Professional Documents
Culture Documents
The Impact of Regulation On Auditor Fees: Evidence From The Sarbanes-Oxley Act
The Impact of Regulation On Auditor Fees: Evidence From The Sarbanes-Oxley Act
171197
The Impact of Regulation on Auditor Fees: Evidence from the Sarbanes-Oxley Act
Aloke Ghosh and Robert Pawlewicz
SUMMARY: We examine changes in auditor fees around the Sarbanes-Oxley Act (SOX). Audit fees are expected to increase after SOX due to an increase in both audit effort and auditors expected legal liability. Our results indicate an economically large increase in audit fees following the enactment of SOX. Controlling for audit and client characteristics, we nd that audit fee levels rose approximately 74 percent in the postSOX period. Nonaudit fees fell signicantly over the same period, but total fees paid to auditors rose because the increase in audit fees more than offset the decline in nonaudit fees. Additionally, we nd that the Big 4 audit rms increased audit fees by 42 percent more than their smaller counterparts. Finally, we nd that while small and large audit rms discount fees on initial engagements to attract new clients in the preSOX period, only small audit rms continue to offer fee discounts for the post-SOX years. Our ndings are robust to a host of sensitivity tests. Keywords: Sarbanes-Oxley Act; audit pricing; auditor fees; Big 4 premium; fee discounting; lowballing. Data Availability: Contact the rst author about availability of the data.
INTRODUCTION hile several studies investigate the cross-sectional determinants of audit fees (e.g., Simunic 1980; Palmrose 1986; Craswell et al. 1995; Bell et al. 2001; Ghosh and Lustgarten 2006), relatively few academic studies examine the intertemporal variation in audit fees. Anecdotal evidence, media coverage, and the popular press take for granted that the additional audit work imposed by the Sarbanes-Oxley Act of 2002 (SOX; U.S. House of Representatives 2002) spurred an increase in audit fees (Ciesielski and Weirich 2006; Wilcox 2007). Drawing heavily on Simunics (1980) audit fee model, we empirically examine the impact of SOX on audit fees.
Aloke Ghosh is a Professor and Robert Pawlewicz is a Ph.D. Student, both at Baruch CollegeCUNY.
We thank Dan Simunic, Jeong-Bon Kim (the discussant), John Elliot, Douglas Carmichael, Donal Byard, Carol Marquardt, Lale Guler, Jianming Ye, Steven Lustgarten, Diana DAmico, two anonymous reviewers, the participants of the 2008 JCAE / AJPT Joint Symposium, and the research seminar at Baruch College, City University of New York, for their helpful comments. Editors note: Accepted by Dan Simunic.
Submitted: August 2007 Accepted: March 2009 Printed Online: November 2009
171
172
SOX redenes the relationship between public companies and their auditors through the formation of a new regulatory body for the public accounting industry. The creation of the Public Company Accounting Oversight Board (PCAOB) has increased both oversight and penalties for audit-related violations. Additionally, several other provisions of SOX are expected to increase audit effort. For example, Section 103 requires audit workpaper retention for a minimum of seven years, Section 203 expands the required communications between the auditor and audit committee, and Section 404 requires that auditors attest to managements assessment of their internal controls.1 In the academic literature, audit fees are frequently modeled as a function of the cost of the audit effort and the auditors expected legal liability (Simunic 1980). Because some key SOX provisions involve substantial increases in audit effort, a growing consensus has emerged among academics and practitioners that audit fees are expected to increase following SOX (e.g., Financial Executives International [FEI] 2004; Grifn and Lont 2007; Iliev 2007). Audit fee increases after SOX could also be driven by a corresponding increase in auditors expected legal liability. Expected legal liability depends on several key factors, including the probability of material misstatements in nancial reports, the probability that the audit would fail to detect a misstatement, and the probability that the auditor would incur a legal liability due to an audit failure (Choi et al. 2008). While the likelihood of material misstatements and audit failure might have declined for the post-SOX period because of enhanced nancial reporting and increased audit effort, expected legal liability may have increased. SOX empowers federal courts and the Securities and Exchange Commission (SEC) to impose equitable remedies for violations of federal securities laws; this suggests the potential for increasing civil monetary penalties levied against the auditor following SOX (Rashkover and Winter 2005). In addition, the creation of a new regulatory body (i.e., the PCAOB) could also increase auditors exposure to legal liability vis-a-vis penalties for audit-related violations.2 ` In summary, audit fees are expected to increase after SOX because of (1) increased audit effort and (2) greater exposure to legal liability. Prior studies conclude that large auditors charge a premium for providing a superior level of audit assurance (e.g., Craswell et al. 1995; Simunic and Stein 1996; Choi et al. 2008). Since SOX is widely expected to increase the audit workload, large auditors providing a higher level of audit assurance are expected to increase audit fees more than small audit rms. The premium associated with large auditors could also increase following SOX if, as some believe, large auditors have a disproportionately greater exposure to litigation risk for the post-SOX period.3 We also examine whether the practice of fee discounting or lowballing on initial audit engagements, documented by prior studies, changed subsequent to SOX (e.g., Simon and Francis 1988; Chan 1999; Craswell and Francis 1999; Ghosh and Lustgarten 2006). Because voluntary auditor switchinga key determinant of fee discounting (Ghosh and
1
Similarly, Section 401 of the Act requires the Financial Accounting Standards Board (FASB) and the SEC to issue new regulations regarding off-balance-sheet activities, including the use of special-purpose entities and pro-forma gures in earnings announcements. These new regulations, including FIN No. 46 and Regulation G, increased required disclosures and the corresponding audit work. SOX 105(c)(4) established monetary penalties of up to $15,000,000 ($750,000) for a public accounting rm (individual) in addition to other sanctions for intentionally violating the act. Further, Cornerstone Research nds that damages claimed by plaintiffs in securities class action lawsuits are trending upward, which is an important explanation for the increase in the average size of the settlements (Cummings 2005). Refer to a discussion of the speech by Conrad Hewitt, the Chief Accountant of the SEC, at the Northwestern University Legal Conference in San Diego on January 25, 2007 in Burns (2007).
November 2009
173
Lustgarten 2006)is expected to decline for the post-SOX years from higher switching costs (Asthana et al. 2004), we posit that fee discounting is lower for the post-SOX years. Using a large sample between scal years 2000 and 2005, we document a signicant increase in audit fees around the time of SOX. Prior to SOX, mean audit fees were $533,360, but this number increased to $1,185,322 in the years following enactment. Thus, audit fees increased more than 122 percent between the pre- and post-SOX years. At the same time, average nonaudit fees declined from $1,259,634 to $671,364. Total fees increased concurrently because the increase in audit fees offset the decline in nonaudit fees. While these results provide preliminary insights into the trends in audit fees, we need to consider several other factors that might explain the results, including changes in client characteristics and sample composition associated with audit fees. Therefore, similar to the approach taken by prior studies (e.g., Craswell and Francis 1999; Ghosh and Lustgarten 2006), we estimate the impact of SOX on audit fees using a regression specication that controls for cross-sectional variation in audit fees. Controlling for the size of the auditor; auditors opinion; market concentration in the audit industry; and client characteristics such as size, risk, and complexity, our results suggest that audit fees rose by approximately 74 percent for the years subsequent to SOX. We also estimate audit fee regressions using a restricted sample that includes rms with available data for all of the relevant years. Our results using the constant sample indicate similar increases in audit fees for the post-SOX period. Because large (i.e., Big 4) auditors, with so-called deep pockets, have more to lose when litigation risk increases, the increase in audit fees for the post-SOX period is expected to be larger for brand-name auditors. Consistent with our predictions, we nd that the average increase in audit fees is 42 percent higher for clients of large audit rms than for clients of small audit rms. We nd that the extent of fee discounts on audit fees for new engagements declined over the post-SOX period. Ghosh and Lustgarten (2006) hypothesize that fee discounting is conned to the audit market segment consisting of small auditors. Because competition among small auditors is high, they are more likely than large auditors to reduce fees to attract new clients. In contrast, because competition for new clients among large auditors is low, client turnover for large auditors is relatively low. Consistent with the ndings of Ghosh and Lustgarten (2006), we nd that small and large auditors discounted fees for new engagements during the pre-SOX years. However, there is no evidence of lowballing among large auditors after enactment. Small auditors give price discounts to new clients for both the pre- and post-SOX years. There are two possible explanations for our results. First, a higher workload might preclude large auditors from fee discounting to attract new clients for the post-SOX years. Second, the increased threat of litigation could also limit large auditors from offering price discounts to new clients on initial engagements for post-SOX years.4 Our study is one of the rst to provide evidence on audit fees around SOX using a rigorous empirical research design. In a recent study, Grifn and Lont (2007) also examine audit fees around SOX, but several key features distinguish our contribution from theirs. First, they focus on residual audit fees, while we analyze audit, nonaudit, and total fees
4
Our results are robust to several added sensitivity analyses, including using levels and changes specications, conducting yearly regressions, excluding the forced switching of former Andersen clients from our sample, including additional controls for client business complexity, deating audit fees to constant 2000 dollars, and including an indicator variable for high-litigation industries.
