Vous êtes sur la page 1sur 46

The determinants of bank loan recovery rates

Hinh Khieu*
University of Southern Indiana
hdkhieu@usi.edu

Donald Mullineaux
University of Kentucky
mullineau@uky.edu

Ha-Chin Yi
Texas State University
hy11@txstate.edu

March 2010

Comments welcome

*Correspondence author: Hinh Khieu, University of Southern Indiana, College of Business, 8600 University
Blvd., OC 3016, Evansville, IN 47725. Email: hdkhieu@usi.edu.
I would like to thank Paul Childs, John Garen, Mark Liu, Kristine Hankins, John Butler, and 2009 FMA
participants for their valuable comments. The authors are responsible for any errors.
The determinants of bank loan recovery rates

Abstract

While there is a very large literature on the determinants of default on various debt instruments, relatively
little is known about the factors which influence recoveries on bank loans in the default state. The issue
has taken on heightened importance since Basel II permits banks to determine required capital holdings
by using model-based estimates of loss given default which depends on the recovery rate. We measure
recoveries using the Ultimate Recovery Database supplied by Moodys and model the recovery rate as
a function of variables reflecting loan and borrower characteristics, industry and macroeconomic
conditions, and several recovery process variables. We find that loan characteristics, such as the presence
of certain types of collateral, are significant determinants of recovery rates, whereas many of the borrower
characteristics before default generally are not. Industry and macroeconomic conditions also are relevant,
as are certain process factors such as prepackaged bankruptcies. Since trading prices on loans
approximately 30 days after default are often used by practitioners (and in some academic studies) as
proxies for the recovery rate, we examine whether this proxy provides a rational estimate of actual
recoveries. We find that the process that drives the 30-day trading price after default differs significantly
from the actual settlement recovery process.

2
The determinants of bank loan recovery rates

1. Introduction

Syndicated loan market has been growing at an exponential rate in worldwide corporate

credit market, reaching to a loan volume of $1.7 trillion in 2006, and academic research has

followed such growth. While loan origination in the credit risk framework is relatively well

understood by research community, the other end of life cycle in syndicated loan market, or loan

mortality is still largely an unexplored area especially, lack of empirical studies due to

incomplete data. Using the Moodys Ultimate Recovery Database, this study empirically

analyzes recovery rates of large syndicated bank loans that have been originated in U.S. in the

long spectrum during 1987-2007. Recovery rates play a critical role in credit risk modeling, and

models that predict recovery rates well are becoming increasingly important because banks are

permitted, and in some cases required, under the 2004 Basel II Accord to use an internal rating-

based approach to estimate loss-given-default. Banks also need to estimate recovery rates by

using the appropriate model to determine the regulatory capital. Recovery rates can be calculated

as one minus loss-given-default. The recovery rate is also an important input in valuing a credit

default swap (CDS). CDS volume has grown very rapidly in recent years. The International

Swaps and Derivatives Association, Inc. (ISDA) released survey results in 2001 showing that the

notional outstanding amount of credit default swaps globally was $631.5 billion. By mid-2008,

ISDA reported CDS volume of $54.6 trillion, resulting in a growth rate of 8,546%. A better

understanding of what drives loan recoveries is therefore a non-trivial issue.

The majority of the existing studies on loan recoveries are published in the practitioner

literature, which is dominated by ratings agencies such as Standard and Poors and Moodys.

Studies in the academic literature are primarily focused on bond recovery rates. There are only a

3
few empirical academic studies of loan recoveries. Two of them consist of case studies using

data extracted from a single lender, while the other mixes both bonds and loans together in a

single sample. While case studies are valuable, it is difficult to know if the results can be

generalized.

In addition, when bonds and loans are included in a single sample, the implicit

assumption is that the same factors influence recovery rates in both markets in identical ways.

The factors that affect loan recoveries may differ from those that influence bond recoveries,

however. For instance, bank loans are more likely to be secured than bonds and secured lenders

recover more, on average, than unsecured creditors. Bank loans also are typically senior to bonds

in a companys capital structure. According to the absolute priority rule, which establishes

priority of claims in liquidation, bank lenders should generally recover more from defaulted debt

than bondholders. Also, theory and evidence argue that bankers in private loan contracting play a

role as delegated monitors and have a comparative advantage over holders of arms length debt

in monitoring borrowers financial and operational behavior (Diamond, 1984). When resolving

defaults, bankers tend to be more involved in the restructuring process by becoming major

stockholders or appointing new directors than bondholders (Gilson, 1990). This should result in

lower agency costs and higher expected recoveries for loans. Bank loans are more likely to be

restructured privately than publicly (Gilson, John, and Lang, 1990) and private resolution tends

to result in higher recoveries (Franks and Torous, 1994). Finally, an increase in bank loan share

in the firms capital structure at the time of filing of bankruptcy is positively related to the

overall recovery rate (Carey and Gordy, 2008). These differences between loans and bonds

appear to warrant an investigation focused solely on loan recovery rates.

4
Research on recoveries employs different measures of recovery rates. One commonly

used measure is the trading price of the debt instrument approximately 30 days after default (see,

for example, Eberhart and Sweeney, 1992; Carty, 1998). In studies by practitioners and

academics alike, this measure serves as a proxy for the ultimate payoff creditors obtain when the

defaulting entity emerges from bankruptcy or is liquidated. The underlying idea is that loan or

bond traders rationally estimate the amount that will eventually be recovered in the future and

then discount that value back to the trading date. However, literatures are not conclusive on the

effectiveness of trading price as an unbiased estimator for ultimate recovery. The alternative to

trading prices is to use more definite measure of ultimate recovery, which we intend to do in this

paper.

The current study has two purposes. First, it investigates the factors that drive actual bank

loan recovery rates (rather than proxies thereof) using Moodys data on U.S. loan defaults over

the period 1987-2007. We focus only on bank loans and our data capture a large number of

different lending institutions. To our knowledge, ours is one of the first academic papers that

uses Moodys Ultimate Recovery Database (URD) for this type of analysis.1 We find that firm

leverage before default negatively affects ultimate recoveries, while firms operating cash flows

are not correlated with the actual settlement value. Firm size influences recoveries differently

across different types of loans. Recoveries from borrowers with a history of defaults are

significantly higher than those of first-time defaulters. A variety of loan contract features are

strongly related to the ultimate payoff to creditors. Prepackaged bankruptcy reorganizations

increase actual settlements. Loan recoveries are nonlinearly related to the length of time to

1
Recently, we become aware of the study by Zhang (2009) that used the Moodys URD data and found the similar
results as ours.

5
emergence. The probability of default measured at the time of loan origination is unrelated to

ultimate recoveries.

Second, we examine how well the market for bank loans anticipates settlement values in

recovery. Since the trading price on loans roughly 30 days after recovery is a commonly used

proxy for ultimate recovery, we test whether this price represents a rational forecast of actual

recovery. A rational forecast is not only unbiased by definition, but also efficient in the sense

that the factors that influence the forecast are identical to those that determine the outcome in

question. Our findings suggest that the 30-day post-default trading price on loans is not a rational

estimate of actual recovery realizations.

The paper proceeds as follows. Section 2 reviews the related literature. Section 3

discusses our model. Section 4 describes our data and presents our econometric specification.

Sections 5, 6, and 7 contain the empirical results. Section 8 concludes.

2. Related Literature

Empirical studies of recovery rates primarily focus on bonds. There are relatively few

papers on bank loans since it is difficult to obtain loan recovery data, given that loans are private

debt contracts.2 On a descriptive level, several empirical studies on defaulted loans show that the

recovery rates exhibit a bimodal distribution (Araten et al. (2004); Asarnow and Edwards (1995);

Schuermann (2004)). That is, many defaults result in full recovery, while the weight of others

has zero or very low recovery rates. Dermine and Neto de Carvalho (2006) find a similar

distribution for loans in a sample drawn from one bank. Other studies do not confirm bimodality,

but instead show that loan recovery rates are skewed to the right, while bond recoveries are

2
Frye (2000) attempts to explain the apparent doubling of loan loss severity over the period 1970-1999, but notes
that the data required for loan defaults are not available and consequently fits his model to bond data instead.

6
skewed to the left (Emery, 2007). The inconsistent results in these studies may be due to

differences in datasets or time periods or both. Other loan mortality studies include works by

Altman and Suggit (2000), and Emery and Cantor (2005), but they mainly report the results on

default rates rather than loan recovery.

Examining Moodys database of ultimate recoveries,3 Emery (2007) finds the recovery

rate on bank loans averages 80% at resolution, compared to only 65% for bonds, and similarly

Emery ad Ou (2004) shows that loss severity in the event of default has been about 2.25 times

greater for bonds than loans, holding ratings constant. Asarnow and Edwards (1995) present a

univariate analysis of bank loan default data on 831 commercial and industrial loans (C&I) and

89 structured loans4 made by Citibank over 24 years and find an average recovery of 65% for

C&I loans and 87% for structured loans. The higher recovery rate on structured loans reflects the

fact that such loans are heavily collateralized and contain many restrictive covenants. Acharya,

Bharath, and Srinivasan (2007) report a recovery rate of 81.12% for bank loans, 59% for senior

secured bonds, 56% for senior unsecured bonds, 34% for senior subordinated bonds, 27% for

subordinated bonds, and 18% for junior subordinated bonds for the period from 1982 to 1999.

Researchers have recently created mathematical models that deal directly with recovery

rates. Guo, Jarrow, and Zeng (2009) extend the reduced-form approach to explicitly model the

recovery rate process in terms of the firms assets and liabilities. While risky debt is priced prior

3
Emery (2007) defines ultimate recoveries as the recovery values that creditors actually receive at the resolution
to default. In a literature review by Altman (2006), ultimate recovery rates refer to the nominal or discounted
value of bonds or loans based on either the price of the security at the end of the reorganization period (usually
Chapter 11) or the value of the package of cash or securities upon emergence from restructuring.
4
Asarnow and Edwards (1995, p. 13) elaborate on the structured loans they study as follows:
The structured loan group has the following characteristics:
The loans are very closely monitored the bank directly controls the companys cash receipts and
disbursements.
The loans are highly structured and contain many restrictive covenants.
The loans are highly collateralized, and lending is done on a formula basis, for example, having a pre-
determined advance rate against customer receivables as collateral.

