Secured Loans 0 The Determinants of Secured Loans John S. Gonas Belmont University Michael J. Highfield Louisiana Tech University Donald J. Mullineaux University of Kentucky Abstract This paper examines the determinants of the secured status of 7,619 commercial loans closed between December 1988, and January 2001. Our main finding, consistent with the argument that collateral helps avoid asymmetric information problems and reduces risk, is that firms with an S&P rating tend to secure loans less often than firms not rated by S&P. In addition, further supporting the notion that banks require collateral in the face of information asymmetry, our results suggest that larger firms, such as those with high sales figures and those listed on a U.S. stock exchange, are less likely to enter a secured loan agreement. Moreover, consistent with a credit risk argument, we find that firms with investment grade S&P senior debt ratings are less likely to secure loans, while those with an S&P rating at or below BB typically offer collateral. We also find that loan size is inversely related to the probability that a loan is secured, but loans with a longer maturity are more likely to be collateralized. Also, the evidence implies that the purpose of the loan and the industry of the borrower are significant in determining the likelihood that a loan is secured. JEL Classification: G20; G21 Keywords: Secured Loans, Collateral, Credit Ratings, Information Asymmetry, Credit Risk We would like to thank Brent Ambrose, Dan Bradley, Mark Carey, Jack Cooney, Marcia Cornett, Dalia Marciukaityte, and seminar participants at the 2002 Midwest Finance Association and 2002 Financial Management Association annual meetings for helpful comments and suggestions. This paper is scheduled for presentation in the Symposium on Financial Institutions at the 2003 Eastern Finance Association annual meeting. This paper has also tentatively been accepted into the Financial Review subject to an expedited review. All errors remain ours. John S. Gonas is an Assistant Professor of Finance at Belmont University. Address: School of Business, 1900 Belmo nt Boulevard, Nashville, Tennessee 37212. Office Phone: 615.460.6907. Office Fax: 615.460.6487. Electronic Mail: gonasj@belmont.edu. Contact Author: Michael J. Highfield is an Assistant Professor of Finance at Louisiana Tech University. Address: Department of Economics, Finance, and Quantitative Analysis, College of Administration and Business, Louisiana Tech University, Post Office Box 10318, Ruston, Louisiana 71272. Office Phone: 318.257.2112. Office Fax: 318.257.4253. Electronic Mail: mikehigh@cab.latech.edu. Donald J. Mullineaux is the dupont Endowed Chair in Banking and Director of the School of Management at the University of Kentucky. Address: Finance Area, Gatton College of Business and Economics, 445 Business & Economics Building, Lexington, Kentucky 40506. Office Phone: 859.257.2890. Office Fax: 859.257.9688. Electronic Mail: mullinea@uky.edu.
Secured Loans 1 The Determinants of Secured Loans Abstract This paper examines the determinants of the secured status of 7,619 commercial loans closed between December 1988, and January 2001. Our main finding, consistent with the argument that collateral helps avoid asymmetric information problems and reduces risk, is that firms with an S&P rating tend to secure loans less often than firms not rated by S&P. In addition, further supporting the notion that banks require collateral in the face of information asymmetry, our results suggest that larger firms, such as those with high sales figures and those listed on a U.S. stock exchange, are less likely to enter a secured loan agreement. Moreover, consistent with a credit risk argument, we find that firms with investment grade S&P senior debt ratings are less likely to secure loans, while those with an S&P rating at or below BB typically offer collateral. We also find that loan size is inversely related to the probability that a loan is secured, but loans with a longer maturity are more likely to be collateralized. Also, the evidence implies that the purpose of the loan and the industry of the borrower are significant in determining the likelihood that a loan is secured. JEL Classification: G20; G21 Keywords: Secured Loans, Collateral, Credit Ratings, Information Asymmetry, Credit Risk
The Determinants of Secured Loans Secured Loans 2 1. Introduction In this paper we examine the relationships between certain firm characteristics and the decision to pledge collateral in a commercial loan agreement. We investigate and expand upon past theory and empirical evidence that ties collateralization to a reduction in: the risk assumed by the lender, information asymmetry problems between the borrower and lender, and the potential for moral hazard problems. Much of this past research has focused on the association between collateral and the borrower s cost of capital, and uses loan rates as a proxy for the intrinsic risk of the borrower. We assume that the decision to secure a loan is determined mutually with other terms of the loan. Our research is primarily motivated by the results of John, Lynch, & Puri (2000). They assume that credit ratings are reliable measures of borrower default risk, and they find that bond issuers are more likely to secure lower-rated public debt. In particular, we first look at whether a borrower has a credit rating or not. We assume that more information is available about firms that have been analyzed by rating agencies. Indeed, the finance literature treats information production as the primary activity of the rating agencies. Next, we examine the borrower s specific senior debt rating from Standard and Poor s (S&P) to determine if ratings have a significant relationship to the decision to secure loans. Controlling for additional firm and loan characteristics, we find that carrying a rating, as well as the specific rating itself, are significantly related to the choice to pledge collateral. Not only are rated firms less likely to pledge collateral than non-rated borrowers, but investment grade firms (S&P BBB or above) are also less
Secured Loans 3 likely to secure loans than non-investment grade firms. These findings are consistent with the notion that credit rating agencies may supplement, or even substitute for, costly bank monitoring. 1 2. Competing Theories and Evidence Three primary rationalizations have been provided for why some bank loans are backed by pledged collateral: (1) signaling, or overcoming information asymmetries and adverse selection problems, (2) managing credit risk, and (3) reducing moral hazard problems. Leeth and Scott (1989) outline several other costs and benefits of collateral to both the borrowing and lending firm: reducing the lender s monitoring and administrative costs, reducing both the borrower s cost of preparing additional reports and the costs associated with more restrictive asset usage, reducing conflicts of interest between unsecured and secured claimants, and limiting the possible dilution of legal claims on a borrowing firm s assets. In the existing literature, there are many different theoretical models and empirical findings that either confirm or weaken the arguments for these rationalizations. 2.1. Adverse Selection and Information Asymmetry The signaling/information asymmetry hypothesis suggests that less risky, high quality firms are more likely to secure loans. If a high quality firm is choosing between a secured and unsecured loan, it may prefer to pledge collateral for two reasons: (1) It will lower its cost of capital since only high quality firms can use collateral to signal reduced risk of default; (2) It will have a lower probability of losing the collateral and thus pledging an asset is accordingly less burdensome. The theoretical models of Townsend (1979), Bester (1985), Besanko and Thakor (1987), and Chan and Kanatas (1987) predict that collateral will be associated with safer, higher quality borrowers. They find that securing a loan is particularly beneficial when it is difficult for a lender to collect pre- 1 Moody s ratings yielded comparable results to Standard & Poors in our estimations.
