chang et al - 2010 - market reaction to auditor switching from big 4 to third-tier small accounting firms

Keywords: market reaction; auditor switching; Big 4; small accounting firms; audit quality.. This is partly because investors have perceived that brand name auditors, i.e., the Big N共Big

Market Reaction to Auditor Switching from Big 4 to Third-Tier Small Accounting Firms Hsihui Chang, C S Agnes Cheng, and Kenneth J Reichelt

SUMMARY: After the demise of Arthur Andersen, the public accounting industry has

witnessed a significant migration of public clients to second-tier共Grant Thornton andBDO Seidman兲 and smaller third-tier accounting firms While prior literature documentsthat smaller auditors are perceived by the stock market as an inferior substitute for aBig 4 auditor, this perception appears to have changed in recent years In this paper, weanalyze market responses to auditor switching from Big 4 to smaller accounting firmsduring 2002 to 2006 We break our sample period into two separate periods共Periods 1and 2兲 based on when regulatory changes occurred These changes includedSarbanes-Oxley 共SOX兲 404 implementation, Public Company Accounting OversightBoard共PCAOB兲 inspections, and a tightened Form 8-K filing deadline We find a rela-tively more positive stock market reaction to clients switching from a Big 4 to a smallerthird-tier auditor in Period 2 This relatively more positive reaction in Period 2 reflectscompanies seeking better services rather than a lower audit fee, when an audit qualitydrop is less likely Overall, our results suggest that companies and investors havebecome more receptive to smaller accounting firms

Keywords: market reaction; auditor switching; Big 4; small accounting firms; audit

quality.

Data Availability: Data are publicly available from the sources identified in the paper.

INTRODUCTION

After the failure of Arthur Andersen, regulators around the world were concerned about the

degree of concentration in the audit profession by the Big 4 In a speech at the AmericanInstitute of Certified Public Accountants共AICPA兲 annual national conference on Decem-ber 11, 2006, Mark Olson, Chairman of the Public Company Accounting Oversight Board, ex-

Hsihui Chang is a Professor at Drexel University, and C S Agnes Cheng is a Professor and Kenneth J Reichelt is an Assistant Professor, both at Louisiana State University.

We gratefully acknowledge the comments and suggestions of Dan Simunic 共the editor兲, two anonymous reviewers, Steven Lin, Kannan Raghunandan, Gordon Richardson, Sam Tiras, Angela Woodland, and workshop participants at Florida International University, Louisiana State University, National Cheng Kung University, the 2007 International Research

Conference for Accounting Educators in Mexico City, and the 2008 Joint Journal of Contemporary Accounting and Economics and Auditing: A Journal of Practice and Theory Symposium, and in particular the discussant at that Sympo-

sium, Clive Lennox We thank Menghistu Sallehu for his excellent research assistance, and directEDGAR for the use of their data and software.

Editor’s note: Accepted by Dan Simunic.

November 2010

pp 83–114

Submitted: September 2007 Accepted: December 2009 Published Online: October 2010

pressed his concerns of audit firm concentration, as well as his hope for improvement from thegrowth in the size and skill of smaller accounting firms共PCAOB 2006兲 Similar confidence hadalso been expressed by Christopher Cox, Chairman of the U.S Securities and Exchange Commis-sion, on December 5, 2005 共SEC 2005兲, and by PCAOB Chairman William McDonough onOctober 17, 2005共Civils 2005兲 Both officials encouraged smaller public companies to considerhiring smaller accounting firms.

Despite the regulators’ encouragement, and their belief that many smaller accounting firmshave improved their competence, the audit market continues to be dominated by the Big 4 audi-tors This is partly because investors have perceived that brand name auditors, i.e., the Big N共Big8/6/5/4兲, provide higher quality audits than smaller accounting firms, as suggested by prior litera-ture共DeAngelo 1981;Dopuch and Simunic 1982;Nichols and Smith 1983;Teoh and Wong 1993兲.Consequently, the stock market has perceived switches from a Big N auditor to a non-Big N as a

“red flag”共Eichenseher et al 1989;Dunn et al 1999;Knechel et al 2007兲

Prior to 2002, some auditor switches, regardless of auditor type, involved companies thatwere experiencing going-concern and/or auditor disagreements These companies were consideredriskier, evident from poorer economic performance共Dhaliwal et al 1993兲, higher auditor litigationrisk 共Krishnan and Krishnan 1997; Shu 2000兲, greater likelihood of violating debt-covenants

共DeFond and Jiambalvo 1993兲, and greater likelihood of a going-concern audit opinion 共Krishnanand Krishnan 1997;Geiger et al 1998兲 From a risk perspective, some of these studies suggest andshow that the stock market responded negatively to the announcement of an auditor switch共e.g.,Dhaliwal et al 1993;Shu 2000兲

From a beneficial reason perspective, recent studies共e.g., Ettredge et al 2007; ruswamy and Whisenant 2004兲 document that auditor switches are undertaken for seeking auditfee savings and/or better services Under such circumstances, a nonnegative or even a positivemarket reaction could be expected for a switch from a Big 4 to a non-Big 4 auditor Specifically,

Sankaragu-if共1兲 investors can justify a net economic benefit from a tradeoff between a perceived downgrade

in audit quality and a combined lower audit fee and better personalized services, or共2兲 the markethas changed its perception about the lower audit quality of non-Big 4 auditors If the first case istrue, the market reaction will:共1兲 neither be negative nor positive when the market perceives anequivalent trade-off between higher audit quality and a combined lower audit fee and improvedpersonalized services, and共2兲 be positive when the net economic benefit is significantly positive