7
to default in existing models, Guo et al. (2009) present a regime-switching model and a jump

diffusion process to quantify defaulted debt prices and realized recovery rates. Their model of

the recovery process differs from other reduced-form models in that it generates a pricing

formula for zero-coupon bond prices both before and after default, as opposed to other models

that price such debt instruments at or before default.

Bakshi, Madan, and Zhang (2006) provide a multifactor model of recovery rates. The

contributions of their model are twofold: (1) credit spreads can be separated into default and

recovery components and (2) recoveries as a fraction of the discounted par value fit a sample of

BBB-rated bonds better than recoveries of face value. The former are found to reduce out-of-

sample pricing errors.

Karoui (2007) also models recovery rates and derives pricing formulas for risky debt

instruments. His model is more tractable than comparable models because recovery rates follow

discrete-time, rather than continuous, adapted processes. He estimates the model using BBB and

B Standard & Poors yield indices and shows that the implied loss given default is approximately

27% for the BBB index and 74% for the B spreads.

Most of the earlier academic studies on credit risk assumed that the probability of default

(PD) and recovery rates (RR) are uncorrelated. 5 There are reasons to doubt this assumption,

however, since studies on credit rating transitions have shown that recovery and default are both

related to conditions external to the firm. For example, Carty (1997) examines 77 years of credit

rating changes from 1920 to 1996 and documents that movements in credit quality are correlated

with macroeconomic, industrial, geographic, and temporal factors. Nickell, Perraudin, and

Varotto (2000) estimate an ordered probit model for credit ratings on long-term bonds between

5
See, for example, Jarrow, Lando, and Turnbull (1997).

8
1970 and 1997 and show that the probability of rating changes depends on the stage of the

business cycle, industry effects, and other factors.

In a recent literature review, Altman (2006) notes that collateral values, which affect

bond and loan recovery rates, decline as economic conditions deteriorate, while at the same time

the number of defaults increases. Altman, Brady, Resti, and Sironi (2005) find a negative relation

between aggregate default rates and recovery rates on bonds over the period 1982-2002. They

show that previous studies, which ignored this correlation, understated both expected and

unexpected losses. Hu and Perraudin (2002) support the findings of Altman et al. (2005).

In a model for implied recoveries6, Das and Hanouna (2007) propose a method based on

different functional relationships among default intensities, default swap spreads, and recovery

rates to derive an implied forward curve of recovery rates, instead of the single rate derived in

other models. Using credit derivative swap spread data from 2000 to 2002, they find that the

interest rate term structure and equity market volatility explain most of the variation in recovery

rates over time and that the recoveries over the period 2000-2002 exhibit a strong negative

correlation with default probabilities.

Jokivuolle and Peura (2000) approach the correlation question indirectly via an option-

pricing-based model for bank loans in which collateral value varies stochastically and is

correlated with PD. They assume that the firms asset value does not determine the RR, but does

have a positive relation with collateral value, and they obtain an inverse relationship between

realized default rates and recovery rates. In short, both recent theoretical models and empirical

analyses point to a likely negative relation between default and recovery rates.

6
Implied recoveries refer to recoveries that are not directly observable, but can be found embedded in a market
transaction such as a credit default swap.

9
The current paper studies factors that affect defaulted bank loan recovery rates. The paper

differs from the existing literature on recovery rates in that (1) it uses discounted settlement rates,

a more direct measure of recovery in Moodys Ultimate Recovery Database instead of trading

prices as proxies for recovery, (2) the data examined include only defaulted bank loans, not

bonds nor a combination of bonds and loans, and (3) the loans were made by a variety of

financial institutions to numerous industrial companies rather than a specific loan data set from a

single bank.

3. The Model

We classify factors that should affect loan recoveries into four groups: 1) loan

characteristics, 2) recovery process characteristics, 3) borrower characteristics, and 4) external

factors, such as macroeconomic conditions and industry characteristics. We also examine

whether the probability of default is related to recovery. We will discuss the prospective roles of

factors within each group, including references to existing empirical evidence, where relevant.

3.1 Loan characteristics

3.1.1 Loan size

Loan size could influence recovery rates. Dermine and Neto de Carvalho (2006) find, for

example, that loan size has a negative association with actual recovery rates accumulated 12 to

48 months after default. The authors argue that banks are likely to delay foreclosure on large

loans because other clients of the bank who conduct business with the large borrower would be

adversely affected by foreclosure and perhaps default on their own loans. This spill-over effect

results in lower recovery rates when these large loans eventually do enter the default state. In

contrast, Acharya et al. (2007) find a positive relationship between recovery rates and issue size

and contend that larger debts should have higher recoveries because larger borrowers have

10
greater bargaining power in bankruptcy. Thorburn (2000) studies Swedish small business

bankruptcies and finds that loan size is not a significant determinant of recovery.

We include loan size as a variable in our model. We are agnostic with respect to the sign,

however. While the relationship aspect of lending suggests a prospective negative sign, the fact

that large creditors could influence their bargaining power in resolving defaults, especially in

bankruptcy, argues for a positive coefficient between loan size and recovery rates.

3.1.2 Loan types

Loan types may affect how much creditors recover when borrowers default. Our

Moodys data, which contain loans to commercial and industrial companies only, permit two

classifications: term loans and revolving credits. A term loan is normally used to finance long

term projects such as the purchase of fixed assets or building a manufacturing plant. A revolving

credit or a line of credit is typically short term in nature, smaller in size, and more often granted

without collateral than a term loan. Whether differences in loan type affect recovery rates

differently is an empirical question and no prior studies examine this question in the recovery

rate context.7 One of the main distinguishing characteristics between term loans and revolvers is

maturity.8 While term loans are more likely to be secured, the longer maturity could enhance the

prospect of deterioration in collateral value, thereby reducing recovery. The short duration of

revolvers also triggers frequent requests for renewals, which allows banks to re-evaluate default,

increasing recovery prospects by limiting loan size or asking for more collateral. However,

outstanding lines of credit may bind banks to supply the borrowing firm with funds up to pre-

arranged amount on demand over the life of the contract (Strahan, 1999). This future obligation

7
There is evidence (Gottesman and Roberts, 2004; Harjoto, Mullineaux, and Yi, 2006) that term loans are priced
differently from revolvers, other things equal.
8
By definition, term loans have a maturity greater than one year. Revolvers can have any maturity, but many have a
maturity of less than a year. Loan commitments of 364 days or less do not need to be backed by bank capital.

11
can be an additional burden to banks in addition to credit risk that the borrowing firm may

default.

We include a binary variable for the loan type in our model, but do not hypothesize a

sign. We also include an interaction term between the size of the borrower and the type of loans.

If large firms tend to use one type of loan rather than the other and large firms are more likely to

fetch higher recoveries than small ones, then the effect on recovery from loan types may also be

shifted according to firm size. We will discuss the potential impact of firm size on loan recovery

in the section about borrower characteristics below.

3.1.3 Collateral

Finance theory suggests that recovery rates should be higher on secured than unsecured

loans since creditors gain a legal right to seize and sell certain designated assets in the event of

default. Collateral has been found to be significantly related to recoveries in both the practitioner

and academic literature (see Araten, Jacobs, and Varshney, 2004; Altman and Kishore, 1996;

VanDeCastle, 1999). For example, VanDeCastle (1999) reports that recovery rates for senior

secured debt is about 25 percent higher than for unsecured senior debt.

Bank loans tend to be more highly collateralized than bonds at origination.9 Banks, unlike

bondholders, also have the option to seek more or better collateral on existing loans when a

borrowers financial conditions deteriorate.10 In contrast, bondholders cannot readily revise the

terms of a loan contract outside of bankruptcy, since the Trust Indenture Act requires that the

majority of the creditors need to assent to any contractual revisions.

9
From Moodys ultimate recovery database, we find that 938 loans out of the total 1,484 loans (63%) have all assets
pledged as collateral versus 134 out of 2,402 (or 5.6%) for bonds. Also, the number of unsecured loans in the
database is 122 out of 1,484 (or 8.2%), while the proportion of unsecured bonds is 1,825 out of 2,402 (or 76%). The
same holds true for all other types of collateral except for pledges by PP&E and other equipment.
10
A necessary condition for a bank to ask for increased collateral is that the borrower violates one or more
covenants, which results in technical default.

12
We include various measures of collateral as dummy variables in our model and expect a

positive relation between the presence of collateral and recovery rates on loans. Since we can

observe different types of collateral, we can also determine whether some types of security yield

higher recovery rates than others.

3.2 Recovery process characteristics

3.2.1 Prepackaged bankruptcy, traditional bankruptcy, and out-of-court restructurings

A default can usually be resolved in two ways: restructuring (or reorganization) or

liquidation. Restructuring or reorganization can be public or private. A private reorganization is

often referred to as a private workout or out-of-court settlement. A public reorganization

involves a Chapter 11 bankruptcy filing. Liquidation can be carried out through a Chapter 7

filing. The process debtors use to resolve financial distress with various classes of claimholders

is likely to affect recovery rates. In the mid-1980s, a hybrid form of financial reorganization

emerged in which a debtor concurrently files a bankruptcy petition and reorganization plan,

known as a prepackaged bankruptcy or prepack.11 The plan needs a confirming vote by two-

thirds of voting creditors in amount and more than 50 percent in number from each class of

creditors, but the vote can be held before or after the plan is filed.12 The pre-petition negotiation

and agreement result in lower legal costs, shorter time in bankruptcy, and fewer bargaining

frictions between shareholders and creditors. This should translate into higher recovery rates on

prepackaged bankruptcies.

