September 20, 2000

Mr. Jonathan G. Katz, Secretary
Securities and Exchange Commission
450 Fifth Street, N.W.
Washington, D.C. 20549-0609

Dear Mr. Katz:

Re: File No. S7-13-00

I am writing in regard to the SEC's proposed rule to prohibit auditors from providing consulting services for their audit clients.

Attached to this e-mail is a WORD file with my research paper titled "Does Size Matter? The Influence of Large Clients on Office-Level Auditor Reporting Decisions."

This paper shows that Big Five auditors report more conservatively for larger clients in their practice offices than they do for smaller (less important) clients. This provides direct evidence that the auditor's fee reliance on large clients does not cause them to be more lenient and less objective with these important clients.

While this study is not about nonaudit services, there is a linkage. Nonaudit services potentially create fee dependence, and therefore potentially more pressure on accounting firms to be lenient in dealing with their large audit clients. Based on my analysis, I find no evidence that dependence on fees from large clients causes a loss of objectivity in auditor reporting decisions.

There may be reasons to support the SEC's proposed rule, but there is no evidence that economic dependence on fees from large clients causes auditors to treat those clients more leniently. This economic dependence problem seems to be at least part of the rationale for the proposed rule. The presumption seems to be that increased financial dependence causes a potential loss of objectivity.

In fact I find the exact opposite: larger clients at the office-level of Big 5 firms have less discretion in accounting for accruals, and are more likely to get a going concern audit report. In other words, auditors are "tougher" on larger clients in their practice offices than they are on their smaller less important clients.

Please feel free to contact me if you have questions about my research.

Sincerely,

Jere R. Francis
KPMG Distinguished Research Professor
College of Business
University of Missouri-Columbia
312 Middlebush Hall
Columbia, Missouri 65211
Phone: 573/882-5156
Fax: 573/882-2437
Email: Francis@Missouri.edu

Does Size Matter?
The Influence of Large Clients on
Office-Level Auditor Reporting Decisions

by

J. Kenneth Reynolds
Louisiana State University
Baton Rouge, LA 70803
email: jkreyn1@lsu.edu

and

Jere R. Francis
University of Missouri - Columbia
Columbia, MO 65211
email: francis@missouri.edu

First draft: August 1999
Current draft: September 14, 2000

Comments Welcome

Please Do Not Cite Without Permission

Does Size Matter?
The Influence of Large Clients on Office-Level Auditor Reporting Decisions

Abstract

The effect of large clients on the reporting decisions of Big Five auditors is investigated. Client size is measured relative to the clientele of the local Big Five office that contracts for the audit and issues the audit report. Large and influential clients in offices create two different incentives for auditors. On one hand, large clients create an economic fee dependence that may cause auditors to compromise independence and issue favorable audit reports in order to retain valuable clients. However, large clients also pose greater risk exposure with respect to litigation and reputation loss in the event of auditor negligence, which creates an incentive to report conservatively. We find no evidence that economic fee dependence causes Big Five auditors to report more favorably for larger clients in their practice offices. This finding supports the accounting profession's claim that existing economic incentives are sufficient to achieve auditor independence and reporting objectivity without additional regulations. However, we do find evidence that reputation protection causes auditors to report more conservatively for larger clients. Specifically, for a sample of potentially distressed firms, larger clients are more likely to receive going concern audit reports; and for a more general sample of firms, larger clients have smaller levels of accounting accruals, which implies less discretion to manage earnings.

JEL classification: M40; M41; L84
Keywords: Auditor Independence, Audit Reports, Conservatism, Going Concern, Accruals

Does Size Matter?
The Influence of Large Clients on Office-Level Auditor Reporting Decisions

1. Introduction

The question examined in this study is how client size influences the reporting decisions of auditors. In a seminal paper, DeAngelo (1981) argues the audit-client relationship is a bilateral monopoly creating an incentive for auditors to compromise their independence and report favorably in order to retain clients. This incentive can be termed economic fee dependence and larger clients create greater fee dependence, ceteris paribus. Mautz and Sharaf (1961) describe the auditor's financial dependence on clients as a built-in anti-independence factor, and the Cohen Commission (AICPA, 1978) observes that an auditor's complete independence is a practical impossibility since the auditor is hired and paid by the client.¹

When auditors have more than one client there is less financial dependence on a single client. Thus a larger portfolio of clients mitigates economic fee dependence and DeAngelo (1981, p. 189) concludes that audit firm size is a good proxy for independence and audit quality since larger accounting firms have more clients than smaller firms and therefore will be less reliant on any single client. DeAngelo's analysis has been used to differentiate the Big Five accounting firms from smaller non-Big Five firms because the large clientele of Big Five firms virtually guarantees no single client is a significant source of firm-wide revenue.

Our study uses the DeAngelo framework to further examine auditor incentives within the individual practice offices of Big Five accounting firms. This is an important contribution as prior auditing research has focused almost exclusively on audit firms as the unit of analysis. The distinction between firms and offices is significant because individual practice offices are responsible for contracting with clients, managing audit engagements, and issuing audit reports (Wallman, 1996). Francis et al. (1999b) report a wide variation in the size of Big Five practice offices. Some offices are small and have only one publicly-listed audit client while larger offices have over one-hundred public clients. In contrast to a firm-wide analysis of Big Five firms where no single client is a significant source of revenue, at the office level a single client has the potential to be very important. For this reason economic fee dependence is more likely to exist at the office level and thus it provides a better setting to test for its effect on auditor reporting. The shift in focus from the overall firm to individual practice offices is also consistent with recent statements by the ACIPA and SEC recognizing that local offices are where potential auditor independence problems exist (AICPA, 1997; Wallman, 1996).

While large clients clearly create the potential for economic fee dependence, large clients also pose a significant risk exposure that may cause auditors to be risk-averse and report conservatively. The risk exposure comes about if a negligent or otherwise questionable audit is performed for large high-profile clients. In such cases an accounting firm is likely to suffer a greater loss in reputation than would be the case for smaller less visible clients (Wilson and Grimlund, 1990). In addition to adverse reputation effects, litigation costs will be greater for large clients (Bonner et al., 1998; Kellogg, 1984; Lys and Watts, 1994; Stice, 1991). To illustrate the magnitude of litigation exposure for large clients, a lawsuit claiming damages for only a modest 10% decline in the market value of a corporate giant like General Motors would exceed the annual global revenue of any Big Five accounting firm.

In this alternative scenario, self-interested auditors have incentives to be independent and perhaps even risk-averse causing them to report conservatively for large clients to protect themselves from the adverse effects of reputation loss and litigation costs. This incentive can be termed reputation protection and is consistent with the claim by Big Five auditors that they engage in risk-management practices (Andersen et al., 1992). As with economic fee dependence, the incentive for reputation protection and auditor reporting conservatism is more appropriately investigated at the office-level of Big Five accounting firms where audit contracting occurs and where audit reports are issued.