If the second case is true, however, it is likely that we will observe a positive market reaction due

to a perceived positive net economic benefit arising from a combined lower audit fee and betterservices, without a compromise in audit quality

We believe the market’s perception of switching from a Big 4 to a Small auditor has changed

in recent years共2004–2006兲 for the following three main reasons

Increase in Seeking Better Services and a Lower Audit Fee from Smaller Auditors

After the demise of Andersen and the enactment of the Sarbanes-Oxley Act共SOX兲, a rary capacity constraint resulted This exogenous shock provided an opportunity for Big 4 auditors

tempo-to rebalance their client portfolios tempo-toward better-aligned former Andersen clients, evident fromincreased sensitivity to client mismatch共Landsman et al 2009兲 Consequently, for smaller clientsBig 4 audit services may have deteriorated and audit fees may have increased so much that smalleraccounting firms appeared more attractive for better services and lower audit fees Consistent with

this notion is a Wall Street Journal article that reports that after SOX, more IPOs are relying on

smaller accounting firms and investors are more receptive to smaller accounting firms 共Reilly

2006兲.1

Decline in Perceived Audit Quality Differences between Big N and Non-Big N Auditors

After the Big N’s involvement in a series of accounting scandals in the early 2000s, investorsviewed a smaller difference in audit quality between Big N and non-Big N auditors This state-ment is consistent with the analysis byChoi et al.共2008兲that the audit quality of the Big N andnon-Big N is expected to converge as the legal and regulatory regime becomes more onerous.During 2004, more onerous legal and regulatory changes were introduced, including SOX 404audit requirements, PCAOB inspections of auditors, and a tighter 8-K filing deadline Evidently,the initial PCAOB inspections of the Big 4 reported disappointing news of the Big 4 audit quality,since 21 of 64 clients examined eventually restated their financial statements.2Thus, a non-Big 4auditor should be an optimal choice in order to avoid paying a brand-name audit fee premium,especially if the brand-name auditor does not provide higher audit quality

Lack of Evidence of Opinion Shopping by Switching to a Small Auditor

Prior studies indicate that opinion shopping is generally unsuccessful 共e.g., Chow and Rice

1982;Krishnan 1994;Krishnan and Stephens 1995; Geiger et al 1998兲; nonetheless, companiesare likely to switch to a non-Big 4 auditor if they believe it will be successful Since PCAOBinspections create quality pressures for non-Big 4 auditors, Small auditors are especially cautious

to not provide a low-quality audit As a result, in recent years it is less likely that opinion shopping

is the motivation for a switch from a Big 4 to a Small auditor

While these three reasons are intuitively appealing, whether investors have actually changedtheir perception remains an empirical issue If switching from a Big 4 to a Small auditor is indeedfor beneficial reasons and the stock market does not perceive a significant drop in audit quality, weshould not see a negative market response Furthermore, assuming public companies can find acompetent Small audit firm, and the company’s management believes that the market will not reactnegatively, we should see an increasing trend of switching from Big 4 to Small audit firms In arecent report, Grant Thornton provides empirical evidence that the market does not respondnegatively to auditor switches from the Big 4 to Grant Thornton共Whisenant 2006兲 However, theGrant Thornton report does not examine Small audit firms, which are different from the Big 4 andMedium 2 in many respects; namely, Small audit firms are better adapted to local markets, aretraditionally viewed as inferior, and are substantially smaller.3Thus, the report’s findings may not

be applicable to switches from the Big 4 to Small audit firms Therefore, our study focuses onauditor switches from Big 4 to Small audit firms We seek to empirically document the marketresponse to these auditor switches and assess whether the evidence supports the regulators’ belief

in the competence of Small audit firms

Analyzing data on auditor changes for the period 2002–2006, we find that the ratio ofswitches to Small audit firms out of all the switches from the Big 4 has been increasing Moreimportantly, the market response to clients switching from a Big 4 to a Small audit firm共BtS兲 isnonnegative and is relatively more positive for the period after August 23, 2004 This relatively

1 On average, we believe Big 4 auditors should provide better audit quality than smaller auditors However, if a Big 4 auditor does not warrant higher audit quality 共e.g., nonexpertise兲, then a switch to a smaller auditor may not be perceived by the market as a bad move.

2 Most of these restatements arose from the balance sheet classification of borrowings from revolving credit agreements

共 PCAOB 2004a , 2004b , 2004c , 2004d 兲.

3 Based on audit fees filed in 2006, the Big 4 occupy about 92.6 percent of the total market share, followed by BDO Seidman 共1.7 percent兲 and Grant Thornton 共1.6 percent兲 The rest 共about 643 firms reported in the audit fee data set from Audit Analytics 兲 are Small audit firms, the highest occupying only 0.2 percent.

