Do Auditor Fees Affect Accruals Quality? Further Evidence

Transcription

1 Center for Corporate Reporting and Governance Working Paper Series Do Auditor Fees Affect Accruals Quality? Further Evidence Myungsoo Son Associate Professor California State University, Fullerton Working Paper CCRG Center for Corporate Reporting and Governance California State University, Fullerton Department of Accounting Fullerton, CA Tel: (657) CCRG Working Papers are circulated to stimulate discussion and comments. They have not been peerreviewed. References in publication to CCRG Working Papers should be cleared with the author or authors.

2 1 How Do Auditor Fees Affect Accruals Quality?: Additional Evidence ABSTRACT This study examines whether the Sarbanes-Oxley Act (SOX) and auditor tenure moderate the association between auditor fees and accruals quality. Using a large sample of firms, we first confirm results in Srinidhi and Gul (2007) showing that in the pre-sox years, higher audit fees (non-audit fees) imply higher (lower) accruals quality. However, we find that, post-sox, these associations disappear. We attribute this to changes brought about by SOX that affected the relative mix of audit effort and rents reflected in auditor fees. These results have important implications for academic studies that use audit fees as a proxy for audit quality. This study also presents the first evidence showing that lengthy auditor tenure attenuates the relationship between audit fees and accruals quality. Keywords: Audit Fees; Nonaudit Fees; Accruals Quality; Sarbanes-Oxley Act; Auditor Tenure Data availability: All data used in this study are publicly available from sources described in the text. 1

3 2 How Do Auditor Fees Affect Accruals Quality?: Additional Evidence INTRODUCTION This study examines the association between accruals quality (and thus audit quality) 1 and non-audit and audit fees. Using pre-sarbanes-oxley (SOX) data, Srinidhi and Gul (2007) find that audit fees (non-audit fees) have a positive (negative) effect on accruals quality. We examine whether the association of the fees with accruals quality continues to hold in the post-sox years. Srinidi and Gul (2007) argue that audit and non-audit fees, both, have components that can increase or decrease accruals quality. On the one hand, high audit fees reflect high audit effort that can increase accruals quality; on the other hand, the fees include client-specific rents that can impair auditor independence and reduce accruals quality. 2 Similarly, non-audit fees can increase accruals quality through knowledge spillovers; however, rents associated with these fees can decrease accruals quality. Srinidhi and Gul s results suggest that audit fees predominantly proxy for audit effort, while non-audit fees predominantly contain rents that impair auditor independence. We posit that after SOX, the relative mix of components reflected in audit and non-audit fees has changed. Audit fees, for example, have increased significantly. Over the period 2000 to 2005, audit fees increased, on average, from $533,360 to $1,185,322, an increase of 122 % (Ghosh and Pawlewicz 2009). In 2004 alone, Section 404 of SOX resulted in fee increases of 86% over the previous year (Raghunandan and Rama 2006). Critics argue that the fee increases represent onerous compliance costs for companies that 1 We use accruals quality and audit quality interchangeably throughout the paper (Srinidhi and Gul 2007). A direct link between high quality audits and high quality accruals has been found, among others, by Caramanis and Lennox (2008), Dechow et al. (2011) and Gunny and Zhang (2009). 2 Economic bonding can result not just from non-audit services but also from audit services (SEC 2001; Kinney and Libby 2002). 2

4 3 do not significantly enhance accruals quality. On the other hand, non-audit fees decreased significantly post-sox due to the prohibition of many non-audit services. As a result, the economic bond from providing these services may have weakened. An interesting empirical question, therefore, is whether post-sox audit fees continue to reflect the same relative mix of effort versus rents, and, non-audit fees rents versus knowledge spillovers. We also examine whether the fees-accruals quality relation varies according to the duration of the auditor-client relationship. As tenure increases, auditors gain more knowledge about their clients business, risks and accounting systems. Therefore, all else constant, as tenure lengthens, we could expect that audits would involve less auditor effort. Interestingly, however, audit fees generally increase with auditor tenure which suggests possibly, that as auditor tenure increases, audit fees contain a relatively higher proportion of rents. This study first confirms the findings of Srinidhi and Gul (2007) that in the pre- SOX years, audit fees are positively associated with accruals quality, while non-audit fees are negatively associated with accruals quality. Interestingly, however, the relation between accruals quality and the fees significantly weakens post-sox, which suggests that audit (non-audit) fees reflect a relatively lower proportion of effort (rents) compared to the pre-sox years. We also document a new result that as auditor tenure increases, the relationship between audit fees and accruals quality weakens. This is consistent with audit fees reflecting proportionately higher levels of rents as tenure increases. We do not find, however, that tenure moderates the association between non-audit fees and accruals quality. 3

5 4 This study contributes to the literature in several ways. Importantly, our findings suggest that high audit fees do not always correlate with high audit quality. This has implications for academic research that uses audit fees as a proxy for audit quality or earnings quality (e.g., Rice and Weber 2012; Higgs and Skantz 2006; Ashbaugh et al. 2003). Furthermore, although audit fees in the post-sox years increased substantially, the fees appear to reflect proportionately lower amounts of audit effort that enhance accruals quality. The results suggest that post-sox, factors unrelated to accruals quality resulted in being priced in audit fees. The remainder of this study comprises four sections. Section 2 reviews the relevant literature and develops testable hypotheses. Section 3 discusses the research design, and Section 4 discusses the empirical results. The last section concludes the study. LITERATURE AND DEVELOPMENT OF HYPOTHESES Auditor Fees and Audit Quality: Pre-SOX Srinidhi and Gul (2007) regard audit fees as a function of two opposing factors audit effort and rents. Higher is the audit effort as reflected in higher audit fees, higher is the accruals quality. On the other hand, higher are the rents reflected in audit fees, higher is the likelihood that auditor independence will be impaired, lowering accruals quality. Prior studies find that audit fees are positively associated with audit quality, which suggests that of the two, audit effort is the dominant factor priced in audit fees. In support, Higgs and Skantz (2006) find higher earnings response coefficients (ERCs) for firms paying high audit fees. Ashbaugh et al. (2003) document that companies compensating their auditors with high audit fees report lower return-adjusted discretionary accruals. Srinidhi and Gul (2007) show that audit fees have a significant and positive association 4

