laurion et al - 2014 - u.s. audit partner rotations

audit clients from 1999 to 2001 obtained from three audit firms, Manry, Mock, and Turner 2008, while not examining the direct effects of audit partner rotation, provide evidence suggesti

U.S Audit Partner Rotations

Henry Laurion henry_laurion@haas.berkeley.edu

Alastair Lawrence*

lawrence@haas.berkeley.edu

James Ryans james_ryans@haas.berkeley.edu

Haas School of Business University of California at Berkeley

2220 Piedmont Avenue Berkeley, CA 94720-1900

October 2014

ABSTRACT: The main purpose of audit partner rotation is to bring a “fresh look” to the audit

engagement while maintaining firm continuity and overall audit quality Despite mandatory audit partner rotation being required in the U.S for over 35 years, to-date there has been limited empirical evidence speaking to the effectiveness of U.S auditor partner rotations given that audit partner information is not disclosed in U.S audit reports Using SEC comment letter correspondences to identify U.S audit partner rotations, we provide initial evidence among publicly-listed companies suggesting that audit partner rotation in the U.S supports a “fresh look” at the audit engagement Specifically, we find that audit partner rotation results in substantial increases in material restatements (129 to 135 percent) and write-downs of impaired assets (one percent of market value) Overall, these findings suggest that audit partner rotation supports auditor independence and is an important component of quality control for U.S accounting firms

KEYWORDS: U.S audit partner rotations; fresh look; material restatements; K/As;

10-Q/As; write-downs

JEL CLASSIFICATION: M41; M42; M48

DATA AVAILABILITY: Data are publicly available from sources identified in the article

We have received valuable comments and suggestions from Hans Christensen, Sunil Dutta, Miguel Minutti-Meza, Alexander Nezlobin, Xiao-Jun Zhang, and seminar participants at the University of California at Berkeley

*Corresponding author

“…we do not believe the identification of the engagement partner will provide

meaningful information to financial statement users…”

Center for Audit Quality Comment Letter to the PCAOB (CAQ 2014)

1 Introduction and background

For over 35 years, audit partner rotation has been an important component of quality control for the vast majority of accounting firms that audit Securities and Exchange Commission (SEC) registrants While U.S audit partner rotation has generally reflected a compromise for full audit firm rotation, the main purpose of audit partner rotation is to bring a

“fresh look” to the audit engagement while maintaining firm continuity and overall audit quality (e.g., SEC 2003) Regulators state that partner rotation requirements must balance the need to achieve a fresh look and a need for the audit engagement team to be composed of competent auditors, and hence, there has been significant tension relating to these needs in regards to who should be subject to rotation and how long the partner should remain on the engagement prior to rotating (SEC 2003)

In 1978, the American Institute of Certified Public Accountants (AICPA) first introduced audit partner rotation by requiring the lead audit partner to rotate off the audit engagement of SEC registrants after seven years with a two-year time out (also known as a “cooling off”) period before the partner may return (AICPA 1978) The Sarbanes-Oxley Act of 2002 (SOX 2002) further requires both the lead partner and the concurring partner (reviewing partner) to rotate off the audit engagement of SEC registrants after five years—thus reducing the partner rotation from seven to five years While SOX is silent concerning the time out period, the SEC adopted rules effective May 6, 2003 requiring a five-year time out period for both lead and concurring partners (SEC 2003) Thus, under current standards lead audit partners must rotate off audit engagements for SEC registrants after five years and then sit out for another five years before returning to the audit engagements

Audit partner rotation addresses a concern that long-term relationships between audit partners and clients can create problems related to partner objectivity and independence Partners who are familiar with the audit client can become complacent over the audit engagement term (e.g., Bamber and Iyer 2007) and new partners may help address such complacence by bringing a renewed sense of skepticism as well as additional insights and expertise to the engagement On the other hand arguments and evidence have been advanced that there are costs to audit partner rotation, including that management may influence the new partner selection (Cohen, Krishnamoorthy, and Wright 2010) and that the new partner may have less industry expertise (Daugherty, Dickins, Hatfield, and Higgs 2012) Despite the perceived importance of partner rotation to the independence and quality control of U.S accounting firms, as only audit firm and not audit partner information is disclosed in U.S audit reports (PCAOB 2013), there has been limited evidence examining the effectiveness of partner rotation in the U.S for publicly-listed firms Of the evidence to-date, it generally concludes that audit partner rotation in the U.S decreases audit quality

