Apparent audit


Published on

Published in: Business, Technology
  • Be the first to comment

  • Be the first to like this

No Downloads
Total views
On SlideShare
From Embeds
Number of Embeds
Embeds 0
No embeds

No notes for slide

Apparent audit

  1. 1. Review of Accounting and FinanceEmerald Article: Apparent audit failures and value relevance of earningsand book valueLi Dang, Kevin F. Brown, B.D. McCulloughArticle information:To cite this document:Li Dang, Kevin F. Brown, B.D. McCullough, (2011),"Apparent audit failures and value relevance of earnings and book value", Reviewof Accounting and Finance, Vol. 10 Iss: 2 pp. 134 - 154Permanent link to this document: on: 29-06-2012References: This document contains references to 36 other documentsTo copy this document: permissions@emeraldinsight.comThis document has been downloaded 973 times since 2011. *Users who downloaded this Article also downloaded: *Imen Khanchel El Mehdi, Souad Seboui, (2011),"Corporate diversification and earnings management", Review of Accounting andFinance, Vol. 10 Iss: 2 pp. 176 - 196 Li, Yiming Qian, (2011),"Outside CEO directors on compensation committees: whose side are they on?", Review of Accountingand Finance, Vol. 10 Iss: 2 pp. 110 - 133 M. Werner, (2011),"The value relevance of pension accounting information: evidence from <IT>Fortune</IT> 200 firms",Review of Accounting and Finance, Vol. 10 Iss: 4 pp. 427 - 458 to this document was granted through an Emerald subscription provided by Universidad Nacional de CordobaFor Authors:If you would like to write for this, or any other Emerald publication, then please use our Emerald for Authors service.Information about how to choose which publication to write for and submission guidelines are available for all. Please for more information.About Emerald www.emeraldinsight.comWith over forty years experience, Emerald Group Publishing is a leading independent publisher of global research with impact inbusiness, society, public policy and education. In total, Emerald publishes over 275 journals and more than 130 book series, aswell as an extensive range of online products and services. Emerald is both COUNTER 3 and TRANSFER compliant. The organization isa partner of the Committee on Publication Ethics (COPE) and also works with Portico and the LOCKSS initiative for digital archivepreservation. *Related content and download information correct at time of download.
  2. 2. The current issue and full text archive of this journal is available at,2 Apparent audit failures and value relevance of earnings and book value134 Li Dang Orfalea College of Business, California Polytechnic State University, San Luis Obispo, California, USA Kevin F. Brown Department of Accountancy, Raj Soin College of Business, Wright State University, Dayton, Ohio, USA, and B.D. McCullough Department of Decision Sciences, LeBow College of Business, Drexel University, Philadelphia, Pennsylvania, USA Abstract Purpose – The purpose of this paper is to examine the value relevance of accounting information in cases of apparent audit failures. Design/methodology/approach – The authors adopt the bootstrapping technique and compare the value relevance of key accounting information across samples of firms experiencing apparent audit failures with matched non-audit failure firms. Findings – Accounting information is found to be less value relevant for firms experiencing apparent audit failures, regardless of auditor reputation. Research limitations/implications – This study has limitations due to the ex ante research approach adopted. Future research could address this issue by possibly incorporating an “intervening” factor into the model to indicate how the market can differentiate audit failure firms from other firms. Originality/value – The paper gives support to the assertion that the market appears to rely less on accounting numbers when audit failures occur, even though formal allegations of audit failure may not appear until years after their occurrence. In addition to contributing to value-relevance research by providing empirical evidence for the market phenomenon around the time of material misstatements, the paper demonstrates an innovative application of bootstrapping to test for differences in R 2. Keywords Auditing, Accounting information, Earnings Paper type Research paper I. Introduction DeAngelo (1981, p. 186) defines audit quality as “the market-assessed joint probability that a given auditor will both (a) discover a breach in the client’s accounting system, and (b) report the breach.” However, assessment of quality for particular audit engagements remains somewhat murky, due to the nature of the audit process, the reporting of audit outcomes, and users’ response to the auditor’s reputation (i.e. a BigReview of Accounting and Finance vs non-Big dichotomous audit quality measurement). Previous studies have examinedVol. 10 No. 2, 2011pp. 134-154 the association between audit quality and earnings response coefficients (Teoh andq Emerald Group Publishing Limited1475-7702 Wong, 1993). The empirical results appear to support that the market respondsDOI 10.1108/14757701111129616 positively to reputable auditors.
  3. 3. However, whether auditor reputation could serve as a reliable proxy for audit Apparentquality may be unknown due to the high-profile audit failures which occurred during the audit failuresdownturn in the financial markets a decade ago. Many notorious firms, for instance,Enron, WorldCom, and Tyco, were audited by Big five auditors with reputations similaror even superior to those of the non-Big five. Thus, one might conclude that an auditor’sreputation may actually hinder the market’s ability to assess the reliability of accountinginformation. 135 The purpose of this paper is to explore whether auditor reputation affects the valuerelevance of accounting information in cases of apparent audit failures. For the purposeof this study, apparent audit failures are defined as instances that an auditor issuesan unqualified opinion on materially misstated financial statements. Since apparentaudit failures indicate poor audit quality, they offer an opportunity to explore theassociation between stock prices and accounting information given low audit quality.Therefore, this study compares ex ante value relevance of accounting information ofpublicly held US firms experiencing apparent audit failures with a matched groupof firms not experiencing such failures. Ex ante value relevance refers to value relevanceof accounting information prior to the discovery of audit failures. Matching auditfailure firms with non-audit failure firms allows audit quality to be evaluated on a“service-by-service” basis, consistent with the suggestion of Lam and Chang (1994). This study differs from previous studies on market reactions to news of auditfailures, accounting scandals, or earnings restatements (Chaney and Philipich, 2002;Dechow et al., 1996; Feroz et al., 1991). Instead of examining stock market reactions to thenews of audit failures, we focus on examining value relevance of accounting informationin the alleged periods of misstatement. Those alleged periods are the reporting periodswhen audited financial statements contain material misstatements. We comparefirms experiencing apparent audit failures with matched firms without audit failures inthe same periods in order to find whether there is a difference in value relevance ofaccounting information prior to audit failures becoming public. Therefore, this is anexploratory study that aims to examine market reactions around the time when financialstatements contain misstatements. In addition, we divided our sample into sub-samplesclassified by auditor types (i.e. Big 8/6/5/4 auditors vs non-Big 8/6/5/4 auditors) andconduct the same comparisons to investigate whether such a difference is conditioned onauditor reputation. Several sources are used to identify apparent audit failures (in these cases, an auditorissues an unqualified opinion on materially misstated financial statements). Thesesources include the US Securities and Exchange Commission’s (SEC’s) Accounting andAuditing Enforcement Releases (AAERs), financial statements restatements because ofpast misstatements, and litigation against auditors due to audit failures. Consistent with prior research (Collins et al., 1997, 1999; Rees, 1999; Rajgopal et al.,2002), value relevance is measured by the explanatory power of contemporaneousearnings and book values for stock prices (i.e. the regression of stock price on earningsand book values). We apply a bootstrapping analysis to test data collected. The results ofthis analysis indicate that the accounting information of firms experiencing apparentaudit failures is generally less value relevant than that of a matched group of firms notexperiencing such failures. Results of sub-sample comparisons further indicate that theaudit reputation may not be essential because the market appears to discount lessreliable accounting information. The remainder of this paper is organized as follows.
