The Impact of PCAOB Auditing Standard No. 5,the PCAOB Inspection Regime,and the Great Recession on Audit Fees and Audit Quality

Abstract

We investigate the coinciding effects of the implementation of Auditing Standard No. 5 (AS5), the change in the Public Company Accounting Oversight Board’s (PCAOB) inspection regime, and the Great Recession on the audit fees and audit quality of accelerated filers. AS5 took effect in November 2007 and promulgated a top-down, risk-based audit approach to SOX 404(b) audits of accelerated filers. Concurrently, the PCAOB adopted a stricter approach to its inspections of audit firms, which encouraged them to improve audit quality and reduce audit fees. Moreover, the Great Recession pressured audit firms to reduce fees. We find that, following the three events, audit fees decreased and quality increased for accelerated filers. We also find that audit fees and audit quality increased for non-accelerated filers, although these filers were not directly affected by AS5.

Keywords

PCAOB Auditing Standard No. 5 PCAOB inspection great recession audit fees audit quality internal control

Introduction

This study examines the joint effect of three coinciding events on audit fees and audit quality of accelerated filer firms (hereafter, accelerated filers) in the 3 years from November 15, 2007, through November 14, 2010.¹ The three events are the adoption of Audit Standard No. 5 (hereafter, AS5) (Public Company Accounting Oversight Board [PCAOB], 2007b), the PCAOB’s stricter inspection approach (Rowe & Sivadasan, 2018), and the 18-month Great Recession (from December 2007 through June 2009) (National Bureau of Economic Research [NBER], 2010). AS5, which replaced AS2, took effect for fiscal year-ends starting November 15, 2007. It is a top-down, risk-based approach to audits of internal control over financial reporting for accelerated filers (PCAOB, 2007a, 2007b). Its promulgation coincided with a stricter, more efficient and focused PCAOB inspection regime that was intended to increase audit quality through improved audit firm quality controls (PCAOB, 2008; Rowe & Sivadasan, 2018). At the same time, the Great Recession led clients to pressure audit firms to reduce fees (Ettredge et al., 2014). Whereas prior studies assess the effects of AS5 by examining a very short (1- to 2-year) event period (Doogar et al., 2010; Krishnan et al., 2011; Wang & Zhou, 2012), we examine a more extended (3-year) event period, which permits us to incorporate the longer-term and joint effects of AS5, the new PCAOB inspection regime, and the Great Recession.

We start by examining whether audit fees decreased during the event period (November 15, 2007, through November 14, 2010). Prior studies restrict their analyses between 6 months and 2 years post-AS5, and find that fees decreased (Doogar et al., 2010; Krishnan et al., 2011; Wang & Zhou, 2012). We examine a much longer period to determine the concurrent and combined effects of AS5, the stricter PCAOB inspection regime, and the Great Recession. We argue that audit fees could have further decreased in the second and third year of the three events. In the first year, audit fees would have been negotiated in the first quarter of 2007 for most clients—well before AS5 was issued and before the recession took hold (Ettredge et al., 2014). After AS5 was approved by the U.S. Securities and Exchange Commission (SEC) in July 2007 (U.S. SEC, 2007), audit firms had less than 1 year to implement the new risk-based methodologies and realize first-year efficiencies for calendar year-end clients.² By the second and third year, audit firms would have reduced fees because they started to improve efficiency through training and process improvements. The PCAOB further facilitated improvements through its inspection comments to audit firms (PCAOB, 2008). In addition, client pressure to reduce fees would have increased in the second year, after the Big Four reputation was threatened by the sanctioning of Deloitte (PCAOB, 2007c), and as the Great Recession resulted after two quarters of negative GDP growth. Alternatively, audit fees may have increased after the first year because the PCAOB inspections demanded stronger quality control procedures from audit firms and the Great Recession increased audit risk.

We also examine whether audit quality changed over the event period. To our knowledge, only one prior study examines the effect of AS5 on audit quality, and it only examines the first 6 months, failing to find any improvement (Wang & Zhou, 2012). However, audit quality may have improved over our longer event period because the PCAOB intended for AS5 to enhance audit quality by placing a greater focus on areas of higher financial reporting risk (PCAOB, 2007a, 2007b). The short (less than 1 year) implementation period would have delayed audit quality improvements into the second year. Concurrently, the PCAOB’s oversight changed to a stricter approach that sanctioned one Big Four firm and publicized inspection details that permit the computation of an “audit failure rate” (Rowe & Sivadasan, 2018). In response to the AS5 mandate and the PCAOB’s stricter oversight, several Big Four firms introduced improvements (Deloitte, 2010; KPMG, 2009; PCAOB, 2013). Consistent with our audit fees analyses, we study a 3-year period where audit quality improvements are more likely to be apparent. Alternatively, audit quality might have decreased because some commentators in the popular press contend that AS5’s top-down, risk-based approach could lead to under-auditing (Berkowitz & Rampell, 2002; Weil, 2004). Moreover, following layoffs in late 2008 and 2009, the PCAOB raised concerns that audit fee pressure could jeopardize audit quality (PCAOB, 2010b). Ettredge et al. (2014) support the PCAOB’s concerns by finding a positive association between fee pressure and restatements in the middle of the recession.

To examine the joint effect of the three events on the audit fees and audit quality of accelerated filers, we compare audit fee and audit quality measures in the 3-year event period to those in the preceding 2-year period (hereafter “pre-period”).³ For robustness, we employ a difference-in-differences design which compares the effects on accelerated filers (the treatment group) with the effects on non-accelerated filers (the control group). We measure audit quality using two output-based measures (DeFond & Zhang, 2014): abnormal accruals and misstatement likelihood. In further robustness tests, we use several alternative methods: controlling for client fixed effects, a matched-pair sample, a balanced panel sample, a size-restricted sample for the abnormal accrual analysis, controlling for PCAOB reports, and an alternative measure of audit quality (meeting or beating consensus analysts’ earnings forecasts).

Our study makes two contributions to the literature. First, we extend prior studies (Doogar et al., 2010; Krishnan et al., 2011; Wang & Zhou, 2012) by examining an extended event period. We find that audit fees decreased for accelerated filers in the event period, compared with the pre-period, and that the decrease was highest in the second and third years. Notably, the clients of annually inspected firms drive these results. We also find that audit fees increased for non-accelerated filers.

Second, we find an increase in the audit quality of accelerated filers audited by annually inspected audit firms. We also find an increase in the audit quality of non-accelerated filers using abnormal accruals and, in the first 2 years of the event period, for misstatement likelihood. However, we cannot conclude from the difference-in-differences tests that audit quality increased more for accelerated filers than for non-accelerated filers because the inferences are not consistent across all tests. Overall, our results suggest that firm-wide improvements to audit quality affected both accelerated and non-accelerated filers and were likely the result of spillovers from implementing AS5 and stricter PCAOB inspections. Therefore, our findings do not support previous criticisms of the PCAOB’s adoption of the AS5 risk-based audit approach (Berkowitz & Rampell, 2002; Weil, 2004). On the contrary, we find that audit quality improved.

Our study differs from recent work by Schroeder and Shepardson (2016), which finds that compared with under AS2, quarterly abnormal accruals quality is lower post-AS5 because we examine audited accrual quality, rather than unaudited quarterly accrual quality. We corroborate our results with two additional measures of audit quality: “Big R” misstatements and meeting or beating consensus analysts’ earnings forecasts.

Background, Related Studies, and Hypothesis Development

Audit Fees, AS5, PCAOB Inspections, and the Great Recession

Following the passage of the Sarbanes-Oxley Act of 2002 (U.S. House of Representatives, 2002) (hereafter, SOX) and the implementation of AS2 (PCAOB, 2004), companies experienced exorbitant audit fee increases (C. Johnson, 2005). After an SEC roundtable in May 2006, the PCAOB announced plans to amend AS2 and emphasize auditor efficiency through its inspection efforts (PCAOB, 2006). In May 2007, the PCAOB issued AS5 to reduce audit costs by employing a top-down, risk-based approach to testing internal controls over financial reporting (PCAOB, 2007a, 2007b). The underlying framework of AS5 allows auditors to focus on the critical financial reporting risks and the associated mitigating internal controls, thus increasing the scalability of audits for client size, complexity, and riskiness (PCAOB, 2007a, 2007b). Some of the expected savings from AS5 included reliance on the work of others (including management and internal auditors), elimination of the requirement to assess management’s evaluation of internal controls, and a reduction of testing when financial reporting risk was low.

