Audit Fees and Auditor Dismissals in the Sarbanes-Oxley Era. Michael L. Ettredge, Chan Li, and Susan Scholz

Transcription

1 Accounting Horizons Vol. 21, No. 4 December 2007 pp Audit Fees and Auditor Dismissals in the Sarbanes-Oxley Era Michael L. Ettredge, Chan Li, and Susan Scholz SYNOPSIS: The accounting scandals and Sarbanes-Oxley Act (SOX) of 2002 resulted in large increases in required audit work, and corresponding increases in audit fees for public companies. This study provides early evidence regarding the relationship between higher audit fees, both levels and changes, and auditor dismissals in the period immediately subsequent to the passage of SOX. We find that clients paying higher fees are more likely to dismiss their auditors. We also find that dismissals are associated with smaller companies, companies with going-concern reports, and companies that later reported material weaknesses in their internal controls. Among dismissing clients, smaller Big 4 clients, paying higher fees, tend to hire non-big 4 successor auditors. This result holds when auditors are divided into Big 4, national, and local tiers. We also find evidence that dismissing clients, in particular clients hiring new non-big 4 auditors, experience smaller fee increases than nonswitching clients in the following year. These results are consistent with the notion that in the immediate post-sox period, some companies dismissed their auditors in expectation of lower fees from the succeeding auditor. Keywords: auditor dismissal; audit fees; SOX Act of 2002; internal control weakness. JEL Descriptor: M420. INTRODUCTION The Sarbanes-Oxley Act (SOX) of 2002 and the preceding accounting scandals significantly changed the relations between many U.S. companies and their external auditors. An important effect of these changes is an upward pressure on audit fees. For example, in response to increased scrutiny and risk, audit firms report increasing their efforts, resulting in higher costs (PricewaterhouseCoopers 2005). Also, further limitations on consulting activities force more audits to be priced as stand-alone services. Finally, the SOX Section 404 report on internal controls (for which preparatory work began during our sample period) adds a new set of procedures and related costs to the standard audit. Deloitte USA reported that its clients fees increased by about 40 percent from 2003 to 2004 (Whitehouse 2005). Other estimates are higher, especially for smaller clients (Solomon Michael L. Ettredge is a Professor and Susan Scholz is an Associate Professor, both at the University of Kansas, Chan Li is an Assistant Professor at the University of Pittsburgh. We appreciate the helpful comments of the AAA auditing mid-year conference participants, and of workshop participants at the University of Kansas, Trinity University, The University of Texas at Dallas, and University of Utah. Submitted: July 2006 Accepted: June 2007 Corresponding author: Susan Scholz sscholz@ku.edu 371

2 372 Ettredge, Li, and Scholz 2005). Importantly, these increased costs are largely because of additional work imposed on clients, rather than demanded by them. That is, while audit fees increased, perceived client benefits arguably did not. We explore the association between audit fees and auditor dismissals in the immediate post-sox time period. Specifically, we investigate whether clients with greater recent fee increases, or those paying higher fees relative to other companies, are more likely to dismiss their auditors following their first post-sox audit. We expect clients to consider lower-cost options following an increase in the cost of audit services. As clients historically receive fee cuts on switching auditors (Simon and Francis 1988; Ettredge and Greenberg 1990; Walker and Casterella 2000), they may reasonably expect to obtain lower fees by dismissing their incumbent auditor and hiring another. However, SOX-induced resource constraints may have reduced the traditional postswitch discount. Also, clients will not switch if expected switching costs are greater than expected fee savings (DeAngelo 1981). We expect clients paying relatively high fees, that is, those with the greatest savings potential, are more likely to switch. Indeed, in a pre-sox setting, Francis and Wang (2005) report that companies paying relatively high fees in the first year of mandated audit fee disclosure (2000) tended to pay lower fees in the subsequent year (2001). Their study focuses on increased client bargaining power, because of newly available, low-cost fee information as the mechanism for realizing lower fees, while we focus on auditor switching. To test our expectation that companies with higher post-sox fees will switch in search of lower fees, we use logistic regression analyses to compare dismissal to nonswitching companies. We find evidence that higher fees are indeed associated with a higher likelihood of auditor dismissal, after controlling for other factors associated with dismissals in previous research (Johnson and Lys 1990; Carcello and Neal 2003). We also investigate the type of auditor hired following a dismissal. The premium charged by Big 4 firms is well-documented (Craswell et al. 1995; Ireland and Lennox 2002; Sankaraguruswamy and Whisenant 2004; Rama and Read 2006). So, on average, Big 4 clients have greater opportunity to realize cost savings by switching to non-big 4 firms, thereby eliminating the Big 4 premium. We find that the majority of Big 4 clients who dismiss their auditors during our sample period subsequently hire non-big 4 auditors, contrary to pre-sox switching patterns. We also find that higher fees are associated with switches to smaller, likely less expensive auditors. Switching to a non-big 4 auditor is not always a viable option for larger clients that require a wider scope of services, and indeed, we find the association between fees and switching to smaller-tier firms is largely attributable to smaller clients. Finally, we find some evidence that dismissing clients do obtain fee relief from their new auditors. As this period is characterized by generally rising fees, we compare relative fee increases in the following year. We report that fees increase less for dismissal clients than for nonswitching clients. Furthermore, among dismissing clients, those switching from Big 4 to non-big 4 auditors have smaller fee increases than those remaining within the Big 4 tier. In sum, our evidence suggests that, in the immediate post-sox period, higher audit fees and the expectation of limiting future fee increases are significant factors in client dismissals of auditors. STUDIES OF POST-SOX AUDITOR-CLIENT RELATIONS Several recent and contemporaneous studies investigate auditor-client relations in the immediate post-sox era. They document an overall increase in audit fees, attributable to the new regulation (Asthana et al. 2004; Griffin and Lont 2007). However, several fee