Secured Loans 4 contract private information regarding the quality of a borrower and its projects. Chan and Kanatas (1985) also argue that securing debt enables a high quality firm to signal its creditworthiness and thus lower its loan rate. The benefits from a lower cost of capital must outweigh the costs associated with pledging collateral; however, Igawa and Kanatas (1990) model how pledging collateral allows a high quality firm to optimize the net benefits gained from securing a loan by overcollateralizing, (the value of pledged collateral exceeds the value of the loan) while simultaneously underinvesting in the maintenance and care of such collateral. Their model implies that the use of collateral can be a sufficient signal to overcome moral hazard problems associated with potentially poor collateral maintenance. Finally, Boot, Thakor and Udell (1991) show how the use of collateral increase in the presence of private information, reinforcing the notion that riskier borrowers pledge more collateral. Additionally, the authors also find that endogenous variables, such as higher interest rate levels, are associated with higher equilibrium collateral requirements for both risky and high quality firms. 2.2. Credit Risk Collateral protects the lender against loss by granting title to specific assets in the event of default. This argument is fundamental in much of the theoretical and empirical research that examines the relationship between firm characteristics and the decision to secure loans. For example, Scott (1977) asserts that because secured claims have priority over unsecured creditors, secured debt may limit the possibility of complete losses in the event of bankruptcy. He demonstrates that the value of a firm with secured debt increases as the possibility of default increases, supporting the argument that the benefits of collateral outweigh the costs for risky firms. There are a number of other theoretical studies that demonstrate that riskier firms are more likely to pledge collateral (Swary and Udell (1988); Boot, Thakor, and Udell (1991); Black and demeza (1992)).
Secured Loans 5 The empirical research on risk reduction shows that, when banks assess the pre-loan risks of prospective borrowers in their credit analysis, firms that are observably riskier are more likely to pledge collateral (Morsman (1986), and Hempel, Coleman, and Simonson (1986)). There are several empirical studies that address firm risk and the likelihood of pledging collateral. Orgler (1970) compiled individual loan data categorizing each as either good or bad based on the opinions of bank examiners. He found a significant positive relationship between the use of collateral and loans that were categorized bad. Hester (1979) regressed a dummy variable, secured/unsecured, on six accounting variables that were proxies for firm risk. He also found that riskier firms were more likely to pledge collateral. Leeth and Scott (1989) found that more collateral is pledged with loans to riskier, small businesses. Lastly, Berger and Udell (1990) found that riskier firms are more likely to borrow on a secured basis, and that the average secured loan in their sample was riskier than the average unsecured loan. 2.3. Moral Hazard The moral hazard hypothesis yields implications similar to the risk reduction hypothesis. Assuming that a lender seeks to reduce moral hazard problems, a secured loan lowers the probability that the borrowing firm will engage in underinvestment, asset substitution, or an inadequate supply of effort. In a model of underinvestment, Myers (1977) demonstrates how the use of collateral eliminates underinvestment in profitable projects and thus reduces the probability of bankruptcy. Stulz and Johnson (1985) show that secured debt enhances firm value because it reduces the incentive to underinvest that arises when a firm uses equity or unsecured debt. Focusing on asset usage and managerial effort, Smith and Warner (1979) predict that collateral can prevent a borrower from consuming a loan or engaging in asset substitution.
Secured Loans 6 3. Hypotheses and Model Specifications In this section we develop testable hypotheses from the conflicting theories and evidence presented in previous literature, and we develop models to facilitate testing these hypotheses. Although discussions of moral hazard problems are ubiquitous in the literature, empirical tests of moral hazard issues are difficult to implement. At this stage of our research, we focus primarily on hypotheses related to information asymmetry and credit risk. We plan to collect additional data to address moral hazard issues. 3.1. Collateral is More Likely in the Presence of Information Asymmetry Problems Information asymmetry problems can be proxied in a variety of ways. First, we suggest that the existence of a credit rating (by Standard & Poor s or Moody s) will be negatively correlated to the need to pledge collateral. We expect that larger firms are most likely to subject themselves to the rating process, given their need for the sizeable amounts of funds supplied by capital markets. Such firms become more informationally transparent, so they pose fewer adverse selection problems and are more easily monitored. Our hypothesis is that rated firms are less likely to pledge collateral than non-rated ones. We assume that a firm s revenues can proxy for its size, and hypothesize that an adverse selection problem is less likely as firm size grows. We consequently expect an inverse relationship between size and collateralization. Larger loans are likely be associated with larger, more established firms, so we hypothesize that larger loans are less like to be collateralized, even though they create more exposure for the bank. Additionally, a larger deal size is also more likely to be syndicated, so that many lenders must consider and agree to the level of risks and exposure associated with these larger loans. Dennis and Mullineaux (1999), find a significantly negative relationship between secured loans and the prospect of syndication, so we expect a negative relationship between loan syndication and loan collateralization.
