Conservatism of the big six audit firms and going-concern modified audit reports.

Geiger, Marshall A. ; Raghunandan, K.

ABSTRACT

This study examines whether Big 6 audit farms exhibit a more conservative reporting posture than non-Big 6 audit firms by examining differences in "reporting errors "for going concern uncertainties. Specifically, based on an analysis of relative error costs, we hypothesize that Big 6 firms are more conservative in their reporting decisions, leading to differences in error proportions for both type 1 (companies receiving going-concern modified opinions that do not fail) and type II (failed companies that did not receive a prior going-concern modification) reporting errors. Evidence from samples of companies with first-time going-concern modified opinions and bankrupt companies is generally consistent with the hypotheses. Big 6 firms were found to have a significantly higher type I reporting error rate when failure was defined as filing for bankruptcy or entering default on debt payments, and a significantly lower type II reporting error rate than non-Big 6 firms. Our findings suggest that reporting decisions are associated with the relative magnitude of the firms' expected losses, resulting in more conservative reporting decisions regarding the issuance of going -concern modified audit reports for Big 6 firms.

INTRODUCTION

Prior research has addressed the existence of differences between Big 6 and non-Big 6 (1) audit firms on a variety of dimensions, including audit quality, audit fees, client retention and industry concentration (DeAngelo, 1981; Palmrose, 1986; Francis & Simon, 1987; Simunic & Stein, 1996; Gul, 1999; Hogan & Jeter, 1999). Some recent studies have also examined whether the largest audit firms are more conservative than smaller audit firms in various settings. Recent papers examining issues related to the market for audit services have documented that Big 6 firms are (a) less concentrated in industries perceived as having high litigation risk compared to those perceived as having low litigation risk (Hogan & Jeter, 1999), (b) less likely to accept clients where the predecessor auditor resigned (Raghunandan & Rama, 1999), and (c) have maintained historically less risky client portfolios than non-Big 6 auditors (Francis & Reynolds, 1999). While these studies have examined several facets of the auditor-client interaction, no study has examined Big 6 firms compared to non-Big 6 firms for conservatism in their resultant reporting decisions--the final outcome of the audit process.

Audit reporting, and particularly reporting on going-concern uncertainties, continues to remain an important issue for U. S. legislators and the public accounting profession. This is evidenced by the fact that the U. S. recently enacted the Private Securities Litigation Reform Act which provides the first and only instance where a specific auditing procedure, related to reporting on going-concern uncertainties, has been mandated by law. This study provides additional evidence on the conservatism of the Big 6 firms by empirically examining reporting decisions regarding going-concern.

An analysis of an audit reporting model, developed below, suggests that rational economic behavior on the part of Big 6 auditors would lead to more conservative audit opinions, after controlling for other going concern related factors. Our findings support this analysis and suggest that Big 6 firms generally do exhibit more conservative reporting decisions related to issuing going-concern modified reports. Specifically, we find that type I reporting errors (defined as instances where a company receives a going-concern modified opinion but does not subsequently fail), as expected, were higher for the Big 6 than the non-Big 6 firms when failure was defined as entering bankruptcy or payment default on debt. We also find that type II reporting errors (defined as instances where a company enters into bankruptcy without a prior going-concern modified opinion), as expected, were significantly lower for the Big 6 firms.

The remainder of the paper is organized as follows: Section II presents the development of the audit reporting model and hypotheses tested, Section III discusses the research method, Section IV presents the results, and Section V provides a summary of the study and a discussion of the research limitations and directions for fixture research.

DEVELOPMENT OF THE AUDIT REPORTING MODEL AND HYPOTHESES

Error Costs and Audit Reporting

Congressional hearings in the U.S. have criticized auditors for not providing adequate early warning of impending client failures (U.S. House of Representatives 1985, 1990). Legislators' continuing concern with this issue is evidenced by the fact that the Private Securities Litigation Reform Act (1995) codified into law the audit reporting requirements related to going-concern uncertainties specified in Statement on Auditing Standards (SAS) No. 59 (AICPA 1988). (2)

SAS No. 59 and the Reform Act require auditors to evaluate the continuing existence of every client for a period of one year from the date of the statements being audited. If, after considering management's plans and mitigating circumstances, the auditor still has substantial doubt about the ability of an entity to continue as a going concern, then the audit report needs to be modified to reflect such uncertainty. Thus, under current professional and legislative reporting requirements, the going-concern assessment remains subject to professional judgment. There are two types of errors that are relevant in the context of audit reporting on going concern uncertainties. A "type I reporting error" occurs when the auditor issues a going-concern modified opinion but the client does not subsequently fail. A "type II reporting error" occurs when a client files for bankruptcy without a prior going-concern modified audit report. Note that under audit reporting standards related to going concern uncertainties (i.e., SAS No. 59), the auditor is not required to predict the future status of the client. Auditors need only identify circumstances that create substantial doubt about the ability of the client to remain viable for the ensuing year based ion available information at the time of the audit report. Consequently, the two types of misclassifications described above are not strictly interpretable as errors. However, as noted by McKeown et al. (1991b) and Chen and Church (1992), clients, financial statement users and the public are likely to perceive both a going-concern modified opinion without subsequent failure, and a bankruptcy without a prior modification, as audit reporting errors.

Clients do not welcome the receipt of a going-concern modified audit report under most circumstances (Kida, 1980; Mutchler, 1984). This is particularly true if these modified reports are viewed by the company as unwarranted based on their continued viability. Thus, while not strictly interpretable as an error, these companies may express displeasure with their audit firm by taking actions such as switching to a different auditor (Mutchler, 1984; Chen & Church, 1992; Geiger et al., 1998). Let the perceived cost associated with this type of reporting error (i.e., issuing a going-concern modified report when the client subsequently does not fail) be [C.sub.I]. We assume that [C.sub.I] increases with the magnitude of the dollar cost directly associated with a type I reporting error, which would include, but is not limited to, lost revenues from providing services to a client who switches to a new auditor. (3)

When a client files for bankruptcy without a prior going-concern modified opinion (i.e., a type II error), there may be a greater likelihood that the auditor will face litigation related costs and suffer a loss of reputation. Carcello and Palmrose (1994) studied bankruptcy related lawsuits against auditors, and found that auditors who had issued modified reports prior to client bankruptcy had the highest dismissal rate and lowest payments, and that auditors who did not issue prior modified reports had the highest payouts. They noted that "overall, our evidence suggests a defensive role for timely modified reporting" (p. 3). Let the perceived cost associated with this type of reporting error (i.e., not issuing a going-concern modified report when the client subsequently fails) be [C.sub.II]. We assume that [C.sub.II] increases with the magnitude of the dollar cost directly associated with a type II reporting error, which would include, but is not limited to, costs associated with litigation against the auditor.

