Exploring the term of the auditor-client relationship and the quality of earnings: a case for...

ABSTRACT: In this study, we document evidence on the relation between auditor tenure and earnings quality using the dispersion and sign of both absolute Jones-model abnormal accruals and absolute current accruals as proxies for earnings quality. Our study is motivated by calls for "mandatory

auditor rotation," which are based on concerns that longer auditor tenure reduces earnings quality. Multivariate results, controlling for firm age, size, industry growth, cash flows, auditor type (Big N versus non-Big N), industry, and year, generally suggest higher earnings quality with longer auditor tenure. We interpret our results as suggesting that, in the current environment, longer auditor tenure, on average, results in auditors placing greater constraints on extreme management decisions in the reporting of financial performance.

Keywords: auditor tenure; earnings quality; audit quality; mandatory rotation.

Data Availability: All data used in this study is publicly available.

I. INTRODUCTION

The issue of earnings quality and the quality of financial statements in general has been the focus of recent Congressional, regulatory, and financial statement user discussions. Recent events have focused the public's eye on earnings quality and on audit quality in general. Numerous legislative proposals at federal and state levels would place limits on the maximum length of a given auditor-client relationship. The proposed limits are based on the notion that extended auditor tenure results in auditor complacency about and possibly complicity in the decisions that management makes regarding the presentation of financial results. While the Sarbanes-Oxley Act of 2002 does not impose mandatory rotation on audit firms, it mandates a study of mandatory rotation by the Comptroller General of the United States to be completed within one year of the passage of the Act.

In this study, we document the relation between earnings quality (using absolute, signed, and raw [unsigned] accruals measures as proxies for earnings quality) and auditor tenure. We assert that earnings quality can be used to draw inferences about audit quality. We do not purport to comment on audit quality in the traditional sense--compliance with Generally Accepted Auditing Standards (GAAS). Rather, we claim that when audit quality is high, auditors constrain the extreme (and presumably self-serving) choices that management would like to make in presenting the financial position of the firm. When audit quality is low, we posit that auditors do not constrain these extreme choices and that in some cases, auditors may even aid management in "pushing the boundaries" of Generally Accepted Accounting Principles (GAAP). (1)

Proponents of mandatory rotation express the belief that poor-quality earnings are associated with extended auditor tenure. Poor-quality earnings are problematic because they can mislead investors, resulting in misallocated resources. In its defense, the accounting profession has argued that mandatory rotation increases audit start-up costs and increases the risk of audit failure because the incoming auditor places increased reliance on the client's estimates and representations in the initial years of an engagement. Over time, auditors can gain firm-specific expertise, which helps them to understand the business and allows them to rely less on management estimates. (2) Managers of firms being audited are also opposed to mandatory rotation because changing auditors is costly. These managers believe that auditors better understand their particular business with experience and managers have concerns about whether a new auditor will have the requisite industry expertise and/or will be able to put forth the additional effort required to audit a new client (Dunham 2002).

In this paper, we provide evidence on the claim made by proponents of mandatory auditor rotation that the current voluntary rotation system is flawed. Because the push for mandatory rotation stems from the belief that extended auditor tenure impairs auditor independence, we assert that understanding the relation between auditor tenure and earnings quality is central to the debate. That is, the veracity of claims made by proponents about the current system should be central to decisions made by regulators, and of interest to auditors and financial statement users. To our knowledge, this is the first broad cross-sectional study to explicitly consider the relation between auditor tenure and earnings quality as proxied for by measures of accounting accruals. We assume that producing higher-quality earnings is the goal of recent legislative proposals and claim that the extent to which auditor tenure either enhances or inhibits earnings quality should be addressed before legislative or regulatory changes impose limits on auditor tenure. Borrowing extensively from previous research that uses accounting accruals as a proxy for earnings quality, we test for an association between auditor tenure and earnings quality for those auditor-client combinations where the auditor-client relationship lasts for at least five years. We do this by examining the relation between absolute, signed, and raw values of both Discretionary and Current Accruals and auditor tenure after controlling for other factors expected to affect the distribution of accruals.

Although we cannot comment on whether mandatory auditor rotation would improve earnings quality, we do provide evidence inconsistent with the claim that earnings quality deteriorates with extended auditor tenure under the current voluntary rotation system. After controlling for firm age, size, industry growth, cash flows, auditor type (Big N versus non-Big N), industry, and year, we find that the magnitude of both Discretionary and Current Accruals decline with longer auditor tenure. Further, we find some evidence that longer auditor tenure is associated with less extreme income-increasing accruals. (3) This suggests that as the relationship lengthens, auditors limit management's ability to use accruals to increase current period earnings. We also find evidence that longer auditor tenure is associated with less extreme income-decreasing accruals. This suggests that as the relationship lengthens, auditors limit management's ability to create reserves to manage future earnings. We find this result particularly interesting because income-decreasing earnings management has received a great deal of attention from regulators and the popular press as of late. (4)

The remainder of the paper is organized as follows. In Section II, we summarize the arguments supporting and opposing mandatory auditor rotation and develop our hypotheses. We discuss our data and empirical results in Section III, and in Section IV we present our conclusions and the implications of our study.

II. HYPOTHESIS DEVELOPMENT

Arguments Supporting Mandatory Auditor Rotation

Whether auditor rotation should be made mandatory is an issue that has been debated for more than 40 years in the U.S. Proponents of mandatory auditor rotation are generally concerned that auditor independence, and thus audit quality, will decrease with increased auditor tenure. Mautz and Sharaf (1961) suggest that extended auditor-client relationships could have a detrimental affect on auditor independence because an auditor's objectivity about a client is reduced with the passage of time. Further, the Metcalf Committee states that "mandatory auditor rotation is a way to bolster auditor independence" (U.S. Senate 1976, 21). More recently, regulators suggest a link between auditor tenure and reductions in earnings quality, and allude to mandatory auditor rotation as a possible solution (U.S. Senate 1977; AICPA 1978; Berton 1991; SEC 1994). It has been suggested that decreased auditor independence could lead to auditors' support for more aggressive accounting choices that "push the boundaries" of GAAP and could ultimately result in a failure to detect material fraud and/or misstatements. Mandatory rotation proponents argue that limiting auditor tenure reduces concerns about deteriorating independence and audit quality. (5) In response to the recent Enron/Anderson audit failure, several bills containing provisions limiting auditor tenure and mandating auditor rotation were proposed in the House and Senate as part of an effort to improve financial reporting and protect investors. (6)

Arguments Opposing Mandatory Rotation

Opponents of mandatory rotation suggest that along with higher audit costs, an increased likelihood of audit failures would result from mandated rotation. New auditors, they argue, lack sufficient knowledge regarding firm-specific risks and, as a consequence, audit failures would likely increase. This argument is consistent with research indicating that a greater proportion of audit failures occur on newly acquired clients (Berton 1991; Petty and Cuganesan 1996) and that auditors' litigation risk is greater in the early years of an engagement (Palmrose 1986b, 1991). Further evidence on auditor tenure and audit quality is provided by the AICPA's Quality Control Inquiry Committee of the SEC Practice Section. The Committee analyzed 406 cases of alleged audit failure between 1979 and 1991 and concluded that allegations of audit failure occur almost three times as often when an audit firm is performing its first or second audit of a given client (AICPA 1992). Finally, the accounting profession argues that uncertainty regarding characteristics of the client increases the potential for audit failures early in the auditor-client relationship (PricewaterhouseCoopers 2002). (7)

Accounting Accruals and Audit Quality

Absent from public discussion and prior research is a cogent picture of the (inter)temporal relation between audit quality and auditor tenure. The results of prior studies, while informative about extreme events (for example, SEC violations and qualified opinions), do not address audit quality for a broad cross-section of firms. However, the relation between auditor tenure and audit quality is of interest in a broader context. By linking earnings quality and audit quality, we attempt to address this relation more generally. While we acknowledge that studying the association between auditor tenure and audit quality in the case of extreme audit failures is important, we suggest that a broader analysis of this relation should be informative to the mandatory rotation debate.

