tanyi - 2010 - audit report lags after voluntary and involuntary auditor changes

Electronic copy available at: http://ssrn.com/abstract=1806345Electronic copy available at: http://ssrn.com/abstract=1806345Audit Report Lags after Voluntary and Involuntary Auditor Chan

Electronic copy available at: http://ssrn.com/abstract=1806345Electronic copy available at: http://ssrn.com/abstract=1806345

Audit Report Lags after Voluntary and Involuntary Auditor Changes

Paul Tanyi, K Raghunandan and Abhijit Barua

Forthcoming in Accounting Horizons

Contact Author: K Raghunandan

School of Accounting Florida International University Miami, FL 33199

Tel: 305-348-2582 Email: raghu@fiu.edu

Paul Tanyi is a Ph.D student, K Raghunandan is a Professor, and Abhijit Barua is an Assistant Professor, all at Florida International University

We thank two anonymous reviewers, and the Associate Editor (Karla Johnstone), for their many useful and constructive comments on earlier versions of this paper

Electronic copy available at: http://ssrn.com/abstract=1806345Electronic copy available at: http://ssrn.com/abstract=1806345

Audit Report Lags after Voluntary and Involuntary Auditor Changes

Abstract

We find that the audit report lag is significantly higher for former Andersen clients (that did not follow their Andersen partner to the new audit firm) than for clients voluntarily changing auditors from another Big 5 predecessor for the fiscal year ended December 31, 2002 (the first year with the new auditor for ex-Andersen clients) The differences in audit reporting lags between the two groups are not significant for fiscal years ended December 31, 2000 (the last year before Andersen’s Enron related problems surfaced) or 2003 (the second year with the successor auditor) We also find that clients with voluntary (i.e., non-Andersen) auditor changes have only marginally higher audit reporting lags compared to clients without auditor changes Our results, focusing on a cost component of involuntary auditor changes, thus provide relevant empirical evidence for debates surrounding mandatory auditor rotation We also find that ex-Andersen clients that followed the Andersen partner to the new audit firm had shorter audit report lags than ex-Andersen clients that did not follow their Andersen partner Our findings highlight the importance of individual relationships in the auditing process, and suggest new avenues for future research

Audit Report Lags after Voluntary and Involuntary Auditor Changes

Section 207 of the Sarbanes-Oxley Act (SOX) (U.S House of Representatives 2002) requires the Comptroller General of the United States to “conduct a study and review of the potential effects of requiring the mandatory rotation of registered public accounting firms.” This requirement was spurred by concerns that the “independence of a public accounting firm … is adversely affected by a firm’s long-term relationship with the client and the desire to retain the client” (GAO 2003)

Some recent studies have sought to derive implications about the effect of mandatory rotation by investigating the association between auditor tenure and various measures of audit quality These studies examine measures such as earnings management by clients (Johnson et al 2002; Myers et al 2003), forecast errors (Ghosh and Moon 2005), and the likelihood of issuing going-concern modified audit opinions (Geiger and Raghunandan 2002; Choi and Doogar 2005) Generally, these studies have concluded that contrary to the concerns expressed by legislators and regulators auditor tenure has a positive association with audit quality and that the empirical evidence does not support calls for mandatory rotation The auditor tenure studies note that their results may be relevant to the debate about mandatory auditor rotation

The studies cited above have examined instances where auditor tenure has been impacted

by voluntary decisions of the client (to dismiss the auditor) or the auditor (to resign from a

client) The controversy about auditor tenures, fueled by legislators’ and regulators’ comments

and actions, relates to mandatory auditor turnover While relevant, the findings of auditor tenure

studies noted above do not directly provide evidence about the consequences of involuntary

auditor changes because of the endogenous nature of the decision in voluntary auditor changes (Nagy 2005; Krishnan et al 2007)

The auditor changes that occurred following the collapse of Arthur Andersen in 2002

provide a unique situation of involuntary auditor changes The indictment of the firm in March

