In- and out-of-court debt restructuring in the presence of

credit default swaps ∗

Mascia Bedendo†

Universita’ Bocconi

Lara Cathcart

Imperial College Business School

Lina El-Jahel

Imperial College Business School

ABSTRACT

This paper investigates whether the availability of credit insurance via credit default swaps

(CDS) has influenced the debt restructuring process in a sample of U.S. reference entities.

Contrary to the predictions of the empty creditors theory, we do not find evidence that the

presence of CDS favors bankruptcy over a workout. Instead we observe that the probability

of filing for bankruptcy in the 2008-2009 crisis is significantly associated with leverage and

short-term debt ratios, a simplified debt structure characterized by a high proportion of secured

debt, and a drop in profitability incurred by the banks lending to the company.

JEL Classification: E51, G32, G33.

Keywords: Credit default swaps, empty creditors, debt restructuring, bank losses.

∗We thank Mark Flannery, Issam Hallak, and Christopher Hennessy for suggestions and comments. We acknowl-edge financial support from Carefin (Bocconi Centre for Applied Research in Finance).

†Corresponding author. Address: Department of Finance, Universita’ Bocconi. Via G. Roentgen,1 - 20136 Milan(Italy). E-mail: mascia.bedendo@unibocconi.it. Ph. +39 0258365973, fax +39 0258362970.

1. Introduction

What determines a distressed firm’s choice between in- and out-of-court debt renegotiation? A

large body of theoretical and empirical research, thoroughly reviewed in the survey by Hotchkiss,

John, Mooradian and Thorburn (2008), has addressed this question. Direct bankruptcy costs, intan-

gible assets, financial leverage, short-term liquidity, and debt structure are widely acknowledged

as key determinants of the outcome of the restructuring process. Although these findings are well

consolidated, the explosive growth of the credit insurance market in the last decade, mainly in the

form of credit default swaps (CDS), has renewed interest in the topic. The availability of credit in-

surance via CDS has been closely associated with the emergence of empty creditors, who may find

it convenient to force distressed firms into bankruptcy even when a workout would be efficient. In

this paper we revisit the initial question from this perspective, and empirically investigate whether

the presence of CDS has played a significant role in determining the outcome of the restructuring

process for a number of reference entities during the 2008-2009 crisis.

In a CDS contract the protection buyer pays a periodic premium to the protection seller, and

receives a settlement equal to the difference between the par and market value of the underly-

ing debt of the reference entity should a default event occur. Typical default events include the

bankruptcy of the debtor or the failure to pay principal / interests on the debt. Out-of-court debt

renegotiations, instead, formally constitute credit events only under particular CDS restructuring

clauses.1 In practice, no workout in the U.S. corporate segment has ever triggered a default in the

CDS market, given the general disagreement about what constitutes a restructuring event.2

In this context, legal scholars Hu and Black (2008a, 2008b) advance the empty creditor hypoth-

esis, and raise concerns that the presence of CDS may separate the creditor’s economic exposure

from his voting rights, thus altering the traditional debtor-creditor relationship. As a result, the in-

sured bondholder of a financially distressed firm has a strong incentive to favor bankruptcy (which

1Modified restructuring (Mod-R) for U.S. contracts, and modified-modified restructuring (Mod-Mod-R) for Euro-pean contracts.

2The Small Bang protocol released by ISDA in July 2009 is the first document to detail the auction settlementfollowing a restructuring event.

1

triggers the payment under his CDS contract) rather than a workout or a debt exchange. Along

the same line, Yavorsky, Bauer, Gates and Marshella (2009) explain why bondholders with CDS

protection and, more in general, investors with negative basis trades, are more likely to take a “hard-

line” in debt renegotiations with the issuer. In a theoretical framework, Bolton and Oehmke (2011)

argue that the presence of CDS can yield a market equilibrium characterized by over-insurance of

bondholders and an inefficiently high incidence of bankruptcy filings compared to the optimum.

Some anecdotal evidence in favor of the empty creditor hypothesis has been produced by

Hu and Black (2008a, 2008b), Bolton and Oehmke (2011), and the financial press (Risk (2009),

Economist (2009)). On the contrary, Mengle (2009) provides a preliminary analysis of the empty

creditor issue and questions its validity due to the lack of compelling evidence. His conclusions,

however, are based solely on the dynamics of the ratio of distressed exchanges to bankruptcies dur-

ing both the 2008-2009 crisis and previous pre-CDS crises, without accounting for other factors.

To our knowledge, no rigorous empirical investigation of the outcome of the debt renegotiation

process in the presence of CDS exists, and this study represents a first contribution in this direction.

A direct empirical verification of the empty creditor hypothesis is unfeasible, as it would require

data on the proportion of the firm’s unsecured debt which is insured through CDS, and such data are

not available.3 Nonetheless, we can provide an indirect verification by investigating the frequency

of bankruptcy filings in CDS reference entities, and to what extent those bankruptcies are driven

by the company’s fundamentals, in line with what we observe in non-reference entities.

Our sample includes all bankruptcy filings and distressed exchanges initiated by non-financial

U.S. rated companies over the period January 2008 - December 2009. This period represents a

good laboratory for looking at distressed firms’ debt restructuring, as it represents the first gen-

eralized crisis since CDS contracts have become liquid. Our empirical investigation is aimed at

verifying whether the presence of CDS significantly increases the likelihood of filing for Chapter

11, once we control for other determinants of the restructuring outcome. The control set includes

3Information on the 1000 most liquid single name CDS volumes outstanding has been provided by the DepositoryTrust and Clearing Corporation (DTCC) since November 2008, but the breakdown of the CDS volumes betweencovered and naked CDS positions is not available. Furthermore, even during a debt restructuring process, bondholdersare not required to disclose whether they hold CDS positions.

2

variables that former literature has found to play a significant role in explaining the restructuring

process for U.S. firms, i.e. operating performance, leverage, short-term debt ratio, as well as com-

plexity and seniority of the debt structure. Special attention is devoted to the impact of bank debt of

distressed firms on their restructuring choice for two reasons. First, the nature of bank debt in our

sample is very different and much more complex than that reported by most related studies (which

date back to the mid 1990s), following the impressive growth of the syndicated loan market in the

past 20 years (Sufi (2007)). Second, since the 2008-2009 crisis originated in the financial sector,

it is interesting to investigate whether the distressed conditions of banks during the crisis have had

an impact on the restructuring outcome of their borrowers.

We conduct both a univariate analysis to highlight any differences in the determinants of the

restructuring method (distressed exchange vs. Chapter 11) between CDS reference entities and

non-reference entities, and a probit regression to test whether the presence of CDS has increased

the probability of filing for bankruptcy for reference entities. We report three main results. First,

we do not find evidence that companies whose bondholders can purchase credit insurance via CDS

are more likely to restructure their debt in court. In fact, the outcome of the debt renegotiation

process seems to be driven by essentially the same variables for both reference entities and non-

reference entities. Second, in agreement with previous literature on debt restructuring, we observe

that firms that file for Chapter 11 report higher leverage, higher short-term debt ratios, and a larger

proportion of secured debt than companies that reorganize out of court. Third, bankrupt firms in our

sample are characterized, on average, by a more simplified debt structure, mainly concentrated on

bank loans. This evidence is in contrast with a large body of existing research, according to which

banks are more likely to overcome lender coordination problems, thus facilitating debt workouts.

However, it is consistent with the peculiarities of a financial crisis, where financial institutions are

the first to be distressed, and may not be in a position to refinance their customers (see Khwaja and

Mian (2008), Chava and Purnanandam (2011) on the impact of banking crises on bank-dependent

borrowers). In line with this explanation, we observe that the likelihood of filing for bankruptcy is

positively related to the profitability shocks experienced by the main bank lenders of the firm.

We perform further tests of the empty creditor hypothesis by looking at the recovery rates of

3

the defaults in the sample. Even though the presence of CDS does not seem to drive the restructur-

ing outcome, insured bondholders may still condition the restructuring process to some extent. If

this is the case, we would expect to observe higher recovery rates for reference entities in both: (i)

distressed exchanges, given the better conditions that tendering debtholders might be able to ne-

gotiate; (ii) bankruptcy, if the pressure of empty creditors forces bankruptcy to occur prematurely,

when the financial resources are still sufficient to ensure satisfactory recoveries. In our sample,

however, the availability of credit insurance does not help explain the recovery prices of either dis-

tressed exchanges or bankruptcies, once the standard determinants of recovery values (Acharya,

Bharath and Srinivasan (2007)) have been taken into account.

As final checks, we compare the dynamics of the CDS basis for the reference entities that went

bankrupt and those that arranged a workout, and we investigate, from a descriptive perspective,

the characteristics of out-of-court debt reorganizations in the sample. Again, our results fail to

uncover a significant role of insured debtholders in the restructuring process. We conclude that,

even though the empty creditor issue might have been a specific concern for some companies,

it was not a phenomenon of significant proportions for the generality of reference entities in the

recent crisis.

Our paper contributes to the existing research in several ways. Our findings on the impact

of CDS on the debt renegotiation process of reference entities enrich both: (i) the fast-growing

literature on empty creditors, which lacks empirical evidence in this respect; and (ii) the empirical

literature on the costs and benefits of CDS for firms, which is still in its infancy, as highlighted by

Stulz (2010).4

At a broader level, we contribute to the more established literature on the determinants of

in- and out-of-court debt restructuring (see Gilson, John and Lang (1990), Franks and Torous

(1994), Asquith, Gertner and Scharfstein (1994), Chatterjee, Dhillon and Ramı́rez (1996), Brunner

and Krahnen (2008), Jostarndt and Sautner (2010)) by: (i) revisiting the role played by classic

4Part of this stream of literature investigates the effects of CDS on the credit market. Notable contributions areby Hirtle (2009), who shows that the presence of CDS yields an increase in bank credit supply and an improvementin credit terms, and by Ashcraft and Santos (2009) who suggest that the introduction of CDS may reduce the cost ofborrowing for safe and transparent firms.

