CCRG Working Papers are circulated to stimulate discussion and comments. They have not been peer-reviewed. References in publication to CCRG Working Papers should be cleared with the author or authors.

Center for Corporate Reporting and GovernanceWorking Paper Series

Do Financial Restatements Lead to Auditor Changes?

Vivek Mande a and Myungsoo Son b

a White Nelson Diehl Evans Professor of Accounting • California State University, Fullertonb Associate Professor • California State University, Fullerton

Working Paper CCRG 2013-05http://business.fullerton.edu/Centers/ccrg/research.htm

Center for Corporate Reporting and GovernanceCalifornia State University, Fullerton

Department of AccountingFullerton, CA 92834-6848

Tel: (657) 278-4414Email: ccrg@fullerton.edu

1

1

Do Financial Restatements Lead to Auditor Changes?

Vivek Mande1 and Myungsoo Son2

February 2012

ABSTRACT

This paper examines whether financial restatements are associated with subsequent auditor changes. A financial restatement represents a breakdown in a company’s financial reporting but importantly also of its audit. We argue that in response to pressure from capital markets, restating firms will dismiss their auditors to increase audit quality and restore reputational capital lost when the restatements are announced to the investing public. Using a large sample of restatements and auditor changes we find that, consistent with our hypothesis, the likelihood of auditor-client realignments increases after firms announce restatements. As expected, we also find that the positive association between restatements and auditor turnovers is more pronounced when restatements are more severe and the quality of corporate governance is high. Finally, we find that stock market returns surrounding auditor changes increase as the severity of restatements increases. The last result supports the idea that stock markets have a positive view of auditor changes following restatements.

Keywords: restatement; auditor switch; dismissal; resignation; abnormal market returns; Data Availability: Data are publicly available from sources identified in the paper.

1 White Nelson Professor of Accounting, Mihaylo College of Business and Economics, California State University, Fullerton, Tel: (657)278-7659, Email:vmande@fullerton.edu. 2 Associate Professor of Accounting, Mihaylo College of Business and Economics, California State University, Fullerton, Tel: (657)278-2732, Email:mson@fullerton.edu.

2

2

Do Financial Restatements Lead to Auditor Changes?

INTRODUCTION

The purpose of this paper is to examine the association between restatements of

financial statements and subsequent auditor changes. We posit that a restatement

destabilizes the relationship between external auditors and their clients because

shareholders view a restatement, at least in part, as an audit failure (Liu et al. 2009;

Raghunandan et al. 2003; Turner 1999). Depending on the incentives and disincentives

of both parties, we argue that a termination of their relationship could occur.

According to Government Accountability Office (GAO 2002), the number of

restatements has not only been steadily increasing, but over the period 1997 to 2002

financial markets lost more than $100 billion in market capitalization due to the

restatements. In a follow-up study, the GAO (2006) documented that the number of

public companies restating financial statements grew from 3.7 percent in 2002 to 6.8

percent in 2005.

Restatements raise questions about management’s integrity, the adequacy of a

firm’s internal controls,3 the effectiveness of the audit committee, and also the external

auditor’s independence and audit quality (Gleason et al. 2008). Kinney et al. (2004),

among others, argue that a financial restatement is viewed by markets not only as a

failure in financial reporting by management, but importantly as an auditing failure.

Dismissing the external auditor following a restatement is a highly visible action

that companies can take to possibly restore market confidence and improve audit

oversight over the financial reporting process. Liu et al. (2009) find that shareholders of

3 A restatement of previously reported financial statements is an indication of a material weakness in internal controls (PCAOB 2007).

3

3

restating companies ratify resolutions against external auditors calling for their dismissal.

However, auditors can also be expected to assign blame for the restatements to their

clients and in some cases will sever ties to preserve the audit firm’s reputation and reduce

its litigation exposure (Feldmann et al. 2009).4 As a result auditor dismissals and

resignations, both, are likely to occur after companies announce financial restatements.

Using a large sample of firms, we find that restatements are an important

predictor of auditor changes in the following year. We find that as the severity of a

restatement increases, the likelihood of an auditor change also increases. Also, consistent

with a restatement constituting an audit failure, we find that restating firms having strong

governance are more likely to change auditors than other firms. These results hold after

controlling for other determinants of auditor changes including the amount of audit fees

paid, the level of auditor-conservatism, and client-firms’ incentives to look for other

auditors.

Similar to restatements, auditor changes are important events that are closely

scrutinized by markets. Because the underlying reasons for an auditor change are often

not publicly disclosed, empirically documenting reasons for client-auditor realignments

increases our understanding of these events. Our study suggests that the recent increases

in auditor changes may be partly attributable to the increases in financial restatements.

When an auditor change occurs (resignations and dismissals), there is generally a

negative stock market reaction to the change. This is because an auditor change often

signals the presence of weak internal controls and/or disagreements with the incumbent

auditor, and because the successive auditor is often of lower quality and prone to making

4Rather than resigning from an engagement, the audit firm, instead, may choose to increase audit fees to compensate for the enhanced risk associated with the restating firm (Feldmann et al. 2009).

4

4

mistakes in the first years of the engagement (Hackenbrack and Hogan 2002; Johnson et

al. 2002; Carcello and Nagy 2004).

However, an auditor change that is in response to an audit failure may not be

accompanied by a negative reaction from markets. To the extent that an auditor change

restores the company’s reputation damaged by a restatement and increases audit

oversight, investors could view the change positively. We explore this idea by examining

abnormal returns surrounding auditor changes of restating firms. We find a positive stock

market reaction which suggests that for restating firms the benefits exceed the generally

high costs associated with auditor changes. This result is consistent with a line of

research showing that firms improve their governance mechanisms following accounting

failures (Farber 2005; Srinivasan 2005; Desai et al. 2006; Wilson 2010).

The remainder of this study comprises four sections. The next section reviews the

relevant literature and develops our hypotheses. This is followed by a discussion of the

research design and the empirical results. The last section concludes the study.

LITERATURE REVIEW AND HYPOTHESES DEVELOPMENT

A restatement occurs when a company, either voluntarily or prompted by auditors

or regulators, revises previously reported financial information. The announcement of a

restatement is made in a press release or on a Form 8-K, and typically results in the filing

of an amended financial report on Forms 10-K/A or 10-Q/A.

Restatements constitute a public acknowledgement that the reported financial

statements are not consistent with generally accepted accounting principles (GAAP) and

represent the most visible evidence of improper accounting (Palmrose and Scholz 2004).

Financial restatements undermine investor confidence in financial reporting and reduce

5

5

market efficiency (SEC 2002). For this reason, restatements come under the scrutiny of

regulators, legislators and the media, and, have been the subject of two GAO studies.

Prior studies show that firms face adverse consequences following financial

restatements. For example, restating firms experience increases in their costs of capital

(Hribar and Jenkins 2004). Palmrose et al. (2004) report a strong negative stock market

reaction (-9.2% return) over a two-day window surrounding announcements of

restatements. The authors find that many restating firms are forced into bankruptcy

and/or are named as defendants in lawsuits filed by various groups including investors,

customers, suppliers and employees.

Prior research also documents that there are negative consequences arising from

restatements to companies’ management and boards. Desai et al. (2006) and Agrawal and

Cooper (2007) show that restating firms experience a higher turnover in top management

relative to other firms. Srinivasan (2005) finds that following a material restatement there

is a higher probability of turnover in the audit committee whose responsibility it is to

oversee the firm’s financial reporting process.

A financial restatement, however, constitutes not just a breakdown in a

company’s financial reporting process but importantly, also in its auditing. For example,

Stanley and DeZoort (2007) argue that financial restatements due to errors or fraud are de

facto auditing failures; Turner (1999), the former chief accountant of the Securities

Exchange Commission (SEC), notes that the SEC considers restatements as constituting

audit lapses; and Larcker and Richardson (2004) argue that a restatement that is the result

of earnings management by the firm constitutes an audit failure because the firm’s

external auditor did not prevent the deception.

6

6

Consistent with this, research finds that restatements lead investors to have a more

negative view about their firms’ auditors (Byrnes et al. 2002). Fuerman (2000) finds

more shareholder litigation against auditors of restating firms, while Liu et al. (2009)

document that shareholders are more likely to vote against the ratification of an auditor of

a restating firm.

Changing auditors would potentially allow a firm to deflect blame for the

restatement to its auditor, restore the firm’s tarnished image, regain the market’s

confidence in its financial reporting and increase audit oversight.5 In addition, restating

firms can possibly mitigate the intensity of SEC enforcement by dismissing their

incumbent auditors. For example, Leone and Liu (2010) note that SEC enforcement and

penalties are less severe when restating firms take corrective actions following

restatements.

On the other hand, the costs of switching audit firms are substantial. Incoming

auditors face a steep learning curve and tend to provide a lower quality of assurance in

the early years of auditing their clients’ operations (Johnson et al. 2002; Carcello and

Nagy 2004).6 Clients also face a limited choice of auditors that specialize in their

industry and are also close to corporate headquarters (Agrawal and Cooper 2007). Finally,

only a low percentage of shareholders actually vote against an auditor’s ratification

(2.15%) and their votes are nonbinding (Liu et al. 2009). Therefore, depending on the

5 However, auditors also can initiate the separation. Auditors actively engage in client portfolio risk management to lower the probability of future audit failures. For instance, auditors adjust their client portfolios to reduce overall exposure to client litigation by simply resigning from engagements that pose a high litigation risk, for example client-firms that have restatements (Krishnan and Krishnan 1997; Shu 2000). Prior studies show that auditors increasingly resign from risky clients rather than simply adjusting audit procedures or audit fees (Johnstone 2000; Bockus and Gigler 1998). 6 The GAO (2003) also reports that it takes an audit firm at least two or three years to become adequately acquainted with a client’s operations.

