What is the Effect of the PCAOB Part II Inspection Report Disclosure of Income Tax. Deficiencies on Auditor-Provided Tax Services?

Transcription

1 What is the Effect of the PCAOB Part II Inspection Report Disclosure of Income Tax Deficiencies on Auditor-Provided Tax Services? Jaehan Ahn Price College of Business University of Oklahoma Herita Akamah College of Business Administration University of Nebraska Lincoln Sydney Shu College of Business Administration San Diego State University September 2016 Acknowledgement: We thank Kecia Smith and Nathan Goldman for their insightful feedback and William Kinney for his institutional material.

2 What is the Effect of the PCAOB Part II Inspection Report Disclosure of Income Tax Deficiencies on Auditor-Provided Tax Services? Abstract: We examine whether quality control deficiencies in auditing income tax accounts publicly released in Deloitte s 2007 PCAOB Part II report impose costs on the accounting firm s auditor-provided tax services (APTS). We document that audit clients reduce reliance on APTS upon the public disclosure of Deloitte s income tax-related audit deficiencies. This suggests that publicly disclosed audit deficiencies in the Part II report have a spillover effect on the inspected firm s non-audit services. We also find that the decision of terminating APTS is predominantly evident in clients who rely more on Deloitte s non-audit, non-tax services, and in those who excessively increase deferred tax valuation allowance and reserve for uncertain tax benefits. This is consistent with our expectation that reduced demand for Deloitte s APTS is driven by stakeholders concern about independence impairment and by clients concern about APTS quality. In additional analyses, we find that Deloitte s audit clients are less likely to retain APTS when the auditors lack sufficient tax expertise. We also find that PCAOB-identified income tax-related audit deficiencies are associated with reduced tax fees for clients that choose to retain APTS. Additionally, we document that reported earnings and income tax expenses are more value relevant for Deloitte s audit clients that drop APTS in response to PCAOB-disclosed deficiencies in auditing income tax accounts. Collectively, our results are consistent with audit clients responding to the PCAOB Part II reports that are publicly disclosed, suggesting that the PCAOB inspection on overall audit quality control has informational value. Specifically, this inspection report incentivizes audit clients to reconsider the benefits/costs of auditor-provided tax services and influences investor perception of financial reporting quality. Keywords: Data: PCAOB Part II Inspection Report; PCAOB; auditor-provided tax service; value relevance of income tax accounts; valuation allowance; uncertain tax benefits. All data used in the study are available from the sources cited in the text.

3 What is the Effect of the PCAOB Part II Inspection Report Disclosure of Income Tax Deficiencies on Auditor-Provided Tax Services? 1. Introduction In the wake of a notable string of financial reporting failures in the 1990 s and 2000 s, Congress empowered the Public Company Accounting Oversight Board (PCAOB) to oversee audit firms, thereby replacing self-regulation with independent inspections. In light of this major shift in the auditing regulatory regime, several parties are interested in and have called for research on the impact of PCAOB inspections (e.g., Nelson 2006; DeFond 2010). For example, DeFond (2010) emphasizes the need for future research to study the consequences of the new PCAOB inspections. Prior research responds to this call through examining the direct effect of PCAOB inspections on audit quality and/or financial reporting quality (e.g., Lennox and Pittman 2010; Church and Shefchik 2011; Abbott, Gunny, and Zhang 2013; Lamoreaux 2016). We complement and extend this literature by exploring whether PCAOB-identified quality control deficiencies in auditing income tax accounts have a spillover effect on the inspected audit firm s non-audit (tax) services. Also, by examining the effect of PCAOB-induced changes in auditor-provided tax services (APTS) on the value relevance of earnings, we are also the first to explore an indirect effect of PCAOB inspections on financial reporting quality. Under Section 104 of the 2002 Sarbanes-Oxley Act (SOX), the PCAOB oversees the quality of external audits through conducting independent periodic inspections on public audit firms. 1 The inspection report generated by the PCAOB includes a portion (Part II) that identifies deficiencies over the audit firm s overall quality control processes. The PCAOB first issues the 1 The inspections are performed annually for audit firms that issue audit reports for more than 100 public firms and triennially for firms that issue audit reports for 100 or fewer public firms (PCAOB 2011). In this study we focus on annually inspected audit firms given the substantial differences between annual firms and triennial firms. 1

