Debt-Equity Simultaneous Holdings and Distress Resolution 1

Transcription

1 Debt-Equity Simultaneous Holdings and Distress Resolution 1 Yongqiang Chu, Ha Diep Nguyen, Jun Wang, Wei Wang, Wenyu Wang This draft: March 2018 Abstract We study the effect of financial institutions simultaneous holdings of loans and equity; bonds and equity; loans and bonds; and loans, bonds, and equity on the resolution outcome of financially distressed firms. Our results show that simultaneous holdings of debt (loans or bonds) and equity are associated with a higher likelihood of out-of-court restructuring versus bankruptcy. Our identification relies on instrumental regressions and using the merger of financial institutions as a source of exogenous shock to the formation of simultaneous holdings. We further show that the effect of simultaneous holdings on the probability of out-of-court restructuring is stronger when these holders have a larger equity stake in the game and when the expected bankruptcy costs are higher. The combined empirical evidence suggests that simultaneous holdings improve the incentive alignment of debt holders and equity holders to facilitate a cost-effective workout. Keywords: distress; bankruptcy; out-of-court restructuring; dual holding; loan; bond; equity JEL Classification: G20, G30, G33 1 We thank Xiaobing Ma for excellent research assistance. Yongqiang Chu: University of South Carolina, yongqiang.chu@moore.sc.edu Ha Diep Nguyen: Indiana University, nguyenhd@indiana.edu Jun Wang: Western University, jun.wang@uwo.ca Wei Wang: Queen s University, wwang@queensu.ca Wenyu Wang: Indiana University, wenywang@indiana.edu 1

2 1. Introduction A financially distressed firm can restructure its debt through either an out-of-court workout or a formal Chapter 11 bankruptcy. Compared to restructuring in court, private workout is an attractive option because it is associated with smaller deadweight costs and less damage to a company s reputation; it affects how customers and suppliers interact with the firm and how employees view the viability of the business (Jensen 1989; Gilson, John, and Lang 1990; Franks and Torous 1994; Hotchkiss et al. 2008; Brown and Matsa 2015) 2. Besides these concerns, firms have strong incentives to avoid bankruptcy as legal proceedings can lead to a substantial increase in the amount of claims, which would not be otherwise immediately payable (e.g., unfunded pensions, trade claims), litigation (e.g., fraudulent transfers and preferences), and the exposure to judges discretion and biases, creating uncertain outcomes (Chang & Schoar, 2006). However, firms face many challenges and obstacles in structuring a successful workout. These are known as creditor coordination and holdout problem (Brown, 1989; Gertner & Scharfstein, 1991; Haugen & Senbet, 1978), conflicts of interest (Bernardo & Talley, 1996; Bulow & Shoven, 1978), information asymmetry (Giammarino 1989; Brown, James, and Mooradian 1993), and tax and other regulatory provisions (Jensen, 1991). These frictions impede a firm s ability to renegotiate its claims out of court, leading to it having to incur the significant costs of bankruptcy. The existing empirical work emphasizes the importance of debt structure and bank participation in resolving creditor coordination and information problems in out-of-court workout. 3 These studies examine the bargaining of secured lenders or public bondholders as a standalone party against other creditors and shareholders. However, the increasingly common practice of financial institutions simultaneously holding different types of securities of the same firm (Chava, Wang, & Zou, 2016; Jiang, Li, & Shao, 2010) suggests that the line between pure debt holding and equity ownership is blurred. The following questions naturally arise: How often and how intense are simultaneous holdings present in distressed firms? What are their effects on the resolution of the financial restructuring of distressed firms? Answering these questions helps further our 2 Bankruptcy costs range from out-of-pocket expenses paid to lawyers and other professionals to a wide range of unobserved opportunity costs, such as loss of customers, suppliers, human capital, and growth opportunities due to a manager s diversion from running the business (e.g. Altman 1984; Opler and Titman 1994; LoPucki and Doherty 2004; Graham et al. 2016). 3 See, e.g., Gilson, John, and Lang (1990); Asquith, Gertner, and Scharfstein (1994); Chatterjee, Dhillon, and Ramirez (1996); James (1996); Demiroglu and James (2015) 2

