Download Accruals, Financial Distress, and Debt Covenants Troy D. Janes

Accruals, Financial Distress, and Debt Covenants
Troy D. Janes
University of Michigan Business School
701 Tappan Street
Ann Arbor, MI 48109
[email protected]
(734) 763-3537
This version: January 2003
This paper is based on my dissertation at the University of Michigan Business School. I
am grateful to my dissertation committee, Patricia Dechow, Ilia Dichev, Tyler Shumway
and Anant Kshirsagar.
This paper has also benefited from helpful comments and
suggestions from Scott Richardson, Judy Day and student and faculty workshops at the
University of Michigan Business School. The author gratefully acknowledges financial
support from the University of Michigan Business School, the William A. Paton
Scholarship Fund and the American Institute of Certified Public Accountants.
Abstract
This paper documents that accruals provide information that is useful for predicting
financial distress and examines the use of this information by commercial lenders in
setting debt covenants. Controlling for the level of earnings, firms with extreme accruals
are more likely to experience financial distress than firms with moderate accruals. Tests
of the relation between accruals and debt covenant tightness show that, as expected, the
debt covenants of borrowing firms with low accruals are set tightly; however, contrary to
expectations, the debt covenants of borrowing firms with high accruals are set relatively
loosely. Since prior research has shown that lenders possess unique information about
borrowers, this result can be interpreted as additional evidence that sophisticated users of
accounting information do not fully utilize the information in accruals. However, it is
important to note that debt covenants reflect only one aspect use of the use of financial
information by lenders, and they may use the information in accruals in other ways.
1.
Introduction
This paper examines whether commercial lenders incorporate the information about
financial distress contained in accruals into debt covenants.
It documents that,
controlling for earnings, accruals provide incremental information over standard
variables used in models for predicting financial distress. It further shows that firms with
extreme accruals are more likely to become distressed than firms with moderate accruals.
This paper also examines one possible use of the information in accruals by commercial
lenders. Results indicate that lenders do not fully consider the relation between accruals
and financial distress when setting the initial tightness of debt covenants. As expected,
debt covenants are set more tightly for borrowing firms with low accruals, regardless of
the level of earnings. However, tests reveal that debt covenants for firms with high
accruals are set more loosely than firms with moderate accruals. Because the initial
tightness of debt covenants is not consistent with the information in accruals about
financial distress, this result adds to prior literature that suggests that sophisticated users
of accounting information do not fully utilize the information in accruals.
Prior research on the information in accruals has shown that high accruals are associated
with declining future performance. Sloan (1996) finds that earnings consisting primarily
of accounting accruals are less persistent than earnings predominantly made up of cash
flows. His results indicate that the performance of firms with extreme accruals tends to
mean-revert more quickly than firms with moderate levels of accruals.
This result
indicates that firms with high accruals will experience lower earnings performance in the
future.
1
Other studies have documented the relation between high accruals and future unfavorable
events.
Changing auditors is generally regarded as a negative signal about firm
performance (Schwartz and Menon, 1985; Johnson and Lys, 1990; Schwartz and Soo,
1995, 1996), and DeFond and Subramanyam (1998) find that firms with high accruals are
more likely to change auditors. Dechow, Sloan and Sweeney (1996) find that firms with
high accruals are more likely to be subject to SEC enforcement actions for violations of
generally accepted accounting principles. High accruals have also been associated with
management's attempts to manipulate earnings to avoid problems such as debt covenant
violations (Dichev and Skinner, 2002; DeFond and Jiambalvo, 1994).
Despite the fact that high accruals have been associated with declining future
performance, there is evidence that sophisticated users of accounting information do not
fully utilize the information in accruals. Sloan (1996) shows that even though high
accruals predict declining performance, stock prices behave as if the market does not
understand this information. A study of analyst forecast errors shows that forecast errors
are larger for firms with high accruals (Bradshaw, Richardson, and Sloan, 2001), and a
study of analyst forecast revisions finds that analysts do not revise their forecasts in
anticipation of predictable accrual reversals (Barth and Hutton, 2001). Ahmed, Nainar
and Zhou (2001) find that analyst forecasts underweight both accrual and cash flow
information, indicating an underutilization of the differing information provided by these
measures. Bradshaw, Richardson and Sloan (2001) examine audit opinions and find that
future earnings reversals driven by high accruals do not affect auditor opinions.1
1
DeFond and Subramanyam (1998) find that firms that change auditors have higher discretionary accruals
than other firms. One interpretation of this finding is that firms that change auditors are “opinion
shopping.” If firms with high accruals respond to potential audit opinion qualifications by changing
auditors, this result may explain why Bradshaw, Richardson and Sloan (2001) find that high accruals do
not affect auditor opinions. That notwithstanding, the fact that these high-accrual firms are able to obtain a
clean opinion from a new auditor indicates that the new auditor either does not understand the information
in high accruals or is willing to look the other way for the sake of new business. Since DeFond and
Subramanyam note that the new auditor is usually smaller than the old auditor (e.g. a change from Big 5
firm to a regional firm), it is possible that the new auditor is less sophisticated than the previous auditor and
does not understand the information in accruals.
2
Richardson (2002) finds that short sellers do not appear to actively trade on the
information in accruals. In contrast to these studies, Collins, Gong and Hribar (2002)
find that institutional investors appear to price accruals more correctly than other investor
groups mentioned above.
A series of prior studies have examined role of commercial lenders as users and
producers of financial information and found that lenders appear to have unique
information about borrowers not available to others. A theoretical study by Campbell
and Kracaw (1980) suggests that an important role of financial intermediaries (e.g.
banks) is the production of information. Empirical studies have found significant market
reactions to announcements about bank loans indicating that financial market participants
behave as though they believe lenders posses unique information. Best and Zhang (1993)
find significant market reactions to the announcement of new loans, and Dahiya, Puri and
Saunders (2002) find that negative stock market returns are associated with sales of loans
by the originating lender. An objective of this study is to add to prior research by
investigating whether commercial lenders, as a sophisticated group of financial
information users, use the information in accruals in setting debt covenants.
A database of private lending agreements, Dealscan, is used to obtain detailed
information on debt covenants. The use of Dealscan has two advantages over prior
studies of debt covenants. First, it allows the study of private debt contracts. Prior to the
release of Dealscan, there was little publicly available information on private debt
contracts.
Consequently, most prior studies of debt covenants examine public debt
contracts (i.e. bonds). Because of the large number of bondholders involved in a public
debt issue, renegotiating a debt contract following a covenant violation can be costly and
difficult. Since there are significantly fewer parties involved, the renegotiation of private
debt contracts following a covenant violation is easier to carry out. As a result, private
3
debt agreements generally contain a greater number of debt covenants than public debt
agreements, and these covenants are set more tightly than those in public debt agreements
(Smith & Warner, 1979; Gopalakrishnan and Parkash, 1995). Therefore, covenant levels
in private debt agreements are likely to be the product of careful analysis by commercial
lenders.
The second advantage to using Dealscan results from the detailed information it provides
on debt covenants. The database generally provides enough information on the level of
debt covenants to allow the calculation of actual debt covenant tightness. Most existing
studies on debt covenants use measures such as total debt or the debt-to-equity ratio to
proxy for covenant tightness because actual data has not been available (Dichev and
Skinner, 2002). Because these proxy measures are noisy, studies using them are difficult
to interpret (e.g. Mohrman, 1993; Ball and Foster, 1982).
The remainder of this paper proceeds as follows: The next section develops testable
hypotheses. Section 3 describes the sample used in testing the relation between accruals
and financial distress and presents the results of those tests. Section 4 examines the
relation between accruals and debt covenant tightness, and Section 5 concludes and
provides suggestions for future work.
2.
Hypothesis Development
2.1
Accruals and Financial Distress
Prior research has shown that high levels of accruals lead to future declines in
performance. However, declining performance does not mean that a firm is financially
distressed.
In order to examine whether commercial lenders should use accrual
4
information in setting debt covenants, a necessary first step is to show that accruals are
useful in predicting financial distress.
Prior research has developed several models for predicting financial distress (in
particular, bankruptcy). Each of these models uses a similar set of accounting ratios to
estimate a firm's risk of bankruptcy.
Using discriminant analysis, Altman (1968)
develops a bankruptcy prediction model (with a summary statistic known as Altman's Zscore) that includes five accounting ratios: working capital to total assets, retained
earnings to total assets, earnings before interest and taxes (EBIT) to total assets, market
value of equity to total liabilities, and sales to total assets.
