Priority Spreading of Corporate Debt *

Transcription

1 Priority Spreading of Corporate Debt * Dominique C. Badoer Robert J. Trulaske, Sr. College of Business University of Missouri Columbia, MO, badoerd@missouri.edu (573) and Evan Dudley Queen s University Kingston, ON K7L 3N6 evan.dudley@queensu.ca and Christopher James # Warrington College of Business Administration University of Florida Gainesville, FL christopher.james@warrington.ufl.edu (352) First draft: October 2016 This draft: July 2017 Debt * A prior version of this paper was titled Volatility and the Priority Structure of Corporate # Corresponding author.

2 Abstract Priority spreading refers to the practice of firms increasing their reliance on secured and junior subordinated debt and reducing their reliance on senior debt as their credit quality deteriorates. We argue that priority spreading occurs, in part, because security provides creditors with greater protection from dilution than covenants that prioritize payments and limit subordination among unsecured creditors. Consistent with this argument, we find that absolute priority rules are more frequently violated among unsecured creditors than among secured creditors. We use volatility as a measure of the potential for dilution and find that increases in the volatility of firms assets are associated with increases in the cost of senior debt relative to both subordinated and secured debt. Consistent with security providing greater protection from dilution than covenants, we also find that exogenous increases in volatility result in greater priority spreading; with firms increasing their reliance on both secured and subordinated debt and reducing their reliance on senior unsecured claims. JEL Classification: G32, G33 Keywords: Debt structure, volatility, leverage, product-market competition, bankruptcy.

3 I. Introduction Firms establish priority among creditors in two basic ways; through granting certain creditors rights to collateral and through covenants that prioritize payments and limit the subordination of unsecured creditor classes. Firms often use both ways to create priority structures. For example, when Caesars Entertainment filed for bankruptcy in 2014, it had first lien and second lien secured debt as well as several classes of senior unsecured debt and subordinated debt outstanding. Rauh and Sufi (2010) document that this type of heterogeneity in priority structure is not unusual, particularly for firms rated BBB or below. Moreover, they find that firms engage in priority spreading following large credit downgrades, with firms increasing their reliance on secured and junior subordinated claims and reducing their reliance on senior unsecured debt. In this paper, we investigate why firms use different ways to prioritize debt claims and provide an explanation for priority spreading based on the effectiveness of different mechanisms for establishing priority. A commonly cited reason for prioritizing creditors claims is to mitigate the potential dilution of existing creditors arising from increases in leverage, the issuance of higher priority claims, or asset substitution (the substituting of riskier assets for safer assets). 1 Drawing on the legal literature on debt priority, we argue that security affords creditors greater protection against dilution than covenants that prioritize payments. As a result, we hypothesize that increases in the potential for dilution of unsecured creditors increase the cost of senior debt relative to secured debt which, in turn, incentivizes firms to shift towards using security as opposed to subordination covenants as way of prioritizing their debt claims. As discussed in detail later, we assume that the potential for dilution is increasing the volatility of the return on firm assets. As a result, if security provides greater protection from dilution we expect increases in asset volatility to be associated with increases the relative cost of senior unsecured debt and a reduction in the relative cost of secured debt. We also expect the cost of subordinated debt relative to senior debt to decline as asset volatility increases. The cost of subordinated debt declines because subordinated creditors hold an option to dilute senior unsecured creditors and the value of this dilution option is increasing in asset volatility. Taken 1 There is an extensive literature on how shareholders can dilute existing creditors. For example, Barclay and Smith (1995), Smith and Warner (1979), Myers (1977) and Bradley and Roberts (2015). As we discuss later in the paper, priority may also affect creditors incentives to monitor and economize on monitoring costs. See Winton (1995) and Park (2000). 1

4 together, we hypothesize that increases in the potential for dilution of unsecured creditors increase the cost of senior unsecured debt relative to secured and junior unsecured debt which, in turn, provides an incentive for firms to engage in priority spreading. As a result, we expect increases in asset volatility to be associated with increases in firms reliance on secured debt and subordinated claims and reductions in their reliance on senior unsecured debt. We refer to this as the priority spreading hypothesis. We begin our empirical analysis by investigating corporate debt restructurings, both inside and outside of bankruptcy. Specifically, we examine the recovery rates of different creditor classes as well as the frequency of violations of absolute priority rules (APR) in debt restructurings. 2 We analyze recovery rates because one alternative explanation for priority spreading is that it enhances the incentives for potentially impaired secured creditors to monitor. 3 However, contrary to this alternative explanation we find that secured creditors are rarely impaired. Specifically, we find that restructurings almost always involve secured creditors recovering 100% of their claims when there are also some more junior claims outstanding. In contrast, senior unsecured creditors rarely fully recover when there are more junior claims outstanding. Moreover, consistent with security providing greater protection from dilution, we find that restructuring outcomes involving departures from absolute priority rules (APR) are significantly more likely for senior unsecured creditors than for secured creditors even when both creditor classes have similar levels of junior debt below them. Next, we examine the impact of changes in asset volatility on the relative cost of senior unsecured debt. We begin by analyzing yield spreads on senior and subordinated bonds in the secondary market. We define volatility as the volatility of the firm s asset returns measured using Bharath and Shumway (2008) s methodology. The effect of innovations in asset volatility on the yields of senior and subordinated bonds will vary with the extent to which absolute priority rules are expected to be violated. The traditional contingent claims approach to valuing risky debt posits that the value of risky debt is equal to the value of a riskless bond less the value of the default 2 The absolute priority rule is based on the fair and equitable requirement in Subsection 1129(b) of the U.S Bankruptcy Code. Under this rule, a Chapter 11 bankruptcy plan is fair and equitable even if a senior class of creditors is not paid in full so long as the classes of claims junior to theirs receive nothing. However, APR applies only if, under the plan, the senior class of creditors is impaired (i.e. not paid in full) and they have not voted in favor of the reorganization plan. See Baird and Bernstein (2006). 3 See, for example, Park (2000). 2

