行政院國家科學委員會專題研究計畫 成果報告
沙氏法案實施後為何會發生大量的會計師解聘? 股票市場
是如何反應的?
研究成果報告(精簡版)
計 畫 類 別 ： 個別型
計 畫 編 號 ： NSC 97-2410-H-004-037-
執 行 期 間 ： 97 年 08 月 01 日至 98 年 09 月 30 日
執 行 單 位 ： 國立政治大學會計學系
計 畫 主 持 人 ： 俞洪昭
計畫參與人員： 此計畫無其他參與人員
報 告 附 件 ： 出席國際會議研究心得報告及發表論文
處 理 方 式 ： 本計畫涉及專利或其他智慧財產權，1 年後可公開查詢
中 華 民 國 99 年 04 月 18 日
1
行政院國家科學委員會補助專題研究計畫
成 果 報 告
□期中進度報告
沙氏法案實施後為何會發生大量的會計師解聘?
股票市場是如何反應的?
計畫類別：
個別型計畫 □ 整合型計畫
計畫編號：NSC 97－2410－H－004－037－
執行期間：97 年 8 月 1 日至 98 年 9 月 30 日
計畫主持人：
俞洪昭
共同主持人：
計畫參與人員：
成果報告類型(依經費核定清單規定繳交)：□精簡報告
完整報告
本成果報告包括以下應繳交之附件：
□赴國外出差或研習心得報告一份
□赴大陸地區出差或研習心得報告一份
□出席國際學術會議心得報告及發表之論文各一份
□國際合作研究計畫國外研究報告書一份
處理方式：除產學合作研究計畫、提升產業技術及人才培育研究計畫、列管
計畫及下列情形者外，得立即公開查詢
□涉及專利或其他智慧財產權，
一年□二年後可公開查詢
執行單位：國立政治大學
中 華 民 國 99 年 1 月 27 日
2
1. INTRODUCTION
The main purpose of this study is to examine whether the non-Big 4’s audit quality increases after the passage of the Sarbanes-Oxley Act (hereafter called SOX).1 This issue is important for two reasons. First, even
though audit quality has received much attention in the auditing literature for many decades,2 the adoption of
the “Big 4 vs. non-Big 4” dichotomy has been extensively used to proxy for audit quality in many empirical auditing studies (e.g., Becker et al. 1998; Behn et al. 2008; Davidson and Neu 1993; DeAngelo 1981; Khurana and Raman 2004; Mansi et al. 2005; Palmrose 1988). The use of such a dichotomy not only overly simplifies the major dimensions of audit quality (Watkins et al. 2004), but also leads to a “research bias” in which the non-Big 4’s audit quality is often overlooked because it is traditionally deemed low. This bias exists even if recent studies have suggested many other measures of audit quality, such as the magnitude of accounting accruals (e.g., Ashbaugh et al. 2003; Becker et al. 1998; Frankel et al. 2002), frequency of restatements (e.g., Kinney et al. 2004), perceived audit quality (e.g., Chaney and Philipich 2002; Khurana and Raman 2004), and industry expertise (e.g., Francis et al. 2005).3
Second, in response to a number of high-profile audit failures occurred in late 2001 and 2002, SOX was passed in July 2002 with an aim to restore investors’ confidence toward the public accounting profession (Oxley 2007). One of the ultimate goals of SOX is to improve audit quality to prevent deceptive accounting practices and, therefore, improve the quality of corporate financial reporting. Because prior studies in audit quality have consistently documented that the Big 4 are associated with high audit quality (e.g., Becker et al. 1998; Craswell et al. 1995; Davidson and Neu 1993; DeFond and Jiambalvo 1991; Francis et al. 1999;
1
Before 1989, there were Big 8 CPA firms (i.e., Authur Andersen, PriceWaterhouse, KPMG, Arthur Young, Coopers and Lybrand, Ernst & Whinney, Deloitte, Haskins & Sells, and Touche Ross). In 1989, the Big 8 reduced to Big 6 because Ernst & Whinney and Arthur Young were merged into the Ernst & Young, and Deloitte, Haskins & Sells and Touche Ross were merged into the Deloitte Touche Tohmatsu. In 1997, the Big 6 further reduced to the Big 5 due to the merger of PriceWaterhouse and Coopers & Lybrand (i.e., the PriceWaterhouseCoopers). In August 2002, because the Authur Andersen was adjudicated by a 5-year probation, the Big 5 reduced to Big 4. To facilitate our discussions, I use Big 4 to represent the Big N whenever it is applicable.
2
See Watkins et al. (2004) for a comprehensive review of prior studies that examine the audit quality issues. 3
Only few studies have examined auditing issues related to the non-Big 4. For example, Louis (2005) finds that, during mergers, the abnormal returns of acquirers audited by non-Big firms outperform those audited by the Big 4 firms. This result implies that non-Big 4 firms may have comparative advantages in assisting their clients in merger activities. In another study, Chang et al. (2008) examines market reaction to announcements of downgrade auditor changes. The empirical results show that market reacts favorably to downgrade auditor changes, given the successor auditors are industry exporters.
3
Palmrose 1986, 1988), the efficacy of SOX to increase audit quality shall be greater for non-Big 4 than for Big 4. Therefore, an examination of the change in non-Big 4’s audit quality before and after SOX provides more prominent evidence about the regulatory consequence of SOX. Currently, such evidence is rare.4
To address the research issue, I first use measures developed in the earnings management literature to capture audit quality. I use earnings management measures to proxy for audit quality because prior studies have found that higher audit quality can effectively mitigate (or suppress) companies’ earnings management (e.g., Ashbaugh et al. 2003; Becker et al. 1998; Frankel et al. 2002). Two measures are used: a traditional measure in which the performance-adjusted discretionary accruals are calculated using the modified Jones model (1995), and a “real” earnings management measure proposed by Roychowdhury (2006). The performance-adjusted discretionary accruals have been extensively used in the earnings management literature during the past few years (e.g., Cohen et al. 2008; Cahan and Zhang 2006; Ferguson et al. 2004; Francis et al. 2005;Ghosh and Moon 2003; Gul et al. 2003; Johnson et al. 2002; Kothari et al. 2005; Myers et al. 2003). However, many researchers have recently begun to argue that companies may employ some real activities (rather than discretionary accruals) to manipulate earnings numbers (Cohen et al. 2008; Roychowdhury 2006). To better understand the association between auditor changes and earnings management, this study contributes to the literature by considering both the discretionary accruals and real earnings management measure to test if companies have changed their methods of managing earnings before and after SOX.
I then evaluate whether there are significant differences in these measures before and after SOX when companies change their auditors. I choose auditor changes because previous research has shown that auditor
4Recent studies have examined the impacts of the potential benefits and costs brought by the SOX on companies’ behavior and market performance. For example, Zhang (2007) investigates the economic consequences of the SOX by examining market reactions to related legislative events prior and subsequent to the passage of SOX. The empirical results report significantly negative cumulative abnormal returns around key SOX events (i.e., restriction of non-audit services and Section 404), suggesting that SOX imposes net costs on complying companies. In another study, Engel et al. (2007) uses a sample of going-private companies from 1998 to 2004 to test the hypothesis that companies go private in response to SOX only if the SOX-imposed costs to the companies exceed the SOX-induced benefits to shareholders, and this difference swamps the net benefit of being a public company prior to the passage of SOX. The empirical findings support the hypothesis and, thus, are consistent with the notion that SOX has affected firms' going-private decisions. Finally, DeFond et al. (2007) examines the influences of SOX on bondholders and finds that bond values decline around the announcement of events leading up to the passage of SOX. This result is consistent with bondholders expecting SOX to impose net costs on them. This study also documents that the decline in bond values is larger among bonds issued by companies that are likely to experience greater changes to their governance systems due to SOX and among bonds with higher default risk. This finding is consistent with bondholders expecting SOX to impose relatively larger net costs on companies that SOX is intended to benefit most.