Zmijewski's (1984) model
includes net income to total assets, total liabilities to total assets, and current assets to
current liabilities. Ohlson (1980) developed a model utilizing firm size (log of total
assets), total liabilities to total assets, net income to total assets, and working capital (or
current liabilities) to total assets. Finally, Shumway (2001) created a hazard model using
some of these accounting ratios together with stock market data. His model includes net
income to total assets, total liabilities to total assets, relative size (relative to the
NYSE/AMEX market), excess returns, and the standard deviation of the firm's stock
returns.
One common factor in each of these models is a measure of earnings. In each of the
models, higher earnings are associated with to a lower risk of bankruptcy. However, in
light of Sloan’s (1996) finding that high accruals are associated with lower future
earnings, considering the level of earnings alone does not give one a complete picture.
All other things being equal, a firm with high earnings and high accruals will experience
a greater decline in future earnings than a firm with high earnings and low accruals. It
follows that accruals provide information over and above that provided by earnings
alone.
5
Although prior research links high accruals and declining earning performance, it does
link accruals and financial distress. However, an analysis of the causes of high accruals
provides a possible link. High accruals resulting from increases in accounts receivable
may indicate that a company is having trouble collecting money owed it. Increases in
inventories may indicate that the company’s sales are lagging. Both of these problems
suggest that the firm may be experiencing liquidity problems that may lead to financial
distress. Or, in a worst case scenario, high accruals may be the result of earnings
management intended to artificially inflate earnings (Dichev and Skinner, 2002; DeFond
and Jiambalvo, 1994).
Likewise, although prior research indicates that low accruals lead to improved earnings
performance (Sloan, 1996), low accruals resulting from increases in accounts payable and
accrued liabilities may also indicate that the company has an inability to pay its debts.
Such liquidity problems may also lead to financial distress. Since it can be argued that
both very high and very low accruals may indicate a liquidity problem that may lead to
financial distress, the first hypothesis of this paper is as follows:
H1:
Holding earnings constant, firms with high absolute accruals are more
likely to experience financial distress than firms with moderate accruals.
2.2
Accruals and Debt Covenants
Early research by Jensen and Meckling (1976) and Smith and Warner (1979) have shown
that borrowers have the incentive and the ability to shift wealth from lenders to
shareholders. In order to facilitate lending, the lender and borrower write covenants into
6
the debt contract that restrict the actions of the borrower and establish monitoring to
ensure that the terms of the debt contract are being met.
These covenants take two forms, sometimes referred to as negative and positive
covenants. Negative covenants generally prohibit certain activities that result in asset
substitution or repayment problems.
Examples of negative debt covenants include
prohibitions on mergers, limits on additional borrowing, restrictions on dividend
payments and excess cash sweeps. Positive covenants require the borrower to take
certain actions, such as insuring assets used as collateral or meeting certain benchmarks
(usually accounting ratios) that indicate financial health. Common examples of positive
debt covenants include minimum or maximum allowable levels of current ratio, leverage
ratios, profitability and net worth ratios.
Debt covenants are used by commercial lenders as early warning systems to signal
impending financial problems among borrowers. When a covenant is violated, lenders
have the option to require immediate repayment of the loan. Most of the time, however,
after reassessing the borrower's situation, the lender waives the violation and resets the
covenant below the current level. If the borrower's performance improves, there is no
further problem. If the borrower's performance continues to deteriorate, the covenant is
again violated, and the lender once again has the opportunity to evaluate the borrower's
performance (Smith, 1993; Chen and Wei, 1993; Gopalakrishnan and Parkash, 1995;
Dichev and Skinner, 2002).
Although enforcement of debt covenants can vary from situation to situation, there is
strong evidence that debt covenants impact firms in several ways. Core and Schrand
(1999) find that firms that are close to violating debt covenants experience a greater
negative stock price reaction to bad news than do firms that are not close to violating
7
covenants. El-Gazzar (1993) finds a negative stock price reaction to the announcement
of new accounting policies that may push firms closer to violating debt covenants.
Finally, Beneish and Press (1993) document costs associated with the violation of
positive debt covenants, referred to as technical default. Frequently, these violations can
be waived or the covenant can be renegotiated, but the borrower incurs costs in doing so,
ranging from the actual costs of negotiation (attorney’s fees, etc.) to the addition of new
covenants.2
Sweeney (1994) and Dichev and Skinner (2002) find evidence that managers take actions
to avoid debt covenant violations, although they are unable to determine whether such
actions are cases of earnings management or “real” actions such as accessing equity,
selling assets, deferring purchases, etc.
Dichev and Skinner also report that debt
covenants in private contracts are used in an active monitoring role, with lenders using
the covenants as an early warning system to inform them of potential problems with the
borrower.
Despite the importance of debt covenants in the lending process and the subsequent
operation of borrowing firms, there have been few studies on the role of accounting in
debt contracts3 (Sloan, 2001).
Existing studies that examine characteristics of debt
contracts and determinants of debt covenants primarily deal with public debt (i.e. bonds).
These studies focus on factors such as the industry in which the borrower operates, the
number of lenders involved in syndicating the loan, leverage, profitability, and
2
See Chen and Wei (1993) for a discussion of the determinants of waivers.
Recent research on the role of accounting in debt contracts has examined the use of performance pricing,
a feature that allows the interest rate charged on a loan to vary based on the borrower’s financial health, as
measured by accounting ratios or credit ratings (Asquith, et al, 2001; Beatty, et al, 2001; Doyle, 2002).
Beatty, et al (2001) finds that performance pricing and covenants are complements rather than substitutes,
particularly when measured over the same variable (e.g. a debt contract that includes performance pricing
based on debt-to-EBITDA as well as a covenant requiring the firm to maintain the a minimum level of the
same ratio). They conclude that performance pricing addresses improvements in firm health (or credit
risk), whereas, debt covenants are used to monitor for declines in firm health.
3
8
probability of bankruptcy (Malitz, 1986; El-Gazaar and Pastena, 1991; Berlin and
Mester, 1992). Additionally, Berlin and Mester (1992) show that the restrictiveness of
debt covenants is decreasing in the credit worthiness of the borrower.4
Evidence discussed above and in Section 1 shows that debt covenants play a significant
role in debt contracting and that lenders have ample opportunity and motivation to use all
available information in setting debt covenants. Therefore, one would expect that the
initial level of debt covenants would reflect information in accruals about financial
distress.5 Stated as Hypothesis 2:
H2:
The initial tightness of debt covenants is a function of the magnitude of
accruals.
The results of tests of the hypotheses developed in this section are presented in the next
two sections.
3.
Relation between Accruals and Financial Distress
3.1
Financial Distress Sample
Tests of Hypothesis 1 involve comparisons of the level of accruals of distressed and nondistressed firms. Although a possible research design in cases such as this is to identify a
sample of distressed firms and compare that sample to a matched sample of nondistressed firms, such non-random sampling can result in biased parameter and
4
Discussions with commercial lenders confirmed that the findings of prior research are consistent with
actual lending practices (Chaika, 2001; Bacevich, 2002).
5
The benchmark contained in the covenant may change over time, generally requiring improving
performance by the borrower. However, because factors unrelated to the lender's analysis of the borrower
(e.g. economic downturns, etc.) may affect debt covenant tightness during the term of the loan, this study
focuses only on the initial tightness of the debt covenant.
9
probability estimates. To avoid such biases, the sample used in this study consists of all
firm-years in the Compustat database with sufficient data to compute the variables of
interest. Note that the requirement for “complete data” may also introduce bias into the
sample, but this bias, in general, does not affect statistical inferences (Zmijewski, 1984).
The sample consists of 36,652 firm-year observations from 7,007 firms during the period
1990-2000. 6 367 of these firms (5.2% of the sample) experience financial distress during
the sample period. Each firm-year observation consists of data for Year t and the two
preceding years.
With the exception of firms identified as distressed, all firm-year
observations with sufficient data to compute the required variables are included in the
sample.
As in prior research (Shumway, 1996; Dichev, 1998), CRSP data was used to identify
financially distressed firms.
There is no generally accepted definition of “financial
distress.” A firm that files for bankruptcy is universally considered to be in financial
distress, but bankruptcy is the extreme manifestation of financial distress. A firm filing
for bankruptcy may have been experiencing financial problems for some time before the
filing, but it is difficult to identify when the period of distress began.7 For purposes of
this study, I use exchange delisting for performance reasons as an indicator of financial
distress. CRSP delisting codes indicate when a firm is delisted from its exchange and for
what reason. Reasons for delisting include bankruptcy, insufficient capital, low stock
price, failure to make SEC and/or exchange-required filings in a timely manner, etc. One
drawback of using delisting to identify financially distressed firms is that delisting is not
always timely. Dichev (1998) cites the case of Continental Airlines, which filed for
6
This period was chosen because cash flow data is not available before 1988. The period also corresponds
with the availability of debt covenant data used in tests of Hypothesis 2.
