the post sox

7/30/2019 The Post Sox

1/27

The Post-SOX Evolution of the

Client Portfolio of the Second Tier:A Focus on Restatement and InternalControl Riskija_454 308..334

R. Mithu Dey and Ashok RobinSaunders College of Business, Rochester Institute of Technology

The Post-SOX era coincides with the rise of the Second Tierauditing firms. Our study tests risk characteristics of new (and

departing) clients versus continuing clients of the Second Tierto judge overall risks faced by this auditing tier. We add to theprior literature by focusing on prevalence and severity ofrestatement and internal control weaknesses (ICW). We showthat the new clients of the Second Tier indeed have higherprevalence and severity of restatements and ICW. Departingclients also show a high risk profile compared to continuingclients, but the difference is perhaps not as strong as thatindicated in the new versus continuing comparison. Weconfirm the widely held belief that the Second Tier is gainingmarket share, but this gain appears to be achieved in asomewhat controlled manner.

Key words: Audit market, auditor choice, internal control, publicaccounting firms, restatements

SUMMARY

Regulators have often voiced concern that theauditing market is concentrated. The recent flowof clients from the Big 4 to the Second Tier hasspawned an additional concern: the Second Tiermight be the recipient of risky clients from the Big4 and that this may place their business at risk. Anearlier and influential study, Hogan and Martin(2009), evaluates the new and departing clients ofthe Second Tier against continuing clients and findsthat (a) new clients, especially those from the Big 4

bring in additional risks mostly because of theirlarger size and (b) these risks are somewhat offsetby the movement of other risky clients from theSecond Tier to lower tier auditors. But this earlierstudy, because of its sampling period, was unable toadequately assess the importance of restatementand internal control weakness (ICW) risks. Our

study fills this void.We show that the new clients of the Second

Tier, when compared to continuing clients, havehigher prevalence and severity of restatementsand ICW. This has two implications. First, thenew clients acquired by the Second Tier are insome sense riskier than implied by an analysisof the traditional variables used in the literature.

Correspondence to: R. Mithu Dey, Saunders College of Business,Rochester Institute of Technology, 105 Lomb Memorial Drive,Rochester, NY 14623, USA. Email: [email protected]

International Journal of Auditing doi:10.1111/j.1099-1123.2012.00454.xInt. J. Audit. 16: 308334 (2012)

ISSN 1090-6738 2012 Blackwell Publishing Ltd


2/27

Second, we contribute to the large literature onauditor switching by demonstrating the potentialimportance of restatement and ICW risks. We arguethat restatement and ICW measures are indicatorsof a perturbed auditing environment which poses

additional risks to the auditor. Thus, traditionalvariables like firm size, return on assets (ROA),loss, accounts receivables and inventory may notfully explain audit risk.

We also show that departing clients of the SecondTier are riskier than their continuing clients,but this result is not as strong as that in the newversus continuing comparison. Combined with theobservation that the number of clients of the SecondTier has held steady in the post-SOX era, thisappears to suggest a deliberate strategy on thepart of the Second Tier to build their clienteleand manage risks prudently. Our sample spans a

number of years in the post-SOX era and thereforeour results go beyond documenting an immediatereaction to Arthur Andersens demise and SOX.

1. INTRODUCTION

The Sarbanes-Oxley Act of 2002 (hereafter, SOX) isthe direct or indirect cause of many noteworthychanges in the auditing market in the UnitedStates. One such change is the rise of the fourso-called Second Tier auditors: Grant Thornton,BDO Seidman, McGladrey, and Crowe. Because ofcapacity constraints at the Big 4, the Second Tiergained market share by signing up clients servicedby the Big 4 as well as other clients who mighthave potentially gone to the Big 4. In the midstof this development, regulators were concernedabout whether Second Tier firms had becomesignificantly risky due to their changing clientele.1

Our study responds to this concern and provides arisk analysis of changes in the Second Tier clientportfolio. Since changes in client portfolios can bebrought about not only by new clients but also bydeparting clients, we perform two comparisons:new versus continuing clients and departingversus continuing clients. Although continuingclients themselves could change in character,an analysis of new and departing clients andespecially a comparison of the two with continuingclients can more directly provide information aboutclient portfolio strategy and the resulting impact onrisk. An earlier study, Hogan and Martin (2009),performed such an analysis, but because of dataconstraints (that study used data ending in 2004),was unable to fully address the impact of two

important risk indicators in the post-SOX era:restatements and internal control weaknesses(ICW). Our study contributes by filling this gapin the literature. We focus on client changesin the Second Tier during a longer post-SOX

period (20042008) which allows us to morecomprehensively compare values of risk variablesespecially those relating to restatements and ICW.

The sampling period is important for evaluatingrisk for the following reasons. The number ofrestatements rose dramatically after SOX andpersisted for a number of years. Academic articles(e.g., Plumlee & Yohn, 2010) and industry reports,2

confirmed by our own check of the restatementsdataset included in Audit Analytics, indicate apeaking of restatements in 20052006 and acontinued level of higher than average annualrestatements. Thus, a sample ending in 2004 such

as the one employed in Hogan and Martin(2009) would not capture many of these post-SOXrestatements. Similarly, because compliance withSection 302 initiated in 2002 and that with Section404 initiated in 2004, the number of firms disclosingICW rose in the years following SOX. Prior to SOX,disclosure of ICW was only triggered by an auditorchange (Krishnan, 2005). As with restatements,a sample ending in 2004 would provide noopportunity to capture the full impact of Section302 and especially Section 404 disclosures relatingto internal control.

Restatements and especially ICW have thepotential to significantly affect auditors riskexposure (e.g., Palmrose & Scholz, 2004; Elder et al.,2009). Further, there is some indication that theserisks are associated with smaller clients, the typeof firms typically audited by the Second Tier. Forexample, Doyle, Ge and McVay (2007) report thatfirms disclosing material weaknesses in internalcontrol are likely to be smaller, younger, financiallyweaker, more complex, growing rapidly orundergoing restructuring. Our own check of theAudit Analytics (AA) restatements database showsthat the distribution of restatements by auditor isskewed toward Second Tier auditors during thetime period we study: their clients account forroughly a fifth of all restatements while their clientmarket share is much lower. Thus an analysis ofclient portfolio risk in the Second Tier withoutconsideration of these factors would be incomplete.We use prevalence as well as severity measures ofrestatements and ICW to examine their associationwith Second Tier client shifts. This is the primarycontribution of our paper.

The Post-SOX Evolution of the Client Portfolio of the Second Tier: A Focus on Restatement and Internal Control Risk 309

Int. J. Audit. 16: 308334 (2012) 2012 Blackwell Publishing Ltd


3/27

Our key results pertain to restatements and ICW.We find that both new and departing clients of theSecond Tier have a significantly higher prevalenceof restatements compared to continuing clients.To measure prevalence of restatements, we use

two variables, an ex-post one and an ex-ante one. Theex-post variable reflects an actual restatementduring the current or previous year. The ex-antevariable reflects the eventual restatement of thecurrent or previous period earnings. Both variablesarguably reflect audit risk. Both variables arestatistically significant in the client switchingmodels that compare characteristics of newor departing clients with those of continuingclients. The results concerning restatements arenoteworthy for two reasons. First, the resultindicating a higher prevalence of restatements fornew clients of the Second Tier differs from Hogan

and Martin (2009), who mostly report insignificantvalues for restatements; at the same time, it issimilar in spirit to findings in Landsman, Nelsonand Rountree (2009) that audit risk variables suchas growth, receivables and issue of going concernopinion are related to Big 4 switches (from Big 4 tolower-tier auditing firms) after SOX and suggestthat restatements are another indication of risk.3

Second, our results also show higher prevalence ofrestatements for departing clients of Second Tierfirms. This suggests that, as the Second Tier firmsgrow, they probably also face capacity constraintsand may therefore actively seek to optimize theirclient portfolios.4

While these are overall results concerning theprevalence of restatements, we also test whetherrestatement severity is related to client switching torule out the possibility that our conclusions aredriven by non-severe restatements such as thoseonly involving reclassification of accounts andnon-core expenses. In models comparing new andcontinuing clients, we report strong evidence thatrestatements that involve core earnings as wellas those that decrease income (our measuresof severity) are associated with switching. Inmodels comparing departing and continuingclients, restatement severity variables are mostlyinsignificant; but when we focus on clientsdeparting to small auditors, we find indication thatseverity matters.

We also find our ICW variables significant inswitches and contribute to the recent and growingliterature concerning internal control. Our firstICW variable captures Section 302 relateddisclosure and our second and main variable

captures Section 404 related disclosure. We findthat the Section 302 (required of all firms) variableis significant in departures from the Second Tierwhile the Section 404 (required of large clientsstaggered over time) variable is significant in new

clients of the Section Tier. As in Elder et al. (2009),we measure severity by taking into account thenumber of weaknesses disclosed and find thatnew clients have more severe ICW than continuingfirms. In contrast, departing firms overall do nothave greater levels of severe ICW, perhaps moreas a reflection of their size rather than their risklevels. Overall, new as well as departing firmsshow greater levels of ICW.

We provide additional tests by dividing switchesinto resignations and dismissals and report thefollowing results. We find that the vast majority ofthe downward switches into the Second Tier are

dismissals (by clients) rather than resignations(by auditors) and that there is no difference insensitivity between risk factors and switching forthe dismissal and resignation samples. The formerresult in particular suggests that the Second Tieris gaining market share not because of clientsdiscarded by the Big 4 but because clients seekthem out. But results from this analysis are onlysuggestive because of possible errors in identifyingwhether an auditor change arises from a dismissalor a resignation.