Thus there are competing incentives regarding the effect of client size on local practice offices of Big Five accounting firms. The purpose of our study is to determine how the tradeoff between economic fee dependence and reputation protection affects auditor decision-making in these offices. Two empirical questions are investigated with respect to auditor reporting decisions. First, does client size affect the auditor's decision to issue an unfavorable going concern audit report? Second, does client size affect the discretion an auditor permits with respect to accounting accruals and, by implication, the discretion to manage reported earnings? Our research design assumes that auditors report objectively for smaller clients in offices, and the analysis examines the effect of increasing client size on auditor reporting decisions. We find that Big Five auditors do not treat large influential clients in their practice offices more leniently than smaller less important clients. Economic fee dependence does not lead to fewer going concern reports or greater discretion with respect to accounting accruals. These results are consistent with a recent white paper issued by the accounting profession (AICPA, 1997). It argued that existing incentives are sufficient to motivate auditors to be independent, despite the presence of economic fee dependence inherent in auditor-client contracting. However, we do find that relatively larger clients in offices are more likely to receive a going concern audit report and have less discretion with respect to accounting accruals. Thus there is evidence that reputation protection leads to auditor reporting conservatism.

The remainder of the paper is organized as follows. Economic incentives of auditors with respect to large clients in the local office context are developed more fully in the next section. Two auditor reporting decisions are analyzed in the study, the issuance of going concern audit reports and the client discretion with respect to accruals. The analysis of going concern audit reports is presented in section 3, and the analysis of accruals is reported in section 4. The study concludes in section 5.

2. Auditor incentives in local offices of big five accounting firms

In prior research, auditor reporting incentives with respect to client size have been conceptualized and measured based on client size relative to the total clientele of an accounting firm (DeAngelo, 1981; Lys and Watts, 1994; Stice, 1991). Such an approach is problematic for multi-office Big Five accounting firms because it does not focus on the relevant decision-making unit within these firms (Wallman, 1996). The Big Five firms have decentralized organizations and ownership structures, and operate through a network of semi-autonomous local offices (Narayanan, 1995). These local offices contract with clients, administer audit engagements, and issue audit reports on behalf of the firm (Francis et al., 1999b).

The distinction between offices and firms is an important one. The loss of any single client has a small impact when the unit of analysis is the Big Five accounting firm as a whole. For example, the largest single client of each Big Five firm represents less than 0.2 % of its firm-wide audited sales, using U.S. company data in our study from Compact Disclosure. However, a different picture emerges when the unit of analysis is the local office with primary engagement responsibility. Based on data in our sample, clients average around 8% of office-level audited sales of clients. The loss of a single large client can cause significant loss of revenue to the local office, resulting in staff reductions and lower partner compensation (to the extent it is tied to the partner's client base). Following the DeAngelo (1981) framework, these economic consequences create strong incentives for local practice offices to be lenient in order to retain large clients.

SEC Commissioner Steven Wallman underscores the importance of evaluating economic fee dependence in the context of the local unit of the accounting firm with client decision-making responsibilities. Wallman (1996, p. 90) adds that there would be independence concerns "where an office or partner was receiving a material percentage of revenues from a single client or group of clients." In their 1997 white paper, the AICPA also indicates that the appropriate measure of economic fee dependence is the magnitude of fees from clients "in relation to particular regions or offices of the firm" (AICPA, 1997, footnote 339). Australia has an audit standard that explicitly cautions auditors to avoid situations in which an "office regularly depends on one audit client or group of connected clients for a significant portion of its total fees" and suggests 15% as a rule-of-thumb limit on the portion of revenues from a single client (Australian Society of Certified Public Accountants Members Handbook, Statement of Auditing Practice AUP 32, paragraph 28).

While economic fee dependence clearly exists for large clients, these large clients also pose a potentially greater risk to the reputation of the office and accounting firm. A negligent or questionable audit of large clients will impair the reputation of the local office involved with the audit and adversely affect its ability to do business in that particular market. The local business community will be aware of the publicity surrounding the audit, perhaps even the names of key accounting firm personnel involved. There is empirical evidence that local offices do experience such negative reputation effects. Wilson and Grimlund (1990) report that local offices associated with audits deemed deficient by the SEC were more adversely affected than other offices of the firm nationally in terms of their ability to retain existing clients and attract new clients.

In sum the local practice offices of Big Five firms face kind of economic fee dependence from large clients described in DeAngelo (1981). When a client is large relative to other clients in the office, this fee dependence is greatest. In addition, larger clients also pose a greater risk to the office and firm in the event of an audit failure, and this creates an alternative incentive to be independent and perhaps even stricter than normal in order to protect the firm's reputation and also to avoid costly litigation.² The next two sections of the paper report empirical tests to determine which of these incentives dominates auditor decision-making with respect to large clients in local practice offices.

3. Going concern audit reports

3.1 Hypothesis development

There are two publicly observable outcomes of an audit, the audited financial statements and the audit report on these statements. Auditing standards specify various types of modified audit reports, in addition to the standard clean opinion. Technically the SEC does not allow financial statements to be filed if the auditor issues a GAAP qualification or adverse opinion, and for this reason research has focused on audit reports that flag material uncertainties or going concern problems. With the adoption of SAS No. 79 eliminating reports for material uncertainties after February 28, 1996, going concern reports remain the only viable topic for audit report research.

SAS No. 59 requires an auditor to evaluate if substantial doubt exists about a firm's ability to continue as a going concern for one year beyond the financial statement date. In deciding to issue a going concern report the auditor faces a tradeoff that parallels the discussion in section 2. Failure to issue a going concern report in the appropriate circumstances may harm the auditor's reputation, resulting in the loss of clients and creating exposure to litigation. A relatively larger client exposes an auditor to more public scrutiny and potentially greater litigation and reputation loss. However, going concern reports also lead to a greater incidence of auditor switching which results in the loss of fees (Krishnan, 1994). This impact will be greater when clients are large relative to the office issuing the audit report, causing auditors to be lenient and report more favorably for large clients.

Hypothesis 1 tests these competing incentives with respect to going concern reports:

H1a: Economic fee dependence causes auditors to be lenient in reporting for large clients. The larger a client relative to the portfolio of clients served by a given office, the less likely the auditor will issue a going concern report.

H1b: Reputation protection causes auditors to be conservative in reporting for large clients. The larger a client relative to the portfolio of clients served by a given office, the more likely the auditor will issue a going concern report.