7

7

relative balance of incentives and disincentives facing client-firms, an auditor change

may or may not occur following a restatement.

Prior research presents mixed and preliminary evidence on whether auditor

changes occur in response to restatements. Agrawal and Cooper (2007) whose main focus

is on top management turnover, examine auditor switches for 518 restating firms using

the GAO’s database for the period 1997-2002. They fail to find consistent evidence that

restating firms are more likely to change their auditors. However, Wallace (2005) who

studies 731 restating firms also from GAO’s database during the same period, reports

univariate statistics showing that there are a greater number of auditor switches for firms

with multiple restatements. Finally, Thompson and McCoy (2008) who analyze Fortune

500 companies during 2001 and 2002 provide univariate evidence only, that firms whose

income is materially reduced by a restatement are more likely to switch their auditors.

A concurrent study to ours that focuses exclusively on auditor changes around

restatement announcements is by Hennes et al. (2011). Our research, however, differs

from Hennes et al. in several ways. First, with regard to the data, Hennes et al. use

restatements obtained from the GAO’s database while our study obtains restatement data

from Audit Analytics. Compared to the GAO, Audit Analytics provides a more

comprehensive coverage on restatements (Scholz 2008). Second, as discussed in a later

section, we hand-collect restatement amounts and classify them according to their effects

on net income. As part of this process, we also identify a group of non-substantive

restatements for use in sensitivity tests to support our argument that only firms having

substantive restatements will switch their auditors. Third, Hennes et al. include auditor

switches 12 months before and after restatements. Lazer et al. (2004), however, find that

8

8

restatements are also initiated at the urging of the successor auditor to clean up

irregularities attributed to the predecessor auditor, whereas we are only interested in

examining whether restatements lead to subsequent auditor changes. In contrast to

Hennes et al., therefore, our sample excludes auditor changes prior to a restatement. 7

Finally, unlike Hennes et al., we examine the role of corporate governance in moderating

the effect of restatements on subsequent auditor turnover.

Based on the discussion above, we state our first set of hypotheses (in the

alternative form) as follows:

H1A: Firms restating their financial statements are more likely than other firms to change their auditors. While restatements are generally viewed negatively by investors, firms with less

severe restatements may not lose sufficient amount of reputational capital to warrant a

change in auditors. However, where the nature of the restatement is severe, an auditor

change may be justified. That is, the benefits from changing the incumbent auditor for

firms with severe restatements may be sufficiently high to cover the substantial costs

associated with a new audit firm. This leads to a follow-up hypothesis stated in the

alternative form:

H1B: As the severity of a financial restatement increases, the likelihood of an auditor change also increases Next, we explore whether corporate governance is a moderating variable in the

association between restatements and subsequent auditor switches. Prior research shows

that firms having effective governance actively monitor financial reporting and auditing

and take timely steps to protect shareholders’ interests (Weisbach 1988; Borokhovich et

7 Not surprisingly, Hennes et al. (2011) report an auditor turnover rate of 28.8% surrounding restatement announcements, while our study documents a rate of 14.5% following restatement announcements.

9

9

al. 1996). More specifically, Farber (2005) argues that corporate governance plays an

important role in restoring financial reporting credibility following financial restatements.

Therefore, if restatements are viewed as audit failures, we should expect that firms

having effective governance will take corrective actions directed at their auditors, to

restore investor confidence and audit quality. This suggests that there could be a higher

likelihood of auditor turnovers in restating firms having effective governance. Our

second hypothesis stated in the alternative form is as follows:

H2: Restating firms having strong governance are more likely than other firms to change their auditors.

RESEARCH DESIGN

Sample Selection and Descriptive Statistics

Figure 1 depicts the sequence of events relevant to our study: upon discovery of a

misstatement or irregularity associated with a past period, a restatement of the financial

statements is announced by the company and, for some firms, the announcements are

followed by auditor changes. As Figure 1 shows, we design our test procedures to focus

on auditor changes in the fiscal year following the year during which the restatement

announcement was made. For restating firms that change their auditors, the median

duration between a restatement announcement and the auditor change is 359 days.

(Insert FIGURE 1 Here)

Panel A of Table 2 summarizes the sample selection process. The initial sample

consists of all active firms on Compustat during 2001 to 2006 (55,525 firm-year

observations). We then eliminate firms that: (a) lack audit fee information in Audit

Analytics, (b) belong to the financial services industry (SIC codes 6000 to 6999), (c) lack

10

10

financial data required for our tests in Compustat and (d) were audited by Arthur

Andersen.

Next, using Audit Analytics, we identify firms from this sample that restated their

financial statements. As noted above, our sample begins in 2001 because financial

restatement data is unavailable on Audit Analytics prior to 2000 and our tests use a lagged

variable for restatements. For the restating firms, we then hand-collect from press

releases and SEC filings (Forms 10-K/A, 10-Q/A, and 8-K), the restatement amounts

(RESTAMT) and the number of prior quarter results that were restated (RESTYR). We

deleted restatements identified in Audit Analytics but for which we could not find SEC

filings and press releases. During this process, we also identify firms with non-

substantive/technical types of restatements.8 These firm-observations were set aside for

use in sensitivity tests. The final sample after these procedures, consisting of both non-

restating firms and firms with substantive restatements, has 31,627 observations.

The above sample includes 2,616 auditor changes (Panel B of Table 2).

Approximately 73% of these changes (1,904 observations) involve dismissals of auditors,

while the remaining 27% (712 observations) represent auditor resignations. Excepting for

2006, auditor dismissals and resignations have been increasing over time.

Panel C of Table 2 provides descriptive data on the 1,807 firm-years in our

sample where restatements occurred. Included are firms with multiple restatement

announcements. Firms with exactly two restatements are 17% of the sample, those with

8 Similar to Audit Analytics, our classification procedure identifies two main sets of reasons for technical/non-substantive restatements: (1) those that are related to reorganization issues (e.g., mergers and acquisitions, discontinued operations, bankruptcy, receiverships or other reorganizations, stock splits and stock dividends) and (2) those related to accounting rules (e.g., voluntary changes in GAAP accounting methods, changes in estimates and clarifications issued by the FASB or SEC on GAAP matters). The majority of the observations (about 56%) were related to the latter category.

11

11

more than two restatements are 2.5% and the remainder (80.5%) consists of firms having

a single restatement. As with auditor changes, the number of restatement announcements

are trending upwards, a pattern also documented in prior studies (Scholz 2008; Files et al.

2011; Burks 2011). Panel C also groups restatements according to whether they decrease

net income (NEGEFFECT), increase net income (POSEFFECT) or have no effect on net

income (NOEFFECT).9 Of the three categories, the largest is NEGEFFECT which is

considered to be the most severe category of restatements (Agrawal and Cooper 2007).

As in Ettredge et al. (2009), we also regard restatements accompanied by negative

stock returns as more severe than those followed by non-negative stock returns.10 Panel D

of Table 2 shows that there are 1,337 firm-years for which stock market returns were

available on the restatement announcement dates. Of these restatements with positive

returns constitute about 42% of the sample, which is comparable to findings in Scholz

(2008).11

(Insert TABLE 2 Here)

Auditor Change Model

We model the likelihood of an auditor change as a function of one-period lagged

restatements and a set of predetermined lagged control variables used by previous

9 We defined a NOEFFECT restatement as occurring when a restatement only involved reclassifications in the income, cash flow or balance sheet statements, footnotes and segment disclosures, where negative and positive changes of net income offset each other, where there was no dollar impact on the income statement, and where specific amounts were not provided in the restatement announcement. Most of the restatements in this category involved reclassifications and changes in footnote or segments disclosures (about 70%). 10 Following Collins et al. (2009), we use the cumulative market-adjusted abnormal returns for the three day window (i.e., [-1, +1]) surrounding the restatement announcement date. We subtract the CRSP equally weighted index (with dividends) from a company’s daily return to obtain the market-adjusted abnormal return for each firm. 11 While it may seem surprising to observe positive returns upon announcements of restatements, Palmrose et al. (2004) suggest that the discovery of the accounting irregularity leading to a restatement could indicate that the company’s internal controls are working and/or there is effective audit committee oversight.

12

12

studies.12 We relate auditor changes to restatements in the previous year rather than the

current year, because we want to rule out the possibility that the restatements may have

occurred at the urging of the new auditor (Lazer et al. 2004). A variation of the model

below considers the severity of restatements.