Our significantly long post-SOX sampling periodallows us to make a secondary contribution:we provide evidence on persistent rather thantemporary shifts in the post-SOX era. Thus we areable to complement the evidence presented inHogan and Martin (2009). Our results show acontinuation of the downshift phenomenon(acquisition of larger clients from the Big 4 anddeparture of smaller clients to smaller auditingfirms) documented by Hogan and Martin.However, there appears to be a deceleration of theexodus of firms from the Big 4 to the Second Tier.Our data show that during 20052008, 256 firms (or60 percent of the new clients of the Second Tierfirms) shifted from the Big 4 to the Second Tier; thiscompares to a figure of 384 firms or 84 percentnoted by Hogan and Martin during the 20012004period. Correspondingly, we note a greater numberof firms shifting from small auditing firms to theSecond Tier, 98 in our time period versus 41 firmsnoted by Hogan and Martin. Overall, there is littlechange in the total number of clients serviced bythe Second Tier firms but the average size of clientscontinues to grow significantly. These results are

310 R. M. Dey and A. Robin



4/27

consistent with (a) client firms possibly seekingbetter pricing and/or service from the Second Tierand (b) an ability of the Second Tier to control clientportfolio risk. The latter result in particular shouldalleviate concerns shown by regulators that the

Second Tier is passively acquiring high risk clientsfrom the Big 4.

Our evidence on the evolution of the auditingmarket in the US has some relevance to certaininternational markets. For example, there is a greatdeal of overlap between UK and US auditorsand both markets continue to be dominated bythe same Big 4 auditors.5 Additionally, in bothcountries, regulators have been concerned aboutthe auditing market in the post-SOX andpost-Andersen era (see, e.g., Abidin, Beattie &Goodacre, 2010). The European Union has alsobeen concerned about audit market concentration

and has been studying the issue in recent years.The 2006 report by the consulting firm Oxera,commissioned by the European Union, suggests arelaxation of ownership rules of auditing firms asa means to lower barriers to entry and to facilitateaccess to financial capital so that entry into largeauditing markets is made possible. The Oxerareport was followed up by another study by theEuropean Union, this time by soliciting commentsfrom market participants.6 The European Unionhas also considered lowering auditor liability as ameans of opening up the auditing market andreducing concentration.7

2. BACKGROUND AND HYPOTHESES

Our paper focuses on shifts in the market forauditing services post-SOX and as such relatesto four streams of literature. The first, auditorswitching, analyzes decisions by firms/auditors toexit/enter relationships. This literature informs usthat audit risks in particular motivate large auditingfirms (Big N) to terminate relationships with clientsand perhaps move these clients downstreamto smaller auditors who do not have as muchreputational capital to lose or deep pocketssusceptible to litigation. It is widely accepted thataudit risks have increased post-SOX and hencethis literature is relevant to our study. The secondand third literature streams separately studyrestatements and internal control. The major themein these literatures is that reporting quality isweakened or perturbed because of restatementsand ICW. Our assumption, one supported bystatements by practitioners (e.g., Turner, Williams

& Weirich, 2005), is that these events areassociated with audit risks which in turn arerelated to auditor switches. The fourth literatureconcerns the market for Second Tier auditors.Major themes in this literature include the quality

of auditing services and the evolution of the clientportfolio (Dey & Robin, 2011). However, there areonly a handful of studies in this literature and wecontribute with additional insights concerning theevolution of the Second Tier market.

Auditor switching

Auditor switching is perhaps viewed best fromthe perspective of an auditing firms portfolio.Johnstone and Bedard (2004) provide anillustration of how this portfolio changes over time.

Changes occur when (a) clients discontinue and(b) when new clients are added. Clients maydiscontinue of their own volition (e.g., when theyperceive fee levels as high and/or services asinadequate) or when the auditing firm perceivesthe client as value reducing (e.g., when risks arehigh and fees are inadequate to cover these risks).From the auditing firms perspective, both clientdiscontinuance and client acceptance decisions arepredicated on the following factors: financial risk,audit risk, auditor business risk and the presentvalue of the future stream of expected audit fees.The first three factors are risks faced by the auditoras a consequence of client characteristics.

Financial risk relates to the financial position oreconomic condition of the client firm and ismeasured using variables such as leverage andROA. Audit risk relates to risks arising from lackof controls, poor management or poor qualityreporting; these risks increase the chance of theauditor not detecting a material misstatement.Auditor business risk refers to risks such as thosestemming from the public or private status of theclient firm: if a client firm is public the auditor facesheightened litigation risk. Overall, many of theserisks map to the litigation risk faced by the auditingfirm. Ultimately, auditing firms are concernedabout shareholder/investor litigation and lossof reputation. In a study of discontinuance/acceptance decisions made by a large auditing firmduring a particular year, Johnstone and Bedard(2004) find that it is audit risk variables that mostlyinfluence the decision.8 This result is of particularrelevance to our study because we use morecontemporary audit risk variables reflecting




5/27

restatement and internal control risks; thesevariables have become important in the post-SOXera.

The importance of risk mitigation by Big N firms(prior to the demise of Arthur Andersen there were

more than four so-called Big firms) is a themefound in earlier papers such as Krishnan andKrishnan (1997). A good example of this literaturestream is Shu (2000), which finds that auditorresignations are related to increases in clientslitigation risk (captured in variables such asinventory, receivables, size, stock volatility andwhether the firm is a technological firm, delistedfirm and received a qualified opinion). Shuconfirms these results by using an event studymethod: stock returns are negative for resignationsand related to changes in litigation risk. Shualso finds that resignations are motivated by

auditor-client fit issues: fit is determined byrelating the probability of being audited by a bigauditor to firm characteristics such as size,acquisitions and new financing. A key result fromShu (2000) that is relevant to our study is thefinding that discontinued firms switch to smallerauditors; furthermore, the greater the increase inlitigation risk, the greater the tendency to switch toa smaller auditing firm. Shu (2000) explains thatsmaller auditors do not risk as much reputationalcapital as larger auditors and also do nothave deep pockets that attract litigation. Thephenomenon of client switches from Big N firms tosmaller auditors was thus established even priorto Arthur Andersen and SOX. This implies thatwhile SOX might have accelerated downward

switches, the long-term post-SOX equilibriummight be characterized by a secular trend ofdownward switching albeit at a lower rate.

A recent study by Landsman, Nelson andRountree (2009) compares auditor switches (from

Big N firms) in the pre- and post-Enron eras. Sincethe post-Enron era is characterized by clients ofArthur Andersen seeking other (usually Big N)auditors, the resulting capacity constraint ishypothesized to change the sensitivity betweenswitches and the twin influencing factors of clientrisk (financial risk, audit risk and auditor businessrisk variables) and client misalignment (size,acquisitions, new financing and so on). Specifically,with the new pool of potential clients following theEnron scandal, Big N firms are perceived to lookmore closely at their current portfolios and pruneout certain clients not aligned with their needs.

This is the insufficient capacity hypothesis. In ananalysis of client switches from Big N to otherauditors, Landsman et al. (2009) show an increasein sensitivity to client misalignment but a decreasein sensitivity to client risk. These results largelysupport the insufficient capacity hypothesis.

Restatements in the post-SOX era

We now turn to a recently acknowledged measureof audit risk: restatements. To emphasize theprevalence and importance of restatements inthe post-SOX era we provide a simple count ofrestatements from the Audit Analytics restatementsdataset. According to Figure 1, the number ofrestatements peaked during 20052007 and is

Figure 1: Number of unique firm restatements by year for Second Tier clients versus non-Second Tier clients.Note: This figure represents the number of unique firm restatements by year for all Second Tier clients versusnon-Second Tier clients. Data source is Audit Analytics.




6/27

double the number during 20022004. This is awell-known fact and has been reported in academicpapers as well as in industry commentary. The highfrequency of restatements triggered a number ofresearch studies inquiring into its causes and

consequences.Interestingly, much of the restatements literature

focuses on its consequences rather than its causes.An exception is Plumlee and Yohn (2010), whoanalyze corporate disclosures and outside newssources related to restatements to ascertain thecauses attributed to restatements. A key finding isthat a majority of restatements during 20032006are attributable to internal company errors. Theauthors state that this finding is consistent with theposition that internal reviews related to SOX areworking. This finding and conclusion appear tosupport the notion that restatements and ICW

flagged by various SOX sections are connected, asconfirmed by our own tests which are reportedlater in this paper.

A large number of studies report on variousconsequences of restatements. The basic conclusionfrom this literature is that restatements matterand that they often have adverse consequencesfor investors, managers, and directors. Specificfindings include: stock prices react negativelyto announcements of restatements (Palmrose,Richardson & Scholz, 2004); labor marketsimpose penalties on directors (Srinivasan, 2005),and managers (Desai, Hogan & Wilkins, 2006). A

particularly interesting stream of research, onerelevant to our work, connects restatements withinformation asymmetry as well as informationrisk. For example, Kravet and Shevlin (2010) findthat a restatement announcement increases the

factor loading on the discretionary information riskfactor and thus increases the cost of capital for afirm. Thus restatements coincide with reportingweakness and have adverse consequences for firmsand their stakeholders; this, in turn, supports theproposition that restatements increase client riskfor auditors. This connection between restatementsand risk for auditors is also supported by studieslinking restatements with auditor change, butmuch of this evidence is preliminary.9

Internal control weaknesses in the

post-SOX eraWe now turn to our next audit risk variable: ICW.Disclosures of ICW increased dramatically in thepost-SOX era (see Figure 2). Although SOX has twoimportant sections pertaining to internal controls,302 and 404, it does not elaborate on the meaningof internal controls. The prior literature on internalcontrols (e.g., Zhang, Zhou & Zhou, 2007) refersto the following definition provided by theCommittee of Sponsoring Organizations (COSO)of the Treadway Commission in their reportpublished in 1992 titled Internal Control Integrated Framework: it is a process, effected by

Figure 2:Number of internal control weaknesses (Section 404) by year for Second Tier clients versus non-SecondTier clients.Note: This figure represents the number of internal control weaknesses, specifically SOX Section 404, by year forall Second Tier clients versus non-Second Tier clients. Data source is Audit Analytics.