6 5 with accruals quality. It is important to note, however, that the above studies draw their conclusions using pre-sox data. Non-audit fees are also modeled as a function of two opposing forces: one increases accruals quality through knowledge spillovers and the other, decreases accruals quality due to economic bonding. On the one hand, the provision of non-audit services helps an audit firm acquire deeper insights into clients operations increasing the quality of an audit (Gifford and Howe 2004; Simunic 1984). 3 On the other hand, non-audit fees primarily have been blamed for impaired auditor independence. It has been suggested, for example, that auditors often lowball audit fees to obtain more lucrative non-audit services (Levitt 2000; DeAngelo 1981). Companies purchasing large amounts of nonaudit services from their auditors are negatively perceived by capital market participants as reflected in lower earnings response coefficients (Krishnan et al. 2005). Srinidhi and Gul (2007) argue that because non-audit services are less regulated and more profitable than audit services, they are likely to contain significant rents. 4 Their results suggest that the economic dependence resulting from non-audit services dominates the positive effect of knowledge spillovers on accruals quality. Auditor Fees and Audit Quality: Post-SOX 3 For example, audit firms that provide tax related services are able to use this knowledge to better estimate tax accruals reported on the financial statements (Higgs and Skantz 2006). Similarly, Kinney et al. (2004) find fewer financial restatements in firms whose auditors also provide tax services evidence that there are knowledge spillovers on accounts of at least certain non-audit services. In contrast, O Keefe et al. (1994) and Davis et al. (1993) do not find that there is a reduction in audit hours as a result of the same auditor providing non-audit services. 4 However, there are also studies that fail to find a positive association between discretionary accruals and non-audit services (Ashbaugh et al. 2003; Chung and Kallapur 2003). Larcker and Richardson (2004) find a negative effect of non-audit services on discretionary accruals only for a small subsample of firms with poor corporate governance. Note, however, that these studies use a different measure of audit quality from that used by Srinidhi and Gul (2007). 5

7 6 Changes in the audit market place post-sox have affected the nature and extent of work performed by auditors (Fung et al. 2012). Section 404 of SOX significantly increased auditor duties, requiring extensive testing and documentation of a company s internal controls. Ghosh and Pawlewicz (2009) note that increased auditor workload also resulted from: Section 103 requiring the retention of audit working papers for a minimum of seven years; Section 204 requiring expanded communications between auditors and audit committees; and Section 401 concerning the auditing of additional disclosures on off-balance-sheet transactions, pro-forma financial reporting and special-purpose entities. Auditor duties also increased as a result of increased oversight by the Public Company Accounting Oversight Board (PCAOB).While higher levels of auditor effort should result in fewer audit failures, critics contend that a significant portion of the additional audit work in the post-sox period relates to increased documentation and compliance that is unrelated to increasing accruals quality (e.g., Krishnan et al. 2008; Montana 2007). Additionally, higher audit fees were reported by companies following SEC (2003) rules that required audit and non-audit fee disclosures in proxy statements. 5 Due to the intense lobbying efforts by audit firms, the definitions of audit (non-audit) services were broadened (narrowed), 6 enabling audit firms to use the revised definitions to their advantage (Abbott et al. 2011; Weil and Rapoport 2003). Ambiguous terms such as normally provided and reasonably related in the rules resulted in inconsistencies in fee classifications and provided opportunities for classifying non-audit fees as audit fees 5 While the new rules were effective for fiscal years ending after December 15, 2003, the SEC encouraged early adoption. Large firms and those having more institutional owner were more likely to adopt early (Ashtana and Krishnan 2006). 6 The revised rule permitted companies to classify any fees for services performed for fulfilling auditor duties under Generally Accepted Auditing Standards (GAAS) as audit fees. Two new disclosure categories of non-audit fees were introduced: audit-related fees and tax fees and one category was prohibited: information systems fees. See the Appendix for a comparison of disclosure rules (SEC 2000 versus SEC 2003). 6

8 7 to avoid adverse market perceptions of non-audit services (Dickins and Higgs 2005; Ashtana and Krishnan 2006). 7 The reclassifications from non-audit fees to audit fees post-sox can potentially weaken the association of audit fees with accruals quality. Finally, Ghosh and Pawlewicz (2009) argue that following the demise of Arthur Andersen, the audit market for Big N auditors was characterized by a reduced level of competition. Big N auditors, as a result, may have been able to charge more rents in the post-sox period, making audit fees a noisier proxy for audit effort. With regard to non-audit fees, SOX prohibits auditors from providing certain types of non-audit services to their audit clients. 8 The prohibition is based on the argument that the provision of both non-audit and audit services impairs auditor independence, lowering accruals quality. Several companies also voluntarily reduced non-audit services to avoid investors negative perceptions about these services (Krishnan et al. 2011). The reduction in non-audit services by companies suggests that there is a decreased opportunity to include rents in post-sox non-audit fees. Based on the discussion above, we propose two hypotheses: H1a: The association between audit (accruals) quality and audit fees weakens in the post-sox period. H1b: The association between audit (accruals) quality and non-audit fees weakens in the post-sox period. Auditor Tenure as a Moderating Variable Research indicates that in addition to the high set up costs, auditors face a steep learning curve in the initial years of an engagement. Caramanis and Lennox (2008) 7 Similarly, Hoitash et al. (2007) argue that, post-sox, some services that were previously included in the non-audit service category were reclassified as audit services. 8 The SEC (2003) banned nine types of non-audit services: bookkeeping, financial information systems design and implementation, appraisal services, actuarial services, internal audit outsourcing services, management functions or human resources, broker or dealer, legal, and any service that the PCAOB determines as compromising auditor independence. 7