3 understanding of how incentive alignment internalizes conflicts of interest between creditors and shareholders or among different classes of creditors in financial distress. This paper provides the first empirical analysis of how financial institutions simultaneous holding facilitates a cost-effective workout. A key challenge to our empirical analysis is to construct a large sample of out-of-court restructurings, partly due to the fact that there are no universal definitions of what types of negotiations and debt contract amendments constitute a formal private workout and there are no standard databases tracing these private restructurings. We follow prior studies (e.g., Gilson et al., 1990; Demiroglu and James, 2015), and we use firms stock returns and financial characteristics, news headlines from RavenPack New Analytics, loan covenant violations, and textual analysis of corporate filings to assemble a comprehensive data set of out-of-court restructurings between 2000 and The 579 unique out-of-court restructuring events we construct and 605 bankruptcy filings we retrieve from bankruptcydata.com make ours one of the largest samples for studying the resolution of financial distress. All our sample firms have book assets of at least $50 million in three consecutive years before the distressed event. We focus on large firms as their debt structure is complex, conflicts of interests are likely to be the most severe, and, thus, simultaneous security holdings can have the most significant impact on distress resolution. To construct simultaneous holdings of loans, public bonds, and equity, we turn to DealScan (for syndicated loan lenders), Thomson s emaxx (for public bondholders) and Thomson Reuters 13F filing database (for institutional equity holders). We first merge these databases with our sample of restructured firms to identify holders of loans, bonds, and equity. We then manually match loan lenders, bondholders, and equity holders by the names and types of the financial institutions in the three databases, carefully adjusting for changes in holder ownership through corporate control events like mergers, acquisitions, or spin-offs. This labor-intensive matching procedure allows us to construct four types of simultaneous holdings loans and equity; bonds and equity; loans and bonds; and loans, bonds, and equity over the entire life span of the sample firms. We find that loan-equity dual holding is the most common type of simultaneous holding in our sample (22% of firms), followed by bond-equity dual holding (10%), loan-bond dual holding (4%), and loan-bond-equity triple holding (4%). Univariate comparison shows that the percentage of firms with dual holdings of loans and equity and bonds and equity, as well as triple holdings of 3

4 loans, bonds, and equity, is much higher in the out-of-court restructuring sample than the bankruptcy sample. The difference in loan-bond dual holding is statistically insignificant between the two subsamples. Furthermore, we find that dual holdings that involve loan (bond) lenders are more likely to occur when loans (bonds) account for a higher fraction of the total debt. All types of simultaneous holding are likely to occur in large firms (larger assets), firms with complex debt structure (lower debt concentration), and firms with higher tangible assets. Our baseline analysis investigates the empirical relation of the four types of simultaneous holding and the likelihood of out-of-court restructuring versus bankruptcy filing. We use either dummy variables or intensity measures (simultaneous holders loan or bond holdings scaled by total loans or bonds outstanding) to measure simultaneous holdings. Our baseline results show that dual holdings of loans and equity and bonds and equity, as well as triple holdings of loans, bonds, and equity, are associated with a higher likelihood of out-of-court restructuring. The economic magnitudes are large. The presence of loan-equity dual holding, bond-equity dual holding, and triple holdings is associated with 33%, 19%, and 45% higher probability of out-of-court restructuring, respectively (compared to the 49% unconditional likelihood of out-of-court restructuring in our sample). We observe a similar economic magnitude when the intensity of simultaneous holding is used. All our regressions control for debt structure, industry and year fixed effects, and other important covariates identified by prior studies. One interesting result that emerges from our baseline regressions is the negative relation between the loan-bond dual-holding dummy and the probability of out-of-court restructuring. However, the statistical significance disappears when debt intensity is used instead. We interpret the result as dual holders of debt claims strengthening their bargaining power against shareholders out of court by not making concessions (James, 1996). However, we do not rule out the possibility that dual holders choose to enter into the capital structure by investing in the most senior claims if they expect that the probability of a firm filing for bankruptcy is high. A potential concern for our baseline regressions is that the empirical relation between simultaneous debt and equity holdings and out-of-court restructuring may be due to omitted timeinvariant unobservable differences between firms that have simultaneous holders and those that do not. To alleviate this concern, we adopt augmented regressions by including lagged simultaneous holdings as a control. Such a specification effectively compares two firms with a comparable 4

5 tendency to have simultaneous holdings in the non-distressed period, but one has simultaneous holding by financial institutions during distress, and the other does not. We find that, with the augmented specifications, the statistical significance and economic magnitudes of our coefficient estimates for either the dummy variable or the holding intensity are qualitatively the same. However, there remains the concern that simultaneous holding during the distress period may be correlated with omitted variables that are specific to corporate financial distress. For example, it is plausible that simultaneous holding tends to appear in firms that are less severely distressed and thus are likely to avoid bankruptcy filing. We apply two identification strategies to address these concerns. First, we adopt the instrumental regression approach, following Jiang et al. (2010). They argue that trading liquidity is a key determinant for institutional equity ownership and has explanatory power for the presence of simultaneous holders of debt and equity. In contrast, stock liquidity, which makes a firm attractive for institutional equity investing, should not directly affect the outcome of distress resolution, conditional upon a firm being in financial distress. We follow their study to use the residual of the Amihud (2002) illiquidity measure regressed on industryadjusted stock returns as our instrumental variable for simultaneous holdings that involve equity (i.e., loan-equity holding, bond-equity holding, and loan-bond-equity holding). We use the Maximum Likelihood Estimation for both binary outcome models with a binary endogenous explanatory variable and treatment regressions for our estimations (Jiang, Li, & Wang, 2012). Our results show that the instrumented dummy variables for loan-equity dual holding, bond-equity dual holding, and loan-bond-equity triple holding are positively associated with the likelihood of out-of-court restructurings. Second, we use mergers between lenders and shareholders as a source of plausibly exogenous shock to formation of simultaneous holding of loans and equity, following (Chu, 2017). The key assumption for this identification strategy is that mergers between financial institutions are often the result of consolidation in the financial sector in response to regulatory changes and are thus exogenous to a firm s specific fundamentals. Our treated sample includes firms whose dual holding increases because their lenders have merged with another institution that holds equity and the mergers occur in the quarter immediately before the onset of distress. For each treated firm, we find a matched firm that has similar firm characteristics as the treated firm but no simultaneous holdings. 5