7/27

an entitys board of directors, management andother personnel, designed to provide reasonableassurance regarding the achievement of objectivesin: (a) effectiveness and efficiency of operations, (b)reliability of financial reporting, and (c) compliance

with applicable laws and regulations. Of particularrelevance to our work is the reference to financialreporting.

SOX Section 302 (Corporate Responsibilityfor Financial Reporting) requires top managers(the so-called signing officers) to certify quarterlyand annual reports and to take responsibility forestablishing and maintaining internal controls.Significantly, the signing officers are responsiblefor the disclosure of significant ICW in quarterlyand annual filings. SOX Section 404 (ManagementAssessment of Internal Controls) is a morestringent requirement concerning internal controls.

It requires each annual report to contain anassessment by the management of the effectivenessof internal control procedures for financialreporting (404a). Furthermore, it requires theauditing firm to attest to and report on theassessment made by managers (404b). Becausecompliance with Section 404b is perceived as costlyand onerous, only accelerated filers (essentiallylarge firms, with market capitalization of at least$75 million) have thus far been subject to it. In fact,404a compliance started in 2004 for acceleratedfilers and 2007 for non-accelerated filers, while404b compliance started in 2004 for accelerate filersand is yet to kick-in for non-accelerated filers.10 Thisis why a study of internal controls should have asignificant sampling period beyond 2004.

What have we learned about ICW disclosuresfrom the prior literature and how does thisevidence connect with our own analysis? Theliterature focuses on two aspects: the nature offirms making these disclosures and the effects ofthese disclosures. Zhang et al. (2007) report thatdeficiencies are negatively related to financialexpertise in the audit committee and positivelyrelated to auditor independence. While this resultdoes not imply causality, it is quite suggestive: thelack of financial expertise creates deficiencies whileindependent auditors encourage their disclosure.Also, this study finds a correlation betweendeficiencies and auditor change as did theAshbaugh-Skaife, Collins and Kinney (2007) study.Another study in this genre, Doyle et al. (2007)focuses only on disclosures of material weaknessesin internal control, and reports that disclosingfirms are likely to be smaller, younger, financially

weaker, more complex, growing rapidly orundergoing restructuring. Together, these resultsimply that firms disclosing ICW pose a high levelof risk to auditors. This understanding, that ICWand auditing risks are related, leads us to insert

ICW as an explanatory variable in our auditorswitching models. Elder et al. (2009) is the firststudy to explicitly associate auditor switches andICW, but we are the first to do so in the context ofswitches involving the Second Tier.

As there is a literature on the consequences ofrestatements, there is a literature focusing on theconsequences of ICW. The finds are quite similar.Ashbaugh-Skaife et al. (2007) report that firmsdisclosing ICW prior to SOX mandates have a moreperturbed operating and reporting environment:they have more complex operations, recentorganizational changes, greater accounting risk,

and more auditor resignations. Two studies by thesame team of authors, Ashbaugh-Skaife et al. (2008,2009) report that firms reporting SOX mandatedICW have higher idiosyncratic risk, systematic risk,and cost of capital; these firms also have lowerquality earnings because the accrual noise is higher(that is, they have larger positive and largernegative accruals). Overall, the literature suggeststhat ICW matters and that it is associated withreporting weaknesses.

The Second Tier auditing marketOur study focuses on new and departing clients ofthe Second Tier auditors. Although the Big 4 firmscontinue to audit most public firms, Second Tierfirms like BDO and Grant Thornton audit anincreasing number of these firms.11 In particular,their share of small cap listed firms has risen inrecent years. Regulators as well as marketparticipants now attach considerable importance tothe health of these auditing firms.12 Despite thehigh level of regulatory and market interest inthem, few studies have focused on Second Tierfirms. One exception is Cassell et al. (2008) whichexamines the perceived quality of audits conductedby Second Tier firms by testing whether theirclients suffer a cost-of-capital penalty; the resultsreveal the absence of a penalty and confirm thenewfound status enjoyed by these auditing firms inthe post-SOX era. Another study offering a similarconclusion is Boone, Khurana and Raman (2010)which reports no differences in actual audit qualitybetween the Big 4 and the Second Tier.




8/27

Our study is closely related to another study ofthe Second Tier, Hogan and Martin (2009), whichalso evaluates client characteristics for Second Tierfirms following SOX. In particular, Hogan andMartin examine: (a) client portfolio changes for the

Second Tier firms (how clients are acquired andlost, with a special focus on clients acquired fromthe Big 4 firms), and (b) changes in the riskcharacteristics of portfolio firms. A key finding inthis paper is that new clients accepted by SecondTier firms are riskier than their continuing clients.But there are also indications that the Second Tierfirms are shedding some risky clients. Overall,however, it appears that some risks faced bySecond Tier auditing firms are rising. This resultappears to vindicate the concern shown byregulators regarding risks faced by Second Tierfirms. But as we argued earlier, because the Hogan

and Martin sample ends in 2004, it does not capturethe importance of restatements or ICW. Our study,because it uses a longer post-SOX sampling period,is able instead to address these issues. Additionally,our study is able to address the persistent ratherthan the temporary effects.

New and departing clients of the Second Tierand restatement/ICW risks

At a general level, our study assesses the evolutionof the client portfolio of the Second Tier: we focuson the risk characteristics of the new and departingclients of the Second Tier in comparison to those ofcontinuing clients. At a specific level, ours is astudy of how restatement and ICW risks affectclient switching concerning the Second Tier. Below,we list and briefly discuss the three hypothesestested in our study.

Hypothesis 1: Clients who switch to the Second Tier(and especially those from the Big 4) have higherlevels of auditing risks as measured by the prevalenceand severity of restatements and ICW compared tocontinuing clients of the Second Tier.

This hypothesis flows mostly from the literatureon auditor switching (e.g., Johnstone & Bedard,2004) in which the switching decision is relatedto audit risks. While the extant literature focuseson traditional variables such as firm size,accounts receivables, and inventory, we focus onrestatements and ICW which have been made moreprominent in the post-SOX era. Implicit in thishypothesis is the assumption that the switches tothe Second Tier are a consequence of the Big 4s

desire to shed risky clients. This assumption isdiscussed in the prior literature (e.g., Hogan &Martin, 2009) and is connected to the demandshock of SOX and the supply shock of Andersen(e.g., Landsman et al., 2009). The desire of the Big 4

to control portfolio risk is more directly reflected inclient firms switching from the Big 4 to the SecondTier, but it may also be indirectly reflected in smallfirm switches to the Second Tier: one might arguethat these new firms are potential clients of the Big4 but turned to the Second Tier because of a lack ofinterest on the part of the Big 4 to pursue them.Thus, this hypothesis is not restricted to switchesfrom the Big 4 only.

The assumption that the Big 4 is shedding (oravoiding) risky clients and that these clients moveto the Second Tier may not be true for two reasons.First, the post-SOX push by the Big 4 to shed risky

clients may have exhausted itself toward the endof our sample period: that is, the effect may onlyhave been a temporary and not a persistent one.Second, because of learning (audit firms becomingmore efficient) and because of the relaxation ofregulations (e.g., Auditing Standard No. 5), the Big4 may no longer feel the need to offload riskyclients to the Second Tier. Thus, it is an empiricalquestion whether client firms that switch areassociated with higher auditing risks (restatementsand ICW).

Hypothesis 2: Clients who switch from the Second

Tier (and especially those who move to smallerauditing firms) have higher levels of auditing risksas measured by the prevalence and severity ofrestatements and ICW compared to continuingclients of the Second Tier.

Hypothesis 2 is similar to hypothesis 1 and couldpossibly be considered its corollary. Just as the Big4 has incentives to offload risky clients to theSecond Tier, the Second Tier may have incentivesto offload risky clients to smaller auditing firms.Such downward shifting from the Second Tier willdepend on (a) the additional risks posed by thelarger and riskier clients acquired from the Big 4,and (b) capacity constraints faced by the SecondTier. If the Second Tier takes in a large number ofrisky clients from the Big 4, they may face increasedpressure to offload other risky clients to balancerisks. This pressure will be intensified if the SecondTier faces capacity constraints. In some respects,therefore, the propensity of the Second Tier to shedrisky clients could be indicative of its marketpower and especially its control over growth. We




9/27

do, however, recognize the potentially reducedrelevance of the ICW variable (especially the onerelated to Section 404) in tests of hypothesis 2. Thisis because of the higher likelihood that clientsswitching from the Second Tier to smaller auditing

firms are small firms. As explained earlier, firmswith a market capitalization of less than $75 millionare deemed non-accelerated filers and are yet tobe required to comply with Section 404b. Therefore,even though our test variable does not distinguishbetween 404a (management report) and 404b(auditor report) because of the manner in whichthe data is reported in Audit Analytics, it mightoffer explanatory power concerning switchesfrom the Second Tier to smaller auditing firms.While the first two hypotheses are our main ones,we conduct further examination of new clientsversus continuing clients by examining resignation

and dismissal firms separately. This hypothesis isstated and explained below.

Hypothesis 3: Resignation firms (firms whoseprevious auditors resigned) will indicate a highersensitivity to auditing risks as measured by theprevalence and severity of restatements and ICWcompared to continuing firms of the Second Tier; incontrast, dismissal firms (firms who dismissed theirprevious auditors) will not show this highersensitivity.

The process by which firms change auditors iscomplex and it is usually difficult to definitivelystate whether a certain change is initiated by theclient firm or by its auditor. Nevertheless,

information is available in the Audit Analyticsdatabase that allows us to categorize changes intoresignations and dismissals. The former is morelikely to represent action initiated by auditing firmsand the latter is more likely to represent action by

client firms. Accordingly, the link between clientswitches and auditing risk variables (restatementsand ICW) is likely to be more pronounced in theresignation sample compared to the dismissalssample.13 This analysis, of course, is constrained bythe accuracy of Audit Analytics in categorizingresignations and dismissals, and, as such, theevidence could be considered tentative.

