搜档网
当前位置:搜档网 › BTBU_2.1 Auditor Fees and Auditor Independence_ Evidence from Going Concern Reporting Decisions

BTBU_2.1 Auditor Fees and Auditor Independence_ Evidence from Going Concern Reporting Decisions

BTBU_2.1 Auditor Fees and Auditor Independence_ Evidence from Going Concern Reporting Decisions
BTBU_2.1 Auditor Fees and Auditor Independence_ Evidence from Going Concern Reporting Decisions

Auditor Fees and Auditor Independence:Evidence from Going

Concern Reporting Decisions*

ALLEN D.BLAY,Florida State University

MARSHALL A.GEIGER1,University of Richmond

Independence is the cornerstone of the accounting profession and one of its most

precious assets.[Robert Mednick Chair,American Institute of Certi?ed Public Accoun-

tants(AICPA)Board of Directors,1997]

1.Introduction

Without independence there is no need for external auditors attesting to the purported accuracy and completeness of company?nancial information(Wallace2004).As a result, the possible adverse effect of auditors providing services to clients who pay them directly and the question of impaired auditor independence,in fact or appearance has histori-cally been a concern of the public accounting profession(Mautz and Sharaf1961; AICPA1978;Simunic1984).This possible adverse effect has also been a longstanding concern of regulators,legislators,and market participants(U.S.Senate1976;Securities and Exchange Commission[SEC]2000a;Sarbanes-Oxley Act[SOX]2002;Zeff2003a; Public Company Accounting Oversight Board[PCAOB]2009).

The association between fees received by audit?rms directly from clients and the pos-sible impairment of auditor independence,particularly with respect to going concern reporting decisions,continues to be of considerable interest to regulators and others.In fact,reporting on?nancially distressed companies has received heightened attention from the Financial Accounting Standards Board(FASB)in their proposal to have management explicitly report on their company’s going concern status(FASB2008,2011),2as well as from the PCAOB in their recent ongoing discussions regarding auditor responsibilities with respect to going concern reporting(PCAOB2009,2011).

In this study,we assess the potential impairment of auditor independence in the context of going concern reporting by examining the association between both current and future audit service and nonaudit service(NAS)fees received by U.S.auditors and the type of opinion rendered to a?nancially distressed client.Clients generally do not welcome a going concern modi?ed(GCM)opinion from their auditor(Kida1980;Geiger,Raghunandan, and Rama1998).Accordingly,if the receipt of current fees or the expectation of future fees *Accepted by Steve Salterio.We gratefully acknowledge the helpful comments from Steve Salterio,the reviewer,Frank He?in,Kenny Reynolds,and workshop participants at Florida State University,the University of Richmond and the2010American Accounting Association Annual Meeting.

1.Marshall A.Geiger is currently an Academic Fellow at the Securities and Exchange Commission.The

Securities and Exchange Commission,as a matter of policy,disclaims responsibility for any private publi-cation or statement by any of its employees.The views expressed herein are those of the author and do not necessarily re?ect the views of the Commission or of the author’s colleagues upon the staff of the Commission.

2.The proposed standard would require preparers to present additional disclosures either when the state-

ments are not prepared on a going concern basis or when there is substantial doubt as to an entity’s abil-ity to continue as a going concern through the next?scal year.

Contemporary Accounting Research Vol.30No.2(Summer2013)pp.579–606óCAAA

doi:10.1111/j.1911-3846.2012.01166.x

580Contemporary Accounting Research

from a client is likely to impair auditor independence,we would expect to?nd auditors giving fewer GCM opinions to a client that pays relatively high total fees(i.e.,a negative association)in order to appease the client(DeFond,Raghunandan,and Subramanyam 2002).

While the presumably more lucrative NAS fees have garnered the recent attention of legislators and researchers(DeFond et al.2002;Feldmann and Read2010;Frankel, Johnson,and Nelson2002;U.S.Senate2002),several commentators have argued that it is not only the receipt of NAS fees but the direct receipt of all types of fee revenue, including basic audit fees,that create the economic bond between the auditor and the client that might impair auditor independence(Mautz and Sharaf1961;SEC2000a; Kinney and Libby2002).Still others have argued that it is not current fee revenues, but the potential for garnering future fees from the client,that pose the greatest threat to auditor independence(DeAngelo1981;Beck,Frecka,and Solomon1988;Public Oversight Board[POB]1995;Levitt2000;Coffee2002;U.S.House of Representatives 2002).

DeAngelo(1981)argues that the ability of the auditor to maintain independence is directly related to auditors’ability to retain the client in order to recover their initial start-up costs on the audit,particularly in cases where the auditor‘‘low balls’’the initial audit fee in order to acquire the client.Therefore,an auditor’s independence is in?uenced by its desire to retain the client and obtain suf?cient subsequent revenues to cover any initial costs.As a result,an auditor might impair its independence in the current period in order to secure audit and NAS fees in subsequent periods(Levitt1998;Huang,Raghunandan, and Rama2009).In the context of our study,if auditors try to appease clients by not ren-dering a GCM audit opinion in hopes of retaining the company as a client in order to gar-ner fees in the future,examination of current reporting decisions and subsequent fees received from these clients represents a more direct test of the impairment of independence associated with possible client retention efforts on the part of auditors(DeAngelo1981; Levitt1998,2000).

Prior research examining NAS fees in the United States has generally not found the expected negative association between current period NAS fees and auditor’s GCM reporting decisions(DeFond et al.2002;Geiger and Rama2003;Robinson2008;Li 2009).The results of these U.S.studies contrast with some of the?ndings in other countries that have publicly disclosed audit fees for some time(Sharma and Sidhu 2001;Firth2002;Basioudis,Papanastasiou,and Geiger2008).These prior U.S.studies have included the initial years of the SEC’s mandatory fee disclosures(1999and2000), which began with proxy statements?led after February5,2001,which also largely coincided with the enactment of the Sarbanes-Oxley Act(hereafter SOX)in August 2002.Additional research suggests that auditor reporting decisions regarding going con-cern may have become less conservative subsequent to the period surrounding the ini-tial fee disclosures and the enactment of SOX(Geiger,Raghunandan,and Rama2005; Fargher and Jiang2008;Li2009;Feldmann and Read2010).In addition,SOX man-dated considerable restrictions on the provision of NAS from audit?rms to clients, thereby signi?cantly reducing the magnitude of NAS fees obtained by?rms from their audit clients(Ghosh and Pawlewicz2009).It remains an empirical question whether these prior results hold in the post-SOX implementation environment due to possible changes in reporting decisions,as well as to the impact of the new NAS fee restrictions and limited revenue opportunities for external auditors.

We examine these issues by conducting a study that differs from prior research in several important ways.To our knowledge,our study is the?rst to use future revenues CAR Vol.30No.2(Summer2013)

Auditor Fees and Auditor Independence581 to empirically test DeAngelo’s1981assertion that independence may be impaired when auditors become too interested in retaining future revenue streams.We also examine fees received by the auditor and reporting decisions in the post-SOX implementation period of2004–2006.These fees represent the restricted levels of allowable NAS services that auditors can provide to audit clients post-SOX.In addition,our examination per-iod represents a relatively stable economic period in the United States,after the tur-moil of the large?nancial frauds surrounding SOX,and prior to the economic downturn and?nancial crisis beginning in2007.Consistent with Callaghan,Parkash, and Singhal2009,our study also employs a more stringent sample of distressed non-GCM control?rms that are more likely to receive extensive consideration of a GCM opinion from their auditor.3Finally,we also replicate prior U.S.results and document that our differences are due to both the time period and the use of a more stringent control sample.