3.2 Sample selection

The initial sample is comprised of 6,747 U.S. companies with sales data and Big Five auditors for fiscal 1996 as reported in the Compact Disclosure database dated October 1997. Sales data is required to construct the measure of client size relative to office-level client portfolios. Previous research indicates that Big Five firms have a reputation advantage over non-Big Five firms, and that the incentives of firms choosing non-Big Five auditors may be different from those of firms audited by the Big Five (Craswell et al., 1995; Francis and Wilson, 1988). To control for potential reputation and incentive differences, the study is restricted to companies audited by Big Five auditors. For simplicity the term Big Five is used even though data in the study predate the 1998 merger that created Pricewaterhousecoopers and reduced the then Big Six group of leading accounting firms to the Big Five.

For each observation in the study, the local office of the Big Five accounting firm with primary engagement responsibility is identified from the office letterhead on which the audit report is written. The audit report is obtained from Compact Disclosure when available, otherwise from 10-K filings in the SEC's EDGAR archives. A total of 499 local offices of Big Five accounting firms conducted audits of the 6,747 companies in the sample. These offices have an average of 13.5 publicly-listed clients per office and a range of 1 to 139 clients per office.

For the going concern report analysis, the sample of 6,747 observations is further reduced to include only those companies that are potentially financially distressed. The reason for this additional screen is to evaluate auditor reporting decisions for a subset of companies in which the going concern report is a more salient decision. Firms are defined as potentially distressed if they have negative net income and/or negative cash flows from operations. This screen results in a sample of 2,439 potentially distressed companies audited by 402 offices of Big Five accounting firms. A going concern report is issued for 9.2% (224) of these companies compared to a going concern rate of only 3.8% for the full sample of 6,747 companies.

3.3 Model specification

A logistic model is estimated in which a categorical audit opinion variable is regressed on a client size variable and a set of other control variables. The dependent variable OPINION is coded one for a going concern audit report and zero for an unmodified clean opinion. The client size variable (INFLUENCE) represents the relative influence a client has on the office performing the audit. The importance of a client is directly proportional to the magnitude of its fees relative to total fees of the office. For each observation in the sample, INFLUENCE is defined as the observation's sales (in natural log form) divided by the sum of sales (in natural log form) of all publicly-listed companies audited by the office. The rationale for using the natural log of sales is its high correlation with audit fees (Craswell et al., 1995; Francis, 1984). INFLUENCE is thus a proxy for the percentage of fees generated by a client relative to the total for the office administering the audit.³

The set of control variables in the study is based on the so-called contrary and mitigating factors identified in SAS No. 59. Contrary factors indicate a going concern report may be appropriate, while mitigating factors are actions or circumstances that would mitigate against a going concern report. Contrary factors include negative financial trends and other indications of possible financial difficulties. The most important contrary factor employed in previous studies is financial distress or the probability of bankruptcy (Bell and Tabor, 1991; Dopuch et al., 1987; Jeter and Shaw, 1995; Krishnan and Krishnan, 1996; Louwers, 1998; Mennon and Schwartz, 1987; Mutchler et al., 1997). We use the Z-score from Altman's (1983) bankruptcy prediction model to measure financial distress (PBANK). Z-scores are increasing in the financial health of companies, and are inversely related to the likelihood of bankruptcy. Thus a lower score for PBANK indicates greater financial distress and a greater likelihood the company will receive a going concern report.⁴

Two additional measures of financial distress are included in our model. Since the Altman model specifies variables only for the current period, the variable LOSS is used to indicate companies with a two-year trend of negative earnings. In addition, Mutchler et al. (1997) provide evidence that debt covenant violations have incremental explanatory power in the going concern decision. Change in debt from the prior year (DEBTCHG) is used to proxy for closeness to debt covenant violation. DEBTCHG is defined as the ratio of total debt to total assets in the prior year (fiscal 1995) minus the same ratio for the test year (fiscal 1996).

Mitigating factors have not been incorporated in previous research (an exception is Mutchler et al. (1997)). If auditors determine that contrary factors exist, they must also evaluate managerial plans to mitigate the effect of contrary factors before concluding that a going concern report is warranted. Managerial actions that mitigate the effect of contrary factors include plans to sell assets, issue new financing or refinance existing debt, and increase ownership equity (SAS No. 59). Because these plans are difficult to observe ex ante, we examine the subsequent 1997 fiscal year financial statements to determine ex post if there were significant sales of assets or the issuance of new debt or equity. The indicator variable MITIGATE is coded one if any of these factors are present and the dollar magnitude exceeds 5% of 1996 assets. This approach assumes auditors had private knowledge of such subsequent actions at the time the going concern report decision was made for the fiscal 1996 financial statements.

The remaining two control variables are absolute company size and the issuance of a prior going concern report. Absolute company size (rather than client size relative to the office responsible for the audit) can affect the auditor's decision to issue a going concern report. In the event of financial distress a large company has greater negotiating power with creditors and more resources to fend off bankruptcy. Absolute company size is measured as log of sales (SALES). Finally, other studies find that the likelihood of a going concern report is greater if there is a prior year going concern report, and the variable PRIORGC is included to control for this effect.

The experimental and control variables are combined in the following logistic regression model:

OPINION = ₀ + ₁INFLUENCE + ₂PBANK + ₃LOSS + ₄DEBTCHG + ₅MITIGATE +

₆SALES + ₇PRIORGC

where:

OPINION	=	1 if going concern audit report, 0 otherwise.
INFLUENCE	=	log client sales/ log sales of each public client of the office issuing the audit report.
PBANK	=	probability of bankruptcy based on the Altman Z-score.
LOSS	=	1 if the company has a loss in the current and previous years, 0 otherwise.
DEBTCHG	=	ratio of total debt/total assets in fiscal 1995 minus same ratio in fiscal 1996.
MITIGATE	=	mitigating factors, equal to 1 if the client sold assets or issued new debt or new equity in the subsequent year.
SALES	=	log of client sales ($000).
PRIORGC	=	1 if the company received a going concern opinion in the prior year, 0 otherwise.

3.4 Results

Table 1 reports descriptive statistics for the sample of 2,439 financially distressed companies. Average sales are $38 million indicating that the sample of potentially distressed firms is smaller on average than the Compact Disclosure population.⁵ The experimental variable INFLUENCE has a mean of 7% for companies with clean audit reports and 8% for companies with going concern reports. The 224 companies receiving a going concern report are significantly different from the 2,215 companies with an unmodified clean opinion on the following dimensions: they have a higher probability of bankruptcy, are more likely to have had losses in the last two years, increased their debt, are smaller in absolute size, and are more likely to have received a prior year going concern report.