3. SAMPLE AND VARIABLES

We obtain our sample for the period 20042008 by

using the following steps. We use the AuditAnalytics Audit Opinions database to identify allfirms audited by Second Tier auditors (GrantThornton, BDO, McGladrey, and Crowe). We thenobtain relevant audit-related data from thefollowing Audit Analytics databases: AuditorChange, Disclosure Control, Internal Controls andRestatements. We then obtained financial data fromCompustat. Our final sample is made up of firmswith non-missing data.

Table 1 provides the number of clients servicedby the Second Tier firms each year during20042008. For each year, the table providesthe number of new clients in the current year(A), those continuing from the previous year

Table 1: New, continuing, and departing clients for Second Tier audit firms

2004 2005 2006 2007 2008 Total

New in current year 131 96 105 94 42621% 16% 17% 15% 17%

Continuing from prior year 497 514 506 519 203679% 84% 83% 85% 83%

Total during current year 607 628 610 611 613 2,456/2,462100% 100% 100% 100% 100%

Departing before next year 110 114 104 92 42018% 18% 17% 15% 17%

Continuing to next year 497 514 506 519 2,03682% 82% 83% 85% 83%

Note: New, continuing, and departing clients are determined using Compustat and Audit Analytics data, asdescribed in the text, and include only publicly traded client firms. New clients are defined as firms previouslyaudited by a non-Second Tier auditor and audited during the current year by a Second Tier auditor. Continuingclients are defined as those continuing with a Second Tier auditor from the prior year. Departing clients aredefined as those departing before the next fiscal year-end to a non-Second Tier auditor. Second Tier audit firmsinclude BDO Seidman; Crowe Horwath; Grant Thornton; and McGladrey & Pullen.




10/27

(Bt-1), total clients during the current year(C = A + Bt-1 = D + Bt), those departing before nextyear (D) and clients continuing to next year (Bt).Since our data start from 2004, items A and B t-1 aremissing for 2004. Although the turnover rate isaround 20 percent, overall, the number of clientsappears to be stable at around 600 firms. Roughly20 percent of the overall client base departseach year and a similar number of new clients areacquired. Our main tests involve the comparisonof new clients with those continuing from theprevious year (that is, we compare A with Bt-1) anddeparting clients with those continuing to the next(D with Bt).

Table 2 provides further information about newand departing clients and allows an understandingof long-term trends. Specifically, we provideevidence on whether the new clients are from Big 4or Small firms (small auditing firms) or othersources. Similarly we provide evidence on whetherdeparting clients go to Big 4 or Small firms, ordepart for other reasons. Table 2 Panel A shows thesource of new clients. Note that the biggest sourceof new clients for the Second Tier firms is the Big 4(256 firms, or 60 percent). However, there is steadytapering off of clients from the Big 4. While therewere 106 new clients from the Big 4 in 2005, there

were only 41 in 2008. This is consistent with asettling down of the effects of SOX and Andersen.

Table 2 Panel B shows information concerningdeparting clients. In addition to client departuresto the Big 4 and Small firms, departures mayalso be induced by bankruptcy, mergers andacquisitions (M&A) or deregistration. The twobiggest categories are departures to Small firmsand departures because of deregistration; theseaccount for almost three-quarters of all departures.Our result confirms the post-SOX pattern ofderegistration (e.g., Leuz, Triantis & Wang, 2008). Atotal of 148 firms (35 percent of departing clients)deregistered during the sample period. We alsonote that the post-SOX deregistration movementappears to have peaked in 2007 (51 percent). Otherthan deregistration, departure of clients to Smallfirms is a major category (160 firms, or 38 percent).

For this sample of Second Tier clients (organizedby client years), we obtain various risk/characteristic variables. Following the priorliterature, we obtain six traditional risk variables asfollows: Assets: total assets in millions (Data6) Leverage: ratio of total liabilities to total assets

(Data181/Data6) ROA: return on assets (Data18/Data6)

Table 2: Further analysis of new and departing clients

Panel A: Sources of new clients

2005 2006 2007 2008 Total

Count % Count % Count % Count % Count %

Big 4 106 81% 56 58% 53 50% 41 44% 256 60%Small firms 14 11% 16 17% 38 36% 30 32% 98 23%Unknown 11 8% 24 25% 14 13% 23 24% 72 17%Total 131 100% 96 100% 105 100% 94 100% 426 100%

Panel B: Destinations of departing clients

2004 2005 2006 2007 Total

Count % Count % Count % Count % Count %

Big 4 7 6% 7 6% 16 15% 10 11% 40 10%Small firms 47 43% 44 39% 46 44% 23 25% 160 38%Bankruptcy 1 1% 0 0% 5 5% 3 3% 9 2%

M&A 13 12% 19 9% 6 6% 5 5% 43 10%Deregistered 38 35% 36 32% 27 26% 47 51% 148 35%Unknown 4 4% 8 14% 4 4% 4 4% 20 5%Total 110 100% 114 100% 104 100% 92 100% 420 100%

Note: This table categorizes new and departing clients. New unknown clients include recently registered firms.For a description of the sample, see Table 1.




11/27

Loss: binary variable, equals 1 if ROA < 0, and 0otherwise

OCF: cash flows from operating activities scaledby total assets (Data308/Data6)

ARInv: ratio of accounts receivable and

inventory to total assets (Data2 + Data3)/(Data6) DSales: percentage change in sales from prior

yearOur main test variables measure prevalence andseverity of restatement and ICW risk. Regardingrestatements, since we are interested in capturingrisks faced by the auditing firm, we consider bothex-ante (eventual restatement of current or previousyears earnings) and ex-post (actual restatementduring the current or previous year) measures.Regarding ICW, we consider both Section 302 andSection 404 disclosures. Below, we define our testvariables (main bullets are prevalence measures,

sub-bullets are severity measures): ICW_302[period]: if a weakness pursuant to SOX

Section 302 is disclosed for the period. ICW_404[period]: if a weakness pursuant to SOX

Section 404 is disclosed for the period. ICW_404COMP[period]: if three or more

weaknesses are disclosed for the period. RST_Exante[period]: if earnings for the period is

eventually restated. RST_Exante_CoreErn[period]: if core earnings

for the period is eventually restated. RST_Exante_Neg[period]: if earnings for the

period is eventually restated negatively. RST_Expost[period]: if there is a restatement

during the period. RST_Expost_CoreErn[period]: if the

restatement involves core earnings. RST_Expost_Neg[period]: if the restatement

decreases earnings.The ICW_302 variable is obtained from theDisclosure Controls dataset in which thereare three binary variables: effective disclosurecontrols, material weakness, and otherdeficiencies: it equals 1 if any of these three binaryvariables indicates a deficiency (NO, YES, or YES,respectively). The ICW_404 variable is obtainedfrom the Internal Controls dataset and equals 1when the binary variable effective internalcontrols is NO.14 As in Elder et al. (2009), we use thenumber of weaknesses to infer severity: three ormore indicates a company-wide problem.

Our restatement variables make use of therestated period end date and the disclosure datefrom the Restatements dataset. RST_Exante equals 1if the restated period end date falls in the specified

period (the disclosure date may be in the sameperiod or in a future period). RST_Expost equals 1 ifthe disclosure date falls in the specified period(the restated period may fall in the same period ormay be in a prior period). We use two approaches

to infer severity. First, following Scholz (2008),we determine whether the restatement concernedcore earnings (that is, a change in revenue orcore-expense). Second, we determine whether arestatement lowered or increased earnings.

For each of the four variables defined above, wetry three different time periods: current year [0],previous year [1] and current or previous year[1,0]. Because we seek to understand switchingbehavior conditioned on prior risks, we reportresults using the [-1] period only.15 Specifically, byusing risk variables measured as of [-1], we canmore confidently assume that observed behavior is

flowing from these risks. We do note, however, thatthe [0] specification yields somewhat weakerresults that nevertheless reinforce the importanceof restatement and ICW risks.

4. RESULTS

Tables 3 and 4 provide means for our key testvariables as well as tests of means between new ordeparting firms on the one hand and continuingfirms on the other. Table 3 compares new clientswith continuing clients while Table 4 comparesdeparting clients with continuing clients. Inboth tables the first column provides values forcontinuing clients and the second, third and fourthcolumns provide values for new or departingclients. We report means and medians and indicatesignificance when values for new or departingclients differ from the values for continuingclients. Overall, the results show the significance ofrestatement and ICW variables.