Our analysis of1,479?nancially distressed companies from2004to2006,including 180with?rst-time GCM opinions,indicates that after controlling for other reporting and fee-related factors,auditors issue signi?cantly fewer GCM opinions in the current period to clients that pay higher subsequent total fees.Speci?cally,auditor reporting decisions in the current period are more favorable for clients that pay higher future fees than for similarly distressed clients that remain viable and stay with their incumbent auditor but pay lower levels of fees in subsequent years.This evidence is consistent with the argument that auditors may appease clients in the current period in order to main-tain future revenue streams from their incumbent clients.Consistent with arguments that auditor reporting may have become less conservative in years following the initial fee disclosure period of1999–2003(Feldmann and Read2010),we?nd that NAS fees in the current year are also signi?cantly negatively associated with GCM audit opinions during the2004–2006period.

As part of our robustness tests,we replicate the DeFond et al.2002and Geiger and Rama2003NAS fee studies and show that our divergent?ndings are due to dif-ferences in auditor reporting decisions in the more recent time period.We also demon-strate the necessity of using a more stringent sample of distressed control?rms when examining GCM opinion decisions in order to achieve adequate power to test our hypotheses.

The remainder of the paper is organized as follows.Section2provides a brief back-ground and literature review and presents the study’s hypotheses.Section3discusses the research method.Section4presents the main results and section5presents robustness and additional tests in support of our main?ndings.Section6concludes the paper with impli-cations,limitations,and future research suggestions.

2.Background,prior research and hypotheses

Audit fees,independence,and auditor going concern judgments

As in most countries,external auditors in the United States collect fees directly from their clients.The possible adverse effect on auditor independence of collecting fees 3.Prior researchers(e.g.,DeFond et al.2002;Fargher and Jiang2008;Lim and Tan2008;Robinson2008)

have identi?ed companies with either negative net income or negative cash?ow from operations as dis-tressed companies in deriving their non-GCM control samples.We only consider companies that have both of these criteria to exhibit suf?cient levels of?nancial distress for their auditors to consider issuing them a GCM opinion.Satisfying these sample selection criteria restricts our control sample to those?nancially stressed companies that would be most likely to be considered candidates for a GCM opinion,but were nonetheless rendered an unmodi?ed opinion.

CAR Vol.30No.2(Summer2013)

582Contemporary Accounting Research

directly from the client has long been an issue of concern for auditors,academicians, regulators,legislators,and the investing public(Mautz and Sharaf1961;Hylton1964; U.S.Senate1976;AICPA1978;Simunic1984;Zeff2003a,2003b).At issue is whether the revenues earned from and paid directly by clients to auditors creates a situation where external auditors become too closely aligned with the client and then begins to lose objectivity,which in turn affects their judgment.In fact,during a recent Standing Advisory Group discussion of the PCAOB’s auditor independence proposals,Lynn Turner,former SEC Chief Accountant,argued that the real problem is that manage-ment is writing the check directly to the auditor(Turner2011).It was concerns like these that led the SEC to adopt a rule requiring public companies to disclose the amount and type of audit and NAS fees paid to their auditor(Securities and Exchange Commission(SEC),2000a,2000b).

Mautz and Sharaf(1961),POB(1995),and Levitt(1998),among others,have argued that the economic bond between auditors and their clients that pay them directly is fraught with dif?culty,particularly in situations where effort and processes are not directly or easily observable,and information asymmetry abounds.Because the ?nal decision regarding what type of audit report to render(GCM or not)to a client is under the control of the auditor,and auditors’decision processes are not publicly observable,Reynolds and Francis(2000),DeFond et al.(2002),and Lim and Tan (2008)argue that reporting decisions represent a direct assessment of auditor indepen-dence,and the possible impairment of that independence associated with receiving larger fees from clients.

Consistent with these arguments,DeAngelo(1981)and Beck et al.(1988)perform ana-lytical examinations of audit fees and the economic bonding between auditor and client and suggest that efforts to retain future fees is potentially a signi?cant impediment to audi-tors maintaining their independence.DeAngelo(1981)argues that an auditor’s objectivity is directly related to the ability to retain the client into the future in order to recover the initial start-up costs on the engagement,including any possible intentional underpricing (i.e.,‘‘low balling’’)of the initial engagement that may have been used in an attempt to acquire the company as a client.Accordingly,an auditor’s independence may be in?u-enced by a desire to retain the client and obtain suf?cient subsequent revenues to recover any initial costs.DeAngelo(1981)argues that threats to independence are driven by the auditor’s expectation of the present value of future‘‘quasi-rents’’(i.e.,the excess of future revenues over future costs)that accrue to incumbent auditors over the duration of the auditor-client relationship.Thus,auditor independence is impaired when an auditor inten-tionally or unintentionally modi?es decisions or decision processes in an attempt to appease the client in order to continue to obtain future fees(AICPA1978;Levitt2000; Huang et al.2009).

In an audit reporting context,if auditors try to appease clients,they would not issue a GCM opinion to their?nancially distressed clients who would otherwise receive one.This is because companies do not welcome the receipt of a GCM opinion and often dismiss the audit?rm following the receipt of a?rst-time GCM opinion(Geiger et al.1998;Carcello and Neal2003).Accordingly,auditors might not issue a GCM opinion in the hope of retaining their incumbent status,and thereby at least temporarily ensure future revenue streams from these engagements.

Prior research in economics and psychology has also demonstrated that expectations of future bene?ts are signi?cant motivators of current decision making(Loewenstein1987; Tversky and Kahneman1991;George and Jones2000;Frederick,Loewenstein,and O’Donoghue2002).In fact,the expectation of future bene?ts can outweigh one’s present position as a motivator for current decision making(Nuttin1985).

CAR Vol.30No.2(Summer2013)

Auditor Fees and Auditor Independence583 Prior research has exclusively examined the effect of current year NAS and audit fees on current year audit reporting decisions(DeFond et al.2002;Geiger and Rama2003) creating a gap between theoretical(e.g.,DeAngelo1981)and applied research.4Thus,in this study we examine the association between current reporting decisions and subsequent total auditor revenues(i.e.,NAS and audit service fees)as a test of the in?uence of future fees on auditor decision making under the assumption that current expectations of subse-quent revenues are re?ected in actual future revenues.5Consistent with the arguments of DeAngelo1981,Beck et al.1988,Gul,Tsui,and Dhaliwal2006,and Lai2009,we expect a negative association between GCM opinion decisions and total fees received in future years.That is,auditors will be less likely to issue GCM opinions in the current period to clients from whom they obtain higher fees in future periods.Thus,our?rst hypothesis(in the alternate form):

Hypothesis1.There is a negative association between going concern modi?ed opinions and the magnitude of subsequent total fees paid to the incumbent audit?rm.

In contrast,there is also an economic incentive for auditors to issue GCM opinions to ?nancially troubled clients.Indicating the auditor’s substantial doubt with respect to the client’s continued viability in the audit opinion helps insulate the auditor from litigation losses in case the client company fails and?les for bankruptcy.Carcello and Palmrose (1994)?nd that in bankruptcy litigation cases against audit?rms,those?rms that rendered GCM opinions to clients before bankruptcy?ling had a signi?cantly lower frequency and magnitude of payouts than those that did not render a GCM opinion prior to bankruptcy?ling.