In our model, and per SAS No. 59, the audit firm compares the likelihood of the client's ability to continue as a going concern against a judgment threshold for "substantial doubt." Let p represent the audit firm's assessed probability of the client's inability to continue as a going-concern (or the likelihood of client failure). If the indifference probability of failure for substantial doubt is [p.sup.*], then the audit firm will issue a going-concern modified opinion if p > [p.sup.*]. If p < [p.sup.*], then the audit firm issues a report without the going-concern modification. If the indifference probability ([p.sup.*]) is lower, then the likelihood of issuing a going-concern modified report, ceteris paribus, is higher. (4)

As noted by Raghunandan and Rama (1995), under such a reporting model [p.sup.*] = 1 / [1 + ([C.sub.II]/[C.sub.I])]. Thus, as C increases or as C decreases, the threshold for issuing a going-concern modified audit report ([p.sup.*]) decreases. Hence, as C increases or as C decreases, the likelihood of issuing a going-concern modified report increases.

Existing audit reporting standards, however, do not specify that [p.sup.*] is to be influenced by the relative magnitudes of an audit firm's perceived error costs. Thus, a strict interpretation of the standards would suggest that [p.sup.*] should be independent of either [C.sub.II] or [C.sub.I]. However, an audit firm trying to minimize its costs/losses would consider the relative magnitude of [C.sub.II] or [C.sub.I] in decisions about going-concern report modifications. We now consider the impact of audit firm size (i.e., Big 6 versus non-Big 6) on the relative magnitude of [C.sub.II] and [C.sub.I], and hence on [p.sub.*].

Reporting Decisions: Big 6 and Non-Big 6 Audit Firms

Previous empirical research on the relation between Big 6/non-Big 6 membership and audit quality has primarily focused on the pricing of audit related services (e.g., Simunic, 1980; Francis & Simon, 1987; Beatty, 1993), on auditor switching behavior (e.g., Nichols & Smith, 1983; Francis & Wilson, 1988; Johnson & Lys, 1990; DeFond, 1992), information integrity (Teoh & Wang, 1993; Francis et al., 1999) and industry concentration (Hogan & Jeter, 1999).

Many prior papers have also examined issues related to audit opinions, including the association between distressed companies, audit opinions, and bankruptcies (e.g., Carcello et al., 1997; Mutchler et al., 1997; Louwers, 1998; Foster et al., 1998). A recent study by Morris and Strawser (1999) found that bank regulators act as if the believe the Big 6 audit firms are more conservative than non-Big 6 audit firms in their reporting decisions regarding troubled banks. However, these studies have not specifically examined whether Big 6 conservatism actually translates into conservative reporting with regard to troubled and bankrupt companies. (5)

There are two reasons why audit firm size may impact going-concern reporting decisions. The reasons are that (1) the "wealth at risk" for the two types of firms is different, and (2) the client risk due to portfolio diversification is different for the two types of firms. These two arguments are discussed in turn below.

Wealth at Risk Argument

The wealth at risk argument suggests that Big 6 audit firms have more wealth with which to pay plaintiffs in lawsuits filed against auditors than do non-Big 6 firms. Accordingly, Big 6 firms have more "wealth at risk" in any particular audit engagement (Simunic & Stein, 1987, 1996; Menon & Williams, 1999). Hence, large audit firms stand to suffer disproportionately greater losses for any given type II error and resultant lawsuit, regardless of the audit fee generated. This argument, then, suggests that, ceteris paribus, Big 6 firms are more likely to suffer greater lawsuit related losses for a given type II reporting error. (6) Consequently, [C.sub.II] from a type II reporting error may be larger for a 6 Big 6 firm for any given client.

Conversely, with respect to a type I misclassification, the revenues from any given client constitute a proportionately lower share of the overall firm revenue for a Big 6 firm than for a non -Big 6 firm. Thus, the loss of the client, or the threat of the loss of a client, upon receipt of what is perceived as an unwarranted going-concern report, would have a proportionately smaller adverse impact on the wealth of a Big 6 firm than on a non-Big 6 firm. Thus, [C.sub.I] from a type I reporting error may be proportionately smaller for a Big 6 firm for any given client.

Together, the above discussion of the wealth at risk argument suggests that the ratio of [C.sub.II]/[C.sub.1] would be greater for a Big 6 firm than for a non-Big 6 firm and hence would lead to p* being lower for a Big 6 firm. This would result in Big 6 audit firms reporting more conservatively and issuing going-concern modified audit reports to clients that are less likely to receive a similar opinion from a non-Big 6 audit firm. Thus, if [p.sup.*] (Big 6) < p < [p.sup.*] (non-Big 6) (where p is the assessed probability of failure for a given client), a Big 6 firm would issue a going-concern qualified audit opinion but a non-Big 6 firm would not issue such a modified opinion.

In the context of going concern there are two possible future outcomes for the client-failure or non-failure. Given the reporting decisions above, if in the future the client does not fail, then the Big 6 firm has a type I reporting error but not the non-Big 6 firm. If the client fails, then the non-Big 6 firm has a type II reporting error but not the Big 6 firm. It follows, then, that the wealth at risk argument leads to the following propositions: (1) the proportion of going-concern modified reports where the client does not subsequently fail (type I error) will be higher for Big 6 audit firms, and (2) the proportion of bankruptcies without a prior going-concern modified opinion (type 11 error) would be lower for Big 6 audit firms.

Portfolio Diversification Argument

Simunic and Stein (1990) suggest that an appropriate analysis of audit risk must also consider the client portfolio of an audit firm. The portfolio diversification argument stems from this analysis and is based on the fact that the Big 6 firms have a larger portfolio of clients. Hence the Big 6 firms may be able to diversify their risk related to any one particular audit client's failure more easily than a non-Big 6 firm. A type II misclassification, with a specific magnitude of litigation loss, can more easily lead to failure of a smaller audit firm since the pool of clients in their portfolio may be less likely to be sufficient to cover the lawsuit related losses. (7) Thus, the same misclassification may be less likely to lead to the failure of the well-diversified Big 6 firm with its large portfolio of clients. Consequently, under the portfolio diversification argument, if the absolute dollar cost of a type II error is the same for a Big 6 and a non-Big 6 firm, the perceived cost [C.sub.II] may be higher for a non-Big 6 firm than for a Big 6 firm, leading to more conservative reporting by the non-Big 6 firms.