Table 3 indicates that the new clients of theSecond Tier are somewhat larger than theircontinuing clients (comparison of the first twocolumns). For example, mean asset values are308.74 for continuing clients and 414.20 for newclients. New clients also appear to be experiencinggreater sales growth. Turing to ICW, we note thatmost specifications show a higher level of risk innew clients. For example, ICW_404[-1] has a valueof 0.07 for continuing clients compared to 0.17for new clients. Similarly, most specifications ofrestatement risk show a high level of risk in newclients. For example, RST_Exante[-1] has a valueof 0.11 for continuing clients compared to 0.18




12/27

Table 3: New versus continuing clients

Variables Continuing clients All new clients New from Big 4 New from small firms

Mean Mean Mean Mean(median) (median) (median) (median)

Assets 308.74 414.20** 379.96 241.12(86.08) (136.49)*** (143.96)*** (60.99)

Leverage 0.67 0.55 0.52 0.60(0.44) (0.49)* (0.46) (0.48)

ROA -0.17 -0.10 -0.07 -0.22(0.01) (0.00) (0.01) (-0.03)***

Loss 0.46 0.48 0.44 0.62***(0.00) (0.00) (0.00) (1.00)***

ARInv 0.30 0.29 0.28 0.29(0.27) (0.24) (0.23) (0.24)

DSales 32.60 40.31 12.45 118.99(8.31) (11.42)** (7.16) (15.70)**

ICW_302[0] 0.29 0.33 0.39*** 0.34(0.00) (0.00) (0.00)*** (0.00)

ICW_302[-1] 0.30 0.26 0.32 0.23

(0.00) (0.00) (0.00) (0.00)ICW_404[0] 0.06 0.10*** 0.13*** 0.08(0.00) (0.00)*** (0.00)*** (0.00)

ICW_404[-1] 0.07 0.17*** 0.20*** 0.15***(0.00) (0.00)*** (0.00)*** (0.00)***

ICW_404COMP[-1] 0.02 0.05*** 0.08*** 0.03(0.00) (0.00)*** (0.00)*** (0.00)

RST_Exante[0] 0.06 0.06 0.06 0.07(0.00) (0.00) (0.00) (0.00)

RST_Exante[-1] 0.11 0.18*** 0.17*** 0.19**(0.00) (0.00)*** (0.00)*** (0.00)**

RST_Exante_CoreErn[-1] 0.05 0.09*** 0.10*** 0.12***(0.00) (0.00)*** (0.00)*** (0.00)***

RST_Exante_Neg[-1] 0.06 0.13*** 0.13*** 0.15***(0.00) (0.00)*** (0.00)*** (0.00)***

RST_Expost[0] 0.07 0.09* 0.10* 0.10(0.00) (0.00)* (0.00)* (0.00)

RST_Expost[-1] 0.13 0.17** 0.16*** 0.15(0.00) (0.00)** (0.00)*** (0.00)

RST_Expost_CoreErn[-1] 0.05 0.09*** 0.10*** 0.09*(0.00) (0.00)*** (0.00)*** (0.00)*

RST_Expost_Neg[-1] 0.07 0.13*** 0.14*** 0.13**(0.00) (0.00)*** (0.00)*** (0.00)**

N 2036 426 256 98

*, **, and *** denote two-tailed significance at the 10, 5, and 1 percent levels, respectively.Note: We use the two-sample t-test to test the differences in mean and the Wilcoxon rank sum test to test the differencesin median. Variables are defined as follows:

Assets: total assets in millions (Data6)Leverage: ratio of total liabilities to total assets (Data181/Data6)ROA: return on assets (Data18/Data6)Loss: binary variable, equals 1 if ROA < 0, and 0 otherwise

ARInv: ratio of accounts receivable and inventory to total assets (Data2 + Data3)/(Data6)DSales: percentage change in sales from prior yearICW_302[period]: if a weakness pursuant to SOX Section 302 is disclosed for the periodICW_404[period]: if a weakness pursuant to SOX Section 404 is disclosed for the periodICW_404COMP[period] : if three or more weaknesses pursuant to SOX Section 404 is disclosed for the periodRST_Exante[period]: if earnings for the period is eventually restatedRST_Exante_CoreErn[period] : if core earnings for the period is eventually restatedRST_Exante_Neg[period]: if earnings for the period is eventually restated and negativeRST_Expost[period]: if there is a restatement during the periodRST_Expost_CoreErn[period]: if the restatement involves core earningsRST_Expost_Neg[period]: if the restatement decreases earnings.




13/27

for new clients. Severity variables such asICW_404COMP[-1] and RST_Exante_Neg[-1]indicate similar results, that is, new firms are riskierthan continuing firms. We also note that the resultsfor all new clients are substantially similar to thosefor new firms from the Big 4 only as well as to thosefor new firms from the small firms only. Somenuances indicated by the subsample analysis: new

clients from small auditing firms are not larger thancontinuing clients and have lower ROA and higherLoss; new clients from the Big 4 have morepronounced ICW.

Table 4 provides a comparison of departingclients with continuing clients. Among thetraditional risk variables, we note that with theexception of ARInv, the rest are significant.

Table 4: Departing versus continuing clients

Variables Continuing clients

All departingclients

Departto Big 4

Depart tosmall firms

Mean Mean Mean Mean

(median) (median) (median) (median)Assets 263.57 149.10*** 527.47** 34.52***

(78.48) (35.45)*** (102.79) (15.38)***Leverage 0.61 1.87*** 0.51 3.69***

(0.42) (0.55)*** (0.39) (0.64)***ROA -0.15 -0.63*** -0.16 -1.43***

(0.01) (-0.04)*** (-0.03) (-0.16)***Loss 0.45 0.59*** 0.60* 0.68***

(0.00) (1.00)*** (1.00)* (1.00)***ARInv 0.30 0.31 0.24* 0.31

(0.27) (0.29) (0.23) (0.29)DSales 35.74 197.95** 17.48 504.00***

(9.90) (5.10)*** (14.75) (2.91)***ICW_302[0] 0.30 0.30 0.35 0.36*

(0.00) (0.00) (0.00) (0.00)*ICW_302[-1] 0.22 0.27** 0.33 0.32***

(0.00) (0.00)** (0.00) (0.00)***ICW_404[0] 0.08 0.05* 0.13 0.04

(0.00) (0.00)* (0.00) (0.00)ICW_404[-1] 0.07 0.05 0.15* 0.03**

(0.00) (0.00) (0.00)* (0.00)**ICW_404COMP[-1] 0.03 0.01 0.03 0.01

(0.00) (0.00) (0.00) (0.00)RST_Exante[0] 0.09 0.05** 0.05 0.05*

(0.00) (0.00)** (0.00) (0.00)*RST_Exante[-1] 0.12 0.16** 0.15 0.19***

(0.00) (0.00)** (0.00) (0.00)***RST_Exante_CoreErn[-1] 0.06 0.06 0.03 0.08

(0.00) (0.00) (0.00) (0.00)

RST_Exante_Neg[-

1] 0.07 0.09 0.08 0.08(0.00) (0.00) (0.00) (0.00)RST_Expost[0] 0.11 0.07** 0.08 0.08

(0.00) (0.00)** (0.00) (0.00)RST_Expost[-1] 0.12 0.16* 0.15 0.17*

(0.00) (0.00)* (0.00) (0.00)*RST_Expost_CoreErn[-1] 0.06 0.07 0.03 0.09*

(0.00) (0.00) (0.00) (0.00)*RST_Expost_Neg[-1] 0.07 0.09 0.05 0.11

(0.00) (0.00) (0.00) (0.00)N 2036 420 40 160

*, **, and *** denote two-tailed significance at the 10, 5, and 1 percent levels, respectively.Note: We use the two-sample t-test to test the differences in mean and the Wilcoxon rank sum test to test thedifferences in median. For variable definitions, see Table 3.




14/27

Departing clients are smaller, have greaterleverage, lower profits (lower ROA and higherLoss) and lower (median) change in sales. Turing

to ICW, we note that ICW_302[-1] is significantlyhigher for departing clients overall, whileICW_404[-1] is lower mainly for clients departingto smaller auditing firms. The lack of a strongfinding concerning ICW_404[-1] is probablyexplained by the threshold for Section 404reporting: many departing clients are probablynon-accelerated filers. Finally, we note thatrestatement risk is higher for departing clients. Forexample, RST_Exante[-1] is 0.12 for continuingfirms and 0.16 for departing firms; this result isstronger if one considers clients departing to smallfirms (value is 0.19). While there is some evidencethat departing clients have higher restatementand ICW risks in terms of prevalence of theserisks, there is almost no evidence that severity isgreater.

The univariate tests suggest strong support forhypothesis 1 (new versus continuing clients) as wellas moderate support for hypothesis 2 (departingversus continuing clients). But more conclusiveresults are only possible using multivariate analysis.Before presenting our multivariate results, weprovide evidence of correlations among restatementand ICW variables in Table 5. We note significantcorrelations between various specifications ofthe restatement and ICW variables. For example,ICW_302[-1] has correlations of 0.38, 0.19 and0.27 with ICW_404[-1], RST_Exante[-1] andRST_Expost[-1], respectively; all values arestatistically significant. This is consistent with theobservation that restatement and ICW variables arein some general sense related to the auditing riskenvironment of client firms. Because of these highcorrelations, although they do not meet the

threshold for inducing severe multi-collinearityproblems, we choose to run cross-sectional models(reported below) using restatement and ICW

variables one at a time.Our key results concerning hypotheses 1 and

2 are found in Tables 6 and 7. We use logisticregressions as in the prior literature on auditorswitching to determine the importance ofrestatement and ICW variables in explainingswitches to and from the Second Tier. Table 6provides a comparison of (all) new and continuingclients. Table 7 provides a comparison of (all)departing and continuing clients.

Table 6 Panel A focuses on ICW. Model 6A1 isthe base model in which only traditional risk (thatis, control) variables are used. In subsequentmodels, ICW variables are added to evaluate theirexplanatory power. The R2 in these models rangefrom 0.026 (model 6A1, the base model) to 0.045(model 6A3 containing ICW_404[-1]). Amongcontrol variables, only two, LnAssets and Loss,appear to be significantly related to auditorswitches. New clients, compared to continuingclients, are larger and are more likely to beloss-making. Turning to ICW, we note that thecoefficient of ICW_404[-1] is significantly positivein model 6A3 indicating that the prevalence of ICWis associated with switches to the Second Tier.Additionally, in model 6A4, we find that severitymatters: the coefficient of ICW_404COMP[-1] issignificantly positive.

Table 6 Panels B and C focus on restatements,the former using ex-ante measures and the latterex-post. The base model is 6A1 and it is not repeatedin these panels. The range of R2 values in Panels Band C are similar to that of Panel A. Overall, thereis strong evidence that the prevalence ofrestatement is associated with switches into the

Table 5: Pearson correlation for internal control and restatement variables

A B C D E F G H I

A. ICW_302[-1] 1 0.38 0.25 0.19 0.15 0.15 0.27 0.18 0.19B. ICW_404[-1] 1 0.55 0.20 0.16 0.16 0.22 0.16 0.17C. ICW_404COMP[-1] 1 0.11 0.09 0.08 0.12 0.10 0.09D. RST_Exante[-1] 1 0.64 0.75 0.62 0.47 0.53E. RST_Exante_CoreErn[-1] 1 0.75 0.48 0.74 0.56F. RST_Exante_Neg[-1] 1 0.53 0.55 0.72G. RST_Expost[-1] 1 0.65 0.75H. RST_Expost_CoreErn[-1] 1 0.76I. RST_Expost_Neg[-1] 1

Note: Significant values (i.e., p < 0.01) are in bold. For variable definitions, see Table 3.




