This item is only available as the following downloads:
1 ASSET SPECIFICITY, I NDUSTRY DRIVEN RECOV ERY RISK AND LOAN PR ICING By ATAY KIZILASLAN A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2012
2 2012 Atay Kizilaslan
3 ACKNOWLEDGMENTS I would like to thank several people who supported me throughout this long journey. Without their help and support, this work would have been impossible. My dissertation chair, advisor, mentor, guide and support, Christopher James. I know that words will be insufficient to express how thankful I am to him, but I will do my best. Many thanks for being an excellent mentor, and guide. I am glad to have a chance to work with you. Remaining faculty members of the department of Finance, Insurance and Real Estat e. Professor Joel Houston, Michael Ryngaert, David Brown, Mahen Nimalendran, Andy Naranjo Thank you for your guidance and supports. My current and old colleagues in the PhD program. We all walk through the same journey. A journey that is full of excitemen t, curiosity, achievements as well as some challenges. I know that without you I could not have walked through this. Ani Manakyan, Aaron Gubin, Sabuhi Sardarli, Alice Bonaime, Ozde Oztekin, and Dominique Badoer, thanks for being there for me. Cem Demiroglu I cant thank you enough. From the first day until today, you were a great friend and mentor! Thank you! My dear friends, Ceylan Yanyali, Gogce Kayihan, Gulver Karamemis, Nail Tanrioven, Sezgin Ayabakan, Dincer Konur, Ece Unur, Mete Yilmaz, Meltem Alemdar Sinem Gokgoz Kilic, Bulent Anil. Throughout my PhD, you were always there with your friendship and help me to achieve this. I w o u l d l i k e t o t h a n k m y p a r e n t s f o r b e i n g t h e r e f o r m e a n d s u p p o r t i n g m e f r o m t h e v e r y f i r s t d a y
4 My dear sister in law Funda Golcuklu. Thank you for supporting me all through this years. I really appreciate your guidance and ad vises, especially when I think there is no end. Last but not least to my brother Eray Kizilaslan. This dissertation will be impossible without your endless support, guidance and help. If I get a chance to be called doctor from now on, it is because of you! Thank you!
5 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 3 LIST OF TABLES ................................ ................................ ................................ ............ 7 ABSTRACT ................................ ................................ ................................ ..................... 8 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ ...... 9 2 LITERATURE SURVEY ................................ ................................ .......................... 15 2.1 Asset Sp ecificity and Lenders ................................ ................................ ........... 17 2.2 Asset Specificity and Fire Sales: Empirical Evidences ................................ ...... 20 2.2.1 Ex Post Evidences ................................ ................................ ................ 20 2.2.2 Ex ante Impacts of Potential Fire Sales ................................ ................. 23 3 DATA DESCRIPTION ................................ ................................ ............................. 25 3.1 Recovery Sample ................................ ................................ .............................. 25 3.2 Compustat Sample ................................ ................................ ........................... 26 3.3 Dealscan Sample ................................ ................................ .............................. 26 3.4 Lines of Credit Sample ................................ ................................ ...................... 27 3.5 In dustry Risk Measures (IDEMs) ................................ ................................ ..... 28 4 SUMMARY STATISTICS ................................ ................................ ........................ 32 5 INDUSTRY RISK, RECOVERY RATES AND THE LIKELIHOOD OF DISTRESS 39 5.1 Industry Risk and Recovery Rates ................................ ................................ .... 39 5.2 Industry Risk and Financial Distress: ................................ ................................ 40 6 INDUSTRY RISK AND BANK LOAN PRICING ................................ ...................... 46 6.1 Industry Risk and Loan Pricing ................................ ................................ ......... 46 6.2 Does the Importance of Industry Risk Vary with Loan Security? ...................... 48 6.3 Does the Importance of Industry Risk Vary with Asset Specificity? .................. 49 6.4 Does the Importance of Industry Risk Vary with Credit Market Conditions? ..... 51 6.5 Industry Risk and Terms of Loan Contract ................................ ........................ 54 7 INDUSTRY RISK AND LIQUIDITY CHOICE ................................ .......................... 63 7.1 Industry Risk and Liquidity Management ................................ .......................... 63 7.2 Cash vs. Lines of Credit ................................ ................................ .................... 64 7.3 Industry Risk and Cash Holdi ngs ................................ ................................ ...... 65
6 8 CONCLUSION ................................ ................................ ................................ ........ 70 APPENDIX: VARIABLE DEFINITIONS ................................ ................................ ......... 71 LIST OF REFERENCES ................................ ................................ ............................... 74 BIOGRAPHICAL SKETCH ................................ ................................ ............................ 78
7 LIST OF TABLES Table page 4 1 Recovery Sample. ................................ ................................ .............................. 35 4 2 De alscan Sample. ................................ ................................ .............................. 36 4 3 Line of Credit Sample. ................................ ................................ ........................ 37 4 4 Correlations. ................................ ................................ ................................ ....... 38 5 1 Recovery Rates and industry distress exposure measures (IDEM). ................... 43 5 2 Likelihood of firm distress during industry distress. ................................ ............ 45 6 1 Loan pricing and IDEM. ................................ ................................ ...................... 56 6 2 Loan pricing and IDEM: Subsample Analysis. ................................ .................... 58 6 3 Loan pricing and IDEM: Subsample Analysis. ................................ .................... 59 6 4 Loan pricing and IDEM: Financial Crisis of 2008. ................................ ............... 60 6 5 Loan pricing and IDEM: Financial Crisis of 2008. ................................ ............... 61 6 6 Terms of Loan Contract and IDEM. ................................ ................................ .... 62 7 1 Liquidity Management and IDEM. ................................ ................................ ....... 68 7 2 Liquidity Management and IDEM: Financial Crises of 2008. .............................. 69
8 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy ASSET SPECIFICITY, I NDUSTRY DRIVEN RECOV ERY RISK AND LOAN PR ICING By Atay Kizilaslan August 2012 Chair: Christopher M. James Major: Business Administration In this downturns, what we refer to as industry risk and bank loan pricing. The basic idea is that if it is costly to redeploy industry assets, then the liquidity of with the industry as well as macroeconomic conditions. Since the cost of bank financing greater exposure to industry downturns (and hence higher expe cted LGD) will find bank loans a more expensive and conditional source of liquidity. We measure industry risk on an industry downturn. We find industry risk is signif icantly related to the recovery rates in bankruptcy and likelihood of the firm experiencing financial distress when its peers are also in distress. More importantly, we find that the spreads on unsecured bank loans are positively related to industry risk m easures. We find that these relationships are stronger for firms with industry specific assets. We also find reliance on cash relative to lines of credit is significantly related to our measures of industry risk exposure.
9 CHAPTER 1 INTRODUCTION A potent assets. If assets are specialized and thus cannot be readily redeployed to industr y to be another firm within the same industry. 1 As a result, asset illiquidity is likely to vary eers because when a distressed firm is trying to sell assets, industry peers may also be facing liquidity and resulting in fire sale discounts. Shleifer and Vishny (1 992) argue that the potential for fire sale discounts may affect ex ante capital structure and liquidity choices. Two recent papers empirically examine these issues. Specifically, Campello and Giambona (2010) examine the relationship between leverage and a sset redeployability. They find that redeployability of tangible assets is an important driver of leverage for financially constrained firms. They also find that asset redeployability affects corporate borrowing more during tight credit market conditions. Related work by Acharya, Almedia and Campello (2010), examines 1 See Shleifer and Vishny (1992) and Williamson ( 1988). Also Shleifer and Vishny (2011) for a recent review of the literature on fire sale discounts.
10 affects the banki The basic idea is that banks are able to create liquidity by pooling idiosyncratic risks. Firms with higher aggregate risk exposure are more likely to demand liquidity when liquidit y is scarce and thus are charged more for insurance. Consistent with this argument, they find that firms with high asset betas rely more heavily on cash as a precautionary hedge against liquidity shocks and pay higher spreads on their bank loans. We add to downturns, what we will refer to as industry risk and the pricing and structure of bank loans and reliance on bank lines of credit as a source of liquidity. The basic idea is tha t loan pricing and the evaluation of credit risk involve both an assessment of the likelihood of default as well as the loss given default (LGD) (Saunders and Allen (2010)). Ceteris paribus, the higher the expected LGD, the higher the credit spread and thu s the higher the expected cost of borrowing under lines of credit. As a result, if industry recovery risk is a substantial component of distress costs, then firms with greater industry risk exposure will face higher credit risk spreads (and perhaps more on erous contract terms in terms of lower advance rates or higher collateral requirements) and therefore rely more heavily on cash as a source of liquidity. Since liquidity risk may affect the costs of raising cash (and the willingness of banks to lend) in a dverse markets, firms with significant industry risk exposure may hold more precautionary cash balance and rely less on bank lines of credit as a source of liquidity. We begin our analysis by examining several potential measures of industry risk based on
11 is related to changes in the value of its industry peers conditional on an industry downt urn. We evaluate potential industry risk measures based on their ability to predict recovery rates (or LGD) and financial distress. A good industry risk measure should have predictive power both in terms of the likelihood of firm distress, conditional on i ndustry distress, and in terms of recovery rates in bankruptcy. We next examine the relationship between various industry risk measures and bank loan pricing, line of credit usage and cash holdings. If industry risk affects expected LGD we would expect th industry risk exposure. For this analysis we use Dealscan data on loan pricing and hand collect data on bank credit line usage and cash holdings for a panel of 348 publicly traded companies over the 1999 to 2009 time period. One challenge in analyzing the importance of industry risk is to distinguish aggregate risk are likely to affect loan pricing for at least two reaso ns. First, previous empirical studies have found that recovery rates are related to the aggregate number of defaults (e.g., Altman, Brady, Resti and Sironi (2005)) and are significantly lower during recessions (e.g., Schuermann (2004)). Lower expected reco aggregate risk exposure should be reflected in credit spreads and therefore may affect the choice between cash and lines of credit. Second, and perhaps more important, egate risk exposure liquidity through lines of credit. Consistent with this argument, Acharya et al. (2010) find
12 that firms with high asset betas rely more heavily on cash a s a precautionary hedge against liquidity shocks and pay higher spreads on their bank borrowing. We examine the importance of industry risk using several measures. Our first timating and adjusting these estimates for leverage. While providing potentially useful industry industry peers conditional on an industry downturn. Our second approach is to estimate istress when its industry peers are also in financial distress). Our tail risk measures are similar in spirit to systemic risk measures used in the banking literature (which focus on an individual our first tail risk measure is based on works by Acharya, Pedersen, Philippon and Richardson (2010) on systemic et returns. We calculate a returns in the worst 5% of industry return months. We call this measure marginal distress estimate (MDE). Our second tail risk measure is the co rrelation between firm and industry returns conditional on negative industry returns. Overall, we find a negative and significant relationship between recovery rates and our industry tail risk measures. We find no significant relationship between recovery rates and aggregate risk measures. We also find that, controlling for aggregate risk,
13 industry risk is positively related to the likelihood that a firm will become distressed during times of industry wide distress. The recovery rate findings and the distr ess analysis indicate that our tail risk measures are better proxies of industry recovery risk than industry beta. We find that all of the industry risk measures are significantly and positively related to the all in drawn spreads charged on bank loans. Consistent with industry risk reflecting higher expected LGD, we find greater effect of industry risk on loan rates for firms in industries with greater assets specificity and during tight credit market related to our industry risk measures. Our study contributes to the literature in several ways. First, we show that ex ante measures of industry risk are related to distress likelihood and recovery rates. Since our industry risk measures are computed before the onset of financial distress they provide a measure of the effect of fire sale externalities on r ecovery values that is independent of contemporaneous industry conditions. Second, we provide evidence that the potential for fire sale discounts is reflected in loan pricing. In particular, we show that borrowing costs are significantly related to our ind ustry risk measures, especially for firms operating in industries with high asset specificity. These findings are consistent with recent empirical work concerning the influence of asset redeployability capital structure decisions (e.g., Benmelech and Bergm an (2009), Rauh and Sufi (2012) and Campello and Giambona (2010)). We extend the results of these papers by showing how asset specificity affects the cost (as opposed to the quantity) of borrowing. Third, we extend recent empirical works by Acharya et al ( 2010) on liquidity management to show that
14 industry risk in addition to aggregate risk affects loan pricing and the reliance on bank lines of credit as a source of liquidity. This last finding provides additional support for the Sufi (2009) argument that c ash and lines of credit are imperfect substitutes for corporate liquidity.
