Accounting Conservatism and Creditor Recovery Rate John Donovan Richard Frankel

advertisement
Accounting Conservatism and Creditor Recovery Rate
John Donovan
johndonovan@wustl.edu
Richard Frankel
frankel@wustl.edu
Xiumin Martin*
martin@wustl.edu
Olin Business School
Washington University in St. Louis
Current version: April 2014
*We thank Brad Badertscher, Bryan Graden, Tim Gray (copy editor), Sudarshan Jayaraman, Jared Jennings, Josh Lee,
Nemit Schroff (discussant) and participants at the 2013 Nick Dopuch Conference and participants at the Kansas
University and Notre Dame accounting seminars.
Abstract
We examine the relation between accounting conservatism and creditor recovery rates for firms
in default. We also test the link between conservatism and the length of distress resolutions. We
find creditors of firms with more conservative accounting before default have significantly
higher recovery rates and shorter bankruptcy resolutions. Conservative firms also are more likely
to violate debt covenants before default and enter bankruptcy more quickly after large negative
shocks. Last, conservative firms show a significantly higher probability of emerging from
bankruptcy. These results suggest accounting conservatism preserves creditor value conditional
on default.
Key words: Accounting conservatism, recovery rate, bankruptcy resolution, covenant
violation
JEL: M4, G32, G33, G34
I. INTRODUCTION
We investigate whether accounting conservatism is associated with lenders’ recovery
rates from borrowers in default. Prior research indicates accounting conservatism facilitates
borrowing.1 Conservatism can reduce borrowing rates i) by quickly transferring control rights to
creditors, thereby facilitating renegotiation of lending terms and forcing changes to borrower
investment and financing practices well before financial distress, and ii) by improving creditor
recovery rates in financial distress. We refer to the first role of conservatism as default-risk
control and the second role as recovery-risk control. Our goal is to isolate accounting
conservatism’s relation to recovery risk.
When financial reporting is conservative, net assets are less likely to overstate liquidation
value. Higher recovery rates could result for several reasons. First, creditors can gain more
timely access to control rights through early covenant violation, preventing wealth transfers to
other parties and the continuation of inept management practices. Second, conservatism can
undermine shareholders’ ability to extract value in negotiations with bondholders. Andersen and
Sundaresan (1996) show that, in the presence of costly liquidation, a creditor may not want to
force liquidation.2 In this circumstance, the borrower may have an incentive to underperform on
the debt contract. However, that incentive can be limited if the creditor can threaten liquidation
when the firm’s liquidation value exceeds the promised payment on the debt. The incentive
conflict can be reduced if the creditor can gain decision rights before the liquidation value falls
below the promised payment on the debt, minimizing bankruptcy costs. Third, conservatism can
reduce information asymmetry between investors with regard to the liquidation value of assets,
facilitating more rapid agreement, thereby limiting direct and indirect costs that grow with the
1
Prior studies document that accounting conservatism is associated with reduced cost of credit (Ahmed et al., 2002;
Wittenberg-Moerman, 2008; Zhang, 2008; Nikolaev, 2010; Brockman et al., 2012), lower default risk (Ahmed et al.,
2002), and the development of credit markets (Ball et al., 2003; Ball et al., 2008).
2
The idea that the threat of liquidation is necessary to deter strategic default by borrowers can also be found in
Harris and Raviv (1991), Bolton and Scharfstein (1996), and Hart and Moore (1998).
1
length of the bankruptcy. A reorganization plan is deemed rejected by a class of creditors if the
class would receive more in liquidation (Bankruptcy Reform Act of 1978, Section 1129 (7) (A)).
Financial statements can serve as a starting point in negotiations. Brown (1989) notes the time
consuming, costly nature of valuation proceedings as courts have little appellate court guidance
and lack expertise in case-by-case fact finding.
Alternatively, conservative accounting might have little effect on creditor collections.
Debt agreements could be altered in light of accounting policies. Guay and Verrecchia (2006)
argue, “Bias can be accommodated readily within many contracting settings by simply adjusting
the parameters of the contract” (p. 156). The net-asset threshold triggering technical default, for
example, could be raised to counteract the effects of less conservative accounting policies. If so,
we expect no relation between conservative accounting and creditor recovery rates. However,
adjustments to book value that convert it to an approximation of liquidation may be difficult to
replicate in a debt covenant. Our study aims to assess the descriptive validity of these competing
stories.
We collect a sample of public firms in default from Moody’s Ultimate Recovery Rate
database over the period of 1987-2011.3 We use five measures of accounting conservatism
(Basu, 1997; Ball and Shivakumar, 2005; Beatty et al., 2008; Zhang, 2008) and one composite
measure based on the average quintile rank of the individual measures. This composite is
negatively correlated with measures of asset write-downs after bankruptcy filing, suggesting that
it identifies firms that are less likely to have overstated net assets. We find a positive relation
between accounting conservatism and creditor recovery rate. The positive relation holds for
3
Moody’s defines default as (1) a missed or delayed disbursement of interest, principal, or both; (2) bankruptcy,
administration, legal receivership, or other legal blocks to the timely payment of interest, principal, or both; or (3) a
distressed exchange. Contemporaneous work by Carrizosa and Ryan (2013) uses credit insurance as a proxy for
expected default probability and expected recovery rates. Our results, using actual recoveries, complement theirs.
2
using both “family recovery rate,” which is the debt-principal weighted recovery rate of each
issue, and “issue-specific recommended discounted recovery rate” while controlling for issue
characteristics, firm characteristics, and macroeconomic factors. In addition, more conservative
firms spend less time in bankruptcy, suggesting that accounting conservatism reduces
information asymmetry, facilitating agreements between borrowers and lenders. The data further
demonstrate that firms with higher levels of financial reporting conservatism are more likely to
violate covenants before distress resolution and more quickly enter bankruptcy proceedings after
large negative shocks. Conservative firms also have higher cash flow-interest expense coverage
and higher cash flow to total assets ratio before default. These results suggest that timely
covenant violation, quicker default, and higher asset productivity are likely to be the mechanisms
through which creditors of conservative borrowers improve collections upon default. We also
find that accounting conservatism is positively associated with the likelihood of emergence from
bankruptcy, implying that conservatism aids the resolution of financial distress. Collectively, our
findings highlight the role of accounting conservatism in controlling recovery risk.
Our paper is related with Zhang (2008). We illustrate the differences between the
populations studied by the two papers in Figure 1. Zhang (2008) studies firms that experience a
negative shock. We focus on the subsample of firms that cannot make promised payments to
creditors. Zhang argues that conservative reporting benefits creditors by providing a timely
signal that accelerates covenant violation, conditional on the occurrence of losses. She suggests
that conservative financial reporting can give lenders a measure of the lower bound of the
borrower’s collateral value. Understated collateral might be associated with a higher recovery
rate in default. We aim to examine this proposition.
[Insert Figure 1 here]
3
To the extent that conservative financial reporting is a management choice, enforced
through reputational concerns, commitment to timely loss reporting might break down if default
is the last period in managements’ game. While management turnover in actual defaults is likely
higher than in technical defaults, it is not universal—e.g., Gilson (1989) finds that the average
turnover rate of CEOs in distressed firms is 71%—and does not mark the end of managers’
careers.4 The possibility that managers will adopt an end-game strategy suggests that an
empirical link between accounting conservatism and creditor recovery in bankruptcy might not
exist. Thus, we compliment Zhang’s (2008) work by investigating whether conservatism
enhances creditor value in the event of shocks that cause default.
We also draw on research examining the relation between creditor rights and borrowers’
financing and investment policies (e.g., Chava and Roberts, 2008; Roberts and Sufi, 2009a,
2009b; Nini et al., 2009). These papers suggest that borrowers adjust policies to favor creditors
following covenant violations. Like these authors, we are reluctant to wrench efficiency
implications from our correlations. Yet Chava and Roberts (2008) note that the severity of
agency problems increases with information asymmetry. They find the transfer of decision rights
to creditors has larger effects on borrower investment when information asymmetry is greater.
Research suggests that conservative accounting reduces information asymmetry between outside
investors and managers (LaFond and Watts, 2008; LaFond and Roychowdhury, 2008). Our
results extend these ideas. Conservatism, by narrowing information asymmetry that surrounds
the liquidation value of net assets, can limit the upheaval surrounding financial distress.
This paper proceeds as follows: Section 2 reviews literature on accounting conservatism
and debt contracts and the institutional details of bankruptcy and develops hypotheses. Section 3
describes our sample and our measures of accounting conservatism. Section 4 presents empirical
4
68% firms in default in our sample (Table 1) emerge from bankruptcy proceedings.
4
results. Section 5 discusses robustness tests. Section 6 concludes.
II. RELATED LITERATURE, INSTITUTIONAL DETAILS, AND HYPOTHESES
DEVELOPMENT
Related literature
This paper relates to prior literature on the contracting role of financial reporting
conservatism (Basu, 1997; Watts, 2003a; Watts, 2003b). Conservatism is defined as the
differential verifiability required for recognition of profits versus losses (Watts, 2003) leading to
timing differences in recognition (i.e., anticipate no profit but all losses). The literature argues
conservative financial reporting reduces information asymmetry and facilitates contracting.
Debtholders’ payoffs are an asymmetric function of firm value. Their entire investment can be
lost when the firm performs poorly, but their payoff is limited to promised interest and principal
payments when the firm performs well. To ensure satisfaction of their claims, creditors focus on
the lower portion of the borrower’s net asset distribution (Townsend, 1979; Watts, 2003). Thus
creditors demand conservatism for early warning of financial deterioration via timely covenant
violations and for measurement of the firm’s ability to pay debts. While both aspects can reduce
creditors’ recovery risk, the former also allows lenders to take actions to minimize default risk
(e.g., forcing changes to borrower investment and financing practices well before financial
distress as in Nini et al., 2009).
Ahmed et al.’s (2002) analysis finds that higher levels of accounting conservatism are
associated with lower cost of debt and that this association increases with borrowers’ leverage
and dividend payments. This suggests that conservatism protects lenders by constraining
dividend payments and mitigating other incentive conflicts between equity and debt holders.
Zhang (2008) finds more conservative firms are more likely to violate a covenant over the life of
5
a loan, conditional on experiencing a negative shock. She also shows that lenders reward more
conservative firms with lower interest rates at loan inception, potentially due to expected benefits
from swift control rights transfers. Tan (2013) examines the impact of a covenant violation on ex
post conservatism and finds financial reporting conservatism increases after debt covenant
violations. This increase is more pronounced when lenders gain greater power by appointing a
chief restructuring officer. Beatty et al. (2008) observe that debt covenants terms alone do not
satisfy lenders’ demand for conservatism. They find that more conservative firms are more likely
to have conservative modifications to covenant calculations in their debt contracts. That is,
asymmetric timeliness and contract modifications complement each other in reducing agency
costs.
Biddle et al. (2013) test the relation between accounting conservatism and bankruptcy
risk and find that asymmetric timeliness is associated with lower estimated bankruptcy risk.
Their sample consists of all firms, regardless of their financial condition. By including
nondistressed firms in their sample, they address how conservatism affects default risk but not
recovery risk.5 Carrizosa and Ryan’s (2013) study is related to ours. Using expected recovery
rates estimated from CDS spreads, they show that implied recovery rates increase with the
combination of conservatism and covenants. Their method requires an estimate of default
probability to infer recovery rates. Our work differs from theirs by using realized recovery rates.
Focusing on actual defaults also enables us to examine the intermediate links of the reduced form
correlation. On the other hand, our approach requires control for selection bias that is absent in
Carrizosa and Ryan’s (2013) design.
Background on bankruptcy
5
Default risk is the probability of default, while recovery risk is the product of default probability and recovery rate
given default.
6
Financial distress can be resolved either inside or outside the bankruptcy court.
Bankruptcy Reform Act of 1978 governs disputes that arise in bankruptcy court. A bankruptcy is
a comprehensive process for the resolution of impaired contractual claims against the firm. The
1978 code provides for either liquidation (Chapter 7) or reorganization (Chapter 11). Outsidethe-court resolution typically involves an agreement between creditors and shareholders to
restructure the firm’s financial claims and is implemented through an exchange offer (distressed
exchange). Chapter 11 of the U.S. Bankruptcy Code allows the firm to continue operating while
seeking to satisfy credit claims.
Research fixes on the direct and indirect costs of bankruptcy and on deviations from
absolute priority rule in resolving financial distress through bankruptcy (Weiss, 1990; Franks and
Torous, 1993; Kalay et al., 2007) where absolute priority rule specifies that senior claimholders
receive the total value of their defaulted claim before junior claimholders or equity holders
receive consideration (Weiss, 1990; Franks and Torous, 1994). These studies indicate that the
direct costs of bankruptcy are relatively low. Weiss (1990) reports that direct costs are
approximately 3% of the firms’ total assets and that deviations from absolute priority are
frequent in bankruptcy resolution. Subsequent papers (e.g., Capkun and Weiss, 2008; Ayotte and
Morrison, 2009) document that increased creditor control over bankruptcy in recent years results
in fewer violations of absolute priority rule and improved creditor recovery rates.
Studies also seek factors associated with creditor recovery rates during bankruptcy.
Acharya et al. (2007) find that creditor recovery rates are significantly lower when an industry is
in distress. Franks and Torous (1994) examine the recovery rates of different classes of creditors
in the event of a distressed exchange or a bankruptcy filing. Bris et al. (2003) examine recoveries
for defaults in Arizona and New York over the period of 1995 to 2001. They find the creditors of
7
firms that emerge from Chapter 11 restructurings recover more than those of firms that are
liquidated through Chapter 7.