However, the intended savings may have taken several years to fully manifest. AS5 was issued on short notice in May 2007 (PCAOB, 2007b) and approved by the SEC in July 2007 (U.S. SEC, 2007), taking effect for accelerated filers with fiscal year-ends on or after November 15, 2007 (AS5 period). To facilitate AS5’s implementation, the PCAOB staff met with the major audit firms at the time of AS5’s adoption to discuss its application and make suggestions (PCAOB, 2008). The PCAOB continued to monitor AS5’s implementation in their inspections (PCAOB, 2008). The implementation period, approximately 5 months for clients with a December fiscal year-end, was likely too short for most benefits to be realized because auditors would need time to learn and adapt to the new audit methodology. Alternatively, audit firms may have anticipated the new audit standard was forthcoming because the PCAOB announced in May 2006 that it planned to amend AS2 to ensure that the auditors’ primary focus was on areas that “pose higher risk of fraud or material error” (PCAOB, 2006).

Two studies show that audit fees decreased in the first 6 to 8 months following the implementation of AS5 (Doogar et al., 2010; Wang & Zhou, 2012), whereas one study shows that audit fees decreased in the first 2 years (Krishnan et al., 2011). In addition, Doogar et al. (2010) find that in the first 8 months following AS5’s implementation, audit fees were not only better aligned with fraud risk, but were also lower than during the last year of AS2, suggesting that AS5 is more efficient. Krishnan et al. (2011) find that audit fees were lower in the first 2 years under AS5 than in the last year under AS2, and audit fees decreased by more in the AS5 era for clients that remediated material weaknesses. Finally, Wang and Zhou (2012) compare an approximate 6-month AS5 period to the last year of AS2 for the same client firms and find that audit fees decreased by 4.02%.⁴

Although two of these studies show that accelerated filers paid lower audit fees in the first year, we posit that further reductions were likely in later years because of delayed fee negotiations, additional efficiencies from the implementation of AS5, a stricter PCAOB inspection approach, and the effects of the Great Recession. Audit fees are traditionally negotiated almost 1 year in advance (Ettredge et al., 2014), and the SEC did not formally approve AS5 until July 2007, so further audit fee reductions may have been forthcoming (2008 and afterward). In addition, audit firms would realize additional efficiencies from fully implementing AS5 with training and audit program improvements. The PCAOB facilitated further efficiency improvements by directing inspections staff in 2008 to continue observing the implementation of AS5 and issue inspection comments consistent with its objectives (PCAOB, 2008).

In December 2007, the PCAOB censured Deloitte which conceivably threatened other Big 4 audit firms (Burns, 2007; PCAOB, 2007c) and harmed the reputation of all Big 4 firms. Although the PCAOB supported the audit firms’ implementation of AS5, its inspection approach became more adversarial (Rowe & Sivadasan, 2018) and conceivably pressured audit fees downward in the second year and after. Just 1 month after AS5’s implementation, and coinciding with the Deloitte censure, the Great Recession began (NBER, 2010). However, clients and auditors may not have understood the severity of the Great Recession until late 2008, after two consecutive quarters of negative GDP growth. In response to the effects of the recession, clients further pressured audit firms to reduce audit fees, as both parties scaled-back operations (Bordonaro, 2008; Plumb, 2009; Rosenberg, 2010).

These events lead to our first hypothesis, stated in the alternative form:

Hypothesis 1 (H1): Ceteris paribus, audit fees decreased for accelerated filers after the coinciding events of AS5 adoption, the stricter PCAOB inspection regime, and the Great Recession.

However, the hypothesis is not without tension. Around the time that the recession ended (i.e., June 2009), the PCAOB’s actions may have had a counteracting effect on audit fees. When PCAOB inspections began in 2004, the PCAOB initially adopted a conciliatory and non-confrontational “supervisory approach” (Wegman, 2008); however, the approach later changed. The PCAOB’s 2007 censure of Deloitte hurt the firm’s ability to retain clients in 2008 and 2009 (Boone et al., 2015; PCAOB, 2007c), and conceivably threatened other Big 4 audit firms (Burns, 2007). In 2010, the PCAOB published, in their 2009 Part I inspection reports, the number of audits with deficiencies and the total number of audits inspected for each audit firm. This allowed for an estimate of the “audit failure rate” (Norris, 2012; Rowe & Sivadasan, 2018). These threats, and the apparent “audit failure rate,” may have led audit firms to improve quality controls further, passing on the associated costs to their clients through increased fees. In addition, in late 2008, the Great Recession may have increased audit risk and auditor effort.

Audit Quality, AS5, the PCAOB Regime, and the Great Recession

Whereas prior studies find that audit fees decreased in the first 1 to 2 years of our event period (Doogar et al., 2010; Krishnan et al., 2011; Wang & Zhou, 2012), only one study examines audit quality in the first 6 months (Wang & Zhou, 2012), and the authors do not find a change in audit quality. We offer two reasons for using a longer (3-year) event period.

First, although AS2 had achieved audit quality improvements (e.g., Ashbaugh-Skaife et al., 2008), AS5 aimed to increase audit quality further. AS5’s streamlined process reduces unnecessary audit procedures to better focus “the auditor’s attention on those matters that are most important to effective internal control” (PCAOB, 2007b). In turn, AS5 should be better able to identify material weaknesses before they cause misstatements (PCAOB, 2007b). Arguably, these improvements should be more apparent over a longer 3-year period.

Second, because the PCAOB took a stricter inspection approach (Rowe & Sivadasan, 2018), audit firms were encouraged to improve audit quality for both accelerated and non-accelerated filers. When the PCAOB censured Deloitte for its audit failure of Ligand Pharmaceuticals (Boone et al., 2015), it stressed that other Big Four audit firms could expect enforcement actions.⁵ By 2010, constituents could compute an “audit failure rate” from the PCAOB 2009 Part I inspection reports (Rowe & Sivadasan, 2018). In response to the PCAOB’s adversarial approach, audit firms started to implement improvements. KPMG announced enhancements in 2009 (KPMG, 2009), Deloitte released a transparency report in January 2010 which stated that it had “instituted additional quality control policies and systems” (Deloitte, 2010), and Ernst & Young noted that it made significant improvements to quality including “new audit tools, additional training and expanded technical guidance” (PCAOB, 2013).

Still, despite these better processes and increased pressure by the PCAOB, audit quality may have decreased or not changed in the event period. First, audit quality may have decreased because risk-based audits lead to under-auditing and are associated with audit failures like those at Enron, WorldCom, and HealthSouth (Berkowitz & Rampell, 2002; Weil, 2004).⁶ Second, audit quality may not have changed, because the PCAOB inspection model is dysfunctional (Glover et al., 2009). In support, Boone et al. (2015) find that audit quality did not change for Deloitte in the 3 years following their 2007 PCAOB censure. Third, audit fee pressure increased during the Great Recession, which incentivized auditors to reduce effort and audit quality to maintain profitability (Ettredge et al., 2014). Relatedly, following layoffs in late 2008 and 2009, the acting PCAOB chairman, D. Goelzer, expressed concerns that audit fee pressure may have jeopardized audit quality (PCAOB, 2010b). Concurrently, the SEC’s Chief Accountant also implied that auditors could be compromising audit quality (U.S. SEC, 2010).⁷

Given that there are arguments in either direction, we hypothesize in the null form:

Hypothesis 2 (H2): Ceteris paribus, audit quality of accelerated filers did not change after the coinciding events of AS5 adoption, the stricter PCAOB inspection regime and the Great Recession.

Sample Selection

Sample Selection and Design

To test H1 and H2, we use a sample period covering a total of 5 years: 2 years before the three events (fiscal year-ends November 15, 2005, through November 14, 2007—the pre-period) and 3 years after (November 15, 2007, through November 14, 2010—the event period). Our sample design compares audit fees and audit quality in each year of the event period with that in the benchmark pre-period. Our variables of interest are three indicator variables each representing 1 year beginning on November 15 and ending on November 14 of the succeeding calendar year: YEAR1 (2007–2008), YEAR2 (2008–2009), and YEAR3 (2009–2010). We refer to these event years as Year 1, Year 2, and Year 3, respectively.

Our sample selection begins with all firms available in the Audit Analytics’ internal control file as of May 2015. We delete foreign clients, firms with zero or negative audit fees, and financial sector firms (SIC 6000–6999). We obtain audit fees, auditor identities, going-concern opinions, internal control opinions, restatements, filing dates, and filer types from Audit Analytics. We obtain all other variables from Compustat and I/B/E/S. Our audit fee sample consists of 10,365 accelerated filers. We then construct a sample of accelerated filers using available data for each audit quality measure: abnormal accruals (n = 9,316) and misstatements (n = 10,365). We refer to these three samples as accelerated filer samples. Finally, we construct three expanded samples that pool accelerated filers and non-accelerated filers for difference-in-differences tests: audit fees (n = 16,553), abnormal accruals (n = 14,981), and misstatements (n = 16,788).⁸ We refer to these samples as the pooled samples.

