francis - 2004 - what do we know about audit quality [aq]

What do we know about audit quality? * Jere R. Francis * University of Missouri—Columbia, 432 Cornell Hall, Columbia, MO 65211, USA University of Melbourne, Victoria, Australia Abstract This paper reviews empirical research over the past 25 years, mainly from the United States, in order to assess what we currently know about audit quality with respect to publicly listed companies. The evidence indicates that outright audit failure rates are infrequent, far less than 1% annually, and audit fees are quite small, less than 0.1% of aggregate client sales. This suggests there may be an acceptable level of audit quality at a relatively low cost. There is also evidence of voluntary differential audit quality (above the legal minimum) along a number of dimensions such as firm size, industry specialization, office characteristics, and cross-country differences in legal systems and auditor liability exposure. The evidence is very positive although there is some indication that audit quality may have declined in the 1990s, in which case there could be merit in recent reforms such as the Sarbanes-Oxley Act of 2002 in the US. However, we do not know from research the optimal level of audit quality and therefore whether we currently have ‘too little’ or ‘too much’ auditing? Despite this lacuna we are entering an era of more mandated auditing in response to high-profile corporate governance failures including the Enron–Andersen affair. Finally, while recent reforms have scaled back the scope of non-audit services due to independence concerns, a case can be made that audit quality will always be somewhat suspect if other services are provided that are perceived to potentially compromise the auditor’s objectivity and skepticism. For this reason public confidence in audit quality may be increased by proscribing all non-audit services for audit clients. Recommendations are also proposed with respect to legal liability reform and changes in partner compensation arrangements. q 2004 Elsevier Ltd. All rights reserved. Keywords: Audit quality; Sarbanes-Oxley Act; Audit failure rates 0890-8389/$ - see front matter q 2004 Elsevier Ltd. All rights reserved. doi:10.1016/j.bar.2004.09.003 The British Accounting Review 36 (2004) 345–368 www.elsevier.com/locate/bar * I appreciate the helpful comments of an anonymous referee and colleagues at University of Missouri— Columbia and University of Melbourne. The paper is based on the author’s plenary lecture at the 2004 annual meeting of the British Accounting Association, York University. * Tel.: C1 573 882 5156; fax: C1 573 882 2437. E-mail address: francis@missouri.edu. This paper reviews research on audit quality from the past 25 years, with a particular focus on empirical research from the United States. In the aftermath of the Enron bankruptcy in 2001 and the related collapse of Arthur Andersen in 2002, it has become fashionable to criticize auditing and to question the quality of audits being performed by accounting firms, especially the large international Big 4 accounting firms. 1 Indeed this criticism motivated recent regulatory changes in the United States brought about by the Sarbanes-Oxley Act of 2002 (Public Law No. 107-204) in which self-regulation by the accounting profession has been replaced with direct regulation by a new independent agency, the Public Company Accounting Oversight Board. My emphasis is on publicly listed companies because the separation of ownership and management control in listed companies makes the independent external audit especially important with respect to corporate governance and the oversight of such companies. The review is not meant to be comprehensive and encyclopedic but is instead a more selective survey whose purpose is to identify and assess a wide range of evidence on audit quality from academic research. It turns out that we know quite a bit, and despite Enron and other high-profile failures, the evidence indicates that the general level of audit quality is satisfactory with very few outright audit failures, although there is some evidence that both earnings quality and audit quality may have declined during the 1990s. 1. What are outright audit failure rates? Audit quality can be conceptualized as a theoretical continuum ranging from very low to very high audit quality. Audit failures obviously occur on the lower end of the quality continuum, and so a good starting point in thinking about audit quality is to ask what the rate of outright audit failure is? An audit failure occurs in two circumstances: when generally accepted accounting principles are not enforced by the auditor (GAAP failure); and when an auditor fails to issue a modified or qualified audit report in the appropriate circumstances (audit report failure). In both cases, the audited financial statements are potentially misleading to users. As a first approximation of audit quality, we can think of audits as either meeting or