9 8 provide direct evidence that auditors expend more effort (i.e., audit hours) in the first years of an audit. Their results are consistent with the GAO s (2003) findings that auditors require at minimum two to three years for becoming familiar with their clients operations. 9 However, as tenure lengthens, auditors acquire a more complete understanding of their client s operations. It follows that, all else constant, audit effort should decrease as tenure increases. Interestingly, however, audit fees generally increase with auditor tenure. We posit, therefore, that audit fees contain increasing amounts of rents as auditor tenure lengthens. This should lead to an attenuation in the fees-accruals quality relationship in firms with long tenured auditors. The prediction is less clear in the case of non-audit services. As auditor tenure lengthens, the auditor s familiarity with the client s business and accounting systems increases and thus the effect of knowledge spillovers can be expected to decline. On the other hand, auditor independence may be most impaired in the early years of the engagement (Gul et al. 2009). Geiger and Raghunandan (2002) also note that auditors are more likely to acquiesce to clients requests in the first few years of the engagement, and that their acquiescence diminishes after six or more years of tenure. We, therefore, do not make any directional prediction for this hypothesis. Our second set of hypotheses are as follows: H2a: The association between accruals quality and audit fees decreases as auditor tenure increases. H2b: The association between accruals quality and non-audit fees can increase or decrease as auditor tenure increases. 9 Short-tenured auditors (tenure of less than two to three years) are also associated with lower-quality audits (Johnson et al. 2002) and higher incidences of fraudulent reporting (Carcello and Nagy 2004). 8

10 9 RESEARCH DESIGN Sample Selection Panel A of Table 2 summarizes the sample selection process. The initial sample, obtained from Audit Analytics, consists of 49,824 firm-year observations on audit and non-audit fees during fiscal years Financial institutions (standard industrial classification (SIC) codes between 6000 and 6999) are excluded from the sample (see, also, Srinidhi and Gul 2007). Companies are deleted from the sample if they do not have sufficient data on Compustat for obtaining the variables used in our regressions. 10 Finally, we retain only Big N clients to ensure that there is relative homogeneity in accruals quality in the sample (Srinidhi and Gul 2007). These steps result in a final sample of 30,474 firm-year observations. Panel B of Table 2 shows the industry distribution using the one-digit SIC code of the companies. The firms are distributed widely among the different industries with a small clustering in manufacturing and business services, which is similar to the distribution of the Compustat population. (Insert TABLE 2 Here) Measure of Accruals Quality Following Srinidhi and Gul (2007), we estimate accruals quality (AQ) 11 using an industry-level pooled cross-sectional model (equation 1) that regresses total current accruals (TCA) on cash flows in the previous year (OCF t-1 ), the current year (OCF t ) and 10 Each firm is required to have five years of historical data for computing the standard deviations of sales revenues and operating cash flows. 11 The measure of accruals quality used in their paper (and our study) was proposed and used in Dechow and Dichev (2002) and Francis et al. (2005). Similar to Srinidhi and Gul (2007), we use this measure because it does not require us to separate the discretionary components from the non-discretionary components of accruals. The use of discretionary accruals obtained from models such as Jones (1991) as a proxy for financial reporting quality has been criticized because of the measurement error associated with separating discretionary from the non-discretionary components (Dechow et al. 1995; Guay et al. 1996; McNichols 2000). 9

11 10 the subsequent year (OCF t+1 ), as well as, on changes in revenue (ΔREV) and property, plant and equipment (PPE). We use all firms on Compustat in a given year within a twodigit SIC industry group that has at least 10 firms in each industry classification. The variable AQ is inverse of the absolute values of the residuals, v t, from equation (1). 12 TCA t = β 0 + β 1 OCF t-1 + β 2 OCF t + β 3 OCF t+1 + β 4 ΔREV t + β 5 PPE t + v t (1) Variable Definition TCA = (change in current assets change in cash (change in current liabilities change in short-term debt included in current liabilities)); OCF = Operating cash flow obtained from cash flow statements; ΔREV = Changes in sales revenues; PPE = Gross values of property, plant and equipment; Regression Model As in Srinidhi and Gul (2007), we then model AQ as a function of the fee variables and other determinants (equation 2). We additionally include an auditor tenure variable in equation 3. AQ t = β 0 + β 1 [AF t, NAF t ] + β 2 SIZE t + β 3 OCFSTD t + β 4 SALESTD t +β 5 OPCYCLE t + β 6 LOSS t + industry dummies t + year dummies t + u t (2) AQ t = β 0 + β 1 [AF t, NAF t ] +β 2 TENURE t + β 3 SIZE t + β 4 OCFSTD t + β 5 SALESTD t +β 6 OPCYCLE t + β 7 LOSS t + Industry dummies t + Year dummies t +u t (3) Variable Definition AQ = (-1)*absolute values of the residual in regressions relating accruals to past, current, and future operating cash flows, changes in sales revenues, and property, plant and equipment; 12 Higher values of v t imply that there is a higher estimation error in the mapping of current accruals into operating cash flow realizations after controlling for firm-specific characteristics. These values most likely represent discretionary accruals that are opportunistic since they are not realized as cash flows (Francis et al. 2005). v t is multiplied by (-1) for exposition purposes. 10