Audit Fees and Audit Quality Measures by Period and Client Size

Table 1, Panels A and B, reports the mean values of audit fees and audit quality measures by period (Pre-period, YEAR1, YEAR2, and YEAR3) for accelerated filers (Panel A) and non-accelerated filers (Panel B). We find that audit fees declined for accelerated filers during the 3-year event period, relative to the 2-year pre-period. We also find that abnormal accruals and “Big R” misstatements declined for accelerated filers during the event period, relative to the pre-period. However, a similar pattern emerges for non-accelerated filers, which were not directly affected by AS5. The result suggests that audit firms may have implemented changes in response to AS5, the stricter PCAOB inspection regime, and the Great Recession that affected both accelerated filers and non-accelerated filers.

Table 1.

Mean Values of Audit Fees and Audit Quality by Period and Filer Type.

Panel A: Accelerated Filers.
Period	Count	AF ($000’s)	\|DACC\|	MISSTATE “Big R”
Pre-period	3,869	2,479.51	0.103	0.046
YEAR1	2,183	2,464.11	0.066	0.025
YEAR2	2,226	2,425.36	0.061	0.022
YEAR3	2,087	2,343.18	0.058	0.022
Total	10,365	2,437.08	0.071	0.031

Panel B: Non-Accelerated Filers.
Period	Count	AF ($000’s)	\|DACC\|	MISSTATE “Big R”
Pre-period	3,066	679.64	0.144	0.038
YEAR1	1,170	367.80	0.140	0.024
YEAR2	957	327.58	0.110	0.025
YEAR3	995	378.14	0.105	0.036
Total	6,188	510.00	0.125	0.032

Note. This table presents mean values of audit fees (AF), the absolute value of abnormal accruals (|DACC|), and “Big R” misstatements by filer type (accelerated filers in Panel A and non-accelerated filers in Panel B) and the period of interest with respective count values. The Pre-period covers fiscal year-ends November 15, 2005, through November 14, 2007, and YEAR1 through YEAR3 cover one year for fiscal years ending November 14, 2008, through November 14, 2010, respectively. Count represents the number of clients in the audit fee sample. MISSTATE is restricted to “Big R” misstatements, which arise from a material misstatement of the financial statements. They are reported on the Form 8-K and require the auditor to re-issue the audit opinion on the restated financial statements.

Descriptive Statistics and Multivariate Analysis

Audit Fees

To test whether audit fees decreased in the event period, we follow specifications in prior studies (Doogar et al., 2010; Elder et al., 2009; Ettredge et al., 2014; E. Johnson et al., 2018; Krishnan et al., 2011; Wang & Zhou, 2012). We estimate the audit fee model using the accelerated filer sample because H1 focuses on accelerated filers. Equation 1, which is estimated using ordinary least squares (OLS) regression, is as follows:

\begin{matrix} \begin{matrix} Log (AF) = α_{0} + α_{1} YEAR 1 + α_{2} YEAR 2 + α_{3} YEAR 3 + α_{4} Log (ASSETS) \\ + α_{5} CHG__AU + α_{6} LOSS + α_{7} GC + α_{8} ICW + α_{9} CURRENT \\ + α_{10} QUICK + α_{11} ROA + α_{12} MERGER + α_{13} LEV + α_{14} FOREIGN \\ + α_{15} Log (SEGMENT) + α_{16} YE + α_{17} TENURE + α_{18} MISSTATEMENT \\ + α_{19} GROWTH + α_{20} DISTRESS + α_{21} MtB + α_{22} ARINV + α_{23} AU__LAG \end{matrix} \\ + Audit firm fixed effects + Industry fixed effects + ε . \end{matrix}

(1)⁹

The dependent variable is the natural logarithm of audit fees in thousands of dollars (Log(AF)). The signs of the coefficients on the three indicator variables (YEAR1, YEAR2, and YEAR3) indicate whether annual audit fees were higher (positive) or lower (negative) in the respective event year, compared with the pre-period. We include audit firm fixed effects to control for unobservable audit firm effects, such as specific audit practices (DeFond & Lennox, 2017). We include industry fixed effects (two-digit SIC codes) to control for unobservable industry effects. ϵ is a random disturbance term.

Table 2 reports descriptive statistics for accelerated filers and non-accelerated filers for the audit fee sample (Panel A), the abnormal accruals sample (Panel B), and the misstatements sample (Panel C). These statistics are comparable to those in prior studies. A comparison between the two filer types reveals that accelerated filers are larger (Assets, Log(ASSETS)), less leveraged (LEV), report fewer losses (LOSS), have more growth options (MtB), report more business segments (Log(Segments)), receive fewer going-concern opinions (GC), and less frequently change auditors (Chg_AU).

Table 2.

Panel A: Sample Descriptive Statistics—Audit Fee Sample.
Variable	Accelerated filer sample N = 10,365			Non-accelerated filer sample N = 6,188
	M	Median	SD	M	Median	SD
Audit Fees (000’s)	2,437.080	1,116.000	4,490.290	509.995	148.000	2,240.440
Log(AF)	7.072	7.018	1.164	5.048	4.997	1.301
Assets (millions)	4,819.870	626.126	17,185.570	1,101.990	26.284	9,286.550
Log(ASSETS)	6.580	6.440	1.887	3.240	3.269	2.340
Chg_AU	0.062	0.000	0.241	0.127	0.000	0.333
LOSS	0.320	0.000	0.466	0.586	1.000	0.493
GC	0.029	0.000	0.167	0.232	0.000	0.422
ICW	0.075	0.000	0.264	NA	NA	NA
CURRENT	2.810	2.019	2.338	2.827	1.824	2.883
QUICK	0.554	0.647	0.830	0.523	0.647	0.645
ROA	0.081	0.067	0.088	0.050	0.000	0.099
MERGER	0.089	0.000	0.284	0.034	0.000	0.181
LEV	0.187	0.128	0.236	0.219	0.010	1.315
FOREIGN	0.347	0.000	0.476	0.244	0.000	0.430
Log(SEGMENT)	0.697	0.693	0.721	0.346	0.000	0.571
YE	0.757	1.000	0.429	0.669	1.000	0.471
TENURE	8.296	6.000	7.148	5.199	3.000	5.422
MISSTATEMENT	0.064	0.000	0.246	0.051	0.000	0.220
GROWTH	0.323	0.080	1.901	0.615	0.034	3.217
DISTRESS	−1.049	−1.577	15.008	79.719	−1.096	3,316.860
MtB	7.387	2.072	441.923	3.134	1.547	243.049
ARINV	0.213	0.183	0.164	0.260	0.205	0.230
AU_LAG	−70.100	−64.000	39.708	−94.450	−86.000	79.770

Note. See the appendix for variable definitions. NA = not applicable.

Panel B: Sample Descriptive Statistics—Abnormal Accruals Sample.
Variable	Accelerated filer sample N = 9,316			Non-accelerated filer sample N = 5,665
	M	Median	SD	M	Median	SD
\|DACC\|	0.071	0.039	0.106	0.125	0.074	0.145
Log(ASSETS)	6.525	6.374	1.885	3.189	3.238	2.103
LOSS	0.328	0.000	0.469	0.592	1.000	0.492
GC	0.028	0.000	0.166	0.221	0.000	0.415
ICW	0.072	0.000	0.259	NA	NA	NA
CFO	0.018	0.091	0.802	−0.823	−0.010	3.290
LEV	0.182	0.118	0.237	0.195	0.009	1.136
LIT	0.265	0.000	0.442	0.293	0.000	0.455
GROWTH	0.322	0.079	1.902	0.564	0.032	3.014
FOREIGN	0.355	0.000	0.478	0.234	0.000	0.424
Log(SEGMENT)	0.691	0.693	0.722	0.339	0.000	0.563
Lag_ACCR	70.599	−0.034	7106.020	0.623	−0.038	328.207
TENURE	8.230	6.000	7.065	5.198	4.000	5.411
FEE_PRESSURE	0.468	0.000	0.499	0.487	0.000	0.500
Chg_AU	0.061	0.000	0.239	0.127	0.000	0.333
RCVB_SD	1428.830	1277.110	1750.590	1252.980	1218.570	1408.680
CFO_SD	1.986	1.966	1.073	2.172	2.061	1.166
Panel C: Sample Descriptive Statistics—Misstatements Sample.
Variable	Accelerated filer sample N = 10,365			Non-accelerated filer sample N = 6,423
	M	Median	SD	M	Median	SD
MISSTATE	0.031	0.000	0.173	0.032	0.000	0.177
Log(ASSETS)	6.580	6.439	1.887	3.234	3.267	2.337
Log(AF)	7.072	7.017	1.164	5.048	4.997	1.299
Log(NON_AF)	10.618	11.776	4.110	7.591	9.645	4.865
LEV	0.187	0.128	0.236	0.288	0.010	5.587
LOSS	0.320	0.000	0.466	0.585	1.000	0.493
MtB	7.391	2.072	442.093	3.481	1.537	242.720
ARINV	0.214	0.183	0.164	0.261	0.206	0.231
NSEG	2.620	2.000	1.942	1.731	1.000	1.354
NEWDEBT	0.536	1.000	0.499	0.397	0.000	0.489
BIG4	0.827	1.000	0.378	0.305	0.000	0.460
ICW	0.075	0.000	0.264	NA	NA	NA
LIT	0.263	0.000	0.440	0.284	0.000	0.451
CHG_AU	0.062	0.000	0.241	0.127	0.000	0.333
FEE_PRESSURE	0.442	0.000	0.497	0.467	0.000	0.499
MEET	0.113	0.000	0.317	NA	NA	NA

