accruals, financial distress, and debt...

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

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.

2

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 1DeFond 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.

3

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

4

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

5

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 Z-

score) 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.

6

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

7

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

8

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. 3 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.

9

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 non-

distressed 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 non-

distressed 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.

10

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.

11

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.

12

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

13

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.

14

(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 I am grateful to Ilia Dichev for suggesting this procedure. 12 Untabulated tests performed on the sample after deleting outliers yield similar results.

15

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

16

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

17

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.

18

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-to-

cash 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.

19

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:

CRCOV

CRCOVCRSLACK −= (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,

20

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):

TABVPREFBVDEBTMVEIOS ++

= (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

21

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.

22

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.

23

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

24

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.

25

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-to-

cash 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-to-

cash 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

26

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.

27

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.

28

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 accounting-

based 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.

29

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 pricing-

based 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.

30

Gopalakrishnan, V., and Mohinder Parkash, 1995, Borrower and lender perceptions of accounting information in corporate lending agreements, Accounting Horizons, 9, 13-26.

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.

31

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.

32

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 0.271 0.256 0.242

Non-distressed 0.272 0.257 0.244 Distressed 0.234 0.129 0.043

t-stat 2.90 *** 9.61 *** 12.90 ***

RE All 0.029 0.008 -0.032

Non-distressed 0.036 0.017 -0.016 Distressed -0.589 -0.962 -1.565

t-stat 12.54 *** 14.98 *** 16.75 ***

EARN All 0.059 0.051 0.044

Non-distressed 0.060 0.053 0.046 Distressed -0.092 -0.167 -0.208

t-stat 14.04 *** 17.03 *** 17.63 ***

MVE All 4.640 4.195 3.938


t-stat 1.55 8.23 *** 12.92 ***

S All 1.235 1.236 1.238


t-stat -0.36 -3.19 ** -6.05 ***

33

TABLE 1 (continued)

Panel B: Other variables

Year -2 Year -1 Year 0

TACC

All -0.043 -0.053 -0.062 Non-distressed -0.042 -0.051 -0.060

Distressed -0.073 -0.168 -0.255 t-stat 3.52 *** 8.46 *** 8.34 ***

TL All 0.491 0.503 0.515


t-stat -3.50 *** -9.58 *** -13.30 ***

CR All 2.600 2.468 2.388


t-stat 1.03 6.91 *** 6.40 *** *, **, *** 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 Working capital divided by total assets RE Retained earnings divided by total assets EARN Earnings before interest and taxes divided by total assets MVE Market value of equity divided by total liabilities S Sales divided by total assets TACC Total accruals divided by total assets TL Total liabilities divided by total assets CR Current ratio

34

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 7330 0.079 0.046 0.028 4 7330 0.080 0.067 0.057 3 7331 0.078 0.070 0.065 2 7330 0.070 0.064 0.059 1 7331 -0.012 0.009 0.008 TACC 5 7330 0.096 -0.018 -0.045 4 7330 -0.003 -0.035 -0.047 3 7331 -0.039 -0.049 -0.053 2 7330 -0.073 -0.063 -0.068 1 7331 -0.193 -0.099 -0.097 CR 5 7330 3.261 3.054 2.900 4 7330 2.914 2.703 2.600 3 7331 2.351 2.220 2.157 2 7330 2.300 2.195 2.170 1 7331 2.177 2.172 2.156

35

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 7330 0.059 0.028 0.012 4 7330 0.059 0.046 0.037 3 7331 0.059 0.052 0.048 2 7330 0.057 0.058 0.055 1 7331 0.048 0.058 0.051 TACC 5 7330 0.098 -0.022 -0.049 4 7330 -0.003 -0.038 -0.053 3 7331 -0.043 -0.053 -0.058 2 7330 -0.085 -0.069 -0.071 1 7331 -0.185 -0.091 -0.091 CR 5 7330 3.319 3.052 2.888 4 7330 3.184 2.914 2.732 3 7331 2.603 2.486 2.433 2 7330 2.301 2.238 2.212 1 7331 2.043 2.059 2.058

36

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 7330 0.081 0.067 0.060 4 7330 0.081 0.076 0.072 3 7331 0.081 0.079 0.078 2 7330 0.081 0.081 0.081 1 7331 0.081 0.084 0.081 TACC 5 7330 0.069 -0.008 -0.022 4 7330 -0.005 -0.027 -0.033 3 7331 -0.038 -0.041 -0.042 2 7330 -0.067 -0.055 -0.054 1 7331 -0.130 -0.074 -0.071 CR 5 7330 2.532 2.413 2.282 4 7330 2.218 2.157 2.056 3 7331 1.881 1.821 1.791 2 7330 1.753 1.733 1.698 1 7331 1.687 1.692 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

