RESEARCH PAPER SERIES

Transcription

1 RESEARCH PAPER NO. 1696R Auditor Independence and Earnings Quality Richard M. Frankel Marilyn F. Johnson Karen K. Nelson August 2001 RESEARCH PAPER SERIES GRADUATE SCHOOL OF BUSINESS STANFORD UNIVERSITY

2 Auditor Independence and Earnings Quality Richard M. Frankel MIT Sloan School of Business 50 Memorial Drive, E52.325g Cambridge, MA (617) Marilyn F. Johnson Michigan State University Eli Broad Graduate School of Management N270 Business College Complex East Lansing, MI (517) Karen K. Nelson Stanford University Graduate School of Business Stanford, CA (650) August 2001 We thank Bill Beaver and participants at the 2001 Stanford Summer Accounting camp for helpful discussions and comments. Karen Nelson acknowledges the support of the Financial Research Initiative at Stanford University Graduate School of Business. We thank First Call for providing the analyst forecast data. Waqas Nazir and Nora Richardson provided valuable data collection assistance.

3 Auditor Independence and Earnings Quality Abstract: We examine whether the provision of non-audit services by auditors is negatively correlated with firm value and the quality of earnings. Because of concerns regarding the effect of non-audit services on auditor independence, the Securities and Exchange Commission recently issued revised auditor independence rules requiring firms to disclose in their annual proxy statement the amount of fees paid to auditors for audit and non-audit services. Using data collected from over 4,000 proxies filed between February 5, 2001 and June 15, 2001, we find a significant negative market reaction to proxy statements filed by firms reporting higher than expected non-audit fees. Our evidence also indicates an inverse relation between the magnitude of non-audit fees and earnings management. Firms purchasing more non-audit services from their auditor are more likely to just meet or beat three earnings benchmarks analysts expectations, prior year earnings, and zero earnings and to report large discretionary accruals. Keywords: Auditor independence, Auditor fees, Earnings management, Discretionary accruals Data Availability: The data used in this study are from the public sources identified in the text.

4 I. INTRODUCTION The purpose of this paper is to examine the relation between non-audit services and earnings quality. On November 15, 2000, the Securities and Exchange Commission (SEC) adopted revised auditor independence rules requiring proxy statement disclosure of all audit and non-audit fees paid by a firm to its auditor. Motivating the new rules is the SEC s belief that the high percentage of revenue auditors derive from the provision of non-audit services poses a threat to auditor independence (e.g., SEC 2000). The audit industry disputes this claim, in part because it is not supported by empirical evidence (Antle, Griffin, Teece, and Williamson 1997). 1 We examine two issues relevant to the ongoing debate over the provision of non-audit services by a firm s financial auditor. First, we measure the market reaction to the disclosure of auditor fees to determine if they convey value-relevant information to investors. Second, we investigate whether the magnitude of non-audit fees explains cross-sectional variation in earnings management. Our analyses are based on audit and non-audit fee data collected from over 4,000 proxies filed with the SEC between February 5, 2001, and June 15, We predict that the market will react negatively to proxy statements disclosing higher than expected non-audit fees. We measure unexpected non-audit fees using two proxies: (i) the ratio of non-audit to total fees, which assumes uniform prior expectations, and (ii) the residual from a model explaining cross-sectional variation in the non-audit fees ratio, which assumes that the publicly available data items in our model are those used by investors to form expectations. Our results indicate that market-adjusted announcement day returns are significantly negative for 1 The audit industry s views are shared by incoming SEC Chairman Harvey L. Pitt. In a 1997 White Paper submitted to the Independence Standards Board, Pitt writes that the overwhelming economic interest of accounting firms in their reputational capital provides a powerful incentive to safeguard independence. Non-audit services increase the firm s investment in reputational capital, contribute importantly to the quality of audit services, and provide other benefits to clients and the public (Pitt and Birenbaum 1997, 4). 1

5 the quartile of firms with the highest unexpected non-audit fees. This finding is robust across both of our proxies for the unexpected component of non-audit fees. Our second set of tests examines whether there is a positive relation between the provision of non-audit services and earnings management. Consistent with predictions, we find that firms with relatively higher non-audit fees are more likely to just meet or beat three earnings benchmarks analysts expectations, prior year earnings, and zero earnings and to report larger income-increasing and income-decreasing discretionary accruals. These findings are robust to several specification checks. In addition, we find no difference in earnings management behavior associated with the two categories of non-audit services for which firms are required to disclose fees, i.e., financial information systems design services and all other types of non-audit services. This evidence is important given ongoing debate about the effect of non-audit services on auditor independence, and ultimately the quality of earnings. Resolution of this debate has been hindered by a lack of direct, large-sample empirical evidence. Our study provides evidence that the recently required auditor fee disclosures are useful to investors in assessing auditor independence and its financial statement effects. 2 This study also contributes to two streams of accounting research. Prior research measures audit quality using an indicator variable for Big Five versus non-big Five auditors (e.g., DeAngelo 1981; St. Pierre and Anderson 1984; DeFond and Jiambalvo 1991,1993; Teoh and Wong 1993; Becker, DeFond, Jiambalvo, and Subramanyam 1998). All audits by a given auditor are assumed to be of equal quality, and audit quality is assumed to be increasing in the size of the auditor. We extend this line of research by showing that the magnitude of non-audit fees relative to total fees is able to differentiate cross-sectional variation in audit quality. Indeed, our results 2