174
paid. Second, their data are limited to large auditors, while we compare the differences in audit fees between large and small auditors around SOX. Finally, we are the rst study to examine fee discounting associated with new engagements and how this phenomenon changed following the acts passage. The remainder of this paper is organized as follows. First we review the related literature on regulation and audit fees. Next we explain the empirical methods and research design, discuss our data, and present descriptive statistics. Finally, we discuss our main results and our sensitivity analyses, and conclude. REGULATION AND AUDIT FEES Background The bulk of the audit fee literature employs some variation of the audit fee model proposed by Simunic (1980) which expresses audit fees (or the cost of an audit) as the sum of the cost of audit effort (c q) and an expected liability loss component (E[L] E[ ]): E[C] c q E[L] E[ ] (1)
where E[ ] represents the expectation operator, C is the total cost of the audit to an auditor, c is the per unit cost of an audit including all normal prots, q is the quantity or effort supplied by an auditor, is the ex post fraction of the losses from lawsuits borne by the auditor, E[L] is the present value of future losses from lawsuits, and E[ ] is the expected E[ ] is the present fraction of the legal losses borne by the auditor. Therefore, E[L] value of future losses expected to be incurred by the auditor if the engagement results in lawsuits. In equilibrium, the total cost of the audit (E[C]) is equal to audit fees. In the Simunic (1980) model, the present value of future losses is a function of the quantity of resources utilized directly by the client rm in operating the internal accounting system (a) and the quantity of resources used by the auditor in performing the audit (q), that is: E[L] F(a, q) (2)
where F denotes a functional form. Much of the audit fee literature uses rm characteristics including size, complexity of operations, and business risk to proxy for the cost of audit effort (Simunic and Stein 1996). Prior audit fee research based on variations of Simunics (1980) audit fee model includes studies that examine how audit fees change with client losses (Simunic 1980), rms switching auditors (e.g., Craswell and Francis 1999; Ghosh and Lustgarten 2006), rms listing on a public exchange (Palmrose 1986), auditors perception of their clients risk (Bell et al. 2001), and audit services for nancial institutions (Fields et al. 2004). Menon and Williams (2001), one the few studies to analyze long-term trends in audit fees, examine whether audit fees changed from 1980 to 1997. Hence, their study does not address whether audit fees changed around SOX. In the following subsections, we analyze the impact of the Sarbanes-Oxley Act on audit fees, the Big 4 premium, fee discounting, and nonaudit fees. Audit Fees: The Sarbanes-Oxley Act and Audit Effort Assuming that the per-unit cost of an audit (c) has remained unchanged over the recent years, audit fees are determined by audit effort (q) and the expected litigation costs for the
November 2009
175
auditor (E[L] E[ ]). Ceteris paribus, several requirements of the Sarbanes-Oxley Act have increased the audit workload, which is expected to lead to higher audit fees. Among others:
Section 101 created the PCAOB to regulate the public accounting industry. Greater oversight of the accounting profession suggests that auditors are likely to be more diligent in their audit work. Consequently, audit effort is expected to increase as auditors spend more time conducting their audits and maintaining extensive records of their audit work for PCAOB examinations. Section 103 requires that audit rms retain audit workpapers for at least seven years, provide a second partner review of the audit report, and describe the extent of testing of the internal control structure of the company. Sections 202 and 204 require audit committee preapproval for services provided by the external auditor, and greater communication between the auditor and audit committee. Preapproval and increased communication, both written and verbal, require greater effort on the auditors part. Section 401 introduces new rules regarding reporting for off-balance-sheet transactions, pro forma nancial reporting, and special-purpose entities.5 These disclosures are expected to increase the effort necessary to both produce and audit nancial statements. Section 404 requires the external audit rm to attest to managements assessment of internal controls as part of the annual audit engagement.
In summary, each of these provisions requires more effort during the nancial statement audit, which is expected to increase audit fees. Audit Fees: The Sarbanes-Oxley Act and Legal Liability In addition to audit effort, in the Simunic (1980) model, the auditors legal liability also plays a crucial role in determining the level of audit fees. In general, an auditors legal liability cost is a function of (1) the probability of a material misstatement, (2) the probability of an audit failure, and (3) the probability of incurring legal costs associated with the failure (Choi et al. 2008). As the quality of nancial statements improves from the enhanced reporting requirements of SOX, maintenance and testing of internal controls, management certication of nancial reports, oversight of the accounting profession by the PCAOB, and enhanced oversight by audit committees, the probability of material misstatements is expected to decline. The incidence of audit failure is also expected to fall subsequent to SOX as auditors exert greater effort to produce their audit opinion and attest to managements assessment of internal controls. Despite an expected decrease in the probability of material misstatements in nancial statements and in the incidence of audit failures, many believe that auditors legal liability might have risen with the passage of SOX due to an increase in legal costs associated with audit failures (Grifn and Lont 2007). In a recent study, PricewaterhouseCoopers (PwC 2008) nds that the average accounting-related settlement (excluding outliers) increased
5
The SEC requires all companies to disclose material nancial results, commitments, transactions, relationships with entities, and uncertainties as part of disclosures with nancial statements. Regulation G, released by the SEC in January 2003, requires non-GAAP- and GAAP-based nancial results to be reconciled in accordance with this section of the act. Additionally, the FASB produced multiple new rules governing the reporting of offbalance-sheet transactions (FASB Interpretation Nos. 45 and 46R).
176
steadily after SOX until peaking in 2005.6 In a related study, Cornerstone Research concludes that the damages claimed by plaintiffs in securities class action lawsuits are trending upward, thereby driving average settlements even higher (Cummings 2005). Although these studies examine securities lawsuits, not specically public accounting lawsuits, the general increasing trend in securities class action settlements suggests an increase in legal liability for auditors.7 Several provisions of SOX are also expected to increase auditors exposure to legal liability. First, the act signicantly expanded the SECs enforcement powers related to accounting matters, especially those involving allegations of accounting fraud and management misconduct (Rashkover and Winter 2005). For example, the Fair Funds provision allows the SEC to direct civil monetary penalties to aggrieved investors rather than to the U.S. Treasury. Further, SOX codies case law empowering federal district courts to impose equitable remedies to benet victims in SEC enforcement actions. While SOX does not change the way civil penalties are determined, it affords the SEC greater discretion to seek larger civil penalties in enforcement actions (Heffes 2005). Additionally, SOX may have led to greater legal liability costs because of new penalties imposed by the PCAOB in cases of audit failure. Collectively, the higher level of penalties for fraud and audit failure under SOX may have led to a perception that auditors legal liability changed after its enactment. Therefore, we expect audit fees to increase subsequent to SOX due to an increase in audit effort and an increase in auditors exposure to legal liability from lawsuits and nes imposed by regulatory bodies. Big 4 Premium and the Sarbanes-Oxley Act Prior studies document that large (Big 4) audit rms charge a fee premium both in the U.S. and around the world (e.g., DeAngelo 1981; Palmrose 1986; Craswell et al. 1995; Choi et al. 2008). We examine whether SOX had any signicant impact on the fee premium charged by large auditors. In a recent study, Choi et al. (2008) provide links among audit pricing, legal liability and the Big 4 fee premium. They nd that audit fees increase monotonically as the strength of a countrys legal liability improves. Since an auditors legal liability is expected to increase with the quality of the legal regime, auditors charge a higher fee as compensation. Further, for a given legal liability regime, Choi et al. (2008) nd that large auditors charge a fee premium. Because legal liability costs are higher for large auditors, they have greater incentives to increase audit effort compared with small auditors, which suggests higher fees. Finally, Choi et al. (2008) argue that the Big 4 fee premium decreases as the legal regime becomes stronger. They reason that because small auditors have a higher audit failure rate than Big 4 auditors, smaller auditors increase audit fees signicantly more to compensate for their increase in legal liability costs. Assuming that SOX ushered in a stricter legal regime, the Choi et al. (2008) ndings suggest a decline in the Big 4 audit fee premium for the post-SOX period.
6
The PwC (2008) study also documents that the average number of cases led post-SOX is lower than the preSOX level. This may indicate a less litigious environment for the post-SOX years, but the size of accountingrelated settlements indicates otherwise. Most of the assertions on perceived changes in auditors litigation risk around SOX are casual observations or claims based on anecdotal evidence. Although a rigorous study examining the inuence of SOX on changes in auditors legal liability is beyond the scope of this paper, we believe such a study is likely to provide valuable insights.
November 2009
177
While Choi et al. (2008) model the Big 4 premium as a function of a countrys legal regime and the audit failure rate, in many countries, such as the United States, the strictness of the legal regime varies by audit rm size. Section 104 of the act directs the PCAOB to conduct annual reviews of all audit rms that audit more than 100 SEC registrants. Consequently, the PCAOB has made public its annual reviews of the Big 4 rms, drawing added scrutiny to their work (Reilly 2007). Additionally, the Big 4 rms audit a disproportionate number of large rms, thereby exposing them to a greater number of large lawsuits (Choi et al. 2008). Consistent with this expectation, in a 2007 speech, SEC Chief Accountant Conrad Hewitt commented that auditor legal liability is worrisome for regulators because the Big 4 accounting rms audit most U.S. public companies. An extension of the Choi et al. (2008) model is that the legal environment may be relatively stricter for the Big 4 auditors after SOX, allowing Big 4 auditors to potentially increase their premium. We empirically test the impact of SOX on the Big 4 premium. Fee Discounting and the Sarbanes-Oxley Act Prior literature nds evidence of fee discounting by audit rms to attract new audit clients (e.g., Simon and Francis 1988; Chan 1999; Craswell and Francis 1999). Ghosh and Lustgarten (2006) hypothesize that fee discounting is more intense among small audit rms than among large audit rms. The audit market can be grouped into two broad categories: a large number of small auditors offering similar products, and a small number of large auditors offering specialized audit services. Since audit quality is considered similar across small auditors, they are expected to compete solely on price and to offer fee discounts to attract new clients. Conversely, large auditors with specialized audit services (the Big 4) are less likely to compete on price because of the fear of price retaliation from rivals. Thus, differences in the degree of competition in the two types of audit markets are considered as a key explanation for fee discounting in Ghosh and Lustgarten (2006). One implication of their study is that fee discounting is expected to continue for small auditors for the post-SOX period because SOX did change the degree of competition among small auditors. While the demise of Andersen reduced the level of competition within large auditors (i.e., the oligopolistic sector), competition within the non-Big 4 audit market (i.e., atomistic sector) remained virtually unchanged. In contrast, fee discounting is expected to diminish for large auditors because the downfall of Andersen in June of 2002 increased the concentration in the market for large auditors. Asthana et al. (2004) nd that there is a decline in the frequency of voluntary auditor switches among the Big 4 auditors because of an increase in switching costs due to the loss of Andersen. Both these arguments suggest fewer incentives for Big 4 auditors to offer fee discounts to attract new clients after SOX. EMPIRICAL METHODS Levels Specications Researchers analyzing fees paid to external auditors typically regress audit fees on variables that proxy for the unobservable audit production function and the auditors litigation risk (e.g., Simon and Francis 1988; Craswell and Francis 1999; Ghosh and Lustgarten 2006). The adjusted R2s from these models are quite high, which indicates a low likelihood that the experimental variables do not proxy for correlated omitted variables (Craswell et al. 1995). We estimate the following multivariate regression model to estimate the inuence of regulation on audit fees.
0 5 8
Herfindahl
7 10
Assets Loss
.