We also examine the relation between current NAS fees received by the auditor and current GCM opinion decisions.6Any intentional fee retention efforts can negatively impact auditor decision making(AICPA1978;Levitt2000),and auditors who perform signi?cant NAS for audit clients have an added economic incentive beyond the audit fee revenues to appease clients in order to maintain these revenue streams.Thus,generation of NAS fees,and any increased effort on the part of auditors in attempting to ensure these revenue streams,may in?uence judgments made on audit engagements.Congress,during the drafting of the SOX legislation,concluded that signi?cant NAS fees adversely impact 4.In a time-related study,Cahan,Emanuel,Hay,and Wong(2008)examined the association of NAS fee

changes and length of years NAS fees were purchased from the auditor leading up to the reporting year and compare these with levels of the client’s reported discretionary accruals in New Zealand.They?nd that,overall,discretionary accruals levels were not signi?cantly related to growth rates in NAS fees lead-ing up to the reporting year,nor were they related to the length of time the client purchased NAS services prior to the reporting year.However,they do?nd some evidence that the interaction of NAS fee duration and client importance measures are positively related to discretionary accruals.

5.The link between fees and auditor independence in the GCM decision assumes that some situations

exist where the auditor might otherwise conservatively modify the audit opinion,but does not do so because of independence threats.This line of research,and all research relating going concern opinions to auditor independence,assume that the auditor has potential?nancial gain from not modifying an audit opinion.This is potentially not true if the client subsequently fails in the next year and therefore pays no future fees.However,auditors do not modify opinions solely to indicate expected client fail-ure.In fact,prior research has found that between80percent and90percent of companies receiving

a GCM opinion do not fail in the subsequent year(e.g.,Hopwood,McKeown,and Mutchler1994).

Thus,expressing substantial doubt does not mean that the auditor would uniformly expect no future fees from a client.

6.For a more detailed discussion of the relation between NAS fees and auditor’s going-concern opinion deci-

sions see DeFond et al.2002,Geiger and Rama2003,DeFond and Francis2005,and Basioudis et al.

2008.

CAR Vol.30No.2(Summer2013)

584Contemporary Accounting Research

auditor independence.This conclusion led to substantial restrictions on allowable NAS engagements for audit clients(U.S.House of Representatives2002;U.S.Senate2002). Accordingly,prior research has examined the association between current NAS fees and various measures of auditor independence in the United States.7,8

Fee disclosures in the United States were?rst required for proxy statements?led on or after February5,2001for the years1999and2000.In an early large sample study of audit and NAS fee disclosures and audit reporting decisions,DeFond et al.(2002)exam-ined U.S.?rms’initial release of audit and NAS fee information for the2000reporting year and concluded that there is no signi?cant association between either audit fees or NAS fees and audit opinions on?nancially stressed companies.A contemporaneous study by Geiger and Rama2003also examined initial fee disclosures for a sample of66GCM companies and a matched sample of66similarly distressed non-GCM companies.They concluded that audit fees were positively associated with going concern modi?cations,but that NAS fees were not signi?cantly associated with GCM decisions.Other studies that include the earliest fee disclosures similarly do not?nd a negative relation between current NAS fees and GCM reporting decisions(Callaghan et al.2009;Robinson2008;Lim and Tan2008;Li2009).

Lai(2009)examined the association between NAS fees and GCM opinions surround-ing the initial fee disclosures and found that audit?rms were more likely to render a GCM opinion after fees were publicly disclosed.However,several studies suggest that auditor reporting in the face of NAS fees may have changed in the post-SOX implementa-tion years.For example,Ghosh and Pawlewicz(2009)present evidence that post-SOX implementation levels of audit fees are signi?cantly higher and levels of NAS fees are sig-ni?cantly lower than pre-SOX implementation levels.The general reduction in overall NAS fees received from clients in the post-SOX environment could lead to less in?uence on auditor decision making simply due to the lower magnitude of the fees,reducing the economic tie between the?rm and that audit client.In contrast,however,reducing the allowable types of NAS services auditors can provide attest clients may produce greater threats to independence.This is due to the audit?rm’s potential desire to retain all the NAS fees available from clients due to the limitation on allowable NAS services,as well as the reduction in the number of clients offering these revenue opportunities to their external auditor(Ghosh and Pawlewicz2009).

7.See Francis2006and Lim and Tan2008for expanded discussions of the research on NAS fees and vari-

ous measures of auditor independence.For example,considerable research has been performed on the association between NAS fees and the levels of discretionary accounting accruals.A substantial disadvan-tage of assessing auditor independence by examining clients’?nancial reporting practices is that auditors have only an indirect effect on the level of?nancial accruals reported by their client.

8.The relation between NAS fees and auditor reporting behavior has also been previously examined in non-

U.S.reporting contexts.For example,research has been performed in Australia where audit fee data have been publicly disclosed since1980.However,the results are somewhat mixed.Wines(1994)presented evi-dence of a negative association between NAS fees and the issuance of quali?ed reports during1980–1989.

Sharma(2001)and Sharma and Sidhu(2001)conclude that higher NAS fees were associated with a lower likelihood of receiving GCM reports.In contrast,Barkness and Simnett(1994)and Craswell(1999)in two large sample studies?nd no signi?cant association between the level of NAS fees and Australian audit report quali?cations.In a study on U.K.?rms,Lennox(1999)examined NAS fees surrounding the initial fee disclosure requirements instituted in1991.His results suggest that NAS fees were not signi?cantly associated with auditors rendering a GCM opinion.Yet Firth(2002)examined auditor decisions to issue quali?ed opinions in the United Kingdom in the year1996and found a signi?cant negative association between NAS fee levels and receiving a quali?ed opinion.Basioudis et al.(2008)examined U.K.auditor reporting for2003and found that?nancially stressed companies with high audit fees were signi?cantly more likely to receive a GCM audit opinion,whereas companies with high NAS fees were signi?cantly less likely to receive a GCM audit opinion.

CAR Vol.30No.2(Summer2013)

Auditor Fees and Auditor Independence585 Fargher and Jiang(2008)and Feldmann and Read(2010)both demonstrate that audi-tor GCM reporting decisions may have changed in the post-SOX period.Speci?cally, Fargher and Jiang(2008)examine GCM reporting in Australia,and Feldmann and Read (2010)examine GCM reporting on U.S.?rms,and they both conclude that audit?rms were initially more conservative and issued GCM opinions with greater frequency right after the enactment of SOX,but then in years2004and after appear to have reversed their reporting postures and become less conservative.Thus,it remains an empirical question whether the prior research results on the relation between NAS fees and GCM reporting surrounding the initial fee disclosures hold true in the later post-SOX auditing and report-ing environment in the United States.

Accordingly,we examine the association of NAS fees and the possible impairment to auditor independence by assessing whether auditors report more favorably on distressed clients from whom they receive higher current NAS fees.9Thus,our second research hypothesis(in the alternate form):

Hypothesis2.There is a negative association between going concern modi?ed opinions and the magnitude of current year nonaudit service fees paid to the audit?rm.

3.Research method

Sample

Using the Audit Analytics database,we?rst identify?rms that received a GCM audit opinion in the years2004–2006.We start with the year2004based on prior research regarding the shift in auditor reporting decisions after SOX(Fargher and Jiang2008; Feldmann and Read2010)and end in2006in order to allow a suf?cient number of years for the subsequent fees analysis.Our examination period represents a relatively stable eco-nomic period in the United States,after the turmoil of the large?nancial frauds surround-ing SOX,and prior to the economic downturn and?nancial crisis beginning near the end of2007.Consistent with prior research,we restrict our analysis to companies that received a?rst-time GCM because issuing a modi?ed opinion in subsequent years is less risky for the auditor and constitutes a different decision model that may not include the perceived risk of losing a disgruntled client(Geiger et al.1998;DeFond et al.2002;Carcello and Neal2003).