15 C HAPTER 2 LITERATURE SURVEY Pledging collateral is an important mechanism to mitigate external financing frictions stemming from various sources. For instance, when adverse selection arise due to asymmetric information, high quality firms can use collateral as a signaling device to differentiate themselves from low quality firms. For high quality firms, pledging collateral is not costly giv en that they are less likely to default and lose the collateral. They can use collateral and raise debt at a lower price. However, low quality firms cannot take such action and have to raise unsecured debt at a high price. On the other hand, when there is a moral hazard problem due to hidden information, lenders ask lower quality firms to pledge collateral. In this case, the role of collateral is that it increases the borrowers' pledgable income and helps low quality firms to obtain debt at a lower price. T he vast amount of empirical evidence supports this channel as the source of financing frictions. Berger and Udell (1990) find that riskier firms are more likely to use collateral when they are raising debt. Jimenez, Salas, and Saurina (2006) also show that lenders are more likely to ask for collateral when the borrower is more likely to default. The redoplayability of the collateral is also important in mitigating the financing frictions related to moral hazard. When firms' assets are more redeployable, the y have alternative uses and higher liquidation values. Besides, these assets are less likely to face fire sale prices when the firm is in distress. Fire sales occur when asset prices move away from their fundamental value if the firm's assets are not easi ly redeployable. Unfortunately, most of the firm's assets are not easily redeployable because they are specialized and should be sold to someone who can use them as efficiently as the seller does. This fact makes the firm's industry peers
16 the best buyer fo r its assets. However, when a firm is in distress and tries to liquidate its assets, its industry peers are more likely to be in financial trouble as well. Since the industry peers cannot raise funds to acquire these assets, the assets are more likely to b e sold to industry outsiders that have less expertise with the assets. The outsiders incur high operational costs when they buy the assets on sale. 1 Therefore, they are willing to buy the assets at valuations that are mu ch lower than the actual value. Ex p ost, the impact of fire sales can be observed through the price at which a firm liquidates its assets in distress. An example is the case of Bear Stearns. In early 2008, the entire financial system was experiencing liquidity problems due to their high expo sure to the declining mortgage backed securities market. The credit spreads were high and financial institutions were having a hard time raising short term debt financing. Bear Stearns was one of those institutions that was in serious trouble due to its hi gh exposure to residential mortgage backed securities. Because Bear Stearns was too interconnected with the rest of the financial institutions, the Federal Reserve Bank of New York did not allow Bear to fail and helped it to be acquired by JPMorgan Chase f or $2 per share. Total value of the transaction was approximately $236. A week before the transaction given before the transaction (and was $20 billion in January 2007), the fire sale envi ronment implied a substantial discount. 1 The outsiders might not know how to operate the assets and have to bear costs of hiring specialist s to operate these assets. In addition, the outsider may have a fear of overpaying since they cannot value the assets properly.
17 Fire sales can also have some ex ante effects on the firm's financial decisions. Lenders take potential for fire sales into consideration and reflect it on terms of the debt contract through the credit spreads or ti ghtness of financial covenants. In return, it would affect both the capital structure as well as the liquidity choice of the firm. Fire sales can also affect distress resolutions to the extent that firms choose between liquidation (Chapter 7) and reorganiz ation (Chapter 11). Our goal in this chapter is to review empirical studies that provide evidence on the impacts of asset specificity on distress outcomes and financial policies of the firm. We begin with the description of asset specificity and an assess point of view. Next, we review the empirical evidence on the impact of asset specificity with regards to fire sales. Then, we discuss how asset specificity affects distress outcomes. Finally, we review the ex ante impacts of as set specificity and potential fire sales on the terms of the debt contract, firm capital structure choices and corporate liquidity policy. 2.1 Asset Specificity and Lenders In debt financing, the lender assumes that the borrower will be able to make regu lar interest payments, comply with the covenants and pay the principal at the maturity date. If the borrower defaults, then the lender exercises its pre emptive claims against the collateral. Ex ante, the lender expects to recover all of its claims and exp erience zero loss given default. The basic definition of loss given default is the ratio of losses to exposure at default. Lenders calculate their loss on principal and carrying cost of non performing loans, as well as the workout expenses when borrower d efaults. The calculation of loss given default is fairly easy when the borrowers default on their bonds. The bank can
18 either observe loss given default from the market price of defaulted bonds or look at credit spreads on the non defaulted risky bonds. Th e bank can also use promised cash flows from the distressed firm to calculate the loss given default. It is a rather complicated method given that lenders should pay close attention to the timing of cash flows and also need to determine the appropriate dis count rate. Various factors affect loss given defaults (and recovery). The first and most important one is the seniority of the debt, which is influenced by the presence of collateral. Being first in line during bankruptcy liquidation is important. In gene ral, unsecured lenders can only recover whatever remains after secured creditors are repaid. Gupton, Gates and Carty (2000) examine the syndicated loan market in the US between 1989 and 2000 and report that senior secured loans have 69.5% recovery rates wh ereas senior unsecured loans have a recovery rate of 52.1%. Altman (2009) examines public corporate bonds between 1978 and 2009. He reports that senior secured bank bonds have a median of 59% recovery rates, whereas senior unsecured bonds have a median rec overy rate of 45.9%. In terms of asset specificity, these findings imply that asset specificity matter most for creditors whose priority makes them more vulnerable to fluctuations in the value of collateral. Secured creditors have the highest claim and ca n impose lower advance rates (i.e. require more collateral per dollar lent) when collateral values are more uncertain. On the other hand, unsecured lenders ask for higher ex ante interest rates on the loans since they are more likely to have lower ex post recovery on their claims.
19 downturns on loan pricing, and find that it has the greatest impact on unsecured loan spreads for firms that operate in industries with greater asset s pecificity. Another important factor that affects recovery rates is the business cycles. A number of empirical studies show that recovery rates are lower during recessions. Altman, Brady, Resti and Sironi (2005) show that recoveries are lower when aggregat e default rates are higher. In addition, Hu and Perraudin (2002) report that the correlation between aggregate defaults and recoveries are 20% in the US. These papers discuss the importance of business cycles on recovery rates but do not focus on how busi ness cycles affect recovery rates. One potential reason why business cycles affect recovery rates is that the likelihood of fire sales increases. When a firm is in distress and wants to liquidate its assets during economic downturns, potential buyers are m ore likely to have their own liquidity problems and might not be able to bid aggressively for these assets. Even worse, they might be trying to sell their own assets. This combination of excess supply and low demand may lead to fire sale prices during econ omic downturns. In addition to this fire sales channel, economic downturns can affect liquidation prices through a decline in the economic prospects of the firm. If fire sales affect liquidation values then the impact should be stronger for firms that ope rate in industries that have fewer deployable assets and when the firm is in distress during industry wide distress. Acharya, Bharath and Srinivasan (2007) examine a broad cross section of industries to disentangle these two potentially interrelated impact s. Using a sample of defaulted firms in the United States between 1982 and 1999, they show that recoveries on defaulted securities are 10 15 cents lower when the firm and its industry are in distress at the same time. They also show that the magnitude of t he discount declines
20 as the seniority of the defaulted security increases. Furthermore, using the industry median of the ratio of machinery and equipment to total assets as a proxy for asset specificity, they show that recoveries decline for firms that ope rate in more asset specific industries. Consistent with the fire sales argument, the recoveries are even lower for firms in more asset specific industries when both the firm and the industry are in distress. They also report that the impact of fire sales v ary with the seniority and presence of collateral. They show that the impact of fire sales are greater on unsecured debt relative to secured debt, and the discount increases when the debt is backed by more asset specific collateral. The third important fac tor that affects recovery rates is the industry of the defaulted firm. Recent empirical studies show that recovery rates vary with the industry of the firm. James and Kizilaslan (2012) show that recovery rates are lower when a firm has greater industry ris k, measured as the exposure of the firm to downturns in its industry. Acharya, Bharath and Srinivasan (2003) examine the cross section of industries and find that recoveries are lower when the industry of the defaulted firm is in distress. They also show t hat the effect of industry distress on recovery rates is greatest for industries that have less easily deployable assets. 2.2 Asset Specificity and Fire Sales: Empirical Evidences 2.2.1 Ex P ost E vidences From a theoretical standpoint, Williamson (1998) an d Shleifer and Vishny (1992) describe asset specificity and its implications very clearly. However, verifying the effects of asset specificity poses a challenge for empirical researchers. One must be able to estimate fundamental values to calculate the fir e sale discount. Unfortunately, it is a
21 tricky task. For this reason, earlier studies focus on stock price reactions. However, they present mixed results. Using 62 assets sales that were completed by financially distressed firms between 1979 and 1988, Bro wn, James and Mooradian (1994) find that asset sales that are used to repay debt lead to lower average share price reactions than sales where the firm retains the proceeds. They suggest that this lower price reaction indicates significant lost option value for shareholders. On the other hand, Lang, Poulsen and Stulz (1995) find a positive and significant stock price reaction when firms use the proceeds of asset sales to repay debt. Pulvino (1998) is one of the first studies to take a direct approach to meas uring fire sale discounts. Using the commercial aircraft transactions that occurred between 1978 and 1991, he shows that financially troubled airlines sell their aircrafts at a 10 15% discount. He defines an airline as financially troubled if its leverage ratio is above the industry median and current ratio is below the industry median. He also shows that discounts are significant only during economic downturns. Furthermore, he reports that financially troubled airlines are more likely to sell their assets to industry outsiders, such as financial institutions. He shows that discounts can increase to 30% during an economic downturn if the buyer is outside the airline industry. This evidence is consistent with the theory presented in Shleifer and Vishny (1992 ) that fire sales lead to costly liquidations for a firm when its industry is in distress. In addition, several studies show that firms prefer debt workout to liquidation when markets are depressed. Asquith, Gertner and Scharfstein (1994) report that the l ikelihood of asset sales is negatively associated with industry leverage and market to
22 book ratio. Almeida, Campello and Hackbarth (2012) develop a model that shows that financially distressed firms may choose to merge with their industry peers even if the re are no synergies associated with the merger. They argue that firms enter such transactions to avoid inefficient liquidations. They call these transactions "liquidity mergers". Consistent with the effect of fire sales, they show that liquidity mergers ar e more likely to occur in more asset specific industries. However, the availability of liquidity within the industry may lead to complications within their model. An acquirer might have its own liquidity problems and be unable to raise funding for the acqu isition. To complete the acquisition, the acquirer needs ex ante liquidity insurance, which could be obtained through a line of credit. To verify this line of reasoning, the authors show that firms prefer to finance liquidity mergers with money raised from a line of credit (rather than cash). Even though firms attempt to avoid fire sales through mergers or debt work outs, fire sales can still create negative externalities for firms, even if they are not in financial distress. Benmelech and Bergman (2011) ar gue that bankrupt firm can create a negative externality for the other firms in the same industry via a decline in their collateral values. They argue that when a firm declares bankruptcy, it is more likely to sell its assets and increase the supply. In ad dition, when the bankrupt firm is trying to sell its assets, the market could be illiquid to the extent that the potential buyers are facing their own liquidity problems and cannot bid aggressively for the assets. This decline in demand, combined with exce ss supply, can reduce the value of collateral for the entire industry. Since collateral values directly affect a firm's ability to raise external debt, such a reduction in collateral values will decrease the firm's debt capacity.
23 Benmelech and Bergman (201 1) find evidence consistent with this notion using data on the pricing of debt tranches secured by commercial aircraft. They find that the bankruptcy of an airline is associated with significant declines in the value of tranches whose underlying collateral is secured by aircraft that are similar to aircraft in the fleet of the bankrupt airline. 2.2.2 Ex ante Impacts of Potential Fire S ales In addition to these ex post externalities, the risk of fire sales could also have an effect on ex ante debt contracts, and in return could affect the financial policy of the firm. For instance, lenders may increase credit spreads when the underlying colla teral is not easily redeployable. Consistent with this idea, Benmelech and Bergman (2009) show that credit spreads are lower on debt tranches of collateralized debt obligations issued by US airlines when they are backed by more redoplayable collateral. The impact is economically sizable. They show that moving from the 25th percentile to 75th percentile of their redeployability measure leads to an approximately 20bps decline in credit spreads. This is almost 10% of the unconditional average spread. In additi on, they find that tranches that are backed by more redeployable collateral are more likely to get higher credit ratings and have lower loan to value ratios. James and Kizilaslan (2012) suggest that firm's exposure to industry downturns affects ex post rec exposure to industry downturns, what they call industry risk, is an important determinant of loan pricing and affects the liquidity policy of the firm. Financial distress is particularly pai nful for firms with assets that are not easily redeployable and firms that are expected to be in financial distress when other firms in the industry are in financial distress. The idea is that those firms are more likely to sell assets at fire sale prices in financial
24 distress. This means that collateral values will decline and recovery values would be lower. Ex ante, banks price this ex post decline in recoveries and charge more on credit lines. In return, firms should rely more on cash as a source of liqu idity since credit lines become a more expensive and possibly less available source of liquidity. Consistent with this idea, they find that recoveries are negatively associated with industry risk, and firms with higher industry risk pay more for their line s of credit. They show that the impact of industry risk is greater for unsecured loans and for firms that operate in more asset specific industries. Furthermore, they show that high industry risk firms prefer cash over bank loans since bank loans are a mor e expensive and less available source of liquidity. Campello and Giambona (2012) also examine asset redeployability and find that the redeployability of tangible assets has a significant effect on the observed leverage ratios. In addition, they show that t his relationship only holds for small and unrated firms and firms with a low dividend payout ratio. This is consistent with the financial friction argument to the extent that asset redeployability only matters for credit constrained firms. They also show t hat asset redeployability helps firms to raise debt when financial markets are tight.