4
changes may be initiated to manage earnings (e.g., Davidson et al. 2005; DeFond and Subramanyam 1998; Kim et al. 2003; Lu 2006). Arguably, if auditor changes are triggered for earnings management purpose, then given the effectiveness of SOX’s provisions, taking as a package, in preventing deceptive accounting practices and the empirical finding that switching costs usually outweigh the agency benefits of changing auditors (Blouin et al. 2007), companies should have weak incentive to switch their auditors. However, Glass Lewis’ research report indicates that more than 1,600 U.S. companies change their auditors in 2004, a 78% jump from 914 companies in 2003. More importantly, the total of 2,514 auditor changes during 2003 and 2004 represent more than one-fourth of the publicly listed companies in the U.S. (Williams 2005). Therefore, this high jump in auditor changes provides a fertile ground to explore the impacts of SOX on the association between auditor changes and earnings management.5
I focus on auditor dismissals (rather than resignations) for two reasons. First, because dismissals represent companies’ voluntary switching of their auditors, the association between auditor changes and earnings management shall be stronger. Second, there are four distinct types of auditor dismissals occurred in practice: changes within Big4, changes within non-Big4, upgrade from non-Big4 to Big 4, and downgrade from Big4 to non-Big4. Therefore, I can evaluate the change in a non-Big 4’s audit quality before and after SOX by examining whether there are significant differences in earnings management measures when this non-Big 4 becomes the successor auditor under the “downgrade” and “changes within non-Big 4” scenarios. The “downgrade” scenario is of particular importance to the regulators and policy makers not only because of the substantial increase of downward auditor dismissals in the post-SOX period, but also because of managers’ weaker incentives to manage earnings due to the numerous provisions rules in SOX (Cohen et al. 2005; Lobo and Zhou 2006). These two counterbalancing forces make the association between auditor dismissals and earnings management not as predictable and straightforward as in prior studies that use samples from the pre-SOX period.
5
Several recent working papers have examined the association between auditor changes and audit fees before and after the SOX (e.g., Griffin and Lont 2005; Ho and Wang 2006). To the best of our knowledge, our study is the first one that investigates the impact of SOX on the association between auditor changes and managers’ earnings management behavior.
5
I use a comprehensive sample of 3,373 auditor dismissals between fiscal years 2001 and 2007, during which auditor dismissals occurred before and after SOX are both included. Descriptive statistics show that downgrade and changes within the non-Big 4 account for 31.38 and 46.30 percents of all auditor dismissals after SOX. In addition, auditor dismissal companies are smaller, less profitable, having more debts and higher growth opportunities, engaging in more merger and acquisition activity, and suffering lower operating cash flows than companies that do not dismiss auditors. Notably, auditor dismissals appear to be motivated by opinion shopping rather than audit fee reduction.
To control for potential self-selection bias of auditor dismissals, I adopt Heckman’s (1979) two-stage estimation procedure. I begin the analyses by using a single dummy variable to proxy for the occurrence of auditor dismissals. The empirical results indicate that companies dismiss their auditors before SOX so that they can use both discretionary accruals and real earnings activities to manage earnings. Importantly, auditor dismissal companies appear to shift their earnings management method from traditional accruals in the pre-SOX period to real manipulation activities in the post-SOX period. I then decompose auditor dismissals into four categories (i.e., upgrade, downgrade, within Big 4, and within non-Big 4) to test whether non-Big 4’s audit quality increase after SOX. The empirical results show that, for downgrade and changes within the non-Big 4, there is a significant association between auditor dismissals and earnings management only in the pre-SOX period. After the passage of SOX, however, non-Big 4 successor auditors show the greatest ability to mitigate companies’ artificial and real earnings management. These conclusions remain valid when I focus on downgrade and changes within non-Big 4 samples separately.
The remainder of this paper proceeds as follows. Section 2 describes the sample selection procedures and research design. Section 3 reports the empirical results and discusses their implications. The paper concludes with a summary of findings in Section 4.
2. RESEARCH DESIGN 2.1 Measurements of Earnings Management:
6
I first use the traditional discretionary accruals to proxy for managers’ earning management. In Cohen et al.’s (2008) term, this is called the artificial earnings management activities. The procedures for estimating the discretionary accruals are as follows. First, I follow Hribar and Collins (2002) by adopting the direct approach to compute the total accruals (TACC). That is, the total accruals equal income before extraordinary items (Compustat item #123) less operating cash flows adjusted for discontinued operations and extraordinary items (#308 #124). Second, I estimate the modified Jones model (1995) on a cross-sectional basis for each Fama and French (1997) industry with 20 or more firms in year t:
, / , 1 1 1( , / , 1 , / , 1) 1( , / , 1)
i t i t i t i t i t i t i t i t
TACC A SALES A REC TA PPE TA , (1)
where
TACC = Operating income less operating cash flows;
SALES = Change in sales from the previous year to the current year;
REC = Change in accounts receivable from the beginning to the end of the year;
PPE = Year-end property, plant and equipment; TA = Total Assets at the end of year t-1;
= the residual term.
Third, I compute the performance-adjusted discretionary accruals (denoted by DA) based on Cahan and Zhang (2006), an alternative approach to control for companies’ performance effect. That is, for each Fama and French (1997), I divide the sample into deciles based on sample companies’ return on assets (ROA). I then adjust each discretionary accrual estimated from Equation (1) by subtracting the median discretionary accruals for the firm’s industry-ROA deciles. Note that, since income-increasing accruals are often used to inflate current earnings and income-decreasing accruals are often used to create cookie jar reserves for future earnings, the distortion in earnings resulting from inappropriate income-increasing and/or income-decreasing accruals is as important as the magnitude of accruals to regulators and investors (Frankel, et al. 2002; Myers et al. 2003). Therefore, I use income-increasing, income-decreasing, and the absolute value of DA to measure the individual and combined effects of managers’ earnings management decisions.
7
Managers may have employed real activities to manipulate earnings numbers as well (Cohen et al. 2008; Roychowdhury 2006). Roychowdhury (2006) identifies three major real manipulation activities that are relatively free of the effects of pure accrual manipulation: (1) accelerate the timing of sales and/or generate additional unsustainable sales through increased price discounts or more lenient credit terms, (2) reduce discretionary expenditures to report higher margins, and (3) overproduce or increase production to report lower cost of goods sold. To detect these real earnings management activities, I use the proxy developed by Roychowdhury (2006), the abnormal production costs (denoted by RPROD), to measure real earnings management. To compute RPROD, I first estimate the following cross-sectional regressions for each Fama and French (1997) industry and year:
, / , 1 0 1(1/ , 1) 2( , / , 1) 3( , / , 1) 4( , 1/ , 1) ,
i t i t i t i t i t i t i t i t i t i t
RPROD A A SALE A SALE A SALE A (2)
where RPROD denotes the production costs in year t, which is defined as the sum of the cost of goods sold and the change in inventories. The abnormal production costs are computed as the difference between the actual values and the normal levels predicted from equations (2). To facilitate the comparisons between DA and
RPROD, I also examine the positive, negative, and absolute values of RPROD in the analyses.