Secured Loans 7 Given the listing requirements of stock exchanges, more established firms are more likely to raise public equity. Such firms are also more informationally transparent, given the SEC s reporting requirements for publicly traded firms. Therefore, we expect that public firms will require less monitoring. Our hypothesis is that exchange traded firms, which we capture with an indicator variable, are less likely to pledge collateral. We also assume that lenders have more difficulty monitoring and gathering information regarding firms with headquarters located outside the United States, and we expect such loans to carry more risks associated with information asymmetry and moral hazard problems. Therefore, we hypothesize that foreign firms should be more likely to pledge collateral. Foreign firms are exposed to more general, exogenous forms of risk such as country specific, economic, political, and exchange rate risk. Such exposure enhances information asymmetry problems with regard to the probability of default. This affects the likelihood that these firms have secured loans. 3.2. Collateralization is More Likely in the Presence of Credit Risk If we assume that secured loans induce the need for more monitoring, it should be the case that the costs associated with such monitoring outweigh the benefits. For example, Rajan and Winton (1995) argue that collateralized debt is likely to be used by firms in greater need of monitoring. Although some of the empirical evidence suggests that collateral can be used to signal a borrower s quality (Besanko and Thakor (1987); Boot, Thakor, and Udell (1991)), we expect to find that lesser known, non-rated firms will be more likely to pledge collateral. Borrowers are more likely to pledge collateral when monitoring costs are high. After isolating a sample that includes only firms that are rated, we would expect lower quality ( high yield ) firms to be more likely to pledge collateral. Even though rated firms are more easily monitored and present fewer information asymmetry, agency, and moral hazard problems, we expect that borrowers with a lower rating will be more leveraged and less able to repay existing
Secured Loans 8 credit. We thus anticipate that they are more likely to pledge collateral. Conversely, we would anticipate that higher quality ( investment grade ) firms would find that the costs (e.g. limited asset control) of securing loans would outweigh the benefits. With regard to maturity, Dennis, Nandy and Sharpe (2000) find a significantly positive relationship between the duration of a revolving credit agreement and its secured status. Assuming that credit risk is at work here, the assumption is that credit risk increases with maturity. For example, Stohs and Mauer (1996) find that lower quality firms tend to issue bonds with longermaturities. This implies that collateralization is positively related to the term of the loan contract. 3.3. Control Variables Given that certain projects are inherently riskier than others, we expect some of a borrowing firm s intended purposes to be correlated to its decision to offer collateral. For instance, if a loan s purpose is to undertake an acquisition, we might anticipate a higher degree of perceived risk, prompting a decision to pledge collateral. Conversely, if a loan s purpose is to purchase highly marketable fixed assets, we might expect the opposite effect. Regardless, it is difficult to hypothesize whether a firm will use collateral based on the loan s purpose. If we were to make an argument(s) for one or several purposes we would have to separately test each type of purpose, controlling for many external variables that are beyond the scope of this research. One expectation is that loans used for refinancing purposes carry a higher degree of repayment risk and therefore warrants the use of collateral. However, such loans can also signal a firms growth potential in that it requires more capital. We hypothesize that the higher probability of default outweighs the expectation of future growth and thus the risks associated with a refinanced loan necessitate the use of collateral. This argument is further supported with the likelihood of an information asymmetry problem, assuming that the lender is unable to adequately appraise a borrower s growth potential.
Secured Loans 9 Some industries are more susceptible to certain exogenous, macroeconomic forms of risk, we might expect that certain industries might be better insulated from such risks. As a result, those industries that have a higher degree of exposure to the influences of seasonal or cyclical economic trends are likely to be required to pledge collateral. For example, the electric utilities industry has historically been highly regulated and thus somewhat less affected by economic shocks. However, specific industries may also be more susceptible to endogenous forms of risk such as asymmetric information and moral hazard problems. Continuing with the highly regulated industries, these firms may be more prone to moral hazard-type risks such as under investment because lenders automatically assume that the government is monitoring the firm, even if the regulators are only monitoring safety requirements or production levels. Alternatively, as another example, recognizing that the financial services industry is associated with off-balance-sheet activities, these types of firms pose both asymmetric information and moral hazard problems to their lenders. 3.4. The Logit Model Controlling for S&P Coverage To test the hypotheses outlined above, we will begin by estimating a model that controls only for the presence of information asymmetry problems and their relationship to collateralization. This model is as follows: SECURED i = β 0 + β 1 SPRATED i + β 2 LNSALES i + β 3 LNDEAL i + β 4 SYNDICATE i + β 5 EXCHANGE i + β 6 FOREIGN i + β PURPOSE VARIABLES i + β INDUSTRY VARIABLES i + ε i [1] where a subscript i indicates that the variable refers to the ith loan agreement, and SECURED is a binary variable equal to one for secured loan agreements and zero for unsecured. With regard to the hypotheses presented earlier, we will use SPRATED to test the hypothesis that borrowers with an S&P rating have less information asymmetry; and therefore have a lower probability of offering
Secured Loans 10 collateral on a loan. SPRATED is a binary variable for firms with a senior debt rating by Standard and Poor s. 2 We expect the coefficient on SPRATED to be negative in this model. As discussed above, information about larger firms is less costly to obtain than for small firms. Thus, assuming that firm size is directly related to firm sales, a borrower with larger sales figures should be less likely to secure a loan because information about the firm is more obtainable. Thus, we expect a negative relationship between the natural logarithm of borrower sales, LNSALES, and the probability that a loan is secured. We will use the variable LNDEAL to test the hypothesis that loan collateralization is directly related to the size of the deal. As noted above, based on our priors, we expect that larger loans are less likely to be secured. Thus, we presume that the coefficient on LNDEAL is negative. SYNDICATE, a binary variable for syndicated loans, is used to test the hypothesis that syndicated loans have less information asymmetry between lenders and borrowers, or that collateral may be a signal of risk, so the loan is harder to sell. Thus, this implies a negative coefficient for the SYNDICATE dummy variable. We also use EXCHANGE, a binary variable for firms listed on a U.S. stock exchange, to test for information asymmetry problems because exchange-listed firms are likely to be better known than the average firm. Moreover, exchange listed firms are forced by regulatory requirements to disclose more financial information than a private firm. Firms with traded stock are more informationally transparent. Conversely, a foreign borrower is likely to be more problematic to the borrower in terms of information asymmetries, and thus firms based outside of the United States are more likely to have to collateralize a loan. Therefore, we expect a positive relationship between the variable FOREIGN, a binary variable for firms located outside of the United States, and the probability that a loan is secured. 2 Please note that the reference variable for SPRATED is SPNR, or firms not rated by Standard and Poor s.