4 Part II report privately to the audit firm and will not publicly release this report unless the audit firm fails to adequately remediate the quality control deficiencies during a one-year remediation window after the initial issuance. In this study, we focus on Deloitte & Touche LLP s 2007 PCAOB Part II inspection report. 2,3 Deloitte s 2007 Part II report is different from reports issued to other annually inspected firms in that this report explicitly addresses the PCAOB s concern about Deloitte s audit procedures performed on income tax accounts. Our study is different from extant research on Deloitte s 2007 PCAOB Part II report in that we examine whether Deloitte s audit quality control weaknesses impose costs associated with Deloitte s non-audit services. Using this unique setting, we investigate whether income tax-related deficiencies disclosed in the Part II report impose costs related to APTS on Deloitte relative to other annually inspected audit firms over a four-year period around the disclosure. More specifically, we examine whether the PCAOB report against Deloitte s audit related to income tax accounts leads to a decrease in demands for Deloitte as a tax service provider among clients that retain Deloitte as their external auditor. Among various influences arising from the PCAOB s Part II report against Deloitte, we explore two potential considerations that likely motivate Deloitte s audit clients to curtail APTS. 4 First, the public disclosure of the Part II report could heighten stakeholders concern over independence impairment associated with the joint provision of Deloitte s audit and tax services. Creating economic bonding between the client and the auditor, APTS has been perceived as 2 Deloitte s 2007 Inspection Report refers to the report on the inspection conducted between March 2007 and November 2007, and was issued by the PCAOB to Deloitte privately on May 19, The PCAOB then determined that Deloitte failed to remediate the specified quality control issues and publicly released the Part II report on the PCAOB website on October 17, Throughout the paper, we refer to Deloitte & Touche LLP s 2007 PCAOB Part II inspection report as Deloitte s 2007 Part II report or the Part II report for brief. 4 It is worth to note the two considerations potentially driving Deloitte audit clients ATS decisions in our study are not mutually exclusive and are not necessarily the only two motives. Instead, we focus on these two factors given concerns over independence in appearance as well as corporate tax avoidance associated with ATS (Schneider, Church, and Ely 2006; Klassen, Lisowsky, and Mescall 2016). 2

5 impairing auditors independence and objectivity (Dopuch, King, and Schwartz 2003; Krishman, Sami, and Zhang 2005; Mishra, Raghunandan, and Rama 2005). The disclosure of unresolved quality control issues related to income tax audits will likely elicit stakeholders adverse perception of Deloitte s APTS. As such, we expect that income tax-related audit deficiencies revealed in the Part II report would motivate Deloitte s audit clients to reduce reliance on APTS in order to bolster the appearance of auditor independence and alleviate stakeholders concerns. Second, the PCAOB s Part II report could negatively influence clients perception of the quality of Deloitte s tax service provision. Deloitte was notified that the 2007 report would be publicly issued after the PCAOB determined that Deloitte had not adequately remediated quality control deficiencies including those related to income tax accounts at the end of the remediation window. Being aware of such negative information, Deloitte increased scrutiny over income tax accounts (Drake, Goldman, and Lusch 2016). Accordingly, we suggest that the tax team may become more risk-averse when recommending tax positions to the client, given the economic bonding created by APTS. For instance, Deloitte s APTS may prompt clients to forego riskier cash saving opportunities that would have been recommended by a non-audit tax serivce provider. Hence, we expect that the Part II report would cause clients to question the quality of Deloitte s tax service, thus leading them to curtail Deloitte s APTS. Nevertheless, it is not ex ante clear that the PCAOB Part II report will reduce audit clients reliance on Deloitte s APTS. The joint provision of audit and tax services has been documented to have a knowledge spillover effect, which enhances audit efficiency and improves actual audit quality (Kinney, Palmrose, and Scholz 2004; Gleason and Mills 2011). Decoupling audit and tax services and subsequently acquiring a new tax service provider entails forgoing such knowledge spillover benefits created by APTS. In particular, the new (non-audit) tax service provider may 3

6 lack familiarity with the client s tax planning opportunities and tax saving strategies (Cook and Omer 2013). Also, the PCAOB s inspectors have been criticized as lacking expertise (Glover, Prawitt, and Taylor 2009; DeFond 2010). Consequently, clients may not consider the Part II report consequential and thus may not find the information in the report useful in choosing tax service providers. To examine whether income tax-related audit deficiencies identified in Deloitte s 2007 Part II report affect the clients decision on APTS, we adopt a difference-in-differences design, with a sample period of two years before and two years after the release date of this report (i.e., October 11, 2011). We document that the likelihood of retaining Deloitte as a tax service provider is 14.2 percent lower than that of other annually inspected audit firms upon the public disclosure of Deloitte s 2007 Part II report. Moreover, our findings are mainly explained by an increase in APTS termination (rather than by a decrease in APTS initiation). In particular, we find that the likelihood of terminating Deloitte as a tax service provider relative to other audit firms is 54.5 percent higher after the Part II report is publicly issued. These results suggest that audit clients respond to income tax-related quality control deficiencies publicly disclosed by the PCAOB through reducing their demand for APTS. Moreover, the results with regard to the two potential explanations for the decreased reliance on Deloitte s APTS are consistent with our expectations. First, we follow prior research and measure stakeholders perception about impaired independence using the ratio of non-audit non-tax fees to total audit fees. We find that the decision of dropping Deloitte s APTS is more pronounced among audit clients with greater concerns about impaired independence. The finding suggests that the perceived costs of retaining Deloitte as a tax service provider are incrementally higher when the auditor s independence is more questionable. As such, Deloitte s clients terminate 4