2.2 Auditor Choice Model
Because an auditor change decision is usually endogenously determined by the managers (Kim et al. 2003; Cahan and Zhang 2006), I adopt Heckman’s (1979) two-stage estimation procedure to control for the self-selection bias. In the first stage, I estimate the following probit model of voluntary auditor changes:
, 0 1 , 2 , 3 , 4 , 5 ,
6 , 7 , 8 , 9 ,
& ,
i t i t i t i t i t i t
i t i t i t i t
CHANGE SIZE MB LEVERAGE ROA LOSS
GC FEE INDSHARE M A (3) where
CHANGE = 1 if a company changes its auditor in year t and 0 otherwise; SIZE = Natural log of total assets at end of year t;
MB = Market value to book value of equity; LEVERAGE = Total debt divided by total assets;
8
ROA = Return on assets, defined as net income before extraordinary items divided by total
assets;
LOSS = 1 if operating income is less than 0 in year t and 0 otherwise;
GC = 1 if the company receives a going concern opinion in the preceding one year and 0
otherwise;
FEE = Audit fee divided by total fees
INDSHARE = Auditor’s market share in the client’s industry, based on the percentage of the
square root of total assets that the auditor audits for all companies in the client’s industry;
M&A = 1 if the company experiences a merger or acquisition in the preceding two years and 0
otherwise; = the residual term.
In estimating Equation (3), the dependent variable, CHANGE, is a dummy variable which equals one if a company changes its auditor during the sample period and zero otherwise. The independent variables include major determinants of voluntary auditor changes documented in prior studies. For example, Francis and Wilson (1988) and Krishnan (1994) find that the costs of changing auditors are higher for larger-size companies. Therefore, they are less likely to dismiss their auditors. I measure a company’s size by the natural log of its total assets (denoted by SIZE) and predict its coefficient to be negative. In addition, Woo and Koh (2001) reports evidence that growing companies are more likely to switch their auditors. Thus, I use the market-to-book ratio (denoted by MB) to control for companies’ growth opportunity and predict the coefficient of MB to be positive.
Since the financial condition of a company usually affects the likelihood that it will change auditor (Krishnan and Stephens 1995), Ie include three measures to proxy for a company’s financial condition: ROA,
LOSS, and LEVERAGE. I predict that the coefficient of ROA (or LOSS) to be negative (or positive) because
profitable (or unprofitable) companies are less (or more) likely to be financially-distressed. Similarly, I expect the coefficient of LEVERAGE to be negative because higher debt levels increase the possibility of financial difficulty.
9
change their auditors (e.g., Chow and Rice 1982; Geiger et al. 1998). Hence, I include going concern opinion as an indicator variable (denoted by CG) and predict its coefficient to be positive. Moreover, companies may have incentives to switch their auditors to reduce audit fees (Schwartz and Menon 1985) because prior research indicates that audit fees tend to be lower in the year of an auditor change than in the prior year (Deis and Giroux 1996). This audit fee-reduction incentive could be even stronger after the SOX because Section 404 requires the auditors to attest on companies’ internal controls over financial reporting during annual audits. As Clark (2005) points out that, on average, public companies have to pay Section 404-type fees to their auditors that are 50 to 100 percent as large as the regular audit fees. Therefore, I control for the possible effect of higher audit fees on companies’ auditor change decision by including a fee ratio variable (denoted by FEE), which is measured by audit fee divided by total fees and predict that companies with higher audit fees in the prior year are more likely to switch their auditors to reduce their audit fee payments.
Auditing studies have documented that companies are less likely to dismiss their auditors who are industry experts or specialists (Williams 1988; Carcello and Neal 2003). I follow Carcello and Neal (2003) by using auditor’s industry share (INDSHARE) as the proxy for industry specialization, which is measured by the percentage of the square root of total assets that the auditor audits for all companies in the client’s industry. Finally, empirical evidence has reported that mergers and acquisitions usually lead to auditor changes (Landsman et al. 2005). Accordingly, I employ M&A as an indicator variable for merger and acquisition activities. Following Collins and Hribar (2002) and Myers et al. (2003), I use Compustat footnote code 1 to identify companies undergoing mergers and acquisitions in the preceding two years before auditor changes.
2.3 Earnings Management Models:
I estimate two earnings management models at the second stage to compare companies with and without auditor dismissals. I first use a single dummy variable CHANGE to test the “overall” association between auditor dismissals and earnings management (to be discussed in section 2.3.1). I then decompose the CHANGE variable into four categories (i.e., upgrade, downgrade, within Big 4, and within non-Big 4) to test whether the association between auditor changes and earnings management vary with the type of auditor switching (to be
10 discussed in section 2.3.2).
2.3.1 Models to test the association between auditor changes and earnings management
In the following equation (4), the dependent variable is measured by DA or RPROD:
, , 0 1 , 2 , 3 , 4 , 5 , 6 , , , 9 , 10 , 7 8 or i t i t i t i t i t i t i t i t i t i t i t i t
DA RPROD SIZE MB LEVERAGE OCF CHANGE SOX
SOX CHANGE LEADER SHAREDECR SHAREINCR
11 , , Lambdai t (4) where
DA = Performance-adjusted discretionary accruals (measured in negative, positive, and
absolute) from equation (1);
RPROD =Abnormal production costs from Equation (2); SIZE = Natural log of total assets;
MB = Market value to book value of equity; LEVERAGE = Total debt dividend by total assets;
OCF = Cash flows from operating activity deflated by beginning total assets; CHANGE = 1 if a company changes audit firm in year t and 0 otherwise;
SOX= 1 for all firm-year observations in 2003 and latter and 0 for observations in 2002 and
2001;
SOXCHANGE = 1 if a company changes audit firm in the post-SOX period and 0 otherwise; LEADER = 1 if the successor (or incumbent) auditor’s industry expertise falls into the
classification presented in Hogan and Jeter (1999), and 0 otherwise;
SHAREDECR = 1 if the company has a decline of more than 10 percent of total outstanding
shares during the year and 0 otherwise;
SHAREINCR = 1 if the company has a increase of more than 10 percent of total outstanding
shares during the year and 0 otherwise;
Lambda = Inverse Mills ratio variable from the Equation (3) regression;
= the residual term.
To address the first research issue, I focus on CHANGE (which captures managers’ earnings management behavior for companies changing their auditors in the pre-SOX periods) and the interaction term
SOXCHANGE (which captures managers’ earnings management activities for companies changing their
11
I also include several control variables that have been found in prior studies to have significant impacts on managers’ earnings management decisions (DeFond and Jiambalvo 1994; Frankel et al. 2002; Matsumoto 2002). For example, I consider company size (denoted by SIZE), which is measured by the natural log of total assets, because larger companies generally face greater political costs and, therefore, have less flexibility and weaker incentives to overstate earnings (Watts and Zimmerman 1978). In addition, empirical evidence shows that managers use discretionary accruals to avoid the violation of debt convents (DeFond and Jiambalvo 1994; Dichev and Skinner 2002). Thus, I control for companies’ financial leverage (denoted by LEVERAGE) and expect its coefficient to be positive. Also, companies with growth opportunity have stronger incentives to avoid negative earnings surprises (Matsumoto 2002) or to have more discretion in terms of accounting choices (Smith and Watts 1992). Similar to Frankel et al. (2002), I use the market-to-book ratio (denoted by MB) to control for firms’ growth opportunity and expect a positive relation between MB and earnings management. Further, prior research has suggested that firms with strong operating cash flow (denoted by OCF) are less likely to employ discretionary accruals to boost earnings (Becker et al. 1998; DeFond and Park 1997). Hence, I include OCF to control for this effect and predict its coefficient to be negative.