Some suggest that more rapid restructuring of distressed firms results in a more efficient
bankruptcy process. Jensen (1991) writes: “[It] often takes years to resolve individual cases. As a
result of such delays, much of the operating value of businesses can be destroyed.” Moreover,
the recently enacted bankruptcy reform legislation, the Bankruptcy Abuse Prevention and
Consumer Protection Act of 2005 (BAPCAP), contains elements designed to expedite
bankruptcies.6 The reason may be that the direct costs of restructuring—such as fees for retaining
investment bankers, attorneys, and restructuring professionals—increase with time. Shorter
workouts also reduce the indirect costs by limiting its impact on firm reputation, freeing
management from drawn-out negotiations, and reducing the extent to which firms forego
investment opportunities. Consistent with this view, Thorburn (2000) finds that the costs of
bankruptcy increase with the time in default. Acharya et al. (2007) find a statistically significant
negative relationship between bond recovery rates and the time spent in default.
Hypotheses development
By applying a higher verification standard for recognizing gains, conservative reporting
gives lenders a measure of the lower bound of a firm’s net assets that closely tracks liquidation
value. Future values of the firm and net assets, for example, are generally not verifiable because
many unrealized or unverifiable gains are contingent on firms’ continued operation. Watts
(2003) argues that lenders are better protected when using the lower bound of the current value
6
Specifically, the BAPCPA limits a debtor’s exclusivity period to file a reorganization plan at 18 months after the
commencement of the bankruptcy case and its exclusivity period to solicit vote on the plan at 20 months after the
filing of a plan (§411). In contrast, the former bankruptcy code allows the court to extend both exclusivity periods
indefinitely so long as the requisite cause is established. These new absolute deadlines may encourage more rapid
plan proposals by debtors and thus lead to shorter proceedings as debtors try to minimize the ability of creditors to
stall so as to cause the loss of exclusivity. Other amendments of the BAPCPA that may also expedite proceedings
include a requirement for the debtor to make faster decisions on unexpired leases (§404) and an absolute plan-filing
deadline for small business cases (§437).
8
of net assets to make lending decisions at the loan inception and to assess the ability of the
borrower to repay. The protection afforded by conservatism becomes even more salient when
firms are in distress. For two firms with the same reported net assets before default, the one with
conservative accounting likely has higher collectible value of net worth. This, in turn, results in
higher recovery rate.
Andersen and Sundaresan (1996) further argue that borrowers can exploit liquidation
costs, because, when liquidation is costly and liquidation value is less than the full debt payment,
the creditor may not want to force liquidation even if the borrower does not fully comply with
his obligations. They show the owner behaves strategically by underperforming on the debt
contract in these circumstances, thus forcing the creditor to accept less than the contract amount.
However, the ability of creditors to credibly threaten shareholders with liquidation can limit the
owner’s strategic debt service when the firm’s liquidation value exceeds the promised payment
on the debt. Under these conditions, the owner bears the full liquidation costs. The incentive
conflict therefore can be reduced if the creditor can gain decision rights before liquidation value
falls below the promised payment on the debt. As conservative financial reporting offers a
reliable indicator of liquidation values, we argue that creditors will promptly exercise their
decision rights before the liquidation value falls below the contract amount, leading to higher
recovery rates when borrowers are conservative.
Last, the information asymmetry between creditors and shareholders with respect to the
firm’s asset values can increase the costs associated with bargaining to reach a reorganization
plan during bankruptcy, and prolonged renegotiation can lower creditors’ recovery rate. If
accounting conservatism reduces such information asymmetry, we expect a positive relation
between accounting conservatism and creditor recovery rate and a negative relation between
accounting conservatism and the duration of financial distress resolution. Based on the above
reasoning, our two hypotheses are stated below:
9
H1: Accounting conservatism is positively associated with creditor recovery rates of
firms in financial distress.
H2: Accounting conservatism is negatively associated with the duration of financial
distress resolution.
III. SAMPLE SELECTION AND MEASURES
Sample selection
We identify firms in bankruptcy using Moody’s Ultimate Recovery Database (URD),
which includes data on 987 firms defaulting on total debt obligations greater than $50 million
over the period of 1987-2011. Moody’s URD identifies a default when a firm files for
bankruptcy, misses an interest payment due on its debt, or enters into a distressed exchange with
existing debtholders.7,8 For each firm included in the URD, Moody’s provides the date of default,
date of emergence, the description and total principal amount of each financial instrument in
default, and other detailed information on each security. The key data item of interest for our
study is the creditor recovery rate. The URD database offers three recovery methods: the
settlement method, the liquidity method, and the trading-price method. In the first, the value of
the settlement instruments is taken at or close to emergence. In the second, the value of the
settlement instruments is estimated at the time of a liquidity event, such as the maturity of the
instrument, the call of the instrument, or a subsequent default event. And in the third, the value is
based on the trading pricing of the defaulted instrument taken at or post-emergence. Moody’s
recommends a method to calculate recovery rate for each default that best represents the actual
recovery. The most common is the settlement method. Furthermore, a firm’s family recovery rate
7
The missed or delayed disbursement of interest, principal, or both includes delayed payments made within a grace
period.
8
Moody’s excludes financial institutions as these firms are highly regulated, making their recoveries somewhat
inconsistent and inappropriate for analyzing loss-given-default.
10
is calculated as the total value distributed to creditors at the date of resolution, relative to the total
liabilities in default at the date of default. The family recovery rate measures the weightedaverage collection percentage across all creditor classes.9,10
We supplement the Moody’s URD using the UCLA Bankruptcy Research Database
(BRD) over the period from 1987-2011.11 This database includes details on 886 firms filing for
bankruptcy; 381 of the observations overlap with the URD dataset. However, creditor recovery
rates are only available for sample firms identified through the Moody’s URD. Our financial and
return data are obtained from the Compustat annual industrial file and CRSP monthly stock file.
Of the total sample of 1,492 bankrupt firms contained in the Moody’s URD and the UCLA BRD,
896 firms contain the necessary data to compute our conservatism measures. Figure 2 shows the
sample composition. We exclude all firms lacking Compustat data to compute control variables.
The final sample for our tests of creditor recovery rates contains 557 firms in financial distress
over the period of 1987-2011.
[Insert Figure 2 here]
Measures of accounting conservatism
To remediate noise in firm-specific measures of conservatism, we employ five measures
of financial reporting conservatism (four used in prior literature) and create a composite measure
based on the average quintile rank of the individual measures. All firm-specific measures of
conservatism are estimated over the ten years beginning in the fiscal year immediately before
default. If financial information required for estimation is unavailable at the measurement date,
9
Moody’s follows the standard groups for credit class set by the US bankruptcy court. In the cases of distressed
exchanges and other types of restructurings, debt is classified in a fashion consistent with that of the bankruptcy
court by a Moody’s analyst.
10
All our results are robust to using issue-specific recommended discounted recovery rates, as well as nominal
recovery rates.
11
We thank Lynn LoPucki for providing his Bankruptcy Research Database, available at
http://lopucki.law.ucla.edu/index.htm.
11
we use the prior period data, where available.
Our first measure of conservatism, Cons_Basu, uses the firm-specific Basu (1997)
measure. Specifically, for each firm in our sample, we estimate the sensitivity of earnings to bad
news relative to good news and the ratio of the two sensitivities serves as the first measure. Our
second measure of conservatism, Cons_r2, is the relative explanatory power of bad news in
earnings versus the explanatory power of good news in earnings following Zhang (2008) and
Basu (1997).
One limitation of these measures is that they rely on the firm’s stock price to identify
news and economic earnings (Givoly and Hayn, 2000). We employ additional measures that do
not rely on stock returns. Following Ball and Shivakumar (2005), our third measure of
conservatism, Cons_BS, estimates the extent to which firms record bad news in earnings through
write-offs and losses in accruals. Our fourth measure, Skewness, captures the difference between
the skewness of operating cash flows and earnings, following Beatty et al. (2008).
Each of these measures attempts to identify asymmetric treatment of bad news and good
news in earnings. Given that each of our sample firms performs poorly, operationally and
financially, before the bankruptcy filing, our interest is the extent to which accounting earnings
captures the deterioration that leads to bankruptcy. Our final individual measure of conservatism,
Special Items Ratio, attempts to measure the recognition of this type of news. We compute the
average special items recorded in the Income Statement (SPI) scaled by the average total assets
over the three years prior to default as a proxy for these recorded losses.12 We then compute
Special Items Ratio as the ratio of special items to cumulative firm stock returns on CRSP in the
three years before default. (Appendix A contains a detailed discussion of the five measures of
12
All results continue to hold if we compute the sum of special items recorded in the Income Statement scaled by
the average total assets over the three-year period prior to default as a proxy for the recorded losses.
12
conservatism.) We standardize each conservatism measure to have a mean of zero and standard
deviation of 1 to make the coefficients of conservatism across different proxies comparable.
Finally, we create a composite measure of conservatism, All Conservatism, based on the average
quintile rank of each individual measure, requiring at minimum two individual conservatism
measures to calculate the composite measure.
IV. EMPIRICAL RESULTS
Descriptive statistics
Table 1 presents descriptive statistics for all sample firms in bankruptcy over the period
of 1987-2011, identified using Moody’s Ultimate Recovery database and the UCLA Bankruptcy
Research database. The number of observations for each dependent variable varies depending on
its availability in each of these two data sources. Each test uses the maximum observations
available. Creditor recovery rates (Recovery Rate) are family recovery rates and are only
available for the 594 sample firms identified in the Moody’s Ultimate Recovery Database.
The mean (median) creditor recovery rate across all creditor classes is 53.39% (52.46%),
indicating that creditors of sample firms lose substantial value on their claims. The average
sample firm in the Moody’s URD defaults on approximately $1.1 billion of total debt,
corresponding to average creditor losses of approximately $513 million.
Factors that can
improve creditor recovery rates therefore can have significant economic value. Asset writedowns after firms enter bankruptcy account for about 0.7% of total assets, while losses reported
after bankruptcy filing are approximately 0.1% of total assets. Approximately 66% of sample
firms violate a debt covenant in the year before default, consistent with the idea that financial
distress triggers debt covenants. On average, 68% of sample firms emerge from bankruptcy, and
sample firms require approximately 422 days (14 months) to resolve financial distress. This
13
resolution period is shorter than prior papers, which indicates the average time from the
bankruptcy filing date to the emergence date approximates two to three years (Weiss, 1990;
Franks and Torous, 1994). This is consistent with shorter resolution periods following the 2005
Act, which is intended to expedite the bankruptcy process. The mean (median) lag between
bankruptcy filing and negative economic news reflected in stock returns is approximately 7.6
(10.0) months. On average, sample firms do not have sufficient operating cash flows to cover
their annual interest expense before bankruptcy, with average (median) Int Coverage of 0.917
(0.390).
Additionally, sample firms display high leverage before filing for bankruptcy. The mean
(median) firm’s total debt is approximately 81% (70%) of the firm’s total assets in the quarter
before default, reinforcing the liquidity constraints indicated by coverage ratios. Sample firms
incur losses prior to default, with mean (median) return on assets of -0.097 ( -0.035), consistent
with these firms experiencing declining operating performance before default. Firms’ Herfindahl
debt concentration ratio is 0.58 (0.56) at the mean (median), indicating that debt is concentrated
with a few banks at default. Bank debt and secured debt constitute 37% and 44%, respectively,
of total debt upon bankruptcy filing. The average firm carries approximately a “B-” S&P credit
rating in the year before default, which suggests that, although the mean firm had the ability to
meet its current obligations in the year before default, its financial condition was vulnerable to a
decline in the near future.13
[Insert Table 1 here]
Table 2 presents Pearson and Spearman correlations among variables below and above
13
Standard & Poor’s Guide to Credit Rating Essentials, available at http://www.standardandpoors.com/ratings/,
defines a firm with a B credit rating as “more vulnerable to adverse business, financial and economic conditions, but
currently has the capacity to meet financial commitments.”
14
the diagonal, respectively, with the significance below 10% level bolded. The table shows that
creditor recovery rate is negatively associated with ex post asset write-downs and the duration of
bankruptcy resolution and positively associated with the likelihood of emergence from
bankruptcy. The five individual measures of accounting conservatism—Cons_r2, Cons_Basu,
Cons_BS, Skewness, and Special Items Ratio—are all positively related with creditor recovery
rates but only statistically significant for the Cons_r2 measure. In addition, the composite
measure of conservatism is positively associated with recovery rate with statistical significance
at the 10% level. Finally, firm size, bank share, and the percentage of secured debt are all
positively associated while leverage is negatively associated with recovery rates. These
univariate results highlight the importance of controlling for these factors while estimating the
relation between accounting conservatism and creditor recovery rates. The correlation
coefficients are low except for that between firm size and debt concentration (-0.658) and
between bank share and percentage of secured debt.