7
An additional confounding factor is that fact that bankruptcy filings may be hastened or delayed for
strategic reasons.
10
bankruptcy protection in December 1990 but was not delisted until 1992.8 Indeed it is
possible that a firm in financial distress may not be delisted at all. However, this
limitation of the CRSP data does not weaken the results of the tests in this paper. Failure
to identify all firms in financial distress merely means a smaller sample of distressed
firms, which lowers the power of the statistical tests and makes it more difficult to find
results.
The variable of interest is the level of accruals found in the firm’s earnings, computed as
earnings before extraordinary items (Compustat Item #18) minus cash flows from
operations from the statement of cash flows (Item #308). Additionally, factors identified
in previous research as having the ability to predict financial distress are included as
control variables (Altman, 1968; Zmijewski, 1984; Ohlson, 1980; Shumway, 2001).
Each of these studies identifies somewhat overlapping sets of factors that predict future
financial distress (specifically, bankruptcy). Altman’s study is the most well known and
often used in practice. Because of this, the factors identified in Altman (1968) are used
as control variables for multivariate tests of the relation between accruals and financial
distress. The Altman model is:
Z = (1.2 x WC) + (1.4 x RE) + (3.3 x EBIT) + (0.6 x MVE) + (0.999 x S)
(1)
The Altman factors are working capital (WC), retained earnings (RE), earnings before
interest and taxes (EARN), market value of equity (MVE), and sales (S). All variables
are scaled by total assets except MVE, which is scaled by total liabilities.9
8
Contrasting cases that are high profile at the time of this writing are those of Enron and WorldCom,
which were both delisted within weeks of filing for bankruptcy in November 2001 and July 2002,
respectively.
9
A firm with a higher Z-score is considered to have a lower probability of bankruptcy. However, recent
research has suggested that Altman's coefficients are outdated (Grice and Ingram, 2001). Begley, Ming,
and Watts (1996) provide re-estimated coefficients that are more accurate when using recent data.
11
3.2
Tests of the Relation between Accruals and Financial Distress
Descriptive statistics for the variables in the financial distress sample are presented in
Table 1. For all firms in the sample, as well as for non-distressed and distressed sample
firms, this table presents the Altman Z-score factors as well as total accruals (TACC),
total liabilities (TL), and current ratio (CR) for the current year, labeled Year 0, and the
two preceding years. In the case of the financially distressed firms, Year 0 represents the
last annual report issued before becoming distressed (i.e. under the definition used in this
study, being delisted). Since each coefficient in the Altman model is positive, the value
of each Altman factor is expected to be higher among the non-distressed firms in the
sample than among the distressed firms.
Table 1 shows that this is the case for all
Altman variables except sales (S). Additionally, Table 1 shows that performance of the
distressed firms, as measured by the Altman factors, declines rapidly over the three years
leading up to becoming distressed, whereas the performance of the non-distressed firm
remains relatively stable. Finally, Table 1, Panel B shows that distressed firms typically
have lower average accruals than non-distressed firms and more debt on average than
non-distressed firms. The current ratios (CR) of distressed and non-distressed firms are
not significantly different in Year -2, but the CR for distressed firms is lower than that of
non-distressed firms in Years -1 and 0.
Table 2 provides additional detail on the sample and univariate evidence about H1. Panel
A of Table 2 presents information on quintile portfolios formed on the level of total
accruals at Year -2. In the case of financially distressed firms, Year -2 represents the
annual report issued approximately three years prior to the firm becoming distressed.
Year -2 accruals were used in forming portfolios because Sloan (1996) reports that high
12
accruals generally revert to the mean over three years. Additionally, the average term of
the loans in the debt covenant sample is about 40 months, or just over three years.10
Using Year -2 data to form portfolios aligns the time horizon of this analysis with the
average time horizon considered by lenders in making lending decisions.
H1 states that accruals provide additional information relevant for predicting financial
distress above that found in earnings alone. The results shown in Table 2 indicate that
accruals do provide additional information regarding the probability of future financial
distress. Panel A of Table 2 presents the mean value of earnings (EARN), total accruals
(TACC), and current ratio (CR) for quintile portfolios formed on total accruals. For all
variables, there is a marked difference between Portfolio 5 (high accruals) and Portfolio 1
(low accruals).
However, consistent with the results in Sloan (1996), the mean
performance of the high accrual portfolio decreases rapidly in the periods following
portfolio formation, while the mean performance of lower accrual portfolios remains
more stable. Also consistent with Sloan, mean earnings performance in Portfolio 1
improves over the three-year period.
In Year -2, mean EARN in Portfolio 5 is 0.079,
compared to a mean EARN of -0.012 in Portfolio 1. By Year 0, however, mean EARN
in Portfolio 5 has dropped to 0.028, and mean EARN in Portfolio 1 has risen to 0.008.
Rapid reversal of accruals in the high accrual portfolio is also evident, with mean total
accruals in Portfolio 5 dropping from 0.096 in Year -2, a number considerably higher
than the mean of other portfolios, to -0.045 in Year 0, which is much closer to the other
portfolio means in Year 0.
To test the hypothesis that, holding earnings constant, firms with higher accruals have a
greater risk of financial distress than firms with moderate accruals, portfolios with similar
mean earnings are created using a two-pass construction. Following Dechow and Dichev
10
Table 4 provides descriptive data for the debt covenant sample and will be discussed later in the paper.
13
(2001), the sample is sorted into decile portfolios based on the level of earnings. Then,
each earnings decile is sorted into quintile portfolios based on the level of total accruals.
Then, portfolios are formed by pooling the subportfolios formed in each decile. The final
result is five portfolios based on total accruals while controlling for earnings. Portfolio 5
is comprised of the highest quintile of total accruals in each earnings decile; Portfolio 1 is
comprised of the lowest quintile of total accruals in each earnings decile, and so on.11
Panel B of Table 2 presents the results of this procedure, with portfolios being formed
using Year -2 values of EARN and TACC. The procedure to hold earnings constant was
successful, with all portfolios but Portfolio 1 reporting mean EARN between 0.060 and
0.061. The mean EARN of Portfolio 1 is 0.051, which indicates that the sample contains
some observations with extremely low earnings.12
Again affirming the findings of Sloan (1996), these results show that, although nearly
equal to the mean EARN of other portfolios in Year -2, the mean EARN of Portfolio 5
drops rapidly from 0.061 in Year -2 to 0.015 in Year 0. This figure is well below the
mean EARN of the other portfolios, which range from 0.041 to 0.057.
It is also
interesting to note that the mean EARN of Portfolio 1 remains fairly constant, increasing
slightly from 0.051 to 0.052 over the same period. As in Panel A, the mean total accruals
of Portfolio 5 drops rapidly after portfolio formation, while the mean total accruals of
Portfolio 1 rises over the same period. Because of the presence of outliers in the sample,
portfolio medians for the same portfolios are reported in Panel C of Table 2. The median
value of EARN in Year -2 is the same (0.081) for all portfolios. It is interesting to note
that the median EARN for Portfolios 1 and 2 remains nearly constant over the three years
11
12
I am grateful to Ilia Dichev for suggesting this procedure.
Untabulated tests performed on the sample after deleting outliers yield similar results.
14
presented, while the median EARN for Portfolio 1 falls from 0.081 in Year-2 to 0.060 in
Year 0, the smallest portfolio mean in Year 0.
The relation between total accruals and the occurrence of financial distress is presented in
Figure 1, which graphically depicts the occurrence of financial distress in the portfolios
discussed above. Figure 1a shows the occurrence of financial distress in portfolios
formed on total accruals.
Among firms with moderate accruals, the occurrence of
financial distress does not vary greatly, ranging from 37 firms (0.5% of the portfolio) in
Portfolio 2 to 55 (0.8%) in Portfolio 4.
Portfolios 1 and 5 have a much greater
occurrence of financial distress with Portfolio 5 (high accruals) containing 88 (1.2%)
distressed firms and Portfolio 1 containing 141 (1.9%) distressed firms. Results are
similar in portfolios based on total accruals while controlling for earn.