To control for auditor’s industry leadership, I follow Hogan and Jeter (1999) by using an indicator variable
LEADER, which equals one if the successor auditor (or incumbent auditor for no-auditor change companies)
industry expertise falls into the classification presented in Hogan and Jeter (1999), and zero otherwise. Additionally, Rangan (1998) and Teoh et al. (1998) show that managers have incentives to use income-increasing discretionary accruals before seasoned equity offers. To take this effect into consideration, I include an indicator variable SHAREINCR and predict that larger increases in outstanding shares are associated with larger and more positive discretionary accruals. Moreover, Becker et al. (1998) shows that managers have incentive to reduce earnings using income-decreasing accruals before share repurchases. Therefore, I also include an indicator variable SHAREDECR to control for larger decreases in outstanding shares. Finally,
Lambda represents the inverse Mills ratios obtained from Equation (3).6
6
12
2.3.2 Models to test the audit quality of non-Big 4
To test whether non-Big 4’s audit quality increases, I decompose CHANGE into four categories: companies that switch from a non-Big 4 auditor and a Big 4 auditor (denoted by UP), companies that switch from a Big 4 auditor and a non-Big 4 auditor (denoted by DOWN), companies that switch auditors within the Big 4 (denoted by WINB4), and companies that switch auditors within the non-Big 4 (denoted by NWINB4). I then estimate the following equation (5): , , 0 1 , 2 , 3 , 4 , 5 , 6 , , 8 , 9 , 10 , , 11 , , 7 or 4 4 i t i t i t i t i t i t i t i t i t i t i t i t i t i t i t
DA RPROD SIZE MB LEVERAGE OCF UP DOWN
WINB NWINB SOX SOX UP SOX DOWN
12 , 13 , , 15 , 16 , 17 , 14 4 4 ,i t i t i t i t i t i t
SOX WINB SOX NWINB LEADER SHAREDECR
SHAREINCR Lambda (5) where
UP = 1 if a company with a non-Big 4 auditor switched to Big 4 in year t and 0 otherwise; DOWN = 1 if a company with a Big 4 auditor switched to a non-Big 4 in year t and 0
otherwise;
WINB4 = 1 if a company with a Big 4 auditor switched to a Big 4 in year t and 0 otherwise; NWINB4 = 1 if a company with a non-Big 4 auditor switched to a non-Big 4 in year t and 0
otherwise.
SOXUP = 1 if a company with a non-Big 4 auditor switched to Big 4 in the post-SOX period
and 0 otherwise;
SOXDOWN = 1 if a company with a Big 4 auditor switched to a non-Big 4 in the post-SOX
period and 0 otherwise;
SOXWINB4 = 1 if a company with a Big 4 auditor switched to a Big 4 in the post-SOX period
and 0 otherwise;
SOXNWINB4 = 1 if a company with a non-Big 4 auditor switched to a non-Big 4 in the
post-SOX period and 0 otherwise.
All other variables are the same as those defined in equation (4).
2.4 Models to Test the Audit Quality in the “Downgrade” and “Changes within non-Big 4” Samples:
In the first stage, I estimate the following probit model of downgrade auditor changes:
the second-stage regressions. I retain the entire sample for equation (3) and, as such, in essence treat the auditor change variable as endogenous. This specification is often referred to in the econometric literature as a “treatment effects” model (Green 2002, 787-789).
13 0 1 2 3 4 5 7 8 9 10 , , , , , , , , , , & , , / 4
i t i t i t i t i t i t i t i t i t i t i t
DOWN SIZE ROA LOSS
FEE INDSHARE M A NWINB MB LEVERAGE GC (6)
where DOWN (or NWINB4) equals one if an auditor change is classified as a downgrade change (or change within non-Big 4) and zero otherwise, and all other variables are the same as those defined in equations (4) and (5). I then include the Lambda estimated from equation (6) into the following equations (7) and (8):
, , 0 1 , 2 , 3 , 4 , 5 , 6 , , , 9 , 10 , 7 8 or i t i t i t i t i t i t i t i t i t i t i t i t
DA RPROD SIZE MB LEVERAGE OCF DOWN SOX
SOX DOWN LEADER SHAREDECR SHAREINCR
11Lambdai t, , (7) , , 0 1 , 2 , 3 , 4 , 5 , 6 , , , 9 , 10 , 7 8 or 4 4 i t i t i t i t i t i t i t i t i t i t i t i t
DA RPROD SIZE MB LEVERAGE OCF NWINB SOX
SOX NWINB LEADER SHAREDECR SHAREINCR
11 , , Lambdai t (8)
where all variables are the same as those defined in equations (4), (5), and (6).
3. EMPIRICAL RESULTS 3.1 Data and Sample Selection:
Our sample consists of auditor dismissals during fiscal year 2001 to 2007 collected in the Audit analytics database. I first use the Audit Analytics to identify companies that change their auditors during the sample periods. I classify each auditor change based on the identity of the predecessor and successor auditors. Next, I exclude all auditor dismissal cases in which Arthur Andersen was the predecessor auditor from our 2002 sample because these companies were forced to change auditors due to Andersen’s 5-year probation (Blouin et al. 2005). I obtain all financial information for both types of companies from the COMPUSTAT annual industrial and research files between 2001 and 2007. Further, I exclude financial institutions (SIC codes 6000-6999) because of its unique operating environment and differences in accounting classifications that make inferences difficult in subsequent analyses. I also restrict the sample to companies whose fiscal year ends on December 31 to make sample companies as homogenous as possible. The final sample consists of 3,373 auditor dismissals. Finally, to control for outlier problem, I follow Kothari et al. (2005) and Cahan and Zhang
14
(2006) by winsorizing observations that fall in the top and bottom 1 percent of the empirical distribution for both the dependent and independent variables.7 Table 1 reports the sample selection procedures.
[Insert Table 1 here]
Panel A of Table 2 shows the distribution of auditor dismissals by years and by types of change. While SOX does not impose mandatory CPA firm rotation, Table 1 indicates that the frequency of voluntary auditor dismissals increase substantially from 374 and 351 in 2001 and 2002, respectively, to 411 and 630 in 2003 and 2004, respectively. Auditor dismissals decrease gradually to 587, 545, and 475 in 2005, 2006, and 2007, respectively. Notably, the highest frequency of auditor dismissals occurred in 2004 (i.e., 18.68% of 3,373), which is the year right after the GAO (2003) report.
[Insert Table 2 here]
Panel A also indicates that downgrade auditor changes account for the highest portion (i.e., 38.06 percent) of all 3,373 auditor dismissals occurred during the sample periods, followed by auditor changes within the non-Big 4 (i.e., 36.19 percent). Upgrade auditor changes comprise only 5.97 percent in the sample. If we focus on the post-SOX periods (i.e., 2003~2007), Panel B of Table 2 reports that the frequency of downgrade auditor changes decreases to 31.38 percent while the frequency of auditor changes within the non-Big 4 increases to 46.30 percent. Overall, Table 2 documents that almost one-third of the auditor dismissals involve downgrade auditor changes in the post-SOX period.
3.2 Descriptive Statistics and Univariate Tests:
Table 3 presents the descriptive statistics of the sample, partitioned by companies without auditor changes (N = 44,941) and companies with auditor changes (N = 3,373). Several findings are worth noting. First, the means (medians) of the performance-adjusted discretionary accruals (DA) are 0.000 (0.000) for auditor-change companies and 0.025 (0.032) for no-auditor change companies, respectively. The differences are significant at the 0.01 level based on the t tests and Mann-Whitney z tests. Similarly, the means (medians) of RPROD are
7
I also trim the observations that fall in the top and bottom 1 percent of the empirical distribution. The OLS results remain unchanged. Therefore, the empirical findings are not sensitive to the way I handle the outliers.
15
0.233 (0.210) for auditor-change companies and 0.126 (0.029) for no-auditor change companies, respectively. The differences are also significant at the 0.01 level based on the t tests and Mann-Whitney z tests. These results imply that companies with auditor changes have significantly larger discretionary accruals and bigger abnormal production costs than companies without auditor changes, suggesting that companies involving auditor dismissals are more likely to engage in earnings management.