Secured Loans 11 The last two sets of variables are collections of binary variables for both the primary purpose of the loaned funds and the industry classification of the borrower. The primary purpose variables will be discussed in more detail in the next section, but broadly speaking, PREF, PCC, PFAB, PGCP, PP, and POTH are binary variables for the purposes of refinancing bank debt, corporate control, fixed-asset backing, general corporate purposes, project finance, and other purposes, respectively. As mentioned above, we expect some of these purposes to be associated with more or less collateral when compared to the reference variable, PCS, capital structure uses. In addition, we will inspect the relationship between collateralization and refinancing through the REFINANCE variable, a binary variable for loans used to refinance other bank debt. As mentioned in the previous section, we believe that refinancing is associated with fewer information asymmetries but more repayment risk. Therefore, we expect a positive relationship between refinancing loans and secured loans. We also use industry variables designated by one-digit SIC codes to control for differences in collateralization across industry groups. The sign and significance of the coefficients on these binary variables for different one-digit SIC codes cannot be determined a priori. However, it is logical to reason that some industries are more risky than others. Therefore, we expect to find that some of these industry groups are more or less risky than typically regulated industries such as transportation and utilities (SIC4), the reference variable. 3 3.5. The Logit Model Controlling for General Credit Quality To test hypothesis that collateralization is more likely in the face of credit risk, we will separate firms rated by S&P into investment grade (AAA, AA, A, and BBB) and high-yield (BB, B, CCC, CC, C, and D). Thus, the reference group of non-rated firms does not change, but SPRATED variable is separated into SPINVEST, a binary variable for investment grade firms, and 3 Regulated industries such as transportation and utilities, designated by the variable SIC4, will serve as the reference variables for the industry variables.
Secured Loans 12 SPHIGHYLD, a binary variable for high-yield firms. The remainder of the model remains unaltered; thus, the logit model controlling for general credit quality is presented in Equation [2] below: SECURED i = β 0 + β 1 SPINVEST i + β 2 SPHIGHYLD i + β 3 LNSALES i +β 4 LNDEAL i + β 5 SYNDICATE i + β 6 EXCHANGE i + β 7 FOREIGN i + β 8 MATURITY i [2] + β PURPOSE VARIABLES i + β INDUSTRY VARIABLES i + ε i A subscript i again indicates that the variable refers to the ith loan agreement, and SECURED is a binary variable for collateralized loans. To examine the relationship between loan maturity and collateral, we will use the variable LNMATURITY, the natural logarithm of the maturity of the loan in months. In accordance with work by Schwartz (1981), Jackson and Kronman (1979), and Stohs and Mauer (1996) we expect to find a positive coefficient on this variable because credit risk increases with maturity. Other than the aforementioned change in the S&P variables and the addition of the maturity variable, we will use the same independent variables used in [1]. 3.6. The Logit Model Controlling for Specific Credit Quality As a robustness check of the previous model, we further disaggregate firms rated by S&P into their individual credit ratings. Thus, we will replace SPINVEST with SPAAA, SPAA, SPA, and SPBBB. Moreover, we will also replace SPHIGHYLD with SPBB, SPB, SPCCC, SPCC, and SPD. 4 Like the second model, the reference group of non-rated firms does not change, and the remainder of the model remains unaltered. Thus, the logit model controlling for specific credit quality is presented in Equation [3] below: SECURED i = β 0 + β 1 SPAAA i + β 2 SPAA i + β 3 SPA i + β 4 SPBBB i + β 5 SPBB i + β 6 SPB i + β 7 SPCCC i + β 8 SPCC i + β 9 SPD i + β 10 LNSALES i + β 11 LNDEAL i + β 12 SYNDICATE i + β 13 EXCHANGE i + β 14 FOREIGN i + β 15 MATURITY i + [3] + β PURPOSE VARIABLES i + β INDUSTRY VARIABLES i + ε i 4 There are no loans in the sample with a borrower rated as C by S&P.
Secured Loans 13 A subscript i again indicates that the variable refers to the ith loan agreement, and SECURED is a binary variable for collateralized loans. Besides the aforementioned change in the S&P variables, we will use the same independent variables used in [2]. 4. Data To apply the models and test the hypotheses presented in the previous section, we collect a sample of 12,685 commercial loans closed between December 1988, and January 2001. The sample was obtained from the Loan Pricing Corporation (LPC) DealScan database. We restrict the sample to loans with complete and confirmed information. After removing loans with missing observations and data not confirmed by LPC, or with obvious data entry errors, the final sample consists of 7,619 loan agreements. In Table 1, we present summary statistics for the variables to be used in the three models. Just over 86 percent of the loans are syndicated, over 73 percent of the loans in the sample are secured, and the average LIBOR spread is about 187 basis points. About 69 percent of the loans involve borrowers listed on a U.S. stock exchange, and approximately 3 percent of the borrowers are based outside the United States. *** Table 1 About Here *** Similar to Kleimeier and Megginson (2000), we have organized the primary purpose of the loan into several large categories, including bank refinancing, corporate control, capital structure, fixed asset backing, and general corporate purposes. We have also added project financing and other financing categories. About 21 percent of the loans were used specifically to refinance debt (PREF), and over 25 percent of the loans were used for corporate control purposes (PCC) such as acquisitions, leveraged buyouts, or employee stock option plans. Similarly, about 48 percent of the loans were used to fund capital structure (PCS) changes such as share repurchases, debtor in
Secured Loans 14 possession financing, partial or full refinancing, or standby commercial paper support. Only 6 percent of the loans in the sample were obtained for fixed asset backing (PFAB), such as mortgage lending or large asset purchases. About 19 percent of the loans in the sample were used for general corporate purposes (PGCP), which includes loans with this as their stated purpose as well as working capital and trade financing. About 1 percent of the loans were used to fund specific projects (PP). This category includes loans where the purpose is stated as project finance or telecom build out. Lastly, about 0.1 percent of the loans in the sample were listed as other purpose, or they did not fall into one of the previous five categories (POTH). The S&P ratings also are presented next in Table 1. Over 74 percent of the sample was not rated and there were no observations for an S&P rating of C.About 12 percent of the borrowers had an investment grade rating (AAA, AA, A, BBB) at the close of the loan, while about 13 percent of the sample had a high-yield rating (BB, B, CCC, CC, C, D) at the close of the loan. The summary statistics for the industry classification variables are also presented in Table 1. 5. Results 5.1. Basic Data Analysis We begin by looking at the basic characteristics of the sample. Table 2 shows the distribution of loan agreements in the sample by secured status across S&P senior debt ratings. *** Table 2 About Here *** There were 9 loans where the borrower had an S&P senior debt rating of AAA and 89 percent were unsecured. Similar results are found for ratings of AA, A, and BBB, but the numbers drop significataly as we proceed down the ratings scale into high-yield debt such as BB, CCC, CC, and D. For example, over 66 percent of the loans made to BB-rated borrowers were secured, 95 percent of loans made to B-rated borrowers were secured, and 100 percent of loans made to CCC-,