To address the relation between auditor tenure and audit quality, we must first provide a plausible definition of audit quality. We borrow from an extensive prior literature to argue that accounting accruals measures are a plausible descriptor of audit quality. Numerous studies have examined the association between various measures of accruals and proxies for audit quality. These proxies include auditor litigation, qualified audit opinions, audit failures, and auditor conservatism. Results suggest that higher accruals levels are positively associated with eventual auditor litigation (Heninger 2001), the issuance of qualified audit opinions (Bartov et al. 2000), audit failures (Geiger and Raghunandan 2002), and auditor changes (DeFond and Subramanyam 1998), while lower accruals levels are associated with greater auditor conservatism, which has been interpreted as suggestive of higher-quality audits (Becker et al. 1998; Francis et al. 1999; Francis and Krishnan 1999). Based on these findings, we posit that high-quality audits mitigate more extreme management reporting decisions, and suggest that accruals can be used to identify these extreme reporting decisions.

Accruals measures have also been used in numerous accounting studies that consider the relation between earnings management and accruals behavior, and earnings quality and accruals behavior. Early studies used the change in Total Accruals as a measure of management discretion (e.g., Healy 1985; DeAngelo 1986), while later studies adopted and modified the "Jones Model" (Jones 1991). (8) Thus, abnormal discretionary accruals (estimated using variations of the Jones Model) became the accepted proxy for extreme managerial discretion. Various accruals measures have also been used to study earnings quality. These studies are important to the audit quality issue because they generally find that information is contained in specific accruals and/or that earnings quality declines with extreme accruals (e.g., Sloan 1996; Thomas and Zhang 2001; Xie 2001; Dechow and Dichev 2002; Richardson et al. 2002). Both the literatures on earnings management and earnings quality provide evidence that suggests that extreme accruals are less desirable, consistent with the audit quality literature cited above. Thus, the extent to which accruals behavior is associated with auditor tenure under the current voluntary rotation system is informative to parties currently debating mandatory rotation.

We follow prior literature and perform analyses of those accruals measures generally considered to be most representative of managements' discretion in the financial reporting process--Discretionary Accruals and Current Accruals. As in prior work, we not only study the raw and absolute values of these measures, but we also posit that studying the association between auditor tenure and signed accruals is important because auditors could constrain income-increasing accruals to a greater or lesser extent than income-decreasing accruals, and auditor tenure could have differential effects on this constraint. (9) By using multiple measures, we better test the relation between auditor tenure and audit quality. We formulate the following alternative hypotheses regarding the relation between auditor tenure and accounting accruals:

[H.sub.rotation proponent]: Audit quality (as measured by accounting accruals) is decreasing in auditor tenure.

[H.sub.rotation opponent]: Audit quality (as measured by accounting accruals) is increasing in auditor tenure.

We test these general hypotheses to determine which is descriptive of the relation between auditor tenure and audit quality (as we have defined it). (10)

III. DATA AND EMPIRICAL RESULTS

Data

Our initial sample consists of all firm-years from 1988 to 2000 inclusive with sufficient data on the 2001 Compustat annual industrial and research files to estimate accruals. In addition, we require at least the six years of prior data to ensure that any abnormal accruals behavior associated with start-up firms (Teoh et al. 1998a, 1998b) is not attributed to short auditor tenure. Because Collins and Hribar (2002) show that estimating accruals is problematic when firms undergo mergers and acquisitions, we also omit these observations. (11) Finally, we limit our analyses to years beginning in 1988 because cash flow from operations was not available before that date. (12)

We calculate auditor tenure as the number of years that the firm has retained the given auditor, and code auditor changes attributable to audit firm mergers as a continuation of the prior auditor. We determine which observations are in the top or bottom 0.5 percent of the distribution of cash flow from operations, Current Accruals or Jones Model residuals annually and delete them to remove outliers. Note that our research question relates to the relation between the accruals (for a given firm) and auditor tenure. To guard against the possibility that observations where firms switch auditors early in the relationship differ systematically from observations where firms retain the auditor for a number of years, we eliminate all firm-year observations where the auditor-client relationship did not last at least five years. That is, when the auditor-client relationship lasts at least five years, we retain all available observations, and when the auditor-client relationship lasts for less than five years, we eliminate all of these auditor-client observations. In this way, the early and later years represent the same auditor-client combinations, thus preserving comparability. Eliminating the quick turnover group reduces the sample size to 42,302 firm-years.

In our analyses, we use two measures of accruals:

Current Accruals = (([DELTA]CA - [DELTA]Cash) - ([DELTA]CL - [DELTA]STD))

and

Discretionary Accruals = Accruals - [[beta].sub.0] + [[beta].sub.1] [DELTA]Revenue + [[beta].sub.2]PPE] (Jones_Model_residuals) (13)

where:

[DELTA]CA = change in current assets (Compustat annual data item 4);

[DELTA]Cash = change in cash and cash equivalents (Compustat annual date item 1);

[DELTA]CL = change in current liabilities (Compustat annual data item 5);

[DELTA]STD = change in short-term notes and current portion of long-term debt (Compustat annual data item 34);

[[beta].sub.i] = coefficient estimates from an OLS regression by industry-year, where an industry is defined by its two-digit SIC code; to reduce sampling error and to be consistent with prior literature, we eliminate industry-years with fewer than eight firms;

Accruals = operating income after depreciation (Compustat annual data item 178)--cash flow from operations (Compustat annual data item 108);

[DELTA]Revenue = change in net sales revenue (Compustat annual data item 12); and

PPE = property, plant, and equipment--net (Compustat annual data item 8).

We scale all variables by average total assets.

Empirical Results

Table 1, Panel A provides descriptive statistics for all firms in our sample where the auditor-client relationship lasts at least five years, while Panel B provides descriptive statistics for the quick-turnover firms (where the auditor-client relationship lasts for less than five years). Within each panel we present descriptive statistics during three stages of the auditor-client relationship: 1 to 2 years; 3 to 4 years; and more than 4 years (when an auditor-client relationship lasts for at least 5 years). (14) We selected these stages (or groupings) for two reasons. First, drafts of the Congressional bills cited earlier provided a range of suggested mandatory rotation periods, and groups of 3 to 5 years and more than five years are representative of the suggested periods. Second, the argument for mandatory auditor rotation hinges, in part, on the potential for reducing extreme management reporting decisions with unbiased oversight by auditors in the initial years of an auditor-client relationship, while the argument against mandatory auditor rotation hinges, in part, on the uncertainty that occurs during the initial years. Our grouping of 1 to 2 years is representative of this initial period.