6 Moreover, we restrict the control group to be firms whose lenders merge, but the merger does not result in any dual holding, to alleviate the concern that the merger of the financial institutions may affect the lender s interaction with the borrowing firm in a way that can alter the distress outcome. We find that the induced-by-merger dual holding of loans and equity is associated with a 46% higher likelihood of out-of-court restructuring. To shed more light on simultaneous holders incentives to internalize the conflicts and avoid the deadweight costs, we analyze whether and how the main relationships differ by the intensity of equity holding. Intuitively, the incentive alignment to facilitate a cost-effective workout is stronger when dual holders have large equity positions. We interact debt-equity dual holders debt intensity with the dummies, indicating the middle tercile and the top tercile of their equity holdings. We find that both the coefficient estimates for the top-tercile dummy of equity holdings and the interaction term of debt intensity and the top-tercile dummy are positive and statistically significant. Our results suggest that simultaneous holders of debt and equity have even stronger incentives to avoid bankruptcy when they have large equity skin in the game as, according to the absolute priority rule, equity is likely to be the first type of security to be wiped out in bankruptcy. We conduct a set of heterogeneity tests to provide further evidence of the underlying mechanism for our empirical results. Specifically, we investigate whether the relation between simultaneous holdings and the likelihood of out-of-court restructuring varies by the expected magnitude of the deadweight costs of bankruptcies. The three proxies for the bankruptcy costs are firm size (complex firms are expected to take longer to restructure in bankruptcy and, thus, incur higher costs), asset specificity (firms with specialized assets find it more costly to implement asset restructuring), and employees outside job options (firms whose employees have outside jobs are more likely to lose labor in bankruptcy). These three measures, therefore, capture both the direct and the indirect costs of bankruptcy related to fire sales and human capital (Altman, 1984; Shleifer and Vishny, 1992; Pulvino, 1999; Graham, Kim, Li, and Qiu, 2016). Our empirical evidence suggests that the positive relation between simultaneous holdings of debt and equity and out-of-court restructuring is stronger (in both statistical significance and economic magnitude) when bankruptcy costs are expected to be high. Interestingly, we find that the negative relation of dual holding of loans and bonds and the probability of out-of-court restructuring is mitigated in firms with high bankruptcy costs, suggesting that dual holders of senior debt claims 6

7 have serious concerns about bankruptcy filings when their holdings are likely to be affected by high bankruptcy costs. In the last part of the analysis, we investigate the effects of simultaneous holdings on real economic outcomes to provide suggestive evidence about efficiency implications whether dual holders or triple holders help rehabilitate economically viable but financially distressed firms. We find that distressed firms with simultaneous holding have higher profitability over the three years after distress. Furthermore, we show that these firms experience higher stock returns after distress resolution. The evidence suggests one channel through which simultaneous holders can capitalize on the value created by cost savings. This paper relates to the literature on the resolution of financial distress. Understanding distress resolution is important on two dimensions. First, the deadweight costs associated with the resolution of distress have important implications for corporate investment and financial policies (Modigliani & Miller, 1958; Stiglitz, 1974). Second, the form of restructuring is closely related to the efficiency of asset allocation in distress (Gertner & Scharfstein, 1991). Prior empirical studies (Gilson et al., 1990; Asquith, Gertner, and Scharfstein, 1994; Chaterjee, Dhillon, and Ramirez, 1996; James, 1996, Demiroglu and James, 2015) document important determinants for restructuring out of court. They find empirical support for the hypothesis that debt structure and bank participation are important factors in resolving creditor coordination and holdout problems. Our paper contributes to this literature in several ways. First, distinct from these studies, we consider the possibility of institutions as holders of multiple classes of security claim. In particular, we examine how simultaneous holdings of debt and equity help align incentives and internalize restructuring costs, to the benefit of both types of holders. Second, we document that both the incentives for, and the ability of, debt holders to influence the resolution of distress matters to restructuring outcomes. Finally, from a methodological prospective, we lay out a framework for identifying and constructing out-of-court restructurings through distressed exchanges and private debt negotiations. This paper also helps improve our understanding of the incentives for simultaneous holdings of debt and equity and their effects on corporate outcomes and security pricing. A growing body of empirical studies documents the fact that simultaneous holding helps mitigate the conflicts of interest between equity holders and debt holders and strengthen incentive alignment. Jiang et al. (2010) find 7