4

determinants of the restructuring process during the 2008-2009 financial crisis; (ii) exploring, in

particular, how the role played by bank lenders has changed in a crisis where the financial sector

has been severely hit and most bank debt is in the form of syndicated loans.

Our research has also interesting policy implications in terms of regulation of the CDS market.

Various measures have been suggested to limit the distortions introduced by CDS on the debt

renegotiation process, such as a revision of the bankruptcy law, or the disclosure of CDS positions

held by bondholders. However, in order to choose the most appropriate measures, it is important

to first understand to what extent the restructuring process of reference entities has been affected

by CDS in practice, and our findings provide useful insights in this respect.

The rest of this paper is organized as follows. Section 2 describes the data and sample collec-

tion. In section 3 we review the determinants of distressed exchange and bankruptcy. We discuss

the methodological steps and the main findings in section 4. Section 5 runs some further tests on

the existence of empty creditors. Section 6 concludes.

2. Data and Sample Selection

Our initial sample includes all bankruptcy filings (under either Chapter 7 or Chapter 11) and dis-

tressed exchanges completed by U.S. rated companies over the period January 2008 - December

2009, as reported by Moody’s Default and Recovery Database. The bankruptcy filings include

prepackaged Chapter 11 cases since they constitute regular credit events in CDS contracts.

According to Moody’s (2000), for a debt restructuring to be classified as a distressed exchange,

two conditions must be satisfied: (i) the issuer offers creditors a new or restructured debt, or a

new package of securities, cash or assets, that amount to a diminished financial obligation relative

to the original obligation, and (ii) the exchange has the effect of allowing the issuer to avoid a

bankruptcy or payment default. Debt exchanges arranged by financially healthy issuers to reduce

their debt levels at attractive market conditions should be excluded by the above definition. Fol-

lowing Mooradian and Ryan (2005), we verify the real nature of the debt restructurings through

5

a detailed search for relevant news on the individual debt exchanges on Factiva, which confirms

that all events classified by Moody’s as distressed exchanges actually refer to financially distressed

firms, and should, therefore, be included in our sample.5

We apply a first filter to discard from the sample all debt exchanges followed by a bankruptcy

filing within six months. Multiple distressed exchanges that occur during the same six-month pe-

riod are instead combined together. We limit our sample to rated companies to ensure a more

meaningful comparison between CDS reference entities (which are always rated given that a sig-

nificant portion of their debt is publicly traded) and other firms in the sample.6

The choice of the sample period is dictated by two considerations. First, the CDS market has

been very liquid over the last five years only, hence the proportion of corporate defaults where

the obligor is a CDS reference entity was negligible prior to 2005. Second, the defaults recorded

during the 2008-2009 crisis have been characterized by an unusually large recourse to distressed

exchanges (Altman and Karlin (2009)), comparable to the debt workouts boom observed in the

1980s. This can be explained with: (i) the generalized shortage of liquidity and, consequently,

of debtor-in-possession (DIP) and exit financing, which has increased the uncertainty linked to

bankruptcy filings; (ii) the approval, in February 2009, of the American Recovery and Reinvest-

ment Act (“The Recovery Act”) which allows the deferral of the income arising from debt can-

cellations arranged in 2009 and 2010 until 2014, and its amortization over the following five-year

period. Given that out-of-court debt renegotiations accounted only for 9.5% of the total number

of corporate defaults over the period 2005-2007, we restrict our sample period to 2008-2009 for

consistency. The overall number of in- and out-of-court restructurings for U.S. rated firms over

this period is equal to 239.

Previous literature has shown that the composition of capital and debt structure plays a major

5As a further check, we compare mean and median values of interest coverage ratio and EBITDA-to-total assetsratio at year-end before default for both distressed exchanges and bankruptcy filings, and we find them not to besignificantly different at 5% confidence level.

6Some previous empirical studies on the determinants of the default method (Gilson, John and Lang (1990),Asquith, Gertner and Scharfstein (1994)) followed a different approach in selecting the sample. First, they iden-tified all financially distressed companies by looking at their stock performance over a certain period. Then theyinvestigated the actions taken by those companies in response to distress. Such an approach is preferable when theanalysis includes non-rated firms, whose defaults are not entirely covered by the Moody’s database.

6

role in explaining the outcome of the debt restructuring process. Detailed information on these

variables is available from the annual 10-K filings, hence we exclude those firms which did not

file a 10-K report at the end of the fiscal year prior to the default event.7 Further, we eliminate

companies that underwent major reorganizations (M&A, spin-offs) between the last 10-K report

date and default. We also drop all financial companies, given the peculiarities of their capital

structure. Our final sample includes a total of 121 default events, 41% of which are distressed

exchanges, and 59% are Chapter 11 filings.8 The sample size is quite small, albeit comparable to

the size of existing related empirical studies (Gilson, John and Lang (1990), Asquith, Gertner and

Scharfstein (1994), Franks and Torous (1994)).

Finally, we need to identify the default events pertaining to CDS reference entities. For this

purpose, we match the firms in our sample with the list of CDS reference entities provided by

MarkIt. We identify 55 matches, which account for 45% of all defaults, thus allowing a meaningful

comparison between default types in reference entities and other firms. An attentive investigation

of the dynamics of the CDS spreads of those companies in the few weeks before the default event

reveals that, in half of the cases, the CDS was actively traded, while in the remaining cases the CDS

was fairly illiquid. Since we cannot think of any a priori reason why the proportion of bondholders

who purchase insurance via CDS should be significantly different in the two cases, we decide to

keep in the sample the reference entities of both liquid and illiquid CDS contracts.9

The composition of our sample of in- and out-of-court restructurings is reported in Table 1.

Most default events (75% of the total) occur in 2009, at the peak of the economic crisis. For the

reasons outlined above, distressed exchanges become more popular in 2009, when they account for

46% of the default events, against 27% in 2008. The industry classification of the defaulted compa-

nies reveals that the sample is distributed across all sectors, with a concentration in manufacturing

7The average number of days between the last report date and the default event is 182 for distressed exchangesand 184 for Chapter 11 filings. Using the last available 10-Q quarterly report prior to default would provide a moreaccurate picture of the financial and economic conditions of the distressed firm, but is unfeasible due to the lack ofdetails on the debt structure for most 10-Q filings.

8There are no Chapter 7 filings in our final sample. Hence, in the rest of the paper, we will use “bankruptcy” and“Chapter 11” interchangeably.

9A separate analysis of the two groups would be desirable, but is unfeasible due to the small sample size.

7

industries, which account for more than 50% of all defaults.10 Around 55% of the default events

are associated with public firms, while the remaining 45% refer to privately held firms with public

debt, most of which had been taken private during the leveraged buyout boom of 2005-2007. The

distribution of default events between distressed exchanges and bankruptcies is not significantly

different across the two sets of firms.

INSERT TABLE 1 HERE

Surprisingly, we find that the incidence of in- and out-of-court restructurings in CDS reference

entities is almost identical to the one observed in other firms: 60% of all defaults of CDS reference

entities are represented by bankruptcy filings, against 58% for other companies. This seems to

contradict the claim that reference entities are more likely to file for Chapter 11 under the pressure

of insured creditors. The frequency of bankruptcies among reference entities with very liquid CDS

is even lower (52%), partly due to the fact that only 15% of the defaults of these firms occurred

in 2008 (when Chapter 11 filings were relatively more frequent, as noted previously), against a

corresponding percentage of 54% for reference entities with less liquid CDS.

One may argue that our final sample of defaults may have been biased by the filtering procedure

and, as a result, may not be representative of the entire universe of the default events pre-screening.

We find this not to be the case, as the proportions of distressed exchanges and Chapter 11 filings

for the default events excluded from our sample are very similar (39% and 61%, respectively) to

the corresponding proportions in-sample (41% and 59%, respectively). In fact, the defaults of the

reference entities included in the sample seem to be biased towards bankruptcy, as 60% correspond

to Chapter 11 filings, against only 47% for the reference entities excluded from the sample.11

10Similar findings were reported by Chatterjee, Dhillon and Ramı́rez (1996).11It is worth noticing that two thirds of the reference entities not included in the sample are financial companies

with widely traded CDS, most of which benefited from government aid during the crisis.

8

3. The Determinants of the Restructuring Method

In assessing whether distressed reference entities are more prone to bankruptcy than other firms, we

need to control for a set of variables that former literature has found to be significant determinants

of the outcome of the restructuring process, namely operating performance, leverage, and debt

structure.12

Data on the values of those variables at the fiscal year-end prior to default are collected man-

ually from the 10-K filings available from the SEC’s EDGAR database. Specifically, data on the

operating income and the financial leverage are obtained from the consolidated financial state-

ments, while the debt structure is detailed in the notes to the financial statements.

3.1. Financial and Economic Distress

The degrees of financial and economic distress have been included in most empirical analyses of

the determinants of the restructuring method. However, while indebtedness level and poor perfor-

mance are essential ingredients for pushing a company into distress, it is not clear whether and how

they should play a role in determining the restructuring outcome. We can expect highly-leveraged

companies with a large proportion of debt due in the short run to face time constraints in attempting

to reorganize their debt out of court, which can lead to a higher chance of filing for bankruptcy. In

addition, we can foresee this effect to be particularly strong during times of generalized liquidity

shortage in the economy. The existing empirical findings on the link between restructuring choices

and financial / economic variables are mixed. Chatterjee, Dhillon and Ramı́rez (1996) document

a lower operating performance and a lower leverage for companies that renegotiate in court, as

well as a higher proportion of short-term debt, which is viewed as a proxy of the firm’s liquidity

needs. Gilson, John and Lang (1990), Franks and Torous (1994), Asquith, Gertner and Scharfstein

(1994) also observe more stringent liquidity needs in firms that file for bankruptcy, whose levels of

12We have not included other variables that may play a role in discriminating between different restructuring meth-ods, such as direct bankruptcy costs or market-based solvency measures, due to data unavailability for the firms in oursample.

9

leverage and operating performance, however, are in line with those of firms that complete a debt

workout.