more positive market response occurs when the predecessor auditor’s quality is lower共the cessor is a nonspecialist and accruals quality is more inferior兲, suggesting that there is less like-lihood for audit quality to drop from switching to a Small auditor We find the relatively morepositive response to a BtS switch is driven by better services,4not by a lower audit fee, when there

prede-is less likelihood for audit quality to drop

Our study extends the literature in several ways We find a relatively more positive marketresponse to clients switching from a low-quality Big 4 auditor to a Small auditor, particularlywhen the Small auditor provides better services, thus extending Sankaraguruswamy andWhisenant共2004兲, who find a relatively more positive response to service and fee reasons withoutexamining the quality of the auditor switch We document a nonnegative market reaction toswitches from a Big 4 auditor to a Small auditor, as well as with switches from a Big 4 to aMedium 2 and to another Big 4 in the post-SOX 404/PCAOB inspection era, thus extendingKnechel et al.共2007兲, who find a negative response to switches from a Big 4 to a non-Big 4 in theperiod 2000 to 2003 In addition, after intended improvements in financial reporting quality andaudit quality共e.g., SOX 404 and PCAOB inspections兲 in recent years, we find that the market doesnot perceive an audit quality drop when companies switched from a “low-quality” Big 4 auditor to

a Small auditor Specifically, we find that the market responded relatively more positive共i.e., lessnegatively兲 to switches from a Big 4 to a Small auditor, especially when the Small auditor isconsidered a specialist and when the client experienced low audit quality from the Big 4 auditor.This finding is consistent with our conjecture that the market is starting to perceive that Smallauditors do not necessarily deliver inferior audit quality Small auditors may even be able toimprove audit quality共possibly as a specialist providing more personal attention兲 when the Big 4predecessor did not provide high audit quality

The remainder of our paper is structured as follows The next section reviews prior literature

of the market reaction to auditor switches and articulates our prediction of the market response toBig 4 to Small auditor switches The third section presents and discusses our empirical results andthe fourth section concludes the paper

PRIOR LITERATURE

In order to gain perspective as to the overall changes in investor attitudes regarding auditorswitches, we must examine the results from both before and after the SOX/Andersen period Earlystudies of stock market reactions to auditor switches provide mixed results Fried and Schiff共1981兲find a negative market reaction to auditor switches between 1972 and 1975 for companiesthat switched, based on a 21-week window However,Nichols and Smith共1983兲find when using

a shorter eight-week window, that there was no significant reaction to auditor switches between

1973 and 1979 Similarly,Johnson and Lys共1990兲fail to find a significant stock price reaction forauditor switches from 1973 to 1982 using a three-day window around the time of the auditorchange announcement

In contrast, later studies tend to document a negative market reaction to auditor switches.Smith共1988兲examines auditor switches between 1975 and 1982, using a one-week event window,and finds a negative and significant reaction when a new auditor has not yet been appointed.Eichenseher et al.共1989兲find a significant negative market reaction to auditor switches from Big

8 to non-Big 8 firms by OTC companies between 1980 to 1982, using a five-week event window.DeFond et al 共1997兲find a significant negative market reaction to auditor resignations between

1982 and 1987, using three event windows: from the auditor resignation date to the 8-K filingreceipt date, from the receipt of the 8-K filing through the following four business days, and the

4 We use a relative measure of Small auditors’ expertise to reflect better services.

sum of these two periods.Wells and Loudder共1997兲find a significant negative market reaction toauditor resignations between 1988 and 1991, using a two-day window Similarly, Dunn et al.共1999兲find a negative market reaction to auditor resignations between 1988 and 1993, on the date

of the resignation letter, and the reaction is more negative from the loss of a Big 6 auditor Inaddition,Shu共2000兲finds a significant negative market reaction to auditor resignations between

1985 and 1996, using a three-day window Whisenant et al 共2003兲 find a significant negativemarket response to auditor switches arising from weaknesses in internal controls and problemsrelated to the reliability of management representations and/or financial statement reliability be-tween 1993 and 1996, and their results are robust using a three-day window and a seven-daywindow Beneish et al 共2005兲 extend Whisenant et al 共2003兲 by controlling for confoundingevents around the time of the auditor change announcement 共e.g., earnings and managementchanges兲 and by limiting the analysis to resignations.Beneish et al.’s共2005兲study uses a three-daywindow from 1994 to 1998 and only finds a negative market response to resignation announce-ments where a reason was provided for the change共disagreements over accounting treatment orover the adequacy of internal controls兲 Resignation announcements without a reason have aninsignificant market response Recently, Knechel et al 共2007兲 report that Big 4 clients whoswitched to a non-Big 4 auditor during the period 2000 to 2003 experienced a negative abnormalreturn

While prior market studies have generally found a negative response to auditor change nouncements, we have seen no studies that document a positive market reaction to the recentmigration of clients from Big 4 to smaller third-tier auditors AlthoughSullivan共2006兲examinesthe trends and determinants of auditor switching behavior to explain that the migration arises fromhigher costs from the SOX Section 404 requirements, the analysis does not include an event study.Louis共2005兲provides indirect evidence from market reactions to merger announcements, findingthat acquiring companies audited by non-Big 4 firms outperform the market reaction to thoseaudited by Big 4 firms.Louis共2005兲suggests that non-Big 4 audit firms have superior knowledge

an-of local markets and they have a better relation with smaller clients However, since the timeperiod studied 共1980–2002兲 predates the SOX/Andersen era and the study does not examineauditor change announcements, the findings are less relevant to the more recent potential changes

in market attitudes Sankaraguruswamy and Whisenant 共2004兲 provide some evidence that themarket response is relatively more positive for service and fee reasons, using a seven-day window,albeit the time period examined, 1993 to 1996, predates the SOX/Andersen era.Whisenant共2006兲examines whether the market responds negatively to auditor switches from a Big 4 or ArthurAndersen to Grant Thornton, during the period 2002 to 2006 The author finds a significantpositive market response共p ⫽ 0.08兲, with an 11-day window, for auditor switches from a Big 4共excluding Arthur Andersen兲 to Grant Thornton However, the study is restricted to switches toGrant Thornton clients, so it does not include switches to Small共third-tier兲 firms, and it uses rawreturns without controlling for the market movement and risk