[Insert Table 2 here]
Results of the relation between accounting conservatism and creditor recovery rate
We hypothesize that accounting conservatism is positively associated with creditor
recovery rates of firms in default in H1. To test this hypothesis, we estimate the following model:
Recoveryi =
α0 + β1Conservatismi + β2Sizeit + β3Leverageit-1 + β4Prior
ROAit + β5Distressed Exchangei +β6Debt Concentrationi +
β7Bank Sharei + β8Secure Debt%i + β9Bankruptcy Periodi +
β10Distressed Exchangei + βkMacroeconomic Controls + εit
(1)
where Recovery is the dependent variable, which is the family recovery rate.14 We follow
14
We also perform this analysis using the issue-specific recommended recovery rate as the dependent variable,
measured at the debt instrument level, while controlling for firm and debt-contract specific variables. The results of
this estimate are qualitatively similar to those presented in Table 4. Specifically, the Cons_Basu measure is
15
Acharya et al. (2007) and Zhang (2009) and select a comprehensive set of control variables in
our test. Appendix B addresses the reasons for including these control variables, consisting of
firm characteristics, debt structures, and macroeconomic variables. We also include FamaFrench 12 industry and year fixed effects in the model.
Table 3 presents the results of this test. Panel A tests the relation between conservatism
and creditor recovery rates using OLS estimation with standard errors clustered by year. The
coefficient on conservatism is positive for all five measures, and all but one is statistically
significant at the 10% level or better. Based on All Conservatism, moving from the first to the
fifth conservatism quintile, creditor recovery rate increases by 15%, approximately 29% relative
to the mean. All other measures demonstrate a similar economic effect: a one standard deviation
increase in conservatism is associated with an increase in creditor recovery rate ranging from
1.0% based on Cons_BS to 3.5% based on Cons_r2.
The signs of the coefficients on control variables are largely consistent with our
predictions. For example, firms that are larger, have a higher percentage of secured debt, and use
distress exchange offers show higher creditor recovery rates. The duration of bankruptcy is
negatively associated with recovery rates. The macroeconomic control variables also load with
expected signs. Overall, we find evidence supporting H1 that higher levels of financial reporting
conservatism are associated with higher creditor recovery rates for firms in financial distress.
Our goal is to study creditor-recovery rates for firms unable to make current or future
promised payments. If the URD dataset is not comprehensive, i.e., it contains a subset of
observations in the darker circle of Figure 1, inferences based on it may be biased. To control for
self-selection, we employ Heckman two-stage estimation. We use firms in the bottom 10% of
insignificant (p-value .21), while the Cons_BS measure is statistically significant at less than the 1% level. All other
results remain unchanged.
16
returns in the intersection of Compustat and CRSP between 1987 and 2011 to approximate the
population of insolvent firms. The first stage includes these firms as well as all firms in
bankruptcy identified from the Moody’s URD dataset and the UCLA Bankruptcy Research
dataset.15 The explanatory variables included in the first stage are Size, Leverage, Debt Rating,
Litigation, and an indicator variable for Big N auditors, as well as the conservatism measures and
macroeconomic control variables. All variables are defined in Appendix C. The percentage of
concordant pairs is 91.1%, and the percentage of discordant pairs is 8.6%, suggesting that the
explanatory variables predict whether a firm enters default reasonably well. Panel B of Table 4
reports the results of the second stage regression. The coefficients on each of our conservatism
measures are both qualitatively and quantitatively similar to that reported in Panel A. Moreover,
the coefficient on the inverse-Mills ratio is positive and statistically significant in some models,
suggesting that the Heckman method is removing a portion of the selection bias. For brevity, we
report subsequent analysis based on this composite measure only.
[Insert Table 3 here]
The results of the relation between accounting conservatism and the duration of financial
distress resolution
We hypothesize that accounting conservatism is negatively associated with the duration
of financial distress resolution in H2. To test this hypothesis, we estimate the following hazard
model:
ln hi(T) =
h0(T) + γ1Conservatismi + γ2Sizeit + γ3Leverageit-1 + γ4Prior ROAit
+ γ5Litigationi + γjDefault Macroeconomic Controls+
γkEmerge Macroeconomic Controls + εit,
(2)
15
The results of Heckman two-stage estimation are not sensitive to the use of this sample. We have also tried other
ways of selecting the first-stage sample firms, including all Compustat firms, all Compustat firms with annual stock
returns in the middle 10 percentile, or with negative cash flows, stock returns, or ROA over the period from of 19872011. The results are robust to each of these samples used in the first stage. We choose the bottom 10 percent of
Compustat firms to report because this yields the highest percentage of concordance pairs among all the alternatives.
17
where hi(T) is the probability that a bankrupt firm i emerges from bankruptcy at t given that the
firm has survived up to T. h0(T) is the underlying hazard rate corresponding to the probability of
emerging from bankruptcy when all the explanatory variables are set to zero. Because the
dependent variable is the time to emergence, which is right skewed due to the presence of
censored observations, we estimate model (2) using hazard model.16,17 The dependent variable is
defined as the time span in days between the bankruptcy filing date and the date of emergence.
If the bankruptcy proceeding is incomplete, we define the dependent variable as the number of
days beginning with the bankruptcy filing date to the end of the sample period. We censor all
observations that did not emerge from bankruptcy. Year and Fama-French 12 industry fixed
effects are included in the model.
Table 4 reports the results of estimating model (3). Column (1) includes all firms that
filed bankruptcy while column (2) excludes firm observations that exist through either an
acquisition or a liquidation. The coefficient on the All Conservatism is positive and statistically
significant at the 1% level for both columns, which indicates that the hazard of an emergence
increases with this measure of conservatism. Overall, we find evidence suggesting that
conservative financial reports facilitate agreement on a plan for reorganization. We also find that
firms with higher leverage and better profitability tend to emerge with a shorter duration while
bankruptcies during high GDP growth periods and periods with higher speculative interest rates
tends to have longer duration.
[Insert Table 4 here]
The results of the relation between accounting conservatism and debt covenant violation
16
Specifically, we estimate the Cox proportional hazard model (Cox, 1972), where the hazard rate does not vary
over time and the functional form of the baseline hazard is not required.
17
Right censoring occurs when firms may emerge after the data collection period or when firms never emerge from
bankruptcy.
18
To understand the institutional mechanisms linking accounting conservatism to higher
creditor collection, we examine whether accounting conservatism is positively associated with
covenant violations. The maintained assumption of this analysis is that control rights are
transferred to creditors upon covenant violation, which will allow creditors to determine whether
to liquidate borrowers in their own best interest.18 Finding such evidence will also corroborate
the results of Zhang (2008), demonstrating that accounting conservatism increases the
probability of covenant violation for firms both bankrupt and solvent firms.
We collect debt covenant data for our sample of default firms by searching and reading
public disclosures in the firm’s Form 10-K, Form 10-Q, and Form 8-K in each quarter over the
five years before the default date using Morningstar’s 10-K Wizard.19,20 Following Nini et al.
(2011), we identify a firm as having violated a covenant if the firm’s disclosures specifically
indicate that it violated a financial covenant, modified or amended a credit agreement to ensure
compliance with covenants, or obtained a waiver for covenant violations during the period.
To test the relation of accounting conservatism with covenant violations before the
default date in a multivariate analysis, we estimate a probit model where the dependent variable
equals 1 if the firm reported a covenant violation within a year before the date of default and 0
otherwise. The variable of interest is our composite measure of accounting conservatism.
Consistent with Zhang (2008), Table 5 shows that accounting conservatism is significantly
18
This assumption is reasonable because Davydenko (2012) finds that at least 8% of all bond payment defaults are
triggered by a covenant violation that prompts senior creditors (i.e., banks) to block a scheduled bond payment,
triggering bankruptcy.
19
We thank Greg Nini, David Smith, and Amir Sufi for providing covenant violations data, available on Amir Sufi’s
website at the University of Chicago at http://faculty.chicagobooth.edu/amir.sufi/data.html. This data set includes
covenant violations for 26 of our sample firms, which we include in our analysis. Please refer to the appendix of
Nini, Smith, and Sufi (2011) for details on information related to this data set. For the remaining 628 firms, we sift
through filings on covenant violations with the aid of 10-K Wizard.
20
Public financial reports are available on 10-K Wizard after January 1, 1994; thus, we restrict our attention to firms
defaulting on their obligations after 1994. Additionally, we eliminate firms with no public disclosures available on
10-K Wizard, which results in a total sample of 654 firms for our analysis.
19
positively associated with covenant violations in the year preceding default at the 1% level,
which indicates that conservative firms are more likely to transfer control rights to creditors over
this period.21
[Insert Table 5 here]
To provide further evidence on the means through which conservatism can affect
recovery rates, we conduct additional analysis. If creditors exercise their control rights to
preserve firm value and exert influence over the bankruptcy decision, we expect conservative
firms will file for bankruptcy on a timelier basis following the realization of negative economic
shocks. Using CRSP stock returns as a proxy for the negative news, we identify the month with
the lowest monthly abnormal stock returns in the three fiscal years before default as the first
month in which the negative shocks drive firms into bankruptcy. The natural log of the number
of months from the first month of shocks to the date of default serves as the dependent variable.
Using the same set of control variables in the covenant violation analysis, results presented in
Table 6 show that accounting conservatism is significantly, negatively associated with the
bankruptcy lag (coefficient estimate = -0.20 and t-statistics = -2.79). Therefore the evidence
suggests that conservative firms preserve firm value by reducing the lag between the realization
of negative shocks and the bankruptcy filing date.
[Insert Table 6 here]
The results of the relation of accounting conservatism with interest coverage and cash flows
prior to default
Conservative firms are more likely to violate covenants after negative outcomes.
Therefore conservative firms are likely to be more solvent and have greater liquidity at default
21
We also split firms into two groups, High Conservatism firms and Low Conservatism firms, based on the firm’s
composite measure of accounting conservatism: All Conservatism. Untabulated results indicate that more High
Conservatism firms in our sample violate financial covenants (257) than Low Conservatism firms (176).
20
for two reasons. First, their managers have less time to destroy or transfer value. Second, their
bankruptcy is more likely to result from a covenant-based transfer of control rights than a missed
interest payment. Testing these conjectures illuminates the mechanism through which accounting
conservatism boosts creditor recovery rates. We estimate the effect of conservatism on
performance before default using the following model:
Performit-1,it-3 =
α0 + ω1Conservatismi + ω2Sizeit + ω3Avg. Leverageit-1, it-3 +
ω4Avg. ROAit-1, it-3 + ω5Avg. Capexit-1, it-3 + ω6 Avg.
Dividendsit-1, it-3 + ω7Litigationi +ωkMacroeconomic
Controls + εit
(3)
We use two different dependent variables. First, we measure the ratio of operating cash
flows to annual interest expense, averaged over the three years before default.22 If the borrower’s
solvency matters in the timing of the bankruptcy filing, we expect a positive coefficient on our
composite measure of conservatism, indicating that conservative firms enter bankruptcy in a
better liquidity position. Table 1 indicates that, on average, sample firms do not have sufficient
cash flows to cover interest expense before default (average Int Coverage is 0.917). Therefore
ensuring the firm has sufficient cash flows to repay interest on its debt significantly improves
creditor collections. Second, we measure the ratio of annual operating cash flows to total assets,
averaged over the three years before bankruptcy (CFO) as a proxy for asset productivity before
default. We include control variables for firm size, leverage, and profitability, as well as capital
expenditures and dividends paid to shareholders; these financial control variables are averaged
over the three years before default, matching the measurement of the dependent variable.
22
Accounting conservatism mechanically reduces earnings through recognition of unrealized losses, as indicated by
the negative and significant correlation between our conservatism measures and ROA before bankruptcy filing in
Table 2. We attempt to capture the firm’s liquidity position before default, and therefore we measure interest
coverage as the ratio of annual operating cash flows to interest expense before default. We use the three-year
average before default to mitigate timing problems inherent in cash flows.
21
Additionally, we control for litigation-prone industries, macroeconomic conditions before
default, and industry and Fama-French 12 industry fixed effects. Standard errors are clustered by
year.
The results of this analysis are presented in Table 7. In column 1 of Panel A, the
dependent variable, Int Coverage, measures the average interest coverage over the three years
before default. The coefficient on All Conservatism is positive and significant at the 10% level.
In column 2, the results confirm our expectation that conservatism is positively related to asset
productivity prior to default. Overall, the results in Table 7 are consistent with the hypothesis
that conservatism preserves firm value by improving the timeliness of bankruptcy filing.
[Insert Table 7 here]
The relation between accounting conservatism and the likelihood of emergence from
bankruptcy
Higher creditor recovery for conservative firms may result either from wealth transfers
from shareholders to creditors or from improvement of efficiency during bankruptcy. We study
emergence from bankruptcy conditional on pre-bankruptcy performance to understand which
explanation the data supports. Emergence from bankruptcy resolution, conditional on pre-filing
ROA, is a measure of performance improvement. If creditors hollow out the firm for their private
benefit, we expect a lower probability of emergence. Table 1 presents descriptive statistics for
firms emerging from bankruptcy in our sample. Approximately 68% of firms in our initial
sample (541 firms) emerge from financial distress. Only 24% of sample firms in Hotchkiss
(1995) emerge from bankruptcy, while 44% of firms reorganize and emerge from bankruptcy in
Kalay et al. (2007).
We test the relation between accounting conservatism and the likelihood of emergence
from bankruptcy using the following probit model:
22
Prob(Emergei =1) = α0 +
4Prior
1Conservatismi
ROAit +
Controls+
+
2Sizeit +
5Litigationi +
kEmerge
3Leverageit-1 +
kDefault
Macroeconomic
Macroeconomic Controls + εit
(4)
The dependent variable in this model, Emerge, is a dummy variable equal to 1 if a firm emerged
from bankruptcy or completed its debt exchange or restructuring and 0 otherwise. All other
variables are defined in Appendix C. If conservatism improves creditor recovery rates by wealth
transfers, then it should not be systematically associated with the likelihood of emergence.