7 APTS in order to restore stakeholders confidence in the auditor s independence. Second, we examine client concerns over the quality of Deloitte s APTS arising from increased scrutiny in auditing of income tax accounts. We find that following the public release of the Part II report, the propensity of terminating Deloitte s APTS is more evident among audit clients experiencing excessive increases in the valuation allowance and reserve for uncertain tax benefits. This is consistent with that Deloitte becomes more conservative in clients tax reserves after deficiencies in auditing income tax accounts identified by the PCAOB, heightening clients concerns about foregone tax-saving opportunities, thereby incentivizing audit clients to curtail APTS. In additional analyses, we document that the reduced reliance on Deloitte s APTS following the public release of the Part II report is more pronounced when the audit office has limited tax expertise. These results suggest that Deloitte s audit clients perceive the costs of terminating the tax services of an auditor with little tax expertise to be relatively lower. Also, we investigate the implications for APTS fees among audit clients choosing to retain APTS. We document lower tax fees for Deloitte s audit clients following the public disclosure of income tax audit deficiencies in the Part II report. 5 In addition, we explore whether changes in APTS following the Part II report provide incremental informational value about tax and non-tax accounts. We find that income tax expenses and after-tax earnings become more value relevant after the public disclosure of the Part II report for Deloitte s audit clients who drop APTS. These results imply that investors perceive the decoupling of audit and tax services to improve the reporting quality of tax and non-tax earnings components. Finally, our results are robust to alternative model specifications and a series of falsification tests. 5 Prior to the release of Deloitte s Part II report, tax fees had declined owing to a disciplinary order against Deloitte by the PCAOB in 2007 (Boone, Khurana, and Raman 2015). Our results of tax fees are noteworthy because they suggest that despite the already low tax fees, the PCAOB disclosure of Deloitte s income tax related quality control deficiencies further depressed tax fees. 5

8 Our study makes several important contributions. First, we contribute to the growing literature on the PCAOB s role in inspecting audit firms. 6 This literature has primarily focused on the effect of PCAOB inspections on audit and financial reporting outcomes (e.g., Abbott, Gunny, and Zhang 2013; Drake et al. 2016). Instead, we provide new evidence on the spillover effect of PCAOB identified audit-related deficiencies on the market for tax services. In particular, our study tests the conjecture in Lennox and Pittman (2010) that clients would respond to PCAOB inspection disclosures if these disclosures provide information about auditors quality control rather than engagement-specific deficiencies. Our findings suggest that clients indeed respond to PCAOB reports on auditors quality control deficiencies by altering their demands for auditor-provided non-audit services. That is, through specifying quality control problems related to the auditing of income tax accounts, the Part II report prompts clients behavior change with respect to their tax service provider choices. Overall our study should be informative to a variety of parties such as regulators, standard setters, and audit committees given their strong interests in the efficacy as well as economic impacts of the PCAOB s inspection results (Compton et al. 2016; Grasso et al. 2016). 7 Second, we expand the scope of research on auditor-provided non-audit services and in particular, APTS. As one of the non-audit services that are still permitted after the SOX, fees generated from APTS represent an important non-audit revenue source for public accounting firms (Cook 2013). Prior research in general concludes that clients consider the benefits arising from APTS (e.g. knowledge spillover effect) to outweigh the costs (e.g. perceived independence impairment) (e.g., Kinney, Palmrose, and Scholz 2004; Paterson and Valencia 2011; Gleason and 6 Abernathy, Barnes, and Stefaniak (2013) provide a comprehensive review of research on the PCAOB s role in supervising the audit profession. 7 For instance, the immediate past Chairman of the FASB, Leslie Seidman, emphasized during the 2016 AAA Annual Meeting that audit committees are strongly interested in the quality control deficiencies identified in the PCAOB s inspection report. 6

9 Mills 2011). 8 In contrast, our study explores a unique setting where the PCAOB report on income tax audit deficiencies likely amplifies the costs associated with APTS. Particularly, the evidence presented in our study suggests that this report elicited concerns about auditors compromised independence and about the quality of APTS. These in turn impose costs on auditors non-audit services in terms of losing tax service market shares. Hence, we offer new insights into the tradeoff audit clients make when making decisions related to APTS. Third, we add to the emerging literature examining market reactions to PCAOB-identified deficiencies (Dee, Lulseged, and Zhang 2011; Offermanns and Peek 2011; Robertson and Houston 2010). Abernathy et al. (2013) in their PCAOB research review, specifically call for more work on the repercussions of the Part II report. We answer the call through investigating the effect of the Part II report on the value relevance of tax and non-tax earnings components. In particular, we provide evidence that PCAOB-induced curtailing of APTS improves the value relevance of accounting information. These results stand in stark contrast to the results documented for non- PCAOB Part II APTS switches (e.g., Krishnan, Visvanathan, and Yu 2013). Additionally, the results indicate that PCAOB-induced termination of APTS has an indirect effect on financial reporting quality, which operates through its influence on investors perception of the overall quality of audited financial statements. Collectively, these results suggest that the relation between APTS and financial reporting quality (as perceived by investors) is dynamic, varying with events that alter investors tradeoff between the spillover benefits and costs of APTS. 9 8 In fact, the professional code of conduct in audit has long recognized the importance of perception management and point out that the perception on independence impairment is as serious as the factual impairment of independence (Francis 2006). 9 At the 2016 AAA Annual Meeting, Kevin Stout from the Office of the Chief Accountant pointed out that the SEC highly values academic research providing data-driven evidence with respect to the economic and capital market impact of the PCAOB s inspection reports. 7