The quick-turnover firms are significantly different from firms where the auditor-client relationship lasts for at least five years (i.e., the sample firms). For example, for the subset of firms that are in their first or second year of the audit engagement, the quick-turnover firms have significantly smaller average ROA and Cash Flow, and are older. Results on size proxies are mixed. Table 1 also reveals that the dispersion of our two accruals measures is significantly greater (in the first two years of the audit engagement) for the quick-turnover firms than for the sample firms. For these reasons, and because including quick-turnover firms does not allow us to adequately address the possible consequences of extended auditor tenure, we eliminate the quick-turnover firms in all subsequent analyses.

Univariate Results

In this section, we present univariate results concerning auditor tenure and our accruals measures. Specifically, in Figure 1 we graph the 90th percentile, mean, median, and 10th percentile of our two accruals measures, Discretionary Accruals and Current Accruals, by tenure decile. Here, the 90th percentile represents extreme positive accruals while the 10th percentile represents extreme negative accruals. (15)

[FIGURE 1 OMITTED]

Results in Panel A of Figure 1 suggest that tenure has little effect on the average reported Current Accruals. That is, mean and median Current Accruals do not appear to differ in the early years and later years of tenure. However, tenure does seem to have significant affect on the extremes of the distribution. The extreme deciles of Current Accruals decline fairly steadily with tenure. In fact, after the third decile of tenure, the effect is monotonic.

Results in Panel B of Figure 1, which examine the relation between Discretionary Accruals and auditor tenure, are similar to those in Panel A, but the affect of tenure on large positive Discretionary Accruals is more ambiguous. That is, the monotonic results for Current Accruals are not evident for large positive Discretionary Accruals. Further, the mean and median Discretionary Accruals are slightly higher for firms with long tenure than for firms with short tenure. The most striking finding, however, is the sharp decline in extreme negative Discretionary Accruals as tenure increases.

Overall, Figure 1 reveals that the negative values of both accruals measures are less extreme (i.e., closer to zero) when auditor tenure is greater. Further, positive Current Accruals also exhibit a similar characteristic. This suggests that as tenure lengthens, auditors constrain management's ability to make extreme reporting decisions. (16)

While these results are consistent with the notion that managers record less extreme accruals as auditor tenure lengthens, we have not yet incorporated other factors that have been shown to influence the cross-sectional distribution of accruals. To ensure that auditor tenure is not a proxy for one or more of these factors, we perform multivariate analyses in the next section.

Multivariate Results

In this section, we document the general association between auditor tenure and accruals. Because our interest is in the cross-section, rather than in cases where extreme audit failures occur, we do not condition on events suggested by prior literature as being associated with extreme accruals. With that in mind, we suggest the following general model. (17)

(1) Accruals = [alpha] + [[beta].sub.1]Tenure + [[beta].sub.2]Age + [[beta].sub.3]Size + [[beta].sub.4]IndustryGrowth + [[beta].sub.5]CashFlow + [[beta].sub.6]AuditorType + [[beta].sub.7-69]Industry + [[beta].sub.1998-1999]Year + [epsilon]

where:

Accruals = Discretionary Accruals or Current Accruals (measured in raw, absolute, and positive and negative values) scaled by average total assets;

Tenure = the number of consecutive years that the firm has retained the auditor;

Age = the number of years for which total assets was reported in Compustat since 1980;

Size = the log of total assets;

Industry Growth = [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] [Sales.sub.i,t]/[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] [Sales.sub.i,t-1] by two-digit SIC code;

Cash Flow = the firm's cash flow from operations divided by average total assets;

Auditor Type = an indicator set to 1 if the auditor is a Big N Auditor, and set to 0 otherwise; (18)

Industry = a control based on two-digit SIC codes (62 industries); and

Year = a control for calendar year.

We are primarily concerned with the sign of the coefficient on Tenure. If proponents of mandatory rotation are correct, then we expect the coefficient on Tenure to be positive (suggesting that auditor independence becomes impaired with longer auditor tenure). However, if opponents of mandatory rotation are correct, then we expect the coefficient on Tenure to be negative (suggesting that auditors place greater constraints on management reporting decisions with greater auditor tenure).

We include control variables in the model to reduce the possibility that our auditor tenure results proxy for other cross-sectional determinants of accruals. Specifically, we include Age because accruals differ with changes in firm life cycle (Anthony and Ramesh 1992; Dechow et al. 2001). Size is included because large firms tend to record larger, more stable accruals (Dechow and Dichev 2002). We control for Industry Growth (measured as growth in industry sales) because growth in the industry should be positively correlated with the accruals. Cash Flow is included because firms with higher cash flow from operations are more likely to be better performers (Frankel et al. 2002) and because and accruals and cash flows are negatively correlated on average (Dechow 1994; Sloan 1996). Because prior research suggests that large audit firms tend to be more conservative and their conservatism tends to limit extreme accruals (Becker et al. 1998; Francis et al. 1999; Francis and Krishnan 1999), we include Auditor Type. Finally, we include Industry because the types of accruals vary by industry (Barth et al. 2001) and the magnitude and type could vary by year.

When performing our analyses, we acknowledge that auditors might require managers to record legitimate asset write-offs and/or impairment charges in the initial year of an auditor-client relationship. These write-offs and charges would create large absolute accruals and large negative accruals in the first year, but we posit that these charges are not earnings management. To ensure that these charges do not influence our interpretation of the relation between auditor tenure and audit quality, we eliminate the first year of the audit in all of our multivariate tests. We estimate Model (1) in Table 2 using all remaining sample observations. We present results on the absolute values in Panel A, on the signed (positive and negative) values in Panel B, and on the raw values in Panel C for our two accruals measures. Panel A reveals that the coefficients for Tenure are negative and significant for both Discretionary and Current Accruals. That is, extended auditor tenure reduces the dispersion of both measures, consistent with the notion that longer auditor tenure results in higher audit quality.

However, regulators are not concerned solely with the dispersion in accruals. Rather, they are concerned about the distortion in earnings that can occur because of inappropriate income-increasing accruals and/or inappropriate income-decreasing accruals. Income-increasing accruals can be used to inflate current earnings, while income-decreasing accruals can be used to create "cookie jar reserves," which allow managers to increase future earnings. Whether extended auditor tenure constrains either or both of these types of accruals should be informative to regulators. To address this, we estimate Model (1) using the signed values for each accruals measure in Panel B of Table 2. (19) The results indicate that extended auditor tenure constrains both income-increasing and income-decreasing accruals. That is, the coefficient for Tenure is negative and significant for positive accruals and positive and significant for negative accruals. This suggests that auditor oversight of management decisions is generally balanced in that auditors appear to constrain extreme income-increasing and income-decreasing accruals.

