1989), Altman, Haldeman and Narayanan(1977), Smith and Warner (1979a) or Chan and Kanatas (1985) find a positive correlationbetween size and the usage of unsecured debt, Jimenez et al. (2006) find the opposite result.

11This definition includes both short-term and long-term debt as opposed to the definitionused in Giambona and Golec (2012), in which the implicit assumption is that all short-termdebt is unsecured.

12Note that the way in which financial strength is defined implied that unsecured debtis determined by some other unsecured source of financing in the capital structure of firms(equity). I want to thank Jose Luis Peydro for pointing this out.

13Evidence from papers using this definition, among many others, can be found in Rauhand Sufi (2010) and Giambona and Golec (2012).

14Evidence from papers using this definition can be found, among many others, in Colla,Ippolito and Li (2012).

15Evidence from papers using these definitions include Rajan and Zingales (1995) or Rauhand Sufi (2010). They find that more profitable and high market-to-book firms use less debt,while firms with higher asset tangibility are more levered.

16SIC codes 2000 - 3999.

6

Page 7: The (Un)secured Debt Puzzle: Evidence from U.S. Public ......5The literature is extense in terms of consumers’ unsecured credit, including papers as Chatterjee, Corbae and Nakajima

(2%) of secured debt holdings over total assets. That is, U.S. public manufac-turing firms show a clear preference for unsecured debt, both in terms of debtand capital structure.

The remaining firm characteristics highlight that the average (median) firm-year observation in the sample has equity in the capital structure of 69% (71%),which denotes a high degree of financial strength, high collateral availabilityequivalent to 26% (23%) and has high investment opportunities 1.86 (1.42). Fi-nally, despite the remarkable heterogeneity across observations, summary statis-tics show large and profitable firms that undertake large investment projects onaverage.

Figure 1 presents the time series evidence on U.S. public manufacturing firms’usage of secured and unsecured debt, both in terms of debt structure and interms of capital structure. The top panel shows the evolution of unsecureddebt in debt structure, while the bottom panel presents a comparison betweensecured and unsecured debt in capital structure. Debt structure exhibits well-defined cyclical properties; it increases during recessions (countercyclical), whilethe capital structure graph shows that both secured and unsecured debt havefollowed a downward trend since the late 90’s, but they peaked again at thebeginning of the 2007 recession17.

The first relevant step for this descriptive section is to see if firms with dis-tinct firm characteristics choose differentiated debt structures. Table 2 presentssummary statistics for firm characteristics of different debt structure definitions,including specialized and mixed debt structures.

Columns 1 and 2 show the summary statistics for those firm-year observationsthat specialize in terms of one type of debt, 100% secured and 100% unsecuredrespectively18. Columns 3-6 contain the summary statistics for those firm-yearobservations with mixed debt structures. Only 27% of the sample specializesin terms of one type of debt, from which only 13.18% choose to specialize insecured debt. In terms of mixed debt structures, the highest concentration islocated in the interval in which firms hold more than 75% in unsecured debtbut less than 100%. The results evidence that 52% of the firm-year observationshave more than 75% of their debt structure in unsecured debt.

From the analysis of firms characteristics of specialized debt structures, we canconclude that firms relying 100% in secured debt are on average less levered(18.6% vs. 22.2%), have a higher financial strength (75.4% vs. 69.1%), are

17This is the effect of a reduced proportion of equity in the capital structure because of thelosses generated in the majority of the manufacturing firms in the 2007-2009, not due to anincrease in leverage.

18Evidence in Colla, Ippolito and Li (2012) suggests that only financially constrained firmstend to specialize in one type of debt. However, they use a different definition for debtstructure heterogeneity.

7

Page 8: The (Un)secured Debt Puzzle: Evidence from U.S. Public ......5The literature is extense in terms of consumers’ unsecured credit, including papers as Chatterjee, Corbae and Nakajima

much smaller in size (189.4 vs. 2166.1) and they are less profitable. Perhapsthis could be evidence for firms relying 100% in secured debt to be financiallyconstrained.

When focusing on mixed debt structures, as firms incorporate more unsecureddebt in their debt structure they become larger and more profitable. However,the maximum average and median financial strength is found in firm-year ob-servations with more than 25% and less than 50% in unsecured debt. This issurprising, as if financial strength determines unsecured debt, we should expectto find the highest financial strength in the last column, more than 75% inunsecured debt. Finally, the highest average and median collateral availabilityis found in firms with less than 25% in unsecured debt. This is the expectedcorrelation, as firms with high secured debt holdings should have, on average,higher collateral requirements.

With the purpose of deepening on the effect of collateral availability and finan-cial strength on debt structure and capital structure choice, Table 3 presentssummary statistics classified according to categories of tangibility and financialstrength: panel a shows summary statistics across the quartiles of the finan-cial strength distribution, while panel b shows the same information across thequartiles of the tangibility distribution.

The most relevant conclusion from panel a is that the data highlights the exis-tence of nonlinearities in the relationship between the percentage of unsecureddebt in debt structure and financial strength: the maximum average (median)unsecured debt holdings in debt structure are achieved in the second quartile ofthe distribution at 71% (93%), but there is a sudden drop when the third andfourth quartiles are considered, down to 56% (55%). Figure 2 graphs this re-lationship, including a local polynomial approximation to reflect the confidenceinterval around the mean.

Additionally, as financial strength increases (leverage tends to zero), averageand median collateral availability, size and profitability decrease sharply. Per-haps this could be indicating that firms in the fourth quartile of the financialstrength distribution are likely to be financially constrained because of the lowaverage (median) collateral availability, 18% (15%).

The rationale for this is straightforward: firms with a high proportion of eq-uity in their capital structure can be either firms with a remarkable balancesheet quality (unconstrained)19 or, have a higher percentage of equity in thecapital structure because they have restricted access to capital markets (con-strained). Appendix C uses the definitions for being financially constrained inAlmeida, Campello and Weisbach (2004) to validate the existence of both finan-cially constrained and unconstrained firms in the upper bound of the financial

19Consistent with the evidence reported in Rauh and Sufi (2010).

8

Page 9: The (Un)secured Debt Puzzle: Evidence from U.S. Public ......5The literature is extense in terms of consumers’ unsecured credit, including papers as Chatterjee, Corbae and Nakajima

strength distribution20.

In terms of panel b, the most relevant conclusion is that financial strengthand collateral availability seem to be to some extent substitutes. As collat-eral availability increases, leverage, size and profitability increase sharply. Justthe opposite of what is been found in the case of financial strength, suggestingthat collateral and financial strength could be substitutes.

Moreover, the summary statistics prove the balance sheet channel mechanism:as collateral availability increases, the degree of financial frictions faced by firmsdecreases and this increases borrowing capacity. However, it increases borrow-ing capacity of both secured and unsecured debt. Therefore, summary statisticsseem to rule out a different mechanism for unsecured debt that could affectinvestment decisions of firms.

The descriptive analysis of the sample performed so far seems to yield the follow-ing relevant conclusions: i) unsecured debt is quantitative more relevant thansecured in terms of debt and capital structure, ii) the majority of the firms showa strong preference towards debt and capital structures with a high proportionof unsecured debt over secured, iii) collateral availability and financial strengthappear to be substitutes and therefore, debt choice is determined by the inter-action between the two determinants and iv) the descriptive evidence seems tovalidate the balance sheet channel: higher collateral implies higher borrowingcapacity.