Specifically, using data of U.S audit clients from 1999 to 2001 obtained from three audit firms, Manry, Mock, and Turner (2008), while not examining the direct effects of audit partner rotation, provide evidence suggesting that audit quality, measured using discretionary accruals, increases with partner tenure Also using proprietary U.S audit firm data—containing engagement-level information from one audit firm for 2002 and 2003—Bedard and Johnstone (2010) find that audit fee realization rates decrease in the year of audit partner rotation, suggesting that new partners invest extra time and effort during the first year on the engagement However, they do not show any tangible benefits to audit partner rotation Moreover, using a proxy for mandatory audit partner rotation for U.S firms who switch audit firms, Litt, Sharma, Simpson, and Tanyi (2014) find lower financial reporting quality, measured using discretionary accruals to meet-or-beat targets and a lower frequency of issuing

going concern opinions, following an estimated audit partner rotation Thus, current U.S evidence examining the effectiveness of partner rotation for U.S publicly-listed firms generally suggests that the fresh look reduces audit quality, supporting the view that audit partner rotation

is bad for audit quality as it results in a loss of client-specific knowledge or industry expertise

Fitzgerald, Omer, and Thompson (2014) also examine the effects of audit partner rotation

in the U.S but among not-for-profit organizations While they do not find any evidence that the likelihood of reporting internal control deficiencies increases in the year of the partner change, they do find some evidence that the likelihood of reporting internal control deficiencies increases in the second year following the partner change In addition, there have been several international studies examining the effects of partner rotation in settings where audit partner data is publicly disclosed (e.g., Australia, China, Germany, Taiwan); however, these studies provide mixed evidence concerning the effects of partner rotation (e.g., Carey and Simnett 2006; Chi and Huang 2005; Chen, Lin, and Lin 2008; Chi, Huang, Liao, and Xie 2009; Gold, Lindscheid, Pott, and Watrin, 2012; Azizkhani, Monroe, and Shailer 2012; Firth, Rui, and Wu 2012; Lennox, Wu, and Zhang 2014) and it is uncertain as to whether the findings from these various international settings generalize to the unique features of the U.S regulations and capital markets

In this paper, we investigate the effects of audit partner rotation among U.S public firms

To identify auditor rotations in the United States, we turn to an unreported fact in the accounting literature that audit partners are copied on correspondences between issuers and the Securities and Exchange Commission (SEC) As stipulated by SOX, the SEC’s Division of Corporation Finance (DCF) now reviews each issuer’s filings at least once every three years, and in practice, reviews are being conducted on average every two years (SEC 2013) As a result, a majority of firms receive SEC comment letters at least once every two years When issuers respond to SEC comment letters, they sometimes copy their audit partner in their

response letters and hence, we identify partner rotations by examining those firms receiving SEC comment letters where different audit partners are copied in adjacent years This framework yields 220 U.S audit partner rotations of 205 public-company clients from 2006 to

2013

Bamber and Bamber (2009) highlight that the effects of audit partner rotation are likely to

be modest and empirical tests must be well-specified in order to detect “an economically material effect, should one exist” In addition, their discussion highlights the importance of analyzing changes in audit quality surrounding the actual audit partner rotation (rather than using partner tenure which often lacks cross-sectional variation) and using sharper measures of audit quality (rather than the typical discretionary accrual and earnings response coefficient measures) when examining the effects of audit partner rotation Thus, our analyses are designed with these considerations in mind, and we use measures of financial reporting quality that we believe will be influenced by the incoming audit partner to directly investigate changes in such measures surrounding the partner rotation

First, if the incoming partner does notice errors or inconsistencies with Generally Accepted Accounting Principles (GAAP) and Generally Accepted Auditing Standards (GAAS) then the frequency of material restatements (e.g., Form 8-K item 4.02) or amendments to previously issued financial statements (e.g., “10-K/A”s, “10-Q/A”s) should increase upon audit partner rotation Moreover, the restatements and amendments should pertain to the prior engagement partner’s financial statements Second, a significant part of the audit process is to enforce impairment standards and that the client is indeed recording timely impairments Hence, if the prior audit partner has become complacent with regards to assessing impairment, the incoming partner will likely have concerns over the impairment of assets Hence, we use material restatements, financial statement amendments, and write-downs of impaired assets to

examine the effects of audit partner rotation, as we believe that these measures will reflect the enhanced efforts and objectivity—should these benefits exist—of the incoming partner

Contrary to prior research, we provide initial evidence among U.S publicly-listed companies suggesting that audit partner rotation in the U.S supports a fresh look at the audit engagement Specifically, we find evidence in the first two years following audit partner rotations suggesting that the new audit partner is responsible for substantial increases in material restatements and write-downs of impaired assets We find that material restatements increase by approximately 129 to 135 percent, and that write-downs of impaired assets increase

by one percent of market value We find that increases in restatements following the partner change relate to asset impairments, deferred taxes, and statement of cash flow classification errors.1 Moreover, we find that increases in write-downs are evident in settings where impairment standards suggest that assets are impaired The foregoing increases in restatements and write-downs of impaired assets suggest that new partners may help address the complacence of the former audit partner by bringing a renewed sense of skepticism as well as additional insights and expertise We also find increases in amendments to previously issued 10-Ks and 10-Qs of approximately 13 to 21 percent; however, these increases appear to be due

to the comment letter process

As we cannot separate mandatory versus voluntary rotations, our findings reflect a combination of both mandatory and voluntary audit partner rotations However, given that audit partner rotation is mandated at least every five years, on average we anticipate that our inferences primarily reflect the effects of mandatory auditor rotations Conversations with U.S audit partners from all four Big 4 firms confirm this notion as they suggest that the vast majority of partner changes are due to mandatory rotations, noting that audit firms work to make a five year commitment because the change can be disruptive to the audit process, both