Note. See the appendix for variable definitions. MEET is based on a sample of 7,277 accelerated filers. NA = not applicable.

Table 3, Column (1), reports the results from estimating Equation 1. T-statistics are estimated from standard errors clustered by client (Petersen, 2009). The adjusted R² is .754 and the coefficients on the control variables are consistent with those in prior studies (Doogar et al., 2010; Elder et al., 2009; Ettredge et al., 2014; E. Johnson et al., 2018; Krishnan et al., 2011; Wang & Zhou, 2012).

Table 3.

OLS Regression of Audit Fees on Event Period. (Dependent Variable = Log(AF)).

Variable	(1) Accelerated filers			(2) Difference-in-differences
	Coefficient		t-stat	Coefficient		t-stat
INTERCEPT	3.505	***	30.21	3.637	***	39.63
YEAR1	−0.026	**	−2.28	−0.027	**	−2.35
YEAR2	−0.055	***	−4.28	−0.063	***	−4.89
YEAR3	−0.113	***	−8.36	−0.117	***	−8.82
NON				−0.427	***	−18.03
YEAR1_NON				0.110	***	4.19
YEAR2_NON				0.112	***	4.01
YEAR3_NON				0.203	***	7.35
Log(ASSETS)	0.481	***	58.99	0.445	***	69.23
CHG_AU	−0.274	***	−7.42	−0.280	***	−10.92
LOSS	0.136	***	6.63	0.152	***	9.12
GC	0.195	***	4.06	0.217	***	7.79
ICW	0.326	***	9.74	0.261	***	11.10
CURRENT	−0.027	***	−5.58	−0.041	***	−11.68
QUICK	0.002		0.24	−0.011		−1.24
ROA	−0.390	***	−3.09	−0.081		−0.98
MERGER	0.051	**	2.34	0.074	***	3.48
LEV	−0.002		−0.03	0.049	***	7.47
FOREIGN	0.139	***	6.61	0.124	***	6.63
Log(SEGMENT)	0.123	***	7.85	0.117	***	8.41
YE	−0.022		−0.90	−0.004		−0.23
TENURE	0.004	**	2.32	0.003	**	2.39
MISSTATEMENT	−0.025		−0.88	−0.066	**	−2.50
GROWTH	0.002		0.53	−0.004	*	−1.96
DISTRESS	0.002	***	5.32	0.000		1.28
MtB	0.000		0.88	0.000		0.58
ARINV	0.686	***	7.59	0.244	***	4.43
AU_LAG	0.000		−0.07	0.000		−0.88
Audit firm fixed effects	Included			Included
Industry fixed effects	Included			Included
F-Test:
YEAR1= YEAR2		***	6.97
YEAR2= YEAR3		***	27.89
YEAR1 + YEAR1_NON				0.083	***	37.23
YEAR2 + YEAR2_NON				0.049	***	28.37
YEAR3 + YEAR3_NON				0.086	***	44.46
N	10,365			16,553
F-Stat	337.65	***		635.44	***
Adjusted R²	.754			.830

Note.***, **, and * denote statistical significance at the 1%, 5%, and 10% levels, respectively, using two-tailed tests. T-statistics are estimated from standard errors clustered by client, following Petersen (2009). The sample covers fiscal year-ends from November 15, 2005, through November 14, 2010. The dependent variable is the log of audit fees. YEAR1 through YEAR3 are indicator variables for the three one-year periods ending November 14, 2008, through November 14, 2010, respectively (i.e., the event period). We predict that audit fees decrease for accelerated filers in the event period, compared with the pre-period. The F-tests test the null hypothesis that the coefficients are equal or that the sum of the coefficients is zero. The appendix defines all other variables. All continuous variables are winsorized at the first and 99th percentiles. OLS = ordinary least squares.

Variables of interest are indicated with bold font.

In support of H1, we find that the coefficients on the variables of interest (YEAR1, YEAR2, and YEAR3) are negative and significant.¹⁰ Thus, controlling for other determinants, audit fees were lower during the 3-year event period, than during the 2-year pre-period. Economically, audit fees decreased 2.6% in Year 1, 5.4% in Year 2, and 10.7% in Year 3, compared with the pre-period.¹¹ A comparison of the coefficients indicates that the audit fee reduction was greater in Year 2 than in Year 1 and greater in Year 3 than in Year 2.

For robustness, in Column (2), we modify Equation 1 using a difference-in-differences design which controls for overall economic trends. Here we compare the difference between the event period and the pre-period for non-accelerated filers versus accelerated filers. Although non-accelerated filers are not directly affected by AS5, they are likely influenced by the changing PCAOB inspection regime and the Great Recession. To identify non-accelerated filers, we add NON, which is equal to 1 for non-accelerated filers and 0 otherwise. To measure the difference-in-differences, we interact each of the event period variables (YEAR1, YEAR2, and YEAR3) with NON, yielding three interaction terms: YEAR1_NON, YEAR2_NON, and YEAR3_NON. The coefficient on YEAR1_NON measures the mean difference between the change in audit fees from the pre-period to Year 1 for the non-accelerated filers and the change in audit fees for the accelerated filers during the same period. A positive coefficient on YEAR1_NON would suggest that the decrease in audit fees from the pre-period to Year 1 for the non-accelerated filers was less than the decrease over this period for the accelerated filers. The coefficient on YEAR2_NON (YEAR3_NON) measures the same effect but between Year 2 (Year 3) and the pre-period. The coefficient on NON measures the mean difference in audit fees between non-accelerated filers and accelerated filers in the pre-period. A negative coefficient would suggest that audit fees were lower for non-accelerated filers than for accelerated filers in the pre-period. The coefficients on YEAR1, YEAR2, and YEAR3 have the same interpretation as in Column (1). The sum of the coefficients on YEAR1 and YEAR1_NON, YEAR2 and YEAR2_NON, and YEAR3 and YEAR3_NON measure the change in audit fees for non-accelerated filers from the pre-period to Year 1, Year 2, and Year 3, respectively.

Column (2) reports the results of the difference-in-differences estimation. The coefficients on YEAR1, YEAR2, and YEAR3 are negative and significant, suggesting that audit fees declined for accelerated filers in the respective event years, compared with the pre-period. This result is consistent with our findings from the accelerated filer sample. The sums of the coefficients on YEAR1 and YEAR1_NON, YEAR2 and YEAR2_NON, and YEAR3 and YEAR3_NON are positive and significant, suggesting that audit fees increased for non-accelerated filers in the respective event years, compared with the pre-period. Most importantly, the coefficients on YEAR1_NON, YEAR2_NON, and YEAR3_NON are positive and significant, indicating that the decrease in audit fees (between the respective event year and the pre-period) was smaller for non-accelerated filers than for accelerated filers.

Overall, our results indicate that audit fees decreased for accelerated filers during the coinciding events of AS5 adoption, the stricter PCAOB inspection regime, and the Great Recession, compared with the pre-period. The decrease in audit fees for accelerated filers was greatest during Year 3 (2009–2010), which coincides with the period after the end of the Great Recession and after audit firms reduced staffing. Fees also increased for non-accelerated filers in each of the 3 years of the event period, compared with the pre-period.

