Are Auditors Compromised by Nonaudit Services? Assessing the Evidence* JERE R. FRANCIS, University of Missouri-Columbia Abstract Ruddock, Taylor, and Taylor (2006) use an earnings conservatism framework to investigate the effects of nonaudit services (NAS) on earnings conservatism, and to test whether audit quality was impaired by NAS in Australia during the 1990s. They find no evidence of differential conservatism conditional on the level of NAS fees paid to auditors, and thus conclude that NAS have no adverse effect on audit quality. While this result may not extrapolate to the U.S. setting due to institutional difference between the two countries, the study does add to a growing body of empirical evidence that questions whether there is any logical rationale for restricting the scope of the services that auditors provide to their audit clients. In reviewing the NAS research literature over the past 40 years, one has to conclude that there is no “smoking gun” evidence linking the provision of nonaudit services with audit failures. However, the literature also finds that NAS can adversely affect the appearance of auditor independence, and this may be more than a “mere perception” problem, because there is also evidence that stock prices are significantly lower for companies that pay their auditors large fees for nonaudit services.

1. Introduction Ruddock, Taylor, and Taylor (2006) (hereafter RTT) report evidence from Australia that audit quality was not compromised during the 1990s by the auditor’s provision of nonaudit services (NAS) to audit clients. The earnings conservatism framework of Basu 1997 is used to measure audit quality, and RTT document that earnings conservatism is unaffected by the level of fees a client pays for nonaudit services. They conclude “that the recent legislative intervention [in the United States] aimed at restricting the supply of NAS is unlikely to result in increased independence” (701–2) As discussed later in this commentary, I am skeptical that evidence from the Australian institutional setting can be directly extrapolated to make such a strong claim about U.S. regulations and policy prescriptions. However, the study does provide evidence that NAS are not problematic in Australia, a country that has institutions similar to those of other common-law countries with developed economies, such as Canada, the United Kingdom, and the United States. Keywords Auditor independence; Audit quality; Conservatism; Nonaudit services JEL Descriptors M41, M42 *

Accepted by Dan Simunic. An earlier version of this paper was presented at the 2004 Contemporary Accounting Research Conference, generously supported by the Canadian Institute of Chartered Accountants and Ordre des comptables agréés du Québec. Professor Francis is also an Honorary Professorial Fellow, University of Melbourne.

Contemporary Accounting Research Vol. 23 No. 3 (Fall 2006) pp. 747–60 © CAAA

748

Contemporary Accounting Research

In a broader sense, RTT add to the growing empirical literature examining the effect of NAS on auditor behavior and whether audit quality is compromised by such services. The Panel on Audit Effectiveness (2000, appendix D) notes that public controversy over the appropriateness of auditor-provided NAS goes back at least to the 1950s. Within the profession, the provision of NAS to audit clients has long been a topic of concern with respect to the appropriateness of providing taxation services to audit clients (Cannon 1952). This matter was officially resolved when the Code of Ethics of the American Institute of Certified Public Accountants (AICPA) was modified to explicitly recognize that tax-related work for clients did not violate independence rules on the grounds that the outcome of such work is subject to independent adjudication by taxation authorities. The provision of NAS continues to be a controversial topic even though the scope of such services has been sharply curtailed by the Securities and Exchange Commission (SEC) 2000b and by additional restrictions imposed by the SarbanesOxley Act of 2002. As an aside, taxation is now the largest nonaudit service activity on most audit engagements, and these services are generally viewed positively as a logical add-on to the audit and even helpful to the audit in verifying tax-related accounts in financial statements.1 However, this view has been recently tarnished by evidence of aggressive tax planning by several of the large accounting firms. As a consequence, the Public Company Accounting Oversight Board (PCAOB) has now issued guidance on the scope of taxation services with respect to tax planning. Despite cutbacks in NAS, it is clear that the appropriateness of auditor-provided NAS continues to be controversial and viewed with skepticism by regulators. For this reason, it is important that there be ongoing research to facilitate wellinformed policy making by regulators with respect to the costs and benefits of restricting the scope of NAS to audit clients. 2. What do we know from extant research? Before discussing RTT, I begin by reviewing what we know from the extant research literature about the effects of NAS on auditor behavior. The first relevant literature is concerned with a general psychological dependence that may exist in the auditor – client relationship and is termed “self-serving” bias by Bazerman, Morgan, and Lowenstein 1997. Drawing on experimental evidence in psychology, and extrapolating this evidence to the auditing setting, Bazerman et al. conclude that auditors have an unconscious bias that prevents impartial audits because of close relationships and repeated interactions with client personnel with whom auditors identify socially. The term “self-serving bias” refers to the cognitive characteristic that individuals cannot separate their own self-interest from that of others in close proximity with whom they interact closely. This idea has been given a fuller analysis by Moore et al. 2006, who use the term “moral seduction” to describe the institutional setting of auditing in which close interactions with clients and common economic interests subvert the possibility of a truly independent and impartial audit. Nelson (2006) argues that the critique in Moore et al. 2006 overstates the case against auditor independence, and indeed a laboratory economics experiment by King 2002 shows that social norms among auditors mitigate the CAR Vol. 23 No. 3 (Fall 2006)

Are Auditors Compromised by Nonaudit Services?