10 This study is organized as follows. The next section provides background information on Deloitte s 2007 PCAOB Part II report and APTS, followed by hypothesis development. Section 3 describes the research design and sample selection. Empirical results are provided in Section 4, and Section 5 concludes. 2. Background and Hypotheses 2.1. Deloitte s 2007 PCAOB Part II Inspection Report The inspection report issued by the PCAOB has a portion (Part I) that describes the audit firm s deficiencies identified at the engagement level inspection using a risk-based approach. 10 Prior research on the PCAOB inspection process has primarily focused on this Part I report (e.g., Abbott, Gunny, and Zhang 2013; Church and Shefchik 2011; DeFond and Lennox 2011; Lamoreaux 2015; Lennox and Pittman 2010). 11 For example, Lennox and Pittman (2010) show that the deficiencies revealed in the Part I report lead to no subsequent auditor switches or changes in audit market shares. They conclude that the Part I report is not informative to investors because it contains no information about the audit firm s overall audit quality. In addition to Part I, the PCAOB s inspection report generally contains a portion (Part II) that discusses deficiencies related to the audit firm s overall quality controls (Center for Audit Quality 2012). 12 The PCAOB issues the Part II report to the audit firm privately at the time the Part I report is issued and provides a 12-month remediation window for the auditor to address quality control issues noted in the Part II report. To the extent that the PCAOB determines that the 10 The presence of an engagement-level deficiency does not necessarily indicate that the audit firm s overall quality control systems have deficiencies. 11 Part I reports are publicly available within a few months after the PCAOB completes the inspection. 12 Upon completion of the inspection, the PCAOB issues the Part II report only if they identify deficiencies in the audit firm s quality control systems 8

11 audit firm does not remediate the deficiencies within the one-year remediation period, the Board will then disclose the deficiencies on their website to the public. There has been a growing interest in the implications of the Part II report. For example, Buslepp and Victoravich (2014) study triennially inspected firms and document that the Part II report leads to a change in auditors. Johnson (2015) shows that upon issuance of the Part II report, audit firms experience reputational damage that leads to lower audit fees. Nagy (2014) examines the change in audit firms market shares following the public disclosure of Part II of the inspection report. He finds that audit firms with an unfavorable Part II report lose a significant amount of market share, consistent with clients viewing the report as a signal of poor audit quality. It is worth noting that a few studies specifically focus on Deloitte s 2007 PCAOB Part II report. Muriel (2013) finds that upon the public disclosure of Deloitte s 2007 PCAOB Part II report, only clients in highly litigated industries and those with higher auditor tenure are more likely to dismiss Deloitte as the external auditor. She finds no evidence that investors react to the public release of Part II. Drake, Goldman, and Lusch (2016) investigate Deloitte s clients financial reporting for income taxes after the PCAOB determined that Deloitte had not successfully remediated audit quality control deficiencies related to income tax accounts. They provide empirical evidence consistent with clients increasing reported valuation allowance on deferred tax assets and reserve for uncertain tax benefits in response to Deloitte s heightened audit scrutiny over income tax accounts. Our study is different from concurrent research in that we examine whether the PCAOB s Part II report on Deloitte s audit quality controls with respect to income tax accounts impose costs with respect to Deloitte s non-audit services. We focus on one type of nonaudit service, tax, because its link with tax-related audit deficiencies is tighter. 9

12 2.2. Auditor-provided tax services and perceptions of independence Auditor-provided tax services (APTS) are one of the few non-audit services that are still permitted after the enactment of the SOX. In fact, most companies continue using APTS in the post-sox period, and tax fees represent the most important non-audit revenue for audit firms (Francis 2006). Prior research shows that investors may perceive the joint provision of audit and non-audit services as impairing auditor independence given their concern about the bond engendered by auditors non-audit fees (Dopuch, King, and Schwartz 2003; Krishman, Sami, and Zhang 2005; Mishra, Raghunandan, and Rama 2005). In particular, Mishra et al. (2005) find that tax service fees are positively related to the proportion of shareholders not voting for auditor ratification. Lassila, Omer, Shelly, and Smith (2010) document that perceived auditor independence (i.e., independence in appearance) reduces clients reliance on APTS. On the other hand, APTS have been documented to provide knowledge spillover benefits that offset the costs arising from the economic bonds created. That is, insights learned from providing tax planning and compliance services can enhance audit effectiveness via better communication between the audit team and the tax team, thus mitigating information asymmetry. Empirical evidence suggests that APTS incentivize such knowledge spillover effect, thereby improving actual audit quality and, by extension, financial reporting quality (e.g., Kinney et al. 2004; Paterson and Valencia 2011; Gleason and Mills 2011; Cook and Omer 2013). For example, Kinney et al. (2004) document a negative relation between APTS fees and accounting misstatements. Cook and Omer (2013) find that non-tax discretionary accruals and the likelihood of restatements do not change after decoupling audit and tax service provision, suggesting that APTS do not result in economic bonding that degrades financial statement quality. 10