8 that syndicated loans of companies with non-commercial banking institutions simultaneous holdings of debt and equity have lower loan spreads than those without simultaneous holders. Relatedly, Chava et al. (2016) find that simultaneous holding significantly reduces the likelihood of having a capital expenditure restriction covenant in loan contracts. Bodnaruk and Rossi (2016) show that dual holding of a target s equity and bonds by financial institutions is associated with a lower merger premium and stronger bond market reactions in M&A transactions. Different from these studies, however, is the fact that we not only study the effect of simultaneous holdings in the sample of distressed firms but also investigate dual holdings of bonds and equity, dual holdings of loans and bonds, and triple holdings of loans, bonds, and equity. Studying how firms resolve their financial distress in the presence of financial institutions simultaneous holding helps further our understanding of the effect of incentive alignment on the mitigation of the deadweight costs of bankruptcy and the effect of the market-based mechanism in resolving financial distress. Our findings even provide implications for prior studies on loan contracting a lower loan yield spread with the presence of simultaneous holders might also be related to the expected avoidance of the deadweight costs of bankruptcy. We structure the rest of the paper as follows: section 2 provides a detailed description of our sample construction, section 3 presents our empirical strategies and main findings, and section 4 concludes. 2. Data and Sample Construction 2.1. Distress and Resolution Identifying private workouts has always been an empirical challenge, partly due to the fact that there are neither universal definitions on what types of negotiations and debt contract amendments constituting a formal private workout nor standard databases tracing these private restructurings. Expanding prior studies (e.g. Gilson et al. 1990; Demiroglu and James, 2015), we employ a multitude of data sources including firm stock returns and financial characteristics, news headlines from RavenPack New Analytics, loan covenant violations, and textual analysis of corporate filings to assemble data on out-of-court restructuring between 2000 and To the best of our knowledge, this sampling approach leads to a private workout sample that is the largest and the most comprehensive in the literature. Appendix A1 presents our sampling algorithm. 8

9 Distressed Firms We use a combination of stock performance, book leverage ratio, and EBITDA/interest ratio (i.e. interest coverage ratio) to determine whether a firm was in potential financial distress in a given year following prior studies (e.g. Gilson, 1989, 1990; Gilson, John and Lang 1990; and Demiroglu and James, 2015). Specifically, we calculate the three-year cumulative returns of common stock for all firms (excluding utilities and financial firms) in the CRSP/Compustat universe and keep those firms that rank in the bottom 15% at the end of each calendar year. We then remove those firms with book leverage below 30% or EBITDA/interest ratio greater than three in any of the preceding three years as these firms are unlikely to be in financial distress despite the poor stock performance. Finally, we remove firms that have their books asset below $50 million in all three consecutive years to keep a consistent size measure with our bankrupt firms. Our sampling procedure results in 4,161 firm years as in potential financial distress. Appendix A2 provides a detailed description on our use of various metrics based on stock performance, leverage ratio, EBITDA/interest ratio, and stock price to identify distressed sample. The combination of using stock returns in the bottom 15%, leverage ratio, and EBITDA/interest ratio captures the highest percentage of bankruptcies (discussed below) in our sample (425/605=70%) as well as the highest percentage of bankruptcies in the distressed firm sample (425/4161=10%) Bankruptcy Sample We draw all bankruptcy filings (both Chapter 11 and Chapter 7 filings) by U.S. firms with $50 million book assets or more at filing from 2000 to 2014 from New Generation Research (NGR) s Bankruptcydata.com. We manually match the bankrupt firms with Compustat using name, location and SIC code, and keep only those filings for which we can identify a valid gvkey in Compustat. This procedure results in an initial sample of 1,004 bankruptcy filings (with 971 Chapter 11 filings and 36 Chapter 7 filings) for the study. We then merge the bankruptcy sample with the CRSP/Compustat merged database. Given that we will use stock performance to identify distressed firms, we keep only CRSP firm years for which we are able to calculate stock returns in a given year and those whose book assets are at least 4 Before determining whether a firm completed an out-of-court restructuring, we collapse distress firm-year observations into distress events. Specifically, we treat firms in distress in consecutive years as one event. For example, if firm A is recognized as being distressed in years and disappears from our distressed firm-year sample after 2003, we collapse the three years into one distress event. We have a total of 2836 unique distress events. 9