However, under the portfolio diversification argument, the effect of differences in [C.sub.I] based on audit firm size works in the opposite direction than [C.sub.II] error costs. Because the portfolio of a Big 6 firm is much more diversified than that of a non-Big 6 firm, the impact of lost revenues from a displeased client will have a lower impact on a Big 6 firm than on a non-Big 6 firm. Hence, the perceived cost [C.sub.I] may be higher for non-Big 6 firms than for Big 6 firms. (8)

Thus, under the portfolio diversification argument, Big 6 audit firms may be less conservative due to their ability to diversify increased risk through a larger portfolio for clients, or may be more conservative due to the reduced influence of a single client on the entire firm. This is because both [C.sub.I] and [C.sub.II] are higher for non-Big 6 firms under the portfolio diversification argument, which in turn means that it is not possible to make a directional prediction of conservatism regarding the differences of the Big 6 and non-Big 6 based on the portfolio diversification argument.

Hypotheses

Ex-ante we believe that the wealth at risk argument would dominate the portfolio diversification argument, resulting in a more conservative reporting posture by the Big 6 audit firms. This is consistent with Menon and Williams (1999) and Morris and Strawser (1999) who argue that litigation risk plays an important role in the audit firm's decisions, and that Big 6 firms face a higher type II error cost than non-Big 6 firms, leading them to make conservative decisions. The empirical evidence in Hogan and Jeter (1999) and Raghunandan and Rama (1999), indicating that Big 6 auditors are less likely to accept risky clients, also support the suggestion that Big 6 firms are more conservative and that the wealth at risk argument may dominate the portfolio diversification argument. Accordingly, with respect to audit opinions, we expect that the wealth at risk of the Big 6 firms will lead to more conservative going-concern reporting by the Big 6 firms compared to the non-Big 6 firms, which leads to our hypotheses:

[H.sub.01]: Type I error rates are higher for going-concern opinion decisions of Big 6 firms.

[H.sub.02]: Type II error rates are lower for bankrupt clients of Big 6 firms.

METHOD

Hypothesis One

Research Design

Hypothesis one is concerned with the resolution of going concern uncertainties for companies receiving going-concern modified opinions, and examines whether subsequent failure rates are different for clients of Big 6 and non-Big 6 audit firms. We would expect that if Big 6 firms are more conservative with respect to their reporting decisions, that they would experience higher type I reporting errors due to their increased propensity to issue modified reports to less stressed companies.

A set of companies receiving first-time going-concern modified audit reports was identified to test hypothesis one. Kida (1980) and Mutchler (1984) have documented the reluctance of audit firms to both issue a going-concern modification, and then to remove the modification once it has been rendered to a specific client. Mutchler and Williams (1990) noted that "the fact that the company already has the qualification becomes a variable in the auditor's decision model, and thus, these observations represent a different type of decision for the auditor. In addition, it can be argued that auditors risk preferences are different for these companies, i.e., the auditor may no longer have to consider the risk of losing the client if a going-concern opinion is issued". Hence, we consider only companies receiving first-time going-concern opinion modifications to test hypothesis one.

Companies in greater financial stress may be more likely to fail or enter into bankruptcy than non-stressed companies. Hence, we use level of financial stress as a control variable. There are many 0different financial distress models in the accounting literature. In this study, we use the probability of bankruptcy (PRB) calculated from Zmijewski's (1984) model as a surrogate for financial stress in evaluating hypothesis one, because (a) it has been used widely in a variety of auditing contexts (Bamber et al., 1993; Chen & Wei, 1993; Wheeler et al., 1993; Carcello et al., 1995, 1999), (b) it is parsimonious, which means fewer observations have to be deleted for lack of available data, and (c) it is free of some statistical problems associated with other models.

Prior research suggests that company size influences the going-concern modified audit opinion decision, and that companies in default may be more likely to receive a going-concern modified audit opinion, even after controlling for financial stress (McKeown et al., 1991a; Chen & Church, 1992). Although we are assessing subsequent failure of firms receiving a first-time going -concern modified report and not the reporting decision itself, we use company size, as measured by the natural log of sales, and default status as control factors in the analysis.

Carcello et al. (1999) examined going-concern modified reporting and partner compensation plans and found a significant positive effect for the implementation of SAS No. 59 on audit reporting decisions. Since our data include going-concern modified opinions from the period before and after the required implementation of SAS No. 59, we also include reporting time period (pre-or post-SAS No. 59) as a control factor in our analysis.

The relationship between subsequent failure and the factors discussed above is examined using a logistic regression to estimate the coeffcients in the following model:

[GC.sub.BANK] = b0 + b1 * PRB + b2 * dft + b3 * LNSL + b4 * TIME + b5 * BIG6

where

[GC.sub.BANK] = Bankrupt (1 if yes, 0 if not) given a first-time going-concern modified opinion,

PRB = Probability of bankruptcy, calculated from Zmijewski's (1984) model,

DFT = 1 if in default, 0 otherwise,

LNSL = Natural log of sales (in thousands of dollars),

TIME = 1 if post-SAS No. 59 time period, 0 otherwise, and

BIG6 = 1 if a Big 6 firm, 0 otherwise.

Filing for bankruptcy is a stringent definition of company failure (Carcello & Palmrose, 1994). Mutchler and Williams (1990) use an alternative definition of failure by also including companies that enter into payment default on debt during the period subsequent to the receipt of their going-concern report. Thus, an additional definition of failure, based on bankruptcy filing or entering into payment default (BAD), was also used to assess these first-time going-concern modified report recipients. Accordingly, we also test hypothesis one by using this broader definition of failure, for the sub-sample of companies not already in payment default at the time of the receipt of the going-concern modified audit report. (10)

Data Collection

A sample of companies receiving first-time going-concern reports is needed to test hypothesis one. It is possible that industry characteristics could influence the relative threshold ([p.sup.*]) for going -concern modified audit reports. In particular, audits of companies in sectors such as banking, other financial services and real estate have been considered to be particularly risky for all audit firms. In addition, prior researchers examining issues related to going-concern modified audit reporting have typically excluded such firms, as well as those in other regulated industries (Mutchler, 1985; McKeown et al., 1991a). Consequently, we limited our analyses to manufacturing firms (SIC codes 20 through 39).

Compact Disclosure (CD)-SEC was our source for identifying all manufacturing companies with first-time going-concern modified reports during the period 1987 through 1994. Financial and default data were obtained from CD-SEC, and from annual reports and 10-ks on Laser Disclosure and LEXIS-NEXIS.. We excluded companies with fiscal year-ends in 1989, however, because audit firms exhibited anomalous reporting behavior in the SAS No. 59 transition period (Carcello et al., 1997) and there were differences in the timing of the adoption of SAS No. 59 by different audit firms (Spires & Williams, 1990).