3 Audit Fees and Auditor Dismissals in the Sarbanes-Oxley Era 373 regularities reported from the pre-sox era persist. Premiums charged by the Big 4 and risk premiums for higher risk clients appear to have remained, and even increased (Asthana et al. 2004; Griffin and Lont 2005; Rama and Read 2006). Lowball fees in the initial year of a new auditor also appear to continue in some post-sox settings (Asthana et al. 2004; Griffin and Lont 2005). Publicly available fee data are available beginning with 2000 audit fees, so research regarding pre-sox audit fees is somewhat limited. 1 Recent studies also document an increase in switching activity, relative to pre-sox periods, in both dismissals and resignations (Griffin and Lont 2005; Cassell et al. 2006; Rama and Read 2006; Ettredge et al. 2007). Some of these studies focus on auditor resignations (Rama and Read 2006) or consider risk primarily from the auditor s perspective (Cassell et al. 2006). Evidence from these studies suggests riskier clients are switching from Big 4 to non-big 4 auditors, leading to an increase in the overall risk profile of non-big 4 client portfolios. This association between higher risk and resignations is consistent with pre-sox studies (Krishnan and Krishnan 1997; Shu 2000). Regarding dismissals, very generally, pre-sox studies indicate that growing companies with increasing agency costs (i.e., size and leverage) tend to realign to larger auditors (Johnson and Lys 1990; DeFond 1992). Our study focuses on the role played by higher fees in post-sox dismissal decisions, incremental to the association between switching and client risk characteristics noted in contemporaneous research. RESEARCH DESIGN We focus on a time period intended to capture the substantive effects of the 2002 SOX Act and its precipitating scandals. Specifically, we study fees paid for audits of fiscal year 2003, which is the first full year following these significant events. 2 Although it is likely that 2002 fees also reflected SOX-related effects, much was unknown about the implementation of SOX regulations at the time of the 2002 audits, particularly at the time of interim fieldwork. Furthermore, research suggests fees for initial audit engagements are atypical (i.e., the lowballing literature), so 2002 fees are also problematic because of Andersen s dissolution that year. We expect 2003 fees to capture increases initially charged in 2002 as well as incremental increases for Our 2003 fee data are from the Audit Analytics database, as reported by clients in We analyze two measures intended to capture alternative aspects of higher fees: fees relative to similar clients and fee increases over time. We expect both measures are associated with dismissals, although for different reasons. First, clients observing similar companies paying lower fees are more likely to reasonably expect that lower fees are attainable, and to find an alternative provider charging such rates. Second, we expect clients with greater fee shocks, that is greater fee increases, are more likely to consider alternative audit providers. While some of these companies may not be able to obtain lower fees, we expect that increased searching by these companies will lead to more dismissals among this group. Our first measure addresses fees relative to similar clients. To do this, we use a standard regression model which explains audit fees for each client. Our model includes all companies with audit fee data from Audit Analytics and necessary financial data from Compustat a total of 5,278 companies. Fee determinants are derived from prior studies 1 In February 2001, the SEC mandated that fees paid to a company s auditor for audits and specified categories of nonaudit services be disclosed in proxy statements. Thus, fees for 2000 audits were reported in many proxies filed in For the 81 percent of our sample with calendar year-ends, this is the year ending December 31, For companies with other fiscal year-ends, we include all fiscal years ending September 2003 through August 2004.

4 374 Ettredge, Li, and Scholz (Francis 1984; Simon and Francis 1988; Craswell et al. 1995; Menon and Williams 2001; Ireland and Lennox 2002; Hay et al. 2006) and include variables to capture each company s size, complexity, and risk factors. We also control for companies filing the first round of SOX 404 reports. While the first full SOX 404 audits were not conducted until the subsequent year, preparatory work began earlier. 3 The difference between actual fees and fees estimated by the regression model for each client is captured in the residual of the regression equation. We use this residual as a measure of abnormal fees (AFEE). This is a summary of our fee model: LNFEE b b LNASSETS b ADJSALES b INVAR b DTRATIO b LOSS b SEGNUM b GOCERN b AUDITOR b DELAY b BUSY b 404RPT INDUST e where: LNFEE ln (audit fees for fiscal 2003 audit work); LNASSETS ln (total assets); ADJSALES sales/total assets; INVAR (total inventory total accounts receivable)/total assets; DTRATIO total debt/total assets; LOSS 1 if company reports a net loss; 0 otherwise; SEGNUM number of reportable segments; GOCERN 1 if company received a going-concern opinion; 0 otherwise; AUDITOR 1 if auditor is one of the Big 4; 0 otherwise; DELAY number of days from a company s fiscal year-end to the date the auditors sign their report; BUSY 1 if client fiscal year-end is between December 1 and March 31; 0 otherwise; 404RPT 1 if client filed a SOX Section 404 report for 2004; 0 otherwise; INDUST industry indicators based on 2-digit SIC codes; and e the model residual and our measure of abnormal fees (AFEE). Results for this model, not tabulated, are consistent with those reported in prior studies using pre-sox data. That is, all coefficients are positive and are significantly associated with LNFEE. Evaluation of the residuals (AFEE) does not suggest any econometric problems; they are normally distributed and there is no evidence of heteroskedasticity. The explanatory power of the model is also consistent with prior studies (adjusted R ). Our second measure of higher fees is the percent change in fees for each client. We use 2001 audit fees as our baseline year because it was the last year ending prior to the 3 Although compliance with Section 404 was not initially anticipated until 2004 (for large companies) or 2005 (for small companies), and was later delayed further, auditors reported adjusting their procedures during 2003 audits to allow for future 404 attestation. Thus, it is likely that 2003 audit fees include some 404-attestation costs. For example, Gullapalli (2004) reports that in , PricewaterhouseCoopers increased its Section 404 compliance staff by 20 percent to 8,000.