2002 created considerable uncertainty about the ability of Andersen to survive which led to forced auditor changes for the vast majority of Andersen’s clients The forced auditor changes that occurred after the demise of Arthur Andersen reflect some elements of the mandatory auditor changes advocated by legislators and regulators This unique setting enables researchers

to provide empirical evidence about auditor decisions and audit quality in the context of

involuntary auditor changes

Some prior studies have examined if successor auditors treated former Andersen clients differently than other clients by focusing on accruals quality and audit opinions (Nagy 2005; Cahan and Zhang 2006; Krishnan et al 2007) These studies compare former Andersen clients with other continuing clients of other Big 4 firms The latter group typically includes only a small proportion of clients that had a voluntary auditor change in the same period as the forced change from Andersen

In this paper we compare the effects of mandatory versus voluntary auditor changes by examining audit report lags—that is, the time lag between the fiscal year end and the date of the audit report We focus on audit report lag (admittedly an imperfect proxy) because it is the only publicly observable quantitative proxy for the extent of auditors’ work.1 Audit report lag, and the associated financial reporting lag, has recently been an issue of significant concern to regulators and the auditing profession (SEC 2002b, 2002c)

1

Audit fees also are publicly observable after February 5, 2001, but audit fees represent a product of both quantity (audit work) and price (dollars per unit of work).

We recognize that clients changing auditors—voluntarily or involuntarily—are quite different from other (continuing) clients Hence, we compare the audit report lags for former Andersen clients and other clients voluntarily changing auditors during 2002 We argue that mandatory auditor changes would lead to higher reporting lags than voluntary auditor changes; hence, the audit reporting lag would be greater for ex-Andersen clients than other clients that voluntarily changed auditors, in the initial year with the successor auditor Our tests thus

examine a cost component of mandatory auditor changes

Some recent studies have sought to differentiate between former Andersen clients by classifying such firms into “followers” (i.e., those that followed their former Andersen partner to

a new audit firm) and “non-followers.” Blouin et al (2007) find that discretionary accruals are greater (that is, higher levels of earnings management are present) for companies that followed their former Andersen partner; further, non-followers were likely to have greater agency and switching related costs Kohlbeck et al (2008) and Vermeer et al (2008) find that audit fees are lower for “follower” former Andersen clients than for non-follower former Andersen clients in the first year with the new auditor In the non-profit sector, Vermeer (2008) finds that switching costs, the financial condition of a non-profit, and the size of the market are associated with the likelihood of a non-profit audit client being a “follower.”

Following prior research, we argue that “follower” former clients of Andersen would be treated differently than the “non-follower” former Andersen clients Further, we argue that

“follower” switches by ex-Andersen clients represent auditor changes in form, but perhaps not in substance This suggests that for a proper examination of the differences arising from voluntary versus involuntary auditor changes the comparisons should be between “non-follower” ex-Andersen clients and non-Andersen clients that switched auditors during the same period In

addition, we also examine the “partner change effect” by comparing the two types of

Andersen clients: we argue that the audit reporting lag will be smaller for “follower”

ex-Andersen clients than for “non-follower” ex-ex-Andersen clients

The latter tests, examining differences between ex-Andersen clients that did or did not follow their Andersen partner to the new audit firm, are also relevant in the context of the

mandatory audit partner change rules of SOX Section 203 of SOX states that:

“It shall be unlawful for a registered public accounting firm to provide audit services to

an issuer if the lead (or coordinating) audit partner (having primary responsibility for the audit), or the audit partner responsible for reviewing the audit, has performed audit services for that issuer in each of the 5 previous fiscal years of that issuer.”