Audit Quality

Abnormal accruals

Following Wang and Zhou (2012), DeFond and Zhang (2014), Chi et al. (2017), and E. Johnson et al. (2018), we measure audit quality using abnormal accruals that approximate within-GAAP earnings management (Kothari et al., 2005). We estimate the magnitude (absolute value) of abnormal accruals (|DACC|) using Kothari et al.’s (2005) performance-adjusted modified Jones model. Table 2, Panel B, reports that the mean of |DACC| is 7.1% for the accelerated filers and 12.5% for the non-accelerated filers, which is comparable to findings in Wang and Zhou (2012). All other variables are similar to those in prior studies.

To test whether audit quality changed in the event period, we estimate Equation 2 using the accelerated filer sample, as follows.

\begin{matrix} | DACC | = Φ_{0} + Φ_{1} YEAR 1 + Φ_{2} YEAR 2 + Φ_{3} YEAR 3 + Φ_{4} Log (ASSETS) \\ + Φ_{5} LOSS + Φ_{6} GC + Φ_{7} ICW + Φ_{8} CFO + Φ_{9} LEV + Φ_{10} LIT \\ + Φ_{11} GROWTH + Φ_{12} FOREIGN + Φ_{13} Log (SEGMENT) \\ + Φ_{14} Lag__ACCR + Φ_{15} TENURE + Φ_{16} FEE__PRESSURE \\ + Φ_{17} CHG__AU + Φ_{18} RCVB__SD + Φ_{19} CFO__SD \\ + Audit firm fixed effects + Industry fixed effects + ε . \end{matrix}

(2)

The variables of interest (YEAR1, YEAR2, and YEAR3) measure the mean difference in |DACC| for each of the three event years, relative to the pre-period. Control variables in Equation 2 are from prior studies. T-statistics are estimated using standard errors clustered by client.

Table 4, Column (1), reports the results from estimating Equation 2. The coefficients on the control variables are generally consistent with those in prior studies. The coefficients on the variables of interest are negative and significant; thus, we reject H2. This result indicates that abnormal accruals are lower for accelerated filers in the event period, relative to the pre-period. Abnormal accruals decreased by 3.9% of lagged assets in Year 1, 4.3% in Year 2, and 3.4% in Year 3, relative to the pre-period.¹² Table 4 also reports that the coefficient on YEAR3 is greater than that on YEAR2, and the coefficient on YEAR1 is not significantly different from that on YEAR2. This suggests that the decrease in abnormal accruals is greatest in Years 1 and 2.

Table 4.

OLS Regression of Abnormal Accruals on Event Period. (Dependent Variable =|DACC|).

Variable	(1) Accelerated filers			(2) Difference-in-differences			(3) Size deciles 1 to 3
	Coefficient		t-stat	Coefficient		t-stat	Coefficient		t-stat
INTERCEPT	0.127	***	9.65	0.107	***	7.42	0.263	***	3.11
YEAR1	−0.039	***	−14.62	−0.039	***	−14.44	−0.073	***	−6.86
YEAR2	−0.043	***	−15.47	−0.042	***	−14.93	−0.055	***	−4.59
YEAR3	−0.034	***	−11.55	−0.033	***	−11.14	−0.056	***	−4.71
NON				0.013	***	3.04	0.003		0.26
YEAR1_NON				0.018	***	3.11	0.033	**	2.46
YEAR2_NON				0.007		1.25	−0.011		−0.79
YEAR3_NON				0.005		0.76	0.012		0.85
Log(ASSETS)	−0.006	***	−6.81	−0.004	***	−5.41	0.012	***	3.47
LOSS	0.004		1.32	−0.004		−1.61	−0.012	**	−2.17
GC	0.033	***	3.23	0.044	***	7.96	0.038	***	4.48
ICW	0.007		1.39	0.015	***	3.90	0.032	***	4.23
CFO	0.001		0.29	−0.002	***	−2.76	−0.004	***	−3.37
LEV	−0.011	**	−2.03	0.001		0.39	0.012		1.55
LIT	0.005		1.16	0.008	*	1.90	0.003		0.36
GROWTH	0.003	***	3.16	0.001	**	2.49	0.002		1.45
FOREIGN	−0.005	**	−2.09	−0.006	**	−2.47	0.004		0.66
Log(SEGMENT)	−0.002		−0.94	−0.004	**	−2.50	−0.005		−1.15
Lag_ACCR	0.000		−1.37	0.000		−0.59	0.000	***	−7.86
TENURE	0.000	*	−1.71	0.000	*	−1.79	−0.001		−1.43
FEE_PRESSURE	−0.016	***	−7.90	−0.016	***	−8.11	0.006		1.15
CHG_AU	0.002		0.31	0.001		0.33	0.016	*	1.72
RCVB_SD	0.000	***	−2.70	0.000	***	−2.63	0.000		−1.37
CFO_SD	−0.002	**	−2.33	−0.002	**	−2.04	−0.009	**	−2.10
Audit firm fixed effects	Included			Included			Included
Industry fixed effects	Included			Included			Included
F-test:
YEAR1=YEAR2			2.43
YEAR2=YEAR3		***	15.29
YEAR1 + YEAR1_NON				−0.022	***	23.45	−0.040	***	31.47
YEAR2 + YEAR2_NON				−0.035	***	54.24	−0.066	***	73.13
YEAR3 + YEAR3_NON				−0.029	***	33.46	−0.044	***	34.95
N	9,316			14,981			4,345
F-Stat	10.78	***		21.46	***		16.87	***
Adjusted R²	.089			.110			.075

Note.***, **, and * denote statistical significance at the 1%, 5%, and 10% levels, respectively, using two-tailed tests. T-statistics are estimated from standard errors clustered by client, following Petersen (2009). The sample covers fiscal year-ends from November 15, 2005, through November 14, 2010. Column (1) is restricted to accelerated filers. Column (2) includes both accelerated and non-accelerated filers. Column (3) includes both accelerated and non-accelerated filers limited to the first through the third smallest deciles of lagged total assets. The dependent variable is |DACC|, the absolute value of abnormal accruals based on the performance-adjusted modified Jones model (Kothari et al., 2005). YEAR1 through YEAR3 are indicator variables for the three one-year periods ending November 14, 2008, through November 14, 2010, respectively (i.e., the event period). We predict that |DACC| should not change for accelerated filers in the event period, compared with the pre-period. The F-tests test the null hypothesis that the coefficients are equal or that the sum of the coefficients is zero. The appendix defines all other variables. All continuous variables are winsorized at the first and 99th percentiles. OLS = ordinary least squares.

Variables of interest are indicated with bold font.

For robustness, we again employ a difference-in-differences design. Column (2) reports that abnormal accruals for accelerated filers are lower in the event period, relative to the pre-period, as indicated by the negative and significant coefficients on YEAR1, YEAR2, and YEAR3. Abnormal accruals also declined for non-accelerated filers, as evidenced by the negative and significant sums of the coefficients on YEAR1 and YEAR1_NON, YEAR2 and YEAR2_NON, and YEAR3 and YEAR3_NON. The decline in abnormal accruals is greater for accelerated filers than for non-accelerated filers only in Year 1, as indicated by the positive and significant coefficient on YEAR1_NON and by the insignificant coefficients on YEAR2_NON and YEAR3_NON.

Table 2, Panel B, reports that mean abnormal accruals are lower for accelerated filers than for non-accelerated filers. To permit a closer comparison of the two filer types, we restrict the sample to the first, second, and third smallest deciles of lagged total assets of the pooled sample. Table 4, Column (3), reports results that are similar to those from the difference-in-differences test using the pooled sample (Column (2)). Thus, our inferences are robust to the use of a sample of clients from the first through the third size deciles.

In summary, the results suggest that abnormal accruals decreased for accelerated filers following the coinciding events of AS5 adoption, the stricter PCAOB inspection regime, and the Great Recession. The magnitude of the decrease in abnormal accruals for accelerated filers is greatest in Years 1 and 2. Abnormal accruals decreased for non-accelerated filers in each of the three event years. The magnitude of the decrease in abnormal accruals is greater for accelerated filers than for non-accelerated filers in Year 1, but there is no difference in the magnitude of decrease between the two filer types in Years 2 and 3. These results suggest that audit firm-wide improvements to audit quality affected accelerated filers and non-accelerated filers, and the magnitude of the improvement is similar between the two filer types over most of the event period.

Misstatements

Our second measure of audit quality is misstatements, which reflects material errors that the auditor failed to initially detect and report upon. This measure is consistent with DeAngelo’s (1981) definition of audit quality and is a common proxy for audit quality (Cassell et al., 2020; Lisic et al., 2019). Misstatements are a more direct and egregious measure of audit quality than abnormal accruals and may indicate errors in auditor judgment; however, the absence of a restatement may also mean that misstatements are undetected (DeFond & Zhang, 2014). According to the PCAOB (2007a, 2007b), AS5 should reduce misstatements by more effectively identifying material weaknesses; however, stricter PCAOB inspections could identify more prior misstatements.