Figure 1b
presents the occurrence of financial distress in portfolios formed on total accruals,
controlling for earnings as done in Table 2, Panel B. Figure 1b shows that financial
distress in the moderate accrual portfolios ranges from 55 occurrences (0.76%) in
Portfolio 2 to 60 (0.82%) in Portfolio 4. Again, the incidence of financial distress is
much higher in the high- and low-accrual portfolios, with Portfolio 1 containing 94
(1.28%) distressed firms, and Portfolio 5 containing 101 (1.41%) distressed firms. These
univariate results indicate that firms with the highest and lowest accruals, holding
earnings constant, are at greater risk of financial distress than firms with more moderate
accruals.
Multivariate tests of the relation between accruals and financial distress are presented in
Table 3, which reports the results of logistic regressions where the dependent variable is
equal to one if the firm is financially distressed following Year 0 and zero otherwise.
Model 1 presents the regression of the distress indicator on earnings and the other factors
in the Altman model. Model 2 is the same regression including indicator variables for
15
low accruals (LOWACC) and high accruals (HIGHACC). LOWACC equals 1 if the
level of accruals for the firm in Year -2 is in the lowest quintile of total accruals, zero
otherwise. HIGHACC equals 1 if the level of accruals for the firm is in the highest
quintile of accruals, zero otherwise. The results are reported for Year 0, Year -1, and
Year -2. As predicted by the Altman model, the coefficient on EARN is negative in both
models. The coefficients on WC, RE, and MVE are negative and statistically significant
in each regression, indicating that increases in each of these factors decreases the
probability of financial distress. The coefficient on S is significant, but in the opposite
direction than predicted. Table 2 indicates that firms with high and low accruals have a
greater risk of financial distress; therefore, the coefficients on LOWACC and HIGHACC
are predicted to be positive. The results indicate that, across all years, LOWACC and
HIACC are significant and of the predicted sign. These results indicate that, when
considered in addition to earnings, low and high accruals indicate a higher risk of
financial distress than more moderate accruals.
A comparison of the explanatory power of Model 1 and Model 2 is made by comparing
the Akaike Information Criterion (AIC) for each model. A lower AIC indicates more
explanatory power (SAS Institute, 1999). For each year, the AIC is lower for Model 2,
indicating that accruals provide incremental explanatory power for predicting bankruptcy
over earnings alone.
In summary, univariate results presented in Table 2 along with the multivariate results in
Table 3 indicate that accruals provide more information for the prediction of financial
distress than using earnings alone. They also show that the relation between accruals and
financial distress is not linear, and firms with the highest and lowest levels of accruals are
more likely to experience financial distress than firms with moderate levels of accruals.
This finding is not surprising in the case of low accruals, where large negative accruals
16
could be associated with increasing liabilities and related claims on future cash flows. In
contrast, high accruals should be associated with increasing cash flows in the future.
However, prior research shows that troubled firms sometimes use income-increasing
accounting choices to mask their financial condition (Dichev and Skinner, 2002; DeFond
and Jiambalvo, 1994).
This use of accruals may prevent the firm from taking
opportunities to work the problems out, diverting the attention of creditors or
shareholders until the firm defaults on a loan, for example (HassabElnaby, 2002). Recent
research on bankruptcy emergence supports this idea. Bryan, Tiras and Wheatley (2002)
find that bankrupt firms that made income-increasing accounting choices prior to
bankruptcy have a lower chance of emerging from bankruptcy. Again, the authors
theorize the use of income-increasing accounting choices delays the filing of bankruptcy
until the firm’s financial problems are deeper, thus resulting in a lower likelihood that the
firm will successfully emerge from bankruptcy.
4
Relation between Accruals and Debt Covenant Tightness
4.1
Debt Covenant Sample
Tests of Hypothesis 2 examine whether commercial lenders understand the implications
of accruals for financial distress. Data on loans is taken from the Dealscan database
provided by LPC Market Access. Dealscan provides a database of over 50,000 loans
dating back to 1986.
Dealscan consists of loan data gathered from SEC filings,
supplemented by research by LPC.
The database includes information on the terms of
the loan (amount, interest rate, length, etc.) as well as the covenants contained in the debt
contract.
17
Loans in the database typically have two or more facilities, or parts. For example, the
loan could include a revolving loan and a term loan. Each facility can have slightly
different terms, such as different interest rates, but the covenants generally apply to all
facilities in the loan. The facility with the longest maturity is assumed to represent the
loan and is considered to be the primary part of the loan in this study. If two facilities
have equal maturities, the facility with the largest principal amount is selected for
inclusion in the sample.
The Dealscan database organizes debt covenant information into 12 positive covenants
and five negative covenants. Recall from previous discussion that positive covenants
generally involve meeting benchmark accounting ratios and negative covenants restrict
specific actions. As discussed in Dichev and Skinner (2002), there is a great deal of
variation in the definitions of the ratios used in debt covenants. For example, in an
examination of Dealscan loans they find over a dozen different ways that the debt-tocash flow ratio is defined in debt contracts. They find similar problems with most other
commonly used covenants. Dichev and Skinner (2002) use the current ratio covenant to
examine debt covenant violations because they find that it is fairly consistently defined.
This allows them to calculate covenant slack using covenant data from Dealscan and data
from the borrower’s financials available from Compustat. Since this study also uses
Compustat data and Dealscan covenant data together, the current ratio covenant is the
primary subject of tests.
For inclusion in the sample, data on loan amount, maturity, and current ratio covenant
must be available from Dealscan. Observations in the sample must also have sufficient
data available on Compustat to calculate the variables used in multivariate tests. The final
sample consists of 1,096 loans originating from 1990-1999.
18
4.2
Variable Measurement
Typically, accounting-based debt covenants establish a minimum level for the ratio in
use. When the lender evaluates the borrower’s financial health, perceived deficiencies
will prompt the lender to set the initial level of the covenant more tightly (Chaika, 2001;
Bacevich, 2002). Doing so gives the lender more advance warning of deterioration in the
borrower’s financial health. A “tight” covenant is one in which the benchmark level in
the covenant is close to the actual level of the measure. The difference between the
actual measure and the covenant benchmark is referred to as “slack.” I use slack at loan
inception as my measure of covenant tightness, with lower slack indicating a tighter
covenant. Initial slack is calculated as:
SLACK =
CR − CRCOV
CRCOV
(2)
where
CR = borrower’s current ratio from the annual report immediately preceding the
loan, calculated as current assets (Item #4) divided by current liabilities
(Item #5)
CRCOV = initial current ratio covenant level per Dealscan
Control variables include the investment opportunity set of the borrower (IOS), the term
(i.e. duration) of the loan (TERM), the indebtedness of the borrower prior to acquiring
new debt (DEBT), the size of the borrower as measured by the log of total assets (SIZE),
and the amount being borrowed (AMOUNT). Smith and Warner (1978) discuss several
opportunities that borrowers have to shift wealth away from the borrower. One of these,
19
referred to as asset substitution, is the ability of the borrower to invest borrowed assets in
riskier projects than those approved of by the lender.
If a project financed through
borrowing is extremely successful, the borrower realizes most of the upside since the
payment to the lender is usually fixed by the debt contract. If the project is a failure, the
borrower may not be able to repay the loan. Therefore, in making loans, the lender
assumes much of the downside risk associated with the assets being loaned.
This
provides incentive to the borrower to invest in riskier projects than it would if using
assets already in place.
Since, the IOS is a proxy for the investment opportunities of a firm, a firm with a greater
IOS has more opportunity to engage in asset substitution than other firms (Skinner,
1993).
It follows that the lender has the incentive to place greater restrictions on
borrowers with a greater IOS. Therefore, IOS is included as a control variable, the
predicted sign on IOS is negative—higher IOS indicates less slack. The measure used as
a proxy for the IOS is that found in Chung and Pruitt (1994):
IOS =
MVE + BVDEBT + BVPREF
TA
(3)
where
MVE = market value of equity (Item #199 x Item #25)
BVDEBT = book value of debt (Item #181)
BVPREF = book value of preferred stock (Item #130)
TA = total assets (Item #6)
El-Gazaar and Pastena (1991) and Malitz (1986) show that loans with longer maturities
should have more restrictive covenants, but Berlin and Mester (1992) states that longer
maturities require looser covenants at inception to account for changes in the firm over
time. Therefore, TERM (loan term in months per Dealscan) is included as a determinant
20
of covenant slack, but the predicted sign on TERM is ambiguous. El-Gazaar and Pastena
(1991) also show that firms with more debt have tighter covenants, so DEBT (Compustat
item #9) is a control variable with a predicted negative sign on the coefficient. Prior
research has indicated that the size of the firm is positively correlated with the ability of
the firm to repay its debts. Therefore, I include FIRMSIZE (Item #6) as an additional
determinant of debt covenant slack, and I predict that the coefficient on FIRMSIZE will
be positive.