5 Audit Fees and Auditor Dismissals in the Sarbanes-Oxley Era 375 climax of the accounting scandals, passage of SOX, and the dissolution of Andersen. It is calculated as (FEE 03 FEE 01 )/FEE 01. We call this variable CHGFEE. 4 Dismissal Sample As noted above, we are interested in the effects of fees charged for 2003 audits, so we examine dismissals announced from January 2004 through December Extending the sample period through December 2004 allows a sufficient window for 2003 audit work to be completed and a potential dismissal decision to be made. 5 Because of our focus on 2003 fees, we do not study all post-sox dismissals. That is, our sample does not include dismissals from late 2002 through It is possible that some of these earlier dismissals are because of higher fees in 2002 or the expectation of higher fees in However, to the degree that this is the case, omitting these switches from our sample should bias against our finding such an association in the later year. Again using Audit Analytics, we identify 635 companies dismissing their auditors during our sample period. These dismissal companies represent 7.9 percent of all Audit Analytics companies reporting revenues during our sample period (635/8,070). In comparison, over a similar period in 2001, dismissals are 4.5 percent (298/6,630) of the 2001 population. This is consistent with studies indicating that switching activity has increased post-sox (Cassell et al. 2006; Rama and Read 2006). Necessary financial data are not available from Compustat for all of these companies, so we analyze a final sample of 428 dismissals. Most of the attrition is because of missing 2001 financial information necessary to calculate several of our model variables. Logistic Regression Model To examine the association between high fees and dismissals, we use a logistic regression model. We include our fee measures, abnormal fees (AFEE), and percent change in fees (CHGFEE), both separately and together, as they measure different attributes. Our analysis compares companies announcing dismissals with those that did not switch auditors, so our dependent variable is an indicator (DISMISS); companies that dismissed their auditor are coded one, and no-switch companies are coded zero. We control for effects of other factors that have previously been shown to be associated with dismissal decisions: client size, financial condition, growth, auditor type, and auditor industry specialization (Johnson and Lys 1990; DeFond and Subramanyam 1998; Carcello and Neal 2003). Since Ettredge et al. (2007) find evidence that adverse SOX 404 reports are associated with auditor dismissals in a later year, our model also includes a variable to indicate companies reporting weak internal controls. These may be indicated in a SOX Section 302 report anytime 4 There are a few companies with extreme values for this measure, due mainly to extremely low fees in To prevent these observations from exerting undue influence on the results, we Winsorize the variable at 99 percent and 1 percent of the distribution. That is, we change the values in the tips of the distribution to equal the values at the above percentiles. Results for the un-winsorized variable are consistent with reported results. We also test other measures of fee increases, including the changes in ln(fees)/ln(sales), fees/ sales, and fees/sales from 2001 to Results do not change. CHGFEE requires data for 2001 audit fees, which are not available for all companies in our sample. Based on an examination of proxy statements, this data limitation is due mainly to companies not reporting fee data in 2001, as well as some companies that did not exist in Reduced sample sizes are noted as appropriate. 5 For calendar year-end companies (81 percent of the sample), the dismissal window ends about nine months after the expected 10-K filing date in March As mentioned previously, our fiscal year-ends range from September 2003 through August Thus, the dismissal window begins about three months after the earliest fiscal year-end (to allow time for the audit to be completed) and ends four months after the last fiscal year-end.