However, if conservatism improves the overall efficiency by allowing creditors to take valuemaximization actions or reduce deadweight losses by mitigating information asymmetry between
shareholders and creditors, we anticipate that accounting conservatism will be positively
associated with the firm’s probability of emergence from bankruptcy.
Table 8 presents the results of estimating model (6) using a probit model. The results
show that accounting conservatism is loaded with a positive coefficient and is statistically
significant at the 10% level after controlling for accounting performance before default. The
coefficient estimates on other control variables indicate that larger firms and firms with higher
leverage and ROA have higher likelihood of emergence. The marginal effect of accounting
conservatism on the probability of emergence is 3.3%.
[Insert Table 8 here]
V. ROBUSTNESS TESTS
Our results suggest that accounting conservatism benefits lenders. However, one might
argue that conservative financial reporting gives too much power to creditors, who can force
healthy firms into bankruptcy to achieve higher recovery. We illustrate this issue in Figure 1 by
drawing the URD circle so that it is not a proper subset of the darker circle. Note the Heckman
23
two-stage procedure discussed in Section 4.3 potentially addresses this selection bias since
accounting conservatism is included as a predictor in the first-stage regression. To illuminate this
issue, we investigate whether a disproportionally high percentage of conservative (less
conservative) firms lies in very high (low) creditor recovery rate region. Finding such evidence
would suggest that conservatism induces unnecessary liquidation. We partition our sample into
high and low conservative firms based on sample median of All Conservatism and plot the
cumulative distribution of recovery rate of high and low conservative firms separately. The
distribution reflects the probability that a realized recovery rate is less than or equal to a specified
threshold. For example, approximately 40% of firms in the High Conservatism subsample have
recovery rates less than 50%. If conservatism pushes healthy firms into bankruptcy
unnecessarily, we would expect the cumulative distribution of recovery rates of High
Conservatism firms to experience a jump in the high creditor recovery rates region (i.e., above
80%). However, the figure shows that there is no difference in the frequency of distribution
between high conservative and low conservative firms in the high creditor recovery rates region
(i.e., above 80%). Furthermore, the frequency difference between the two groups of firms mainly
lies in the relatively low recovery rates region. We perform a Kruskal Wallis test to determine
whether the recovery rate distribution differs between high and low conservative firms.
Untabulated results indicate no statistically significant difference between the two groups (pvalue = 0.15), and the observed difference between the two groups mainly resides in the low
recovery rate region. Thus conservatism does not appear to push healthy firms into bankruptcy.
[Insert Figure 3 here]
Furthermore, we check the relation between stock-price reaction at bankruptcy filing
announcements and accounting conservatism. If conservatism allows creditors to push healthy
24
firms into bankruptcy or otherwise transfer wealth from shareholders, we expect a negative
relation between accounting conservatism and share price changes. We fail to find a systematic
relation, suggesting that shareholders do not expect to be harmed by greater conservatism.
Moreover, shareholders are able to anticipate payouts (Gilson et al. 1990). We find a significant,
positive relation between stock reactions at bankruptcy filings and subsequent creditor recovery
rates. This evidence again reinforces the view that higher creditor recovery results from reduced
losses in reorganization rather than at the expense of shareholders.
Last, we exclude firms with creditor recovery rates above 80% from the sample and reestimate model (1) based on this smaller sample. The firms excluded are relatively healthier and
thus may have been pushed by a creditor into bankruptcy resolution inefficiently. All results are
qualitatively similar.
Robustness tests
Validation of the measures of accounting conservatism
To validate our measures of accounting conservatism, we correlate our composite
measure with post-bankruptcy filing asset write-downs. By law and rule, within 15 days of the
filing of a bankruptcy petition, firms must file schedules of the debtor’s assets and liabilities with
the bankruptcy court. Failure to file the appropriate schedules quickly results in dismissal of the
bankruptcy filing.23 A reasonable conjecture is that court supervision and creditor pressure force
firms to more conservative reporting during bankruptcy proceedings. Thus bankruptcy filing can
cause firms that were less conservative before bankruptcy to write down assets. If so, we predict
that higher (lower) levels of accounting conservatism before bankruptcy filing are associated
23
Bankrupt publicly traded companies are not relieved of their SEC-related financial reporting requirements.
However, the SEC generally will accept the monthly reports a public company must file with the bankruptcy court
in lieu of its quarterly and annual reports (Form 10-K and Form 10-Q), respectively. The company must file each
monthly report with the SEC using Form 8-K within 15 calendar days after the monthly report is due to the
bankruptcy court.
25
with lower (higher) asset write-downs during bankruptcy proceedings. Such correlation will
support the internal validity of our accounting conservatism measures.24
We measure ex post write-downs in two ways: (i) by searching for the cumulative loss
reported in the monthly operating report or SEC filing available on Compustat and (ii) by
examining the change in the firm’s total assets following bankruptcy filing. First, we measure
Ex-Post Loss as (-1) multiplied by the firm’s net income scaled by total assets after default. We
also scale the total loss by the number of days over the write-down measurement period, because
each monthly operating report filed with the bankruptcy court is not standardized with respect to
the number of days covered in each report.25 Second, we measure Ex-Post Asset WD as (-1)
multiplied by the percentage change in the firm’s total assets from the last balance sheet before
default to the first balance sheet following default. Total write-downs are then scaled by the
number of days between the default date and the post-filing balance sheet date, because firms
with a greater lag between the default date and the balance sheet date are more likely to record
greater asset write-downs.
Table 9 reports the results of this analysis with ex post asset write-downs as the
dependent variable, with All Conservatism as the variable of interest. In both columns, the
coefficient on our composite measure of accounting conservatism is negative and statistically
significant at conventional levels. The evidence suggests that more conservative firms experience
lower asset write-downs after bankruptcy filing and provides validation to our measures of
conservatism.
24
To conduct this test, we hand-collect monthly operating reports filed with the bankruptcy courts from public
disclosures available in firms’ Form 10-K, Form 10-Q, and Form 8-K during the bankruptcy resolution period using
Morningstar’s 10-K Wizard for the URD sample for firms whose first SEC filing is unavailable on Compustat
following default and before emergence from bankruptcy. This procedure results in a total sample of 217 firms with
sufficient financial data to compute asset write-downs following default.
25
The average length between the bankruptcy filing date and the date of first monthly operating or quarterly
financial report is approximately 48.3 days. The short period means that write-downs recorded following bankruptcy
filings reflect pre-bankruptcy losses.
26
[Insert Table 9 here]
Control for growth opportunities, prepackaged bankruptcy, and alternative measure of
bankruptcy risk
We employ several robustness tests using additional control variables and econometric
designs to ensure that our results are not driven by correlated omitted variables. We first include
the firm’s growth opportunities prior to bankruptcy. Untabulated results indicate that high
growth opportunities (MTB) are insignificantly associated with creditor recovery rates. Most
importantly, the coefficient on our conservatism measures remains positive and statistically
significant at the 10% level or better except for the skewness measure, and our composite
measure of conservatism remains significant at the 5% level.
Under a prepackaged bankruptcy agreement, the firm and its creditors jointly determine
the allocation of firm value to claimants and specify this allocation in a contract before the firm
enters Chapter 11. Prepackaged bankruptcies may significantly improve the efficiency of
resolution by eliminating negotiations during the proceedings and may result in higher creditor
recovery rates than standard Chapter 11 proceedings. We include prepackaged bankruptcy as a
dummy variable in estimating model (1).26 The untabulated results show that prepackaged
bankruptcies are significantly positively associated with creditor recovery rates. Yet the impact
of conservatism on recovery rates remains quantitatively and qualitatively similar after
controlling for prepackaged bankruptcy. Finally, we include additional control variables MTB
and Prepackage at the same time in estimating model (1). Our results are robust to this estimate.
In untabulated analysis, we employ an alternative measure to firm leverage, distance to
default, to estimate a volatility adjusted measure of the remaining slack the firm has in its
26
We first identify prepackaged bankruptcy filings using the UCLA Bankruptcy Research Database (BRD). We can
match 342 firms from our sample to the BRD data; for the remaining 215 firms in our sample, we assume a
prepackaged filing if the firm’s resolution period is less than four months (excluding successful distressed
exchanges), following Ayotte and Morrison (2009).
27
covenants.27 Higher distance to default levels implies lower default likelihood, and thus we
expect a positive relation between distance to default and creditor recovery rates. The results for
the accounting conservatism measures are qualitatively similar when we replace leverage with
distance to default. Finally, our results continue to hold when we cluster standard errors by
industry or industry-year.
VI. CONCLUSION
We examine the relation between accounting conservatism and creditor recovery rates.
We hypothesize that accounting conservatism preserves firm value by facilitating timely
transfers of decision rights from shareholders to creditors through covenant violations. We also
hypothesize that conservative financial reporting improves the efficiency in bankruptcy
resolution by mitigating information asymmetry between shareholders and creditors and
facilitating agreement on a reorganization plan among these parties.
We find evidence that accounting conservatism is positively associated with creditor
recovery rates. We also find that conservative firms have shorter bankruptcies, consistent with
our argument that conservative reports mitigate information asymmetry. Next, we show that
conservative firms have a higher likelihood of violating debt covenants before default and file for
bankruptcy sooner after negative economic shocks. We also demonstrate that conservative firms
have better solvency and asset productivity before default. Therefore we not only demonstrate
the robustness of the findings in Zhang (2008) for defaulted firms but also suggest that timely
covenant violation, quicker default, and higher asset productivity are likely to be the mechanisms
through which creditors of conservative borrowers improve collections upon default. We also
find that conservative firms have higher likelihood of emergence from bankruptcy, further
27
We thank Tyler Shumway for SAS code used to calculate distance to default for all sample firms
available at http://www-personal.umich.edu/~shumway/papers.dir/nuiter99_print.sas.
28
suggesting that conservatism improves the efficiency of resolution. Finally, our results are robust
to a number of alternative estimation procedures such as Heckman two-stage estimation. They
also hold after controlling for additional factors that might be correlated omitted variables.
Overall, our study suggests that conservative accounting benefits creditors of borrowers in
financial distress and improves the efficiency of bankruptcy resolution, which likely benefits all
stakeholders of the firm.
29
References
Acharya, V., S. Bharath, and A. Srinivasan. 2007. “Does Industry-wide Distress Affect
Defaulted Firms? Evidence from Creditor Recoveries.” Journal of Financial Economics
85 (3) (September): 787–821.
Ahmed, A., B. Billings, R. Morton, and M. Stanford-Harris. 2002. “The Role of Accounting
Conservatism in Mitigating Bondholder-Shareholder Conflicts over Dividend Policy and
in Reducing Debt Costs.” Accounting Review 77 (4) (October): 867–890.
Anderson, R. W., and S. Sundaresan. 1996. “Design and Valuation of Debt Contracts.” Review of
Financial Studies 9 (1) (January 1): 37–68.
Ayotte, K., and E. Morrison. 2009. “Creditor Control and Conflict in Chapter 11.” Journal of
Legal Analysis 1 (2) (June 20): 511–551.
Ball, R., A. Robin, and G. Sadka. 2008. “Is Financial Reporting Shaped by Equity Markets or by
Debt Markets? An International Study of Timeliness and Conservatism.” Review of
Accounting Studies 13 (2-3) (September 1): 168–205.
Ball, R., A. Robin, and J. Wu. 2003. “Incentives Versus Standards: Properties of Accounting
Income in Four East Asian Countries.” Journal of Accounting and Economics 36 (1–3)
(December): 235–270.
Ball, R., and L. Shivakumar. 2005. “Earnings Quality in UK Private Firms: Comparative Loss
Recognition Timeliness.” Journal of Accounting and Economics 39 (1) (February): 83–
128.
Basu, S. 1997. “The Conservatism Principle and the Asymmetric Timeliness of Earnings.”
Journal of Accounting and Economics 24 (1) (December): 3–37.
Beatty, A., J. Weber, and J. Yu. 2008. “Conservatism and Debt.” Journal of Accounting and
Economics 45 (2–3) (August): 154–174.
Bharath, S., and T. Shumway. 2008. “Forecasting Default with the Merton Distance to Default
Model.” Review of Financial Studies 21 (3): 1339-1369.
Biddle, G., M. Ma, and F. Song. 2013. “Accounting Conservatism and Bankruptcy Risk.”
Working paper, SSRN eLibrary.
Bolton, P. and Scharfstein, S., 1996, Optimal Debt Structure and the Number of Creditors."
Journal of Political Economy 104, 1-25.
Bris, A., I. Welch, and N. Zhu. 2006. “The Costs of Bankruptcy: Chapter 7 Liquidation Versus
Chapter 11 Reorganization.” The Journal of Finance 61 (3): 1253–1303.
Brockman, P., T. Ma, and X. Martin. 2012. “CEO Compensation Structure and Asymmetric
Timely Loss Recognition: An Empirical Analysis from Debt Contracting Perspective.”
Working paper, SSRN eLibrary.
Brown, D., 1989, Claimholder incentive conflicts in reorganization: The role of bankruptcy law,
Review of Financial Studies 2, 109-123.
Capkun, V., and L. Weiss. 2008. “Bankruptcy Resolution and the Restoration of Priority of
Claims.” American Law & Economics Association Annual Meetings. American Law &
Economics Association 18th Annual Meeting. Working Paper 43.
Carrizosa, R., and S. Ryan. 2013. “Conservatism, Covenants, and Recovery Rates.” Working
paper, SSRN eLibrary.