Table 1 Descriptive statistics for 2,439 potentially distressed companies in fiscal 1996 (U.S. companies on Compact Disclosure with net loss or negative cash flow from operations, and audited by the Big Five accounting firms)
Variable	Mean	Std. Dev.	Median	Lower Quartile	Upper Quartile
Companies with an unmodified clean audit report (n=2,215)
INFLUENCE	0.07	0.10	0.03	0.02	0.07
PBANK	1.37*	4.02	1.32	0.28	2.57
LOSS	0.34*	0.47	0	0	1
DEBTCHG	-0.01*	0.55	0	-0.02	0.05
MITIGATE	0.44	0.50	0	0	1
SALES	10.63*	2.24	10.63	9.29	12.09
PRIORGC	0.02*	0.15	0	0	0
Companies receiving a going concern audit report (n=224)
INFLUENCE	0.08	0.14	0.03	0.02	0.08
PBANK	-5.65*	12.96	-2.32	-7.08	0.29
LOSS	0.61*	0.49	1	0	1
DEBTCHG	0.06*	1.53	0.04	0	0.27
MITIGATE	0.41	0.49	0	0	1
SALES	9.16*	2.51	8.90	7.49	11.11
PRIORGC	0.44*	0.50	0	0	1
* difference in means between going concern and non going concern samples is significant at 0.01.
Variable Definitions: INFLUENCE = log of client sales/ log (sales) of all public clients of the office issuing the audit report. PBANK = probability of bankruptcy measured by the Altman Z-score (0.717*Net Working Capital/Assets + 0.847*Retained Earnings/Assets + 3.107*Earnings Before Interest and Taxes/Assets + 0.42*Book Value of Equity/Liabilities + 0.998*Sales/Assets). LOSS = 1 if the company has a loss in the current and previous years, 0 otherwise. DEBTCHG = change in ratio of total debt to total assets from 1995 to 1996. MITIGATE = mitigating factors equal to 1 if the client sold assets or issued new debt or new equity in subsequent year, 0 otherwise. SALES = log of client sales ($000). PRIORGC = 1 if company received a going concern audit report in the prior year, 0 otherwise.

The experimental variable INFLUENCE has a low Pearson product correlation with the other independent variables: the largest correlation is +0.10 between relative size (INFLUENCE) and absolute size (SALES). The largest correlation among the remaining independent variables is -0.25 between PBANK and PRIORGC. To further assess potential multicollinearity in the data, the going concern model was also estimated using OLS regression to derive variance inflation factors. None of the variance inflation factors exceed 1.5 indicating that multicollinearity is not a concern in the model estimation (Greene, 1993).

The logistic regression model used to test H1 is reported in Table 2. The model Chi-square is significant at p<. 01 with a pseudo r-square of 34%.

Table 2
Logistic regression estimate for 2,439 potentially distressed firms (dependent variable =1 if going concern audit report and 0 otherwise)
Variable	Predicted Sign	Estimate	Chi-square	p-value
Intercept	+/-	-3.1581	41.54	0.0001
INFLUENCE	+/-	1.1848	3.53	0.0601
PBANK	-	-0.2072	96.23	0.0001
LOSS	+	0.2988	2.58	0.1078
DEBTCHG	+	0.1395	1.92	0.1653
MITIGATE	-	-0.3996	4.70	0.0301
SALES	-	0.0213	0.25	0.6200
PRIORGC	+	3.1262	173.54	0.0001
Model chi-square = 501.38, p<.0001. Pseudo-R2 = 34%.
Variable Definitions: INFLUENCE = log of client sales/ log (sales) of all public clients of the office issuing the audit report. PBANK = probability of bankruptcy measured by the Altman Z-score (see Table 1). LOSS = 1 if the company has a loss in the current and previous years, 0 otherwise. DEBTCHG = change in ratio of total debt to total assets from 1995 to 1996. MITIGATE = mitigating factors equal to 1 if the client sold assets or issued new debt or new equity in subsequent year, 0 otherwise. SALES = log of client sales ($000). PRIORGC = 1 if company received a going concern audit report in the prior year, 0 otherwise.

The experimental variable INFLUENCE is positive and significant at p=. 06.⁶ For the sample of potentially distressed companies, relatively larger clients in offices are more likely to receive a going concern report, after controlling for financial distress, debt changes, mitigating factors, absolute company size, and prior going concern reports. This finding indicates that auditors in local offices report more conservatively for larger clients that are potentially distressed, a finding consistent with reputation protection and litigation avoidance (H1b) driving auditor behavior rather than economic dependence on fees (H1a).

Results for the control variables are significant (p<.10) and in the expected direction except for SALES and DEBTCHG which are insignificant.⁷ Companies are more likely to receive a going concern report if they are financially distressed, and PBANK and LOSS are significant in the predicted directions. Companies are also more likely to receive a going concern report if they received a prior going concern report (PRIORGC). Finally, the presence of mitigating factors (MITIGATE) reduces the likelihood of a going concern report. This last finding contrasts with Mutchler et al. (1997) who report mixed results for mitigating factors.

3.5 Sensitivity analysis

Despite the low correlation between absolute size (SALES) and the office-level variable INFLUENCE, the following test was undertaken to assure that INFLUENCE does not proxy for or interact with absolute client size. A term is added to the model in Table 2 interacting INFLUENCE with a dichotomous variable for absolute size (coded 1 if the observation is above median sales in the sample). The interaction term will indicate if the variable INLFUENCE differs for the upper and lower halves of absolute client size in the sample. The interaction term is insignificant (p=.42), and INFLUENCE is significant at p<.05. This provides further assurance that statistical significance of INFLUENCE in Table 2 is not affected by the absolute size of clients in the sample.

As noted in section 3.2 the sample of offices of Big Five accounting firms has a considerable range in terms of the number of publicly-listed clients per office. To determine if the results reported in Table 2 are driven by observations from extremely small or large offices we delete such observations and re-estimate the model. When observations are deleted from offices having less than three clients, INFLUENCE is positive and significant at p=.05. When observations are deleted from offices having 100 or more clients, INFLUENCE is positive and significant at p=.06. When both groups are deleted at the same time INFLUENCE is positive and significant at p=.05. On the basis of this analysis we conclude that the results in Table 2 are robust with respect to office size.

The results in Table 2 also illustrate the importance of using the local office unit of analysis. Absolute client size (SALES) has no statistical association with going concern reports. The model in Table 2 was also estimated dropping the test variable INFLUENCE, and absolute client size variable (SALES) remained insignificant. These results demonstrate that absolute client size has no association with auditor reporting decisions in the sample of potentially distressed firms used in the study. Inferences about client size would therefore be misleading if based solely on absolute size. Client size does matter, but this is evident only when client size is defined relative to the office issuing the audit report.

As a final test, the experimental variable INFLUENCE is redefined at the firm level rather than office level. As expected the redefined variable is insignificant. The reason is that even the largest client is quite small relative to the firm's entire U.S. clientele. This analysis further demonstrates the importance of using offices as the unit of analysis in audit research rather than the firm as a whole in order to understand economic incentives facing auditors.