In Table 2, Panel C we estimate Model (1) using the raw values of our accruals measures. The results are mixed and highlight the need to conduct the analyses presented in Panels A and B. The coefficient for Tenure using Discretionary Accruals is significant and positive suggesting that as auditor tenure increases, the level of Discretionary Accruals increases. This is consistent with arguments made by proponents of mandatory rotation. The coefficient for Tenure using Current Accruals is negative but not significant, suggesting no relation between auditor tenure and the level of Current Accruals. However, we posit that regulators (and users of financial statements) should not be as concerned about the level of accruals as about the extremes of the distribution and the directional affect. Thus, looking only at levels (in Panel C) provides a less-than-satisfactory indication of the relation between auditor tenure and audit quality.

In sum, the results in Panels A and B of Table 2 suggest that extended auditor tenure has a beneficial effect on the dispersion of accruals (i.e., accruals are less extreme with extended auditor tenure) and that auditors constrain both income-increasing and income-decreasing accruals as auditor tenure increases.

With respect to the control variables, our analyses reveal no consistent results for Age or Industry Growth, but we do find that as firm size increases, accruals are more positive (consistent with Dechow and Dichev [2002]), and that as firm size increases, both positive and negative accruals are smaller. The coefficient on Cash Flow is significantly negative in all specifications (other than for negative Current Accruals, in which case it is positive), consistent with findings in Dechow (1994) and Sloan (1996). Further, the coefficient for Auditor Type is negative for all Discretionary Accruals measures but is mixed for Current Accruals measures. Our results for the absolute Discretionary Accruals and raw Discretionary and Current Accruals measures are consistent with prior research suggesting that Big N auditor clients report more conservative earnings than do non-Big N auditor clients. On the other hand, results for the absolute Current Accruals and positive and negative Discretionary and Current Accruals measures suggest that Big N auditors constrain the reporting choices of their clients less than non-Big N auditors do.

Next, we re-estimate Model (1) omitting observations in the extreme (top and bottom) deciles of ROA (measured by year and industry) to control for the possibility that the association between auditor tenure and accruals is the result of poor financial performance early in the auditor's tenure (DeFond and Park 1997; DeFond and Subramanyam 1998). These results are presented in Table 3, Panel A (absolute values), Panel B (signed values), and Panel C (raw values) for our two accruals measures. While our results on Tenure for income-decreasing accruals are generally consistent with the results in Table 2, we find insignificant results on income-increasing Discretionary Accruals for the non-extreme ROA subsample. Thus, we conclude that our main results on income-decreasing accruals are not driven by extreme performance, but our main results on income-increasing accruals are influenced by firms with extreme financial performance.

Sensitivity Analyses

To test whether the association between auditor tenure and accruals is affected by our design choices and/or sample composition, we perform a number of additional tests. First, to test whether industry specialization is important, we form an indicator for specialization and interact it with the tenure variable. In this way, we attempt to address whether an auditor's reduced knowledge regarding the client's industry in the early years of an auditor-client relationship (20) has an affect on the association between tenure and earnings quality. Our specialist definition is based on that in Palmrose (1986a). Here, we code an auditor as a (two-digit SIC code) industry specialist if the auditor audits the largest percentage of sales in the industry (i.e., is the lead auditor) or audits sales within 15 percent of the lead auditor. Our analyses reveal no significant differences (in the association between tenure and earnings quality) for specialist versus non-specialist auditors. Thus, our findings on specialization are not consistent with claims made by opponents of mandatory auditor rotation.

Next, to test whether period-specific characteristics around the time of the auditor change are driving our results, we eliminate firm-years representing the last year with the predecessor auditor. Again, our results are robust to this change.

In analysis not reported, we replicate all of our tests without requiring that the auditor audits the client in question for at least five years. The results obtained using this larger sample are generally consistent with those reported but are stronger. Any difference exists because the levels and dispersion of accruals are smaller for the long-tenure group (those where the relationship lasts for at least five years) than for the quick-turnover group (those where the relationship does not last for at least five years) for most measures of accruals. This magnifies our result on the tenure variable, but does not allow us to assess the relation between tenure and accruals as cleanly as does our chosen design.

To ensure that legitimate asset write-downs and impairment charges required by an incoming auditor in the initial year of the engagement do not explain the association between tenure and income-decreasing accruals, we eliminated the first year of an auditor-client relationship. As a sensitivity test, we also estimate Discretionary Accruals after eliminating Special Items. The results are consistent with the results reported in Tables 2 and 3.

Finally, the results in Panel B of Tables 2 and 3 are the result of estimating Model (1) using a truncated regression approach. In analysis not reported, we also estimate the model using ordinary least squares. The results are similar to those presented in the tables but confirm the bias due to truncation above (below) zero.

IV. CONCLUSIONS AND IMPLICATIONS

In this study, we examine the cross-sectional association between auditor tenure and audit quality. We use two measures of earnings quality to proxy for audit quality. These measures of earnings quality are drawn from prior literature that examines how accruals vary around events such as SEC violations and qualified opinions. These studies find that the level of accruals around the examined events is generally higher than in the absence of these events. We investigate the extent to which auditor tenure is associated with the dispersion in accruals and whether the recognition of income-increasing or income-decreasing accruals varies with auditor tenure. We contribute to the literatures on earnings quality and audit quality by documenting that longer auditor tenure is generally associated with less dispersion in the distributions of Discretionary Accruals and Current Accruals and that auditors appear to place greater constraints on both income-increasing and income-decreasing accruals as the auditor-client relationship lengthens.

Our study also provides useful evidence to regulators, members of the accounting profession, and financial statement users as they deliberate on the costs and benefits of mandatory auditor rotation. Specifically, our results suggest that under the current system, increased auditor tenure does not lead to reduced audit and earnings quality. If deteriorating earnings and/or audit quality are the motivation for calls for mandatory rotation, then our results do not support such an argument. However, our results also do not imply that forcing firms to remain with the same auditors would improve earnings quality or audit quality. Rather, Johnson and Lys (1990) show that allowing auditor turnover can improve audit quality because clients may switch to a more efficient provider (i.e., a new auditor) as their needs change. In this case, requiring a client to retain an auditor who is unable to provide adequate services could reduce audit quality. Further, we do not provide evidence on whether mandatory auditor rotation would improve or hamper audit quality. We merely provide evidence that under the current system of voluntary auditor rotation, audit quality does not appear to deteriorate with tenure.