15/27


16/27

Second Tier. In model 6B1, we find that thecoefficient of RST_Exante[-1] is significantlypositive. Similarly, in model 6C1, we find that thecoefficient of RST_Expost[-1] is significantlypositive. In addition to these results concerning theprevalence of restatements, we also note that clientsswitching to the Second Tier are more likely tohave severe restatements. In models 6B26B3and 6C26C3, we find that the coefficients of thecore earnings and negative restatement variables(in the ex-ante as well as the ex-post specifications)are significantly positive. Thus, clients switchingto the Second Tier are more likely to haverestatements that involve core accounts anddecrease earnings. Overall, the evidence in Table 6supports hypothesis 1.16,17

Table 7 compares departing firms withcontinuing firms. In general, the R2 values arehigher in this table than in Table 6. To calibrateour results, as in Table 6, we start with the basemodel 7A1. This base model indicates that clients

departing from the Second Tier tend to be smallerfirms, with greater losses and with greaterleverage. Thus, riskier firms, at least from atraditional auditing risk perspective, are departing.Panel A evaluates the ICW risk of departing firmsversus continuing firms. As expected, we find thatthe coefficient of ICW_302[-1] is significantlypositive (model 7A2). However, because departingfirms are smaller firms (recall that the coefficientof LnAssets is significantly negative) with a higherlikelihood of being non-accelerated filers, ICW_404variables are insignificant (models 7A3 and 7A4).

Table 7 Panels B and C show the associationbetween restatement risk and departures from theSecond Tier, the former using ex-ante measures andthe latter ex-post. The range of R2 values in Panel Bis similar to that of Panel A. Overall, there isstrong evidence that the prevalence of restatementis associated with switches out of the Second Tier.In model 7B1, we find that the coefficient ofRST_Exante[-1] is significantly positive. Similarly,

Table 6: Continued

Panel C: Restatements ex post

Variable Model 6C1 Model 6C2 Model 6C3

Intercept -2.84*** -2.85*** -2.85***(141.96) (142.12) (142.01)

LnAssets 0.22*** 0.22*** 0.22***(33.00) (33.70) (32.53)

Leverage -0.04 -0.04 -0.05(0.24) (0.19) (0.40)

ROA -0.02 -0.02 -0.02(0.03) (0.02) (0.03)

Loss 0.23* 0.24* 0.22*(3.41) (3.69) (3.08)

ARInv 0.24 0.25 0.23(0.87) (0.92) (0.80)

DSales 0.00 0.00 0.00(0.18) (0.16) (0.18)

RST_Expost[-1] 0.36**

(5.82)RST_Expost_CoreErn[-1] 0.57***

(7.80)RST_Expost_Neg[-1] 0.80***

(20.72)

Pseudo-R2 0.030 0.031 0.039N(total) 2365 2365 2365N(new) 392 392 392N(continuing) 1973 1973 1973

*, **, and *** denote two-tailed significance at the 10, 5, and 1 percent levels, respectively.Note: This table presents the logit regression results for new firms to the Second Tier versus continuing clients.The dependent variable equals 1 when the client is new and 0 when continuing. The chi-square statistics arereported in parentheses. For variable definitions, see Table 3.




17/27

Table 7: Logistic regression analyses of departing clients compared to continuing clients

Panel A: Internal control weakness

Variable Model 7A1 Model 7A2 Model 7A3 Model 7A4

Intercept -1.02*** -1.00*** -1.02*** -1.04***(21.04) (20.37) (21.01) (21.66)

LnAssets -0.23*** -0.24*** -0.22*** -0.22***(33.71) (36.85) (32.09) (31.89)

Leverage 0.12** 0.12** 0.12** 0.12**(5.65) (5.38) (5.65) (5.72)

ROA 0.11 0.11 0.11 0.11(2.48) (2.58) (2.47) (2.47)

Loss 0.38*** 0.34*** 0.38*** 0.39***(9.62) (7.57) (9.63) (10.09)

ARInv 0.24 0.21 0.24 0.25(0.88) (0.62) (0.88) (0.90)

DSales 0.00 0.00 0.00 0.00(1.59) (1.66) (1.59) (1.59)

ICW_302[-1] 0.31**

(5.86)ICW_404[-1] -0.04

(0.03)ICW_404COMP[-1] -0.43

(0.94)

Pseudo-R2 0.060 0.064 0.060 0.061N(total) 2366 2366 2366 2366N(departing) 401 401 401 401N(continuing) 1965 1965 1965 1965

Panel B: Restatements ex ante

Variable Model 7B1 Model 7B2 Model 7B3

Intercept -1.03*** -1.02*** -1.02***

(21.36) (21.12) (20.99)LnAssets -0.23*** -0.23*** -0.23***(34.86) (33.97) (34.57)

Leverage 0.12** 0.12** 0.12**(5.18) (5.60) (5.42)

ROA 0.10 0.11 0.11(2.12) (2.50) (2.49)

Loss 0.36*** 0.38*** 0.37***(8.41) (9.50) (9.11)

ARInv 0.23 0.24 0.24(0.77) (0.86) (0.83)

DSales 0.00 0.00 0.00(1.33) (1.60) (1.61)

RST_Exante[-1] 0.36**(5.38)

RST_Exante_CoreErn[-

1] 0.18(0.60)RST_Exante_Neg[-1] 0.31

(2.32)

Pseudo-R2 0.064 0.060 0.062N(total) 2366 2366 2366N(departing) 401 401 401N(continuing) 1965 1965 1965




18/27

in model 7C1, we find that the coefficient ofRST_Expost[-1] is significantly positive. In additionto these results concerning the prevalence ofrestatements, we provide results concerning theirseverity. We note weaker results in this regard. Inmodel 7C3, we note a significant coefficient forRST_Expost_Neg[-1], but this is the only significantvariable among severity variables. Principallybecause of the significance of prevalence measures(of ICW and restatements), we conclude that ourresults are supportive of hypothesis 2.18

Tables 8 and 9 provide further evidence onhypotheses 1 and 2. Here, we focus on the keysubsets of new and departing clients: new from theBig 4 (Table 8) and departing to small auditingfirms (Table 9).

Results in Table 8 (new from the Big 4 versuscontinuing) are similar to those reported earlier inTable 6 (all new clients versus continuing). The R2

values and the behavior of controls are roughlycomparable. With regard to controls, we again find

that new firms tend to be larger than continuingfirms. However, in Table 8, we do not find thecoefficient of Loss to be significant; instead, wefind that the coefficient of DSales is positive andmarginally significant. Results concerning ICWand restatements are largely the same. New clientsfrom the Big 4 have greater companywide ICW.Also, new clients have a greater prevalence ofrestatements (coefficients of both RST_Exante[-1]and RST_Expost[-1] are significantly positive) andthere is also strong indication that new clients havegreater levels of severe restatements.

Table 9 reports the comparison between clientsdeparting to small auditing firms and continuingclients. These results are similar to those reportedearlier in Table 7. As in Table 7, we find thatthe coefficienst of ICW_302[-1], RST_Exante[-1]and RST_Expost[-1] are significantly positive,indicating the prevalence of these risk factors. Adifference is that restatement variables (especiallyin Panel C) show greater levels of significance.

Table 7: Continued



Intercept -1.02*** -1.02*** -1.02***(20.94) (21.21) (21.11)

LnAssets -0.23*** -0.23*** -0.23***(35.46) (34.30) (34.74)

Leverage 0.12** 0.12** 0.12**(5.15) (5.61) (5.07)

ROA 0.11 0.11 0.11(2.55) (2.53) (2.46)

Loss 0.36*** 0.37*** 0.37***(8.71) (9.24) (9.07)

ARInv 0.23 0.24 0.24(0.79) (0.85) (0.89)

DSales 0.00 0.00 0.00(1.62) (1.60) (1.61)

RST_Expost[-1] 0.32**

(4.25)RST_Expost_CoreErn[-1] 0.33

(2.33)RST_Expost_Neg[-1] 0.34*

(2.94)

Pseudo-R2 0.063 0.062 0.062N(total) 2366 2366 2366N(departing) 401 401 401N(continuing) 1965 1965 1965

*, **, and *** denote two-tailed significance at the 10, 5, and 1 percent levels, respectively.Note: This table presents logit regression results for clients departing from the Second Tier versus continuingclients. The dependent variable equals 1 when the client is departing and 0 when continuing. The chi-squarestatistics are reported in parentheses. For variable definitions, see Table 3.