6 indicate that the distinction between Big Five and non-big Five auditors is not significant after controlling for the relative size of non-audit fees. Our study is also related to the earnings management literature. Although considerable prior research examines earnings management incentives, few studies investigate factors that limit earnings management (Healy and Wahlen 1999). We show that there is less earnings management when non-audit fees are relatively low. The remainder of the paper is organized as follows. Section II discusses the SEC s new rules on auditor independence and develops our hypotheses. Section III describes the sample and data. Section IV presents the empirical results. Section V summarizes and concludes. II. BACKGROUND AND HYPOTHESIS DEVELOPMENT SEC disclosure requirements The SEC has long recognized the importance of auditor independence. For example, Accounting Series Release (ASR) 296 requires that auditors not have any financial or other interest which would create the perception of impaired independence. A primary focus of the SEC s independence rules has been the provision of non-audit services. In 1978, concerns about the expansion of non-audit services provided by public accounting firms led the SEC to issue ASR 250 requiring disclosure of information about non-audit services. However, the Public Oversight Board (POB) of the American Institute of Certified Public Accountants (AICPA) subsequently concluded that available empirical evidence does not reveal any actual instances where the furnishing of MAS [management advisory services] has impaired independence (POB 1979, 35), and the rule was rescinded in Related evidence in Gore, Pope, and Singh (2001) for a sample of U. K. firms suggests that earnings management activity is generally positively associated with the provision of non-audit services for firms with non-big Five auditors, but not for firms with Big Five auditors. 3

7 In the 1990 s, a dramatic increase in the proportion of revenues derived from non-audit services renewed concerns about auditor independence. For example, former SEC Chairman Arthur Levitt (2000) states that: auditing no longer dominates the practices of the largest firms. It accounts for just 30 percent of total revenues down from 70 percent in Consulting and other management advisory services now represent over half up from 12 percent in 1977 As the firms expand their product lines, consulting and other services may shorten the distance between the auditor and management. Independence if not in fact, then certainly in appearance becomes a more elusive proposition. In response to this trend, the SEC issued revised auditor independence rules on November 15, 2000 after extensive public comment and debate. The rules require companies to disclose in proxy statements filed on or after February 5, 2001, information regarding fees billed by the auditor for the most recent year. 3 Specifically, the rules require companies to disclose: (1) under the caption Audit Fees the aggregate fees billed for professional services rendered for the audit of the annual financial statements and for the reviews of the quarterly financial statements; (2) under the caption Financial Information Systems Design and Implementation Fees the aggregate fees billed for professional services rendered for (i) directly or indirectly operating, or supervising the operation of, the company s information system or managing the company s local area network, or (ii) designing or implementing a hardware or software system that aggregates source data underlying the financial statements or generates information that is significant to the company s financial statements taken as a whole; and (3) under the caption All Other Fees the aggregate fees billed for all services rendered by the auditor other than those described in (1) and (2). Hypothesis development Audits are a device to enhance the credibility of financial statements. Independent auditors facilitate contracting between investors and management, and hence the operation of 3 In addition, the rules liberalize restrictions regarding investments by auditors and their family members in audit clients and the employment of auditors and their family members by audit clients. The rules also identify certain non-audit services that, regardless of the size of fees they generate, impair independence. 4

8 capital and labor markets. Because actual independence is not observable, shareholders rely on signals such as an auditor s reputation, the degree of oversight by professional and regulatory bodies, and the degree of economic bonding between auditor and client to assess auditor independence. 4 Both regulators and academic researchers have posited that the provision of non-audit services to an audit client can impair auditor independence by creating an economic bond between auditor and client. For example, the POB (POB 2000, 119) states: Of fundamental importance in understanding the conflict of interest that arises from the provision of non-audit services to audit clients is the fact that in so doing the audit firm is really serving two different sets of clients: management in the case of management consulting services, and the audit committee, the shareholders and all those who rely on the audited financials and the firm s opinion in deciding whether to invest, in the case of the audit It is obvious that in serving these different clients the firm is subject to conflicts of interest that tear at the fragile fabric of loyalty owed to one client or the other. And it is equally obvious that the existence of dual loyalties creates a serious appearance problem, regardless of whether, in particular cases, the fabric actually tears apart or not. Simunic (1984) models the joint demand for audit and non-audit services. He shows that when both services are provided by the same auditor and the auditor retains a portion of the cost savings that arise from knowledge spillovers, then the auditor will be economically bonded to the client. 5 The new auditor independence rules adopted by the SEC require firms to disclose audit and non-audit fees to provide investors with the information to evaluate for themselves whether the proportion of fees for audit and non-audit services causes them to question the auditor s independence (SEC 2000, III.c.5). 4 Antle (1984) notes the problems associated with defining auditor independence without reference to a formal theoretical context. In a principal-agent framework, he defines independence as the extent to which managers and auditors collude in pursuit of their own self-interests. Simunic (1984) describes independence as the likelihood that the auditor will distort his findings. Kinney (1999) discusses alternative approaches to defining auditor independence. 5 The multiperiod horizon of many audit engagements is another source of economic bonding. DeAngelo (1981) presents a model in which auditors have an incentive to lowball initial engagements with the intent of making up 5