Currentratio Leverage
14
Receivableratio
11
Profitability
15
12
Opinion
Segments
Foreign
(3)
Audit fees is the natural logarithm of fees paid to the external audit rm for the scal years nancial statement audit (from Audit Analytics), and SOX is an indicator variable that equals 1 if the scal year-ends after July 30, 2002, 0 otherwise.8 A positive coefcient on SOX ( 1) indicates that, controlling for other factors, average audit fees increased after SOX. The control variables are dened as follows: Auditor is an indicator variable that equals 1 when the external auditor is a Big 4 or a Big 5 auditing rm (auditor information is obtained from Audit Analytics), 0 otherwise; Herndahl is the normalized HerndahlHirschman Index (HHI) within each two-digit standard industry classication (SIC) code for each scal year;9 Assets is the natural logarithm of total assets (Compustat item #6); Currentassets is the ratio of current assets (Compustat item #4) to total assets; Currentratio is the ratio of current assets to current liabilities (Compustat item #5); Receivableratio is the ratio of total receivables (Compustat item #2) to total assets; Inventoryratio is the ratio of inventory (Compustat item #3) to total assets; Leverage is the ratio of total debt (Compustat items #9 and #34) to total assets; Protability is the ratio of income before extraordinary items (Compustat item #18) to total assets; Loss is an indicator variable that equals 1 for rms with negative income before extraordinary items, 0 otherwise; Merger is an indicator variable that equals 1 if the rm has a nonzero number in the Acquisition/Merger Pretax eld (Compustat item #360), 0 otherwise; Opinion is an indicator variable that equals 1 if the audit opinion (Compustat item #149) is not listed as unqualied, 0 otherwise; Segments is the number of business segments reported in the Compustat Segments le; and Foreign is the ratio of sales revenue generated outside of the U.S. to total reported revenue in the Compustat Segments le.10 All the variables are measured as of the end of the scal year. Because large audit rms charge a premium for providing higher quality audits (e.g., DeAngelo 1981; Palmrose 1986; Craswell et al. 1995; Choi et al. 2008), we expect a positive relationship between Auditor and Audit fees. We include Herndahl to control for the concentration in the audit industry; we expect a positive relationship between Herndahl and Audit fees. We control for client rm size (Assets) because rm size is positively associated with Audit fees (Simunic 1980). We use Currentassets, Currentratio, Leverage, Protability, Loss, and Opinion as proxies for audit risk, which are also positively correlated with audit fees (Simunic 1980; Craswell et al. 1995; Seetharaman et al. 2002). Finally, we include Receivableratio, Inventoryratio, Merger, Segments, and Foreign as proxies for audit complexity (Ghosh and Lustgarten 2006).
8
10
We use July 30, 2002, as the cutoff date because President George W. Bush signed the bill into law on this date, with the provisions of the act becoming effective after this date. Herndahl is computed as the sum of the market shares squared for each audit rm for each industry, which is then normalized by subtracting 1 / n and dividing the difference by 1 (1 / n), where n is the number of audit rms in the industry. Refer to the U.S. Department of Justice website (http: / / www.usdoj.gov / atr / public / testimony / hhi.htm) for further details. Segment information is only available for a subsample of 11,257 observations. Therefore, all the tests using these variables are conned to the smaller subsample.
November 2009
179
Because large audit rms with brand-name concerns are potentially more likely to be affected by the SOX provisions, we also interact SOX with Auditor (SOX * Auditor) to investigate whether the increase in audit fees for the post-SOX period is bigger for large audit rms. Specically, we estimate the following augmented model. Audit fees
0 5 8 1
SOX
6
SOX * Auditor
7
Auditor
Herfindahl
Assets
Currentassets
9 12 16
Currentratio
10
Inventoryratio
13
Leverage
14
11 15
Loss
Merger
Opinion (4)
Foreign
A positive coefcient on SOX * Auditor ( 2) indicates that, controlling for other factors, the average increase in audit fees is larger for the post-SOX period for Big 4 audit rms compared with small audit rms. To further investigate the impact of SOX on nonaudit fees and total fees, we modify Equation (4) by replacing Audit fees with Nonaudit fees and Total fees (both logarithmic transformations). Nonaudit fees
0 5 8 1
SOX
6
SOX * Auditor
7
Auditor
Herfindahl
Assets
Currentassets
9 12 16
Currentratio
10
Inventoryratio
13
Leverage
14
11 15
Loss
Merger
Opinion (4a)
Foreign
.
3 7
Total fees
0 5 8
SOX
SOX * Auditor
Auditor
Herfindahl
Assets
Currentassets
9 12 16
Currentratio
10
Inventoryratio
13
Leverage
14
11 15
Loss
Merger
Opinion (4b)
Foreign
Because changes in fees around SOX could reect changes in sample composition, we construct a restricted sample with relevant data for all the years from 2000 through 2005. Changes Specications As in Ghosh and Lustgarten (2006), we analyze audit fees for new engagements using a changes specication rather than a levels specication because Switch is inherently a change variable. Another advantage of using a changes specication is that it mitigates biases related to omitted correlated variables. Assuming that the omitted correlated variables are rm-specic and constant over time, using a changes specication reduces some of the bias problem in cross-sectional regressions (Ghosh and Lustgarten 2006). We estimate the following change in audit fees model to examine fee discounting.
0 5 8 11 14 16
Switch
7
Herfindahl Leverage
Assets
Currentratio
10 13
Inventoryratio
Loss-to-Noloss
15
Noloss-to-Loss (5)
Nomerger-to-Merger
17
Merger-to-Nomerger
Unqualified
where Switch is an indicator variable that equals 1 if the external audit rm for the current year is not the same as the prior year, 0 otherwise; Loss-to-Noloss (Noloss-to-Loss) is an indicator variable that equals 1 for rms with negative income before extraordinary items for the prior (current) year but not for the current (prior) scal year, 0 otherwise; Nomergerto-Merger (Merger-to-Nomerger) is an indicator variable that equals 1 if the Acquisition/ Merger Pretax number is nonzero (zero) for the current scal year but not for the prior (current) year, 0 otherwise; Qualied (Unqualied) is an indicator that equals 1 if the auditor issues a qualied (unqualied) opinion for the current year but not for the prior year, 0 otherwise; and represents a change operator dened (difference between the current year and the prior year). All other variables are as previously dened. A negative coefcient on Switch ( 3) is consistent with fee discounting for the preSOX years. A positive (negative) coefcient on SOX * Switch ( 2) indicates whether there is a decline (increase) in fee discounting for the post-SOX period compared with the preSOX period. Because Ghosh and Lustgarten (2006) nd that fee discounting varies according to Auditor, we analyze whether fee discounting associated with initial audit engagements changed following SOX by partitioning the auditor switch sample into two broad groups: (1) when the incoming audit rm is small and (2) when the incoming audit rm is large.11 Therefore, we modify Equation (5) as follows. Audit fees
0 4 7 10 13 16 18 1
SOX
SOX * SwitchAtomistic
5
SwitchAtomistic Assets
8
SwitchOligopolistic
9
Currentassets
11 14
Currentratio
12 15
Inventoryratio
Loss-to-Noloss
17
Noloss-to-Loss (6)
Nomerger-to-Merger
.
Unqualified
A negative coefcient on SwitchAtomistic ( 4) is consistent with fee discounting in the atomistic market (i.e., when the incoming auditor is small) for the pre-SOX years. Similarly, a negative coefcient on SwitchOligopolistic ( 5) indicates fee discounting in the oligopolistic market (i.e., when the incoming auditor is large) for the pre-SOX years. A positive (negative) coefcient on SOX * Switch (for 2 or 3) indicates whether there is a decline
11
Ghosh and Lustgarten (2006) partition auditor switches into four categories: (1) prior auditor was large but the current auditor is small, (2) prior auditor was small but the current auditor is large, (3) both the prior and current auditors are large, and (4) both the prior and current auditors are small. Because we have a small number of observations in two of these categories, we focus on the incoming auditor.
November 2009
181
(increase) in fee discounting in the post-SOX period compared with those in the pre-SOX period for the different market segments. DATA AND STATISTICS Sample Selection We obtain fee information for 2000 through 2005 from the Audit Analytics database, which compiles audit fee information based on proxy statements led with the SEC. We also obtain the name of the audit rm and the audit opinion (Opinion) from Audit Analytics. Financial data are obtained from the Compustat Annual and Segments les. The full sample consists of 23,273 rm-year observations from 2000 through 2005 with the necessary data. For additional tests, we compile two separate subsamples. In the constant subsample, we require a rm to have data for all six years to be included; in the segments subsample, we require rms to have segment data. The constant subsample includes 11,763 rm-year observations; the segments subsample includes 11,257 rm-year observations. Because our analysis also includes auditor switching, which requires two consecutive years of audit fee data, we create a third subsample to analyze audit fees for rms switching auditors after deleting the initial observation for each company. The subsample analyzing auditor switching or audit fee changes consists of 17,602 observations. We winsorize all continuous variables at the 1 and 99 percent levels to reduce the inuence of extreme observations. Descriptive Statistics We report the descriptive statistics in Table 1. For the full sample (Panel A), the average (median) client rm pays just under $1,765,138 ($550,000) in Total fees to the audit rm. The mean (median) Audit fees are $1,007,155 ($321,200), and Nonaudit fees are $832,125 ($143,000). The average (median) rm in our sample operates in an industry where the audit rms have Herndahl of 0.23 (0.22); owns $2.89 billion ($248 million) in assets; and holds Currentasset of 0.49 (0.49), Currentratio of 2.81 (1.83), Receivableratio of 0.15 (0.12), Inventoryratio of 0.10 (0.05), and Leverage of 0.26 (0.19). The average (median) Protability is 0.14 (0.02), with approximately 40 percent of the observations showing a loss for the scal year. The median rm is U.S.-based, did not engage in merger activity for the scal year, did not switch auditors from the previous scal year, employs a Big 4 (or Big 5) auditor and received an unqualied audit opinion. In our constant sample (Table 1, Panel B), we nd that the average (median) client rm pays just under $1,968,020 ($645,000) in Total fees to the audit rm. The mean (median) Audit fees are $1,091,414 ($362,000) and Nonaudit fees are $985,704 ($177,500). The average (median) rm in our constant subsample operates in an industry where the audit rms have Herndahl of 0.23 (0.22); owns $2.87 billion ($296 million) in assets; and holds Currentasset of 0.49 (0.48), Currentratio of 2.86 (1.88), Receivableratio of 0.15 (0.13), Inventoryratio of 0.10 (0.05), and Leverage of 0.25 (0.20). The average (mean) Protability is 0.11 (0.02), with approximately 39 percent of the observations showing a loss for the scal year. Again, the median rm is U.S.-based, did not engage in merger activity for the scal year, did not switch auditors from the previous scal year, employs a Big 4 (or Big 5) auditor and received an unqualied audit opinion. The results reported in Panel C for the subsample with segment data are very similar to those in Panels A and B. Thus, the summary statistics do not indicate substantial differences among the three subsamples.