  4. 4. RAF Section II describes prior research pertaining to market perceptions of audit quality.10,2 Section III contains the development of the research hypothesis. Section IV discusses data collection and the research methodology. Section V presents statistical analysis and empirical results. Section VI provides concluding remarks.136 II. Prior research Previous studies of audit failures have focused on market reactions to the announcements of earnings restatements, fraud class actions, and SEC enforcement actions (Liu et al., 2009; Akhigbe et al., 2005; Chaney and Philipich, 2002; Dechow et al., 1996; Bhagat et al., 1994; Francis et al., 1994; Kellogg, 1984). Overall, these studies have documented a negative market reaction to such announcements. For example, Francis et al. (1994) find a negative market response to the corrective disclosure of firms experiencing class action litigation. Feroz et al. (1991) document a 10 percent decline in stock prices at announcements of accounting violations. Akhigbe et al. (2005) find that earnings restatements lead to a negative market response and that the negative response is conditioned on the content of earnings management. Liu et al. (2009) find that shareholders are more likely to vote against the ratification of the auditor following an audit failure. In spite of these findings, none of these studies examine market reactions before accounting violations or audit failures became public knowledge. Chaney and Philipich (2002) examine the reputation effect of the Enron audit failure. Specifically, they investigate the impact of auditor reputation on the market prices of Arthur Andersen’s clients around the Enron bankruptcy. After Andersen’s reputation was damaged, its clients experienced a significant drop in their stock prices. The result of Chaney and Philipich (2002) appears to indicate that market perception of reliability of accounting information is conditioned on market perception of auditor reputation. Recently, Griffin et al. (2004) investigate the impact of class action litigation of audit failures on stock returns and earnings-returns relation in the alleged periods. They find that the market responds positively to releases of misleading accounting information. As Griffin et al. point out, the primary reason for this result is that the market initially accepts the misstated information. Furthermore, the ex post negative reaction to class action lawsuit announcement is conditioned on the returns-earnings relation in the alleged periods. In summary, except for Griffin et al (2004), most previous studies have focused on market reactions around news releases of audit failures instead of examining the market reaction to misstated earnings in the alleged periods. To yield additional insights for the literature, the purpose of this study is to explore value relevance of accounting information when financial statements contain material misstatements. Specifically, we examine the contemporaneous association between stock prices and misstated accounting information in the periods with alleged misstatement. This study differs from other audit failure studies in the following aspects: . Instead of focusing on ex post market reactions to audit failures, this study examines the market perception of the information quality ex ante. . Instead of using an event study approach, this study examines the association between stock prices and accounting information in a longer window. . This study also provides evidence on whether market perception of information quality in the alleged periods is conditioned on auditor reputation.
  5. 5. III. Development of hypothesis ApparentBecause apparent audit failures are less ambiguous, they provide a unique way of audit failurestesting the market’s perception of accounting information quality ex ante driven by pooraudit quality. Using firms not experiencing apparent audit failures as benchmarks, thisstudy examines whether the value relevance of accounting information is different forfirms experiencing apparent audit failures from those of benchmarked firms. In thisstudy, value relevance indicates the extent to which accounting information explains the 137variation of market prices. When financial statement users perceive higher informationquality, therefore more reliable financial statements, accounting information such asearnings and book values should explain more of the variation in stock prices. The Ohlson (1995) model is used to test the value relevance of accountinginformation (Amir, 1996; Amir and Lev, 1996; Collins et al., 1997, 1999; Rees,1999; Rajgopal et al., 2002). The Ohlson model relates a firm’s market value to itscontemporaneous accounting information. Specifically, this model provides a structure tostudy the relationship between equity values and earnings, as well as its relationship withbook values (Stober, 1999). The degree of value relevance is measured by the R 2 of theOhlson model (Collins et al., 1997; Rees, 1999; Rajgopal et al., 2002). The R 2 measuresthe degree of the variation in the dependent variable explained by independent variables.If the perceived information quality (i.e. audit quality) is low, the association between stockprice and accounting information, hence the R 2 of the Ohlson model, should be weakbecause accounting information will be viewed as less reliable. Owing to the exploratory nature of this study, we state our hypothesis in its nullform: H0. There is no difference between the explanatory power of earnings and book values for stock prices of firms experiencing apparent audit failures and firms not experiencing apparent audit failures.IV. Data collection and methodologyCases of apparent audit failures and the matched control groupFirms included in this study are selected from the SEC’s AAERs and restatementsof financial statements found in the Wall Street Journal Index and Lexis-NexisNews Library for fiscal years ending between 1980 and 2000. The initial sample isthe combination of these two data sources. AAERs indicate firms whose financialstatements containing misstatements documented in the SEC sanctions against firms ortheir auditors. We exclude AAER cases where quarterly misstatements are the focus ofthe sanction because quarterly statements are only reviewed, but not audited,by auditors. Restatements consist of firms that restated prior years’ financial statementsbecause of significant misstatements. These represent apparent audit failures becauseauditors did not detect and/or report those material misstatements initially. We searchedkeywords with the root “restat-” in Wall Street Journal Index and Lexis-Nexis NewsLibrary. Quarterly restatements are also excluded from our sample. To ensure that thesearch of the above two sources did not omit any audit failures; we also searched, as asecondary source, news accounts of auditor litigation including the auditor litigationdatabase compiled by Palmrose (1999). Cases of litigation against auditors containallegations that auditors failed to detect and report material misstatements. Such casesare included in this analysis due to possible lags in SEC scrutiny which would ordinarilyresult in enforcement activity and/or restatements of prior financial statements.