749

propensity for self-serving bias. While this research stream raises basic and fundamental questions about the very possibility of auditor independence, as an empiricist I am skeptical about claims that an unconscious (and therefore unmeasurable?) bias exists. Nelson (2006) also questions the usefulness of the conscious/unconscious dichotomy as well as the implied dichotomy of psychological versus economic incentives, and suggests instead that the central issue is simply how auditors’ incentives (including social and economic pressures) affect their decisions.2 I begin with this broader research on the possibility of auditor independence because it is puzzling why the NAS issue has been so compartmentalized by both policymakers and scholars. By “compartmentalized” I mean that the NAS problem is typically treated in isolation relative to the broader contracting environment in which auditing occurs. This is also how RTT treat the topic in their study. There are two broad conjectures as to why NAS are inherently problematic for auditors. First is the possibility that such services fundamentally change the auditor’s role from outside skeptical reviewer to inside adviser and decision maker, and that this change compromises the auditor’s ability to be an independent outsider. Second is that the increasing fee reliance on NAS creates an economic bond that compromises auditor independence. The simple fact is that auditors are hired by clients and paid to render an “impartial” opinion. Thus the inherent contracting environment creates the potential for moral seduction described in Moore et al. 2006. However, in this regard it seems that NAS are simply another layer of relations that complicates auditor independence, rather than something entirely different or unique. The same is true of the economic dependence created by fees. All fees create a fee dependence on the client, not just NAS fees. Nearly 30 years ago the Commission on Auditors’ Responsibilities (1978, 93) observed that a completely independent audit is, by definition, impossible due to the fee dependence inherent in audit contracting. Despite the close similarities between auditing and NAS, the tendency has been to view NAS in isolation and to compartmentalize them relative to other aspects of the auditor–client relationship and contracting environment. Like Nelson 2006, I believe that the analysis of auditor independence requires a more comprehensive analysis of incentives and the institutional setting in which audit contracting takes place, and therefore I question whether a stand-alone marginal analysis of the NAS issue can really inform us of very much without taking in account this larger context in which auditors and their clients operate. An obvious implication is that researchers should examine the total fee dependence on a client as well as the marginal fee dependence effects from additional nonaudit service activities (e.g., Reynolds and Francis 2000). Historically, the professional accounting literature on auditor independence has stipulated that auditors must be independent both in fact and in appearance. Carey (1970) points out that this distinction first emerged in the 1930s, following the adoption of the Securities Act of 1933 and the Securities and Exchange Act of 1934, which banned certain common practices of the time, including serving on a client’s board of directors. With these new independence requirements, the presumption was that an auditor was “factually” independent of a client, and professional CAR Vol. 23 No. 3 (Fall 2006)

750

Contemporary Accounting Research

standards subsequently focused on appearance-related issues with an extensive set of rules prohibiting actions or relations with clients that might create the appearance of impropriety. It is in this context that NAS emerge as a potentially problematic practice, at least in appearance, and the professional code of conduct has engaged in “perception management” over the years by setting a boundary on the scope of nonaudit work that can be performed for audit clients. Even so, it is clear that the appearance problem cannot be entirely eliminated through ethics rules alone. As the Panel on Audit Effectiveness (2000, 115) observes, “So long as auditors provide non-audit services to audit clients, there will be at least an issue with respect to the appearance of independence.” The research literature has also been influenced by the independence in fact/ independence in appearance only dichotomy. There is quite a lengthy survey literature in which financial statements users were asked their opinions or perceptions of the effects of NAS on auditor independence. The earliest of these studies date from the 1960s, and they consistently find that a significant minority of the respondents view the provision of NAS as a potential threat to auditor independence (e.g., Schulte 1965; Hartley and Ross 1972; Titard 1971; Pany and Reckers 1984; Earnscliffe Research & Communications 1999). There has also been experimental research in which the level of NAS is manipulated to determine its effect on perceptions and hypothetical decision making by financial statement users. Most of these studies report evidence that NAS impair the perception of auditor independence and undermine the reliability of audited information (Lavin 1976; Shockley 1981; Pany and Reckers 1984), although Pany and Reckers (1987) caution that these results could be driven by demand effects, because the studies with significant results on NAS manipulations are typically within-subject designs. In contrast, studies using a between-subjects design, where demand effects are less likely to occur, have not found significant results on NAS manipulations, although this could also be the consequence of the inherently lower power of between-subject designs (McKinley, Pany, and Reckers 1985; Pany and Reckers 1987).3 In sum, the perception research discussed above, while somewhat dated, points to potential independence problems. Survey research is viewed with some skepticism in the accounting research literature, and experiments may not necessarily have external validity with respect to decision tasks, auditor incentives, or institutional settings. However, recent empirical research provides additional support for this earlier perception literature. Two recent studies re-examine the NAS perception issue by studying the stock market’s reaction to the disclosure of NAS fees. These studies show that the negative perceptions of NAS found in surveys appear to be more than “mere” perceptions. Krishnan, Heibatollah, and Zhang (2005) and Francis and Ke (2006) both report significant and negative stock price reactions for U.S. firms with higher levels of NAS fees. The research design in both studies uses an earnings response coefficient (ERC) framework to examine whether the market reaction to quarterly earnings surprises following the initial fee disclosures (beginning in February 2001) was affected by the level of reported NAS fees. Krishnan et al. (2005) find that the market response to earnings surprises was lower in 2001 for firms with higher levels of fees measured three ways: CAR Vol. 23 No. 3 (Fall 2006)