We measure firm leverage with the ratio of total debt to total assets, and operating performance

with the ratio of EBITDA to total assets. To correct for industry-specific effects on capital structure

and operating performance, we also compute industry-adjusted indicators of firm’s leverage and

EBITDA, by subtracting the industry median leverage and the industry median EBITDA-to-total

assets from the raw measures. The industry medians are based on 3-digit SIC codes provided that

there are at least five firms in the industry, excluding the sample firm. When there are less than

five companies at the 3-digit SIC level, industry adjustments are made at the 2-digit SIC level. The

industry constituents, as well as their values of leverage and operating performance, are taken from

Compustat.

The ratio of debt due within a year (short-term debt) to total debt is adopted as a measure of a

firm’s immediate liquidity needs.

3.2. Debt Structure

The most relevant determinant of the choice between in-court and out-of-court restructuring is the

firm’s debt structure, in the form of both debt seniority and debt composition.

3.2.1. Debt Seniority

The priority structure of debt claims plays a crucial role in the debt restructuring process. Secured

creditors are unlikely to become significantly impaired in bankruptcy, while they may be worse off

in a distressed exchange when new issues with similar priority are offered to more junior creditors.

As a result, they tend to express a preference for in-court reorganizations. On the contrary, un-

secured and subordinated creditors may be more likely to agree to the debt exchange offer, given

their nature of residual claimants in the bankruptcy process. Asquith, Gertner and Scharfstein

(1994) and Chatterjee, Dhillon and Ramı́rez (1996) report a higher likelihood of bankruptcy when

10

the proportion of secured debt is larger. Asquith, Gertner and Scharfstein (1994) also suggest that

a higher fraction of secured debt may indicate a relatively low proportion of intangible assets in

the company, which translates into less costly bankruptcy proceedings.

Based on the information provided in the financial footnotes on the debt structure, we classify

each debt issue into one of three groups: (i) secured, if any firm’s assets were pledged as collateral

for the debt issue; (ii) senior unsecured; (iii) subordinated (which includes junior subordinated,

subordinated, or senior subordinated). For each firm, we compute the percentages of total debt

represented by secured, senior unsecured and subordinated debt. We also record the total number

of tiers in the debt structure as a proxy for the complexity of the priority structure of the company.

3.2.2. Debt Composition

In the presence of a single lender, complete debt contracts, and symmetric information, a dis-

tressed exchange is always more efficient than a costly bankruptcy. In a more realistic setting,

severe information asymmetries between creditors and firm’s managers may undermine attempts

to renegotiate out of court (Mooradian (1994)). Additionally, when the firm’s debt is held by a

multitude of creditors, successful debt workouts may become more problematic due to both hold-

out problems and conflicts of interest (Gertner and Scharfstein (1991)). In this respect, a strand of

the existing literature (Gilson, John and Lang (1990), Brunner and Krahnen (2008)) claims that a

large proportion of bank debt is beneficial in avoiding bankruptcy given that: (i) banks are better

informed than other creditors; (ii) different banks lending to the same firm normally share similar

interests and coordinate their response in case of financial distress.

This claim, however, is debatable under several aspects. First, virtually all the bank debt of

financially distressed firms is secured (James (1995)), hence banks have no incentive to make

concessions in favor of an out-of-court restructuring (see for example the evidence provided by

Asquith, Gertner and Scharfstein (1994) and Chatterjee, Dhillon and Ramı́rez (1996)). Second, as

first hinted by Franks and Torous (1994), when bank loans are syndicated, coordination problems

may arise among the participants to the syndicate, which include both banks and institutional

11

investors. Third, during a banking crisis the distressed conditions of financial institutions can

adversely affect bank-dependent borrowers (Khwaja and Mian (2008), Chava and Purnanandam

(2011)). More generally, banks tend to contract their credit supply in recession, thus inducing a

substitution effect from bank loans to bonds in borrowers who have access to bond markets (Becker

and Ivashina (2011)). In recent study, Li and Srinivasan (2011) document that, in fact, U.S. banks

do not provide any (price or non-price) benefits to help their borrowers in distress.

For the above reasons, the direction in which debt composition affects the restructuring out-

come remains, to a large extent, an empirical question. We analyze the debt structure by looking

at the composition of total debt between public debt and bank loans, which jointly represent, on

average, 95% of the total debt of the companies in the sample. The 10-K financial notes are gen-

erally not very helpful in identifying public debt issues. For this purpose, we consider both the

debt issues that were originally registered with the SEC, and those issued under Rule 144A and

subsequently registered with the SEC.13 Data on the securities registrations are available from the

SEC’s EDGAR database. We also record, for each company in the sample, the number of public

issues outstanding at the time of default, as a further proxy of potential creditors’ coordination

problems.

The notes to the SEC financial statements reveal the overall amount of bank debt, the proportion

of bank debt which is secured, and how bank loans are divided between revolving lines of credit

and term loans, but do not provide information on lenders. To better investigate the structure of

bank debt, we restrict our attention to the loans covered in detail by the DealScan’s database,

which includes 99.3% of all bank loans in our sample. The following indicators are computed to

proxy for loan complexity and lenders’ coordination problems: 1) proportion of bank debt in the

form of institutional term loans (B-term, C-term or D-term), which are designed to be placed with

nonbank institutional investors, such as investment banks, pension funds, hedge funds, insurance

companies; 2) number of lead arrangers, i.e. number of banks that establish a relationship with

the firm and organize the loan; 3) number of participants, i.e. number of banks and/or institutional

13In line with previous evidence on the registrations of 144A offerings (Fenn (2000) and Huang and Ramı́rez(2010)), we find that 87% of the Rule 144A issues in our sample are later registered with the SEC and, therefore, aremore similar to bonds than to private placements held by a handful of institutional investors.

12

investors that, in a syndicated loan, join the lead arrangers to fund part of the loan.

As a measure of reputation of the lead arrangers we calculate an indicator variable (or an

average indicator when there is more than one lead) that equals 1 if the lead arranger was in the top

5th percentile for amounts arranged as a lead arranger in the year prior to the borrower’s default.

Information on the identity of the lead arrangers and their market share is taken from DealScan.

Since Gopalan, Nanda and Yerramilli (2011) document a significant reputation damage for lead

arrangers following the bankruptcy filing of their borrowers, we may expect reputable arrangers to

favor workouts against bankruptcy. As the authors note, however, this effect may turn out to be

negligible in times of crisis, or for very reputable arrangers.

To assess whether the distressed conditions of bank lenders have had any impact on the restruc-

turing outcome of their borrowers, we compute measures of profitability and profitability shocks,

i.e. the annual return-on-assets (ROA) and the change in annual ROA, of the lead arranger at the

end of the fiscal period prior to the borrower’s default.14 Data on banks’ ROA are taken from

BankScope.

As a final measure of the overall complexity of the debt structure, we include firm’s size,

computed as the natural logarithm of total assets. Large firms are likely to experience more severe

creditors’ coordination problems. However, very large companies can benefit from the too big to

fail effect in times of crisis.

4. Empirical Findings

4.1. Distressed Exchange vs Bankruptcy: Univariate Analysis

Tables 2 and 3 contrast mean and median values of size, leverage, operating performance, debt

structure, and bank debt characteristics for firms that restructured out of court against those that

14Average values are calculated when there are multiple lead arrangers. Other indicators of distress, such as Tier 1capital ratios or net charge-off ratios were considered, but had to be excluded due to the lack of data for a significantnumber of the lead arrangers in our sample.

13

filed for bankruptcy: Panel A reports the findings for the entire sample, while Panel B refers to

reference entities only and Panel C to non-reference entities only.15 The last two columns report the

values of the t-test for difference in means and the non-parametric Wilcoxon test for difference in

medians. In discussing the findings we mainly focus on the medians due to the skewed distribution

of most variables.

INSERT TABLES 2 AND 3 HERE

4.1.1. Degree of Financial and Economic Distress

Looking at Panel A of Table 2, we observe that companies that file for bankruptcy are more fi-

nancially distressed and face more immediate liquidity needs than those that manage to restructure

their debt out of court. The median value of the industry-adjusted leverage ratio for the former

reaches 0.52, against a value of 0.33 for the latter, and the median ratios of short-term debt to total

debt are equal to 0.50 and 0.02, respectively. The high proportion of short-term debt for bankrupt

firms is largely explained by the covenant violations recorded on some companies’ debt before de-

fault: following a covenant violation, the debt of a distressed firm often becomes due immediately,

and it is then reported as short-term debt in the SEC filings.

No significant differences arise between the two groups in terms of operating performance,

which is also in line with the median operating performance of the corresponding industries. This

indicates that, in general, at year-end prior to the default event, the companies were not in a situa-

tion of economic distress relative to other firms in their industry, which confirms the predominantly

financial nature of the 2008-2009 crisis.

4.1.2. Characteristics of the Debt Structure

In line with both our predictions and previous empirical findings, firms that renegotiate their debt

under Chapter 11 during the sample period are characterized by higher proportions of secured15The difference in the size of the subsamples between Tables 2 and 3 is due to a small number of firms that have

zero bank debt, which are included in Table 2 but not in Table 3.

14

debt to total debt, and lower proportions of senior unsecured debt to total debt (median values of

0.59 and 0.15), than companies that complete a debt workout (median values of 0.44 and 0.42,

respectively). More in general, we find that firms with a larger number of debt priority tiers have

better chances to restructure their debt out of court.

With respect to debt composition, we observe that successful debt workouts are more likely

when large proportions of debt are publicly held (in terms of both overall percentage of public

debt to total debt, and number of public issues outstanding), rather than granted by banks. These

findings are consistent with the peculiarities of the crisis, which originated in the financial sector

and subsequently spread to other sectors. In this context, many banks incurred severe losses and

liquidity shortages, and were not in a position to refinance their financially distressed borrowers.

Therefore, firms that relied heavily on bank debt had lower chances to restructure out of court than

companies with more diversified sources of public debt. In line with such explanation, we observe

bankruptcies to be more common among firms whose main bank lenders experienced, on average,

more severe shocks in profitability. Instead, we do not find the loan syndicate composition or the

reputation of the main bank lenders to significantly affect the restructuring outcome.