  6. 6. RAF Accordingly, litigation cases that provide substantial evidence of audit failure are10,2 included as apparent audit failures. While lawsuits seeking damages from an auditor are not always indicative of the auditor’s failure to adhere to professional standards, such suits may reveal audit failures before subsequent issuance of an AAER. A careful review of the litigation news accounts helped to ensure that only apparent audit failures are captured in the sample.138 In order to perform hypothesis testing, all accounting data for the years of the alleged audit failures that are necessary for testing the proposed relation are extracted from the COMPUSTAT database. All stock price information is obtained from the Center for Research in Security Prices (CRSP). The matched control group is selected based on established matching criteria. These criteria include the year of financial statements, auditor size (Big 8/6/5/4 vs non-Big 8/6/5/4), client industry, and client size. The required accounting data and price information are compiled for the matched control group from COMPUSTAT and CRSP, respectively. Model specification The model used to test the hypothesis is Ohlson’s (1995) valuation model (Amir, 1996; Amir and Lev, 1996; Collins et al., 1997, 1999). The Ohlson model is expressed as follows: P it ¼ a1 EPS it þ a2 BVPS it þ 1it ð1Þ where: Pit ¼ closing stock price of firm i’s equity three months after fiscal year end t[1]; EPSit ¼ firm i’s reported earnings per share before extraordinary items for period t; BVPSit ¼ firm i’s book value of equity per share at time t; 1it ¼ random error term with mean 0 and variance 1; and a1 and a2 ¼ the regression coefficients. Testing of hypothesis To test the hypothesis, comparisons are made between the apparent audit failure group of firms and the matched control group of firms not experiencing audit failures. For each comparison, two OLS regressions of the Ohlson model are performed: one regression for the audit failure group, and the other for the matched control group. If the R 2 for the audit failure group is statistically significantly lower than the R 2 for the matched control group, it suggests that value relevance of accounting information is lower in the alleged period and that the market appears to rely less on accounting information in the alleged period when there are material misstatements. If the R 2 for the audit failure group is equal to or higher than the R 2 for the matched control group, it suggests that the market may be “fooled” in the alleged period. Moreover, we examine whether value relevance of accounting information in the alleged period is conditioned on auditor reputation. For example, there might be a greater difference in value relevance of accounting information for firms audited by Big 8/6/5/4 auditors than those audited by non-Big 8/6/5/4 auditors. Comparisons we use to test the hypothesis are summarized in Table I. Comparison 1 tests whether there is a difference in value relevance of accounting information between the audit failure group and the matched group, in general.
  7. 7. Comparisons 2 and 3 test whether value relevance differs within Big 8/6/5/4 and Apparentnon-Big 8/6/5/4 groups. Comparisons 4 and 5 directly test whether the difference in value audit failuresrelevance of accounting information between the audit failure group and the matchedgroup is conditioned on auditor size.V. Analysis and resultsSample selection and characteristics 139We searched both the SEC’s online archives and the Lexis-Nexis News Library forAAERs. Between 1982 and 2000, AAERs No. 1 through No. 1357 were released. Thereare 559 unique firms identified from those AAERs[2]. Of these 559 firms, 383 uniquefirms are selected and the remaining 176 firms are deleted because the AAERs: (1) do not indicate the years misstated; (2) indicate the misstatement years are before 1980; (3) pertain to quarterly financial statements; or (4) indicate that the auditors are not responsible for the misstated financial statements.Our search for financial restatements in the Wall Street Journal Index and theLexis-Nexis News Library yields 462 unique firms with restatements. Of those 462 firms,273 unique firms meet the requirements for our analysis. The other 189 firms are deletedfor reasons discussed in (1) through (4) above. We then combine the 383 firms sanctionedin the AAERs with the 273 firms disclosing subsequent restatements and find that42 firms exist in both data sources. Therefore, we identify 614 firms that experiencedapparent audit failures from these two data sources. Since our analysis requires accounting data from COMPUSTAT, we first searchedfor the ticker identification numbers for those 614 firms by their firm namesin COMPUSTAT. A total of 154 firms were deleted because they do not haveticker identification numbers[3], leaving 460 firms remaining for further analysis.An additional 297 firms that meet our data requirements are found in the auditorlitigation database compiled by Palmrose (1999)[4]. Since 74 of these firms are alreadyincluded in our 460-firm sample, our overall sample increased to 683 unique firms.All these firms have materially misstated annual financial statements within the periodfrom 1980 to 2000, and their COMPUSTAT ticker identification numbers are available. As this study defines apparent audit failures as cases in which auditors provideunqualified opinions on financial statements that contain material misstatements,a search of the COMPUSTAT database[5] for those firms receiving unqualified opinionswas performed. Of the 683 firms, 442 firms (848 firm/years) have audit opinionComparison Failure group (AF) auditors Non-failure group (NAF) auditors1 Both Big 8/6/5/4 and non- Both Big 8/6/5/4 and non-Big 8/6/5/4 auditors Big 8/6/5/4 auditors2 Big 8/6/5/4 auditors Big 8/6/5/4 auditors3 Non-Big 8/6/5/4 auditors Non-Big 8/6/5/4 auditors Table I.4 Big 8/6/5/4 auditors Non-Big 8/6/5/4 auditors Comparisons used to test5 Non-Big 8/6/5/4 auditors Big 8/6/5/4 auditors the hypothesis