[Insert Table 3 here]
Second, I find evidence that auditor dismissal companies are smaller (SIZE), having more debts (LEVERAGE) and higher growth opportunities (MB), engaging in more merger and acquisition activity (M&A), and suffering lower operating cash flows (OCF) and more operating losses (LOSS) than control sample that do not dismiss auditors. These findings indicate that auditor dismissal companies are in significantly worse financial conditions than no-auditor dismissal companies. I also find that auditor dismissal companies have higher audit fees to total fees ratio (FEE) and are more likely to have received a going concern opinion from their predecessor auditors (GC), implying that auditor dismissals appear to be motivated by opinion shopping (Chow and Rice 1982 and Geiger et al. 1998) rather than audit fee reduction. Finally, auditor dismissal companies have larger decrease and increase in outstanding shares (SHAREDECR and SHAREINCR), suggesting that these companies are more likely to use income-increasing and income-decreasing accruals before seasoned equity offerings and share repurchase, respectively. Interestingly, while companies with and without auditor dismissals appear to hire incumbent auditors that are industry experts (INDSHARE), auditor dismissal companies are more willing to hire industry experts as their successor auditors (LEADER).
3.3 Multivariate Analysis of Auditor Changes:8
The first column of Table 4 reports the results of the first-stage auditor change model. This column indicates that companies having certain characteristics, as have been documented in Table 3, are more likely to dismissal their incumbent auditors. Notably, the significance of the coefficient of GC supports my conjecture that auditor dismissals are motivated by opinion shopping.
8
16
[Insert Table 4 here]
The second column of Table 4 reports the results of the second-stage model using traditional discretionary accruals as the measure of earnings management. The coefficients of CHANGE are significantly positive for absolute DA (coefficient 0.045, t = 4.68, two-tailed p < 0.000) and positive DA (coefficient 0.184, t = 6.36, two-tailed p < 0.000), and significantly negative for negative DA (coefficient 0.029, t = 4.25, two-tailed p < 0.000). That is, in the pre-SOX period, auditor dismissal companies have significantly more income-increasing and income-decreasing discretionary accruals than companies that do not dismiss their auditors. This finding implies that, taken as a whole, auditor dismissal companies change their auditors before SOX to engage in earnings management. In contrast, the coefficients of SOXCHANGE are only marginally significant for three
DA measures.9 More importantly, the signs of these coefficients reverse. Two possible SOX effects drive this
result: a direct effect in which SOX, taking as a package, mitigates auditor dismissal companies’ earnings management, and an indirect effect in which successor auditors suppress auditor dismissal companies’ earnings management due to SOX. As shown in the second column of Table 4, the signs and significance of the three coefficients of SOX support the direct effect. To examine whether the indirect effect also exists, a partition of auditor dismissal types is necessary. I will address this issue in the next section.
The third column of Table 4 reports the results of the second-stage model using abnormal production costs as the measure of earnings management. The coefficients of CHANGE are significant for positive RPROD (coefficient 0.288, t = 11.27, two-tailed p < 0.000) and negative RPROD (coefficient 0.077, t = 2.93, two-tailed p < 0.000), indicating that companies dismiss their auditors before SOX so that they can use both discretionary accruals and real earnings activities to manage earnings. Note that the coefficients of SOX are significant for positive RPROD (coefficient 0.031, t = 3.99, two-tailed p < 0.000) and negative RPROD (coefficient 0.032, t = 4.16, two-tailed p < 0.000), implying that SOX, taking as a package, is not effective in mitigating auditor dismissal companies’ use of real earnings activities to distort earnings. This finding, together with the existence of SOX’s direct effect reported in the second column, suggests that auditor dismissal
17
companies appear to shift their earnings management method from traditional accounting accruals in the pre-SOX period to real manipulation activities in the post-SOX period. This is consistent with recent trend that companies’ top management has become more conservative by reducing artificial earnings management to avoid penalty and fine generating from the CEO/CFO certification requirement in SOX Section 302 (Lobo and Zhou 2006). Real earnings management activities are more prevalent because they are generally more difficult to detect and, therefore, expose the managers to lesser liabilities (Cohen et al. 2005).
Since SOX is ineffective in mitigating real manipulation activities, the significantly negative coefficient of
SOXCHANGE (i.e., 0.076, t = 2.78, two-tailed p < 0.000) manifests the existence of SOX’s indirect effect.
Again, I will address this issue in the next section. Finally, the coefficients of Lambda are all significant at the 1% significance level (except for the absolute DA), indicating that I have successfully controlled for the self-selection bias.
3.4 Auditor Change Types and Earnings Management:
Table 4 shows that companies dismiss their auditors before SOX to engaging in earnings management using both discretionary accruals and abnormal production costs. A follow-up question would be: which type of successor auditors “connive” such earnings management behavior? Also, the existence and extent of SOX’s indirect effect on audit quality deserve further exploration. In this section, I decompose the CHANGE variable into four category dummies (i.e., upgrade, downgrade, within Big 4, and within non-Big 4) to examine whether the association between auditor dismissals and earnings management vary with types of auditor switching.
The first three column of Table 5 reports that, before SOX, the Big 4 effectively suppress companies’ artificial earnings management in auditor dismissals involving upgrade (coefficient of UP for negative DA is 0.053, t = 2.52, two-tailed p < 0.012) and changes within Big 4 (coefficient of WINB4 for positive DA is 0.115,
t = 2.13, two-tailed p < 0.033). Also, the last column of Table 5 shows that the Big 4 successfully mitigate
companies’ negative real earnings management for upgrade auditor dismissals (coefficient 0.223, t = 4.31, two-tailed p < 0.000). Different from the Big 4’s results, the signs and significance of the coefficients of
18
levels of discretionary accruals and abnormal production costs than non-auditor dismissal companies in the pre-SOX period. These imply that it is those auditor dismissals involving changes within non-Big 4 that are associated with earnings management, leading to the significance of CHANGE in Table 4. Given audit quality can be captured by measures of earnings management (Ashbaugh et al. 2003; Becker et al. 1998; Frankel et al. 2002), these results suggest that, before SOX, audit quality of the non-Big 4 is significantly lower than that of the Big 4. This finding is consistent with the well-documented empirical research that uses the Big N vs. non-Big N dichotomy to proxy for audit quality (e.g., DeAngelo 1981; Mansi et al. 2005; Palmrose 1988).
[Insert Table 5 here]
Contrary to the pre-SOX finding, the signs of the coefficients of SOXNWINB4 reverse and are significant for absolute (coefficient 0.057, t = 3.37, two-tailed p < 0.000), positive (coefficient 0.305, t = 6.92, two-tailed p < 0.000), and negative DA (coefficient 0.062, t = 4.74, two-tailed p < 0.000). The same results can be found for positive (coefficient 0.415, t = 10.01, two-tailed p < 0.000) and negative RPROD (coefficient 0.059, t = 6.66, two-tailed p < 0.000). Comparing these results with the coefficients of NWINB4 across two earnings management measures indicates that it is those auditor dismissals involving changes within non-Big 4 that drive the strongest indirect effect of SOX, leading to significant coefficients of SOXCHANGE in Table 4.