Chava, S., and M. Roberts. 2008. “How Does Financing Impact Investment? The Role of Debt
Covenants.” The Journal of Finance 63 (5): 2085–2121.
Davydenko, S. 2012. “Insolvency, Illiquidity, and the Risk of Default.” Working Paper.
University of Toronto.
30
Franks, J., and W. Torous. 1994. “A Comparison of Financial Recontracting in Distressed
Exchanges and Chapter 11 Reorganizations.” Journal of Financial Economics 35 (3)
(June): 349–370.
Gilson, S. 1989. “Management Turnover and Financial Distress.” Journal of Financial
Economics 25 (2) (December): 241–262.
Gilson, S., K. John, and L. Lang. "Troubled Debt Restructuring: An Empirical Study of Private
Reorganization of Firms in Default." Journal of Financial Economics 27 (2): 315-353.
Givoly, D., and C. Hayn. 2000. “The Changing Time-series Properties of Earnings, Cash Flows
and Accruals: Has Financial Reporting Become More Conservative?” Journal of
Accounting and Economics 29 (3) (June): 287–320.
Guay, W., and R. Verrecchia. 2006. “Discussion of an Economic Framework for Conservative
Accounting and Bushman and Piotroski (2006).” Journal of Accounting and Economics
42 (1–2) (October): 149–165.
Harris, M. and Raviv, A., 1991. "The Theory of Capital Structure." Journal of Finance 46, 297355.
Hart, O. and Moore, J., 1998. "Default and Renegotiation: A Dynamic Model of Debt."
Quarterly Journal of Economics 113, 1-41.
Heckman, J. 1979. “Sample Selection Bias as a Specification Error.” Econometrica 47 (1): 153161.
Hotchkiss, E. 1995. “Postbankruptcy Performance and Management Turnover.” The Journal of
Finance 50 (1): 3–21.
Hotchkiss, E., K. John, K. Thorburn, and R. Mooradian. 2008. “Bankruptcy and the Resolution
of Financial Distress.” Handbook of Corporate Finance: Empirical Corporate Finance.
Vol. 2, Chapter 14.
Jensen, M. 1991. “Corporate control and the politics of finance.” Journal of Applied Corporate
Finance 4: 13–33
Kalay, A., R. Singhal, and E. Tashjian. 2007. “Is Chapter 11 Costly?” Journal of Financial
Economics 84 (3) (June): 772–796.
Lafond, R., and S. Roychowdhury. 2008. “Managerial Ownership and Accounting
Conservatism.” Journal of Accounting Research 46 (1): 101–135.
LaFond, R., and R. Watts. 2008. “The Information Role of Conservatism.” The Accounting
Review 83 (2) (March): 447–478.
Nikolaev, V. 2010. “Debt Covenants and Accounting Conservatism.” Journal of Accounting
Research 48 (1): 51–89.
Nini, G., D. Smith, and A. Sufi. 2009. “Creditor Control Rights and Firm Investment Policy.”
Journal of Financial Economics 92 (3) (June): 400–420.
Roberts, M., and A. Sufi. 2009a. “Control Rights and Capital Structure: An Empirical
Investigation.” The Journal of Finance 64 (4): 1657–1695.
Roberts, M., and A. Sufi. 2009b. “Renegotiation of Financial Contracts: Evidence from Private
Credit Agreements.” Journal of Financial Economics 93 (2) (August): 159–184.
Roberts, M., and A. Sufi. 2009c. “Financial Contracting: A Survey of Empirical Research and
Future Directions.” Annual Review of Financial Economics 1: 1-20.
Tan, L. 2013. “Creditor Control Rights, State of Nature Verification, and Financial Reporting
Conservatism.” Journal of Accounting and Economics 55 (1) (February): 1–22.
Thorburn, K. 2000. “Bankruptcy Auctions: Costs, Debt Recovery, and Firm Survival.” Journal
of Financial Economics 58 (3) (December): 337–368.
31
Townsend, R., 1979, Optimal Contracts and Competitive Markets with Costly State Verification,
Journal of Economic Theory 21, 265-293.
Watts, R. 2003a. “Conservatism in Accounting Part I: Explanations and Implications.”
Accounting Horizons 17 (3) (September): 207–221.
Watts, R. 2003b. “Conservatism in Accounting Part II: Evidence and Research Opportunities.”
Accounting Horizons 17 (4) (December): 287–301.
Weiss, L. 1990. “Bankruptcy Resolution: Direct Costs and Violation of Priority of Claims.”
Journal of Financial Economics 27 (2) (October): 285–314.
Wittenberg-Moerman, R. 2008. “The Role of Information Asymmetry and Financial Reporting
Quality in Debt Trading: Evidence from the Secondary Loan Market.” Journal of
Accounting and Economics 46 (2–3) (December): 240–260.
Zhang, J. 2008. “The Contracting Benefits of Accounting Conservatism to Lenders and
Borrowers.” Journal of Accounting and Economics 45 (1) (March): 27–54.
Zhang, Z. 2009. “Recovery Rates and Macroeconomic Conditions: The Role of Loan
Covenants.” Working paper, SSRN eLibrary.
32
Figure 1: Populations Examined
This figure illustrates the population of firms we wish to study. This population is shown as the darker inner circle. The lighter outer
circle represents the population of firms studied by Zhang (2008). Inferential issues are illustrated by fact that the dashed circle from
which we draw our data does not perfectly coincide with the population of firms unable to make current or future promised payments
on debt. It could contain firms forced by creditors into bankruptcy at the expense of shareholders (A) and could exclude firms unable
to make current or future payments, omitted from the URD database either because of Moody’s data collection procedures or because
these firms still maintain liquidity to avoid default (B).
Firms that experience a negative shock
Firms unable to make current and/or future
promised debt payments
B
Firms in Moody’s URD database
A
33
Figure 2: Sample Composition
This figure describes sample composition over the period of 1987-2011. Panel A shows the initial sample comes from Moody’s
Ultimate Recovery Rates dataset (URD) of 987 firms and UCLA Bankruptcy Research Database (BRD) of 886 firms. The intersection
of these two datasets holds 381 sample observations. Panel B shows that, after deleting all firms without financial data and
conservatism measures available for the empirical analysis, the final sample holds 896 firms; 557 firms are in the URD dataset, and
681 are in the BRD dataset, with an intersection of 342 firms.
Panel A: Initial full sample
Moody’s URD
N=
UCLA BRD
(N = 987)
381
(N=886)
B: Final sample with available financial data
Moody’s URD
(N = 557)
N=
342
UCLA BRD
(N=681)
N = 896
34
Figuree 3: Cumula
ative distribution of reccovery ratess for high an
nd low consservative firrms
This figure plots the cumulative distribution
d
fun
nction of recov
very rates of saample firms inn bankruptcy oover the periodd of 1987-2011
1,
identiified from Moo
ody’s URD database. The sam
mple is partitio
oned into Highh Conservatism
m and Low Connservatism firm
ms based on the
sample median of th
he composite measure
m
of acccounting conseervatism: All C
Conservatism. T
The compositee measure is crreated based on
n
the avverage quintilee rank of the ind
dividual conseervatism measu
ures: Cons_r2, Cons_Basu, C
Cons_BS, Skewnness, and Speccial Items Ratio
o.
The ddistribution reeflects the prob
bability that a realized reco
overy rate is lless than or eequal to a speecified threshoold within each
h
conseervatism group
ping. A differeence in the diistribution of recovery ratess for High Coonservatism annd Low Conseervatism firmss,
speciffically, a disco
ontinuity in thee high end (i.e.,, above 80%) of
o the creditor recovery rate ddistribution forr High Conservvatism firms, is
i
consistent with acco
ounting conserrvatism forcing
g healthy firms into bankrupttcy unnecessariily. The observved difference in the recovery
y
rates for High Consservatism and Low Conservaatism firms lies mainly in thee lower end off the distribution, reflecting tthat accounting
g
conseervatism does not
n force health
hy firms into unnecessary
u
ban
nkruptcy.
35
Table 1: Summary statistics
This table reports descriptive statistics for the total sample of bankrupt firms over the period of 1987-2011, identified from Moody’s
URD database and the UCLA BRD. The number of observations for each dependent variable and key control variables varies
according to its availability in these two data sources. We identify 896 bankrupt firms in these databases with sufficient data to
compute firm-specific measures of accounting conservatism as described in Figure 2. Creditor recovery rates (Recovery Rate) are only
available for firms identified in the Moody’s database. All other tests use the maximum number of observations available combining
both sources of bankruptcy data. All variables are defined in Appendix C.
Descriptive Statistics
N
Mean
Q1
Median
Q3
Std.
Dev.
Dependent Variables
Recovery Rate
Ex-Post Loss
Ex-Post Asset WD
Violate
Bankruptcy Lag
Int Coverage
CFO
Bankruptcy Period
Emerge
594
217
206
654
598
746
769
750
794
53.392
0.001
0.007
0.662
2.028
0.917
0.006
422.855
0.681
29.950
0.000
0.000
0.000
1.099
-0.310
-0.021
133.000
0.000
52.465
0.001
0.001
1.000
2.303
0.390
0.020
305.000
1.000
76.140
0.001
0.004
1.000
2.944
1.260
0.052
582.000
1.000
28.931
0.002
0.025
0.473
1.125
12.976
0.125
409.328
0.466
Conservatism Measures
Cons_r2
Cons_Basu
Cons_BS
Skewness
Special Items Ratio
All Conservatism
382
434
784
881
398
896
0.000
0.000
0.000
0.000
0.000
2.005
-0.017
-0.077
-0.037
-0.641
-0.200
1.500
0.021
-0.060
-0.037
-0.005
-0.141
2.000
0.052
-0.035
-0.037
0.735
0.002
2.500
1.000
1.000
1.000
1.000
1.000
0.836
Firms-specific Control
Variables
Size
Leverage
Prior ROA
Debt Concentration
Bank_Share
Secured Debt %
Debt Rating
896
896
896
594
594
594
889
6.379
0.805
-0.097
0.579
0.375
0.437
15.871
5.433
0.454
-0.085
0.352
0.098
0.130
14.000
6.185
0.699
-0.035
0.556
0.311
0.407
16.000
7.217
0.959
-0.012
1.000
0.590
0.692
17.000
1.529
0.724
0.244
0.347
0.318
0.344
3.260
36
Table 2: Correlation coefficients
This table reports Pearson and Spearman correlations between key dependent and independent variables used in the analysis. Pearson correlations are presented below the
diagonal, and Spearman correlations are presented above the diagonal. Correlation coefficients significant at the .10 level or lower are bolded. All variables are defined in
Appendix C.