1 We highlight that the increases are not related to fraud restatements

from the clients’ and auditors’ perspectives, and clients do not appreciate unnecessary audit partner rotations Partner retirements, relocations, medical emergencies, and new public-client responsibilities were stated as the other reasons for audit partner rotation although the audit partners indicated that these rotations are not very common Hence, it appears that our main inferences are reflective of exogenous rather than endogenous audit partner rotations

We perform a series of robustness analyses to collaborate our main findings First, the increase in restatements following the new audit partner can relate to errors missed by the former audit partner or errors missed by the new audit partner Hence, we re-run our analyses

by only including restatements that relate to restatements of the former audit partner’s financial statements, finding similar inferences suggesting that the increase in restatements in the first two years of the new audit partner appear to relate to issues missed by the former audit partner and not restatements relating to inexperience on behalf of the new audit partner Second, we examine whether the audit partner rotations reflect the effects of switching from a non-specialist office to a specialist office We find that our main inferences hold when only include audit partner rotations that do not involve audit office switches, suggesting that our inferences

do not relate to the effects of specialist auditors Third, we use several different control groups for our analyses to ensure that our main inferences are not a result of the comment letter process or a specific set of control firms

Taken together, the study makes the following main contributions First, it contributes to the prior literature by providing initial public-company evidence highlighting that partner rotation in the U.S appears to support a fresh look at the audit engagement This evidence is based on direct measures of audit quality, including restatements and write-downs Audit studies that use discretionary accruals as indicators of audit quality should account for the fact that abnormal accruals that arise through transactions such as write-downs may indicate that audit quality is improving instead of the assumption that such an abnormal accrual (in the

absolute value) indicates a decrease in audit quality Second, this study highlights a framework for future research to identify audit partner identities and rotations in the U.S setting using SEC comment letter correspondences Third, the study speaks to the recent U.S debate over audit firm turnover, highlighting that audit firm rotation is not the only solution to obtaining a fresh look for an audit engagement and that the current fresh look mechanism appears to be working to some degree Lastly, the study’s findings relate to the PCAOB’s recent rulemaking proposals (PCAOB 2009, 2011, and 2013), which propose the disclosure of audit partner names

in U.S audit reports Our analyses highlight that audit rotation is supports a fresh look in settings where the partners generally believe that they will not be publicly identified as the client’s auditor, and hence, our findings cannot speak to whether including audit partners names in the audit report would “prompt engagement partners to perform their duties with a heightened sense of accountability” (PCAOB 2013) However, contrary to the arguments of the Center for Audit Quality, our findings do suggest that audit partner rotations and audit partner identification does provide meaningful information which would be useful to investors, and hence supports the PCAOB’s proposal to disclose audit partner names in U.S audit reports The remainder of this paper is organized as follows Section 2 summarizes the prior literature and develops the main hypotheses Section 3 outlines our research design and data Section 4 reports the results of our empirical analyses, and Section 5 concludes

2 Prior Literature and Hypothesis Development

2.1 Prior Literature

Efforts to improve audits have regularly considered a number of requirements designed to improve audit quality by addressing issues of transparency, accountability, and independence Recent legislation and proposed regulations have included requirements for audit engagement partners to sign audit reports (PCAOB 2009, 2011, and 2013), for an increase in the frequency

of mandatory audit partner rotation in the U.S (SOX, Section 203), and studies to consider the impact of mandatory audit firm rotation (SOX, Section 207) These three practices: audit partner identification, audit partner rotation, and audit firm rotation, are interrelated issues that have been debated and studied for decades with varying degrees of generalizability because of differing institutional factors of the countries and time periods examined

Audit firm rotation in the United States is endogenous by its voluntary nature, and because the costs and benefits of audit firm rotation differ from audit partner rotation, the net benefits of firm rotation and partner rotation are likely not directly comparable Nevertheless, audit firm rotation has been studied both because regulators have considered the requirement in order to improve auditor independence and also because audit firm switches are observable.2These studies generally find that audit firm tenure improves audit quality, concluding that the costs of audit firm rotation likely outweigh the benefits However, a notable exception to voluntary audit firm rotation in recent years in the U.S was the dissolution of Arthur Anderson Research examining firms switching from Arthur Anderson to other audit firms provides contrary inferences to the foregoing studies, suggesting a significant improvement in financial statement quality when firms switched auditors (e.g., Blouin, Grein, and Rountree 2007)