25 C HAPTER 3 DATA DESCRIPTION We collected data on recovery rates, industry risk, loan pricing and line of credit usage from several sources. The sample we use in our various analyses depends on the availability of data. 3 .1 Recovery Sample To investigate whether creditor losses during financial distress are related to our risk measures (estimated before the firm enters into financial distress) we use data on rec overy rates from Bankruptcydata.com. We first identify a list of firms that filed for bankruptcy during the January 1998 to February 2010 time period using Bankruptcydata.com. Bankruptcydata.com provides information on the bankruptcies of public firms as w ell as selected private companies that have public debt or are deemed significant or newsworthy (Ayotte and Morrison (2009)). Bankruptcydata.com provides data for 1,396 bankruptcies during the 1998 to 2010 time period. We hand match this list to Compustat by firm name and require that the bankrupt firm file at least one annual financial statement with the Securities and Exchange Commission during the two fiscal years preceding the bankruptcy filing. Overall, 781 firms satisfy this condition. Next, we searc h Banktrupctydata.com for the reorganization plans of these firms. We found reorganization plans for 309 of the public firms meeting our selection criteria. 1 1 Reorganization plans (that include debt recovery rates) are available from Bankruptcydata.com for less than 50% of the bankru ptcies Reorganization plans are missing for (1) firms that were acquired or liquidated while in bankruptcy, (2) firms whose bankruptcies are still in progress, and (3) some firms that successfully exited bankruptcy. Using the universe of firms that succes sfully exited bankruptcy, we examine whether there are any differences between firms with and without reorganization plans in Bankruptcydata.com (not tabulated for brevity).
26 We eliminate all non debt related claims (i.e. equity, fee, administrative and compensation relate d claims). This process yields 2,371 debt claims issued by 309 firms. 3 .2 Compustat Sample To examine whether our industry risk measures predict firm/industry distress, we construct a sample of firm year observations from Compustat during the period 1999 t o 2009. We exclude financial firms and utilities (SIC codes 6000 6999 and 4900 4999). We also exclude any observations with missing data on total assets. Stock return data for our sample of Compustat firms are obtained from the CRSP monthly stock price dat abase. We merge the CRSP and Compustat data using the historical file from CRSP. Our final sample consists of firm year observations in the intersection of our Compustat sample and CRSP. 3 .3 Dealscan S ample For our loan pricing analysis we use data from Dealscan. In particular, we obtained from Dealscan a list of US dollar denominated lines of credit taken out by all industrial US firms during the 1987 through 2009 time period. 2 From that list, we selected loans to firms with non missing financial inform ation in Compustat at the end of 1 Relative to firms whose reorganization plans are missing, firms with reorganization plans are larger, have higher cash flows, hold more tangible assets, and have higher leverage. Also, reorganization p lans are more likely to be available for firms that operate in more asset specific industries. However, there is no significant difference in the industry distress exposure measures of the two groups. 2 Given our focus on liquidity management, we examine the relationship between the pricing of lines of credit and industry risk. However, we obtain qualitatively similar results if we include all short term bank loan agreements in our sample.
27 the fiscal year preceding the loan date. For these loans, we obtained information on the amount, pricing and maturity of the credit facility. We also obtain information from Dealscan on the number of financial covenants ass ociated with the loan facilities. 3 Our Dealscan sample consists of 26,378 loans taken out by 6,232 unique firms. 3 .4 Lines of C redit S ample To examine the relationship between cash holdings, line of credit usage and industry risk, we hand collected informa tion on line of credit use for a random sample of firms. 4 We start by randomly selecting 500 US industrial firms 5 from Compustat with non missing information on cash flows, cash holdings, and with stock price information available for fiscal years 2005, 20 06 and 2007. We apply these screens to ensure that firms are public and report the key financials that we need for our analysis. Then we annual reports (10 Ks) of the firms in this sample for the period 1999 to 2009. Following Ks. We obtain information on the size as well Ks. We eliminate 3 As Chava a nd Roberts (2008) and Demiroglu and James (2010) discuss, Dealscan data on loan covenants is often missing. Following previous studies, we exclude from our analysis of covenant structure observations with missing data and specifically do not assume that wh en covenant information is missing a specific covenant was not imposed. 4 Information on line of credit use is not available from Compustat or other machine readable sources. 5 We restrict our sample to industrial firms in order to avoid capital structur es governed by regulatory environments, such as financial and utility firms.
28 firms that do not provide sufficiently detailed information on the presence, size, and utilization of their lines of credit. This procedure yields a sample of 348 firms that we follow from 1999 to 2009. 3 .5 Industry Risk M easures (IDEMs) Following other studies that examine the effect of industry conditions and peer group effects, we def digit SIC Code (Acharya et al. (2007) and Leary and Roberts (2010)). We also replicate our analysis using 2 digit SIC Code classifications and Fama French (48) industry classifications and find qualitatively similar results. 6 We examine three industry risk or distress exposure measures (IDEMs). Our first betas using stock price data, but since there is mechanical relationship between beta and leverage we use de levered industry betas. We estimate industry betas using stock return data for our sample obtained from the Center for Research in Security Prices (CRSP) monthly stock price database. Next, we calculate yearly industry betas us ing the following two factor model: ( 3 1) 6 A recent study by Hoberg and Phillips (2010) finds that more refined industry definitions based on detailed information (i.e. SEC filings) do not explain financial and investm ent choices better than SIC codes.
29 where R i,t is return on stock i in month t R m,t is the market return in month t and R I,t is the industry return in month t We obtain the market return from CRSP index files and calculate the value weighted return on the industry portfolio excluding firm i set industry returns to missing if the industry portfolio has less than 5 firms. We estimate the model above using the monthly stock return over the prior 60 months for each firm fiscal year observation. Then, we unlever both betas using the ratio of market value of equity to face value of debt plus market value of equity. We ca lculate the face value of debt for each firm by summing the total book value of short term debt and one half of the book value of long term debt (Acharya et al. (2010)). To eliminate the impact of outliers, we winsorize market and industry asset betas at t he bottom and top 2.5%. Although adjusting betas for leverage should eliminate the mechanical relationship between leverage and beta, our estimated betas may still suffer from measurement error. In order to mitigate this problem, we employ the approach su ggested by Griliches and Hausman (1986) and Acharya et al. (2010) and instrument the endogenous variable (asset betas) using a linear combination of the prior two year lags. We disentangle the effects of aggregate and industry risk by controlling for aggre gate risk in our analysis. We control for aggregate risk using a market beta that is based on the two factor model described above in equation (1). Similar to industry beta, market beta is unlevered and we instrument for it using its own lags in order to m itigate the impact of measurement error.
30 peers both in industry upturns and downturns. However, Longin and Solnik (2001) find that downside correlations are much higher than up side correlations. 7 Thus, industry beta may underestimate exposure to industry distress in the tail. In order to address the distress exposure. The first measure is the corr industry return when industry returns are less than zero. We calculate conditional correlations using up to 120 monthly returns for each firm fiscal year observation. 8 average stock return on worse industry return days. This measure is based on recent work by Acharya, Pedersen, Philippon and Richardson (2010) on systemic risk in the banking industry. Specifically, collapse of the financial sector) by the bank stock returns on the worst 5% of market outcomes. The idea is that returns on moderately bad days are informative of what would happen during extreme events. We adopt this method in our empirical analysis, worst industry return days. Specifically, using up to 120 monthly returns, we take the 7 Ang and Chen (2002) show that conditional correlations on the downside are approximately 12% higher than correlations implied by a normal distribution. 8 Since we examine the impact of conditional correlations using b oth cross sectional and panel data, we standardize returns using the mean values and standard deviations before we calculate conditional correlations.
31 worst 5% of months for the industry return for each firm fiscal year observation and compute the average return of those months for each firm fiscal year observation. 9 We also compute aggregate tail risk measures; these are computed in the same way as our industry tail risk measures except that we use the CRSP market return instead of the industry return when computing aggregate tail risk measures. Since our findings are similar whether we measure aggregate risk by conditional correlation or market MDE we focus on Market MDE. 9 We expand the calculation period because we calculate both conditional correlations and MDE condition a l on downturns and thus number of observations per year are smaller than the number of observations used to compute unconditional measures. However, we obtain similar results if compute tail risk measures using up to 60 months of returns.
32 CHAPTER 4 SUMMARY STATIS TICS In Table 4 1 we provide summary statistics f or the recovery sample. In Table 4 2 and Table 4 3, we present the summary statistics for the Dealscan sample and the lines of credit sample, re spectively. As shown in Table 4 1 the average claim size in bankruptcy is $369.5 million and the average recovery rate is 74.1%. The recovery rate is higher than 51.1% reported by Acharya et al. (2007). Their sample ends in 1999 and recent work by Bharath, Panchapegesan and Werner (2010) finds a significant decline in the duration of bankruptcy during the last two decades. Quicker exits from bankruptcy are likely to be associated with lower bankruptcy costs and thus higher recovery va lues. 1 We find that the recovery rate is 53% for unsecured claims and 97.3% for secured ones. These results are similar to the findings of Baird, Bris and Zhu (2007) concerning recovery rates for secured and unsecured creditors. Note that a relatively few (6.3%) secured claims recover less than 100% of face value. This is consistent with lenders setting advance rates and collateral coverage requirements so that conditional on distress, secured creditors are unimpaired. If advance rates and collateral cover age requirements are set based on expected loss severity, then for secured claims, credit risk spreads may not vary much with industry risk. In other words, for secured claims, 1 In our sampl e, the average (median) number of months spent in Chapter 11 is 12.55 (9) months. Consistent with the notion that quicker exits from bankruptcy are positively associated with higher recovery rates, the average recovery rate on debt claims is 76% for firms who spend less than a year in Chapter 11, whereas the average recovery rate on debt claims drops to 70% for firms who spend more than a year in Chapter 11.
33 industry risk is likely to affect lending terms primarily through advance rates rather than loan spreads. We investigate this issue later in the paper. As shown in Table 4 2 the median price on credit lines is 150 bps (over LIBOR) and average maturity of loans in the Dealscan just over 3 years (37.9 months). On average loans in the Dealscan sample have 2.7 financial covenants. In addition, 72.1% of these credit lines are secured. Later in the paper we examine the relationship between industry risk and the covenant and collateral of the loans in our sample. The median loan in our line of credit size reported by Sufi (2009) and Lins et al. (2010). As explained in Sufi (2009) firms determine their cash holding and credit line usage jointly. This joint determination creates a mechanical negative relation between any measure scaled by total assets and the availability and use of credit lines. Thus, we scale cash flows, tangible assets, net worth, and market value of assets by book value of non cash assets. We provide a detailed description of each of the variables in the Appendix. 2 As shown in Panel Table 4 3 consistent with the notion that lines of credit are an important source of liquidity, we find that 67.2% of firm years have lines of credit. In addition, we fin d that 44% (38%) of total firm liquidity comes from total (unused) lines of credit. 2 To reduce the impact of outliers, we winsorize all financial ratio s at the 2.5 st and 9 7.5 th perce ntile
34 In Table 4 4 we examine the correlation among the various IDEMs, as well as aggregate risk measures. Since the correlation matrices are similar across all samples, we rep ort the correlations for the Compustat sample (the broadest of the samples that we use). All the IDEMs are positively correlated with one another. However, conditional correlations and MDE measure are more highly correlated with each other and both are not highly correlated with industry beta. The high correlation between tail measures is not surprising since both the conditional correlations and marginal distress are computed conditional on an industry downturn. Finally, the correlations between the market beta and industry tail risk measures are quite low suggesting the industry tail risk is not simply measuring market risk. Note however that the correlation between MDE and market MDE is much higher (.61) than the correlation with market beta. This is perh aps not surprising given that low industry returns are likely to occur when the overall market is down.