Are Auditors Compromised by Nonaudit Services?

751

as the ratio of nonaudit fees to total fees, as the dollar magnitude of nonaudit fees, and as the dollar magnitude of total fees. Using a somewhat different design, Francis and Ke (2006) compare the market response to quarterly earnings surprises for firms both before and after the initial fee disclosures that began in February 2001. They find that firms with high levels of NAS fees have lower ERCs following fee disclosure than in the immediately preceding pre-disclosure period. In addition, they find that ERCs in the post-disclosure period are lower for firms with high NAS fees than for firms with low NAS fees. On average, the magnitude of the reduction in ERCs is 43 percent for firms with high NAS relative to other firms. Francis and Ke (2006) define high NAS firms in various ways, and their sensitivity analyses suggest that the dollar magnitude of nonaudit fees matters more than the relative relation between nonaudit and audit fees. In their study, nonaudit fees appear to become problematic somewhere between $369,000 and $738,000 (Francis and Ke 2006, footnote 20). Overall, the foregoing evidence consistently points to the likelihood that users/investors may perceive that NAS negatively affect auditor independence and lower the quality of audits. Given that much of this evidence predates the rise of the consulting culture in the 1990s, it is quite possible that these studies understate the level of anxiety over the provision of NAS (Levitt 1998; Securities and Exchange Commission 2000a). Of course, we have no way of knowing whether these negative perceptions are truly justified, but the profession has long acknowledged that the perception of an independence problem is just as serious as the factual impairment of independence. I have argued elsewhere that the perception problem will always exist and cause periodic crises of confidence in auditing, and for this reason the accounting profession may be better off by voluntarily eliminating the provision of NAS for audit clients (Francis 2004).4 Let us turn now to research on NAS and the factual impairment of auditor independence. An auditor that knowingly is not independent of a client is in outright violation of professional standards and, for listed securities, would also be in outright violation of the SEC’s independence rules. In such cases, it is likely that there is also fraud in terms of the deliberate reporting of materially misleading information. Evidence on the factual impairment of auditor independence is very limited and there are few proven cases of auditor fraud or the deliberative violation of auditor independence. Indeed, this was one of the basic arguments the profession used in opposing the SEC’s 2000 proposal to ban most nonaudit services. The profession claimed that there is no “smoking gun” evidence that the provision of NAS to audit clients has ever resulted in a single audit failure. However, as the Panel on Audit Effectiveness (2000, 110) observes, there is unlikely to ever be such a “smoking gun” because it is the unconscious loss of objectivity by an auditor that is more likely to be the underlying problem with NAS, and this subjective dimension cannot be readily observed with real-world data. For this reason, I believe there is a need for new and rigorous experimental studies that probe these more subjective dimensions of auditor behavior that are not easily detected using archival data. Despite reservations about the ability to detect the subjective impact of NAS on auditor behavior using real-world archival data, there is a recent stream of CAR Vol. 23 No. 3 (Fall 2006)