Surveys of auditors concerning audit partner rotation provide conflicting evidence, with some studies providing evidence that audit partner rotation can provide independence and

“fresh look” benefits (e.g., Bamber and Iyer 2007; Beasley, Carcello, Hermanson, and Neal 2009) while other research suggests that management may influence the audit partner rotation process (e.g., Cohen et al 2010) and that partner rotation could actually harm audit quality as incoming partners are more likely to audit firms in unfamiliar industries (e.g., Daugherty et al 2012) and reduce their time spent on activities related to audit quality (e.g., Winn 2014)

2 Such studies include: DeFond and Subramanyam (1998); Gieger and Raghunandan (2002); Johnson, Khurana, and Reynolds (2002); Myers, Myers, and Omer (2003); GAO (2003); Carcello and Nagy (2004); Mansi, Maxwell,

Very few studies have empirically examined audit partner rotation in the United States, primarily because audit partner names are not disclosed, and hence research cannot directly observe audit partner changes separately from audit firm changes Audit partner signatures allow researchers and investors to observe changes in audit partners when there is no corresponding change in firm, but the audit partner signature can also induce reputational incentive effects on audit partner behavior For example, Carcello and Li (2013) indicate improved audit quality through a decline in earnings management after the introduction of engagement partner signatures in the United Kingdom

To proxy for audit partner changes in the United States, where the audit partner changes are not generally observable, Litt et al (2014) use the fifth consecutive year after an audit firm change to proxy for an audit partner change, finding lower earnings quality following a presumed partner rotation evidenced by an increased propensity to meet or beat earnings forecasts using discretionary accruals and decreased propensity to issue going-concern opinions Because firms are selected based on having changed audit firms in order to join the sample, these results may lack generalizability Manry, Mock, and Turner (2008) study audit partner tenure and discretionary accruals, using audit records hand collected from a sample of

90 firms, finding that discretionary accruals decline with partner tenure for small firms, but are not significant for large firms

An issue with both Litt et al (2014) and Manry et al (2008) is the use of accruals as the measure of audit quality, which Bamber and Bamber (2009) note are unsatisfying proxies for audit quality One study that addresses this concern using internal control deficiencies as a measure of audit quality is Fitzgerald et al (2014) in the United States not-for-profit setting, which also provides limited generalizability to the for-profit sector Finally, while not looking

at audit quality per se, Bedard and Johnstone (2010) use a proprietary sample from a single United States audit firm, and find that audit partners invest a significant amount of additional

time following audit partner rotation, although whether the additional partner time transpires to improved audit quality remains an empirical question

Studies of audit partner rotation are facilitated in countries where the audit partner is required to be identified, a factor that intertwines audit partner rotation with audit partner identification Partner identification has led to studies of the effects of rotation using data from Australia (e.g., Carey and Simnett 2006; Fargher, Lee, and Mande 2008; Azizkhani et al 2013), Taiwan (e.g., Chi and Huang 2005; Chen et al 2008; Chi et al 2009), Germany (e.g., Gold et al 2012), and China (e.g., Firth et al 2012; Lennox et al 2014) Institutional differences between these countries and the United States make generalization of these studies’ inferences difficult Carey and Simnett (2006) study audit partner rotation by considering a single year’s cross section of Australian audit partner tenure data from 1995, finding that longer audit partner tenure is negatively associated with financial reporting quality Azizkhani et al (2013) study the relation between audit partner tenure and rotation, and the cost of capital, finding that audit partner rotation is associated with a higher cost of capital Generalization of the Australian studies to the United States is difficult, primarily because audit partner rotation was not mandatory until 2003 Azizkhani et al (2013) only report on audit partner tenure (not rotation) during the post-2003 period, and find insignificant results, and they do not have sufficient data to test the effects of audit partner rotation post-2003

Taiwanese studies also present institutional differences compared to the United States setting Taiwan requires audits to be signed by two engagement partners, whereas firms in the United States have a single engagement partner Audit partner rotation was not mandatory in Taiwan until 2004, whereas in the United States audit partners have been required to rotate since 1978 Chi and Huang (2005) consider the association between audit firm and audit partner tenure using discretionary accruals, finding that earnings quality appears to improve for approximately five years, but then declines with further increases in tenure However, since Chi

and Huang (2005) consider data from 1998 to 2001, a period of time when audit partner rotation was not mandatory in Taiwan audit partner changes may present endogeneity concerns Chen et al (2008) find that longer tenure is associated with lower discretionary accruals, however their data is also from a period when audit partner rotation was voluntary Interestingly, Chen et al (2008) note that more than half of audit partner rotations result in the partner rotating back to the client after a single year time-out period Chi et al (2009) consider the first year of mandatory audit partner rotation in Taiwan (i.e., 2004) and note only weak evidence of lower quality financial reporting Similar to the limitations of related studies, their small sample size and use of accruals-based audit quality proxies appear to provide limited power