Thus, we also provide empirical evidence about the costs associated with the mandatory audit partner rotation rules of SOX

We test our hypotheses by comparing the audit report lags for the following three groups: ex-Andersen clients that followed their partner to the new audit firm (“followers”), ex-Andersen clients that did not follow their Andersen partner to the new audit firm (“non-followers”) and clients that switched auditors during 2002 from a Big 5 predecessor other than Andersen (“non-Andersen switchers”) Given the rapid demise of Andersen, we limit our analysis to firms with fiscal year ends of December 31, 2002

We find that non-follower ex-Andersen clients had longer audit report lags (62.57 days

vs 56.08 days) than clients of other Big 5 auditors who switched to a new auditor in 2002 This provides empirical evidence related to the differences between mandatory and voluntary auditor changes Further, we find also that audit report lags are, on average, 4.56 days (7.8 percent) lower for “follower” clients than for “non-follower” clients This finding quantifies some of the benefits associated with partner familiarity, or the costs associated with mandatory audit partner changes

Finally, we compare clients without an auditor change against the three types of clients with an auditor change discussed above We find that clients with voluntary auditor changes

have only marginally higher audit reporting lags (p = 10) compared to clients without auditor

changes; in contrast, both types of clients with mandatory auditor changes (ex-Andersen

“followers” and ex-Andersen “non-followers”) have significantly higher (p < 01) audit report

lag than clients without auditor changes Overall, our results provide evidence that voluntary auditor changes lead to modest increases in audit report lags, and that mandatory auditor changes significantly increase the audit report lag when compared with voluntary auditor changes

We recognize that the post-Andersen auditor changes are different from those under a mandatory auditor rotation regime First, under mandatory rotation, everyone knows ahead of time when a change is scheduled Second, incoming auditors following mandated changes also would know that their tenure is limited to a maximum period specified by the rotation regime rules Yet, the failure of Andersen represents a unique situation that captures some elements of a mandatory auditor rotation regime in that the clients were involuntarily required to go with a new auditor at a specified time

Our paper adds to the auditing literature along several different streams First, given the interest of legislators and the public in mandatory auditor rotation, we provide empirical

evidence about differences in audit report lags arising from voluntary and involuntary auditor changes Second, audit report lags have been of recent interest to the SEC, auditors, and public companies; we provide empirical evidence about audit report lags following three different types

of auditor changes: voluntary auditor changes (by clients of other Big 5 firms), involuntary auditor changes in form, but perhaps not in substance (“follower” ex-Andersen clients), and involuntary auditor changes in form and substance (“non-follower” ex-Andersen clients) Third,

our paper adds to research about the personal nature of auditing relationships—namely, clients’ associations with specific audit partners and the production efficiencies arising from such

relationships

The next section discusses the background and develops the hypotheses This is followed

by a discussion of data and method The results follow, and the paper ends with a summary and discussion

BACKGROUND AND HYPOTHESES Mandatory Rotation of Auditors

More than a quarter century before SOX, the Metcalf Committee report had expressed similar concerns about the effects of long tenure on auditor judgments The Staff Report of the Committee on Government Affairs (U.S Senate 1976) noted:

“Long association between a corporation and an accounting firm may lead to such close identification of the accounting firm with the interests of its client’s management that truly independent action by the accounting firm becomes difficult One alternative is mandatory change of accountants after a given period of years…” (p 21)

Even earlier, nearly 50 years ago, Mautz and Sharaf (1961) suggested that long associations with the same client can lead to problems with independence Though Mautz and Sharaf (1961, 208) did not call for mandatory auditor rotation, they noted that “the greatest threat to his [the

auditor’s] independence is a slow, gradual, almost casual erosion of his ‘honest

disinterestedness’.”