Finally, similar to the finding that firms with more debt have tighter
covenants, El-Gazaar and Pastena (1991) find that debt covenant tightness is increasing
in the size of the loan. This prompts the inclusion of LOANSIZE (principal amount of
the loan per Dealscan) as a control variable with a negative predicted sign.
Finally, to separate the effects of low and high accruals, I use indicator variables for low
and high accruals (LOWACC and HIGHACC). These indicator variables are computed
in the same manner as those in Table 3, with HIGHACC representing borrowers in the
highest quintile of total accruals, and LOWACC representing borrowers in the lowest
quintile of total accruals.
4.3
Tests of the Relation between Accruals and Debt Covenant Tightness
Descriptive data for this sample is found in Table 4. For comparison purposes, the last
column of Table 4 presents the mean value of each variable measured over the broader
sample used to test Hypothesis 1. Borrowing firms have higher earnings and accruals
than the average firm. The IOS for the borrowing firms is lower, suggesting that firms
that use commercial loans have lower prospects for growth. Borrowing firms are also
smaller and have more debt than the average firm.
21
As with the financial distress sample, the debt covenant sample has been divided into
quintile portfolios based on the level of total accruals reported in the annual report
preceding the loan closing. Table 5, Panel A presents these results. Similar to the results
of tests on the financial distress sample, the firms in Portfolio 5 have the highest earnings
(EARN), greatest IOS, which is primarily a function of market value, and highest current
ratio (CR). Additionally, firms in Portfolio 5 enjoy the highest slack among the quintile
portfolios. The last column in Table 5, Panel A shows that slack increases monotonically
from a low of 0.343 in Portfolio 1 to 0.695 in Portfolio 5. This relation is shown
graphically in Figure 2a.
Panel B of Table 5 presents the results of forming portfolios on total accruals while
holding earnings constant, using the same two-pass construction used to form the
portfolios in Table 2, Panel B. As shown in the EARN column, the mean earnings of
each portfolio are similar, except for Portfolio 1. Mean earnings in Portfolios 2 to 5
range from 0.033 to 0.042, but the mean of Portfolio 1 earnings is 0.003. This result is
presented graphically in Figure 2b. Again, the low mean of Portfolio 1 suggests the
presence of extremely low earnings in the sample.13 Panel C of Table 5 presents median
values of the variables for the same portfolios as in Panel B. The median earnings for
each portfolio varies from 0.043 to 0.044. Whether examining mean or median values,
the portfolio slack increases monotonically as accruals increase.
The results of testing H1 suggest that the relation between the level of accruals and
financial distress is non-linear, with high and low accruals leading to greater incidence of
financial distress than moderate levels of accruals. However, univariate tests of the debt
covenant sample presented in Table 5 indicate that, contrary to expectations, firms with
high accruals have looser debt covenants. Table 6 presents the results of multivariate
13
Untabulated tests on a sample with outliers deleted yield similar results.
22
tests of the relation between accruals and covenant slack. Model 1 of Table 6 shows the
results of including the HIACC and LOWACC indicator variables in a regression of debt
covenant slack on earnings. The coefficient on earnings is 0.567, which is significant at
the 1% level, indicating a positive relation between earnings and slack. The coefficient
on LOWACC is –0.140, which is statistically significant at the 5% level, which shows
that low accruals are associated with lower slack. The coefficient on HIGHACC (0.102)
indicates that high accruals are associated with higher levels of slack. However, this
coefficient is not significant. The results of this test indicate that, while lenders correctly
associate low accruals with greater risk of financial distress, they do not make the same
association with high accruals.
Model 2 in Table 6 runs the same regression but includes other factors that have been
shown to affect debt covenant slack. In Model 2, the coefficient on earnings (0.303) is
still positive, but no longer statistically significant. The coefficient on LOWACC (0.136) remains significantly negative, and the coefficient on HIGHACC (0.013) remains
statistically insignificant. Of the control variables, only IOS and DEBT have significant
coefficients.
The coefficient of 0.061 on IOS indicates that a greater investment
opportunity set is associated with higher slack. The sign on this variable was predicted to
be negative using the reasoning that more investment opportunities were associated with
greater opportunities to shift risk to the lender. However, it appears that lenders value the
borrower’s opportunities to invest in many projects more than they fear any additional
risk the increase in investment opportunity may bring. The coefficient on DEBT is also
significant. The coefficient of -0.857 indicates that borrowers with higher ex ante debt
levels are subjected to a greater level of monitoring.
The above tests show that, in setting debt covenants, commercial lenders fail to fully
utilize the information in high accruals for future financial performance. To further
23
examine the effectiveness of the current ratio debt covenant in light of the information in
accruals, I examine the number of firms that violate the current ratio debt covenant in the
portfolios formed on total accruals. To determine whether a firm violated the current
ratio debt covenant, current ratio data was collected from Compustat for the periods
following loan inception. Data was collected for all years in the loan term up to and
including the year 2000. Firms for which current ratio data was not available for all
years were dropped from the sample for this test, leaving a sample of 756 borrowing
firms. A borrower whose reported current ratio dropped below the covenant benchmark
was considered to be in violation of the loan covenant.
The lighter shaded bars in Figure 3 show the results of this analysis. Nearly 45% of the
borrowers in Portfolio 1 violated the current ratio debt covenant at some point during the
term of the loan. The other portfolios, including Portfolio 5, the high accrual portfolio,
had considerably fewer violators, with around 25% of the firms violating the debt
covenant in each of the other portfolios. Given the fact that Portfolio 1 borrowers were
given considerably less slack than other borrowers, it appears that the number of violators
in Portfolio 1 may be due to the lack of debt covenant slack in that portfolio. To address
this question, the slack for the borrowers in each portfolio is set equal to the average
slack of Portfolio 1, 0.343. The number of violators in each portfolio is then calculated.
The results, represented by the darker bars in Figure 3, show that all portfolios now have
a similar number of violators. This result suggests that violation of debt covenants is a
function of covenant slack; therefore, borrowers in the high accrual portfolio, that receive
greater covenant slack despite their higher bankruptcy risk are not monitored enough.
Taken together, the results discussed in this section indicate that the debt covenants
included by commercial lenders in debt contracts do not reflect an understanding of the
information in accruals for future financial distress.
24
4.4
Other Debt Covenant Samples
In addition to the current ratio debt covenant sample, analyses have been performed on
two additional debt covenants, the debt-to-cash flow covenant and the net worth
covenant. The debt-to-cash flow covenant sets a ceiling level for the debt-to-cash flow
ratio of the borrower, whereas the net worth covenant sets a floor for the borrowing
firm’s net worth (assets minus liabilities). The floor established by net worth covenants
typically increases each year by a percentage of the borrower’s net income for that year.
Tests similar to those documented in this paper for the current ratio covenant were
performed on the debt-to-cash flow and net worth covenant data. For the sake of brevity,
the results of these tests are not tabulated. For both samples, the tests did not show a
significant relation between accruals and covenant slack.
Although these findings provide additional evidence that lenders do not use the
information in accruals in setting debt covenants, there are a few alternative explanations
for these results. First, the samples were relatively small, so the tests may have lacked
power. The debt-to-cash flow sample contained 109 observations, and the net worth
sample consisted of 176 observations compared to over 1000 in the current ratio sample.
Second, the risk of measurement error was particularly great for the debt-to-cash flow
sample. The current ratio is fairly unambiguously defined, and on its face, the debt-tocash flow sample is rather straightforward, too. However, the definitions of debt and
cash flow vary from contract to contract. In an examination of Dealscan loans, Dichev
and Skinner (2002) find over a dozen different ways that the debt-to-cash flow ratio is
defined in debt contracts.
Although I was careful to delete any observations with
ambiguous definitions from the sample, it is very possible that the measure of the debt-tocash flow ratio constructed using Compustat data did not match up well with the
definition the covenant benchmark was based on. Finally, accruals may not be used the
25
same way by lenders in setting various covenants. In my sample data, total accruals and
current ratio are significantly, positively correlated, with a Pearson correlation coefficient
of 0.22 (p<0.0001). The Pearson correlation coefficient on accruals and the debt-to-cash
flow ratio is 0.06 (p=0.554), and the coefficient on accruals and net worth is 0.05
(p=0.488). If lenders understand this and do not consider accruals when setting these two
covenants, one would expect to find no relation.
5.