We then obtain our distressed non-GCM?rms by identifying?rms in COMPUSTAT with year-ends from2004–2006that had both negative income and cash?ows from opera-tions in the same year but did not receive a GCM opinion.Following the arguments in Callaghan et al.2009,requiring our distressed control?rms to exhibit both of these?nan-cial stress criteria more appropriately restricts our control sample to the?rms that are most likely to be considered for a GCM opinion by their auditors.10Our selection procedure,while more restrictive than prior studies,provides a more representative control 9.Prior researchers and regulators(Abdel-Khalik1990;SEC2000b;DeFond et al.2002;Whisenant,Sankar-

aguruswamy,and Raghunandan2003;Basioudis et al.2008)have argued that examining the magnitude or impact of NAS fees cannot be properly accomplished without also concurrently assessing the level of audit service fees paid by clients.Therefore,we also examine the association between current audit service fees and auditor reporting decisions.However,it is not the focus of our study and we do not make predic-tions about the direction of the effect.

10.McKeown,Mutchler,and Hopwood(1991)and Hopwood et al.(1994)classify a company as being?nan-

cially distressed if one of four?nancial stress signals were exhibited in the?scal year of the audit opinion being examined or in the preceding two years.DeFond et al.(2002)require a?rm to exhibit either nega-tive income or negative cash?ows from operations to be included in their distressed non-GCM control sample.

CAR Vol.30No.2(Summer2013)

586Contemporary Accounting Research

sample of highly?nancially distressed non-GCM?rms for which auditors must exercise professional judgment regarding a GCM opinion decision.

Consistent with prior going concern reporting studies(Behn,Kaplan,and Krumwiede 2001;Blay and Geiger2001;Carcello and Neal2003;Geiger and Rama2003),we restricted our analysis to companies in the manufacturing sector(SIC20–39)to eliminate confounding industry effects and to keep the?nancial ratios consistent with the develop-ment of the?nancial stress models we employ.We then required?rms to have complete audit report,audit?rm,and fee information for the current year available in the Audit Analytics database,as well as all necessary?nancial and share price data available in the COMPUSTAT and CRSP databases.Results of our screening and data requirements yielded1,479observations:180companies receiving a?rst-time going concern modi?ed audit opinion(the GCM sample),and1,299distressed companies that did not receive a going concern modi?ed opinion(the non-GCM sample).Among the observations are854 unique?rms.Of the non-GCM sample,362?rms appear more than once in the analysis.11 Of the GCM sample,71?rms received a clean opinion in at least one other year in the sample period,with13of these?rms receiving a clean opinion following the modi?cation. No?rms appear in the GCM sample more than once because we include only?rst-time GCM observations.

Research model

Following prior research by DeFond et al.2002and Callaghan et al.2009we test our hypotheses by estimating the coef?cients of the following logistic regression that models the auditor’s probability of issuing a going concern modi?ed audit opinion to a?nancially distressed client:12,13

OPINION?b0tb1eSIZETtb2ePROBANKZTTtb3eAGETtb4eBETAT

tb5eRETURNTtb6eVOLATILITYTtb7eLEVTtb8ePLOSST

tb9eROATtb10eINVESTMENTSTtb11eCFFOT

tb12eISSUE DEBTTtb13eISSUE EQUITYTtb14eSELL ASSETST

tb15eDISC OPERTtb16eBIG4Ttb17eREPORTLAGTtb18eDFTT

tb19elogeFUTURE FEESTTtb20elogeCUR TOTAL FEETTtb21elogeCUR AUDIT FEETT

tb22elogeCUR NAS FEETTtb23elogeCUR FEERATIOTTtee1T; where:

OPINION=an indicator variable equal to1for?rms with going concern modi?ed opin-ions,0otherwise

SIZE=natural logarithm of total assets at the end of the year measured in millions of dollars

PROBANKZ=probability of bankruptcy score from Zmijewski1984

AGE=natural logarithm of the number of years the company has been listed on a stock exchange

BETA=the?rm’s beta estimated using a market model over the?scal year

11.Because observations for speci?c?rms could be included in the sample more than once,all p-values pre-

sented in multivariate models control for?rm-clustered error terms.

12.We also include yearly dummy variables(not reported in the model speci?cations or in the tables)for the

models estimated in the paper in order to control for any differences among the years in our examination period.

13.We primarily use the DeFond et al.2002model in our main analyses for comparability to prior research

that found no signi?cant relation between NAS fees and GCM opinions.In sensitivity testing,we apply other models to demonstrate the robustness of our?ndings.

CAR Vol.30No.2(Summer2013)

Auditor Fees and Auditor Independence587 RETURN=the?rm’s stock return over the?scal year

VOLATILITY=the variance of the residual from the market model over the?scal year LEV=total liabilities over total assets at the end of the?scal year

PLOSS=an indicator variable equal to1if the?rm reports a bottom-line loss in the previous year,0otherwise

ROA=return on total assets at the end of the year

INVESTMENTS=short-and long-term investment securities(including cash and cash equivalents)de?ated by total assets at year-end

CFFO=cash?ows from operations de?ated by total assets at the end of the year ISSUE_DEBT=an indicator variable equal to1when the?rm issues new debt during the year,0otherwise

ISSUE_EQUITY=an indicator variable equal to1when the?rm issues new equity dur-ing the year,0otherwise

SELL_ASSETS=an indicator variable equal to1when the?rm sells?xed assets during the year,0otherwise

DISC_OPER=an indicator variable equal to1when the?rm indicates a discontinued operation in the current year,0otherwise

BIG4=an indicator variable equal to1when the auditor is a member of the Big4,0 otherwise

REPORTLAG=number of days between?scal year-end and the audit report date

DFT=an indicator variable equal to1when the?rm is in default on debt,0otherwise log(FUTURE_FEES)=the natural logarithm of the total fees paid to the incumbent auditor in the subsequent two years

log(CUR_TOTAL_FEE)=the natural logarithm of the total fees paid to the incumbent auditor

log(CUR_AUDIT_FEE)=the natural logarithm of the audit fees paid to the incumbent auditor

log(CUR_NAS_FEE)=the natural logarithm of the non-audit fees paid to the incumbent auditor

CUR_FEERATIO=the ratio of non-audit fees to total fees paid to the incumbent auditor

Variables of interest

In the model above,log(FUTURE_FEES),log(CUR_NAS_FEE),and CUR_FEERATIO are our variables of interest and are added to the model in various combinations to test our hypotheses.In our?rst hypothesis,we examine the association of current GCM decisions and the fees received by incumbent auditors in subsequent years(FUTURE_ FEES).Prior research suggests that individuals put a heavily disproportionate weight on expected bene?ts in the immediate future compared to expected bene?ts to be received at later times(Chapman2000;Frederick et al.2002;Elbert2010).The disproportionate dis-counting of expected future bene?ts,especially for bene?ts expected beyond the very near term,is referred to as‘‘temporal discounting’’.Temporal discounting is the cognitive bias to drastically discount,much more than any rational expectations of discounting would dictate,the expected future bene?ts that are not anticipated to be received in the very immediate future(Chapman1996,2000).Temporal discounting bias goes far beyond both individual preferences for present or immediate reward(i.e.,grati?cation),as well as the uncertainty of the attainment of the bene?ts in the future(Frederick et al.2002). Accordingly,for our main analyses we use a relatively short subsequent fee period of two years.14

14.We then report similar results for varying subsequent time periods in our later analyses.