4.1.3. Presence of Credit Default Swaps

In Panels B and C of Tables 2 and 3 we replicate the comparison among firms that arranged out-of-

court versus in-court restructuring for both CDS reference entities and non-reference entities. The

purpose is to highlight potential differences in the two groups that could help us shed some light

onto any peculiarities that may characterize the restructuring choice of reference entities.

The underlying intuition is the following: if we observed significant differences in the degree

of economic / financial distress or in the debt structure between firms that complete a distressed

exchange and firms that file for Chapter 11 within non-reference entities, but no significant dif-

ferences within reference entities, then additional factors not included in our analysis would be

responsible for the restructuring choice of the latter, and the presence of empty creditors may be

one of those. If the proportion of insured creditors in CDS reference entities was sufficiently high

15

to influence the outcome of the debt renegotiation, we would not expect such outcome to depend

significantly on the company’s fundamentals, as bankruptcy would be primarily the (inefficient)

result of the presence of empty creditors.16 Further, we would expect the debt priority structure to

be a less significant determinant of the restructuring method, as a substantial portion of unsecured

creditors would behave like secured debtholders, finding it convenient to force the company into

Chapter 11.

Our empirical findings reveal that most of the factors that significantly affect the choice be-

tween distressed exchange and bankruptcy are common to both groups of reference entities and

non-reference entities. In both cases, companies that file for bankruptcy are more financially dis-

tressed than those that renegotiate out of court, have more immediate liquidity needs in order to

repay short-term debt, a higher proportion of secured debt, a lower proportion of senior unsecured

debt, and a less diversified debt structure, with a lower number of public debt issues outstanding.

Firms that restructure under Chapter 11 are also smaller and more (less) reliant on bank (public)

debt, although these determinants are statistically significant, in both mean and median, only for

reference entities. The distressed condition of the lead bank lenders seems instead to play a signif-

icant role only for non-reference entities. This can be explained with the fact that the difference in

exposure to bank debt between bankrupt firms and other firms is significantly larger for reference

entities than for non-reference entities. Therefore, we do not expect profitability shocks in banks

lending to less bank-dependent reference entities to significantly impact the debt renegotiation.

All in all, the traditional determinants of the restructuring method remain significant also for CDS

reference entities, and we do not find evidence suggesting that the presence of empty creditors may

be an important omitted variable that would play a major role in this respect.

In order to legitimate, from a methodological perspective, both the comparative study presented

in Panels B and C and the probit analysis that follows, we contrast in Table 4 the mean and median

characteristics of CDS reference entities and non-reference entities at year-end prior to default.

Such comparison is also interesting per se, since no attempts have been made so far in the liter-

16It is worth reminding that when bankruptcy represents the efficient restructuring outcome (e.g. when justified bypoor economic fundamentals, or low bankruptcy costs), the presence of insured creditors becomes irrelevant and noempty creditor issue arises.

16

ature to assess in what respect CDS reference entities differ from other firms with similar credit

quality. As expected, we find reference entities to be significantly larger than other firms in the

sample (median log asset size of 7.97 and 6.37, respectively), as CDS contracts are typically avail-

able on large companies with widely traded debt securities, for the purpose of hedging or trading

positions in those securities. As a direct consequence, syndicated loans arranged for reference en-

tities involve a significantly larger number of lead banks and participants, and are organized by the

main players of the syndicated loan market. Reference entities have larger proportions of senior

unsecured debt compared to non-reference entities, given that it is the standard class of reference

obligations underlying CDS contracts. Moreover, we record a significantly higher number of tiers

in the debt structure, and of public issues outstanding for reference entities compared to other

firms. Interestingly, however, the overall proportions of bank debt and public debt in reference

entities are not significantly different from those in other companies in the sample. Similarly, the

proportions of secured and subordinated debt are similar across the two groups of firms. Contrary

to the general perception, reference entities are not significantly more leveraged than other firms.

However, in our sample, they report a lower (at 10% confidence level) operating performance,

highlighting a more severe condition of economic distress prior to default.

INSERT TABLE 4 HERE

In essence, our analysis does not uncover distinctive features of reference entities that can be

unambiguously associated, a priori, with a particular reorganization method. On the one hand,

some of the characteristics (high number of tiers and public debt issues, large proportion of unse-

cured debt, large size) might have favored out-of-court restructurings during the crisis, due to the

availability of diversified sources of debt, as well as to the benefits of a too big to fail effect. On the

other hand, those features might have been offset, in some cases, by worse fundamental economic

conditions or by severe creditors’ coordination problems.

17

4.2. Distressed Exchange vs Bankruptcy: Probit Analysis

In order to formally test whether CDS reference entities experience a higher probability of filing

for bankruptcy than other firms, we estimate a simple probit model where the dependent variable

(DEF) equals zero if the company restructures out of court and equals one if it renegotiates under

Chapter 11. A dummy variable (CDS) indicates whether the firm is a CDS reference entity. In

a first specification, we regress the dependent variable on the CDS dummy for all default events

i = 1 . . .n:

DEFi =c+α1CDSi (1)

The estimates (Table 5, column 2) confirm that the presence of CDS is not a significant deter-

minant of the debt restructuring outcome in our sample.

In a second specification we add a selection of variables from the univariate analysis to control

for the effect of financial leverage, economic distress, and debt structure on the choice between

the two restructuring methods. Specifically, we control for the size (Size), the industry-adjusted

measures of leverage (Lev IA) and operating performance (EBITDA IA), the proportion of debt

due within a year (Short Debt), the ratio of secured debt to total debt (Sec Debt), the number of

tiers in the debt structure (Tiers) and the change in annual ROA of the firm’s principal bank lenders

(∆Prof ).

We apply the following probit specification to our sample of default events i = 1 . . .n:

DEFi =c+β1Sizei +β2Lev IAi +β3EBIT DA IAi +β4Sec Debti +β5Short Debti+

+β6Tiersi +β7∆Pro fi +β8CDSi (2)

The results from the probit regression are reported in Table 5, column 3. In line with the find-

ings from the univariate analysis, we observe that the probability of restructuring in court is higher

18

for firms with higher book leverage, more short-term financing needs, and a larger fraction of se-

cured debt. We also find bankruptcy to have a significant positive association with the profitability

shocks incurred by the main bank lenders. Again, we do not observe a higher likelihood of filing

for bankruptcy in reference entities.17

In order to assess whether the probit model in (2) applies to both CDS reference entities and

non-reference entities, we re-estimate the model on the two subsamples of firms (Table 5, columns

4-5). The key bankruptcy determinants are the same for both groups, with the exception of the

profitability shocks in the principal bank lenders, for the reasons outlined previously. The hypoth-

esis that the coefficients of the two regressions are not significantly different cannot be rejected

either at a pairwise level (Table 5, column 6), or at a joint level according to a likelihood-ratio

Chow test.18

Based on the model estimates for the subsample of non-reference entities, we then derive pre-

dicted probabilities of distressed exchange and bankruptcy both in-sample for non-reference en-

tities, and out-of-sample for CDS reference entities. If the latter were more prone to bankruptcy

due to the pressure of empty creditors, we would expect to observe larger errors in predicting debt

workouts for those firms. Specifically, in such cases, a company would be forced into bankruptcy

even though the financial and economic fundamentals are sound enough to yield a forecast of

distressed exchange. Our findings (Table 5, column 5) do support this claim. In CDS reference

entities, the percentage of correct predictions of debt workouts (i.e. the ratio between correctly pre-

dicted distressed exchanges and total number of predicted distressed exchanges) is indeed lower

than the percentage of correct predictions of Chapter 11 filings (68% against 75%). However, this

also occurs in sample for non-reference entities, and the ratios of the percentage of correctly pre-

dicted workouts to the percentage of correctly predicted bankruptcies are virtually the same for

17The findings are robust to alternative specifications of the model, including: (i) the usage of a different selectionof independent variables; (ii) the inclusion in the model of a dummy variable controlling for the introduction of theRecovery Act in February 2009, which provides for tax incentives for distressed exchanges. Additional results areavailable from the authors upon request.

18The comparison of probit coefficients across groups is biased if there are significant differences in the degree ofresidual variation. The estimates from a heteroskedastic probit model reveal that, in our sample, the residual varianceof the non-reference entities group is only 20% lower than the residual variance of reference entities, and the two arenot significantly different. The results are available from the authors upon request.

19

both reference entities and other firms.

INSERT TABLE 5 HERE

A number of recent default events have gained considerable attention from the press, to quickly

become anecdotal evidence in favor of the empty creditor theory. Some of those defaults are

included in our sample, specifically the bankruptcies of Lear Corp., LyondellBasell Industries, and

Six Flags Inc., as well as the distressed exchanges of Harrah’s Entertainment Inc, Unsys Corp.,

and YRC Worldwide Inc. From the estimates of (2), we derive predicted probabilities of filing

for Chapter 11 for these companies. We find high probabilities for the three companies that went

bankrupt (79% for Lear Corp., 82% for LyondellBasell Industries, and 64% for Six Flags Inc.)

which suggests that the restructuring outcome was actually in line with the prediction based on

firms’ fundamentals. Lower probabilities are associated with the three firms that restructured out-

of-court (40% for Harrah’s Entertainment Inc, 29% for Unsys Corp., and 37% for YRC Worldwide

Inc.). Again, we do not find evidence in favor of the empty creditor issue.

In the light of the empty creditor debate, there are two potential explanations to our findings: (i)

the proportion of insured bondholders is not sufficiently large to affect the outcome of the restruc-

turing process; (ii) the proportion of insured creditors is large enough to impede the negotiation

of a workout, yet creditors prefer to support it, instead of forcing the company into bankruptcy.

This behavior can be ascribed to several factors, such as reputational issues for the creditors, or

the uncertainty related to the bankruptcy process.19 Unfortunately, the unavailability of data on

CDS volumes owned by bondholders makes it impossible to empirically verify which of those

conjectures applies in practice.