19/27

Table 8: Logistic regressions analyses of new Big 4 clients compared to continuing clients



Intercept -2.96*** -2.97*** -2.92***(99.04) (99.12) (92.24) -2.93***

LnAssets 0.22*** 0.22*** 0.18*** (94.90)(21.39) (20.66) (14.18) 0.21***

Leverage -0.18 -0.18 -0.10 (18.16)(1.08) (1.06) (0.41) -0.17

ROA 0.13 0.14 0.19 (0.95)(0.22) (0.25) (0.56) 0.12

Loss 0.12 0.11 0.04 (0.19)(0.58) (0.42) (0.05) 0.07

ARInv -0.19 -0.20 -0.20 (0.17)(0.33) (0.37) (0.34) -0.21

DSales 0.00* 0.00* 0.00* (0.40)(3.63) (3.62) (3.56) 0.00*

ICW_302[-1] 0.11 (3.27)

(0.52)ICW_404[-1] 0.99***

(28.81)ICW_404COMP[-1] 1.14***

(16.57)

Pseudo-R2 0.031 0.031 0.053 0.043N(total) 2221 2221 2221 2221N(new Big 4) 248 248 248 248N(continuing) 1973 1973 1973 1973



Intercept -3.02*** -3.01*** -2.99***

(100.90) (100.24) (99.56)LnAssets 0.22*** 0.22*** 0.22***(21.07) (20.66) (20.14)

Leverage -0.21 -0.21 -0.21(1.38) (1.37) (1.37)

ROA 0.14 0.15 0.15(0.26) (0.29) (0.27)

Loss 0.11 0.14 0.12(0.43) (0.69) (0.56)

ARInv -0.20 -0.18 -0.20(0.37) (0.29) (0.34)

DSales 0.00* 0.00* 0.00*(3.79) (3.81) (3.72)

RST_Exante[-1] 0.59***(10.50)


1] 0.86***(13.18)RST_Exante_Neg[-1] 0.78***

(13.26)

Pseudo-R2 0.039 0.041 0.041N(total) 2221 2221 2221N(new Big 4) 248 248 248N(continuing) 1973 1973 1973




20/27

Our final table, Table 10, presents evidence onhypothesis 3. Although hypothesis 3 pertains to thedeparting versus continuing as well as the newversus continuing comparisons, because of spaceconstraints we chose to focus on the new versuscontinuing comparisons. Also, for the same reason,we report prevalence measures only. Panel Acompares the dismissals sample of new firmswith continuing firms and Panel B compares theresignations sample of new firms with continuingfirms. Significantly, we note that Panel A contains85 resignations and Panel B contains 248 dismissals:the new clients of the Second Tier are more likelyto have dismissed their auditors. According tohypothesis 3, restatement and ICW risks are morepronounced in the resignations rather than thedismissals sample. We do not find evidenceconsistent with this hypothesis. Both the dismissalsand resignations samples show similar levels ofsignificance for the restatement and ICW variables.For example, in both Panels A and B, we find that

ICW_404[-1], RST_Exante[-1] and RST_Expost[-1]are significant. We do, however, note that themagnitude of the coefficients are greater in Panel Aconsistent with the hypothesis. In Panel C, we run adirect comparison of resignations with dismissals.Resignation firms are significantly smaller andthere is no indication that their restatement andICW risks are potentially higher. This is consistentwith results reported in Panels A and B that bothresignations and dismissals show sensitivity torestatement and ICW.

5. DISCUSSION AND CONCLUSION

The Post-SOX era coincides with the rise of theSecond Tier auditing firms. Since many of theirnew clients are large firms from the Big 4, and sincelarge firms are assumed to carry higher level oflitigation risks with them, regulators have beenconcerned about client portfolio risks of the SecondTier. An earlier and influential study, Hogan and

Table 8: Continued



Intercept -3.00*** -3.00*** -3.00***(100.24) (100.19) (100.11)

LnAssets 0.22*** 0.22*** 0.22***(20.50) (20.91) (20.33)

Leverage -0.20 -0.20 -0.22(1.33) (1.29) (1.45)

ROA 0.14 0.14 0.14(0.24) (0.25) (0.23)

Loss 0.10 0.11 0.10(0.41) (0.48) (0.39)

ARInv -0.19 -0.19 -0.19(0.31) (0.31) (0.33)

DSales 0.00* 0.00* 0.00*(3.54) (3.68) (3.61)

RST_Expost[-1] 0.48***


(9.90)RST_Expost_Neg[-1] 0.85***

(16.97)

Pseudo-R2 0.036 0.038 0.044N(total) 2221 2221 2221N(new Big 4) 248 248 248N(continuing) 1973 1973 1973

*, **, and *** denote two-tailed significance at the 10, 5, and 1 percent levels, respectively.Note: This table presents the logit regression results for new firms from the Big 4 versus continuing clients. Thedependent variable equals 1 when the client is a new from the Big 4 and 0 when continuing. The chi-squarestatistics are reported in parentheses. For variable definitions, see Table 3.




21/27

Table 9: Logistic regression analyses of departing clients to small firms compared to continuing clients



Intercept -0.48 -0.41 -0.50 -0.48(2.08) (1.54) (2.23) (2.03)

LnAssets -0.65*** -0.70*** -0.65*** -0.66***(92.90) (103.36) (86.97) (90.95)

Leverage 0.03 0.03 0.03 0.03(0.54) (1.03) (0.52) (0.54)

ROA 0.15* 0.16* 0.14* 0.15*(3.16) (3.53) (3.13) (3.16)

Loss 0.33 0.23 0.34 0.32(2.56) (1.23) (2.70) (2.51)

ARInv 0.19 0.02 0.19 0.19(0.23) (0.00) (0.22) (0.23)

DSales 0.00 0.00 0.00 0.00(1.97) (2.47) (1.97) (1.97)

ICW_302[-1] 0.81***

(16.35)ICW_404[-1] -0.28

(0.28)ICW_404COMP[-1] 0.09

(0.01)

Pseudo-R2 0.181 0.198 0.182 0.181N(total) 2117 2117 2117 2117N(depart to SF) 152 152 152 152N(continuing) 1965 1965 1965 1965



Intercept -0.48 -0.47 -0.47

(2.02) (1.99) (2.02)LnAssets -0.67*** -0.66*** -0.66***(95.35) (93.43) (93.73)

Leverage 0.03 0.03 0.03(0.64) (0.60) (0.61)

ROA 0.14 0.15* 0.15*(2.53) (3.26) (3.22)

Loss 0.27 0.31 0.31(1.72) (2.34) (2.30)

ARInv 0.12 0.17 0.17(0.09) (0.19) (0.18)

DSales 0.00 0.00 0.00(1.58) (2.00) (1.99)

RST_Exante[-1] 0.74***(9.91)


1] 0.57*(2.76)RST_Exante_Neg[-1] 0.37

(1.28)

Pseudo-R2 0.191 0.184 0.183N(total) 2117 2117 2117N(depart to SF) 152 152 152N(continuing) 1965 1965 1965




22/27

Martin (2009), evaluates the new and departingclients of the Second Tier against continuing clientsand finds that: (a) new clients, especially those fromthe Big 4 bring in additional risks mostly because oftheir larger size, and (b) these risks are somewhatoffset by the movement of other risky clients fromthe Second Tier to other auditors. But this earlierstudy, because of its sampling period, was unable toadequately assess the importance of restatementand ICW risks. Our study fills this void. Weexplicitly formulate hypotheses relating theswitching behavior of the Second Tier to these newrisk factors and test the hypotheses using a samplethat fully exploits the restatement and ICWdisclosures in the post-SOX era.

We show that the new clients of the SecondTier have higher prevalence and severity ofrestatements and ICW. This has two implications.First, this implies that the new clients acquired bythe Second Tier are in some sense riskier thanimplied by an analysis of the traditional variables

used in the literature. Second, we contribute tothe large literature on auditor switching bydemonstrating the importance of restatement andICW risks. We argue that restatement and ICWmeasures are indicators of a perturbed auditingenvironment, which poses additional risks to theauditor. Thus, traditional variables like firm size,ROA, loss, accounts receivables and inventory maynot fully explain audit risk.

We also show that departing clients of the SecondTier are somewhat riskier than continuing firms.The comparison of all departing and continuingfirms shows some evidence of differences in risk;however, more risks are evident in the subset offirms departing to small auditors. Combined withthe observation that the number of clients of theSecond Tier has held steady in the post-SOX era,this appears to indicate a somewhat deliberatestrategy on the part of the Second Tier to build theirclientele and manage risks prudently. Our samplespans a number of years in the post-SOX era and

Table 9: Continued



Intercept -0.47 -0.48 -0.49(2.04) (2.04) (2.15)

LnAssets -0.67*** -0.67*** -0.67***(97.48) (93.88) (96.13)

Leverage 0.03 0.03 0.03(1.02) (0.59) (1.00)

ROA 0.15* 0.15* 0.15*(3.38) (3.29) (3.36)

Loss 0.29 0.31 0.31(2.05) (2.23) (2.24)

ARInv 0.15 0.18 0.19(0.14) (0.20) (0.22)

DSales 0.00 0.00 0.00(2.01) (1.99) (1.99)

RST_Expost[-1] 0.66***


(6.84)RST_Expost_Neg[-1] 0.67**

(5.11)

Pseudo-R2 0.189 0.188 0.186N(total) 2117 2117 2117N(depart to SF) 152 152 152N(continuing) 1965 1965 1965

*, **, and *** denote two-tailed significance at the 10, 5, and 1 percent levels, respectively.Note: This table presents logit regression results for clients departing to small firms versus continuing clients. Thedependent variable equals 1 when the client is departing to small firms and 0 when continuing. The chi-squarestatistics are reported in parentheses. For variable definitions, see Table 3.