5 option. 4 If priority rules are expected to be adhered to, the default option associated with senior debt should be less valuable than the default option associated with subordinated debt. As a result, we would expect secondary market yields on subordinated debt to be more sensitive to changes in volatility than yields on senior debt. If on the other hand, asset volatility is related to the likelihood of dilution arising from the violation in priority rules, then yields on senior and subordinated debt will converge as volatility increases. The likelihood of violations of priority rules among unsecured creditors may be positively related to volatility because of how volatility affects the value of the rights afforded junior creditors in bankruptcy. Specifically, in Chapter 11 junior creditors generally have the right to insist on a valuation hearing to determine the value of the enterprise and thus the value of the securities that are distributed to creditors under the reorganization plan. 5 This right, together with other ways in which junior creditors can delay a bankruptcy proceeding, provide an option-like claim to junior creditors. We refer to this as the dilution option, the value of which is increasing in the volatility of the firm s assets. The existence of the dilution option implies that even if creditors are certain that the bankruptcy court, if called upon, will apply APR, bankruptcy and restructuring outcomes can nevertheless deviate substantially from APR. As discussed later, this is because the value of the option to delay accrues to junior creditors but the costs of delay are born by senior unsecured creditors. Since rational creditors will price the dilution option, the cost of borrowing will reflect the potential for restructuring outcomes that violate priority rules. In order to examine the relationship between yields on senior and subordinated debt and volatility we obtain data on yields from secondary market transactions from the Trade Reporting and Compliance Engine (TRACE) database. As expected, we find that secondary market yields are significantly higher on subordinated bonds and are increasing in lagged volatility. However, consistent with junior creditors having a dilution option, we find that yield spreads between subordinated and senior bonds are decreasing in lagged volatility. We also examine the impact of changes in asset volatility on syndicated loan spreads at the time of issuance relative to the credit default swap (CDS) spreads of the issuing firms. We compare loan spreads to CDS spreads of the same issuer because syndicated loans are typically secured, 4 Longstaff and Schwartz (1995) and Smith and Warner (1979). 5 Ayotte and Morrison (2009) find that junior creditors frequently exercise this right. In particular, they find junior creditors object in over 50% of the cases in their sample. 3

6 whereas CDS contracts are written on senior unsecured bonds and therefore the CDS spread approximates the par yield spread of a senior unsecured debt security of an issuer. 6 We document that the difference between CDS spreads and loan spreads increases with lagged volatility, which is consistent with our hypothesis that security provides greater protection from dilution. 7 If, as we hypothesize, an increase in the likelihood of dilution increases the relative cost of senior unsecured debt, then we would also expect the reliance on senior unsecured debt to be negatively related to asset volatility. Moreover, if security affords greater protection from dilution than covenants, we expect that firms will rely more on secured debt as volatility increases. However, identifying the causal effect of volatility on debt priority structure presents at least two empirical challenges. First, volatility is likely related to profitability and as a result, determining a causal impact of volatility on priority structure requires disentangling the effect of volatility on profitability from the direct effect of volatility. Second, both volatility and debt priority structure are likely to be endogenous. In other words, volatility may be correlated with priority structure without having a causal effect. For example, management may engage in asset substitution that increases volatility and at the same time issue secured debt as a way of transferring wealth from existing unsecured bondholders to shareholders. We address these challenges by identifying exogenous shocks to volatility and profitability through using lagged changes in tariffs and industry-level trade weighted foreign exchange rates as instruments for both profitability and volatility. Tariffs and trade weighted exchange rates have been used extensively in the economics and finance literature as instruments to identify exogenous shocks to product market competition. 8 We argue that shocks to product market competition, in turn, directly affect firms profitability and increase the volatility of firms asset returns but are unlikely to otherwise directly affect firm-level debt priority structure choices. Moreover, because we use lagged industry-level measures as instruments, it is unlikely that firms contemporaneous debt priority structure choices will directly influence either of our instruments. 9 6 See, for example, Duffie (1999) and Hull, Predescu, and White (2004). 7 Unlike for bonds, we do not have data on secondary loan market trading and thus endogenity is a greater concern in the loan spread/ CDS analysis. As we discuss below, in our other tests we address these concerns using an instrumental variables approach. However, given the small number of firms with loan spread and CDS information the instrumental variables approach is not feasible. We therefore view these results merely as supportive of our hypothesis rather than causal. 8 See, for example, Bertrand (2004), Fresard (2010), Valta (2012), and Xu (2012). 9 Our use of tariffs and trade weighted exchange rates is also motivated by findings that the volatility of output in manufacturing industries is negatively related to trade barriers. See for example di Giovanni and Levchenko (2009). 4