Prob(Auditor Changet)= f (Restatementt-1, Controlst-1) (1)

The first set of control variables relates to auditor characteristics: 1) auditor going

concern and modified opinion (GC and MODOP), 2) short and long auditor tenure

(STTEN and LTTEN), 3) audit fees (AUDFEE), and 4) industry expertise (EXPERT). We

expect auditor changes to occur more frequently when: audit opinions are not clean

(Krishnan and Krishnan 1997; Johnstone and Bedard 2004), auditor tenure is either too

short or too long (Hennes et al. 2011),13 auditor fees are high (Woo and Koh 2001), and

the auditor does not have industry expertise (Landsman et al. 2009).14

We also include controls for clients’ financial risks. Because less profitable firms

(proxied by return on assets, ROA), firms with losses (LOSS) and highly leveraged firms

(LVRG) are considered risky (Schloetzer 2007), we predict a higher likelihood of an

auditor change in these companies. Therefore, LOSS and LVRG are expected to have

positive coefficients, while a negative coefficient is predicted for ROA. Stice (1991)

argues that high growth firms (GROWTH) pose additional risks for auditors because these

12 Note that all the independent variables are measured one year prior the year of the auditor change to control for potential endogeneity or simultaneity bias. We assume that the economic determinants were in place before the auditor change was made. 13 Auditor tenure is denoted as short-term (STTEN) if it is less than 4 years and long-term (LTTEN) if it is longer than 8 years, following Johnson et al. (2002). 14 Following Knechel et al. (2007), we measure each auditor’s industry market share using clients’ sales for each two-digit primary Standard Industry Classification (SIC) code in Compustat . Prior research has used 10 percent, 15 percent, and 20 percent thresholds during years in which there were Big 8, Big 6, and Big 5 auditors present, respectively (Knechel et al. 2007). We use a 30 percent threshold which ensures that the all of the Big 4 firms are not classified as industry specialists (e.g., Mayhew and Wilkins 2003). We find qualitatively similar results to those reported when we use thresholds of 35 percent and 40 percent.

13

13

firms tend to have less effective internal control systems. Therefore, GROWTH is

expected to be positively related to auditor changes.

Client-firms having large negative discretionary accruals, DA (a proxy for auditor

conservatism) have incentives to dismiss the incumbent auditors in the hope of finding a

more “reasonable” successor auditor (DeFond and Subramanyan 1998). This predicts a

negative association between auditor changes and DA.15

Following Landsman et al. (2009), we control for mergers and acquisitions

(M&A). Because companies are more likely to change auditors after M&A events, we

expect a positive coefficient on this variable. The more diverse and complex are the

operations of a client, the greater is the likelihood of material errors occurring and the

greater is audit effort that is required (Bamber et al. 1993). Therefore, with increased

audit complexity, the probability of a auditor-client realignment occurring is also greater.

Audit complexity is proxied by number of segments (SEG) measured as the number of

reportable segments of a client and as the ratio of foreign sales to total sales (FRGN).

Because the costs of changing auditors are higher for large clients (DeAngelo 1981), we

predict a negative coefficient on firm size (SIZE). Finally, year and industry dummies are

included as controls in the model.

EMPIRICAL RESULTS Descriptive Statistics

15 We calculate performance matched discretionary accruals following Kothari et al. (2005). First, we obtain discretionary accruals from the Jones model. Then, we match each firm-year observation with an observation belonging to a firm from the same two-digit SIC code and year and having the closest return on assets. We define performance matched discretionary accruals as each firm’s discretionary accruals minus its matched counterpart’s discretionary accruals.

14

14

Panel A of Table 3 presents a comparison of the variables used in our tests for

restating and non-restating firms.16 Firms announcing restatements appear to have riskier

and more complex operations, and also appear to pose higher risks for the auditor.

Specifically, these firms are associated with more unclean auditor opinions, shorter

auditor tenure and have higher values for: audit fees, losses, leverage, growth, number of

segments and negative discretionary accruals. Restating firms also tend to be larger in

size than the control group of non-restating firms. Interestingly, in the pre-SOX period,

restatement firms were smaller than the average firm on Compustat (Scholz 2008).

Panel B of Table 3 provides a comparison of variables for restating firms that

change auditors and those that do not. Consistent with prior research, firms that change

auditors are smaller in size, are less profitable and are more risky (i.e., greater leverage).

As expected, these firms also appear to have greater incentives to change their auditors.

Specifically, they have: a higher frequency of going concern opinions, pay higher audit

fees, fewer industry expert auditors and more negative discretionary accruals.

Panel B also shows that restating firms that change auditors report larger amounts

of restatements. The mean restatement amounts (RESTAMT) expressed as a percentage of

total assets are -2.32% versus -1.04% for restating firms that change auditors and those

that do not, respectively.17 Restating firms that change auditors are also associated with

larger negative abnormal stock market returns (CARREST) during a three day window

around the restatement announcements compared to those that keep their auditors (means 16 All continuous variables in Table 3 are winsorized at the 1% and 99% levels to reduce the effects of extreme values on the test results. 17 Restatement amounts (RESTAMT) are measured as the cumulative earnings effect of the restatement scaled by total assets as of the fiscal year-end prior to the restatement announcement (Palmrose et al. 2004; Files et al. 2009). The mean (median) amount of income decreasing restatements is $ 15.88 ($9.32) million, while mean (median) amount of income increase restatements is $15.57 ($11.54) million. Following Myers et al. (2009), we measure restatement periods (RESTYR) as the total number of misstated years corrected by the restatement, assigning a value of 0.25 for each quarter restated.

15

15

of CARREST are -0.0227 and -0.0087, respectively). These results are consistent with the

idea that as the severity of restatements increases, the likelihood of an auditor change also

increases. For both groups, on average, a restatement announcement resulted in

corrections to five prior quarters of financial statements (RESTYR). There is no statistical

difference between the two groups in the number of prior quarters subject to restatements.

Panel C of Table 3 shows the mean auditor turnover for restating firms in the

years before and after a restatement announcement. The mean auditor turnover in the

years t-4 through t-2 for restating firms is lower than the overall mean auditor turnover

rate of 9.2% for the non-restating (control) firms. However, in the year preceding the

restatement (t-1) and in the years following (until year t+4) the mean auditor turnover of

restating firms is higher than this overall mean. The higher auditor turnover in year t-1

supports results in Lazer et al. (2004) who find that restatements occur in the year

following an auditor switch at the urging of the new auditor. Our study, however,

documents a new result, namely that a higher level of auditor turnover also occurs

following a restatement, from year t+1 (14.5%) through year t+4 (10.3%).18

In Panel D of Table 3, we present univariate test results regarding our hypotheses.

In support of our main hypothesis H1A, we find that firms having financial restatements

are more likely to change auditors. The mean of the auditor switch variable for the

restating firms in year t+1 (14.5%) is statistically significantly higher at the 1% level

18 The results for years t+2 onwards have to be interpreted with the caveat that we lose observations when we lead and lag years for auditor changes relative to a restatement announcement. This is because our dataset consists of restatement announcements during 2000-2005 and auditor changes during 2001-2006. For example, with regard to a restatement announcement in 2005, we only have subsequent auditor change data for 2006 (year t+1). Auditor change data for years t+2, t+3 and t+4 are not in our dataset for 2005 restatement announcements. We also note that as we move farther away from the restatement announcement year, the confounding effects from other factors on auditor changes increase. For both of these reasons, in our main tests, we only consider auditor changes in fiscal year t+1. We thank a reviewer for pointing out this issue.

16

16

from the overall mean auditor turnover rate for the non-restating firms (9.2%). The severe

restatement group of firms, defined using income and stock return proxies, is linked to

higher auditor changes which supports hypothesis H1B. Finally, supporting hypothesis

H2, there are more auditor switches in restating firms having higher quality corporate

governance (14%) than in the remainder of the firms (4.7%).19

Columns A and B of Panel E provide evidence on the direction of Big N auditor

switches: lateral (Big N to Big N) and downward (Big N to non-Big N). If the purpose of

an auditor change following a restatement is to restore the client-firm’s tarnished

reputation, we could expect to see fewer downward auditor switches. Consistent with this,

we find that the proportion of restating firms with lateral auditor changes is larger

(12.2%) than the proportion of restating firms with downward changes (9.3%) and this

difference is statistically significant at the 10% level of testing. 20

(Insert TABLE 3 Here)

Logistic Regression Results on Auditor Turnovers

Table 4 presents estimation results for four logistic regressions that test

hypotheses H1A and H1B. The overall fit of all models is statistically significant at the

1% level. The variance inflation factors (VIF) are well below 10.00 alleviating concerns

about multicollinearity. Because there are multiple observations across time for a given

firm, we present robust standard errors in our statistical tests of significance of the

coefficients.

19 Higher (lower) quality governance is defined using a governance index developed by DeFond et al. (2005). The governance index is discussed in a later section. 20 The passage of Section 404 of Sarbanes-Oxley led to many Big N to Non-Big N auditor switches. Also, as shown in Panel E there is also a much smaller sample of firms that were audited by a non-Big auditor. With regard to these auditor switches (Columns D and E), we should expect a larger proportion of upward than lateral changes. However, we do not find the differences in proportions to be statistically different.

17

17

Restatement Announcements and Auditor Changes

Model 1 estimation results show support for the main hypothesis, H1A. We find

that restatements are positively associated with auditor switches: the coefficient on REST,

a dummy variable that takes value of 1 if there is a restatement and zero otherwise, is

positive and statistically significant at the 1% level. The marginal effect of REST on the

likelihood of an auditor switch is 0.0397, which suggests that the announcement of a

restatement increases the probability of an auditor switch by approximately 4 percent.21

The computed marginal effect potentially understates the effect of a restatement on

auditor turnover because our tests only analyze auditor changes in the subsequent year

(t+1).

Restatements For Non-substantive/Technical reasons

Using firms with restatements for technical reasons as a test group and non-

restating firms as a control group, we similarly examine whether restatement

announcements also lead to a higher frequency of auditor changes. We find an

insignificant association between these types of restatements and auditor switches (results

are untabulated) which provides additional support for hypothesis H1A that only

restatements due to substantive reasons result in auditor switches.