35 Table 4 1. Recovery Sample Summary Statistics N Mean 1st Quartile Median 3rd Quartile Standard Deviation Recovery Rate (%) 1958 74.1 39.1 100 100 39.87 Dummy: Unsecured Claim 2371 21.00% ----Dummy: Secured Claim 2371 23.50% ----Dummy: Subordinated Claim 2371 7.30% ----Recovery Rate on Unsecured Claims (%) 383 53.11 10 47.5 100 42.5 Recovery Rate on Secured Claims (%) 465 97.3 100 100 100 13.74 Recovery Rate on Subordinated Claims (%) 134 20.19 0 0 33.65 34.49 Claim Amount 1482 369.5 1.1 25.95 214.1 1517.75 Assets Cash 300 2366.87 258.7 653.28 1925.16 5202.49 Cash/Assets 300 0.07 0.01 0.03 0.09 0.09 Market Beta 176 0.17 0.03 0.10 0.23 0.30 Industry Beta 176 0.18 0.10 0.16 0.26 0.19 Conditional Correlation 207 0.38 0.21 0.37 0.53 0.21 Marginal Distress Estimate (MDE) 184 15.00% 6.60% 13.50% 20.80% 14.00% This table presents summaries for public firms that filed bankruptcy between 1998 and 2010. Information on third party claims are obtained from reorganization plans that are reported at bankruptcydata.com. All firm financials are measured at the last fiscal year available before the bankruptcy filing date and are obtained from Compustat. All industry variables are measured in the year of bankruptcy. Variable definitions appear in the Appendix.
36 Table 4 2. Dealscan Sample Summary Statistics N Mean 1st Quartile Median 3rd Quartile Standard Deviation All in drawn (bps) 23717 179.49 75 150 255 129.26 Secured 16936 72.1% ----Assets Cash 26346 2842.58 101.29 458.66 2083.44 5909 Cash/Assets 26345 0.08 0.01 0.04 0.10 0.11 Deal Amount / Assets 26367 0.24 0.07 0.16 0.32 0.24 Deal Maturity 24349 37.9 13 36 60 22.3 Market Beta 17599 0.39 0.12 0.32 0.62 0.45 Industry Beta 17599 0.44 0.18 0.42 0.67 0.39 Conditional Correlation 19258 0.4 0.24 0.39 0.54 0.21 Marginal Distress Estimate (MDE) 18546 10.70% 5.40% 9.90% 15.50% 9.90% Number of Financial Covenants 13253 2.67 2 3 3 1.16 Covenant Intensity 26378 1.54 0 1 2 1.75 This table tabulates summaries for the Dealscan sample which includes lines of credit that are completed between 1987 and 2009. Information on lines of credit and firm financials are obtained Dealscan database and Compustat, respectively. All firm financials and industry variables are measured at the last fiscal year available before the loan completion date. Variable definitions appear in the Appendix.
37 Table 4 3. Line of Credit Sample Summary Statistics N Mean 1st Quartile Median 3rd Quartile Standard Deviation Dummy: Lines of Credit 3828 67.20% ----Total Line / (Total Line + Cash) 3713 0.44 0 0.41 0.84 0.39 Unused Line / (Unused Line + Cash) 3712 0.38 0 0.31 0.75 0.37 Cash / Assets 3713 0.24 0.04 0.13 0.37 0.26 Assets Cash 3714 2219.12 39.73 184.86 947.34 5713.65 Market Beta 3055 0.57 0.19 0.47 0.90 0.67 Industry Beta 3055 0.48 0.17 0.43 0.74 0.48 Conditional Correlation 3127 0.31 0.17 0.29 0.44 0.19 Marginal Distress Estimate (MDE) 3238 12.20% 5.10% 11.00% 18.40% 13.10% This table presents summaries for the lines of credit sample (panel data set of 348 public firms). Information on the debt structure of public firms is hand collected from 10 Ks and firm financials are obtained from Compustat. Variable definitions appear in the Appendix.
38 T able 4 4. Correlations Industry Beta Conditional Correlation Marginal Distress Estimate Market Beta Market Conditional Correlation Market Marginal Distress Estimate Stock Return Volatility Industry Beta 1.00 ------Conditional Correlation 0.26 1.00 -----Marginal Distress Estimate 0.30 0.52 1.00 ----Market Beta 0.55 0.06 0.04 1.00 ---Market Conditional Correlation 0.04 0.48 0.39 0.16 1.00 --Market Marginal Distress Estimate 0.14 0.28 0.61 0.23 0.54 1.00 -Stock Return Volatility 0.18 0.01 0.32 0.14 0.05 0.46 1.00 This table present s correlations among distress exposure measures in the Compustat Sample. Variable definitions appear in the Appendix.
39 CHAPTER 5 INDUSTRY RISK, RECOV ERY RATES AND THE LI KELIHOOD OF DISTRESS We hypothesize that if industry risk is related to the quality of collateral then we would expect that industry risk to be negatively related to recovery rates. We also investigate the extent to which industry risk measures are related and the likelihood of distress controlling for aggregate risk as well as fi rm and industry characteristics. We employ multivariate tests to address these questions. 5 .1 Industry R isk and Recovery R ates We model recovery rates as a function of firm financials prior to the bankruptcy filing, industry risk and market risk. We also control for claim size because larger claims may have higher recovery rates since a larger stakeholder may have greater bargaini ng power in bankruptcy. In addition, we control for growth in GDP because recovery rates are lower during recessions (see Scheurmann (2004) and Chen (2010)). Our recovery rate a nalysis is presented in Table 5 1 As shown in regressions (1) through (3), ne ither industry nor market beta are significantly related to recovery values. Indeed, both industry and market beta are positivel y related to recovery rates. We also test whether industry betas and market betas are jointly significant using an F test. We ca 0.64). To the extent that these risk measures are related to loan pricing and liquidity choice, the evidence in Table 5 1 suggests it is not through a relationship wit h LGD. In contrast, as shown in columns (4) and (5), our industry tail risk measures (conditional correlations and MDE) are negative and significantly related to recovery rates. The relationships are economically significant as well. For example, a one sta ndard
40 deviation increase in the conditional correlation is associated with a 3.5 % decrease in recovery rates. We also calculated conditional correlations and MDE using market returns instead of industry returns. If we observe the impact of industry distr ess exposure only at the tail, then we might also observe the impact of market risk exposure only at the tail. However, as shown in columns (6), we find no significant relationship between recovery rates and market MDE. Overall, these results indicate that our industry tail risk measures provide a better measure of the quality of collateral (in terms of recovery rates) than unconditional risk measures or market based risk measures. As shown in Table 5 1 recovery rates are significantly lower for unsecured and subordinate claims relative to senior claims (the omitted category) and secured claims. Recall that for secured claims recovery rates are generally 100%. If lenders adjust collateral coverage ex ante to account for industry risk then recovery rates on secured claims would not be expected to vary with industry risk. This argument suggests that industry driven recovery risk is likely to have the greatest impact of unsecured claims. Consistent with this argument, in untabulated results, we find that the es timated coefficient on the conditional correlation and MDE are higher for the unsecured sample (.25 and .40 respectively and coefficients are significant at the .05 level or higher). 5 .2 Industry Risk and Financial D istress: Following the literature o n financial distress, we employ four different distress definitions (Gilson, John and Lang (1990), Opler and Titman (1994), Asquisth et al. (1994) and Acharya et al. (2007)). Our first measure of industry distress is based on the median annual stock return for the industry. In particular, an industry is defined as being in distress if the median stock returns of all the firms in the same three digit SIC code
41 are less than 30% in a given year. Similarly, the firm is defined as distressed if the stock return of the firm is less than 30% in a given year. We investigate whether our industry risk measures are related to the likelihood that a firm will be in distress during an industry downturn using multivariate probit regressions. A potential concern with usin g stock returns as a measure of distress is that stock returns reflect both economic as well as financial distress and thus are likely to be systematically related to leverage. More important, since our industry risk measures are based on stock returns, th ere may be a mechanical relationship between our industry risk measures and the likelihood of distress when distress is defined by stock returns. To address these concerns, we also define alternative distress measures based on sales growth (Acharya et al. (2007)) and coverage ratio (Asquisth et al. (1994)). Since our empirical findings are not sensitive to these alternative distress measures we use, we report results using the stock return based distress measure. 1 Table 5 2 provides estimates of probit mod els in which the dependent variable is an indicator variable equal to one if the firm is in distress and its industry is in distress and zero otherwise. As shown, controlling for firm financial characteristics and GDP growth, the likelihood of firm distres s conditional on industry distress is positively and significantly related to all of the industry risk. Each measure is economically significant as well. For example, holding everything else in the model constant, a one standard deviation increase in the m arginal distress estimate (conditional correlation) is 1 The results using these alternative measures are available from the authors on request.
42 associated with a 13.1% (8.4%) increase in the likelihood of firm distress when its industry is in distress In regressions (6) and (7) of Table 5 2 we examine the relationship between firm distress an d MDE controlling for market beta and market MDE (the results are similar if we use conditional correlation as our industry risk measure). As shown we find a positive and statistically relationship between the likelihood of firm distress and MDE controlli ng for market beta and market MDE. These findings are consistent with industry risk being different from aggregate risk and an important predictor of firm distress during industry wide distress. The relationship between the likelihood of firm distress con ditional on industry distress and firm financial characteristics are consistent with finds of prior studies of financial distress. For example, consistent with Shumway (2001) we find a negative relationship between the likelihood of distress and cash flows In addition, consistent with Zmijewski (1984) we find that leverage is a strong predictor of distress. 2 2 In untabulated results, we repeat analysis using Altman (1968) control variables including retained earnings over assets, mar ket equity of total liabilities and sales of assets. variables we find qualitatively and quantitatively similar results to those reported in Table III
43 Table 5 1.Recovery Rates and industry distress exposure measures (IDEM) Dependent Variable: Recovery Rate Estimation Method: OLS (1) (2) (3) (4) (5) (6) Industry Beta 0.05 -0.11 ---(0.53) -(0.93) ---Market Beta -0.05 0.10 ----(0.75) (1.11) ---Conditional Correlation ---0.16* -----( 1.80) --MDE ----0.27** -----( 2.02) -Market MDE -----0.21 -----( 1.23) Change in GDP 0.06* 0.06* 0.06** 0.02 0.01 0.03 (1.92) (1.90) (2.01) (0.52) (0.24) (1.04) Log (Assets) 0.03 0.03* 0.03* 0.02 0.03* 0.02 (1.60) (1.70) (1.69) (1.34) (1.71) (1.34) Log (Claim) 0.04*** 0.04*** 0.04*** 0.04*** 0.04*** 0.04*** ( 8.53) ( 8.47) ( 8.15) ( 9.01) ( 7.75) ( 7.94) EBITDA / Assets 0.23 0.20 0.25 0.19 0.06 0.15 (1.32) (1.14) (1.40) (1.14) (0.32) (0.87) Tangible / Assets 0.10 0.09 0.10 0.17* 0.08 0.11 (1.04) (0.99) (1.05) (1.84) (1.03) (1.29) Long Term Debt / Assets 0.05 0.05 0.04 0.14** 0.06 0.08 ( 0.74) ( 0.80) ( 0.63) ( 2.20) ( 1.01) ( 1.24) Dummy: Unsecured Claim 0.22*** 0.22*** 0.22*** 0.23*** 0.22*** 0.23*** ( 6.06) ( 5.99) ( 5.96) ( 6.14) ( 6.02) ( 6.16) Dummy: Secured Claim 0.15*** 0.15*** 0.15*** 0.16*** 0.15*** 0.16*** (5.04) (5.05) (5.06) (5.31) (5.19) (5.35)
44 Table 5 1. Continued. Dependent Variable: Recovery Rate Estimation Method: OLS (1) (2) (3) (4) (5) (6) Dummy: Subordinated Claim 0.42*** 0.42*** 0.42*** 0.40*** 0.40*** 0.41*** ( 6.52) ( 6.49) ( 6.48) ( 5.96) ( 6.10) ( 6.66) Number of observations 701 701 701 807 737 781 Adjusted R 2 34.7% 34.7% 34.8% 36.2% 34.0% 34.4% Year Fixed Effects Yes Yes Yes Yes Yes Yes Industry Fixed Effects Yes Yes Yes Yes Yes Yes This table presents regressions that examine how market and industry distress exposure affect the recovery of claims in bankr uptcy. The period of analysis is from 1998 to 2010. Information on bankruptcies and firm financials are obtained from bankruptcydata.com and Compustat, respectively. We report the coefficient estimates and t statistics (based on standard errors clustered by firm). We use ***, **, and to denote that the co efficient estimate is different from zero at the 1%, 5% and 10% levels (two tailed), respectively. Variable definitions appear in the Appendix.