Auditing: A Journal of Practice & Theory November 2009 American Accounting Association
182
TABLE 1 Summary Statistics Panel A: Full Samplea Variable Fees Audit fees Nonaudit fees Total fees Firm Characteristics Auditor Herndahl Assets (millions) Currentassets Currentratio Receivableratio Inventoryratio Leverage Protability Loss Merger Opinion Observations Q1 $127,000 $38,550 $199,810 1.0000 0.1961 $46.506 0.2863 1.1482 0.0575 0.0033 0.0192 0.0883 0.0000 0.0000 0.0000 Median $321,200 $143,000 $550,000 1.0000 0.2217 $248.06 0.4910 1.8321 0.1236 0.0540 0.1916 0.0216 0.0000 0.0000 0.0000 23,273 Mean $1,007,155 $832,125 $1,765,138 0.7974 0.2315 $2,890.3 0.4940 2.8066 0.1492 0.1029 0.2605 0.1431 0.3960 0.0379 0.0007 Q3 $915,000 $517,000 $1,531,000 1.0000 0.2589 $1,203.5 0.6976 3.1590 0.2043 0.1594 0.3694 0.0647 1.0000 0.0000 0.0000
Panel B: Constant Subsamplea Variable Fees Audit fees Nonaudit fees Total fees Firm Characteristics Auditor Herndahl Assets (millions) Currentassets Currentratio Receivableratio Inventoryratio Leverage Protability Loss Merger Opinion Observations Q1 $151,932 $52,400 $247,400 1.0000 0.1904 $67.474 0.2837 1.1973 0.0626 0.0035 0.0231 0.0818 0.0000 0.0000 0.0000 Median $362,000 $177,500 $645,000 1.0000 0.2199 $296.38 0.4785 1.8781 0.1253 0.0491 0.1966 0.0219 0.0000 0.0000 0.0000 11,766 Mean $1,091,414 $985,704 $1,968,020 0.8466 0.2266 $2,873.6 0.4881 2.8615 0.1491 0.0963 0.2478 0.1088 0.3899 0.0378 0.0004 Q3 $1,023,280 $611,000 $1,757,000 1.0000 0.2544 $1,336.0 0.6904 3.2693 0.2020 0.1493 0.3636 0.0634 1.0000 0.0000 0.0000
Panel C: Segments Subsamplea Variable Fees Audit fees Nonaudit fees Total fees Q1 $205,436 $75,766 $346,000 Median $526,000 $275,000 $938,200 Mean $1,462,664 $1,278,614 $2,609,435 Q3 $1,512,000 $916,000 $2,619,000
November 2009
The Impact of Regulation on Auditor Fees TABLE 1 (continued) Variable Firm Characteristics Auditor Herndahl Assets (millions) Currentassets Currentratio Receivableratio Inventoryratio Leverage Protability Loss Merger Opinion Segments Foreign Observations
a
183
Q1 1.0000 0.2008 $98.986 0.3425 1.3172 0.0839 0.0127 0.0168 0.0475 0.0000 0.0000 0.0000 1.0000 0.1784
Median 1.0000 0.2217 $464.15 0.5073 1.9675 0.1438 0.0834 0.1766 0.0278 0.0000 0.0000 0.0000 3.0000 0.4474 11,257
Mean 0.8769 0.2316 $4,089.3 0.5157 2.8073 0.1613 0.1098 0.2233 0.0625 0.3580 0.0452 0.0003 3.3468 0.4993
Q3 1.0000 0.2575 $2,148.2 0.6912 3.2213 0.2118 0.1668 0.3318 0.0706 1.0000 0.0000 0.0000 5.0000 0.8811
Audit fees are fees paid to the external audit rm for the scal years nancial statement audit, Nonaudit fees are fees paid to the external audit rm for services not related to the audit, and Total fees are the amount of fees paid to the external audit rm for the scal year. Audit, Nonaudit and Total fees are obtained from Audit Analytics. Auditor is an indicator variable that equals 1 when the external auditor is one of the Big 4 or Big 5 auditing rms, 0 otherwise. Herndahl is the normalized Herndahl Index within each two-digit standard industry classication (SIC) code. Assets is total assets (in millions). Currentassets is the ratio of current assets to total assets. Currentratio is the ratio of current assets to current liabilities. Receivableratio is the ratio of total receivables to total assets. Inventoryratio is the ratio of inventory to total assets. Protability is income before extraordinary items divided by total assets. Leverage is the ratio of total debt to total assets. Loss is an indicator variable that equals 1 for rms with negative income before extraordinary items, 0 otherwise. Merger is an indicator variable that equals 1 if the Acquisition / Merger Pretax number is non-zero, 0 otherwise. Opinion is an indicator variable that equals 1 if the audit opinion is not listed as unqualied, 0 otherwise. Segments is the number of business segments reported. Foreign is the ratio of sales revenue generated outside of the United States to total reported segment revenue. All variables are measured as of the scal year-end. The full sample (Panel A) consists of 23,273 observations, the constant subsample (Panel B) consists of 11,766 observations, and the segments subsample (Panel C) consists of 11,257 observations over the years 2000 to 2005.
RESULTS Univariate Tests Figure 1 presents the trends in average fees for the years 2000 through 2005. The graph illustrates that audit fees increased monotonically over the sample period. While the increase in audit fees between 2000 and 2001 is small, there is a substantial rise in audit fees in each of the subsequent years.12 The monotonic increase in audit fees suggests that the costs of implementing the various requirements of the Sarbanes-Oxley Act were incurred over multiple years. Costs related to many sections of the act, including those associated with audit workpaper retention and review (Section 103), preparing audit workpapers for PCAOB review (Section 101), audit partner rotation (Section 203), and audit committee preapproval and communication (Sections 202 and 204), are ongoing costs likely to increase audit fees every year after SOX. Also, new reporting requirements under Section 401,
12
Studies examining the relationship between audit fees and SOX may suffer from biases because of changes in audit fee denitions from 2003 (SEC 2003). However, we believe that such a change is unlikely to bias our results. The mean increase in audit fees from 2001 to 2002 and from 2002 to 2003, which coincides with changes in the audit fee denition, is very modest. The largest audit fee increases for our sample occurs between 2003 and 2005, after the 2003 rule had already taken effect.
184
$2,000,000
Fees Paid
$1,500,000
$500,000
$0 2000 (n=2,202) 2001 (n=3,409) 2002 (n=4,258) 2003 (n=4,773) 2004 (n=4,799) 2005 (n=3,832)
Fiscal Year
including off-balance-sheet liability reporting under FIN45 (effective December 31, 2002) and special-purpose entity reporting under FIN46R (effective December 31, 2003), demand greater audit effort from their effective dates forward. Finally, Section 404 became effective on November 30, 2004, for large accelerated lers. Firms and their auditors were incurring costs up to and beyond that date preparing for the internal control documentation and testing requirements. In contrast, nonaudit fees declined monotonically over the same period. The decline in nonaudit fees is steeper for the years after 2001, which is consistent with the regulatory restrictions imposed on audit rms providing nonaudit services.13 Total fees (the sum of audit and nonaudit fees) declined from 2000 to 2002, with a slight increase in 2003, but large gains subsequent to 2003 (the post-SOX period). In Table 2 we report the magnitude of the levels and changes in audit fees around SOX. For the full sample (Panel A), average Audit fees increased from $533,360 for the pre-SOX period to $1,185,322 for the post-SOX period. The increase in Audit fees of $651,962 (122 percent) over the two periods is statistically signicant. In contrast, Nonaudit fees decreased from $1,259,634 to $671,364 for the same periods. The decline in Non13
Auditors are forbidden from providing the following services to their audit clients: bookkeeping or nancial statement preparation; nancial information system design and implementation; appraisal or valuation services, fairness opinions, or contribution in-kind reports; actuarial services; management or human resources functions; broker / dealer, investment advisor, or investment banking services; and legal services or expert services related to the audit.
November 2009
185
TABLE 2 Audit Fees around the Sarbanes-Oxley Act Panel A: Full Sample Overall Mean Audit fees Nonaudit fees Total fees Observations $1,007,155 $832,125 $1,765,138 23,273 Pre-SOX Mean $533,360 $1,259,634 $1,595,982 6,360 Post-SOX Mean $1,185,322 $671,364 $1,828,747 16,913 Difference $651,962** $588,270** $232,765**
Panel B: Big 4 Auditors Overall Mean Audit fees Nonaudit fees Total fees Observations $1,216,517 $1,025,730 $2,149,264 18,559 Pre-SOX Mean $596,081 $1,438,447 $1,807,372 5,516 Post-SOX Mean $1,478,905 $851,189 $2,293,853 13,043 Difference $882,824** $587,258** $486,480**
Panel C: Non-Big 4 Auditors Overall Mean Audit fees Nonaudit fees Total fees Observations $182,896 $69,904 $252,834 4,714 Pre-SOX Mean $123,443 $90,994 $214,432 844 Post-SOX Mean $195,862 $65,305 $261,209 3,870 Difference $72,419** $25,689* $46,777*
*, ** Indicate signicance at the 0.05 percent and 0.01 percent levels, respectively. Audit fees are fees paid to the external audit rm for the scal years nancial statement audit, Nonaudit fees are fees paid to the external audit rm for services not related to the audit, and Total fees are the amount of fees paid to the external audit rm for the scal year. Audit, Nonaudit, and Total fees are obtained from Audit Analytics.
audit fees of $588,270 ( 47 percent) between the two periods is also signicant. Total fees increased from $1,595,982 to $1,828,747 over the two periods, which is a 15 percent increase. For rms using large auditors (Table 2, Panel B), these increases are even more dramatic. We nd that average Audit fees increased from $596,081 over the pre-SOX period to $1,478,905 over the post-SOX period, an increase of 148 percent. In contrast, Nonaudit fees decreased from $1,438,447 to $851,189, a decline of 41 percent. Total fees increased from $1,807,372 to $2,293,853 around SOX, which translates to a 27 percent increase. For rms using small auditors, the increases in Audit fees are much less pronounced. In Table 2, Panel C we nd that average Audit fees increased from $123,443 over the preSOX period to $195,862 over the post-SOX period. Thus, Audit fees increased by 59 percent between the two periods. In contrast, Nonaudit fees decreased from $90,994 to $65,305, which translates to a 28 percent decline between the two periods. Total fees increased from $214,432 to $261,209 (22 percent) over the two periods. Overall, the univariate results indicate that Audit fees increased while Nonaudit fees decreased following SOX for both large and small auditors.