6 376 Ettredge, Li, and Scholz through 2004, or in an adverse SOX 404 report for the 2004 financial statements. This is a summary of our model: where: DISMISS b b HIGH FEES b CSIZE b LOSS b GOCERN b LEVERAGE b CHGOCF b CHGDEBT b GROWTH b OLDBIG4 b SPECIALIST b ICMW e DISMISS 1 if auditor was dismissed; 0 no switch; HIGH FEES CHGFEE or AFEE, or both; CSIZE ln (client revenue) 6 LOSS 1 if client reported negative earnings; 0 otherwise; GOCERN 1 if client received a going-concern opinion; 0 otherwise; LEVERAGE total liabilities/total assets; CHGOCF (operating cash flow 03 /assets 03 ) (operating cash flow 01 /assets 01 ); CHGDEBT (debt 03 /assets 03 ) (debt 01 /assets 01 ); GROWTH (sales 03 sales 01 )/sales 01 ; OLDBIG4 1 if incumbent/pre-switch auditor is Big 4; 0 otherwise; SPECIALIST 1 if the auditor audits at least 25 percent of the client s industry s revenues; 0 otherwise; and ICMW 1 if client reports a material internal control weakness in its SOX 302 report or receives an adverse SOX 404 report through 2004; 0 otherwise. RESULTS Descriptive Statistics Our primary analysis compares the 428 dismissal companies discussed above with 4,516 companies with necessary financial data that did not switch auditors during our sample period. 7 Table 1 compares means and medians for model variables between the two groups. On average, the percent change in fees for dismissal companies is significantly higher for dismissal than for no-switch clients. Fee levels (AFEE) are also significantly higher for dismissal companies. Among control variables, dismissal companies tend to be smaller, report more losses, receive more going-concern and adverse internal control reports, and have higher leverage and greater increases in debt levels. Clients of Big 4 and specialist auditors are less likely to dismiss. Results of mean and median tests are consistent for all continuous variables except CHGFEE, where only mean tests are significant, and GROWTH, where only median tests indicate slower growing companies are more likely to dismiss their auditors. Logistic Regression Analysis for Auditor Dismissals Table 2 presents results for three versions of our basic model using CHGFEE only, AFEE only, and both CHGFEE and AFEE as measures of higher fees. In each, the fee 6 In this model, we use ln(revenues) to measure client size because of relatively high total asset balances among the financial firms in our sample. Our results do not change if we substitute industry-adjusted ln(assets), if we include industry indicator variables or if we exclude financial firms from all analyses. 7 We exclude companies with auditor resignations during our sample period from our analyses. As resignations are initiated by the auditor, rather than the client, we do not expect the same association between higher fees and resignations that we do for dismissals.

7 Audit Fees and Auditor Dismissals in the Sarbanes-Oxley Era 377 TABLE 1 Descriptive Statistics for Dismissal and No-Switch Companies Means/ Percentages Medians Dismiss No-Switch t-statistic/ Chi-sq Dismiss No-Switch z-score n 428 4, ,516 Variables CHGFEE 78.3% 68.2% 2.056** 44.8% 47.6% AFEE *** ** CSIZE *** *** LOSS 49.5% 30.1% 8.311*** GOCERN 17.5% 3.8% *** LEVERAGE *** *** CHGDEBT *** *** CHGOCF GROWTH *** OLDBIG4 63.6% 81.1% 8.664*** SPECIALIST 29.7% 36.1% 2.654*** ICMW 18.5% 12.2% *** *, **, *** significant at.10,.05,.01 levels, respectively, based on t-statistics and Wilcoxon z-scores for tests of continuous variables, Chi-square results for categorical variables. Variable Definitions: CHGFEE percent change in fees from 2001 to 2003 (fees 03 fees 01 /fees 01 ) (based on a reduced sample size of 321 dismissal and 3,672 no-switch companies); AFEE abnormal audit fees for 2003 (residuals derived from standard fee model); CSIZE ln(revenue); LOSS 1 if company reports a loss; 0 otherwise; GOCERN 1 if company received a going-concern opinion; 0 otherwise; LEVERAGE total liabilities 03 /total assets 03 ; CHGDEBT (debt 03 /assets 03 ) (debt 01 /assets 01 ); CHGOCF (operating cash flow 03 /assets 03 ) (operating cash flow 01 /assets 01 ); GROWTH (sales 03 sales 01 )/sales 01 ; OLDBIG4 1 if incumbent/pre-switch auditor is Big 4; 0 otherwise; SPECIALIST 1 if auditor audits at least 25 percent of the industry revenues; 0 otherwise; and ICMW 1 if client reports a material internal control weakness in its 302 report or receives an adverse 404 report through its 2004 financial statements; 0 otherwise. variables are significantly positive, indicating that both higher relative fees and larger increases in fees are associated with clients subsequently dismissing their auditors. Control variable results consistently indicate companies dismissing their auditors are smaller, more likely to have received going-concern opinions, and more likely to later receive adverse internal control reports than are nonswitching companies. 8 In the model with only relative fees (AFEE), we also find some evidence that companies reporting losses, having greater increases in operating cash flows and debt, and employing industry specialist 8 Although this variable is defined as one for both adverse SOX 302 and 404 reports, the vast majority of companies are coded one because of adverse SOX 404 reports, which first appeared in the subsequent year s financial statements.