Recent literature suggests a nonnegative or even a positive market response to a switch from

a Big 4 to a Small auditor may be imminent when the Big 4 predecessor does not provide highaudit quality In recent years, anecdotal evidence in the financial press共e.g.,Reilly 2006兲 suggeststhat investors are more receptive to companies using smaller auditors Evidence of a higher auditfee demanded by Big 4 over Small auditors 共Ho and Wang 2007兲 and service-related reasons

共Louis 2005兲 suggest that Small accounting firms may be an attractive audit alternative When theswitch is not likely to reduce audit quality, the market response should be nonnegative

EMPIRICAL RESULTS Frequency and Trend of Switching from Big 4 to Small Accounting Firms

Based on 8-K filings compiled by Audit Analytics, we first analyze the frequency and trend ofauditor switching for the years 2000 through 2006 Since this study focuses on switches from aBig 4 to a Small accounting firm, we consider only clients that already have an auditor before thechange and exclude initial public offerings Table1 summarizes the auditor switching frequencyand trend.5Panel A presents the number of switches between size categories of accounting firms.Panel B reports the number of switches by year, across 2000 to 2006, and the number of switchesfrom Big N to Small auditors Similar to Panel B, Panel C reports for publicly traded companiesthat have filed an auditor change

Table1, Panel A, reports that more auditor switches are from Small auditors共48 percent兲 than

from Big 4 auditors 共33 percent兲, and more switches are to Small auditors 共62 percent兲 than to Big

4 auditors共30 percent兲 Panels B and C indicate that the total number of auditor switches washighest in 2002, when Arthur Andersen failed However, the number increased from 2003 to 2004,when the SOX Section 404 requirements became effective, and it further increased from 2004 to

2005 for publicly traded companies Panel C also reports that in 2006 most Big 4 departures are

to a Small auditor共52 percent兲, and the percentage has been increasing since 2002

Final Sample with Return Data

To evaluate the market reaction to auditor switches, we collect daily returns from the Centerfor Research in Security Prices共CRSP兲 for each case of an auditor change filed after the demise

of Arthur Andersen共January 2002兲.6

Our sample period contains two separate periods: Period 1 isprior to August 23, 2004, and Period 2 is post-August 23, 2004 Our choice of these two periods

is motivated by three regulatory events: the implementation of the SOX Section 404 requirements

on auditing of internal controls over financial reporting 共ICOFR兲, the first PCAOB inspectionreports, and the change in the Form 8-K filing deadline

Auditing Standard No 2共for SOX 404 audits兲 was issued by the PCAOB and approved by theSEC on June 17, 2004 Accelerated filers 共those in excess of $75M market capitalization兲 arerequired to engage an auditor to audit their ICOFR, beginning with fiscal years ending November

15, 2004 Since auditing of ICOFR is very costly, clients may switch auditors Hence, the reasonfor switching may differ when firms consider the Section 404 requirements

The first PCAOB inspection reports were announced August 26, 2004 These inspections arerequired by Section 101 of the SOX to ensure a minimum共and a likely higher兲 level of auditquality The PCAOB is required to conduct inspections of firms with over 100 clients every yearand inspections of firms with 100 or fewer clients every three years Hence, the market perception

of audit quality of Small auditors may have changed due to the awareness that Small auditors aresubject to the PCAOB’s tight inspections.7

Effective August 23, 2004, the SEC tightened its 8-K filing deadline from five to four businessdays for notifying an auditor change and several other events Studies have shown that the com-pliance rate was low under the five-day deadline requirement共Schwartz and Soo 1996;Carter andSoo 1999兲 After the SOX Section 404 requirements took effect, we suspect that the compliancerate improved, since auditors may consider compliance with filing deadlines part of ICOFR, and

5 Data are from Audit Analytics as of January 2007.

6 Grant Thornton’s study by Whisenant 共2006兲 uses observations from January 1, 2002, to May 23, 2006 Our sample

period is similar except we exclude January filings since they include switches to Arthur Andersen, and we extend our

period to December 31, 2006.

7 We do not claim that the PCAOB has improved the audit quality of Small auditors We simply suggest that there may

be a potential change in the market perception.

penalties for noncompliance have increased Using August 23, 2004, as the cut-off date to separateour sample into two periods roughly coincides with the legislative events discussed in this and thetwo previous paragraphs To keep our observations homogenous, we use August 23, 2004, as ourcut-off date.

TABLE 1 Analysis of Auditor Switches Panel A: Number of Auditor Switches between Size Categories of Accounting Firms

% Departed

Switches from Big 4 to Small

Switches from Big 4 to Small

Our final sample requires observations to have both departed and engaged auditors, a filingdate, a return, and market value data.8This limitation generates a total of 1,824 observations forour final sample This sample is used for our multiple regression analyses and is not limited toswitches from a Big 4 accounting firm, but includes all other categories of audit switches Table2reports the number of switches by period, industry, and auditor switching category For Period 1,there are 1,121 switches; ignoring 666 mandatory switches from Arthur Andersen, there are only