13 2.3. Hypotheses We investigate whether the PCAOB Part II report on Deloitte s audit deficiencies related to income tax accounts significantly decrease audit clients reliance on Deloitte as their tax service provider. While the PCAOB has publicly issued up to ten Part II reports for annually inspected audit firms as of year 2015, the uniqueness about the 2007 Deloitte s Part II is that the PCAOB explicitly identifies the auditor s quality control weaknesses attributed to auditing on income tax accounts. In other words, Deloitte s 2007 Part II report is the only report pertaining to the PCAOB s concerns about the firm s audit procedures of income tax accounts. This unique feature provides a proper setting to examine non-audit costs arising from the PCAOB inspections on audit quality control and in particular, costs related to tax service provision. In this study, we are interested in whether the PCAOB s Part II report against Deloitte leads to a decrease in the accounting firm s APTS. Specifically, given the auditor s quality control problems related to income tax accounts revealed by the PCAOB, we explore two considerations that could potentially discourage Deloitte s audit clients from employing APTS. First, audit clients declined reliance on Deloitte s tax services could be driven by stakeholders heightened concern about APTS as a threat to auditor objectivity. APTS increase the auditor s economic dependence on the client and can be perceived as impairing auditor independence (Mishra et al. 2005; Dopuch et al. 2003; Krishman et al. 2005). 13 This suggests that stakeholders perception of income tax audit will likely influence their confidence about the overall audit quality (Cook and Omer 2013). Consequently, deficiencies in auditing income tax accounts identified in the Part II report will likely elicit stakeholders concern over compromised auditor 13 SEC Rule S defines independence as a mental state of objectivity and lack of bias and emphasizes the importance of maintaining independence in appearance (Securities and Exchange Commission 2000). 11

14 independence and cause greater doubt about the financial reporting quality of firms that use Deloitte s joint audit and tax services. In addition, with more interaction opportunities provided by APTS, partners on the audit team are typically involved in tax services provided by the tax team (De Simone, Ege, and Stomberg 2015; Gleason and Mills 2011). As such, not only could revealed quality control problems damage Deloitte s reputation for auditing income tax accounts, but these deficiencies could also undermine its reputation for tax service provision, thereby impeding stakeholders perceived quality of Deloitte clients tax compliance and planning activities. 14 Therefore, we expect that identified deficiencies related to income tax audit in the PCAOB Part II report will motivate Deloitte s audit clients to reduce reliance on APTS, in order to bolster the appearance of audit independence and alleviate stakeholders adverse perception. Second, a reduced demand for Deloitte s APTS could be driven by their audit clients concern over the quality of the auditor s tax service. As discussed earlier, Deloitte has a 12-month remediation window after the PCAOB identifies and issues the Part II report privately. Once the PCAOB deems that Deloitte has not adequately remediated quality control deficiencies including those related to income tax audit, the report becomes publicly available and Deloitte will be notified of such public disclosure. Drake et al. (2016) document that Deloitte s failed remediation of quality control deficiencies related to auditing of income tax accounts elicits audit scrutiny over clients financial reporting of tax-related accounts. This suggests that Deloitte has a stronger incentive to rectify its income tax-related quality control issues and restore its reputation, leading to greater skepticism and more diligent attention to the income tax accounts of clients. 14 Tarnished reputation leads to a decreased demand for services (Klein and Leffler 1981; Wang, Lo, and Hui 2003). 12

15 To maximize profitability, a firm is expected to pursue all opportunities that reduce tax liabilities, as long as the tax positions offer incremental benefits (Slemrod 2004). Nevertheless, with more interaction opportunities created by APTS, Deloitte s tax team may accede to the audit team s wishes and become considerably cautious when proposing tax positions to clients. In other words, given failed remediation of quality control deficiencies over income tax audit, Deloitte s tax team will likely prompt clients to forgo riskier, firm value enhancing, and cash saving opportunities that would have been recommended by a non-audit tax service provider (either internal or external to the client). Hence, we expect that heightened audit scrutiny over income tax accounts induced by the PCAOB s Part II report will cause audit clients to doubt the quality of Deloitte s tax service and reduce their reliance on Deloitte as their tax service provider. Nevertheless, it is ex ante unclear as for whether quality control deficiencies on income tax audit identified in the PCAOB s Part II report will affect audit clients decisions related to Deloitte s APTS. Empirical evidence suggests that APTS provide the auditor with superior knowledge that enhances audit effectiveness and thus improves the quality of the audited financial statements (Kinney et al. 2004; Gleason and Mills 2011). When clients dismiss APTS to avoid the appearance of compromised independence, they forgo the knowledge spillover benefit. Even if the incoming tax service provider has a similar level of expertise, it still requires time to understand the client s tax saving opportunities and design/implement corresponding strategies. Also, the shift from old AICPA peer reviews to the new PCAOB inspection represents a trade-off of expertise for independence (DeFond 2010; Kinney 2016). In fact, prior research has questioned the efficacy of the PCAOB inspection process (e.g., Glover, Prawitt, and Taylor 2009; Lennox and Pittman 2010). Consequently, the clients and/or their stakeholders may perceive PCAOB s Part II report as an inconsequential, noisy signal of Deloitte s audit quality. 13