Restatements and Type of Auditor Change

We also estimate a multinomial logistic regression to examine whether different

types of auditor changes have different determinants. Using the “no-auditor change” as

the reference category, we allow our dependent variable to have the following four

outcomes based on the four different types of auditor changes: Big N to Big N, Big N to

21 The marginal effect is the change in the estimated probability of an auditor switch corresponding to a unit change in a variable, holding all other variables constant at their sample mean values.

18

18

Non-Big N, Non-Big N to Big N, and Non-Big N to Non-Big N. We find positive and

statistically significant coefficients (1% level of testing) on REST for the two outcomes

representing the most frequently encountered types of auditor changes: Big N to Big N

and Non-Big N to Non-Big N (results are not tabulated). Regarding the coefficients on

the other determinants, we found results generally consistent with expectations, excepting

for some associations relating to Big N to Non-Big N auditor changes.22

Overall, the above results support hypothesis, H1A. However, we acknowledge

that the choice of the subsequent auditor may be endogeneous to our regression model(s),

which represents a potential weakness of our study.

Severity of Restatements and Auditor Changes

While all substantive restatements represent a breakdown in the financial

reporting process, those having a more negative effect on net income (Agrawal and

Cooper 2007) and those followed by a more negative stock market reaction are regarded

as more severe (Ettredge et al. 2009). The sample used for these tests consists only of

restating firms with substantive restatements. In support of hypothesis H1B, estimation

results for Models 2, 3 and 4 in Table 4 indicate that the more negative the restatement

announcement return and the more negative the effect of the restatement on net income,

the higher is the frequency of an auditor switch in the following year: the coefficients on

CARREST and RESTAMT are negative and statistically significant when included

individually and together, at least at the 10% level of significance.23

22 For example, we expect, in general, that high leverage will be positively associated with auditor turnover. However, we find that leverage is negatively associated with a Big N to Non-Big N change, suggesting that risky firms tend to stay with their incumbent auditor rather than make a downward change. Similarly, we found the “opposite” associations for EXPERT, DA, GROWTH and ROA. 23 As a test of sensitivity (results not tabulated), we include dummy variables for negative and positive stock market return (NEGRET and POSRET, respectively) and dummy variables denoting the negative (NEGEFFECT) and positive (POSEFFECT) effects on net income. While the coefficients on both

19

19

We also consider that there is a more severe breakdown in financial reporting and

auditing, in the group of firms having multiple restatements (Files et al. 2011). However,

in both univariate and the regression tests, we do not find that auditor turnover in these

firms is statistically different from that in firms with single restatements (results not

tabulated).

As for the control variables, with the exception of return on assets and leverage,

we find associations that are consistent with expectations. Specifically, we find that

switching firms are smaller, and tend to have: a higher frequency of going

concern/modified opinions, a lengthier relationship with their auditors, higher audit fees,

a non-industry expert as their auditor, more M&A activity and more segments.24

(Insert TABLE 4 Here)

Logistic Regressions for Corporate Governance and Auditor Changes

Firms are reluctant to switch auditors because there are adverse consequences

associated with auditor changes. Therefore, on the one hand, it is possible that firms

having strong governance will seek to avoid the negative consequences that follow

auditor changes and will be less likely to dismiss their auditors.25 On the other hand, we

should expect well-governed firms to take prompt action to restore their firms’

reputations by dismissing the auditors for their failure to detect accounting irregularities

(Weisbach 1998; Borokhovich et al. 1996).

NEGRET and POSRET are positive and statistically significant, and the coefficient on NEGRET is more than twice as large and is statistically significantly larger at the 1% level in a Chi-square test. Additionally, only the coefficient on NEGEFFECT is statistically significant at the 1 % level and the difference in the coefficients for NEGEFFECT (most severe restatement) and POSEFFECT (least severe restatement) is statistically significant at the 1% level. 24For firms during the period 2004-2006, we perform additional tests by including an additional control variable, ICMW, a dummy variable coded 1 if a firm reports a material weakness in internal controls under SOX 404, and 0 otherwise. We find that coefficient of ICMW is positive and statistically positive (p=0.0001); all other results are qualitatively similar (results not tabulated). 25 Hoitash and Hoitash (2009) find that stronger audit committees are more likely to retain their auditors.

20

20

Before we discuss our tests, there is a potential endogeneity issue to be addressed.

While we argue that increased governance should lead to more dismissed auditors, it is

also possible that higher governance should also lead to fewer restatements. However,

Larcker et al. (2007) document that accounting restatements are not related to governance

proxies that include board characteristics, stock ownership of insiders, institutional

ownership, activist stockownership, executive compensation, and anti-takeover variables.

On the other hand, in an unpublished working paper, Baber et al. (2009) document that

restatements occur more frequently in firms with weak corporate governance.

There is, therefore, mixed evidence on the endogeneity between the restatements

and corporate governance. Additionally, in contrast to the above research, we are

interested in examining the responsiveness of corporate governance after an irregularity

occurs. We argue that once an irregularity is discovered, companies with higher quality

governance can be expected to take corrective actions by dismissing their auditors. We

measure the effectiveness of corporate governance at the start of the year during which

there was an auditor change. Our proxy, therefore, should reflect changes to external and

board governance that may have occurred since the restatement, mitigating the impact of

the endogeneity, if any, on our results.26

Our proxy for measuring effective governance is a governance index (GOV)

proposed by DeFond et al. (2005). The following six variables are included in the index:

1) board size; 2) board independence; 3) audit committee size; 4) audit committee

26 As a test of sensitivity, we also include a dummy variable proxying for CEO and/or CFO turnover during the year of the auditor change. This variable was obtained from ExecuComp. We do not report results including this variable because the coefficient on this variable was not statistically significant and because this variable is only available for a subset of firms in our sample; ExecuComp appears to only provide comprehensive coverage mostly for the S&P 500 firms.

21

21

independence; 5) institutional ownership; and 6) the G-index.27 A score of one is

assigned to each of the six variables when a governance attribute is present and 0 when it

is absent. The scores are then aggregated to obtain a composite governance index. The

merit of this index is that it captures not only the independence and effectiveness of the

full board and its audit committee but also the monitoring by institutional investors and

the extent of management’s entrenchment (e.g., protections against hostile takeovers).

Panel A of Table 5 provides descriptive statistics on the governance index and its

components. The mean and median values of GOV are 3.14 and 3, respectively. Firms

with good governance are defined as those having GOV values that are greater than the

median value of 3. For these tests, we use a sub-sample (3,689 observations) for which

the governance variables are available on the RiskMetrics.28

Model 1 in Panel B shows estimation results when the governance index is

included in the logistic regression.29 As expected, the coefficient on the test variable,

REST, is positive and statistically significant at the 1 % level of testing. Since

RiskMetrics only covers the S&P 1500 firms, our main hypothesis, H1A, therefore,

appears to be also confirmed for the larger publicly traded firms. 30

Effective corporate governance can make it difficult for management to dismiss

their auditors opportunistically (Agrawal and Cooper 2007). Consistent with this, we find 27 The G index, developed by Gompers et al. (2003), is a composite index of 24 provisions that represents the level of shareholder protection or conversely the level of managerial entrenchment. High values of the G index represent a high level of management entrenchment, for example protection to management from takeovers (i.e., weak governance). 28 From the RiskMetrics database, we select variables representing the various governance characteristics of a firm. In addition, we obtain institutional ownership from Thompson Reuters (CDA/Spectrum) database. 29 Following Engel et al. (2010), we exclude utilities since regulated firms have different corporate governance structures than firms in non-regulated industries. However, the inclusion of the industry does not alter significantly any results documented. 30The results from using the S&P 1500 firms may not be generalizable to all U.S. public firms to the extent that the S&P 1500 firms differ systematically from all other firms. Firms on the S&P, however, represent approximately 85% of market capitalization of all publicly traded firms (Baber et al. 2009). Our tests, therefore, include firms that represent an economically significant portion of the entire market.

22

22

that GOV is negatively related to auditor turnover (p<0.0001). More importantly,

however, in support of Hypothesis H2, we find a positive and statistically significant

coefficient on REST*GOV (p=0.0021), which suggests that the association between

restatements and subsequent auditor turnovers is more pronounced in firms with more

effective governance.

In Model 2, we use individual components of the governance index as

independent variables. Again, we find that the coefficient on REST is statistically

significant (p=0.0307). The most interesting result here concerns the monitoring provided

by the audit committee. We find a negative and statistical significant coefficient on

RACBD which indicates that large audit committees are associated with a lower

frequency of auditor switches. This is consistent with large audit committees providing

better monitoring of their companies’ financial reporting and auditing, in turn reducing

the likelihood of opportunistic dismissals of auditors by management. However, after a

restatement announcement, supporting hypothesis, H2, larger audit committees are

associated with a higher frequency of auditor changes (REST*RACBD).31

Overall, results in Table 5 suggest that firms having good governance and

effective audit committees will take corrective steps to restore auditor oversight and their

companies’ reputations tarnished by the restatements.

(Insert TABLE 5 Here)

Stock Market Reaction to Auditor Changes

31 The only other coefficients that are statistically significant are on NUMBD and REST*GINDEX. Consistent with expectations, they suggest that bigger boards are associated with less effective monitoring (Yermack 1996) and that stronger shareholder rights are positively associated with auditor switches following the restatements.