  8. 8. RAF information available for the specified financial statement years. An additional 26 firms10,2 (69 firm/years) were deleted because auditors issued qualified opinions to those firms. Therefore, 416 firms (779 firm/years) are eligible for inclusion as apparent audit failures. The sample for this study was reduced further because of the absence of other data required to test the hypothesis. To test the hypothesis, earnings per share, book value of stockholders’ equity, and stock price data must be available from the COMPUSTAT and140 CRSP databases. Given these considerations, the resulting sample of apparent audit failures is 346 firms (616 firm/years). Table II reports the sample size and the industry distribution information of sample firms. Given the matched-pairs design used in this study, each audit failure firm in the sample is matched with a control firm based on year of misstated financial statements, industry (SIC code), company size (total assets), and auditor type (Big 8/6/5/4 vs non-Big 8/6/5/4)[6]. To get the matched pairs, we searched for similar size firms (with total assets within 10 percent) in the same industry (two- to four-digit SIC code) in the alleged years[7]. If there is more than one similar size firm available, we select the one with the closest SIC code. If there is no firm of similar size available in the same industry, we select the firm with the closest total assets. In case there is no matched pair available according to our matching criteria, we drop the audit failure observation. Hypothesis testing Prior accounting research has used a model’s R 2 statistic to measure the value relevance of accounting information (Lang et al., 2003; Sami and Zhou, 2004; Francis and Schipper, 1999; Nwaeze, 1998; Collins et al., 1997; Amir and Lev, 1996; Harris et al., 1994). These studies have measured value relevance as the R 2 resulting from the regressions of stock prices on per share values of earnings and book values of equity. These studies compare value relevance measured by R 2 either over different time periods or across different samples. In this study, R-squares are compared across audit failure and non-audit failure groups to investigate whether there is a difference in value relevance between the group of firms experiencing apparent audit failures and a matched group of firms that do not experience Panel A: sample determination for apparent audit failure cases Number of firms Number of firm/years Audit failure cases 416 779 Less: observations without price data 70 163 Sample for testing 346 616 Panel B: industry distribution for apparent audit failure cases Industry SIC code Number of observations Percentage Agriculture, forestry, fishing 01-09 1 0.29 Mining 10-14 15 4.34 Construction 15-17 7 2.02 Manufacturing 20-39 127 36.71 Transportation and public utilities 40-49 31 8.96 Wholesale trade 50-51 12 3.47 Retail trade 52-59 38 10.98 Finance, insurance, real estate 60-67 28 8.09Table II. Services 70-89 82 23.70Sample determination Public administration 90-99 5 1.45and industry distribution Total 346 100.00
  9. 9. apparent audit failures. Except in Harris et al. (1994), Lang et al. (2003) and Sami and Zhou Apparent(2004), value-relevance studies have not included a formal test for the difference of R 2. Harris et al. (1994), Lang et al. (2003) and Sami and Zhou (2004) use the procedure audit failuresdemonstrated in Cramer (1987) to obtain the mean and the variance of R 2. Then, theyconduct a z-test to compare the means of two R 2. A major problem in using the Cramermethod in this context of a z-test is that it depends on the assumption of the normaldistribution of R 2. However, the distribution of R 2 is not normal, even asymptotically 141(Anderson, 2003, p. 155). As an example, Figure 1 shows an R 2 distribution given an R 2of 0.9 for a regression model with ten independent variables (sample size ¼ 100). Visualinspection of this figure reveals that the distribution of R 2 is not normal. An additionalcomplication is that calculation of the moments of R 2, which is required to applyCramer’s method, is prone to computational difficulties that may produce incorrectresults, even when the requisite formulae seem to be programmed correctly. Because of the difficulty and possible unreliability of using the Cramer procedure, thisstudy uses the bootstrap method to create tests for the difference in R 2. Bootstrapping isa resampling method that requires fewer assumptions than traditional methods.In general, it also provides more accurate results. For example, the bootstrap methoddoes not require normality of the distribution of the R 2 and it can provide a fasterconvergence to the expected value of the parameter of interest. Since the models cannot be nested to formally test H0: R 2 ¼ R 2 against Ha: R 2 – R 2 , a b a bwe employ the usual approach of comparing the confidence interval for R 2 and the aconfidence interval for R 2 . Barr (1969) illustrates that the length of the confidence bintervals for the two-interval test must be constructed with the multiplier: pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 0 n1 þ n2 z ¼ pffiffiffiffiffi pffiffiffiffiffi z0:975 n1 þ n2if the significance level of 0.05 is desired. When the sample sizes of the two samples arethe same, i.e. both have n observations, the multiplier becomes: pffiffiffiffiffi pffiffiffi 0 2n 2 z ¼ pffiffiffi z0:975 ¼ z0:975 : 2 n 2As a result, the percentage of the confidence intervals is 83.4 percent. Therefore,comparing two 95 percent intervals is incorrect if a ¼ 0.05 is required[8]. 20 15 10 5 Figure 1. An example of an R 2 distribution 0.86 0.88 0.92 0.94
  10. 10. RAF Descriptive statistics10,2 The sample selection procedure yields 616 firm/year observations of apparent audit failures, which are matched with a control group in order to test the hypothesis. In the audit failure group, 502 (114) firm/years were audited by Big 8/6/5/4 auditors (non-Big 8/6/5/4 auditors). Table III illustrates the observations included in each comparison[9]. Table IV presents descriptive statistics, including the mean, median, and standard142 deviation for total assets (TA), stock prices (P), earnings per share (EPS), and book value of equity per share (BVPS) for both groups in each comparison. Table IV also includes a comparison of the means of these variables. Since the distributions of these variables might not be normal, both a two-sample t-test and a nonparametric Wilcoxon test are conducted. In general, the audit failure group and the matched control group are not significantly different in total assets, which reflects a successful control for company size. Because of the matching requirements, the audit failure and non-audit failure groups exhibit different characteristics in firm size, as shown in comparison 4. As stated earlier, in cases where there are no companies available in the same industries with the similar firm sizes as audit failure observations, we select matched pairs with the closest total assets. Consistent with the tendency of larger firms to have Big 8/6/5/4 auditors, comparison 4 shows that the average company size of the audit failure group is significantly larger than that of the matched control group. In comparison 5, the Wilcoxon test results show a significant difference in company size, while the two-sample t-test does not indicate a significant difference. The control firms also appear to earn more than the audit failure firms. With the exception of comparison 3, the matched control groups have a higher EPS than the audit failure group. Audit failure group Matched control group Comparison 1 Number of observations 616 616 Big 8/6/5/4 502 502 Non-Big 8/6/5/4 114 114 Comparison 2 Number of observations 502 502 Big 8/6/5/4 502 502 Non-Big 8/6/5/4 0 0 Comparison 3 Number of observations 114 114 Big 8/6/5/4 0 0 Non-Big 8/6/5/4 114 114 Comparison 4 Number of observations 463 463 Big 8/6/5/4 463 0 Non-Big 8/6/5/4 0 463Table III. Comparison 5Number of firm/year Number of observations 114 114observations in each Big 8/6/5/4 0 114comparison Non-Big 8/6/5/4 114 0