10 $50 million in any of the preceding three years before bankruptcy filing from 1998 to We remove companies that are utilities (SIC codes ) or financials (SIC codes ). This filtering process results in a total number of 53,204 annual observations. We are able to match 605 out of the 1,004 bankruptcy filings (with 596 Chapter 11 filings and 9 Chapter 7 filings) in the sample period. This makes our final sample of bankruptcy filings for the study Out-of-Court Restructuring We adopt a two-step procedure to identify all out-of-court restructurings in our sample. First, we follow the definition provided by Gilson, et al. (1990) to identify distressed exchange offers and debt maturity extensions. 6 Specifically, we require that the due interest or principal payments be reduced or waived or the creditors exchange their debt claims into new debt securities and/or equity securities with concessions not receiving par payment. We also consider the extension of original debt maturity as a form of restructuring. Unlike prior studies that rely on news manual search through Down Jones Newswires and LexisNexis, which typically result in a small study sample, we resort to RavenPack News Analytics to identify out-of-court restructuring. 7 We consider news headlines that are in the following categories: bankruptcy-fears, credit-extension-recipients, debt, debt-extension-recipients, debtextension-provider, debt-increase, debt-reduction, debt-renegotiation, debt-restructuring, debtrestructuring-approval, debt-restructuring-complete, debt-restructuring-considered, debtrestructuring-failed, debt-restructuring-rejection, debt-shelf-registration, loan-recipient, mixedshelf-registration, note-acquisition, and note-sale. We are able to retrieve a total of 225,791 news headlines from 2000 to August Merging our distressed events with RavenPack results in 1,362 relevant news headlines for 1,336 distressed events. We then read the original news articles and search firm s SEC filings (8-K, 10-Q or 10K) to verify whether the firm went through an out-ofcourt restructuring. We are able to identify 394 cases, out of which 240 went through a distressed exchange and 154 received debt maturity extension through private negotiations. Appendix A3 6 Their definitions and classifications of out-of-court restructuring are used by several studies (e.g. Demiroglu and James, 2015). 7 RavenPack News Analytics is a comprehensive news database that monitors and analyzes real-time firm-specific business news from more than 19,000 news sources, including Dow Jones Newswires, Wall Street Journal, Barron s and MarketWatch. It covers more than 40,000 entities in over 200 countries. Ravenpack and has been used by several studies (e.g. Gao, Parsons, Shen, 2017). 10

11 provides a detailed illustration on our procedure using RavenPack to identify out-of-court restructurings. Second, we resort to identifying private renegotiations using loan covenant violations and searching corporate filings for all distressed events for which we cannot clearly identify the completion of out-of-court restructuring in the first step. We choose covenant violations because they act as necessary conditions for default in private debt contracts. This step is to augment our sampling using RavenPack search given its limited coverage on smaller firms as presented in Appendix A3. We start with firms from the distressed sample that have 4 consecutive quarters of covenant violation as a sign of worsening performance and financial difficulties. We use a threshold of 4 quarters to ensure that these covenant violations are not mild violations, which are quickly waived and do not lead to distress. Out of 2143 distressed firm-years that remains after matching with RavenPack, we identify 990 firm-years with 4 consecutive quarters of covenant violation 8. We then pull firm s 10Ks from SEC EDGAR and search for any mentioning of default, fail to pay, extend maturity, renegotiate and similar terms in the Liquidity part of Management Discussion & Analysis section. We exclude technical defaults by removing search results that explicitly state in their 10Ks that their covenant violations are not related to payment. Our goal is to find evidence of actual default or default that would otherwise occur if firms did not take any measures. We then manually check their filings (10Ks, 10Qs, 8Ks) to determine if the firms have privately renegotiated their debt with creditors. 9 Appendix A4 provides selected examples of language used in 10Ks regarding the classification. This sampling method results in 185 private renegotiations, of which 81 are payment waiver and 104 are maturity extension. Therefore, in total our sample consists of 579 out-of-court restructurings. [Figure 1 here] 8 Data on covenant violations is from Chava & Roberts (2008) and Nini, Smith, & Sufi (2012). We thank Amir Sufi and Michael Roberts for making this data set publicly available on their websites. 9 The distressed events that do not lead to bankruptcy, out-of-court restructuring, or exit from the sample because of merger or going-private transactions are regarded as self-recovery (i.e., a firm still exists in Compustat within two years after the end of the distress event).there are two reasons why self-recovery firms exist in our sample. First, the classification of distress event is imperfect, and there are misclassified firms that have no financial difficulty. Second, some firms are able to recover through asset restructuring such as divesting assets or receiving capital injection from existing security holders, which our data does not capture. 11

12 Figure 1 presents the frequency of bankruptcy and out-of-court cases over the period with the bankruptcy sample ending in The first observation that stands out from the graph is the large spikes in number of distresses around the 2001 Dot.com crisis and 2008 financial crisis. Second, over time as out-of-court restructuring becomes more common, the number of bankruptcy cases significantly drop, closing the gap between the two types of restructuring. While in the 2001 crisis, there were three times more bankruptcy filings than out-of-court restructurings, the occurrence of the two types are approximately equal during the recession. Interestingly, the practice of simultaneous holding started to take off around early 2000s after a series of deregulations in financial industry at both state and federal levels and became prevalent in the second half of our sample period Simultaneous Holdings of Debt and Equity We identify four types of simultaneous holdings at the holder level. The first type is the loanequity dual holding, i.e., institution investors simultaneously holding equity and loans of the same firm. The second type is the bond-equity dual holding, i.e., institutional investors simultaneously holding equity and bonds of the same firm. The third type is the loan-bond dual holding, i.e., institutional investors holding loans and bonds of the same firm. The fourth type is the loan-bondequity triple holding, i.e., institutions holding loans, bonds and equity of the same firm. Holding measures are first matched at holding level (facilities and issues in case of loan and bond respectively), then accumulated to holder level to identify simultaneous holding and finally aggregated to firm level for analysis. We rely on Thomson Reuter s 13F filing database to identify institutional shareholders, Reuters Loan Pricing Corporation s (LPC) DealScan syndicated loan database to classify loan lenders and Thomson Reuters emaxx ASCII Bond Holding Data Feed 10 to identify bondholders. All three types of holders are manually matched on names, types of business structure and addresses (wherever possible). Firms are matched on GVKEY from Dealscan by Michael Robert s matching database and then historical CUSIP to 13F and emaxx through CRSP/Compustat. We classify a lender as a simultaneous holder if either the lending institution itself or at least one of its first-level subsidiaries holds a significant equity and/or bond position in the borrowing 10 The emaxx database has been used by several studies e.g. Hasan, et al. (2013) and Bodnaruk and Rossi (2016) 12