3 Results: The (Un)secured Debt Puzzle

3.1 Determinants of Debt Structure Choice

This section aims to deepen on the determinants of debt structure by meansof regression analysis in order to achieve robustness in terms of the conclu-sions derived in the descriptive analysis previously performed. The empiricalspecification is defined as follows:

Unsecuredi,tTotalDebti,t

= α+θi+φt+γF inStrengthi,t+δCollaterali,t+X ′i,tβ+ϵi,t (1)

where i denotes a firm, t denotes a year, and α a constant. My focus is on theimportance and robustness of the estimates of financial strength, FinStrengthi,t

and collateral Collaterali,t. The regressions also contain a set of control vari-ables X ′

i,t, including the log of size, market-to-book and profitability. The em-pirical specification is estimated using simple OLS for the sample of manufac-turing firms over the period 1994-2010. All the specifications are estimated

20The definitions include dividend payouts, size, S&P Bond Rating and the Ka-plan&Zingales Index (1997).

9

Page 10: The (Un)secured Debt Puzzle: Evidence from U.S. Public ......5The literature is extense in terms of consumers’ unsecured credit, including papers as Chatterjee, Corbae and Nakajima

using firm-fixed effects, θi, to control for possible simultaneity biases from un-observed individual heterogeneity and year-fixed effects, φt, and finally includ-ing heterokedasticity-consistent errors clustered at a firm level, as in Petersen(2009).

The hypothesis tested is γ > 0 and δ < 0. The OLS specification may notdo a good job in capturing nonlinearities present in the data because as Fig-ure 2 showed, the percentage of unsecured in debt structure of firms exhibits anonlinear relationship with respect to financial strength and there is a suddenreduction in unsecured debt holdings for firms with financial strength above71% (the sample median). Table 4 reports the results for the determinants ofdebt structure for two different samples. The first sample gathers firm-yearobservations with financial strength below the median (71% ), columns (1)-(4),while the second sample captures those above, columns (5)-(8)21.

The estimate of 0.1750 on financial strength in column 4 (first sample) exhibitsthe appropriate sign: financial strength is a determinant for unsecured debtholdings. Firms appear to adjust their debt structure towards more unsecureddebt in response to positive changes in financial strength. Moreover, it is bothstatistically and economically significant.

On the other hand, the coefficient on available collateral, -0.2753, suggests thata 1% increase in tangibility, generates a decrease in unsecured debt equal to0.2753%. That is, firms appear to adjust their debt structure towards moresecured debt in response to positive changes in collateral availability. Thesefindings are consistent with the convention regarding the role of collateral: thehigher the available collateral, the higher the secured debt holdings.

The comparison of estimated coefficients from samples 1 and 2, denotes thatsample 2 has a negative coefficient for financial strength. The interpretationfor this sign is that as firms incorporate a higher percentage of equity in thecapital structure, it evidences a higher degree of financial constraints faced andtherefore, they will exhibit a negative and very sensitive (-0.4070) reaction toincreases in financial strength.

Moreover, it is worth mentioning that the sensitivity of constrained with re-spect to increases in collateral is lower that than of unconstrained (-0.2144 vs.-0.2753), again highlights the restricted access of constrained firms to capitalmarkets: they tend to adjust debt structure less towards more secured whencollateral availability increases. Finally, note that as firms get larger in size,

21Appendix D includes the regression results for the complete sample. The estimate of-0.1483 on financial strength suggests that a 1% increase in financial strength, generates areduction in unsecured debt equal to 0.1483% and firms appear to adjust their debt structuretowards more secured debt in response to positive changes in financial strength. These findingsare inconsistent with the hypothesis that unsecured debt is determined by financial strengthor the quality of the balance sheet of the firm.

10

Page 11: The (Un)secured Debt Puzzle: Evidence from U.S. Public ......5The literature is extense in terms of consumers’ unsecured credit, including papers as Chatterjee, Corbae and Nakajima

they tend to incorporate higher unsecured debt in their debt structure, consis-tent with the (un)secured debt puzzle.

Controlling for nonlinearities in terms of collateral availability, in addition tothe existing ones in terms of financial strength, seems coherent in order to knowif financial strength and collateral availability jointly determine debt structurechoice. Table 5 examines the relationship of financial strength and available col-lateral with mean and median unsecured debt holdings by means of a two-waysorting procedure based on the quartiles of the financial strength and tangibilitydistributions22.

Two important conclusions can be derived from the analysis in Table 5. First,those firms with the highest unsecured debt holdings in their debt structure arelocated in the second quartile of the financial strength distribution (0.58-0.71)and the tendency towards high proportions of unsecured debt is independent ofcollateral availability.

Second, those firm-year observations with the lowest proportion of unsecureddebt are located in the fourth quartile of the financial strength distribution(where the financially constrained firms are located) and in the first quartile ofthe collateral availability distribution, with an average 23% in unsecured debt.As collateral availability increases, surprisingly, firms rather prefer to incor-porate more unsecured debt in their debt structure. However, if the medianholdings are considered for financially constrained firms, they exhibit no un-secured debt holdings: the median financially constrained firm does not haveaccess to unsecured debt markets, independent of the collateral holdings.

Summing up, the most relevant conclusions from the debt structure deter-minants regression are first, collateral availability and financial strength seemsubstitutes. Moreover, for firms that are unconstrained, collateral is irrele-vant. However, other things being equal, debt structure seems more sensitive tochanges in collateral than to changes in financial strength. Second, firms thatare financially constrained, on average, have access to unsecured debt and ascollateral availability increases, they prefer to incorporate more unsecured debt.However, when the median is considered, firms specialize in 100% secured debtas they have no access to unsecured debt.

3.2 Determinants of Capital Structure Choice

This section aims to deepen on the determinants of capital structure by meansof regression analysis in order to achieve robustness in terms of the conclu-sions derived in the descriptive analysis previously performed. The empirical

22Note that firms along the fourth quartile of financial strength gather firm-year observa-tions proved to be financially constrained according to the definitions provided in Almeida,Campello and Weisbach (2004).

11

Page 12: The (Un)secured Debt Puzzle: Evidence from U.S. Public ......5The literature is extense in terms of consumers’ unsecured credit, including papers as Chatterjee, Corbae and Nakajima

specifications are defined as follows:

DebtTypei,tTotalAssetsi,t

= α+θi+φt+γF inStrengthi,t+δCollaterali,t+X ′i,tβ+ϵi,t (2)

where i denotes a firm, t denotes a year, and α a constant. DebtTypei,tin the dependent variable can be either total debt, secured debt or unsecureddebt. My focus is on the importance and robustness of the estimates of financialstrength, FinStrengthi,t and collateral Collaterali,t

23.

The hypothesis being tested is δsec > 0 in the secured debt over total assetsregression and δunsec < 0 in the unsecured debt over total assets regression.This would provide sufficient evidence for the existence of an additional chan-nel, complementary to the balance sheet channel: the (un)secured channel24.Table 6 reports the estimation results for the determinants of the capital struc-ture regression. Column 1 reports the estimated coefficients for total debt overtotal assets as the dependent variable, while columns (2)-(3) and (4)-(5) reportthose for secured over total assets and unsecured over total assets respectively.

By looking at the estimated coefficients for financial strength in columns (1)-(5),we can derive a straightforward conclusion: incorporating more equity in thecapital structure of the firm, reduces debt holdings for both secured and unse-cured, however, unsecured debt is more sensitive than secured to the increasesin financial strength. A 1% increase in financial strength, reduces secured debtby 0.23% (column 3), while unsecured debt reacts more negatively than double,-0.48% (column 5). This is consistent with the evidence reported in Rauh andSufi (2010)25.