11
This prepack feature was in fact one of the most important innovations under the Bankruptcy Reform Act of 1978
because of its time and cost efficiency with more flexible voting requirement (Altman and Hotchkiss, 2006).
12
A completely private workout before a bankruptcy filing would be successful only if 100 percent of the creditors
agreed.

13
Also, firms that adopt such a procedure appear to be more operationally sound than those

that do not (Ryan, 2008). Firms that use distressed exchanges13, which are informal and private

workouts, are normally more solvent at the time of reorganization than bankrupt firms (Franks

and Torous, 1994). As a result, creditors may not have to accept large haircuts (discounts)

from such firms in the recovery process (Ryan, 2008). Franks and Torous (1994) find that the

average recovery rate is 80.1% in private workouts, compared with 50.9% in formal Chapter 11

procedures. McConnell, Lease, and Tashjian (1996) document that the average recovery rate for

debtors that reorganize through prepacks from January 1980 to June 1993 (72.9%) lies between

those rates in the traditional (non-prepackaged) Chapter 11 proceedings and private workouts.

Izvorski (1997) shows that average recoveries on 281 defaulted U.S. corporate bonds for the

period 1983-1993 are 30.36% for debtors entering Chapter 11 and 36.23% for firms going

through informal reorganizations.

To account for the effects of different forms of restructuring or reorganization on

recovery, we include a dummy variable for our sample firms that went through Chapter 11 or

Chapter 7 bankruptcy procedures with a prepack, a dummy variable for those that restructured

outside of court, and a dummy variable for other types of restructuring.14 The formal non-

prepackaged bankruptcy dummy variable serves as the reference group. We expect the included

dummies to enter our model positively.

13
According to Keenan (2000), a distressed exchange occurs (1) when the distressed issuer makes a tender offer to
pay cash for all or part of an outstanding debt, which is usually at a price higher than the trading price but lower than
the face amount; or (2) the issuer makes an exchange offer to replace the outstanding debt securities with a new
package of securities such as cash, new bonds, stocks, or some combination.
14
In Moodys ultimate recovery database containing loans alone, there are 14 observations for firms that defaulted
and cured their default by paying back their obligations and 12 observations for other types of restructuring. We
have no hypothesis for this type of default, but still include them in our sample for completeness in our binary
variable constructions.

14
3.2.2 Time to emergence

The time it takes a defaulting borrower to emerge from bankruptcy or restructuring

should also affect recovery rates. The shorter this time horizon, the lower will be the costs the

firm incurs in its legal battle for a new life. Also, a longer time to emergence may indicate that

different stakeholders are having difficulty identifying the fair value of the defaulting firms

assets relative to its liabilities. Carty (1998) argues that uncertainty about the eventual settlement

date makes bankrupt debt less attractive to investors, who demand a premium on such

investments. He also points out that both secured and unsecured claimholders may not be paid

interest during the workout period, so neither would prefer a long, drawn-out path to resolution.

These factors should imply lower recoveries as the time it takes to emerge from financial distress

increases.

Other studies have considered this factor in their models. Franks and Torous (1994)

examine bond and loan recovery rates for firms using Chapter 11 and those engaged in distressed

exchanges over the period 1983 to 1988. They find that firms that restructured through distressed

exchange offers spent less time in reorganization than those that did so through a formal Chapter

11 filing and therefore creditors recovered significantly more from the former group. We

therefore hypothesize a negative relationship between the observed time to emergence and

recovery rates.

We also interact the time to resolution with the existence of a prepackaged bankruptcy.

Since prepacks can reduce the time to emerge and both variables are included in the model, we

are interested in the marginal effect of prepacks on recovery rates, holding the time to emergence

constant.

15
Finally, Covitz, Han, and Wilson (2006) find that U.S. bond recovery rates from 1983 to

2002 are nonlinearly related to the time in default. In particular, they document that bond

recovery rates increase with time in default for the first one and a half years post-default, but

then start to decrease. To account for this prospect, in our estimations, we include a quadratic

term for our TIMETOEMERGE variable. Since the current study involves only bank loans, we

are aware that the effect of this quadratic term may be different from that in bond studies. Bank

holders are not as dispersed as bondholders, so negotiations with defaulting firms may proceed

more quickly in the case of loans.

3.3 Borrower characteristics

Borrower characteristics are commonly included as explanatory variables in research

involving bond recoveries (Acharya et al., 2007; Covitz and Han, 2004). However, existing loan

recovery studies include only a limited number of firm characteristics.15 The question at hand is:

can accounting information about borrowers near the time of default help predict recovery rates

to lenders?

3.3.1 Firm size

The impact of firm size on recoveries is ambiguous ex ante. On the one hand, large firm

size may proxy for high bankruptcy costs, which may in turn result in lower recovery rates. On

the other hand, larger firms presumably present less severe information asymmetry problems to

investors. If this is the case, the restructuring process is likely to occur more quickly for large

firms than for small ones, which may in turn improve recoveries for lenders. Gilson, John, and

15
In their bank loan recovery analysis, Grunert and Weber (2009) include firm size measured as the natural
logarithm of total assets. Dermine and Neto de Carvalho (2006) have only firm age as a firm characteristic.
Thorburn (2000) studies debt recovery rates from Swedish defaults without specifying whether the focus is on bonds
or loans and includes size and profit margins as the only two firm characteristics.

16
Lang (1990) points out that creditors are more willing to accept a restructuring plan from

stockholders who can provide more transparent information about the value of the firm.

We include firm size as a variable in our model. We have no hypothesized sign for the

variable. While information asymmetry problems suggest a prospective positive sign, the fact

that large debtors incur high direct bankruptcy costs in resolving defaults argues for a negative

coefficient. We measure firm size as the natural logarithm of the market value of firm assets one

year before default, where the market value of firm is computed as the book value of debt plus

the market value of equity.

3.3.2 Cash flows, asset tangibility, and leverage

Like Acharya et al. (2007), we include cash flows, asset tangibility, and firm leverage

one year before default as borrower-related explanatory variables in the recovery model. Higher

cash flows, which we measure as the EBITDA one year before default scaled by total book-value

assets, should positively affect recoveries. Acharya et al. (2007) argue that more cash-rich firms

can attract higher prices for their assets from potential buyers, holding other variables constant.

In addition, more operationally sound firms that enter the default state are more likely to

negotiate outside of court systems to resolve distress than firms with weak cash flows before

default (Ryan, 2008; Frank and Torous, 1994), which lowers costs and enhances recovery.

Higher asset tangibility should potentially increase recovery value for creditors.

Williamson (1988) argues that debt holders recovery depends on the degree to which the

borrowers assets are not designated to specific uses or are re-deployable. We measure asset re-

deployability or tangibility as the sum of net property, plant, and equipment as a ratio of total

assets one year before default. Conditional on the impact of any assets pledged as collateral on

17
recovery, this variable should enter our model with a positive coefficient, reflecting the marginal

effect of tangible assets on which lenders can exercise a claim.

Two strands of literature express opposite views on the potential relation between

recovery and leverage. An argument for a positive link between leverage and recovery suggests

that higher leverage prompts increased monitoring of managers behavior, which in turn results

in higher asset quality. Therefore, higher leverage should vary positively with recovery rates

when default occurs, since high-quality assets are liquidated.

However, some prior studies find a significant negative relationship between leverage

and recovery (Acharya et al., 2007; Varma and Cantor, 2004).16 Acharya et al. (2007) argue that

higher-leverage firms tend to have greater dispersion of debt ownership. This complicates

restructuring negotiations to resolve the default and results in lower recovery rates. Carey and

Gordy (2008) study firm-level recovery of firms defaulting over the period 1987-2006 and find

that recovery rates are reduced by approximately five percentage points when the median

leverage ratio increases by one unit.

We measure leverage as the sum of total long-term debt and debt in current liabilities

scaled by the borrowers book-value assets one year before default. We hypothesize a negative

sign for the impact of leverage on loan recoveries.

3.3.3 Prior defaults

Loans to borrowers with a history of prior defaults may produce higher recovery rates,

ceteris paribus. Such borrowers may have had to provide lenders with more collateral or be

subject to greater scrutiny in the loan review process. This factor is not considered in the existing

academic literature. We hypothesize that loans from borrowers with prior defaults will have

higher recovery rates. We include a binary variable for this factor and expect a positive sign.
16
Their data include both bonds and loans, while our sample deals with bank loans only.

18
3.4 Macroeconomic conditions, industry characteristics, and probability of default

3.4.1 Macroeconomic conditions

Numerous studies show that recoveries are lower during recessions than during economic

expansions. Business cycles, not surprisingly, are also linked to the probability of default

(Gersbach and Lipponer, 2003). Frye (2000), fitting his model to Moodys data on U.S.

corporate bonds, estimates that recoveries during downturns decline 20% to 25% relative to

normal-year recoveries. Caselli, Gatti, and Querci (2008) use a large sample of 11,649 loans to

households and small and medium enterprises (SME) from the five largest Italian banks between

1990 and 2004 and find that LGD depends on various macroeconomic variables.

We include a measure of GDP growth as a proxy for the state of the economy and expect

a positive sign for this variables estimate. Although there is a debate on whether GDP growth is

a good proxy for overall economic conditions, we include this measure in our baseline model and

examine alternative measures in our robustness checks.

3.4.2 Industry characteristics

Two types of industry characteristics may be relevant to loan recoveries: industry distress

and cross-sectional variation across industries themselves.