Auditing: A Journal of Practice & Theory November 2009 American Accounting Association
186
Multivariate Tests Levels Specications We examine whether the increase in audit fees subsequent to SOX remains signicant after controlling for other factors that also inuence audit fees, including rm size, auditor size, and audit complexity. Regression 1 results of Table 3 are consistent with those of prior studies. The parameter estimates for Assets, Currentassets, Receivableratio, Loss, and Merger are positive and signicant at the 5 percent level, while the estimates for Currentratio, Inventoryratio, and Protability are negative and signicant. Similar to prior studies, the adjusted R2 is high (above 75 percent). More importantly, the parameter estimate of 54.95), implying that audit fees are approximately 74 percent SOX ( 1) is 0.55 (t-stat
TABLE 3 Audit Fees and SOX Independent Variables Intercept SOX SOX * Auditor Control Variables Auditor Herndahl Assets Currentassets Currentratio Receivableratio Inventoryratio Leverage Protability Loss Merger Opinion Segments Foreign Adjusted R2 (%) Observations Coefcient (t-stat) Regression 1 8.86 (317.88)** 0.55 (54.95)** Coefcient (t-stat) Regression 2 9.11 (280.04)** 0.25 (10.63)** 0.35 (13.50)** 0.07 0.12 0.54 0.80 0.05 0.70 0.30 0.02 0.20 0.17 0.16 0.18 ( 3.02)** ( 2.00)* (179.34)** (27.34)** ( 26.18)** (14.58)** ( 7.46)** (1.23) ( 21.72)** (15.43)** (6.77)** ( 1.13) Coefcient (t-stat) Regression 3 8.98 (143.92)** 0.35 (6.89)** 0.25 (4.86)** 0.01 ( 0.12) 0.02 ( 0.18) 0.54 (120.38)** 0.80 (17.08)** 0.06 ( 17.19)** 0.66 (8.18)** 0.25 ( 3.91)** 0.03 (1.12) 0.29 ( 12.03)** 0.11 (6.89)** 0.11 (3.76)** 0.79 ( 2.20) 0.04 (15.22)** 0.05 (2.69)** 75.28 11,257
0.21 0.12 0.54 0.79 0.05 0.71 0.28 0.02 0.20 0.17 0.16 0.20
(14.45)** ( 1.87) (180.17)** (27.00)** ( 26.04)** (14.66)** ( 7.11)** (1.02) ( 21.20)** (15.31)** (6.61)** ( 1.22)
75.43 23,273
75.58 23,273
*, ** Indicate signicance at the 0.05 percent and 0.01 percent levels, respectively. The dependent variable is Audit Fees. Audit fees is the natural logarithm of fees paid to the external audit rm for the scal years nancial statement audit. The explanatory variables are dened as follows. SOX is an indicator variable that equals 1 if the scal year-ends after July 30, 2002, 0 otherwise. Auditor is an indicator variable that equals 1 when the external auditor is one of the Big 4 or Big 5 auditing rms, 0 otherwise. Herndahl is the normalized Herndahl Index within each two-digit standard industry classication (SIC) code. Assets is the natural logarithm of total assets. Currentassets is the ratio of current assets to total assets. Currentratio is the ratio of current assets to current liabilities. Receivableratio is the ratio of total receivables to total assets. Inventoryratio is the ratio of inventory to total assets. Protability is income before extraordinary items divided by total assets. Leverage is the ratio of total debt to total assets. Loss is an indicator variable that equals 1 for rms with negative income before extraordinary items, 0 otherwise. Merger is an indicator variable that equals 1 if the Acquisition / Merger Pretax number is non-zero, 0 otherwise. Opinion is an indicator variable that equals 1 if the audit opinion is not listed as unqualied, 0 otherwise. Segments is the number of business segments reported. Foreign is the ratio of sales revenue generated outside of the United States to total reported segment revenue. All variables are measured as of the scal year-end. t-statistics are calculated using White (1980)-corrected standard errors. The full sample consists of 23,273 observations and the segments subsample consists of 11,257 observations over the years 2000 to 2005.
November 2009
187
higher for the average rm after SOX.14 Thus, our results are consistent with the hypothesis that audit fees increased subsequent to the Sarbanes-Oxley Act. In Regression 2, we interact SOX with Auditor. The coefcient on SOX indicates the magnitude of the increase in audit fees over the post-SOX period for small auditors. The coefcient on SOX * Auditor ( 2) indicates the incremental increase in audit fees for large audit rms over the same period. A positive coefcient suggests that the fee increases for large auditors over the post-SOX period exceeded those of small audit rms. As before, the coefcient on SOX in Regression 2 continues to be positive and signif0.25, t-stat 10.63), indicating that even small auditors charged higher fees icant ( 1 after SOX (fees were 28 percent higher). Further, the increase in audit fees for the postSOX period is signicantly higher for large audit rms. The coefcient on SOX * Auditor 0.35, t-stat 13.50). The coefcient estimates indicate is positive and signicant ( 2 that, after controlling for other factors (including the market concentration for audit services), large audit rms on average charge 42 percent higher audit fees than small audit rms in the post-SOX period. Our results indicate that large auditors are more likely to be affected by the SOX provisions than small auditors. In Regression 3, we include additional controls for rm complexity which are expected to be positively related to Audit fees. The results using the segment data are similar to those 0.35, t-stat 6.89) and SOX from the full sample. Both the coefcients on SOX ( 1 0.25, t-stat 4.86) remain positive and signicant. By truncating the * Auditor ( 2 sample and including additional controls for rm complexity, the variables for Auditor and Herndahl lose their signicance. The coefcients for the additional complexity measures are positive and signicant, as expected. In Table 4 we present the results for Total fees and Nonaudit fees for both the full sample and the subsample including. The results for Nonaudit fees provide consistent evidence of a signicant decrease in nonaudit fees for the post-SOX period for both small and large audit rms.15 These results indicate that while, on average, audit rms lost nonaudit services from their audit clients, large (Big 4) auditors lost signicantly larger amounts. The coefcient for SOX is positive and signicant in the regressions when we use Total fees as the dependent variable. For both the full sample ( 1 0.15, t-stat 6.47) and the segments data subsample ( 1 0.22, t-stat 4.40), only the SOX variable is positive and signicant at the 0.05 level. The results from these regressions indicate that total fees collected are greater for large auditors (Auditor, 3) and that total fees increased signicantly after SOX (SOX, 1). The coefcients on SOX * Auditor suggest that the gap in total fees between large and small auditors did not widen after SOX. We also report the results using a restricted constant sample in Table 5 to mitigate the concern that a shift in sample composition might drive our results. Consistent with our prior results, both for the full sample and for the segments sample, we nd that (1) the coefcients on SOX are positive and signicant for Audit fees and Total fees regressions, but negative for the Nonaudit fees regression, and (2) the coefcients on SOX * Auditor
14
15
For a regression where the dependent variable is the logarithmic transformation, the coefcient on a binary explanatory variable (such as SOX) needs to be transformed as e 1 to calculate a percentage change in the base dependent variables (audit fees) when the binary variable changes from 0 to 1 (see Kennedy 1992, 223). All regression t-statistics are calculated using White (1980) heteroscedasticity-corrected standard errors. In the regression using the full sample, the coefcients on SOX and SOX * Auditor type are both negative and statistically signicant ( 1 0.19, t-stat 4.46; 2 0.39, t-stat 8.44). For the regression based on the sample with segments data, both coefcients of interest are again negative and statistically signicant at the 0.01 level ( 1 0.19, t-stat 2.08; 2 0.42, t-stat 4.50).
188
TABLE 4 Nonaudit Fees, Total Fees, and SOX Dependent Variable Independent Variables Intercept SOX SOX * Auditor type Control Variables Auditor Herndahl Assets Currentassets Currentratio Receivableratio Inventoryratio Leverage Protability Loss Merger Opinion Segments Foreign Adjusted R2 (%) Observations Nonaudit Fees Coefcient (t-stat) 7.59 (132.70)** 0.19 ( 4.46)** 0.39 ( 8.44)** 0.63 0.07 0.66 1.01 0.04 0.75 0.53 0.05 0.24 0.13 0.39 0.31 (14.68)** ( 0.65) (138.27)** (20.96)** ( 13.72)** (9.22)** ( 7.68)** (1.83) ( 15.45)** (7.35)** (10.32)** (1.94) 7.60 (70.72)** 0.19 ( 2.08)** 0.42 ( 4.50)** 0.82 0.04 0.64 0.91 0.05 0.41 0.50 0.18 0.26 0.07 0.34 0.53 0.04 0.11 (9.38)** ( 0.21) (90.61)** (12.20)** ( 10.08)** (3.16)** ( 4.57)** (3.79)** ( 8.15)** (2.52)* (7.34)** ( 0.47) (9.84)** (3.55)** 62.71 11,257 Total Fees Coefcient (t-stat) 9.36 (296.29)** 0.15 (6.47)** 0.01 (0.41) 0.22 0.10 0.58 0.88 0.05 0.64 0.44 0.04 0.22 0.16 0.24 0.02 (9.29)** ( 1.75) (207.47)** (31.77)** ( 26.11)** (14.06)** ( 11.53)** (2.51)* ( 25.19)** (15.25)** (11.35)** (0.07) 9.30 (157.08)** 0.22 (4.40)** 0.08 ( 1.55) 0.34 0.02 0.57 0.86 0.05 0.53 0.42 0.08 0.25 0.10 0.21 0.66 0.04 0.06 (6.90) (0.19) (139.05)** (19.13)** ( 17.53)** (6.97)** ( 7.02)** (2.88)* ( 12.54)** (6.49)** (7.63)** ( 1.14) (14.08)** (3.05)** 78.38 11,257
61.97 23,273
78.61 23,273
*, ** Indicate signicance at the 0.05 percent and 0.01 percent levels, respectively. The dependent variables are Nonaudit Fees and Total Fees. Nonaudit fees is the natural logarithm of fees paid to the external audit rm for services not related to the audit. Total fees is the natural logarithm of total amount of fees paid to the external audit rm for the scal year. The explanatory variables are dened as follows. SOX is an indicator variable that equals 1 if the scal year-ends after July 30, 2002, 0 otherwise. Auditor is an indicator variable that equals 1 when the external auditor is one of the Big 4 or Big 5 auditing rms, 0 otherwise. Auditor is an indicator variable that equals 1 when the external auditor is one of the Big 4 or Big 5 auditing rms, 0 otherwise. Herndahl is the normalized Herndahl Index within each two-digit standard industry classication (SIC) code. Assets is the natural logarithm of total assets. Currentassets is the ratio of current assets to total assets. Currentratio is the ratio of current assets to current liabilities. Receivableratio is the ratio of total receivables to total assets. Inventoryratio is the ratio of inventory to total assets. Protability is income before extraordinary items divided by total assets. Leverage is the ratio of total debt to total assets. Loss is an indicator variable that equals 1 for rms with negative income before extraordinary items, 0 otherwise. Merger is an indicator variable that equals 1 if the Acquisition / Merger Pretax number is non-zero, 0 otherwise. Opinion is an indicator variable that equals 1 if the audit opinion is not listed as unqualied, 0 otherwise. Segments is the number of business segments reported. Foreign is the ratio of sales revenue generated outside of the United States to total reported segment revenue. All variables are measured as of the scal year-end. t-statistics are calculated using White (1980)-corrected standard errors. The full sample consists of 23,273 observations and the segments subsample consists of 11,257 observations over the years 2000 to 2005.