45 Table 5 2. Likelihood of firm distress during industry distress Dependent Variable: Dummy: Both firm and indus try is in distress Estimation Method Probit (1) (2) (3) (4) (5) (6) (7) Industry Beta 0.03*** -0.05*** ----(6.68) -(8.70) ----Market Beta -0.00 0.02*** --0.00 --( 0.17) (5.34) --( 1.05) -Conditional Correlation ---0.05*** ------(6.13) ---MDE ----0.11*** 0.18*** 0.04*** ----(8.54) (9.53) (2.82) Market MDE ------0.12*** ------(6.79) Change in GDP 0.10*** 0.10*** 0.10*** 0.07*** 0.06*** 0.10*** 0.06*** ( 40.96) ( 40.95) ( 40.43) ( 14.95) ( 14.17) ( 40.19) ( 14.18) LN (Assets) 0.01*** 0.01*** 0.01*** 0.00*** 0.00*** 0.01*** 0.01*** (5.05) (6.09) (4.69) (4.81) (5.14) (5.37) (5.88) EBITDA / Assets 0.21*** 0.22*** 0.21*** 0.15*** 0.13*** 0.21*** 0.12*** ( 19.11) ( 19.51) ( 18.37) ( 12.88) ( 11.87) ( 18.00) ( 11.58) Net Worth /Assets 0.02** 0.03*** 0.01 0.02*** 0.01* 0.02** 0.01* (2.10) (2.93) (1.16) (3.32) (1.89) (2.22) (1.86) Market to Book 0.01*** 0.01*** 0.01*** 0.01*** 0.00*** 0.01*** 0.00*** (4.98) (5.60) (3.94) (6.49) (5.56) (5.66) (5.60) Number of observations 20,028 20,028 20,028 24,335 23,919 19,448 23,793 Pseudo R 2 38.2% 38.0% 38.3% 42.1% 41.7% 38.4% 41.9% Year Fixed Effects Yes Yes Yes Yes Yes Yes Yes Industry Fixed Effects Yes Yes Yes Yes Yes Yes Yes This table reports the regressions that examine the likelihood of firm distress when its industry is also in distress. The in dustry is defined as distressed if the median stock returns of all the firms in the three digit SIC code of the firm is less than 30% in a given year. Similarly, the firm is defined as distressed if the stock return of the firm is less than 30% in a given year. The dependent variable is an indicator variable that is equal to one if both the firm and industry are in distress. The sam ple period is from 1999 to 2009. We report marginal effects of the coefficient estimates, as well as t statistics based on robust standard errors. We use *** ** and to denote that the coefficient estimate is significantly different from zero at the 1%, 5%, and 10% level, respectively.
46 CHAPTER 6 INDUSTRY RISK AND BA NK LOAN PRICING 6 .1 Industry Risk and L oan P ricing In the previous section we documented a significant relationship between our tail risk measures and both the likelihood of firm distress in an industry downturn and recovery rates. These findings suggest that higher industry risk, if economically important to lenders, should be associated with ex ante higher interest rates on bank loans. Moreover, industry risk should be a more important component for firms in industry that employ specific or difficult to redeploy assets. We investigate these issues using loan pricing information from Dealscan. The dependent variable in our loan pricing regression analysis is the all in drawn spread. Dealscan defines all in drawn spread as the amount the borrower pays in basis points over LIBOR for each dollar drawn down. T he All in Drawn Spread thus includes the interest cost of borrowing plus any fees associated with the line of credit and line usage. We model this spread as a function of lending standards, firm financial characteristics including size, cash flow, net wort h, market to book, tangible assets and industry cash flow volatility. These controls are motivated by previous empirical work by Strahan (1999) and others who find that large, profitable and high market to book firms, along with firms with more tangible as sets, pay less on their bank loans. We also control for other loan characteristics including the size of the credit line scaled by total assets, the natural logarithm of loan maturity and dummy variables for deal purpose. More
47 creditworthy firms are more l ikely to obtain larger loans with longer maturity. However, credit risk may also be increasing in the maturity of the loan. Thus the impact of maturity is ambiguous. 1 Our loan pricing results are presented in Table 6 1 As shown in regression (1) higher industry beta is associated with significantly higher all in drawn spreads. In terms of economic significance, one standard deviation increase in industry beta is associated with 3.5 bps increase in loan price. As shown in column (2) we find no statistica lly significant relationship between loan rates and market beta. Indeed, the coefficient estimate is negative. As shown in regression (4) and (5), firms with greater industry tail risk pay more for their credit lines. For example, a one standard deviation increase in marginal distress exposure is associated with a 7.5 bps increase in loan price. The significant relationship between the industry tail risk and loan pricing is robust to including market based risk measures. For example, as shown in column (6) beta we continue to find a positive and statistically significant relationship between all in drawn spreads and marginal distress exposure. Finally, in column (7) we report the results of the credit spread regression inc luding both industry and market MDE as explanatory variables. As shown, both tail risk measures are significantly related to credit spreads. This result is perhaps not surp ri sing 1 Loan maturity and deal size are almost certainly endogenous. However, our primary interest is in whether industry risk is related to loan pricing. Since industry risk is measured prior to the deal, it is exogenous to loan pricing. We obtain similar result s when we exclude maturity and deal size from the pricing regressions.
48 given (as shown in Table 4 4 ) industry and market MDE are positively correlat ed. More important, finding that loan rates are increasing in both aggregate and industry tail risk is consistent with Acharya et. al. (2010) argument that lenders price liquidity risk (arising from an increase in the demand for bank provided liquidity du ring in macro downturns) and in banks pricing industry risk recovery risk because LGD increases during industry downturns. Specifically these results, together with our findings concerning industry tail risk and recovery rates, are consistent with loan rat es reflecting ind ustry recovery risk. Table 6 1 also provides some insights into the relationship between all in drawn spreads and firm financials and other loan characteristics. Borrowers pay less for larger and longer term loans. In terms of firm charact eristics, not surprisingly, we find that larger, more profitable and less levered firms borrow at lower rates. Finally, greater industry cash flow volatility is associated with significantly higher spreads. 6 .2 Does the Importance of I ndustry R isk V ary wi th L oan S ecurity? Acharya et al. (2007) argue that industry driven recovery risk should matter most in the loan pricing of unsecured claims. The basic idea is that, because of the priority of their claim, unsecured creditors have a greater exposure to vari ations in asset values than secured creditors. Moreover since secured creditors can impose lower advance rates (i.e. require more collateral per dollar lent) when collateral values are more uncertain, we would expect industry risk to have the greater effec t on unsecured loan spreads. Consistent with this argument, Acharya et al. (2007) find that the recovery rates on unsecured claims (particularly senior unsecured) are significantly more sensitive to industry distress than the recovery rates on secured clai ms. Benmelech and assets matters most for the junior tranches of airline CDOs.
49 To investigate the relationship between loan pricing, collateral, assets specificity and in dustry risk, we obtained information from Dealscan on the whether or not a loan in our sample was secured. Using this information, we test whether the importance of industry risk varies with whether a loan is secured or unsecured. For brevity we report the results using MDE as a measure of industry risk and note where there results are sensitive to the industry risk measures used. As shown in Table 6 2 industry risk matters the most for unsecured loans. The effect of industry risk is almost five times as l arge for the unsecured as for secured lending. Indeed we find no significant relationship between loan rates and industry risk for secured loans in our sample. The economic impact of industry risk differs significantly between the secured and unsecured sam ples. For example, a one standard deviation increase in marginal distress estimate corresponds to a 5.7 bps increase in the all in drawn spread for unsecured loans. On the other hand, a one standard deviation increase in marginal distress estimate implies only a 1.9 bps increase for secured loans. This is consistent with the recovery risk being more important for unsecured loans because secured lenders can vary advance rates to account for differences in recovery rates. In addition we find no significant re lationship between the spreads on unsecured loans and market risk. 6.3 Does the Importance of I ndustry R isk V ary with A sset S pecificity? Previous empirical studies find that LGD is lower for firms operating in industries with more readily deployable asse ts. Ex ante, lower expected LGD should be reflected
50 in lower loan rates. We investigate this issue by sorting firms based on their industry asset specificity and then examining whether the effect of industry risk on loan pricing varies with asset specifici ty. 2 Following the literature (see Berger, Ofek and Swary (1996), Stromberg (2001) and Acharya et al. (2007)), we measure asset specificity as the ratio of the book value of machinery and equipment relative to book value of total assets. We measure industr y asset specificity as the median of specific assets of all firms in that industry over the entire sample period. 3 Columns (1) and (2) of Table 6 3 provides estimates of the loan spread regression for subsamples of unsecured loans grouped by whether the f irm operates in an industry with high (above the median) or low (below the median) asset specificity. As shown, we find a significant relationship between loan spreads and MDE only for unsecured loans of firms operating in industries with high asset specif icity. The impact of industry risk is economically significant as well. For example, a one standard deviation increase in marginal distress estimate increases the cost of unsecured loans by 8.2 bps for firms in the high asset specificity sample, whereas th e same impact increases the loan price by 2 Consistent with this argum ent, Benmelech and Bergman (2011 ) find that credit spreads are lower on debt tranches that are secured with more re deployable collateral. 3 Previous studie s use the net value of machinery and equipment (ppenme), but this variable is missing after 1999. We use the net value for the years it is available and use gross value of machinery and equipment (fate) for the rest. In unreported results, we use only gro ss values of machinery and equipment when it is available. Results are qualitatively and quantitatively similar if we use those alternative definitions
51 only 0.7 bps for firms in the low asset specificity sample. 4 This finding is consistent with argument that our industry risk measures reflect expected loss given default since as unsecured creditors are most likely to bear the brunt of the cost associated with variability in asset values in financial distress. These findings add to those of Acharya et al. (2007) and Benmelech and Berman (2009) who find that industry distress affects recovery rates more for firms wit h difficult to redeploy assets. 6.4 Does the Importance of I ndustry R isk V ary with C redit M arket C onditions? Arguably, creditors should be more concerned with industry risk during tight credit market conditions and during economic downturns because the overall likelihood of cre dit market conditions, agency problems become more acute and thus lenders are likely to place greater emphasis on the quality of collateral when making loan decisions. Consistent with this argument, Campello and Giambona (2010) find that asset redeployabil ity has a greater impact on amount of borrowing during tight credit market conditions. We examine whether the importance of industry risk for loan pricing varies during tight credit market conditions. We measure lending standards by the net percentage of banks reporting an increase in lending standards on Commercial and Industrial loans in 4 We test the equality of average industry distress exposure measure between high and low asset speci ficity samples using and F test. We reject the null hypothesis that asset specificity has no impact on average industry distress exposure measures at the .01 level.
52 5 Lown and Morgan (2006) find that net percentage of banks tightening standards is more informative of changes in bank lending than are changes in Fed Funds rates or changes in loan rates. We define economic downturns as recessions based on the NBER. We investigate the relationships between loan rates, industry risk and lending standards by dividing the sample into two groups using the sample median of tightness of lending standards. As shown in Table 6 3 in columns (3) and (4), industry distress exposure has a significantly greater impact on loan pricing during tight credit markets. For example, the coefficient on m arginal distress estimate almost doubles in tight market conditions. In terms of economic significance, a one standard deviation increase in MDE implies a 11.2 bps increase in borrowing costs during tight credit markets, while it is only 7.3 bps during loo se credit markets. Tight credit market conditions occur primarily during economic downturns. There are 47 tight credit market quarters in our sample (out of 92). Fifteen of these quarters are also recession quarters as defined by the NBER. When we divide our sample into recession and non recession quarters (not tabulated), we also find that the effect of industry distress exposure is significantly greater in recession quarters than in expansion quarters. Since our sample extends through the financial crisi s, we are able to examine the extent to which the financial crisis affec ted loan pricing. In Table 6 4 we examine 5 We measure lending standards in the quarter of the initiation date of credit lines.
53 univariate differences in loan pricing for loans originated in 2008 and after and line usage based on whether the firm year observation is 2 008 and after. The average all in drawn spread increased to 238 bps from an average of 178 bps during the pre crisis period. In terms of line of credit access we find no significant difference in the proportion of firms with lines of credit pre and post financial crisis. This finding is consistent with the survey evidences of Campello, Graham, Giambona and Harvey (2011) that find declines in access to credit lines is limited to smaller private firms. In Table 6 5 we examine whether the financial crisis that began in late 2007 had a differential impact on the cost of borrowing based on the size, profitability and industry risk exposure of the firm prior to the crisis. Greater loss aversion among lenders or supply constraints may result in higher risk pre mia for observably similar firms during a financial crisis. Consistent with this argument, as shown in Table 6 5 the size of the coefficients on our industry risk measures increase significantly for crisis years. For example, in the pre crisis period, a o ne standard deviation increase in our marginal distress estimate increases loan spread by 7.1 bps, whereas one standard deviation increase in our marginal distress estimate increases loan spread by 16.0 bps during the crisis. On the other hand, the impact of firm financial characteristics decreases significantly during the crisis. For example, in the pre crisis period, a one standard deviation increase in firm size (profitability) is associated with a 77.1 (24) bps decrease in spread, whereas the same chang e decreases loan spread by only 32.6 (21) bps during the recent financial crisis. These results suggest an overall increase in credit risk spreads during the recent financial crisis.