23

23

While the underlying reasons are rarely revealed in a Form 8-K, auditor changes

generally occur after “negative events” including a client’s refusal to accept a qualified

audit opinion, evidence of an illegal act, or impaired auditor independence. In support,

prior studies find a negative market reaction to an auditor change (Griffin and Lont 2010;

Knechel et al. 2007).32 However, an auditor change following a restatement could be a

positive (or at least non-negative) event for investors if it represents the company’s

attempt to restore its tarnished reputation in the market place.33 And, although the cost

associated with an auditor change is high, as the severity of a restatement increases, the

benefits to changing the auditor could exceed these costs.

To explore this idea, we estimate an OLS model using the stock market returns

surrounding an auditor change for a restating firm.34 The sample used for this test

consists of firms that announced restatements and subsequently changed their auditors.

Following Knechel et al. (2007), we eliminated seven observations for which there were

confounding/concurrent news during a seven-day period (-3, 3) surrounding the date of

auditor switch. These are firms that disclosed, during this window, their quarterly or

annual earnings, or other significant corporate news.35

As restatements increase in severity, we expect that the net benefits to changing

the auditor will also increase. Specifically, we predict that as the abnormal returns around 32 However, there are some studies that find no evidence of a market reaction (Johnson and Lys 1990; Klock 1994). Klock (1994) find that the market does not react to 8-K’s filed for a change in certifying accountant in a standard event-time test using data during 1986-1987. Johnson and Lys (1990) do not find a significant market reaction to disclosures of voluntary auditor changes. 33 Similar to our study, there is other research that finds positive abnormal returns on auditor switch dates (Carter and Soo 1999). For example, Johnson and Lys (1990) suggest that auditor changes by management (especially dismissals) could be regarded as a positive event by investors to the extent that managers act in the best interest of shareholders 34 Following Knechel et al. (2007), we use auditor dismissal dates (as reported in Audit Analytics) rather than filing dates of 8-Ks to avoid potential confounding effects due to other events that are simultaneously reported in 8-K forms. 35 Other news includes shareholder lawsuits, officer/director changes, product introductions, merger related news and dividend announcements.

24

24

restatement announcements (CARREST) and the restatement amounts (RESTAMT)

become more negative, the abnormal returns surrounding the auditor change will become

more positive.36

CARACt =β0 + β1 {CARRESTt-1 and/or RESTAMTt-1}+ β2 RPTEt + β3 Q4t + β4

RSGNt + β5 NONBIGt-1 + β6 SIZEt-1 + β7 GCt-1 +year dummies +industry dummies + ε (2)

CARAC is the cumulative abnormal stock return over a three-day window

surrounding the auditor change (-1, 1). We calculate abnormal stock returns by

subtracting the CRSP equal-weighted market return from the firm’s holding returns on

each day and summing these returns over three days (Griffin and Lont 2010; Scholz

2008).37 As for the control variables, a negative market reaction is expected for firms

having reportable disclosures, RPTE (Griffin and Long 2010; Whisenant et al. 2003).38

Auditor changes late in the year (Q4) are predicted to be accompanied by a negative

market reaction (Knechel et al. 2007; Hackenbrack and Hogan 2002). Griffin and Lont

(2010) find that auditor resignations (RSGN) experience more negative market returns

than dismissals. Similar to Hennes et al. (2011), we include a dummy variable for an

audit performed by a Non-Big N auditor (NONBIG) prior to the auditor change. Hennes

et al. argue that there is greater information asymmetry in client-firms having Non-Big N

auditors and, therefore, an auditor change following a restatement for these firms will

36 We also included two additional proxies for severity of a restatement: multiple restatement firms and the lag between the restatement announcement and auditor change. Both of the variables were not statistically significantly related to CARAC (results not tabulated) 37 We find similar results when abnormal returns are based on a market model estimated using the CRSP value-weighted return over days -220 to -20 relative to the auditor turnover date (e.g., Baik et al. 2008). 38 According to Audit Analytics, reportable disclosures (RPTE) include internal control reportable condition, scope limitation, financial restatement, audit opinion concerns, management not reliable, illegal acts, SEC investigation, SEC banned auditor, SEC inquiry regarding company or auditor, lack of independence, bankruptcy, existing public audits, PCAOB registration, incoming auditor will re-audit, fee reduction, incoming auditor approved by board, consulted with incoming auditor, and agreement or disagreement in auditor letter.

25

25

lead to a stronger market reaction.39 The coefficient on GC (going concern opinion) is

predicted to be negative. We do not have any prediction for the sign on firm size (SIZE).

Panel A of Table 6 shows univariate statistics for the variables used in the

regression. The mean return around the auditor change is positive as expected. On

average, each firm has approximately one reportable event. 31% of the observations

consist of auditor resignations. 27% of auditor changes for the restating firms involve the

Non-Big N auditors. Consistent with Hennes et al., we also find that there is higher

auditor turnover in Non-Big auditors (22%) compared to Big 4 auditors (11%).40 Finally,

the univariate statistics show that about 8% of the firms have going concern opinions.

Our regression results in Panel B show that, consistent with expectations, the

coefficients on CAREST and RESTAMT are negative and statistically significant.

Therefore, as the restatements increase in severity, investors view an auditor switch as

being beneficial for the company. This is consistent with Wilson (2010) who finds that

investors’ confidence in the reported financial information increases when firms take

corrective actions following restatements, thus demonstrating to investors their

commitment to high-quality financial reporting.41 Regarding the other variables, larger

firms and restating firms that change auditors in the fourth quarter experience a more

negative stock market reaction.

(Insert TABLE 6 Here)

39 As a test of sensitivity we also include the following four variables: B_B (Big N to Big N), B_N (Big N to non-Big N), N_N (non-Big N to non-Big N), are expected to be negative, while the coefficient on N_B (non-Big N to Big N) is expected to be positive. We find statistically insignificant coefficients on all of these variables (results not tabulated). 40 We obtained these percentages as follows: Of the sample of all restating firms, there were 524 firms audited by Non-Big N auditors and 117 of these switched their auditors (22%), while the remainder 1,283 restating firms were audited by Big N auditors and 145 of these switched their auditors (11%). 41 Specifically, Wilson (2010) finds that earnings response coefficients (ERC) for the group without auditor dismissal are significantly lower than the group with auditor dismissal for the period following restatements.

26

26

Resignations versus Dismissals

Rather than be dismissed, the external auditor could resign from an engagement to

protect the audit firm’s reputation and attempt to deflect blame on the company’s

management. There are at least two reasons why investors may not discriminate between

an auditor resignation and a dismissal. First, to the extent that shareholders view a

restatement as a breakdown in auditing, the form of the auditor change (resignation or

dismissal) should not be relevant. Second, Lee et al. (2004) argue that there is often no

difference between an auditor resignation and a dismissal because an audit firm can

preemptively resign from an engagement rather be dismissed at a later date by the audit

committee. Alternatively, however, if investors view a resignation differently and assign

blame to management rather than the auditor, we could observe a negative investor

reaction to a resignation, in contrast to a positive reaction to a dismissal. The results in

Panel B of Table 6, however, show that the coefficient on RSGN is statistically

insignificant suggesting that the market does not distinguish between a resignation and a

dismissal.42

CONCLUSION

Restatements and auditor changes, both, are events that draw enormous public

attention. Restatements constitute a failure of the accounting and audit functions to

produce reliable financial statements, creating the need for changes in the firm’s

42 We also estimate the model using just the dismissal firms and find qualitatively similar results (results untabulated). However, the model using just resignation firms did not have a good fit (i.e., negative adjusted R-squared and a statistically insignificant F-value for the model).

27

27

governance mechanisms. One aspect of governance that is destabilized by a restatement

is the relationship between restating firms and their auditors.

In support, we find that disclosures of restatements are positively correlated with

auditor switches in the following year. This result holds after controlling for previously

identified determinants of auditor changes. We find that restatements due to technical

reasons—namely those not related to auditing and financial reporting failures—do not

lead to auditor changes, reinforcing our argument that only substantive financial

misstatements result in a separation between auditor and client.

As the severity of restatements increases, so does the likelihood of auditor

changes. Specifically, we find that as the impact of a restatement on net income and the

company’s stock price become more negative, the likelihood of an auditor change

increases. We also find that restating firms with more effective corporate governance are

more likely to switch auditors, possibly to restore market confidence that has been lost

due to the lack of audit oversight. Finally, we argue that the benefits to a firm from

changing auditors can be expected to increase as the severity of a restatement increases.

Consistent with this, we find that stock market returns around the auditor turnover dates

are positively related to the severity of a restatement. Future research could further

explore this issue by examining whether there are differences in the quality of financial

reporting of restating firms that changed auditors and those that did not.