  11. 11. Apparent Auditor failure group Matched control group Compare meansa Mean Median SD Mean Median SD t-test Wilcoxon test audit failuresComparison 1TA 2,421.1010 135.5100 8,309.7470 1,860.2880 114.5100 6,508.7950 0.1875 0.332P 17.4288 12.2500 17.7024 18.3725 12.0630 22.4992 0.4135 0.9602EPS 20.0065 0.3450 7.0179 0.7458 0.5000 5.7049 0.0392 0.0022 143BVPS 7.6564 12.2500 9.7234 8.6550 5.7230 16.2675 0.1913 0.2421Comparison 2TA 2,958.4940 224.6000 9,121.2050 2,243.6640 182.2900 7,122.9120 0.1667 0.2256P 19.4984 14.7500 17.9248 21.2554 15.2500 23.8012 0.1867 0.5052EPS 0.01631 0.4900 7.7616 0.9306 0.7850 6.2850 0.0405 0.0007BVPS 8.7918 6.2610 10.2356 9.9054 6.9360 17.6025 0.2208 0.3822Comparison 3TA 54.6836 16.7150 138.4634 172.0858 15.0150 1,447.4700 0.3904 0.9080P 8.3157 4.0000 13.3492 5.6772 3.4380 6.6272 0.0605 0.2113EPS 20.1072 2 0.0100 0.9703 20.0679 0.0200 1.0853 0.7734 0.8574BVPS 3.1489 1.1916 4.3721 2.6570 1.7497 5.3825 0.4496 0.0890Comparison 4TA 3,008.3940 211.5500 9,391.1080 1,393.2470 64.5000 9,913.3340 0.0111 0.0000P 19.6246 14.5000 18.3666 15.1947 8.8750 22.0622 0.0009 0.0000EPS 20.0472 0.4700 8.0709 0.5576 0.3700 1.6394 0.1147 0.9536BVPS 8.5995 6.1680 10.4589 7.8357 4.9530 13.6130 0.3386 0.0002Comparison 5TA 54.6836 16.7150 138.4634 46.9001 23.3100 70.6154 0.5932 0.0426P 8.3157 4.0000 13.3492 10.3906 6.7500 14.9525 0.2691 0.0038EPS 20.1072 2 0.0100 0.9703 0.1656 0.0900 1.3441 0.0796 0.0309BVPS 3.1489 1.1916 4.3721 4.7300 3.3794 5.4881 0.0018 0.0000Notes: All p-values less than or equal to 0.05 are shown in italics; aboth p-values of the two-sample t-test Table IV.and the nonparametric two-sample Wilcoxon test are given; TA, total asset at the end of the fiscal year t; Descriptive statistics forP, stock price, three month after the end of the fiscal year t; EPS, earnings per share excluding observations included inextraordinary items for fiscal year t; BVPS, book value of equity per share at the end of the fiscal year t hypothesis testingHypothesis testing using bootstrap confidence intervalsFor the purpose of the current study, 1,000 bootstrap resamples were created bysampling with replacement from each of the original samples. Each bootstrap resampleis the same size as the original sample[10]. For each bootstrap resample, there is one R 2generated from the bootstrap regression, which is called R 2 *. Therefore, the 1,000bootstrap resamples generated 1,000 R 2 *s[11]. To test whether R 2 of the audit failure group and the matched control group differ, wecompare the 83.4 percent bootstrap percentile confidence intervals of R 2. Figures 2-6show the histograms about the distribution of R 2 based on bootstrap resamples. Thehistograms of R 2 *s sometimes indicate approximate normality, but other times show anapproximately normal central region with non-normality in the tail region. Moreimportantly, there are three cases where even the central region is obviously non-normal.Of these three, two cases are the distributions of R 2 for the audit failure groups auditedby non-Big 8/6/5/4 auditors in comparisons 3 and 5, which exhibit an asymmetricpattern because most R 2 from the bootstrap resamples are very close to zero. The thirdcase appears in comparison 5, where R 2 for the matched control group audited by Big
  12. 12. RAF 1510,2 10144 Density 5 0 0.10 0.15 0.20 0.25 R2 star 15 10 Density 5 0 0.20 0.25 0.30 0.35 0.40 0.45 0.50Figure 2. R2 star Notes: R2 *: audit failure group – both Big 8/6/5/4 and non-Histograms for Big 8/6/5/4; R2 *: matched control group – both Big 8/6/5/4comparison 1 and non-Big 8/6/5/4 8/6/5/4 auditors appear to have a bimodal distribution, a probability distribution characterized by two humps rather than the more common single hump that characterizes the normal distribution. One hump is near 0.15 and the other is near 0.46. Since we run bootstrap regressions by resampling the residuals from the original regression, the bimodal distribution of R 2 appears to be driven by an outlier in the residuals. Examining the regression output, we find one residual with an extreme value. The bimodality exists when a highly influential point is included in some bootstrap resamples but not in others. When we delete this outlier, the bimodal distribution of R 2 disappears. The histogram of R 2 after removing the outlier is shown in Figure 6.
  13. 13. 15 Apparent audit failures 10 Density 145 5 0 0.10 0.15 0.20 0.25 0.30 R2 star 15 10 Density 5 0 0.25 0.30 0.35 0.40 0.45 0.50 0.55 R2 star Figure 3. Notes: R2 *: audit failure group – Big 8/6/5/4; R2 *: matched Histograms comparison 2 control failure group – Big 8/6/5/4As stated earlier, one advantage of bootstrap methods is that they do not requiredistributions to be normal. Hence, the above descriptions of a non-normal pattern are nottroublesome. To create 83.4 percent bootstrap percentile confidence intervals, the 1,000 R 2 *sfrom bootstrap regressions were sorted in an ascending order. The lower value of theconfidence interval is the 83rd (0.083*1,000) R 2 * and the upper value is the 917th(0.917*1,000) R 2 *. The results of the bootstrap percentile confidence intervals are shown inTable V. Comparisons of bootstrap percentile confidence intervals indicate that there aresignificant differences in R 2 for the audit failure groups and the matched control groups[12].Hypothesis testing using the Cramer procedureTo be consistent with prior literature (Harris et al., 1994; Lang et al., 2003; Sami and Zhou,2004), we perform the test using the Cramer procedure. Cramer (1987) provides a procedure
  14. 14. RAF 3010,2 25 20146 Density 15 10 5 0 0.0 0.1 0.2 0.3 0.4 R2 star 15 10 Density 5 0 0.1 0.2 0.3 0.4 0.5 0.6Figure 4. R2 star Notes: R2 *: audit failure group – non-Big 8/6/5/4; R2 *: matchedHistograms comparison 3 control group – non-Big 8/6/5/4 to calculate the first moment and second moment of R 2. Once we have the mean and variance of R 2, we can calculate the z-statistic as in Harris et al. (1994) and Lang et al. (2003) and compare means of R 2 across two samples. The expression of a z-test can be modified as follows since the number of observations of R 2 is only one for each sample: E R2 2 E R 2 1 2 z ¼ rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi V R 2 þ V R2 1 2
  15. 15. 15 Apparent audit failures 10 147 Density 5 0 0.10 0.15 0.20 0.25 0.30 R2 star 15 10 Density 5 0 0.2 0.3 0.4 0.5 0.6 R2 star Figure 5. Notes: R2 *: audit failure group – Big 8/6/5/4; R2 *: matched Histograms comparison 4 control group – non-Big 8/6/5/4The results using the Cramer procedure and z-test are presented in Table VI. These resultsare similar to those reported by employing the bootstrap method. Therefore, our resultsare robust[13]. As discussed in McCullough and Vinod (1993), caution should be givenwhen using the Cramer procedure to calculate the mean and standard deviation of the R 2.For instance, we encountered computer operation overflows, resulting in incorrectoutputs, before we modified our programming to resolve such problems.