Two findings related to the RPROD measures are worth noting. First, only for negative RPROD that the coefficients of UP (i.e., 0.223, t = 4.31, two-tailed p < 0.000), WINB4 (i.e., 0.388, t = 8.89, two-tailed p < 0.000), and SOXWINB4 (i.e., 0.139, t = 2.88, two-tailed p < 0.000) are significant. These results imply that the Big 4 are able to detect and suppress companies’ negative abnormal production costs when there are upgrade and within-Big 4 auditor changes in the pre-SOX and post-SOX periods, respectively. Second, for positive
RPROD, the coefficients of DOWN (i.e., 0.0.097, t = 2.05, two-tailed p < 0.041) and SOXDOWN (i.e., 0.124, t
= 2.42, two-tailed p < 0.016) are both positive and significant. These results suggest that the non-Big 4 appear to allow companies’ positive abnormal production costs when there are downgrade auditor changes before and after SOX.
19
between auditor dismissals and earnings management only for companies that change their auditors within the non-Big 4 in the pre-SOX period. After the passage of SOX, however, non-Big 4 successor auditors show the greatest ability to mitigate companies’ artificial and real earnings management when these companies dismiss their non-Big 4 predecessor auditors. These findings have not been documented in prior literature in auditor changes and earnings management.
3.5 “Downgrade” and “Within Non-Big 4” Auditor Changes and Earnings Management:
Table 2 reveals that there are surprisingly large numbers of auditor dismissals involving downgrade (account for 31.38%) and within non-Big 4 (account for 46.30%) in the post-SOX periods. To further explore whether these prevailing “downgrade” and “within non-Big 4” auditor dismissals are driven for earnings management purpose, I compare the “downgrade” / “within non-Big 4” group with the other three groups using the same statistical analysis procedures used in previous sections.
3.5.1 “Downgrade” auditor changes and earnings management
The first column of Table 6 presents the results of first-stage probit regression, showing that downgrade change companies have higher profitability (ROA and LOSS) and less merger and acquisition activities (M&A), and are more likely to dismiss auditors for audit fee reduction (FEE) rather than opinion shopping.
[Insert Table 6 here]
The second column of Table 6 indicates that the coefficients of DOWN are significantly positive for positive DA (i.e., 0.180, t = 1.34, one-tailed p < 0.090) and significantly negative for negative DA (i.e., 0.039,
t = 1.80, two-tailed p < 0.072), indicating that downgrade auditor change companies have significantly higher
(lower) level of income-increasing (income-decreasing) DA than other three types of auditor change companies in the pre-SOX period. The third column of Table 6 also shows that the coefficient of DOWN is significantly positive for positive RPROD (i.e., 0.316, t = 2.37, two-tailed p < 0.018). These results imply that downgrade auditor changes are associated with earnings management before SOX.
In contrast, the signs of the coefficients of SOXDOWN reverse and are significant for absolute (coefficient 0.052, t = 1.44, one-tailed p < 0.075), positive (coefficient 0.164, t = 1.34, one-tailed p <
20
0.090), and negative DA (coefficient 0.029, t = 1.51, one-tailed p < 0.066). The same results can be found for positive (coefficient 0.231, t = 1.89, two-tailed p < 0.059) and negative RPROD (coefficient 0.274, t = 3.23, two-tailed p < 0.000). These results suggest that non-Big 4 successor auditors are able to suppress artificial and real earnings manipulations done by companies that dismiss their Big 4 predecessor auditors.
Two possibilities may explain the above findings. First, the non-Big 4 are concerned that a “downgrade” auditor change may attract market participants’ attention on it and, therefore, react more conservatively. Second, the passage of SOX motivates the non-Big 4 to improve their audit quality to a higher level. Since Table 5 has shown that auditor changes within non-Big 4 drive the strongest indirect effect of SOX, it appears that the second possibility is more feasible to explain downgrade auditor changes.
3.5.2 “Within non-Big 4” auditor changes and earnings management
The first column of Table 7 indicates that companies that change their auditors within non-Big 4 are smaller in size (SIZE), having more debts (LEVERAGE), less profitable (ROA and LOSS), and are more likely to dismiss auditors for opinion shopping (GC).
[Insert Table 7 here]
Similar to the results documented in Table 5, Tables 7 shows that the signs and significance of the coefficients of NWINB4 support my previous finding that auditor dismissals involving changes within non-Big 4 are associated with earnings management in the pre-SOX period. Conversely, SOX’s provisions induce the non-Big 4 to improve their audit quality, leading to significant reductions in artificial and real earnings management in the post-SOX period.
3.6 Sensitivity Analyses: (results are not tabulated)
3.6.1 Alternative measure of discretionary accruals
To ensure that the empirical results are not sensitive to the choice of discretionary accruals measures, I also calculate performance-adjusted discretionary accruals based on two-digit SIC codes, years, and lagged ROA. Moreover, I estimate performance-adjusted discretionary accruals by including current and lagged ROA in the modified Jones model. The empirical results remain the same under these alternative procedures.
21
The statistical tests conducts in the preceding sections assume that financial statements of firms with fiscal years ending in or after 2003 are subject to the jurisdiction of SOX. To examine the sensitivity of the results to this assumption, I exclude year 2003 and redo all the analyses. The empirical results are relatively insensitive to this alternative classification.
3.6.3 Matched-pairs analyses
I compare the relation between auditor changes and earnings management using another set of control firms. As an additional analysis, each auditor change company is matched with a non-change company based on year, industry, and size. The matched firms are chosen from the COMPUSTAT in the same year and two-digit SIC as the auditor change companies. In addition, each non-change company is within 30 percent of the total assets amount for corresponding auditor change companies10. For matched-pairs sample, I also
conduct level and changes analysis after controlling self-selection bias. The results of these analyses are similar to those reported earlier.
4. SUMMARY AND CONCLUSIONS
Professional institutions and the public press have reported a dramatically higher increase of auditor changes in the post-SOX periods than in the pre-SOX periods (Jean 2004; Williams 2005; Yoon 2004). However, few attempts have been made to explore the possible reasons underlying this increase. Following the framework of earnings management literature, I focus on auditor dismissals and posit that the non-Big 4’s audit quality has increased after SOX.
A comprehensive sample of 3,373 auditor dismissals between fiscal years 2001 and 2007 is collected and analyzed. Descriptive statistics show that downgrade and changes within the non-Big 4 account for 31.38 and 46.30 percents of all auditor dismissals after SOX. Auditor dismissal companies are generally smaller, less profitable, having more debts and higher growth opportunities, engaging in more merger and acquisition activity, and suffering lower operating cash flows than companies that do not dismiss auditors. To control for
10
There is no difference in the mean size between auditor change firms (measured by total assets) and non-change firms, suggesting that I have successfully matched on firm size. To conduct a model that is consistent with the model in the full sample analyses, I include firm size as a control variable in the matched sample analyses. If I exclude the firm size from matched sample analysis, I obtain the similar results.
22
potential self-selection bias resulting from auditor dismissals, I adopt Heckman’s (1979) two-stage estimation procedure. Two important findings are documented. First, companies dismiss their auditors before SOX so that they can use both discretionary accruals and real earnings activities to manage earnings. More importantly, auditor dismissal companies appear to shift their earnings management method from traditional accruals in the pre-SOX period to real manipulation activities in the post-SOX period. Second, after decomposing auditor dismissals into four categories, the empirical results show that, for downgrade and changes within the non-Big 4, earning management occurs only in the pre-SOX period. After SOX, non-Big 4 successor auditors show the greatest ability to mitigate companies’ artificial and real earnings management. These conclusions remain valid when downgrade and changes within non-Big 4 samples are examined separately.
