Audit Market Response to PCAOB Censures of US Big 4 Firms

Abstract

Subsequent to the first-ever Public Company Accounting Oversight Board (PCAOB) censure of a US Big 4 firm (Deloitte) in December 2007, there were two other PCAOB US Big 4 firm censures as of 2016 year-end. We examine whether these two post-2007 PCAOB censures of US Big 4 firms conveyed new information to the audit market. For both censures, we find little or no evidence of any change in the factual audit quality of the censured firm over a three-year window surrounding the censure. Our findings suggest that the quality control deficiencies (identified during inspection of specific audit engagements) that triggered the PCAOB censure were isolated occurrences rather than systemic to the firm at large, i.e., the censures do not imply an impairment in the US Big 4 firm's overall factual audit quality. We also find that the negative response of investors and audit committees documented in prior research for the 2007 Deloitte censure disappeared for the later US Big 4 firm censures. Given that the PCAOB inspects (and can censure) non-US auditors who audit US-listed foreign companies, our findings are of potential interest to regulators, investors and audit committees outside the US.

Keywords:

• PCAOB
• Audit firm inspections
• Censures of US Big 4 firms
• US Big 4 audit quality

Acknowledgements

Authors' names are listed alphabetically. We thank the editor and two reviewers for their helpful comments and suggestions. We also thank workshop participants at The University of Texas at San Antonio and attendees at the AAA 2015 Annual Meeting. K. K. Raman acknowledges support from the Ramsdell Endowed Chair for Accounting at The University of Texas at San Antonio.

Notes

1 Consistent with PCAOB terminology, we refer to the auditor as the firm and the auditee as the client or company. Also, consistent with prior literature we use the terms audit firm and auditor interchangeably.

2 https://pcaobus.org/Enforcement/Pages/default.aspx.

3 As of 2016 year-end, the three US Big 4 firm censures were that of Deloitte in December 2007, EY in February 2012, and Deloitte in October 2013. A fourth US Big 4 firm censure (that of PwC) occurred in August 2017, a date too recent to include/examine in our study for data availability reasons. Given the size of the US Big 4 firms and their dominance of the US audit market, we focus on the Big 4 firms.

4 For the three censures we examine in our study, the largest monetary penalty was $2 million which is likely not material to the US Big 4 firms given their annual US revenues of $ 20 billion or more. Hence, for PCAOB censures to have the intended punitive effect (“real consequences” in PCAOB parlance), the censures need to be associated with reputational harm to the censured firm as reflected in lower audit fee growth rate and/or loss of clients.

5 Thus, from the start of PCAOB inspections in 2004 until the end of the time period of our study (31 March 2016), there were only three PCAOB censures of US Big 4 audit firms (Deloitte 2007, EY 2012, and Deloitte 2013). By contrast, during the same time period, there were 13 SEC enforcement actions against all the US Big 4 firms: PwC May 2004, KPMG October 2004, PwC January 2005, KPMG April 2005, Deloitte April 2005, KPMG October 2005, KPMG Feb 2006, EY March 2007, EY Aug 2008, EY Dec 2009, KPMG Jan 2014, EY July 2014, and Deloitte July 2015. Further, between March 31, 2016 (the end of our study period) and Dec 2017, there were 4 additional SEC enforcement actions against US Big 4 firms (EY Sept 2016, EY Oct 2016, KPMG Dec 2016 and KPMG Aug 2017) but only 1 additional PCAOB censure of a US Big 4 firm (PwC in 2017).

6 For clarity, we refer to SEC and PCAOB disciplinary actions as enforcement actions and censures, respectively.

7 The US Big 4 firms are inspected annually and have been found to have engagement-level audit deficiencies every year since the start of PCAOB inspections in 2004 as publicly disclosed in Part-I (the public portion) of the PCAOB inspection reports, albeit without identifying the client involved. Further, if the PCAOB inspectors identify firm-level quality control defects, the quality control criticisms are reported in Part-II (i.e., the non-public portion) of the inspection report. The Part II of the inspection report is subject to public release only if the audit firm fails to remediate to the satisfaction of the PCAOB. Given the 12 month window to remediate quality control criticisms as well as subsequent due process procedures, the time lag between the year of the inspection report and Part-II release can vary substantially.