9 Proponents of the SEC s new disclosure requirements argue that the provision of nonaudit services results in lower audit quality. DeAngelo (1981) defines audit quality as the probability that the auditor will both detect and report errors. The detection and reporting of errors requires auditor independence. The provision of non-audit services can impair independence in two ways. First, an auditor concerned about the possible loss of non-audit fee revenue may be less likely to disagree with management s accounting choices. Second, the auditor may be reluctant to criticize the work of the firm s consulting division. As a result, the audit is less effective as a monitoring mechanism, causing contracting costs to increase and firm value to decrease. Opponents of the rule changes counter that the threat of litigation provides a strong incentive to maintain independence. 6 In addition, audit partners have large financial and reputational investments in their firms that would be reduced or eliminated in the event of an audit failure. 7 Opponents also argue that economies of scope in the bundling of audit and nonaudit services would be eliminated by the SEC s rule changes. Perhaps the most compelling argument offered against the rule change is that there is no evidence to date that indicates an independence problem exists. For example, in a report commissioned by the AICPA, Antle, Griffen, Teece, and Williamson (1997, ii) conclude that: There is no evidence that the supply of non-audit services threatens auditor independence. The supply of non-audit services is not a significant factor in auditors losses in litigation or in pricing their liability insurance. There is no evidence that investors are concerned that the supply of non-audit services impairs independence. the initial losses in later periods. An independence problem arises if an auditor is unwilling to risk losing the future, higher fees by displeasing the client. 6 By one estimate, Big Five audit firms spent $2.4 billion on legal costs from , or approximately 10% of gross assurance service revenues. (Mayer, Brown & Platt, letter of June 3, 1994, to Mr. Walter Schuetze, Chief Accountant, Securities and Exchange Commission.) 7 In 1996, total partner capital invested in what was then the Big Six exceeded $3.5 billion (Antle, Griffin, Teece, and Williamson 1997). 6

10 Specifically, Palmrose (1999) finds no evidence that the provision of non-audit services is part of the basis of complaint in litigation against auditors, suggesting that investors do not share regulators concerns about the provision of non-audit services. In addition, empirical evidence that the provision of non-audit services creates an economic bond between auditor and client is weak. Although Simunic (1984) finds that audit fees are higher when an auditor provides nonaudit services to its audit client, consistent with a knowledge spillover argument, Palmrose (1986) shows that audit fees are also higher when non-audit services are provided by a party other than the client s auditor. The new fee disclosures provide an opportunity to re-examine the effect of non-audit services on auditor independence. If the provision of non-audit services impairs auditor independence, then the disclosure of these fees will inform investors about the potential for noise or bias in the reported financial statements. Thus, we predict there will be a negative stock price reaction to news that the level of non-audit services is higher than expected: H1: Ceteris paribus, stock prices will respond negatively to news that non-audit services are higher than expected. Failure to observe a significant negative stock price reaction is consistent with the argument that the provision of non-audit services does not impair auditor independence. However, it may also indicate that our tests are not sufficiently powerful to detect a stock price effect. For example, we are unlikely to detect a significant market reaction if our proxy for the news contained in the disclosures regarding the level of non-audit services is a poor measure of investors expectations. The latest revision of the auditor independence rules is part of the SEC s effort to curb a perceived increase in earnings management activity (Levitt 1998). Consolidation among the largest accounting firms and the rapid growth of non-audit services have magnified the SEC s 7

11 concerns regarding the effect of auditor independence on earnings quality. 8 Several highly publicized accounting irregularities have focused additional attention on this issue. 9 Underlying the SEC s position is the belief that the provision of non-audit services impairs independence and thus compromises audit quality. A higher quality audit increases the likelihood that questionable accounting practices will be detected, that the auditor will object to their use, and that the auditor will qualify the audit report if the practices are not modified by management. High-quality audits deter earnings management (or, in the extreme, accounting fraud). Therefore, if the provision of non-audit services impairs independence resulting in a lower quality audit, then we expect earnings management activity to increase with the level of non-audit services provided by the auditor: H2: Ceteris paribus, there is a positive relation between the provision of non-audit services and earnings management. III. DATA Sample selection Our initial sample consists of all definitive proxy statements on the SEC s EDGAR database that were filed between February 5, 2001 and June 15, We require that the proxy statement disclose audit fees, financial information systems fees, other fees, and the firm s auditor. We exclude firms that changed auditors during the year. We also exclude financial institutions (SIC codes ) because estimation of discretionary accruals is problematic for these firms. These procedures yield an initial sample of 4,225 observations. Because of the 8 These concerns echo a long history of SEC statements linking accounting quality and auditor independence. For example, ASR 296 states that the quality of information disseminated in the securities markets and the continuing conviction of individual investors that such information is reliable are key to the formation and effective allocation of capital. Accordingly, the audit function must be meaningfully performed and the accountants independence not compromised. 9 See, e.g., SEC List of Accounting-Fraud Probes Grows, Wall Street Journal, July 6, 2001, C1. 8