Before concluding, it is worth noticing that throughout our investigation we have focused on

distressed exchanges and bankruptcies, while dismissing the default events due to missed payments

that sometimes precede debt renegotiations. After the expiry of a grace period, missed payments

19In this respect, it is worth reminding that involuntary bankruptcy cases initiated by creditors can be challenged bythe debtor. Moreover, when disputes between creditors and obligors are taken to court, judges can investigate in detailthe economic interests of bondholders by requiring, for example, to disclose their positions in credit derivatives on thefirm’s debt, as in the LyondellBasell case.

20

of interests or principal are considered default events by Moody’s and reported in the Default and

Recovery Database. A missed payment also produces a credit event (“failure to pay”) under CDS

contract specifications and may therefore have an impact on our analysis. Once a failure to pay

occurs, insured bondholders can claim from protection sellers the payment of the face amount of

debt, net of the recovery value, upon delivery of the bonds. If all bondholders with a position in

CDS claim their payoffs, the protection sellers become the new bondholders who will participate

to the renegotiation talks. At this stage, in the absence of insured creditors with an incentive to

file for Chapter 11, the firm becomes essentially comparable to a non-reference entity. One way to

correct our sample to take the issue into account would be to remove, for CDS reference entities,

all debt workouts and bankruptcy filings that were preceded by a missed payment. However, since

the percentage of those bankruptcies is higher than the percentage of the distressed exchanges

(28% against 9%), this would bias the remaining sample of reference entity defaults towards debt

workouts. Alternatively, one could remove all renegotiations preceded by missed payments for

both reference entities and other firms, but the size of the residual sample would be too small.

Hence, we decide not to apply any correction in this respect: this is unlikely to introduce biases

in our study, as the percentages of debt workouts and bankruptcy filings anticipated by missed

payments are similar for both reference entities and other companies.

4.3. Evidence from Recovery Rates

Another avenue for the empty creditor issue to manifest itself is through the recovery rates of

firms in distressed exchange or bankruptcy. Even though the presence of insured bondholders does

not seem to have an obvious impact on the restructuring outcome, it may still translate into higher

recovery rates for reference entities than for other companies. This may occur in both: (i) distressed

exchanges, given the better conditions that tendering debtholders might be able to negotiate; (ii)

bankruptcy, if the pressure of empty creditors forces bankruptcy to occur prematurely, when the

assets’ value is still sufficient to ensure satisfactory recoveries. In order to test whether the presence

of CDS has had an effect on recovery rates, we run two separate regressions to explain recoveries

21

following distressed exchanges and Chapter 11 filings:

RECOVi =c+ γ1Log Issuei + γ2Sen Classi + γ3Sizei + γ4Levi + γ5EBIT DAi ++γ6Tangi

+ γ7Med Ind Qi + γ8Ind Distressi + γ9CDSi + γ10Indi (3)

Our measure of recovery rate (RECOV) is the price of the defaulted security recorded 30 days

after the default event occurs. In addition to the CDS dummy (CDS), we add some control variables

that the existing empirical literature (see, for example, Acharya, Bharath and Srinivasan (2007))

has found to be significant determinants of recovery rates. We expect the seniority of the security

in default (Sen Class) to be positively associated with recovery rates. To account for seniority, we

define a variable that equals 0 for senior secured debt, 1 for senior unsecured and 2 for subordi-

nated. All else equal, the size of the issue in default (Log Issue) can also have a positive impact

on recoveries, given the higher bargaining power of debtholders of large issues. We include the

ratio of EBITDA to total assets (EBITDA) as an indicator of the operating profitability for the firm,

which may have a positive effect on recovery rates, through an increase in the overall firm’s value.

We control for leverage (Lev), measured as the ratio of total debt to total assets, since debt restruc-

turing may be harder to achieve in highly-leveraged firms, and recoveries may be lower. Firm’s

size (Size) can have a negative impact on the amount recovered due to coordination problems that

may arise among creditors. Asset tangibility (Tang), measured as the ratio of net property, plant

and equipment to total assets, is instead expected to increase recovery rates. Finally, we control

for the effect of industry-related variables. We include a dummy (Ind Distress) to identify firms

that belong to a distressed industry (defined as an industry where the median stock return of the

constituents is lower than −30% in the year of default), as their recoveries may be lower. The

median industry Tobin’s Q (Med Ind Q) accounts for the potential effect of growth prospects of

the firm’s assets on the recovery value. The median industry Q is computed as median of the ratio

of market value of the firm (estimated as book value of total assets−book value of total equity +

market value of equity) to the book value of the firm, and is taken over all firms (excluding the

22

one in default) in the same 3-digit SIC code. Industry dummies (Ind), defined according to the

first digit SIC code, are also added to capture any residual industry-related effects. Information

on recovery rates, seniority, size and number of the assets in default is taken from the Moody’s

Default and Recovery Database. Firm-specific variables are obtained from the 10-K filings, while

industry characteristics are taken from CRSP/Compustat.

INSERT TABLE 6 HERE

We estimate the model in (3) separately for distressed exchanges and bankruptcy, using ordi-

nary least squares, with standard errors adjusted for heteroscedasticity and clustered at firm’s level.

The results are reported in Table 6 (columns 2 and 4), and are in line with our previous findings

regarding the lack of evidence of the empty creditor issue: we do not find the presence of CDS to

have a significant impact on recovery rates of either distressed exchanges or bankruptcies. Instead,

we observe that recovery values in debt workouts are positively associated with debt seniority,

firm’s profitability, asset tangibility and median industry Q. Firm’s leverage has a negative impact

on recoveries. Surprisingly, firms that belong to distressed industries experience higher recovery

values in distressed exchanges.20 Once a company files for bankruptcy, the recovery values are

positively related to the seniority of the debt in default, and negatively related to the size of the

company. As for distressed exchanges, firms with high leverage ratios tend to recover less.

As a final check, we assess the effects of the presence of CDS on the recovery rates of different

seniority classes of debt. Since most CDS are written on senior unsecured debt, we might find a

significant effect on the recovery rates of this class only, which may disappear when all seniority

classes are pooled together. For this purpose, we run similar regressions as in (3), but we include

in the model three interaction dummies that take the value of 1 if the issue in default has a given

seniority (secured, senior unsecured, or subordinated) and the issuer is a CDS reference entity,

and 0 otherwise. The results are reported in Table 6, columns 3 and 5. The interaction dummy

variables are all insignificant in explaining recoveries in both distressed exchanges and Chapter 11

20We explain this result by observing that, in our sample, distressed industries are also the industries with a highfraction of tangible assets, hence this could represent a residual effect not entirely accounted for by the asset tangibility.

23

filings. To conclude, our investigation of recovery rates has not provided any evidence in favor of

the empty creditor issue. However, it has shed some light on the determinants of recovery rates in

the recent crisis.

5. Further Evidence and Robustness Checks

5.1. Characteristics of Distressed Exchanges

The presence of hostile creditors capable of affecting the debt restructuring outcome and / or the

recovery rates is not supported by the results of our empirical analysis. For completeness, we

search for any evidence of their existence by looking at the characteristics of the out-of-court debt

restructuring in the sample. If empty creditors existed, debt workouts would have some distinctive

characteristics, the completion of the debt restructuring process might take longer, and a significant

number of amendments and / or extensions might be necessary for the exchange offer to be finally

accepted. In essence, the presence of insured creditors can be viewed as the equivalent of a severe

holdout problem.

For this purpose, we collect extensive information on the distinctive features of the individual

distressed exchanges in the sample from Factiva. The relevant characteristics of the workouts are

reported in Table 7, distinguishing between restructurings completed by CDS reference entities

and by other firms. First, we classify debt workouts according to the particular type of exchange

offered. We find that cash tender offers, where the distressed firm offers to purchase a portion of

its debt, represent 41% of all offers for reference entities, against 32% for other companies. This

might be consistent with the presence of empty creditors, in line with the intuition that offers to

repurchase the debt in cash provide a good solution to avoid bankruptcy when holdout problems

are severe.21 However, debt-for-equity offers, which normally exacerbate holdout issues, constitute

23% of all offers for reference entities, against 7% for the remaining firms. Debt exchange offers,

21Chatterjee, Dhillon and Ramı́rez (1995) also find that tender offers are more common in the presence of holdoutissues.

24

where existing debt is exchanged for new debt, alone or combined with equity and / or cash,

represent the vast majority of workouts for non-reference entities (61%), and only 36% of all

offers for reference entities. In both groups of companies the new securities are typically more

senior, and have longer maturity than the tendered ones.

INSERT TABLE 7 HERE

Second, we look at the seniority of the debt tendered in the offer. In this respect, if an empty

creditor problem existed, we would expect to see reference entities targeting mainly junior claims,

who stand to benefit more from an out-of-court restructuring, instead of secured and senior unse-

cured (but covered by CDS insurance) debt. However, our evidence does not support this predic-

tion, as we find that the proportions of junior and senior debt targeted by reference entities and

other firms are fairly comparable (the percentages for junior / convertible debt are equal to 42%

and 46%, respectively).

Further, previous research (Chatterjee, Dhillon and Ramı́rez (1995)) has discussed how co-

ercive techniques such as exit consent solicitations can be usefully employed to solve holdout

problems.22 In our sample, however, exit consent solicitations characterize 39% of the distressed

exchanges completed by non-reference entities, but only 14% of the workouts of reference entities.

Also, we do not find significant differences either in the time between the exchange offer is made

and its completion for the two groups of firms (average number of days equal to 39 for reference

entities and 44 for other companies), or in the average number of extensions / amendments to the

original offer (0.97 and 1.15, respectively).

5.2. The CDS Basis

Bondholders with CDS protection typically build their positions as negative basis trades, whereby

the CDS premium they are required to pay is lower than the bond’s yield to maturity (net of

22Exit consent solicitations involve the consent of tendering bondholders to eliminate restrictive covenants from thedebt. The completion of the workout is conditional on its approval by a given majority of bondholders.