4. Accounting accruals

4.1 Hypothesis development

The issuance of a going concern audit report is a serious decision, but it affects less than 4% of U.S. companies in our sample audited by Big Five accounting firms. A less extreme judgment that auditors make, and one that applies to companies across the full spectrum of size and financial condition, is the discretion clients are permitted with respect to accounting accruals. For these reasons accruals provide an alternative and perhaps better setting than the going concern decision to evaluate the effects of economic dependence and reputation protection on auditor decision-making.

Prior research shows that managers use accruals to systematically manage reported earnings (DeFond and Park, 1997; Jones, 1991; Healy, 1985). How are accounting accruals affected by the economic fee dependence and reputation protection? Economic fee dependence will cause an auditor to be lenient and allow large clients more discretion with respect to accounting accruals, which implies more potential to manage earnings. Alternatively, the fear of reputation loss and litigation will cause an auditor to limit discretion with respect to accounting accruals for large clients. Less discretion means less potential for earnings management by clients and therefore less likelihood of an ex post claim of audit failure with respect to reported earnings. Lys and Watts (1994) report that high levels of accruals are associated with auditor lawsuits, and SEC actions against companies are largely related to misstated accruals (Feroz et al., 1992).

Hypothesis 2 tests these competing incentives with respect to accruals:

H2a: Economic fee dependence causes auditors to be lenient in reporting for large clients. The larger a client relative to the portfolio of clients served by a given office, the more discretion a client has with respect to accounting accruals.

H2b: Reputation protection causes auditors to be conservative in reporting for large clients. The larger a client relative to the portfolio of clients served by a given office, the less discretion a client has with respect to accounting accruals.

When there are no directional hypotheses for accruals, Warfield et al. (1995) and Francis et al. (1999a) argue that the extent to which companies use accruals to manage earnings is best measured by the absolute (unsigned) value of accruals. The magnitude of unsigned accruals measures a company's success in managing earnings either up or down, as needed, depending on year-specific situations (DeFond and Park, 1997; Healy, 1985). Additional tests are also reported using signed accruals.

A second way of evaluating discretion is to examine the cross-sectional variation of signed accruals. Companies that make greater use of accounting accruals to manage earnings should on average have greater cross-sectional variation in their accruals than companies that do not. If auditors allow larger clients more (less) discretion, larger clients should have greater (less) variance than smaller clients.

Hypothesis 3 tests these competing incentives with respect to the variance of signed accruals:

H3a: Economic fee dependence causes auditors to be lenient in reporting for large clients. For larger clients in the portfolio of clients served by a given office, the variance of accounting accruals is greater than for smaller clients.

H3b: Reputation protection causes auditors to be conservative in reporting for large clients. For larger clients in the portfolio of clients served by a given office, the variance of accounting accruals is less than for smaller clients.

4.2 Model specification

Two measures of accounting accruals (scaled by lagged assets) are used to test H2 and H3. The first measure is the absolute value of total accruals. Total accruals are defined as operating net income minus cash flows from operations, scaled by lagged total assets.

The second measure is the absolute value of discretionary accruals, scaled by lagged total assets. Discretionary accruals, denoted DA_ijt for firm i in industry j for year t, are computed using the industry cross-sectional variation of Jones' (1991) model (DeFond and Jiambalvo, 1994; DeFond and Subramanyam, 1998). Industry is defined by two-digit SIC code. Discretionary accruals are defined as the difference between total accruals and estimated non-discretionary accruals.

The estimation of non-discretionary or expected accruals proceeds in a two-step process. First, the following model is estimated for the full sample by regressing total accruals on the change in revenues from the prior year and the level of property plant and equipment to control for the economic determinants of expected accruals:

TA_ijt/A_ijt-1 = _jt[1/A_ijt-1] + _0jt[REV_ijt/A_ijt-1] +_1jt[PPE_ijt/A_ijt-1]+e_ijt

where:

TA_ijt = total accruals (net income from continuing operations, minus operating cash flows) for company i in industry j for year t.

A_itj-1 = total assets for company i in industry j for year t-1.

REV_ijt = change in revenues from prior year for company i in industry j for year t.

PPE_ijt = gross PP&E for company i in industry j for year t.

e_ijt = error term for company i in industry j for year t.

Next the appropriate industry-specific model parameters from this estimation are used to calculate a value for each observation, i.e., a value for each company i in industry j for year t, scaled by lagged total assets.. This calculation is an estimate of the observation's non-discretionary or expected accruals. Discretionary accruals (DA_ijt) are then defined as total accruals (TA_ijt) minus the calculated value for non-discretionary accruals.

Previous research identifies several additional factors that may influence the magnitude of accruals as defined above. The most important of these are operating cash flows, absolute company size, and leverage (Becker et al., 1998). Operating cash flows (OCF) are defined as cash flows from operations scaled by lagged total assets and have been shown to vary inversely with discretionary accruals (Dechow et al., 1995). Company size is measured as log of sales (SALES) and may be correlated with operating characteristics that cause large companies to have systematically smaller discretionary accruals, even though accruals are scaled by lagged assets. Companies with higher debt levels have a greater incentive to use accruals to increase earnings due to closeness to debt covenant constraints. The variable DEBT is defined as the ratio of total debt/total assets. Similarly, financially distressed companies also have an incentive to use accruals to increase earnings. The Altman Z-score (PBANK) is included in the model and is expected to be negatively associated with accruals (recall that a lower Z-score indicates greater financial distress).

The following OLS regression model is used to test the hypothesized relationship between accruals and client influence at the office level of Big Five accounting firms:

ACCRUAL = ₀ + ₁INFLUENCE + ₂OCF + ₃SALES + ₄DEBT + ₅PBANK + e

where:

ACCRUAL = either the absolute value of total accruals scaled by lagged total assets (ABSACCR), or the absolute value of discretionary accruals scaled by lagged total assets (ABSDA).

INFLUENCE = log of client sales/ log sales of each client of the office issuing the audit report.

OCF = operating cash flows scaled by lagged total assets.

SALES = log of client sales.

DEBT = ratio of total debt/total assets.

PBANK = probability of bankruptcy based on the Altman Z-score.

White's (1980) heteroskedasticity-consistent covariance matrix adjustment is used for t-statistics on all of the coefficients in the regression models reported in Tables 4 and 5.