Periodically, the SEC continued to express its concerns about the possible adverse effects from long auditor tenures (SEC 1994; Turner and Godwin 1999) However, the SEC did not take any regulatory action related to mandatory auditor rotation

Proponents of auditor rotation suggest that long-term relationships between the auditor and the client could undermine perceptions of auditor independence (U.S Senate 2002) Since the incentives associated with keeping a particular client are smaller, and since there would be another audit firm reviewing the work within a specified period of time, auditors “might be less likely to succumb to management pressure” (GAO 2003)

The opposition to mandatory auditor rotation is based on the fact that effective audits require a thorough understanding of the client’s business and processes; such understanding develops over time and there is a steep learning curve that lasts a year or more Hence, audit quality is likely to be lower in the initial years of an audit (GAO 2003) Along these lines, Loebbecke et al (1989) find that irregularities are more likely in the initial years of an audit engagement

A related point is that the disruption to the client caused by rotation leads to non-trivial commitment of resources (personnel and financial) in educating the auditor about the client’s operational and accounting matters In addition, Geiger and Raghunandan (2002) note an

incentive-related argument: incumbent auditors earn quasi-rents due to the start-up costs

associated with audits and low-balling (the pricing of initial-year audits below economic cost) arises as a natural phenomenon Hence, auditors would be interested in recouping their initial-year losses in the future; consequently, the threats to auditor independence may in fact be higher

in the initial years of an audit engagement

Legislators considered imposing mandatory auditor changes during the hearings that preceded SOX; many former SEC chairs supported such a rule (U.S Senate 2002) Nevertheless,

in the face of opposition from auditors and others, the final rules adopted a compromise With respect to audit firm rotation, Section 207 of SOX required the GAO to “conduct a study and

review of the potential effects of requiring the mandatory rotation of registered public accounting firms” and report back to Congress within one year of enactment of the law However, Section

203 of SOX (cited earlier) required mandatory rotation of the audit partner every five years

Related Research

Some recent studies have examined the association between auditor tenure and various measures of audit quality Johnson et al (2002) and Myers et al (2003) use clients’ abnormal accruals as a measure of audit quality and document a positive association between tenure and audit quality Davis et al (2009) find that firms with both short and long tenure are more likely

to use discretionary accruals to meet or beat earnings forecasts, suggesting that audit quality is lower in firms with short or long tenures Ghosh and Moon (2005) show a positive association between auditor tenure and investors’ perceptions of earnings quality (as measured by the

earnings response coefficient) Geiger and Raghunandan (2002) find, for a sample of bankrupt firms, that going-concern modified audit opinions are positively associated with auditor tenure However, Choi and Doogar (2005) use a general sample of stressed firms and find that there is

no association between auditor tenure and the likelihood of a going-concern opinion

The auditor tenure variable in all of the above studies is impacted by voluntary decisions made by the clients (to dismiss the auditor) or the auditor (to resign) Voluntary auditor changes are quite different from involuntary auditor changes, such as those arising from mandatory auditor turnover (Nagy 2005; Krishnan et al 2007) It is this difference which makes the auditor changes that occurred in the aftermath of Andersen’s failure unique The forced auditor change for ex-Andersen clients provides a unique opportunity to empirically examine issues related to involuntary auditor changes

Audit Report Lag, Auditor Changes, and Andersen’s Demise

Audit report lag, and the associated financial reporting lag, has recently been an issue of significant concern to regulators and the auditing profession (SEC 2002b, 2002c).2 As part of the changes made to financial reporting subsequent to SOX, the SEC reduced the 10-K filing lag (i.e., the number of days from the fiscal year end to the filing of the 10-K with the SEC) starting

in 2003 (SEC 2002c) Krishnan and Yang (2009) find that audit report lags have increased in recent years, both prior to the introduction of new rules shortening the 10-K filing lag and also after the introduction of the new rules

Any auditor change creates disruption; both the client and the auditor incur substantial switching costs In the aftermath of an auditor switch the client “spends a significant amount of resources—both financial and human—educating the new auditor about company operations and accounting matters” (GAO 2003) The new auditor has to learn about the business practices, operations, and financial reporting systems of the client, and incur substantial extra effort

becoming familiar with the client in year following an auditor change (Flanigan 2002) Hence, the audit risk is also higher in an initial year engagement; in turn, this suggests that the auditor would likely perform more work in an initial audit engagement