CAR Vol.30No.2(Summer2013)

588Contemporary Accounting Research

If the expectation of the receipt of these future fees impairs auditor independence in the current period,we expect a negative coef?cient on log(FUTURE_FEES).Similarly,if the receipt of current year NAS fees impairs auditor independence,we expect a negative coef?cient on log(CUR_NAS_FEE)and on CUR_FEERATIO.Because of con?icting prior research,we cannot offer an a priori expectation for GCM opinions and current year audit fees(CUR_AUDIT_FEE)or total fees(CUR_TOTAL_FEE).

Control variables

Following prior research our model includes controls for both contributing and mitigating factors found to be related to auditor GCM report decisions.As noted in SAS No.59,the level of?nancial distress is an important factor in the auditor’s decision,and,although we have tried to include only highly stressed companies in our samples,we provide additional control for?nancial distress by including the probability of bankruptcy in our model. PROBANKZ is the probability of bankruptcy score from Zmijewski1984,with higher val-ues indicating a higher probability of bankruptcy.The log(AGE)variable is also included because younger?rms are more prone to failure(Dopuch,Holthausen,and Leftwich1987; DeFond et al.2002).We include three additional market-based measures,RETURN, BETA,and VOLATILITY.RETURN,the stock price return over the?scal year,and BETA,the systematic risk of the?rm’s stock return,are expected to be negatively associ-ated with GCM opinions.VOLATILITY,the return volatility exhibited by the?rm’s stock over the?scal year,is expected to be positively associated with GCM opinions.Other con-tributing factors in our model include LEV and PLOSS because?rms with higher leverage and those that have reported net losses in the past are more likely to receive a going con-cern modi?ed report(Kida1980;Reynolds and Francis2000).We include CFFO(operat-ing cash?ows divided by total assets)because poor operating cash?ows are often associated with the probability of bankruptcy,and our bankruptcy score does not include cash?ows.

In addition,the model includes several additional factors that are likely to mitigate the probability of receiving a going concern opinion.The SIZE variable,measured as log(assets)at the end of the year,is included because large?rms have more negotiating power in the event of?nancial dif?culties and hence are more likely to avoid bank-ruptcy(Reynolds and Francis2000).We include ROA because more pro?table?rms are less likely to receive GCM opinions.The variable INVESTMENTS is included because?rms with large cash and investment securities have more resources to rely on in the event of?nancial dif?culties(De Fond et al.2002).We also include separate variables for ISSUE_DEBT,ISSUE_EQUITY,and SELL_ASSETS because?rms that are able to raise additional funding during the year from increased borrowing,issuing stock(public or private),or selling assets or operating units are less likely to receive a GCM opinion from their auditor(Mutchler,Hopwood,and McKeown1997;Behn et al.2001;Geiger and Rama2003),and we expect a negative coef?cient on these fac-tors.15Similarly,we expect a negative association between DISC_OPER and GCM opinions because these?rms have modi?ed their former?rm strategy(Behn et al. 2001)and also have additional resources to sustain their ongoing operations(Mutchler et al.1997).We also include the indicator variable BIG4because prior research argues that Big4auditors are more likely to issue going concern audit opinions to stressed companies(Mutchler et al.1997).We include REPORTLAG because audits of ?nancially troubled companies are often more time-consuming and going concern 15.We use a change of10percent in each of these balance sheet components for the reporting year as indica-

tors of raising additional funds through selling assets,equity or increased borrowing.

CAR Vol.30No.2(Summer2013)

Auditor Fees and Auditor Independence589 modi?cations have been found to be associated with longer audit reporting lags (Chen and Church1992;Behn et al.2001;Geiger et al.2005).Finally,we include DFT in our model because clients in default on debt are most likely to receive going concern modi?cations(Chen and Church1992;Geiger and Rama2003;Menon and Williams2010).16

4.Results

Descriptive statistics

Table1presents the descriptive statistics for our full sample of1,479?rms on the variables used in the going concern model in both(1)and other models presented later,as well as a comparison of variable means between the GCM and non-GCM sub-samples.17Our sam-ple?rms have an overall smaller mean FEERATIO at0.15than those examined in DeFond et al.2002having a mean FEERATIO of0.49.This decrease in proportionate NAS fees is indicative of SOX limiting the amount of NAS fees that can be provided to audit clients,and is consistent with prior research on the reduction in NAS fees post-SOX (Ghosh and Pawlewicz2009).

Overall,the?rms in our study have mean total assets of$217.53million,which is lar-ger than the?rms examined in Geiger and Rama2003of$118million and smaller than those in DeFond et al.2002of$813million.In addition,the mean PROBANKZ for our ?rms is0.70,which is very similar to the?rms examined in Geiger and Rama2003of 0.70,but is substantially higher than for the?rms examined in DeFond et al.2002of0.22. This high mean PROBANKZ suggests that our sample selection procedures have identi?ed ?rms that are most likely to be considered for a possible GCM opinion by their audit?rm. The remaining control variable means for our overall sample are similar to those reported in prior studies on?nancially distressed?rms and GCM opinion recipients(DeFond et al. 2002;Geiger and Rama2003;Callaghan et al.2009).

The second panel in Table1presents an examination of the mean differences between the180GCM?rms and the1,299non-GCM?rms.With respect to our fee measures,we ?nd that non-GCM?rms have higher mean CUR_NAS_FEE and CUR_FEERATIO levels in the current year(p-value=.06and.02,respectively).However the levels of FUTURE_ FEES,CUR_TOTAL_FEE,and CUR_AUDIT_FEE are not signi?cantly different between the two groups.In addition,as expected,we?nd that GCM?rms on average have higher stock price declines,volatility,leverage,and reporting lags,and are more likely to be in default on their debt.We also?nd that GCM?rms have,on average,lower betas,fewer investments,and less cash?ow from operations,and that they issue equity less often and are less likely to have a Big4audit?rm.Of particular note,however,is the similarity between the GCM?rms and the non-GCM?rms in terms of SIZE,AGE,PLOSS,ROA, and PROBANKZ.These univariate results reinforce the ef?cacy of our sample selection 16.In a recent study,Menon and Williams(2010)suggest that market reaction to GCM reporting is largely

driven by default status and the amount of institutional ownership of the audited?rm.While our models incorporate default status,if we include the percentage of institutional ownership into(1)modeling GCM opinions and into our model of expected current audit fees we?nd that these variables are not signi?cant in any of the models at conventional levels(p-value>.10),and do not substantively alter any of our reported regression results.We also test whether our models are sensitive to our choice of control vari-ables.We?nd that the results presented are not in?uenced by the removal of nonsigni?cant control vari-ables and also do not appear to be sensitive to the singular removal of other signi?cant control variables in our models.Accordingly,our results appear generally robust with respect to the selection of control variables.

17.All continuous variables are winsorized at the1and99percent levels.However,using the raw unwinsor-

ized data produces essentially the same results.