7 We establish the relevance of our instrumental variables using panel data for a large sample of rated manufacturing firms. We find that lagged levels of tariffs are negatively related to volatility and positively related to profitability, and the point estimates are significant at the 1% level. Most important, we test for the joint significance of the instruments in each first stage regression and reject the null hypothesis of no relationship. Next, using an instrumental variables approach, we document that volatility is significantly related to the priority structure of firms debt. Specifically, consistent with our hypotheses we find, conditional on issuing debt, a positive and significant relationship between volatility and both the likelihood of issuing secured bank debt and subordinated debt. In contrast, increases in volatility are associated with a significant decrease in the likelihood of issuing senior debt. Additionally, we find that volatility is negatively related to the proportion of senior debt used and positively related to the proportion of secured and unsecured debt. Overall, we interpret our results as consistent with the hypotheses that increases in volatility increase creditors concerns regarding dilution and violations of priority rules. This affects the cost of senior unsecured claims relative to secured and subordinated claims, which in turn, induces firms to engage in priority spreading. Our findings are also consistent with Welch (1997) s theory as to why bank debt is secured. His theory is based on the idea that the deadweight costs of litigation and influence can be lowered by awarding potentially stronger creditors ex-post the strongest priority position ex-ante. 10 Banks, he argues, are better organized, more sophisticated and have greater incentives to build expertise and reputations as tough bargainers than diffuse and more transient investors in publicly traded debt. Since secured priority affords creditors the strongest priority position ex-ante, bank debt is expected to be secured. 11 An alternative explanation for our findings is that increases in volatility lead to a priority structure that is designed to maximize banks incentives to undertake monitoring that is socially beneficial for all creditors. This explanation is based on the theory proposed by Park (2000), which Rauh and Sufi (2010) propose as a potential explanation for the inverse relationship between credit ratings and debt heterogeneity. The basic idea behind Park s theory is that banks have a stronger incentive to monitor if their claim is smaller, and the presence of junior debt serves to reduce the 10 Litigation and influence costs include not just direct legal expenses but also expenses associated with organizing a creditor class, free rider problems, management time and hassle and 11 Less diffusely held bank debt also economizes on costs of monitoring collateral by reducing inefficient duplication of effort (see Winton 1995). 5

8 size of the bank claims. While this explanation is consistent with our findings, according to Park (2000) s theory, for banks to have an incentive to engage in monitoring that is beneficial for other creditors, the bank s claim must be at least large enough to be impaired in liquidation. However, for firms with subordinated debt we find that secured lenders are typically unimpaired and recover the full nominal value of their claims. We contribute to the literature on the structure of corporate debt in several ways. First, we provide empirical support for predictions made in the legal literature that priority is more effectively maintained through collateral grants than though subordination clauses. Second, consistent with junior creditors holding a valuable dilution option, we find that the relative cost of subordinated debt is decreasing in volatility. Third, we provide evidence of priority spreading in response to increases in potential for dilution of senior unsecured creditors by subordinated creditors. The rest of the paper is organized as follows. In Section II we review the theoretical and legal literature relating to debt priority structure, and discuss how our measure for asset volatility is related to the potential for dilution. Section III provides evidence that secured creditors are typically unimpaired in bankruptcies and that deviations of APR are more frequent for senior unsecured creditors than for senior secured creditors. Section IV examines the relation between asset volatility and the cost of senior unsecured debt relative subordinated debt and secured debt. Section V presents our analysis of the relationship between volatility, debt capacity, and the priority structure of corporate debt and addresses endogeneity concerns using an instrumental variables framework. Section VI provides a summary and concludes. II. Background A. The mechanism for creating priority One purpose of the priority structure of a firm s debt is to mitigate the dilution of existing creditors. Dilution can occur through the issuance of additional debt claims, the issuance of higher priority claims, or through asset substitution. How priority is established among creditors has been analyzed extensively in the legal literature but has received relatively little attention in the finance 6