12 interest in the new auditor fee disclosures, we provide some preliminary descriptive statistics for this broad sample before imposing any additional data requirements. Table 1 reveals that audit fees comprise, on average, slightly more than half of all fees billed by auditors to their audit clients. Less than 25 percent of firms engaged their auditor to provide financial information systems design and implementation services. However, fees for these services are relatively large. The smallest financial information systems fee reported by our sample, $116,000, is not far below the median fee for either audit or other non-audit services. In contrast to the infrequent occurrence of financial information systems fees, over 95 percent of firms report fees for other non-audit services. The average fee for other services is nearly $1 million. 10 Finally, mean (median) non-audit fees as a percent of total fees is 47 (48) percent. Table 2 provides summary fee data for each of the Big Five audit firms, and for all other audit firms combined. Non-audit fees comprise the majority of fees billed by Big Five auditors, while audit fees comprise the majority of fees billed by non-big Five auditors. Audit fees as a percent of total fees are nearly twice as large for the non-big Five auditors as for any of the Big Five auditors. In addition, the composition of fees varies among the Big Five auditors. PricewaterhouseCoopers, the largest auditor in terms of total fees billed, derives approximately three-fourths of its total fee revenue from non-audit services, compared to approximately twothirds for the other Big Five auditors. Ernst & Young earns only six percent of its revenue from financial information services, but bills more fees for other non-audit services (64 percent) than any other Big Five auditor. 10 Approximately one-fourth of the firms in the initial sample provide information on the composition of other nonaudit fees. Untabulated descriptive statistics indicate that firms with large fees for other services, as a percent of total fees, are more likely to disclose additional information on the composition of other fees. Controlling for the ratio of other fees to total fees, firms audited by Ernst & Young are more likely to disaggregate other fees, while firms audited by non-big Five auditors are less likely to disaggregate other fees. 9

13 To determine the primary sample for our empirical tests, we merge the fee data with Compustat. There are 2,506 firms that also have total assets for fiscal year 2000 on Compustat. Panel A of Table 3 details the frequency of proxy filings by month for this sample. Firms must file their proxy statement prior to the annual shareholders meeting which typically occurs three to four months after the fiscal year end. The majority of the firms in our sample filed their proxy statement in March and April, corresponding to a December or January fiscal year end. Despite the time-period clustering of our sample, Panel B of Table 3 reveals that the industry composition of our sample is similar to that of the 2000 Compustat database. Variable measurement and descriptive statistics Our hypotheses predict that the provision of non-audit services is negatively correlated with firm value and earnings quality. We measure the provision of non-audit services using the ratio of non-audit fees to total fees (FEERATIO). This variable is consistent with definitions used by prior researchers (Scheiner and Kiger 1982; Glezen and Millar 1985; Parkash and Venable 1993; Firth 1997), and reflects the SEC s stated purpose in requiring these disclosures, i.e., that investors will consider the proportion of fees for audit and non-audit services in evaluating whether the auditor is independent (SEC 2000). The descriptive statistics reported in Table 3, Panel C indicate that the mean (median) FEERATIO is 50 (53) percent, slightly higher than the percentages reported in Table 1. Untabulated statistics reveal that median non-audit fees as a percent of net sales (average total assets) is 0.11 (0.10) percent, indicating that non-audit fees do not have a material direct effect on firms operations. Our first hypothesis predicts that, certeris paribus, the market reaction to the public disclosure of auditor fees is inversely related to the unexpected portion of non-audit services. To identify the earliest date at which the fee information became public, we searched the EDGAR 10

14 database to determine if the firms in our sample filed a preliminary proxy statement disclosing auditor fees. 11 If we find a preliminary proxy containing the necessary fee data then we use the release date of the preliminary proxy as our event date; if not, we use the release date of the definitive proxy. We obtain returns data from Datastream, and calculate market-adjusted returns (ARET) by subtracting either the S&P 500 Index return (for firms with a market capitalization greater than $10 billion) or the S&P Midcap 400 Index (for firms with a market capitalization less than $10 billion) from the firm-specific raw return on the date that the proxy statement was filed. 12 Table 3, Panel C reveals that mean (median) ARET is 0.09 ( 0.17) percent. Because stock prices will react to the new information contained in the auditor fee disclosures, tests of our first hypothesis require a proxy for the unexpected component of the non-audit fee ratio. Our first proxy, FEERATIO, assumes uniform prior expectations. This proxy is appropriate if the non-audit fee ratio contains a shock that is large relative to its expected firm-specific component. Anecdotal evidence suggests that investors were surprised by the magnitude of non-audit fees (e.g., Weil and Tannenbaum 2001), consistent with the notion of a large, positive unexpected shock. Our second proxy is the residual (FEERESIDUAL) from a model that assumes that non-audit fees vary inversely with agency costs (Parkash and Venable 1993; Firth 1997). The Appendix provides details on the empirical estimation of the model. This alternative proxy assumes that the publicly available data items in our model are those used by investors to form expectations of non-audit fees Approximately 9 percent (143 firms) of the 1,538 definitive proxies in our returns sample were preceded by the release of a preliminary proxy. 12 The proper cut-off is unclear given that the indices have some market capitalization overlap. Genzyme is the largest company in the S&P Midcap 400 Index with a market capitalization of $10.1 billion. American Greetings is the smallest firms in the S&P 500 Index with a market capitalization of $700 million. 13 Although firm-level data on audit and non-audit fees have not been publicly disclosed in the United States since ASR 250 was rescinded in 1982, it is likely that investors formed expectations on the magnitude of non-audit fees based on similar disclosures in the U.K. and Australia. Gore, Pope and Singh (2001) indicate that non-audit fees as a percent of total fees are not nearly as material in the U.K. as our evidence indicates they are in the U.S. 11