CAR Vol.30No.2(Summer2013)

TABLE1

Descriptive statistics and comparisons of180going concern modi?ed and1,299?nancially distressed non–going concern modi?ed?rms during2004–2006a

Variables

Full sample(n=1,479)

Mean

differences in subsamples Mean SD Min Max

GCM

(n=180)

Non-GCM

(n=1,299)p-value

ASSETS($000,000)217.531,040.040.3518,112.00229.87215.830.89 PROBANKZ0.700.380.00 1.000.730.690.19 AGE(in years)13.058.39 1.0056.0013.4013.000.64 BETA 1.090.71-0.60 3.000.96 1.120.01 RETURN)0.010.78-0.88 3.84)0.160.010.03 VOLATILITY0.000.000.009.240.000.000.00 LEV0.500.490.007.500.980.440.00 PLOSS0.850.350.00 1.000.840.860.68 ROA)0.360.75)500.10515.00)0.42)0.350.61 INVESTMENTS0.480.310.000.970.340.500.00 CFFO)0.360.51)3.390.92)0.83)0.300.00 ISSUE_DEBT0.300.460.00 1.000.330.300.32 ISSUE_EQUITY0.740.440.00 1.000.550.770.00 SELL_ASSETS0.050.210.00 1.000.030.050.33 DISC_OPER0.120.320.00 1.000.090.120.31 BIG40.610.490.00 1.000.370.640.00 REPORTLAG73.3826.1524.00213.0082.2472.160.00 DFT0.080.280.00 1.000.280.060.00 BM0.250.62)2.70 1.73)0.150.310.00 RESTATE0.160.370.00 1.000.120.160.07 INVRECRATIO0.210.200.000.820.230.200.07 SEGNUM 1.58 1.30 1.0012.00 1.54 1.580.62 sqrt(EMPLOYEES)0.500.720.0013.570.450.500.34 PENSION0.100.300.00 1.000.070.100.24 FORGN0.490.050.00 1.000.500.490.52 BADCONTROLS0.070.250.00 1.000.070.060.70 ACCEL_FILER0.720.450.00 1.000.530.750.00 ACQ0.090.290.00 1.000.080.090.46 FUTURE_FEES($000)b1329.684263.100.0098,765.001,219.881,344.900.71 CUR_TOTAL_FEE($000)727.822,501.02 2.0067,300.00670.32735.780.71 CUR_NAS_FEE($000)116.98458.630.0010,900.0077.70122.420.06 CUR_AUDIT_FEE($000)610.84220.93 2.0059,500.00592.62613.370.89 CUR_FEERATIO0.150.160.000.900.120.160.02 Notes:

a Financial distress is de?ned as having both negative net income and cash?ow from operations in

the same year.

b For1,469distressed?rms that did not?le for bankruptcy in the subsequent year.

All p-values are two-tailed.

(The table is continued on the next page.) 590Contemporary Accounting Research

CAR Vol.30No.2(Summer2013)

Auditor Fees and Auditor Independence591 TABLE1(Continued)

Variables are de?ned as follows:

ASSETS=total assets at the end of the year

PROBANKZ=probability of bankruptcy score from Zmijewski1984

AGE=number of years the company has been listed on a stock exchange

BETA=the?rm’s beta estimated using a market model over the?scal year

RETURN=the?rm’s stock return over the?scal year

VOLATILITY=the variance of the residual from the market model over the?scal year

LEV=total liabilities over total assets at the end of the?scal year

PLOSS=an indicator variable equal to1if the?rm reports a bottom-line loss in the previous year,0otherwise

ROA=return on total assets at the end of the year

INVESTMENTS=short-and long-term investment securities(including cash and cash equivalents) de?ated by total assets at year-end

CFFO=cash?ows from operations de?ated by total assets at the end of the year

ISSUE_DEBT=an indicator variable equal to1when the?rm issues new debt during the year,0 otherwise

ISSUE_EQUITY=an indicator variable equal to1when the?rm issues new equity during the year,0otherwise

SELL_ASSETS=an indicator variable equal to1when the?rm sells?xed assets during the year, 0otherwise

DISC_OPER=an indicator variable equal to1when the?rm reports a discontinued operation in the current year,0otherwise

BIG4=an indicator variable equal to1when the auditor is a member of the Big4,0otherwise REPORTLAG=number of days between?scal year-end and the audit report date

DFT=an indicator variable equal to1when the?rm is in debt default during the year,0otherwise BM=book value of equity divided by the market value of equity at the end of the year RESTATE=an indicator variable equal to1when the?rm announces or restates prior?nancial information during the year,0otherwise

INVRECRATIO=inventory plus accounts receivable divided by total assets at the end of the year SEGNUM=number of reported business segments

sqrt(EMPLOYEES)=square root of the number of employees employed by the company PENSION=an indicator variable equal to1if the?rm offered an employee pension or post-retire-ment plan,0otherwise

FORGN=an indicator variable equal to1if the?rm had foreign operations during the year,0 otherwise

BADCONTROLS=an indicator variable equal to1if the?rm reported an internal control weak-ness(SOX404or302)during the year,0otherwise

ACCEL_FILER=an indicator variable equal to1if the?rm was an accelerated?ler,0otherwise

(The table is continued on the next page.)

CAR Vol.30No.2(Summer2013)

592Contemporary Accounting Research

TABLE1(Continued)

ACQ=an indicator variable equal to1if the?rm acquired another company during the year,0 otherwise

FUTURE_FEES=the total fees paid to the incumbent auditor in the subsequent two years

CUR_TOTAL_FEE=the total fees paid to the incumbent auditor in the current year

CUR_NAS_FEE=the nonaudit fees paid to the incumbent auditor in the current year

CUR_AUDIT_FEE=the audit fees paid to the incumbent auditor in the current year

CUR_FEERATIO=the ratio of nonaudit fees to total fees paid to the incumbent auditor in the current year

procedures in identifying an equivalent control sample of non-GCM?rms in terms of size, age,pro?tability,and?nancial stress that appear most likely to be considered by their auditor for a GCM opinion.Table2presents the Pearson correlation matrix for the vari-ables used in the study.18

Multivariate analyses

Table3presents the results of estimating the logistic model in(1)using future and current year fee measures to test our hypotheses.Following DeFond et al.2002,Model1presents a baseline case of our going concern model without including any of the fee variables,and Models2–6introduce various combinations of our fee variables.The results indicate that Model1does a reasonably good job of explaining the going concern decision.The pseudo R2is30percent,and we?nd signi?cance in the predicted direction at p-value<.05for the coef?cients on SIZE,VOLATIVITY,LEV,INVESTMENTS,CFFO,ISSUE_ EQUITY,REPORTLAG,and DFT.Similar to prior research,we do not?nd signi?cance in the predicted direction for the coef?cients on AGE,BETA,ISSUE_DEBT,SELL_ ASSETS,and DISC_OPER(Dopuch et al.1987;Callaghan et al.2009).Also of note is our lack of signi?cance on the PROBANKZ variable,reinforcing the?nding that our sam-ple selection procedures provided a similarly distressed sample of non-GCM control?rms in comparison to our GCM sample?rms.

GCM opinions and subsequent fees–Tests of Hypothesis1

In order to examine our?rst hypothesis with respect to the relation between current reporting decisions and fees received in subsequent years,we analyzed the sum of total fees received by the incumbent auditor for the two years immediately after the current year.We eliminated ten?rms(8GCM?rms and2non-GCM?rms)from the sample for the subsequent fees analysis because they?led for bankruptcy in the following year and, accordingly,did not remain viable for at least one subsequent year.

Table3,Model2presents the results of estimating the logistic model in(1)using our subsequent fee measure(FUTURE_FEES)to test the?rst hypothesis.The results indicate a negative association between current year auditor GCM reporting decisions and subse-quent fees collected by the incumbent auditor(p-value=.04).These results provide support for Hypothesis1.We?nd that auditors are less likely to issue a GCM opinion to clients that pay higher subsequent total fees.