not meeting minimum legal and professional requirements. Audit quality is inversely related to audit failures: the higher the failure rate, the lower the quality of auditing. What do we know about audit failure rates? Outright audit failures are difficult to determine with certainty but can be inferred from several sources including auditor litigation and business 1 The studies cited in this review date to earlier periods when the dominant group was the Big 8 prior to 1989, the Big 6 from 1989–1997, Big 5 from 1998 to 2001, and the Big 4 since the collapse of Arthur Andersen in 2002. For convenience the term ‘Big 4’ will generally be used through the paper to refer to all of these groupings. The Big 8 firms were Arthur Andersen, Arthur Young, Coopers and Lybrand, Deloitte Haskins and Sells, Ernst and Whinney, KPMG, Price Waterhouse, and Touche Ross. The Big 6 came into being in 1989 when Ernst and Whinney merged with Arthur Young in the US to become Ernst and and Young, and Deloitte Haskins and Sells merged with Touche Ross to become Deloitte Touche. The Big 5 came into being in 1997 when Coopers and Lybrand merged with Price Waterhouse in the US to become Pricewaterhousecoopers. Note that in some countries these mergers played out differently in terms of who merged with whom. J.R. Francis / The British Accounting Review 36 (2004) 345–368346 failures, investigations by the Securities and Exchange Commission (SEC), and earnings restatements. Each of these is now reviewed. Arguably the most convincing evidence of an outright audit failure occurs when there is litigation against auditors (Palmrose, 1988). It turns out that the number of lawsuits against auditors in the United States is small, despite the often-heard claim about rampant litigation (Andersen et al., 1992). 2 Using a comprehensive dataset, Palmrose (2000) documents around 1000 lawsuits against the large national accounting firms over the period 1960–1995, or an average of only 28 lawsuits per year. Given a population of around 10,000 publicly listed companies in the United States, 28 lawsuits per year imply an annual audit failure rate of 0.28%, i.e. 28 hundredths of one per cent. The number of successful lawsuits is even smaller and is generally estimated to be around 50% of total lawsuits after excluding cases successfully defended by accounting firms and non- meritorious cases dismissed by courts (Palmrose, 1997). The bottom line is that the number of proven audit failures is so small as to approach a rate of zero, and so it is difficult to imagine what could be done to change auditing practices or the regulatory environment that would result in a significantly lower audit failure rate. A broader definition of auditor failure could be based on business failure rates. It turns out that failures of publicly listed companies in the United States are also small in number, averaging around 40 per year (Francis and Krishnan, 2002). However, it may be wrong to presume an audit failure has occurred just because a business failure occurs, and this is borne out by the fact that auditors of bankrupt companies are sued only 25% of the time (Palmrose, 1987). 3 Regardless, the audit failure rate implied by bankruptcy rates is very small and comparable to the low failure rate based on auditor litigation. Since 1982, the SEC has issued Accounting and Auditing Enforcement Releases (AAERs) which report outcomes of its investigations against companies and auditors. AAERs are the result of consent decrees in which companies and/or their auditors do not formally plead guilty to misdeeds, but instead accept an administrative action such as a fine, or other reprimand, and agree that they will not engage in the kind of behavior described in the AAERs which the SEC deems unacceptable (Feroz et al., 1991). Arguably, these consent decrees could be interpreted as evidence of ‘audit failures’ even though they are not legally described as such. Dechow et al. (1996) examine the first 436 AAERs issued from 1982 to 1993. Over this 12-year period there were 165 actions describing auditor deficiencies representing about 14 actions per year. Given the US population of approximately 10,000 publicly listed companies which are subject to SEC regulations, 14 yearly actions against auditors is an annual audit failure rate of 0.14% (i.e. 14 hundredths of one per cent) from 1982 to 1993. More recently, 1485 new AAERs have been issued from 1994 to 2003, or an average of 149 per year, and this may reflect an 2 The concern for litigation is not so much the frequency of lawsuits, but the potential magnitudes of lawsuits and the possibility that one or a few large lawsuits have the potential to bankrupt an accounting firm as happened in 1991 with Laventhol Horwarth, the 7th largest US accounting firm at the time. 3 An auditor is responsible for issuing a going concern report if the auditor believes a company may not survive 12 months from the balance sheet date. However, the auditor is not responsible