8 378 Ettredge, Li, and Scholz TABLE 2 Logistic Regression: Analysis of Association between Audit Fees and Dismissals Dependent Variable DISMISS DISMISS DISMISS Dismissal n No-switch n 3,672 4,516 3,672 Coef. p-value Coef. p-value Coef. p-value CHGFEE *** ** AFEE *** ** CSIZE *** *** *** LOSS * GOCERN *** *** *** LEVERAGE CHGOCF ** CHGDEBT * GROWTH OLDBIG ** SPECIALIST * ICMW *** *** *** Intercept * ** * Percent correctly classified 92% 91% 92% Model Chi-Square *** *** *** Pseudo R *, **, *** significant at.10,.05,.01 levels, respectively. Variable Definitions: DISMISS 1 if dismissal firm; 0 otherwise. See Table 1. auditors are more likely to dismiss. However, dismissals are less likely for companies employing Big 4 auditors. 9 In summary, our results suggest that in the immediate post-sox period, even after controlling for important client characteristics and risk factors, companies experiencing larger fee increases from the pre-sox period, or unusually high fee levels relative to other companies, are more likely to dismiss their auditors. Additional Analyses For further evidence regarding the association of dismissals and fees, we conduct two additional analyses. We examine the direction of the auditor switches to see if dismissal clients tend to select their new auditor from a relatively smaller, lower-cost auditor tier. We also examine fees paid in the following year to see if dismissal clients realize fee savings. 9 As noted in Table 1, the size difference between dismissal and nonswitching companies is quite large. We repeat this analysis using a matched sample design to be sure our regression model adequately controls for size effects. In this analysis, dismissal companies are matched with no-switch companies based on industry and size. Using a conditional regression model (see Hosmer and Lemeshow 2000; Agrawal and Chadha 2005), results for the fee variables are nearly identical. Among control variables, neither size nor going-concern opinion is significant, and clients of Big 4 auditors are more likely to dismiss.

9 Audit Fees and Auditor Dismissals in the Sarbanes-Oxley Era 379 New Auditor Selection First, we analyze the type of auditor selected after the dismissal. The premium charged by Big 4 auditors has been well-documented (Craswell et al. 1995; Ireland and Lennox 2002; Sankaraguruswamy and Whisenant 2004; Rama and Read 2006). So, on average, a client switching from a Big 4 to a non-big 4 firm can reasonably expect to pay a lower audit fee. Therefore, if dismissals are due in part to an attempt to reduce future fees, we should observe a meaningful number of Big 4 clients switching to non-big 4 auditors to realize fee savings. If so, this contrasts with switching evidence documented prior to SOX, in which the new auditor following a dismissal is most often either the same type or a larger auditor (Francis and Wilson 1988; Johnson and Lys 1990; DeFond and Subramanyam 1998; Wolk et al. 2001; General Accounting Office 2003; Landsman et al. 2005). To confirm this prior pattern with our dataset, we examine Audit Analytics data for 2001, and estimate that 62 percent of clients dismissing a Big 4 auditor in our baseline year subsequently selected another Big 4 firm. As shown in Table 3, post-sox switching patterns between auditor types do appear to differ from pre-sox periods. Of the 428 dismissal companies in our sample, 64 percent (272) dismissed Big 4 auditors, and 69 percent (187) of those switched to a non-big 4 successor. (In contrast, among companies dismissing non-big 4 auditors, only 9 percent hired a Big 4.) This pattern is consistent with concurrent research which also documents a post-sox increase in switches from Big N to non-big N auditors (Cassell et al. 2006). Some clients may change to a smaller auditor because Big 4 auditors will not accept them. Therefore, we directly examine the association between higher fees and the choice of new auditor in a logistic regression model that controls for client risk factors including poor financial condition and weak internal controls. We first focus our switching model on the 272 clients dismissing a Big 4 auditor. Our dependent variable equals one if the new auditor is a non-big 4 firm, zero otherwise. Since all companies previously employed Big 4 auditors, there is no variable to indicate the type of auditor dismissed. We expect the association between higher fees and switching to a smaller auditor is stronger for smaller clients. This is because smaller auditors are more viable alternatives for smaller clients, since larger clients often require the wider scope of services typically offered by Big 4 audit firms. Therefore, we add an interaction between higher fees TABLE 3 Type of New Auditor Switch Direction n % Big 4 to Big % Big 4 to non-big % Total Dismissing Big 4 firms % Non-Big 4 to Big % Non-Big 4 to non-big % Total Dismissing Non-Big 4 firms % Total Dismissal Sample % Chi-Square (p-value)

10 380 Ettredge, Li, and Scholz and size to the model. 10 We report only models using AFEE, aschgfee is not significant in these models when included. In results reported in Table 4, AFEE now represents the relation between higher fees and choosing a non-big 4 auditor for smaller clients. The coefficient is positive and significant, suggesting that an unusually high fee paid by smaller clients to a former Big 4 auditor is associated with a switch to a non-big 4 firm. When larger clients dismiss a Big 4 firm, there does not appear to be a significant association between higher fees and selecting a new non-big 4 auditor. The interaction term between fees and size (AFEE*CSIZE) is negative, indicating the association between higher Dependent Variable TABLE 4 Logistic Regression Analysis: Successor Auditor Type Clients Dismissing Big 4 Auditors NEWNB4 All Clients Dismissing Auditors SMALLER Model N Dependent Variable Coef. p-value Coef. p-value AFEE ** *** CSIZE *** *** AFEE*CSIZE * * LOSS * * GOCERN ** LEVERAGE CHGOCF ** CHGDT GROWTH OLDBIG4 n/a *** OLDBIG4*CSIZE n/a *** SPECIALIST * * ICMW ** * Intercept *** *** Correctly classified 70% 73% Model Chi-Square *** *** Pseudo R *, **, *** significant at.10,.05,.01 levels, respectively. Variable Definitions: NEWNB4 1 if new auditor is not one of the Big 4; 0 otherwise; and SMALLER 1 if new auditor is from a smaller tier than the dismissed auditor; 0 otherwise. See Table Here CSIZE is changed to a variable that equals one if the company is larger than the median and zero if smaller. This is to facilitate interpretation of the interaction terms.