Some features of the study point to several directions for future research and caveats. First, further work is warranted on testing how the capital market interprets the dramatic increase of downward auditor changes in the post-SOX periods. To the extent that the capital participants can really “monitor” these downgrade auditor change companies, companies’ opportunistic behavior shall be mitigated. Second, reasons other than earnings management may also explain the remarkable increase of auditor changes in the post-SOX periods. An understanding of these other reasons may provide the regulators with insights into the effectiveness of SOX in improving audit quality. Finally, I focus on auditor dismissals. Therefore, the empirical results could not be applied to explain the huge increase in auditor resignations brought by the Big 4 vs. non-Big 4 auditors. Since SOX has changed the legal environment imposed on the auditing profession, more studies are needed to further investigate resignation decisions made by CPA firms with differential audit quality in the post-SOX periods.
23
REFERENCE
American Institute of Certified Public Accountants, Commission on Auditors’ Responsibilities: Report Conclusions, and Recommendations (New York: AICPA 1978).
Ashbaugh, H., R. LaFond, and B.W. Mayhew. 2003. Do nonaudit services compromise auditor independence? Further evidence. The Accounting Review 78 (July): 611-640.
Becker, C., M. L. DeFond, J. Jiambalvo, and K. R. Subramanyam. 1998. The effect of audit quality on earnings management. Contemporary Accounting Research 15 (Spring): 1-24.
Behn, B.K., J.H. Choi, and T. Kang. 2008. Audit quality and properties of analysts to reduce earnings forecasts.
The Accounting Review 83 (March): 327-350.
Betty, R. 1989. Auditor reputation and the pricing of initial public offerings. The Accounting Review 64 (October): 693-709.
Blouin, J., B. Grein, and B. R. Rountree. 2005. An analysis of forced auditor rotation: The case of former Arthur Andersen clients. Working paper, University of Pennsylvania.
Cahan, S. F. and W. Zhang. After Enron: Auditor conservatism and ex-Andersen clients. The Accounting
Review 81 (January): 49-82.
Carcello, J. V. and T. L. Neal. 2003. Audit committee characteristics and auditor dismissals following “new” going-concern reports. The Accounting Review (January): 95-117.
Chaney, P.K. and K.L. Philipich. 2002. Shredded reputation: The cost of audit failure. Journal of Accounting
Research 40 (September): 1221-1245.
Chang, H.H., C.S. Agnes Cheng, and K.J. Reichelt. Market reaction to auditor switching from Big Four to small audit firms. Working Paper. Louisiana State University.
Chow, C. and S. Rice.1982. Qualified audit opinions and auditor switching. The accounting Review 1 (April): 326-335.
Clark, R. C. 2005. Corporate governance changes in the wake of the Sarbanes-Oxley Act: A morality tale for policymakers too. Harvard Law and Economics Discussion Paper: http://www.law.harvard.edu /programs/ olin_center/.
Cohen, D., A. Dey, and T. Lys. 2008. Real and accrual-based earnings management in the pre- and post-Sarbanes-Oxley periods. The Accounting Review 83 (May): 757-788.
Craswell, A. T., J. R. Francis and S. L. Taylor.1995. Auditor brand name reputations and industry specializations.
Journal of Accounting and Economics 20 (December): 297-322.
Craswell A. T., and J. R. Francis. 1999. Pricing initial audit engagements: a test of competing theories. The
Accounting Review 74 (Arial): 201-216.
Davidson, R. A., and D. Neu. 1993. A note on the association between audit firm size and audit quality.
Contemporary Accounting Research 9 (Spring): 479-488.
Davidson, W. N. III, P. Jiraporn and P. DaDalt. 2005. Causes and consequences of audit shopping: An analysis of auditor opinions, earnings management, and auditor changes. Working paper, Southern Illinois University.
DeAngelo, L.E. 1981. Auditor size and audit quality. Journal of Accounting and Economics 3 (1): 183-199. Dechow, P., R. G. Sloan, and A. P. Sweeney. 1995. Detecting earnings management. The Accounting Review 70
24
DeFond, M. L., and J. Jiambalvo. 1991. Incidence and circumstances of accounting errors. The Accounting
Review 66 (July): 643-655.
DeFond, M. L., and J. Jiambalvo. 1994. Debt covenant violation and manipulation of accruals. Journal of
Accounting and Economics 25 (February): 35-67.
DeFond, M. L., and C. W., Park. 1997. Smoothing income in anticipation of future earnings. Journal of
Accounting and Economics 23 (June): 115-139.
DeFond, M. L., and K. R. Subramanyam.1998. Auditor changes and discretionary accruals. Journal of
Accounting and Economics 25 (February): 35-67.
DeFond, M. L., and J. R. Francis. 2005. Audit research after Sarbanes-Oxley. Auditing: A Journal of Practice &
Theory 24 (supplement): 5-30.
DeFond, M.L., Hung, M., Karaoglu, E., Zhang, J., 2007. Was the Sarbanes-Oxley Act good News for corporate bondholders? Working paper, University of Southern California.
Deis D. R. Jr, Giroux G. A. 1996. The effect of auditor changes on audit fees, audit hours, and audit quality.
Journal of Accounting and Public Policy 15 (Spring): 55-76.
Dichev, I. D. and D. J. Skinner. 2002. Large-sample evidence on the debt covenant hypothesis. Journal of
Accounting Research 40 (September): 1091-1123.
Engel, E., Hayes, R.M., Wang, X., 2007. The Sarbanes-Oxley Act and firms’ going-private decisions. Journal of
Accounting and Economics 44: 116-145.
Fama, E. F., and D. R. French. 1997. Industry costs of equity. Journal of Financial Economics 43 (February): 153-193.
Francis, J. R., and E. R. Wilson. 1988. Auditor changes: A joint test of theories relating to agency costs and auditor differentiation. The Accounting Review 63 (October): 663-682.
Francis, J.R., and J. Krishnan. 1999. Accounting accruals and auditor reporting conservatism. Contemporary
Accounting Research 16 (Spring): 135-165.
Francis, J. R., E. L. Maydew, and H. C. Sparks. 1999. The role of Big 6 auditors in the credible reporting of accruals. Auditing: A Journal of Practice & Theory 18 (Fall): 17-34.
Francis, J. R. 2004. What do we know about audit quality? The British Accounting Review 36 (December): 345-368.
Francis, J.R., K. Reichelt, and D. Wang. 2005. The Pricing of National and City-Specific Reputations for Industry Expertise in the U.S. Audit Market. The Accounting Review 80 (January): 113-36.
Francis, J. R., R. LaFond, P. Olsson, and K. Schipper. 2005. The market pricing of accruals quality. Journal of
Accounting Economics 39 (June): 295-327.
Frankel, R. M., M. F. Johnson, and K. K. Nelson. 2002. The relation between auditor fees for no audit services and earnings management. The Accounting Review 77 (Supplement): 71-105.
Geiger, M., K. Raghunandan, and D. V. Rama. 1998. Costs associated with going-concern modified audit opinions: An analysis of auditor changes, subsequent opinions, and client failures. Advances in
Accounting 16: 117-139.
Ghosh, A., and D. C. Moon. 2003. Does auditor tenure impair audit quality? Working paper, Office of Economic Analysis, Securities and Exchange Commission.
Greene, W. 2002. Econometric analysis. 5red. Upper Saddle River, NI: Prentice-Hall.
25
based on SEC disclosures under Sarbanes-Oxley. Working paper, University of California, Davis. Gul, F. A., C. J. P. Chen, and J. S. L. Tsui. 2003. Discretionary accounting accruals, managers’ incentives, and
audit fees. Contemporary Accounting Research 20 (Fall): 441-464.
Gullapalli, D. 2005. Number of firms that switched auditor jumped 78% in 2004. The Wall Street Journal (February, 17).
Heckman, J. 1979. Sample selection bias as a specification error. Econometrica 47: 153-161.
Ho, J. and J. Wang. 2006. Examination of audit fee premiums and auditor switching pre and post the demise of Arthur Andersen and the enactment of Sarbanes-Oxley Act. Working paper, University of California, Irvine.