25

financing costs). Negative basis trades represent very popular arbitrage strategies that investors

normally unwind at a profit when the basis converges to zero or turns positive. If the reference

entity defaults those trades deliver a significant profit, hence recent research (Yavorsky, Bauer,

Gates and Marshella (2009)) has conjectured that bondholders with a negative basis trade would

prefer a formal restructuring to an out-of-court renegotiation.

In this respect, if insured bondholders actually played a role in the outcome of the renegotiation

process, we would expect to observe a negative CDS basis prior to default for reference entities that

restructured under Chapter 11. Intuitively, the more negative the CDS basis, the higher the number

of investors for whom it is convenient to enter negative basis trades and, in case of financial distress

of the reference entity, to favor bankruptcy.

We compute the CDS basis as the difference between the 5-year CDS spread and the 5-year

asset swap spread, normalized by dividing for the CDS spread. Data on the spreads are provided

by Bloomberg only for the most liquid CDS, hence we limit our analysis to the subsample of

reference entities with widely traded CDS. Figure 1 shows the dynamics of the average CDS basis

over the six months prior to default, for the 13 companies that filed for bankruptcy and the 13

companies that completed a debt workout. The basis is, in fact, more negative for the former,

which may suggest the presence of a larger number of insured creditors for those firms. However,

the sign of the basis turns positive in the three months prior to default and, in the last two months,

the average basis becomes significantly larger than the corresponding one for firms that restructure

out of court.23 Therefore, most of the investors with a negative basis trade could have already

unwound their positions as soon as the basis converged, well ahead of the default event. If this is

the case, the proportion of insured creditors at default may not be sufficiently large to affect the

restructuring choice.

INSERT FIGURE 1 HERE23Mengle (2009) also documents the prohibitive cost of credit insurance close to default.

26

5.3. Robustness Checks

A crucial component in the above analysis is the identification of the CDS reference entities. As

explained, we regard as reference entities all firms for which MarkIt provided a valid CDS spread

at the time of default, regardless of whether the CDS had been frequently or infrequently traded

in the period prior to default. One may argue that a different approach may be more appropriate

in selecting reference entities that may be affected by empty creditors. On the one hand, we

may expect the number of insured creditors to be significant only in firms with frequently traded

CDS, as the CDS liquidity enables investors to easily create and unwind insured positions. On the

other hand, most of the volumes in widely traded CDS typically reflect purely speculative “naked”

positions, whereas positions in less liquid CDS might be mainly associated with “buy-and-hold”

hedges for bond portfolios.

To address these concerns, we replicate the probit analysis on the determinants of the restruc-

turing outcome using two different identification criteria for reference entities. In the first case,

we label as reference entities only firms with liquid CDS, and we pool together non-reference en-

tities and firms with illiquid CDS. In the second case, we remove from the sample the firms with

widely traded CDS and we identify as reference entities those with less liquid CDS, which are then

compared to non-reference entities. The results from the probit estimates are reported in Table 8.

INSERT TABLE 8 HERE

Again, we do not find evidence that reference entities, however defined, have a higher prob-

ability of filing for bankruptcy. The main determinants of the restructuring choice remain firm’s

leverage, the proportion of short-term debt and of secured debt, and the profitability shocks in the

principal bank lenders.

27

6. Concluding Remarks

Recent contributions from both legal and financial literature have described the availability of credit

insurance via CDS as detrimental to efficient debt renegotiations in conditions of financial distress,

with direct policy implications in terms of regulation of the credit derivative market. In this paper

we look into this claim by investigating, from an empirical perspective, whether the presence of

CDS has in fact played a significant role in determining the debt renegotiation outcome during the

2008-2009 crisis. Our findings, from both univariate and probit analyses, indicate that reference

entities do not seem more likely to file for Chapter 11 than other firms, and that the determinants

of the restructuring method are the same for both groups of firms. We conjecture two reasons

that could support these results. First, the proportion of insured bondholders is not sufficiently

large to affect the outcome of the restructuring process. Second, this proportion is large enough to

affect debt renegotiations, however the insured bondholders prefer not to force the company into

bankruptcy. Since no information is available on CDS volumes owned by bondholders for hedging

purposes, we are unable to draw firm conclusions. In this respect, regulators should introduce

measures aimed at increasing transparency on who holds CDS protection.

Our research also uncovers some interesting facts regarding the role played by the traditional

determinants of the debt restructuring choice during the 2008-2009 period. In line with what was

documented for previous economic downturns, we observe that high leverage and short-term debt

ratios and a large proportion of secured debt increase the probability of filing for Chapter 11. In

contrast with previous findings, bankruptcy is more likely for companies with a less diversified

and dispersed debt structure, mainly concentrated on bank debt. This suggests that, during the

crisis, the traditional creditors’ coordination problems had less severe effects than banks’ liquidity

shortages on the chances of a successful debt workout.

28

References

Acharya, V., Bharath, S. and Srinivasan, A.: 2007, Does industry-wide distress affect defaulted

firms? evidence from creditor recoveries, Journal of Financial Economics 85, 787–821.

Altman, E. I. and Karlin, B.: 2009, The re-emergence of distressed exchanges in corporate restruc-

utrings, The Journal of Credit Risk 5, 43–56.

Ashcraft, A. B. and Santos, J. A.: 2009, Has the CDS market lowered the cost of corporate debt?,

Journal of Monetary Economics 56, 514–523.

Asquith, P., Gertner, R. and Scharfstein, D.: 1994, Anatomy of financial distress: An examination

of junk-bond issuers, Quarterly Journal of Economics 109, 625–658.

Becker, B. and Ivashina, V.: 2011, Cyclicality of credit supply: Firm level evidence. Working

Paper, Harvard University.

Bolton, P. and Oehmke, M.: 2011, Credit default swaps and the empty creditor problem, Review of

Financial Studies, Forthcoming .

Brunner, A. and Krahnen, J. P.: 2008, Multiple lenders and corporate distress: Evidence on debt

restructuring, Review of Economic Studies 75, 415–442.

Chatterjee, S., Dhillon, U. S. and Ramı́rez, G. G.: 1995, Coercive tender and exchange offers in

distressed high-yield debt restructurings: An empirical analysis, Journal of Financial Eco-

nomics 38, 333–360.

Chatterjee, S., Dhillon, U. S. and Ramı́rez, G. G.: 1996, Resolution of financial distress: Debt re-

structurings via Chapter 11, prepackaged bankrupcties and workouts, Financial Management

25, 5–18.

Chava, S. and Purnanandam, A.: 2011, The effect of banking crisis on bank-dependent borrowers,

Journal of Financial Economics 89, 116–135.

Economist: 2009, CDSs and bankruptcy: No empty threat.

Fenn, G. W.: 2000, Speed of issuance and adequacy of disclosure in the 144A high-yield debt

market, Journal of Financial Economics 56, 383–405.

29

Franks, J. R. and Torous, W. N.: 1994, A comparison of financial recontracting in distressed

exchanges and Chapter 11 reorganizations, Journal of Financial Economics 35, 349–370.

Gertner, R. and Scharfstein, D.: 1991, A theory of workouts and the effects of reorganization law,

Journal of Finance 46, 1189–1222.

Gilson, S. C., John, K. and Lang, L. H.: 1990, Troubled debt restructurings: An empirical study of

private reorganization of firms in default, Journal of Financial Economics 27, 315–353.

Gopalan, R., Nanda, V. and Yerramilli, V.: 2011, Does poor performance damage the reputation

of financial intermediaries? evidence from the loan syndication market, Journal of Finance,

Forthcoming .

Hirtle, B.: 2009, Credit derivatives and bank credit supply, Journal of Financial Intermediation

18, 125–150.

Hotchkiss, E. S., John, K., Mooradian, R. M. and Thorburn, K. S.: 2008, Bankruptcy and the reso-

lution of financial distress, in B. E. Eckbo (ed.), Handbook of Empirical Corporate Finance,

Vol. 2, Elsevier B.V.

Hu, H. T. and Black, B.: 2008a, Debt, equity, and hybrid decoupling: Governance and systemic

risk implications, European Financial Management 14, 663–709.

Hu, H. T. and Black, B.: 2008b, Equity and debt decoupling and empty voting II: Importance and

extensions, University of Pennsylvania Law Review 156, 625–739.

Huang, R. and Ramı́rez, G. G.: 2010, Speed of issuance, lender specialization, and the rise of the

144A debt market, Financial Management 39, 643–673.

James, C.: 1995, When do banks take equity in debt restructurings and the composition of ex-

change offers in financial distress, Review of Financial Studies 8, 1209–1234.

Jostarndt, P. and Sautner, Z.: 2010, Out-of-court restructuring versus formal bankruptcy in a non-

interventionist bankruptcy setting, Review of Finance 14, 623–668.

Khwaja, A. I. and Mian, A.: 2008, Tracing the impact of bank liquidity shocks: Evidence from an

emerging market, American Economic Review 98, 1413–1442.

30

Li, Y. and Srinivasan, A.: 2011, Relationship bank behavior during borrower distress and

bankruptcy. Working Paper.

Mengle, D.: 2009, The empty creditor hypothesis, ISDA Research Notes, International Swaps and

Derivatives Association.

Moody’s: 2000, Moody’s approach to evaluating distressed exchanges, Special Comment,

Moody’s Investors Service.

Mooradian, R. M.: 1994, The effect of bankruptcy protection on investment: Chapter 11 as a

screening device, Journal of Finance 49, 1403–1430.

Mooradian, R. M. and Ryan, H. E. J.: 2005, Out-of-court restructurings and the resolution of

financial distress: Section 3(a)(9) compared to investment-bank-managed exchange offers,

Journal of Business 78, 1593–1624.

Risk: 2009, The CDS curse.

Stulz, R. M.: 2010, Credit default swaps and the credit crisis, Journal of Economic Perspectives

24, 73–92.

Sufi, A.: 2007, Information asymmetry and financing arrangements: Evidence from syndicated

loans, Journal of Finance 62, 629–668.

Yavorsky, A., Bauer, G., Gates, D. and Marshella, T.: 2009, Analyzing the potential impact of

credit default swaps in workout situations, Special Comment, Moody’s.