Summary and Future Work
This paper examines whether a firm’s accounting accruals provide information that is
useful in predicting financial distress. It also examines whether commercial lenders use
the information in accruals for predicting financial distress as reflected in the initial
tightness of debt covenants. Tests of the relation between accruals and financial distress
indicate that accruals provide information for the prediction of financial distress above
that found in earnings alone. Further, firms with extreme accruals are more likely to
experience financial distress. Tests of the relation between accruals and debt covenant
tightness show that although borrowing firms with low accruals have tighter debt
covenants, borrowers with high accruals do not. Overall, the tests in this paper suggest
that accruals provide useful information for predicting future financial distress, but
lenders do not incorporate the information in accruals into debt covenants.
As a caveat, it is important to note that debt covenants represent only one way in which
lenders might use the information in accruals about financial distress. For examples,
lenders may respond to the information in accruals by increasing the interest rate on the
loan and price-protecting itself. Future work in this area could include an examination of
the relation between the information in accruals and interest rates charged on loans.
26
Additional work in this area could incorporate a measure of accrual quality into the
analysis of the relation between debt covenant slack and accruals. It may be that the
relations reported in this study are not so much due to the level of accruals as they are
due to the quality of the accruals. Dechow and Dichev (2001) provide a measure of
accrual quality that relates accruals to future realizations of cash flows. This measure
may be useful in extending this study.
27
References
Ahmed, Anwer S., S.M. Khalid Nainar, and Jian Zhou, 2001, Do analysts forecasts fully
reflect the information in accruals?, Syracuse University Working Paper.
Altman, Edward I., 1968, Financial ratios, discriminant analysis and the prediction of
corporate bankruptcy, The Journal of Finance, 23, 589-609.
Asquith, Paul, Anne Beatty, and Joseph Weber, 2001, Performance pricing in debt
contracting, Pennsylvania State University Working Paper.
Bacevich, Mike, Personal interview, December 12, 2002.
Ball, Ray and George Foster, 1982, Corporate financial reporting: A methodological
review of empirical research, Journal of Accounting Research, 20, 161-234.
Barth, Mary E., and Amy P. Hutton, 2001, Financial analysts and the pricing of accruals,
Stanford University Working Paper.
Beatty, Anne, Ilia D. Dichev, and Joseph Weber, 2001, The role and characteristics of
accounting-based performance pricing in private debt contracts, University of
Michigan Working Paper.
Begley, Joy, Jin Ming, and Susan Watts, 1996, Bankruptcy classification errors in the
1980s: An empirical analysis of Altman's and Ohlson's models, Review of
Accounting Studies, 1, 267-284.
Beneish, Messod D. and Eric Press, 1993, Costs of technical violation of accountingbased debt covenants, The Accounting Review, 68, No. 2, 233-257.
Berlin, Mitchell and Loretta J. Mester, 1992, Debt covenants and renegotiation, Journal
of Financial Intermediation 2, 95-133.
Best, Ronald and Hang Zhang, 1993. Alternative information sources and the information
content of bank loans, The Journal of Finance, XLVIII, No. 4, 1507-1522.
Bradshaw, Mark T., Scott A. Richardson, and Richard G. Sloan, Do analysts and auditors
use the information in accruals?, Journal of Accounting Research, 39, 45-74.
Bryan, Daniel, Samuel L. Tiras and Clark M. Wheatley, 2002. The association of audit
opinion, auditor change and accounting choice with bankruptcy emergence, State
University of New York at Buffalo working paper.
Campbell, Tim S. and William A. Kracaw, 1980. Information production , market
signaling, and the theory of financial intermediation, The Journal of Finance, 35, No.
4, 863-882.
28
Chaika, David, 2001, Personal interview, September 24, 2001.
Chen, Kevin C. W., and K.C. John Wei, 1993, Creditors decisions to waive violations of
accounting-based debt covenants, The Accounting Review, 68, 218-233.
Chung, Kee H. and Stephen W. Pruitt, 1994, A simple approximation of Tobin’s q,
Financial Management, 23, 70-74.
Collins, Daniel W., Guojin Gong and Paul Hribar, 2002, Investor sophistication and the
mispricing of accruals, University of Iowa working paper.
Core, John E. and Catherine M. Schrand, 1999, The effect of accounting-based debt
covenants on equity valuation, Journal of Accounting and Economics 27, 1-34.
Dahiya, Sandeep, Manju Puri and Anthony Saunders, 2002. Bank borrowers and loan
sales: New evidence on the uniqueness of bank loans, Forthcoming in Journal of
Business.
Dechow, Patricia M., and Ilia D. Dichev, 2001, The quality of accruals and earnings: The
role of accrual estimation errors, University of Michigan working paper.
Dechow, Patricia M, Richard G. Sloan, and Amy P. Sweeny, 1996, Causes and
consequences of earnings manipulations: An analysis of firms subject to enforcement
actions by the SEC, Contemporary Accounting Research, 13, 1-36.
DeFond, Mark L. and James Jiambalvo, 1994, Debt covenant violation and manipulation
of accruals, Journal of Accounting and Economics, 17, 145-176.
DeFond, Mark L., and K.R. Subrmanyam, 1998, Auditor changes and discretionary
accruals, Journal of Accounting & Economics, 25, 35-68.
Dichev, Ilia D., 1998, Is the risk of bankruptcy a systematic risk?, The Journal of
Finance, LIII, 1131-1147.
Dichev, Ilia D. and Douglas J. Skinner, 2002, Large-sample evidence on the debt
covenant hypothesis, forthcoming in Journal of Accounting Research.
Doyle, Jeffrey T., 2001, Credit risk measurement and pricing in performance pricingbased private debt contracts, University of Michigan Working Paper.
El-Gazzar, Samir M., 1993, Stock market effects of the closeness to debt covenants
resulting from capitalization of leases, The Accounting Review, 68, No. 2, 258-272.
El-Gazzar, Samir M. and Victor Pastena, 1991, Factors affecting the scope and initial
tightness of covenant restrictions in private lending agreements, Contemporary
Accounting Review 8, No. 1, 132-151.
29
Gopalakrishnan, V., and Mohinder Parkash, 1995, Borrower and lender perceptions of
accounting information in corporate lending agreements, Accounting Horizons, 9, 1326.
Grice, John Stephen and Robert W. Ingram, 2001, Tests of the generalizability of
Altman's bankruptcy prediction model, Journal of Business Research, 54, 53-61.
HassabElnaby, Hassan R., 2002. The dynamic relation between earnings management
and the violation of debt covenants. Virginia State University working paper.
Jensen, Michael C., and William Meckling, 1976, Theory of the firm: Managerial
behavior, agency costs and ownership structure, Journal of Financial Economics,
305-360.
Johnson, W.B. and T. Lys, 1990, The market for audit services: Evidence from voluntary
auditor changes, Journal of Accounting and Economics, 12, 282-308.
Malitz, Ileen, 1986, On financial contracting: The determinants of bond covenants,
Financial Management 15, No. 2, 18-25.
Mohrman, Mary Beth, 1993, Debt contracts and FAS No. 19: A test of the debt covenant
hypothesis, The Accounting Review, 68, 273-288.
Ohlson, James A., 1980, Financial ratios and the probabilistic prediction of bankruptcy,
Journal of Accounting Research, 18, 109-131.
Richardson, Scott, 2002, Accruals and short selling. An opportunity foregone?,
University of Pennsylvania working paper.
SAS Institute, Inc., 1999, SAS OnlineDoc, Version 8.
Schwartz, K. and K. Menon, 1985, Auditor switches by failing firms, The Accounting
Review, 60, 248-261.
Schwartz, K. and B. Soo, 1995, An analysis of Form 8-K disclosures of auditor changes
by firms approaching bankruptcy, Auditing: A Journal of Practice and Theory, 14,
125-136.
Schwartz, K. and B. Soo, 1996, The association between auditor changes and reporting
lags, Contemporary Accounting Research, 13, 353-370.
Shumway, Tyler, 1996, The Premium for Default Risk in Stock Returns, Ph.D.
dissertation, University of Chicago.
Shumway, Tyler, 2001, Forecasting bankruptcy more accurately: A simple hazard model,
Journal of Business, 74, 101-124.
30
Skinner, Douglas J., 1993, The investment opportunity set and accounting procedure
choice, Journal of Accounting and Economics, 16, 407-445.
Sloan, Richard G., 1996, Do stock prices fully reflect information in accruals and cash
flows about future earnings?, The Accounting Review, 71, No. 3, 289-315.
Sloan, Richard G., 2001, Financial accounting and corporate governance: a discussion,
Journal of Accounting and Economics, 32. 335-347.
Smith, Clifford W., 1993, A perspective on accounting-based debt covenant violations,
The Accounting Review 68, 289-303.
Smith, Clifford W. and Jerold B. Warner, 1979, On financial contracting: An analysis of
bond covenants, Journal of Financial Economics 7, 117-161.