are positive (negative) for the Audit fees (Nonaudit fees) regressions, but insignicant for the Total fees regression. Thus, our results using the restricted sample are very similar to those using the full sample. Changes Specications Similar to Ghosh and Lustgarten (2006), we employ a changes specication to investigate whether auditors on initial engagements discount audit fees. We report the regression results of changes in audit fees around SOX for rms switching auditors in Table 6. We
Auditing: A Journal of Practice & Theory American Accounting Association November 2009
189
TABLE 5 Audit Fees, Nonaudit Fees, Total Fees, and SOX: Constant Subsample Dependent Variable Audit Fees Independent Variables Intercept SOX SOX * Auditor type Control Variables Auditor Herndahl Assets Currentassets Currentratio Receivableratio Inventoryratio Leverage Protability Loss Merger Opinion Adjusted R2 (%) Observations Coefcient (t-stat) 9.08 (208.87)** 0.39 (11.40)** 0.32 (8.68)** 0.09 0.10 0.55 0.74 0.05 0.74 0.04 0.05 0.26 0.13 0.19 0.34 ( 2.78)** ( 1.28) (130.96)** (18.14)** ( 19.88)** (10.84)** ( 0.73)** ( 1.84) ( 16.65)** (8.57)** (5.83)** (1.64) 75.80 11,763 Nonaudit Fees Coefcient (t-stat) 7.49 (95.84)** 0.12 ( 1.89)** 0.50 ( 7.58)** 0.58 0.14 0.69 0.92 0.04 0.79 0.33 0.02 0.35 0.07 0.34 1.26 (9.92)** (0.95) (102.24)** (13.55)** ( 8.88)** (6.85)** ( 3.29)** ( 0.49) ( 14.54)** (2.64)** (6.41)** (3.44)* 60.84 11,763 Total Fees Coefcient (t-stat) 9.29 (215.39)** 0.26 (6.20)** 0.04 ( 0.97) 0.19 0.04 0.60 0.83 0.05 0.69 0.20 0.05 0.28 0.12 0.23 0.90 (5.81)** ( 0.49) (155.41)** (20.80)** ( 18.51)** (10.33)** ( 3.54)** ( 1.99)* ( 18.91)** (8.49)** (7.82)** (3.08)** 78.30 11,763
*, ** Indicate signicance at the 0.05 percent and 0.01 percent levels, respectively. The dependent variables are Audit Fees, Nonaudit Fees, or Total Fees. Audit fees is the natural logarithm of fees paid to the external audit rm for the scal years nancial statement audit. Nonaudit fees is the natural logarithm of fees paid to the external audit rm for services not related to the audit. Total fees is the natural logarithm of total amount of fees paid to the external audit rm for the scal year. Audit, Nonaudit and Total fees are obtained from Audit Analytics. Auditor is an indicator variable that equals 1 when the external auditor is one of the Big 4 or Big 5 auditing rms, 0 otherwise. Herndahl is the normalized Herndahl Index within each two-digit standard industry classication (SIC) code. Assets is the natural logarithm of total assets. Currentassets is the ratio of current assets to total assets. Currentratio is the ratio of current assets to current liabilities. Receivableratio is the ratio of total receivables to total assets. Inventoryratio is the ratio of inventory to total assets. Protability is income before extraordinary items divided by total assets. Leverage is the ratio of total debt to total assets. Loss is an indicator variable that equals 1 for rms with negative income before extraordinary items, 0 otherwise. Merger is an indicator variable that equals 1 if the Acquisition / Merger Pretax number is non-zero, 0 otherwise. Opinion is an indicator variable that equals 1 if the audit opinion is not listed as unqualied, 0 otherwise. All variables are measured as of the scal year-end. t-statistics are calculated using White (1980)-corrected standard errors. The constant subsample consists of 11,766 observations for the companies that remained in our sample every year from 2000 to 2005.
report the results without including the segment data to maximize the number of observations. In Regression 1, SOX ( 1 0.18, t-stat 21.50) continues to be positive and signicant at the 1 percent level. Our results suggest that controlling for changes in client characteristics, audit fees for continuing engagements increase an additional 18 percent in the post0.12, t-stat SOX period. The coefcient for Switch is negative and signicant ( 3 7.13), indicating fee discounting of approximately 12 percent for companies that switched auditors. These results are similar to those reported in Ghosh and Lustgarten (2006). In Regression 2, we investigate whether the extent of lowballing changed following 0.21, t-stat 5.16), but the SOX. The coefcient on Switch ( 3) is negative ( 3
Auditing: A Journal of Practice & Theory November 2009 American Accounting Association
190
TABLE 6 Audit Fee Changes and SOX Independent Variables Intercept SOX SOX * Switch Control Variables Switch Herndahl Assets Currentassest Currentratio Receivableratio Inventoryratio Leverage Protability Loss-to-Noloss Noloss-to-Loss Nomerger-to-Merger Merger-to-NoMerger Qualied Unqualied Adjusted R2 (%) Observations Coefcient (t-stat) Regression 1 0.09 (12.81)** 0.18 (21.50)** Coefcient (t-stat) Regression 2 0.09 (13.56)** 0.18 (20.70)** 0.10 (2.71)** 0.21 0.01 0.33 0.09 0.00 0.09 0.25 0.03 0.10 0.03 0.01 0.05 0.00 0.01 0.01 ( 5.16)** ( 0.23) (22.60)** ( 1.90)** ( 1.46)* (1.19) (2.51)** (1.05) ( 9.68)** ( 2.27)* (0.50) (2.20)* ( 0.03) (0.09) (0.11) 8.31 17,602
0.12 0.01 0.33 0.09 0.00 0.09 0.25 0.03 0.10 0.03 0.01 0.05 0.00 0.01 0.01
( 7.13)** ( 0.21) (22.57)** ( 1.90)* ( 1.45) (1.16) (2.50)** (1.06) ( 9.67)** ( 2.26)* (0.52) (2.21)* ( 0.03) (0.10) (0.12) 8.29 17,602
*, ** Indicate signicance at the 0.05 percent and 0.01 percent levels, respectively. The dependent variable is the percentage change in Audit Fees between the current year and the prior year. Audit fees is obtained from Audit Analytics. The explanatory variables are dened as follows. SOX is an indicator variable that equals 1 when the scal year-end date is after July 30, 2002, 0 otherwise. SOX*Switch is an interaction of the SOX and Switch variables. Switch is an indicator variable that equals 1 if the external audit rm for the current year is not the same as the prior year, 0 otherwise. Herndahl is the normalized Herndahl Index within each two-digit standard industry classication (SIC) code. Assets is the natural logarithm of total assets. Currentassets is the ratio of current assets to total assets. Currentratio is the ratio of current assets to current liabilities. Receivableratio is the ratio of total receivables to total assets. Inventoryratio is the ratio of inventory to total assets. Protability is income before extraordinary items divided by total assets. Leverage is the ratio of total debt to total assets. Loss-to-Noloss is an indicator variable that equals 1 for rms with negative (positive) income before extraordinary items for the prior (current) scal year, 0 otherwise. Noloss-to-Loss is an indicator variable that equals 1 for rms with positive (negative) income before extraordinary items for the prior (current) scal year, 0 otherwise. Nomerger-to-Merger is an indicator variable that equals 1 if the Acquisition / Merger Pretax number is zero (non-zero) for the prior (current) scal year, 0 otherwise. Merger-to-NoMerger is an indicator variable that equals 1 if the Acquisition / Merger Pretax number is non-zero (zero) for the prior (current) scal year, 0 otherwise. Qualied (Unqualied ) is an indicator that equals 1 if the auditor issues a qualied (unqualied) opinion for the current year but not for the prior year, 0 otherwise. is a change operator dened as the difference between the number for the current year and those of the prior year. t-statistics are calculated using White (1980)-corrected standard errors. The sample consists of 17,602 observations over the years 2001 to 2005.
coefcient on SOX * Switch ( 2) is positive ( 2 0.10, t-stat 2.71). Our results suggest that while incoming auditors were discounting audit fees by 22 percent for the pre-SOX period relative to what the predecessor auditors were charging, the number declined to 11 0.21 0.10) for the post-SOX period. The coefcient on SOX ( 1), which percent ( captures the change in fees for continuing audit engagements after the act, continues to be 0.18, t-stat 20.70). Thus, controlling for other factors, positive and signicant ( 1
Auditing: A Journal of Practice & Theory American Accounting Association November 2009
191
auditors increased audit fees by an additional 18 percent per year for continuing engagements after SOX. The results in Table 7 provide further insights on the impact of SOX on fee discounting within the two market segments (large and small audit rms). In Regression 1, only the
TABLE 7 Audit Fee Changes, Fee Discounting, and SOX Independent Variables Intercept SOX SOX * SwitchAtomistic SOX * SwitchOligopolistic Control Variables SwitchAtomistic SwitchOligopolistic Herndahl Assets Currentassets Currentratio Receivableratio Inventoryratio Leverage Protability Loss-to-Noloss Noloss-to-Loss Nomerger-to-Merger Merger-to-NoMerger Qualied Unqualied Adjusted R2 (%) Observations Coefcient (t-stat) Regression 1 0.09 (12.43)** 0.18 (21.82)** Coefcient (t-stat) Regression 2 0.10 0.17 0.01 0.23 0.29 0.19 0.05 0.33 0.08 0.00 0.09 0.24 0.03 0.10 0.03 0.01 0.05 0.00 0.01 0.02 (13.62)** (20.64)** (0.17) (4.10)**
0.28 0.01 0.05 0.33 0.08 0.00 0.09 0.24 0.04 0.10 0.03 0.01 0.05 0.00 0.01 0.03
( 11.40)** (0.53) ( 0.77) (22.43)** ( 1.80)* ( 1.49)* (1.17) (2.48)** (1.44) ( 9.59)** ( 2.08)* (0.69) (2.12)* ( 0.26) (0.10) (0.23) 9.12 17,602
( 3.67)** (3.86)** ( 0.85) (22.47)** ( 1.80)* ( 1.48)* (1.19) (2.50)** (1.41) ( 9.60)** ( 2.08)* (0.69) (2.11)* ( 0.26) (0.10) (0.22) 9.20 17,602
*, ** Indicate signicance at the 0.05 percent and 0.01 percent levels, respectively. The dependent variable is the percentage change in Audit Fees between the current year and the prior year. Audit fees is obtained from Audit Analytics. The explanatory variables are dened as follows. SOX is an indicator variable that equals 1 when the scal year-end date is after July 30, 2002, 0 otherwise. SwitchAtomistic (SwitchOligopolistic) is an indicator variable that equals 1 if an auditor switch is in the atomistic (oligopolistic) sector, and 0 otherwise. Auditor is an indicator variable that equals 1 when the external auditor is one of the Big 4 or Big 5 auditing rms, 0 otherwise. Herndahl is the normalized Herndahl Index within each two-digit standard industry classication (SIC) code. Assets is the natural logarithm of total assets. Currentassets is the ratio of current assets to total assets. Currentratio is the ratio of current assets to current liabilities. Receivableratio is the ratio of total receivables to total assets. Inventoryratio is the ratio of inventory to total assets. Protability is income before extraordinary items divided by total assets. Leverage is the ratio of total debt to total assets. Loss-to-Noloss is an indicator variable that equals 1 for rms with negative (positive) income before extraordinary items for the prior (current) scal year, 0 otherwise. Noloss-to-Loss is an indicator variable that equals 1 for rms with positive (negative) income before extraordinary items for the prior (current) scal year, 0 otherwise. Nomerger-to-Merger is an indicator variable that equals 1 if the Acquisition / Merger Pretax number is zero (non-zero) for the prior (current) scal year, 0 otherwise. Merger-to-NoMerger is an indicator variable that equals 1 if the Acquisition / Merger Pretax number is non-zero (zero) for the prior (current) scal year, 0 otherwise. Qualied (Unqualied ) is an indicator that equals 1 if the auditor issues a qualied (unqualied) opinion for the current year but not for the prior year, 0 otherwise. is a change operator dened as the difference between the number for the current year and those of the prior year. t-statistics are calculated using White (1980)-corrected standard errors. The sample consists of 17,602 observations over the years 2001 to 2005.