752

Contemporary Accounting Research

empirical research that does attempt to evaluate whether NAS are associated with observable differences in the quality of reported earnings. Much of this literature is reviewed by RTT, so I will be brief here. While this literature is important, it does not — and cannot — determine whether auditor independence is factually impaired by the presence of NAS. Instead, what these studies document is the extent to which there is measurable variation in earnings quality attributes that is statistically associated with the level of NAS fees. The most prominent of these studies is by Frankel, Johnson, and Nelson 2002, a study that was cited (as a working paper) during the congressional hearings leading up to the passage of the SarbanesOxley Act in 2002. Frankel et al. (2002) document two primary findings: companies that pay their auditors relatively more for NAS (a) have larger abnormal accruals and (b) are more likely to meet or beat analysts’ earnings forecasts. Together, these two results are consistent with the possibility of more aggressive use of a firm’s discretion over accounting accruals to manage its earnings in order to meet benchmark earnings targets, with the implication being that auditors are compromised as NAS fees increase and thus are more likely to go along with such opportunistic behavior by their clients. While important, this type of research cannot determine whether generally accepted accounting principles (GAAP) are actually violated; rather, the research can simply document the extent to which earnings attributes vary cross-sectionally as a function of auditor-provided NAS. Levitt (1998) argues that earningsmanagement games may, at the limit, constitute materially misleading financial reporting and therefore may violate securities laws. However, from a legal viewpoint, this argument remains only conjectural at this time. The results in Frankel et al. 2002 have been reexamined in several subsequent studies, most of which are cited in RTT. Collectively, these studies show that the results in Frankel et al. are fragile and sensitive to sample selection and model specification. At a minimum, these subsequent studies demonstrate that Frankel et al.’s results are not robust and therefore cannot be relied upon as the sole basis for policy making with respect to such an important decision as restricting the scope of economic activity between auditors and their clients. For the accounting research community, there is an even more important lesson. Replication is at the heart of science, and a single study proves little by itself without independent verification by other studies. Yet accounting scholarship has moved away from the publication of replications. One cannot help but wonder how many other fragile results exist in the published research literature that might not hold up to the kind of rigorous replication undertaken with respect to Frankel et al. 2002. Summing up, then, what do we know about the effects of NAS on auditor independence from the research literature? The perception research gives reason to believe that nonaudit service activities may create at least the perception of diminished auditor independence and objectivity. However, there is no smoking-gun evidence linking NAS with audit failures. There is no compelling body of evidence that NAS impair audit quality that allows us to empirically observe the impairment of audit quality through proxies such as earnings attributes, nor other empirical approaches such as the likelihood of an auditor issuing a going-concern report CAR Vol. 23 No. 3 (Fall 2006)

Are Auditors Compromised by Nonaudit Services?

753

(DeFond, Raghunandan, and Subramanyam 2002) or the likelihood of having earnings restated (Kinney, Palmrose, and Scholz 2004) when NAS fees are high. On the other hand, given the inherent difficulty of empirically observing the impairment of audit quality, the failure to find smoking-gun evidence should not give us too much comfort that NAS are unproblematic for the accounting profession. Despite this caution, empirical research does show that firms hiring their auditors for NAS consistently have higher audit fees (Simunic 1984; Francis, Reichelt, and Wang 2005), and that audit effort is higher (more engagement hours are invoiced) when clients purchase NAS (Davis, Ricchiute, and Trompeter 1993). In other words, firms pay their auditors more for audits and auditors expend more effort when NAS are performed. To the extent that higher fees and more effort imply higher-quality audits, the empirical evidence supports the profession’s claim that the provision of NAS may actually improve audit quality (e.g., AICPA 1997; Panel on Audit Effectiveness 2000, 127–32). So, in spite of the rather gloomy picture in Moore et al. 2006 about the possibility of independent audits, the scientific evidence does not support the conclusion that the factual impairment of auditor independence occurs when NAS are performed for audit clients. However, evidence accumulated over the past 40 years does show that the provision of NAS to audit clients is widely perceived as potentially undermining the auditor’s ability to be independent and objective. The unanswered question is whether there is simply a negative perception but not a “real” factual independence problem with respect to NAS, or if auditor behavior really is compromised by the provision of NAS to audit clients. Given the inherent limitations of archival-based research in identifying the kinds of subjective effects/ biases that NAS are alleged to engender, I think the answer to this question can only come from carefully designed experimental studies. More generally, I agree with Nelson 2006 that experimental research is needed to investigate the broader claims of Moore et al. 2006 that the institutional structure of auditing (including NAS) is deeply problematic in terms of inhibiting unbiased and objective audits, and that research is needed to examine the effectiveness of recent reforms in mitigating these potential threats to auditor independence. 3. Comments on Ruddock, Taylor, and Taylor RTT is a competent and carefully done study, and my comments focus on three aspects of their research design. First, I discuss institutional differences between the United States and Australia that may limit the extrapolation of the study to a U.S. setting. Second, I discuss some general reservations about the earnings conservatism framework used in the study. Third, I discuss some issues regarding the definition and measurement of nonaudit fees in the study. There are important differences between Australia and the United States that may limit the relevance of RTT with respect to the U.S. audit market, or the generality of their results beyond Australia. This is a particularly important issue because this is basically a “no results” paper, and also because the authors do make rather strong claims about U.S. audit-market regulation based on their Australian data. CAR Vol. 23 No. 3 (Fall 2006)