Sweeney, Amy P., 1994, Debt covenant violations and managers’ accounting responses,
Journal of Accounting and Economics 17, 281-308.
Zmijewski, Mark E., 1984. Methodological issues related to the estimation of financial
distress prediction models, Journal of Accounting Research, 22, 59-82.
31
TABLE 1
Comparison of predictors of financial distress
in distressed and non-distressed firms
36,652 firm-year observations from the period 1990-2000
Panel A: Altman Z-score factors
Year -2
Year -1
Year 0
WC
All
Non-distressed
Distressed
t-stat
0.271
0.272
0.234
2.90
***
0.256
0.257
0.129
9.61
***
0.242
0.244
0.043
12.90
RE
All
Non-distressed
Distressed
t-stat
0.029
0.036
-0.589
12.54
***
0.008
0.017
-0.962
14.98
***
-0.032
-0.016
-1.565
16.75
***
EARN
All
Non-distressed
Distressed
t-stat
0.059
0.060
-0.092
14.04
***
0.051
0.053
-0.167
17.03
***
0.044
0.046
-0.208
17.63
***
MVE
All
Non-distressed
Distressed
t-stat
4.640
4.645
4.120
1.55
4.195
4.214
2.320
8.23
***
3.938
3.963
1.475
12.92
***
S
All
Non-distressed
Distressed
t-stat
1.235
1.235
1.248
-0.36
1.236
1.235
1.356
-3.19
**
1.238
1.235
1.505
-6.05
***
***
32
TABLE 1 (continued)
Panel B: Other variables
Year -2
Year -1
Year 0
TACC
All
Non-distressed
Distressed
t-stat
-0.043
-0.042
-0.073
3.52
***
-0.053
-0.051
-0.168
8.46
***
-0.062
-0.060
-0.255
8.34
***
TL
All
Non-distressed
Distressed
t-stat
0.491
0.491
0.538
-3.50
***
0.503
0.501
0.646
-9.58
***
0.515
0.512
0.772
-13.30
***
CR
All
Non-distressed
Distressed
t-stat
2.600
2.602
2.458
1.03
***
2.388
2.396
1.615
6.40
***
*, **, ***
2.468
2.475
1.794
6.91
Difference between non-distressed and distressed
significant at the 10%, 5%, or 1% level, respectively.
Sample consists of 36,652 firm-years during the period 1990-2000,
including 367 observations from distressed firms.
Distressed firms are firms identified by CRSP as having been delisted
for performance reasons during the sample period. Year 0 is the firm's annual
report immediately preceding delisting.
Variables:
WC
RE
EARN
MVE
S
TACC
TL
CR
Working capital divided by total assets
Retained earnings divided by total assets
Earnings before interest and taxes divided by total assets
Market value of equity divided by total liabilities
Sales divided by total assets
Total accruals divided by total assets
Total liabilities divided by total assets
Current ratio
33
TABLE 2
Factors predicting financial distress
Portfolios formed on total accruals at Year -2
36,652 firm-year observations from the period 1990-2000
Panel A: Variable Means. Portfolios formed on total accruals at Year 2
(1=Lowest)
# of
Variable
Port. Obs.
Year -2
Year -1
Year 0
EARN
5
4
3
2
1
7330
7330
7331
7330
7331
0.079
0.080
0.078
0.070
-0.012
0.046
0.067
0.070
0.064
0.009
0.028
0.057
0.065
0.059
0.008
TACC
5
4
3
2
1
7330
7330
7331
7330
7331
0.096
-0.003
-0.039
-0.073
-0.193
-0.018
-0.035
-0.049
-0.063
-0.099
-0.045
-0.047
-0.053
-0.068
-0.097
CR
5
4
3
2
1
7330
7330
7331
7330
7331
3.261
2.914
2.351
2.300
2.177
3.054
2.703
2.220
2.195
2.172
2.900
2.600
2.157
2.170
2.156
34
TABLE 2 (continued)
Panel B: Variable Means. Portfolios formed on total accruals controlling
for
earnings at Year -2 (1=Lowest)
# of
Variable
Port.
Obs.
Year -2
Year -1
Year 0
EARN
5
4
3
2
1
7330
7330
7331
7330
7331
0.059
0.059
0.059
0.057
0.048
0.028
0.046
0.052
0.058
0.058
0.012
0.037
0.048
0.055
0.051
TACC
5
4
3
2
1
7330
7330
7331
7330
7331
0.098
-0.003
-0.043
-0.085
-0.185
-0.022
-0.038
-0.053
-0.069
-0.091
-0.049
-0.053
-0.058
-0.071
-0.091
CR
5
4
3
2
1
7330
7330
7331
7330
7331
3.319
3.184
2.603
2.301
2.043
3.052
2.914
2.486
2.238
2.059
2.888
2.732
2.433
2.212
2.058
35
TABLE 2 (continued)
Panel C: Variable medians. Portfolios formed on total accruals controlling
for earnings at Year -2 (1=Lowest)
# of
Variable
Port.
Obs.
Year -2
Year -1
Year 0
EARN
5
4
3
2
1
7330
7330
7331
7330
7331
0.081
0.081
0.081
0.081
0.081
0.067
0.076
0.079
0.081
0.084
0.060
0.072
0.078
0.081
0.081
TACC
5
4
3
2
1
7330
7330
7331
7330
7331
0.069
-0.005
-0.038
-0.067
-0.130
-0.008
-0.027
-0.041
-0.055
-0.074
-0.022
-0.033
-0.042
-0.054
-0.071
CR
5
4
3
2
1
7330
7330
7331
7330
7331
2.532
2.218
1.881
1.753
1.687
2.413
2.157
1.821
1.733
1.692
2.282
2.056
1.791
1.698
1.681
Sample consists of 36,652 firm-years during the period 1990-2000, including 367
observations from distressed firms. Distressed firms are firms identified by CRSP
as having been delisted for performance reasons during the sample period. Year 0
is the annual report immediately preceding delisting. In Panel B, earnings are
controlled for by first forming decile portfolios based on EARN, then forming five
subportfolios within each earnings portfolio based on TACC. Portfolios in Panel B are
formed by grouping together the TACC subportfolios. For example, Portfolio 5 consists
consists of the five TACC subportfolios with the highest accruals.
Variables:
EARN
Earnings before interest and taxes divided by total assets
TACC
Total accruals divided by total assets
CR
Current ratio
36
TABLE 3
Logistic regressions of occurrence of financial distress
on earnings, accruals, cash flows, and control variables
36,652 firm-year observations from 1990-2000
Intercept
WC
RE
Control variables
MVE
S
EARN
Pred. Sign
-
-
-
-
-
-
Year 0
Model 1
Model 2
-4.790 ***
-4.987 ***
-1.800 ***
-1.806 ***
-0.472 ***
-0.446 ***
-0.228 ***
-0.232 ***
0.477 ***
0.448 **
-1.996 ***
-1.926 ***
0.492 **
0.546 ***
853.65 *** 3267.89
873.95 *** 3251.60
Year -1
Model 1
Model 2
-4.680 ***
-4.869 ***
-1.515 ***
-1.594 ***
-0.409 ***
-0.378 ***
-0.106 ***
-0.105 ***
0.381 ***
0.342 ***
-3.106 ***
-2.964 ***
0.566 ***
0.509 ***
620.97 *** 3500.58
642.26 *** 3483.29
Year -2
Model 1
Model 2
-4.737 ***
-4.927 ***
-0.435 *
-0.543 **
-0.380 ***
-0.350 ***
-0.026 *
-0.026 **
0.243 *
0.197 *
-3.059 ***
-2.856 ***
0.625 ***
0.489 **
322.59 *** 3798.95
346.19 *** 3779.36
Period
***, **, *
Accruals
LOWACC
HIGHACC
+
Likelihood
Ratio
AIC
+
Significant at <0.0001, 0.01, 0.05, respectively
Above are the results of logistic regressions. The dependent variable is an indicator variable
equal to 1 if the firm was financially distressed in Year 0. A firm is considered to be distressed
if CRSP indicates that it's stock was delisted for financial reasons. Year 0 is the annual
report preceding the delisting.