4.3 Results

The initial sample for the analysis of accruals is the 6,747 companies described in section 3.2. Companies with missing data are eliminated. In addition, companies in industries with less than five companies are omitted because of the sensitivity of the cross-sectional discretionary accruals model to a small number of observations. These screens yield a final sample of 4,952 companies. Table 3 reports descriptive statistics for these 4,952 companies. Sales average $141 million, which is more typical of average firm size in the Compact Disclosure population than the sample of potentially distressed firms in section 2 whose sales averaged only $38 million. Correlations between the experimental variable INFLUENCE and the other independent variables are all less than +/-0.10. The largest correlations among the other independent variables are DEBT and PBANK (-0.54), and SALES and OCF (+0.32), with other correlations less than +/- 0.12. Variance inflation factors are less than 1.5 in all models indicating that multicollinearity is not a concern in the model estimations.

The OLS regression estimations are reported in Table 4 using unsigned discretionary accruals (ABSDA) as the dependent variable and Table 5 using unsigned total accruals (ABSACCR) as the dependent variable. Both models are significant at p<.01 and have r-squares of 7%. The reported p-values for coefficient t-statistics in Tables 4 and 5 are based on the two-tail probability of rejecting a null hypothesis of no association with the dependant variable.

Table 3 Descriptive statistics for accruals tests, fiscal 1996 (U.S. companies on Compact Disclosure with Big Five auditors)
Variable	N	Mean	Std. Dev.	Min	Max
ABSDA	4,952	0.0872	0.1127	0	1.009
ABSACCR	4,952	0.0984	0.1253	0	0.998
INFLUENCE	4,952	0.0755	0.1141	0	1
OCF	4,952	0.0236	0.4008	-10.712	12.379
SALES	4,952	11.856	2.2133	0	18.916
DEBT	4,952	0.5898	0.4862	0.006	16.283
Variable Definitions: ABSDA = absolute value of discretionary accruals, scaled by lagged assets. ABSACCR = absolute value of total accruals, scaled by lagged assets. INFLUENCE = log of client sales/ log (sales) of all public clients of the office issuing the audit report. OCF = operating cash flows, scaled by lagged assets. SALES = log of client sales ($000). DEBT = ratio of total debt/total assets

Table 4 Regression estimate for test of discretionary accruals scaled by lagged assets (dependant variable is absolute value of discretionary accruals, n=4,952)
Variable	Predicted Sign	Estimate	t-statistic*	p-value
Intercept	+/-	0.2167	21.146	0.0001
INFLUENCE	+/-	-0.0386	-3.815	0.0001
OCF	-	-0.0265	-3.044	0.0023
SALES	-	-0.0101	-12.287	0.0001
DEBT	+	-0.0047	-0.598	0.5502
PBANK	-	-0.0018	-1.735	0.0827
R-square = 7.2%	F-ratio = 77.183 p<.0001
Variable Definitions: INFLUENCE = log of client sales/ log (sales) of all public clients of the office issuing the audit report. OCF = operating cash flows, scaled by lagged assets. SALES = log of client sales ($000). DEBT = ratio of total debt/total assets PBANK = probability of bankruptcy measured by the Altman Z-score (see Table 1). *Reported t-statistics are based on the White (1980) heteroscedasticity-corrected covariance matrix.

Table 5 Regression estimate for test of total accruals scaled by lagged assets (dependant variable is absolute value of total accruals, n=4,952)
Variable	Predicted Sign	Estimate	t-statistic*	p-value
Intercept	+/-	0.2476	20.175	0.0001
INFLUENCE	+/-	-0.0377	-3.007	0.0026
OCF	-	-0.0118	-0.966	0.3440
SALES	-	-0.0118	-12.059	0.0001
DEBT	+	-0.002	-0.024	0.9805
PBANK	-	-0.0032	-2.54	0.0109
R-square = 6.8%	F-ratio = 71.955, p<.0001
Variable Definitions: INFLUENCE = log of client sales/ log (sales) of all public clients of the office issuing the audit report. OCF = operating cash flows, scaled by lagged assets. SALES = log of client sales ($000). DEBT = ratio of total debt/total assets PBANK = probability of bankruptcy measured by the Altman Z-score (see Table 1). *Reported t-statistics are based on the White (1980) heteroscedasticity-corrected covariance matrix.

The experimental variable INFLUENCE is negative and significant at p<.01 in Table 4 using discretionary accruals, and is negative and significant at p<.01 in Table 5 using total accruals. Similar results are obtained when controls for industry sector are added to the models using the approach described in footnote 6, and when extremely small/large office are deleted as described in section 3.5. All of the control variables in Tables 4 and 5 are significant in the expected direction except debt level (DEBT) in both tables and operating cash flows (OCF) in Table 5.⁸

Relatively larger clients in offices of Big Five firms have lower levels of both discretionary and total accruals after controlling for operating cash flows, absolute company size, leverage, financial distress, industry sector, and extreme office size. These results support H2b and imply that auditors report more conservatively for larger clients in offices - more conservative in the sense that larger clients have less discretion with respect to accruals.

Table 6 reports the mean and variance of signed total accruals and signed discretionary accruals for high and low influence clients. High influence clients are defined as those above the full sample (n=6,747) median value of INFLUENCE and low influence clients are those below the sample median value. An F-test determines that the variances of discretionary and total accruals are significantly different at p<.01 between high and low influence clients. The direction of these differences indicates that high influence clients have a lower variance in accruals which is consistent with less discretion for large clients. This finding supports H3b and is consistent with other tests indicating that reputation protection leads to auditor conservatism.

Table 6 Test of variance of signed accruals
	Sample Size	Mean	Variance
Discretionary Accruals:
Low Influence Clients **	2,443	-0.0009	0.0244*
High Influence Clients **	2,509	-0.0040	0.0163
Total Accruals:
Low Influence Clients **	2,443	-0.0195	0.0287*
High Influence Clients **	2,509	-0.0210	0.0214
Notes: *Variance in the low influence client segment is significantly different from the variance in the high influence client segment based on an F-test (p<.01). The means in signed accruals are not significantly different between the two groups. **Low influence clients are defined as those in which the variable INFLUENCE is below the full sample median. High influence clients are those in which the variable INFLUENCE is above the full sample median value. The slight difference in the number of observations per group is due to the deletion of observations with missing values for variables in the accrual tests in Tables 4 and 5.

In sum, the tests of H2 and H3 parallel the results for H1 in section 3. Big Five auditors do not report more leniently for larger clients in local practice offices. Instead they appear to report more conservatively. Relatively larger clients in offices have smaller accruals, a result that implies auditors allow these clients less discretion with respect to accruals and, by implication, less discretion in managing reported earnings. These findings are consistent with reputation protection and the avoidance of litigation driving auditor decision-making rather than economic fee dependence

4.4 Further analysis

A further analysis is undertaken to assure that statistical results for the test variable INFLUENCE are not driven by an association between the accruals variables and absolute firm size (SALES), even though SALES is included as a control variable in the regressions and correlations between the accruals variables and SALES are less than 0.10. The models in Tables 4 and 5 are re-estimated separately for the upper and lower half of company size (split at the median value of SALES for the 4,952 observations used for the accruals tests). The re-estimation of Table 4 shows that INFLUENCE remains negatively associated with discretionary accruals at p<.01 for both the smaller and upper halves of the sample. The re-estimation of Table 5 shows that INFLUENCE remains negatively associated with total accruals at p<.01 for the smaller half of the sample, and at a somewhat weaker level (p<.09, two-tail) for the upper half of the sample.