15 Our second hypothesis predicts that, certeris paribus, there is a positive relation between the provision of non-audit services and earnings management. We use two complementary approaches to identify firms managing earnings. Prior research (e.g., Burgstahler and Dichev 1997; Burgstahler and Eames 1998; Degeorge, Patel, and Zeckhauser 1999) documents a significantly higher than expected frequency of firms with slightly positive earnings surprises, earnings changes, and earnings levels, consistent with firms managing earnings to meet simple benchmarks. Both the SEC and the POB expressed particular concern that the pressure for public companies to meet analysts expectations and project a smooth earnings path creates pressure on auditors to enable their clients to meet those objectives (SEC 2000; POB 2000). Thus, our first test of the earnings management hypothesis examines whether firms that just meet or beat three earnings benchmarks, i.e., (i) analysts expectations, (ii) prior period earnings, and (iii) zero earnings, purchase more non-audit services than other firms. Using data obtained from First Call, we calculate quarterly earnings surprises as the difference between actual earnings per share and the last available consensus (median) forecast calculated prior to the announcement of earnings. We identify firms that just meet or beat analysts expectations with an indicator variable (SURPRISE) equal to one if the firm reports a earnings surprise of 0.00 or 0.01 in at least three of the four quarters in fiscal 2000, and zero otherwise. Similarly, following Burgstahler and Dichev (1997), we identify firms reporting small increases in earnings (INCREASE) and small profits (POSITIVE) by scaling net income and the annual change in net income by beginning of the year market value of common equity. We consider firms with scaled changes in earnings in the interval [0.00, 0.02) as reporting a small increase in earnings, and firms with scaled earnings in the interval [0.00, 0.04) as reporting a 12

16 small profit. 14 The statistics in Table 3, Panel C indicate that 18 percent of the sample firms report a small positive forecast error, 18 percent report a small increase in earnings over the prior year, and 15 percent report a small profit. We estimate the following logit model, where BENCHMARK indicates the three earnings benchmarks, SURPRISE, INCREASE, or POSITIVE: Prob β 0 + β FEERATIO + β BIGFIVE + β LITRISK ( BENCHMARK ) = F + β4m/b + β5%inst + β6 LOGMVE + u (1) Our tests control for other factors that are expected to influence whether a firm just meets or beats earnings benchmarks. Prior research finds that Big Five auditors are less likely to allow earnings management than non-big Five auditors (e.g., DeFond and Jiambalvo 1991,1993; Becker, DeFond, Jiambalvo, and Subramanyam 1998). Thus, we include a Big Five indicator variable (BIGFIVE) to control for a relation between earnings management and auditor size. Table 3, Panel C shows that Big Five firms audit 92 percent of our sample. 15 Matsumoto (1999) reports that firms with high litigation risk, growth prospects, and institutional ownership are more likely to be concerned about missing earnings benchmarks. Poorly performing firms and firms with a sudden stock price drop due to a negative earnings surprise are at higher risk of a shareholder lawsuit, and thus have greater incentives to meet earnings benchmarks. We measure litigation risk (LITRISK) as an indicator variable equal to one if the company operates in an industry at high risk of securities litigation (SIC codes , , , , and ). In our sample, 39 percent of the firms operate in high litigation risk industries. Additionally, high growth firms are more likely to 14 We base the width of the intervals on the recommendation in Degeorge, Patel, and Zeckhauser (1999) that interval width be positively related to the variability of the data and negatively related to the number of observations. Specifically, they suggest an interval width of 2(IQR)n -1/3, where IQR is the sample interquartile range of the variable and n is the number of available observations. 13