The dependent variable in our multivariate analysis, MISSTATE, equals 1 if the client makes material misstatements in the fiscal period and 0 otherwise. We obtain misstatements from the Audit Analytics Advanced Restatement Feed on February 3, 2016, using restated fiscal years from November 15, 2005, through November 14, 2010. To ensure that the dependent variable is measuring material misstatements, we only code “Big R” misstatements as 1. “Big R” misstatements occur when the auditor restates the audit opinion. Clients report “Big R” restatements in a Form 8-K filing (item 4.02).¹³ Table 2, Panel C, indicates that the mean of MISSTATE is 0.031 for accelerated filers and is 0.032 for non-accelerated filers, which is comparable to figures in Rowe and Sivadasan (2018).

To test whether the likelihood of misstatement changed in the event period, compared with the pre-period, we estimate Equation 3 using logistic regression and the accelerated filer sample.

\begin{matrix} P (MISSTATE = 1) = φ_{0} + φ_{1} YEAR 1 + φ_{2} YEAR 2 + φ_{3} YEAR 3 + φ_{4} Log (ASSETS) \\ + φ_{5} Log (AF) + φ_{6} Log (NON__AF) + φ_{7} LEV + φ_{8} LOSS + φ_{9} MtB \\ + φ_{10} ARINV + φ_{11} NSEG + φ_{12} NEWDEBT + φ_{13} BIG 4 + φ_{14} ICW \\ + φ_{15} LIT + φ_{16} CHG__AU + φ_{17} FEE__PRESSURE \\ + Audit firm fixed effects + Industry fixed effects + ε . \end{matrix}

(3)

We include control variables following prior studies (Ettredge et al., 2014; E. Johnson et al., 2018; Newton et al., 2013). The descriptive statistics (Table 2, Panel C) are similar to prior studies.

Table 5, Column (1), reports the results from estimating Equation 3. Z-statistics are estimated from standard errors clustered by client. Control variable coefficients are generally consistent with those from prior studies. The coefficients on YEAR1, YEAR2, and YEAR3 are negative and significant, and thus we reject H2. Therefore, the likelihood of material misstatement for accelerated filers is lower in the event period, compared with the pre-period. The results are economically significant. The likelihood of a material misstatement decreased by between 1.6% in Year 1 to 2.1% in Year 3, compared with the pre-period.¹⁴

Table 5.

Logistic Regression of Misstatements on Event Period. (Dependent Variable: MISSTATE).

Variable	(1) Accelerated filers			(2) Difference-in-differences
	Coefficient		Z-stat	Coefficient		Z-stat
INTERCEPT	−3.172	***	−5.30	−2.652	***	−4.94
YEAR1	−0.593	***	−4.11	−0.605	***	−4.44
YEAR2	−0.693	***	−4.14	−0.731	***	−5.24
YEAR3	−0.686	***	−3.84	−0.734	***	−5.25
NON				−0.670	***	−3.56
YEAR1_NON				−0.164		−0.52
YEAR2_NON				−0.059		0.10
YEAR3_NON				0.423	**	2.16
Log(ASSETS)	−0.131	*	−1.74	0.069	*	1.77
Log(AF)	−0.016		−0.15	−0.195	**	−2.42
Log(NON_AF)	0.012		0.63	0.004		−0.31
LEV	0.488	**	2.02	−0.003		0.02
LOSS	0.214		1.35	0.202		1.26
MtB	0.000		−0.16	0.000		−0.08
ARINV	−0.292		−0.58	0.081		−0.32
NSEG	−0.022		−0.45	−0.013		−0.62
NEWDEBT	0.120		0.83	0.134	*	1.69
BIG4	−0.807	**	−2.52	−1.000	***	−3.44
ICW	1.373	***	8.43	1.112	***	5.87
LIT	0.333		1.13	0.103		0.83
CHG_AU	−0.285		−1.17	−0.007		−0.02
FEE_PRESSURE	0.045		0.38	−0.003		−0.06
Audit firm fixed effects	Included			Included
Industry fixed effects	Included			Included
F-test:
YEAR1=YEAR2			0.43
YEAR2=YEAR3			0.00
YEAR1 + YEAR1_NON				−0.768	***	14.99
YEAR2 + YEAR2_NON				−0.790	***	14.63
YEAR3 + YEAR3_NON				−0.311	*	3.40
N	10,365			16,788
Likelihood Ratio	307.23	***		302.12	***
Pseudo R²	.100			.062

Note.***, **, and * denote statistical significance at the 1%, 5%, and 10% levels, respectively, using two-tailed tests. Z-statistics are estimated from standard errors clustered by client, following Rogers (1993). The sample covers fiscal year-ends from November 15, 2005, through November 14, 2010. The dependent variable (MISSTATE) is an indicator variable equal to 1 if the financial report was restated by a material amount (excluding “Little R” misstatements) and 0 otherwise. YEAR1 through YEAR3 are indicator variables for the three one-year periods ending November 14, 2008, through November 14, 2010, respectively (i.e., the event period). We predict that the probability of MISSTATE=1 should not change for accelerated filers in the event period, compared with the pre-period. The F-tests test the null hypothesis that the coefficients are equal or that the sum of the coefficients is zero. The appendix defines all other variables. All continuous variables are winsorized at the first and 99th percentiles.

Variables of interest are indicated with bold font.

Column (2) reports the results from the difference-in-differences estimation. The likelihood of material misstatement for accelerated filers is lower in the event period, compared with the pre-period, as evidenced by the negative and significant coefficients on YEAR1, YEAR2, and YEAR3. The likelihood of material misstatement is also lower for non-accelerated filers in Years 1 and 2, based on the negative and significant sums of the coefficients on YEAR1 and YEAR1_NON and YEAR2 and YEAR2_NON. However, in Year 3, the likelihood of material misstatement is not significantly different from the pre-period. The magnitude of the decrease in the likelihood of material misstatement is only greater for accelerated filers than non-accelerated filers in Year 3, as evidenced by the positive and significant coefficient on YEAR3_NON. The magnitude of the decrease is not significantly different between the two filer types in Years 1 and 2.

In summary, our results indicate that the likelihood of material misstatement decreased for accelerated filers following the coinciding events of AS5 adoption, the stricter PCAOB inspection regime, and the Great Recession. The likelihood of material misstatement decreased for non-accelerated filers in Years 1 and 2, but did not change in Year 3. The magnitude of the decrease is greater for accelerated filers than for non-accelerated filers only in Year 3. These results suggest that audit firm-wide improvements to audit quality affected accelerated and non-accelerated filers, and the magnitude of the improvement is similar between the two filer types during most of the event period.

Robustness Tests

We conduct several robustness tests, but for brevity, we focus on the accelerated filers.¹⁵ Specifically, we discuss the coefficient signs on YEAR1, YEAR2, and YEAR3 in difference-in-differences tests and regression tests using only the accelerated filer sample.

Client Fixed Effects

To control for time-invariant omitted variables, we employ client fixed effects instead of industry fixed effects. A trade-off of using client fixed effects is that the sample size is substantially reduced (and potentially less representative) when the dependent variable is MISSTATE (16,788 to 1,398), and statistical power decreases. We employ a difference-in-differences design, as in our main analysis (Column (2) of Tables 3, 4, and 5), and we replace industry fixed effects with client fixed effects (untabulated). For the audit fee analysis, we find that audit fees decreased for accelerated filers in Year 3 but there is no change in Years 1 and 2. For the audit quality analysis, we find that abnormal accruals and misstatement likelihood decreased for accelerated filers in Years 1, 2, and 3.¹⁶ In summary, except for Years 1 and 2 in the audit fee analysis, our audit fee and audit quality inferences are robust to replacing industry fixed effects with client fixed effects.

PCAOB Triennially Inspected Firms and Annually Inspected Firms—Audit Fees

E. Johnson et al. (2018) find that audit fees decrease for annually inspected audit firms that eventually had their PCAOB Part II report made public for significant quality control deficiencies. To control for the effects of the eventual Part II report public release, we partition our sample between clients of triennially inspected audit firms (100 or fewer issuer clients) and annually inspected audit firms (more than 100 issuer clients). We employ a difference-in-differences design (untabulated). For triennially inspected audit firms, we find that audit fees did not change for accelerated filers in Years 1, 2, and 3. However, for annually inspected audit firms, we find that audit fees decreased for accelerated filers in Years 1, 2, and 3.