for predicting bankruptcy per se, and it is possible for companies to fail for reasons an auditor could not have reasonably anticipated 12 months in advance. J.R. Francis / The British Accounting Review 36 (2004) 345–368 347 increase in questionable financial reporting practices during the 1990s. While there has been no formal study (to date) of these AAERs actions, my informal review based on a random sample indicates that a majority of these AAERS are not directed toward auditors. The bottom line is that the annual audit failure rate implied by AAERs is well under 1% even since the mid-1990s when the number of AAERs began to increase. Another potential source of audit failure data comes from earnings restatements filed with the SEC and recently summarized in a report by the Government Accounting Office (GAO, 2003). Arthur Levitt, former SEC Chairman, was concerned by the increase in earnings restatements in the late 1990s, and was convinced they indicated past accounting (and auditing) failures due to the kind of aggressive earnings management he warned of in a 1998 speech entitled ‘The Numbers Game’ (Levitt, 1998). GAO (2003) covers earnings restatements from 1997–2002, and restatements increased monotonically over the period from 92 (1997) to 250 (2002), with a sharp increase in 1999 coinciding with Chairman Levitt’s concerns. However, my informal review of these restatements indicates a majority are straightforward adjustments of accounting ‘estimations’ in prior-year financial statements and therefore (arguably) not audit failures. Indeed in about one-fourth of the cases the auditor identified the problem and initiated the re-statement, and in the majority of cases there was no related SEC action against companies or litigation against auditors, which means that most earnings restatements are not interpretable as audit failures. In sum, the ex post evidence of audit failures from SEC sanctions, litigation rates, business failures and earnings restatements all point to a very low failure rate, much less than one per cent annually. This may understate the true failure rate if there are undetected audit failures. However, if an audit failure goes undetected, it most likely means that the company and its stakeholders have not suffered adverse economic effects such as bankruptcy or financial distress arising from the audit failure. What we can more properly conclude is that known audit failures with material consequences are relatively infrequent. While there may be other audit failures with lesser consequences, we have no evidence on the rate of these occurrences. 2. How costly is auditing? The evidence reviewed so far indicates that outright audit failures with material consequences are very infrequent, fewer than 100 per year in a population of over 10,000 listed companies. However, in evaluating audit quality it is important to assess both the benefits and costs of auditing. For example, while audit failure rates are low, if audit fees are large it is possible that too much investment is being made in audit quality relative to the benefits achieved. Audit costs are analyzed using 2002–2003 audit fee disclosure data in the United States. I made the following calculations for 5500 large US publicly listed companies, having aggregate sales of $8177 billion and aggregate market value of $9912 billion (note all monetary amounts are in US dollars). Aggregate audit fees for these 5500 companies totalled $3.4 billion which means audit fees represented only 0.04% of sales and 0.03% of market value. I also analyzed audit fees for each decile of companies from the smallest decile of firm size to the largest decile. As expected average fees as J.R. Francis / The British Accounting Review 36 (2004) 345–368348 a percentage of sales decrease as firm size becomes larger. For the smallest decile, audit fees average 2% of sales, but for the largest decile audit fees average less than 1/100 of one per cent of sales. So auditing appears to be a relative ‘bargain’ in the sense that audits cost a relatively small fraction of client sales. However, the low cost of auditing does not necessarily mean that audit quality is low since the outright audit failure rate is also quite low. While this data may give some comfort that the social benefit of auditing (as suggested by low failure rates) is achieved at a reasonable social cost, this kind of cost–benefit analysis does not tell us anything about audit quality for the vast majority of companies which have ‘legally’ satisfactory audits. 4 Remember, audit quality is conceptualized as a continuum from very low to very high quality, and outright failures occur on the extreme low end of quality. The remainder of this paper surveys what we know about audit quality over the remainder of the quality continuum and focuses primarily on the two empirical observables in auditing: (1) auditor–client alignments, or who audits whom; and (2) audit outcomes which include audit reports and the audited financial statements which are a joint outcome of auditor– client negotiations (Antle and Nalebuff, 1991). 