Secured Loans 15 CC-, and D-rated borrowers were secured. Interestingly, only 80.27 percent of loans made to nonrated borrowers were secured. 5.2. The Logit Regression Controlling S&P Coverage Using the first logit model described in Section II, we estimate the regression predicting the probability of a secured loan. These results are provided in Table 3 below. *** Table 3 About Here *** With regard to the information transparency of a firm, we find that a borrower with an S&P rating is less likely to secure a loan than a similar borrower without an S&P rating. Consistent with our hypothesis, banks appear to use collateral to overcome information asymmetries. When borrowers have been reviewed and rated by Standard and Poor s, information about the firm is more transparent and collateral is not as likely. When we proxy for borrower size with the borrower s annual sales figures, we find that larger firms are less likely to secure a loan than their smaller counterparts. We also find that larger loans are less likely to be secured. In line with the findings of Dennis and Mullineaux (2000), we find a significant negative relationship between collateralization and syndication. Likewise, we also find that firms listed on a U.S. stock exchange are less likely to collateralize a loan than a firm that is not listed on an exchange. This supports the hypothesis that the listing requirements of U.S. exchanges increase the likelihood that a borrower is a proven, relatively well-established firm. In addition, because of the regulations governing U.S. securities trading, exchange listed firms involve fewer information asymmetry problems given required disclosure through annual statements and the active eye of analysts and the market itself. The nationality of the firm is not a significant factor in determining loan collateralization. We now turn our attention to the hypothesis that the purpose of the loan affects the secured status of the loan. As shown, relative to capital structure uses, we find that loans used to back fixed
Secured Loans 16 assets (PFAB) and for corporate control purposes (PCC) are more likely to be secured. 5 Loans used to purchase fixed assets are probably tied to the actual asset being purchased; thus, monitoring is relatively easy and the cost to the borrower is small. With regard to corporate control loans, the positive relationship supports the hypothesis that moral hazard can be a significant problem because firms may not be successful in their attempts to takeover or buyout a firm. Moreover, the proceeds from these types of loans are typically invested in marketable assets that can significantly decrease in value in a short period of time. Thus, these loans may also be secured to prevent the lender from realizing a total loss in the event of a market downturn. We find that loans used for general corporate purposes are less likely to be secured, and the results also support the notion that the borrower s industry affects loan collateralization. The findings suggest that three industries tend to secure loans less often than regulated industries (SIC4) like transportation and utilities: mining and construction (SIC1), manufacturing (SIC3), and financial institutions (SIC6). Since each one of these variables has a negative, statistically significant coefficient, we conclude that these industries are either less risky than traditionally regulated industries, or another unknown factor is at play. 5.3. The Logit Model Controlling for General Credit Quality We next estimate the second logit model described in Section II, which identifies the probability of a secured loan, controlling for general credit quality. These results are provided in Table 4 below. *** Table 4 About Here *** When we look at S&P senior debt ratings, we find that borrowers with an investment-grade S&P senior debt rating are less likely to secure loans than similar borrowers without an S&P rating. 5 The coefficient on loans used for other purposes is positive, but it is only significant at the 10 percent level.
Secured Loans 17 A borrower with a high-yield S&P senior debt rating is more likely to collateralize a loan than a similar, non-rated counterpart. Thus, the conclusion from the basic data analysis stands: if collateralization is costly to the borrower, then a borrower would prefer to be non-rated by S&P than to be issued a high-yield rating. In support of this possibility, John, Lynch, & Puri (2000) find that the yield differentials between secured and unsecured public debt are higher for firms with low credit ratings as well as for firms with newly issued debt, compared to more seasoned issues. Turning our attention to the maturity variable, like Dennis, Nandy, and Sharpe (2000), we find that long-term loans are more likely to be secured than their short-term counterparts. One explanation for this finding is that long-term loans present the bank with moral hazard problems because the borrower has more opportunities to misuse funds after the loan is obtained. Thus, collateral serves to reduce the prospect of borrowers engaging in underinvestment or asset substitution. The remainder of the model changes quantitatively, but not in statistical significance. Thus, the results and conclusions drawn from [1] are robust. Information asymmetry continues to play a large role in collateralization, but we now find that credit risk also figures into the secured decision. 5.4. The Logit Model Controlling for Specific Credit Quality Finally, using the third logit model described in Section II, we have estimated the regression model that predicts the probability of a secured loan, controlling for specific S&P senior debt ratings. These results are provided in Table 5 below. *** Table 5 About Here *** As shown, our findings in Table 5 are very similar to those in Tables III and IV. All investment grade S&P ratings (AAA, AA, A, and BBB) are negative and significant at the 10 percent level. Moreover, the latter three are negative and significant at the 1 percent level. Thus, like the findings from [2], loans made to borrowers with an S&P investment-grade senior debt
Secured Loans 18 rating are less likely to secure a loan than a similar non-rated firm. Also in line with the results from [2], we find positive coefficients on all of the S&P high-yield senior debt ratings in the sample (BB, B, CCC, CC, D). However, only the coefficient on the B rating is significant at the 1 percent level. The results relating to senior debt ratings clearly show that investment-grade firms are not as likely to collateralize as high-yield firms. We find that the remainder of the model changes quantitatively, but not in statistical significance. Thus, the results and conclusions from both [1] and [2] continue to be robust to the specification of the credit rating variables. 6 Information asymmetry and credit risk are considered in the decision to secure a loan. 6. Conclusion This paper examines the determinants of the secured status of commercial loans obtained by borrowers both in the United States and abroad. We examine 7,619 commercial loans closed between December, 1988 and January, 2001. Our main finding, consistent with the argument that collateral helps avoid asymmetric information problems and facilitates risk reduction, is that firms with an S&P rating are less likely to secure loans than firms not rated by S&P. Credit risk also is a factor influencing the decision to secure a loan since lenders collateralize loans less often when the borrower has an investment-grade S&P senior debt rating as opposed to having no rating or a highyield rating. Lenders almost always secure a loan when the borrower s S&P senior debt rating is a high-yield rating of BB or lower. High quality firms without an S&P senior debt rating may avoid having to secure loans by obtaining an S&P investment grade rating of BBB or better. Thus, the firm is signaling to the bank that it is indeed a high quality firm and does not pose enough risk to necessitate collateralization of loan agreements. 6 We also performed the same analysis on only rated firms. Again, the results are robust.