11 Audit Fees and Auditor Dismissals in the Sarbanes-Oxley Era 381 fees and switching to a non-big 4 auditor is significantly less for large companies. Furthermore, the sum of the AFEE and interaction coefficients is not significant. 11 The negative coefficient for CSIZE also suggests larger clients are less likely to switch to a non-big 4 auditor. To explore these results further, we increase the number of auditor groups to three: Big 4 firms, firms with a national presence, and firms with regional/local scope. 12 We calculate a second dependent variable, SMALLER, which equals one for dismissing clients switching to any firm belonging to a smaller tier than the dismissed auditor. OLDBIG4 is included in this model as a control variable, as is an interaction term between OLDBIG4 and CSIZE, to capture the different scope requirements of larger clients. These results are shown in the second set of columns in Table 4. Results for the fee variables are very similar to the previous model. That is, smaller clients paying higher fees are more likely to switch to a smaller auditor, while larger clients are not. Not surprisingly, smaller clients previously employing a Big 4 auditor are more likely to switch to a smaller auditor (OLDBIG4 is positive and significant), while larger clients of the Big 4 are less likely to do so (the interaction term is significantly negative). 13 We also find evidence that clients dismissing Big 4 auditors in favor of non-big 4 firms are more likely to receive going-concern reports, to report losses and more negative changes in operating cash flows, and to have previously employed an industry specialist. 14 A possible explanation for the latter result could be that industry specialist auditors also tend to charge a premium (Craswell et al. 1995; DeFond et al. 2000). Thus, companies can avoid two premium charges by dismissing a Big 4 industry specialist and hiring a non-big 4 auditor. Control variable results for clients switching to any smaller auditor are similar, although going-concern reports and operating cash flow changes are not significant. In both analyses, clients later reporting internal control weaknesses are less likely to switch to a smaller auditor. This result may be related to company size, since only larger clients were required to issue SOX 404 reports in To summarize the association between higher fees and new auditor choice, smaller (but not larger) companies tend to switch from larger to smaller-tier auditors if they have been paying unusually high audit fees. This association is consistent when auditors are grouped into either the usual Big 4/non-Big 4 dichotomy or a three-tier split. Subsequent Fees As higher fees are often associated with higher-risk clients, it is possible that the results presented above are because of Big 4 auditors weeding out riskier clients particularly as the SOX mandates push auditors to capacity limits. Auditor-initiated switches should be reported as resignations, which are excluded from our study. However, both clients and 11 We also analyze this effect by partitioning the sample at the median of CSIZE. Results for the fee variables are qualitatively the same. AFEE is significant in the partition of smaller companies, but not in the group of larger clients. 12 Based on an analysis of total client revenue audited, using data from Audit Analytics, we classify six firms as national level: BDO Seidman LLP, BKD LLP, Crowe Chizek & Company LLC, Grant Thornton LLP, McGladrey & Pullen LLP and Moss Adams LLP. 13 We also investigate whether there are differences for firms dismissing Big 4 firms in favor of national versus regional/ local tier auditors. We find that smaller clients paying higher fees are associated with choosing both national and local level firms, while larger clients are not associated with choosing either. Results also suggest that smaller clients, and those receiving going-concern reports are more likely to choose a local firm, but clients having internal control material weaknesses are more likely to choose a national firm. 14 Forty-four percent of observations with Big 4 auditors are also classified as being audited by a specialist. Nearly all specialist auditors are Big 4 firms (99 percent).

12 382 Ettredge, Li, and Scholz auditors have incentives to disguise some resignations, so our sample likely includes some implicit resignations. In all our models, we attempt to address this by controlling for important measures of client risk (LOSS, GOCERN, and ICMW). 15 In this section, we analyze subsequent fees for evidence that switching firms do obtain lower fees. If higher risk were driving our results, switching clients would be less likely to realize fee savings, as the new auditor would likely also charge a risk premium. This should especially be true in our post-sox sample period (Rama and Read 2006). Since we expect audit fees, on average, increased from 2003 to 2004, we follow Kohlbeck et al. (2006) and test the association between dismissals and the change in fees. Thus, the dependent variable is the change in the natural log of fees from 2003 to The test variable is DISMISS, coded one for switching clients. If such clients successfully obtain lower fees, we expect DISMISS to be associated with lower fee increases. That is, we expect it to have a negative coefficient. Our fee change model controls for changes in client characteristics (assets, sales, inventory and accounts-receivable ratio, debt ratio, number of segments, and the delay in filing 10-Ks) from 2003 to It includes changes in loss and going-concern status and whether an adverse SOX 404 opinion was reported for Other audit fee determinants (whether the client s audit is during the busy season and whether the client filed a SOX 404 report for 2004) are also included in the model. Results of this analysis are shown in Table 5. The first regression includes all companies in our sample with available data for The coefficient for DISMISS is significantly negative, indicating that clients dismissing their auditors have smaller fee increases in the following year than do nonswitching companies. Next, we test the model on clients dismissing Big 4 auditors, looking for evidence that clients selecting new non-big 4 firms have smaller fee increases than those remaining with a Big 4 auditor. Again, the significantly negative coefficient for NEWNB4 suggests smaller fee increases for these clients. In summary, this evidence suggests that clients dismissing their auditors, and particularly clients dismissing Big 4 auditors in favor of non-big 4 auditors, have lower fee increases in the next year than companies not switching auditors or auditor types. This is consistent with the notion that, during this period, companies dismissed their auditors in expectation of lower future fees. CONCLUSIONS, LIMITATIONS, AND FUTURE RESEARCH We investigate the association between auditor fees and dismissals following significant shocks to the audit market because of passage of SOX in 2002 and the preceding accounting scandals. Based on historical pricing patterns, clients might reasonably expect to pay lower fees after an auditor switch. Therefore, we investigate whether clients paying higher fees after the fee increases surrounding SOX 2002 are more likely to dismiss their auditors. We consider two measures of higher fees: the percent increase in fees for each client and abnormal fees (residuals from a regression of fees on client characteristics associated with fee levels in prior studies). Controlling for client and auditor characteristics previously found to affect auditorswitching decisions, we find that auditor dismissals are associated with both steeper fee 15 We also test the Altman z-score using the subsample of nonfinancial firms. It is insignificant across all models and does not affect any of our reported results. Furthermore, we test a count of business segments and an indicator of foreign sales as additional measures of audit complexity. They are also nearly always insignificant and do not affect any of our reported results.