754

Contemporary Accounting Research

The audit market in Australia is qualitatively different from that in the United States, as are auditor incentives. With regard to the market, the data come from 1990s when there were six or five dominant accounting firms.5 In Australia the Big 6/5 audit about 60 percent of listed companies, compared with the United States, where the same accounting firms audit about 85 percent of the COMPUSTAT population of U.S. companies.6 Importantly, the level of nonaudit fees in Australia is much lower than that in the United States. RTT report a median ratio of nonaudit fee to total fees (RNAS ) of 27 percent. In the United States, nonaudit fees are nearly double this amount for the 2000 fiscal year, and Frankel et al. (2002) report a ratio of 51 percent for their sample. So it is entirely possible that the “no results” finding in RTT is simply the consequence of far fewer consulting dollars on the table and therefore less temptation to favor clients paying high fees. The regulatory and litigation environment in Australia might also explain why there are no significant results. There is less scrutiny of auditors by regulators in Australia (the Australian Securities and Investments Commission [ASIC]) than in the United States (the SEC), and there is nothing comparable in Australia to Accounting and Auditing Enforcement Releases (AAERs), in which auditors are routinely punished and sanctioned by the SEC. With regard to litigation, the level in Australia is nowhere near the level in the United States. If auditor reputations are less likely to be tarnished in Australia by regulators or by litigation, as appears to be the case, there is also less incentive for auditors to perform high-quality audits in order to protect their reputations. What this means is that audit quality is more likely to be uniform (and lower) in Australia than in the United States, and that U.S. auditors, with greater reputation capital, have an incentive to be more careful and to report more conservatively in order to avoid adverse reputation effects from regulatory scrutiny or litigation (Francis and Krishnan 1999). Khurana and Raman (2004) report evidence linking auditor reputation, litigation risk, and cost of capital that is consistent with this conjecture. Specifically, they find that firms with Big 4 auditors have a lower cost of capital than firms with non – Big 4 auditors in the United States, but not in other common-law countries with less litigation risk (that is, Australia, Canada, and the United Kingdom). They interpret this result as suggesting that litigation risk drives audit-quality differences in the United States, whereas the absence of cost-of-capital differences in the other countries implies that there are no differences in audit quality. My point is that the “no results” in RTT may simply reflect the lack of meaningful differential audit quality in Australia due to the underlying institutional incentives, in which case it is not surprising to observe no cross-sectional differences in the degree of earnings conservatism conditional on auditor type. Next I want to talk briefly about the earnings conservatism (EC) framework. The premise of EC is asymmetric recognition of gains and losses: firm-specific bad news is recognized more quickly in accounting earnings than is good news. The paper uses two tests from Basu’s 1997 seminal paper on EC. The level of nonaudit fees is used to proxy for auditor quality, and the premise is that if NAS are “bad” and impair audit quality, then we should see less EC when there are high levels of nonaudit service fees. Since accounting rules with respect to accounting conservaCAR Vol. 23 No. 3 (Fall 2006)

Are Auditors Compromised by Nonaudit Services?

755

tism are “fixed” at any point in time within a country, if we observe cross-sectional differences in conservatism, then such differences are assumed to be driven by the auditor’s enforcement of EC rules, which of course implies differential audit quality. Regarding the EC story, you’re either a believer or a skeptic, and I admit to being a skeptic. Serious econometric and other validity problems have been identified with respect to EC tests based on the so-called reverse regression of earnings on returns (Dietrich, Muller, and Riedl 2005; Givoly, Hayn, and Natarajan 2005). However, RTT also use Basu’s 1997 time-series model of earnings changes, and these results are consistent with the reverse regressions, so this does mitigate concerns over use of the reverse regression design.7 However, my concern with the EC framework is more fundamental. Specifically, I question the premise that “bad news” is incorporated into accounting earnings by GAAP more quickly than “good news”. In other words, the asymmetry in losses versus gains recognition may be overstated. Yes, there are specific loss recognition rules relating to inventory (lower-of-cost-or-market rule) and asset impairments (for example, SFAS No. 142 and SFAS No. 144) that are consistent with the EC story. But there are not many other specific EC rules one can point to. A secondary question thus becomes how many of these asset write-downs actually occur, and how timely are these writedowns? Anecdotally, one has the impression that firms often delay asset write-downs, which undermines the timely recognition story. On the gain side, it turns out that there are rules that do require the timely recognition of unrealized gains; for example, there are a number of areas in which mark-to-market accounting for financial instruments is used in which unrealized gains are given timely recognition. One could even make a counterclaim that the very foundation of accrual-based accounting is inherently anti-conservative. The reason is that unrealized revenues are accrued before cash flows are received, and expense recognition is deferred with respect to asset-related cash expenditures through yearly depreciation/ amortization charges. Both of these practices are central features of accrual accounting and are fundamentally unconservative relative to a cash basis of accounting that defers revenue recognition until cash is collected and writes off expenses immediately when cash is spent. So I believe it is an open question whether earnings conservatism and asymmetric loss recognition are inherent within GAAP. In addition to overstating the degree of asymmetry in GAAP with respect to loss / gain recognition, there is a further aspect of EC research that is unsettling. The EC testing framework does not explicitly document the magnitude of earnings conservatism, nor does it show the precise channel or mechanism through which earnings conservatism occurs. In contrast, while the abnormal accruals literature has been heavily criticized, it does nevertheless attempt to show how discretion over accruals is used to achieve strategic reporting objectives and to measure the dollar magnitude of this effect through abnormal (unexpected) accruals. So the mechanism through which earnings are affected is explicitly analyzed in accruals research, while the mechanism remains a black box in EC research. Finally, I have some reservations over the measurement of nonaudit fees in RTT. Their primary test variable is derived from unexpected fees based on the residuals of a first-stage model of expected NAS fees. Nonaudit services are voluntary in CAR Vol. 23 No. 3 (Fall 2006)