Secured Loans 19 Supporting the notion that banks require collateral in the face of information asymmetries, our results suggest that larger firms, such as those with high sales figures and those listed on a U.S. stock exchange, are less likely to negotiate a secured loan agreement. We also find that loan size is inversely related to the probability that a loan is secured. This may be because only larger companies are receiving larger loans. We also find that syndicated loans, which tend to be large loans, are more likely to be unsecured than non-syndicated loans, supporting some previous research. We also find that loans with longer maturity are more likely to be collateralized, which supports the hypothesis that a long-term loan is associated with additional default risk. Although the models used in this paper are highly significant and very good at predicting the probability that a loan is secured, it should be noted that this line of research has many unexplored avenues. 7 We plan to examine the sample of firms not rated by S&P more fully. Although these firms should be more problematic with regard to information asymmetries, they appear to be collateralized less often than risky, high yield firms. We also plan to examine whether moral hazard is a factor influencing the collateralization decision. Indeed, there is still much more to be done to truly understand why lenders and borrowers agree to secure loan agreements. 7 Logit models used in this paper are 78.6, 81.5, and 82.5 percent concordant in prediction power.
Secured Loans 20 REFERENCES Berger, A., and G. Udell, 1990. Collateral, loan quality, and bank risk, Journal of Monetary Economics 25, 21-42. Berger, A., and G. Udell, 1995. Relationship lending and lines of credit in small firm finance, Journal of Business 68, 351-381. Besanko, D., and A. Thakor, 1987. Collateral and rationing: Sorting equilibria in monopolistic and competitive credit markets, International Economic Review 28, 671-689. Bester, H., 1985. Screening versus rationing in credit markets with imperfect information, American Economic Review 75, 850-855. Black, J., and D. De Meza, 1992. Diversionary tactics: Why loans to small businesses are so safe. Working Paper. Exeter: University of Exeter. Boot, A., and A. Thakor, 1994. Moral hazard and secured lending in an infinitely repeated credit market game, International Economic Review 35, 899-920. Boot, A., A. Thakor, and G. Udell, 1991. Secured lending and default risk: Equilibrium analysis, policy implications, and empirical results, Economic Journal 101, 458-472. Chan, Y. S., and G. Kanatas, 1985. Asymmetric valuations and the role of collateral in loan agreements, Journal of Money, Credit, and Banking 17, 84-95. Chan, Y. S., and G. Kanatas, 1987. Collateral and competitive equilibria with moral hazard and private information, Journal of Finance 42, 345-364. Dennis, S., and D. Mullineaux, 2000. Syndicated Loans, Journal of Financial Intermediation 9, 404-426. Dennis, S., D. Nandy, and I. Sharpe, 2000. The determinants of contract terms in bank revolving credit agreements, Journal of Financial and Quantitative Analysis 35, 87-110. Hempel, G., A. Coleman, and D. Simonson, 1986. Bank Management, Wiley, New York, New York. Hester, D., 1979. Customer relationships and terms of loans: Evidence from a pilot survey: Note, Journal of Money, Credit, and Banking 11, 349-357. Igawa, K., and G. Kanatas, 1990. Asymmetric information, collateral, and moral hazard, Journal of Financial and Quantitative Analysis 4, 469-490. Jackson, T. H., and A. T. Kronman, 1979. Secured financing and priorities among creditors, The Yale Law Journal 88, 1143-1182.
Secured Loans 21 John, K., A. Lynch, and M. Puri, 2000. Credit ratings, collateral and loan characteristics: Implications for yield, Working Paper, New York University Stern School of Business. Klapper, L., 2000. The uniqueness of short-term collateralization. Working Paper. The World Bank. Leeth, J. and J. Scott, 1989. The incidence of secured debt: Evidence from the small business community, Journal of Financial and Quantitative Analysis 24, 379-393. Morsman, E., 1986. Commercial loan structuring, Journal of Commercial Bank Lending 68, 2-20. Myers, S. C., 1977. Determinants of Corporate Borrowing, Journal of Financial Economics 5, 147-175. Orgler, Y., 1970. A credit scoring model for commercial loans, Journal of Money, Credit, and Banking 2, 435-445. Rajan, R., and A. Winton, 1995. Covenants and collateral as incentives to monitor, Journal of Finance 50, 1113-1146. Schwartz, A., 1981. Security interests and bankruptcy priorities: A review of current theories, The Journal of Legal Studies 1981, 1-37. Scott, J. H., 1977. Bankruptcy, secured debt, and optimal capital structure, Journal of Finance 32, 1-19. Smith, C. W., Jr., and J. B. Warner, 1979. Bankruptcy, secured debt, and optimal capital structure: Reply, Journal of Finance 35, 247-251. Stohs, M. H., and D. C. Mauer, 1994. The Determinants of Corporate Debt Maturity Structure, Journal of Business 69, 279-312. Stulz, R., and H. Johnson, 1985. An analysis of secured debt, Journal of Financial Economics 14, 501-521. Swary, I., and G. Udell, 1988. Information production and the secured line of credit. Working Paper. New York University. Townsend, R., 1979. Optimal contracts and competitive markets with costly state verification, Journal of Economic Theory, October, 265-293.