3.4.2.1 Industry distress

Whether asset values of defaulting firms change with industry conditions has received

both theoretical and empirical analyses. Shleifer and Vishny (1992) present a market-equilibrium

model focused on asset liquidation. They argue that industry distress matters because firm assets

will sell at low values in default, given that the most likely buyers are firms in the same industry

who are also in distress. Acharya et al. (2007) empirically find that industry distress is negatively

related to recoveries from defaulted debt instruments. We measure the presence of industry

19
distress with an indicator variable that takes a value of one when a given firm is in an industry

whose median stock return in the year of default is -30% or less and zero otherwise. We expect a

negatively signed coefficient for this variable.

3.4.2.2 Cross-section variation across industries

Whether there are industry effects on recoveries, as there are in the case of default

prospects, is a matter of debate in the empirical literature. Academic studies (for example,

Altman and Kishore, 1996; Dermine and Neto de Carvalho, 2006) find there is a cross-sectional

variation in recoveries across different industries, while the impact of industry on recovery is not

confirmed in a number of practitioner studies (see Araten et al., 2004; Gupton, Gates, and Carty,

2000).We include binary variables for all industries using the classifications employed by

Moodys. Where there are too few observations in any industry for meaningful regression

analyses, we aggregate and label them Other Industry.

3.4.3 Probability of default

The probability of default should be an important factor in determining the recovery rates

for creditors. Numerous empirical studies have established a direct and indirect link between the

probability of default (PD) and recovery rates (RR). Altman et al. (2005) find a negative

association between PD and RR using defaulted bond data over the period from 1982-2002. Hu

and Perraudin (2002) document a similar negative relationship from Moodys default data

covering 22 years from 1971 to 2000.

We include a proxy for default prospects in our model and expect a negatively signed

coefficient. The proxy is the all-in-spread on the original and drawn loans that later default. The

spread is measured as the amount the borrower pays in basis points over the London Inter-Bank

Offered Rates (LIBOR) for each dollar drawn down, including any annual fees, to the lender.

20
The underlying idea is that the spread captures the lenders best estimate of the likelihood of

default at the time when the loans were originated.

4. Sample data and econometric specification

4.1 Sample data

We use bank loan recovery data from Moodys Ultimate Recovery Database, which

contains loan defaults from 1987 to early 2007, as well as a number of loan characteristics. The

defaulted loans included in the database are loans to commercial and industrial companies. The

borrowers are North American firms with total debt value at the time of default greater than $50

million. Moodys data sources include SEC filings, news, and bankruptcy documents. We

manually match these firms with Standard & Poors Compustat firms by borrower names to

obtain firm characteristics. The U.S. GDP data are from the U.S. Department of Commerce

Bureau of Economic Analysis. The stock return data are from the Center for Research in Security

Prices (CRSP). The all-in-spread for drawn funds is from Loan Pricing Corporation (LPC)

DealScan database. Our final sample contains 1,364 observations for settlement value recoveries.

We use Moodys discounted settlement value as an ultimate measure of recovery as the

dependent variable. Moodys defines its ultimate recovery as:

The nominal settlement recovery amount discounted back from each settlement
instruments trading date to the last date cash was paid of the individual defaulted
instruments, using the defaulted instruments effective interest rate.17

where the nominal settlement recovery is the value of the settlement instruments received at or

close to emergence and expressed as a percentage of the principal amount at default.

17
Moodys Ultimate Recovery Databases technical specifications, p. 4.

21
4.2 Econometric specification

We have identified several groups of factors that may exert different impacts on cross-

sectional variation in recovery rates. We next estimate the relation between those factors and the

recovery rates on defaulted loans. We test the following specification:

RR = 0 + 1 * Loan Characteristics
+ 2 * Recovery Process Characteristics
(1)
+ 3 * Firm Characteristics
+ 4 * Macroeconomic Conditions, Industry Characteristics, and PD +

We apply two estimation methods to our models. Following Acharya et al. (2007), we

apply the OLS method with Huber-Whites (1980) heteroskedasticity-consistent standard errors,

assuming a linear relationship between recovery rates and the explanatory variables. We also use

the OLS approach, but assume a log-linear relationship by using the natural logarithm of the

recovery rates as the dependent variable, as shown in equation (2) below.

Ln(RR) = 0 + 1 * Loan Characteristics


+ 2 * Recovery Process Characteristics
(2)
+ 3 * Firm Characteristics
+ 4 * Macroeconomic Conditions, Industry Characteristics, and PD +

Since our recovery data are mostly bounded between 0 and 1, with a large number of

recovery rates equal to one, applying the OLS method to linear equation (1) may suffer from the

same drawback as using the linear probability model for binary data. That is, the impact of any

explanatory variable may not be constant throughout its range of values when 0 RR 1 and the

predicted value is not guaranteed to be bounded between 0 and 1. Following Papke and

Wooldridge (1996), we apply the quasi-maximum likelihood estimation (QMLE) method to

estimate the following non-linear model:

22
0 + 1 * Loan Characteristics

+ 2 * Recovery Process Characteristics
E(RR | X ) = G (3)
+ * Firm Characteristics
3
+ * Macroeconomic Conditions, Industry Characteristics, and PD
4

e X
where G () is the logistic function of the form: G ( X ) = . The QMLE of is consistent
1 + e X

and N -asymtotically normal irrespective of the distribution of the RR conditional on.18

5. Empirical results

5.1 Univariate analysis results

Descriptive statistics for the entire sample are in Table 1. 19 The average (median)

recovery rate is 84.14% (100%) for the settlement value. The sample is split almost evenly

between revolvers and term loans (54% versus 46%). As far as pre-default negotiation is

concerned, 19% of the settlement recoveries have some sort of reorganization plan approved by

shareholders before or at the time of filing for bankruptcy. The average length of time a sample

firm stays in default is 14 months with the maximum being almost 10 years. All of the firm

characteristics in our baseline analyses are measured one year before default. The mean (median)

cash flows relative to total assets for the settlement sample firm are 5.29% (7.04%).

[Table 1]

Table 2 shows ultimate recoveries by default year. A low default rate during the second

half of 1990s is consistent with economic expansion and a strong performance in equity market

during the same period. Syndicated loan volume culminated in 2001 but economic downturn in

mid-2001 curtailed loan activities and loan defaults rose. In year 2002 alone, our sample shows

18
Papke and Wooldridge (1996), p. 622.
19
In a practitioner study, Emery (2007) used the Ultimate Recovery Database from Moodys and provided simple
descriptive statistics. We have one more year of defaults than he does, which translates into 39 more observations.
Our results are consequently very similar to his.

23
241 defaults occurred, almost 8 times greater than those in year 1995 when a bull market started

in stock market. Starting in year 2004, loan defaults diminished greatly.

[Table 2]

As shown in Table 3, the majority of the firms in the sample (67%) defaulted in a non-

prepackaged bankruptcy, while a small number went through prepackaged bankruptcy (19%) and

private workouts (12%). The mean recovery rates for loans with a reorganization plan lie

between those for loans resolved in the traditional bankruptcy and those for loans going through

the other forms of default resolution, which is consistent with McConnell, Lease, and Tashjian

(1996). Table 3 also examines the time to resolution in our two samples by types of default

resolution. Debtors with a prepackaged negotiation resolve their default in about four months, on

average, compared with 19 months for non-prepackaged bankrupt cases. This difference is

significant at the conventional level.

[Table 3]

Table 4 presents a summary of recovery rates by industry using Moodys classifications.

The INDUSTRIAL sector has the highest mean ultimate recoveries (100.2%) 20 , which are

significantly different from the two sectors with the lowest recoveries (OTHERINDUSTRY and

TELECOM) at the 1 percent level. However, the number of defaulted loans in INDUSTRIAL is

quite low (8 loans) compared with TELECOM (134 loans). DISTRIBUTION, though having the

most observations in the sample, has significantly lower recoveries than INDUSTRIAL, which

has the least number of observations. Among those with the highest volume of recoveries

(DISTRIBUTION, MANUFACTURING, and CONSUMER), recoveries are not significantly

20
One reason why the recovery rate can exceed 100% of the amount at default is that creditors may receive equity in
addition to cash and securities in exchange for the defaulted loans being restructured. The equity trades ex post at a
higher price, which causes the total restructuring package to exceed the amount owed at default.

24
different from one another. Overall, our univariate analysis appears to show that there is some

significant cross-sectional variation in recoveries across different industries.

[Table 4]

5.2 Multivariate regression results

Table 5 reports the main multivariate regression results. Model 1 shows the OLS

regression result with heteroskedasticity-consistent standard errors. Model 2 is similar to Model

1, except that the dependent variable is the natural logarithm of recovery rates. Model 3 presents

a quasi-maximum likelihood estimate of recovery rates based on the logistic function,

e X
E(RR | X ) = G ( X ) , whereG ( X ) = .
1 + e X

The sign of the LOANSIZSE coefficient is consistent with the findings of Dermine and

Neto de Carvalho (2006), Carty and Lieberman (1996), and Emery (2004). However, the fact

that the coefficient estimate is insignificant corroborates with Asarnow and Edwards (1995),

Thorburn (2000), and Altman and Kishore (1996). Acharya et al. (2007) propose that the

bargaining power of large creditors may increase recoveries for large debts, while Dermine and

Neto de Carvalho (2006) argue that delayed foreclosures on large loans may decrease subsequent

recoveries. Neither argument appears to hold in the case of our models. However, while we

consider ultimate recoveries only on U.S. defaulted loans, Acharya et al. (2007) study trading

prices at emergence for a mixed sample of bonds and loans and Dermine and Neto de Carvalho

(2006) focus on defaulted loans from a single Portuguese bank.

25
On average, a term loan recovers about 4 cents less on a dollar than a revolver, when firm

size is evaluated at its mean value.21 The result suggests that the short-term nature of revolving

lines of credit may subject borrowers to increased monitoring from lenders, which in turn limits

the loss in the event of default, compared with the long term nature of term loans. This is

consistent with the theoretical model by Rajan and Winton (1995) that short-term debt gives

lenders greater flexibility and control to effectively monitor managers with minimum efforts

because loan covenants dictate such a way that loans are needed to be payable on demand.