2.2 Hypothesis Development

To study the impact of audit partner rotation in the United States, we turn to a unreported fact that audit partners are often copied on correspondence between issuers and the SEC When issuers respond to SEC comment letters issued by the SEC’s DCF, they sometimes copy a number of parties in their response letters, including lawyers, SEC staff members, and auditors Previous SEC comment letter research has focused on aspects of the comment letter process such as the determinants of receiving a comment letter (e.g., Cassell, Dreher, and Myers 2013), the benefits of comment letter disclosures (e.g., Bozanic, Dietrich, and Johnson 2013; Johnston and Petacchi 2014), and the information content of comment letters (Dechow, Lawrence, and Ryans 2014)

heretofore-To move beyond the use of discretionary accruals and earnings response coefficients as proxies for audit quality, suggested by Bamber and Bamber (2009), we consider the observable consequences of auditors’ questioning of management estimates One such area where a “fresh look” by a new audit partner could have a material impact is correcting for errors in the previous financial statements of the former partner Restatements can be achieved with

different levels of severity At the lowest level, a restatement can be achieved by filing an amended financial statement (e.g., a 10-K/A) to correct a prior deficiency More serious errors

to financial statements resulting in material misstatements require an announcement that previously issued financial statements can no longer be relied upon (8-K Item 4.02), and the previous financial statements must be restated (ASC 250, FASB 2009e) Hence, we consider the issuance of amendments and material restatements to prior financial statements as our first measure of audit quality (e.g., Palmrose, Richardson, and Scholz 2004; Hribar and Jenkins 2004; Kinney, Palmrose, and Scholz 2004, Liu, Raghunandan, and Rama 2009)

The “fresh look” provided by a new audit partner is not limited to restatements and amendments, and another area where the independent auditor can exert influence is by identifying the need to impair assets when managers exercise discretion to delay write-downs (e.g., Ramanna and Watts 2012) Using the major impairment provisions of GAAP, Lawrence, Sloan, and Sun (2013) summarize the relation between assets write-downs and the ratio of the carrying value to the fair value of the underlying asset (BTM) required by GAAP They highlight that for assets with strict fair value impairment tests, inventories (ASC 330, FASB 2009a) and indefinite-lived intangibles (ASC 350, FASB 2009b), write-downs are only required by GAAP when the BTM exceeds one, and beyond a BTM of one, asset write-downs are a linear function of BTM However, they note that impairment standards with less strict impairment standards—those for PP&E and finite-lived intangibles (ASC 360, FASB 2009d) and investments (ASC 320, FASB 2009c)—a buffer zone is created and a write-down may not

be required until the BTM is substantially above one Hence, Lawrence et al (2013) operationalize their analyses using the firms’ aggregate BTM, and suggest that write-downs are mandated by GAAP when the firms’ BTM is greater than one and a significant portion of the firms’ assets reflect either inventory or indefinite-lived intangibles They document that the vast majority of significant asset impairments result for firms with BTM ratios greater than one

and significant intangible assets, which they suggest highlight settings where GAAP and GAAS require impairments Hence, following Lawrence et al (2013) we expect a fresh look to

be most important for write-downs in such settings where impairment standards suggest that assets are impaired (i.e., firms with a BTM greater than one and significant intangible assets)

Opponents of audit partner rotation argue that if the new partner has limited industry knowledge and is unable to appropriately challenge management estimates as the outgoing partner, then audit quality will decrease following the audit partner rotation However, if audit industry knowledge does not suffer and the partner rotation provides additional independence and the “fresh look” that the policy is designed to achieve, then audit quality should increase

We agree with Bamber and Bamber (2009)’s suggestion that the effects of audit rotation may

be somewhat modest and empirical tests must be well-specified in order to detect such effects Consistent with the proposed regulatory benefits of the fresh look perspective of audit partner rotation, we expect that if the prior audit partner has become complacent in key components of the audit, and the new partners exhibit independent judgment and rigor, they will find errors or inconsistencies with the financial reporting practices of the client In turn we should see increases in restatements and amendments of prior financial statements as well as increases in GAAP and GAAS mandated write-downs following partner rotation Hence, our two hypotheses are as follows (stated in alternative form):

H1: Audit partner rotation is associated with an increase in financial statement restatements and amendments

H2: Audit partner rotation is associated with an increase in write-downs of impaired assets