Table 1 Descriptive Statistics The sample contains 7,619 commercial loans arrangements closed between December 1988, and January 2001. VARIABLE N MEAN STD DEV MINIMUM MAXIMUM SECURED 7619 0.7323796 0.4427476 0 1.0000000 LIBORSP 7619 187.2371210 108.3665199 0 1400.00 LNMATURITY 7619 3.6641936 0.7166013 0 5.3752784 SYNDICATE 7619 0.8627116 0.3441741 0 1.0000000 EXCHANGE 7619 0.6933981 0.4611129 0 1.0000000 LNSALES 7619 19.5340380 1.9571471 6.7373702 29.4910339 FOREIGN 7619 0.0357002 0.1855539 0 1.0000000 LNDEAL 7619 18.5716848 1.5741766 13.8155106 23.2882152 REFINANCE 7619 0.2086888 0.4063982 0 1.0000000 PCC 7619 0.2560704 0.4364898 0 1.0000000 PCS 7619 0.4773592 0.4995199 0 1.0000000 PFAB 7619 0.0637879 0.2443908 0 1.0000000 PGCP 7619 0.1890012 0.3915353 0 1.0000000 PP 7619 0.0123376 0.1103946 0 1.0000000 POTH 7619 0.0014438 0.0379719 0 1.0000000 SPAAA 7619 0.0011813 0.0343514 0 1.0000000 SPAA 7619 0.0103688 0.1013047 0 1.0000000 SPA 7619 0.0395065 0.1948094 0 1.0000000 SPBBB 7619 0.0714004 0.2575095 0 1.0000000 SPBB 7619 0.0421315 0.2009023 0 1.0000000 SPB 7619 0.0799317 0.2712053 0 1.0000000 SPCCC 7619 0.0094501 0.0967573 0 1.0000000 SPCC 7619 0.0006562 0.0256107 0 1.0000000 SPC 7619 0 0 0 0 SPD 7619 0.0023625 0.0485514 0 1.0000000 SPNR 7619 0.7430109 0.4370020 0 1.0000000 SPINVEST 7619 0.1224570 0.3278344 0 1.0000000 SPHIGHYLD 7619 0.1345321 0.3412455 0 1.0000000 SPRATED 7619 0.2569891 0.4370020 0 1.0000000 ƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒ
Table 2 Distribution of Commercial Loans by Secured Status Across Senior Debt Ratings The sample contains 7,619 commercial loans arrangements closed between December 1988, and January 2001. This table shows the distribution of the loans by Moody s bond rating and collateral classification. S&P Classification Senior Debt ƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒ Total Rating Secured (Percent) Unsecured (Percent) AAA 1 11.11% 8 88.89% 9 AA 7 8.86 72 91.14 79 A 23 7.64 278 92.36 301 BBB 116 21.32 428 78.68 544 BB 212 66.04 109 33.96 321 B 582 95.57 27 4.43 609 CCC 72 100.00 0 0.00 72 CC 5 100.00 0 0.00 5 C 0 --- 0 --- 0 D 18 100.00 0 0.00 18 NR 4,544 80.27 1,117 19.73 5,661 Total 5,580 73.24 2,039 26.76 7,619
Table 3 Logit Model Predicting a Secured Loan Controlling for S&P Coverage The sample contains 7,619 commercial loans arrangements closed between December 1988, and January 2001. A binary variable representing a secured loan (SECURED) is regressed on a binary variable for firms rated at the close of the loan by S&P (SPRATED), the natural logarithm of the annual sales of the borrower (LNSALES), the natural logarithm of a the loan size (LNDEAL), a binary variable for syndicated loans (SYNDICATE), a binary variable for firms listed on a U.S. stock exchange (EXCHANGE), a binary variable for foreign firms (FOREIGN), a set of binary variables for the purpose of the loan including bank refinancing, corporate control, fixed asset backing, general corporate purposes, project financing, and other unclassified purposes (PREF, PCC, PFAB, PGCP, PP, and POTH, respectively), and a set of binary variables for the industry of the borrower based on one-digit SIC codes (SIC0, SIC1, SIC2, SIC3, SIC5, SIC6, SIC7, SIC8, and SIC9). Please note that the reference variable for the purpose binary variables is capital structuring (PCS), and the reference variable for the industry binary variables is traditionally regulated industries such as transportation and utilities (SIC4). SIGN STANDARD PARAMETER DF EXPECTED ESTIMATE ERROR CHI-SQUARE PR > CHISQ INTERCEPT 1 12.9809 0.5122 642.1818 <.0001 SPRATED 1-0.3950 0.0716 30.4619 <.0001 LNSALES 1-0.2646 0.0199 177.6388 <.0001 LNDEAL 1-0.3240 0.0294 121.7273 <.0001 SYNDICATE 1-0.2154 0.1248 2.9799 0.0843 EXCHANGE 1-0.3556 0.0690 26.5876 <.0001 FOREIGN 1-0.0371 0.1477 0.0629 0.8019 PREF 1 /+ 0.2317 0.0737 9.8749 0.0017 PCC 1 /+ 0.7881 0.0767 105.6794 <.0001 PFAB 1 /+ 0.4956 0.1344 13.5915 0.0002 PGCP 1 /+ -0.4281 0.0829 26.6485 <.0001 PP 1 /+ 0.8037 0.3043 6.9755 0.0083 POTH 1 /+ 0.9073 0.8666 1.0962 0.2951 Testing Global Null Hypothesis: BETA=0 Test Chi-Square DF Pr > ChiSq Likelihood Ratio 1548.7629 21 <.0001 Score 1477.8170 21 <.0001 Wald 1175.1674 21 <.0001