28

28

REFERENCES Agrawal, A., and T. Cooper. 2007. Corporate governance consequences of accounting scandals: evidence from top management, CFO and auditor turnover. Working paper, University of Alabama. Baber, W. R., S. Kang, L. Liang, and Z. Zhu. 2009. Shareholder rights, corporate governance and accounting restatement. Working paper, George Washington University and Syracuse University. Baik, B., B. K. Billings, and R. M. Morton. 2008. Reliability and transparency of non- GAAP disclosures by real estate investment trusts (REITs). The Accounting Review 83(2): 271-301. Bamber, E., L. Bamber, and M. Schoderbek. 1993. Audit structure and other determinants of audit report lag: an empirical analysis. Auditing: A Journal of Practice & Theory 12: 1–23. Bockus, K., and F. Gigler. 1998. A theory of auditor resignation. Journal of Accounting Research 36(Autumn): 191-208. Borokhovich, K., R. Parrino, and T. Trapani. 1996. Outside directors and CEO selection. Journal of Financial and Quantitative Analysis 31 (September): 337-355. Burks, J. J. 2011. Are investors confused by restatements after Sarbanes-Oxley? The Accounting Review 86(2): 507-539. Byrnes, N., M. McNamee, D. Brady, L. Lavelle, and C. Palmeri. 2002. Accounting in crisis: Reform is urgent. Business Week (January 28): 44-49. Carcello, J., and A. Nagy. 2004. Audit firm tenure and fraudulent financial reporting. Auditing: A Journal of Practice & Theory 23(2): 55-69. Carter, M. E., and B. S. Soo. 1999. The relevance of form 8-K reports. Journal of Accounting Research 37(1): 119-132. Collins, D., A. Masli, A. L. Reitenga, and J. M. Sanchez. 2009. Earnings restatements, the Sarbanes-Oxley Act, and the disciplining of chief financial officers. Journal of Accounting, Auditing & Finance 24(1): 1-34. DeAngelo, L. E. 1981. Auditor size and audit quality. Journal of Accounting and Economics 3: 183-199. DeFond, M. L., R. Hann, and X. Hu. 2005. Does the market value financial expertise on audit committees of boards of directors? Journal of Accounting Research 43(2): 153-193.

29

29

DeFond, M. L., and K. R. Subramanyam. 1998. Auditor changes and discretionary accruals. Journal of Accounting and Economics 25(February): 35-67. Desai, H., C. Hogan, and M. Wilkins. 2006. The reputational penalty for aggressive accounting: earnings restatements and management turnover. The Accounting Review 81 (1): 83-112. Engel, E., R. M. Hyes, and X. Wang. 2010. Audit committee compensation and the demand for monitoring of the financial reporting process. Journal of Accounting and Economics 49: 136-154. Ettredge, M., Y. Huang, and W. Zhang. 2009. Restatement disclosures and subsequent accounting conservatism. Working paper, University of Kansas and University of Texas at Dallas. Farber, D. B. 2005. Restoring trust after fraud: Does corporate governance matter? The Accounting Review 80(2): 539-561. Feldmann, D. A., W. J. Read, and M. J. Abdolmohammadi. 2009. Financial restatements, audit fees, and the moderating effect of CFO turnover. Auditing: A Journal of Practice & Theory 28(1): 205-223. Files, R., N. Y. Sharp, and A. M. Thompson. 2011. Why do firms restate repeatedly? Working paper, University of Texas at Dallas and Texas A&M University. Files, R., E. P. Swanson, and S. Tse. 2009. Stealth disclosure of accounting restatements. The Accounting Review 84 (5): 1495-1520. Fuerman, R. D. 2000. Auditors and the post-litigation reform act environment. Research in Accounting Regulation 14: 199-218. Gleason, C. A., N. T. Jenkins, and W. B. Johnson. 2008. The contagion effects of accounting restatements. The Accounting Review 83(1): 83-110. Gompers, P. A., J. L. Ishii, and A. Metrick. 2003. Corporate governance and equity prices. Quarterly Journal of Economics 118: 107-155. Griffin, P. A., and D. H. Lont. 2010. Do investors care about auditor dismissals and resignations? What drives the response? Auditing: A Journal of Practice & Theory 29(2): 189-214. Hackenbrack, K. E., and C. E. Hogan. 2002. Market response to earnings surprises conditional on reasons for an auditor change. Contemporary Accounting Research 19(2): 1995-223.

30

30

Hennes, K. M., A. J. Leone, and B. P. Miller. 2011. Auditor dismissals around accounting restatements. Working paper, University of Oklahoma, University of Miami, and Indiana University. Hoitash, R., and U. Hoitash. 2009. The role of audit committees in managing relationships with external auditors after SOX: Evidence from the USA. Managerial Auditing Journal 24(4): 368-397. Hribar, P., and N. T. Jenkins. 2004. The effect of accounting restatements on earnings revisions and estimated cost of capital. Review of Accounting Studies 9: 337-356. Johnson, V., I. Khurana, and J. Reynolds. 2002. Audit-firm tenure and the quality of financial reports. Contemporary Accounting research 19 (4): 637-660. Johnson, W. B., and T. Lys. 1990. The market for audit services. Journal of Accounting and Economics 12(January): 281-308. Johnstone, K. M. 2000. Client-acceptance decisions: simultaneous effects of client business risk, audit risk, auditor business risk, and risk adaptation. Auditing: A Journal of Practice & Theory 19(1): 1-25. Johnstone, K. M., and J. C. Bedard. 2004. Audit firm portfolio management decisions. Journal of Accounting Research 42(4): 659-690. Kinney, W. R. Jr., 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(3): 561-588. Klock, M. 1994. The stock market reaction to a change in certifying accountant. Journal of Accounting Auditing and Finance 9: 339-347. Knechel, W. R., V. Naiker, and G. Pacheco. 2007. Does auditor industry specialization matter? Evidence from market reaction to auditor switches. Auditing: A Journal of Practice & Theory 26(1): 19-45. Kothari, S. P., Leone, A. J., Wasley, C.E., 2005. Performance matched discretionary accrual measures. Journal of Accounting Economics 39, 163-197. Krishnan, J., and J. Krishnan. 1997. Litigation risk and auditor resignations. The Accounting Review 72(October): 539-560. Landsman, W. R., K. K. Nelson, K. K., and B. R. Rountree, B. R. 2009. Auditor switches in the pre- and post-Enron eras: risk or realignment? The Accounting Review 84 (2): 531-558.

31

31

Larcker, D. F., and S. A. Richardson. 2004. Fees paid to audit firms, accrual choices, and corporate governance. Journal of Accounting Research 42(3): 625-658. Larcker, D. F., S. A. Richardson, and I. Tuna. 2007. Corporate governance, accounting outcomes, and organizational performance. The Accounting Review 82(4): 963-1008. Lazer, R., J. Livnat, and C. E. L. Tan. 2004. Restatements and accruals after auditor changes. Working paper, New York University. Leone, A. J., and M. Liu. 2010. Accounting irregularities and executive turnover in founder-managed firms. The Accounting Review 85(1): 287-314. Lee, H., V. Mande, and R. Ortman. 2004. The Effect of audit committee and board of director independence on auditor resignation. Auditing: A Journal of Practice & Theory 23(2): 131-146. Liu, L., K. Raghunandan, and D. Rama. 2009. Financial restatements and shareholder ratifications of the auditor. Auditing: A Journal of Practice & Theory 28(1): 225-240. Mayhew, B. W., and M. S. Wilkins. 2003. Preview audit firm industry specialization as a differentiation strategy: Evidence from fees charged to firms going public. Auditing: A Journal of Practice & Theory 22(2): 33-52. Myers, L. A., S. Scholz, and N. Y. Sharp. 2009. Restating under the radar: Determinants of restatement disclosure choices and the related market reactions. Working paper, University of Arkansas, University of Kansas, and Texas A&M University. Palmrose, Z.-V., V. Richardson, and S. Scholz. 2004. Determinants of market reactions to restatement announcements. Journal of Accounting and Economics 37(1): 1-31. Palmrose, Z-V., and S. Scholz. 2004. The circumstances and legal consequences of non-GAAP reporting: Evidence from restatements. Contemporary Accounting Research 21(1): 139-180. Public Company Accounting Oversight Board (PCAOB). 2007. Auditing Standard No. 5: An audit of internal control over financial reporting that is integrated with an audit of financial statements. Raghunandan, K., W. J. Read, and J. S. Whisenant. 2003. Initial evidence on the association between nonaudit fees and restated financial statements. Accounting Horizons 17(3): 223-234. Schloetzer, J. D. 2007. Arthur Andersen, SOX section 404 and auditor turnover: theory and evidence. Working paper, University of Pittsburgh.

32

32

Securities and Exchange Commission (SEC). 2002. Proposed rule: Framework for enhancing the quality of financial information through improvement of oversight of the auditing process. Securities and Exchange Commission, Washington, DC. Shu, S. 2000. Auditor resignations: Clientele effects and legal liability. Journal of Accounting and Economics 29(April): 173-205. Scholz, S. 2008. The changing nature and consequences of public company financial restatements: 1997-2006. Treasury Department Report. Srinivasan, S. 2005. Consequences of financial reporting failure for outside directors: Evidence from accounting restatements and audit committee members. Journal of Accounting Research 43(2): 291-334. Stanley, J. D., and F. T. DeZoort. 2007. Audit firm tenure and financial restatements: An analysis of industry specialization and fee effects. Journal of Accounting and Public Policy 26: 131-159. Stice, J. D. 1991. Using financial market information to identify pre-engagement factors associated with lawsuits against public accountants. The Accounting Review 66(July): 516-533. Thompson, J., and T. McCoy. 2008. An analysis of restatements due to errors and auditor changes by Fortune 500 companies. Journal of Legal, Ethical and Regulatory Issues 11(2): 45-57. Turner, L. 1999. 20th Century Myths. Speech delivered at New York University (November 15). Available at http://www.sec.gov/news/speech/speecharchive/1999/spch323.htm U.S. Government Accounting Office (GAO). 2002. Financial Statements Restatement: Trends, Market Impacts, Regulatory Responses and Remaining Challenges. Report 03-138. Washington D.C.: Government Printing Office. U. S. Government Accounting Office (GAO), 2003. Public Accounting Firms: Required Study on the Potential Effect on Mandatory Audit Firm Rotation. Report 04-216. Washington D.C.: Government Printing Office. U.S. Government Accountability Office (GAO). 2006. Financial Restatements: Update of Public Company Trends, Market Impacts, and Regulatory Enforcement Activities. Report 06-678. Washington D.C.: Government Printing Office. Wallace, W. A. 2005. Auditor changes and restatements. The CPA Journal 75(3): 30-33. Weisbach, M. 1988. Outside directors and CEO turnover. Journal of Financial Economics 20(January-March): 431-460.