The following sources were used to identify the subsequent year resolutions of going concern uncertainties faced by the firms in our sample: 1) Wall Street Journal Index, 2) CD-SEC, 3) Bloomberg Financial Services, 4) Predicast's Index of Corporate Change, 5) LEXIS-NEXIS, and 6) a list of public company bankruptcies obtained from New Generation Research Inc., publishers of the yearly Bankruptcy Almanac. Only firms for which subsequent financial statements were found, and not in bankruptcy, were designated as non-failed firms. Based on these data requirements, our sample consisted of 533 companies with first-time going-concern modified reports.

As noted earlier, we also examined hypothesis one by using an expanded definition of failure that includes bankruptcy and entering into payment default on debt (BAD). Since our initial sample includes companies that were in payment default at the time they received their initial going-concern modified report, we eliminated these 98 companies for this test. Accordingly, this analysis was performed on the 435 companies not in payment default at the time they received their first-time going-concern modified report.

Hypothesis Two

Research Design

Hypothesis two is concerned with the proportion of bankruptcies with prior going-concern modified reports. We would expect that if Big 6 firms were more conservative in reporting on going concern that they would have fewer type II errors than non-Big 6 firms due to issuing more going-concern modified opinions prior to bankruptcy. To test this hypothesis, a set of bankrupt companies was identified. Again, to control for other audit reporting related factors, a multivariate logistic regression was used.

Prior research suggests that a going-concern modified audit report is more likely when (1) there is greater financial stress, (2) there is debt default, (3) the bankruptcy lag (time period from audit report date to bankruptcy date) is smaller, (4) the audit reporting lag (time period from fiscal year end to audit report date) is larger, and (5) the company is smaller (McKeown et al., 1991a; Chen & Church, 1992; Raghunandan & Rama, 1995; Mutchler et al., 1997). Hence, these variables are included as control factors. In summary, to test hypothesis two we estimate the coefficients in the following logistic regression:

[BANK.sub.GC] = b0 + b1 * PRB + b2 * DFT + B3 * BKLAG * b4 * REPLAG + b5 * LNSL + b6 * BIG6,

where

[BANK.sub.GC] = 1 if audit report prior to bankruptcy is going-concern modified; else 0,

PRB = Probability of bankruptcy, calculated from Zmijewski's (1984) model,

DFT = 1 if company was in default at the time of opinion; 0 otherwise,

BKLAG = Time, in square root of days from audit report date to bankruptcy date,

REPLAG = Time, in square root of days from fiscal year end to audit report date,

LNSL = Natural log of sales (in thousands of dollars), and

BIG6 = 1 if the auditor was a Big 6 firm, 0 otherwise.

Data Collection

To test hypothesis two, a fist of public company bankruptcies was obtained from New Generation Research Inc. Relevant financial statement and auditor data were obtained from CD -SEC, and from annual reports and 10-k filings (on Laser Disclosure). We deleted companies in the banking, other financial, real estate and regulated industries. This is because such companies have unique financial characteristics and are not amenable to the measure of financial stress used in our analyses, and/or the error costs for auditors could be different in regulated industries compared to other industries. We were able to obtain complete data for 247 companies which entered into bankruptcy over the period 1991-1995. (11)

RESULTS

Hypothesis One

Table 1 presents descriptive statistics for the sample of companies with first-time goingconcern modified reports that were also used to test hypothesis one. The Big 6 audited 343 companies in our sample, compared to 190 audited by the non-Big 6. As seen in the table, only 12.4 percent (66 of 533) of the companies that received a first-time going-concern modified opinion entered into bankruptcy in the subsequent year. This finding is consistent with the survival rates found in prior studies examining going-concern modified report recipients (Altman, 1982; Mutchler & Williams, 1990). The mean PRB scores for clients of Big 6 and non-Big 6 auditors were 0.66 and 0.73, respectively. The proportions of companies in default at the time they received their first-time going-concern opinion were 45.8 percent for the Big 6 and 26.8 percent for the non-Big 6. In addition, the LNSL values for clients of Big 6 and non-Big 6 auditors were 9.48 and 7.71, respectively. Fifty-one of the 343 companies (14.8 percent) receiving a first-time going-concern modified audit report from a Big 6 firm entered into bankruptcy within one year. The corresponding bankruptcy proportion for the 190 clients of non-Big 6 firms was 7.9 percent (15 companies).

The last two columns of Table 1 provide descriptive information on the companies not in payment default when they received their initial going-concern audit report that were used in the BAD analysis. Of the 435 initial non-payment-default companies, 24.8 percent (85 companies) entered into either bankruptcy or payment default within one year. Forty-seven of the 271 companies (17.3 percent) audited by the Big 6 entered bankruptcy or payment default, while 38 of the 164 companies (23.2 percent) audited by the non-Big 6 entered into bankruptcy or payment default within one year.

Results from the logistic regression for bankruptcy are presented in Table 2. The overall bankruptcy regression is significant (chi-square = 34.25 with 5 d.f., significant at p<.0001) and the c-statistic, measuring the extent of association between predicted probabilities and observed responses, is 0.71. In the regression, the coefficients on the PRB and LNSL variables are in the expected directions and are statistically significant. The coefficient for the DFT variable is in the expected direction but is not significant for this group of already financially stressed companies. The coefficient for the TIME variable (for pre-post-SAS No. 59) is positive and significant, consistent with the findings of Carcello et al. (1999). The Big 6 variable, however, is not significant in this analysis of company failure defined as subsequently filing for bankruptcy.

Results of the logistic regression considering BAD status (entering bankruptcy or payment default) as the definition of company failure for the reduced sample of 435 companies is also reported in Table 2. As seen in the table, the model chi-square of 39.28 is significant at p < .0001 with 4 d.f, and has a c-statistic of 0.71. In addition, and consistent with the bankruptcy analysis, all of the remaining control variables are positive and significant at p-values below .05. Using BAD as a definition of company failure, the BIG6 variable was negative and significant (p = .0061). This result indicates that the Big 6 audit firms had a lower proportion of going-concern report recipients who entered bankruptcy or payment default, and thus had a higher proportion of type I errors. Thus, when failure is defined as entering into bankruptcy or newly entering into payment default, type I errors are higher for Big 6 audit firms relative to non-Big 6 audit firms. This finding is consistent with hypothesis one and suggests that the Big 6 audit firms have higher type I error rates than non-Big 6 firms for this broader definition of company failure. (12)

Hypothesis Two

Table 3 provides descriptive statistics about the set of bankrupt companies used in this study. As seen in the last row of Table 3, 57 percent (121 of 211) of the bankrupt companies with a Big 6 auditor received a prior going-concern modified audit report, and 42 percent (15 of 36) of the bankrupt companies with a non-Big 6 auditor received a prior going-concern modified audit report.