Another size-related test is undertaken by adding an additional term to the models in Tables 4 and 5. This term is created by interacting INFLUENCE with an indicator variable for absolute client size (coded 1 if sales are greater than median sales in the sample). The interaction term is not significant (p>.10) for either total accruals or discretionary accruals while INFLUENCE remains significant at the levels reported in Tables 4 and 5. This test shows that the relation between INFLUENCE and accruals is consistent across the lower and upper halves of absolute size in the sample.

These additional estimations establish the robustness of the results across the spectrum of company size. After controlling for absolute client size in the accruals models in a number of different ways, the experimental variable INFLUENCE is consistently significant which indicates that Big five auditors report more conservatively for relatively larger clients in their practice offices.

An additional analysis of discretionary accruals in Table 4 is made by partitioning observations into two subsamples, those with positive (income-increasing) discretionary accruals and those with negative (income-decreasing) discretionary accruals. The model in Table 4 is then re-estimated on these two subsamples. For observations having positive discretionary accruals, INFLUENCE is negative and significant at p<.05 which means that larger clients have smaller income-increasing accruals. For observations having negative discretionary accruals, INFLUENCE is positive and significant at p<.03 which means larger clients have smaller income-decreasing accruals. If we assume firms with positive (negative) discretionary accruals are managing earnings up (down) then the evidence indicates auditors constrain large clients more than small clients in achieving these earnings management objectives.

A final test is made regarding the robustness of the results to financial distress, even though the Z-score (PBANK) is in the models to control for financial distress. We partition the sample at the median value of PBANK and re-estimate the accruals models in Tables 4 and 5 separately for the upper and lower halves of PBANK. For the more distressed half of the sample, INFLUENCE is significant at p=.03 for discretionary accruals and p=.01 for total accruals. For the less distressed half of the sample, INFLUENCE is significant at p=.01 for discretionary accruals and p=.06 for total accruals. Thus the association between INFLUENCE and accruals is robust across the full spectrum of financial condition in the sample.

5. Discussion

Other studies document that Big Five auditors report more conservatively than non-Big five auditors (Francis and Krishnan, 1999; Basu et al., 2000) and allow clients less discretion with respect to accounting accruals (Becker et al., 1998; Francis et al., 1999a). Our study extends this research by restricting the analysis to Big Five audited companies and redefining the unit of analysis as the local offices of Big Five accounting firms.

This is the first study to analyze the effect of client size on auditor decision-making at the office level where audit contracting occurs and where decisions are made with respect to the audited financial statements (Wallman, 1996). We find no evidence that economic fee dependence causes auditors to be lenient and report more favorably for larger clients relative to smaller clients in offices. In terms of regulation and public policy, our findings support the profession's claim that current incentives with respect to reputation protection and litigation avoidance are sufficient to over-ride the possible impairment of objectivity by the economic fee dependence inherent in auditor-client contracting.

However, auditors do appear to report more conservatively for larger clients in local practice offices. For a sample of potentially distressed firms, larger clients in offices are more likely to receive going concern audit reports. For a more general sample, relatively larger clients in offices have smaller accruals which implies less discretion to manage earnings. Healy and Palepu (1993) argue that managerial discretion is important in facilitating the communication of financial performance. Our findings raise the question of whether auditor conservatism is excessive and exists largely in response to legal liability exposure. This has been suggested by the Big Five accounting firms (Anderson et al., 1992), and Basu et al. (2000) report evidence consistent with more auditor conservatism in periods of greater liability exposure. There is also evidence that some companies respond to what they perceive to be excessive conservatism by switching auditors.⁹ DeFond and Subramanyam (1998) report that companies switching auditors were treated conservatively by their auditor with respect to accruals, and that these companies did have some success in finding a less conservative auditor. Krishnan (1994) finds that companies receiving going concern reports were also treated conservatively by their auditors and likely to switch auditors, though these companies did not succeed in getting a more favorable report from their new auditors (Krishnan and Stephens, 1995).

Our study also has implications for the SEC's contention that consulting services provided by accounting firms for their audit clients may compromise auditor independence and objectivity (Sutton, 1997; Turner and Godwin, 1999; Levitt, 1999; Securities and Exchange Commission, 2000). Consulting services can be viewed as another source of economic fee dependence in addition to audit fees, and several studies report a positive association between audit fees and consulting fees (Simunic, 1984; Craswell et al., 1995). Fee dependence is inherent in auditor-client contracting, and therefore larger clients represent greater economic fee dependence for accounting firms - irrespective of whether the source of fees is auditing or consulting. The evidence here is that economic dependence on large clients at the office level does not compromise auditor decision-making, suggesting there is no obvious need to regulate auditor-client contracting in terms of fee or service restrictions.¹⁰

Finally, the study demonstrates how local offices of Big Five accounting firms can be used effectively as the unit of analysis in auditing research. As Wallman (1996) and others have noted, decision-making with respect to audit clients occurs in local offices of accounting firms. Engagement partners in local offices contract with clients on behalf of the firm, oversee the delivery of audits including coordination with other offices where applicable, and issue the final audit report. Prior audit research has used aggregated firm-level data and therefore is removed from the local office context in which audit decisions are made. Research on a wide range of topics including auditor-client alignments, auditor switching, opinion shopping, audit pricing, and audit reports could benefit from a re-examination using the local office perspective. Finally, given the Big Five domination of the audit market, inter-office differences within the Big Five firms seems a much more promising approach to research on auditor differentiation than the large firm/small firm dichotomy that has dominated prior research.

Footnotes

References

Altman, E., 1983. Corporate Financial Distress: A Complete Guide to Predicting, Avoiding, and Dealing with Bankruptcy. Wiley, New York City.

American Institute of Certified Public Accountants. 1978. The Commission on Auditors' Responsibilities: Report, Conclusions, and Recommendations. American Institute of Certified Public Accountants, New York City.

American Institute of Certified Public Accountants. 1997. Serving the Public Interest: A New Conceptual Framework for Auditor Independence. American Institute of Certified Public Accountants, New York City.

Andersen et al. 1992. The liability crisis in the united states. A joint letter signed by Arthur Andersen, Coopers & Lybrand, Deloitte Touche, Ernst & Young, KPMG Peat Marwick, and Price Waterhouse. Reprinted in Journal of Accountancy (November 1992), 19-23.