If we focus on the results for the secured debt over total assets regressions,column (3), we see that the results for the capital structure still validate theconvention regarding the role of collateral: more collateral availability also in-creases secured debt holdings in the capital structure of the firm. Concretely,a 1% increase in collateral availability, increases secured debt holdings in thecapital structure by 0.084%.

On the other hand, the results for the unsecured debt over total assets regression,column 9, shows that higher collateral availability does not contribute to moreunsecured debt holdings once we control for unobserved variation at a firm levelusing firm-fixed effects. Firms appear to adjust their capital structure towards

23The regressions maintains the same set of controls, X′i,t, as in the debt structure regres-

sions: log of size, market-to-book and profitability. As well as firm-fixed effects, year-fixedeffects and clustered standard errors at a firm level (as in Petersen (2009).

24Note that γ < 0 for any type of debt being considered. That is, if more equity is incorpo-rated in the capital structure, necessarily the amount of leverage should be reduced.

25Firms with more unsecured debt in their debt structure tend to have more equity intheir capital structure, and therefore, a further increase in equity will reduce more thanproportionally the holdings of unsecured debt.

12

Page 13: The (Un)secured Debt Puzzle: Evidence from U.S. Public ......5The literature is extense in terms of consumers’ unsecured credit, including papers as Chatterjee, Corbae and Nakajima

less unsecured debt as their collateral availability increases26. This result is veryinteresting from a balance sheet channel perspective. That is, more collateralavailability decreases the degree of financial frictions faced and this increasessecured borrowing capacity of firms, but not unsecured borrowing capacity.

The above result would suggest the existence of a different mechanism, in addi-tion to the conventional collateral channel, which would operate through unse-cured debt and could generate very different dynamics in terms of investment:the unsecured channel.

The last hypothesis that aims to be tested is whether we can confirm the exis-tence of a pecking-order in terms of unsecured debt as suggested by Giambonaand Golec (2012). The debt structure regression results showed that collateralseems to have a greater impact, in absolute value and other things being equal,over unsecured debt than financial strength. Therefore, I would like to test howthe sensitivity of each type of debt changes as we vary the opposite determinant.

I define financial strength and collateral availability categories in 20% inter-vals27 and I perform the secured debt over total assets regression across thefinancial strength categories and the unsecured debt over total assets regres-sion across the collateral availability categories. The empirical specificationsare displayed below:

Unsecuredj,i,tTotalAssetsj,i,t

= Ωj+γj,1FinStrengthj,i,t+δj,1Collateralj,i,t+X ′j,i,tβj,1+ϵj,i,t

(3)Securedk,i,t

TotalAssetsk,i,t= Ωk+γk,1FinStrengthk,i,t+δk,1Collateralk,i,t+X ′

k,i,tβk,1+ϵk,i,t

(4)where j are the collateral categories, k are the financial strength categories andΩj , Ωk, gather the constant, the firm-fixed effects term and the year-fixed effectsterm in the two specifications.

The above specifications will allow for two effects: effect1 will be the sensi-tivity of secured debt to collateral across financial strength categories, δk, whileeffect2 will be the sensitivity of unsecured debt to financial strength across col-lateral availability categories, γj . Then, if effect 1 dominates over effect 2, otherthings being equal, the pecking-order hypothesis will be proven. Figure 3 and4 graph the estimated coefficients in both regressions (effects 1 and 2 ) and Ap-pendix E shows all the regression results28.

26Note that the evidence from debt structure suggested that only firms that are financiallyconstrained increase unsecured debt holdings as collateral availability increases.

27We will have 5 different financial strength categories, while only 4 will be available interms of collateral availability, as there are no observations for firms with tangible assets overtotal assets above 80%.

28The same procedure has been undertaken using categories defined as a function of thequartiles of the distribution of financial strength and collateral availability and the results

13

Page 14: The (Un)secured Debt Puzzle: Evidence from U.S. Public ......5The literature is extense in terms of consumers’ unsecured credit, including papers as Chatterjee, Corbae and Nakajima

As Figure 3 evidences, the contribution of collateral to secured debt decreases aswe move along the financial strength categories. This is consistent with the factthat debt structure is determined by the interaction between financial strengthand collateral availability, and collateral and financial strength act as substi-tutes.

On the other hand, Figure 4 shows that the contribution of financial strengthto unsecured debt decreases as we increase collateral availability. However, notethat the sensitivity of secured debt to changes in financial strength is higherthan the sensitivity of unsecured debt to changes in collateral availability. Thatis, there is a pecking-order for unsecured debt.

To conclude the actual section on the evidence on the (un)secured debt puz-zle, first note that financial strength and collateral availability determined un-secured debt and secured debt respectively. Second, the increase in collateralavailability does not increase borrowing capacity as evidenced by the balancesheet channel once we account for unobserved individual heterogeneity, but onlysecured debt borrowing capacity. Moreover, firms relying on secured debt tendto be more financially constrained, while those firms relying on unsecured debttend to be unconstrained. This implies that debt structure acts as a signalingdevice of borrower’s quality and even firms that are financially constrained havea preference to incorporate more unsecured debt as their collateral availabilityincreases. Finally, financial strength and collateral availability can act as sub-stitutes but given the pecking-order evidence for unsecured, financial strengthprimarily determines debt choice as collateral only plays a role when access tounsecured debt is restricted.

4 The Mechanism: Price Discrimination in theUnsecured Debt Market

It remains a question the mechanism behind the strong preference for unsecureddebt. According to Giambona and Golec (2012), there is a pecking order forunsecured debt because it allows firms to maintain spare collateral capacity thatcan be used for other purposes other than investment (i.e. risk management).

Rampini, Sufi and Viswanathan (2013) argue that collateral is a scarce resourcethat is pledged for risk management and investment purposes. As firms be-come more financially constrained, they will tend to sacrifice risk managementin order to pledge the existing collateral to finance investment. However, theevidence presented throughout the paper indicates that firms try to pledge col-lateral as infrequently as possible.

remain the same.

14

Page 15: The (Un)secured Debt Puzzle: Evidence from U.S. Public ......5The literature is extense in terms of consumers’ unsecured credit, including papers as Chatterjee, Corbae and Nakajima

Moreover, the survey by Graham and Leary (2011) shows that capital struc-ture is designed so as to optimize financing costs, while Graham and Harvey(2001) determine that the most relevant factors affecting debt policy choiceare financial flexibility29, credit ratings30 and interest rates. This last factor isprecisely the mechanism that I will hypothesize it is behind the preference forunsecured: borrowing on an unsecured basis allows to minimize the total costsof financing.

The conventional wisdom regarding interest rates is that secured debt contractsshould have a lower interest rate attached than unsecured debt contracts. Thisshould be the case because the ex-ante risk that unsecured debt contracts havefor financial intermediaries is so high due to the lack of collateral pledged, thatthe interest rate consistent with the risk assumed would be very large. However,in practice, financial intermediaries set the interest rate for unsecured debt suchthat it is competitive.

Berger and Udell (1990) analyze the commercial and industrial loans marketin the U.S.31 and controlling for loan characteristics, as well as macroeconomicconditions, they conclude that when risk is observable secured debt is riskier,evidenced by a higher interest rate premium than unsecured debt contracts.While Berger and Udell (1990) control for loan characteristics, my interest ison interest rate and borrower characteristics at origination of the loans, to un-derstand if there is descriptive evidence that can validate the hypothesis thatunsecured debt contracts have a lower interest rate attached as a function offirm characteristics.