16 Based on the preceding discussion, it remains an open empirical question as for whether income tax-related deficiencies disclosed in Deloitte s 2007 PCAOB Part II report impose costs with regard to the auditor s APTS. Our first hypothesis therefore is stated in the null form as follows: H1: The PCAOB Part II Report quality control deficiencies related to the auditing of income tax accounts do not affect audit clients reliance on Deloitte s APTS. To examine the two potential channels that discourage audit clients from using Deloitte s APTS given the quality control deficiencies on income tax audit, we state the two parts of our second hypothesis in the directional form: H2a: The effect of quality control deficiencies related to the auditing of income tax accounts on audit clients reliance on Deloitte s APTS is more pronounced when stakeholders have greater concern about impaired auditor independence. H2b: The effect of quality control deficiencies related to the auditing of income tax accounts on audit clients reliance on Deloitte s APTS is more pronounced when clients have greater concern about APTS quality. 3. Research Design 3.1. Model specification Estimation period The first time the public became aware of Deloitte s audit performance issues related to income taxes is October 17, Since we are interested in understanding the effect of changes in stakeholders perception around this PCAOB disclosure, we study the four-year period around 14

17 this disclosure date. That is, the period before (after) the event is fiscal years 2009 and 2010 (2011 and 2012). We carefully determine the time period for the following reasons. First, a post-period of one year may not sufficiently capture clients behavior. Specifically, clients are bound by contracts that are in place as of the disclosure date. Hence, these clients might not have the ability to switch their tax service provider within the first fiscal year after the disclosure date. Second, we choose two years before and after the event to balance the sample period. Third, we do not extend the period beyond two years because the PCAOB disclosed another Deloitte inspection report on November 21, Thus, our choice of two years avoids contamination arising from this subsequent event Likelihood of Retaining the Auditor as the Tax Service Provider (H1) Our first hypothesis examines whether Deloitte s audit clients are less likely to retain their principal auditor as their tax service provider after the disclosure of Deloitte s income tax related deficiencies in the PCAOB Part II report. To test this hypothesis, we restrict our sample to companies who do not switch auditors throughout our sample period (i.e., ) and estimate the following logistic regression model: APTS i,t = α 0 + α 1DT i,t + α 2POST + α 3DT i,t POST + β ncontrols i,t + Ԑ i,t. (1) In equation (1), APTSi,t equals one if the client uses its principal auditor as its tax service provider in year t, zero otherwise. DTi,t equals one if the client s principal auditor is Deloitte in year t, zero otherwise. POST equals one if the fiscal year is ending after October 17, 2011, and 15

18 zero, otherwise. Our variable of interest is the interaction term (DTi,t POST) in equation (1). Consistent with H1, we expect a negative coefficient, α3, on this interaction term. 15 To control for other determinants of the propensity to retain the principal auditor as the tax service provider, we select a set of control variables following Omer et al. (2006), Milles et al. (1998), and Lassila et al. (2010). These variables control for complexity (SIZE and LEV); audit tenure (TENURE); tax avoidance (ETR and TAXAVOID); merger and acquisition activity (MA); foreign operations (FORIN); net operating loss carryforward (NOL); capital intensity (CAP); perceived auditor independence (NTAFEE); and audit fees (AFEE). We also control for industrywide and economy-wide determinants using industry (based on two-digit SIC codes) and year fixed effects, respectively. We winsorize all continuous variables at the 1% and 99% level to reduce the effect of extreme values and consistent with Boone, Khurana, and Raman (2015), we cluster the standard errors by auditor Reputational Concerns from Using the Auditor as the Tax Service Provider (H2a) Recall from our main arguments that Deloitte s clients might have a stronger incentive to terminate auditor provided tax services when they would otherwise suffer reputational damage. Our measure of reputational concerns is based on the idea that external stakeholders can perceive tax-related deficiencies as a signal of Deloitte s impaired independence. Consistent with Lassila et al. (2010), we measure perceived independence impairment using the ratio of non-audit non-tax fees to total audit fees in year t-1 (lagged NTAFEE). We consider clients with above the median 15 While Ai and Norton (2003) demonstrate that the coefficient on the interaction term in a nonlinear model might have an incorrect sign, Puhani (2012) demonstrates that this concern is alleviated in the difference-in-differences model that we use in our paper. Hence, we do not make any adjustments to address this concern in our model. 16 Our inferences are unchanged when we instead cluster the standard errors by client (i.e., the coefficient on the test variables of interest have a sign and significance level that yields the same inferences as those reported in the paper). 16

19 of lagged NTAFEE as those with high reputational concerns, and create HNTAFEE equals 1 (0 otherwise) for clients with lagged NTAFEE above the annual median. We estimate equation (1) separately within sub-samples of clients with high and low reputational concerns. Consistent with H2a, we expect the negative coefficient on the interaction term (DTi,t POST) in equation (1) to be more pronounced when HNTAFEE equals Tax Quality Concerns from Retaining the Auditor as the Tax Service Provider (H2b) Our measure of tax quality concern is based on excessive auditor conservatism in tax service provision. Clients may believe that Deloitte will become excessively conservative in its provision of tax services following the negative PCAOB report. Specifically, in order to avoid additional negative PCAOB inspection reports, Deloitte might pass up opportunities to generate legitimate tax savings, which is undesirable from the perspective of Deloitte s clients. This perspective is consistent with Maydew and Shackelford (2007). To capture this concept of excessive conservatism in tax service provision and consequently high tax quality concerns, we use abnormal changes in the deferred tax valuation allowance (HIGHVA) and reserve for uncertain tax benefits (HIGHUTB). In particular, we use data on the valuation allowance from the 10-K tax footnotes. We then calculate HIGHVA which equals 1 (0 otherwise) for clients with: (1) changes in the valuation allowance from year t-2 to t-1 above the industry median (two-digit SIC code) and (2) year t-1 valuation allowance (scaled by assets) above the industry median (two-digit SIC code). Also, we compute HIGHUTB which equals 1 (0 otherwise) for clients with: (1) changes in reserve for uncertain tax benefits from year t-2 to t-1 above the industry median (two-digit SIC code) and (2) year t-1 reserve for uncertain tax benefits (scaled by assets) above the industry median (twodigit SIC code). We estimate equation (1) separately within sub-samples of clients with high and 17