18.Tests for multicollinearity on all multivariate models indicate that the highest variance in?ation factor

(VIF)was4.05and in most cases are less than2.Therefore,multicollinearity does not appear to be a sig-ni?cant impediment to the interpretation of our models.

CAR Vol.30No.2(Summer2013)

TABLE2

Pearson correlation matrix a

V1

ASSETS V2V3V4V5V6V7V8V9V10V11 V2:PROBANKZ0.369**

V3:AGE0.114**)0.089**

V4:BETA0.366**0.300**0.018

V5RETURN0.024)0.042)0.0060.056*

V6:VOLATILITY)0.270**)0.147**)0.029)0.108**0.021

V7:LEV)0.165**0.315**0.086**0.009)0.0110.036

V8:PLOSS)0.0070.114**)0.099**0.091**0.0220.074**)0.02

V9:ROA0.181**)0.102**0.065*0.0430.039)0.069**0.008)0.108**

V10:INVESTMENTS0.0240.229**)0.209**0.111**0.045)0.055*)0.301**0.205**)0.146**

V11:CFFO0.451**)0.216**0.135**0.061*0.116**)0.138**)0.367**)0.121**0.174**)0.151**

V12:ISSUE_DEBT0.153**0.146**0.0090.0120.007)0.0480.227**)0.0050.005)0.158**0.025 V13:ISSUE_EQUITY0.152**0.090**)0.0360.116**0.023)0.077**)0.192**0.118**)0.0160.194**0.009 V14:SELL_ASSETS)0.0220.062*0.0200.091**)0.046)0.0300.0390.025)0.0100.004)0.049 V15:DISC_OPER0.109**)0.0370.197**)0.0090.019)0.0040.046)0.079**0.056*)0.176**0.089** V16:BIG40.542**0.375**)0.0240.256**)0.024)0.116**)0.138**0.062*0.0100.295**0.120** V17:REPORTLAG)0.0180.020)0.022)0.080**)0.057*0.070**0.167**)0.010**0.015)0.172**)0.023 V18:DFT)0.056*)0.0140.107**)0.079**)0.0220.0400.210**)0.074**0.039)0.255**)0.026 V19:CUR_TOTAL_FEE($000)0.818**0.388**0.112**0.275**)0.034)0.171**0.065*)0.0030.107**0.0040.272** V20:CUR_NAS_FEE($000)0.392**0.187**0.074**0.133**)0.009)0.041)0.008)0.0250.084**)0.0460.120** V21:CUR_AUDIT_FEE($000)0.797**0.389**0.120**0.273**)0.038)0.180**)0.067**0.0120.100**0.0250.265** V22:CUR_FEERATIO0.244**0.115**0.056*0.078**)0.004)0.005)0.006)0.0220.075**)0.0410.073** V23:FUTURE_FEES($000)0.276**0.171**0.0190.137**0.041)0.088*)0.063*0.0280.0130.095**0.071** V24:BM0.209**)0.199**0.145**)0.036)0.094**)0.032)0.539**)0.061*)0.089**)0.0460.285** V25:RESTATE0.010)0.015)0.0200.0160.0310.009)0.023)0.005)0.0390.0220.001 V26:INVRECRATIO)0.130**)0.180**0.187**)0.143**)0.0420.0190.21)0.257**0.181**)0.684**0.134**

(The table is continued on the next page.)Auditor Fees and Auditor Independence 593

CAR Vol. 30 No. 2 (Summer 2013)

TABLE2(Continued)

V1ASSETS V2V3V4V5V6V7V8V9V10V11 V27:SEGNUM0.209**)0.0220.207**0.0470.0360.0340.039)0.134**0.068**)0.287**0.145** V28:sqrt(EMPLOYEES)0.578**0.167**0.218**0.175**)0.012)0.117**0.102**)0.117**0.130**)0.259**0.202** V29:PENSION0.511**0.206**0.224**0.172**0.016)0.115**0.129**)0.079**0.115**)0.184**0.168** V30:FORGN0.0390.038)0.064*0.0380.0160.053*0.018)0.0180.0210.0430.023 V31:BADCONTROLS0.239**0.057*0.087**0.089**)0.058*)0.0240.037)0.0380.063*)0.120**0.104** V32:ACCEL_FILER)0.263**0.198**0.0140.242**0.074**)0.147**)0.083**0.048*0.061*0.173**0.087** V33:ACQ0.239**0.0070.0330.058*0.010)0.029)0.003)0.093*0.084**)0.206**0.136**

V12V13V14V15V16V17V18V19V20V21V22

V13:ISSUE_EQUITY)0.019

V14:SELL_ASSETS0.032)0.012

V15:DISC_OPER0.015)0.078**0.062*

V16:BIG40.051*0.147**0.0310.018

V17:REPORTLAG0.089**)0.079**0.0110.073**0.007

V18:DFT0.086**)0.084**0.057*0.065*)0.115**0.141**

V19:CUR_TOTAL_FEE($000)0.139**0.132**0.075**0.157**0.543**0.103**)0.007

V20:CUR_NAS_FEE($000)0.0430.057*0.056*0.054*0.308**0.027)0.0090.449**

V21:CUR_AUDIT_FEE($000)0.134**0.134**0.062*0.156**0.535**0.095**)0.0020.979**0.347**

V22:CUR_FEERATIO0.0190.0370.0430.0160.215**0.009)0.0130.275**0.977*0.171**

V23:FUTURE_FEES($000)0.0110.075**)0.018)0.0350.185**)0.130**)0.073**0.250**0.161**0.241**0.118** V24:BM)0.095**0.071**0.053*0.0310.056*)0.100**)0.0400.074**0.0420.076**0.053* V25:RESTATE)0.0270.0160.0170.024)0.0060.001)0.0010.0210.0260.0080.004 V26:INVRECRATIO0.031)0.190**0.062*0.105**)0.259**0.108**0.175**)0.067*0.015)0.083**)0.099** V27:SEGNUM0.046)0.083**0.0150.207**)0.0490.100**0.067**0.204**0.111**0.193**)0.010 V28:Sqrt(EMPLOYEES)0.154**)0.057*0.0230.185**0.231**0.112**0.095**0.562**0.261**0.547**0.113** V29:PENSION0.104**)0.0490.059*0.161**0.216**0.0250.075**0.452**0.240**0.434**0.089** V30:FORGN0.030)0.0260.010)0.125**0.056*)0.0050.0140.007)0.0200.0120.081**

(The table is continued on the next page.)594 Contemporary Accounting Research

CAR Vol. 30 No. 2 (Summer 2013)

TABLE2(Continued)

V12V13V14V15V16V17V18V19V20V21V22

V31:BADCONTROLS0.064*)0.0130.0460.075**0.0110.241**0.062*0.339**0.072**0.348**)0.050 V32:ACCEL_FILER0.0270.081**0.0140.0270.187**)0.107**)0.0400.224**0.0340.230**0.084** V33:ACQ0.070*0.0430.0630.067*0.075**0.055*0.0240.223**0.1740.202**0.011 V23V24V25V26V27V28V29V30V31V32 V24:BM0.053*

V25:RESTATE0.004)0.030

V26:INVRECRATIO)0.099**0.046)0.024

V27:SEGNUM)0.0100.051*0.0090.157**

V28:sqrt(EMPLOYEES)0.113**0.0490.0200.159**0.355**

V29:PENSION0.089**0.019)0.0020.103**0.241**0.575**

V30:FORGN0.081**)0.012)0.021)0.0210.0090.0160.015

V31:BADCONTROLS)0.0500.0510.075**0.088**0.138**0.305**0.209**0.012

V32:ACCEL_FILER0.084**0.0190.049)0.124**0.0200.118**0.132**0.073**0.067**

V33:ACQ0.0110.111**)0.0340.063*0.137**0.191**0.203**0.0140.104**0.016

Notes:

a Variables are as de?ned in Table1.