3. What do we know from audit report research? Since 1989 there are effectively two types of audit reports issued in the United States: the standard clean unmodified report and a modified report for going concern uncertainty. 5 Butler et al. (2004) calculate that 6.6% of US listed companies received a going concern reports during the period 1994–1999. Another study documents yearly going concern reporting rates ranging from 9 to 5%, with a monotonic decline over the period 1990–1997 (Francis and Krishnan, 2002). Even though modified going concern reports represent less than 10% of total audit reports, one way of determining if audits are of high quality is to determine if investors respond to going concern reports in a manner consistent with such reports conveying ‘bad news’ about the client. If auditing is of high quality then going concern audit reports should convey useful information; however, if auditing is of low quality, then modified audit reports would have little or no informational value to investors. Before examining the empirical evidence, it is useful to first consider ‘misreporting’ rates. Misreporting comes about from false negatives (a type 2 statistical error), i.e. issuing a clean audit report when in fact a going concern report was appropriate, and from false positives (a type 1 statistical error), i.e. issuing a going concern report for firms that do not subsequently fail or otherwise become financially distressed. Both false negatives and false positives are 4 A referee points out that individual accounting firms may be highly profitable and may earn economic rents due to the accounting profession’s exclusive monopoly over the supply of audits. From this viewpoint it is possible that audits currently cost more than they would in a more competitive market setting for the existing level of audit quality. However, absent firm-level production cost data it is not possible to know if this is the case. 5 Qualified and adverse opinions cannot be issued for SEC registrants, and disclaimers are rarely issued. There are other kinds of ‘technical’ modified opinions such as a change in accounting methods. However, the essential reporting choice is now between a standard clean opinion and a going concern report. J.R. Francis / The British Accounting Review 36 (2004) 345–368 349 (arguably) reporting failures in the sense that the right audit report was not issued which reduces the informational value of audit reports due to noise. 3.1. False negatives and false positives in audit reporting With regard to false negatives, Carcello and Palmrose (1994) find that only 30% of bankruptcies are preceded by a going concern audit report. In other words, 7 in 10 bankruptcies have a ‘false negative’ or clean audit report. Of course, as already noted, there are only around 40 bankruptcies of publicly listed companies per year, so the absolute number of false negatives per year is quite small. Strictly speaking, bankruptcies not preceded by going concern audit reports are not necessarily audit failures since the objective of an audit is not the prediction of bankruptcy. However these cases are widely viewed as audit failures and for this reason false negatives (clean reports) for bankrupt companies can create potentially significant litigation risk for auditors. Carcello and Palmrose (1994) document that auditors not issuing a going concern report before bankruptcy are sued twice as often (64 versus 36%), have lower lawsuit dismissal rates, and higher resolution payments (around $10 million versus $1 million). Settlements can also increase the cost of future insurance premiums. There may also be a cost to investors who are less well informed about bankruptcy risks than they might otherwise have been. Consistent with this, Chen and Church (1996) find the market response to a bankruptcy announcement is less negative (by 13%) when the auditor has previously issued a going concern report, indicating that bankruptcy is less of a surprise to investors. In general a large negative stock market reaction is also more likely to trigger an investor lawsuit against auditors, so this is less likely if auditors give an early warning of bankruptcy risk by issuing a going concern report. With regard to false positives, Francis and Krishnan (2002) document that around seven going concern reports were issued (nZ1003) per bankrupt company (nZ143) in their sample over the period 1990–1994, which means that on average six out of seven going concern reports were false positives. This implies that auditors are conservative in the sense of ‘over issuing’ going concern reports. 6 However, despite over issuing going concern reports, auditors’ diagnostic skills apparently fail to get it right when it matters and companies actually fail, i.e. as already indicated most bankruptcies are not preceded by a going concern report. The low cost of false positives relative to false negatives may explain why there are so many going concern audit reports. 