Table 4 provides the results from the logistic regression with audit report type as the dependent variable. The overall model is significant (chi-square = 156.0, with 6 d.f., significant at p < .0001) and the c-statistic is 0.91. The coefficients for financial distress, default status, bankruptcy lag, and company size are in the expected directions and are significant. These results indicate that auditors were more likely to have issued prior going-concern modified reports for companies in greater financial stress, companies in default, companies with a shorter bankruptcy lag and smaller companies. The coefficient for the BIG6 variable is significant (p = .031), which supports the second hypothesis. Big 6 audit firms were more likely than non-Big 6 firms to have issued a prior going-concern modified report for companies that subsequently declared bankruptcy. (13)

Additional Analyses

ANCOVA Model

If the Big 6 firms are more conservative, and hence more likely to issue a going-concern modified audit opinion (i.e., exhibit a lower [p.sup.*]), then it is likely that firms receiving a first-time going-concern modified audit opinion would, on average, exhibit lower financial stress. To examine if companies receiving a going-concern modified opinion from Big 6 and non-Big 6 firms may be different with respect to their level of financial stress, we performed an unbalanced ANCOVA analysis on the sample of firms receiving a first-time going-concern modified audit opinion. The dependent variable was the probability of bankruptcy (PRB), Big 6 classification was the independent class variable, and the log of sales (LNSL) and default status (DFT) were used as two covariates.

Results of the ANCOVA indicate that the overall model is significant (p<.01), along with both covariates, and that the mean covariate adjusted PRB score for clients of the Big 6 firms (.666) was significantly lower than that for clients of non-Big 6 firms (.720) (p < .05). Thus, these results provide further evidence indicating that, ceteris paribus, Big 6 firms were generally more conservative and had lower indifference probabilities ([p.sup.*]) than non-Big 6 firms with respect to companies receiving their first-time going-concern modified audit report.

Strategic Bankruptcies

Our analysis for hypothesis two includes all bankruptcies, irrespective of the nature of the bankruptcy. It is possible that some bankruptcies may be strategic, and not as a result of financial distress. To ensure that the results are not being driven by such strategic bankruptcies, we deleted companies which did not have at least one of the following indicators of financial stress: (1) negative net income, (2) negative working capital and (3) negative stockholder's equity. (14) When the analysis was performed with this sub-sample, the BIG6 variable continued to remain significant (p < .05), and the significance of the other variables was substantively similar to the results presented in Table 4.

SUMMARY AND DISCUSSION

This study examined whether the Big 6 audit firms in the U. S. are more conservative than non-Big 6 firms in their reporting decisions regarding going concern. An analysis of the subsequent viability status of 533 first-time going-concern recipient companies indicates that when client failure is defined as bankruptcy only, there was no significant difference in type I error rates between Big 6 and non-Big 6 firms. However, when client failure is defined as bankruptcy or entering into payment default, the Big 6 firms had significantly higher type I error rates. Results from the analysis of 247 bankrupt firms indicates that, after controlling for financial and reporting factors, Big 6 audit firms were more likely to have issued a prior going-concern modified opinion. (15) Additional analyses also found that the mean levels of financial distress were lower for companies receiving a first-time going-concern modified opinion from a Big 6 firm than from a non-Big 6 firm.

Together, our results suggest that the Big 6 audit firms are generally more conservative than non-Big 6 firms in their audit report decisions regarding going-concern modifications and the related differences in reporting errors were empirically consistent with the hypotheses. Our results also support the conclusions of Morris and Strawser (1999) that financial regulators in the U.S. behave as if the Big 6 audit firms report more conservatively than the non-Big 6 firms when issuing modified reports to troubled companies.

The interpretation of our results is subject to the following limitations. First, we only examine differences between the Big 6 and non-Big 6 firms. While this is consistent with the approach taken by many previous auditing researchers, we do not examine the overall size effects of audit firms by using a continuous audit firm size measure, or whether there exist differences between national non-Big 6 firms (i.e., second-tier firms) and local/regional firms. Second, the support for the first hypotheses is dependent on the definition of firm failure. Some might suggest that a move from non-default status to default status is more serious than the move from default to bankruptcy, because entering default could lead to many adverse consequences (such as initial loss of control or re-negotiation of debt on adverse terms). However, since bankruptcy is the most extreme manifestation of possible adverse consequences, others would suggest that the move to bankruptcy (from default status) is more serious than the move to default (from clean status). Given that our results indicate significance for only one of these two definitions of company failure, our results should be interpreted as offering modest support for the first hypothesis.

The results presented in this paper generally support the wealth at risk argument, which suggests that the Big 6 firms would be more conservative because of their greater wealth at risk in an audit of any given client. Our findings are consistent with those of prior researchers who have found evidence that the Big 6 firms are more conservative than non-Big 6 firms in a variety of auditor-client interactions. Together the extant research suggests that the Big 6 firms in the U.S. generally appear to be (1) more conservative in their client selection and retention decisions and (2) more conservative in their reporting decisions on those clients.

One reason for the dominance of the wealth at risk argument and the conservative posture of the Big 6 may be the litigation environment in the U.S. In particular, as reflected by the large magnitude of judgments in some recent litigation involving audit firms, very few lawsuits may be sufficient to cause the failure of even a large audit firm. In other words, the catastrophic nature of litigation losses may overwhelm the benefit of client portfolio diversification.

In the context of initial public offerings, Clarkson and Simunic (1994) show that the different litigation environments of the U.S. and Canada lead to significant differences in the market for audit services. Specifically, they find that large auditors were more likely to serve as auditors for initial public offerings involving riskier clients in Canada but not in the U.S. Thus, an interesting area for future research is to examine differences in audit reporting in countries where auditors are subject to a less hostile litigation environment.

REFERENCES

American Institute of Certified Public Accountants. (1988). Statement on Auditing Standards No. 59, The Auditor's Consideration of an Entity's Ability to Continue as a Going Concern. New York: AICPA.

Altman, E. I. (1982). Accounting implications of failure prediction models. Journal of Accounting Auditing and Finance, (Fall), 4-19.