12 11 AF = Natural logarithm of audit fees; NAF = Natural logarithm of non-audit fees; TENURE = Number of years as auditor with a current client; SIZE = Natural logarithm of a firm s total assets; OCFSTD = Volatility in operating cash flows (OCF) measured as the standard deviation in quarterly OCF for the prior 20 quarters; SALESTD = Volatility in sales revenues measured as the standard deviation in quarterly sales for the previous 20 quarters; and OPCYCLE = Operating cycle measured by 360/(Sales/average accounts receivables) + 360/(cost of goods sold/average inventory); LOSS = 1 if the net income is less than 0 and 0 otherwise. AQ is expected to be lower for smaller firms and for those having greater cash flow volatility, sales volatility, longer operating cycles and higher incidences of losses (Dechow and Dichev 2002). The coefficients of main interest, however, are those on AF and NAF. In the pre-sox era, we predict that high audit fees will be associated with small accrual errors (or high AQ), while high non-audit fees will be associated with high accrual errors (or low AQ). These associations are expected to weaken in the post-sox era. Further, we hypothesize that auditor tenure will moderate the association between auditor fees and AQ. We test these predictions using interaction terms: between auditor fees and SOX in equation (4) and between auditor fees and auditor tenure in equation (5). 13 AQ t = β 0 + β 1 [AF t, NAF t ] +β 2 SOX t + β 3 [SOX*AF t, SOX*NAF t ] + β 4 SIZE t + β 5 OCFSTD t + β 6 SALESTD t +β 7 OPCYCLE t + β 8 LOSS t + Industry dummies t + Year dummies t +u t (4) AQ t = β 0 + β 1 [AF t, NAF t ] +β 2 TENURE t + β 3 [TENURE*AF t, TENURE*NAF t ] + β 4 SIZE t + β 5 OCFSTD t + β 6 SALESTD t +β 7 OPCYCLE t + β 8 LOSS t + Industry dummies t + Year dummies t +u t (5) 13 Following Gul et al. (2009), auditor tenure is calculated as the consecutive number of years auditing a client as reported in Compustat. As a test of sensitivity, this study also uses the natural log of tenure and finds qualitatively similar results. Similarly, results do not change when we alternatively measure AF and NAF as audit fees (non-audit fees) scaled by total assets. 11

13 12 EMPIRICAL RESULTS Descriptive Statistics Table 3, Panel A shows trends in audit and non-audit fees for the sampled years. In 2000, 2001 and 2002, non-audit fees are larger than audit fees. However, consistent with prior studies (e.g., Ghosh and Pawlewicz 2009), this trend reverses in the following years. The reversal is primarily due to the regulatory prohibition on certain non-audit services and the increased demand for audit services, post-sox. The mean audit fees increased by more than 500% between fiscal years 2000 ($551,143) and 2008 ($2,951,582), while the mean non-audit fees declined by approximately 50% over the same period (from $1,323,786 to $676,549). Panel B presents means, medians, and standard deviations of the variables used in our regressions. 14 The mean (median) value of AQ is ( ) which suggests that accruals, on average, contain positive residuals that are not realized into cash flows. This is comparable to findings in Francis et al. (2005). The mean and median values of TENURE are 9.6 years and seven years, respectively. This variable also has a distribution comparable to that documented by prior studies (e.g., Gul et al. 2009). (Insert TABLE 3 Here) Table 4 presents the Pearson correlations between the variables. AQ increases with: tenure, audit and non-audit fees, and firm size. On the other hand, AQ decreases with: the volatilities of cash flows and sales, the duration of operating cycles and the incidences of losses. TENURE is positively associated with audit and non-audit fees, firm size and the operating cycle. It is negatively associated with volatilities in cash flows and 14 All continuous variables were winsorized at both the 1% and 99% levels to reduce the effects of extreme values. 12

14 13 sales, and with incidences of losses. All of these results are consistent with the expectations, excepting non-audit fees which have a positive correlation with AQ. (Insert TABLE 4 Here) Accruals Quality and Auditor Fees Figure 1, Panel A graphs the coefficients of year-by-year regressions of AQ on audit fees and non-audit fees without considering other controls. The panel shows that the coefficients on audit fees are the largest (smallest) in the pre-sox (post-sox) years. This is consistent with audit fees reflecting a higher proportion of audit effort relative to rents in the pre-sox years. With regard to coefficients on non-audit fees, we find only marginally lower (higher) magnitudes in the pre-sox (post-sox) years. Panel B graphs the coefficients of regressions of AQ on audit fees and non-audit fees according to auditor tenure. The more positive coefficients on audit fees in the early years of an engagement suggest that in those years, audit fees proxy to a greater extent for audit effort. However, as the auditor tenure passes year seven, audit fees increasingly appear to represent higher rents, as evidenced by the smaller coefficients. In contrast, no clear pattern is observed in the coefficients on non-audit fees. (Insert FIGURE 1 Here) Pre- and Post-SOX Analyses Table 5 models empirically test for differences in the associations of accruals quality and auditor fees over pre- and post-sox periods. 15 The cross-sectional variation explained by these models (adjusted R 2 s) is above 22%, which is comparable to results in 15 All regression models presented in this paper include two-digit SIC industry and year dummies. The reported standard errors are robust with respect to firm clustering and all tests of significance reported are one-tailed where the direction of the association is expected and are two-tailed otherwise. 13