31

32

Table 1 – Summary Statistics on Default Events Table 1 reports summary statistics on the default events recorded over the period Jan 08 – Dec 09 for non-financial U.S. rated companies with 10-K filings available at year-end prior to default. Data on the default events and their classification into distressed exchanges and Chapter 11 filings are obtained from the Moody’s Default and Recovery Database. Information on which defaulted companies are CDS reference entities is provided by MarkIt.

Number of DefaultsOverall

Defaults (N=121)

Distressed Exchanges

(N = 50)

Chapter 11 Filings (N = 71)

Overall Defaults

Distressed Exchanges

Chapter 11 Filings

Year2008 30 8 22 25% 27% 73%2009 91 42 49 75% 46% 54%

IndustryMining and construction 10 5 5 8% 50% 50%Manufacturing 64 22 42 53% 34% 66%Transportation, Communication and Utilities 19 12 7 16% 63% 37%Trade: Retail and Wholesale 11 5 6 9% 45% 55%Services 17 6 11 14% 35% 65%

Company typePublic 67 26 41 55% 39% 61%Private with public debt 54 24 30 45% 44% 56%

Presence of CDSNon-reference entities 66 28 38 55% 42% 58%Reference entities: 55 22 33 45% 40% 60%- liquid CDS 27 13 14 22% 48% 52%- less liquid CDS 28 9 19 23% 32% 68%

Percentages

33

Table 2 – Debt Workout vs Chapter 11: Univariate Analysis Table 2 presents average and median values of firm characteristics for companies that completed a distressed exchange and companies that filed for Chapter 11. The last two columns report the values of the t-test for difference in means and the Wilcoxon test for difference in medians. Panel A reports findings for all companies in the sample, while Panel B only includes CDS reference entities and Panel C only includes non-reference entities. Size is the log of total assets. Leverage is computed as the ratio of total debt to total assets, and Operating income as the ratio of EBITDA to total assets. Industry-adjusted measures are computed by subtracting the industry median (based on 3-digit SIC codes) leverage and EBITDA-to-total assets from the raw measures. Short-term debt is debt due within one year. Public debt includes both debt issues originally registered with the SEC and Rule 144A issues later registered with the SEC. Data on firm characteristics are obtained from the 10-K filings at year-end before default. Information on which defaulted companies are CDS reference entities is provided by MarkIt. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively.

(continued on next page)

Mean Median Mean Median t -test Wilcoxon

Size 7.36 6.98 6.88 6.74 1.91* 1.69*Leverage 0.72 0.67 1.02 0.83 -3.26*** -3.06***Leverage - industry adjusted 0.42 0.33 0.72 0.52 -3.31*** -3.23***Operating income 0.08 0.09 0.09 0.09 -0.55 -0.27Operating income - industry adjusted 0.01 -0.01 0.01 0.01 0.01 -0.34Short term debt / Total debt 0.09 0.02 0.51 0.50 -6.93*** -4.93***Bank debt / Total debt 0.36 0.38 0.48 0.51 -2.39** -2.33**Public debt / Total debt 0.60 0.59 0.46 0.45 2.76*** 2.48**Secured debt / Total debt 0.44 0.44 0.61 0.59 -3.33*** -3.24***Senior unsecured debt / Total debt 0.42 0.42 0.23 0.15 3.20*** 3.18***Subordinated debt / Total debt 0.14 0.07 0.16 0.00 -0.38 0.26Number of tiers 2.88 3.00 2.48 2.00 2.34** 2.34**Number public issues outstanding 3.70 2.50 2.44 2.00 1.85* 2.36**# Observations 50 50 71 71

Mean Median Mean Median t -test Wilcoxon

Size 8.36 8.24 7.58 7.59 2.12** 2.03**Leverage 0.76 0.69 1.08 0.81 -1.90* -1.67*Leverage - industry adjusted 0.45 0.34 0.76 0.51 -1.93* -2.00**Operating income 0.05 0.07 0.08 0.09 -0.75 -0.88Operating income - industry adjusted -0.03 -0.03 0.00 0.01 -0.85 -1.39Short term debt / Total debt 0.10 0.02 0.43 0.21 -4.03*** -2.65***Bank debt / Total debt 0.32 0.30 0.46 0.51 -2.11** -2.17**Public debt / Total debt 0.63 0.62 0.50 0.45 2.05** 2.18**Secured debt / Total debt 0.42 0.42 0.55 0.58 -1.83* -2.10**Senior unsecured debt / Total debt 0.49 0.50 0.31 0.28 2.15** 2.39**Subordinated debt / Total debt 0.09 0.04 0.14 0.05 -1.23 -0.51Number of tiers 3.23 3.00 2.88 3.00 1.30 1.47Number public issues outstanding 6.23 5.00 3.88 3.00 1.85* 2.69***# Observations 22 22 33 33

Panel A: Overall Sample

Panel B: CDS Reference Entities

Distressed Exchanges Chapter 11 Filings

Distressed Exchanges Chapter 11 Filings

34

Table 2 (continued) – Debt Workout vs Chapter 11: Univariate Analysis

Mean Median Mean Median t -test Wilcoxon

Size 6.57 6.49 6.27 6.27 1.39 1.19Leverage 0.69 0.65 0.97 0.86 -3.01*** -2.58***Leverage - industry adjusted 0.39 0.32 0.69 0.57 -3.21*** -2.84***Operating income 0.09 0.11 0.09 0.09 0.15 0.23Operating income - industry adjusted 0.03 0.00 0.01 0.00 0.70 0.77Short term debt / Total debt 0.09 0.01 0.57 0.93 -5.71*** -4.36***Bank debt / Total debt 0.39 0.40 0.49 0.45 -1.68* -1.36Public debt / Total debt 0.58 0.57 0.43 0.44 1.95* 1.59Secured debt / Total debt 0.45 0.46 0.66 0.65 -2.83*** -2.64***Senior unsecured debt / Total debt 0.37 0.31 0.16 0.00 2.51** 2.50**Subordinated debt / Total debt 0.18 0.11 0.17 0.00 0.19 0.51Number of tiers 2.61 3.00 2.13 2.00 2.45** 2.38**Number public issues outstanding 1.71 1.00 1.18 1.00 2.02** 1.83*# Observations 28 28 38 38

Distressed Exchanges Chapter 11 Filings

Panel C: Non-Reference Entities

35

Table 3 – Debt Workout vs Chapter 11: Bank Debt Characteristics Table 3 presents average and median values of bank debt characteristics for companies that completed a distressed exchange and companies that filed for Chapter 11. The statistics only refer to firms with non-zero bank debt. Number of lead arrangers and participants refer to the origination of the existing syndicated loans of the company. Market quote of the lead arranger is an indicator variable that equals 1 if the lead arranger was in the top 5th percentile for amounts arranged as lead arranger during the year prior to default, and 0 otherwise. Profitability and ∆ Profitability of lead arranger refer, respectively, to annual ROA and change in annual ROA of the lead arranger computed at quarter-end before default. In case of multiple lead arrangers, we consider average values for market quote, profitability, and ∆ in profitability. Information on loans and bank profitability is obtained from Dealscan and Bankscope. Information on which defaulted companies are CDS reference entities is provided by MarkIt. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively.

Mean Median Mean Median t -test Wilcoxon

Bank debt / Total debt 0.38 0.38 0.50 0.51 -2.28** -2.22**Term Loans B / Bank debt 0.49 0.74 0.50 0.58 -0.08 0.38Secured bank debt / Bank debt 0.89 1.00 0.98 1.00 -2.06** -2.82***Number lead arrangers 1.75 2.00 1.69 2.00 0.38 0.66Number participants 7.04 5.00 5.78 4.00 0.75 0.92Market quote lead arranger 0.59 0.50 0.56 0.50 0.43 0.31Profitability lead arranger 0.12 0.33 -0.16 0.28 1.40 0.79∆ Profitability lead arranger -0.25 -0.42 -0.69 -0.50 2.16** 1.50# Observations 47 47 68 68

Mean Median Mean Median t -test Wilcoxon

Bank debt / Total debt 0.36 0.30 0.47 0.51 -1.87* -2.05**Term Loans B / Bank debt 0.56 0.74 0.51 0.72 0.39 0.78Secured bank debt / Bank debt 0.93 1.00 0.96 1.00 -0.52 -1.21Number lead arrangers 2.05 2.00 2.00 2.00 0.22 0.33Number participants 10.75 5.50 7.09 5.50 1.01 0.50Market quote lead arranger 0.77 1.00 0.64 0.73 1.37 1.03Profitability lead arranger -0.06 0.30 -0.07 0.31 -0.10 -0.10∆ Profitability lead arranger -0.35 -0.60 -0.56 -0.46 0.56 0.20# Observations 20 20 32 32

Mean Median Mean Median t -test Wilcoxon

Bank debt / Total debt 0.40 0.40 0.51 0.45 -1.70* -1.30Term Loans B / Bank debt 0.45 0.00 0.49 0.54 -0.39 -0.25Secured bank debt / Bank debt 0.86 1.00 1.00 1.00 -2.22** -2.74***Number lead arrangers 1.52 1.00 1.42 1.00 0.68 0.84Number participants 4.30 4.00 4.61 3.00 -0.33 1.09Market quote lead arranger 0.46 0.50 0.49 0.50 -0.29 -0.34Profitability lead arranger 0.27 0.52 -0.25 0.20 1.81* 1.33∆ Profitability lead arranger -0.14 -0.34 -0.81 -0.52 2.25** 1.83*# Observations 27 27 36 36

Panel C: Non-Reference Entities

Distressed Exchanges Chapter 11 Filings

Panel A: Overall Sample

Distressed Exchanges Chapter 11 Filings

Panel B: CDS Reference Entities

Distressed Exchanges Chapter 11 Filings

36

Table 4 – CDS Reference Entities vs Non-Reference Entities: Univariate Analysis Table 4 presents average and median values of firm characteristics for CDS reference entities and non-reference entities included in our sample of defaulted firms. The last two columns report the values of the t-test for difference in means and the Wilcoxon test for difference in medians. For a description of the variables, see Tables 2 and 3. Bank debt characteristics (bottom panel) are computed only for firms with non-zero bank debt. Data on firm characteristics are obtained from the 10-K filings at year-end before default. Data on bank loans and bank characteristics are obtained from Dealscan and Bankscope. The identification of CDS reference entities is based on information provided by MarkIt. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively.