Bamber, M., L. Bamber & M. Schoderbek. (1993). Audit structure and other determinants of audit report lag: An empirical analysis. Auditing: A Journal of Practice and Theory, (Spring), 1-23.

Beatty, R.P. (1993). The economic determinants of auditor compensation in the initial public offerings market. Journal of Accounting Research,(Autumn), 294-302.

Carcello, J., D. Hermanson & F. Huss. (1995). Temporal changes in bankruptcy-related reporting. Auditing: A Journal of Practice and Theory, (Fall), 133-143.

Carcello, J., D. Hermanson & F. Huss. (1997). The effect of SAS No. 59: How treatment of the transition period influences results. Auditing: A Journal of Practice and Theory, (Spring), 114-123.

Carcello, J., D. Hermanson & F. Huss. (1999). Going-concern opinions: The effects of partner compensation plans and client size. Auditing: A Journal of Practice and Theory, (forthcoming).

Carcello, J. & Z. Palmrose. (1994). Auditor litigation and modified reporting on bankrupt clients. Journal of Accounting Research, (Supplement): 1-30.

Chen, K. C. W. & B.K. Church. (1992). Default on debt obligations and the issuance of goingconcern opinions. Auditing: A Journal of Practice and Theory, (Fall), 30-49.

Chen , K. C. .W. & Wei. (1993). Creditors' decisions to waive violations of accounting-based debt covenants. The Accounting Review, (April), 218-232.

Chow, C. W. & S. Rice. (1982). Qualified Audit Opinions and Auditor Switching. The Accounting Review, (April), 326-335.

Clarkson, P. & D. A. Simunic. (1994). The association between audit quality, retained ownership and firm-specific risk in US vs. Canadian IPO markets. Journal of Accounting and Economics, 17, 207-228.

DeAngelo, L. E. (1981). Auditor size and audit quality. Journal of Accounting and Economics, 3, 183-199.

DeFond, M. L. (1992). The association between changes in client firm agency costs and auditor switching. Auditing: A Journal of Practice and Theory, (Fall), 16-31.

Foster, B.P., T.J. Ward & J. Woodroof. (1998). An analysis of the usefulness of debt defaults and going-concern opinions in bankruptcy risk assessment. Journal of Accounting, Auditing, and Finance, (Summer), 331-371.

Francis, J. R. (1984). The effect of audit firm size on audit prices: A study of the Australian market. Journal of Accounting and Economics, (August), 133-151.

Francis, J. R. & D. T. Simon. (1987). A test of audit pricing in the small client segment of the U.S. audit market. The Accounting Review, (January), 145-157.

Francis, J. R. & E.R. Wilson. (1988). Auditor changes: A joint test of theories relating to agency costs and auditor differentiation. The Accounting Review, 63 (October), 663-682.

Francis, J. R. & J. K. Reynolds. (1999). Evidence on risk-based client screening by Big Six audit firms. Working paper, University of Missouri-Columbia.

Francis, J. R., E. L. Maydew & H. C. Sparks. (1999). The role of Big 6 auditors in the credible reporting of accruals. Auditing: A Journal of Practice and Theory, (Fall), 17-34.

Geiger, M. A., K. Raghunandan & D. V. Rama. (1998). Cost associated with going-concern modified audit opinions: An analysis of auditor changes, subsequent opinions, and client failures. Advances in Accounting, 16, 117-139.

Gul, F. A. (1999). Audit prices, product differentiation and economic equilibrium. Auditing: A Journal of Practice and Theory, (Spring), 90-100.

Hopwood, W., J. McKeown & J. Mutchler. (1989). A test of the incremental explanatory power of opinions qualified for consistency and uncertainty. The Accounting Review, (January), 28-48.

Hogan, C. E. & D. C. Jeter. (1999). Industry specialization by auditors. Auditing: A Journal of Practice and Theory, (Spring), 1-17.

Johnson, W. B. & T. Lys. (1990). The market for audit services: Evidence from voluntary auditor changes. Journal of Accounting and Economics, (January), 281-308.

Kasznik, R. & B. Lev. (1995). To warn or not to warn: Management disclosures in the face of earnings surprise. The Accounting Review, (January), 113-134.

Louwers, T. (1998). The relation between going-concern opinions and the auditor's loss function. Journal of Accounting Research, (Spring), 143-156.

McKeown, J. C., J. F. Mutchler & W. Hopwood. (1991a). Towards an explanation of auditor failure to modify the audit opinion of bankrupt companies. Auditing: A Journal of Practice and Theory, (Supplement): 1-13.

McKeown, J. C., J. F. Mutchler & W. Hopwood. (1991b). Reply. Auditing: A Journal of Practice and Theory, (Supplement): 21-24.

Menon, K. & D. D. Williams. (1999). Error cost and auditors' termination decisions. Journal of Accounting, Auditing & Finance, (Spring), 95-123.

Morris, M. H. & W. D. Nichols. (1988). Consistency exceptions: Materiality judgments and audit firm structure. The Accounting Review, (April), 237-254.

Morris R. E. & J. R. Strawser. (1999). An examination of the effect of CPA firm type on bank regulators' closure decisions. Auditing: A Journal of Practice & Theory, (Fall), 142-158.

Mutchler, J. (1984). Auditor's perceptions of the going-concern opinion. Auditing: A Journal of Practice & Theory, (Spring), 17-30.

Mutchler, J. (1985). A multivariate analysis of the auditor's going-concern opinion decision. Journal of Accounting Research, (Autumn), 668-682.

Mutchler, J., W. Hopwood & J.C. McKeown. (1997). The influence of contrary information and mitigating factors on audit opinion decisions on bankrupt companies. Journal of Accounting Research, (Autumn), 295-310.

Mutchler, J. & D. D. Williams. (1990). The relationship between audit technology, client risk profiles, and the going-concern opinion decision. Auditing: A Journal of Practice & Theory, (Fall), 39-54.

Nichols, D. R. & D. B. Smith. (1983). Auditor credibility and auditor changes. Journal of Accounting Research, (Autumn), 534-544.

Ohlson, J. A. (1980). Financial ratios and the probabilistic prediction of bankruptcy. Journal of Accounting Research, (Spring), 109-131.

Palmrose, Z. V. (1986). Audit fees and auditor size: Further evidence. Journal of Accounting Research, (Spring), 97-110.

Pratt, J. & J.D. Stice. (1994). The effects of client characteristics on auditor litigation risk judgments, required audit evidence, and recommended audit fees. The Accounting Review, (October), 639-656.

Private Securities Litigation Reform Act: 1995. Public Law 104-67. Washington, D.C.:GPO.