756

Contemporary Accounting Research

nature and in fact are zero for many firms. The NAS models do not have the same degree of explanatory power as do audit fee models, so low statistical power is a potential concern, particularly in a “no results” study. Second, if NAS create an economic bond, it seems more likely that the absolute fee level is what would matter, rather than “unexpected” high or low fee levels relative to a predictive model. In addition, RTT arbitrarily drop the middle third of the sample and code firms as “high” NAS firms if the first-stage residuals are in the upper third, and as “low” NAS firms if the residuals are in the lower third. RTT (see note 24) test an alternative specification based on the ratio of nonaudit fees to total fees that uses the full sample. However, this measure still does not capture the absolute dollar magnitude of nonaudit fees. I would also like to see a test using a continuous measure of the dollar magnitude of NAS fees, because it is possible that auditor independence is compromised only when fees reach some threshold level. Additional sensitivities along these lines would give the reader more confidence in the study.8 As I also suggested earlier, it is not clear why NAS should be compartmentalized and treated in isolation from other factors, such as the magnitude of audit fees, that may affect auditor incentives. Perhaps it is the total fee dependence that matters, not just the marginal effect of NAS. This also opens up the broader question of the appropriate unit of analysis in examining auditor incentives. RTT classify firms as “high” and “low” NAS firms based on a fee model using the entire sample of companies. However, auditor incentives are more likely to be affected by the client’s relative importance to the auditor, which suggests that total fees may be a better gauge of the economic bond between a client and the auditor. In addition, the relative importance of a single client may be contextualized in different ways: relative to the accounting firm’s entire clientele; relative to the engagement office administering the audit (Reynolds and Francis 2000); or even the importance of a client to the specific engagement partner that signs the audit report, which is publicly disclosed information in Australia. While there is a practical limit to how much can be done in a single study, there is other evidence in the literature to suggest that officelevel clienteles or even partner-level clienteles may be a better unit of analysis with respect to auditor incentives, and perhaps future research can explore these issues. 4. Conclusions RTT use an earnings conservatism framework to investigate the effects of high versus low levels of NAS on earnings conservatism, and to test whether audit quality was impaired by NAS in Australia during the 1990s. They find no evidence of differential conservatism conditional on the level of NAS fees paid to auditors. While there are institutional reasons that may limit the ability to extrapolate their results to the U.S. regulatory context (including the fact that Australian NAS fees are about half the level of U.S. NAS fees), the study adds to a growing body of empirical evidence questioning whether there is a legitimate reason to restrict the scope of services that auditors provide to their clients. From a review of the NAS research literature, there is no direct evidence that auditor independence is compromised by the provision of NAS to audit clients. This is hardly surprising, because the profession itself has used the absence of CAR Vol. 23 No. 3 (Fall 2006)

Are Auditors Compromised by Nonaudit Services?

757

smoking-gun evidence to argue against restrictions on the scope of nonaudit services (AICPA 1997). The closest thing to a smoking gun might be the recent study by Frankel et al. 2002, who report evidence of more aggressive accounting with respect to accruals when auditors provide higher levels of NAS. However, this kind of archival research cannot determine whether aggressive accounting used with respect to accruals is actually in breach of GAAP. In addition, several follow-up studies either are unable to replicate the results in Frankel et al. 2002 or show that their results are not robust to alternative samples and model specifications. The bottom line is that the profession is justified in saying there is no evidence linking NAS with audit failures. However, this may be of little comfort, either to the profession or to regulators. A body of research spanning 40 years has consistently shown that there is a widely held perception among users of financial statements that NAS have the potential to compromise auditor independence. Recent research further suggests that such negative perceptions pose more than a public relations problem for auditors and their clients, and that there are real economic consequences associated with NAS in terms of lower stock prices for companies that pay their auditors high levels of fees for NAS. Endnotes 1. On the basis of U.S. fee data disclosures beginning with the 2000 annual reports, the median ratio of nonaudit fees to total fees for audit clients has declined steadily from 2000 to the present. Data currently available in Audit Analytics show that the median ratio of nonaudit fees to total fees declined monotonically from 0.48 in 2000 to 0.43 in 2001, to 0.31 in 2002, to 0.25 in 2003, to 0.20 in 2004, and to 0.18 in 2005. With the prohibition of some types of NAS by the SEC and the Sarbanes-Oxley Act, the percentage of NAS represented by taxation services has also grown steadily over time. In 2000, taxation services represented on average 64 percent of total nonaudit services, but by 2005 they represented 70 percent. 2. The arguments in Moore et al. 2006 imply that complete auditor independence/ objectivity is the desired goal. It is likely that the costs of attaining such a goal far outweigh the benefits in terms of improved audit quality. For example, governmentappointed auditors who serve a nonrenewable one-year term with fees set by a government agency would avoid moral seduction and achieve complete auditor independence in the Moore et al. framework. However, audit costs would be substantially higher due to start-up costs, and because of learning-curve effects auditors would be more likely to make errors. The current institutional arrangements are not perfect, but it is entirely possible that they are optimal, even though complete independence/objectivity does not occur. 3. The insignificance of NAS is also supported in experimental economics research by Dopuch and King 1991, who create an audit market and find that the provision of NAS does not seem to adversely affect audit quality. 4. Accounting firms could still sell nonaudit services to other (nonaudit) clients. The main negative consequence would be potential lost economies of scope in cross-selling a bundle of audit and other services. However, the existence of such economies is dubious given the specialization and division of labor in large accounting firms, where CAR Vol. 23 No. 3 (Fall 2006)