The information on the interest rates of loans comes from LPC’s Dealscan32, adatabase of loans to large firms. The data in Dealscan comes primarily fromSEC filings and includes most loans made to large publicly traded companies(e.g. the Forbes 500) but there is very little information, however, on lending tosmall and middle-market firms. This is a drawback in order to analyze interestrates for both types of debt contracts as we are not able to cover the completeCompustat manufacturing sample considered in the previous analysis.

Nevertheless, there is a reason why this should not be a problem and we couldstill derive consistent relationships between interest rates attached to debt con-tracts and the associated firm characteristics. Dealscan contains loan informa-tion from the largest public firms in the U.S., which are most likely uncon-strained. Therefore, there is no reason to believe that the interest rates onsecured and unsecured debt for unconstrained firms should be significantly dif-

29Defined as not having enough internal funds to finance investment.30We already know from Rauh and Sufi (2010) that as firms improve their credit quality,

they tend to incorporate more senior unsecured in their debt structure and have a higherproportion of equity in their capital structure.

31Their sample covers each quarter from 1977 to the first half of 1988.32Dealscan hereafter.

15

Page 16: The (Un)secured Debt Puzzle: Evidence from U.S. Public ......5The literature is extense in terms of consumers’ unsecured credit, including papers as Chatterjee, Corbae and Nakajima

ferent.

Table 7 shows the summary statistics, for interest rates and firm character-istics, from all debt contracts signed by U.S. public manufacturing firms duringthe period 1994-2010, classified as secured and unsecured bank debt contracts.Appendix E contains detailed information on how the the sample for debt con-tracts from Dealscan has been constructed.

As expected, the proportion of bank debt contracts signed on a secured basis ismuch higher than that of unsecured, 70.85%33. Surprisingly and in contradic-tion with the intuitive idea that unsecured debt contracts should have a higherinterest rate attached, the mean (standard deviation) basis points in addition tothe reference rates attached to secured debt contracts is 247.26 (129.04), whilethose of unsecured debt contracts is 85.42 (82.932). That is, unsecured debtcontracts do have a lower interest rate attached or secured debt borrowers tendto be riskier34.

Furthermore, when analyzing debt and capital structure at the date of origina-tion of secured and unsecured debt contracts, the average (standard deviation)percentage of unsecured debt in debt structure for secured debt contracts was46% (0.40) as opposed to the 89% (0.25) found for unsecured.

When considering secured and unsecured debt holdings in the capital struc-ture, secured debt contracts exhibit a lower degree of specialization in termsof debt types with 14% (0.15) over total assets in each type of debt. On theother hand, unsecured debt contracts show that these firms tend to specialize interms of unsecured debt, 19.4% (0.13), while holding 1.7% (0.06) in secured debt.

The rest of the firm characteristics analyzed denote that unsecured debt con-tracts tend to be firms with average sizes larger (5,547 vs. 1,326), higher finan-cial strength (0.67 vs. 0.58), higher collateral availability (0.28 vs. 0.25) andlarger investment projects (231.18 vs. 48.34).

All these results are consistent with conclusions previous derived when ana-lyzing debt and capital structure determinants: firms that are larger, with highfinancial strength and high collateral availability tend to have more unsecureddebt than secured, both in the debt structure and in the capital structure.Therefore, the descriptive evidence provided allows to conclude that there is apecking-order for unsecured debt because it allows to minimize total costs offinancing and financial intermediaries are willing to offer lower interest rates forunsecured debt contracts because, ex-ante, firms that borrow on an unsecured

33The evidence in Bolton and Freixas (2000) and Rauh and Sufi (2010) suggests that ascredit quality increases firms tend to substitute bank debt for nonbank debt (i.e. mediumterm notes). This implies that the majority of larger firms borrow on an unsecured basis butthrough medium term notes.

34Consistent with the evidence in Berger and Udell (1990).

16

Page 17: The (Un)secured Debt Puzzle: Evidence from U.S. Public ......5The literature is extense in terms of consumers’ unsecured credit, including papers as Chatterjee, Corbae and Nakajima

basis tend to have a better quality balance sheet and a built-in reputation ofrepayment.

One could definitely argue that unsecured debt contracts tend to have lowerinterest rates attached because of the timing in which they have been origi-nated. That is, unsecured debt contracts tend to be originated at the beginningof expansions when lending standards soften. In addition to his, the low inter-est rates for so long in the 2002-2006 expansion could have motivated the lowerinterest rates for unsecured debt contracts.

In order to prove the above possibility, Figure 5 gathers in panel a the evo-lution for interest rates on unsecured and secured debt contracts (top panel),as well as the evolution of the number of contracts signed35 for unsecured andsecured (bottom panel).

Two conclusions can be derived from the analysis of these graphs. First, inter-est rates on unsecured debt are systematically lower than secured debt interestrates, except for years 1996-2000 and in 2005. In addition to this, secured debtinterest rates tend to be more sensitive to fluctuation in the business cycle, whileunsecured debt interest rates remained relatively stable over time until the 2007recession. Second, focusing on the evolution of debt contracts originated duringthe 2002-2006 expansion, it shows that the number of unsecured debt contractsincreased slightly, while secured debt contracts decreased. however, the evi-dence reported does not imply that unsecured debt spreads were low becauseof timing of issuance: unsecured debt spreads have been low and below securedbeyond the 2002-2006 expansion.

5 Conclusions

The purpose of this paper was to, first, to empirically prove the “(un)secureddebt puzzle” by performing a descriptive analysis of the key firm characteristicsdefined; debt and capital structure, financial strength, collateral and size andto shed light on the possible mechanisms behind the puzzle, by identifying adifferent mechanism that could make unsecured debt more attractive to mostproductive firms. Second, to analyze the role of collateral for U.S. public man-ufacturing firms in order to understand when exactly collateral will be pledgedand the relevance of collateral in defining debt and capital structure. Third, tounderstand the implications of credit quality on shaping debt structure and cap-ital structure of firms. And finally, to test the pecking-order hypothesis and toprovide descriptive evidence on the mechanism behind the puzzle: interest rates.

The motivation for this paper was that, so far, the economics and finance lit-

35These facilities are the mean for new loans and not the mean for loans outstanding atthat point in time, which allows for the comparison of firm characteristics at origination dateat each point in time.

17

Page 18: The (Un)secured Debt Puzzle: Evidence from U.S. Public ......5The literature is extense in terms of consumers’ unsecured credit, including papers as Chatterjee, Corbae and Nakajima

erature has not focused on the role of unsecured debt, while a lot of attentionhas been devoted to the role of collateral in solving market imperfections, im-plicitly focusing on secured debt borrowing, which according to the convention,it is determined by the level of collateral the firm has. This is surprising asunsecured debt is as important in the financing structure of firms as secureddebt, provided that 64% of total debt of U.S. public firms is unsecured, usingS& P’s Compustat database from 1994 to 2010. Oppositely, back since the late70’s, the law literature has emphasized about the relevant role that unsecureddebt plays in the context of firm’s financial and investment policies and, oncreditor’s bargaining process upon default, by introducing the popular conceptof the “secured debt puzzle”.

According to the results derived in the present paper, the role of unsecuredis relevant and there is enough descriptive evidence to justify an unsecuredchannel affecting firm investment, in addition to the collateral channel whichdates as far back as Fisher (1933). First, I show that debt structure is notsolely determined by collateral, but by the interaction between collateral andfinancial strength, which determines unsecured debt. In addition to this, I showthat collateral only plays a role for those firms that are financially constrained.Second, I prove that higher collateral does not increase borrowing capacity bylowering the financial frictions faced, but only secured debt borrowing capacity.This result allows for a complementary channel to the so-called collateral chan-nel, the unsecured channel. Moreover, I test the pecking-order hypothesis andconclude that firms have a clear preference for unsecured debt because it allowsto minimize total costs of financing.