192
coefcient on SwitchAtomistic is signicant ( 4 0.28, t-stat 11.40), while that on SwitchOligopolistic is insignicant ( 5 0.01, t-stat 0.53). More importantly, the coefcient on SOX continues to be positive and signicant, suggesting that audit fees increased for the post-SOX period. Our results suggest that small audit rms discount audit fees to attract new clients, while large audit rms do not. In Regression 2, we interact SwitchAtomistic and SwitchOligopolistic with SOX to investigate whether fee discounting changed after SOX. We nd that the coefcients on SwitchAtomistic and SwitchOligopolistic are both negative and signicant, which suggests evidence of lowballing by large and small audit rms to attract new clients in the pre-SOX period. When we interact the two indicator variables with SOX, we nd that only SOX * SwitchOligopolistic is signicant ( 3 0.23, t-stat 4.10). Thus, while large audit rms might have discounted audit fees for the pre-SOX period, there is no evidence of lowballing in the post-SOX 0.04). The coefcient on SOX * SwitchAtomistic is insignicant ( 2 0.01, period ( 3 5 0.17), which suggests that small audit rms were discounting audit fees for new t-stat clients during the pre-SOX period and that this phenomenon did not change for the postSOX period. However, the results on the SwitchAtomistic and SwitchOligopolistic (pre-SOX estimates) need to be interpreted with caution because of the small sample size. Because of econometric concerns from estimating a pooled regression where the error term might be serially correlated, we also estimate Regression 1 of Table 7 (without the SOX variable) for each of the years from 2001 to 2005 (untabulated). Consistent with the 4.86) pooled results, we nd that SwitchOligopolistic is only negative and signicant (t-stat in 2001 but insignicant for all of the subsequent years, indicating that large auditors were offering fee discounts in 2001 only. In contrast, the coefcient on SwitchAtomistic is negative and signicant at less than the 1 percent level for all six years, indicating that small auditors were offering fee discounts for the entire sample period (both pre- and post-SOX). Overall, our regression results provide compelling evidence to indicate that audit fees increased following SOX. This increase in fees is independent of the number of control variables included in the regression and does not depend on whether we use a levels or changes specication. Further, we nd strong evidence of fee discounting in the pre-SOX period, but this phenomenon declined after SOX. Finally, most of our results suggest that only small audit rms discount fees to attract new clients while large audit rms do not. Explanations for Audit Fee Increases One possible explanation for the increase in audit fees for the post-SOX years is linked to a stricter legal environment. Related and important questions are whether (1) auditors increased audit fees by the same quantity for all clients for the post-SOX years because of a stringent legal environment and/or (2) the increase in audit fees resulted from a structural shift in the audit fee structure at the rm level (i.e., client attributes including size, complexity, and risk have become relatively more important determinants of audit fees for postSOX years). We test these related conjectures by estimating Regression 1 from Table 3 separately for the pre- and post-SOX years.16 We then compare the estimated coefcients between the two periods. A larger intercept term ( 0) for the post-SOX years, compared with that of the pre-SOX years, indicates that all clients are equally affected by the provisions of SOX and are paying higher audit fees during this period. In contrast, larger slope coefcients ( 1
16
To maximize the number of observations, we estimate the regression using the unrestricted sample. Because there are no observations with a qualied audit opinion for the pre-SOX years, we do not include Opinion in the regression.
November 2009
193
to 11) for the post-SOX years, compared with those of the pre-SOX years, are consistent with the premise that there is a structural change in audit pricing at the rm level following SOX. Panel A of Table 8 presents the audit fee regression results for the pre- and post-SOX years. In the Difference column, we test whether the estimated coefcients are statistically
TABLE 8 Abnormal Audit Fees and SOX Panel A: Estimated Regression Coefcients for Pre-SOX and Post-SOX Years Independent Variables Intercept Auditor Herndahl Assets Currentassets Currentratio Receivableratio Inventoryratio Leverage Protability Loss Merger Adjusted R2 (%) Observations Coefcient (t-stat) Pre-SOX
0 1 2 3 4 5 6 7 8 9 10 11
Coefcient (t-stat) Post-SOX 9.28 0.21 0.21 0.56 0.92 0.05 0.64 0.43 0.03 0.19 0.18 0.02 (262.72)** (12.76)** ( 2.41)* (154.62)** (25.96)** ( 19.97)** (10.84)** ( 8.68)** (1.88) ( 17.79)** (12.94)** (6.59)** 75.07 16,913
Difference (t-stat) 0.03 0.14 0.19 0.07 0.41 0.00 0.23 0.45 0.01 0.04 0.03 0.09 ( 52.87)** (409.46)** ( 149.99)** (1,112.22)** (627.47)** (49.94)** ( 219.97)** ( 506.88)** (17.01)** (203.77)** (104.21)** ( 151.43)**
9.31 0.08 0.01 0.49 0.51 0.05 0.87 0.03 0.03 0.24 0.15 0.11
(211.12)** (3.14)** (.0.17) (94.06)** (10.38)** ( 17.38)** (11.38)** (0.44) (0.84) ( 13.44)** (8.81)** (2.51)* 74.68 6,360
Panel B: Abnormal Audit Fees for Post-SOX Years Post-SOX Mean Reported Predicted Abnormal 12.94 12.40 0.54 (90.59)** Median 12.90 12.42 0.48 (25.30)**
*, ** Indicate signicance at the 0.05 percent and 0.01 percent levels, respectively. The Pre-SOX (Post-SOX) regression includes all observations from the full sample where the scal year-ends on or before (after) July 30, 2002. The dependent variable Audit fees is the natural logarithm of fees paid to the external audit rm for the scal years nancial statement audit. The explanatory variables are dened as follows. Auditor is an indicator variable that equals 1 when the external auditor is one of the Big 4 or Big 5 auditing rms, 0 otherwise. Herndahl is the normalized Herndahl Index within each two-digit standard industry classication (SIC) code. Assets is the natural logarithm of total assets. Currentassets is the ratio of current assets to total assets. Currentratio is the ratio of current assets to current liabilities. Receivableratio is the ratio of total receivables to total assets. Inventoryratio is the ratio of inventory to total assets. Protability is income before extraordinary items divided by total assets. Leverage is the ratio of total debt to total assets. Loss is an indicator variable that equals 1 for rms with negative income before extraordinary items, 0 otherwise. Merger is an indicator variable that equals 1 if the Acquisition / Merger Pretax number is non-zero, 0 otherwise. All variables are measured as of the scal year-end and regression t-statistics are calculated using White (1980)-corrected standard errors. In Panel B we compute predicted or counterfactual audit fees for the post-SOX years by interacting the estimated coefcients from the pre-SOX period with all the rm-year specic values for the post-SOX years.