15 14 Srinidhi and Gul (2007). Estimation results for the pooled regression (Model 1) show positive and negative coefficients on AF and NAF, respectively. These are statistically significant at least at the 5% level. 16 The significant associations of AF and NAF found in the pre-sox period, however, disappear in the post-sox period (Models 2 and 3). Consistent with this, Model 4 results show negative and positive coefficients on SOX*AF and SOX*NAF. In the post-sox years, therefore, there is an attenuation in the relationship between fees and accruals quality. The results support hypotheses H1a and H1b. With respect to control variables, AQ is increasing in firm size (SIZE), decreasing in cash flow volatility (OCFSTD), sale volatility (SALESTD), and incidences of losses (LOSS). All of these associations are consistent with prior results. (Insert TABLE 5 Here) Auditor Tenure as a Moderating Variable Table 6 models include TENURE as an additional control variable. In Model 1 that includes the full sample, positive and negative coefficients are found on AF and NAF, respectively. Both coefficients are statistically significant at the 10% level. TENURE has a positive and statistically significant coefficient, suggesting that AQ is increasing with auditor tenure. In Model 2 that only includes firms whose auditors tenure is below the median tenure (short-tenure), we find a negative and statistically significant coefficient on NAF. This is consistent with non-audit fees having an adverse effect on auditor independence in the early years of an audit. In addition, as expected, a positive and a marginally 16 Since we successfully replicate Srinidhi and Gul (2007), we believe that our results for the post-sox years cannot be attributable to issues regarding research design, sample selection or lack of statistical power. 14

16 15 significant coefficient of AF is found, suggesting that high audit fees lead to higher AQ in the initial years of the engagement. With regard to Model 3 that includes only longer tenured auditors (tenure equal to or above median tenure), we observe insignificant coefficients on audit fees and nonaudit fees. It appears that neither audit fees nor non-audit fees have any effect on AQ in the longer auditor-tenured firms. In Model 4 that includes two interaction terms, we find a negative coefficient on TENURE*AF which suggests that audit fees are not as effective a proxy for audit effort as auditor tenure lengthens. However, auditor tenure does not moderate the association between non-audit fees and AQ, as evidenced by statistically insignificant coefficient on TENURE*NAF. Overall, the results generally support H2a, but not H2b. (Insert TABLE 6 Here) Sensitivity Tests (Results Untabulated) Using Discretionary Accruals Quality As our first test of sensitivity, we replace accruals quality (AQ) with discretionary accruals quality (DISAQ). 17 Following Francis et al. (2005), we decompose AQ into innate and discretionary components (DISAQ). 18 As expected, we find that audit fees (non-audit fees) are positively (negatively) associated with DISAQ only in the pre-sox years. Additionally, the effects of audit fees and non-audit fees on DISAQ are statistically significant only in the short auditor-tenured group of companies. As tenure lengthens, the 17 Similar to Srinidhi and Gul (2007) we use AQ instead of DISAQ in our main tests. By using AQ we avoid separating the discretionary from the non-discretionary (or innate) components of accruals. The separating procedure can result in erroneous assignments of non-discretionary items as discretionary items and vice versa. 18 The model used is: AQ t = β 0 + β 1 SIZE t + β 2 OCFSTD t + β 3 SALESTD t +β 4 OPCYCLE t + β 5 LOSS t + industry dummies t + year dummies t + u t. The inverse measures of the residuals are defined as discretionary AQ (DISAQ). 15

17 16 positive effect of audit fees in reducing discretionary estimation errors declines. Substituting discretionary accruals quality in place of AQ does not, therefore, qualitatively alter our conclusions. Using Abnormal Audit Fees Next, we attempt to disentangle rents from audit effort by calculating abnormal or excess audit fees. Abnormal audit fees are likely to reflect rents rather than audit effort. Similar to prior studies (e.g., Francis and Wang 2005; Hay et al. 2006), we first estimate an audit fees model using pre-sox data. 19 The coefficients obtained from the audit fees model are then used to predict post-sox audit fees (prior to the implementation of SOX 404). 20 Abnormal fees in the post-sox years are defined as actual audit fees in the post- SOX period minus the predicted audit fees. Post-SOX accruals quality (AQ) is then regressed on both abnormal audit fees and predicted audit fees along with other controls. We find that the coefficient on predicted audit fees is positive but statistically insignificant. However, the interesting finding here is that in contrast to Table 5 results showing an insignificant coefficient on audit fees, we find that the coefficient on abnormal audit fees is negative and statistically significant (p<0.10). These results are consistent with the argument that in the post-sox period, rents (abnormal audit fees) reflected in audit fees rather than audit effort (predicted audit fees) are more strongly related to accruals quality. 19 The audit fees model estimated using pre-sox data is AF t = β 0 + β 1 SIZE t + β 2 FRGN t + β 3 SEG t + β 4 RECV t + β 5 INVT t + β 6 OPIN t + β 7 DY t + β 8 LVRG t + β 9 LOSS t +β 10 CATA t + β 11 EBIT t +β 12 QUICK t + industry dummies + year dummies + ε, where AF is the natural log of audit fees, SIZE is firm size, FRGN indicates the presence of foreign transactions, SEG is the number of segments, RECV are accounts receivables, INVT is inventory, OPIN denotes unqualified opinions, DY is an indicator variable for firms having a December fiscal year-end, LVRG is leverage, LOSS is an indicator for firms reporting losses, CATA are current assets, EBIT is earnings before interest and taxes, and QUICK is the quick ratio. 20 We limit the post-sox period to years prior to SOX 404 because audit fees following SOX 404 also include fees paid for audits of internal control. SOX 404 became effective for large accelerated filers starting fiscal years ending on or after November 15,