The results regarding the importance of the role of collateral require furthercomments. According to the results derived, on average, collateral is irrelevantacross debt structure and capital structure categories, except perhaps for firmsthat are financially constrained. Moreover, collateral plays no role for firms sat-isfying the “(un)secured debt puzzle”. The second dimension, financial strength,determines when collateral will be pledged as this is only the case when accessto unsecured debt is restricted.

If unsecured debt provides a different mechanism, then: i) all the literatureregarding optimal capital structure should be updated in order to account forthe effects and implications that unsecured debt might have in the optimalcapital structure choice (results show that there is a high positive correlationbetween strong financial strength and a high percentage of unsecured debt). ii)The sectorial composition of the economy would matter, both from a demandperspective and from a supply perspective. iii) The reaction to new informationarrival would be different and this could have important implications for mon-etary policy for instance. Finally, iv) it could affect aggregate investment andbusiness cycle dynamics. Financial frictions literature gives collateral a centralrole in the propagation and amplification of shocks through (a) the balancesheet channel and (b) the bank-lending channel. However, financial accelerator

18

Page 19: The (Un)secured Debt Puzzle: Evidence from U.S. Public ......5The literature is extense in terms of consumers’ unsecured credit, including papers as Chatterjee, Corbae and Nakajima

types of models will be missing some important part of the story, the unsecuredchannel.

The present paper could be improved or extended in many ways. The resultscould be extended to all sectors in the economy36in order to know if sectorialcomposition of the economy matters. Additionally, it would be helpful to definean accurate measure for collateral in the selected sample, from 10ks and 10qs inSEC filings in order to be sure about the results regarding the role of collateral.This would allow not only to have the appropriate measure of collateral, butalso to properly define the variable for encumbered collateral.

On the other hand, it would be interesting to gather debt covenant informa-tion attached to debt contracts from the SEC filings for the selected sample inorder to determine whether covenants of unsecured debt are looser or stricterand to quantify the impact of debt covenants attached to debt contracts oninvestment policy of firms. Moreover, by incorporating interest rate informa-tion to the analysis, we could empirically prove whether price discrimination inthe unsecured debt market takes place and to what extend price discriminationdetermines debt structure of U.S. public firms. Moreover, using regression dis-continuity design, we could empirically prove the unsecured upgrade by definingthresholds for both, collateral availability and financial strength.

The endogeneity problem present among all relevant firm characteristics de-fined also requires attention. Clearly, debt structure, capital structure andinvestment policy of firms co-determine and reversed causality could be presentin any regression analysis performed regarding the defined firm characteristics.For instance, in results estimated for the investment regression, the quantitativeimpact of secured and unsecured debt could be biased. Another econometricapproach is needed in order to disentangle the real importance of financing pol-icy in affecting investment decisions of firms: GMM-IV estimation, so as toeliminate the endogeneity bias present.

The future research possibilities regarding unsecured debt are clearly large asthis is a new strand in the literature. For instance, the analysis in the presentpaper could be also performed for private firms in order to determine whetherthe results for U.S. public manufacturing firms can also be extended to privatefirms. In addition to this, if unsecured debt provides a different mechanism at across-sectional level, one could expect to find a similar pattern in the aggregate.Preliminary research undertaken, shows that unsecured debt is counter-cyclical.That is, it increases during recessions. Studying aggregate dynamics on un-secured debt and its effect over aggregate investment and over the businesscycle would prove useful in identifying an additional mechanism, besides thatof collateral, which could play a role in the propagation and amplification of

36In fact, the “(un)secured debt puzzle” also holds when considering all sectors in the econ-omy.

19

Page 20: The (Un)secured Debt Puzzle: Evidence from U.S. Public ......5The literature is extense in terms of consumers’ unsecured credit, including papers as Chatterjee, Corbae and Nakajima

exogenous shocks to the real economy.

References

Almeida, H., Campello, M. and MS. Weisbach, 2004. ”The Cash-FlowSensitivity of Cash,” Journal of Finance, vol. 59, No. 4, pp.1777-1804.

Almeida, H., and M. Campello, 2007. ”Financial constraints, assettangibility, and corporate investment,” Review of Financial Studies.

Baird, D., and R. Rasmussen. 2006. ”Private Debt and the MissingLever of Corporate Governance,” University of Pennsylvania Law Review154:1209-1252.

Barclay, M. J., and C. W. Smith, Jr., 1995. ”The priority structureof corporate liabilities,” Journal of Finance 50, 899-916.

Berger, A., and G. Udell. 1990. Collateral, Loan Quality, and BankRisk, Journal of Monetary Economics.

Bernanke B., Gertler M., and S. Gilchrist, 1996. ”The Financial Ac-celerator and the Flight to Quality”, The Review of Economics andStatistics, Vol. 78, No. 1. (Feb., 1996), pp. 1-15.

Bolton, P., and X. Freixas, 2000. ”Equity, bonds and bank debt: Capitalstructure and financial market equilibrium under asymmetric information,”Journal of Political Economy 108, 324-351.

Boot, A., and A. Thakor. 1997. ”Financial System Architecture,” Re-view of Financial Studies 10:693-733.

Carey, M., M. Post, and S. A. Sharpe, 1998. ”Does corporate lendingby banks and finance companies differ? Evidence on specialization inprivate debt contracting,” Journal of Finance 53, 845- 878.

Carey, M, Prowse, S., Rea, J., and G. Udell, 1993. The Economics ofPrivate Placements: A New Look, Financial Markets, Institutions andInstruments.

Chava, S. and M.R. Roberts. 2008. ”How Does Financing Impact In-vestment? The Role of Debt Covenant Violations,” Journal of Finance63:2085-2121.

Colla, P., F. Ippolito and K. Li 2012. ”Debt Specialization,” Journalof Finance 68, 1783-1807.

20

Page 21: The (Un)secured Debt Puzzle: Evidence from U.S. Public ......5The literature is extense in terms of consumers’ unsecured credit, including papers as Chatterjee, Corbae and Nakajima

DeMarzo, P. M., and M. J. Fishman, 2007. ”Optimal long-term fi-nancial contracting,” Review of Financial Studies 20, 2079-2128.

Diamond, D. 1991a. ”Debt Maturity and Liquidity Risk,” QuarterlyJournal of Economics 106:709-737.

Diamond, D. 1991b. ”Monitoring and Reputation: The Choice be-tween Bank Loans and Privately Placed Debt,” Journal of PoliticalEconomy 99:689-721.

Diamond, Douglas W., 1993. ”Seniority and maturity of debt con-tracts,” Journal of Financial Economics 33, 341-368.

Giambona, E., and J. Golec 2012. ”The Growth Opportunity Chan-nel of Debt Structure”. Unpublished paper.

Graham, J. R., and M. T. Leary, 2011. ”A review of empirical capi-tal structure research and directions for the future,” Annual Review ofFinancial Economics 3, 1-37.

Hennessy, C. 2004. ”Tobins Q, Debt Overhang, and Investment,” Journalof Finance 59:1717-1742.

Hennessy, C. and T. Whited. 2005. ”Debt Dynamics,” Journal of Fi-nance 60:1129-1165.

Houston, J. and C. James. 1996. ”Bank Information Monopolies andthe Mix of Private and Public Debt Claims,” Journal of Finance 51:1863-1889.

Houston, J. and C. James. 2001. ”Do Relationships have Limits? BankingRelationships, Financial Constraints, and Investment,” Journal of Business74(3):347-374.