*Signi?cant at the.05level.

**Signi?cant at the.01level.Auditor Fees and Auditor Independence 595

CAR Vol. 30 No. 2 (Summer 2013)

TABLE3

Regression results for going concern modi?cation opinion models for180going concern modi?ed and1,299?nancially distressed non–going concern modi-?ed?rms during2004–2006a

Parameter Exp.

sign

Model1Model2b Model3Model4Model5Model6 Coef?cient p-value Coef?cient p-value Coef?cient p-value Coef?cient p-value Coef?cient p-value Coef?cient p-value

Intercept 1.200.06)0.490.12)1.500.21)2.400.08)0.900.09)1.470.17 SIZE))0.440.00)0.510.00)0.460.00)0.490.00)0.420.00)0.460.00 PROBANKZ+)0.140.370.080.38)0.140.36)0.130.38)0.150.32)0.160.31 log(AGE)))0.210.14)0.240.14)0.210.14)0.220.13)0.240.07)0.240.07 BETA)0.130.210.150.210.130.210.120.220.130.210.130.20 RETURN))0.150.18)0.100.19)0.150.18)0.150.19)0.140.14)0.140.14 VOLATILITY+113.800.0192.450.01112.600.01112.800.01119.500.02117.600.02 LEV+0.830.000.610.000.830.000.820.000.830.000.830.00 PLOSS+0.140.310.240.300.140.310.110.350.060.420.060.42 ROA)0.130.150.180.130.130.150.150.130.130.190.130.19 INVESTMENTS))1.230.00)1.340.00)1.230.00)1.320.00)1.360.00)1.370.00 CFFO))0.920.00)0.940.00)0.910.00)0.920.00)0.960.00)0.950.00 ISSUE_DEBT)0.020.46)0.010.460.020.47)0.020.470.010.48)0.010.48 ISSUE_EQUITY))0.580.00)0.560.01)0.590.00)0.580.00)0.580.01)0.580.00 SELL_ASSETS))0.750.18)0.620.18)0.750.18)0.710.19)0.690.10)0.690.10 DISC_OPER))0.210.26)0.320.24)0.210.25)0.230.24)0.240.24-0.240.23 BIG4+0.290.120.350.120.280.140.300.130.290.120.270.15 REPORTLAG+0.010.040.000.050.010.040.010.040.010.020.010.03 DFT+ 1.630.00 1.720.00 1.620.00 1.590.00 1.580.00 1.570.00

(The table is continued on the next page.)596 Contemporary Accounting Research

CAR Vol. 30 No. 2 (Summer 2013)

TABLE3(Continued)

Parameter Exp.sign

Model1Model2b Model3Model4Model5Model6 Coef?cient p-value Coef?cient p-value Coef?cient p-value Coef?cient p-value Coef?cient p-value Coef?cient p-value

log(CUR_AUDIT_FEE)?0.150.31

log(CUR_NAS_FEE)))0.050.02

CUR_FEERATIO))1.700.01)1.720.01 Pseudo R20.300.400.310.310.310.31 Notes:

a Financial distress is de?ned as having both negative net income and cash?ow from operations in the same year.p-values are one-tailed for signed expec-

tations,two-tailed for unsigned expectations.Variables are as de?ned in Table1.

b For1,469distressed?rms that did not?le for bankruptcy in the subsequent year.Auditor Fees and Auditor Independence 597

CAR Vol. 30 No. 2 (Summer 2013)

598Contemporary Accounting Research

To determine whether our subsequent fee results are sensitive to the length of the sub-sequent fee period assessed(i.e.,two years),we examine varying lengths of time for the subsequent fee period.Speci?cally,when we expand our subsequent fee examination per-iod to three years,four years,or all available years of subsequent fees through?scal year 2010,we obtain results consistent with those presented.In each of these analyses we con-tinue to?nd a signi?cant negative association between subsequent total fees received by the auditor and current GCM opinions(p-value<.05).In addition,examining only the fees in the next year,or taking a present value of future fees by using a.10discount rate (based on DeAngelo1981),for each of the subsequent periods assessed(i.e.,two years, three years,four years,and all available years),produces substantively the same results as reported.Further,we do not?nd any association if we analyze audit and nonaudit fees separately,indicating that it is the summation of all future fees that is associated with current GCM opinion decisions and not necessarily how those future fees are obtained (i.e.,NAS versus audit service fees).

GCM opinions and current year fees–Tests of Hypothesis2

Models3–6introduce combinations of our current year fees to examine Hypothesis2.The results of Model3examining the association between OPINION and log(CUR_ TOTAL_FEE)indicates that there is no signi?cant overall association between total fees paid in the current year and GCM opinion decisions(p-value>.10).These?ndings are consistent with prior research(DeFond et al.2002).19

However,a different picture emerges once we separate the current year’s audit service fees and NAS fees.Model4separates log(CUR_NAS_FEE)and log(CUR_AUDIT_FEE) and indicates that there is no signi?cant positive association between GCM opinions and audit fees(p-value=.3),as in DeFond et al.2002.However,we do?nd a signi?cant negative association between current NAS fees and receipt of a GCM opinion(p-value =.02).While these NAS fee results are consistent with our hypotheses,they are in stark con-trast to those of prior research in the United States.Our analysis reveals a signi?cant reduc-tion in the probability of receiving a GCM opinion when current NAS service fees are high.

Our signi?cant current NAS fee?ndings are also echoed in our NAS fee ratio analyses. In Models5and6we assess the CUR_FEERATIO,representing the proportion of current NAS fees to total fees paid to the company’s auditor in the current year with and without the addition of CUR_TOTAL_FEE as an additional control.20Results of both of these mod-els indicate that a higher CUR_FEERATIO is correlated with a lower probability of receiv-ing a GCM opinion(p-value=.01).Thus,unlike research using initial fee disclosure periods,we?nd a signi?cant negative association between current NAS fee levels and the receipt of a GCM opinion in the post-SOX reporting environment of2004–2006. Controlling for expected fees and endogeneity

As noted by DeFond et al.2002and Callaghan et al.2009,a potential limitation of the analysis of current year fees is that auditor independence may be in?uenced by the amount of client fees relative to their expected amounts,rather than the nominal amounts.As a result,auditors may be in?uenced by whether the client is a source of unusually high or low NAS and audit service fees in the current year.Therefore,we model audit and NAS fees in order to develop an expected fee level that will then 19.Geiger and Rama(2003)present an examination of audit fees and NAS fees separately,but did not exam-

ine the combined total fees paid to the auditor.

20.Examining the ratio of current NAS fees to audit service fees,or the ratio of the log(CUR_NAS_FEE)to

the log(CUR_TOTAL_FEE),or the ratio of the log(CUR_NAS_FEE)to the log(CUR_AUDIT_FEE)pro-duces similar results to those presented in the paper for CUR_FEERATIO.

CAR Vol.30No.2(Summer2013)

相关主题