17 be punished by the market for missing earnings benchmarks (Skinner and Sloan 1999). We use the market-to-book ratio at the beginning of the year (M/B) to proxy for growth. Mean (median) M/B is 6.78 (2.34). Because institutional investors are believed to focus on near-term earnings performance, firms with high institutional ownership are expected to have greater incentives to meet earnings benchmarks. We measure institutional ownership (%INST) as the percentage of shares held by institutions at the beginning of the year (as reported by Spectrum), and find that mean (median) %INST for our sample is 36 (31) percent. Finally, following Matsumoto (1999), we control for firm size with the log of the market value of equity (LOGMVE). Table 3, Panel C reveals that the distribution of market value is highly skewed, with the mean ($3,204 million) nearly ten times larger than the median ($304 million). Our second approach to identifying firms managing earnings uses discretionary accruals estimated with a cross-sectional modified Jones model (Dechow, Sloan, and Sweeney 1995). Specifically, discretionary accruals (DACC) are calculated as follows: ( ˆ α ˆ β1[ ] ˆ β2 ) DACC = TA + REV REC + PPE (2) where all variables are scaled by total assets at the beginning of the year, and TA is total accruals, defined as net income less cash from operations, REV is the change in net revenues, REC is the change in net receivables, PPE is gross property, plant, and equipment, and the estimates of α, β 1, and β 2 are obtained from the following model: TA = α + β REV + β PPE + e (3) 1 2 In his discussion of earnings management, Levitt (1998) identifies several techniques that allow firms to either overstate or understate current earnings, including big bath charges, 15 The data in Table 2 indicate that 85 percent of firms are audited by a Big Five auditor. Thus, the additional data requirements for our primary sample exclude a disproportionate number of firms audited by non-big Five auditors. 14

18 cookie jar reserves, and premature recognition of revenue. Ceteris paribus, independent auditors should be a deterrent to both earnings overstatements and understatements. We use the absolute value of discretionary accruals (ASDACC) to measure the combined effect of incomeincreasing and income-decreasing earnings management decisions. Table 3, Panel C indicates that mean (median) ABSDACC is 13 (7) percent of total assets. We estimate the following model to examine the relation between discretionary accruals and auditor independence: ABSDACC = α + β FEERATIO + β BIGFIVE + β CFO + β ABSCFO + β ACC + β ABSACC + β LOGASSET + β ACQUISTION 5 + β FINANCING + β LEVERAGE + ε (4) Following Becker, DeFond, Jiambalvo, and Subramanyam (1998), we control for several other potential determinants of discretionary accruals. Similar to the earnings benchmark model, we include a Big Five indicator variable (BIGFIVE) to control for a relation between earnings management and auditor size. In addition, prior research shows that discretionary accruals models fail to completely extract non-discretionary accruals that are correlated with firm performance. 16 Thus, we include four measures of firm performance as control variables, all of which are deflated by average total assets: cash from operations (CFO), the absolute value of cash from operations (ABSCFO), total accruals (ACC) and the absolute value of total accruals (ABSACC). The log of total assets (LOGASSET) is used to control for firm size. We include indicator variables for whether the firm acquired another company during the year (ACQUISTION) or issued debt or equity (FINANCING) because managers are expected to manage earnings to influence the terms of these transactions (e.g., Erickson and Wang 1999; Teoh, Welch, and Wong 1998a,b). The data for these variables were obtained from Securities 15

19 Data Corporation. In our sample, 45 percent of the firms acquired another company during 2000, and 26 percent issued debt or equity. Finally, we control for leverage, measured as the ratio of total liabilities to total assets (LEVERAGE), because managers of highly leveraged firms may have incentives to overstate earnings to avoid violating debt covenants (DeFond and Jiambalvo 1994) or, alternatively, to understate earnings due to contractual negotiations by distressed firms (DeAngelo, DeAngelo, and Skinner 1994). Mean (median) LEVERAGE is 0.52 (0.50). IV. EMPIRICAL RESULTS Market reaction to the disclosure of auditor fees Table 4 presents summary statistics from an event study that examines the mean market reaction to the release of proxy statements containing auditor fee disclosures. We assess the statistical significance of the results using a t-test to determine whether the mean market reaction in each portfolio is significantly negative (t µ<0 ), and a sign test to determine whether the proportion of firms with positive abnormal returns in each portfolio is significantly less than 50 percent (sign test %<50% ). Because the observations used in the tests are drawn from a four and one-half month period in the beginning of 2001, our results may be biased due to cross-sectional dependence in the returns data. However, Bernard (1987) finds that cross-sectional dependence problems are less severe for estimations using daily returns data. Moreover, we attempt to control for common factors that predict returns and are related to FEERATIO by including these variables (size, market-to-book, and past returns) in the estimation of the unexpected non-audit fee ratio discussed in the Appendix For example, Dechow, Sloan, and Sweeney (1995) show that discretionary accruals are biased for extreme values of earnings and cash flows. For further discussion of related research design issues, see McNichols (2000). 17 The release of CRSP data for our sample period will allow us to adopt more sophisticated methodologies for analyzing stock price reactions. 16