Kanatas, G. and D. Besanko. 1993. ”Credit Market Equilibrium withBank Monitoring and Moral Hazard,” Review of Financial Studies 6:213-232.

Kaplan, S. and L. Zingales. 1995. Do Investment-Cash Flow Sensitivi-ties Provide Useful Measures of Financing Constraints? Quarterly Journalof Economics. 115:169-215.

Kiyotaki, N., and J. Moore, 1997 ”Credit Cycles”

LoPucki, L., 2004. ”The Unsecured Creditor’s Bargain,” Virginia Law

21

Page 22: The (Un)secured Debt Puzzle: Evidence from U.S. Public ......5The literature is extense in terms of consumers’ unsecured credit, including papers as Chatterjee, Corbae and Nakajima

Review, Vol. 80, No. 8, Symposium on the Revision of Article 9 of theUniform Commercial Code, pp. 1887-1965.

Mann, R., 2005. ”Explaining the Pattern of Secured Credit”, HarvardLaw Review, Vol. 110, No. 625, pp. 625-83, 1997

Modigliani, F., and M. Miller, 1963. ”Corporate income taxes andthe cost of capital: A correction,” American Economic Review 53, 433-443.

Myers, S. C., 1977. ”The determinants of corporate borrowing,” Journal ofFinancial Economics 5,147- 175.

Nini, G., D. Smith, and A. Sufi, 2009. ”Creditor control rights andfirm investment policy,” Journal of Financial Economics 92, 400-420.

Park, C., 2000. ”Monitoring and the structure of debt contracts,”Journal of Finance 55, 2157-2195.

Rampini, A., Sufi, A., and S. Viswanathan, 2013. ”Dynamic RiskManagement”

Rauh, J. D., and A. Sufi, 2010. ”Capital structure and debt struc-ture,” Review of Financial Studies 23, 4242-4280.

Rajan, R. G., and L. Zingales, 1995. ”What do we know about capi-tal structure: Some evidence from international data,” Journal of Finance50, 1421-1460.

Roberts, M., and A. Sufi, 2009. ”Financial contracting: A survey ofempirical research and future directions,” Annual Review of FinancialEconomics, 207-226.

Roberts, M., and A. Sufi, 2009b. ”Renegotiation of financial contracts:Evidence from private credit agreements,” Journal of Financial Economics93, 159-184.

Roberts, M., and A. Sufi, 2009a. ”Control rights and capital struc-ture: An empirical investigation,” Journal of Finance 64, 1657-1695.

Smith, C.W. and J.B. Warner. 1979. ”On Financial Contracting: AnAnalysis of Bond Covenants”, Journal of Financial Economics 7:117-161.

22

Page 23: The (Un)secured Debt Puzzle: Evidence from U.S. Public ......5The literature is extense in terms of consumers’ unsecured credit, including papers as Chatterjee, Corbae and Nakajima

6 Appendix

6.1 Appendix A: Variable Description

• Percentage of Debt Unsecured in Debt Structure: Total Debt mi-nus Mortgages and Other Secured Debt (item 9 - item 241) over TotalDebt (item 9).

• Percentage of Debt Unsecured in Capital Structure: Total Debtminus Mortgages and Other Secured Debt (item 9 - item 241) over TotalAssets (item 6).

• Percentage of Debt Secured in Capital Structure: Mortgages andOther Secured Debt (item 241) over Total Assets (item 6).

• Financial Strength (book value): Equity (item 6 - item 181) overEquity plus Total Debt (item 6 - item 181 + item 9). Equity is computesas Total Assets minus Total Liabilities.

• Tangibility or Collateral Availability: Property, Plant and Equip-ment, Net (item 8) over Total Assets (item 6).

• Size: Total assets (item 6), total assets in million USD.

• Investment or Capital Expenditures: Capital Expenditures (item128).

• Profitability: Operating income before depreciation (13) over Total as-sets (6).

• Market-to-Book: Market Value of Equity plus Total debt plus Pre-ferred stock liquidating value (10) minus Deferred taxes and investmenttax credit (35) over Total assets (6).

6.2 Appendix B: Sample corrections

I start with U.S. firms traded on the AMEX, NASDAQ, and NYSE, and coveredby S&P’s database Compustat, from 1994 to 2010. First, I remove firm-yearobservations whose percentage of debt unsecured is outside the unit circle37 andend up with 89,684 firm-year observations. This adjustment becomes necessarydue to the way in which Compustat classifies debt into short-term and long-termsecured38. I remove utilities (SIC codes 4900-4949), financial firms (SIC codes6000-6999) and public administrations (SIC codes above 9000), to end up with68,561 firm-year observations.

37All firm-year observations outside (0,1) have been ruled out.38Whenever a debt contract does not specify whether secured debt is short-term or long-

term, Compustat assigns debt to long-term debt. Therefore, you find cases in which the levelof long-term secured exceeds the level of total long-term debt.

23

Page 24: The (Un)secured Debt Puzzle: Evidence from U.S. Public ......5The literature is extense in terms of consumers’ unsecured credit, including papers as Chatterjee, Corbae and Nakajima

I further remove i) firm-years with missing, negative or zero values for totalassets (68,543 observations remaining); ii) firm-years with missing, negative orzero common equity as I am only interested on studying observations from firmsthat are not financially distressed39 (61,636 observations remaining); iii) firm-years with missing, negative or zero values for total sales (60,221 observationsremaining); iv) firm-years with missing, negative or zero values for net prop-erty, plant and equipment (60,024 observations remaining) and v) firm-yearswith missing, negative of zero investment (59,124 observations remaining).

Then, I rule out firm-year observations which are involved in substantial M&Aactivity, by eliminating all firm-year observations with the percentage of amountspent on acquisitions over total assets exceeding 15%. This correction eliminates6,231 observations from my sample. Finally, I windsorize all key firm charac-teristics at the 1st and 99th percentiles (52,846 observations remaining).

My final sample for all sectors in the economy comprises 52,846 observations. Ithen merge the resulting sample of the Compustat leveraged firms with CapitalIQ40, which will allow the decompositions of long-term debt secured and un-secured from Compustat into a broader classification of secured and unsecuredlong-term debt by instrument type. I rule out firm-years for which the differ-ence between total debt as reported in Compustat and the sum of debt types asreported in Capital IQ exceeds 10% of total debt (as in Colla, Ippolito and Li(2012)). However, the present study will present results from the manufactur-ing sector, so I keep all firm-year observations belonging to the manufacturingsector (SIC codes 2000-3999). My final sample for the manufacturing sectorcontains 25,096 observations. Firm-level characteristic variables are from Com-pustat, while firm-level debt structure variables are from Capital IQ. AppendixA provides a detailed description of the variables used in the analysis. Table1 presents descriptive statistics for the manufacturing sample for the period1994-2010.

6.3 Appendix C: Definitions for Financially Contrained inAlmeida, Campello and Weisbach (2004)

39This sample correction is also related to the data assignment problem in Compustat citedbefore, regarding short-term and long-term secured debt. When a firm defaults on a secureddebt payment, the contract is automatically reclassified as short-term debt. Therefore, duringrecessions or when firms face financial distress, secured debt “disappears” from the securedlong-term debt item and it is reclassified as short-term debt. This represents an importantsource of bias if we do not rule out firms that are financially distressed.