18 17 Allowing Structural Changes from Pre- to Post-SOX We also allow each variable in our regressions in Table 5 to have a different coefficient before and after SOX. This specification is intended to address the possibility that there may have been structural changes in the determinants of accruals quality after SOX (Ghosh and Pawlewicz 2009). Our results show that the coefficients on audit and non-audit fees are statistically significantly positive and negative, respectively only in the pre-sox period. The interactive effect of audit fees with auditor tenure is negative and statistically significant in both pre and post-sox years. Finally, the interactive effect of non-audit fees with auditor tenure is positive and statistically significant only in the pre- SOX period. Overall, these results are similar to those documented in Table 5. Incorporating Litigation Risk Premium A factor affecting audit fees not considered by Srinidhi and Gul (2007) is audit risk from potential lawsuits. Krishnan et al. (2012) argue that auditors include a risk premium in audit fees when a point is reached where mitigation of potential litigation losses is not possible through additional audit effort. Ghosh and Pawlewicz (2009) document that the litigation risk premium priced in audit fees has increased in the post- SOX years. To the extent that post-sox audit fees reflect, to a proportionately greater degree, factors other than effort (e.g., risk premium) there would be an attenuation in their relationship with accruals quality. The attenuation would be pronounced in high litigation-risk industries that face bigger increases in risk premiums post-sox (Ghosh and Pawlewicz 2009). Consistent with this expectation, we find larger negative coefficients on SOX*AF for the high litigation-risk group. 21 Overall, results support our 21 High-litigation industries are those with the following SIC codes: 2,833 2,836, 3,570 3,577, 3,600 3,674, 5,200 5,961, and 7,370 7,374 (Francis et al. 1994). About 35% of our sample firms belong to these 17

19 18 basic argument that the relative proportion of audit effort reflected in post-sox audit fees has decreased. CONCLUSION This study examines the relationship between fees paid to auditors and accruals quality. We find results suggesting that audit fees reflect higher audit effort which in turn enhances accruals quality. However, we document two situations where higher audit fees do not lead to higher accruals quality: (1) in the years following SOX and (2) in firms with long-tenured auditors. We also find that the adverse effects of non-audit fees on accruals quality are present only in the pre-sox years and for firms having shorter tenured auditors. The present study contributes to the auditor fees literature by documenting that the relationship between audit and non-audit fees and accruals quality depends on the time period examined and the length of auditor tenure. Investors, audit committees and academics should not assume that high audit fee payments will necessarily lead to high quality audits, especially in the post-sox years and in firms having a lengthy relationship with the same audit firm. industries. The remainder are classified as low litigation industries. The coefficients on SOX*AF for the high and low litigation groups are and , respectively. 18

20 19 APPENDIX Fees Categories Disclosed in Proxy Statements Prior to and After December 15, 2003 Proxy Statements Filed after February 5, 2001but prior to December 15, 2003 (SEC 2000) (a) Audit Fees Defined as fees paid for the annual audit and review of financial statements for the most recent fiscal year. (b) Financial Information and System Design and Implementation Fees (c) All Other Fees Defined as the aggregate fees billed for specified information technology services rendered by the outside auditor during the most recent fiscal year. Specified information technologies services include the operation or management of a company s information system or local area network, and the design and implementation of a hardware or software system that aggregates data underlying the financial statements or otherwise generates information significant to a company s financial statements. Defined as the aggregate fees for services rendered by the principal accountant other than those disclosed in items (a) and (b) above for the most recent fiscal year. Fiscal Year Ending after December 15, 2003 (SEC 2003) (a) Audit Fees (b) Audit- Related Fees (c) Tax Fees (d) All Other Fees Defined as the aggregate fees billed for each of the last two fiscal years for professional services rendered by the principal accountant for the audit registrant s annual financial statements and review of financial statements included in the registrant s Form 10Q or 10- QSB or services that are normally provided by the accountant in connection with statutory and regulatory filings and engagements for those fiscal years. Defined as the aggregate fees billed in each of the last two fiscal years for assurance and related services by the principal accountant that are reasonably related to the performance of the audit or review of the registrant s financial statements. Defined as the aggregate fees billed in each of the last two fiscal years for professional services rendered by the principal accountant for tax compliance, tax advice, and tax planning. Defined as the aggregate fees billed in each of the last two fiscal years for other products and services provided by the principal accountant. 19

21 20 REFERECENS Abbott, L.J., S. Parker, and G. F. Peters Does mandated disclosure induce a structural change in the determinants of non-audit service purchases? Auditing: A Journal of Practice & Theory 30(2): Ashbaugh, H., R. LaFond, and B. Mayhew Do non-audit services compromise auditor independence? Further evidence. The Accounting Review 78: Asthana, S., and J. Krishnan Factors associated with early adoption of the SEC's revised auditor fee disclosure rules. Auditing: A Journal of Practice & Theory 25 (2): Caramanis, C.. and C. Lennox Audit effort and earnings management. Journal of Accounting and Economics 45: Carcello, J., and A. Nagy Audit firm tenure and fraudulent financial reporting. Auditing: A Journal of Practice & Theory 23: Charles, S. L., S. M. Glover, and N. Y. Sharp The association between financial reporting risk and audit fees before and after the historic events surrounding SOX. Auditing: A Journal of Practice & Theory 29 (1): Chung, H., and S. Kallapur Client importance, non-audit services, and abnormal accruals. The Accounting Review 78 (4): Davis, L. R., D. N. Ricchiute, and G. Trompeter Audit effort, audit fees, and the provision of non-audit services to audit clients. The Accounting Review 68(1): DeAngelo, L Auditor size and audit quality. Journal of Accounting and Economics 3: Dechow, P. M., and I. Dichev The quality of accruals and earnings: The role of accrual estimation errors. The Accounting Review 77: Dechow, P. M., W. Ge, C. Larson, and R. Sloan Predicting material accounting Misstatements. Contemporary Accounting Research 28(1): Dechow, P., R. Sloan, and A. Sweeney Detecting earnings management. The Accounting Review 70: DeFond, M. L., K. Raghunandan, and K. R. Subramanyam Do non-audit service fees impair auditor independence? Evidence from going-concern audit opinions. Journal of Accounting Research 40(4):