8 On the PCAOB website, EY's first Part-II public release is labeled “2010_Ernst_Young” but 2010 refers to year of the inspection report. The first page of the document makes it clear that the Part-II public release is dated 23 May 2013. We verified information about this and other Part-II releases with PCAOB staff via email correspondence.

9 Once again, on the PCAOB website Deloitte's second Part-II public release is labeled “2009_Deloitte” but 2009 refers to the year of the inspection report. The first page of the document makes it clear that the Part-II public release is dated 21 November 2013.

10 For completeness, we note that PwC experienced Part-II releases in May 2009 and August 2010 but did not experience its first censure until August 2017 which suggests that the censure was triggered by a quality control deficiency not previously identified in a Part-II release. Also, KPMG has had two Part-II releases (both in October 2014) but no censures to date. In other words, Part-II public releases and censures do not appear to be necessarily related, i.e., a Part-II public release is not necessarily preceded by or followed by a PCAOB censure, and vice versa. To wit, it appears that some unremediated quality control defects trigger a Part-II release, while other quality control defects (potentially viewed by the Board as more “serious”) appear to trigger a censure (PCAOB 2011, p. 1). Underlining the fact that the Part-II releases and the censures relate to separate (i.e., unrelated) quality control defects, the Part-II releases and censure announcements do not refer to each other.

11 KPMG-Brazil was censured in March 2017. However, the Big 4 firms are network firms rather than a single legal entity, i.e., each national Big 4 firm is legally a separate and independent member of the firm's global network. We examine the sensitivity of our results to the use of KPMG as a comparison group and find that our inferences generally remain unchanged when PwC is used as a comparison group and also when the analysis is repeated without the use of a comparison group.

12 We replicate the BKR (2015) study with respect to the 2007 Deloitte censure to confirm that our findings are consistent with the prior study particularly since our comparison now is with KPMG rather than all the other US Big 4 firms.

13 Simunic (1980) suggests that the value of a US Big 4 audit derives from its two bundled components, assurance and insurance. However, as noted previously, the largest monetary penalty to date has been the $2 million imposed on EY and Deloitte as part of the February 2012 and October 2013 censures, and for US Big 4 firms (annual revenues of about $20 billion each), these monetary penalties are likely immaterial. Consequently, the monetary penalty imposed by the PCAOB censures likely had no impact on the insurance value of the censured firm's audits.

14 By contrast, we note that much more research has been done on the efficacy of PCAOB inspections for small auditors (i.e., audit firms with fewer than 100 public clients). Thus, Abbott, Gunny, and Zhang (2013), Gunny and Zhang (2013), and Gramling, Krishnan, and Zhang (2011) suggest that PCAOB inspections are able to distinguish actual audit quality among small auditors, and that audit quality for these small auditors (as measured by the propensity to issue going-concern opinions for distressed clients) improves subsequent to their being cited by the PCAOB for audit deficiencies. More broadly, DeFond and Lennox (2011) suggest that average audit quality for small auditors has increased post-SOX because the threat of PCAOB inspections drove nearly half the number of small auditors in the US out of the audit market for public company clients.

15 Similarly, Christensen, Glover, Omer, and Shelley (2014) quote a sophisticated investor in a management role as noting that auditors are performing more audit work (and carefully documenting the additional work) not to increase audit quality but to demonstrate compliance out of fear of the PCAOB.

16 Similar arguments can be made for gains due to new clients. For brevity, we do not formally state a hypothesis for client gains. However, we report the results for client gains in section 3.6.3.

17 PwC is a less desirable comparison group because the firm was censured by PCAOB in August 2017, whereas KPMG, while subject to enforcement action by the SEC, remains uncensured by the PCAOB. Non-Big 4 firms are also a less desirable comparison group because of well-documented differences in pricing and audit quality vis-à-vis the Big 4.

18 To test the robustness of our results, we also measure audit quality using going-concern audit reports as well as the number of accounting issues per 10-K raised in SEC comment letters and the length of time taken to close the comment letter conversation. Results are discussed in Section 3.6 below.