40Regulation S-X of the Securities Act of 1933 requires firms to detail their long-term debtinstruments. Regulation S-K of the same act requires firms to discuss their liquidity, capitalresources, and operating results. Firms often also provide information on notes payable withina year (Rauh and Sufi (2010)). The SEC mandated electronic submission of SEC filings in1996. Capital IQ has been compiling detailed information on capital structure and debtstructure by going through financial footnotes contained in firms 10K SEC filings since then.However, coverage by Capital IQ is comprehensive only from 2002 onwards.

24

Page 25: The (Un)secured Debt Puzzle: Evidence from U.S. Public ......5The literature is extense in terms of consumers’ unsecured credit, including papers as Chatterjee, Corbae and Nakajima

# Observations % Observations1) Dividend Payout

Financially Constrained 4,731 51%Unconstrained 4,568 49%

2) Total Assets

Financially Constrained 2,897 31%Unconstrained 6,402 69%

3) Kaplan & Zingales Index

Financially Constrained 1,444 16%Unconstrained 7,855 84%

4) Credit Rating

Financially Constrained 8,891 100%Unconstrained 0 0%

6.4 Appendix D: Debt Structure Determinants Regres-sion Results, Complete Manufacturing Sample

% Unsecured over Total Debt(1) (2) (3) (4)

Financial Strength -0.1583** -0.1522** -0.1503** -0.1483**(0.0235) (0.0234) (0.0235) (0.0239)

Tangibility -0.2125** -0.1855** -0.1889** -0.1912**(0.0483) (0.0484) (0.0487) (0.0489)

Log (Size) 0.0361** 0.0355** 0.0364**(0.0086) (0.0087) (0.0088)

Market-to-book -0.0027 -0.0025(0.0029) (0.0029)

Profitability -0.0187(0.0286)

Firm Fixed Effects yes yes yes yesYear Fixed Effects yes yes yes yesClustered St. Errors firm firm firm firm

# Observations 25,096 25,096 25,096 25,096R2 0.639 0.641 0.641 0.641

25

Page 26: The (Un)secured Debt Puzzle: Evidence from U.S. Public ......5The literature is extense in terms of consumers’ unsecured credit, including papers as Chatterjee, Corbae and Nakajima

6.5 Appendix F: LPC Dealscan Sample Construction

The data on Dealscan are organized by ”Deal” and by ”Facility”. A deal definesa contract signed between a borrower and a lender (or lenders) at a particulardate. Each deal is comprised of one or more facilities (debt contracts). Duringthe 1994-2010 period, there were 5,266 facilities on Dealscan. That is, 5,266distinct debt contracts signed by manufacturing firms. Interest rate informa-tion on debt contracts is obtained from variable ”allindrawn”41 in ”CurrentFacility Pricing”’. While Dealscan has very good information on loan contractfeatures, it has very little information about the borrower and therefore, bor-rower attributes (firm characteristics) come from the previous manufacturingCompustat sample. I merge the LPC Dealscan data on debt contract and in-terest rates with Compustat to gather the firm characteristics the borrower hadat the date of origination of the debt contract.

41This variable considers the basis points above reference rate for each debt contract, whichin the majority of the cases happens to be the LiBOR rate.

26

Page 27: The (Un)secured Debt Puzzle: Evidence from U.S. Public ......5The literature is extense in terms of consumers’ unsecured credit, including papers as Chatterjee, Corbae and Nakajima

7 Tables

Table 1: Manufacturing Sample Overview

This table presents mean, median and standard deviation for key firmcharacteristics for U.S. public manufacturing firms (SIC codes 2000-3999) from1994 to 2010. Appendix A provides a detailed description of the variables usedin the analysis, while Appendix B contains a description of sample corrections.

Mean Median St. Dev.

% Unsecured (Total Debt) 0.64 0.79 0.37% Unsecured (Total Assets) 0.15 0.11 0.15% Secured (Total Assets) 0.08 0.02 0.12

Financial Strength 0.69 0.71 0.22Tangibility 0.26 0.23 0.17

Size 1370 160 3610Market-to-book 1.86 1.42 1.36Profitability 0.05 0.11 0.20

Investment (Level) 68.73 6.79 207.19# Observations 25,096

27

Page 28: The (Un)secured Debt Puzzle: Evidence from U.S. Public ......5The literature is extense in terms of consumers’ unsecured credit, including papers as Chatterjee, Corbae and Nakajima

Table 2: Summary Statistics by Debt Structure Category

This table contains mean and [median] of key relevant firm characteris-tics and controls by reliance on debt types for U.S. public manufacturingfirms (SIC codes) from 1994 to 2010. The first two columns contain the 100%secured and 100% unsecured debt structures respectively. For the rest of thecolumns, column [0-25%) for instance, contains firm-year observations whichhave a percentage of debt unsecured higher than zero but lower or equal to25%. Appendix A provides a detailed description of the variables used in theanalysis, while Appendix B includes a description of sample corrections.

Specialized Mixed100% 100% (0, (25%, (50%, (75%,Sec Unsec 25%] 50%] 75%] 100%)(1) (2) (3) (4) (5) (6)

% Unsecured 0.00 1.00 0.1 0.38 0,63 0,93(Total Debt) [0.00] [1.00] [0.09] [0.38] [0.62] [0.95]% Unsecured 0.00 0.22 0.02 0.08 0.14 0.23(Total Assets) [0.00] [0.21] [0.01] [0.06] [0.12] [0.22]

% Secured 0.19 0.00 0.20 0.13 0.08 0.02(Total Assets) [0.14] [0.00] [0.18] [0.10] [0.07] [0.01]

Financial 0.75 0.69 0.70 0.73 0.70 0.66Strength [0.81] [0.70] [0.74] [0.80] [0.73] [0.67]

Tangibility 0.25 0.26 0.27 0.25 0.25 0.26[0.22] [0.23] [0.24] [0.22] [0.22] [0.22]

Size 189 2,166 333 430 644 2,266[70] [547] [81] [59] [81] [351]

Market-to-book 1.57 1.58 1.55 1.79 1.64 1.52[1.10] [1.20] [1.09] [1.23] [1.12] [1.11]

Profitability 0.04 0.09 0.05 0.01 0.03 0.06[0.10] [0.13] [0.10] [0.08] [0.09] [0.11]

# Observations 903 5,947 4,664 3,398 3,136 7,048

28

Page 29: The (Un)secured Debt Puzzle: Evidence from U.S. Public ......5The literature is extense in terms of consumers’ unsecured credit, including papers as Chatterjee, Corbae and Nakajima

Table 3: Summary Statistics by Debt Choice Determinants

This table contains mean and [median] of key relevant firm characteris-tics and controls by quartiles of the distribution of financial strength (panela) and collateral availability (panel b) for U.S. public manufacturing firms(SIC codes 2000-3999) from 1994 to 2010. Appendix A provides a detaileddescription of the variables used in the analysis, while Appendix B includes adescription of sample corrections.