758

5.

6.

7.

8.

Contemporary Accounting Research such services are typically performed by separate groups that have little or no interaction with one another. The data in the study also preceded the accounting scandals in the United States in 2001 and 2002 relating to Enron and WorldCom, and the HIH failure in Australia in 2001. There is little doubt that these corporate failures tarnished auditor reputations, at least in the short run (Chaney and Philipich 2002). Another importance institutional difference is that mining companies comprise nearly half of listed companies in Australia. It’s unclear what exactly this might imply for the study, but it is an important feature of the Australian audit market. In addition, they use a test from Ball and Shivakumar 2005 that examines the contemporaneous association of accruals and operating cash flows as an alternative test of conservatism. This test also fails to find differences in conservatism as a function of NAS fee levels. There has been a tendency to study NAS using linear models in which a response variable is tested as a linear function of increasing levels of NAS fees. However, it may be that NAS impair auditor objectivity only at very high or extreme levels. At a minimum, future research should explore the sensitivity of this issue, especially if there are not significant results using a linear specification of NAS fee levels.

References American Institute of Certified Public Accountants (AICPA). 1997. Serving the public interest: A new conceptual framework for auditor independence. New York: AICPA. Ball, R., and L. Shivakumar. 2005. Earnings quality in UK private firms: Comparative loss recognition timeliness. Journal of Accounting and Economics 39 (1): 83–128. Basu, S. 1997. The conservatism principle and the asymmetric timeliness of earnings. Journal of Accounting and Economics 24 (1): 3–37. Bazerman, M., K. Morgan, and G. Lowenstein. 1997. The impossibility of auditor independence. Sloan Management Review 38 (4): 89–94. Cannon, A. 1952. Tax pressures on accounting principles and accountants’ independence. The Accounting Review 27 (4): 419–26. Carey, J. L. 1970. The rise of the accounting profession: To responsibility and authority, 1937 – 1969. New York: American Institute of Certified Public Accountants. Chaney, P., and K. Philipich. 2002. Shredded reputation: The cost of audit failure. Journal of Accounting Research 40 (4): 1221–45. Commission on Auditors’ Responsibilities. 1978. Report, conclusions, and recommendations. New York: American Institute of Certified Public Accountants. Davis, L., D. Ricchiute, and G. Trompeter. 1993. Audit effort, audit fees, and the provision of nonaudit services. The Accounting Review 68 (1): 135–50. DeFond, M. L., K. Raghunandan, and K. R. Subramanyam. 2002. Do non-audit service fees impair auditor independence? Evidence from going concern audit opinions. Journal of Accounting Research 40 (4): 1247–74. Dietrich, J. R., K. A. Muller, and E. J. Riedl. 2005. Asymmetric timeliness tests of accounting conservatism. Working paper, Ohio State University. Dopuch, N., and R. King. 1991. The impact of MAS on auditors’ independence: An experimental markets study. Journal of Accounting Research 29 (Supplement): 60–98. CAR Vol. 23 No. 3 (Fall 2006)

Are Auditors Compromised by Nonaudit Services?