A former chief accountant of the SEC noted in congressional testimony that “on average, you provide about a third additional hours in the first year” (Turner 2002) To the extent at least

a part of such additional work is done at the end of the fiscal year (as opposed to interim work), the audit reporting lag should be higher following auditor changes This argument also suggests

2

For example, the SEC (2002b) sought to shorten the filing due dates for annual and quarterly reports filed with the commission “as a step in modernizing the periodic reporting system and improving the usefulness of quarterly and annual reports to investors.” Following concerns expressed by auditors and others, the SEC (2002c) modified the proposals but noted that technological advances have made it easier for registrants to “capture, process and disseminate” financial information

that if ex-Andersen clients are to be compared with other clients on audit-related issues in

general, or audit reporting lag in particular, then the appropriate comparison group should be clients that also had (voluntarily) an auditor change around the same time as the (involuntary) auditor changes that occurred due to Andersen’s demise

On October 16, 2001, Enron announced major restatements; events unfolded rapidly, with Enron declaring bankruptcy on December 2, 2001 Almost immediately, legislators and the media began asking questions about the role of Andersen in Enron’s failure Andersen admitted

to shredding Enron-related documents on January 10, 2002 and was indicted for such document destruction on March 14, 2002; the SEC (2002a) issued a special release in which it required companies to obtain extra assurances from Andersen for audit reports signed after March 14,

2002 The special release made references to clients who may wish to leave Andersen and to the need for orderly transitions to new auditors Andersen was found guilty of criminal wrong doing

on June 15, 2002, and the firm responded by formally announcing it would cease to practice.3

Thus, Andersen clients had to involuntarily switch auditors and do so quite quickly Such switches were made somewhat easier by the fact that many Andersen offices were purchased by the other large audit firms (see Kohlbeck et al 2008 for a more detailed description of the

acquisition process) Many former Andersen clients followed their former Andersen partner to the new audit firm, thereby easing the transition

Blouin et al (2007), Kohlbeck et al (2008), Vermeer (2008) and Vermeer et al (2008) use publicly available data about the purchase of former Andersen offices to document

differences between ex-Andersen clients who (likely) followed their former partner to the

successor auditor (“follower”) and those who did not follow their former Andersen partner

3

On May 31, 2005, the U.S Supreme Court reversed Andersen’s conviction

(“non-followers”).4 In the context of our study, it is likely that the typical extra work associated with a new client would be much less for a “follower” client than for a “non-follower” client This is because the prior experience with a “follower” client means that for such clients the audit team will not have to navigate the steep learning curve typically experienced in initial audit engagements That is, for the “follower” ex-Andersen clients the auditor change after Andersen’s demise can be viewed as an auditor change in form, but not in substance This in turn means that the extra audit effort associated with new audit clients should be either non-existent or lower if the client followed the Andersen partner than if the client switched to a different firm with a new audit partner and team

Hypotheses

In this study, we examine if the audit reporting lag differs for clients with mandatory auditor changes when compared clients with voluntary auditor changes.We do this by comparing the audit reporting lags for ex-Andersen clients that did not follow their Andersen partner against audit reporting lags for other clients that changed auditors during the same period This leads to the first hypothesis (in the null form):

H 1 : There would be no differences in audit reporting lag, in the first fiscal year with the new auditor, between “non-follower” ex-Andersen clients and non-Andersen clients changing auditors

The “follower” Andersen clients can be viewed as simply having an auditor change in form, but not in substance Therefore, we expect that “follower” ex-Andersen clients would have

4

We use the qualifier “likely” because publicly available data about the purchase of Andersen offices are only at the firm level; it is possible that a former Andersen partner decided not to go to the new firm

However, such actions by individual partners would introduce a bias against finding differences between

follower and non-follower ex-Andersen clients

a lower audit reporting lag in the first year with a new audit firm than “non-follower”

ex-Andersen clients This leads to the following hypotheses (in the null form):