54 6.5 Industry R isk and T erms of L oan C ontract Previous research on the t erms of loan contracts finds collateral is more likely to be required by lenders on loans with higher default risk (see Berger and Udell (1995), Jimenez, Salas and Saurina (2006) and Booth and Booth (2006)). If firms with greater industry risk are perceiv ed by lenders as being riskier borrowers (because the LGD is higher) we would expect that loans to these borrowers are more likely to require collateral. Moreover, in the previous section we found that higher industry risk is associated with higher credit spreads only in the unsecured loan sample. This finding suggests that high industry risk borrowers may attempt to mitigate the higher cost of unsecured borrowing by borrowing on secured terms. To investigate whether the likelihood of loan security increas es with the industry distress exposure risk, we estimate Probit regressions in which the dependent variable is an indicator variable that is equal to one if the lines of credit is secured and zero otherwise. We model this variable as a function of industr y tail risk as well as firm financials and credit market conditions. We also examine the relationship between industry risk and the restrictiveness of financial covenants associated with bank loan contracts. Alternatively, lenders can use financial covenan ts. Financial covenants enable lenders to set operational or balance sheet benchmarks that borrowers must comply with. Covenants provide lenders with state contingent control rights in that conditional on default they enable lenders to modify the loan cont Chava and Roberts (2008)). Demiroglu and James (2010) show that number of financial covenants and the tightness of financial covenants depends on contracting parties' expectation on future changes in the borrowers' financials as well as the borrower's risk
55 characteristics. These arguments suggest that the restrictiveness of financial covenants may vary with the industry risk exposure of the borrowing firm. We examine the relationship betwee n the covenant structure associated with the loan and the industry risk using an index of covenant intensity developed by Bradley and Roberts (2004). Specifically, Bradley and Roberts construct a covenant intensity index that is defined as sum of the follo wing six covenant indicators: collateral, dividend restriction, having more than two financial covenants, asset sales, debt issuance and equity issuance sweeps. Since the index takes on integer values between 0 and 6, we examine the relationship between co venant intensity and industry risk using a Poisson model. Table 6 6 presents coefficient estimates for Probit and Poisson models. As shown in regressions (1), MDE has a significant impact on the likelihood of loan security. For example, holding everything else constant, a one standard deviation increase in MDE implies an 11.8% increase in the likelihood of loan security. As shown in regression (2), MDE also has significant impact on the covenant intensity index.
56 Table 6 1.Loan pricing and IDEM Dependent Variable: All in Drawn Estimation Method: OLS (1) (2) (3) (4) (5) (6) (7) Industry Beta 8.93*** -9.74*** ----(3.27) -(3.04) ----Market Beta -3.38 1.22 --3.62 --( 1.44) (0.44) --( 1.49) -Conditional Correlation ---9.44* ------(1.75) ---MDE ----76.44*** 84.62*** 47.85*** ----(6.50) (7.01) (3.72) Market MDE ------69.55*** ------(4.77) Log (Deal Maturity) 4.27*** 4.34*** 4.28*** 5.13*** 5.08*** 4.69*** 5.00*** ( 2.83) ( 2.87) ( 2.84) ( 3.52) ( 3.48) ( 3.08) ( 3.43) Deal Amount / Assets 70.22*** 70.26*** 70.17*** 69.46*** 67.85*** 67.45*** 66.76*** ( 13.20) ( 13.19) ( 13.19) ( 14.33) ( 13.47) ( 12.53) ( 13.22) Lending Standards 33.77*** 33.71*** 33.78*** 30.76*** 33.37*** 33.73*** 32.68*** (3.69) (3.67) (3.69) (3.44) (3.75) (3.65) (3.68) Ln (Assets Cash) 37.56*** 37.06*** 37.60*** 37.23*** 37.11*** 37.02*** 36.59*** ( 46.25) ( 46.70) ( 46.27) ( 48.22) ( 48.09) ( 46.26) ( 46.64) Net Worth/(Assets Cash) 107.49*** 104.85*** 107.80*** 101.67*** 105.23*** 104.72*** 104.65*** ( 22.08) ( 21.71) ( 21.72) ( 22.85) ( 22.96) ( 21.80) ( 22.94) Market to Book 9.74*** 9.15*** 9.83*** 9.41*** 9.91*** 9.08*** 9.79*** ( 6.83) ( 6.43) ( 6.90) ( 6.73) ( 6.79) ( 6.33) ( 6.69) Tangible /(Assets Cash) 7.25 8.28 7.33 5.38 6.46 7.73 7.37 (1.35) (1.54) (1.36) (1.05) (1.25) (1.45) (1.42) ROA, Cash Adjusted 233.16*** 235.24*** 232.72*** 223.28*** 227.35*** 232.46*** 225.53*** ( 19.03) ( 19.22) ( 18.96) ( 18.54) ( 17.88) ( 18.74) ( 17.67) Industry CF Volatility 2.27*** 2.30*** 2.27*** 2.14*** 2.18*** 2.25*** 2.07*** (9.10) (9.21) (9.11) (8.97) (9.02) (9.02) (8.58) Number of observations 14,274 14,274 14,274 15,474 15,119 13,800 15,051 Adjusted R 2 54.7% 54.6% 54.7% 53.9% 54.8% 55.2% 55.0% Year Fixed Effects Yes Yes Yes Yes Yes Yes Yes Industry Fixed Effects Yes Yes Yes Yes Yes Yes Yes Deal Purpose Dummies Yes Yes Yes Yes Yes Yes Yes
57 This table presents regressions that examine how industry distress exposure affects the spreads of lines of credit. Informati on on lines of credit report the coefficient estimates and t statistics (based on standard errors clustered by firm). The period of analysis is from 1987 to 2009. We use ***, **, and to denote th at the coefficient estimate is different from zero at the 1%, 5% and 10% levels ( two tailed), respectively. Variable definitions appear in the Appendix.
58 Table 6 2.Loan pricing and IDEM: Subsample Analysis Dependent Variable: All in Drawn Estimation Method: OLS Sample: Collateral No Yes MDE 78.20*** 15.67 (5.19) (1.33) Log (Deal Maturity) 7.41*** 12.01*** ( 4.16) ( 4.42) Deal Amount / Assets 14.68** 80.76*** ( 2.22) ( 12.80) Lending Standards 33.99*** 34.27** (2.69) (2.43) Ln (Assets Cash) 18.24*** 28.59*** ( 17.14) ( 22.02) Net Worth/(Assets Cash) 19.26*** 98.46*** ( 2.98) ( 17.02) Market to Book 10.64*** 8.24*** ( 6.05) ( 4.41) Tangible /(Assets Cash) 14.09*** 14.08** ( 2.58) (2.01) ROA, Cash Adjusted 60.37*** 218.26*** ( 3.07) ( 15.36) Industry CF Volatility 0.58** 1.28*** (2.37) (3.77) Number of observations 3,603 6,614 Adjusted R 2 40.4% 37.1% Year Fixed Effects Yes Yes Industry Fixed Effects Yes Yes Deal Purpose Dummies Yes Yes This table presents regressions that examine how the impact of industry distress exposure on the spreads of lines of credit varies with the existence of collateral. In order to examine how impact of industry distress exposure on the spreads of lines of credit vary with the existence of collateral, we divide the sample into two based on the existence of loan security. Information on lines of credit and firm financials are obtained estimates and t statistics (based on standard errors clustered by firm). The period of analysis is from 1987 to 2009. We use ***, **, and to denote that the coefficient estimate is different from zero at the 1%, 5% and 10% levels (two tailed), respectively. Variable definitions appear in the Appendix.
59 Table 6 3. Loan prici ng and IDEM: Subsample Analysis Dependent Variable: All in Drawn Estimation Method: OLS Sample: Unsecured Loans Asset Specificity Lending Standards Low High Loose Tight MDE 8.17 123.65*** 98.56*** 135.70*** (0.36) (6.64) (4.38) (6.68) Log (Deal Maturity) 10.49*** 6.96*** 11.27*** 4.69** ( 3.37) ( 3.18) ( 6.12) ( 2.43) Deal Amount / Assets 32.11*** 4.05 28.36*** 28.90*** ( 2.85) ( 0.50) ( 3.66) ( 2.82) Ln (Assets Cash) 21.18*** 17.39*** 26.87*** 26.78*** ( 13.20) ( 13.05) ( 24.69) ( 21.22) Net Worth/(Assets Cash) 4.34 36.45*** 67.91*** 69.12*** ( 0.45) ( 4.51) ( 8.72) ( 8.04) Market to Book 7.83*** 12.01*** 10.73*** 7.99*** ( 2.69) ( 5.49) ( 4.61) ( 3.53) Tangible /(Assets Cash) 13.25 21.46*** 2.59 8.26 (1.20) ( 3.56) (0.38) ( 1.09) ROA, Cash Adjusted 79.25** 68.18*** 101.93*** 190.52*** ( 2.51) ( 2.66) ( 4.00) ( 7.00) Industry CF Volatility 0.36 0.98*** 1.60*** 1.87*** (0.83) (3.48) (5.97) (5.57) Number of observations 1,096 2,552 3,911 4,594 Adjusted R 2 44.2% 40.8% 44.8% 47.8% Year Fixed Effects Yes Yes Yes Yes Industry Fixed Effects Yes Yes Yes Yes Deal Purpose Dummies Yes Yes Yes Yes This table presents regressions that examine how the impact of industry distress exposure on the spreads of lines of credit varies with the asset specificity of the industry for unsecured loans. In order to examine how impact of industry distress exposure on spreads of line of credits varies with the asset specificity of the industry, we divide the unsecured loan sample into two (high and low) using the sample median of industry asset specificity. In order to examine how impact of industry of distress exposure on spreads of line of credits varies across credit market conditions, we divide the un secured loan sample into two (high and low) using the sample median of lending Dealscan database and Compustat, respectively. We report the coefficient estimates and t st atistics (based on standard errors clustered by firm). The period of analysis is from 1987 to 2009. We use ***, **, and to denote that the coefficient estimate is different from zero at the 1%, 5% and 10% levels (two tailed), respectively. Variable defin itions appear in the Appendix.
60 Table 6 4. Loan pricing and IDEM: Financial Crisis of 2008 Financial Crisis of 2008 Before Crisis Dealscan Sample N Mean Median N Mean Median All in Drawn 694 237.50 *** 225 *** 23023 177.70 150 Deal Amount / Assets 750 0.19 *** 0.13 *** 25617 0.24 0.16 Deal Maturity 741 37 36 23608 37.98 36 Secured 562 74.20% 16374 72.00% Assets Cash 750 4302.5 *** 1078.49 *** 25596 2799.8 445.80 ROA, Cash Adjusted 750 0.13 0.13 25416 0.13 0.13 Line of Credit Sample Dummy: Lines of Credit 696 65.90% -3132 67.50% -Total Line / (Total Line + Cash) 620 0.43 0.43 3093 0.44 0.41 This table presents how terms and usage of lines of credit vary between before and during the financial crisis of 2008. In order to examine how impact of industry distress exposure on terms and usage of lines of credit varies during the financial crisis of 2008, we divide Dealscan sample into two based on whether loans are originated in 2008 and after, and line of credit sample into two based on whether the firm year observations are in 2008 or after. Information on lines of credit and firm financials are o public firms is hand collected from 10 Ks. The period of analysis is from 1987 to 2009 for Dealscan sample, and the period analysis is from 1999 to 2009 for line of credit sample. We test the null that the means (medians) during and before financial crises are equal using t tests (Wilcoxon rank sum tests).We assume unequal variances for t tests. We use ***, **, and to denote that the null is rejected at the 1%, 5%, and 10% level, respectively. Variable definitions appear in the Appendix.