23/27

Table 10: Resignations versus dismissals

Panel A: New clients whose auditors resigned versus continuing clients

Variable Model 10A1 Model 10A2 Model 10A3 Model 10A4 Model 10A5

Intercept -2.99*** -2.99*** -2.80*** -3.07*** -3.03***(42.06) (42.03) (35.71) (43.26) (42.47)

LnAssets -0.02 -0.02 -0.09 -0.02 -0.02(0.04) (0.08) (1.27) (0.06) (0.09)

Leverage -0.18 -0.18 -0.13 -0.22 -0.20(0.72) (0.72) (0.51) (0.93) (0.84)

ROA -0.03 -0.02 0.00 -0.04 -0.03(0.01) (0.01) (0.00) (0.04) (0.01)

Loss 0.26 0.23 0.13 0.22 0.23(1.09) (0.85) (0.27) (0.81) (0.83)

ARInv -0.39 -0.43 -0.43 -0.42 -0.40(0.51) (0.59) (0.60) (0.58) (0.54)

DSales 0.00 0.00 0.00 0.00 0.00(0.01) (0.01) (0.00) (0.01) (0.01)

ICW_302[-1] 0.17(0.51)

ICW_404[-1] 1.39***(23.89)

RST_Exante[-1] 0.83***(9.31)

RST_Expost[-1] 0.62**(5.12)

Pseudo-R2 0.006 0.006 0.039 0.019 0.013N(total) 2058 2058 2058 2058 2058N(resign) 85 85 85 85 85N(continuing) 1973 1973 1973 1973 1973

Panel B: New clients who dismissed their auditors versus continuing clients

Variable Model 10B1 Model 10B2 Model 10B3 Model 10B4 Model 10B5

Intercept -3.30*** -3.30*** -3.25*** -3.36*** -3.33***

(127.02) (126.49) (119.66) (129.50) (128.02)LnAssets 0.22*** 0.23*** 0.19*** 0.23*** 0.22****

(24.09) (24.11) (17.15) (24.30) (23.61)Leverage -0.03 -0.03 0.00 -0.04 -0.04

(0.09) (0.09) (0.00) (0.14) (0.13)ROA -0.04 -0.04 -0.01 -0.05 -0.04

(0.06) (0.06) (0.01) (0.08) (0.06)Loss 0.20 0.21 0.13 0.18 0.19

(1.84) (1.86) (0.76) (1.47) (1.53)ARInv 0.29 0.29 0.28 0.29 0.30

(0.82) (0.84) (0.75) (0.82) (0.86)DSales 0.00 0.00 0.00 0.00 0.00

(0.12) (0.12) (0.19) (0.15) (0.14)ICW_302[-1] -0.03

(0.03)ICW_404[-1] 0.93***

(24.46)RST_Exante[-1] 0.49***

(6.92)RST_Expost[-1] 0.33*

(3.22)

Pseudo-R2 0.023 0.023 0.042 0.028 0.026N(total) 2221 2221 2221 2221 2221N(dismiss) 248 248 248 248 248N(continuing) 1973 1973 1973 1973 1973




24/27

therefore our results do not just indicate animmediate reaction to Andersen and SOX. Rather,our results seem indicative of a longer-term trendin the auditing market where the Second Tier isgaining market share in a steady and deliberatefashion. This conclusion is also supported by ourresults concerning resignations and dismissals. Notonly do we find that most of the switches to theSecond Tier are dismissals rather than resignations(that is, in most cases, the auditor is fired), we alsofind that the sensitivities of switches to restatementand ICW risks are no different between theresignation and dismissal samples.

NOTES

1. The GAO report, dated January 2008 (GAO,2008), identifies several constraints faced by

Second-Tier firms including: (a) the abilityto hire and retain employees, (b) lack ofreputation, (c) lack of capabilities in multiplecountries to service multinationals, and (d) theability to deal with the litigation risk arisingfrom large clients.

2. A report issued by Audit Analytics in February2009 entitled 2008 Financial Restatements:An Eight Year Comparison indicates a sharprise in the number of restatements during20052006. While the level is also quite high in2007, it represents a fall from 2006. An evenlower level of restatements in 2008 may suggesta trend. Overall, our sample period has a highnumber of restatements.

3. There is also a stream of papers that solely focuson restatements and provide insights for theauditing context. For example, Abbott, Parker

Table 10: Continued

Panel C: Resignations versus dismissals of new clients

Variable Model 10C1 Model 10C2 Model 10C3 Model 10C4 Model 10C5

Intercept 0.46 0.47 0.53 0.39 0.43(0.67) (0.71) (0.88) (0.48) (0.57)

LnAssets -0.29*** -0.30*** -0.32*** -0.29*** -0.29****(7.97) (8.49) (9.05) (7.84) (8.12)

Leverage -0.17 -0.18 -0.16 -0.24 -0.22(0.30) (0.35) (0.27) (0.57) (0.48)

ROA 0.10 0.12 0.15 0.06 0.07(0.03) (0.05) (0.08) (0.01) (0.02)

Loss 0.11 0.07 0.06 0.10 0.09(0.11) (0.05) (0.04) (0.09) (0.08)

ARInv -0.62 -0.57 -0.59 -0.59 -0.61(0.93) (0.74) (0.82) (0.81) (0.88)

DSales 0.00 0.00 0.00 0.00 0.00(0.04) (0.03) (0.01) (0.02) (0.02)

ICW_302[-1] 0.25(0.71)

ICW_404[-1] 0.43(1.62)

RST_Exante[-1] 0.43(1.79)

RST_Expost[-1] 0.39(1.44)

Pseudo-R2 0.056 0.059 0.063 0.064 0.062N(total) 333 333 333 333 333N(resign) 85 85 85 85 85N(dismiss) 248 248 248 248 248

*, **, and *** denote two-tailed significance at the 10, 5, and 1 percent levels, respectively.Note: This table presents results from logit regression. The dependent variable equals one if the client is a new clientof the Second Tier from the Big 4 or small firms and zero otherwise. We exclude new clients from unknown sincethey are mainly IPO firms. Panel A only includes new clients whose previous auditor resigned and Panel B only

includes new clients who dismissed their previous auditor. In Panel C, the dependent variable equals 1 if the auditorresigned and 0 if the auditor was dismissed from the Big 4 or small firms. The chi-square statistics are reported inparentheses. For variable definitions, see Table 3.




25/27

and Peters (2004) relate restatements toweaknesses in the auditing committee. Thislink suggests that restatements may be used asa proxy for poor controls and therefore is anindicator of audit risk.

4. Furthermore, client firms may themselvesplay a role in auditor switches: there is apossibility that clients grappling with earningsrestatements are active in searching for a newauditor. To test for this possibility, we alsopresent tests based separately on dismissals(client initiated) and resignations (auditorinitiated).

5. Hasan et al. (2005) survey Big 5 and non-Big 5audit firms in 11 non-US countries to examinethe differences between the types of assuranceservices and levels at which theses assuranceservices are provided by the two categories of

firms.6. The initial Oxera report and the subsequent

European Union study may be foundat: http://ec.europa.eu/internal_market/auditing/docs/market/oxera_report_en.pdfand http://ec.europa.eu/internal_market/auditing/docs/market/consultation2008/summary_report_en.pdf, respectively.

7. For this and other developments in theEuropean Union, see http://ec.europa.eu/internal_market/auditing/communications/index_en.htm.

8. An alternate way for auditing firms to deal withclient risk is to increase fees or to chargerisk-adjusted fees. Few studies compare thealternatives of fee increases and resignations.An exception is Elder et al. (2009), whichstudies auditor decisions in the context ofinternal control weaknesses. In fact, threepossible responses are considered: a modifiedopinion, a fee increase, or resignation. Thischoice is empirically modeled by using theordered logit method (see Table 10).Nevertheless, the switching literature focuseson auditor change as opposed to fee increases(or modified opinions) while acknowledgingthis limitation (e.g., Johnstone & Bedard, 2004,see Note 6).

9. In an earlier study, Wallace (2005) shows thatthe percentage of auditor changes is greaterfor restating firms when compared tonon-restating firms. But this relation appears tobe driven by client firms changing their auditormany times. Once the serial changers wereremoved, there appears to be no link between

restatements and auditor changes. Despite this,there appears to be a belief among practitionersthat restating firms will seek new auditors: forexample, Linn and Diehl (2005) argue that thecost of fixing a restatement may drive some

clients to the Second Tier. Tyranski (2008) alsonotes that restatement is one of the manyreasons for auditor switches.

10. The Dodd-Frank Act of 2010 has indefinitelypostponed and therefore essentially eliminatedthe need for small firms (market capitalizationbelow $75 million) to comply with SOX 404b.But these small firms were required to complywith SOX 404a starting in 2007.

11. A number of articles in the popular press hasremarked on this phenomenon. A majorityof these articles appeared around 20052006and focused on the SOX-related movement

of clients from the Big 4 to the Second Tier.One example is Byrnes (2005) which describeshow BDO won the contract to audit a publicfirm with more than $1 billion in sales. Anotherexample is Reilly (2006). The market shareof the Big 4 continues to be big, however:Tyranski (2008) notes that the share of publiclylisted clients had only fallen from a high of 98%to 94% despite many clients shifting to non-Bigfirms.

12. Regulators remain concerned aboutconcentration in the auditing market. This isa recurring theme in various GovernmentAccountability Office (GAO) reports on theauditing market, the most recent of which,dated January 2008 (GAO, 2008), identifiesconcentration as a problem that has not yetadversely affected prices or quality but has thecapacity to do so in the future. Considerableconcern is shown in the report concerningthe constraints faced by Second Tier firmsin auditing large listed firms. Some of theconstraints identified in this report include: (a)the ability to hire and retain employees, (b) lackof reputation, (c) lack of capabilities in multiplecountries to service multinationals, and (d) theability to deal with the litigation risk arisingfrom large clients.

13. Hypothesis 3 can also be tested by directlycomparing resignations firms with dismissalfirms. We follow both approaches.

14. As explained earlier, 404a compliance beginsin 2004 for accelerated filers and 2007 fornon-accelerated filers, while 404b complianceis only required of accelerated filers starting




26/27

in 2004. Since our sample contains non-accelerated filers and covers years prior to 2007,there is a downward bias in the ICW_404variable. But because we do have observationsfrom 20072008 as well as accelerated filers in

the sample, we are still able to constructmeaningful tests using the ICW_404 variable.

15. Compliance with SOX Section 404 started in2004 for accelerated firms. The measurement ofICW_404 as of [-1] means that we impart adownward bias in the measurement of thisvariable for the departing and continuingsamples (first year of data is 2004) but not forthe new sample (first year of data is 2005). Sincethis bias is against finding significance for theICW_404 variable, we make no adjustment.However, as a robustness check, we re-runtests by excluding year 2004 observations and

find similar results.16. We also conducted year-by-year comparisons

of new versus continuing and departing versuscontinuing. These results are not reportedin the paper

the post sox

Documents