20 We partition the sample into quartiles based on our proxies for unexpected non-audit fees. The results for our first proxy, FEERATIO, are reported in Panel A of Table 4. The median FEERATIO in the quartile containing firms with the highest unexpected non-audit fees is 0.77, compared to 0.20 in the quartile with the lowest. If the market is surprised by the magnitude of non-audit fees revealed in the proxy statement and believes the provision of nonaudit services impairs auditor independence, then we expect abnormal returns to be most negative in the quartile containing the firms with the highest unexpected non-audit fees. As predicted, the results indicate that the market reaction is negative and significant in the top quartile. The mean return on the date these firms filed proxies was 0.44 percent, significant at the 0.05 level. In contrast, the mean return is positive, but statistically insignificant, in the other three quartiles. Similarly, 43 percent of the returns in the top quartile are positive, significant at the 0.01 level, compared to 47 to 51 percent of the returns in the other three quartiles. To examine the robustness of our results, we partition the sample into quartiles based on our second estimate of unexpected non-audit fees (FEERESIDUAL) in Panel B of Table 4. The median FEERESIDUAL in the top quartile is 0.21, compared to 0.24 in bottom quartile. The results of the market reaction tests are similar to those reported in Panel A. Specifically, in the top quartile, the mean market return is 0.40 percent, significant at the 0.05 level, and the proportion of firms with positive returns is 45 percent, significant at the 0.01 level. None of the results in the other three quartiles is significant. Overall, the results reported in Table 4 support the hypothesis that investors responded negatively to revelations that fees for non-audit services were higher than expected. Because financial statements may be less reliable or unbiased for firms purchasing more non-audit services, investors may reduce their expectations of discounted future cash flows. However, the 17

21 disclosed fee information also may have implications for the market s expectations of future cash flows unrelated to the quality of the audited financial reports. For example, unexpectedly high fees for non-audit services may signal poor management quality or that the firm faces difficulties previously unknown by investors. Association between non-audit services and earnings management Our first test of the earnings management hypothesis examines whether firms that just meet or beat earnings benchmarks purchase more non-audit services. We predict a positive relation between FEERATIO and firms that just meet or beat each of the three earnings benchmarks, SURPRISE, INCREASE, and POSITIVE. Table 5 reports summary statistics from the empirical estimation of equation (1). Consistent with predictions, the coefficient estimate on FEERATIO is positive and significant in all three estimations. However, the result is relatively less significant (p-value of 0.07) in the POSITIVE estimation compared to the SURPRISE and INCREASE estimations (p-values < 0.01). These results suggest that firms paying relatively more non-audit fees are more likely to meet earnings benchmarks, particularly analysts expectations and prior period earnings, consistent with the concerns expressed by the SEC (2000) and POB (2000). Regarding the control variables, BIGFIVE is insignificant in all estimations in Table 5. The results for the remaining control variables are generally consistent with Matsumoto (1999). Litigation risk, LITRISK, is positive and significant in the SURPRISE and POSITIVE models, but not in the INCREASE model. These findings suggest that firms at high risk of securities litigation have greater incentives to meet earnings benchmarks. In addition, LOGMVE and %INST are positive and significant in all estimations, suggesting that larger firms and those with greater institutional ownership are more likely to just meet or beat earnings benchmarks. 18

22 Our second test of the earnings management hypothesis examines the relation between discretionary accruals and the purchase of non-audit services. Table 6 reports summary statistics from the empirical estimation of equation (4). Consistent with predictions, the coefficient estimate on FEERATIO is positive and significant. Firms purchasing more non-audit services exhibit greater earnings management. Controlling for the level of non-audit fees, there is no evidence that Big Five auditors limit discretionary accruals more than non-big Five auditors. Our results also indicate that discretionary accruals are correlated with firm performance, firm size, and acquisition and financing activity. 18 Sensitivity tests Estimations using alternative specifications indicate that our findings are robust. First, to further ensure that our results are not driven by cross-sectional variation between firms with a Big Five and non-big Five auditor, we repeat the analyses in Tables 5 and 6 using only the sample of firms with a Big Five auditor. The untabulated results are similar to those reported above. Second, we examine the association between the provision of non-audit services and the direction of earnings management for this sample of firms. Prior research suggests that earnings overstatements are of greater concern to auditors because investors frequently sue auditors when a client overstates income, but rarely, if ever, sue when a client understates income (e.g., St. Pierre and Anderson 1984). Because it is inappropriate to partition the sample on the dependent variable, we use a reverse regression procedure to examine the association between the purchase of non-audit services and the direction of earnings management. We obtain a measure of discretionary 18 We also estimate equation (4) using the absolute value of discretionary working capital accruals, defined as total accruals less depreciation expense, and the absolute value of discretionary accruals obtained from a cross-sectional 19