194
different for the post-SOX years. Our results suggest that although the intercept coefcient 0.03 / 9.31), declined for the post-SOX years, the decline is less than 0.5 percent ( which is economically insignicant. In contrast, seven of the 11 slope coefcients are larger for the post-SOX years. Other than Auditor, the variables with larger estimated coefcients (Assets, Currentassets, Inventoryratio, Leverage, Protability, and Loss) are generally considered various constructs for audit risk. Thus, the relative importance of audit risk in inuencing audit pricing increased signicantly following SOX. Further, Big 4 auditors were charging higher audit fees during this period. To assess the aggregate impact of the structural shifts in the estimated coefcients on audit fees, we compute abnormal audit fees for the post-SOX years using the counterfactual approach as in Chaney et al. (2004). The counterfactual approach estimates what audit fees for the post-SOX years would have been had the audit fee pricing environment not changed. Specically, we compute predicted audit fees for the post-SOX years by interacting the estimated coefcients from the pre-SOX years with the rm-specic values from the postSOX years. Abnormal audit fees for the post-SOX years are dened as the difference between the reported and predicted audit fees. Panel B of Table 8 presents reported audit fees, predicted audit fees, and abnormal audit fees. We nd that mean and median abnormal audit fees are positive and highly signicant, where abnormal audit fees are the difference between reported audit fees and predicted fees. Our results suggest that structural shifts in the audit fee structure at the rm level, especially from changes in the pricing of audit risk, resulted in abnormally high audit fees for the post-SOX years. SENSITIVITY TESTS In this section we conduct several sensitivity tests. For brevity, we discuss the results of the tests without reporting them in separate tables. Audit Fees for Large Auditors from Yearly Regressions A more rened test for increases in audit fees is to perform yearly regressions (i.e., estimate Equation (3) without SOX) and focus on the intertemporal variations in the estimated coefcients on Auditor over the sample period. If auditor fees increased following SOX, we expect the magnitude of the coefcient on Auditor to increase over successive years. Consistent with our expectations, we nd that the coefcients on Auditor are 0.09, 0.10, 0.18, 0.38, and 0.47 for the years 2001 to 2005, respectively. The yearly regressions indicate that audit fees for large auditors increased monotonically but that the biggest increases began in 2002. Former Andersen Clients Andersen ofcially exited the market for audit services after its conviction for obstruction of justice on June 13, 2002, which is very close to the date used to partition our sample into two groups (pre- and post-SOX). The exit of Andersen might confound our fee discounting results because it occurred concurrently with the enactment of SOX. To control for the effect of forced auditor switches, we exclude the switches of former Andersen clients that happened after the rms conviction from our sample and reestimate Regression 2 from Table 7. When we eliminate former Andersen client switches, our results remain qualitatively 0.05, unchanged. The coefcient for SOX * SwitchAtomistic remains insignicant ( 2 0.61), while the coefcient for SOX * SwitchOligopolistic is positive and signicant t-stat 0.24, t-stat 2.95) ( 3 0.22, t-stat 3.77). The coefcients for SwitchAtomistic ( 4
Auditing: A Journal of Practice & Theory American Accounting Association November 2009
195
and SwitchOligopolistic ( 5 0.20, t-stat 3.71) remain negative and signicant. By eliminating clients forced to hire new auditors after Andersens conviction, we control for the possibility that forced auditor switches might be an explanation for our ndings. Complexity/Audit Effort Relationship To ensure that the results for SOX in our levels regressions are not driven solely by the complexity of the clients business, we conduct two additional analyses. First, using a segment subsample, we construct an indicator variable (Complex) that equals 1 if a rm has Receivableratio, Inventoryratio, Segments, and Foreign values above the median values for each variable, 0 otherwise. We include Complex as an added explanatory variable and 0.19, reestimate Regression 3 from Table 3; we nd that the coefcients for Complex ( 6.83), SOX ( 0.39, t-stat 7.79), and SOX * Auditor type ( 0.27, t-stat t-stat 5.15) are all positive and signicant. Second, we reestimate Regression 3 from Table 3 after including Complex and SOX * Complex; we nd that only SOX * Complex remains signicant. More importantly, SOX ( 0.38, t-stat 7.68) and SOX * Auditor ( 0.27, t-stat 5.10) remain positive and signicant. These sensitivity results indicate that companies with more complex operations were paying signicantly higher fees and that the amount of additional fees charged for complexity has grown since SOX. Even controlling for the additional increase in audit fees for complex businesses, we nd additional evidence that SOX increased audit fees and the Big 4 premium. Constant Dollar Audit Fee Regressions Because Audit Analytics presents audit fees directly from company disclosures, we control for the possibility that the results for SOX are partially driven by general price increases over our sample period. We scale the audit fees by the GDP deator published by the United Nations for each year and reestimate Regressions 1 and 2 from Table 3.17 For Regression 1, the coefcient for SOX remains positive and signicant ( 1 0.51, 48.55) and for Regression 2, the coefcients for both SOX ( 1 0.21, t-stat t-stat 8.08) and SOX * Auditor type ( 2 0.36, t-stat 12.70) are also positive and signicant. Our ndings that SOX increased audit fees, as well as the Big 4 premium, remain robust when controlling for increases in general prices. High-Litigation Industries Several studies nd that rms in some industries are more likely to be sued than others (e.g., PwC 2008). To examine whether the impact of SOX on audit fees is stronger for rms in more litigious industries, we reestimate Regressions 2 and 3 from Tables 3 and 5 by including an indicator variable for rms in high-litigation industries. As in prior research (e.g., Francis et al. 1994; Frankel et al. 2002; Ashbaugh et al. 2003), we dene Litigation as equaling 1 when rms are from any one of the following industries: biotechnology (SIC codes 28332836 and 87318734), computers (SIC codes 35703577 and 73707374), electronics (SIC codes 36003674), and retailing (SIC codes 52005961); 0 otherwise. The results show that the Litigation variable is insignicant in all regressions. However, the coefcients on SOX and SOX * Auditor continue to be positive and signicant. Thus, our tests indicate that SOX did not alter the increase in audit fees differentially between high- and low-litigation industries.
17
Data obtained from the United Nations Statistics Division website (http: / / unstats.un.org / ): 2000 102.4; 2002 104.1; 2003 106; 2004 108.8; 2005 111.8.
100; 2001
196
CONCLUSION This paper studies fees paid to public accounting rms around the Sarbanes-Oxley Act. Drawing heavily on the audit fee model developed by Simunic (1980), we hypothesize that key provisions of SOX increase audit effort, expected auditor legal liability, and, hence, fees. Using a large sample drawn from Audit Analytics between 2000 and 2005, we nd a large increase in audit fees following SOX (i.e., for scal years ended after July 30, 2002). Controlling for other auditor and client characteristics, our results suggest that average audit fees increased by approximately 74 percent following SOX. We also nd evidence that the post-SOX fee increase is even larger for the Big 4 audit rms. While audit fees increased sharply after SOX, there was a marked decline in nonaudit fees. Overall, total fees (the sum of audit and nonaudit fees) rose, indicating that the total revenue package from the average audit client increased after SOX, even though public accountants were precluded from providing specic nonaudit services. We also nd that the fee discounting (lowballing) associated with initial engagements documented by prior researchers (e.g., Craswell and Francis 1999, Ghosh and Lustgarten 2006) abated in the post-SOX period. Similar to Ghosh and Lustgarten (2006), we nd evidence of fee discounting in the atomistic market segment for the post-SOX years. While we nd some evidence of fee discounting by large audit rms in the pre-SOX years, we document no such evidence for the post-SOX years. Overall, we nd that SOX not only increased audit fees, but also enhanced the large auditor premium, while reducing fee discounts for new audit clients.
REFERENCES
Ashbaugh, H., R. LaFond, and B. Mayhew. 2003. Do nonaudit services compromise auditor independence? Further evidence. The Accounting Review 78 (3): 611639. Asthana, S., S. Balsam, and S. Kim. 2004. The effect of Enron, Andersen and Sarbanes-Oxley on the market for audit services. Working paper, Temple University. Bell, T., W. Landsman, and D. Shackelford. 2001. Auditors perceived business risk and audit fees: Analysis and evidence. Journal of Accounting Research 39 (1): 3543. Burns, J. 2007. SECs top accountant favors limiting auditors liability, reduce complexity of rules. Associated Press, January 25, San Diego, CA. Chan, D. 1999. Lowballing and efciency in a two-period specialization model of auditing competition. Contemporary Accounting Research 16 (4): 5172. Chaney, P., D. Jeter, and L. Shivakumar. 2004. Self-selection of auditors and audit pricing in private rms. The Accounting Review 79 (1): 51. Choi, J., J. Kim, X. Liu, and D. Simunic. 2008. Audit pricing, legal liability regimes, and Big 4 premiums: Theory and cross-country evidence. Contemporary Accounting Research 25 (1): 55 99. Ciesielski, J., and T. Weirich. 2006. Ups and downs of audit fees since the Sarbanes-Oxley Act. The CPA Journal 76 (10): 2835. Craswell, A., J. Francis, and S. Taylor. 1995. Auditor brand name reputations and industry specializations. Journal of Accounting and Economics 20: 297322. , and . 1999. Pricing initial audit engagements: A test of competing theories. The Accounting Review 74 (2): 201216. Cummings, J. 2005. Sarbanes-Oxley fails to slow fraud. Business Finance 11: 9. DeAngelo, L. 1981. Auditor size and audit quality. Journal of Accounting and Economics 3: 183 199. Fields, L., D. Fraser, and M. Wilkins. 2004. An investigation of the pricing of audit services for nancial institutions. Journal of Accounting and Public Policy 23 (1): 5377. Auditing: A Journal of Practice & Theory American Accounting Association November 2009
197
Financial Executives International (FEI) 2004. FEI Special Survey on Sarbanes-Oxley Section 404 Implementation Executive Summary. Available at: http: / / www.404institute.com / docs / SOXSurveyJuly.pdf. Francis, J., D. Philbrick, and K. Schipper. 1994. Shareholder litigation and corporate disclosures. Journal of Accounting Research 32 (2): 137164. Frankel, R., M. Johnson, and K. Nelson. 2002. The relation between auditors fees for nonaudit services and earnings management. The Accounting Review 77 (Supplement): 71105. Ghosh, A., and S. Lustgarten. 2006. Pricing of initial audit engagements by large and small audit rms. Contemporary Accounting Research 23 (2): 333368. Grifn, P., and D. Lont. 2007. An analysis of audit fees following the passage of Sarbanes-Oxley. Asia-Pacic Journal of Accounting and Economics 14: 161192. Heffes, E. M. 2005. SEC enforcement division: Ratcheting up. Financial Executive 21: 1416. Iliev, P. 2007. The effect of the Sarbanes-Oxley Act (Section 404) on audit fees, accruals and stock returns. Working paper, Brown University. Kennedy, P. 1992. A Guide to Econometrics. 3rd edition. Cambridge, MA: MIT Press. Menon, K., and D. Williams. 2001. Long-term trends in audit fees. Auditing: A Journal of Practice & Theory 20 (1): 115136. Palmrose, Z.-V. 1986. Audit fees and auditor size: Further evidence. Journal of Accounting Research 24 (1): 97110. PricewaterhouseCoopers LLP (PwC). 2008. 2007 Securities Litigation Study. New York, NY: PwC Advisory Crisis Management. Rashkover, B., and C. Winter. 2005. The impact of Sarbanes-Oxley on SEC enforcement in public company disclosure cases. International Journal of Disclosure and Governance 2: 312324. Reilly, D. 2007. Audit watchdog report raps Ernst, KPMG. The Wall Street Journal (January 12): C3. Securities and Exchange Commission (SEC). 2003. Final Rule: Strengthening the Commissions Requirements Regarding Auditor Independence Requirements. Release No. 33-7919. Washington, D.C.: Government Printing Ofce. Seetharaman, A., F. Gul, and S. Lynn. 2002. Litigation risk and audit fees: Evidence from U.K. rms cross-listed on U.S. markets. Journal of Accounting and Economics 33: 91115. Simon, D., and J. Francis. 1988. The effects of auditor change on audit fees: Tests of price cutting and price recovery. The Accounting Review 63 (2): 255269. Simunic, D. 1980. The pricing of audit services: Theory and evidence. Journal of Accounting Research 18 (1): 161190. , and M. Stein. 1996. The impact of litigation risk on audit pricing: A review of the economics and the evidence. Auditing: A Journal of Practice & Theory 15 (Supplement): 119134. U.S. House of Representatives. 2002. The Sarbanes-Oxley Act of 2002. Public Law 107-204 [H.R. 3763]. Washington, D.C.: Government Printing Ofce. White, H. 1980. A heteroskedasticity-consistent covariance matrix estimator and a direct test for heteroskedasticity. Econometrica 48: 817838. Wilcox, K. 2007. Dealing with Sarbox. The Wall Street Journal (June 1): A13.
Copyright of Auditing is the property of American Accounting Association and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.