Recovery Ex-Post Ex-Post
Bankruptcy
Int
Rate
Asset WD
Loss
Violate
Lag
Coverage
Recovery Rate
0.023
0.025
-0.340
-0.216
0.204
Ex-Post Asset WD
0.062
-0.073
-0.160
0.412
-0.172
Ex-Post Loss
0.008
-0.189
0.148
-0.173
-0.148
Violate
0.022
-0.054
-0.172
-0.089
0.086
Bankruptcy Lag
0.004
0.044
0.031 -0.0934
0.001
Int Coverage
-0.018
-0.089 -0.0040
-0.013
0.114
CFO
-0.065
-0.028
0.120
-0.221
0.1231
0.419
Days in
-0.037
0.0324
0.057
-0.179
-0.135
-0.084
Emerge
-0.050
-0.088
0.0404
0.014
0.249
-0.099
Cons_r2
-0.043
0.017 -0.0390
0.028
0.009
0.118
Cons_Basu
0.056
-0.012
-0.048
0.0532
-0.049
0.398
Cons_BS
0.010
-0.033
-0.009
0.0449
-0.039
-0.004
Skewness
0.054
-0.081
0.0531
0.057
-0.187
-0.144
Special Items Ratio
0.057
-0.065
-0.039
0.128 -0.0960
0.121
All Conservatism
0.049
0.077
-0.167
-0.199
0.1021
-0.136
Size
-0.078 -0.1956
-0.055
0.130
-0.240
0.083
Leverage
0.049
0.046
0.0201
0.043
-0.119
-0.084
Prior ROA
-0.101
0.048 -0.0032
0.062
0.002
0.122
Debt
0.075
-0.004 -0.0134
0.037
-0.104
-0.083
Bank_Share
-0.078
0.044
0.054
0.143
-0.116
0.2939
Secured Debt %
0.070
-0.062
-0.006
0.215
0.1804
0.111
Debt Rating
0.003
-0.004
-0.039
0.0062
-0.038
-0.092
Days in
Special Items
All
CFO
Bankruptcy Emerge Cons_r2 Cons_Basu Cons_BS Skewness
Ratio
Conservatism
0.098
-0.020
0.064
0.083
0.219
-0.261
0.247
0.132
0.072
-0.091
-0.080
-0.016
-0.066
-0.013
-0.079
-0.129
-0.131
-0.256
-0.072
-0.083
-0.051
0.007
-0.004
-0.149
-0.133
-0.123
-0.123
0.053
0.040
-0.004
0.047
0.056
0.034
0.129
0.137
0.094
0.002
-0.038
0.046
0.021
-0.099
-0.158
-0.085
-0.138
-0.134
0.037
0.057
-0.013
0.034
0.884
0.091
0.071
-0.086
0.121
0.037
0.027
0.023
-0.060
-0.007
0.052
0.114
0.141
0.042
-0.073
0.003
0.035
-0.007
-0.077
-0.033
-0.086
-0.076
0.004
0.014
0.040
0.066
-0.075
0.093
0.059
-0.015
0.021
-0.026
-0.041
0.019
0.069
0.130
0.525
0.042
-0.084
0.009
0.007
0.012
0.079
-0.085
0.531
0.021
-0.007
0.038
-0.043
-0.022
0.051
0.007
0.495
-0.006
-0.028
0.030
0.010
0.130
0.084
0.179
0.667
0.064
0.022
0.046
-0.022
0.037
-0.008
0.122
0.580
-0.043
0.068
0.059
0.096
0.247
0.111
0.636
0.197
-0.002
0.025
-0.050
0.003
-0.022
-0.026
0.180
0.177
0.070
0.007
-0.031
0.011
0.005
0.049
-0.242
-0.129
0.071
0.086
0.042
0.046
-0.008
0.040
0.003
0.120
-0.138
-0.205
-0.110
0.025
0.069
-0.036
0.013
-0.014
-0.130
-0.213
-0.090
0.128
0.035
0.029
-0.090
-0.055
0.063
-0.003
0.017
0.128
-0.098
-0.001
-0.031
-0.093
-0.068
0.002
-0.042
-0.036
0.074
-0.079
-0.005
-0.004
-0.007
-0.003
-0.123
-0.243
0.140
0.057
0.101
Size
Leverage
0.119
-0.130
-0.333
0.194
0.077
-0.135
-0.164
0.094
-0.029
0.035
0.156
-0.169
-0.043
0.085
0.160
-0.269
0.065
0.184
-0.076
0.007
0.045
-0.034
0.057
-0.122
-0.002
0.069
0.011
0.234
-0.039
0.115
-0.281
-0.308
0.184
-0.29
-0.579
0.09
0.090
-0.11
0.009
-0.07
-0.175
0.22
Prior
Debt
ROA
Concentration
0.252
-0.099
-0.199
0.205
-0.009
0.096
0.011
-0.065
0.026
0.011
0.109
-0.101
0.122
-0.083
0.001
-0.184
0.084
-0.110
-0.008
-0.115
-0.105
0.091
-0.027
0.082
-0.049
-0.099
-0.023
-0.187
-0.028
-0.106
0.198
-0.658
0.067
-0.247
-0.109
-0.080
-0.011
-0.073
0.006
-0.019
0.013
-0.102
Bank
Secured
Debt
Share
Debt %
Rating
0.018
0.172
0.223
-0.139
-0.045
-0.048
-0.053
-0.032
-0.124
-0.016
0.326
0.201
0.049
-0.032
0.114
0.068
0.192
-0.181
0.061
0.156
-0.128
0.043
0.073
-0.227
-0.083
-0.082
0.155
0.037
-0.060
-0.012
0.029
0.072
0.029
0.042
-0.076
-0.092
0.025
0.046
0.082
0.022
-0.033
0.137
0.023
-0.027
0.100
0.056
0.163
-0.145
-0.038
-0.104
0.322
0.007
-0.004
-0.178
-0.058
0.008
-0.148
0.634
-0.157
-0.04823
0.620
-0.044
-0.143
37
Table 3: The relation between accounting conservatism and creditor recovery rates
This table reports the results of testing the relation between accounting conservatism and creditor recovery rates of firms in
bankruptcy using all firms with available data in the Moody URD over the period from of 1987-2011. Panel A presents the results of
ordinary least squares estimation. Panel B provides the results of the second stage from Heckman two-stage estimation. The first stage
estimates the probability that a distressed firm will be included in our bankruptcy sample, and the first-stage sample firms include
bankruptcy firms identified from Moody’s URD and the UCLA BRD and firms in the intersection of Compustat and CRSP with
annual stock returns in the bottom ten percentile. Each test uses the maximum number of observations available for each conservatism
measure. Five individual accounting conservatism measures and a composite measure, All Conservatism, based on the average quintile
rank of the individual measures are used in the test. With the exception of Special Items Ratio, firm-specific conservatism measures
are calculated over the ten years before the default date, while the Special Items Ratio measure is estimated over the three years before
default. Fama-French 12 industry and year fixed effects are included and standard errors are clustered by year. All variables are
defined in Appendix C.
Recoveryi = α0 + β1Conservatismi + β2Sizeit + β3Leverageit-1 + β4Prior ROAit + β5Distressed Exchangei + β6Debt
Concentrationi + β7Bank Sharei + β8Secure Debt%i + β9Bankruptcy Periodi + β10Distressed Exchangei
+ βkMacroeconomic Controls + εit
38
Table 3 (continued)
Panel A: OLS estimation
Dependent Variable = Recovery Rate
Accounting Conservatism Measures
Intercept
Conservatism
Firm specific controls
Size
Leverage
Prior ROA
Distressed Exchange
Debt Concentration
Bank_Share
Secured Debt %
Bankruptcy Period
Macroeconomic Controls
S&P 500
Spread
Spec_rate
GDP Growth
Lag_S&P 500
Lag_Spread
Lag_Spec
Lag_GDP
Number Obs
R-Square
Cons_r2
(t-stat)
14.54
(0.32)
3.541*
(1.75)
Cons_Basu
(t-stat)
45.89
(1.61)
2.848***
(3.62)
Cons_BS
(t-stat)
11.98
(0.53)
1.044***
(3.49)
Skewness
(t-stat)
5.337
(0.24)
2.138*
(1.77)
Special Items
Ratio
(t-stat)
56.01
(1.14)
3.959
(1.33)
All Conservatism
(t-stat)
-6.641
(-0.30)
3.932**
(2.68)
3.656
(1.65)
0.624
(0.39)
1.290
(0.08)
19.58***
(3.35)
5.504
(0.83)
11.25
(1.46)
15.92**
(2.47)
-0.00949
(-1.27)
7.779
(0.36)
7.563
(1.31)
-571.0
(-0.49)
152.4
(0.57)
-11.79
(-0.81)
-5.658
(-0.47)
173.4*
(1.74)
265.7
(1.34)
1.987
(1.54)
4.438**
(2.13)
16.13***
(3.27)
35.35***
(5.05)
11.17*
(1.82)
19.39*
(1.90)
6.762
(0.89)
-0.00821
(-0.97)
-11.06
(-0.53)
6.150
(1.00)
-1123.7**
(-2.30)
330.1*
(1.75)
-10.54
(-0.84)
13.00
(1.34)
110.2
(1.10)
4.440
(0.02)
2.463**
(2.58)
-0.196
(-0.11)
4.224
(0.81)
22.27***
(6.84)
3.924
(1.15)
5.031
(0.91)
14.87***
(4.06)
-0.0108**
(-2.63)
7.652
(0.51)
1.738
(0.30)
943.9*
(1.73)
223.2
(1.13)
-18.06**
(-2.53)
-1.428
(-0.32)
-39.28
(-0.81)
-70.64
(-0.68)
2.672***
(3.23)
-0.510
(-0.29)
5.627
(1.23)
4.686
(1.46)
22.58***
(6.90)
4.136
(0.94)
15.68***
(4.92)
-0.00987**
(-2.42)
8.319
(0.62)
0.806
(0.15)
974.2*
(1.90)
160.0
(0.98)
-18.27**
(-2.47)
0.615
(0.13)
-33.23
(-0.92)
-23.04
(-0.24)
2.779**
(2.38)
-3.114
(-0.88)
8.911
(1.64)
4.588
(0.97)
28.35***
(4.36)
10.45
(1.17)
8.198
(0.86)
-0.00884
(-1.49)
5.307
(0.29)
10.98**
(2.56)
-1726.7
(-1.64)
133.2
(0.66)
-10.87
(-0.85)
22.33**
(2.80)
-3.838
(-0.07)
-38.84
(-0.26)
2.738***
(3.15)
-0.350
(-0.19)
6.708
(1.49)
23.29***
(7.59)
4.461
(1.33)
3.716
(0.74)
15.12***
(4.07)
-0.00970**
(-2.50)
11.44
(0.83)
2.270
(0.41)
1059.0*
(1.97)
187.4
(1.15)
-19.03**
(-2.70)
-1.433
(-0.30)
-23.05
(-0.55)
-51.55
(-0.48)
215
0.436
233
0.472
533
0.356
594
0.360
251
0.495
557
0.363
***, **, * indicate statistical significance at the .01, .05 level, and .10 level, respectively.
39
Table 3 (continued)
Panel B: Heckman two-stage estimation
Dependent Variable = Recovery Rate
Accounting Conservatism Measures
Intercept
Conservatism
Firm specific controls
Size
Leverage
Prior ROA
Distressed Exchange
Debt Concentration
Bank_Share
Secured Debt %
Bankruptcy Period
Macroeconomic Controls
S&P 500
Spread
Spec_rate
GDP Growth
Lag_S&P 500
Lag_Spread
Lag_Spec
Lag_GDP
Lambda
Number Obs
Cons_r2
(t-stat)
2.470
(0.02)
3.722**
(2.34)
Cons_Basu
(t-stat)
33.63
(0.70)
2.952*
(1.70)
Cons_BS
(t-stat)
10.05
(0.23)
1.104
(1.06)
Skewness
(t-stat)
3.215
(0.07)
1.960*
(1.86)
Special Items
Ratio
(t-stat)
59.59
(0.73)
3.811**
(2.46)
All Conservatism
(t-stat)
-10.57
(-0.24)
4.250***
(3.13)
4.959***
(3.48)
3.702
(1.49)
6.781
(0.60)
18.57***
(3.23)
4.954
(1.01)
6.272
(0.87)
17.59***
(2.97)
-0.0120***
(-2.58)
3.006*
(1.74)
6.672***
(2.66)
15.18**
(2.39)
34.46***
(5.87)
11.18**
(2.20)
16.28**
(2.09)
8.413
(1.18)
-0.0109**
(-2.11)
2.809***
(2.81)
0.807
(0.45)
4.316
(1.10)
22.25***
(6.50)
3.671
(1.15)
4.328
(0.98)
14.96***
(3.78)
-0.0116***
(-3.71)
2.999***
(3.22)
0.518
(0.30)
5.676
(1.50)
22.53***
(7.03)
4.487
(1.51)
3.468
(0.84)
15.75***
(4.23)
-0.0107***
(-3.59)
2.986*
(1.87)
-2.603
(-0.52)
8.511*
(1.73)
28.55***
(5.62)
4.656
(0.98)
10.34
(1.51)
8.141
(1.31)
-0.00914**
(-2.23)
3.203***
(3.38)
1.092
(0.63)
6.816*
(1.79)
22.96***
(6.89)
4.075
(1.32)
2.541
(0.59)
15.41***
(4.01)
-0.0108***
(-3.59)
3.299
(0.15)
5.082
(0.76)
-323.1
(-0.12)
153.4
(0.53)
-10.41
(-0.84)
-12.25
(-1.27)
175.3**
(2.05)
241.5
(1.56)
8.987***
(2.71)
-16.87
(-0.76)
4.056
(0.54)
-1032.8
(-1.15)
338.2
(1.08)
-7.134
(-0.48)
10.97
(0.99)
116.6
(1.10)
26.08
(0.14)
6.334*
(1.72)
6.265
(0.49)
1.165
(0.26)
865.4
(0.84)
238.5
(1.36)
-17.92**
(-2.10)
-1.999
(-0.34)
-35.18
(-0.60)
-61.76
(-0.56)
2.071
(0.92)
6.981
(0.58)
0.211
(0.05)
903.4
(0.89)
175.0
(1.06)
-18.11**
(-2.29)
0.0833
(0.02)
-28.88
(-0.51)
-14.79
(-0.14)
2.045
(0.99)
4.683
(0.23)
11.18*
(1.66)
-1951.7
(-0.85)
157.0
(0.60)
-10.26
(-0.84)
22.61**
(2.49)
10.06
(0.12)
-19.21
(-0.13)
0.537
(0.13)
9.616
(0.76)
1.400
(0.31)
915.6
(0.90)
202.5
(1.18)
-18.21**
(-2.22)
-1.740
(-0.31)
-18.80
(-0.33)
-37.00
(-0.35)
2.919
(1.36)
215
231
529
590
249
553
***, **, * indicate statistical significance at the .01, .05 level, and .10 level, respectively.
40
Table 4: The relation between accounting conservatism and bankruptcy resolution
This table reports the results of testing the relation between accounting conservatism and the length of bankruptcy resolution. We
estimate a Cox proportional hazard model, where hi (T) is the instantaneous risk of a bankrupt firm emerging from bankruptcy at time
T giving the firm survives to time T. All firms that do not emerge from bankruptcy are censored. Column (1) uses all sample firms in
bankruptcy with available data identified from Moody’s URD and the UCLA BRD over the period of 1987-2011 and column (2)
excludes bankrupt firms existing through an acquisition or liquidation. Models include Fama-French 12 industry and year fixed
effects. All variables are defined in Appendix C.
ln hi (T) =
h0(T) + γ1Conservatismi + γ2Sizeit + γ3Leverageit-1 + γ4Prior ROAit + γ5Litigationi + γjDefault
Macroeconomic Controls+ γkEmerge Macroeconomic Controls + εit
[1]
All Conservatism
p-value
Hazard Ratio
Firm specific controls
Size
p-value
Hazard Ratio
Leverage
p-value
Hazard Ratio
Prior ROA
p-value
Hazard Ratio
Litigation
p-value
Hazard Ratio
Macroeconomic Controls
S&P 500
p-value
Hazard Ratio
Spread
p-value
Hazard Ratio
Spec_rate
p-value
Hazard Ratio
GDP Growth
p-value
Hazard Ratio
Emerge S&P 500
p-value
Hazard Ratio
Emerge Spread
p-value
Hazard Ratio
Emerge Spec
p-value
Hazard Ratio
Emerge GDP
p-value
Hazard Ratio
Number Obs
[2]
0.182***
0.00
1.20
0.157***
0.01
1.17
-0.041
0.30
0.96
0.201***
<.0001
1.22
0.523**
0.04
1.69
-0.018
0.93
0.98
-0.209***
<.0001
0.81
0.200***
0.00
1.22
0.348
0.15
1.42
0.114
0.60
1.12
0.22
0.61
1.25
-0.099
0.63
0.91
-15.439***
<.0001
0.00
-16.98***
0.00
0.00
-0.15
0.80
0.86
-0.143
0.59
0.87
50.531**
0.04
8.82E+21
2.785
0.74
16.20
0.311
0.45
1.37
-0.218
0.30
0.80
-18.807***
<.0001
0.00
-19.255***
0.00
0.00
-0.316
0.61
0.73
-0.229
0.37
0.80
50.127**
0.04
5.89E+21
14.67*
0.09
2.34E+4
750
498
***, **, * indicate statistical significance at the .01, .05 level, and .10 level, respectively.
41
Table 5: The relation between accounting conservatism and covenant violation
This table reports the results of testing of the relation between accounting conservatism and the probability of covenant violation in the
year before default using a probit model. This test uses all sample firms in bankruptcy identified from Moody’s URD and the UCLA
BRD over the period of 1994-2011. Covenant violations are identified by manually searching public disclosures in sample firms’
Form 10-K, 10-Q, and 8-K. Fama-French 12 industry and year fixed effects are included and standard errors are clustered by year. All
variables are defined in Appendix C.