Basu, S., Hwang, L., Jan, C. 2000. Differences in conservatism between big eight and non-big eight auditors. Working Paper (Baruch College, City University of New York).

Becker, C., DeFond, M., Jiambalvo, J., Subramanyam, K. 1998. The effect of audit quality on earnings management. Contemporary Accounting Research 15, 1-24.

Bell, T., Tabor, R. 1991. Empirical analysis of audit uncertainty qualifications. Journal of Accounting Research 29, 350-370.

Bonner, S., Palmrose, Z., Young, S. 1998. Fraud type and auditor litigation: an analysis of sec accounting and auditing enforcement releases. The Accounting Review 73, 503-532.

Burroughs, G., Black, C. 1998. Profit sharing in australian big 6 accounting firms: an exploratory study. Accounting Organizations and Society 23, 517-530.

Craswell, A., Francis, J., Taylor, S. 1995. Auditor brand name reputations and industry specializations. Journal of Accounting and Economics 20, 297-322.

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

Dechow, P., Sloan, R., Sweeney, A., 1995. Detecting earnings management. The Accounting Review 70, 193-225.

DeFond, M., Jiambalvo, J., 1994. Debt covenant violation and manipulation of accruals. Journal of Accounting and Economics 17, 145-176.

DeFond, M., Park, C., 1997. Smoothing income in anticipation of future earnings. Journal of Accounting and Economics 23, 115-139.

DeFond, M., Subramanyam, K., 1998. Auditor changes and discretionary accruals. Journal of Accounting and Economics 25, 35-68.

Dopuch, N., Holthausen, R., Leftwich, R., 1987. Predicting audit qualifications with financial and market variables. The Accounting Review 62, 431-454.

Francis, J., 1984. The effect of audit firm size on audit prices: a study of the australian market. Journal of Accounting and Economics 6, 133-151.

Francis, J., Krishnan, J., 1999. Accounting accruals and auditor reporting conservatism. Contemporary Accounting Research 16, 135-165.

Francis, J., Maydew, E., Sparks, H.C., 1999a. The role of big 6 auditors in the credible reporting of accruals. Auditing: A Journal of Practice and Theory 18, 17-34.

Francis, J., Reynolds, J.K., 2000. Do Large Accounting Firms Screen Out Risky Clients? Working paper (University of Missouri).

Francis, J., Stokes, D., Anderson, D., 1999b. City markets as a unit of analysis in audit research and the re-examination of big 6 market shares. Abacus 35, 185-206.

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

Greene, W. 1993. Econometric Analysis. Macmillan, New York City.

Healy, P., 1985. The effect of bonus schemes on accounting decisions. Journal of Accounting and Economics 7, 85-107.

Healy, P., Palepu, K. 1993. The effect of firms' financial disclosure strategies on stock prices. Accounting Horizons 7, 1-11.

Jones, J., 1991. Earnings management during import relief investigations. Journal of Accounting Research 29, 193-228.

Jeter, D., Shaw, P., 1995. Solicitation and auditor reporting decisions. The Accounting Review 70, 293-315.

Kellogg, R., 1984. Accounting activities, securities prices, and class action lawsuits, Journal of Accounting and Economics 6, 185-204.

Krishnan, J., 1994. Auditor switching and conservatism. The Accounting Review 69, 200-215.

Krishnan, J., Krishnan, J., 1996. The role of economic trade-offs in the audit opinion decision: an empirical analysis. Journal of Accounting Auditing and Finance 11, 565-586.

Krishnan, J., Stephens, R., 1995. Evidence on opinion shopping from audit opinion conservatism. Journal of Accounting and Public Policy 14, 179-201.

Levitt, A., 1999. Remarks to the panel on audit effectiveness of the public oversight board. Securities and Exchange Commission, www. sec.gov/news/speeches/spch301.htm.

Louwers, T., 1998. The relation between going-concern opinions and the auditor's loss function. Journal of Accounting Research 36, 143-156.

Lys, T., Watts, R., 1994. Lawsuits against auditors. Journal of Accounting Research 32 (supplement), 65-102.

Mautz, R., Sharaf, H., 1961. The Philosophy of Auditing. American Accounting Association, Sarasota Springs.

Mennon, K., Schwartz, K., 1987. An empirical investigation of audit qualification decisions in the presence of going-concern uncertainties. Contemporary Accounting Research 4, 301-315.

Miller, T., 1992. Do we need to consider the individual auditor when discussing auditor independence? Accounting Auditing and Accountability Journal 5, 74-84.

Mutchler, J., Hopwood, W., McKeown, J., 1997. The influence of contrary information and mitigating factors on audit opinion decisions on bankrupt companies. Journal of Accounting Research 35, 295-310.

Narayanan, V., 1995. Moral hazard in repeated partnerships. Contemporary Accounting Research 11, 895-917.

Feroz , E., Park, K., Pastena, V. 1992. The financial and market effects of the sec's accounting and auditing enforcement releases. Journal of Accounting Research 30 (Supplement), 107-148.

Press, E., Weintrop, J., 1990. Accounting-based constraints in public and private debt agreements: their association with leverage and impact on accounting choice. Journal of Accounting and Economics 12, 65-95.

Securities and Exchange Commission, 2000. The Commission's Proposal to Modernize the Rules Governing the Independence of the Accounting Profession, issued June 27, 2000.

Simunic, D. 1984. Auditing, consulting, and auditor independence. Journal of Accounting Research 22, 679-702

Statement on Auditing Standards No. 59, 1988. American Institute of Certified Public Accountants, New York City.

Statement on Auditing Standards No. 79, 1996. American Institute of Certified Public Accountants, New York City.

Stice, J., 1991. Using financial and market information to identify pre-engagement factors associated with lawsuits against auditors. The Accounting Review 66, 516-533.

Sutton, M., 1997. Auditor independence: the challenge of fact and appearance. Accounting Horizons 11, 86-91.

Turner, L., Godwin, J., 1999. Auditing, earnings management, and international accounting issues at the securities and exchange commission. Accounting Horizons 13, 281-297.

Wallman, S., 1996. The future of accounting, part iii: reliability and auditor independence. Accounting Horizons 10, 76-97.

Warfield, T., Wild, J., Wild, K., 1995. Managerial ownership, accounting choices, and informativeness of earnings. Journal of Accounting and Economics 20, 61-91.

White, H., 1980. A heteroskedasticity-consistent covariance matrix estimator and a direct test for heteroskedasticity. Econometrica 48, 817-838.

Wilson. T., Grimlund, R., 1990. An examination of the importance of an auditor's reputation. Auditing: A Journal of Practice and Theory 9, 43-59.