3 Research Design and Data

3.1 Research Design

As outlined in the previous sections, we examine the effect of audit partner changes on firm’s financial reporting and audit quality using the frequency of restatements and the amount

of write-downs To examine the effect of audit partner changes on restatements, we estimate the following regression model for the two years before and two years after the auditor rotation:

RESTATEMENT i,t = α 0 + α 1 NEW_PARTNER i,t + α 2 BTM i,t + α3 LEV i,t + α4 ROA i,t

+ Year Fixed Effects + εi,t ,

(1)

where RESTATEMENT is measured using the number (AMEND) and incidence (AMENDD) of amended 10-K and 10-Q filings or the number (REST) and incidence (RESTD) of Audit Analytics restatements which we label as material restatements NEW_PARTNER equals “1” if

the fiscal year corresponds to the first or second year following the partner rotation, and “0”

otherwise We include the book-to-market ratio (BTM), leverage (LEV), and return on assets (ROA) to control for previously documented determinants of restatements Evidence of a

positive and significant coefficient on 𝛼! is consistent with H1 that audit partner rotation is

associated with an increase in financial statement restatements, suggesting that new audit partners catch errors in prior financial statements that were not caught by the previous audit

partner

It is possible that the restatements are related not to the observed partner change, but to the fact that the firm received comment letters from the SEC—the events that allowed us to observe the partner change To confirm that the result is not due to the receipt of comment letters (or time-period effects), we also implement a difference-in-differences strategy, using five matched firms closest in market value, in the same Fama-French 49 industry that also received a comment letter in the new partner year as control firms We use the control firms to calculate the difference in restatements between the new partner (or treatment) group and the control group (the first difference), during the pre-partner change years and during the post-partner change years Next we compare the change in this difference between the pre- and post-partner change periods to obtain the difference-in-differences estimate for the effect of the new partner We highlight that to avoid excessively limiting the pool of suitable control firms, we

match firms that switch auditors to firms where we are unsure whether auditor rotation has occurred (i.e., in our main analyses, we don’t require control firms to have year-to-year SEC comment letters indicating no auditor rotation).3 Given that it is possible that the control firm does in fact have an unobserved audit partner change, the differenced model should bias any observe auditor rotation effects towards zero We estimate the foregoing difference-in-differences model as follows:

DRESTATEMENT i,t = α 0 + α 1 NEW_PARTNER i,t + α 2 BTM i,t + α3 LEV i,t + α4 ROA i,t + Year

Fixed Effects + εi,t ,

(2)

where DRESTATEMENT is calculated as the difference in the number (AMEND_d) and incidence (AMENDD_d) of amended 10-K and 10-Q filings or the number (REST_d) and incidence (RESTD_d) of material restatements between the audit partner change firm and the

average of the five matched firms.4 Evidence of a positive and significant coefficient on 𝛼!suggests that the effect of the new partner is incremental to effects of the SEC’s comment letter review and time-period effects in bringing about financial statement amendments or

restatements noted in Equation (1)

To examine the effect of audit partner changes on the amount of write-downs (WD), we

estimate the following regression model for the two years before and two years after the auditor rotation:

WD i,t = α0 + α1 NEW_PARTNER i,t + α2 BTMD i,t-1 + α3 INTA i,t-1 + α4 BTMD i,t-1 * INTA i,t-1

+ α 5 NEW_PARTNER i,t * BTMD i,t-1 + α 6 NEW_PARTNER i,t * INTA i,t-1

+ α7 NEW_PARTNER i,t * BTMD i,t-1 * INTA i,t-1 + Year Fixed Effects + εi,t ,

The foregoing specification is motivated by Lawrence et al (2013) who highlight that assets appear to be impaired under GAAP when the firm’s opening BTM ratio is greater than

one (BTMD) and a significant portion of the firms’ assets reflect either inventory or

indefinite-lived intangibles They find that the majority of all significant impairments occur among firms with BTM ratios greater than one and with significant intangibles The idea is that the fraction

of goodwill and intangibles to total assets (INTA), which are subjected to a strict fair value test,

appear impaired when they compose a significant portion of the book value and the market value is trading significantly below book value Hence, we model write-downs as a function of

NEW_PARTNER, BTMD, and INTA, and the respective interactions between these three

variables We expect that if the new partner requires downs of assets then the downs will be most evidence where assets appear impaired, which we capture using the three-

write-way interaction term (NEW_PARTNER x BTMD x INTA) Hence, a positive and significant

coefficient on 𝛼! is consistent with H2 that audit partner rotation is associated with an increase

in write-downs of impaired assets

Again, it is possible that the write-downs are related not to the observed partner change, but to the fact that the firm received a comment letter To confirm that the findings are not driven by the SEC comment letter review process, as with amendments and restatements we implement a difference-in-differences design, where we difference the write-downs of the audit partner change firm with those of five matched control firms using the procedure described above for Equation (2) We estimate the foregoing difference-in-differences model as follows:

DWD i,t = α 0 + α 1 NEW_PARTNER i,t + α 2 BTMD i,t-1 + α3 INTA i,t-1 + α4 BTMD i,t-1 * INTA i,t-1

+ α 5 NEW_PARTNER i,t * BTMD i,t-1 + α 6 NEW_PARTNER i,t * INTA i,t-1

+ α 7 NEW_PARTNER i,t * BTMD i,t-1 * INTA i,t-1 + Year Fixed Effects + εi,t ,

(4)

where DWD indicates that write-downs is calculated as the difference in write-downs between

the audit partner change firm value and those of the matched firms Evidence of a positive and

significant coefficient on 𝛼! suggests that the SEC’s comment letter review or time-period effects are not responsible for any documented increase in write-downs of impaired assets

noted in Equation (3)

3.2 Data

We determine partner rotation using the Audit Analytics Comment Letter Database of SEC comment letter correspondences spanning from 2006 through to 2013 We use the Audit

Analytics field WEB_GRP_PPL_COPIED, which provides a parsed list of the names and

associated firms of those copied on the correspondence, where available We manually identify audit firms and code an observation as having a partner rotation if the copied audit partner name changes from the prior year, while the copied audit firm name stays the same Using audit fee data from Audit Analytics, we identify 17 audit firms with market share ranked in the top ten from 2000 to 2013 Using a text search for the names of these audit firms, we identify 5,988 comment letter responses that contain the name of an auditor In some cases there are multiple names from the same audit firm copied on one letter

We assume that if the company copies one or more individuals from its audit firm, that the lead audit engagement partner will be one of the copied individuals Therefore the date of the comment letter response is assigned as the “Service Date” for the individual who is copied

We manually check a sample of 50 identified names, using Google and LinkedIn, to confirm that the names are in fact audit partners We use the Compustat annual file to assign firms’ fiscal years to as many Service Dates as possible For every firm year with at least one Service Date, we create a list of all names copied in comment letter responses during that year We assume these are the individuals serving the company for that fiscal year (quarterly reviews, internal control evaluations, and year-end audits) In order to define a within-firm partner change, “Partner Change”, we require that the list of names in one firm-year does not share any names with another firm-year and that all names belong to the same accounting firm We

identify 109 cases of a Partner Change from one fiscal year to the next and 111 cases where there is a Partner Change with one fiscal year separating a partner change.5 We assume that for the Partner Changes with an information gap year in the middle, the old partner’s last year is the year of the information gap and the new partner’s first year is the year in which her name is

copied We create a variable called NEW_PARTNER, which is equal to “0” in the two years

preceding a new partner and “1” in the new partner’s first year and the following year Appendix B provides an example of how we use the comment letter correspondence to identify audit partner rotations Conversations with auditors confirm that clients “cc” the audit partner who will be responsible for the subsequent 10-K audit opinion

We obtain write-down and fundamental data from Compustat, including total assets (AT), beginning of year book value (SEQ), leverage (LEV), defined as the sum of current and long- term debt divided by total stockholders' equity ((DDL t + DLTT t ) / SEQ t)), and market value

(PRCC_F * CSHO) Our write-down variable, WD, is computed from Compustat as the sum of

asset and intangible write-downs plus goodwill impairments, scaled by the beginning of year

market value, and we multiply by -1 so that write-downs are positive (-(WDP t + GDWLIP t) /

(CSHO t-1 * PRCC_F t-1)) We obtain restatement data from two sources Amended filings are identified from the SEC index files, where we record the dates of all Form 10-K/As and 10-Q/As Because amended filings may contain either restated financial information or minor changes to disclosures, we consider this the least severe type of restatement Secondly, we consider restatements from the Audit Analytics Non-Reliance Restatements database, which includes items from both Form 8-K, item 4.02 disclosures, as well as other filings such as Forms 8-K, 8-K/A, 10-K, 10- Q, 10-Q/A, 10-K/A, that contain variations of the word “restate” and upon examination are confirmed to involve a restatement of previously issued financial statements Because Audit Analytics’ definition of restatements includes revisions to financial

5

statements, including non-reliance announcements, we consider these restatements to be more material restatements than amendments

4 Results

4.1 Descriptive Statistics and Univariate Results

Table 1 provides descriptive statistics for the audit partner change firms and for Compustat firms as a comparison We note that the representative partner change firm is

considerably larger than the all-Compustat sample, with mean (median) total assets (ASSETS)

in millions of $51,939 ($3,635) for the partner change sample and $11,465 ($344) for the Compustat sample (unreported) This difference corresponds to results noted in Cassell et al (2013) and Dechow et al (2014), that larger firms are more likely to receive a comment letter, perhaps due to the reporting complexity of larger firms and to the fact that the SEC aims to conducts more frequent examinations of economically significant firms As a result, we expect

all-to observe comment letters two years in a row more often for larger firms To provide a closer comparison in Table 1, we compare the descriptive statistics for the partner-change sample to all Compustat firms with a market value greater than $1 billion The mean (median) total assets