7 The largest potential cost of false positives is client dissatisfaction, and Krishnan (1994) documents a higher rate of auditor switches for firms receiving false positives, around 22% compared to a base line switching rate of 6%. There is also a social cost of auditor switches as the client and new auditor incur transaction and start-up costs related to the switch. In addition, there is evidence that ‘new audits’ may be of lower quality due to a learning curve effect (Johnson et al., 2002), in which case 6 The ‘over issuing’ rate is lower if the number of unique firms receiving one or more going concern reports (729) is compared to the 143 bankrupt firms, i.e. about five reports per bankruptcy. 7 Another explanation is that auditor reporting conservatism is induced by the US legal system and the auditor’s exposure to litigation risk. This is discussed later in the paper. J.R. Francis / The British Accounting Review 36 (2004) 345–368350 a reduction in audit quality may occur when false positives induce auditor changes. Finally, another cost of false positives is the self-fulfilling prophecy problem. When auditors issue a going concern report, it may trigger a sequence of events that has the unintended consequence of pushing a client into bankruptcy. For example, banks or other lenders may not extend or renew lines of credit, and suppliers may change their terms in order to speed up cash payments. To sum up, auditors are not always accurate in their reporting choices and this can potentially reduce audit quality. They report conservatively (too many false positives) but more often than not they fail to get it right when it matters (too many false negatives). While the existence of false positives and false negatives creates noise and reduces the informativeness of audit reports, it does not necessarily eliminate their informational value altogether. I now review research which indicates that modified audit reports do have informational value, despite the presence of false positives. 3.2. Do modified audit reports really matter? The informativeness of audit reports is difficult to assess for two reasons. First, audit reports are typically released concurrently with the 10-K annual report filed with the SEC and therefore it is virtually impossible to separate information in the audit report from the overall set of information in the 10-K. Second, most going concern reports are repeat offenders and so there is little or no surprise value in the audit report. For this reason, much of the going concern literature examines first-time report modifications, and there is evidence that first-time going concern reports which come as a surprise do result in the expected negative stock market reaction (Dodd et al., 1984; Loudder et al., 1992). Another approach to informational value examines if modified audit reports have predictive power. Raghunandan (1993) investigates the predictive ability of audit reports modified for contingent losses arising mainly from lawsuits. These audit reports were required prior to 1989, and he finds they were more predictive of actual litigation outcomes than were financial statements having footnote-only disclosures of loss contingencies. In other words, when auditors commented directly on loss contingencies in the audit report, there was a significantly greater likelihood that a litigation-related loss would be incurred, which provides evidence that modified audit reports are informative. Initial public offerings (IPOs) are a particularly rich setting to investigate audit reports because less is known about companies due to greater information asymmetry and therefore the reports of auditors may potentially convey important information to investors. A study which exploits this setting is Weber and Willenborg (2003). They examine a sample of small microcap IPOs with market capitalization under $10 million. Surprisingly they document that around 25% of microcap IPOs go public with an audit report modified for going concern, and find that the pre-IPO audit reports have predictive ability with respect to both future stock returns and subsequent delistings. In sum, there is not a large amount of research on the informativeness of modified audit reports, in part because of the difficulty in developing a research design that can tease out the informativeness of such reports. However, there is evidence that modified audit reports are informative and have predictive ability, despite the noise created by false positives. J.R. Francis / The British Accounting Review 36 (2004) 345–368 351 4. Auditor differentiation and audit quality It is difficult to assess audit quality ex ante because the only observable outcome of the audit is the audit report which is a generic template and the overwhelming majority of reports are standard clean opinions. While it is possible to assess audit quality ex post in the case of outright audit failures, as already noted, these are relatively infrequent occurrences. An important development in audit quality research is based on the premise that ‘differences’ in audit quality exist and can be inferred by comparing different groups or classes of auditors. This research implicitly assumes that all audits meet minimum legal and professional standards (except of course the cases of outright audit failure), and therefore the research focuses on differential audit quality above and beyond the legal minimum. At an intuitive level we observe that there are many different kinds of audit firms which suggest there is a supply of differential auditing demanded by different clienteles. 