13 Audit Fees and Auditor Dismissals in the Sarbanes-Oxley Era 383 TABLE 5 OLS Regression Analysis: Change in Audit Fees from 2003 to 2004 Dependent Variable Dismissal and No-Switch Clients FEE Clients Dismissing Big 4 Auditors FEE n 4, Coef. p-value Coef. p-value DISMISS *** NEWNB *** LNASSETS *** ADJSALES ** INVAR ** ** DTRATIO * LOSS *** SEGNUM *** GOCERN *** DELAY *** *** BUSY *** RPT *** *** ICMW *** LNASSET *** LNFEE *** *** Intercept *** *** Model F-statistic *** *** Adj R *, **, *** significant at.10,.05,.01 levels, respectively. Variable Definitions: FEE ln(audit fee 04 ) ln(audit fee 03 ); DISMISS 1 if client dismissed auditor in 2004; 0 otherwise; NEWNB4 1 if client hired new non-big 4 auditor in 2004; 0 otherwise; LNASSETS ln(total assets 04 ) ln(total assets 03 ); ADJSALES (sales 04 /assets 04 ) (sales 03 /assets 03 ); INVAR ((inventory 04 accts rec 04 )/assets 04 ) ((inventory 03 accts rec 03 )/assets 03 ); DTRATIO (total debt 04 /assets 04 ) (total debt 03 /assets 03 ); LOSS 1 if client reports initial loss for 2004; 0 otherwise; SEGNUM number of business segments 04 number of business segments 03 ; GOCERN 1 if client reports initial going-concern report for 2004; 0 otherwise; DELAY auditor reporting lag 04 auditor reporting lag 03 ; BUSY 1 if client s fiscal year end is between December 1 and March 31; 0 otherwise; 404RPT 1 if client filed a SOX 404 report for 2004; 0 otherwise; ICMW 1 if client reports an internal control material weakness through 2004; 0 otherwise; LNASSET03 ln(total assets 03 ); and LNFEE03 ln(audit fees 03 ). increases and higher relative fees. Companies dismissing auditors also tend to be smaller, and to receive more going-concern and adverse internal control reports. On average, most auditors likely increased their fees during this period because of increases in effort and resource constraints. So, meaningful fee savings are most likely when the client switches from a Big 4 auditor to a smaller-tier/lower-cost auditor. Indeed,

14 384 Ettredge, Li, and Scholz in our sample, most clients dismissing Big 4 clients choose a new non-big 4 auditor, and we document an association between higher fees and subsequent switches to smaller-tier auditors, after controlling for several important client-risk characteristics. This appears to be attributable mainly to the subset of Big 4 clients that are small enough to feasibly hire a non-big 4 auditor. Most of the significant control variables in our model have coefficients that are consistent with studies conducted prior to SOX (i.e., clients remaining with a Big 4 auditor tend to be larger, and those switching to smaller auditors tend to receive more going-concern opinions and report losses). One finding contrary to previous studies is that clients are more likely to dismiss industry specialist auditors. We speculate that companies are trying to avoid an additional premium charged by such auditors. We also study the subsequent year s fees, where we find evidence that fee increases are lower for dismissing clients than nonswitching clients. Similarly, clients dismissing Big 4 auditors and hiring non-big 4 auditors have lower fee increases than those choosing a different Big 4 firm. Again, these results are consistent with clients dismissing auditors in expectation of fee relief. An important limitation of our study is that we focus on a unique period: the audit of the first full fiscal year subsequent to the 2002 accounting scandals and passage of SOX. We do not address the association between fees and dismissals prior to this tumultuous year; neither do we investigate possible changes from the pre- to post-sox audit market. Furthermore, it is likely that as auditors and clients gain experience with post-sox audit and assessment requirements, efficiencies and cost savings will limit future fee increases or even partially reverse recent increases (PricewaterhouseCoopers 2006). In this case, feemotivated auditor switches would likely decrease. For example, Francis and Wang (2005) report that when fees disclosure was mandated, adjustment to the new information was largely complete in the subsequent year. Studies of additional periods, both past and future, will be necessary for a more complete understanding of the role of fees in auditor dismissals. Finally, our results appear to depend on a specific subset of smaller Big 4 clients for which switching to a lower-cost auditor is a realistic alternative. Further research can investigate whether this represents a permanent realignment of the audit market, or whether such clients will re-hire Big 4 auditors as the market eventually settles into a post-sox equilibrium. REFERENCES Agrawal, A., and S. Chadha Corporate governance and accounting scandals. The Journal of Law and Economics 48 (2): Asthana, S., S. Balsam, and S. Kim The effect of Enron, Andersen, and Sarbanes-Oxley on the market for audit services. Working paper, Temple University. Available at: ssrn.com /abstract Carcello, J., and T. Neal Audit committee characteristics and auditor dismissals following new going-concern reports. The Accounting Review 78 (1): Cassell, C., G. Giroux, L. A. Myers, and T. C. Omer Audit firms client portfolio decisions in the face of increased regulation and constrained resources. Working paper, Texas A&M University. Available at: Craswell, A. T., J. R. Francis, and S. L. Taylor Auditor brand name reputation and industry specializations. Journal of Accounting and Economics 20 (3):