Panel A: Quartiles of Financial StrengthQ1 Q2 Q3 Q4

% Unsecured (Total Debt) 0.65 0.71 0.65 0.56[0.80] [0.93] [0.84] [0.55]

% Unsecured (Total Assets) 0.28 0.19 0.10 0.02[0.31] [0.22] [0.10] [0.01]

% Secured (Total Assets) 0.15 0.08 0.05 0.02[0.08] [0.02] [0.02] [0.01]

Financial Strength 0.38 0.63 0.80 0.95[0.41] [0.63] [0.80] [0.96]

Tangibility 0.30 0.30 0.27 0.18[0.27] [0.27] [0.23] [0.15]

Size 1,780 1,869 1,218 560[288] [339] [135] [83]

Market-to-book 1.29 1.31 1.58 2.19[0.94] [1.03] [1.20] [1.67]

Profitability 0.06 0.09 0.07 -0.00[0.10] [0.12] [0.12] [0.08]

# Observations 6,261 6,295 6,285 6,255

Panel B: Quartiles of Collateral AvailabilityQ1 Q2 Q3 Q4

% Unsecured (Total Debt) 0.65 0.65 0.64 0.64[0.77] [0.81] [0.79] [0.78]

% Unsecured (Total Assets) 0.12 0.14 0.15 0.17[0.05] [0.10] [0.12] [0.16]

% Secured (Total Assets) 0.05 0.06 0.08 0.10[0.01] [0.02] [0.02] [0.04]

Financial Strength 0.76 0.71 0.67 0.63[0.83] [0.74] [0.69] [0.64]

Tangibility 0.07 0.17 0.29 0.51[0.07] [0.17] [0.29] [0.48]

Size 895 1,299 1,477 1,758[78] [146] [201] [315]

Market-to-book 1.99 1.68 1.47 1.22[1.45] [1.22] [1.09] [0.94]

Profitability -0.03 0.05 0.09 0.10[0.05] [0.11] [0.12] [0.13]

# Observations 6,259 6,296 6,288 6,253

29

Page 30: The (Un)secured Debt Puzzle: Evidence from U.S. Public ......5The literature is extense in terms of consumers’ unsecured credit, including papers as Chatterjee, Corbae and Nakajima

Table 4: Debt Structure Determinants Regression Results, TwoSamples

This table presents regression results to examine the relation betweendebt structure, financial strength and collateral availability as determinants,along with usual controls in the literature for U.S. public manufacturing firms(SIC codes 2000-3999) from 1994 to 2010. Columns (1)-(4) show the results forsample1, for financial strength below the sample median(71%), while Columns(5)-(8) show the results for sample2, financial strength above the samplemedian(71%). Definitions of the variables are provided in Appendix A, whilesample corrections are described in Appendix B. All specifications includefirm- and year-fixed effects and robust standard errors are clustered at the firmlevel (as in Pertersen (2009)) and reported in parentheses. **, and * denotestatistical significance at the 5% and 1% level, respectively.

% Unsecured over Total DebtFinancial Strength ≤ 71% Financial Strength > 71%

(1) (2) (3) (4) (5) (6) (7) (8)

Financial 0.194** 0.146** 0.150** 0.175** -0.560** -0.401** -0.405** -0.407**Strength (0.033) (0.030) (0.029) (0.030) (0.059) (0.057) (0.057) (0.057)Tangibility -0.282** -0.326** -0.293** -0.275** -0.116** -0.225** -0.224** -0.214**

(0.042) (0.036) (0.036) (0.036) (0.045) (0.040) (0.040) (0.041)

Log (Size) 0.072** 0.074** 0.080** 0.068** 0.068** 0.070**(0.002) (0.002) (0.002) (0.002) (0.002) (0.002)

Market-to 0.006 0.007 -0.005 -0.005-book (0.005) (0.005) (0.003) (0.003)Profitability -0.095* 0.015

(0.057) (0.038)

Firm FE yes yes yes yes yes yes yes yesYear FE yes yes yes yes yes yes yes yesClustered SE firm firm firm firm firm firm firm firm

# Obs. 12548 12548 12548 12548 12548 12548 12548 12548R2 0.74 0.74 0.74 0.74 0.66 0.66 0.66 0.66

30

Page 31: The (Un)secured Debt Puzzle: Evidence from U.S. Public ......5The literature is extense in terms of consumers’ unsecured credit, including papers as Chatterjee, Corbae and Nakajima

Table 5: Two-way sorting of Percentage of Debt Unsecured, byFinancial Strength and Tangibility

This table presents the relation between the percentage of debt unsecured,financial strength and tangibility for U.S. public manufacturing firms from1994 to 2010. Two-way sorting is carried out year by year and then aggregatedacross years. Each cell in the table presents mean [median] percentage ofdebt unsecured in debt structure. Definitions of the variables are provided inAppendix A, while sample corrections are described in Appendix B.

Tangibility

Financial Strength Q1 Q2 Q3 Q4

Q1 0.65 0.67 0.69 0.61[0.91] [0.88] [0.90] [0.75]

Q2 0.71 0.74 0.69 0.70[0.99] [0.98] [0.95] [0.94]

Q3 0.55 0.51 0.51 0.57[0.77] [0.60] [0.62] [0.76]

Q4 0.23 0.27 0.39 0.42[0.00] [0.00] [0.00] [0.08]

31

Page 32: The (Un)secured Debt Puzzle: Evidence from U.S. Public ......5The literature is extense in terms of consumers’ unsecured credit, including papers as Chatterjee, Corbae and Nakajima

Table 6: Capital Structure Determinants Regression Results

This table presents regression results to examine the relation betweencapital structure, financial strength and collateral availability as determinants,along with usual controls in the literature for U.S. public manufacturing firmsfrom 1994 to 2010. Columns (1) shows the results for the total debt over totalassets regression, (2)-(3) shows the results for the secured debt over total assetsregression and (4)-(5) shows the results for the unsecured debt over total assetsregression. Definitions of the variables are provided in Appendix A, whilesample corrections are described in Appendix B. All specifications includefirm- and year-fixed effects and robust standard errors are clustered at the firmlevel (as in Pertersen (2009)) and reported in parentheses. **, and * denotestatistical significance at the 5% and 1% level, respectively.

% Debt over Total AssetsTotal Secured Unsecured(1) (2) (3) (4) (5)

Financial Strength -0.706** -0.228** -0.231** -0.479** -0.475**(0.005) (0.009) (0.009) (0.010) (0.010)

Tangibility 0.024** 0.087** 0.086** -0.075** -0.061**(0.007) (0.015) (0.015) (0.015) (0.015)

Log (Size) 0.014** -0.005* 0.019**(0.001) (0.003) (0.003)

Market-to-book 0.000 -0.000 0.000(0.000) (0.001) (0.001)

Profitability 0.006 0.018* -0.012(0.004) (0.008) (0.009)

Firm FE yes yes yes yes yesYear FE yes yes yes yes yesClustered SE firm firm firm firm firm

# Obs. 25,096 25,096 25,096 25,096 25,096R2 0.97 0.72 0.72 0.80 0.80

32

Page 33: The (Un)secured Debt Puzzle: Evidence from U.S. Public ......5The literature is extense in terms of consumers’ unsecured credit, including papers as Chatterjee, Corbae and Nakajima

Table 7: Summary Statistics, Interest Rates and Firm Characteristics,for Secured and Unsecured Debt Contracts

This table presents the comparison of spreads over reference rate andfirm characteristics of secured and unsecured debt contracts at date of origina-tion for U.S. public manufacturing firms from 1994 to 2010. Definitions of thevariables are provided in Appendix A, sample corrections for the Compustatsample are described in Appendix B, while Appendix F contains the procedureto derive the LPC Dealscan sample (interest rate spreads).

Secured Bank Debt Unsecured Bank Debt

Mean St. Dev. Mean St. Dev.

Spread 247.25 129.04 85.41 82.93

% Unsecured (Total Debt) .455 .402 .894 .251% Unsecured (Total Assets) .147 .165 .194 .127% Secured (Total Assets) .142 .150 .017 .060

Financial Strength .576 .265 .670 .199Tangibility .251 .164 .281 .163

Size 1,326.2 5,110.1 5,547.5 13,308.0Market-to-bookProfitabilityInvestment (Level) 48.3 192.6 231.1 544.0# Debt Contracts 3,731 1,535

33