37

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

Control variables Accruals Likelihood Period Intercept WC RE MVE S EARN LOWACC HIGHACC Ratio AIC

Pred. Sign - - - - - - + +

Year 0 Model 1 -4.790 *** -1.800 *** -0.472 *** -0.228 *** 0.477 *** -1.996 *** 853.65 *** 3267.89 Model 2 -4.987 *** -1.806 *** -0.446 *** -0.232 *** 0.448 ** -1.926 *** 0.492 ** 0.546 *** 873.95 *** 3251.60 Year -1 Model 1 -4.680 *** -1.515 *** -0.409 *** -0.106 *** 0.381 *** -3.106 *** 620.97 *** 3500.58 Model 2 -4.869 *** -1.594 *** -0.378 *** -0.105 *** 0.342 *** -2.964 *** 0.566 *** 0.509 *** 642.26 *** 3483.29 Year -2 Model 1 -4.737 *** -0.435 * -0.380 *** -0.026 * 0.243 * -3.059 *** 322.59 *** 3798.95 Model 2 -4.927 *** -0.543 ** -0.350 *** -0.026 ** 0.197 * -2.856 *** 0.625 *** 0.489 ** 346.19 *** 3779.36 ***, **, * 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.

38

TABLE 3 (continued) Explanatory Variables: EARN Earnings before interest and taxes divided by total assets LOWACC Dummy variable equal to one if the firm's total accruals are in the lowest quintile of total accruals (as of Year 2) HIGHACC Dummy variable equal to one if the firm's total accruals are in the highest quintile of total accruals (as of Year 2) WC Working capital divided by total assets RE Retained earnings divided by total assets MVE Market value of equity divided by total liabilities S Sales divided by total assets

39

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 FIRMSIZE 251.702 525.565 38.880 93.730 250.260 1,658.600 LOANSIZE 0.389 1.253 0.128 0.242 0.432 n/a TERM 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 Variable Definitions: EARN Net income scaled by total assets. TACC Total accruals at fiscal year-end preceding loan inception, scaled by total assets. CR Current ratio of borrowing firm. IOS Investment opportunity set of borrowing firm at fiscal year-end preceding loan inception. 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. TERM Length of loan term in months. CRCOV Initial level of current ratio debt covenant. SLACK (CR -CRCOV)/CRCOV

40

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

41

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

42

TABLE 6 Regression of current ratio covenant slack on

determinants of debt covenant slack 1,096 loans from the period 1990-1999

Predicted Model 1 Model 2

Variable Sign Coef. t stat Coef. t stat INTERCEPT + 0.604 17.49 *** 0.867 8.01 *** EARN + 0.567 2.70 ** 0.303 1.46 LOWACC - -0.140 -1.96 * -0.136 -1.97 * HIGHACC - 0.102 1.49 0.013 0.20 IOS - 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.

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 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

43

Figure 1aFinancially Distressed Firms by Portfolio Formed on Total Accruals

36,652 firm-year observations from 1990-2000

0

0.5

1

1.5

2

2.5

1 2 3 4 5

Portfolios (7,330 or 7,331 observations each) formed on total accruals(1=lowest)

Fina

ncia

lly d

istre

ssed

firm

s (as

% o

f por

tfol

io)

44

Figure 1bFinancially Distressed Firms by Portfolio Formed on

Total Accruals after Controlling for Earnings36,652 firm-year observations from 1990-2000

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

1 2 3 4 5

Portfolios (7,330 or 7,331 observations each) formed on total accruals controlling for earnings(1 = lowest)

Fina

ncia

lly d

istre

ssed

firm

s (as

% o

f por

tfol

io)

45

Figure 2a Current Ratio Covenant Slack by Portfolio Formed on Total Accruals

1,096 loans from 1990-1999.

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

1 2 3 4 5

Portfolios formed on total accruals (1=lowest)

Cur

rent

rat

io c

oven

ant s

lack

46

Figure 2b Current Ratio Covenant Slack by Portfolio Formed on

Total Accruals Controlling for Earnings1,096 loans from 1990-1999

0.000

0.100

0.200

0.300

0.400

0.500

0.600

0.700

0.800

1 2 3 4 5

Portfolios formed on total accruals controlling for earnings(1=lowest)

Cur

rent

rat

io c

oven

ant s

lack

47

Figure 3Current Ratio Covenant Violations by Portfolio Formed on Total Accruals at Loan Inception

756 loans from the period 1990-1999

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

0.45

0.5

1 2 3 4 5

Portfolios Formed on Total Accruals at Loan Inception (1=Lowest)

Vio

lato

rs (A

s % o

f Por

tfol

io)

Actual Covenant Standardized Covenant

accruals, financial distress, and debt...

Documents