Raghunandan, K. & D. V. Rama. (1995). Audit reports for companies in financial distress: Before and after SAS No. 59. Auditing: A Journal of Practice and Theory, (Spring), 50-63.

Raghunandan, K. & D. V. Rama. (1999). Auditor resignations and the market for audit services. Auditing: A Journal of Practice and Theory, (Spring), 124-134.

Simon D. & J.R. Francis. (1988). The effects of auditor change on audit fees: Test of price cutting and price recovery. The Accounting Review, (April), 255-269.

Simunic, D. (1980). The pricing of audit services: Theory and evidence. Journal of Accounting Research, (Spring), 161-190.

Simunic, D. & M. T. Stein. (1987). Product differentiation in auditing: Auditor choice in the market for unseasoned new issues, Research Monograph No. 13 Vancouver: The Canadian Certified General Accountants' Research Foundation).

Simunic, D. & M. T. Stein. (1990). Audit risk in a client portfolio context. Contemporary Accounting Research, (Spring), 329-343.

Simunic, D. & M. T. Stein. (1996). The impact of litigation risk on audit pricing: A review of the economics and the evidence. Auditing: A Journal of Practice and Theory, (Supplement): 119-134.

Spires, E. & D. Williams. (1990). Auditors' adoption of SAS 58 audit reports. Accounting Horizons, (September), 76-82.

Teoh, S. & T. Wong. (1993). Perceived auditor quality and the earnings response coefficient. The Accounting Review, 68, (April), 346-367.

U.S. House of Representatives. (1985). Hearings Before the Subcommittee on Oversight and Investigations of the Committee on Energy and Commerce, (February 20), No. 99-17.

U.S. House of Representatives. (1990). Hearings Before the Subcommittee on Telecommunications and Finance of the Committee on Energy and Commerce, (August 2), No.101-196.

Wheeler, S., K. Pany & E. Chewning. (1993). Inter-Firm differences in propensities to modify audit opinions for pre-SAS No. 58 accounting principles changes. Accounting Horizons, (September), 46-54.

Zmijewski, M. E. (1984). Methodological issues related to the estimation of financial distress prediction models. Journal of Accounting Research, (Supplement), 59-82.

ENDNOTES

(1) A large majority of these studies have been performed before the mergers among the largest accounting firms 1 reduced their number from eight to six to five. This study includes data from the time period before the number of the largest firms was reduced to five, so we refer to the "Big 6" in this paper.

(2) The Private Securities Reform Act (1995) has an audit requirement section related to going-concern reporting. 2 Specifically, the Act requires that

"Each audit ..of the financial statements of an issuer by an independent public accountant shall include, in accordance with generally accepted auditing standards, as may be modified or supplemented from time to time by the Commission (SEC) ... an evaluation of whether there is a substantial doubt about the ability of the issuer to continue as a going concern during the ensuing fiscal year."

This is the only instance where Congress has mandated a specific auditing standard for financial statement audits since its inception under the Securities Exchange Act of 1934.

(3) Note that the revenues lost include not only audit fees, but also revenues from the supply of non-audit services. 3 In addition, there also may be costs associated with other displeased existing clients who may switch to another audit firm, or from potential clients who choose a different audit firm.

(4) The firm view is appropriate for understanding going-concern modified audit reporting because more extensive 4 consultation with others within the audit firm is likely to occur in the case of distressed companies (such as those receiving a going-concern modified audit opinion). Furthermore, at a minimum, second partner reviews are required for public companies, requiring at least one additional firm partner to concur with the audit opinion rendered.

(5) Mutchler (1984) examined audit firm size as a factor in assessing the issuance of a going-concern qualified 5 opinion. However, her analysis focused only on the issuance of going-concern qualified opinions to stressed companies, and did not assess the resolution of the going-concern opinions to the companies that received them, or prior audit opinions for bankrupt companies.

(6) This argument assumes that lawsuit related losses of audit firms are not linear functions of the client revenue 6 derived from the audit engagement being litigated.

(7) For example, Pratt and Stice (1994) note that "partners from Laventhol & Horwath, previously the seventh 7 largest accounting firm in the U.S., cited litigation claims against their firm in their decision to declare bankruptcy" (p. 639).

(8) In addition a Big 6 firm may charge higher audit fees and also may have greater non-audit service related 8 revenues from a given client. Francis and Simon (1987) note that the average Big 6 fee premium in various studies ranges from 16 to 19 percent. However, the size of the smallest Big 6 firm is many times greater than that of the largest non-Big 6 firm. Consequently, while Big 6 fee premia may mitigate the lower [C.sub.I] for a Big 6 firm, they may not completely eliminate this phenomenon.

Further, it may be infeasible for some clients to switch auditors from a Big 6 firm to a non-Big 6 firm, due in part to the pressures from investors and creditors to maintain a Big 6 auditor. This may afford the Big 6 firms greater leverage to issue a going-concern modified report with a reduced fear of losing the client. These factors could then lead to a relatively lower [C.sub.I] for a Big 6 firm and result in more conservative reporting.

(9) Our analysis focuses on differences in the perceived costs associated with the two types of audit reporting errors across Big 6 and non-Big 6 audit firms. That is, audit firms' reporting decisions are influenced by their perceptions of the future consequences of their decisions. It is plausible to suggest that audit firms may have considered the costs associated with reporting errors and accordingly priced risk into their engagement fee. However, to make specific predictions about the impact of audit firm size on audit reporting and pricing, there have to be systematic differences in the relative magnitudes of risk premia (if any) incorporated into audit fees by Big 6 and non-Big 6 firms. This is an unanswered empirical question, and arguments can be marshaled on both sides. Consequently, we cannot make ex-ante predictions in a specific direction related the impact of risk premia on audit reports. Further, Simunic and Stein (1996) found that in the context of one large public accounting firm, auditors responded to client-specific increases in litigation risk by increasing the audit fee; however, this was largely due to changes in audit effort as opposed to the audit firm charging a higher price premium for client risk.

(10) As discussed later, an even broader definition of failure would include bankruptcy as well as both payment and 10 technical default. We also performed analyses using this alternative definition of BAD. The results from such additional analyses were substantively the same as those reported in the paper.

(11) SAS No. 59 specifically noted that the term "reasonable period of time" meant one year from the date of the 11 financial statements. Our sample includes companies where the audit report on the preceding financial statements was dated more than one year prior to the bankruptcy date. When we restricted our analysis to only those companies where the bankruptcy occurred within one year of the preceding fiscal year end, the results were substantively the same as those reported in this paper.