(ASSETS) in millions for the market value greater than $1 billion Compustat sample is $38,663

(4,625) (hereinafter “Compustat firms”)

The fact that the partner-change firms are more similar to large firms adds to the economic significance of our results, since the average effect of audit partner rotation is not confined to smaller firms, who although are great in number have a limited economic impact

when measured by market value Mean (median) market values (MV) are slightly lower for

partner change firms than Compustat firms with mean (median) values of $11,671 ($2,486) and

$12,403 ($3,563), respectively The percentage of Compustat firms with Big 4 auditors (BIG4)

is 78.4 percent whereas the percentage of partner-change firms with Big 4 auditors is 90.9

percent, which is not surprising given that we restrict our analysis to audit firms with market shares in the top ten from 2000 to 2013 The closing mean (median) book-to-market ratios

(BTM t) are 0.754 (0.794) for the partner change firms, higher than the 0.660 (0.669) observed

for Compustat firms—inferences concerning opening book-to-market ratios (BTM t-1) are

similar The fraction of intangibles to market value (INTA t-1) is slightly lower for partner change firms with mean (median) values of 0.149 (0.068) for partner change firms and 0.157

(0.073) for Compustat firms Leverage ratios (LEV t) are also slightly lower for partner change firms with mean (median) values of 0.827 (0.374) for partner change firms and 0.890 (0.503)

for Compustat firms Partner change firms are less profitable (ROA t), with the mean (median) return on assets for partner change firms being 0.031 (0.036) versus 0.057 (0.050) for Compustat firms

Material restatements are similar for partner change firms and Compustat firms, with mean (median) values of 0.061 (0) and 0.064 (0) Audit Analytics restatements per firm-year

(REST t) for partner change and Compustat firms, respectively, and mean (median) values of

0.060 (0) and 0.059 (0) for the incidence of Audit Analytics restatements (RESTD t) for partner change firms and Compustat firms, respectively The number of amended filings per firm-year

(AMEND t), is slightly higher for Compustat firms than for partner change firms, with mean (median) values of 0.250 (0) and 0.236 (0) for Compustat and partner change firms,

respectively The incidence of amendment filings (AMENDD t) however, is lower for Compustat firms than for partner change firms, with mean (median) values of 0.170 (0) and

0.184 (0) for Compustat and partner change firms, respectively Lastly, write-downs (WD t) are larger for partner change firms than for Compustat firms, with mean (median) write-downs of 0.018 (0) for partner change firms and 0.008 (0) for Compustat firms

Table 2 provides univariate results illustrating an increase in restatements, amendments, and write-downs between the pre-partner change and post-partner change periods Specifically,

Audit Analytics restatements (REST) increase from 0.041 per firm-year prior to the partner change to 0.084 per firm-year after the audit partner change, an increase of 102 percent (p < 0.01) Moreover, the incidence of Audit Analytics restatements (RESTD) increase from a mean

value of 0.039 firm-years prior to the partner change to 0.084 firm-years after the audit partner

change, an increase of 114 percent (p < 0.01) The rate of amended filings (AMEND) increase

from 0.204 per firm-year prior to audit partner changes to 0.271 per firm-year after the audit

partner change, an increase of 33 percent (p < 0.05, one-tailed) The incidence of amended filings (AMENDD) increase from 0.165 prior to audit partner changes to 0.205 after the audit partner change, an increase of 24 percent (p < 0.10, one-tailed) Univariate results therefore

indicate that there is a significant increase in both restatements and amended filings after an audit partner change Panels A and B of Figure 1, graphically illustrate these findings for Audit

Analytics restatements (REST) and amended filings (AMEND), respectively

Considering the impact of a partner change on write-downs as a fraction of market value,

the level of write-downs (WD) increases on average from 0.016 prior to the partner change to

0.021 after the partner change, an increase of 32 percent, though this increase is not statistically

significant (p = 0.22, one-tailed) While the magnitude of the effect appears economically

significant, the limited sample size appears to limit the power to be able to observe a statistically significant effect of a partner change on write-downs on a univariate basis Moreover, we only expect write-downs of impaired assets to significantly increase so we lose power when we examine total write-downs Panel C of Figure 1, graphically illustrates the level

of write-downs relative to the audit partner change year, with elevated write downs in the partner change year and the first year after (years 0 and 1)

To ensure that our findings are robust to the effect of receiving a comment letter on restatements, amendments, and write-downs, we also study these variables by differencing the partner change firms’ variable with the mean of five control firms who also receive comment