13 firm. We rely on two sources to identify all first-level subsidiaries of the lenders in our sample. The first source is lending institutions' (if they are publicly traded companies) 10K filings from various years; and the second source is Federal Financial Institutions Examination Council's (FFIEC) website if the 10K filings are not available or if the organization structure is not provided in the 10K filings. Dealscan database identifies the direct lenders and ultimate parent of the lenders to a loan. However, matching either lender or ultimate parent names in Dealscan directly to names in 13f is insufficient and may lead to false negatives because direct matching does not take into account the dynamic expansions and restructurings of financial institutions over time, especially through acquisitions and spin-offs. We use SDC Platinum M&A database to track changes in lender ownership. However, SDC database only has information at branch level. Therefore, for loans that Dealscan does not identify lenders to branch level, we collect information on mergers at holding company level from institution history section of (FFIEC) database or company website. For bond-equity holders, we adopt a similar approach. emaxx database contains detailed information on holdings of institutional fixed income investors at both fund level and managing firm level, on a quarterly basis. We manually identify the ultimate parent of the institutional bond investors and then match to the parent firm of institutional stock investors from 13F by firm name. We also consider the changes of parent firms due to dynamic expansions and restructurings of financial firms over time. All simultaneous holdings are measured over at least four consecutive quarters from the last quarter of a distress event which is the quarter we identify resolution outcome. Specifically, we back track the simultaneous holder s holdings from the end of distress event and keep only those that cover at least 4 consecutive quarters (or the whole distress event period if the whole period is less than 4 quarters). Alternatively, we use a total measure by requiring that the simultaneous holder s holdings cover at least 4 quarters (or the whole distress period) during the distress event and do not require consecutive quarters or the coverage starts from the first or the last year-quarter. 11 We employ three measures to capture variations in simultaneous holding from different perspective. The first one, Dummy, is a dummy variable that takes value of 1 if a firm has dual 11 Results using this alternative total measure are reported in Appendix C1. 13

14 holders that satisfy the timing conditions discussed above during the distress event and 0 otherwise. The second one, Debt intensity, is a continuous measure that exploits variation in the magnitude of dual holders debt position to capture the idea that the bigger positions dual holders have, the more influential they are in the resolution process. Specifically, for a particular firm in a quarter, we calculate the intensity measure as total debt held by the dual holder as a fraction of total loan and bond outstanding Summary Statistics Distressed Firms and Resolution outcomes We remove 188 distressed events that are either too short (less than two quarters from the start to the end of distress) or we cannot identify information on firm fundamentals. This leaves us with 996 events for all of our empirical analysis. Panel A of Table 1 shows that our sample of distressed firms are well balanced in terms of resolution outcome, 49.6% of the distressed firms went through out-of-court workout and the rest resort to court-supervised reorganization. The statistics is generally in line with those documented in previous studies such as 42.7% in Gilson et al. (1990) or 50.1% in Demiroglu and James (2016). Panel B presents the summary statistics of simultaneous holdings for all the distressed firms. On average, loan-equity dual holders appear in 21.7% of all the distressed firms and 29.0% of distressed firms that have syndicated loans, implying that this type of investor is relatively common. In addition, out of all the distressed firms, 9.7% have bond-equity dual holders, 4.3% have loanequity dual holders and 4.4% have triple dual holders. Conditional on having simultaneous holding, these simultaneous holders collectively hold 38.5% of an average firm s total loan outstanding, and 34.3% of firm s total bond outstand. For loan-bond dual holders and loan-bond-equity triple holders, this variable is 10.8% and 18.9%, respectively. These numbers indicate a significant position that dual holders represent in a firm s debt structure. [Table 1 here] When we further compare the group of distressed firms that go through workout vs. the group of distressed firms that go to bankruptcy in Table 2, the analysis shows that the average firm in 12 Similarly, we construct an alternative measure, Debt Count, calculated as number of dual holders as a fraction of total loan lenders and bondholders. The empirical results using this measure are presented in Appendix C2. 14