455 switches For Period 2, there are 703 switches.9

From Table 2, Panel A, we observe that durable manufacturers, financial, and related firms account for approximately 50 percent of the final sample in both periods Panel Bindicates that switches from the Big 4 constitute the largest number of switches: 378 out of 455non-Arthur Andersen switches in Period 1, and 571 out of 703 switches in Period 2 Among theswitches from Big 4 auditors, the majority of the switches are from a Big 4 to another Big 4auditor in Period 1 However, in Period 2, switches to a Small auditor outnumber switches to aMedium 2 and a Big 4 auditor

computer-Univariate Analysis of Market Return during the Event Window

To measure whether a company’s stock market return resulted from a triggering event and notfrom a change in the aggregate stock market level, we use the market-adjusted return10 for ourunivariate analysis Market-adjusted return is defined as the buy and hold daily return less thevalue-weighted market daily return, accumulated over the event window We focus our analyses ofmarket responses using a five-day event window共⫺4,0兲, where day 0 is the filing date Ideally, weshould identify the first day that the market is aware of the auditor change共e.g., news announce-ment兲 and then design an event window surrounding the news announcement date 共e.g., ⫺1 to ⫹1day兲 We searched the Form 8-K auditor change filings and find that a small number 共n ⫽ 152, or

8 percent兲 had filed a press release We find that the mean filing date is 1.47 days after the newsannouncement date共standard deviation ⫽ 2.48兲 Our event window 共4,0兲 is based on the fact thatfour days after the news announcement falls within one standard deviation, and that the SECrequires that the 8-K should be filed within four business days of the auditor change.11,12Table3reports market responses to auditor switches from Big 4 to Small auditors共BtS兲, Big

4 to Medium 2共BtM兲, and Big 4 to Big 4 共BtB兲 for the two periods For each switch group, Table

3 provides the number of observations, the mean daily return, the mean market-adjusted return,and their respective p-values

Panel A indicates that in Period 1 the market responded nonnegatively to switches from BtS,BtM, and BtB The mean daily return and the mean market-adjusted return are not significant atconventional levels, except for the BtB mean daily return共significant at 10 percent, one-tailed兲.Panel B indicates a nonnegative market response in Period 2 BtS auditor switches earn a 0.8percent mean daily return共significant at a 10 percent level, one-tailed兲, BtM earn 0.7 percent 共notsignificant兲, and BtB earn 0.5 percent 共significant at a 10 percent level, one-tailed兲; however, themean market-adjusted return is not significant at conventional levels for all three groups As well,the BtS switch group reports a higher mean daily return and a higher mean market-adjusted return

8 We deleted one observation with the most extreme positive return.

9 When we drop switches from Arthur Andersen, our main conclusions do not change.

10 We also use size-adjusted and market model-adjusted abnormal returns; our conclusions are not altered.

11 For robustness purposes, we use a longer event window 共⫺11,0兲, and we find our results are stronger for Period 1 but weaker for Period 2 These findings are consistent with our conjecture that there is improved compliance in Period 2, which deserves further investigation.

12 When the news announcement date is substituted for the filing date of a smaller sample of Big 4 to Small auditor switches 共n ⫽ 23兲, daily returns and market-adjusted returns are virtually unchanged.

TABLE 2 Analysis of Final Sample Panel A: Number of Observations by Industry and Period

Panel B: Number of Observations by Switching Category and Period

a Industry groups are based on the following SIC codes: Agriculture 共0100–0999兲, Mining and Construction 共1000–1999兲, Food 共2000–2111兲, Textiles and Printing 共2200–2799兲, Chemicals 共2800–2824, 2840–2899兲, Pharmaceuticals 共2830–

2836 兲, Extractive 共1300–1399, 2900–2999兲, Durable Manufacturers 共3000–3999兲, Transportation 共4000–4899兲, Utilities 共4900–4999兲, Retail 共5000–5999兲, Financial 共6000–6999兲, Services 共7000–8999兲, Computers 共3570–3579, 3670–3679, 7370–7379 兲, and Other is all other SIC groups.

Period 1 is February 1, 2002, to August 23, 2004 Period 2 is August 24, 2004, to December 31, 2006.

Big 4 accounting firms are Deloitte, Ernst & Young, KPMG, and PricewaterhouseCoopers Medium 2 are BDO Seidman and Grant Thornton Small accounting firms are all other firms except Andersen.

than the other groups, but the differences are not significant Results in Table 3suggest that themarket response to switches from Big 4 to other audit firms is nonnegative in our sample periods.Our univariate analysis is restricted to a small sample and does not control for the underlyingreasons for auditor switching We conduct multiple regression analyses by using the whole sampleafter we examine the underlying reasons and firm characteristics for the switches in Periods 1 and2.

Comparison of Reasons and Firm Characteristics

Since most company 8-K auditor change filings do not report the auditor change reasons,13weemploy observable measures as surrogates for our major reasons of interest共i.e., audit fee reduc-tion and better services兲 We also select reported 8-K reasons that may imply opinion shopping if

we find a significant number of observations共⬎5 percent兲

We propose that the market will not respond negatively to an auditor switch if it is not likely

to imply a decrease in future audit quality Under this situation, the market may respond relativelymore positive to BtS switches if the switch is associated with an audit fee reduction or betterservices

13 We find that 22 percent 共Period 1: 13 percent, Period 2: 38 percent兲 of our sample observations report either related reasons 共e.g., going-concern, internal controls, financial restatements, audit opinion concerns兲, auditor-related reasons 共e.g., scope limitations, lack of independence, and auditor merger兲, or fee-related reasons in the auditor change 8-K filings.

client-TABLE 3 Market Responses to Auditor Switches—Univariate Analysis

Mean Market-Adjusted Return is the mean Daily Return less the mean CRSP value-weighted market return.

p-values are two-tailed tests of no difference from 0.