(12) We performed sensitivity analysis by using an even broader definition of failure, by also considering technical 12 default. Specifically, we considered a company to have failed if any one of the following conditions were met: (a) filing for bankruptcy, (b) entering payment default or (c) entering technical default. Analysis with this expanded definition was performed on those companies not already in technical or payment default (or bankruptcy) at the time they received the initial going-concern modified audit opinion. The logistic regression results for this reduced sample (n=325) were substantively similar to those presented in Table 2. Specifically, all control variable coefficients were positive and significant (p<.05), and the coefficient on the Big 6 variable was negative and significant (p <.01).

(13) We performed sensitivity analysis by including a dummy variable for firms in high-tech industries, as defined 13 in Kasznik and Lev (1995). In such analysis, the results remained substantively similar to those presented in Table 4 and the Big 6 variable continued to remain significant (p < .05).

(14) The financial stress indicators are based upon the measures used by prior researchers, such as McKeown et al. 14 (1991a). Note that the third factor used here, negative stockholder's equity, is a more stringent measure than the negative retained earnings measure used by McKeown et al. (1991a).

(15) Mutchler et al. (1997) found that their Big 6 variable was not significant in explaining differences in audit 15 opinions prior to bankruptcy. One possible reason for our different results may be the different nature of the samples. In particular, Mutchler et al. (1997) include only companies traded on the NYSE or AMEX, while our sample includes a substantial number of companies (191 out of 247) not traded on these exchanges. Companies listed on the NYSE and AMEX may be different along many dimensions than companies not traded on such exchanges.

Marshall A. Geiger, University of Richmond

K. Raghunandan, Texas A & M International University

TABLE 1: Descriptive Statistics--Going Concern Modified Opinion
Sample (Means)

 Auditor (n = 533) Bankrupt (n = 533)

 Non-Big 6
Item Big 6 (n = 343) (n = 190) Yes (n = 66) No (n = 467)

PRB 0.66 0.73 0.76 0.67
DFT 157 (45.8%) 51 (26.8%) 39 (59.1%) 169 (36.2%)
LNSL 9.48 7.71 10.13 8.67
BKT 51 (14.9%) 15 (7.9%) -- --
BAD * 47 17.3% 38 (23.2%) -- --

 BAD * (n = 435)

Item Yes (n = 85) No (n = 350)

PRB 0.74 0.66
DFT 31 (36.5%) 79 (22.6%)
LNSL 9.42 8.40
BKT -- --
BAD * -- --

Legend:

PRB = Probability of bankruptcy, calculated from Zmijewski's (1984)
model.

DFT = In default when receiving modified report.

LNSL = Natural log of sales (in thousands of dollars).

BKT = Bankrupt within one year from the financial statement date.

BAD = Bankrupt or in payment default within one year from the
financial statement date.

* Note that the BAD sample includes only companies which were
not in payment default at the time of receiving the initial
going-concern modified audit opinion. There were 435 such
companies, of which 271 were audited by the Big 6.

TABLE 2: Subsequent Resolution Model--Going-Concern Modified Sample

Logistic Regressions: [GC.sub.BANK]/[GC.sub.BAD] =
f(PRB, DFT, LNSL, TIME, BIG6)

 Bankruptcy Analysis (n = 533)

Variable Coefficient Chi-square Signif. *

Intercept -5.62 50.27 .0001
PRB 0.91 4.31 .0189
DFT 0.21 0.47 .2461
LNSL 0.22 10.11 .0008
TIME 0.71 3.75 .0264
BIG6 0.40 1.41 .1178
Model chi-sq. 34.25
p-value .0001
c-statistic 0.71

 BAD Analysis (n = 435)

Variable Coefficient Chi-square Signif. *

Intercept -4.35 45.34 .0001
PRB 0.75 3.78 .0261
DFT -- -- --
LNSL 0.23 14.73 .0001
TIME 1.10 11.98 .0003
BIG6 -0.70 6.27 .0061
Model chi-sq. 39.28
p-value .0001
c-statistic 0.71

Legend:

[GC.sub.Bank] = 1 if entered bankruptcy, 0 otherwise.

[GC.sub.BAD] = 1 if entered bankruptcy or payment default,
0 otherwise.

PRB = Probability of bankruptcy, calculated from Zmijewski's
(1984) model.

DFT = 1 if in default, 0 otherwise.

LNSL = Natural log of sales (in thousands of dollars).

TIME = 1 if post-SAS No. 59, 0 otherwise.

BIG6 = 1 if Big 6 auditor, 0 otherwise.

* Significance levels are one-tail, except for the intercept term.

TABLE 3: Descriptive Statistics--Bankrupt Sample (Means)

 Auditor Audit Report

 Big 6 Non-Big 6 GC NGC
Item (n = 211) (n = 36) (n = 136) (n = 111)

LNSL 11.39 9.88 10.78 11.68
BKLAG 14.30 13.51 13.28 15.22
REPLAG 8.90 9.45 9.28 8.60
PRB 0.68 0.53 0.83 0.43
Proportion in default 60% 39% 76% 34%
Proportion with 57% 42% -- --
 going-concern
 modified opinion

Legend:

LNSL = Natural log of sales (in thousands of dollars).

BKLAG = Time, in sq. root of days from audit report date to
bankruptcy date.

REPLAG = Time, in sq. root of days from fiscal year end to
audit report date.

PRB = Probability of bankruptcy, calculated from Zmijewski's
(1984) model.

TABLE 4: Audit Report Model--Bankrupt Sample

Logistic Regression: [BANK.sub.GC] = f(PRB, DFT, BKLAG, REPLAG,
LNSL, BIG6)

 Expected
Variable Sign Coefficient Chi-square Significance *

Intercept 3.94 3.77 .052
PRB + 4.30 43.97 .0001
DFT + 2.04 22.87 .0001
BKLAG - -0.20 11.53 .0007
REPLAG + 0.11 0.70 .401
LNSL - -0.61 25.51 .0001
BIG6 + 1.27 4.65 .031

Model Chi-square = 156.0, 6 df., p < .0001; c-statistic = 0.91

Legend:

[BANK.sub.GC] = 1 if going-concern modified, 0 otherwise.

PRB = Probability of bankruptcy, calculated from Zmijewski(1984).

DFT = 1 if company in financial default, 0 otherwise.

BKLAG = Time, in sq. root of days from audit report date to
bankruptcy date.

REPLAG = Tim e, in sq. root of days from fiscal year end to audit
report date.

LNSL = Natural log of sales (in thousands of dollars).

BIG6 = 1 if Big 6 auditor, 0 otherwise.

* Significance levels are one-tail, except for the intercept term.