33

33

Whisenant, J. S., S. Sankaraguruswamy, and K. Raghunandan. 2003. Market reactions to disclosure of reportable events. Auditing: A Journal of Practice & Theory 22(1): 181-194. Wilson, W. M. 2010. An empirical analysis of the decline in the information content of earnings following restatements. The Accounting Review 83(2): 519-548. Woo, E-S., and H. C. Koh. 2001. Factors associated with auditor changes: a Singapore study. Accounting and Business Research 31(2): 133-144. Yermack, D.1996. Higher market valuation for firms with a small board for director. Journal of Financial Economics 40: 185-211.

34

34

Figure 1: The Time-Line Used in Our Study Showing Misstatements, Restatement Announcements and Subsequent Auditor Changes

Fiscal year t Fiscal year t+1

Misstatements/Irregularities Restatement Announcements

Auditor Changes

Mean: 208 days Median: 359 days

35

35

Table 1: Definitions of Variables Used in our Main Tests

Variable Definition AUDITOR CHANGE

1 if the firm changes its auditor, and 0 otherwise;

REST 1 if the firm restates its financial statements for other than technical reasons, and 0 otherwise;

GC 1 if the audit opinion is going concern, and 0 otherwise; MODOP 1 if the audit opinion is modified for reasons other than going concern, and 0 if it is

unqualified; STTEN 1 if the auditor tenure is less than 4 years; LTTEN 1 if the auditor tenure is longer than 8 years; AUDFEE Previous year’s audit fee (i.e., old auditor’s fees) scaled by total assets; EXPERT 1 if an auditor has 30 percent or more market share in an industry as defined using

the two-digit SIC code, and 0 otherwise; ROA Return on assets, defined as net income before extraordinary items divided by total

assets; LOSS 1 if earnings before extraordinary items is less than 0, and 0 otherwise; LVRG Ratio of long term debt to total assets; GROWTH Percentage changes in sales; DA Performance matched discretionary accruals; MNA 1 if the client had a merger or acquisition in the two previous years, and 0 otherwise; SEG Number of reportable segments; FRGN Ratio of foreign sales to total sales; SIZE Natural logarithm of the market value of equity; GOV Corporate governance index used in DeFond et al. (2005); GINDEX G index developed by Gompers et al. (2003); NUMBD Number of board members; RINDBD Ratio of independent to full board members; RACBD Ratio of audit committee to full board members; RINDAC Ratio of independent audit committee members to total audit committee members; %INOWN Institutional ownership; RESTAMT Cumulative earnings effect of the restatement scaled by total assets as of the fiscal

year-end prior to the restatement announcement;

RESTYR Total number of misstated years corrected by the restatement, assigning a value of 0.25 for each quarter restated; and

CARREST Three-day abnormal market returns surrounding restatement announcement date.

36

36

Table 2: Sample Selection Procedure and Sample Distribution Panel A: Sample Selection Procedures

# of Firm Years Active firms on Compustat (2001- 2006) 55,525 Less: firms without data on Audit Analytics for audit fees needed in our tests 14,390 firms in financial industries with SIC codes 6000-6999 7,507 firms missing financial data on Compustat needed for our tests 736 firms audited by Arthur Andersen 917 firms for which SEC filings could not be found 125 firms announcing restatements for technical reasons 223 Final sample (2001-2006) 31,627

Panel B: Distribution of Auditor Changes by Year Auditor Changes

Total Dismissals Resignations 2001 173 156 17 2002 259 210 49 2003 498 368 130 2004 661 450 211 2005 629 422 207 2006 396 298 98 Total 2,616 1,904 712

Panel C: Distribution of Restatements According to their Impact on Net Income

Total Negative Impact Positive Impact No Impact 2001 81 57 15 9 2002 177 107 29 41 2003 256 117 50 89 2004 363 161 66 136 2005 402 155 64 183 2006 528 194 62 272 Total 1,807 791 286 730

Panel D: Distribution of Restatements According to the Sign of Abnormal Market Returns around Restatement Announcements

Total Negative Return Positive Return 2001 66 38 28 2002 127 81 46 2003 178 96 82 2004 245 138 107 2005 283 165 118 2006 438 253 185 Total 1,337 771 566

37

37

Table 3: Descriptive Statistics for Variables Used in our Tests Panel A: A Comparison of Variables for Restating and Non-Restating firms

Variable Firms With Restatements

(1,807 firm years) Firms Without Restatements

(29,820 firm years) Differences in Means

Mean Median Std Mean Median Std t-statistic GC 0.1501 0.0000 0.3573 0.1387 0.0000 0.3456 1.41

MODOP 0.3944 0.0000 0.4889 0.2589 0.0000 0.4381 13.06***

STTNR 0.4368 0.0000 0.4961 0.4112 0.0000 0.4921 2.22***

LTTNR 0.2821 0.0000 0.4501 0.2980 0.0000 0.4574 1.48 AUDFEE 0.0449 0.0321 0.0385 0.0282 0.0202 0.0273 25.40***

EXPERT 0.4081 0.0000 0.4949 0.3905 0.0000 0.4879 0.98

ROA -0.4042 -0.0066 1.3964 -1.0939 0.0043 2.7933 10.76***

LOSS 0.5222 1.000 0.4996 0.3985 0.0000 0.4895 10.75***

LVRG 0.3453 0.2348 0.4974 0.2675 0.1471 0.4664 7.07***

GROWTH 0.1572 0.0604 0.9068 0.0415 0.0468 0.8736 5.62***

DA -0.0516 -0.0131 0.2368 -0.0921 -0.0243 0.2693 6.43***

MNA 0.0021 0.0000 0.0455 0.0015 0.0000 0.0397 0.52

SEG 2.2955 1.0000 1.9140 2.0953 1.0000 1.8201 4.67*** FRGN 0.0613 0.0000 0.1696 0.0594 0.0000 0.1653 0.59 SIZE 5.0474 5.2388 2.5906 4.4287 4.8592 3.3152 8.04***

Panel B: A Comparison of Variables for Restating Firms that Switch Auditors and Restating Firms that do not Switch Auditors

Variable Restating Firms that switch auditors

(262 firm years) Restating Firms that do not switch

auditors (1,545 firm years) Differences in Means

Mean Median Std Mean Median Std t-statistic GC 0.2571 0.0000 0.4378 0.1319 0.0000 0.3386 5.46***

MODOP 0.3179 0.0000 0.4665 0.4074 0.0000 0.4915 -2.84***

STTNR 0.4107 0.0000 0.4928 0.4413 0.0000 0.4967 -0.95

LTTNR 0.2714 0.0000 0.4455 0.2839 0.0000 0.4510 -0.43 AUDFEE 0.0492 0.0357 0.0421 0.0443 0.0314 0.0379 1.99**

EXPERT 0.3321 0.0000 0.4718 0.4243 0.0000 0.4970 -3.52***

ROA -0.8449 -0.0470 2.1853 -0.3296 -0.0034 1.1979 -5.76***

LOSS 0.5821 1.0000 0.4941 0.5121 1.0000 0.5000 2.17**

LVRG 0.3913 0.2146 0.6170 0.3375 0.2366 0.4739 1.67*

GROWTH 0.2289 0.0503 1.2311 0.1450 0.0620 0.8394 1.43

DA -0.0845 -0.0306 0.2742 -0.0461 -0.0116 0.2295 -2.52**

MNA 0.0036 0.0000 0.0597 0.0018 0.0000 0.0426 0.60

SEG 2.1643 1.0000 1.7825 2.3617 1.0000 1.9351 -1.24 FRGN 0.0654 0.0000 0.1761 0.0711 0.0000 0.1809 -0.49 SIZE 3.9151 4.1666 2.5551 5.2393 5.4185 2.5479 -8.04***

RESTAMT -0.0232 -0.0142 0.0417 -0.0104 0.0089 0.0297 -6.23*** RESTYR 1.2450 1.0000 1.0879 1.2345 1.0000 1.2682 0.57 CARREST -0.0227 -0.0140 0.0982 -0.0087 -0.0085 0.0946 -1.79*

38

38

Panel C: Mean Auditor Turnover Rates for Restating Firms in the Years Preceding and Following a Restatement Announcement (t) (Compared to the Mean Auditor Turnover Rate of 0.0924 for Non-Restating Firms during the Sample Period)

t-4 t-3 t-2 t-1 T t+1 t+2 t+3 t+4 Mean

auditor turnover

rate

0.0851 0.0637 0.0849 0.1232 0.1805 0.1449 0.1217 0.1267 0.1033

Panel D: Univariate Tests of Hypotheses H1a, H1b and H2

Hypothesis Variable Restating firms Non-Restating firms

t-statistic (test of

differences in means)

H1a Mean rate of Auditor Turnover 0.1449 0.0924 7.61***

Hypothesis Variable Income decreasing restatements

Income increasing restatements

t-statistic (test of

differences in means)

H1b Mean rate of Auditor Turnover 0.2023 0.1119 3.44***

Hypothesis Variable Negative

announcement returns

Positive announcement

returns

t-statistic (test of

differences in means)

H1b Mean rate of Auditor Turnover 0.1388 0.1060 1.79*

Hypothesis Variable High quality governance

Low quality governance

t-statistic (test of

differences in means)

H2 #Auditor Changes following Restatements/#Restatements 0.1401 0.0465 2.65***

Panel E: Direction of Auditor Changes

Variable Row #

Big N to Big N

(A)

Big N to Non-Big

N (B)

t-statistic (test of means)

(C)

Non-Big N to Big

N (D)

Non-Big N to Non-

Big N (E)

t-statistic (test of

differences in

means) (F)

#Auditor Changes following Restatements

I 59 86 11 106

# Auditor Changes II 481 922 109 1,104 I / II 0.1227 0.0933 1.69* 0.1009 0.0960 -0.93 *, **, and *** represent 10%, 5%, and 1% statistical significance respectively, using two-tailed tests. See Table 1 for definitions of the variables. The quality of governance is measured using an index suggested by DeFond et al. (2005). High (low) quality governance is above (below) the median value of the index.