  16. 16. RAF 30 1510,2 25 20 10 Density Density 15148 10 5 5 0 0 0.0 0.1 0.2 0.3 0.4 0.0 0.1 0.2 0.3 0.4 0.5 0.6 R2 star R2 star 15 10 Density 5 0 0.1 0.2 0.3 0.4 0.5 0.6Figure 6. R2 star Notes: R2 *: audit failure group – non-Big 8/6/5/4; R2 *: matched control group – Big 8/6/5/4; R2 *:Histograms comparison 5 matched control group – Big 8/6/5/4; after removing the outlier Sensitivity analysis To ensure that our analyses are not affected by potentially confounding factors such as negative earnings or financial distress, we perform several additional analyses. First, as shown in the descriptive statistics, the audit failure group on average exhibits a lower EPS compared with the matched control group. The frequency of losses in the audit failure group is higher compared with the matched control group. Prior literature (Collins et al., 1999; Burgstahler and Dichev, 1997; Hayn, 1995) has suggested that losses are less informative for prices than profits. Therefore, the relation between stock price and earnings and book values of firms in financial distress may be fundamentally different from non-distressed firms. To ensure that our results are not driven by the relatively higher frequency of losses or a greater degree of financial distress among the firms in the audit failure group, we carry out two additional analyses. First, we performed a second matching of non-audit failure firms to the audit failure group, adding two additional matching criteria, return on assets (ROA) and leverage, to control for the influence of financial distress. We also analyze the data after removing all these observations with negative EPS and their pairs in the matched control group.
  17. 17. ApparentComparison Failure group Non-failure group audit failures1 Auditor Both Big and non-Big 8/6/ Both Big and non-Big 8/6/ 5/4 5/4 83 percent bootstrap percentile (0.1344, 0.2234) (0.3555, 0.4825) CI2 Auditor Big 8/6/5/4 Big 8/6/5/4 149 83 percent bootstrap percentile (0.1175, 0.2072) (0.3520, 0.4877) CI3 Auditor Non-Big 8/6/5/4 Non-Big 8/6/5/4 83 percent bootstrap percentile (0.0068, 0.1161) (0.2640, 0.5030) CI4 Auditor Big 8/6/5/4 Non-Big 8/6/5/4 83 percent bootstrap percentile (0.1229, 0.2207) (0.3965, 0.6054) CI5 Auditor Non-Big 8/6/5/4 Big 8/6/5/4 83 percent bootstrap percentile (0.0068, 0.1161) (0.2045, 0.3894) CI Table V. Comparisons of bootstrapNotes: All p-values less than or equal to 0.05 are shown in italics; regression model: P it ¼ a1 EPS it percentile confidenceþa2 BVPS it intervals for R 2Also, in our original analyses, we include material misstatements that received both“unqualified opinion” and “unqualified opinion with explanatory language” as auditfailure cases. Some of the “explanatory language” cases refer to an auditor’s substantialdoubts about a firm’s ability to continue as a going concern. Although the “goingconcern” opinions are included in the audit failure group, insofar as the auditor’s opinionon the fairness of the firm’s financial statements is not qualified, it is possible thatfirms that receive going concern opinions are fundamentally different from the rest ofthe firms in the sample. While we already control for financial distress factor by addingROA and financial leverage as two additional matching criteria, we perform additionalanalyses after removing all firms with going concern opinions and their matched pairs inthe control group, in order to exclude the possibility that our results are driven by someother confounding factor associated with the going concern opinion. Additionally, we find that a few audit failure firms changed their auditors during theperiod their financial statements were misstated. It is possible that these firms might bedifferent from other firms because they might be priced differently due to the change inauditors. Previous research (Teoh, 1992) has documented that auditor changes mightaffect stock prices and therefore the association between stock prices and accountinginformation. Schwartz and Menon (1985) examine the motivations for auditor switchingand consider financial distress as a factor that affects auditor switching. Also, there areother factors, such as audit opinion qualification (Chow and Rice, 1982) that mighttrigger auditor switching. Therefore, in addition to controlling for financial distress, wealso analyze the data without those firms that changed auditors during theirmisstatement years. The results of all these additional analyses are consistent with ouroriginal analyses, which support our hypothesis[14].