To assess if a switch is not likely to imply a decrease in audit quality, we examine the auditinferiority of the predecessor auditor We presume that if the audit quality of the predecessor islow, then there is little room for the successor to lower the quality There may even be greaterroom for the successor to improve audit quality Consequently, we suggest that predecessor auditinferiority is consistent with a low likelihood of a drop in future audit quality We use twovariables to measure predecessor audit inferiority: nonexpertise and accruals inferiority.

To examine if the market response to a BtS switch is relatively more positive due to a loweraudit fee or better services, given a low likelihood of an audit quality drop, we condition themarket response to an audit fee reduction or better services on accruals inferiority and nonexper-tise of the predecessor Below we discuss these measures

associ-a Big 4 speciassoci-alist This implies thassoci-at if associ-a client switches from associ-a Big 4 nonspeciassoci-alist to associ-a Smassoci-allauditor, it is likely that the loss of audit quality共if any兲 will be lower than a switch from a Big 4specialist to a smaller auditor Hence, we use nonexpertise of the predecessor auditor to surrogatefor the low likelihood of an audit quality drop after the switch

We define an industry specialist as the audit firm with the greatest market share in audit feesfor a particular city, two-digit SIC and year, based onFrancis et al.共2005b兲 Non_Specialist equals

1 if the predecessor auditor is not an industry specialist, 0 otherwise We use the city of the auditorstated on the auditor’s report and the audit fee, from Audit Analytics, to construct this variable.Table4 provides some evidence that BtS switchers are less likely to reflect an audit qualitydrop than BtM and BtB switchers in Period 2 BtS switchers have a relatively higher ratio ofnonspecialist predecessors in Period 2 and the smallest decrease over the two periods In Period 2,BtS have the largest ratio of nonspecialist predecessors共0.407兲, which is significantly greater thanfor BtB共0.300兲 Comparing Period 2 to Period 1 for BtS, the Non_Specialist ratio decreases by

0.030共not significant兲, which is less than the decrease for BtM 共0.107兲 and BtB 共0.129兲

Accruals Inferiority Prior to Audit Switch

A second variable to measure the low likelihood of an audit quality drop after a switch is theclient’s accruals inferiority Numerous studies have used accruals quality to measure audit quality共e.g.,Becker et al 1998;Francis et al 1999;Chung and Kallapur 2003;Srinidhi and Gul 2007兲.These studies assume that the client’s accruals quality is associated with audit quality sinceaccruals are audited However, accruals may measure audit quality with error For example, aclient with poor accruals quality may hire a brand-name Big 4 auditor to convey that their earningsquality is high 共Francis et al 1999兲 If the Big 4 auditor fails to detect accruals-based earningsmanagement when the client has a higher propensity to manage earnings 共Chung and Kallapur

2003兲, we may observe higher reported accruals Thus, the propensity to manage earnings createsmeasurement error for accruals quality in measuring audit quality

There is little debate that auditor expertise measures auditor quality, hence, audit quality, so

we use the Non_Specialist variable as one surrogate of audit quality While prior studies show that

accruals quality is associated with auditor expertise共e.g.,Balsam et al 2003;Krishnan 2003兲, theclient’s propensity to manage earnings may reduce accruals quality, thus creating measurement

error in the Non_Specialist variable to measure audit quality We believe that both a nonspecialist

variable and an accruals inferiority variable measure audit quality with error, so we propose to use

TABLE 4 Comparison of Selected Mean Reasons and Firm Characteristics between Big 4 Switching

Groups Panel A: Period 1 Means

BtS (n ⴝ 71) (n BtM ⴝ 81) (n BtB ⴝ 226) BtS less BtM BtS less BtB

*,**,***Significant at 0.10, 0.05, and 0.01, respectively All tests are two-tailed.

See the Appendix for variable definitions.

an interaction variable共i.e., Non_Specialist ⴱ Accruals_Inferiority兲 Our results are stronger using

this interaction variable and our main regression analysis only reports results using this interactionvariable to measure the low likelihood of an audit quality drop

Our accruals inferiority measure derives fromDechow and Dichev共2002; hereafter DD兲 Intheir model, accruals quality is the extent to which accruals map into operating cash flow realiza-tions This measure has been used to assess audit quality共e.g.,Srinidhi and Gul 2007;Chen et al

2008兲 Audit quality reflects the auditor’s role in reducing estimation errors of accruals Oneproblem with this measure is if the variation contained inDD’saccruals quality measure stemsfrom client operational characteristics, then it will reflect the firm’s underlying economics and willnot measure accruals quality properly.Francis et al 共2005a; hereafter FLOS兲 point out that theDD-based accruals quality measure consists of two components: an innate accruals quality com-ponent resulting from client operational characteristics, and a discretionary accruals quality com-ponent resulting from managerial earnings opportunism共and noise兲 We use their method and baseour audit inferiority measure on the portion of accruals inferiority that is less likely from theclient’s operational characteristics

FollowingFLOS, Accruals_Inferiority is defined as the residual of annually estimating the

DD-based accruals quality measure共AQ兲 from the following OLS model 共firm subscripts are not

shown for brevity兲:

AQ t=␭0+␭1lAssets t+␭2␴共CFO兲 t+␭3␴共Sales兲 t+␭4OperCycle t+␭5NegEarn t + u t 共1兲where:14

lAssets t ⫽ the natural logarithm of total assets 共#6兲 in year t;

␴共CFO兲 t⫽ the standard deviation of operating cash flow 共#308兲 divided by average assets

over years t−9 to t共with at least five years兲;