20 low tax quality concerns. Consistent with H2b, we expect the negative coefficient on the interaction term (DTi,t POST) in equation (1) to be more pronounced when HIGHVA (HIGHUTB) equals one Sample selection We merge Audit Analytics with Compustat, yielding an initial sample of 24,092 clientyear observations over the period We further exclude observations with missing APTS fees in year t and t-1 and with missing control variables in main model (equation 1). Also, to better isolate switching risk related to only tax services from an overall switching risk, we exclude clients that switch auditors during our sample period, and clients that show up or disappear during our sample period. Finally, to hold audit firms incentives/pressure under PCAOB s scrutiny constant, we restrict our sample to the 9 audit firms that are subject to the PCAOB s annual inspection (Drake et al. 2016), resulting in 10,118 client-year observations for testing H1 and H2a. 17 To examine the effect of excessive conservatism in tax service provision (H2b), we collect data on the valuation allowance from S&P Capital IQ (CIQ) and reserve for uncertain tax benefits (UTB) from Compustat. After matching our sample of 10,118 client-year observations to clients with nonmissing and non-zero valuation allowance (UTB) data, we obtain 5,056 (5,020) observations for testing H2b. 17 Audit firms that are subject to PCAOB s annual inspections are PricewaterhouseCoopers, Deloitte & Touche, Ernst & Young, KPMG, BDO, Grant Thorton, McGladrey, Crowe Horwath, and MaloneBailey. 18

21 4. Empirical Results 4.1. Descriptive Statistics Table 1, panel A (B) reports descriptive statistics of the variables used in our main analysis (equation 1). Panel A (Panel B) presents the statistics for clients before (after) Deloitte s Part II report. 18 Within each panel, we further split the sample into Deloitte s clients (DT sample) and non-deloitte s clients (non-dt sample). Across both panels, approximately 70% of firms use APTS. In addition, based on the Kolmogorov-Smirnov test in column (a), we note that the distribution of our dependent variable (APTS) is not significantly different between DT and non- DT samples. Similarly, from the Levenes test in column (b), the variance of APTS is not significantly unequal between DT and non-dt samples. Although we observe that a few control variables (e.g., NTAFEE, SIZE, LEV, and AFEE) have a significantly different distribution across DT and non-dt samples, we further note that these differences persist across periods. To expand upon this observation, we present differences in the means of all variables between DT and non- DT samples across periods in Table 1, panel C (i.e., differences-in-differences). From column (c), no one pair of difference-in-differences is significant. Hence, the treatment (DT sample) and control (non-dt sample) groups follow a similar trend across the pre- and post-periods, validating the parallel trend assumption. Table 2 presents the correlation matrix of variables used in the regression models, but our main hypothesis is not directly testable from this table. We observe some high correlations between control variables and our treatment variable (DT). Although these correlations might indicate serious multicollinearity concerns, we find that the highest variance inflationary factor (VIF) is 18 The number of client-year observations is 5,368 in pre-period of issuance of the part II report. This is not the same as the number of client-year observations (4,750) in post-period of issuance of the part II report. The reason is that we include full fiscal years of 2009 to 2012, while we cut off the sample period into pre- and post-issuance of the part II report based on the actual report issuance date (i.e., October 17, 2011). 19

22 4.51 (which is less than 10). Thus, there is no evidence of severe multicollinearity concerns (Kutner, Nachtsheim, and Neter 2004) Hypothesis testing Likelihood of Retaining the Auditor as the Tax Service Provider (H1) Our first hypothesis examines whether Deloitte s clients are less likely to retain Deloitte as their tax service provider following the PCAOB s part II report. Table 3 reports the results from estimating equation (1), where the coefficient of interest is DTi,t POST. Table 3 column (1) reports the results for the full sample where the dependent variable captures the existence of auditor provided tax services (APTS). In this first column, the coefficient on DTi,t POST is significantly negative ( 0.142, z-stat = 3.4) and suggests that following the PCAOB report of Deloitte s income tax related deficiencies, Deloitte s clients are 14.2% less likely to retain APTS than are clients of other auditors. Table 3 column (2) reports the results for the restricted sample of only clients whose auditors provided tax services in the prior year, and the dependent variable captures clients termination of auditor provided tax services (APTS_STOP). In this second column, the coefficient on DTi,t POST is significantly positive (0.545, z-stat = 4.4), suggesting that subsequent to the PCAOB report of Deloitte s income tax related deficiencies, Deloitte s clients are 54.5% more likely to stop using APTS than are clients of other auditors. In Table 3 column (3), we restrict the sample to only clients whose auditors did not provide tax services in the prior year, and the dependent variable captures clients initiation of APTS (APTS_START). In this third column, the coefficient on DTi,t POST is not significant (0.037, z-stat = 0.4). Together, these results suggest 20