22 21 Dickins, D., and J. Higgs Interpretation and use of auditor fee disclosures. Financial Analysts Journal 61(3): Francis, J. R., and B. Ke Disclosure of fees paid to auditors and the market valuation of earnings surprises. Review of Accounting Studies 11: Francis, J. R., R. LaFond, P. Olsson, and K. Schipper The market pricing of accruals quality. Journal of Accounting and Economics 39(2): Francis, J. R., D. Philbrick, and K. Schipper Shareholder litigation and corporate disclosures. Journal of Accounting Research 32 (2): Francis, J. R., and D. Wang Impact of the SEC s public fee disclosure requirement on subsequent period fees and implications for market efficiency. Auditing: A Journal of Practice & Theory 24: Frankel, R., M. Johnson, and K. Nelson The relation between auditors fees for non-audit services and earnings management. The Accounting Review 77: Fung, S.Y., F. A. Gul, and J. Krishnan City-level auditor industry specialization, economies of scale, and audit pricing. The Accounting Review 87(4): Geiger, M. A., and K. Raghunandan Auditor tenure and audit reporting failures. Auditing: A Journal of Practice & Theory 21: Ghosh, A., and R. Pawlewicz The impact of regulation on auditor fees: Evidence from the Sarbanes-Oxley Act. Auditing: A Journal of Practice & Theory 28(2): Gifford, R. H., and H. Howe Regulation and unintended consequences: Thoughts on Sarbanes-Oxley. The CPA Journal 74(6): Government Accounting Office (GAO) Public Accounting Firms: Required Study on the Potential Effect on Mandatory Audit Firm Rotation. Report Washington D.C.: Government Printing Office. Guay, W., S. P. Kothari, and R. Watts A market-based evaluation of discretionary accrual models. Journal of Accounting Research 34: Gunny, K., and T. Zhang PCAOB inspection reports and audit quality. Available at: Gul, F. A., S. Y. Fung, and B. Jaggi Earnings quality: Some evidence on the role of auditor tenure and auditors industry expertise. Journal of Accounting and Economics 47:

23 22 Hay, D., W. R. Knechel, and N. Wong Audit fees: A meta-analysis of the effect of supply and demand attributes. Contemporary Accounting Research 23(1): Higgs, J. L., and T. R. Skantz Audit and non-audit fees and the market s reaction to earnings announcement. Auditing: A Journal of Practice & Theory 25(1): Hoitash, R., A. Markelevich, and C. A. Barragato Auditor fees and audit quality. Managerial Auditing Journal 22(8): Johnson, V., I. Khurana, and J. Reynolds Audit-firm tenure and the quality of financial reports. Contemporary Accounting Research 19: Kinney Jr., W. R., and R. Libby Discussion of the relation between auditors fees for non-audit services and earnings management. The Accounting Review 77: Kinney Jr., W. R., Z. Palmrose, and S. Scholz Auditor independence, non-audit services, and restatements: Was the U.S. government audit right? Journal of Accounting Research 42: Krishnan, G. V., M. Pevzner, and P. Sengupta How do auditors view managers voluntary disclosure strategy? The effect of earnings guidance on audit fees. Journal of Accounting and Public Policy 31: Krishnan, J., D. Rama, and Y. Zhang Costs to comply with SOX Section 404. Auditing: A Journal of Practice & Theory 27(1): Krishnan, J., H. Sami, and Y. Zhang Does the provision of non-audit services affect investor perceptions of auditor independence? Auditing: A Journal of Practice & Theory 24: Krishnan, J., L. Su, and Y. Zhang Non-audit services and earnings management in the pre-sox and post-sox eras. Auditing: A Journal of Practice & Theory 30(3): Larcker, D. F., and S. A. Richardson Fees paid to audit firms, accrual choices, and corporate governance. Journal of Accounting Research 42(3): Levitt, A Speech by SEC chairman: renewing the covenant with investors. New York University Center for Law and Business. Available at: McNichols, M Research design issues in earnings management studies. Journal of Accounting and Public Policy 19: Montana, J The Sarbanes-Oxley Act: Five years later. Information Management Journal 41(6):