Finally, our study does not address all instances of earnings management and our tests do not condition on managers' incentives to manage earnings. Therefore, additional avenues exist to provide evidence on the affects of extended auditor tenure on the quality of financial reporting. (21)

TABLE 1
Descriptive Statistics by Tenure Group

                             Cash
                ROA          Flow         MVE         Assets
                Mean         Mean         Mean         Mean
Tenure        (Median)     (Median)     (Median)     (Median)
Group           [n]          [n]          [n]          [n]

Panel A: Omitting Observations Where the Auditor-Client
Relationship Lasts for Less than Five Years

Group 1       -0.0324       0.0394       781.11      1,153.58
  1 or 2      (0.0237)     (0.0635)     (32.868)     (41.201)
  years       [2,637]      [2,637]      [2,174]      [2,637]
Group 2       -0.0407       0.0293       991.15      1,463.79
  3 or 4      (0.0249)     (0.0631)     (41.135)     (49.897)
  years       [3,730]      [3,730]      [3,115]      [3,730]
Group 3       -0.0193       0.0477      1,788.15     2,329.66
  >4 years    (0.0316)     (0.0723)     (93.283)    (129.472)
              [35,935]     [35,935]     [31,124]     [35,935]

Panel B: Observations Where the Auditor-Client Relationship
Lasts for Less than Five Years

Group 1       -0.0939       0.0017      1,024.55     1,721.13
  1 or 2      (0.0077)     (0.0457)     (29.176)     (46.152)
  years       [6,844]      [6,844]      [5,851]      [6,844]
Group 2       -0.1048      -0.0080       827.15      1,246.54
  3 or 4      (0.0043)     (0.0388)     (22.586)     (33.953)
  years       [1,869]      [1,869]      [1,594]      [1,869]

Panel C: Tests of Differences between Observations Where the
Auditor-Client Relationship Lasts at Least Five Years and
Observations Where the Auditor-Client Relationship Lasts for
Less than Five Years

t-statistics for Tests of Differences in Means

(Z-statistics for Tests of Differences in Medians)

Group 1       7.04 ***     7.91 ***    -1.98 *       -3.13 ***
  1 or 2     (8.65 ***)   (5.90 ***)   (1.76)       (-1.71)
  years
Group 2       5.90 ***     5.13 ***     0.97         0.83
  3 or 4     (8.35 ***)   (6.82 ***)   (6.64 ***)   (4.78 ***)
  years

                                           Absolute       Absolute
                                         Discretionary     Current
               Tenure          Age         Accruals       Accruals
                Mean          Mean           Mean           Mean
Tenure        (Median)      (Median)       (Median)       (Median)
Group            [n]           [n]            [n]            [n]

Panel A: Omitting Observations Where the Auditor-Client
Relationship Lasts for Less than Five Years

Group 1        1.6011        10.7144        0.0918         0.1582
  1 or 2         (2)          (10)         (0.0553)       (0.0702)
  years        [2,637]       [2,637]        [2,637]        [2,637]
Group 2        3.5357        11.4434        0.0917         0.1565
  3 or 4         (4)          (11)         (0.0524)       (0.0677)
  years        [3,730]       [3,730]        [3,730]        [3,730]
Group 3        10.4634       12.1084        0.0812         0.1313
  >4 years      (10)          (11)         (0.0484)       (0.0592)
              [35,935]      [35,935]       [35,935]       [35,935]

Panel B: Observations Where the Auditor-Client Relationship
Lasts for Less than Five Years

Group 1        1.3494        11.4566        0.1091         0.1763
  1 or 2         (1)          (10)         (0.0622)       (0.0793)
  years        [6,844]       [6,844]        [6,844]        [6,844]
Group 2        3.3087        11.6249        0.1163         0.1992
  3 or 4         (3)          (11)         (0.0723)       (0.0817)
  years        [1,869]       [1,869]        [1,869]        [1,869]

Panel C: Tests of Differences between Observations Where the
Auditor-Client Relationship Lasts at Least Five Years and
Observations Where the Auditor-Client Relationship Lasts for
Less than Five Years

t-statistics for Tests of Differences in Means

(Z-statistics for Tests of Differences in Medians)

Group 1       22.59 ***     -8.84 ***     -6.55 ***       -2.26 **
  1 or 2     (22.25 ***)   (-1.37)       (-3.27) ***     (-3.64) ***
  years
Group 2       16.87 ***     -1.53         -6.76 ***       -3.62 ***
  3 or 4     (16.07 ***)   (-1.83 ***)   (-7.52 ***)     (-3.78 ***)
  years

*, **, *** Indicates significant at .10, .05, and .01, respectively,
using a two-tailed test.

Observations relating to a given auditor-client combination will
appear in Panel A if and only if that combination lasted for at
least five years. However, if the auditor-client combination lasted
ten years, the first two years will be included in the Group 1
observations, the next two years will be included in Group 2
observations, and the final six years will be included in Group 3
observations. Observations relating to a given auditor-client
combination will appear in Panel B if and only if that combination
lasted for less that five years. The groupings can be interpreted
as in Panel A.

TABLE 2
Regressions of Accruals Measures on Auditor Tenure and Control
Variables All Observations in the Sample

Accruals = [alpha] + [[beta].sub.1]Tenure + [[beta].sub.2]Age +
[[beta].sub.3]Size + [[beta].sub.4]IndustryGrowth +
[[beta].sub.5]CashFlow + [[beta].sub.6]AuditorType
+ [[beta].sub.7-69]Industry + [[beta].sub.1988-1999]Year
+ [epsilon]

Panel A: Absolute Value Measures

Variable          Discretionary Accruals   Current Accruals

Intercept                  0.19275               0.25324
                        [< 0.0001]            [< 0.0001]
Tenure                    -0.00037              -0.00142
                          [0.0041]              [0.0004]
Age                       -0.00038              -0.00393
                          [0.0121]            [< 0.0001]
Size                      -0.01148              -0.01032
                        [< 0.0001]            [< 0.0001]
Industry Growth           -0.00305               0.00432
                          [0.2348]              [0.5807]
Cash Flow                 -0.05453              -0.18027
                        [< 0.0001]            [< 0.0001]
Auditor Type              -0.00356               0.01283
                          [0.0146]              [0.0039]
n                         41,250                41,250

Panel B: Signed Measures

              Positive        Negative       Positve      Negative
            Discretionary   Discretionary    Current      Current
Variable      Accruals        Accruals       Accruals     Accruals

Intercept        0.19450        -0.5468       -0.2555       0.6487
              [< 0.0001]        [0.4689]    [< 0.0001]     [0.7444]
Tenure          -0.00146         0.0534       -0.0293       0.2463
                [0.0502]      [< 0.0001]    [< 0.0001]   [< 0.0001]
Age              0.00004         0.0803       -0.0526       0.4188
                [0.9610]      [< 0.0001]    [< 0.0001]   [< 0.0001]
Size            -0.09200         0.8027       -0.1102       1.1682
              [< 0.0001]      [< 0.0001]    [< 0.0001]   [< 0.0001]
Industry        -0.00830         1.0595        0.0986       0.1010
  Growth      [< 0.0001]        [0.0006]      [0.0284]     [0.7942]
Cash Flow       -0.11950        -0.1302       -0.6248       0.9005
              [< 0.0001]        [0.0709]    [< 0.0001]   [< 0.0001]
Auditor         -0.00882        -0.4710        0.0658      -0.7464
Type            [0.2650]      [< 0.0001]     [0.0124]    [< 0.0001]
n               21,449          19,801       21,619        19,631

Panel C: Raw Measures

Variable          Discretionary Accruals   Current Accruals

Intercept                 -0.03712             -0.00154
                          [0.0039]             [0.9619]
Tenure                     0.00038             -0.00043
                          [0.0330]             [0.3359]
Age                        0.00269              0.00009
                        [< 0.0001]             [0.8581]
Size                       0.00077              0.00387
                          [0.0228]           [< 0.0001]
Industry Growth            0.00142             -0.00121
                          [0.6869]             [0.8914]
Cash Flow                 -0.07104             -0.10073
                        [< 0.0001]           [< 0.0001]
Auditor Type              -0.02205             -0.01294
                        [< 0.0001]             [0.0103]
n                         41,250               41,250

Each cell contains a coefficient estimate and p-values in square
brackets. Omitting observations where the auditor-client
relationship lasts for less than five years, observations where
firms reported a merger or acquisition in the year, and the first
year of an auditor-client relationship.