9 literature. 12 Schwartz (1997) identifies three primary ways of creating priority among debtors: issuing secured debt, issuing subordinated debt, and through covenants that result in later creditors subordinating their claims to earlier creditors. The latter two ways of creating priority are referred to in the legal literature as subordination priority or negative pledge covenant priority. While the bankruptcy code recognizes priority established through a security interest and through subordination, how priority is established has important implications for the costs of enforcing priority and the ability of creditors to mitigate future dilution. Priority established through a security interest is governed by Article 9 of the Uniform Commercial Code, Secured Transactions. Under Article 9, two key events are generally required for the creation of a secured claim: attachment, which is an agreement to create a security interest in an asset or a group of assets, and perfection. Perfection usually occurs when a financing statement is filed in the appropriate public record. Perfection serves both as a notice function (a means for other creditors to determine whether there is a security interest), and as a way of establishing priority concerning the underlying collateral. Generally, absent a lien subordination agreement to the contrary, temporal rules of lien priority apply, where the first creditor to perfect its interest has first priority; the second creditor to prefect its interest has a second lien and so on. 13 Lien priority means first lien holders are paid in full out of the proceeds of the sale or liquation of the collateral before second lien holders receive any payment. In the event that the first lien holder is impaired, in the sense that the proceeds from liquidating the collateral are insufficient to fully repay the first lien debt, then the unpaid balance of the first lien holder and all of the amount owed to more junior lien holders becomes a senior unsecured claim of the debtor in bankruptcy. There are both advantages and disadvantages of establishing priority through a security interest rather than through negative pledge covenants. The decision to create priority through security can be analyzed in the context of what Smith and Warner (1979) and Bradley and Roberts (2015) call the Costly Contracting or Agency Theory of covenants. These theories posit that contract choice involves weighing the benefits of mitigating potential agency costs through 12 There are several notable exceptions. Smith and Warner (1979) examine how contractual restrictions in debt contracts mitigate stockholder-bondholder conflicts. Hackbarth and Mauer (2012) use conflicts between stockholders and bondholders to explain variations in priority structure, and Ravid et al. (2015) use conflicts between creditor classes to explain the sequencing of debt issues of different priority over time. 13 See UCC Article 9, Secured transactions (1998) Summary at; aspx?title=ucc%20article%209,%20secured%20transactions%20(1998). Also see Mann (1997). 7

10 contractual restrictions against potential contracting costs (including potential ex-post inefficiencies associated with contract enforcement). In terms of benefits, the law and economics literature has identified a number of potential benefits of creating priority through security. 14 One obvious advantage of granting collateral is that a security interest provides a higher priority claim on the assets that serve as collateral than a senior unsecured claim on the overall cash flows of the firm. A secured creditor has a security interest in a specific asset or group of assets. In contrast, senior unsecured contracts provide senior creditors priority claims on all the unencumbered assets of the borrower, but only provide them with a junior claim relative to secured claims on any encumbered assets. To the extent that secured lien holders claims are impaired, secured creditors share pari-passu with other senior creditors in the unsecured assets of the borrower. A second advantage of creating priority through security is that it increases the effectiveness of the lenders efforts to limit risk shifting behavior and dilution by narrowing the focus of the lenders monitoring efforts. As Mann (1997) points out When a lender has an effective lien on a particular asset [or group of assets] it can focus its monitoring on that asset [or group of assets] secure in the knowledge that repayment is likely so long has the liquidation value of the asset remains greater that the outstanding amount of the loan. 15 In other words, to ensure their claims are unimpaired, secured creditors need to insure only that the liquidation value of the collateral exceeds the amount they are owed. The other unencumbered assets of the borrower serve as a secondary source of repayment in liquidation. A third advantage of secured debt over subordination priority is that temporal rules of priority generally do not apply to senior unsecured claims. As a result, later senior lenders are not bound by the earlier loan contract even if later advances (or repayments) violate the seniority covenant. 16 In short, violation of a seniority covenant still typically conveys priority. 17 As a result, a borrower s commitment to abide by priority created through subordination is less credible than a commitment 14 Mann (1997) and Schwartz (1997) provide excellent reviews of the legal advantages and disadvantages of creating priority through security rather that strictly by contract. 15 Mann (1997) pp See Schwartz (1997) and LoPucki (1994). 17 In the event of a breach the lenders must proceed against the breaching borrower. See Schwartz (1997). If a creditor receives payment that violates the priority of senior creditors, and if the borrower is in the so called zone of insolvency, then such payments may be considered a voidable preference in bankruptcy. Also, if a security interest or senior interest is granted in violation of a negative pledge to an existing junior creditor that grant may be considered a voidable preference if it occurs within 90 days of a bankruptcy filing. See LoPucki (1994) 8