␴共Sales兲 t⫽ the standard deviation of sales revenue 共#12兲 divided by average assets over

years t−9 to t共with a least five years兲;

OperCycle t⫽ the natural logarithm of the operating cycle length 共sum of days receivables, and

days inventory兲 in year t;

NegEarn t ⫽ the number of years out of the past ten in year t where net income before

extraordinary items共#123兲 ⬍ 0;

u t⫽ the error term;

AQ t ⫽ accruals quality in year t, the standard deviation of the residual from estimating

the following Equation共2兲by year t and two-digit SIC:

TACC t = b0+ b1CFO t−1 + b2CFO t + b3CFO t+1 + b4⌬REV t + b5PPE t+␧t 共2兲

where:

TACC t⫽ total accruals 共net income before extraordinary items 共#123兲 less operating cash

flow共#308兲兲 in year t divided by average total assets 共average of current and

prior year #6兲 in year t;

CFO t ⫽ operating cash flow 共#308兲 in year t divided by total assets in year t;

⌬REV t ⫽ the change in sales revenue 共#12兲 between year t−1 and t divided by total assets

in year t;

14 The Compustat Annual variable number is in parentheses.

PPE t ⫽ gross value of PPE 共#7兲 in year t divided by total assets in year t; and

␧t⫽ the error term

AQ tis the standard deviation of the residuals from Equation 共2兲 from years t−4 to t and is computed for each firm in year t 共with at least three observations available兲 Year t is the last fiscal

year audited by the predecessor auditor

A larger standard deviation of residuals 共AQ t兲 from Equation 共2兲 indicates poorer accrualsquality, since there is greater uncertainty of total accruals mapping into operating cash flow

realizations Innate_Accruals_Inferiority is the predicted value of accruals quality from Equation

共1兲, the accruals quality predicted by client operational characteristics Accruals_Inferiority is the

residual from Equation共1兲, the unpredicted accruals quality potentially due to managerial earningsopportunism

Table4, Panels A and B, report that BtS switchers have similar Accruals_Inferiority to BtM

switchers, and to BtB switchers in both Periods 1 and 2 There is no significant difference in meanaccruals inferiority in both periods.15

Audit Fee Decrease

Prior literature has shown that client characteristics affect the size of the audit fee共e.g., Hay

et al 2006;Ettredge et al 2007兲 Typical determinants of audit fees are client size, complexity, andaudit risk We construct an audit fee decrease variable based on the residual of an audit feeestimation model共Ettredge et al 2007兲 Specifically, we estimate the audit fee by year t from the

following model共firm subscripts are not shown for brevity兲:

laf t=␤0+␤1lta t+␤2segment t+␤3for_sales_pc t+␤4rec t+␤5inv t+␤6roa t+␤7loss t

+␤8opinion t+␤9ltd t+␤10dec t+␤11opinion_lag t + industry dummies t + e t 共3兲

where:

laf t ⫽ the natural logarithm of audit fee in year t;

lta t ⫽ the natural logarithm of total assets 共#6兲 in year t;

segment t ⫽ the natural logarithm of the number of business segments in year t;

for_sales_pc ⫽ foreign sales as a percentage of total sales in year t; t

rec t ⫽ total receivables 共#2兲 divided by total assets in year t;

inv t ⫽ inventory 共#3兲 divided by total assets in year t;

roa t ⫽ earnings before extraordinary items 共#18兲 divided by total assets in year t; loss t ⫽ 1 if net income 共#172兲 in year t ⬍ 0, 0 otherwise;

opinion t ⫽ 1 if the auditor issues a going-concern opinion in year t, 0 otherwise;

ltd t ⫽ long-term debt 共#9兲 divided by total assets in year t;

dec t ⫽ 1 if the fiscal year t ends December 31, 0 otherwise;

opinion_lag t ⫽ the number of days between the fiscal year-end t and the audit report date; industry

dummies t ⫽ indicator variables from two-digit SIC categories in year t; and

e t⫽ the error term

15 Innate accruals inferiority does significantly differ between switching groups 共see Table 4 兲, since client operational characteristics vary.

Audit fee, audit opinion, and audit report date are obtained from Audit Analytics, and all othervariables are from Compustat Annual and Compustat Annual Segment data Following prior

studies, client size is estimated by lta, complexity is estimated by segment, for_sales_pc, and dec, and risk is measured by the remaining variables Fee_Decrease equals 1 if the actual audit fee in

the last year of the predecessor auditor is greater than model共3兲 predicts and if the actual audit fee

in the first year of the successor auditor is less than model共3兲 predicts, 0 otherwise.16

Table4reports that BtS switchers have significantly more fee decreases than BtB switchers inboth Periods 1 and 2

Better Services

Prior literature finds that clients switch from larger auditors to seek better services共guruswamy and Whisenant 2004兲 Big 4 auditors are reputed to favor larger and more lucrativeclients, and consequently neglect smaller clients共Louis 2005兲 A Wall Street Journal article 共Ber-ton 1994兲 provides insight from complaints by former small clients of Big N firms, which includeslow/delayed responses to clients’ questions, less experienced and less cooperative audit staff,higher turnover of audit staff, and poorer tax advice

Sankara-Compared to a Big N auditor, a Small auditor has fewer layers of management, allowingprompter client advice For publicly traded clients, Small auditors are more likely to staff theengagement with more experienced staff that pay more personal attention These factors create abetter working relationship, improving the auditor’s knowledge of the client, building the auditor’sexpertise, and improving audit quality.17