       Accruals = one of the accruals measures presented in
                  Figure 1;
         Tenure = the number of consecutive years that the firm
                  has retained the auditor;
            Age = the number of years of data available on
                  Compustat since 1980;
           Size = the log of total assets;
Industry Growth = [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]
                  by two-digit SIC code;
      Cash Flow = the firm's cash flow from operations divided by
                  average total assets;
   Auditor Type = an indicator set to 1 if the auditor is a Big N
                  auditor, and set to 0 otherwise;
       Industry = a control based on two-digit SIC codes (62
                  industries); and
           Year = a control based on calendar year.

TABLE 3
Regressions of Accruals Measures on Auditor Tenure and Control
Variables Eliminating Observations with Extreme ROA

Accruals = [alpha] + [[beta].sub.1]Tenure + [[beta].sub.2]Age
+ [[beta].sub.3]Size + [[beta].sub.4]IndustryGrowth +
[[beta].sub.5]CashFlow + [[beta].sub.6]AuditorType +
[[beta].sub.7-69]Industry + [[beta].sub.1988-1999]Year
+ [epsilon]

Panel A: Absolute Value Measures

Variable          Discretionary Accruals   Current Accruals

Intercept                  0.10165               0.19663
                        [< 0.0001]            [< 0.0001]
Tenure                    -0.00029              -0.00093
                          [0.0004]              [0.0034]
Age                       -0.00066              -0.00296
                        [< 0.0001]            [< 0.0001]
Size                      -0.00608              -0.00542
                        [< 0.0001]            [< 0.0001]
Industry Growth           -0.00093               0.00580
                          [0.5687]              [0.3524]
Cash Flow                 -0.03480              -0.31820
                        [< 0.0001]            [< 0.0001]
Auditor Type               0.00170               0.00570
                          [0.0764]              [0.1208]
n                         37,571                37,571

Panel B: Signed Measures

                 Positive        Negative       Positive     Negative
               Discretionary   Discretionary    Current      Current
Variable         Accruals        Accruals       Accruals     Accruals

Intercept         -0.01210        -0.03000       -0.2749       8.9784
                  [0.6106]        [0.7367]     [< 0.0001]    [0.0106]
Tenure            -0.00012         0.00364       -0.0367       0.1771
                  [0.7020]        [0.0021]     [< 0.0001]   [< 0.0001]
Age                0.00153         0.01160       -0.0536       0.4614
                [< 0.0001]      [< 0.0001]     [< 0.0001]   [< 0.0001]
Size              -0.01200         0.06760       -0.0868       0.8708
                [< 0.0001]      [< 0.0001]     [< 0.0001]   [< 0.0001]
Industry           0.01450         0.08300        0.1406       0.5380
  Growth          [0.0590]        [0.0044]     [< 0.0001]    [0.2992]
Cash Flow         -0.61090        -0.82370       -2.4576      12.3452
                [< 0.0001]      [< 0.0001]     [< 0.0001]   [< 0.0001]
Auditor Type      -0.00151        -0.04730       -0.0145      -1.2009
                  [0.6378]      [< 0.0001]       [0.1179]     [0.0011]
n                 18,550          19,021         19,958       17,613

Panel C: Raw Measures

Variable          Discretionary Accruals   Current Accruals

Intercept                 -0.03388               0.03202
                        [< 0.0001]              [0.2360]
Tenure                     0.00041              -0.00032
                          [0.0002]              [0.3792]
                           0.00222               0.00033
Age                     [< 0.0001]              [0.4406]
Size                       0.00361               0.00318
                        [< 0.0001]            [< 0.0001]
Industry Growth            0.00306               0.00767
                          [0.1555]              [0.2828]
Cash Flow                 -0.36939              -0.40922
                        [< 0.0001]            [< 0.0001]
Auditor Type              -0.00688              -0.00450
                        [< 0.0001]              [0.2852]
n                         37,571                37,571

Each cell contains a coefficient estimate and p-values in square
brackets.

Omitting observations where the auditor-client relationship
lasts for less than five years, observations where firms
reported a merger or acquisition in the year, and the first
year in the auditor-client relationship.

       Accruals = one of the accruals measures presented in
                  Figure 1;
         Tenure = the number of consecutive years that the
                  firm has retained the auditor;
            Age = the number of years of data available on
                  Compustat since 1980;
           Size = the log of total assets;
Industry Growth = [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]
                  by two-digit SIC code;
      Cash Flow = the firm's cash flow from operations divided by
                  average total assets;
   Auditor Type = an indicator set to 1 if the auditor is a Big N
                  auditor, and set to 0 otherwise;
       Industry = a control based on two-digit SIC codes (62
                  industries); and
           Year = a control based on calendar year.

We thank Mark Bradshaw, Mark Peecher, Marjorie Shelley, and two anonymous reviewers for their insightful comments. All errors and omissions are our own.

Editor's note: This paper was accepted by Terry Shevlin, Senior Editor.

(1) For a more thorough discussion of audit quality and auditor independence, see Levitt (2000).

(2) For example, while Chairman of the SEC, Wallman wrote that "[p]eriodically rotating the audit firms of a public company seems contrary to the notion of learning as much as possible about an audit client. It also would appear to be remarkably inefficient." (See Wallman, 1996, 93.)

(3) Specifically, we find a significantly negative association between tenure and Positive Current Accruals in all specifications. We also find a significantly negative association between tenure and Positive Discretionary Accruals when all sample firm-years are included in our analyses, but when we omit firms with extremely positive and negative return on assets, the sign on tenure remains negative but the result is not significant.

(4) See, for example, Levitt (1998).

(5) Research by Deis and Giroux (1992) suggests a negative relation between auditor tenure and audit quality, consistent with concerns about the current system raised by proponents of mandatory auditor rotation. However, their results may have little bearing on the current debate because their sample includes only small CPA firms and the audit clients represent quasi-governmental entities in the public sector.

(6) See, for example, the Integrity in Auditing Act of 2002 (which suggests that auditors should not be considered independent if they have audited a firm for more than seven consecutive years), the Comprehensive Investor Protection Act of 2002 (which suggests that non-independence is a problem when the auditor has consecutively audited a firm's financial statements for more than four years), and the Truth and Accountability in Accounting Act of 2002 (which states that an auditor should not be considered independent if it has audited a firm's financial statements for more than seven consecutive years).