Table 4 Logit Regression Predicting a Secured Loan Controlling for General Credit Quality The sample contains 7,619 commercial loans arrangements closed between December 1988, and January 2001. A binary variable representing a secured loan (SECURED) is regressed on a binary variable for firms rated as investment grade at the close of the loan by S&P (SPINVEST), a binary variable for firms rated as high yield at the close of the loan by S&P (SPHIGHYLD), the natural logarithm of the annual sales of the borrower (LNSALES), the natural logarithm of a the loan size (LNDEAL), a binary variable for syndicated loans (SYNDICATE), a binary variable for firms listed on a U.S. stock exchange (EXCHANGE), a binary variable for foreign firms (FOREIGN), the natural logarithm of the term of the loan in months (LNMATURITY), a set of binary variables for the purpose of the loan including bank refinancing, corporate control, fixed asset backing, general corporate purposes, project financing, and other unclassified purposes (PREF, PCC, PFAB, PGCP, PP, and POTH, respectively), and a set of binary variables for the industry of the borrower based on one-digit SIC codes (SIC0, SIC1, SIC2, SIC3, SIC5, SIC6, SIC7, SIC8, and SIC9). Please note that the reference variable for the purpose binary variables is capital structuring (PCS), and the reference variable for the industry binary variables is traditionally regulated industries such as transportation and utilities (SIC4). SIGN STANDARD PARAMETER DF EXPECTED ESTIMATE ERROR CHI-SQUARE PR > CHISQ Intercept 1 10.1910 0.5602 330.9350 <.0001 SPINVEST 1-1.9020 0.1111 292.8657 <.0001 SPHIGHYLD 1 + 1.0163 0.1102 85.1255 <.0001 LNSALES 1-0.2040 0.0209 95.1707 <.0001 LNDEAL 1-0.3223 0.0323 99.3344 <.0001 SYNDICATE 1-0.4332 0.1275 11.5408 0.0007 EXCHANGE 1-0.1811 0.0736 6.0521 0.0139 FOREIGN 1 0.0422 0.1573 0.0720 0.7885 LNMATURITY 1 + 0.4352 0.0456 91.0203 <.0001 PREF 1 /+ 0.2299 0.0821 7.8479 0.0051 PCC 1 /+ 0.7578 0.0850 79.5042 <.0001 PFAB 1 /+ 0.4635 0.1403 10.9077 0.0010 PGCP 1 /+ -0.3273 0.0876 13.9466 0.0002 PP 1 /+ 0.2882 0.3278 0.7731 0.3793 POTH 1 /+ 1.4092 0.9598 2.1556 0.1420 Testing Global Null Hypothesis: BETA=0 Test Chi-Square DF Pr > ChiSq Likelihood Ratio 2319.3343 23 <.0001 Score 2336.2519 23 <.0001 Wald 1424.1272 23 <.0001
Table 5 Logit Regression Predicting a Secured Loan Controlling for Specific S&P Ratings The sample contains 7,619 commercial loans arrangements closed between December 1988, and January 2001. A binary variable representing a secured loan (SECURED) is regressed on a set of binary variables for the S&P senior debt rating of the firm at the time of the loan closing (SPAAA, SPAA, SPA, SPBBB, SPBB, SPB, SPCCC, SPCC, SPD), the natural logarithm of the annual sales of the borrower (LNSALES), the natural logarithm of a the loan size (LNDEAL), a binary variable for syndicated loans (SYN DICATE), a binary variable for firms listed on a U.S. stock exchange (EXCHANGE), a binary variable for foreign firms (FOREIGN), the natural logarithm of the term of the loan in months (LNMATURITY), a set of binary variables for the purpose of the loan including bank refinancing, corporate control, fixed asset backing, general corporate purposes, project financing, and other unclassified purposes (PREF, PCC, PFAB, PGCP, PP, and POTH, respectively), and a set of binary variables for the industry of the borrower based on one-digit SIC codes (SIC0, SIC1, SIC2, SIC3, SIC5, SIC6, SIC7, SIC8, and SIC9). Please note that the reference variable for the purpose binary variables is capital structuring (PCS), and the reference variable for the industry binary variables is traditionally regulated industries such as transportation and utilities (SIC4). SIGN STANDARD PARAMETER DF EXPECTED ESTIMATE ERROR CHI-SQUARE PR > CHISQ Intercept 1 9.6618 0.5680 289.3781 <.0001 SPAAA 1-2.3811 1.3925 2.9240 0.0873 SPAA 1-2.0111 0.4130 23.7186 <.0001 SPA 1-2.7293 0.2337 136.3401 <.0001 SPBBB 1-1.7060 0.1230 192.3513 <.0001 SPBB 1 + 0.0944 0.1392 0.4598 0.4977 SPB 1 + 1.9504 0.2059 89.7659 <.0001 SPCCC 1 + 14.4281 264.8 0.0030 0.9565 SPCC 1 + 13.7307 1039.9 0.0002 0.9895 SPD 1 + 14.6005 533.0 0.0008 0.9781 LNSALES 1-0.1901 0.0210 82.0339 <.0001 LNDEAL 1-0.3000 0.0328 83.6753 <.0001 SYNDICATE 1-0.4778 0.1279 13.9616 0.0002 EXCHANGE 1-0.1518 0.0745 4.1543 0.0415 FOREIGN 1 + 0.0429 0.1563 0.0755 0.7835 LNMATURITY 1 + 0.3980 0.0462 74.2988 <.0001 PREF 1 + 0.2398 0.0828 8.3804 0.0038 PCC 1 +/ 0.7617 0.0857 78.9513 <.0001 PFAB 1 +/ 0.4771 0.1397 11.6695 0.0006 PGCP 1 +/ -0.3244 0.0884 13.4766 0.0002 PP 1 +/ 0.3441 0.3322 1.0727 0.3003 POTH 1 +/ 1.5435 0.9172 2.8321 0.0924 Testing Global Null Hypothesis: BETA=0 Test Chi-Square DF Pr > ChiSq Likelihood Ratio 2446.6655 30 <.0001 Score 2396.9571 30 <.0001 Wald 1383.6976 30 <.0001