We develop a measure of better services based on the Small auditor’s expertise Differentfrom the measure of expertise for the Big 4, we measure expertise based on the Small auditor’swebsite and industry concentration of the number of switches to the same Small auditor Thismeasure cannot be compared directly to the expertise measure we use for the Big 4共i.e., Smallauditor expertise does not necessarily imply better expertise than a Big 4 nonspecialist兲; it is ameasure to distinguish among Small auditors who will more likely provide better services

We examine the auditor’s website to determine whether the auditor indicates that they haveexpertise in the client’s industry If there is not enough information on the website, we contactedthe auditor’s office by telephone For Small auditors with six or more BtS client switches, weexamine whether there is industry concentration among the switches to a particular Small auditor.The rationale is that if more switches come from clients in the same industry, then it is likely thatthe Small audit firm is an industry specialist We use different benchmark concentration ratios forSmall auditors with different numbers of switches to determine if the auditor is likely an industryspecialist For Small auditors with six to ten BtS switches, we require at least a 50 percent industryratio共i.e., at least three to five switches are from clients in the same industry兲; for auditors with 11

to 20 BtS switches, we require at least a 40 percent industry ratio; and for auditors with more than

20 BtS switches, we require at least a 35 percent industry ratio.18The variable pert equals 1 if the successor Small auditor is identified as an industry specialist for BtS switches,

Small_Auditor_Ex-0 otherwise Table4, Panels A and B, report the mean Small_Auditor_Expert ratio is 0.437 and

0.350 for all BtS switches for Periods 1 and 2, respectively

16 Our results are qualitatively similar when we use a continuous measure instead of an indicator variable of fee decrease.

17 Informal interviews with several CFOs from publicly traded companies who switched from a Big 4 auditor to a small auditor confirm that their working relationship significantly improved with the smaller auditor, and the improved working relationship will likely improve audit quality.

18 Our design of the benchmark ratio is based on the absolute number of clients in the same industry rather than a strict ratio We also vary our benchmark ratio, and our results get weaker when our benchmark ratio is too low.

8-K Reported Reasons

Previous research共e.g.,Geiger et al 1998兲 suggests that clients attempt to switch auditors inorder to opinion shop, albeit unsuccessfully We previously discussed that accruals inferioritymeasures the low likelihood of an audit quality drop Accruals inferiority may also reflect opinionshopping: clients with higher accruals inferiority may seek a more tolerant auditor If a switchsuggests opinion shopping, then the market should respond negatively Table4reports that accru-als inferiority decreased from Period 1 to Period 2, for the BtS switchers and for the BtMswitchers, possibly reflecting less opinion shopping intention There is no significant change forthe BtB switchers

In addition to the accruals inferiority measure, we also extract the 8-K reasons that may relate

to opinion shopping 共qualified audit opinions, internal controls, and financial restatements兲 andwhich also have a significant number of observations in our sample共at least 5 percent兲.19

These8-K reasons are required disclosures for audit switches Based on Audit Analytics-coded 8-K

reasons, we identified three opinion-shopping measures: GCAO, AAIC, and AAFR.20GCAO equals

1 if there was not an unqualified audit opinion on the financial statements共adverse, disclaimer,qualified, or modified, including going-concern兲 in the past two fiscal years, 0 otherwise AAIC

equals 1 if there was an internal control issue, 0 otherwise AAFR equals 1 if a restatement of the

financial statements occurred or will occur, 0 otherwise Table4 reports that the percentage of

AAIC and AAFR reasons has increased from Period 1 to Period 2, likely due to the coincident

change in the audit environment共i.e., PCAOB inspections and SOX 404 audits兲 These increases

are statistically significant for all audit switch groups GCAO has increased but not significantly

for BtS and BtM groups However, the increases for these three variables are the lowest for theBtS group

Firm Characteristics

We evaluate various firm characteristics that previous studies 共e.g., Johnson and Lys 1990;Shu 2000兲 have employed to help explain the underlying reasons for auditor switches Thesecharacteristics are size, growth, complexity, need for external financing, and profitability Clientsare more likely to switch to a smaller共larger兲 auditor if they are smaller 共larger兲 in size, havelower共higher兲 growth, less 共more兲 complexity, are in less 共more兲 need of external financing, andhave poorer共better兲 profitability We follow previous studies to measure these characteristics Forsize, we include market capitalization For growth, we include the average past three years’ salesgrowth, and the book-to-price ratio For complexity, we include the natural logarithm of thenumber of business segments and the foreign sales ratio For need for financing, we includeleverage and the change in financing after the auditor switch For profitability, we add a lossindicator and ROA in the last year of the predecessor auditor The variables are defined as follows,

where year t is the last fiscal year audited by the predecessor auditor:

Size⫽ market capitalization 共$billions兲 during the auditor switch, the stock price pershare times the number of shares outstanding共in billions兲;

SG ⫽ average sales growth for years t−2 to t 共sales t/ salest−1+ salest−1/ salest−2

+ salest−2/ salest−3兲 / 3 where salestis sales revenue共#12兲 in year t;

BP ⫽ the book to price ratio, book value 共#216兲/market value 共#199 ⴱ #25兲, in year t; Segment ⫽ the natural logarithm of the number of business segments in year t;

FORSA ⫽ the ratio of foreign sales to total sales revenue in year t;

19 We obtained the 8-K reasons from Audit Analytics and verified them against our hand-collected source.

20 We excluded redundant reasons, such as “Reportable Conditions,” or those which are too infrequent, such as “Consulted with Incoming Auditor.”