759

Earnscliffe Research & Communications. 1999. Report to the United States Independence Standards Board: Research into perceptions of auditor independence and objectivity. (Cited in Panel on Audit Effectiveness 2000, 113.) Financial Accounting Standards Board (FASB). 2001a. Statement of Financial Accounting Standards No. 142: Goodwill and other intangible assets. Norwalk, CT: FASB. Financial Accounting Standards Board (FASB). 2001b. Statement of Financial Accounting Standards No. 144: Accounting for the impairment or disposal of long-lived assets. Norwalk, CT: FASB. Francis, J. 2004. What do we know about audit quality? British Accounting Review 36 (4): 345 – 68. Francis, J. R., and B. Ke. 2006. Disclosure of fees paid to auditors and the market valuation of earnings surprises. Review of Accounting Studies 11 (4) (forthcoming). Francis, J., and J. Krishnan. 1999. Accounting accruals and auditor reporting conservatism. Contemporary Accounting Research 16 (1): 135–65. Francis, J., K. Reichelt, and D. Wang. 2005. The pricing of national and city-specific reputations for industry expertise in the U.S. audit market. The Accounting Review 80 (1): 113–36. Frankel, R. M., M. F. Johnson, and K. K. Nelson. 2002. Auditor independence and earnings quality. The Accounting Review 77 (Supplement): 71–105. Givoly, D., C. Hayn, and A. Natarajan. 2005. Measuring reporting conservatism. Working paper, Pennsylvania State University. Hartley, A., and T. Ross. 1972. MAS and audit independence: An image problem. Journal of Accountancy (November): 42–51. Khurana, I., and K. Raman. 2004. Litigation risk and the financial reporting credibility of Big 4 versus non-Big 4 audits: Evidence from Anglo-American countries. The Accounting Review 79 (3): 473–95. King, R. 2002. An experimental investigation of self-serving biases in an auditing trust game: The effect of group affiliation. The Accounting Review 77 (2): 265–84. Kinney, W., Z. Palmrose, and S. Scholz. 2004, Auditor independence, non-audit services and earnings restatements: Was the U.S. government right? Journal of Accounting Research 42 (3): 561–88. Krishnan, J., S. Heibatollah, and Y. Zhang. 2005. Does the provision of nonaudit services affect investor perceptions of auditor independence? Auditing: A Journal of Practice & Theory 24 (2): 111–35. Lavin, D. 1976. Perceptions of auditor independence. The Accounting Review 51 (1): 41–51. Levitt, A. 1998. The numbers game. Remarks delivered at the NYU Center for Law and Business, New York, September 28. Available online at http://www.aaahq.org/newsarc/ pr101898.htm. McKinley, S., K. Pany, and P. Reckers. 1985. An examination of the influence of CPA firm type, size, and MAS provision on loan officer decisions and perceptions. Journal of Accounting Research 23 (2): 887–96. Moore, D., P. Tetlock, L. Tanlu, and M. Bazerman. 2006. Conflicts of interest and the case of auditor independence: Moral seduction and strategic issue cycling. Academy of Management Review 31 (1): 10–29.

CAR Vol. 23 No. 3 (Fall 2006)

760

Contemporary Accounting Research

Nelson, M. 2006. Ameliorating conflicts of interest in auditing: effects of recent reforms on auditors and their clients. Academy of Management Review 31 (1): 30–42. Panel on Audit Effectiveness. 2000. The Panel on Audit Effectiveness: Report and recommendations. New York: Public Oversight Board of the American Institute of Certified Public Accountants. Pany, K., and P. Reckers. 1984. Non-audit services and auditor independence — A continuing problem. Auditing: A Journal of Practice & Theory 3 (2): 89–97. Pany, K., and P. Reckers. 1987. Within- vs. between-subjects experimental designs: A study of demand effects. Auditing: A Journal of Practice & Theory 7 (1): 39–53. Reynolds, J. K., and J. Francis. 2000. Does size matter? The influence of large clients on office-level auditor reporting decisions. Journal of Accounting and Economics 30 (3): 375–400. Ruddock, C., S. J. Taylor, and S. L. Taylor. 2006. Nonaudit services and earnings conservatism: Is auditor independence impaired? Contemporary Accounting Research 23 (3): 701–46. Sarbanes-Oxley Act of 2002. 2002. Pub. L. no. 107–204. Schulte, A. 1965. Compatability of management consulting and auditing. The Accounting Review 40 (3): 587–93. Securities and Exchange Commission (SEC). 2000a. Proposed rule: Revision of the commission’s auditor independence requirements. 17 CFR parts 210 and 240. Release nos. 33-7870; 34-42994; 35-27193; IC-24549; IA-1884; file no. S7-13-00. Washington, DC: SEC. Securities and Exchange Commission (SEC). 2000b. Revision of the commission’s auditor independence requirements. Release nos. 33-7919; 34-43602; file no. S7-13-00. Washington, DC: SEC. Shockley, R. A. 1981. Perceptions of auditors’ independence: An empirical analysis. The Accounting Review 56 (4): 785–800. Simunic, D. 1984. Auditing, consulting, and auditor independence. Journal of Accounting Research 22 (2): 670–702. Titard, P. 1971. Independence and MAS — Opinions of financial statement users. Journal of Accountancy (July): 47–52.

CAR Vol. 23 No. 3 (Fall 2006)