24 23 O Keefe., T. B., D. A. Simunic, and M. T. Stein The production of audit services: Evidence from a major public accounting firm. Journal of Accounting Research 32(2): Raghunandan, K., and D. V. Rama SOX Section 404 material weakness disclosures and audit fees. Auditing: A Journal of Practice & Theory 25(1): Raghunandan, K., W. J. Read, and J. S. Whisenant Initial evidence on the association between non-audit fees and restated financial statements. Accounting Horizons 17(3): Rice, S. C., and D. P. Weber How effective is internal control reporting under SOX 404? Determinants of the (non-)disclosure of existing material weaknesses. Journal of Accounting Research 50(3): Securities and Exchange Commission (SEC) Final Rule: Revision of the Commission's Auditor Independence Requirements. Release No Washington, D.C.: Government Printing Office. Securities and Exchange Commission (SEC) Final Rule: Revision of the Commission's Auditor Independence Requirements. Release Nos ; ; ; IC-24744; IA-1911; FR-56; File No. S Washington, D.C.: Government Printing Office. Securities and Exchange Commission (SEC) Final Rule: Strengthening the Commission's Requirements Regarding Auditor Independence. Release Nos ; ; ; IC-25915; IA-2103, FR-68, File No. S Washington, D.C.: SEC. Simunic, D. A Auditing, consulting, and auditor independence. Journal of Accounting Research 22: Srinidhi, B. N., and F. A. Gul The differential effects of auditors non-audit and audit fees on accrual quality. Contemporary Accounting Research 24(2): Weil, J., and M. Rapoport New SEC definition may cloud audit fees. Wall Street Journal 22 January, C1. 23

25 24 Figure 1: Regressions Coefficients of Accruals Quality on Audit and Non-audit Fees Panel A: Regression Coefficients by Year Panel B: Regression Coefficients by Auditor Tenure The horizontal axis represents fiscal year in Panel A and the number of years of auditor tenure in Panel B. The vertical axis represents coefficients that are obtained from the following regression of accruals quality (AQ) on audit fees (AF) and non-audit fees (NAF) separately by year (Panel A) and by length of auditor tenure (Panel B). The model used is AQ t = β 0 + β 1 AF t, + β 2 NAF t + u t 24

26 25 Table 1: Variable Definitions Variable Definition AQ = (-1)*absolute values of the residual in regressions of accruals on past, current and future operating cash flows, changes in sales revenues, and property, plant, and equipment; TENURE = Auditor tenure; AF = Natural logarithm of audit fees; NAF = Natural logarithm of non-audit fees; SIZE = Natural logarithm of total assets; OCFSTD = Volatility in operating cash flows (OCF) measured as the standard deviation in quarterly OCF for the past 20 quarters; SALESTD = Volatility in sales revenues measured as the standard deviation in quarterly sales for the past 20 quarters; OPCYCLE = Operating cycle measured as: 360/(Sales/average account receivables) + 360/(cost of goods sold/average inventory); and LOSS = 1 if net income is less than 0 and 0 otherwise. 25

27 26 Table 2: Sample Selection and Industry Distribution Panel A: Selection Procedures Firms with audit fees 49,824 Less: firms in financial industries 4,608 firms without financial data 4,643 firms with non-big 4 auditors 10,099 Sample 30,474 Panel B: Industry Distribution SIC Codes Industry # Observations Percent Mining, Construction 1, Manufacturing food, textiles, lumber, chemicals 5, Manufacturing Rubber, metal, machinery, 9, equipment Transportation, Communication, Utilities 4, Wholesale, Retail 3, Services 5, Total 30,

28 27 Table 3: Descriptive Statistics Panel A: Mean Auditor Fees over the Sample Period Year # Observations Audit Fees ($) Non-audit Fees ($) Total Fees ($) , ,143 1,323,786 1,874, , ,945 1,275,422 1,856, , , ,142 1,888, ,034 1,140, ,829 1,953, ,736 1,940, ,021 2,711, ,433 2,373, ,727 3,056, ,202 2,722, ,739 3,390, ,011 2,823, ,088 3,532, ,887 2,951, ,549 3,628,131 Panel B: Summary Data on Variables Used in the Regressions (N=30,474) Variable Mean Std P25 P50 P75 AQ TENURE AF NAF SIZE OCFSTD SALESTD OPCYCLE LOSS

29 28 Table 4: Pearson Correlation Matrix AQ TENURE AF NAF SIZE OCFSTD SALESTD OPCYCLE TEN- URE AF NAF SIZE OCF- STD SALE- STD OP- CYCLE (0.01) LOSS See Table 1 for variable definitions. 28

30 29 Table 5: Regressions of Accruals Quality on Auditors Fees, Pre- and Post-SOX This table reports estimation results of regressions relating accruals quality to auditor fees. We partition the sample into the pre-sox (2000 to July 30, 2002) and post-sox (August 1, 2002 to 2008) years. The dependent variable, AQ, is measured as (-1)*the absolute values of residuals from regressions of accruals on past, current and future operating cash flows, changes in sales revenues, and property, plant, and equipment. All standard errors are robust with respect to firm clustering. T-values are based on one-tailed tests if the coefficient has an expected sign and on two-tailed tests otherwise. ***, **, and * denotes 1%, 5%, and 10% significance, respectively. See Table 1 for variable definitions. Variable Pooled (1) Pre-SOX (2) Post-SOX (3) Interaction (4) Intercept (-13.82)*** (-10.15)*** (-7.44)*** (-13.73)*** AF (2.41)** (6.11)*** (-1.18) (2.35)** NAF (-2.11)** (-3.07)*** (-0.05) (-1.76)* SOX (-0.97) SOX*AF (-4.39)*** SOX*NAF (4.71)*** SIZE (9.86)*** (0.45) (11.70)*** (10.08)*** OCFSTD (-42.42)*** (-18.26)*** (-38.41)*** (-42.19)*** SALESTD (-4.02)*** (-2.38)** (-2.61)*** (-3.86)*** OPCYCLE (0.40) (0.36) (0.01) (0.42) LOSS (-14.56)*** (-9.07)*** (-11.17)*** (14.50)*** Industry Controlled Controlled Controlled Controlled Year Controlled Controlled Controlled Controlled N 30,474 6,958 23,516 30,474 F-value *** 37.08*** *** *** Adj-R