Prob(Violatei)
Dependent Variable
Intercept
= α0 + φ1Conservatismi + φ2Sizeit + φ3Leverageit-1 + φ4Prior ROAit + φ5Litigationi + φ6Interest
Missi + φkMacroeconomic Controls + εit
Violate
1.261
(0.237)
All Conservatism
0.214***
(0.000)
Firm specific controls
Size
-0.115**
(0.011)
Leverage
-0.0227
(0.787)
Prior ROA
0.181
(0.398)
Litigation
-0.365*
(0.090)
Interest Miss
0.000772
(0.599)
0.251
Macroeconomic Controls
S&P 500
(0.726)
0.0378
Spread
(0.877)
Spec_rate
(0.280)
-36.75
7.027
GDP Growth
(0.466)
Number Obs
654
Pseudo R-Square
0.102
***, **, * indicate statistical significance at the .01, .05 level, and .10 level, respectively.
42
Table 6: The relation between accounting conservatism and bankruptcy filing lag
This table reports the results of testing of the relation between accounting conservatism and the timing of bankruptcy filing relative to
economic news reflected in stock returns. We measure monthly abnormal stock returns using CRSP stock returns based on the
estimated CAPM beta over the 60 months ending in the preceding calendar year. We identify the month with the lowest monthly
abnormal stock returns in the three fiscal years before default (i.e., periods t-3 through t-1) as the month in which the bad news driving
firms into bankruptcy arrives. The dependent variable in this model, Bankruptcy Lag, is the natural log of the number of months
between the date of default and the month with the lowest monthly abnormal stock returns in the three fiscal years before default.
This test uses all sample firms in bankruptcy identified from Moody’s URD and the UCLA BRD over the period of 1987-2011. Model
includes industry and year fixed effects, with standard errors clustered by year. All variables are defined in Appendix C.
Bankruptcy Lagi = α0 + δ1Conservatismi + δ2Sizeit + δ3Leverageit-1 + δ4Prior ROAit + δ5Litigationi +
δkMacroeconomic Controls + εit
Dependent Variable
Bankruptcy
Lag
Intercept
4.362***
All Conservatism
-0.200**
(5.96)
(-2.79)
Firm specific controls
Size
-0.0934**
(-2.08)
Leverage
0.107
(1.31)
Prior ROA
0.333
(1.33)
Litigation
0.172
(1.11)
Macroeconomic Controls
S&P 500
-0.795*
(-1.91)
Spread
0.0224
(0.12)
Spec_rate
-34.36***
(-5.43)
GDP Growth
10.04
(1.27)
Number Obs
R-Square
598
0.192
***, **, * indicate statistical significance at the .01, .05 level, and .10 level, respectively.
43
Table 7: The relation between accounting conservatism and ex-ante performance
Panel A reports the results of testing the relation between accounting conservatism and cash flow-interest coverage in the three years
before bankruptcy. In column 1, the dependent variable, Int Coverage, is the ratio of annual operating cash flows to total interest
expense, averaged over the three years before bankruptcy filing to determine the association between conservatism and liquidity and
solvency. In column 2, the dependent variable is an indicator variable equal to 1if the firm’s ratio of annual operating cash flows to
total interest expense over the three years before bankruptcy is greater than one and 0 otherwise; a probit model is estimated to
determine whether the firm has sufficient cash flows to cover average annual interest expense before default. In column 3, the
dependent variable is the ratio of annual operating cash flows to total assets, averaged over the three years before bankruptcy to
determine the association between conservatism and asset productivity. These tests use all sample firms in bankruptcy identified from
Moody’s URD and the UCLA BRD over the period of 1987-2011. Model includes industry and year fixed effects, with standard errors
clustered by year. All variables are defined in Appendix C.
Performit-1,it-3 = α0 + ω1Conservatismi + ω2Sizeit + ω3Avg. Leverageit-1, it-3 + ω4Avg. ROAit-1, it-3 + ω5Avg. Capexit-1, it-3
+ ω6 Avg. Dividendsit-1, it-3 + ω7Litigationi +ωkMacroeconomic Controls + εit
Dependent Variable
Intercept
Conservatism
Firm specific controls
Size
Avg. Leverage
Avg. ROA
Avg Capex
Avg. Dividend
Litigation
Macroeconomic Controls
S&P 500
Spread
Spec_rate
GDP Growth
Number Obs
R-Square / Pseudo R-Square
Int Coverage
-2.444
(-0.46)
1.598*
(2.07)
CFO
-0.0791
(-1.35)
0.0236***
(3.30)
0.298
(1.03)
12.58
(1.63)
-1.040
(-0.67)
4.258
(1.16)
2.198
(0.97)
-2.548
(-0.88)
0.00841**
(2.54)
0.279**
(2.73)
0.00737
(0.41)
0.0666**
(2.38)
0.0369
(0.64)
-0.00532
(-0.17)
8.288
(1.25)
-1.416
(-0.74)
17.51
(0.80)
-61.24
(-0.86)
747
0.113
0.00331
(0.09)
-0.00866
(-0.31)
-2.212***
(-5.29)
0.280
(0.44)
769
0.334
***, **, * indicate statistical significance at the .01, .05 level, and .10 level, respectively.
44
Table 8: The relation between accounting conservatism and the likelihood of emergence
This table reports the results of testing the relation between accounting conservatism and the likelihood of emerging from bankruptcy.
We estimate a probit model, where the dependent variable is a dummy variable, coded as 1 if a firm emerges from bankruptcy and 0
otherwise. This test uses all sample firms in bankruptcy with available data identified from Moody’s URD and the UCLA BRD over
the period 1987-2011. Model includes industry and year fixed effects, with standard errors clustered by year. All variables defined in
Appendix C.
Prob(Emergei) = = α0 +
1Conservatismi
+
Macroeconomic Controls+
Dependent Variable
Intercept
All Conservatism
Firm specific controls
Size
Leverage
Prior ROA
Litigation
Macroeconomic Controls
S&P 500
Spread
Spec_rate
GDP Growth
Emerge S&P 500
Emerge Spread
Emerge Spec
Emerge GDP
Number Obs
Pseudo R-Square
2Sizeit +
kEmerge
3Leverageit-1 +
4Prior
ROAit +
5Litigationi +
kDefault
Macroeconomic Controls + εit
Emerge
-8.048***
(0.000)
0.106*
(0.087)
0.145***
(0.000)
0.229*
(0.061)
0.448**
(0.011)
-0.229
(0.102)
1.144**
(0.042)
0.169
(0.475)
-2.512
(0.478)
-7.129
(0.136)
-0.0496
(0.926)
-0.0742
(0.745)
341.5***
(0.000)
-2.236
(0.858)
794
0.123
***, **, * indicate statistical significance at the .01, .05 level, and .10 level, respectively.
45
Table 9: The relation between accounting conservatism and write-downs following default
This table reports the results of testing the relation between accounting conservatism and the magnitude of ex post write-downs
following default using all firms with available data in the Moody URD over the period from of 1987-2011. In column 1, the
dependent variable, Ex-Post Loss, is measured as (-1) multiplied by of the firm’s net income scaled by total assets after default; we
scale ex post return on assets by the number of days over the income measurement period because each monthly operating report filed
with the bankruptcy court is not standardized with regard to the number of days covered in each report. In column 2, the dependent
variable, Ex-Post Asset WD, is measured as (-1) multiplied by the percentage change in the firm’s total assets from the last balance
sheet before default to the first balance sheet after default. We scale total write-downs by the number of days between the default date
and the ex post balance sheet date to standardize this measurement period, because firms with a greater lag between the default date
and the balance sheet date are more likely to record greater asset write-downs. Financial information following bankruptcy is obtained
from the monthly operating reports filed with the bankruptcy court, manually collected from 8-K filings on Morningstar’s 10-K
Wizard. If monthly operating reports are not available, we obtain data from the first quarterly financial report following default
available on Compustat. All variables are defined Appendix C.
Writedowni = α0 + ψ1Conservatismi + ψ2Returnsit-1 + ψ3Sizeit + ψ4Leverageit-1 + ψ5Debt
Concentrationi + ψ6Bank Sharei + ψ7Secure Debt%i + ψMacro Controls + εit
Dependent Variable
Intercept
All Conservatism
Stock Returns t-1
Size
Leverage
Debt Concentration
Bank_Share
Secured Debt %
S&P 500
Spread
Spec_rate
GDP Growth
Number Obs
R-Square
Ex-Post Loss
0.00329***
(3.52)
-0.000254*
(-1.83)
0.000348
(1.15)
-0.000344***
(-4.03)
-0.0000651
(-0.29)
0.000109
(0.33)
-0.00145***
(-3.24)
0.00131***
(3.15)
-0.000251
(-0.33)
-0.0000114
(-0.03)
0.00890*
(1.96)
0.000465
(0.05)
Ex-Post Asset
WD
0.0398**
(2.53)
-0.00471**
(-2.02)
0.00388
(0.75)
-0.00111
(-0.76)
0.00207
(0.55)
-0.00366
(-0.67)
-0.00309
(-0.41)
-0.00304
(-0.43)
-0.00700
(-0.53)
-0.00543
(-0.94)
-0.0501
(-0.66)
-0.0675
(-0.45)
217
0.167
206
0.055
***, **, * indicate statistical significance at the .01, .05 level, and .10 level, respectively.
46
Appendix A: Measures of Accounting Conservatism
This appendix describes the five individual measures of accounting conservatism in detail. Our
first measure of conservatism, Cons_Basu, uses the firm-specific Basu (1997) measure.
Specifically, for each firm in our sample, we estimate the sensitivity of earnings to bad news
relative to good news. The ratio of the two sensitivities then serves as the measure. We employ
the following regression:
α0 + α1DRit + β0Rit + β1Rit*DRit + εit
Earnit =
where Earnit equals income before extraordinary items for firm i in period t, scaled by the firm’s
market value of equity at the end of the prior period; Rit equals annual returns of firm i, measured
over the period ending three months after the firm’s fiscal year-end; and DRit is an indicator
variable equal to 1 if the firm’s returns are negative (Rit < 0) and 0 otherwise. Thus the
coefficient β0 estimates the sensitivity of accounting earnings to good news, while the sum of
coefficients (β0 + β1) estimates the sensitivity of accounting earnings to bad news. Our measure
captures the sensitivity of earnings to bad news relative to the sensitivity to good news:
Cons_Basu = (β0 + β1) / β0, where higher values of Cons_Basu represent more conservatism.28
Our second measure of conservatism, Cons_r2, is the relative explanatory power of bad news in
earnings versus the explanatory power of good news in earnings following Zhang (2008) and
Basu (1997): Cons_r2 = R2(bad news) / R2 (good news). Positive returns represent good news,
and negative returns represent bad news. R2 measures the R-squared from an earnings-returns
regression for the good and bad news subsamples. Thus higher values of Cons_r2 represent more
conservative firms.
We employ additional two measures of conservatism that do not rely on stock returns.
Following Ball and Shivakumar (2005), our third measure of conservatism, Cons_BS, estimates
the extent to which firms record bad news in earnings through write-offs and losses in accruals.
Ball and Shivakumar (2005) modify the Basu (1997) framework by regressing accruals on
negative operating cash flows to determine the extent to which firms record losses. For each firm
in our sample, we estimate the following regression model:
Accrualsit =
α0 + α1DCFOit + β0CFOit + β1CFOit*DCFOit + εit
where all variables are scaled by total assets at the beginning of the period.29 DCFO is an
indicator variable equal to 1 if the firm has negative annual operating cash flows in period t and
zero otherwise. Ball and Shivakumar (2005) report that accrued losses are more likely to occur in
periods of negative cash flows. Therefore our measure of conservatism, Cons_BS, is equal to the
β1, where higher values of Cons_BS represent more conservative firms. Our fourth measure of
conservatism, Skewness, captures the difference between the skewness of operating cash flows
and earnings, following Beatty et al. (2008). We require at least three annual observations to
compute Skewness and estimate this measure over a maximum of ten years before default. When
bad news is recognized in earnings more quickly than good news, earnings will be negatively
skewed relative to the firms’ cash flows. Thus higher levels of Skewness represent more
61 firms in Moody’s sample have a negative β0. The tabulated results are based on unadjusted β0. However, our
results hold when these observations are deleted from the sample.