There appears to be a differential effect of firm size across the two types of loans. In fact,

recoveries increase with firm size in the case of term loans. The finding confirms the results of

Strahan (1999), who documents that large firms tend to use more revolving lines of credit and

fewer term loans than small firms. He notes that large firms normally have access to public debt

markets, use it to finance long-term projects, and therefore go to banks mainly for revolving lines

of credit associated with working capital needs, while small borrowers depend more on long-

term loans from banks because they have limited access to the bond markets.

Consistent with prior literature on both bonds and loans, our results show that loans

supported by a pledge of any type of assets have higher recovery rates than unsecured loans, all

else constant. In addition, we find loans collateralized by inventories or accounts receivable or

both provide about 30% more recovery than non-collateralized credits, all else constant.

Turning to the process variables, the coefficient estimate of PREPACK is significant,

albeit only at the 10 percent level, and positively signed, but those of the other recovery process

variables are insignificant. The result suggests that restructuring with prepackaged negotiations

21 E (RR | x )
= 2 + 3 FirmSize . The marginal impact above was evaluated at the average FIRMSIZE of
LoanType
3.2826. If the average firm size is computed based on only those firms whose LOANTYPE = 1, the magnitude of
the marginal impact is similar. The partial derivative of RR with respect to LOANTYPE, a binary variable, as shown
above, is just a Taylor-series linear approximation over the range between zero and one.

26
yields higher recovery than the non-prepackaged bankruptcies, all else constant. In particular,

recoveries increase, on average, by about 11 cents on a dollar when TIMETOEMERGE is

evaluated at its average of 14 months. The effect of PREPACK is significantly higher than that

of RESTRUCTURE, which is inconsistent with McConnell, Lease, and Tashjians (1996)

finding that higher recoveries come from private workouts.22

Table 5 results show that the longer the time spent in default, the lower is the recovery

rate. Our model also has both a quadratic term and an interaction term that appear to yield

interesting results. The PREPACK x TIMETOEMERGE interaction term and the quadratic term,

TIMETOEMERGESQ, have positive and significant coefficient estimates. The differential

impact of TIMETOEMERGE on recoveries is thus not constant across the two types of default

resolution one with a plan of restructuring and one without it. By reducing the time in

bankruptcy, defaulting firms with an advance plan of reorganization produce higher recoveries

for their creditors.

The quadratic term coefficient is positive and strongly significant, so we confirm the non-

linear relationship between recovery rates and time to resolution that Covitz, Han, and Wilson

(2006) document in their study of bond recovery rates. However, the sign of the quadratic term

estimate in our loan study is opposite to their bond findings. Covitz, Han, and Wilson (2006) find

a concave behavior of recoveries with respect to the time in bankruptcy. In particular, they

document that recoveries increase with the time to resolution for up to one and a half years and

the relationship reverses afterwards. Our finding suggests a convex behavior: ultimate recoveries

for loans decrease with bankruptcy duration out to 33 and a half months when defaulting firms

22
McConnell, Lease, and Tashjian (1996) compute recoveries as the payoff to a particular class of creditors on a
face value basis. This computed measure is almost equivalent to Moodys ultimate recoveries that we use in our
current study except that Moodys further discounts the recoveries back to the last time when the borrowers made
their payments to the loans and that our RRs are on an instrument level.

27
had a prepack and out to 14 months when defaulting firms had no prepacks. Beyond these

periods, recovery rates rise.

Our findings related to borrower characteristics, macroeconomic conditions, and industry

factors are mostly consistent with the literature on bond and loan recoveries. For example, the

negative and significant LEVERAGE coefficient estimate supports the argument of Acharya et

al. (2007) that higher leverage results in greater difficulties among bargaining parties to resolve

default, which in turn lowers recoveries. GDP growth is positively correlated with ultimate

recovery rates, which is in line with Frye (2000) and Sabato and Schmid (2008). Industry distress

negatively influences them, a finding consistent with Acharya et al. (2007).

As expected, there is strong evidence that some prior history of defaults is associated

with higher recovery rates than first-time defaults. In fact, the positive and strongly significant

coefficient estimate on EVERDEFAULTED indicates that the former bring an average of 11

cents more on a dollar than the latter. Lenders presumably negotiate for enhanced protection

from borrowers who have a history of defaults.

Higher cash flows one year before default do not translate into higher recoveries, all else

constant. It could be that more solvent firms tend to restructure privately to speed up the

resolution of their financial distress. Yost (2002) documents that firms with a high ratio of

operating income to total assets are more likely to restructure out-of-court than in traditional

Chapter 11 bankruptcy. So, the inclusion of a dummy variable in the model for whether firms

restructure outside of courts could make FIRMCF a redundant variable. However, the polychoric

correlation between the two variables is only 21.5%. We further examine the prospect of

multicollinearity by removing all dummies related to restructuring methods and re-estimating the

regression. We find that FIRMCF remains insignificant (results not reported).

28
Our proxy for the probability of default at the time of origination, the all-in-spread (AIS)

over LIBOR for each defaulted loan, shows an unexpected positive, but insignificant estimate.

On one hand, there is no link between default prospects initially and ultimate recoveries, which

we consider the better measure of recovery than the measure using trading prices after default.

On the other, our results suggest that the higher the spread at origination, and thus the higher the

perceived credit risk, the higher the recoveries expected by distressed loan traders, all else

constant.

Model 2 of Table 5 contains the results of estimating the natural logarithm of settlement

value using the OLS method. 23 The significance of all the variables in Model 2 and the

magnitude of the impact are similar to those in Model 1. For example, the median recoveries on

term loans is 7 percent lower than on revolvers when firm size is evaluated at its mean value,

compared with 4 percent in Model 1.24

Model 3 of Table 5 estimates the ultimate recovery determinants again, using a quasi-

maximum likelihood estimation method proposed by Papke and Wooldridge (1996).25 The signs,

statistical significance, and economic magnitude of most of the coefficients are the same as in the

linear model estimated with OLS methods, though the quasi-maximum likelihood approach

generally offers some advantages over the OLS method as discussed earlier.

[Table 5]

23
We lost a number of observations in Model 2 due to the fact that 19 of RR settlement values have a value of zero.
24
Since LOANTYPE is a dummy variable, the semi-elasticity in the log-linear model is interpreted as the median,
not mean, recovery.
25
A number of settlement recoveries exceeds 1 and must therefore be constrained to 1 so that RR is a fraction
bounded between zero and one.

29
6. How well does the trading price serve as a proxy for ultimate recoveries?

In this section, we examine how well trading prices about 30 days after default serve as

proxies for the actual payoffs creditors receive. In particular, we examine whether the trading

price is an unbiased and efficient predictor of ultimate recovery.

The rating agencies, including Moodys, have conducted a number of studies on recovery

and usually employ the price of the loan at the default date or one month afterwards as a measure

of recovery. For example, Emery (2004), among others, uses the average bid price on the loan

one month after default to calculate the recovery rate. The economic rationale for this approach

is that this price reflects the expected present value of recovery or the best proxy for the

ultimate realization of value for each particular debt class.26 Acharya et al. (2007) assume the

discounted price at default is an unbiased estimate of the recovery at emergence. Eberhart and

Sweeney (1992) apply event-study-like methods to the bond market and find results supporting

market efficiency. In particular, they find that the trading price reasonably reflects the ultimate

payoff to bondholders following the resolution of default. However, Emery (2007) regresses the

ultimate recovery on the trading price with both bonds and loans in the sample and finds that the

latter explains only approximately 50 percent of the variation in the former.

There are several practical shortcomings to the use of the trading price as a proxy for

ultimate recovery. First, the measurement requires that the defaulted loans or bonds trade in a

market where prices can be observed. Not all loans trade. Second, the observed price may not

reflect recoveries on the original principal because of partial loan prepayments before and during

bankruptcy.27 Third, the trading price may be depressed by supply or demand fluctuations or

26
Bonelli, Grossman, and OShea (2001).
27
Bonelli, Grossman, and OShea (2001).

30
both which are unrelated to recovery in any precise way.28 For example, Altman et al. (2005)

argue that in a period of tight credit,29 demand for defaulted loans or bonds will be low, while

supply is high. In this case, the market value of the instrument may not reflect the actual

recovery for a lender working towards resolution. Fourth, the market value one month after the

default does not incorporate direct expenses, such as legal fees, that must be incurred during the

workout period and will lower the ultimate recovery. Finally, the 30-day cut-off point for which

trading prices are collected is somewhat arbitrary. It is the scarcity of default loan trading that

prevents analysts from gathering loan prices longer than one month post-default. On average, the

time to resolution of loan defaults ranges from 1.5 to 2.4 years (Gupton, Gates, Carty, 2000;

Araten, Jacobs, and Varshney, 2004; Altman and Hotchkiss, 2006), which is significantly longer

than 30 days.

Methodologically, we take the ultimate recovery as the benchmark. We examine whether

the trading price serves as a reasonable proxy relative to this benchmark. We approach this task

in three ways. We first examine a simple test of the significance of the mean difference between

the ultimate settlement values and the trading prices. Second, we regress the ultimate recovery

solely on the trading price recovery.30 If the trading price is an unbiased predictor of the ultimate

settlement value, the intercept from the regression should be zero and the coefficient of the

trading measure should be one. Finally, using the difference between the two values as the

dependent variable, we estimate a multivariate regression on one of the same specifications we

estimated above. If the secondary loan market is efficient, the 30-day post-default trading prices

28
Altman et al. (2005).
29
Altman et al. (2005) state: For example, in a recession period with increasing default rates, recovery rates would
decrease, leading to higher credit losses. This, in turn, would lead to higher capital requirements and,
correspondingly, possibly to a decrease in the supply of bank credit to the economy, thereby exacerbating the
recession.
30
Emery (2007) conducts this test, but his sample consists of both bonds and loans, while ours has only loans.