(7) Prior research suggests two reasons why auditors are likely to be less aggressive in their oversight of management early in an auditor's tenure. First, Geiger and Raghunandan (2002) suggest that one reason new auditors are less aggressive in their oversight is because they are attempting to recoup losses from the competitive practice of low-bailing. However, while the practice of low-bailing has been documented by Simon and Francis (1988), Ettredge and Greenberg (1990), and Deis and Giroux (1996), the literature suggests that low-bailing and the recouping of losses does not impair auditor independence (DeAngelo 1981, 113; Lee and Gu 1998, 534). Thus, if auditor independence is not impaired, auditor oversight of managements' decisions should not be affected regardless of tenure. Second, Geiger and Raghunandan (2002) also suggest that new auditors are likely to be less aggressive because they lack knowledge of client-specific risks. Although they find that auditors are less likely to issue going concern opinions for new clients, this does not provide convincing evidence on either the relation between low-bailing and audit quality or the lack of client-specific knowledge. That is, their result can be interpreted that the auditors failed to issue going concern opinions to increase the likelihood that they would recoup some or all of the fees lost in the initial engagement period or that new auditors were unaware of the client-specific risks and this lack of knowledge lead to the issuance of nonqualified opinions.

(8) For an extensive review of the earnings management literature, see Healy and Wahlen (1999) and Dechow and Skinner (2000).

(9) For example, the perception that litigation is more likely to occur when income is overstated might lead auditors to constrain income-increasing accruals more severely than income-decreasing accruals even if auditors become less independent of their clients as the relationship lengthens. Tests on the signed accruals rather than on their absolute values allow us to examine this possibility.

(10) In related work, Davis et al. (2002) provides results suggesting that discretionary accruals increase with auditor tenure. However, differences in design (noted below) make it difficult to compare this study with ours. First, we examine the absolute value of two accruals measures and separately examine the positive and negative components of these measures. Second, we do not require that firms survive throughout our entire sample period, thus eliminating the survivorship bias in Davis et al. (2002). Third, we delete observations that limit our ability to generalize our findings. These include: (1) very young firms, so that abnormal accruals behavior associated

Footnote 10, continued

   with start-up firms (Teoh et al. 1998a, 1998b) is not attributed
   to short auditor tenure; (2) firms undergoing mergers and
   acquisitions, so that the difficulties in reliably estimating
   accruals for these firms (Collins and Hribar 2002) do not
   influence our findings; and (3) observations where firms change
   auditors in the first five years of an audit engagement,
   because their inclusion does not allow us to directly address
   the question of what happens to accruals with lengthy auditor
   tenure.

(11) To identify firms that undergo mergers and acquisitions, we follow Collins and Hribar (2002) and use the footnote codes appearing in Compustat.

(12) The number of new auditor-client combinations each year ranges between 306 (in 1994) and 527 (in 2000). The number of new auditor-client combinations for Big N auditors is 3,590 versus 1,561 for non-Big N auditors.

(13) We estimate the residuals and the coefficients contemporaneously. That is, for each year, we estimate the model within two-digit SIC industries, and use the residuals from this regression as our estimate of discretionary accruals.

(14) Note that observations relating to a given auditor-client combination will appear in Panel A if and only if that combination lasted for at least five years. However, if the auditor-client combination lasted for ten years, the first two years will be included in the Group 1 observations, the next two years will be included in the Group 2 observations, and the final six years will be included in the Group 3 observations. Observations relating to a given auditor-client combination will appear in Panel B if and only if that combination lasted for less that five years. The groupings can be interpreted as in Panel A.

Omitting observations where the auditor-client relationship lasts for less than five years.

Current Accruals = (([DELTA]CA - [DELTA]Cash) - ([DELTA]CL - [DELTA]STD))

where:

[DELTA]CA = change in current assets (Compustat annual data item 4);

[DELTA]Cash = change in cash and cash equivalents (Compustat annual date item 1);

[DELTA]CL = change in current liabilities (Compustat annual data item 5); and

[DELTA]STD = change in short-term notes and current portion of long-term debt (Compustat annual data item 34).

Discretionary Accruals = Accruals - [[beta].sub.0] + [[beta].sub.1][DELTA]Revenue + [[beta].sub.2]PPE

where:

Accruals = operating income after depreciation (Compustat annual data item 178)--cash flow from operations (Compustat annual data item 108);

[[beta].sub.i] = coefficient estimates from an OLS regression by industry-year, where an industry is defined by its two-digit SIC code;

[DELTA]Revenue = change in net sales revenue (Compustat annual data item 12); and

PPE = property, plant, and equipment--net (Compustat annual data item 8).

All variables are scaled by average total assets.

(15) We group firms by tenure decile (rather than by year of tenure) because there are fewer observations with very long tenure (i.e., greater than 15 years) than with very short tenure (i.e., 1 to 5 years) leading to differing precision levels for the accruals estimates of short- and long-tenure firms. A small number of very long-tenure firms would lead to high variance in the different measures of accruals. Aggregating into deciles mitigates this problem. The lower bound for each decile (1 through 10) is 1, 3, 5, 6, 7, 9, 10, 11, 13, and 16 years, respectively.

(16) Further, the mean absolute Discretionary (Current) Accruals as a percentage of Total Assets decline with auditor tenure. For the 1st, 5th, and 10th deciles of auditor tenure, the mean absolute Discretionary (Current) Accruals are 0.09 (0.16), 0.09 (0.14), and 0.07 (0.08), respectively.

(17) Firm and year subscripts are suppressed for presentation.

(18) This measure varies with mergers of the large accounting firms; the Big N is the largest number of major accounting firms considered.

(19) Here we use a truncated regression approach because the ordinary least squares (OLS) estimates are generally biased toward zero when a sample is truncated. Thus, we estimate a maximum likelihood truncated regression model to arrive at an unbiased coefficient on Tenure. See Greene (2000, 682-690) or Maddala (1977, 269-273) for a description of the truncated regression approach.

(20) Recall that this is one of the arguments against mandatory rotation.

(21) For example, are restatements and Enforcement Actions (which have been interpreted as evidence of previous audit failures) more or less likely with longer audit tenure? Does audit tenure affect management's ability to manage earnings to avoid debt covenant violations, or to avoid losses, earnings declines, or missing analyst earnings forecasts?

REFERENCES

American Institute of Certified Public Accountants (AICPA). 1978. The Commission on Auditors Responsibilities: Report, Conclusions and Recommendations. New York, NY: AICPA.

--. 1992. Statement of Position Regarding Mandatory Rotation of Audit Firms of Publicly Held Companies. New York, NY: AICPA.

Anthony, J. H., and K. Ramesh. 1992. Association between accounting performance measures and stock prices: A test of the life cycle hypothesis. Journal of Accounting and Economics 15 (June/September): 203-227.

Barth, M. E., D. P. Cram, and K. K. Nelson. 2001. Accruals and the prediction of future cash flows. The Accounting Review 76 (January): 27-58.

Bartov, E., F. A. Gul, and J. S. L. Tsui. 2000. Discretionary-accruals models and audit qualifications. Journal of Accounting and Economics 30 (December): 421-452.