11 based on security. While a distressed publicly traded company s attempt to deliberately breach of seniority clause by issuing new debt is likely to be quickly detected and stopped by senior creditors (through a declaration of default), unsecured creditors still face a significant threat of dilution from what legal researchers refer to as reluctant or involuntary creditors. 18 Categories of reluctant creditors include (1) trade creditors (to the extent they do not secure their claims), (2) product liability and warranty claimants, (3) victims of anti- trust violations, patent, trademark and copyright infringement, (4) environmental agencies (for clean- up costs), (5) pensioner and the Pension Benefit Guarantee Corporation (in the case of under- funded pension plans) and certain wage claims. 19 Reluctant creditors are typically treated as senior creditors in bankruptcy and thus dilute the claims of unsecured but not unimpaired secured creditors. As LoPucki (1994) points out Any debtor who either has or expects to have involuntary unsecured creditors will find an economic advantage to selling secured status to voluntary creditors 20 The absence of temporal priority for unsecured claims implies that the advantage of secured debt is expected to be greater for financially weak firms. A fourth advantage of creating priority through security is that unsecured creditors bear greater risks and costs in bankruptcy. In bankruptcy the debtor may use and possibly consume a secured creditor s collateral only if it provides the secured creditor adequate protection against loss. If secured creditors can demonstrate that they will not be adequately protected they can insist that the collateral backing their loan be liquated to pay off their claim. 21 In contrast, unsecured creditors are not entitled to adequate protection payments. In addition, in bankruptcy unimpaired secured creditors typically continue to receive interest payments while senior unsecured creditors do not. 22 As a result, the cost of bankruptcy in terms of the risk of asset dissipation and delay are greater for unsecured creditors. As previous researchers have pointed out, creditors incentives to participate in out of court restructurings, including subordinating their claims, depends, in part, on a creditor s expected costs 18 See Sullivan, Warren and Westbrook (1989) and LoPucki (1994) 19 See Campello et al. (2016) for a discussion of the workers claims in bankruptcy. Campello et al. (2016) argue that organized labor representation on the unsecured creditors committee leads to significantly greater bankruptcy costs. Consistent with this argument they find negative and statistically significant abnormal bond returns associated with close but successful union elections. 20 See LoPucki (1994), pp See Baird and Bernstein (2006). 22 See Luehrman and Techner (1992). 9

12 of bankruptcy. 23 All else equal, the greater a creditor s expected cost of bankruptcy the lower its reservation price in an out of court restructuring. Thus, a security interest provides a stronger commitment to preserve seniority and enforce covenants than priority established through subordination covenants. 24 As a result, secured debt affords secured creditors much greater leverage (in terms of the ability to credibly commit to exercise control rights) over borrowers than senior unsecured creditors and therefore creates stronger incentives to repay secured debt. The idea is that secured creditors can credibly threaten to liquidate collateral in the event of a default. If the collateral is worth more to the borrower than in liquidation and the liquidation value of the collateral exceeds the amount of the secured loan, the borrower (and potentially unsecured creditors) bears the potential loss in user-specific value. The value of the right to adequate protection and interest payments increases with the volatility of the borrowing firm s asset value, because increases in volatility increase the value of junior unsecured creditor s option to delay. As Baird and Bernstein (2006) explain, in a bankruptcy impaired creditors have the right to insist on a valuation hearing. Even if the court is expected to make an unbiased appraisal, junior creditors who expect to recover little in a prompt resolution have an incentive to delay by insisting on a valuation hearing in hopes that the value of the firm s assets increases in the interim. 25 By delaying the reorganization, junior impaired creditors potentially gain from any increase in value. Note that delay also increases bankruptcy costs but these costs will be borne by impaired senior unsecured creditors. 26 While collateral provides benefits there are two ways in which secured debt potentially increases the costs of establishing priority. First, the origination and administration costs are generally greater for secured than unsecured loans. For example, creating a security interest 23 See Asquith, Gertner and Sharfstein (1994), Gilson, John and Lang (1990), and James (1995). 24 Unsecured creditors will subordinate their claim in a restructuring if the loss from dilution through subordination is less that their expected costs of bankruptcy. The potential for dilution is particularly acute when the senior unsecured claim is the fulcrum security in bankruptcy. The fulcrum security in a bankruptcy is the security that is impaired but is expected to be partially repaid if the firm is liquidated and absolute priority rules are adhered to. Senior debt is likely to be the fulcrum security when an insolvent firm has secured debt and very little subordinated debt. When senior debt is the fulcrum security the bankruptcy costs are borne principally by senior debt holders (in terms of administrative and delay costs and any dissipation of assets). Given senior creditors bear these costs they may be willing to subordinate a portion of their claim to subordinated creditors to speed up the bankruptcy proceeding which undermines credibility of covenants that protect against subordination. 25 Asset volatility is likely to be related to uncertainty concerning the outcome of the valuation process. Even if the courts appraisal is unbiased on average an increase in the standard error of the estimate creates option value for junior creditors. 26 See Luehrman and Techner (1992). 10