23 that the decreased likelihood of retaining auditor provided tax services (Table 3, column 1) is driven by increased APTS termination rather than decreased APTS initiation. 19 Overall, the results in Table 3 reject the hypothesis 1 prediction of no effect of the PCAOB report on APTS. Instead, we provide evidence consistent with clients perceiving the costs of retaining the tax services of an auditor with known income tax related deficiencies to exceed any potential knowledge spillover benefits from these services. Hence, unlike the PCAOB disclosures of audit firms engagement weaknesses (Part I reports) which elicit limited response from clients, the public disclosure of auditors quality control problems (Part II reports) elicits significant client response (Lennox and Pittman 2010). In particular, we provide empirical evidence that the public disclosure of tax related quality control problems is associated with reduced client demand for auditor provided tax services Reputational Concerns from Retaining the Auditor as the Tax Service Provider (H2a) Hypothesis 2a predicts that Deloitte s clients with greater concern about their auditor s impaired independence are less likely to retain Deloitte as their tax service provider following the PCAOB s part II disclosure. Table 4 reports the results from estimating equation (1) separately on sub-samples derived from a median split of the ratio of non-audit non-tax fees to total audit fees in year t-1 (NTAFEE). Table 4 column (1) presents results for clients with low (i.e., below-median) NTAFEE and Table 4 column (2) presents results for clients with high (i.e., above-median) NTAFEE. Within clients with low (high) NTAFEE, the coefficient on DTi,t POST is significantly positive (negative). Most importantly, using a Wald test, we find that the difference in this coefficient across the two subsamples is statistically significant ( 0.423, p-value < 0.01), implying 19 We find similar results (in sign and statistical significance) when we identify a client as having APTS if tax fees are greater than $50,

24 that the propensity to dismiss Deloitte as the tax service provider is more pronounced among clients with greater reputational concerns. Overall, the results in Table 4 support hypothesis 2a. In particular, the evidence suggests that Deloitte s clients perceive the costs of retaining the tax services of an auditor with known income tax related deficiencies to be incrementally higher when the auditors heavily rely on nonaudit services for revenue. Since such reliance fuels stakeholders perception of auditor independence impairment, these clients have stronger incentives to dismiss the auditor as their tax service provider in order to restore the appearance of auditor independence. Focusing on clients with greater reputational concerns, as in Table 4 column (2), it is worth noting that the significantly negative coefficient on POST suggests a negative spillover effect of Deloitte s Part II report on the APTS of other audit firms. Also, the coefficient on DTi,t is not significant, implying that there is no evidence of a difference in APTS between Deloitte and non- Deloitte clients before the public becomes aware of Deloitte s income tax related deficiencies. However, the significantly negative coefficient on DTi,t POST ( 0.337, p-value < 0.01) suggests that once these deficiencies become publicly known, clients respond negatively by curtailing APTS. Hence, public disclosure of PCAOB quality control inspection findings improves the informational value of PCAOB reports Tax Quality Concerns from Retaining the Auditor as the Tax Service Provider (H2b) Hypothesis 2b predicts that Deloitte s clients with greater concern about APTS quality are less likely to retain Deloitte as the tax service provider following the PCAOB s part II disclosure. Table 5 panel A reports the results from estimating equation (1) separately on sub-samples derived from median splits of abnormal changes in the deferred tax valuation allowance (columns 1 and 2) 22

25 and abnormal changes in reserve for uncertain tax benefits (3 and 4). Clients have greater concern about tax service quality when the change in valuation allowance (reserve for UTB) is high. Consistent with our expectations, using a Wald test for differences across regressions (HIGHVA = 0 vs. HIGHVA = 1), the coefficient on DTi,t POST is significantly more negative ( 0.661, p- value < 0.05) when the change in valuation allowance is abnormally high. In addition, based on a Wald test for differences across regressions (HIGHUTB = 0 vs. HIGHUTB = 1), we find that the difference in the coefficient on DTi,t POST across the two subsamples is statistically significant ( 0.545, p-value < 0.01). Overall, the results in Table 5 panel A provide support for hypothesis 2b. Specifically, the evidence suggests that Deloitte s clients perceive the costs of retaining the tax services of an auditor with known income tax related deficiencies to be incrementally higher when the auditor exercises excessive conservatism with respect to either deferred tax allowances or reserve for uncertain tax benefits. Since excessive conservatism in tax service provision elicits clients concern about tax service quality, clients have stronger incentives to dismiss the auditor as their tax service provider in order to mitigate these tax costs Sensitivity Tests Alternative Model Specifications In this section, we re-examine our main model (equation 1) using two alternative model specifications. First, our main model uses observations at the client-level but observations at the audit-office-level are also appropriate. Therefore, as a sensitivity test, we re-estimate the model in Table 3, where all variables in this model are averaged at the audit-office-year-level. In untabulated analysis, we find similar results (in sign and significance) as those presented in Table 3. Second, 23