Becker, C. L., M. L. DeFond, J. J. Jiambalvo, and K. R. Subramanyam. 1998. The effect of audit quality on earnings management. Contemporary Accounting Research 15 (Spring): 1-24.

Berton, L. 1991. GAO weighs auditing plan for big banks. Wall Street Journal (March 27): A3.

Collins, D. W., and P. Hribar. 2002. Errors in estimating accruals: Implications for empirical research. Journal of Accounting Research 40 (March): 105-134.

Davis, L. R., B. Soo, and G. Trompeter. 2002. Auditor tenure, auditor independence, and earnings management. Working paper, Boston College, Chestnut Hill, MA.

DeAngelo, L. E. 1981. Auditor independence, "low-bailing" and disclosure regulation. Journal of Accounting and Economics 3 (August): 113-127.

--. 1986. Accounting numbers as market valuation substitutes: A study of management buyouts of public stockholders. The Accounting Review 61 (July): 400-420.

Dechow, P. M. 1994. Accounting earnings and cash flows as measures of firm performance: The role of accounting accruals. Journal of Accounting and Economics 18 (July): 3-42.

--, and D. Skinner. 2000. Earnings management: Reconciling the views of accounting academics, practitioners, and regulators. Accounting Horizons 14 (June): 235-250.

--, S. A. Richardson, and I. Tuna. 2001. Are benchmark beaters doing anything wrong? Working paper, University of Michigan, Ann Arbor, MI.

--, and I. D. Dichev. 2002. The quality of accruals and earnings: The role of accrual estimation errors. The Accounting Review 77 (Supplement): 35-59.

DeFond, M. L., and C. W. Park. 1997. Smoothing income in anticipation of future earnings. Journal of Accounting and Economics 32 (July): 115-139.

--, and K. R. Subramanyam. 1998. Auditor changes and discretionary accruals. Journal of Accounting and Economics 25 (February): 35-67.

Deis, D. R., and G. A. Giroux. 1992. Determinants of audit quality in the public sector. The Accounting Review 67 (July): 462-479.

--, and --. 1996. The effect of auditor changes on audit fees, audit hours, and audit quality. Journal of Accounting and Public Policy 15 (Spring): 55-76.

Dunham, K. J. 2002. Firms that want to switch auditors find it takes time, money and faith. Wall Street Journal (March 15).

Ettredge, M., and R. Greenberg. 1990. Determinants of fee cutting on initial audit engagements. Journal of Accounting Research 28 (Spring): 198-210.

Francis, J. R., and J. Krishnan. 1999. Accounting accruals and auditor reporting conservatism. Contemporary Accounting Research 16 (Spring): 135-165.

--, E. L. Maydew, and H. C. Sparks. 1999. The role of Big 6 auditors in the credible reporting of accruals. Auditing: A Journal of Practice & Theory 18 (Fall): 17-34.

Frankel, R. M., M. F. Johnson, and K. K. Nelson. 2002. The relation between auditors' fees for nonaudit services and earnings quality. The Accounting Review 77 (Supplement): 71-105.

Geiger, M., and K. Raghunandan. 2002. Auditor tenure and audit quality. Auditing: A Journal of Practice & Theory 21 (March): 187-196.

Greene, W. H. 2000. Econometric Analysis. Englewood Cliffs, NJ: Prentice Hall.

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

--, and J. A. Wahlen. 1999. A review of the earnings management literature and it implications for standard setting. Accounting Horizons 13 (December): 365-383.

Heninger, W. G. 2001. The association between auditor litigation and abnormal accruals. The Accounting Review 76 (January): 111-126.

Johnson, W. B., and T. Lys. 1990. The market for audit services: Evidence from voluntary auditor changes. Journal of Accounting and Economics 12 (January): 281-308.

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

Lee, C. J., and Z. Gu. 1998. Low balling, legal liability and auditor independence. The Accounting Review 73 (October): 533-555.

Levitt, A. 1998. The numbers game. Remarks delivered at the NYU Center for Law and Business, NY, September 28. Available at: http://www.sec.gov/news/speech/speecharchive/1998/ spch220.txt.

--. 2000. Renewing the covenant with investors. Remarks delivered at the NYU Center for Law and Business, NY, May 10. Available at: http://www, sec.gov/news/speech/spch370.htm.

Maddala, G. S. 1997. Introduction to Econometrics. New York, NY: MacMillan Publishing Company.

Mautz, R. K., and H. A. Sharaf. 1961. The Philosophy of Auditing. American Accounting Association Monograph No. 6. Sarasota, FL: American Accounting Association.

Palmrose, Z-V. 1986a. Audit fees and auditor size: Further evidence. Journal of Accounting Research 24 (Spring): 97-110.

--. 1986b. Litigation and independent auditors: The role of business failures and management fraud. Auditing: A Journal of Practice & Theory (Fall): 90-113.

--. 1991. Trials of legal disputes involving independent auditors: Some empirical evidence. Journal of Accounting Research (Supplement): 149-185.

Petty, R., and S. Cuganesan. 1996. Auditor rotation: Framing the debate. Australian Accountant 66 (May): 40-41.

PricewaterhouseCoopers LLP. 2002. Mandatory Rotation of Audit Firms: Will It Improve Audit Quality? New York, NY: PricewaterhouseCoopers LLP.

Richardson, S., R. G. Sloan, M. T. Soliman, and I. Tuna. 2002. Information in accruals about the quality of earnings. Working paper, University of Michigan, Ann Arbor, MI.

Securities Exchange Commission (SEC). 1994. Staff Report on Auditor Independence. Washington, D.C.: Government Printing Office.

Simon, D. T., and J. R. Francis. 1988. The effects of auditor change on audit fees: Tests of price cutting and price recovery. The Accounting Review 63 (April): 255-269.

Sloan, R. 1996. Do stock prices fully reflect information in accruals and cash flows about future earnings? The Accounting Review 71 (July): 289-315.

Teoh, S. H., I. Welch, and T. J. Wong. 1998a. Earnings management and the long-run market performance of initial public offerings. The Journal of Finance 53 (December): 1935-1974.

--, T. J. Wong, and G. Rao. 1998b. Are accruals during initial public offerings opportunistic? Review of Accounting Studies 3: 175-208.

Thomas, J. K., and H. Zhang. 2001. Inventory changes and future returns. Working paper, Columbia Business School, New York, NY.

U.S. Senate. 1976. The Accounting Establishment. Prepared by the Subcommittee on Reports, Accounting and Management of the Committee on Governmental Affairs. Washington, D.C.: Government Printing Office.

--.1977. Improving the Accountability of Publicly Owned Corporations and Their Auditors. Prepared by the Subcommittee on Reports, Accounting and Management of the Committee on Governmental Affairs. Washington, D.C.: Government Printing Office.

Wallman, S. 1996. The future of accounting, Part III: Reliability and auditor independence. Accounting Horizons 10 (December): 76-97.

Xie, H. 2001. The mispricing of abnormal accruals. The Accounting Review 76 (July): 357-373.

Submitted June 2002

Accepted February 2003

James N. Myers

Linda A. Myers

University of Illinois at Urbana-Champaign

Thomas C. Omer

University of Illinois at Chicago