15 bankruptcy group is larger, has larger fraction of tangible assets (mean difference in tangibility for two groups is 4.5%), and higher book leverage (mean difference is 0.107). These stylized facts are consistent with a priori as large firms and firms with high leverage tend to have complex structure of claimants and, thus, are harder to have successful out-of-court negotiation (Moulton and Thomas 1993, Campbell 1996). Firms with small portion of tangible asset are less likely to preserve values of their asset through bankruptcy, therefore tend to work out out-of-court. Moreover, firms in outof-court group spend more in R&D, have higher cash holding as well as higher debt concentration ratio (Debt HHI). High R&D expense indicate higher intangible assets; therefore it is hard for firms to preserve values during bankruptcy process. Higher debt concentration ratio indicates that the debt claims are clustered in small number of debt claimants, thus firms can mitigate holdout and coordination problems and are more likely to successfully restructure out-of-court. [Table 2 here] When we compare the dual holding variables for firms that restructure out of court vs. those in bankruptcy, the results in Table 2 again show that the out-of-court group has a significantly higher percentage of firms with simultaneous holdings. For example, on average 30.4% out of court sample firms have loan-equity dual holders while only 13.1% bankruptcy sample firms have loan-equity dual holders. The difference is significant at the 1% level. Consistently, bond-equity dual holders appear in 13% out of court firms, but only appear in 6.6% bankruptcy firms, and in terms of triple dual holders, the different is also significant Comparison of firms with and without dual holding To compare distressed firms with and without dual holding, we perform multivariate regressions on the determinants of dual holding. Panel A of Table 3 shows that firm characteristics such as size, tangibility, and leverage are strongly correlated with the presence of dual holders in distressed firms. Specifically, larger firms, firms with more tangible assets, lower leverage and more complex debt structure are more likely to have dual holders. This suggests that the two sets of firm are indeed different, and likely not only in observable characteristics but also along the unobservable dimensions as well. Therefore, it is critical that our regression control for firm characteristics and our empirical strategies take measures to address this issue. [Table 3 here] 15

16 Panel B of Table 3 reports correlation coefficients for different types of dual holding measures. While correlation among measures within one type is high which is trivial, measures across different types of holders are weakly correlated. We observe investor market specialization, to certain extent, in our sample i.e. we rarely observe the same institutional investor participating in both syndicated loan and bond market at the same time. 3. Empirical Results 3.1. Baseline Results The baseline specification examines to what extent the presence of dual holding explains the likelihood of private workout. We first run ordinary least square regressions of distress outcome on our measures of dual holders as follows: Outcome i = βle i + θloan i + γ k X k,i + Year FE + Industry FE + ε i Outcome i = βbe i + ρbond i + γ k X k,i + Year FE + Industry FE + ε i Outcome i = βlb i + θloan i + ρbond i + γ k X k,i + Year FE + Industry FE + ε i Outcome i = βlbe i + θloan i + ρbond i + γ k X k,i + Year FE + Industry FE + ε i (1A) (1B) (1C) (1D) In the above equations, Outcome equals one when the distress firm is restructured out of court, and zero if the firm filed for Chapter 11 bankruptcy. LE, BE, LB, and LBE are measures of four types of simultaneous holding: loan-equity, bond-equity, loan-bond, and loan-bond-equity holding, respectively. Specifically, for each of the four simultaneous holder types, we employ both Dummy and Debt Intensity measures. The two measures are constructed on a quarterly basis and then collapsed to distress event level by taking average over the distress period. We include Loan and Bond the total amount of loan and bonds scaled by the total liability of the firm in these regressions to alleviate the concern that our results may be driven by the effect of loan holders or bond holders rather than dual holders. By including loan and bond amount, we are identifying the effect of dual holdings above and beyond the effect of loans and bonds themselves. We motivate our control variables following prior studies on out-of-court restructuring. These variables act as proxies for factors that influence (1) different classes of claimant incentives to agree to a private resolution and (2) the ease to reach an agreement, which together determine whether financial distress can be resolved privately. A priori, firm that are expected to incur large costs of bankruptcy have strong incentives to achieve workouts, while firms that expect to benefit 16

17 from specific provisions of the bankruptcy code are more likely to restructure in court. Moreover, the lower-cost alternative will be adopted only if claimholders can agree on how to share the cost savings which in turns depends on the complexity of the bargaining process as well as relative bargaining powers of related parties. The control variables are: Firm Size log of sales (SALE), Cash holding Cash (CHE) scaled by total assets (AT), Cash Flow operating cash flow (OIBDP) divided by total assets (AT), Leverage total liability (DLC + DLTT) scaled by total assets (AT), Tangibility total property, plant, and equipment (PPENT) scaled by total assets, R&D spending R&D expenditure (XRD) divided by total assets (AT) 13, Missing R&D A dummy variable that takes 1 if R&D expenditure (XRD) is missing, Debt concentration the Hirfindahl index of debt structure (Colla, Ippolito, and Li 2013) is used to proxy for the coordination problem. In addition to above mentioned control variables, we include year fixed effects and industry (SIC-2 digit) fixed effects to account for market-wide and industry-wide conditions that can have an effect on liquidation value and creditor willingness to negotiate out of court. For example, market illiquidity and worsen industry condition can severely reduce the liquidation value of some industryspecific assets, making creditors more willing to accept renegotiation. The estimates of Equations (1A) (1D) are presented in Table 4 for Dummy and Debt Intensity measures respectively. Model (1) in Panel A shows that firms with loan-equity dual holders (Dummy) are 33% more likely to go through out-of-court restructuring instead of bankruptcy, controlling for the loan fraction of total debt, and firm characteristics. When we consider bondequity dual holders (Model 2, BE) and loan-bond-equity triple holders (Model 4, LBE), the numbers are 19.5% and 44.8%, respectively. The presence of LB holders, on the other hand, is associated with an increase of 15% in probability of filing for bankruptcy. We interpret this finding as follows: since dual holders of debt claims have stronger bargaining power against shareholders in bankruptcy as well as a smaller expected gap between in and out of chapter 11 restructurings (James, 1996), they are more willing to enter bankruptcy. However, we do not rule out the possibility that dual holders choose to enter into the capital structure by investing in the most senior claims if they expect the probability of firm filing for bankruptcy is high. [Table 4 Here] 13 Treating missing XRD as zero. 17