19 ABSDACC is the absolute value of the difference between total accruals and the fitted normal accruals estimated by using the following modified Jones (1991) model by fiscal year and two-digit industry SIC code: $\frac{T{A}_{it}}{Asset{s}_{i,t-1}}=\alpha \frac{1}{Asset{s}_{i,t-1}}+{\beta }_1\frac{\mathrm{\Delta }SALE{S}_{it}-\mathrm{\Delta }A{R}_{it}}{Asset{s}_{i,t-1}}+{\beta }_2\frac{PP{E}_{it}}{Asset{s}_{i,t-1}}+{\beta }_3\frac{I{B}_{it-1}}{Asset{s}_{i,t-1}}+{\epsilon }_{i,t},$ where TA is total accruals calculated as income from continuing operations less operating cash flows from continuing operations, ΔSALES is change in sales revenue, ΔAR is the change in accounts receivables, PPE is gross property and equipment, IB is income before discontinued operations and extraordinary items, and the subscripts i and t denote client and year, respectively. Asset deflated income, or ROA, is included to performance-adjust abnormal accruals as suggested by Kothari et al. (2005).

20 We omit the control variable ISSUE to avoid loss in sample size arising from the use of leading (t+1) values required to operationalize the variable. BKR (2015) define ISSUE as 1 if external financing (stock and long-term debt issues) during the current and following fiscal year exceeds 5% of the current year ending total assets; 0 otherwise. BKR (2015) did not find ISSUE to be statistically significant (p > .10) in the accruals model. We also omit the control variable AUDIT_FEE because it induces severe multicollinearity in our sample. Specifically, when AUDIT_FEE appears as a control variable in the model, VIF on SIZE is around 20 and the VIF on AUDIT_FEE is around 25; when AUDIT_FEE is omitted as a control variable, VIF on SIZE is around 2.3 and the largest VIF in the model is around 2.6. We note that when AUDIT_FEE is included as an explanatory variable, the variable loads with an insignificant coefficient in each of the Table 3 Panel B regressions and the difference-in-differences estimates are as follows. For the first Deloitte censure, the difference-in-differences estimate is 0.001 with a t-statistic of 0.28 in the model with AUDIT_FEE versus an estimate of 0.002 and a t-statistic of 0.29 in the model without AUDIT_FEE. For the EY censure, the difference-in-differences estimate is −.001 with a t-statistic of −0.18 in the model with AUDIT_FEE versus an estimate of −.001 with a t-statistic of -0.18 in the model without AUDIT_FEE (i.e., results are unchanged after the coefficient estimate and t-statistic are rounded to 3 and 2 digits, respectively). For the second Deloitte censure, the difference-in-differences estimate is 0.007 with a t-statistic of 1.55 in the model with AUDIT_FEE versus an estimate of 0.007 with a t-statistic of 1.55 in the model without AUDIT_FEE (i.e., results are unchanged after the coefficient estimate and t-statistic are rounded to 3 and 2 decimal places, respectively).

21 To remain consistent with Dee et al. (2011), we use clients of the other US Big 4 firms as an event period comparison group.

22 The Sefcik and Thompson (1986) weighted return approach is similar to a cross-sectional regression of abnormal returns on clients’ characteristics at each event date. However, the Sefcik and Thompson (1986) approach accounts for cross-sectional heteroscedasticity and cross-correlation of the residuals, providing a correct estimate of the standard error.

23 Zmijewski's financial distress score is measured as −4.3–4.5*(Net income/Total assets) + 5.7(Total liabilities/Total assets) − 0.004(Current assets/Current liabilities).

24 Clients may dismiss an auditor in anticipation of a resignation to avoid the negative reaction by the market or regulators. Moreover, auditors may force a client dismissal by increasing fees to an unrealistic level for that particular client. They also may prefer to be dismissed because a resignation could harm their ability to attract new clients. In other words, the real reason for the auditor change could be different from the reason provided.

25 EY sustained a client loss of .01292 while second-censure Deloitte sustained a client loss of .01296, which rounded yields a loss rate of 13 per 1000 clients.

26 As discussed previously, the learning explanation is that the audit market learnt that PCAOB censures of US Big 4 firms are unrelated to the factual audit quality of the censured firm. Although it would be reasonable to expect no change in the fee growth rate for EY following its censure, we believe that it is not inconsistent with the learning explanation for EY to experience the fee growth we see. In other words, the audit market may have learnt to ignore the PCAOB censure (because the censure is unrelated to EY's factual audit quality) but it is still possible for EY's fee growth rate to increase for reasons unrelated to the censure.