31
should incorporate all available information about anticipated recoveries once defaulting firms

emerge from bankruptcy or distress.

We perform a test of the joint hypotheses that the model of bank loan recoveries is

correctly specified and the forecast of the actual recovery is rational. More specifically, if the

trading price is a rational estimate of ultimate recovery, the trading price model should be

identical to the ultimate recovery model. Consequently, in a regression of the difference, which

is simply the observed forecast error, on the true ultimate recovery model, none of the

coefficients should be significant. If any variable or variables remain significant in the

regression, it implies that the forecast error for ultimate recovery could be improved by taking

better account of available information.

Table 6 shows the descriptive statistics of the two measures and their mean and median

differences. The trading price is highly positively correlated with the settlement recovery (82%).

However, the trading price is, on average, 9 cents below the ultimate recovery dollar and the

difference is statistically significant at the 1 percent level. This implies that the average loan is

underpriced with respect to recovery shortly after default.

[Table 6]

The unreported results of our second experiment, the regression of settlement values on

trading prices, show that the intercept is not equal to zero and is strongly significant, while the

coefficient of the settlement measure is not equal to one and is also highly significant. The

adjusted R-squared is 67.33%, which is higher than the 50% R-squared in Emery (2007). The

finding suggests that the trading price is a biased predictor of ultimate recovery.

Finally, we estimate a regression of the difference between the two measures of recovery

on the explanatory variables used in our previous estimations. A number of variable coefficients

32
are significant, which implies these variables maintain predictive power in explaining the

variation in the forecast error.31 In particular, firm leverage one year before default, loan spread,

industry distress, and whether the defaulted firm is in leisure industry, manufacturing industry,

service industry, technology industry, and telecommunications industry are all statistically

significant at the 10 percent or lower level. We conclude that the 30-day trading price, though

highly correlated with the ultimate recovery, is not a rational predictor of settlement value. This

may reflect the fact that factors other than those relevant to recovery are affecting the observed

trading price post-default. Alternatively, investors may be overly pessimistic in a systematic

way with respect to anticipated recovery, given the large uncertainties about the outcomes of the

resolution of financial distress.

7. Robustness checks

The sample in our study includes recovery rates from defaulted bank loans, which can be

observed only when borrowers default. A sample selection bias may arise and least squares

regression using only the observed recovery rates yields inconsistent estimates.32 To account for

possible sample selection bias, we perform Heckmans (1979) two-step procedure. In the second

stage estimation where the inverse Mills ratio was included, we find that the coefficient estimate

of the ratio (LAMBDA) is statistically insignificant (results not reported). The hypothesis tested

by LAMBDA is that there is no selection bias, or non-random selection, or otherwise put,

correlation between the disturbances in the selection and outcome equations. This can be

interpreted as a test of omitted variable bias (LAMBDA being omitted) or of simultaneity bias

between outcome and selection. The statistical insignificance of the coefficient of LAMBDA

means that the hypothesis of no selection bias cannot be rejected in the data. Therefore,

31
Results are not reported to save space.
32
Greene (2003), p. 783.

33
maintaining the hypothesis of no selection bias, least-squares estimation of recovery

rates, RRit = x' + uit , still yields consistent estimators without LAMBDA.

Another econometric issue that stems from the standard process of writing loan

agreement contracts needs some discussion. Most loans contain cross default clauses such that

if a firm defaults on any one loan to a given lender, the firm is automatically in default on

certain other debt instruments specified in the loan agreement contract. This situation can result

in a perfect correlation among certain observations on a given default date. Since we use

instrument-level recovery rates, the sample data are thus clustered among the defaulting

borrowers.

There are 583 distinct borrowers among the 1,364 defaulted loans in the settlement value

sample with the maximum number of loans per borrowers being 14. It is obvious that the number

of clusters far exceeds the loan units within each cluster. In addition, we can assume

independence across borrower clusters. Therefore, the OLS method still produces consistent

estimates and only the asymptotic variance matrix needs to be corrected for clustering effects.33

The results accounting for clustering effects, which are not reported, show that there are no

changes in the magnitude or significance of the variable estimates that were reported in Table 5.

Therefore, we conclude that the factors we identified in our model as drivers of ultimate and

trading price recoveries are robust to sample clustering.

We also examined model re-specifications that apply alternative measures of some of the

variables. First, there may be other time-related effects that affect recovery rates and are not fully

captured by GDP growth, which was the only control for the macroeconomic conditions in Table

5. We therefore add year dummy variables to our base model and re-estimate it using the OLS

33
Wooldridge (2002), p. 135.

34
method. Many year dummies are significant, suggesting that GDP does not fully capture time-

related influences. However, all of the other coefficient estimates except PREPACK remain

unchanged. Although the coefficient of PREPACK is no longer significant in the presence of

year dummies, its interaction with TIMETOEMERGE maintains its sign and statistical

significance. Therefore, whether a borrower had a prepack before entering into a formal or

informal default state still has an effect on recoveries, at least through its impact on the time the

borrower spends in default.

Second, Sabato and Schmid (2008) find GDP growth to be a weak proxy for the state of

the economy in the context of recovery rates. The main reason is that there is a time lag in the

effects of economic conditions on recoveries. Izvorski (1997) finds that the yield on the three-

month Treasury bill (TB3M) is negatively related to the economic growth. He also finds that the

difference between the yield on an AAA corporate bond and a ten-year Treasury bond

(SPREADCORP) has a positive correlation with the economic condition. In place of GDP, we

use TB3M and SPREADCORP separately as two alternative proxies for the state of the

economy. The data used to construct these variables are from the U.S. Department of Commerce

Bureau of Economic Analysis. TB3M is negatively signed and significant as expected.

SPREADCORP also has the expected sign and is strongly significant at the 1 percent level. The

other coefficient estimates remain qualitatively similar (not reported). Therefore, estimates of all

three alternative proxies suggest that higher recovery rates are correlated with economic growth.

Finally, we also re-examined our models with alternative constructions of other variables:

industry financial distress, probability of default, and loan size. The results, which are not

reported for brevity, are consistent with those reported in our main findings in Table 5.

35
8. Conclusion

There are a few academic studies on recovery rates on defaulted debt instruments.

However, they either include bonds and loans together in the sample or base their results on data

from case studies. In addition, much more research has been devoted to default than to recovery.

Since the Basel II regulations tie required bank capital to loss given default, it is important to

understand the factors that influence recovery as well as default. This paper investigates those

factors, using Moodys ultimate recovery database on U.S. loans in default over the period 1987-

2007.

The results show that firm leverage before default negatively affects ultimate recoveries,

while cash flows do not. A variety of loan contract features are strongly related to the ultimate

payoff for creditors. Secured loans have higher recoveries and among the types of collateral,

inventories and accounts receivable result in the highest recoveries. Prior defaults yield higher

recoveries than first-time defaults. Prepackaged bankruptcy reorganization increases actual

settlements. Loan recoveries vary significantly with the length of time to emerge in a nonlinear

manner. Recoveries first decrease and then increase as the time to resolution lengthens. Industry

distress does have a negative impact on settlement recovery rates and there is cross-sectional

variation in recoveries with respect to the type of industry the defaulting borrowers operate in.

Macroeconomic conditions do have a positive impact on recoveries. Probability of default

measured at the time of loan origination is unrelated to ultimate recoveries.

We also examine how well the market for bank loans anticipates ultimate settlement

values for creditors. Our findings suggest that 30-day post-default trading prices are highly

correlated with settlement recoveries, but that using the trading price as a predictor produces a

biased and inefficient estimate.

36
References

Acharya, V., Bharath, S., Srinivasan, A., 2007. Does industry-wide distress affect defaulted
firms? Evidence from creditor recoveries. Journal of Financial Economics 85, 787-821.

Altman, E., 2006. Default recovery rates and LGD in credit risk modeling and practice: An
updated review of the literature and empirical evidence. New York University working paper.

Altman, E., Hotchkiss, E. 2006. Corporate financial distress and bankruptcy, 3rd edition. John
Wiley & Sons (New Jersey).

Altman, E., Brady, B., Resti, A., Sironi, A., 2005. The link between default and recovery rates:
Theory, empirical evidence, and implications. Journal of Business 78 (6), 2203-2227.

Altman, E., Kishore, V., 1996. Almost everything you wanted to know about recoveries on
defaulted bonds. Financial Analysts Journal, November/December, 57-64.

Altman, E., Suggit, H., 2000. Default rates in the syndicated bank loan market: a mortality
analysis. Journal of Banking and Finance 24, 229 - 253.

Araten, M., Jacobs, M., Varshney, P., 2004. Measuring LGD on commercial loans: an 18-year
internal study. RMA Journal, May, 28-35.

Asarnow, E., Edwards, D., 1995. Measuring loss on defaulted bank loans: a 24-year study.
Journal of Commercial Lending, March, 10 (2), 11-23.

Bakshi, G., Madan, D., Zhang, F., 2006, Understanding the role of recovery in default risk
models: empirical comparisons and implied recovery rates. The University of Maryland working
paper.

Bonelli, S., Grossman, R., OShea, S., 2001. Bank loan and bond recovery study: 1997-2000.
Fitch IBCA, Duff & Phelps Loan Products Special Report.

Carey, M., Gordy, M. 2008. The bank as grim reaper: debt composition and recoveries on
defaulted debt. Federal Reserve Systems working paper.

Carty, L., 1997. Moodys rating migration and credit quality correlation, 1920-1996. Moodys
Investors Services, Global Credit Research, Special Comment.

Carty, L., 1998. Bankrupt bank loan recoveries. Moodys Investors Services, Global Credit
Research, Special Comment, 1-16.