39

39

Table 4: Logistic Regression Estimation Results of Models of Auditor Changes

Variable Expected Sign

Model 1 (H1a) Model 2 (H1b) Model 3 (H1b) Model 4 (H1b) Coefficient p-value Coefficient p-value Coefficient p-value Coefficient p-value

INTERCEPT -2.3540 <.0001 -0.1706 0.7125 -0.3760 0.4239 -0.3551 0.4510 REST + 0.4425 <.0001 CARREST - -1.4733 0.0853 -1.4720 0.0887 RESTAMT - -4.6767 0.0162 -4.6506 0.0169 GC + 0.7460 <.0001 0.1724 0.4298 0.1314 0.5496 0.1205 0.5835 MODOP + 0.2634 <.0001 0.0181 0.9131 0.0324 0.8456 0.0227 0.8916 STTNR + -0.0437 0.3814 -0.2489 0.1260 -0.2518 0.1229 -0.2533 0.1209 LTTNR + 0.1311 0.0204 0.1696 0.3609 0.1735 0.3510 0.1805 0.3331 AFEE + 6.3920 <.0001 5.3033 0.0050 5.2963 0.0050 5.3549 0.0047 EXPERT - -0.1570 0.0016 -0.0881 0.5857 -0.0886 0.5845 -0.0894 0.5812 ROA - 0.1122 <.0001 0.0085 0.8766 0.0160 0.7713 0.0156 0.7772 LOSS + 0.1152 0.0175 -0.2144 0.1725 -0.2228 0.1578 -0.2235 0.15162 LVRG + -0.1262 0.0004 -0.1243 0.3404 -0.1282 0.3268 -0.1242 0.3425 GROWTH + -0.0074 0.7166 0.0993 0.1064 0.0933 0.1315 0.0944 0.1269 DA - -0.1070 0.2093 -0.0006 0.9983 0.0117 0.9669 0.0076 0.9787 MNA + 0.5814 0.0974 0.2892 0.8059 0.4752 0.6867 0.4817 0.6827 SEG + 0.0370 0.0068 0.0537 0.1848 0.0549 0.1760 0.0548 0.1766 FRGN + 0.1237 0.3606 0.3861 0.3457 0.4195 0.3061 0.3753 0.3606 SIZE - -0.2498 <.0001 -0.2200 <.0001 -0.1982 <.0001 -0.1999 <.0001 Wald Chi-Square

1659.21 <.0001 106.15 <.0001 109.48 <.0001 111.77 <.0001

Max-rescaled R2

0.1169 0.1029 0.1052 0.1077

N 31,627 1,337 1,807 1,337 See Table 1 for definitions of the variables.

40

40

Table 5: Tests of Auditor Changes and Corporate Governance Panel A: Descriptive Statistics on Corporate Governance Variables Used in our Tests (N=3,689)

Variable Mean Std. Dev. P25 Median P75 GOV 3.1369 1.2024 2.0000 3.0000 4.0000 NUMBD 8.5724 2.0231 7.0000 8.0000 9.0000 RINDBD 0.7050 0.1413 0.6250 0.7143 0.8333 RACBD 0.3902 0.1109 0.3333 0.3846 0.4444 RINDAC 0.9584 0.1199 1.0000 1.0000 1.0000 GINDEX 11.4827 3.4402 9.0000 12.0000 15.0000 %INOWN 0.4462 0.3749 0.0000 0.5712 0.7823 Panel B: Logistic Regression Results of Models of Auditor Changes that include Corporate Governance Variables

Variable Expected Sign

Model 1 (H2) Model 2 (H2) Coefficient. p-value Coefficient. p-value

INTERCEPT 2.1322 0.0261 1.3738 0.1794 REST + 0.7800 0.0052 0.6566 0.0307 GOV - -0.3513 <.0001 REST*GOV + 0.6430 0.0021 NUMBD + 0.5516 0.0252 RINDBD - -0.1517 0.4630 RACBD - -1.6648 <.0001 RINDAC - -0.0575 0.8082 GINDEX + 0.2444 0.1930 %INOWN - -0.0552 0.7666 REST*NUMBD - 0.0978 0.8753 REST*RINDBD + 0.2694 0.6824 REST*RACBD + 1.5689 0.0064 REST*RINDAC + 0.1173 0.8755 REST*GINDEX - -0.9990 0.0597 REST*%INOWN + -0.1746 0.7610 GC + -1.6679 0.3311 -1.1156 0.4314 MODOP + -0.1935 0.2984 -0.1269 0.4917 STTNR + -0.3100 0.2462 -0.3171 0.2346 LTTNR + -0.4266 0.0485 -0.4033 0.0626 AFEE + 19.1615 <.0001 16.3248 <.0001 EXPERT - -0.2406 0.1715 -0.2276 0.1958 ROA - -0.9351 0.1084 -0.9448 0.0889 LOSS + -0.3833 0.1559 -0.3643 0.1781 LVRG + 0.7496 0.1387 0.7506 0.1383 GROWTH + -0.2670 0.3869 -0.3121 0.3160 DA - 0.0122 0.9856 0.2629 0.6927 MNA + -7.1263 0.9840 -6.9699 0.9847 SEG + -0.0352 0.3648 -0.0143 0.7112 FRGN + 0.4141 0.2313 0.6231 0.0756 SIZE - -0.3929 <.0001 -0.3520 <.0001 Wald Chi-Square 652.96 <.0001 675.72 <.0001 Max-rescaled R2 0.4186 0.4423 N 3,689 3,689 See Table 1 for definitions of the variables.

41

41

Table 6: Stock Market Reaction to Auditor Changes of Restating Firms Panel A: Descriptive Statistics of Variables Used (N=171)

Variable Mean Std. Dev. P25 Median P75 CARAC 0.0039 0.0661 -0.0297 -0.0046 0.0322 CARREST -0.0411 0.1284 -0.0624 -0.0152 0.0191 RESTAMT -0.0124 0.0303 -0.0099 -0.0005 0.0000 RPTE 1.3743 1.6089 0.0000 1.0000 2.0000 Q4 0.1111 0.3152 0.0000 0.0000 0.0000 RSGN 0.3099 0.4638 0.0000 0.0000 0.0000 NONBIG 0.2690 0.4446 0.0000 0.0000 1.0000 SIZE 3.7111 2.4099 0.8526 3.8015 5.9085 GC 0.0819 0.2750 0.0000 0.0000 0.0000 Panel B: OLS Regression Results for Stock Market Reaction (CAR) surrounding Auditor Changes Model 1 Model 2 Model 3 Variable Expected

Sign Coefficients t-value Coefficients t-value Coefficients t-value

INTECEPT +/- -0.0684 2.71** 0.0376 1.20 0.0641 2.54** CARREST - -0.0763 -1.82* -0.0768 -1.82* RESTAMT - -0.3377 -1.72* -0.3354 -1.74* RPTE - -0.0054 -1.43 -0.0055 -1.42 -0.0063 -1.63

Q4 - -0.0390 -2.30** -0.0381 -2.19** -0.0393 -2.31**

RSGN - 0.0089 0.78 0.0063 0.54 0.0101 0.88

NONBIG + -0.0139 -0.98 -0.0126 -1.07 -0.0164 -1.15

SIZE +/- -0.0104 -2.88*** -0.0107 -2.74*** -0.0093 -2.58**

GC - -0.0038 -0.20 -0.0099 -0.49 -0.0119 -0.60

F-value 1.87* 1.86* 1.91* Adj R-Sq 0.0270 0.0270 0.0378 N 171 171 171 *, **, and *** represent 10%, 5%, and 1% statistical significance, respectively using two-tailed tests.

Variable Definition CARAC Three day cumulative abnormal returns surrounding the auditor switch date; CARREST Three-day abnormal market returns surrounding restatement announcement date; RESTAMT Cumulative earnings effect of the restatement scaled by total assets as of the fiscal

year-end prior to the restatement announcement; RPTE Number of reportable events disclosed upon an auditor change; Q4 1 if a firm switches its auditor in the fourth quarter of the fiscal year, and 0

otherwise; RSGN 1 if the predecessor auditor resigns, and 0 otherwise; NONBIG 1 if the firm was audited by a non-Big N auditor, and 0 otherwise; SIZE Natural logarithm of the market value of equity; and GC 1 if the firm receives a going concern opinion, and 0 otherwise.