4.1. The big firm–small firm dichotomy The first wave of auditor-differentiation research focused on the dichotomy between large and small firms as a basis for differential audit quality. DeAngelo (1981) argues that accounting firm size is a proxy for quality (auditor independence) because no single client is important to a large auditor and the auditor has a greater reputation to lose (their entire clientele) if they misreport. By contrast, an accounting firm with only one client may logically conclude that they have more to gain by going along with their client and misreporting than by being tough and potentially getting fired. A related line of research argued that the large ‘Big 8’ international accounting firms had established brand name reputations and therefore had incentives to protect their reputation by providing high- quality audits (Simunic and Stein, 1987; Francis and Wilson, 1988). These arguments do not necessitate that Big 8 (now Big 4) audits are always superior. Individual audit failures by Big 4 firms can and do occur. Rather, the arguments simply mean that audits of Big 4 firms as a group will, on average, be of higher quality than other (smaller) accounting firms. There are several streams of research motivated by the big firm-small firm dichotomy. Many studies document that Big 4 audits around the world carry a premium relative to the audits of other firms, after controlling for client characteristics affecting audit fees such as size, complexity and auditor-client risk sharing (Simunic, 1980). On average the Big 4 premium has been around 20%. Moizer (1997) provides a review of earlier studies, and for more recent evidence see DeFond et al. (2000), Ferguson et al. (2003). A higher audit fee implies higher audit quality, ceteris paribus, either through more audit effort (more hours) or through greater expertise of the auditor (higher billing rates). 8 The question then arises as to whether or not there is a demand for differential audit quality given that any licensed auditor can legally satisfy the requirement to have an audit. 8 A higher audit fee per se does not necessarily ensure a higher quality audit, particularly if accounting firms have pricing power over clients. However, the evidence on audit outcomes reviewed later in the paper provides confirmatory evidence that accounting firms charging higher fees (such as the Big 4) also provide higher quality audits on average. J.R. Francis / The British Accounting Review 36 (2004) 345–368352 In other words, why do firms voluntarily pay more for a higher quality audit when lower- priced and legal alternatives exist? Research has focused on those clienteles that logically might be expected to demand higher quality audits. Specifically, the evidence indicates that firms with greater monitoring needs due to higher agency costs are more likely to use Big 4 auditors (Francis and Wilson, 1988; DeFond, 1992; Francis et al., 1999). A second line of argument is that firms with greater inherent uncertainty (and greater information asymmetry between the firm and outsiders) have an incentive to communicate their intrinsic quality by hiring a more credible, high-quality auditor. This argument has mostly been made in the context of IPOs and the evidence indicates there is reduced information asymmetry (i.e. less underpricing) when going public with larger brand name auditors (Beatty, 1989; Willenborg, 1999). Corroborative evidence from other research supports the notion that larger accounting firms supply higher quality audits. For example, the Big 4 firms are sued relatively less frequently after controlling for clientele size, and Big 4 firms are sanctioned less frequently by the Securities and Exchange Commission (Palmrose, 1988; Feroz et al., 1991). A counter explanation is that the large accounting firms are not really better, they just have more resources to fight lawsuits and regulators. However, research on audit outcomes, which is discussed next, rebuts this view with evidence of higher quality audits by larger (Big 4) accounting firms. There are two observable audit outcomes, audit reports and audited financial statements. Evidence from audit report research supports that Big 4 auditors are of higher quality. Francis and Krishan (1999) show that Big 4 auditors