Audit quality effects of the PCAOB's 2006 tax service restrictions

Abstract

In 2004 and 2005, use of aggressive tax services provided by a company's auditor had become so broadly concerning that it was the focus of a PCAOB roundtable and discussed in a Congressional subcommittee investigation report (PCAOB, 2004 and US Senate, 2005). Although auditor provision of these and other nonaudit services to issuer audit clients was restricted in 2006, research on the effectiveness of these restrictions finds that they did not impact audit quality (Notbohm, Paterson and Valencia, 2015 and Lennox, 2016). We reexamine this issue with a focus on the audit quality effects for the engagements most impacted by the restrictions-Big 4 audit clients with pre-restriction tax service fees of at least $100,000 that fell by at least 75% following the restrictions. Using a difference-in-difference framework and two proxies for audit quality, we find evidence of the effectiveness of the PCAOB's 2006 restrictions among those clients. Additionally, we find these results are sensitive to the level of pre-restriction tax service fees, with the restatement (going concern) effect of the restrictions strengthening (weakening) in the pre-restriction tax service fee level. We also find that the effects of these restrictions are concentrated among clients of Big 4 auditors rather than clients of the 2nd tier or 3rd tier auditors. Results of additional analyses indicate that audit quality, as measured by the probability of restatement, was lower in the pre-restriction period for purchasers than for non-purchasers. Our results are robust to a barrage of sensitivity tests. Our findings contribute to the continued regulator discussion about the proper level and types of allowable tax nonaudit services.

Introduction

Announcements of high-profile corporate tax reporting investigations (e.g., Caterpillar, Hewlett Packard) involving nonaudit tax services provided by the company's auditor have renewed discussions about the types of tax services auditors should provide their clients (Mclaughlin, 2017; Novak, 2012b). Lawmakers investigating these companies have expressed concern about potential conflicts of interest when the auditing firm designs and sells questionably aggressive tax structures to audit clients and subsequently audits the financial reporting for these structures (U.S. Senate, 2012, U.S. Senate, 2014). The Public Company Accounting Oversight Board (PCAOB) and Securities Exchange Commission (SEC) also expressed interest in reexamining the potential independence effects of tax and other nonaudit services (Harris, 2014; PCAOB, 2015; Tysiac, 2013).

In the early 2000s, the PCAOB, SEC and many in practice had similar concerns about the auditor independence effects of auditor provision of personal tax services and aggressive tax consulting services to audit clients (PCAOB, 2004). In 2006 the PCAOB restricted these types of auditor-provided tax services by implementing rules 3521, 3522, and 3523 (PCAOB, 2021; Securities Exchange Commission, 2006).^1,2 These rules reaffirmed the existing restrictions on contingent fees and added further restrictions on the auditor provision of certain aggressive tax consulting services and certain personal tax preparation services for audit clients. The key purpose behind these restrictions was to improve auditor independence by restricting certain nonaudit services that may impair independence.

However, recent empirical research finds little evidence of these rules' effectiveness (Lennox, 2016; Notbohm, Paterson, & Valencia, 2015). If the 2006 restrictions on tax nonaudit services had little or no effect on audit quality, then some may question if further restrictions on tax nonaudit services should be considered. Our study reexamines this research question with a focus on the engagements that were likely most impacted by the restrictions. Based on comments in a 2004 PCAOB Roundtable discussion, U.S. Senate Subcommittee Investigation Report, U.S. Department of the Treasury Investigation Report, and a financial press article on the subject, we expect that the aggressive tax strategy services restricted by Rule 3522 were provided primarily by the largest auditing firms (U.S. Department of the Treasury, 1999; Novak, 1998; PCAOB, 2004; U.S. Senate, 2005). Additionally, because auditor-provided tax services that generate large fees likely create greater conflicts of interest than if the same services were provided for small fees, we expect the restricted tax services created the greatest conflicts of interest when the services generated large fees. Using both financial statement restatements and going concern modified wording audit opinions as indicators of audit quality, we estimate difference-in-difference regressions, similar to those estimated in Lennox (2016), to test the audit quality effects of these restrictions on suspected purchasers of the restricted services.

We find that the probability of restatement (going concern opinion) was significantly reduced (increased) by the 2006 restrictions among Big 4 audit clients that significantly reduced their tax service fees. Additionally, we find that the restatement result gets stronger as the level of pre-restriction tax service fee increases. However, we also find that the going concern result fades in the level of pre-restriction tax service fee. We expect this fading going concern result is due to the inability of financially distressed companies to benefit from expensive tax nonaudit services. In additional analyses we find that prior to implementing the 2006 restrictions the probability of restatement was higher for suspected purchasers than for non-purchasers, and we also find that the audit quality effects of the restrictions were concentrated among Big 4 firms. Our results are robust to alternative definitions of the treatment variable, matched pairs balanced sample analysis, exclusion of clients audited by KPMG, inclusion of financial firms, and are partially robust to an alternative control for internal control strength.

The primary difference between our study and Lennox (2016) is that our study focuses on the engagements that are likely most impacted by the 2006 restrictions. This alternate focus impacts our analyses in two significant ways. First, we focus our sample on audit clients of the Big 4. We impose this condition because many practitioner, regulator, and legislator comments about the “promoters” of the services that were restricted by PCAOB Rule 3522 are consistent with only the largest public accounting firms. Second, in our main tests we limit our treatment firms to those with at least $100,000 of pre-restriction period tax nonaudit fees. This condition ensures that we focus our tests on only purchasers of significant levels of aggressive tax services, contingent fee services, and/or personal tax services.

Evidence of the effectiveness of the 2006 PCAOB restrictions is important because it adds to the ongoing debate about the proper levels and types of allowable auditor-provided tax services. If the 2006 restrictions improved auditor independence with respect to purchasers of the restricted services, then other tax service restrictions may yield further independence improvements. Additionally, our results may benefit audit committees, boards of directors, owners, management and other decision makers charged with deciding which types of nonaudit services their company should purchase from their auditor. Evidence that the purchase of the restricted tax services harmed auditor independence and audit quality may caution these decision makers about approving the purchase of permitted tax services from their auditor.³

The remainder of the paper is organized as follows: Section Two discusses important background information related to the 2006 tax service restrictions and relevant prior research. In Section Three we motivate our hypothesis. Section Four includes our sample selection criteria, description of our data, and discussion of our research methodologies. In Section Five we discuss our results. In Section Six we present the results of our additional analyses. We conclude in Section Seven.