Hribar, P., and D. W. Collins. 2002. Errors in estimating accruals: Implications for empirical research. Journal
of Accounting Research 40 (March): 105-134.
Jean, S. 2004. Overwhelmed accounting firms focus on big clients, leaving smaller ones behind. Saint Paul
Pioneer Press (October 19).
Jones, J. 1991. Earnings management during import relief investigations. Journal of Accounting Research 29 (Autumn): 193-228.
Johnson, V. E., I. K. Khurana, and J. K. Reynolds. 2002. Audit-firm tenure and the quality of financial reports.
Contemporary Accounting Research (Winter): 637-660.
Kim, J-B, R. Chung, and M. Firth. 2003. Auditor conservatism, asymmetric monitoring, and earnings management. Contemporary Accounting Research 20 (Summer): 323-359.
Kinney, W.R., Z-V. Palmrose, and S. Scholz. 2004. Auditor independence, non-audit services, and restatements: Was the U.S. government right? Journal of Accounting Research 42: 561-588.
Kothari, A. Leone, and C. E. Wasley. 2005. Performance matched discretionary accrual measures. Journal of
Accounting and Economics 39 (February): 173-197.
Krishnan, J. 1994. Auditor Switching and Conservatism. The Accounting Review 69 (January): 200-215.
Krishnan, J., and R. G. Stehpens. 1995. Evidence on opinion shopping from audit opinion conservatism. Journal
of Accounting and Public Policy 14 (Autumn): 179-201.
Khurana, I.K. and K.K. Raman. 2004. Litigation Risk and the Financial Reporting Credibility of Big 4 versus Non-Big 4 Audits: Evidence from Anglo-American Countries. The Accounting Review 79 (April): 473-95.
Landsman, W. R., K. K. Nelson, and B. R. Rountree. 2006. An empirical analysis of Big N auditor switches: Evidence from the pre- and post-Enron Eras. Working Paper, University of Florida.
Lennox, C. 2000. Do companies successfully engage in opinion-shopping?: Evidence from the UK. Journal of
Accounting and Economics 29 (June): 321-337.
Lobo, G. J., and J. Zhou. 2006. Did conservatism in financial reporting increase after the Sarbanes-Oxley Act? Initial evidence. Accounting Horizons 20 (March): 57–73.
Louis, H. 2005. Acquirers’ abnormal returns and the non-Big 4 auditor clientele effect. Journal of Accounting
and Economics 40: 75-99.
Lu, T. 2006. Does opinion shopping impair auditor independence and audit quality? Journal of Accounting
Research 44 (June): 561-583.
Mansi, S.A., W.F. Maxwell, and M.P. Darius. 2004. Does Auditor Quality and Tenure Matter to Investors? Evidence from the Bond Market. Journal of Accounting Research 42 (September): 755-793.
26
Matsumoto, D. A. 2002. Management’s incentives to avoid negative earnings surprise. The Accounting Review 77 (July): 483-514.
Myers, J., L. Myers, and T. Omer. 2003. Exploring the term of the auditor client relationship and the quality of earnings: A case for mandatory auditor rotation. The Accounting Review 78 (July): 779-799.
Oxley, M.G., 2007. The Sarbanes-Oxley Act of 2002—restoring investor confidence. Current Issues in Auditing 1: C1–C2.
Palmrose, Z. 1986. Litigation and independent auditors: The role of business failures and management fraud.
Auditing: A Journal of Practice and Theory 64 (Spring): 90-113.
Palmrose, Z. 1988. An analysis of auditor litigation and audit service quality. The Accounting Review 63 (January): 5573
Plitch, P. and L. L. Wei. 2004. Moving the market: Auditor-client breakups rise, while disclosure often langs.
The Wall Street Journal (August, 3).
Pittman, J. A. and S. Fortin. 2004. Auditor choice and the cost of debt capital for newly public firms. Journal of
Accounting and Economics 37 (February): 113-136.
Rangan, S. 1998. Earnings management preceding management buyout offers. Journal of Financial Economics 50 (October): 101-122.
Roychowdhury, S. 2006. Earnings management through real activities manipulation. Journal of Accounting and
Economics 42 (December): 335-370.
Schwartz, K. B., and K. Menon. 1985. Auditor switches by failing firms. The Accounting Review 60 (April): 248-261.
Teoh, S. H., and T. J. Wong. 1993. Perceived auditor quality and the earnings response coefficient. The
Accounting Review 68 (April): 346-366.
Teoh, S. H., I. Welch, and T. J. Wong. 1998. Earnings management and the underperformance of seasoned equity offerings. Journal of Financial Economics 50 (October): 63–99.
U. S. General Accounting Office (GAO). 2003. Public Accounting Firms: Required Study on the Potential Effects of Mandatory Audit Firm Rotation. Washington, D.C.: Government Printing Office.
U.S. House of Representatives. 2002. The Sarbanes-Oxley Act of 2002. Public Law 107–204 [H. R. 3763]. Washington, D.C.: Government Printing Office.
Watkins, A., W. Hillison and S. Morecroft. 2004. Audit quality: A synthesis of theory and empirical evidence.
Journal of Accounting Literature 23:153-193.
Williams, D. D. 1988. The potential determinants of auditor change. Journal of Business Finance and
Accounting 15 (Summer): 243-256.
Williams, J. 2005. Auditor Turnover Gains Momentum in 2004. Yellow Card Trend Alert. Glass Lewis & Co., LLC.
Watts, R. L., and J. Zimmerman. 1978. Toward a positive theory of the determination of accounting standards.
The Accounting Review 53 (January): 112–134.
White, H., 1980. A heteroskedasticity-consistent covariance matrix estimator and a direct test for heteroskedasticity. Econometrica 48 (May): 817-838.
Woo, E-S. and H. C. Koh. 2001. Factors associated with auditor changes: a Singapore study. Accounting and
Business Research 31 (Summer): 133-144.
Yoon, L. 2004. Public companies left the largest auditors in droves last year in favor of smaller firms. CFO.com (February 11).
27
Zhang, X., 2007. Economic consequences of the Sarbanes-Oxley Act of 2002. Journal of Accounting and
28 TABLE 1
Sample Selection Procedure
All auditor dismissals during fiscal year 2001 to 2007 are collected from the Audit analytics database 9,791
Less: All auditor dismissal cases involving Arthur Andersen (2,377)
Less: Financial institutions (SIC codes 6000-6999) (2,169)
Less: Observations missing in the Compustat (1,872)
29 TABLE 2
Frequencies of Auditor Changes in the Sample (N = 3,373) By Year and Auditor Change Types
Sample Year Upgrade to Big 4 a Downgrade to non-Big 4 Switch within Big 4 Switch within non-Big 4 Total Number of Auditor Changes Percentage of Auditor Changes in the Sample
Panel A: Full Auditor Change Sample (Years 2001 ~ 2007)
2001 34 88 141 111 374 11.09% 2002 30 84 68 169 351 10.41% 2003 24 113 77 197 411 12.18% 2004 24 239 96 271 630 18.68% 2005 18 219 98 252 587 17.40% 2006 28 154 127 236 545 16.16% 2007 36 106 63 270 475 14.08% Total Number of Auditor Changes 194 1,003 670 1,506 3,373 Percentage of Auditor Changes in the Sample 5.97% 38.06% 19.78% 36.19%
Panel B: Auditor Change Sample in the Post-SOX period (Years 2003 ~ 2007)
Total Number of Auditor Changes 130 831 461 1,226 Percentage of Auditor Changes in the Sample 4.91% 31.38% 17.41% 46.30%