Mean Median Mean Median t -test Wilcoxon

Size 7.89 7.97 6.39 6.37 7.05*** 6.09***Leverage 0.95 0.77 0.85 0.75 0.94 0.51Leverage - industry adjusted 0.64 0.45 0.56 0.43 0.69 0.00Operating income 0.07 0.08 0.09 0.10 -1.17 -1.30Operating income - industry adjusted -0.01 -0.01 0.02 0.00 -1.46 -1.72*Short term debt / Total debt 0.30 0.05 0.37 0.08 -0.90 -1.11Bank debt / Total debt 0.41 0.40 0.45 0.41 -0.82 -0.45Public debt / Total debt 0.55 0.54 0.49 0.52 1.17 0.76Secured debt / Total debt 0.50 0.54 0.57 0.59 -1.48 -1.44Senior unsecured debt / Total debt 0.38 0.37 0.25 0.05 2.3** 3.14***Subordinated debt / Total debt 0.12 0.05 0.18 0.00 -1.46 -0.15Number of tiers 3.02 3.00 2.33 2.00 4.31*** 4.11***Number public issues outstanding 4.82 3.00 1.41 1.00 5.26*** 6.15***

# Observations 55 55 66 66

Bank debt / Total debt 0.43 0.41 0.47 0.40 -0.75 -0.33Term Loans B / Bank debt 0.53 0.73 0.47 0.53 0.72 0.42Secured bank debt / Bank debt 0.95 1.00 0.94 1.00 0.34 0.18Number lead arrangers 2.02 2.00 1.46 1.00 4.08*** 3.84***Number participants 8.50 5.50 4.48 4.00 2.56** 2.56**Market quote lead arranger 0.69 0.90 0.47 0.50 2.96*** 2.78***Profitability lead arranger -0.07 0.31 -0.03 0.28 -0.18 -1.03∆ Profitability lead arranger -0.50 -0.54 -0.53 -0.43 0.13 -0.24

# Observations 52 52 63 63

CDS Reference entities Non-reference entities

37

Table 5 – Debt Workout vs Chapter 11: Probit Analysis Table 5 presents estimates of probit models, where the dependent variable equals 1 if the company files for Chapter 11 and equals 0 if the company restructures out of court. For a description of the variables, see Tables 2 and 3. CDS Reference Entity dummy equals 1 if the firm is a reference entity, according to MarkIt. Standard errors are reported in brackets. The last column reports values of χ2 test of individual coefficient equality between regressions (a) and (b), with p-values in brackets. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively.

Independent variablesEntire sample

Entire sample

CDS reference

entities (a)

Non-reference

entitites (b)

Coefficient equality in (a) and (b)

Size -0.015 -0.068 0.268 0.98(0.16) (0.20) (0.27) (0.32)

Leverage - industry adjusted 1.163** 0.899* 2.011*** 1.25(0.45) (0.53) (0.68) (0.26)

Operating income - industry adjusted -1.336 -1.677 -1.194 0.02(1.42) (2.91) (1.52) (0.88)

Secured debt / Total debt 1.267** 1.335* 1.262* 0(0.52) (0.72) (0.74) (0.94)

Short term debt / Total debt 1.855*** 1.739** 2.057*** 0.11(0.48) (0.75) (0.78) (0.73)

Number of tiers -0.298 -0.275 -0.196 0.05(0.18) (0.29) (0.23) (0.82)

∆ Profitability lead arranger -0.312** -0.173 -0.514** 1.36(0.12) (0.13) (0.20) (0.24)

CDS reference entity dummy 0.062 0.353(0.23) (0.39)

Constant 0.191 -0.896 0.003 -3.501* 1.79(0.15) (1.10) (1.69) (1.92) (0.22)

# Observations 115 115 52 63Pseudo R2 4.00E-04 0.337 0.241 0.448

Chow test of coefficient equality in (a) and (b) - χ2(8) 4.470p-value 0.812

% correct predicted probabilities of Chapter 11 filings for non-reference entities (in-sample) 86%

% correct predicted probabilities of distressed exchanges for non-reference entities (in-sample) 79%

% correct predicted probabilities of Chapter 11 filings for reference entities (out-of-sample) 75%

% correct predicted probabilities of distressed exchanges for reference entities (out-of-sample) 68%

38

Table 6 – Determinants of Recovery Prices Table 6 presents OLS estimates of determinants of recovery prices for both distressed exchanges and Chapter 11 filings. Seniority class equals 0 if the issue in default is secured, 1 if it is senior unsecured, 2 if it is subordinated. Asset tangibility is the ratio of property, plant and equipment to total assets. Median industry Q is the median of the ratio of market value of the firm to book value, for all firms in the 3-digit SIC code of the defaulted firm. Industry distress dummy equals 1 if the median stock return of all firms in the 3-digit SIC code of the defaulted firm is less than -30% and 0 otherwise. The interaction dummies in regressions (b) equal 1 if the issue in default has a given seniority and the issuer is a CDS reference entity, and 0 otherwise. For a description of the other variables, see Table 2. All regressions include industry dummies. Standard errors (in brackets) are corrected for heteroscedasticity and clustered at firm’s level. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively.

Independent variables

Distressed Exchanges

(a)

Distressed Exchanges

(b)

Chapter 11 Filings

(a)

Chapter 11 Filings

(b)

Log (Issue size) -0.005 -0.010 0.009 -0.001(0.02) (0.02) (0.02) (0.02)

Seniority class -0.075* -0.091* -0.264*** -0.232***(0.04) (0.05) (0.03) (0.04)

Size -0.019 -0.016 -0.062*** -0.049**(0.02) (0.02) (0.02) (0.02)

Leverage -0.295** -0.289** -0.062* -0.065*(0.12) (0.11) (0.04) (0.04)

Operating income 0.544* 0.544* 0.166 0.199(0.28) (0.31) (0.20) (0.21)

Asset tangibility 0.324** 0.284* 0.001 0.013(0.14) (0.14) (0.11) (0.10)

Median industry Q 0.958*** 0.938*** -0.101 -0.094(0.16) (0.14) (0.12) (0.11)

Industry distress dummy 0.160** 0.162** -0.053 -0.057(0.07) (0.07) (0.05) (0.05)

CDS reference entity dummy 0.001 -0.062(0.07) (0.07)

CDS dummy * Secured dummy 0.111 0.112(0.15) (0.10)

CDS dummy * Sen unsecured dummy -0.032 -0.042(0.08) (0.07)

CDS dummy * Subordinated dummy 0.051 0.089(0.11) (0.09)

Constant -0.391 -0.339 0.844*** 0.847***(0.29) (0.31) (0.25) (0.24)

Industry class dummies Yes Yes Yes Yes

# Observations 160 160 202 202R2 0.422 0.440 0.443 0.457

39

Table 7 – Characteristics of Distressed Exchanges Table 7 presents the characteristics of the distressed exchanges completed by CDS reference entities and non-reference entities in the sample. Information on the characteristics of the distressed exchanges is obtained from news search on Factiva.

CDS Reference

Entities

Non-Reference

Entities

CDS Reference

Entities

Non-Reference

Entities

Type of exchange offerCash tender offers 9 9 41% 32%Debt-to-equity swaps 5 2 23% 7%Debt exchanges 8 17 36% 61%- pure debt exchanges 4 6 50% 35%- with cash 3 6 38% 35%- with equity 1 5 13% 29%Overall sample 22 28

Notes tendered- Secured / senior unsecured 18 15 58% 54%- Junior 6 8 19% 28%- Convertible 7 5 23% 18%

Distressed Exchanges Percentages

40

Table 8 – Robustness Checks: Probit Analysis Table 8 presents robustness checks for the probit models of Table 5. The dependent variable equals 1 if the company files for Chapter 11 and equals 0 if the company restructures out of court. For the models in columns 2-3, the CDS Reference Entity dummy equals 1 if the firm is a reference entity with liquid CDS before default, and 0 otherwise. For the models in columns 4-5, the CDS Reference Entity dummy equals 1 if the firm is a reference entity with illiquid CDS before default, and 0 if it is not a reference entity (reference entities with liquid CDS are excluded). For a description of the variables, see Tables 2 and 3. Standard errors are reported in brackets. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively.

Independent variables

Size 0.031 0.180(0.17) (0.23)

Leverage - industry adjusted 1.289*** 1.591***(0.45) (0.55)

Operating income - industry adjusted -1.741 -1.900(1.40) (1.49)

Secured debt / Total debt 1.269** 1.009**(0.53) (0.50)

Short term debt / Total debt 1.881*** 1.753***(0.49) (0.50)

Number of tiers -0.272 -0.280(0.19) (0.22)

∆ Profitability lead arranger -0.309** -0.376***(0.12) (0.14)

CDS reference entity dummy -0.344 0.114 0.374 0.139(0.27) (0.45) (0.30) (0.45)

Constant 0.298** -1.220 0.191 -2.270(0.13) (1.16) (0.15) (1.65)

# Observations 115 115 90 90Pseudo R2 0.010 0.331 0.013 0.363

Reference entities with liquid CDS only

Reference entities with illiquid CDS only

41

Figure 1 – CDS Basis Figure 1 compares the average CDS basis for the reference entities that completed a distressed exchange and those that filed for bankruptcy, over the six months prior to the default event. The CDS basis is calculated as the difference between the 5-year CDS spread and the 5-year asset swap spread, normalized by the CDS spread. CDS spreads and asset swap spreads are from Bloomberg. The CDS basis is computed for reference entities with very liquid CDS.

-15%

-10%

-5%

0%

5%

10%

15%

20%

25%

30%

-6 -5 -4 -3 -2 -1

CDS

basi

s (%

of s

prea

d)

Months to default event

Average CDS Basis

Chapter 11

Distressed Exchange