15 Audit Fees and Auditor Dismissals in the Sarbanes-Oxley Era 385 DeAngelo, L. E Auditor size and audit quality. Journal of Accounting and Economics 3 (3): DeFond, M. L The association between changes in client firm agency costs and auditor switching. Auditing: A Journal of Practice & Theory 11 (1): , and K. R. Subramanyam Auditor changes and discretionary accruals. Journal of Accounting and Economics 25 (1): , J. R. Francis, and T. J. Wong Auditor industry specialization and market segmentation: Evidence from Hong Kong. Auditing: A Journal of Practice & Theory 19 (1): Ettredge, M., and R. Greenberg Determinants of fee cutting on initial audit engagements. Journal of Accounting Research 28 (1): , Heintz, C. Li, and S. Scholz Auditor realignments accompanying implementation of SOX 404 reporting requirements. Working paper, University of Kansas. Available at: http: //ssrn.com/abstract Francis, J. R The effect of audit firm size on audit prices: A study of the Australian market. Journal of Accounting and Economics 6 (2): , and E. R. Wilson Auditor changes: A joint test of theories relating to agency costs and auditor differentiation. The Accounting Review 63 (4): , and D. Wang Impact of SEC s disclosure requirement on subsequent period fees and implications for market efficiency. Auditing: A Journal of Practice and Theory 24 (Supplement): General Accounting Office Public Accounting Firms: Mandated Study on Consolidation and Competition. Washington, D.C.: Government Printing Office. Griffin, P. A., and D. H. Lont The effects of auditor dismissals and resignations on audit fees: Evidence based on SEC disclosures under Sarbanes-Oxley. Working paper, University of California, Davis. Available at: An analysis of audit fees following the passage of Sarbanes-Oxley. Asia-Pacific Journal of Accounting and Economics 14 (2): Gullapalli, D Grasping internal controls. Wall Street Journal (November 3). Hay, D., C. R. Knechel, and N. Wong Audit fees: A meta-analysis of the effect of supply and demand attributes. Contemporary Accounting Research 23 (1): Hosmer, D., and S. Lemeshow Applied Logistic Regression. New York, NY: John Wiley & Sons. Ireland, J. C., and C. S. Lennox The large audit firm fee premium: A case of selectivity bias? Journal of Accounting, Auditing and Finance 17 (1): Johnson, W. B., and T. Lys The market for audit services: Evidence from voluntary auditor changes. Journal of Accounting and Economics 12 (1 3): Krishnan, J., and J. Krishnan Litigation risk and auditor resignations. The Accounting Review 72 (4): Kohlbeck, M., B. Mayhew, P. Murphy, and M. S. Wilkins Competition for Andersen s clients. Working paper, University of Wisconsin Madison. Available at: Landsman, W. R., K. K. Nelson, and B. R. Rountree An empirical analysis of Big N auditor switches. Working paper, University of North Carolina. Menon, K., and D. Williams Long-term trends in audit fees. Auditing: A Journal of Practice & Theory 20 (1): PricewaterhouseCoopers Litigation study. CFODirect Network (June 28) DataLine : Highlights of the 2006 SEC/ PCAOB roundtable on second-year experiences with Section 404. CFODirect Network (May 12). Rama, D. V., and W. J. Read Resignations by the Big 4 and the market for audit services. Accounting Horizons 20 (2): Sankaraguruswamy, S., and J. S. Whisenant An empirical analysis of voluntarily supplied client-auditor realignment reasons. Auditing: A Journal of Practice & Theory 23 (1):

16 386 Ettredge, Li, and Scholz Simon, D. T., and J. R. Francis The effects of auditor change on audit fees: Tests of price cutting and price recovery. The Accounting Review 63 (2): Shu, S. Z Auditor resignations: Clientele effects and legal liability. Journal of Accounting and Economics 29 (2): Solomon, D At what price? Wall Street Journal (October 17). Walker, P. L., and J. R. Casterella The role of auditee profitability in pricing new audit engagements. Auditing: A Journal of Practice & Theory 19 (1): Whitehouse, T Audit pricing played a part in increased 2004 audit costs. Compliance Week (May 17). Wolk, C. M., S. E. Michelson, and C. E. Wootton Auditor concentration and market shares in the U.S.: a descriptive note. British Accounting Review 33 (2):