61 Table 6 5 Loan pricing and IDEM: Financial Crisis of 2008 Dependent Variable: All in Drawn Estimation Method: OLS Loans Originated: 2008 2009 1987 2007 MDE 187.05*** 71.73*** (3.33) (6.20) Log (Deal Maturity) 2.87 6.90*** (0.36) ( 4.80) Deal Amount / Assets 43.45* 64.16*** ( 1.73) ( 12.60) Ln (Assets Cash) 19.55*** 37.60*** ( 5.85) ( 49.06) Net Worth/(Assets Cash) 100.59*** 104.35*** ( 4.45) ( 23.06) Market to Book 23.39*** 9.47*** ( 3.20) ( 6.54) Tangible /(Assets Cash) 38.65* 4.63 (1.77) (0.89) ROA, Cash Adjusted 189.20*** 231.42*** ( 2.94) ( 18.63) Industry CF Volatility 1.51 2.21*** (1.31) (9.07) Number of observations 591 15,291 Adjusted R 2 51.8% 54.3% Year Fixed Effects Yes Yes Industry Fixed Effects Yes Yes Deal Purpose Dummies Yes Yes This table reports the coefficient estimates and t statistics (based on standard errors clustered by firm). The period of analysis is from 1987 to 2009. We use ***, **, and to denote that the coefficient estimate is different from zero at the 1%, 5% and 10% levels (two tailed), respectively. Variable definitions appear in the Appendix.
62 T able 6 6 Terms of Loan Contract and IDEM Dependent Variable: Dummy: Loan is secured Covenant Intensity Estimation Method: Probit Poisson MDE 0.38*** 0.24** (4.79) (2.33) Log (Deal Maturity) 0.06*** 0.26*** (5.79) (16.33) Deal Amount / Assets 0.11*** 0.03 ( 3.06) ( 0.69) Lending Standards 0.08 0.14 (1.23) (1.61) Ln (Assets Cash) 0.17*** 0.16*** ( 24.80) ( 22.17) Net Worth/(Assets Cash) 0.43*** 0.42*** ( 14.48) ( 12.59) Market to Book 0.05*** 0.12*** ( 4.74) ( 9.42) Tangible /(Assets Cash) 0.04 0.13*** (1.06) ( 2.81) ROA, Cash Adjusted 0.84*** 0.08 ( 9.14) (0.96) Industry CF Volatility 0.01*** 0.01*** (6.42) (3.19) Number of observations 10,562 16,137 Adjusted R 2 35.3% 20.2% Year Fixed Effects Yes Yes Industry Fixed Effects Yes Yes Deal Purpose Dummies Yes Yes This table presents regressions that examine how industry distress exposure affects the collateral and covenant structure in lines of credit. Information on lines of credit and firm For the probit estimation, we report marginal effects of the coefficient estimates, as well as t stati stics based on robust standard errors. For the poisson estimation, we report the coefficient estimates, as well as t statistics based on robust standard errors. The period of analysis is from 1987 to 2009. We use ***, **, and to denote that the coefficie nt estimate is different from zero at the 1%, 5% and 10% levels (two tailed), respectively. Variable definitions appear in the Appendix.
63 CHAPTER 7 INDUSTRY RISK AND LI QUIDITY CHOICE 7 .1 Industry R isk and L iquidity M anagement The evidence in the previous sections suggests that greater industry exposure is associated with lower debt recovery rates during industry downturns and that creditors build this into the price of credit ( pricing channel ). The pricing of credit lines, in turn, should determin e how much the firm relies on cash versus lines of credit for liquidity management purposes. Creditors may also respond to industry downturns by denying the firm credit or ore contingent access to bank credit during downturns by increasing their cash holding ( availability channel ). In particular, Sufi (2009) finds that the use of cash flow based covenants by bank lenders leads to lines of credit being a close substitute for cash only for firms with high expected future cash flows. Moreover, Chava and Roberts (2008) and Nini, Smith and Sufi (2009) find that technical covenant violations lead to tighter lending terms and less availability. As a result, cash flow covenants toget her with net worth and borrowing base covenants that tie credit availability to firm cash flows and the value of collateral may make lines of credit a poor substitute for cash reserves for firms with significant industry risk exposure. Given our the findin gs that covenant intensity is substitute for cash for firms with higher industry risk. Finally, cash holdings may be related to industry risk because firms with great er industry risk exposure hold more precautionary cash balances. In particular, firms with greater industry risk exposure have less liquid operating assets and thus may choose to
64 hold more cash as a hedge against operating cash flow shocks (see Shleifer an d Vishny (2011) for a discussion of the impact of potential fire sale discounts on cash holdings). In this section, we address these questions by examining whether the proportion of bank provided liquidity to total liquidity and the ratio of cash to asset s are related to we measure industry risk by MDE or conditional correlation, for brevity we report our results only using MDE. 7 .2 Cash vs. L ines of C redit We measure b ank define total liquidity as the sum of cash and bank liquidity. The bank liquidity ratio equals zero for firms with no lines of credit. Since the bank liquidity variable is left truncated, w e examine the determinants of the bank liquidity ratio using a one sided Tobit regression. We estimate the regressions using control variables including lending standards, firm size, cash flow, net worth, market to book, tangible assets, industry cash flow volatility, age and year and industry fixed effects. Credit market conditions may also affect the access to and utilization of credit lines, especially for smaller firms (Demiroglu, James and Kizilaslan (2012)). Our findings are presented in Table 7 1 Si nce we are using panel data, we cluster standard errors by firm in all of the regressions to account for within firm correlation of regression residuals. As shown in column (1), controlling for firm financials and credit market conditions, firms rely more on cash and less on credit lines as MDE increases. Since Acharya et al. (2010), find that firms with higher market betas hold more cash Acharya et al., we find a posit ive relationship between cash holdings and market beta.
65 However, controlling for aggregate risk exposure, we continue to find a positive and significant relationship between cash holdings and our industry tail risk measures. The findings reported in Table 7 1 also provide some insights into the relationship between liquidity choice and firm financial characteristics. Consistent with the findings of previous empirical studies, we find older firms and firms with fewer growth opportunities rely more heavily on credit lines as a source of liquidity (Sufi (2009)). Note that the coefficient on cash flow has the expected sign but is not statistically significant when we use total lines of credit to define bank liquidity. The lack of a significant relationship betwe en line usage and cash flows appears to because our sample period is different In Table 7 2 we divide our sample into two based on whether the firm year observation is before or after 2008. As shown in regression (1) and (2) industry distress exposure measure is negative and statistically significant in both the pre crises period and during th e crises. While the absolute value of the point estimate is on MDE is higher during the crisis period, the coefficient estimates for the pre and post crisis periods are not significantly different from one another. 7.3 Industry R isk and C ash H oldings In t he previous section we showed that firms with greater industry risk exposure rely more heavily on cash relative to lines of credit for their liquidity needs. We next examine whether these firms also hold higher cash reserves as a form of protection from li quidity shocks. Following the literature on cash holdings, we measure cash reserves as the ratio of cash to the book value of assets (see for example Almeida et al. (2004) and Bates, Kahle and Stulz (2009)). If firms with potentially less liquid operating
66 assets face potentially greater external financing frictions, we would expect these firms to hold more precautionary cash balances. Our regression model assumes that cash holdings are a function of firm characteristics, credit market conditions and industr y distress exposure measures. As shown in column (3) of Table 7 1 firms with high MDE hold significantly more cash relative to total assets. In column (4), we include both MDE and market risk together and continue to find a negative and statistically sign ificant relationship between cash holdings and both the industry MDE and market beta. 1 One potential concern with the analysis thus far is that our IDEMs, as well as our cash flows or operating performance. This is plausible since overall volatility in returns may affect the likelihood of distress and the external financing costs that a firm may face. Moreover, Bates et al. (2009) find that the increase in cash holdings s ince 2000 is related to an increase in the volatility of cash flows and an increase in idiosyncratic risk. To address this issue, in the regressions reported in Table 7 1 we control for industry cash flow volatility, which should be related to the volatil flows and is a standard control variable used in the literature on cash holdings. As a further robustness check, we compute the volatility of individual firm monthly returns over the prior 60 months as well as the volatilit y of cash flows (using yearly data for the 1 In untabulated results, we control for the Bank MDE. Bank MDE is equal to the average of monthly returns in the worst 5% of bank return months After controlling for Bank MDE, we continue to find a negative and significant relationship between MDE and bank liquidity ratio.
67 prior 10 years). After controlling for the volatility of individual firm returns and cash flows, we continue to find a negative and significant relationship between both industry as well as aggregate risk measures
68 T able 7 1 Liquidity Management and IDEM Dependent Variable: Total Line / (Total Line + Cash) Cash / Assets Estimation Method: Tobit OLS MDE 0.37*** 0.39*** 0.17*** 0.19*** ( 3.99) ( 3.27) (4.63) (4.56) Market Beta -0.07*** -0.03*** -( 3.01) -(3.97) Lending Standards 1.26 1.79** 0.13 0.16 ( 1.60) ( 2.36) (0.63) (0.63) Ln (Assets Cash) 0.00 0.00 0.02*** 0.02*** ( 0.47) ( 0.42) ( 5.63) ( 5.33) Net Worth/(Assets Cash) 0.21*** 0.19*** 0.10*** 0.10*** ( 4.22) ( 4.01) (11.89) (11.64) Market to Book 0.05*** 0.05*** 0.02*** 0.02*** ( 5.27) ( 4.70) (7.54) (6.94) Tangible /(Assets Cash) 8.29*** 8.11*** 0.97 0.85 ( 3.76) ( 3.65) ( 1.34) ( 1.16) ROA, Cash Adjusted 0.07 0.05 0.01 0.01 (1.49) (1.22) ( 1.17) ( 0.87) Industry CF Volatility 0.02*** 0.02*** 0.01*** 0.01*** ( 6.24) ( 5.87) (6.60) (6.15) Log (Age) 0.05* 0.05 0.02** 0.02 (1.86) (1.54) ( 2.01) ( 1.63) Number of observations 3,227 2,988 3,227 2,988 Pseudo R 2 / Adjusted R 2 37.9% 37.7% 69.6% 70.4% Year Fixed Effects Yes Yes Yes Yes Industry Fixed Effects Yes Yes Yes Yes This table reports regressions that examine the impact of industry distress exposure measures on the liquidity management of the firms. Information on debt structure of public firms is hand collected from 10 Ks. The period of analysis is from 1999 to 2009. The financials of the firms are obtained from Compustat. We report the coefficient estimates and t statistics (based on standard errors clustered by firm). We use ***, **, and to denote that the coefficient estimate is different from zero at the 1%, 5% and 10% levels (two tailed), respectively. Variable definitions appear in the Appendix.
69 Table 7 2. Liquidity Management and IDEM: Financial Crises of 2008 Dependent Variable: Total Line / (Total Line + Cash) Cash / Assets Estimation Method: Tobit OLS Sample 2008 2009 1999 2007 2008 2009 1999 2007 MDE 0.40** 0.37*** 0.20*** 0.16*** ( 2.10) ( 4.07) (3.59) (4.25) Lending Standards 0.05* 0.21*** 0.01 0.06*** (1.73) (6.25) (0.69) ( 4.54) Ln (Assets Cash) 0.00 0.00 0.02*** 0.02*** ( 0.30) ( 0.42) ( 5.26) ( 4.79) Net Worth/(Assets Cash) 0.22*** 0.21*** 0.10*** 0.10*** ( 4.17) ( 3.94) (8.43) (10.56) Market to Book 0.05** 0.05*** 0.02*** 0.02*** ( 2.48) ( 5.26) (4.56) (6.54) Tangible /(Assets Cash) 6.77** 7.95*** 1.13 1.01 ( 2.08) ( 3.49) ( 0.94) ( 1.44) ROA, Cash Adjusted 0.01 0.09* 0.00 0.02* (0.11) (1.96) (0.24) ( 1.68) Industry CF Volatility 0.02*** 0.02*** 0.01*** 0.01*** ( 5.72) ( 5.96) (5.32) (6.82) Log (Age) 0.09** 0.03 0.02 0.02* (2.03) (1.28) ( 1.23) ( 1.81) Number of observations 600 2,627 934 2,293 Pseudo R 2 / Adjusted R 2 36.7% 37.0% 68.2% 70.1% Year Fixed Effects Yes Yes Yes Yes Industry Fixed Effects Yes Yes Yes Yes This table reports regressions that how the impact of industry distress exposure measures on the liquidity management of the firms varies during the financial crises of 2008. We divide line of credit sample into two based on whether firm year observations are in 2008 and after. Information on debt structure of public firms is hand collected from 10 Ks. The period of analysis is from 1999 to 2009. The financials of the firms are obtained from Compustat. We report the coefficien t estimates and t statistics (based on standard errors clustered by firm). We use ***, **, and to denote that the coefficient estimate is different from zero at the 1%, 5% and 10% levels (two tailed), respectively. Variable definitions appear in the Appe ndix.