Carty, L., Lieberman, D., 1996. Defaulted bank loan recoveries. Moodys Investors Services,
Global Credit Research, Special Comment.

37
Caselli, S., Gatti, S., Querci, F., 2008. The sensitivity of the loss given default rate to systematic
risk: new empirical evidence on bank loans. Journal of Financial Services Research 34, 1-34.

Covitz, D., Han, S., 2004. An empirical analysis of bond recovery rates: exploring a structural
view of default. Federal Reserve Board working paper.

Covitz, D., Han, S., Wilson, B.A., 2006. Are longer bankruptcies really more costly? Federal
Reserve Board working paper.

Das, S., Hanouna, P., 2007. Implied recoveries. Villanova University working paper.

Dermine, J., Neto de Carvalho, C., 2006. Bank loan losses-given-default: a case study. Journal of
Banking and Finance 30, 1219-1243.

Diamond, D.W., 1984. Financial intermediation and delegated monitoring. Review of Economic
Studies 51, 393-414.

Eberhart, A., Sweeney, R., 1992. Does the bond market predict bankruptcy settlements? Journal
of Finance 47(3), 943-980.

Emery, K., 2004. Recovery rates on North American syndicated bank loans, 1989-2003. Special
Comment, Moodys Investors Service, Global Credit Research.

Emery, K., 2007. Moodys ultimate recovery database. Special Comment, Moodys Investors
Service, Global Credit Research.

Emery, K., Cantor, R., 2005, Relative default rates on corporate loans and bonds. Journal of
Banking and Finance 29, 1575-1584.

Emery, K., Ou, S., 2004. Credit loss rates on similarly rated loans and bonds. Special Comment,
Moodys Investors Service, Global Credit Research.

Franks, J., Torous, W., 1994. A comparison of financial recontracting in distressed exchanges
and Chapter 11 reorganizations. Journal of Financial Economics 35, 349-370.

Frye, J., 2000. Depressing recoveries. Federal Reserve Bank of Chicago Policy Studies.

Gersbach, H., Lipponer, A., 2003. Firm defaults and the correlation effect. European Financial
Management 9 (3), 361-377.

Gilson, S.C., John, K., Lang, L. 1990. Troubled debt restructurings: an empirical study of private
reorganization of firms in default. Journal of Financial Economics 27 (2), 315-353.

Gilson, S.C.1990. Bankruptcy, boards, banks, and blockholders: evidence on changes in


corporate ownership and control when firms default. Journal of Financial Economics 27 (2), 355-
387.

38
Gottesman, A., Roberts, G., 2004. Maturity and corporate loan pricing. Financial Review 39, 55-
77.

Grunert, J., Weber, M., 2009. Recovery rates of commercial lending: empirical evidence for
Germany.Journal of Banking and Finance 33, 505-513.

Guo, X., Jarrow, R., Zeng, Y., 2009. Modeling the recovery rate in a reduced form model.
Mathematical Finance 19, 73-97.

Gupon, G., Gates, D., Carty, L., 2000. Bank-loan loss given default. Moodys Investors Service,
Global Credit Research, 69-92.

Harjoto, M., Mullineaux, D., Yi, H-C., 2006. A comparison of syndicated loan pricing at
investment and commercial banks. Financial Management 35, 49-70.

Hu, Y., Perraudin, W., 2002. The dependence of recovery rates and defaults. Birkbeck College
working paper.

Izvorski, I., 1997. Recovery rates and survival times for corporate bonds. International Monetary
Fund working paper.

Kao, D., 2000. Estimating and pricing credit risk: an overview. Financial Analysts Journal 56
(4), 50-66.

Jarrow, R., Lando, D., Turnbull, S., 1997. A Markov model for the term structure of credit risk
spreads. Review of Financial Studies 10, 481-523.

Jokivoulle, E., Peura, S., 2000. A model for estimating recovery rates and collateral haircuts for
bank loans. Bank of Finland discussion paper 2/2000.

Keenan, S., 2000. Moodys approach to evaluating distressed exchange. Moodys Investors
Service, Global Credit Researchs Special Comment, 1-26.

McConnell, J., Lease, R., and Tashjian, E., 1996. An empirical analysis of prepackaged
bankruptcies. Journal of Financial Economics 40 (1), 135-162.

Nickell, P., Perraudin, W., Varotto, S., 2000. Stability of rating transitions. Journal of Banking
and Finance 24, 203-227.

Rajan, R., Winton, A., 1995. Covenants and collateral as incentives to monitor, Journal of
Finance 50, 11131146.

Ryan, V., 2008. Mastering the turnaround. What it really takes to survive a corporate bankruptcy.
CFO magazine (July/August), 63-68.

39
Sabato, G., Schmid, M., 2008. Estimating conservative loss given default. Swiss Institute of
Banking and Finance, University of St. Gallen working paper.

Schuermann, T., 2004. What do we know about loss given default? Wharton Financial
Institutions Center working paper.

Shleifer, A., Vishny, R., 1992. Liquidation values and debt capacity: a market-equilibrium
approach. Journal of Finance 47, 1343-1366.

Strahan, P., 1999, Borrower risk and the price and nonprice terms of bank loans, Boston College
working paper.

Thorburn, K., 2000. Bankruptcy auctions: costs, debt recovery, and firm survival. Journal of
Financial Economics 58, 337-368.

VanDeCastle, K., 1999, Recovering your money: insights into losses from default, Standard &
Poors.

Varma, P., Cantor, R., 2004. Determinants of recovery rates on defaulted bonds and loans for
North American corporate issuers: 1983-2003. Moodys Special Comment, 1-24.

White, H., 1980. A heteroscedasticity-consistent covariance matrix and a direct test for
heteroscedasticity. Econometrica 27, 817-838.

Williamson, O., 1988. Corporate Finance and Corporate Governance. Journal of Finance 43(3),
567-591.

Yost, K. 2002. The choice among traditional Chapter 11, prepackaged bankruptcy, and out-of-
court restructuring. Auburn University working paper.

Zhang, Z., 2009. Recovery rates and macroeconomic conditions: The role of loan covenants,
Boston College working paper.

40
Table 1: Descriptive statistics of discounted settlement and trading price recoveries.
Recovery and loan data come from Moodys Ultimate Recovery Database (1987-2007) and Loan Pricing
Corporations DealScan databse, respectively. Firm data are from Compustat and stock return data are
from Center for Research in Security Prices (CRSP). SETTLEMENT is the discounted recovery amount
as a percentage of the principal amount at default. The discount rate is the defaulted instruments effective
interest rate. LOANSIZE: the dollar amount in millions of the facility at the time of issuance.
LOANTYPE: a dummy variable equal to one if the loan is a term loan and zero if it is a revolver.
ALLASSETCOLL: a dummy variable equal to one if the loan is secured by all firm assets and zero
otherwise. INVENTRECEIVECOLL: a dummy variable equal to one if the loan is secured by inventory
or accounts receivable or both and zero otherwise. OTHERCOLL: a dummy variable equal to one if the
loan is secured differently from the other types and zero otherwise. UNSECURED: a dummy variable
equal to one if the loan is unsecured and zero otherwise. This variable serves as the reference group for
collateral types. COLLATERAL: a binary variable which equals 1 if the loan is collateralized and zero
otherwise. PREPACK: a dummy equal to one if the bankruptcy is through a pre-packaged bankruptcy and
zero otherwise. RESTRUCTURE: a dummy variable equal to one if default is resolved by out-of-court
restructuring, including distressed exchange offers and zero otherwise. OTHERDEFAULT: a dummy
variable, equal to one if default is resolved by other methods than an out-of-court restructuring,
prepackaged formal bankruptcy, or non-prepackaged formal bankruptcy and 0 otherwise.
BANKRUPTCY: a dummy variable equal to one if default occurs only in the context of a non-prepack
formal bankruptcy filing and zero otherwise. TIMETOEMERGE: the length of time in months between
bankruptcy or restructuring and emergence, often known as resolution time. FIRMSIZE: the market
value of firm-level assets one year before default. The market value of firm-level assets is calculated as
the book value of long-term and short-term debt plus the number of common shares outstanding
multiplied by the price per share. FIRMPPE: firm asset tangibility, measured as net property, plant, and
equipment over total book assets one year before default. FIRMCF: firm cash flows, measured as
EBITDA over total book bassets one year before default. FIRMLEV: firm leverage, measured as total
long-term debt plus debt in current liabilities over total book assets one year before default.
EVERDEFAULTED: a dummy variable equal to one if the firm has defaulted before and zero otherwise.
INDDISTRESS: a dummy variable equal to one if the industry median stock returns in the year of default
is less than -30% and zero otherwise. The stock returns are calculated without the defaulting firms and the
industry is defined according to the three-digit SIC codes. GDP: annual GDP growth rate from previous
year. It is calculated at year t-1 where t is the year when the firm defaults. AIS: the spread (in percentage)
at the time of loan origination over LIBOR of the drawn loan that was defaulted

41
Variable N Mean Median Std Dev Minimum Maximum
SETTLEMENT 1364 0.8414 1 0.2822 0 1.4021
LOANSIZE ($M) 1355 142 70 230 0 209
LOANTYPE 1364 0.4611 0 0.4987 0 1
ALLASSETCOLL 1364 0.6261 1 0.4840 0 1
INVENTRECEIVECOLL 1364 0.0242 0 0.1537 0 1
OTHERCOLL 1364 0.2427 0 0.4289 0 1
UNSECURED 1364 0.0770 0 0.2667 0 1
PREPACK 1364 0.1862 0 0.3894 0 1
RESTRUCTURE 1364 0.1246 0 0.3304 0 1
OTHERDEFAULT 1364 0.0191 0 0.1368 0 1
BANKRUPTCY 1364 0.6701 1 0.4704 0 1
TIMETOEMERGE 1364 13.8156 10.2667 14.8546 0 110.1667
FIRMSIZE ($M) 849 1,176 16 10,302 1 157,459
FIRMPPE 838 0.3949 0.3703 0.2339 0 0.9586
FIRMCF 827 0.0529 0.0704 0.1678 -3.0783 0.6880
FIRMLEV 836 0.8156 0.7183 0.4782 0 3.7418
EVERDEFAULTED 1364 0.1261 0.0000 0.3321 0 1.0000
GDP 1364 0.0486 0.0469 0.0132 0.0317 0.0769
AIS 1364 0.0330 0.0275 0.0340 0 0.7500
INDDISTRESS 893 0.1680 0 0.3741 0 1

Table 2: Ultimate recovery rates by default year.