As a further test (untabulated), we restrict the sample to annually inspected audit firms and add an indicator variable to control the eventual release of a Part II report (REMEDY). We also control for the same variable in the following year with an indicator variable (POST_I). Deloitte, PwC, and Grant Thornton are the only annually inspected firms in the event period where REMEDY=1. We find that audit fees decreased for accelerated filers in Years 1, 2, and 3. In summary, clients of annually inspected audit firms drive our main audit fee results, and our audit fee inferences are robust to controlling for Part II reports that were later released to the public for annually inspected audit firms.

PCAOB Triennially Inspected Firms and Annually Inspected Firms—Audit Quality

Next, we partition the samples of abnormal accruals and misstatements between triennially inspected audit firms and annually inspected audit firms and employ a difference-in-differences design (untabulated). For triennially inspected audit firms, we find that abnormal accruals and misstatement likelihood did not change for accelerated filer clients in Years 1, 2, and 3. For annually inspected audit firms, abnormal accruals and misstatements decreased for accelerated filer clients in Years 1, 2, and 3.¹⁷

As a further test (untabulated), we restrict the sample to annually inspected audit firms and add a variable for the number of PCAOB Part I report deficiencies reported in the year to the audit firm (PUBP1), following E. Johnson et al. (2018). We find comparable results to our test of annually inspected audit firms. In summary, clients of annually inspected audit firms drive our main audit quality results, and our audit quality inferences are robust to controlling for the number of Part I deficiencies for annually inspected audit firms.

Matched-Pair Test—Audit Fees and Audit Quality

Next, we employ a matched-pair sample and a difference-in-differences design, which better ensures that audit fee trends in the pre-period are the same for accelerated filers and non-accelerated filers (untabulated). We match the two filer types (with replacement) based on the change in audit fees in the pre-period (within 10%) and on two-digit SIC code. For the audit fee analysis, we find that audit fees decreased for accelerated filers in Year 3, but there was no change in Years 1 and 2. For the audit quality analysis, we find that abnormal accruals and misstatement likelihood decreased for accelerated filers in Years 1, 2, and 3.¹⁸ In summary, except for Years 1 and 2 in the audit fee analysis, our audit fee and audit quality inferences are robust to using a matched-pair sample.

Meeting or Beating Consensus Analysts’ Earnings Forecasts

Next, as an additional measure of audit quality, we assess the likelihood of meeting or beating consensus analysts’ earnings per share forecasts within one penny. We follow specifications in prior studies (Ettredge et al., 2014; E. Johnson et al., 2018; Wang & Zhou, 2012) and estimate the following logistic regression model:

\begin{matrix} P (MEET = 1) = λ_{0} + λ_{1} YEAR 1 + λ_{2} YEAR 2 + λ_{3} YEAR 3 + λ_{4} Log (ASSETS) \\ + λ_{5} LOSS + λ_{6} GC + λ_{7} ICW + λ_{8} CFO + λ_{9} LEV + λ_{10} LIT \\ + λ_{11} GROWTH + λ_{12} FOREIGN + λ_{13} Log (SEGMENT) \\ + λ_{14} TENURE + λ_{15} Log (NUMEST) + λ_{16} STDEV \\ + λ_{17} CHG__AU + λ_{18} FEE__PRESSURE + Audit firm fixed \\ effects + Industry fixed effects + ε . \end{matrix}

(4)

MEET is equal to 1 if reported earnings meet or beat consensus analysts’ earnings per share forecasts by one penny or less, and 0 otherwise. We estimate Equation 4 only for accelerated filers (n = 7,277, untabulated) because the non-accelerated filer sample (n = 832) lacks a sufficient number of comparable control firms to conduct a valid difference-in-differences test. We find that the coefficients on YEAR2 and YEAR3 are negative and significant, whereas the coefficient on YEAR1 is not significant. In summary, except for in Year 1, our audit quality inferences are robust to using a consensus analysts’ earnings forecast benchmark as a measure of earnings quality.

Balanced Panel Design

Finally, to control for attrition of clients, we estimate Equations 1, 2, and 3 using a balanced panel of accelerated filers (untabulated). For the audit fee analysis, we find that audit fees decreased for accelerated filers in Years 2 and 3 but there was no change in Year 1. For the audit quality analysis, we find that abnormal accruals and misstatement likelihood decreased for accelerated filers in Years 1, 2, and 3. In summary, except for Year 1 in the audit fee analysis, our audit fee and audit quality inferences are robust to the use of a balanced panel sample.

Conclusion

We investigate whether the concurrence of AS5, the stricter PCAOB inspection regime, and the Great Recession affected audit fees and audit quality of accelerated filers in the 3 years from November 15, 2007, through November 14, 2010. Using a larger sample window than prior studies, we compare the 3-year event period to the 2-year pre-period. We find a decrease in audit fees and an increase in audit quality (abnormal accruals and “Big R” misstatement likelihood) in our main tests.

Our findings are, however, subject to at least two limitations. First, we cannot directly measure audit quality; thus, we rely on the literature to support the validity of multiple proxies. Second, we cannot rule out other possible reasons for improved audit quality, such as the concurrent adoption of different PCAOB audit standards and the private release of Part II reports to audit firms. However, we mitigate the effects of two major regulatory events by ending our sample period in November 2010, prior to the implementation of Audit Standard 12, Identifying and Assessing Risks of Material Misstatement (PCAOB, 2010a) and the first public release of a PCAOB Part II report of an annually inspected audit firm (Norris, 2011). In addition, in our robustness tests for the audit fee analysis, we control for the private release of PCAOB Part II reports to annually inspected audit firms that were eventually released to the public.

Overall, our results contribute to the literature by suggesting that the concurrent events of AS5, the stricter PCAOB regime, and the Great Recession coincided with lower audit fees and higher audit quality for accelerated filers of annually inspected audit firms. Therefore, our evidence supports claims made by the PCAOB and the SEC that AS5 protects investors’ interests while reducing compliance costs (U.S. SEC, 2007).

Footnotes

Appendix

Acknowledgements

We would like to thank the editor, Linda Myers, anonymous reviewers, Seunghwa Rho for her econometric assistance, Laura Alford, Jasmijn Bol, Mark DeFond, Kris Hoang, Rani Hoitash, Natalia Kochetova-Kozloski, Jagan Krishnan, David Lont, Bill Messier, Santanu Mitra, Gary Monroe, Stephen Rowe, Padmakumar Sivadasan, Dechun Wang, Tulane University workshop participants, Louisiana State University workshop participants, and participants at the 2014 American Accounting Association Annual Conference, the 2015 European Accounting Association Annual Congress, and the 2017 JAAF Conference for their helpful comments. Kenneth Reichelt and Jared Soileau acknowledge funding from Louisiana State University. Elizabeth Johnson acknowledges funding from Florida Gulf Coast University and Louisiana State University.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

Kenneth J. Reichelt

Data Availability

All data are available from public sources identified in the text.

Notes

References

Ashbaugh-Skaife

Collins

D. W.

Kinney

W. R.

Jr. LaFond

(2008). The effect of SOX internal control deficiencies and their remediation on accrual quality. The Accounting Review, 83(1), 217-250.

Bell

T. B.

Doogar

Solomon

(2008). Audit labor usage and fees under business risk auditing. Journal of Accounting Research, 46(4), 729-760.

Berkowitz

Rampell

(2002, August 9). Drive-by audits have become too common and too dangerous. The Wall Street Journal. http://online.wsj.com/article/SB1028822538710052160.html

Boone

J. P.

Khurana

I. K.

Raman

K. K.

(2015). Did the 2007 PCAOB disciplinary order against Deloitte impose actual costs on the firm or improve its audit quality? The Accounting Review, 90(2), 405-441.

Bordonaro

(2008, September 22). Deloitte mum on layoffs—Big four accounting firm won’t explain the impact on Connecticut. Hartford Business.com. http://www.hartfordbusiness.com/article/20080922/PRINTEDITION/309229995/deloitte-mum-on-layoffs–big-four-accounting-firm-wont-explain-impact-on-connecticut

Burns

(2007, December 11). Deloitte receives $1 million fine. The Wall Street Journal, C8.

Cassell

C. A.

Hansen

J. C.

Myers

L. A.

Seidel

T. A.

(2020). Does the timing of auditor changes affect audit quality? Evidence from the initial year of the audit engagement. Journal of Accounting, Auditing & Finance, 35(2), 263-289.

Chi

Myers

L. A.

Omer

T. C.

Xie

(2017). The effects of audit partner pre-client and client-specific experience on audit quality and on perceptions of audit quality. Review of Accounting Studies, 22(1), 361-391.