have lower thresholds for issuing modified audit reports, which indicates greater reporting conservatism for a given set of client characteristics. Lennox (1999) finds that Big 4 auditors report with greater accuracy in the United Kingdom, and Weber and Willenborg (2003) find that the pre-IPO audit reports of large national and international (Big 4) accounting firms have more predictive accuracy than smaller accounting firms with respect to future stock returns and subsequent delistings. The evidence from financial statements also supports that Big 4 audits are of higher quality. Using the abnormal accruals paradigm (Jones, 1991), the evidence indicates that clients of Big 4 audited companies have lower abnormal accruals which implies less aggressive earnings management behavior and therefore higher earnings quality (Becker et al., 1998; Francis et al., 1999). 9 Consistent with these findings, Nelson et al. (2002) report evidence from one Big 4 accounting firm that auditors detect earnings management attempts (especially income-increasing attempts) and require clients to make appropriate adjustments. In addition, Teoh and Wong (1993) document that the earnings surprises of Big 4 audited companies are valued more highly by the stock market which is consistent with higher earnings quality when a Big 4 firm is the auditor. In sum, despite some recent high-profile cases, the collective evidence is strongly supportive that audits of large (Big 4) accounting firms are of higher quality. There is one 9 The accrual paradigm calculates unexpected or abnormal accruals relative to an expectation model of normal accruals. Abnormal accruals imply that managerial discretion over accounting is used to distort reported earnings for private benefits, and that managed earnings are different from the outcome of a neutral application of generally accepted accounting principles (Schipper, 1989). J.R. Francis / The British Accounting Review 36 (2004) 345–368 353 troubling aspect to this line of research. An alternative explanation may simply be that ‘good’ companies are more likely to select Big 4 auditors, are less likely to manage earnings, and in general are more likely to have higher quality earnings. In other words, it’s not high-quality auditing that causes the observed audit outcomes; rather, auditor choice is endogenous and it may simply be that good firms with good earnings quality hire high-quality auditors. Even though prior research has controlled for systematic clientele differences between Big 4 and non-Big 4 auditors, endogeneity and selection cannot be entirely ruled out as an alternative explanation and more work is needed on this important topic. A few studies have examined endogeneity and these studies are generally supportive of the research results cited above. See Hogan (1997), Ireland and Lennox (2002), and Weber and Willenborg (2003). 5. Moving beyond the big firm-small firm dichotomy The first wave of research described above viewed the Big 4 as a homogenous group of firms. The second wave of research relaxes this assumption and has begun examining potential sources of differentiation in audit quality within the dominant Big 4 group of accounting firms. 10 Three primary sources of differentiation have been investigated to date: differences due to industry specialization, differences across individual practice offices (cross-city differences), and institutional differences across countries (cross- country differences). 5.1. Industry expertise Big 4 accounting firms promote their industry expertise and we observe empirically that industry market shares are not evenly distributed among the large accounting firms. The linkage between industry market share and industry expertise is as follows. Solomon et al. (1999) argue that industry experts have a deeper knowledge than non-experts due to greater experience in the industry which enables experts to make more accurate audit judgments. If accounting firms have more clients/fees in an industry, then they have more opportunities for their auditors to acquire the kind of deep industry knowledge which leads to industry expertise. Francis et al. (2005) use the new US audit fee disclosures for 2000–2001 and calculate industry fee leaders for 63 non-financial industries based on two-digit SIC industry codes. On average they find that industry leaders have 50% of industry fees, while the number two firm has only 22% of industry fees. Industry leadership in the 63 industries is widely dispersed among firms and distributed as follows: Arthur Andersen (14), Deloitte Touche (5), Ernst and Young (16), KPMG (9), and Pricewaterhousecoopers (19). The evidence in support of industry expertise parallels the research on Big 4 audit quality. Audit fees are higher for industry leaders implying higher audit quality, 10 Another reason for this development is that Big 4 market shares continue to expand globally and now exceed 90% of publicly listed companies in the US. From a practical viewpoint this means there is low power in research designs of studies comparing large and small auditors because there is such low variance in the experimental variable, i.e. most observations are audited by large (Big 4) auditors. J.R. Francis / The British Accounting Review 36 (2004) 345–368354 [...]