37
TABLE 3 (continued)
Explanatory Variables:
EARN
LOWACC
HIGHACC
WC
RE
MVE
S
Earnings before interest and taxes divided by total assets
Dummy variable equal to one if the firm's total accruals are in the lowest quintile of
total accruals (as of Year 2)
Dummy variable equal to one if the firm's total accruals are in the highest quintile of
total accruals (as of Year 2)
Working capital divided by total assets
Retained earnings divided by total assets
Market value of equity divided by total liabilities
Sales divided by total assets
38
TABLE 4
Descriptive statistics on 1,096 loans containing current ratio covenants
from the period 1990-1999
Fin. Distress
Standard
Lower
Upper
Sample
Variable
Mean
Deviation Quartile Median Quartile
Mean
EARN
0.030
0.134
0.006
0.043
0.083
0.010
TACC
-0.036
0.128
-0.086
-0.031
0.023
-0.064
CR
2.167
1.186
1.379
1.918
2.556
2.462
IOS
1.548
2.014
0.830
1.174
1.743
1.814
DEBT
0.294
0.230
0.111
0.265
0.433
0.241
251.702
525.565
38.880
93.730
250.260
1,658.600
0.389
1.253
0.128
0.242
0.432
n/a
40.720
30.181
18.000
36.000
60.000
n/a
CRCOV
1.413
0.447
1.100
1.300
1.500
n/a
SLACK
0.558
0.802
0.103
0.357
0.794
n/a
FIRMSIZE
LOANSIZE
TERM
Variable Definitions:
EARN
TACC
CR
IOS
DEBT
FIRMSIZE
LOANSIZE
TERM
CRCOV
SLACK
Net income scaled by total assets.
Total accruals at fiscal year-end preceding loan
inception, scaled by total assets.
Current ratio of borrowing firm.
Investment opportunity set of borrowing firm at fiscal year-end
preceding loan inception.
Total indebtedness of borrowing firm at fiscal year-end
preceding loan inception, scaled by total assets.
Total assets of borrowing firm at fiscal year-end
preceding
loan inception.
Dollar amount of loan scaled by total assets of borrowing firm.
Length of loan term in months.
Initial level of current ratio debt covenant.
(CR -CRCOV)/CRCOV
39
TABLE 5
Descriptive statistics by portfolio formed on total accruals at loan inception
1,096 loans from 1990-1999
Panel A: Portfolio means: Portfolio formed on total accruals at Year -2 (1=lowest)
Portfolio
N
TACC
EARN
IOS
TERM
DEBT
FIRMSIZE
LOANSIZE
CR
SLACK
5
219
0.116
0.080
2.062
37.416
0.232
157.926
0.396
2.464
0.695
4
219
0.011
0.049
1.583
37.164
0.266
242.769
0.345
2.393
0.657
3
219
-0.032
0.044
1.470
46.315
0.308
316.264
0.465
2.216
0.541
2
219
-0.074
0.036
1.314
43.290
0.314
320.021
0.367
2.141
0.557
1
221
-0.203
-0.058
1.312
39.683
0.349
222.495
0.376
1.634
0.343
Panel B: Portfolio means: Portfolios formed on total accruals controlling for earnings at Year -2 (1=lowest)
Portfolio
N
TACC
EARN
IOS
TERM
DEBT
FIRMSIZE
LOANSIZE
CR
SLACK
5
219
0.105
0.038
1.930
36.417
0.258
178.355
0.563
2.469
0.730
4
219
0.009
0.042
1.611
37.868
0.268
240.937
0.343
2.394
0.658
3
219
-0.037
0.035
1.363
44.427
0.314
272.893
0.293
2.164
0.511
2
219
-0.078
0.033
1.424
43.736
0.337
346.166
0.324
2.044
0.469
1
221
-0.178
0.003
1.409
41.081
0.292
219.442
0.425
1.772
0.421
40
TABLE 5 (continued)
Panel C: Portfolio medians: Portfolios formed on total accruals controlling for earnings at Year -2 (1=lowest)
Portfolio
N
TACC
EARN
IOS
TERM
DEBT
FIRMSIZE
LOANSIZE
CR
SLACK
5
219
0.093
0.043
1.095
32.000
0.242
63.955
0.249
2.136
0.485
4
219
0.013
0.043
1.262
35.000
0.258
86.555
0.224
2.060
0.458
3
219
-0.025
0.043
1.117
36.000
0.294
100.290
0.217
1.919
0.361
2
219
-0.062
0.044
1.222
36.000
0.305
154.725
0.224
1.801
0.331
1
221
-0.128
0.043
1.239
36.000
0.258
93.160
0.299
1.636
0.271
Variable Definitions:
TACC
Total accruals at fiscal year-end preceding loan
inception, scaled by total assets
EARN
Earnings for fiscal year-end preceding loan inception,
scaled by total assets
IOS
Investment opportunity set of borrowing firm at fiscal year-end preceding loan
inception
TERM
Length of loan term in months
DEBT
Total indebtedness of borrowing firm at fiscal year-end
preceding loan inception, scaled by total assets
FIRMSIZE
Total assets of borrowing firm at fiscal year-end preceding
loan inception
LOANSIZE
Dollar amount of loan scaled by total assets of borrowing firm
CR
Current ratio at fiscal year-end preceding loan inception
SLACK
(CR - CRCOV)/CRCOV
41
TABLE 6
Regression of current ratio covenant slack on
determinants of debt covenant slack
1,096 loans from the period 1990-1999
Predicted
Variable
Model 1
Sign
Coef.
t stat
INTERCEPT
+
0.604
17.49
EARN
+
0.567
LOWACC
-
HIGHACC
-
IOS
Model 2
Coef.
t stat
***
0.867
8.01
2.70
**
0.303
1.46
-0.140
-1.96
*
-0.136
-1.97 *
0.102
1.49
0.013
0.20
-
0.061
4.81
TERM
?
-0.0003
-0.04
DEBT
-
-0.857
-6.42 ***
FIRMSIZE
+
-0.022
-1.07
LOANSIZE
-
0.019
1.00
ADJ R2
0.022
***
***
0.107
***, **, * Significant at <0.0001, 0.01, 0.05, respectively
The dependent variable in each model is the level of initial slack in the current ratio debt covenant.
Variable Definitions:
SLACK
Initial level of slack in the current ratio debt covenant (CR - CRCOV)
EARN
Earnings for fiscal year-end preceding loan inception, scaled by total assets
LOWACC
Dummy variable equal to one if the firm's total accruals are in the lowest quintile of
total accruals immediately prior to loan inception
HIGHACC
Dummy variable equal to one if the firm's total accruals are in the highest quintile of
total accruals immediately prior to loan inception.
Investment opportunity set of borrowing firm at fiscal year-end preceding
IOS
loan
inception
TERM
Length of loan term in months
DEBT
Total indebtedness of borrowing firm at fiscal year-end
preceding loan inception, scaled by total assets
SIZE
Total assets of borrowing firm at fiscal year-end preceding loan inception
AMOUNT
Principal amount of the loan scaled by total assets at fiscal year end preceding
loan inception
42
Figure 1a
Financially Distressed Firms by Portfolio Formed on Total Accruals
36,652 firm-year observations from 1990-2000
Financially distressed firms (as % of portfolio)
2.5
2
1.5
1
0.5
0
1
2
3
4
5
Portfolios (7,330 or 7,331 obse rvations e ach) forme d on total accruals
(1=lowe st)
43
Figure 1b
Financially Distressed Firms by Portfolio Formed on
Total Accruals after Controlling for Earnings
36,652 firm-year observations from 1990-2000
1.6
Financially distressed firms (as % of portfolio)
1.4
1.2
1
0.8
0.6
0.4
0.2
0
1
2
3
4
5
Portfolios (7,330 or 7,331 obse rvations e ach) forme d on total accruals controlling for e arnings
(1 = lowe st)
44
Figure 2a
C urre nt Ratio Cove nant Slack by Portfolio Forme d on Total Accruals
1,096 loans from 1990-1999.
0.8
0.7
Current ratio covenant slack
0.6
0.5
0.4
0.3
0.2
0.1
0
1
2
3
4
5
Portfolios forme d on total accruals (1=lowe st)
45
Figure 2b
Current Ratio Covenant Slack by Portfolio Formed on
Total Accruals Controlling for Earnings
1,096 loans from 1990-1999
0.800
0.700
Current ratio covenant slack
0.600
0.500
0.400
0.300
0.200
0.100
0.000
1
2
3
4
5
Portfolios formed on total accruals controlling for earnings
(1=lowest)
46
Figure 3
Current Ratio Covenant Violations by Portfolio Formed on Total Accruals at Loan Inception
756 loans from the period 1990-1999
0.5
0.45
Violators (As % of Portfolio)
0.4
0.35
0.3
0.25
0.2
0.15
0.1
0.05
0
1
2
3
4
5
Portfolios Formed on Total Accruals at Loan Inception (1=Lowest)
Actual Covenant
Standardized Covenant
47