  18. 18. RAF Comparison Failure group Non-failure group10,2 1 Auditor Both Big and non-Big 8/6/5/4 Both Big and non-Big 8/6/5/4 n 616 616 R2 0.1714 0.4179 Mean (R 2) 0.1735 0.4190150 SD (R 2) 0.0264 0.0270 z-statistic 6.4977 2a Auditor Big 8/6/5/4 Big 8/6/5/4 n 502 502 R2 0.1555 0.4129 Mean (R 2) 0.1582 0.4142 SD (R 2) 0.0286 0.0210 z-statistic 6.1826 3b Auditor Non-Big 8/6/5/4 Non-Big 8/6/5/4 n 114 114 R2 0.0266 0.3555 Mean (R 2) 0.0353 0.3624 SD (R 2) 0.0369 0.0651 z-statistic 4.3668 4a Auditor Big 8/6/5/4 Non-Big 8/6/5/4 n 463 463 R2 0.1628 0.4827 Mean (R 2) 0.1656 0.4839 SD (R 2) 0.0301 0.0291 z-statistic 7.6015 5b Auditor Non-Big 8/6/5/4 Big 8/6/5/4 n 113c 113c R2 0.0284 0.2862 Mean (R 2) 0.0381 0.2946 SD (R 2) 0.0369 0.0663 z-statistic 3.3805 Notes: All p-values less than or equal to 0.05 shown in italics; regression model: P it ¼ a1 EPS it þ a2 BVPS it ; ain comparisons 2 and 4, audit failure group is the same group except the number of observations differs; failure group in both comparisons represents firms audited by Big auditors; in comparison 4, we dropped some failure group firms since we cannot find the matched pairs audited by non-Big auditors; bin comparisons 3 and 5, audit failure group is the same group since itTable VI. represents audit failure firms audited by non-Big auditors; cthe number of observations is due toComparison of R 2 deleting an outlier existing in the matched control group and its pair in the audit failure group; similarusing a z-test based to the bootstrap testing results, z-statistic is not significant before deleting the outlier because the largeon the Cramer procedure magnitude of the variance of the R 2 for the matched control group, 0.1939 VI. Limitations and conclusions Limitations This study has limitations due to the ex ante research approach adopted. Because of this ex ante approach, our matching criteria have to rely on the “noisy” accounting measure, total assets, in the alleged years. Therefore, the matching might not be effective in controlling for size. Although our primary objective is to explore the market phenomenon around the time of material misstatements, exploring why the market discounted accounting information reported by audit failure firms even the news of these misstatements is also
  19. 19. an important question. Future research could address this issue by possibly incorporating Apparentan “intervening” factor into the model to indicate how the market can differentiate audit audit failuresfailure firms from other firms. Finally, our analysis is based on the pre-Sarbanes-OxleyAct (SOX) period and therefore caution must be exercised when generalizing the findingsof this study to the post-SOX period.Conclusions 151In this study, we examine whether there is a difference in value relevance of accountinginformation between firms experiencing apparent audit failures and firms that have notexperienced apparent audit failures. The bootstrap method is used to test for differencesin R 2 across samples. The results suggest that the value relevance of earnings and bookvalues of equity is lower for firms experiencing apparent audit failures than those forfirms that have not experienced apparent audit failures. Although the market mayexhibit a lower level of confidence in accounting information from firms audited bynon-Big 8/6/5/4 auditors, the results provide evidence that the market generally exhibitslower confidence in firms experiencing audit failure, regardless of their auditors’ size.These empirical results show that the explanatory power of accounting informationof firms experiencing apparent auditor failures is lower than that of firms that have notexperienced apparent auditor failures. In our view, the market appears to rely less on accounting information reported byfirms experiencing apparent audit failures regardless of auditor size, even though anapparent audit failure may not be confirmed by future actions and events until well afterits occurrence. Moreover, in comparison 4, accounting information provided by firmsexperiencing audit failures and audited by Big 8/6/5/4 auditors is less value relevantcompared with firms audited by non-Big 8/6/5/4 auditors that have not experiencedaudit failures. This result suggests the difference in value relevance between theaudit failure group and the matched groups is not conditioned on auditor size. Thus,as illustrated in comparisons 1 through 5, the results indicate that there is a difference invalue relevance between the audit failure and the matched group ex ante. In addition to providing evidence suggesting that auditor size does not impact thevalue relevance of accounting information in a dysfunctional manner, this studydemonstrates an innovative application of bootstrapping to test for differences in R 2 inthe context of value-relevance research. We believe this method may allow researchersto make more precise conclusions when conducting future value-relevance studies.Notes 1. According to SEC requirements, before September 2002, Form 10-K had to be filed within 90 days after the end of the company’s fiscal year. 2. Several AAERs do not identify firm names. Further, some firms received multiple AAER sanctions. There are also some cases where the alleged impropriety involves a governmental entity. 3. Having ticker identification numbers does not necessarily mean that these firms have all required data for specific years in question available in COMPUSTAT. 4. Palmrose’s (1999) database provides tickers for those companies that have ticker identification numbers in COMPUSTAT. We exclude those cases of apparent audit failure which do not have tickers in the Palmrose (1999) database.
  20. 20. RAF 5. COMPUSTAT data item no. 149 provides both auditor and audit opinion information. However, it does not provide reasons why the qualified audit opinions were issued. We10,2 include auditor opinions with a value 1 (unqualified opinion) and 4 (unqualified opinion with explanatory language). 6. The matched-pairs control group also excludes firms with qualified audit opinions. 7. By selecting matched pairs according to such criteria, we control for confounding factors152 such as size, industry, and time period. Since total assets reported by audit failure firms in the alleged years could be noisy, control for size might not be very effective. However, we do not expect that any discrepancy in matching would cause significant problems given that “size” is matched within a ^10 percent range. 8. For example, if the sample size for two samples are both 300, the multiplier for the confidence interval is: pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffi pffiffiffiffiffiffiffi pffiffiffi 0 2 £ 300 2 · 300 2 z ¼ pffiffiffiffiffiffiffi · z0:975 ¼ pffiffiffiffiffiffiffi · z0:975 ¼ · z0:975 ¼ 0:834: 2 300 2 300 2 9. Note that in comparison 4 in Table II, 39 of the 502 audit failure cases are dropped due to inability to match these firm/years with those of a non-Big 8/6/5/4 control group. 10. For example, for the original sample size of 616 firm/years, the bootstrap resamples also will contain 616 observations, but will not be identical to the set of observations in the original sample. 11. See McCullough and Vinod (1993) for details on implementing the bootstrap procedure. 12. In comparison 5, the result is not significant before the outlier is removed. The confidence interval (0.0822, 0.5369) is wide when the outlier is included in the sample because of the bimodal distribution of R 2. 13. If the results are different using these two methods, the bootstrap method should be more reliable since the z-test requires normality of the R 2 and the distribution of R 2 is not exactly normal. 14. As a result of the more strict matching criteria, the sample size decreased in the second match. We deleted 195, 128, and 61 firm/years with negative EPS, with explanatory language and multiple auditors during the misstatement years, respectively. Results from these additional analyses are not tabulated. References Akhigbe, A., Kudla, R.J. and Madura, J. (2005), “Why are some corporate earnings restatements more damaging?”, Applied Financial Economics, Vol. 15 No. 5, pp. 327-36. Amir, E. (1996), “The effect of accounting aggregation on the value-relevance of financial disclosures: the case of SFAS No. 106”, The Accounting Review, Vol. 71 No. 4, pp. 573-90. Amir, E. and Lev, B. (1996), “Value-relevance of nonfinancial information: the wireless communications industry”, Journal of Accounting and Economics, Vol. 22, pp. 3-30. Anderson, T.W. (2003), An Introduction to Multivariate Analysis, Wiley, New York, NY, pp. 149-57. Barr, D.R. (1969), “Using confidence intervals to test hypotheses”, Journal of Quality Technology, Vol. 1, pp. 256-8. Bhagat, S., Brickley, J.A. and Coles, J.L. (1994), “The costs of inefficient bargaining and financial distress”, Journal of Financial Economics, Vol. 35 No. 2, pp. 221-47.