Count of
Year Mean Median recoveries
1987 86.40 100.00 5
1988 87.39 100.00 9
1989 78.75 100.00 27
1990 85.41 100.00 40
1991 88.76 100.00 58
1992 88.36 100.00 63
1993 86.79 100.00 33
1994 91.59 100.00 29
1995 89.38 100.00 37
1996 89.86 100.00 36
1997 92.94 100.00 28
1998 77.22 100.00 20
1999 86.90 100.00 74
2000 79.64 100.00 131
2001 75.30 100.00 193
2002 81.23 100.00 241
2003 82.34 100.00 167
2004 90.38 100.00 96
2005 100.00 100.00 46
2006 99.46 100.00 24
2007 100.00 100.00 7

42
Table 3: Ultimate recovery rates by types of default.

p-value
Non- Other for mean
Prepackaged prepackaged default & difference
formal formal Out-of-court restructuring of time to
bankruptcy bankruptcy restructuring types resolution

(1) (2) (3) (4) (1) - (2)


Mean 90.00 79.83 97.36 92.00 <.0001
Ultimate
recovery Median 100.00 100.00 100 100
1,364 26
N (%) (100) 254 (19) 914 (67) 170 (12) (2)

Table 4: Ultimate recovery rates by industry (Moodys classifications).

Industry Mean Median N


TELECOM 68.49 100.00 134
SERVICE 81.03 100.00 94
TECH 83.72 100.00 66
HEALTH 73.54 100.00 67
OTHERINDUSTRY 52.25 36.83 15
DISTRIBUTION 81.56 100.00 219
MANUFACTURING 82.79 100.00 153
CONSUMER 87.45 100.00 163
ENERGY 91.25 100.00 78
LEISURE 89.58 100.00 98
AUTO 93.92 100.00 42
CHEMICAL 89.35 100.00 25
MEDIA 96.26 100.00 65
TRANSPORT 100.00 100.00 43
CONSTRUCTION 91.19 100.00 14
INDUSTRIAL 100.20 100.00 8
METAL 88.84 100.00 55
NATPRODUCT 93.02 100.00 20
PACKAGING 100.00 100.00 5

43
Table 5: Multivariate regression results.
Recovery and loan data come from Moodys Ultimate Recovery Database (1987-2007) and Loan Pricing
Corporation, respectively. Firm data are from Compustat and stock return data are from Center for
Research in Security Prices (CRSP). Model 1 is an OLS regression result with heteroskedasticity-
consistent standard errors. Model 2 is similar to Model 1 except that the dependent variable is the natural
logarithm of recovery rates. Model 3 is a quasi-maximum likelihood estimate based on the logistic
function. Settlement is the discounted recovery amount as a percentage of the principal amount at default.
The discount rate is the defaulted instruments effective interest rate. LOANSIZE: the natural logarithm
of dollar amount of the facility at the time of issuance. The amount is scaled by the contemporaneous
market value of the firm assets. The market value of firm-level assets is calculated as the book value of
long-term and short-term debt plus the number of common shares outstanding multiplied by the price per
share. LOANTYPE: a dummy variable equal to one if the loan is a term loan and zero if it is a revolver.
ALLASSETCOLL: a dummy variable equal to one if the loan is secured by all firm assets and zero
otherwise. INVENTRECEIVECOLL: a dummy variable equal to one if the loan is secured by inventory
or accounts receivable or both and zero otherwise. OTHERCOLL: a dummy variable equal to one if the
loan is secured differently from the other types and zero otherwise. UNSECURED: a dummy variable
equal to one if the loan is unsecured and zero otherwise. This variable serves as the reference group for
collateral types. COLLATERAL: a binary variable which equals 1 if the loan is collateralized and zero
otherwise. PREPACK: a dummy equal to one if the bankruptcy is through a pre-packaged bankruptcy and
zero otherwise. RESTRUCTURE: a dummy variable equal to one if default is resolved by out-of-court
restructuring, including distressed exchange offers and zero otherwise. OTHERDEFAULT: a dummy
variable, equal to one if default is resolved by other methods than an out-of-court restructuring,
prepackaged formal bankruptcy, or non-prepackaged formal bankruptcy and 0 otherwise.
BANKRUPTCY: a dummy variable equal to one if default occurs only in the context of a non-prepack
formal bankruptcy filing and zero otherwise. This binary variable serves as the reference group.
TIMETOEMERGE: the length of time in months between bankruptcy or restructuring and emergence,
often known as resolution time. PREPACK x TIMETOEMERGE: a term of interaction between
prepackaged bankruptcy dummy variable and time to emergence. FIRMSIZE: the natural logarithm of the
market value of firm-level assets one year before default. The market value of firm-level assets is
calculated as the book value of long-term and short-term debt plus the number of common shares
outstanding multiplied by the price per share. FIRMPPE: firm asset tangibility, measured as net property,
plant, and equipment over total book assets one year before default. FIRMCF: firm cash flows, measured
as EBITDA over total book bassets one year before default. FIRMLEV: firm leverage, measured as total
long-term debt plus debt in current liabilities over total book assets one year before default.
EVERDEFAULTED: a dummy variable equal to one if the firm has defaulted before and zero otherwise.
INDDISTRESS: a dummy variable equal to one if the industry median stock returns in the year of default
is less than -30% and zero otherwise. The stock returns are calculated without the defaulting firms and the
industry is defined according to the three-digit SIC codes. GDP: annual GDP growth rate from previous
year. It is calculated at year t-1 where t is the year when the firm defaults. AIS: the spread (in percentage)
at the time of loan origination over LIBOR of the drawn loan that was defaulted. p-values are in italicized.
***, **, and * indicate 1%, 5%, and 10% levels of significance, respectively. Industry dummy results are
omitted for brevity.

44
(1) (2) (3)
Loan characteristics
LOANSIZE -0.0007 0.0069 -0.0734
0.927 0.660 0.410
LOANTYPE -0.1126 *** -0.1918 ** -0.9357 ***
0.002 0.024 0.000
LOANTYPE x FIRMSIZE 0.0218 *** 0.0376 * 0.1642 ***
0.011 0.067 0.014
ALLASSETCOLL 0.2179 *** 0.3073 *** 1.8415 ***
<0.001 0.001 <0.001
INVENTRECEIVECOLL 0.2994 *** 0.4810 *** 3.7396 ***
<0.001 <0.001 <0.001
OTHERCOLL 0.2031 *** 0.3273 *** 1.7203 ***
<0.001 <0.001 <0.001
Recovery process characteristics
PREPACK 0.0515 * 0.1754 *** 0.3068
0.094 0.007 0.617
RESTRUCTURE -0.0020 0.0373 0.3612
0.964 0.641 0.723
OTHERDEFAULT 0.0136 0.1058 -1.2618
0.791 0.305 0.165
TIMETOEMERGE -0.0067 *** -0.0094 *** -0.0568 *
0.001 0.013 0.079
TIMETOEMERGESQ 0.0001 *** 0.0001 *** 0.0009 *
<0.001 0.001 0.081
PREPACK x TIMETOEMERGE 0.0039 *** 0.0033 0.0794
0.008 0.251 0.418
Borrower characteristics
FIRMSIZE -0.0043 -0.0065 -0.0880
0.623 0.672 0.372
FIRMCF 0.0082 0.0616 0.3749
0.840 0.535 0.706
FIRMPPE 0.0706 0.1162 0.8246
0.142 0.222 0.237
FIRMLEV -0.0626 *** -0.0871 * -0.5358 *
0.006 0.064 0.064
EVERDEFAULTED 0.1075 *** 0.1612 *** 15.7309 ***
0.001 0.002 <0.001
Macroeconomic & industry conditions
GDP 2.4900 *** 3.3172 *** 24.8111 **
<0.001 0.006 0.017
INDDISTRESS -0.0765 *** -0.1566 *** -0.6263
0.011 0.010 0.113
Probability of default
AIS 0.3144 0.3331 0.3846
0.398 0.648 0.935
INTERCEPT 0.3329 * -1.0788 *** -0.2167
0.079 0.004 0.917
N 793 786 793
R-squared (Pseudo R-squared) 0.2510 0.1797 (0.3184)34

34
It should be noted that the pseudo R-squared in Model 3 cannot be used to compare with the R-squared in Models
1 and 2 since the models are completely different.

45
Table 6: Descriptive statistics of ultimate recovery, trading prices and their differences.
Recovery and loan data come from Moodys Ultimate Recovery Database (1987-2007) and Loan Pricing
Corporation, respectively. Settlement is the discounted recovery amount as a percentage of the principal
amount at default. The discount rate is the defaulted instruments effective interest rate. Trading price is
the 30-day post-default trading price of a defaulted loan.

Panel A: N Mean Median Stdev Min Max


Settlement 1364 0.8414 1 0.2822 0 1.4021
Trading price 385 0.6467 0.7 0.2779 0.03 1.0550
Panel B:
Difference =
Settlement
Trading 329 0.0914 0.0682 0.1724 -0.3577 0.7430
Panel C:
Correlation Settlement recovery Trading price
Settlement 1
Trading price 0.8212 1

46