DeAngelo

L. E.

(1981). Auditor size and audit quality. Journal of Accounting & Economics, 3(3), 183-199.

10.

DeFond

M. L.

Lennox

C. S.

(2017). Do PCAOB inspections improve the quality of internal control audits? Journal of Accounting Research, 55(3), 591-627.

11.

DeFond

M. L.

Zhang

(2014). A review of archival auditing research. Journal of Accounting & Economics, 58(2–3), 275-326.

12.

Deloitte. (2010). Advancing quality through transparency. www.cs.trinity.edu/rjensen/temp/DeloitteTransparency%20Report.pdf

13.

Doogar

Sivadasan

Solomon

(2010). The regulation of public company auditing: Evidence from the transition to AS5. Journal of Accounting Research, 48(4), 795-814.

14.

Elder

Zhang

Zhou

(2009). Internal control weaknesses and client risk management. Journal of Accounting, Auditing & Finance, 24(4), 543-579.

15.

Ettredge

Fuerherm

E. E.

(2014). Fee pressure and audit quality. Accounting, Organizations, and Society, 39(4), 247-263.

16.

Glover

S. M.

Prawitt

D. F.

Taylor

M. H.

(2009). Audit standard setting and inspection for U.S. public companies: A critical assessment and recommendations for fundamental change. Accounting Horizons, 23(2), 221-237.

17.

Johnson

(2005, April 9). Public companies complain to SEC about audit costs. The Washington Post, E01. http://www.washingtonpost.com/wp-dyn/articles/A38809-2005Apr8.html

18.

Johnson

Reichelt

K. J.

Soileau

J. S.

(2018). No news is bad news: Do PCAOB part II reports have an effect on annually inspected firms’ audit fees and audit quality? Journal of Accounting Literature, 41, 106-126.

19.

Kothari

S. P.

Leone

A. J.

Wasley

C. E.

(2005). Performance matched discretionary accrual measures. Journal of Accounting and Economics, 39(1), 163-197.

20.

KPMG. (2009). Our audit methodology. http://www.kpmg.com/us/en/services/audit/pages/default.aspx

21.

Krishnan

Song

(2011). The effect of Auditing Standard No. 5 on audit fees. Auditing: A Journal of Practice & Theory, 30(4), 1-27.

22.

Lisic

L. L.

Myers

L. A.

Seidel

T. A.

Zhou

(2019). Does audit committee accounting expertise help to promote audit quality? Evidence from auditor reporting of internal control weaknesses. Contemporary Accounting Research, 36(4), 2521-2553.

23.

National Bureau of Economic Research. (2010, September 20). The business cycle dating committee. http://www.nber.org/cycles/sept2010.html

24.

Newton

N. J.

Wang

Wilkins

M. S.

(2013). Does a lack of choice lead to lower quality? Evidence from auditor competition and client restatements. Auditing: A Journal of Practice & Theory, 32(3), 31-67.

25.

Norris

(2011, October 17). Accounting board criticizes Deloitte’s auditing system. The New York Times. http://www.nytimes.com/2011/10/18/business/accounting-board-criticizes-deloittes-auditing-system.html

26.

Norris

(2012, August 23). Bad grades are rising for auditors. The New York Times. http://www.nytimes.com/2012/08/24/business/bad-grades-rising-at-audit-firms.html?_r=0

27.

Petersen

M. A.

(2009). Estimating standard errors in finance panel data sets: Comparing approaches. The Review of Financial Studies, 22(1), 435-480.

28.

Plumb

(2009, May 11). Big accounting firms trim their local payrolls. Washington Business Journal. http://www.bizjournals.com/washington/stories/2009/05/11/story6.html?page=all

29.

Public Company Accounting Oversight Board. (2004). Auditing Standard No. 2. An audit of internal control over financial reporting performed in conjunction with an audit of financial statements (PCAOB Release No. 2004-001). https://pcaobus.org/Standards/Archived/Pages/Auditing_Standard_2.aspx

30.

Public Company Accounting Oversight Board. (2006, May 17). Board announces four-point plan to improve implementation of internal control reporting requirements [News release]. https://pcaobus.org/News/Releases/Pages/05172006_FourPointPlan.aspx

31.

Public Company Accounting Oversight Board. (2007a, June 12). Auditing Standard No. 5—An audit of internal control over financial reporting that is integrated with an audit of financial statements and related independence rule and conforming amendments [PCAOB Release No. 2007-05A]. https://pcaobus.org/Rulemaking/Docket%20021/2007-06-12_Release_No_2007-005A.pdf

32.

Public Company Accounting Oversight Board. (2007b, May 24). Board approves new audit standard for internal control over financial reporting and, separately, recommendations on inspection frequency rule [News release]. http://pcaobus.org/News/Releases/Pages/05242007_BoardApprovesNewAuditStandard.aspx

33.

Public Company Accounting Oversight Board. (2007c, December 10). PCAOB issues disciplinary orders against Deloitte & Touche LLP and a former audit partner [News release]. https://pcaobus.org/News/Releases/Pages/12102007_DisciplinaryOrders.aspx

34.

Public Company Accounting Oversight Board. (2008, March 31). Implementing Auditing Standard No 5 (Speech by Daniel L. Goelzer, Board Member, PCAOB AS5 training event, Washington, D.C.). https://pcaobus.org/News/Speech/Pages/03312008_GoezlerAuditingStandardNo5.aspx

35.

Public Company Accounting Oversight Board. (2010a). Auditing Standard No. 12. Identifying and assessing risks of material misstatement (PCAOB Release No. 2010-004). https://pcaobus.org/Standards/Archived/PreReorgStandards/Pages/Auditing_Standard_12.aspx

36.

Public Company Accounting Oversight Board. (2010b, December 7). The public company accounting oversight board—Recent accomplishments and 2011 agenda [Speech by Daniel L. Goelzer, Acting Chairman, AICPA national conference on SEC and PCAOB developments, Washington, D.C.). http://pcaobus.org/News/Speech/Pages/12072010_GoelzerAICPAConference.aspx

37.

Public Company Accounting Oversight Board. (2013, May 23). In the matter of Ernst & Young LLP’s quality control remediation submission (PCAOB Release No. 104-2013-087). https://pcaobus.org/Inspections/Documents/05232013_EYReportStatement.pdf

38.

Rogers

W. H.

(1993). SG17: Regression standard errors in clustered samples. Stata Technical Bulletin, 13, 19-23.

39.

Rosenberg

(2010, March 9). Recession flattens accounting firms in ’09. CPA Trendlines. http://cpatrendlines.com/2010/03/09/recession-flattens-accounting-firms-in-09/

40.

Rowe

S. P.

Sivadasan

(2018). Higher audit quality and higher restatement rates: An examination of big four auditee restatements [Working paper]. Tulane University.

41.

Schroeder

J. H.

Shepardson

M. L.

(2016). Do SOX 404 control audits and management assessments improve overall internal control system quality? The Accounting Review, 91(5), 1513-1541.

42.

Tan

C. E. L.

Young

S. M.

(2015). An analysis of “Little r” restatements. Accounting Horizons, 29(3), 667-693.

43.

U.S. House of Representatives. (2002). Sarbanes-Oxley Act of 2002. Government Printing Office.

44.

U.S. Securities and Exchange Commission. (2007, July 25). SEC approves PCAOB Auditing Standard No. 5 regarding audits of internal control over financial reporting; adopts definition of “significant deficiency” (News Release 2007-144). https://www.sec.gov/news/press/2007/2007-144.htm

45.

U.S. Securities and Exchange Commission. (2010, December 6). Speech by SEC staff: Remarks before the 2010 AICPA National Conference on Current SEC and PCAOB developments (James L. Kroeker, Chief Accountant, Washington, D.C.). https://www.sec.gov/news/speech/2010/spch120610jlk.htm

46.

Wang

Zhou

(2012). The impact of PCAOB Auditing Standard No. 5 on audit fees and audit quality. Accounting Horizons, 26(3), 493-511.

47.

Wegman

(2008). Government regulation of accountants: The PCAOB enforcement process. Journal of Legal, Ethical and Regulatory Issues, 11(1), 75–94. https://www.abacademies.org/articles/jlerivol11no12008.pdf

48.

Weil

(2004, March 25). Behind wave of corporate fraud: A change in how auditors work; “Risk based” model narrowed focus of their procedures, leaving room for trouble. The Wall Street Journal. https://www.wsj.com/articles/SB108017085819664786

49.

Zmijewski

M. E.

(1984). Methodological issues related to the estimation of financial distress prediction models. Journal of Accounting Research, 22(Suppl.), 59-82.