13 involves appraisal and title costs. 27 A second and potentially greater cost of secured debt arises from the greater leverage security provides lenders over borrowers. As Diamond (1993), Rajan (1992) and others point out, providing lenders strong control rights can distort investment incentives since lenders focus is on maximizing the likelihood of repayment which can conflict with shareholder and overall enterprise value maximization. B. The relationship between dilution and asset volatility Overall, if collateral is more effective at creating priorities and limiting dilution than subordination by contract, then as the potential for dilution of senior unsecured claims increases we expect relative costs to change in a way that leads firms to reduce their reliance on senior unsecured debt and increase their reliance on secured and junior subordinated debt. To empirically test this prediction, we need a measure of the potential for dilution of unsecured creditors. In our empirical tests we assume that firms incentives to engage in priority spreading are increasing in volatility of their asset returns. We measure asset volatility following the methodology of Bharath and Shumway (2008). This method is based on observed monthly equity returns over the prior 12 months, an estimate of the volatility of debt and the firm s current capital structure. 28. We use asset volatility as a measure of the potential for dilution for several reasons. First, an increase in volatility, ceteris paribus, will increase the potential dilution of existing creditors because it makes existing debt riskier. With risky debt outstanding, shareholders have a preference to finance new projects with debt of equal or greater priority than the firm s existing risky debt. Doing so reduces the value of the new projects that accrues to existing debt holders and thus increases the share of project value accruing to shareholders. 29 Second, the potential for asset 27 See Mann (1997) and Smith and Warner (1979) for an analysis of the differences in transactions and monitoring costs associated with secured versus unsecured lending. 28 Bharath and Shumway (2008) describe their volatility measure as a naïve estimate of the asset volatility relative to the measure proposed by Merton (1974). Their measure assumes that the market value of debt is equal to the face value of debt and that the volatility of debt is a linear function of the volatility of equity. The volatility of the firm s asset returns is assumed to be a weighted average of the volatility of debt and equity, using the proportion of equity and debt relative to firm value as the weight. As they point out, it is relatively easy to criticize the assumptions on which their measure is based. Nevertheless, they find very little difference in default predictions based on their measure of volatility and a more computationally complex measure based on direct measures of the market value of debt and the volatility of return of debt. We use the Bharath and Shumway measure due to limits on the availability of data on the market value and volatility of debt. 29 Indeed, a borrower s inability to issue senior claims can lead to an underinvestment problem where borrowers with risky debt outstanding forego risk reducing positive net present value investments. In particular, Johnson and Stulz (1985) show that, in the absence of the ability to issue senior claims firms may pass up risk reducing but value 11

14 substitution is likely to increase with the riskiness of a borrowing firm s debt since losses associated with the risky new project that occur in the default states associated with the firm s existing projects are borne by creditors and not shareholders. Third, an increase in volatility increases the likelihood that priority rules among unsecured creditors will be violated in a restructuring of a firm s debt claims, because it increases the value of junior creditors dilution option. III. Secured debt and adherence to absolute priority rules We examine whether security affords creditors stronger claim to priority than do covenants by analyzing the frequency of priority rule violations for secured and unsecured creditors in both Chapter 11 bankruptcy outcomes and out of court restructurings. While a number of previous studies document that Chapter 11 bankruptcy outcomes often depart significantly from APR these studies focus on deviations from priority rules where shareholders receive payments before other creditors are paid in full. 30 In contrast our focus is on the relative frequency of violations involving secured versus senior unsecured creditors. We examine recovery rates of secured and unsecured creditors, as well as deviations from priority rules for a sample of corporate debt restructurings by manufacturing firms (SIC codes 2000 to 3999) between 1987 and The data on corporate debt restructuring and recovery rates are from the Moody s Default and Recovery Database (DRD), and the events we consider are bankruptcies that don t result in a liquidation of the firm, distressed exchanges, default and cure events, as well as other restructuring events resulting in a loss as identified by the Moody s DRD. For each restructuring event, we aggregate all debt securities of the corresponding firm by their respective creditor class, based on the priority of the debt securities. We calculate the recovery enhancing projects because of the wealth transfer from shareholders to creditors. As a result, a blanket prohibition on issuing senior debt is typically not optimal. 30 See for example Franks and Torous (1989), Eberhart, Moore and Roenfeldt (1990) and more recent studies by Carapeto (2006) and Ayotte and Morrison (2009). For a review of the empirical research on out-of-court and bankruptcy restructurings see Hotchkiss et al. (2008). 31 As discussed in greater detail in Section V, the main sample for our empirical analysis is based on manufacturing firms between 1980 and 2005 because we rely, in part, on changes in tariffs to identify exogenous shocks to volatility. We restrict the sample of debt restructurings accordingly in order to be as consistent as possible. Because our copy of the Moody s DRD only covers restructuring events from 1987 onwards there is a slight difference to the date rage of the sample used for the later empirical analysis of the paper. 12