29
We measure accruals using the balance sheet approach as ∆Current Assets - ∆Current Liabilities - ∆Cash and
Cash Equivalents + ∆Current Debt – Depreciation. We measure operating cash flows using the figures reported in
the statement of cash flows, where available (OANCF). Otherwise, we measure operating cash flows as Total Funds
from Operations - ∆Current Assets - ∆Debt in Current Liabilities - ∆Current Liabilities + ∆Cash and Cash
Equivalents. All of our results are qualitatively similar when we measure total accruals as net income minus
operating cash flows reported in the statement of cash flows.
47
28
conservative firms.
Our final individual measure of conservatism, Special Items Ratio, attempts to measure the
recognition of large declines in operating performance before the bankruptcy filing.
Conservative firms are more likely to record losses and asset write-downs as special items before
bankruptcy. Therefore we compute the average special items recorded in the income statement
(SPI) scaled by the average total assets over the three years before default as a proxy for these
recorded losses.30 We then compute Special Items Ratio as the ratio of special items to
cumulative firm stock returns on CRSP in the three years before default. We require negative
stock returns over this period to ensure that higher values of Special Items Ratio represent more
conservatism.
30
All results continue to hold if we compute the sum of special items recorded in the income statement scaled by the
average total assets over the three years before default as a proxy for the recorded losses.
48
Appendix B: Discussion of control variables in the test of the relation between accounting
conservatism and creditor recovery rates
Larger firms with more assets in the period of default can sell assets to improve liquidity; thus we
include Size equal to the natural log of the firm’s total assets in the quarter of default as a control
variable and expect a positive relation between firm size and creditor recovery rates. Leverage plays
an important role as a monitoring mechanism for creditors and other stakeholders of the firm
(Zhang, 2009). Highly levered firms also have fewer assets relative to the debt held by creditors and
thus have less ability to generate cash flows from assets to settle with creditors in bankruptcy. We
measure Leverage (debt divided by total assets) in the quarter before default and expect a negative
relation between firm leverage and creditor recovery rates. We also control for ex ante profitability,
and expect a positive relation between Prior ROA and recovery rates.
In addition, Zhang (2009) demonstrates that the nature and concentration of the firms’
creditors have a significant impact on the recovery rates of creditors in bankruptcy. Successful
resolution of financial distress through Chapter 11 requires the firm to develop a reorganization
plan that specifies what each class of claimants will receive in exchange for their pre-bankruptcy
claims (Hotchkiss et al., 2008). The bankruptcy courts require approval by at least one-half in
number and two-thirds in value of the total creditors in each impaired class. Therefore coordination
among creditor classes and between creditors and the firm plays a critical role in the affirmation of
the reorganization plan and bankruptcy resolution. We measure Debt Concentration as a proxy for
coordination among parties using the Herfindahl-Hirschman (HH) index of the face value of the
firm’s debt across different lenders. Following Zhang (2009), we measure Debt Concentration as:
∑
∑
where Dij is the face value of the j-th loan of firm i at the date of default. The HH index captures the
concentration of the firm’s debt among all creditors. It is equal to one when all of the firm’s debt is
maintained by a single loan and approaches zero as the number of lenders holding financial
instruments with similar face values increases. We anticipate that better coordination among
creditors increases recovery rates, and thus we expect a positive relation between Debt
Concentration and family recovery rates. Prior literature also indicates that banks monitoring and
screening actions can increase creditor recovery.
Prior literature suggests that secured creditors experience significantly higher recovery rates
than other creditor classes in bankruptcy due to their ability to exert control over the bankruptcy
proceedings (Capkun and Weiss, 2008). Furthermore, secured creditors may have less demand for
accounting conservatism, especially if the value of the assets securing their claim is significantly
higher than the total value of the loan. Therefore we include the total percentage of debt held by
secured creditors (Secure %) at the date of default in estimating model (1). Additionally, the time
spent in default and the speed of bankruptcy resolution has a significant effect on creditor recovery
rates (Acharya, Bharath, and Srinivasan, 2004). We therefore also control for the length of time
between the date of default and the emergence date (Bankruptcy Period). Finally, Zhang (2009)
indicates that factors such as improved business conditions (i.e., macroeconomic factors) lead to
higher creditor recovery rates. Our tests include macroeconomic variables measured at the default
(period t) and measured at the loan inception (period τ) to control for this effect.31
31
Zhang (2009) provides empirical evidence that creditor recovery rates are negatively associated with lagged
macroeconomic control variables, measured at the date of the loan origination. We estimate lagged macroeconomic
control variables for each firm at the inception date of the firm’s largest loan (in dollars).
49
Appendix C: Variable definitions
This appendix provides variable definitions for all variables used in all tests.
Variable
Definition
Avg. Leverage
Average quartile ranking of the Cons_r2, Cons_Basu, Cons_BS, Skewness, and Special Items Ratio
measures of conservatism.
Average capital expenditures scaled by total assets, measured over the three year period prior to
default. If capital expenditures are missing on Compustat, we code Avg. Capex as zero.
Average total dividends scaled by total assets, measured over the three year period prior to default. If
dividends are missing on Compustat, we code Avg. Dividend as zero.
Average leverage measured over the three year period prior to default.
Avg. ROA
Average return on assets measured over the three year period prior to default.
Bank_Share
Percentage of the firm’s debt held by banks at the time of default.
Bankruptcy Lag
The natural log of the number of months between the bankruptcy filing date and the month with the
lowest abnormal stock returns over the period in the three fiscal years prior to the bankruptcy filing
date.
Bankruptcy Period
Total time spent in financial distress, measured as the number of months from the date of obligor
default to the date of emergence as indicated in the Moody’s Ultimate Recovery Database (URD) and
UCLA Bankruptcy Research Database (BRD).
All Conservatism
Avg Capex
Avg. Dividend
Big N
CFO
Cons_Basu
Indicator variable equal to 1 if the firm has a Big N auditor, 0 otherwise
Ratio of annual operating cash flows to total assets, averaged over the three year period prior to
bankruptcy filing.
Sensitivity of earnings to bad news relative to good news using the following regression:
Earnit = α0 + α1DRit + β0Rit + β1Rit*DRit + ε, where Earn equals income before extraordinary
items (ib) for firm i in period t, scaled by the firm’s market value of equity at the end of the prior
period; Rit equals annual returns of firm i, measured over the period ending three months after the
firm’s fiscal year-end; and DRit is an indicator variable equal to 1 if the firm’s returns are negative
(Rit < 0), and 0 otherwise. Thus, the coefficient β0 estimates the sensitivity of accounting earnings to
good news, while the sum of coefficients (β0 + β1) estimates the sensitivity of accounting earnings to
bad news. Our measure of conservatism captures the sensitivity of earnings to bad news relative to the
sensitivity to good news, where higher values represent more conservative firms. We estimate this
firm-specific measure of conservatism over the ten year period prior to default.
Sensitivity of earnings to bad news relative to good news following Ball and Shivakumar (2005),
where the sign of operating cash flows represents good and bad news, respectively. We modify the
Basu (1997) regression following Ball and Shivakumar (2005), by regressing accruals on positive and
negative cash flows, where our measure of conservatism captures the sensitivity of accruals to
negative cash flows relative to the sensitivity to positive cash flows. Accruals are defined as change
in current assets , minus change in current liabilities, minus change in cash and cash equivalents, plus
changes in current debt, minus depreciation, scaled by total assets. Higher values of Cons_BS
represent more conservative firms. We estimate this firm-specific measure of conservatism over the
ten year period prior to default.
Cons_BS
50
Appendix C: Variable definitions (continued)
Variable
Definition
Days in Bankruptcy
Total time spent in bankruptcy, measured as the number of days from the date of obligor default to the date of
emergence as indicated in the Moody’s Ultimate Recovery Database (URD) and UCLA Bankruptcy Research
Database (BRD).
Debt Concentration
Herfindahl-Hirschman index of the firm’s debt concentration across bank lenders at the time of default.
Debt Rating
S&P Credit Rating available on Compustat, scaled from 1 to 23 where higher values indicate lower credit
ratings. If missing, we impute debt rating following Beatty et al. (2008) by regressing debt rating on return on
assets, leverage, size, subordinated debt, dividends paid and an indicator variable for loss firms; we use the
fitted value from this model as the estimated credit rating.
Distressed Exchange
Emerge GDP
Indicator variable = 1 if the firm completed a distressed exchange, 0 otherwise
Indicator variable equal to 1 if the firm successfully emerged from bankruptcy, 0 if the firm was liquidated or
acquired in bankruptcy resolution.
Trailing four quarter U.S. GDP growth rate, measured at the date of emergence.
Emerge S&P 500
Trailing twelve month returns of the S&P 500 index, measured at the date of emergence.
Emerge Spec
Trailing twelve month Moody’s speculative grade corporate default rates, measured at the date of emergence.
Emerge Spread
Bond yield spread between Moody’s BAA-rated and AAA-rated corporate bonds, measured at the date of
emergence.
Emerge
Ex-Post Asset WD
Ex-Post Loss
GDP Growth
Int Coverage
Interest Miss
Lag_GDP
Lag_S&P 500
Lag_Spec
Lag_Spread
Total asset write-downs recorded after bankruptcy measured as (-1) multiplied by the percentage change in
the firm’s total assets from the last balance sheet prior to default to the first balance sheet following default.
We scale total write-downs by the number of days between the default date and the ex-post balance sheet date
to standardize this measurement period because firms with a greater lag between the default date and the
balance sheet date are more likely to record greater asset write-downs. Financial information following
bankruptcy is obtained from the monthly operating reports filed with the bankruptcy court manually collected
from 8-K filings on Morningstar’s 10-K Wizard. If monthly operating reports are not available, we obtain data
from the first quarterly financial report following default on Compustat.
Total losses recorded after bankruptcy measured as (-1) multiplied by of the firm’s net income scaled by total
assets after default. We measure net income following bankruptcy using data available in monthly operating
reports manually collected using 10-K Wizard. If the monthly operating report is unavailable, we use the first
balance sheet date available on Compustat following default. We scale total losses by the number of days in
the measurement period, because each monthly operating reports filed with the bankruptcy court are not
standardized with regards to the number of days in each report.
Trailing four quarter U.S. GDP growth rate, measured at the date of default.
Ratio of annual operating cash flows to total interest expense, averaged over the three year period prior to
default.
Indicator variable equal to 1 if the firm missed an interest payment on its debt obligations in the year prior to
bankruptcy, 0 otherwise.
Trailing four quarter U.S. GDP growth rate, measured at the inception of the firm's largest (in dollars) debt
contract.
Trailing twelve month returns of the S&P 500 index, measured at the inception of the firm's largest (in dollars)
debt contract.
Trailing twelve month Moody’s speculative grade corporate default rates, measured at the inception of the
firm's largest (in dollars) debt contract.
Bond yield spread between Moody’s BAA-rated and AAA-rated corporate bonds, measured at the inception
of the firm's largest (in dollars) debt contract.
51
Appendix C: Variable definitions (continued)
Variable
Definition
Leverage
Total debt divided by total assets in the quarter prior to default from data available on Compustat.
Litigation
Indicator variable equal to 1 if the firm is in a high litigation risk industry, 0 otherwise. SIC codes 2833-2836,
3570-3577, 3600-3674, 5200-5961, 7370-7374 are deemed high litigation risk industries following Beatty et.
al (2008).
Prior ROA
Income before extraordinary items divided by total assets, measured one year prior to default from data
available on Compustat.
Recovery Rate
Firm-wide recovery rates, calculated as the percentage of total value distributed to creditors in bankruptcy
resolution relative to the total debt outstanding at the default date, available from the Moody’s Ultimate
Recovery Database (URD).
S&P 500
Trailing twelve month returns of the S&P 500 index, measured at the date of default.
Secured Debt %
Percentage of firm debt at default classified as secured in the Moody’s Ultimate Recovery Database (URD)
Size
Natural log of total assets, measured in the quarter of default from data available on Compustat.
Skewness
The difference between the skewness of annual operating cash flows and earnings, following Beatty et al.
(2008). We require at least three observations to compute Skewness, and estimate this measure over a
maximum of ten years prior to default.
Solvent
Indicator variable equal to one if the firm's average ratio of annual operating cash flows to total interest
expense over the three-year period prior to default is greater than one, and zero otherwise.
Spec_rate
Trailing twelve month Moody’s speculative grade corporate default rates, measured at the date of default.
Special Items Ratio
Ratio of the average special items recorded in the Income Statement (Compustat SPI) scaled by the average
total assets over the three year period prior to default, to cumulative stock returns over the three year period
prior to default. We require negative total stock returns over the three year period prior to default, to ensure
higher values of Special Items Ratio indicates more conservatism.
Bond yield spread between Moody’s BAA-rated and AAA-rated corporate bonds, measured at the date of
default.
Spread
Violate
Indicator variable equal to 1 if the firm violated a covenant in the year prior to bankruptcy, 0 otherwise.
52
Download