H 2 : Audit reporting lag in the first fiscal year with the new audit firm would not be higher for “non-follower” ex-Andersen clients than for “follower” ex-Andersen clients

Thus, hypothesis one examines the effect of mandatory, as opposed to voluntary, auditor changes Hypothesis two examines the “partner familiarity effect” and thus provides evidence related to mandatory audit partner rotation

METHOD AND DATA

We compare the audit reporting lags for fiscal year 2002, which is the first year with the successor auditor following Andersen’s demise Further, since Andersen was a member of the Big 5, in our tests we restrict the control group to those clients that had another (non-Andersen) Big 5 firm as the predecessor auditor and had an auditor change in 2002 We use the same model

to explain audit report lags as Krishnan and Yang (2009) and add one specific variable to the regression model depending on the research question of interest.5 Thus, to test the first

hypothesis we use the following regression model:

SQLAG = α 0 + α 1 AA + α 2 EXTRAORD + α 3 SEGMENTS + α 4 FOREIGN + α 5 HIGROWTH

+ α 6 HILITIG + α 7 HITECH + α 8 FINCOND + α 9 LOSS + α 10 GC + α 11 LNTA +

SQLAG = Square root of the number of days between fiscal year end and the date of the

audit report

AA = 1 if the firm is an ex-Andersen client, 0 otherwise

EXTRAORD = 1 if the firm has extra-ordinary items on its financial statement, 0 otherwise

SEGMENTS = square root of the number of business segments

FOREIGN = 1 if the firm has foreign operations, 0 otherwise

HIGROWTH = 1 if the firm belongs to high-growth industries (2 digit SIC codes 35, 45, 48, 49,

FINCOND = Probability of bankruptcy, estimated from Zmijewski’s (1984) model for

non-financial firms (calculated as of the end of the fiscal year)

LOSS = 1 if the firm reports a loss before extra-ordinary items, 0 otherwise

GC = 1 if the firm receives a going concern opinion, 0 otherwise

LNTA = natural log of total assets (measured as of the end of the fiscal year)

Since we are using the same model as Krishnan and Yang (2009), our discussion of the control variables is relatively brief Briefly, audit report lag is expected to be higher for firms

with more complex operations, so we include EXTRAORD, SEGMENTS and FOREIGN in the

model Financial problems and going-concern uncertainties add to the auditor’s work, so we

expect that FINCOND, LOSS and GC will have positive coefficients High growth is typically

with more changes, leading to more new things to audit so we expect the coefficient of

HIGROWTH to be positive We also control for the fact that the auditor’s business risk varies

across industries, and hence include HILITIG and HITECH We include our final control variable, LNTA, for two reasons First, larger clients can exercise greater influence on their

auditor to complete the audit quicker (e.g., Ashton et al 1989) Second, almost all prior

accounting and auditing research includes client size as a control factor

The “treatment” sample for testing hypothesis one includes ex-Andersen clients that did not follow their former Andersen partner to the new audit firm The control sample includes clients that changed auditors from another (other than Andersen) Big 5 auditor during 2002

Our second hypothesis compares audit reporting lags for follower and non-follower Andersen clients We use the same model as for hypothesis one but with two changes First, the sample now includes all (and only) ex-Andersen clients Second, since we are only comparing

ex-two groups of ex-Andersen clients, we replace the AA variable in the model with a variable called FOLLOW that takes a value of 1 when the client follows the former Andersen partner to a

new audit firm, and is 0 otherwise

Data

The Andersen sample consists of all non-financial firms that had a December 31 fiscal year end in 2002, and have data available about (a) audit report lag, audit firm name and audit opinion type for fiscal 2002 available in the Audit Analytics database, and (b) relevant financial and other control variables in the Compustat database The control sample consists of all firms that (a) had a December 31 fiscal year end for 2002, (b) had an auditor change during 2002 from

a Big 5 predecessor other than Andersen, and (c) meet the same data requirements as the

Andersen sample (i.e., data available in Audit Analytics and Compustat databases) We use a