18 In Panel B, the coefficients of Debt Intensity are statistically and economically significant for loan-equity dual holding, bond-equity dual holding, and loan-bond-equity triple holding. The coefficient estimate for LBE Debt Intensity is the largest: a one standard deviation increase in LBE Debt Intensity is associated with 25% increase in the probability of workout. It should be noted that as the likelihood of an institution holding loan, bond and equity of the same firm at the same time is small, firms tend to have substantially fewer LBE holders and LBE holders also make up a smaller portion of firm s debt than other types of dual holder. However, conditional on having simultaneous holdings, this type of investors can mitigate the conflicts among all three types of investors and are, therefore, in the strongest position to influence the resolution process. Overall, results in panel B confirm that the larger positions dual holders hold in a firm s debt structure, the stronger their ability to influence the restructuring process and, thus, a higher likelihood for firms to resolve privately. The coefficients for control variables are also consistent with our predictions and previous literature (Gilson et al., 1990, Moulton & Thomas, 1993, Campbell, 1996, Chatterjee et al., 1996). The coefficient of Firm size is significantly negative at the 1% level for all the four models, which is consistent with the fact that larger firms generally have more complex structure of claimants, therefore, are likely to experience more severe creditors coordination problems and harder to reach an agreement by negotiation, resulting in a tendency to resort to court-supervised reorganization. Tangibility and Cash holding both have significantly negative coefficient, as firms with low tangibility and low cash holding incur higher cost in bankruptcy and, therefore, have higher incentives to preserve values of their assets through out-of-court restructuring. High R&D expense indicate higher intangible assets, and it is hard for firms to preserve values during bankruptcy process, thus the probability of successful workout is increasing. Finally, firms with high leverage are harder to achieve out-of-court restructuring debt restructuring, as shown by the significant negative coefficient. So far we present the results based on linear probability regressions with fixed effects and standard errors clustered at industry level to correct for cross-sectional correlation and heteroskedasticity inherent in linear probability model (Petersen, 2009). In terms of econometric specification, there are trade-offs between bias and efficiency. Binary dependent variables such as Outcome require special consideration. Models that can accommodate their special scale, such as probit model are subject to inconsistent estimate caused by incidental parameters problem if the models include fixed effects (Greene, 2004). In contrast, models that give unbiased estimates with 18

19 fixed effects such as linear model may not be efficient for such types of data. As robustness check we also use probit regression without fixed effects and the results (in Appendix E3) are qualitatively and quantitatively similar as our linear regressions. In general, our baseline results show a strong relationship between the presence of simultaneous holdings of LE, BE and LBE and a more cost-effective distress resolution, reflected in higher probability of private workouts instead of bankruptcy Augmented Regressions One caveat of a cross-sectional regression as in the baseline regression is that it cannot control for unobservable firm characteristics. As presented in Table 3, firms with and without dual holding are different in terms of their observable attributes, suggesting they may be different along the unobservable dimensions as well. For instance, dual holders may target firms with better governance or less information problem and in turn firms with those characteristics are more likely to restructure privately (Gilson et al. 1990, and Hoshi, Kashyap, & Scharfstein, 1990). Consequently, the observed positive relationship between tendency to workout privately and the presence of dual holders may be due to their correlation with the unobservable firm characteristics. Such a potential endogeneity problem can lead to inconsistent estimates and affect statistical inference. To overcome the above-mentioned shortcoming of cross-sectional regression, we augment the set of control variables with additional variables that can account for the unobservables. We utilize information on the firm's dual holding over time as a proxy for firms' unobservable characteristics that correlates with the likelihood of having dual holding. By controlling for level of dual holding in the past, specification (2) is effectively comparing two firms with the same tendency to have dual holding, the only difference is the timing of dual holding in relation to distress event: one happens to have dual holding during distress while the other does not. Outcome i = βle i + δpastle i + θloan i + γ k X k,i + Year FE + Industry FE + ε i Outcome i = βbe i + δpastbe i + ρbond i + γ k X k,i + Year FE + Industry FE + ε i Outcome i = βlb i + δpastlb i + θloan i + ρbond i + γ k X k,i + Year FE + Industry FE + ε i Outcome i = βlbe i + δpastlbe i + θloan i + ρbond i + γ k X k,i + Year FE + Industry FE + ε i (2A) (2B) (2C) (2D) 19