70 CHAPTER 8 CONCLUSION In this paper, we calculate several potential industry risk measures and examine their relationship to recovery rates in financial distress and the likelihood of financial distress. We find that industry tail risk measures are significantly related to both the likelihood that a firm becomes distressed during an industry downturn and what reliance on bank lines of credit as a source of liquidity are significantly related to our industry risk measures. Our results are robust to including aggregate risk measures and other controls. Our findings are consistent with the argument that greater industry risk is associated with higher LGD which, in turn, leads to higher credit spreads on bank loans. Moreover, higher industry risk may affect the conditionality of lines of credit because, conditional on distress, high industry risk borrowers may be forced to renegotiate their lines when industry conditions are unfavorable and collateral values are low. Consistent with this argument, we find that firms with greater industry risk exposure rely more heavily on cash and less on bank lines of credit as a source of liquidity. Overall, these results suggest that t he potential for fire sale discounts affects that ex ante pricing and structure of bank loans. To the best of our knowledge, ours is the affecting loan pricing and f irm liquidity choices.
71 APPENDIX: VARIABLE DEFINITIONS Industry Beta = the asset (unlevered) industry beta, calculated from the equity (levered) industry beta. Equity industry beta is obtained from a two factor model in which firm return is regressed on market return (value weighted return of CRSP universe vwretd) and in dustry return (value weighted return of all the firms in the three digit SIC code of the firm). The asset industry beta is obtained by multiplying equity industry beta with equity / (debt + equity) where equity is the market capitalization of the firm and debt is equal to the sum of short term debt and long term debt multiplied by 0.5. Conditional Correlation = correlation between the normalized stock return of the firm and the normalized return on the value weighted portfolio of all the firms in the same 3 digit SIC code of the firm given the normalized return on the value weighted portfolio of all the firms in the same 3 digit SIC is less than zero. Marginal Distress Estimate (MDE) = average stock return of the firm given that the return on the value wei ghted portfolio of all the firms in the same 3 digit SIC code of the firm is below its 5 th percentile. Market Beta = the asset (unlevered) market beta, calculated from the equity (levered) market beta. Equity market beta is obtained from a two factor mode l where firm return is regressed on market return (value weighted return of CRSP universe vwretd) and industry return (value weighted return of all the firms in the three digit SIC code of the firm). The asset market beta is obtained by multiplying equit y market beta with equity / (debt + equity) where equity is the market capitalization of the firm and debt is equal to the sum of short term debt and long term debt multiplied by 0.5. Market Marginal Distress Estimate (MDE) = average stock return of the firm given that market return (value weighted return of CRSP universe vwretd) is below its 5 th percentile. Total Line / (Total Line + Cash) = total amount of lines of credit / total amount of lines of credit +cash (che) Unused / (Unused Line + Cash) = un used amount of lines of credit / unused amount of lines of credit + cash (che) Cash / Assets = che / at Debt /Assets = (dlc + dltt) / at
72 Net Debt /Assets = (dlc + dltt che) / at LN (Assets Cash) = ln(at che) Net Worth/(Assets Cash) = ceq /(at che) Market to book = (at ceq + prcc_f*csho) / at Tangible/(Assets Cash) = ppent/(at che) ROA, Cash Adjusted = oibdp / (at che) Industry Cash Flow Volatility = the median 10 year moving standard deviation of cash flow over lagged assets (oibdp(t) / at(t 1)) of all the firms in the 3 digit SIC of the firm. LN (age) = Log (current fiscal year (fyear) first fiscal year of available accounting data (year1)) Asset Specificity = the median ratio of machinery and equipment (net = ppenme, gross = fate) to total as sets, using one digit SIC code over sample period. Lending Standards = the percentage of (large) banks tightening lending standards in the All in drawn = spread paid by firms over L IBOR for each dollar drawn down from the loan. Deal Amount / Assets = ratio of loan amount to total assets. LN (Deal Maturity) = natural logarithm of the loan maturity reported in Dealscan. Covenant Intensity Index = sum of the following six covenant indic ators: collateral, dividend restriction, having more than two financial covenants, asset sales, debt issuance and equity issuance sweeps. Recovery Rate = the estimated recovery rate on the claim. LN (Claim) = the natural logarithm of the amount of the clai m. LN (Assets) = ln(at) EBITDA / Assets = oibdp/at Tangible / Assets = ppent/at Long Term Debt / Assets = dltt /at Industry MTB = the median of the ratio of market value of the firm (book value of assets (at) book value of equity (ceq) + market value of equity (prcc_f*csho)) to book value of assets of all the firms in the 3 digit SIC.
73 Change in GDP = annual GDP percent change ba sed on 2005 dollars, obtained from www.bea.gov ).
74 LIST OF R EFERENCE S Acharya, V., S. Bharath, and A. Srinivasan, 2007, Does industry wide distress affect defaulted firms? Evidence fr om creditor recoveries, Journal of Financial Economics 85, 787 821. Acharya, V., L. H. Pedersen, T. Philippon, and M. Richardson, 2010, Measuring systemic risk, Working Paper, NYU Stern. Acharya, V., H. Almeida, and M. Campello, 2010, Aggregate risk and the choice between cash and lines of credit, Working Paper, University of Illinois at Urbana Champaign. Almeida, H., M. Campello, and M. S.Weisbach, 2004, The cash flow sensitivity of cash, Journal of Finance 59, 1777 1804. Altman, E. I., 1968, Financial ratios, discriminant analysis and the prediction of corporate bankruptcy, Journal of Finance 23, 589 609. Altman, E. I., B. Brady, A. Resti, and A. Sironi, 2005, The link between default and recovery rates: Theory, empirical evidence and implications Jour nal of Business 78, 2203 2227. Altman, E., 2009, Default Recovery Rates and LGD in Credit Risk Modeling Practice, Advances in Credit Risk Modeling and Corporate Bankruptcy Prediction, ed. by S. Jones and D. Hensler, Cambridge University Press, Cambridge, England, 2009. Ang, A., and J. Chen, 2002, Asymmetric correlations of equity portfolios, Journal of Financial Economics 63, 443 494. Asquith, P., R. Gertner, and D. Scharfstein, 1994, Anatomy of financial distress: An examination of junk bond issuers, Quar terly Journal of Economics 109, 625 658. Ayotte, K. M., and E.R. Morrison, 2009, Creditor control and conflict in Chapter 11, Journal of Legal Studies 1, 511 551. Bates, T.W., K. M. Kahle, and R. Stulz, 2009, Why do U.S. firms hold so much more cash than t hey used to?, Journal of Finance 64, 1985 2021. Baird, D., A. Bris, and N. Zhu, 2007, The dynamics of large and small Chapter 11 cases: An empirical study, Working Paper, University of Chicago. Benmelech E., and N. K. Bergman, 2009, Collateral pricing, Jo urnal of Financial Economics 91, 339 360. Benmelech E., and N. K. Bergman, 2011, Bankruptcy and the collateral channel, Journal of Finance 66, 337 378.
75 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 firms, Journal of Business 68, 351 382. Berger, P., E. Ofek, and I. Swary, 1996, Investor valuation of the abandonment option, Journal of Financial Economics 42, 257 287. Bernanke, B., and Gertler M., Inside the black box: The credit channel of monetary policy transmission, Journal of Economic Perspectives 9, 27 48. Bharath,S. T., V. Panchapegesan, and I.M. Werner, 2010, The changing nature of Chapter 11, Working Paper, Ohio State University Booth, J. R., and L. C. Booth, 2006, Loan collateral decisions and corporate borrowing costs, Journal of Money, Credit and Banking 38, 67 90. Bradley, M., and M. Roberts, 2004, The structure and pricing of corporate debt cov enants, Working Paper, Duke University. Brown D., C. James, and B. Mooradian, 1994, The information content of Distressed Restructuring involving public and private debt claims, Journal of Financial Economics 33, 93 118. Campello, M., E. Giambona, J. Gra ham, and C. R. Harvey, 2011, Liquidity management and corporate investment during the crisis, Review of Financial Studies 24, 1944 1979. Campello, M., and E. Giambona, 2010, Capital structure and the redeployability of tangible assets, Working Paper, Unive rsity of Illiniois at Urbana Champaign. Chava, S., and M.R. Roberts, 2008, How does financing impact investment? The role of debt covenants, Journal of Finance 63, 2085 2121. Chen, H., 2010, Macroeconomic conditions and the puzzles of credit spreads and ca pital structure, Journal of Finance 65, 2171 2212. Demiroglu, C., C. James and A. Kizilaslan, 2012, Bank lending standards and access to line of credit Journal of Money, Credit and Banking forthcoming. Demiroglu, C., and C. James, 2010, The information c ontent of bank loan covenants Review of Financial Studies 23, 3700 3737 Fama, E., and K. R. French, 2002, Testing tradeoff and pecking order predictions about dividends and debt, Review of Financial Studies 15, 1 33.
76 Gilson, S. C., K. John, and L. Lang, 1 990, Troubled debt restructurings: An empirical investigation of private reorganization of firms in default, Journal of Financial Economics 27 315 353. Griliches, Z., and J. Hausman, 1986, Errors in variables in panel data, Journal of Econometrics 31, 93 118. Gupton, G., D. Gates, and L. Carty, 2000, Bank loan loss given default, Moody's Special Comment November, 69 92. Hu, Y. and W. Perraudin, 2002, The dependence of recovery rates and default, Mimeo, BirBeck College Hoberg, G., and G. Phillips, 2010, D ynamic product based industry classifications and endogenous product differentiation, Working Paper, University of Maryland. James,C., and A. Kizilaslan, 2012, Asset Specificity, Industry Driven Recovery Risk and Loan Pricing, University of Florida, Worki ng Paper. Jimenez, G., V. Salas, and J. Saurina, 2006, Determinants of collateral, Journal of Financial Economics 81, 255 281. Leary M. T., and M.R. Roberts, 2010, Do peer firms affect corporate financial policy?, Working Paper, Wharton. Lins, K. V., H. S ervaes, and P. Tufano, 2010, What drives corporate liquidity? An international survey of cash holdings and lines of credit, Journal of Financial Economics 98,160 176. Longin F., and B. Solnik, 2001, Extreme correlation of international equity markets, Jour nal of Finance 56, 649 676. Lown, C., and D. Morgan, 2006, The credit cycle and the business cycle: New findings using the loan officer opinion survey, Journal of Money Credit and Banking 38, 1575 97. Nini, G., D. C. Smith, and A. Sufi, 2009, Creditor cont rol rights and firm investment policy, Journal of Financial Economics 92, 400 420. Opler, T., and S. Titman, 1994, Financial distress and corporate performance, Journal of Finance 49, 1015 1040. Pulvino, T. C., 1998, Do fire sales exist? An empirical inve stigation of commercial aircraft transactions, Journal of Finance 53, 939 978. Rauh, J. and A. Sufi, 2012, Explaining corporate capital structure: Product markets, leases and asset similarity, Review of Finance 16, 115 155.
77 Saunders, A., and L. Allen, 2010 Credit risk management in and out of the financial crisis: New approaches to value at risk and other paradigms, 3rd edition, John Wiley and Sons. Schuermann, T., 2004, What do we know about loss given default?, Credit Risk: Models and Management, 2 nd Edition, Risk Books. Shleifer, A., and R. Vishny, 2011, Fire sales in Finance and Macroeconomics, Journal of Economic Perspectives 25, 29 48. Shleifer, A., and R. Vishny, 1992, Liquidation values and debt capacity: A market equilibrium approach, Journal of Finance 47, 1343 1366. Shumway, T., 2001, Forecasting bankruptcy more accurately: A simple hazard model, Journal of Business 74, 101 124. Strahan, P., 1999, Borrower risk and the price and nonprice terms of bank loans, Working Paper, New York Fed. Strom berg, P., 2001, Conflict of interest and market liquidity in bankruptcy auctions: Theory and tests, Journal of Finance 55, 2641 2691. Sufi, A., 2009, Bank lines of credit in corporate finance: An empirical analysis, Review of Financial Studies 22, 1057 108 8. Williamson, O., 1988, Corporate finance and corporate governance, Journal of Finance 43, 567 591. Zmijewski, M. E., 1984, Methodological issues related to the estimation of financial distress prediction models, Journal of Accounting Research 22, 59 82.
78 BIOGRAPHICAL SKETCH Atay Kizilaslan is a Ph.D. student at the University of Florida. In March 2012, he started to work as an Associate at Cornerstone Research in New York City. He holds an M.A. in economics from Georgia Institute of Technology and B. A. in Business Administration from Middle East Technical University in Ankara. Atay's primary research interests include loan contracting, financial institutions and asset backed securities.