15 rate of a creditor class as the average nominal recovery rate across all the debt securities in the same class, where nominal recovery rates generally reflect the value of the settlement instruments received for each debt security expressed as a percentage of face value. 32 For each creditor class we further determine whether it was negatively affected by a violation of priority rules through the outcome of the restructuring. We consider a creditor class to be negatively affected by a violation of priority rules if its nominal recovery rate is less than 100% and a creditor class of lower priority has a positive nominal recovery rate. Note that by definition a creditor class can only suffer from a priority violation if there is at least one creditor class of lower priority to it that would stand to benefit. Our final sample for this analysis of recovery rates consists of 214 debt restructuring events by 200 firms and involves a total of 516 different creditor classes across all restructuring events. A total of 178 restructuring events are bankruptcies, 31 are distressed exchanges, 3 are default and cure events, and only 2 are considered other restructurings. Priority violations are relatively common with 42 restructuring events involving at least one creditor class that was negatively affected by a violation of priority. In our sample deviations from priority rules are slightly less frequent (on a relative basis) for bankruptcies than they are for out of court restructurings. Specifically, 34 bankruptcies (about 19% of all bankruptcy events) are associated with APR violations compared to 8 out of court restructurings (about 22% of all out of court restructurings). Recovery rates for secured and senior unsecured claims are likely to differ because secured and senior unsecured creditors are positioned differently in the priority structure. Specifically, secured claims may have greater credit support, in terms of the amount of debt outstanding that is junior to their claim, than senior unsecured claims. However, if security affords greater priority protection, we expect, controlling for the amount of credit support, violations in priority rules to be less frequent for secured than for unsecured senior creditors. Therefore, to test whether security affords creditors greater priority protection, we control for the level of credit support provided to senior creditors. Table 1 provides summary statistics for the restructuring events in our sample. Panel A presents the average recovery rates across the different creditor classes in our sample of debt restructurings. Note that across the full sample, secured creditor classes, when compared to 32 For a more detailed description of how creditor classes are determined and for more details on recovery rate calculations please refer to Appendix A 13

16 unsecured creditor classes, are rarely impaired. The average recovery rate of secured creditors is over 93% and over 75% of secured creditors make a full recovery where the nominal recovery rate is at least 100%. By comparison, only about 17% of unsecured creditor classes make a full recovery and the average recovery rate of 45% is considerably lower as well. These results however, are unconditional on whether the firm has additional debt securities outstanding at the time of the restructuring that would be of lower priority and could provide additional credit support to more senior classes. We therefore split the sample by whether creditor classes are the most junior in terms of priority (Panel B) and by whether there are additional creditor classes of lower priority that could absorb any losses first (Panel C). Not surprisingly, as shown in Panels B and C of Table 1, in the absence of additional subordinated creditor classes recovery rates are lower for both secured and unsecured creditor classes, while they are higher for both when there are additional classes of lower priority. The results in Panel C are particularly interesting as they show that, if a company has debts of lower priority outstanding secured creditors recover on average over 99% of the face value of their claims and in 83% of cases make a full recovery while senior unsecured creditors make a full recovery only about 30% of the time. This finding suggests that secured creditors are typically not impaired and thus have no incentive to scale down their claims in restructurings. 33 Panel D of Table 1 provides evidence that APR is violated more frequently for senior unsecured creditors than for secured creditors. Specifically, conditional on having at least one more junior creditor class of lower priority, almost 40% of senior unsecured creditor classes are negatively affected by deviations from APR compared to only 9% of secured creditor classes. To test the significance of this difference we estimate probit models where the dependent variable takes a value of one if a creditor class in a debt restructuring was negatively affected by an APR violation. 34 We include indicator variables in the regressions for whether a creditor class is secured and additionally include a variable for the level of credit support provided by more junior creditor classes. Specifically, for each creditor class in our sample we calculate a variable Fraction of debt below which represents the fraction of total face value debt with lower priority to the corresponding creditor class. For example, for a firm with secured debt of face value $10 and 33 This result is consistent with James (1995) finding that secured bank lenders do not scale back their claims unless they are impaired. 34 Our results are similar when we estimate the likelihood of APR violations using a linear probability model. 14

17 junior claims of face value $90, the fraction below the secured creditor class would be 0.9. This is admittedly a somewhat crude measure for the amount of credit support for senior creditors, in that it ignores the claims of reluctant creditors which potentially erode the credit support provided by more junior creditors. As shown in column (1) of Panel A of Table 2, controlling for credit support, secured creditor classes are significantly less likely to suffer from APR violations than unsecured creditor classes. Specifically, the corresponding marginal effect is a 19% reduction in the likelihood of suffering from an APR violation for secured creditor classes and this difference is significant at the 1%- level. Not surprisingly, the likelihood of violations of APR is decreasing in the credit support provided to more senior creditors. The coefficient estimate on Fraction of debt below indicates that an increase in credit support of 10% reduces the likelihood of a priority violation by roughly 3%. In column 2, we test whether the impact of credit support varies with whether priority is established through security or the contract. As shown, credit support affords secured creditors significantly greater protection from APR violations than senior unsecured creditors. Specifically, an increase in credit support of 10% for secured creditors reduces the likelihood of an APR violation by roughly 4%, while for senior unsecured creditors we find no statistically significant relation between APR violations and credit support (the point estimate is not statistically significant). As shown in columns (3) and (4) of Table 2 our findings are robust to the inclusion of industry fixed effects, and as displayed in Panel B we obtain similar results if restrict the sample to restructurings that occur in bankruptcy. Overall the results in Table 2 are consistent with the hypothesis that security affords greater protection from dilution than covenants that prioritize unsecured claims. IV. Volatility and the cost of debt In this section we examine how increases in the potential for dilution of senior unsecured creditors affects the relative cost of senior unsecured debt. If, as the results in Section III suggest, secured debt provides creditors with greater protection from dilution that unsecured debt, we would expect that the cost of senior unsecured debt should increase relative to secured debt when firms asset volatility increases. Moreover, to the extent that subordinated creditors hold a dilution option we would also expect that the cost of subordinated debt and senior unsecured debt converges with increasing volatility. We empirically test these predictions in this section. 15