... and the United Kingdom (e.g see Fearnley and Beattie) 7 What do we know about audit quality? Summing up, it turns out that we know more about audit quality than we might have originally suspected: † † † † † † Auditing is relative inexpensive, less than 1/10 of one percent of aggregate client sales; Outright audit failures with material economic consequences are very infrequent; Audit reports are informative,... as the United Kingdom While these findings collectively suggest that audit quality may be at a socially desirable level, there are some fundamental and important things we do not know about audit quality First, we do not know if the US evidence on audit quality generalizes to audits in other countries that have different legal systems and in particular to non-common law countries J.R Francis / The British... significant cross-city variation in auditing and support that the office-level of analysis is an important development in research on audit quality 5.3 Other approaches to the study of audit quality There are other approaches to the study of auditor quality including research on audit tenure, non -audit fees, audit committees, accounting firm alumni, and the effect of legal systems on auditor incentives... lack of knowledge, the new US audit requirements in Sarbanes-Oxley are expected to have a large impact on audit fees with expected increases of 50% or more; yet, we have no compelling reason to believe that audit quality will necessarily be improved or that audit failures will be significantly reduced by these more costly audits Third, we do not know what the optimal social arrangement is for auditing... restrictions in Sarbanes-Oxley on the so-called revolving door or ‘outplacement’ of accounting firm personnel to clients What do we know about this phenomenon from empirical studies? Lennox (2004) finds that auditors are less likely to issue modified audit reports for clients with accounting firm alumni in top-level management positions, and Menon and Williams (2004) document that abnormal accruals are larger... Review 36 (2004) 345–368 357 An additional argument against rotation is that audit quality may be lower initially for new audit engagements while auditors acquire knowledge of the client Two recent studies provide evidence on auditor tenure and audit quality by examining the association between accounting accruals and tenure Meyers et al (2003) find no evidence that long tenure impairs audit quality and... that actual audit quality is high However, there continues to be a widely held perception, rightly or wrongly, that audit quality is impaired by non -audit services and this suspicion has deepened post-Enron These concerns led to restrictions on non -audit services by the SEC in 2000 and further restrictions by SarbanesOxley in 2002; however, companies continue to buy large amounts of non -audit services.15... objectivity toward high-risk clients There is some suggestion that this may have occurred in the Arthur Andersen audit of Enron Sarbanes-Oxley addresses this problem, in part, by prohibiting the use of non -audit fees in performance-based contracts Apparently it was common to reward audit partners based on the amount of non -audit services purchased by audit clients While non -audit fees are off the table... The British Accounting Review 36 (2004) 345–368 361 with weaker investor protection and less ability to sue auditors for negligence and misconduct Second, we do not know how much auditing is optimal or if requiring ‘more’ or ‘better’ auditing to reduce the risk of an audit failure is a good cost–benefit trade-off While audits are relatively inexpensive and the outright audit failure rate is low, when... this does not mean we have achieved nirvana with respect to audit quality Further, there is a public perception that audit quality may have declined in the 1990s, a view that is supported by some recent research To restore public confidence in auditing, and to possibly increase actual audit quality as well the public’s perception of quality, accounting firms can signal their unambiguous commitment to audit . Kingdom (e.g. see Fearnley and Beattie). 7. What do we know about audit quality? Summing up, it turns out that we know more about audit quality than we might have originally suspected: † Auditing. lower the quality of auditing. What do we know about audit failure rates? Outright audit failures are difficult to determine with certainty but can be inferred from several sources including auditor. high audit quality. Audit failures obviously occur on the lower end of the quality continuum, and so a good starting point in thinking about audit quality is to ask what the rate of outright audit