23 accruals that is orthogonal to the control variables in equation (4), and then partition the observations into firms with income-increasing discretionary accruals (DACC + ) and firms with income-decreasing discretionary accruals (DACC ). We regress FEERATIO on these two variables and find a positive and significant association between FEERATIO and DACC + and a negative and significant association between FEERATIO and DACC. These results confirm the finding in Table 6 that firms purchasing more non-audit services report relatively higher incomeincreasing and income-decreasing discretionary accruals. In addition, the magnitude of the coefficient estimates on DACC + and DACC is not statistically different, suggesting that for a given level of non-audit services, auditors are equally likely to prevent earnings overstatements and understatements. Third, we disaggregate FEERATIO into two components the ratio of financial information systems design fees to total fees (ISRATIO) and the ratio of all other non-audit fees to total fees (OTHRATIO). Because of concerns that the provision of information technology services poses a particularly strong threat to independence, the preliminary independence rules issued by the SEC on June 27, 2000 prohibited auditors from providing these services to their audit clients. The final rules relaxed the ban, but require separate disclosure of the information technology services component of non-audit fees. To examine whether information technology services pose a stronger threat to auditor independence than do other types of non-audit services, we repeat the analyses in Tables 5 and 6 substituting ISRATIO and OTHRATIO for FEERATIO. The untabulated results indicate that both fee variables are positive and statistically significant in all estimations except for OTHRATIO in the POSITIVE earnings benchmark test. In addition, the coefficient estimates on the two variables are statistically indistinguishable in all estimations. Jones model (Jones 1991). Our untabulated findings are similar to those reported in Table 6. 20

24 These results indicate that the two categories of non-audit fees required to be disclosed in firms annual proxy statements are equally important to an assessment of earnings quality. Fourth, the results in Table 4 indicate that the market reaction to the release of the auditor fee information is only significant in the quartile with the highest unexpected non-audit fees. To control for similar non-linearities in the relation between non-audit fees and earnings management, we repeat the analyses in Tables 5 and 6 using an indicator variable equal to one if the firm is in the highest quartile of FEERATIO, and zero otherwise. All results are robust to this alternative specificiation. Finally, we repeat the analyses in Tables 5 and 6 using a common sample of 1,209 firms with no change in inferences. V. SUMMARY AND CONCLUSION This study examines the implications of non-audit fees for shareholder wealth and the quality of earnings. The SEC recently revised its auditor independence rules to protect investors who rely on the integrity of financial statements audited by public accountants. Of particular concern to the SEC is the effect of non-audit services on auditor independence. We examine the market reaction to the disclosure of auditor fee data required by the new independence rules. We find a significant negative stock price response for the quartile of firms with the highest unexpected non-audit fees. We also study the relation between the provision of non-audit services and two measures of earnings management the likelihood of meeting earnings benchmarks and the magnitude of discretionary accruals. We find that firms with a high ratio of non-audit fees to total fees are more likely to report small positive earnings surprises, small increases in earnings, and small profits. In addition, firms with a high ratio of non-audit fees to total fees report larger income-increasing and income-decreasing discretionary accruals. Disaggregating total non-audit fees into fees from information technology services and fees for 21

25 all other types of non-audit services reveals that the provision of both types of non-audit services increases earnings management activity. This study is a first step in examining the implications of non-audit services for auditor independence and the quality of financial reporting. Although our evidence indicates that firms purchasing non-audit services are more likely to engage in earnings management, we do not address the related issue of whether the provision of non-audit services influences an auditor s incentives to detect and report financial statement fraud. Auditors legal liability as well as their financial and reputational investments may serve as effective constraints in this setting. Nelson, Elliot, and Tarpley (2000) examine several factors that affect auditors decisions to waive firms earnings management attempts, but do not consider the effect of the provision of non-audit services on this decision. The SEC disclosure requirements may also alter the existing equilibrium in the audit services market. For example, the disclosure of fee data could increase the competitiveness of the audit market by reducing the cost to firms of making price comparisons and negotiating fees. In addition, firms may reduce the purchase of non-audit services from their auditor to avoid the appearance of independence problems. 22

26 APPENDIX Estimation of Unexpected Non-Audit Fees In this appendix, we develop a model to explain cross-sectional variation in FEERATIO. The purpose of this model is to estimate the unexpected portion of non-audit fees for use in our market reaction tests. We base our model on Parkash and Venable (1993) and Firth (1997) who examine a company s decision to purchase non-audit services from their auditor, conditional on the level of auditing services purchased. They argue that companies facing high agency costs will demand high quality audits and restrict the purchase of other services from their auditor. We estimate the following model using all observations in our primary sample with the necessary data: 13 FEERATIO = α i INDi i= 1 + β BIGFIVE + β ROA + β LOSS + β ANNRET + β VARRET + β %INST_END + β LEV_END + β INVREC + β ACQUISITON 6 + β FINANCING + β MVE_END + β M/B_END + υ (A1) We include separate industry intercepts, IND i, for each of the i = 1 to 13 industries identified in Panel B of Table 3. An indicator variable (BIGFIVE) controls for the possibility that Big Five auditors are able to charge a premium for their services, whether audit or non-audit. Consistent with agency incentives, Parkash and Venable (1993) and Firth (1997) find that nonaudit fees (relative to audit fees) are increasing in firm profitability and institutional ownership, and decreasing in leverage. We include two measures of accounting performance, ROA, defined as net income divided by average total assets, and LOSS, an indicator variable equal to one if the firm reports a net loss. We also include two measures of stock return performance, ANNRET, the compounded CRSP monthly return in calendar 2000, and VARRET, the variance of CRSP monthly returns for calendar Our control for institutional ownership is %INST_END, the 23