18 A. Analysis of secondary market bond yields If investors expect that priority rules will be adhered to, yields of subordinated debt claims are expected to be more sensitive to changes in volatility than yields of senior debt claims. Intuitively, all else equal, for bonds that are not in default, the default option associated with subordinated debt is closer to being in the money than the default option associated with senior debt. As a result, the yields on subordinated debt claims should be higher and more sensitive to changes in volatility than the yields on senior debt. 35 However, if priority rules are likely to be violated, we would expect the yields on subordinated debt to converge to the yield on senior debt as volatility increases. To examine the relation between bond yields and asset volatility we obtain secondary bond market data from the enhanced Trade Reporting and Compliance Engine (TRACE) database provided by the Financial Industry Regulatory Authority (FINRA) for the time period of July 2002 through December The enhanced TRACE database contains daily pricing, yield, and trade size information on all corporate bond trades in the United States from July 2002 onwards. 36 We apply the filters suggested by Dick-Nielsen (2014) to clean erroneous trades, subsequent trade corrections, and cancellations, and merge the TRACE data with bond information from the Mergent Fixed Income Securities Database and with quarterly firm-level data from Compustat. In order to be consistent with the rest of our analysis we restrict the sample to daily trades of bonds that are rated, are not in default, have positive amounts outstanding, and are issued by manufacturing firms (SIC codes ). Because we are interested in yield differences between senior unsecured and subordinated bonds, we restrict the sample to fixed-rate coupon bonds and exclude all asset backed, credit enhanced, convertible and exchangeable bonds as well as secured lease obligations. Because secondary bond markets tend to be less liquid than stock markets and because the average execution costs are substantially smaller for large trades, Bessembinder et al. (2009) propose using daily trade-weighted prices when using TRACE data (since this will put more 35 In the context of Black Scholes, the derivative of a put with respect to volatility (Vega) increases as the value of the underlying approaches the strike price from above. 36 While public disclosure of trading data through TRACE was phased in over four different phases between July 2002 and February 2005, the transaction level data were still reported to TRACE regardless of whether public disclosure was mandated at the time and are available to researchers through the enhanced TRACE database. See Asquith, Covert and Parag (2013) for a discussion of the enhanced TRACE database. 16

19 weight on large institutional trades that should more accurately reflect the underlying price). Since our analysis involves yields, we calculate daily trade-weighted yields for each bond instead of prices. We subtract the corresponding maturity matched Treasury rate from each bond s daily trade-weighted yield to calculate daily yield spreads. 37 Since not all bonds in our sample trade on a daily basis, we aggregate the daily bond trades at the monthly level by calculating the average monthly yield spread for each bond in our sample. 38 Finally, since a firm can have multiple bonds of the same or of different priority outstanding and trading each month, we further aggregate our sample at the creditor class level for each firmmonth. Specifically, for each firm-month we classify the firm s traded bonds as either senior unsecured or subordinated. We define senior unsecured bonds as bonds classified by Mergent as Senior and subordinated bonds as bonds classified by Mergent as Senior Subordinated, Subordinated, Junior or Junior Subordinated. For each firm-month we then calculate the value-weighted yield spread of all senior bonds and the value-weighted yield spread of all subordinated bonds of the firm, where the weights correspond to the remaining principal amount outstanding of each bond. Our final dataset is a monthly panel of monthly average yield spreads for each firm and creditor class. It consists of 7,103 observations involving 280 different manufacturing firms. Note that because we mostly observe distinctions between senior and subordinated debt among firms A rated or below, the yield analysis is based on a sample that consists primarily of firms with BBB ratings or below. An advantage of focusing on secondary yields as opposed to yields in the primary market is that the potential endogeneity of asset volatility is less of a concern. In order to further mitigate concerns relating to endogeneity between secondary market yields and asset volatility, we use onequarter lagged values of volatility in our regression analysis. Table 3 presents estimates of regressions relating yield spreads to an indicator variable as to whether the bond is subordinated or senior, lagged volatility, lagged profitability, interactions with subordination and both volatility and profitability, bond maturity, trading activity and a lagged volatility interacted with maturity. 37 We use daily constant maturity Treasury rates from the Federal Reserve s website. If a bond s remaining maturity falls between two Treasury maturities we calculate the implied Treasury rate through linear interpolation. Because Treasury rates are only available for 1 month to 30 years we exclude all bonds with remaining maturities less than one month or exceeding 31 years. For bonds with maturities between 30 and 31 years we assign the 30-year Treasury rate. 38 We winsorize our sample by eliminating bond-months where the average yield spread is negative or the average yield spread is above 25%. 17