  21. 21. Burgstahler, D.C. and Dichev, I.D. (1997), “Earnings, adaptation and equity value”, Apparent The Accounting Review, Vol. 72 No. 2, pp. 187-215. audit failuresChaney, P.K. and Philipich, K.L. (2002), “Shredded reputation: the cost of audit failure”, Journal of Accounting Research, Vol. 40 No. 4, pp. 1221-45.Chow, C.W. and Rice, S.J. (1982), “Qualified auditor opinions and auditor switching”, The Accounting Review, Vol. 57 No. 2, pp. 326-35.Collins, D.W., Maydew, E.L. and Weiss, I.S. (1997), “Changes in the value-relevance of earnings 153 and book values over the past forty years”, Journal of Accounting and Economics, Vol. 24, pp. 39-67.Collins, D.W., Pincus, M. and Xie, H. (1999), “Equity valuation and negative earnings”, The Accounting Review, Vol. 74 No. 1, pp. 29-61.Cramer, J.S. (1987), “Mean and variance of R 2 in small and moderate samples”, Journal of Econometrics, Vol. 35, pp. 253-66.DeAngelo, L.E. (1981), “Auditor size and audit quality”, Journal of Accounting and Economics, Vol. 3, pp. 183-99.Dechow, P.M., Sloan, R.G. and Sweeney, A.P. (1996), “Causes and consequences of earnings manipulation: an analysis of firms subject to enforcement actions by the SEC”, Contemporary Accounting Research, Vol. 13 No. 1, pp. 1-36.Feroz, E.H., Park, K. and Pastena, V.S. (1991), “The financial and market effects of the SEC’s accounting and auditing enforcement releases”, Journal of Accounting Research, Vol. 29 No. 3, pp. 107-42.Francis, J. and Schipper, K. (1999), “Have financial statements lost their relevance?”, Journal of Accounting Research, Vol. 37, pp. 319-52.Francis, J., Philbrick, D. and Schipper, K. (1994), “Shareholder litigation and corporate disclosures”, Journal of Accounting Research, Vol. 32 No. 2, pp. 137-64.Griffin, P.A., Grundfest, J.A. and Perino, M.A. (2004), “Stock price response to news of securities fraud litigation: an analysis of sequential and conditional information”, ABACUS, Vol. 40 No. 1, pp. 21-48.Harris, T.S., Lang, M. and Moller, H.P. (1994), “The value relevance of German accounting measures: an empirical analysis”, Journal of Accounting Research, Vol. 32, pp. 187-209.Hayn, C. (1995), “The information content of losses”, Journal of Accounting and Economics, Vol. 20, pp. 125-53.Kellogg, R.L. (1984), “Accounting activities, security prices, and action lawsuits”, Journal of Accounting and Economics, Vol. 6 No. 3, pp. 185-204.Lam, S. and Chang, S. (1994), “Auditor service quality and auditor size: evidence from initial public offerings in Singapore”, Journal of International Accounting Auditing and Taxation, Vol. 3 No. 1, pp. 103-14.Lang, M., Raedy, J.S. and Yetman, M.H. (2003), “How representative are firms that are cross-listed in the United States? An analysis of accounting quality”, Journal of Accounting Research, Vol. 41, pp. 363-86.Liu, L., Raghunandan, K. and Rama, D. (2009), “Financial restatements and shareholder ratifications of the auditor”, Auditing: A Journal of Practice Theory, Vol. 28 No. 1, pp. 225-40.McCullough, B.D. and Vinod, H.D. (1993), “Implementing the single bootstrap: some computational considerations”, Computational Economics, Vol. 6 No. 1, pp. 1-15.
  22. 22. RAF Nwaeze, E.T. (1998), “Regulation and the valuation relevance of book value and earnings: evidence from the United States”, Contemporary Accounting Research, Vol. 15 No. 4,10,2 pp. 547-73. Ohlson, J.A. (1995), “Earnings, book values, and dividends in equity valuation”, Contemporary Accounting Research, Vol. 11 No. 2, pp. 661-87. Palmrose, Z. (1999), Empirical Research in Auditor Litigation: Considerations and Data,154 American Accounting Association, Sarasota, FL. Rajgopal, S., Venkatachalam, M. and Kotha, S. (2002), “The value-relevance of network advantages: the case of e-commerce firms”, Journal of Accounting Research, Vol. 41, pp. 135-62. Rees, W.P. (1999), “Influences on the value-relevance of equity and net income in the UK”, Managerial Finance, Vol. 25 No. 2, pp. 58-65. Sami, H. and Zhou, H. (2004), “A comparison of value relevance of accounting information in different segments of the Chinese stock market”, International Journal of Accounting, Vol. 39 No. 4, pp. 403-27. Schwartz, K.B. and Menon, K. (1985), “Auditor switches by failing firms”, The Accounting Review, Vol. 60 No. 2, pp. 248-61. Stober, T.L. (1999), “Empirical applications of the Ohlson (1995) and Feltham and Ohlson (1995, 1996) valuation models”, Managerial Finance, Vol. 25 No. 12, pp. 3-16. Teoh, S.H. (1992), “Auditor independence, dismissal threats, and the market reaction to auditor switches”, Journal of Accounting Research, Vol. 30, pp. 1-25. Teoh, S.H. and Wong, T.J. (1993), “Perceived auditor quality and the earnings response coefficient”, The Accounting Review, Vol. 68 No. 2, pp. 346-66. Further reading Dechow, P.M. (1994), “Accounting earnings and cash flows as measures of firm performance: the role of accounting accruals”, Journal of Accounting and Economics, Vol. 18, pp. 3-42. About the authors Li Dang is an Associated Professor at California Polytechnic State University-San Luis Obispo. Her research interests and publications focus on audit quality, international accounting, and accounting information systems. She teaches financial accounting and accounting information systems courses. Li Dang is the corresponding author and can be contacted at: Kevin F. Brown is an Associate Professor in the Department of Accountancy of the Raj Soin College of Business at Wright State University in Dayton, Ohio. He has published his research in several scholarly and professional journals, including Issues in Accounting Education, Journal of Accounting Education, and The CPA Journal. He has presented his research at numerous national conferences, including Annual Meetings of the American Accounting Association. His primary area of research is in auditing and his teaching areas include auditing and financial accounting. B.D. McCullough is a Professor of Decision Sciences at Drexel University. Before he joined the faculty at Drexel University, he was an Assistant Professor of Economics at Fordham University from 1989 to 1994 and a Senior Economist at the Federal Communications Commission from 1995 to 2000. He is the Associate Editor for five journals and has authored or co-authored more than 50 scholarly publications. To purchase reprints of this article please e-mail: Or visit our web site for further details: