PCAOB Inspections and Audit Fees: An Analysis of Inspection Rounds of Small Audit Firms

Abstract

This study investigates the impact on audit pricing of PCAOB inspection outcomes on the quality control systems of small audit firms. Quality control deficiencies (QCDs) provide a strong signal about audit quality, which is expected to lead to price changing. We first document that audit fees in the pre-inspection period were higher for small audit firms without QCDs, implying that quality differences were already partially known and priced in the market. Next, we find evidence that in the post-inspection period audit fees further increase for firms without QCDs, while there is limited evidence for fee increases for firms with remediated QCDs, suggesting changes in audit firm reputation. Interestingly, we do not find support for fee changes for audit firms with disclosed quality control deficiencies. Finally, the magnitude of change in audit fees decreases over time. Collectively, these findings contribute to our understanding of audit quality differences amongst small audit firms and the functioning of the small audit firm market.

Keywords:

• PCAOB inspections
• Audit fees
• Quality control deficiencies
• Small audit firms

1. Introduction

In the past 15 years we have seen a shift worldwide from the dominant model of self-regulation of the auditing profession to the installment of public oversight. This profound change in audit regulation was triggered by high-profile financial reporting scandals in which auditors were also blamed (e.g., Enron and Worldcom in the US, Lernaut and Hauspie in Belgium, Parmalat in Italy) (Offermanns & Vanstraelen, 2014). The US was the frontrunner in creating a public oversight authority, the Public Company Accounting Oversight Board (PCAOB), by means of the Sarbanes-Oxley Act in 2002. Many other countries subsequently followed the US by establishing public oversight bodies. For example, the European Commission required member states to install independent oversight over financial reporting and auditing in its revised Eighth Directive of 2006. In contrast to the US, public oversight in Europe relates to audit firms with and without public interest entities (PIEs), and the large majority of clients of PIE audit firms in Europe are private companies which are subject to a mandatory audit. One of the major tasks of public oversight bodies is to conduct periodic independent inspections of audit firms. The aim of the inspections is to improve audit quality and protect the interests of investors. Given the amount of resources that these inspections require, investigating whether inspections are effective in achieving this mission and identifying their economic consequences are important questions for academics, practitioners, regulators and society at large. This explains the growing body of research on public oversight. Interestingly, there is substantial variation in inspection regimes across countries, also within the EU, including whether or not inspection results are publicly disclosed (e.g., Carson et al., 2017; Florou & Shuai, 2017; García Osma et al., 2017). Since the US is one of the few public oversight bodies that publicly discloses inspection findings at the individual audit firm level, research on public oversight mainly relates to the PCAOB.¹ Furthermore, the PCAOB already has a long history of public disclosure of inspection reports, which allows to investigate economic consequences of public oversight over time. In this study, we make use of this data of multiple inspection rounds to investigate the impact on audit fees of PCAOB inspection outcomes on the audit firm's quality control system. Taking into account institutional differences, investigating disclosure of audit firm's inspection findings over time in the US specific setting is relevant as a benchmark for other countries that consider or have implemented changes to improve their national inspection regime. For example, a number of European countries have made changes in the past years in the organization of public oversight, which make their inspection regime more similar to the inspection regime of the PCAOB (e.g., change from inspections through oversight to direct inspections).

Prior research on the PCAOB has focused primarily on annually (i.e., large) inspected US and foreign audit firms and how these inspections affect perceived and actual audit quality output factors including financial reporting quality and auditor reporting (e.g., Aobdia, 2019a; Boone et al., 2018; Carcello et al., 2011; Fung et al., 2017; Gipper et al., 2017; Krishnan et al., 2017; Lamoreaux, 2016). Next to audit output factors, recent research started to investigate audit input factors (i.e., audit hours and fees) for annually inspected audit firms by the PCAOB (Acito et al., 2018; Aobdia, 2018). Overall, there is growing evidence to support the theory that PCAOB inspections improve actual and perceived audit quality of annually inspected audit firms.

Research on annually inspected audit firms essentially relates to engagement related deficiencies (i.e., Part I findings) identified by the PCAOB given the little variation in quality control related deficiencies (i.e., Part II findings), which are identified for nearly all of the annually inspected audit firms. In contrast, there is more variation in whether or not quality control related deficiencies are identified among triennially (i.e., small) inspected audit firms by the PCAOB. Compared to engagement specific deficiencies, quality control findings are more pervasive findings and likely affect all engagements since the quality control system can be considered the foundation for the way audits are performed within the firm. Nagy (2014) documents that audit firms lose market share in a 12-month period after the public disclosure of the quality control deficiencies, suggesting reputation damage. However, it remains unclear how the quality control deficiencies affect the clients that do not switch audit firms. The purpose of this paper is to investigate the impact on audit pricing of PCAOB inspection outcomes on the quality control systems of small audit firms, while controlling for the inspection outcomes at the engagement level. Audit pricing is an important feature of the functioning and level of competition of the audit market, and we know relatively little about small audit firms, even though they play a significant role in the competitive landscape of local markets (Bills & Stephens, 2016). Compared to the oligopolistic sector dominated by the big audit firms, price competition is much more intensive in the atomistic market constituted by a large number of small audit firms (Ghosh & Lustgarten, 2006). Moreover, small audit firms are perceived as having a higher risk of losing independence due to the high competition (Shockley, 1981). While small audit firms are commonly treated as a homogeneous group, PCAOB inspection reports on the quality control systems of small audit firms provide a strong signal about their audit quality (Aobdia, 2020). A credible audit quality signal likely affects the audit firm's reputation, which is expected to lead to price changing in a competitive market.

We start our analyses with examining variation in audit fees prior to the commencement of PCAOB inspections, grouping our small auditors conditional on inspection outcomes. Next, we investigate whether inspections affect audit fees charged to clients of small audit firms, how this effect changes over time, and whether inspections have a long-term impact on audit fees. We argue that PCAOB inspections can lead to a change in audit fees of small audit firms for at least two reasons. First, the cost of remaining within the market for public company audits increases after the installment of the PCAOB inspections. This is proportionally more costly for small audit firms, and even more so for lower-quality audit firms, as it requires more effort from these firms to comply with the strict PCAOB quality control standards (DeFond & Lennox, 2011). While the costs of complying with PCAOB quality control standards may mainly relate to the beginning period of inspections, the increase in engagement costs is likely structural since more work is needed on each engagement to meet PCAOB regulatory standards. Second, audit firms face regulatory sanctions and penalties if serious deficiencies are not corrected, which may be particularly concerning for low-quality small audit firms where the likelihood that the PCAOB uncovers misconduct is higher.² Given that the market in which small audit firms operate is highly competitive, the increased costs that these firms face as a result of PCAOB inspections would be expected to lead to an increase in audit fees, and this effect may arguably even be more pronounced for low quality audit firms with quality control deficiencies.

At the same time, a number of studies show the importance of auditor reputation for providing firms with incentives to supply high quality audits (Craswell et al., 1995; Francis et al., 2005; Skinner & Srinivasan, 2012; Weber et al., 2008). For audit firms with deficient quality control inspection reports, it has been documented that an auditor's reputation can be harmed, resulting in an adverse effect on clients’ valuations (Dee et al., 2011), or the firm may exit the audit market (DeFond & Lennox, 2011). From this point of view, although they are faced with increased costs, audit firms with deficient quality controls will have difficulties in passing on these additional costs to their clients by increasing audit fees. On the contrary, PCAOB inspection reports provide audit firms without quality control deficiencies opportunities to signal their high quality. This positive reputation signal arguably increases their abilities to charge higher audit fees.

We make use of PCAOB inspection reports for small audit firms for inspection rounds published from 2005 to 2014. During this period, most of the small audit firms were inspected for at least three rounds. We manually coded all the inspection reports included in our analyses based on Part II PCAOB findings and noted whether or not the audit firm successfully remediated the identified quality control deficiencies within one year as required by the PCAOB.

An important feature of the PCAOB inspections of small audit firms is the natural staggered setting, in which the PCAOB inspects small audit firms at different points in time. This specific setting not only allows us to compare the audit fee change from pre- to post- inspection for those client companies whose auditors are already being inspected by the PCAOB, but also enables us to benchmark the fees charged to companies whose auditor is inspected at a later point in time in our sample. Both benchmarks help us to control for contemporaneous effects of other changes in the audit market that are unrelated to the publication of the PCAOB inspection reports. Moreover, we also include year- and industry-fixed effects to account for unobserved time- and industry-invariant characteristics.

Our results show that before the publication of the inspection reports, audit fees were higher on average for audit firms without quality control deficiencies. This suggests that PCAOB-identified quality control differences were already partially known and priced in the audit market before the start of the inspections. Next, we find that audit fees increase after the PCAOB inspections, and show that this increase is largely associated with audit firms without any quality control deficiencies, suggesting positive reputation effects. We find only limited evidence for fee increases for firms with remediated quality control deficiencies, implying that these audit firms experience difficulties in passing on the incurred additional costs to clients. Interestingly, we do not find evidence of fee changes for audit firms without remediated (and thus disclosed) quality control deficiencies. To further corroborate these results, we find in an additional analysis that audit firms that remediated their identified quality control deficiencies have increased their number of CPAs after the second-round inspection, which provides some evidence that these firms seek to increase audit effort. However, combined with the main findings that fee increases for these firms are limited, this would suggest that PCAOB-identified quality control deficiencies limit their ability to charge for the additional effort, arguably due to reputation damage. Interestingly, we find that while audit firms without quality control deficiencies are able to increase audit fees, their number of public clients apparently decreases after the third-round inspection. If a clean PCAOB inspection report is an indication of higher audit quality, this finding would provide some indication that in the small audit firm market segment, client companies focus on lowering audit fees rather than seeking higher audit quality. Finally, we document that the magnitude of change in audit fees decreases after multiple inspection rounds.

Our findings contribute to the relatively limited number of studies on the effect of PCAOB inspections of triennially inspected audit firms on audit quality by focusing on the impact on audit pricing of inspection outcomes on the quality control systems of small audit firms. Furthermore, our insights contribute to our understanding of the functioning of the small audit firm market, which is also relevant for other parts of the world where small audit firms have an important share in the statutory audit market. The insights of our study relating to triennially inspected audit firms in the US may serve as a benchmark for future studies relating to other institutional settings to the extent that data are or become available. For example, Sweden and Norway publicly disclose inspection results at the individual audit firm level as well as the UK for annually inspected audit firms and since 2014 also the Netherlands for triennially inspected Big 4 audit firms. The remainder of the paper proceeds as follows. Section 2 provides the background of the study. Next, we develop our hypotheses in Section 3. The research design is outlined in Section 4, which is followed by a discussion of the results in Section 5. Section 6 contains additional analyses and Section 7 provides conclusions and limitations of the study.

2. Background

Under the provisions of the Sarbanes-Oxley Act of 2002 (SOX), the PCAOB conducts annual inspections of firms that audit more than 100 issuers, and triennial inspections of audit firms with fewer than 100 registrant clients (the latter referred to as ‘small audit firms’ hereafter). Along with the evaluation of an audit firm's quality control policies and procedures, the inspection process involves a review of some audits that were carried out by the firm, which are chosen based on characteristics of the client, its industry, practice office, partner, or prior inspection results (PCAOB, 2009). The results of the inspection process are publicly disclosed in a report for each audit firm. While it does not disclose the identity of inspected clients, Part I of the inspection report contains information about engagement-specific deficiencies and Part II details any quality control deficiencies that have been identified. Details about quality control deficiencies are made available to the public only if the audit firm does not sufficiently address the PCAOB's concerns within a one-year period. The PCAOB is authorized to conduct disciplinary proceedings, impose sanctions, and communicate inspection results to other regulatory agencies (Ege et al., 2019; Gunny & Zhang, 2013). The PCAOB has demonstrated its power to impose sanctions for violations of standards detected via inspections by revoking the registration of audit firms and censuring, suspending, or barring auditors (Gilbertson & Herron, 2009; PCAOB, 2011).

From a conceptual point of view, researchers and practitioners have argued both for and against the effectiveness of the inspection process, i.e., whether the process is able to systematically identify meaningful audit deficiencies in a way that can lead to an improvement in audit quality. Some argue that the PCAOB inspection process is superior to the older peer review system because it is independent and objective, has better access to auditor documentation, and has more resources available for inspectors (Carcello et al., 2011; Gunny & Zhang, 2013). Others criticize the inspection process because of limited staff and expertise, inadequate transparency of procedures and inspection outcomes, the slow timing of feedback, and the small and risk-based samples leading to inspection outcomes which are arguably not representative for the audit firm portfolio (DeFond, 2010; Glover et al., 2009; Peecher et al., 2013).

This conceptual debate has served as motivation for a number of empirical studies examining the association between inspection outcomes and various proxies for audit quality. The large majority of these studies relate to annually inspected audit firms and generally show positive effects of PCAOB inspections. For example, both financial reporting quality and investors’ reactions to releases of earnings announcements of clients of large audit firms increase after the installment of PCAOB inspections (e.g., Carcello et al., 2011; Gipper et al., 2017). In addition, Aobdia (2018) recently shows that audit effort increases on inspected and non-inspected engagements of offices or partners of annually inspected audit firms with engagement quality deficiencies. Johnson et al. (2018) find that annually inspected audit firms with remediated quality control deficiencies increased audit effort in the following 12 month period but decreased audit fees at the same time due to reputation damage. However, audit fees increase again three years after the release of the Part II report.

In contrast to the growing body of research on the impact of PCAOB inspections on large audit firms, there is much less research on small audit firms and the limited evidence is also less conclusive. On one hand, there is evidence that small audit firms deregister from the PCAOB in response to inspections (DeFond & Lennox, 2011). For small audit firms that remain in the market, Gramling et al. (2011) find that those firms with PCAOB deficiencies are more likely to issue a GCO for financially distressed clients subsequent to their PCAOB inspection than prior to their inspection. Furthermore, there is some evidence that audit firms lose market share following the public disclosure of quality control deficiencies (Abbott et al., 2013; Daugherty et al., 2011; Nagy, 2014). These findings would support the notion that PCAOB oversight and inspections are effective. On the other hand, it has been documented that small audit firms do not perceive that the inspection process improved audit quality or public confidence in the audit profession (Daugherty & Tervo, 2010). Tanyi and Litt (2017) show that annually inspected non-Big 4 audit firms are more selective in their choice of new clients in the post-inspection period, compared with triennially inspected non-Big 4 audit firms. This would suggest that the impact of the PCAOB may not be equivalent between large and small audit firms.

The purpose of this study is to contribute to research on the effects of PCAOB inspections of small audit firms by investigating how inspection outcomes on the quality control systems of small audit firms affect audit fees. In contrast to annually inspected audit firms, inspection reports without any identified quality control deficiencies are more common for small audit firms. Since quality control deficiencies relate to the audit firm level, they are perceived to be more severe as they likely affect all engagements. Hence, quality control deficiencies arguably provide a stronger signal of audit quality compared to engagement-specific deficiencies. Empirical research also shows that disclosure of quality control deficiencies results in adverse effects on clients’ valuations (Dee et al., 2011), and there is some evidence of loss of market share (Abbott et al., 2013; Daugherty et al., 2011; Nagy, 2014). Since price competition is an important characteristic of the atomistic small audit firm market, a strong signal about audit quality is expected to lead to price changing if it affects the audit firm's reputation. This is especially salient given that small audit firms are commonly perceived and treated in research as a homogenous group. Motivated by the importance of small audit firms in the competitive landscape of local audit markets as well as the observed variation in quality control inspection findings amongst small audit firms, we investigate whether and how these inspection findings impact audit fees of small audit firms, while controlling for engagement-specific deficiencies.

3. Development of Hypotheses

An audit firm's quality control system aims to provide reasonable assurance that the firm's personnel comply with applicable professional standards and the firm's standards of quality (PCAOB, 2003, QC Section 20.03). The PCAOB's evaluation of a firm's system of quality control typically includes a review of policies, procedures and practices concerning audit performance, training, compliance with independence requirements, client acceptance and retention, and the establishment of policies and procedures (PCAOB, 2012). During the fieldwork, the PCAOB inspectors also dissect the audit work papers, interact frequently with the engagement team to improve their understanding of the work completed during the audit and to determine whether any engagement-specific deficiencies are related to a firm-wide quality control problem (Aobdia, 2018). As discussed above, compared to engagement level deficiencies, quality control deficiencies are identified at the audit firm level and are arguably more likely to represent audit quality and audit effort. Hence, the quality control system can be considered the foundation for the way audits are performed within the firm. In addition, compared to audit firms without quality control deficiencies, audit firms with disclosed quality control deficiencies are usually smaller in size, in terms of for example the number of partners, partners to public clients ratio and total professionals to public clients ratio (Hermanson and Houston, 2008).³ These figures indicate a potential understaffing situation, which might negatively affect audit quality. Since the market in which smaller audit firms compete is highly fragmented and competitive, audit fees are expected and have been documented to reflect audit effort (e.g., Bell et al., 2001; Menon & Williams, 2001; Niemi, 2004; Schelleman & Knechel, 2010; Simunic, 1980). Research further shows that there is a quality differentiated audit demand, and that clients interested in a high-quality audit choose a high-quality auditor and are willing to pay a fee premium (DeFond & Zhang, 2014).

As a result, we expect that audit fees are lower for clients of auditors with quality control deficiencies in the pre-inspection period (i.e., before the first inspection) compared to clients of auditors without any quality control deficiencies before the first inspection. We formally this test with our first hypothesis:

Hypothesis 1: Audit fees in the pre-inspection period are higher for clients of audit firms without quality control deficiencies.

The PCAOB's strict enforcement of compliance with auditing standards drives the costs upwards for audit firms choosing to remain auditing public clients, which is especially salient for smaller audit firms (DeFond & Lennox, 2011). Stricter compliance requires auditors to invest in a variety of practice areas that are closely monitored by the PCAOB, such as procedures for client acceptance and retention, partner compensation and review, auditor independence and staff training. Moreover, the percentage of engagements being inspected is much larger for small audit firms (Lennox & Pittman, 2010) compared with big audit firms. As PCAOB inspections have a disruptive impact on auditors’ normal activities, the examination of a higher proportion of their clients imposes a relatively greater cost on small auditors (DeFond & Lennox, 2011). Furthermore, PCAOB inspections are expected to increase the cost of an engagement, since more work is required to meet PCAOB standards. As a result, audit firms that remain in the market and audit public clients are likely facing increased costs, which could result in higher audit fees. At the same time, audit fees are not only reflective of audit effort but also of reputation (e.g., Francis et al., 2017; Hay et al., 2006; Larcker & Richardson, 2004; Lyon & Maher, 2005; Nelson, 2006; Reynolds & Francis, 2001; Schelleman & Knechel, 2010; Seetharaman et al., 2002). We argue that the extent to which an audit firm can pass on these increased costs to a client will depend on the impact of the inspection report on the audit firm's reputation.⁴ An audit firm without any quality control deficiencies is arguably more able to pass on increased costs to clients or even charge an audit fee premium, as it provides a strong positive signal about the audit firm's quality, which we test by means of the following hypothesis:

Hypothesis 2a: Clients of audit firms without identified quality control deficiencies are associated with an increase in audit fees in the post-inspection period, compared to the pre-inspection period.

For audit firms with remediated quality control deficiencies, it is harder to predict the impact on audit fees. Remediation of quality control deficiencies requires additional investments from the audit firm and higher audit effort, as the firm has to establish and implement quality control upgrades that were agreed upon as part of its settlement with the PCAOB. Higher audit effort and investments made to address identified quality control deficiencies arguably provide the audit firm with a convincing argument to increase audit fees. At the same time, it remains an empirical question whether clients would accept this approach in a highly competitive market. Indeed, even if the PCAOB does not publicly disclose the specific quality control deficiencies in the initial inspection reports, as firms are given the opportunity to remediate their quality control deficiencies, clients are still able to distinguish whether any deficiencies were identified during the inspection. This could lead to reputation damage even for the audit firms that remediated the deficiencies following the inspections.⁵ In this regard, the PCAOB encourages audit committees to discuss inspection findings with their auditor and what the audit firm is doing to remedy the problems (e.g., PCAOB, 2004, 2006a, 2006b, 2012). Prior research shows that an audit firm experiences economic losses, including lower fees, after incurring damage to its reputation (e.g., Boone et al., 2015; Davis & Simon, 1992; Francis et al., 2017; Johnson et al., 2018). Thus, instead of choosing to increase the audit effort and pay higher audit fees, client companies may focus on negotiating lower audit fees and make use of the reputational damage caused by a deficient report, especially in the small audit firm market. Hence, while audit firms with remediated deficiencies are arguably more likely to increase audit effort, it remains unclear whether they will be able to pass these higher costs to their clients because of reputation loss. Therefore, we formulate a non-directional hypothesis for audit firms with remediated quality control deficiencies:

Hypothesis 2b: There is a difference in audit fees for clients of audit firms with remediated quality control deficiencies in the post-inspection period, compared to the pre-inspection period.

For audit firms with publicly disclosed quality control deficiencies, it would seem very difficult, if not impossible, to charge higher audit fees. On the contrary, clients of these firms are likely to try to negotiate lower fees because of loss of reputation in combination with a lower perceived likelihood that these firms will increase effort, as suggested by their failure to remediate the identified quality control deficiencies. This would result in lower audit fees, which we formally test with the following hypothesis:

Hypothesis 2c: Clients of audit firms with disclosed quality control deficiencies are associated with a decrease in audit fees in the post-inspection period, compared to the pre-inspection period.

Finally, we examine the impact of inspections on audit fees over time. As discussed earlier, inspections are expected to result in increased costs. First, engagement costs are expected to increase to comply with PCAOB standards, including documentation requirements. Second, small audit firms likely need to invest in their internal quality control system to meet PCAOB standards. While the increase in engagements costs is arguably structural in nature, these costs may not necessarily further increase after each inspection round unless the PCAOB becomes stricter with each inspection round, resulting in higher compliance costs over time. Furthermore, investments in the audit firm's internal quality control system are not expected to further increase over time once an appropriate system is in place. Overall, this would imply that the impact of inspections on audit fees decreases over time. Hence, we hypothesize that:

Hypothesis 3: The magnitude of change in audit fees from pre- to post- inspection decreases over time.

4. Research Design

4.1. Sample Selection

The sample selection is based on the first three rounds of inspection of US audit firms that are inspected on a triennial basis (<100 registrant clients). We include all inspection reports available on the PCAOB website as of December 2014. The inspected audit firms are matched with their respective audit clients in Audit Analytics and financial information is retrieved from Compustat for the years 2003 through 2015. The final sample consists of observations contained in the intersection of these three data sources.⁶

To test our first hypothesis, only observations before the first inspection are included.⁷ Table 1, Panel A provides detailed information about our sample selection procedure. We start with 761 PCAOB inspection reports relating to all first-time inspected audit firms and merge them with Audit Analytics to obtain the clients sample. Next, we retrieve all the client financial information from Compustat and exclude all observations that have missing values for calculating the variables in our model. Subsequently, we delete client observations with a financial year-end after the first inspection and observations classified as financial institutions (SIC codes 6000–6700) or utilities (SIC codes 4000–4900). This results in a sample of 4970 client-year observations for 1811 clients and 417 audit firms. Table 1, Panel B displays the composition of the sample for testing our second set of hypotheses. We start again with 761 inspection reports for the first round, and further identity 537 inspection reports for the second round and 373 for the third round.⁸ After excluding audit firms that do not have data available in Audit Analytics, this results in a sample of 664 audit firms with 9628 clients for the first inspection; 485 audit firms with 8785 clients for the second inspection; and 343 audit firms with 7720 clients for the third inspection. Next, we retrieve financial data from Compustat and exclude all observations with missing values for variables in our empirical models. This yields a sample of 503 audit firms with 3271 clients for the first round; 400 audit firms with 3083 clients for the second round; and 287 audit firms with 2767 clients for the third round. For each round of inspections, we further exclude the client-year observations that have a financial year-end before the previous round and after the next round, and exclude financial institutions (SIC codes 6000–6700) and utilities (SIC codes 4000–4900).⁹ To ensure proper representation of client firms in all time periods, only auditor-client combinations that have at least one financial year-end before and after the inspection are included in the samples for testing our hypotheses. This yields a final sample of 4963 client-year observations for 1131 clients and 323 audit firms for the first-round inspection; 3886 client-year observations for 963 clients and 259 audit firms for the second-round inspection; and 3103 client-year observations for 756 clients and 174 audit firms for the third-round inspection. In addition, we construct a constant sample containing only clients that do not switch audit firm from at least one year before the publication of the first inspection report to one year after the publication of the third-round inspection report to test our third hypothesis. This sample consists of 2588 client-year observations for 275 clients and 115 audit firms.

Table 1. Sample selection procedure.

Number of Audit firms (number of clients)	Pre-first inspection
Panel A: Sample for H1
Originally from PCAOB website	761
- without data from Audit Analytics	665 (9625)
- without data from Compustat	577 (4117)
- missing values for variables used in the model	504 (3268)
- observations with year-end after the first inspection	443 (1989)
- SIC 6000–6700 and SIC4000–4900	417 (1811)
Number of Audit firms (number of clients)	First inspection	Second inspection	Third inspection
Panel B: Samples for H2
Originally from PCAOB website	761	537	373
- without data from Audit Analytics	664 (9628)	485 (8785)	343 (7720)
- without data from Compustat	576 (4120)	450 (3889)	328 (3510)
- missing values for variables used in the model	503 (3271)	400 (3083)	287 (2767)
- observations with year-end after the next inspection and before previous inspection	485 (2592)	369 (2189)	251 (1736)
- SIC 6000–6700 and SIC4000–4900	463 (2372)	358 (2041)	241 (1637)
- observations without at least one fiscal year before and one fiscal year after the inspection	323 (1131)	259 (963)	174 (756)

For each inspection report, we manually code the inspection reports as ‘NON_QCD’ if no quality control deficiency is identified during the inspection, ‘QCD’ if any quality control deficiency is identified, ‘QCD_D’ if identified quality control deficiencies are disclosed as they were not remediated within one year, and ‘QCD_ND’ if identified quality control deficiencies are not disclosed as they were remediated within one year. We also create a variable ‘POST’ to indicate whether the observation belongs to the period before or after the publication of the inspection reports.

4.2. Empirical Models

We start our analyses with a benchmark model in the pre-inspection period to examine whether audit fees differ conditional on the ex-post PCAOB inspection outcomes. Following Francis et al. (2005) and Hay et al. (2006), we use the following audit fee model using ordinary least squares regression with robust standard errors: (1) $\begin{array}{rl}LA{F}_{it}& ={\alpha }_0+{\alpha }_{1}NON\mathrm{\_}QCD+{\alpha }_{2}LOGASSET{S}_{it}+{\alpha }_{3}LEVERAG{E}_{it}+{\alpha }_{4}INVER{E}_{it}\\ & +{\alpha }_{5}RO{A}_{it}+{\alpha }_{6}LOS{S}_{it}+{\alpha }_{7}FOREIG{N}_{it}+{\alpha }_{8}BUS{Y}_{it}+{\alpha }_{9}OPINIO{N}_{it}+{\alpha }_{10}LOGSE{G}_{it}\\ & +{\alpha }_{11}SHOR{T}_{it}+{\alpha }_{12}LOGAVG\mathrm{\_}ASSET+{\alpha }_{13}LOGTOTAL\mathrm{\_}FEE+{\alpha }_{14}DEFPERCENT\\ & +{\alpha }_{15}LOGLAG+{\alpha }_{16}ISP+{\alpha }_{17}AS2+{\alpha }_{18}AS5+{\alpha }_{19}ACCER+\text{Year fixed effects}\\ & +\text{Industry fixed effects}+\epsilon \end{array}$(1) Where LAF is measured as the natural logarithm of audit fees as reported in Audit Analytics, adjusted for inflation.¹⁰ NON_QCD is our variable of interest, which is an indicator variable equal to 1 for clients of audit firms without any deficiencies identified in Part II of the inspection reports. If audit fees were already higher before the inspection for clients of auditors without Part II deficiencies, the coefficient on NON_QCD will be positive. We include LOGASSETS, the natural logarithm of total assets adjusted for inflation to control for size. To account for client risk, we include LEVERAGE, the sum of the company's current and long-term debt divided by total assets, and INVERE, the sum of inventories and receivables scaled by total assets. We expect both of them to be positively related to audit fees, as they indicate higher audit risk. We include client performance variables ROA, measured as net income divided by total assets, and LOSS, a dummy variable for a loss in the current year. As less profitable companies exhibit more financial risk, we expect audit fees to decrease with ROA and to increase with LOSS. Client complexity is measured by LOGSEG, the natural logarithm of the number of business segments reported, and we expect it to be positively related to audit fees. Additional dummy variables include OPINION, FOREIGN and BUSY, where OPINION equals 1 when a going-concern opinion is issued, FOREIGN equals 1 whenever foreign income is earned, and BUSY is set to 1 for audits where the financial year-end is in December. We expect all of them to be positively associated with audit fees. We control for the first year of the auditor-client relationship by including SHORT, which equals 1 in the first year of the auditor-client relationship, and we add two variables LOGAVG_ASSET and LOGTOTAL_FEE to control for audit firm size, calculated as the natural logarithm of the average client size of the audit firm and the natural logarithm of the total fee collected by the audit firm, respectively. Both variables are adjusted for inflation. We also include DEFPERCENT, measured by the number of deficient engagements divided by the total number of engagements inspected by the PCAOB, to control for the effect of Part I deficiencies. We further include LOGLAG, measured by the natural logarithm of the number of days between a company's fiscal year end and the date of the audit report, and expect it to be positively related to audit fees. We control for auditor industry specialization, which is expected to be positively associated with audit fees, by means of ISP which equals 1 if clients are in the two-digit SIC industry in which their audit firm has their largest portfolio share (measured by audit fees).¹¹ During our sample period, the most significant changes of auditing standards include the installments of PCAOB Auditing Standard No. 2 (referred to as AS2 hereafter) and No.5 (referred to as AS5 hereafter), which require management and the auditor to report on internal controls over financial reporting. It has been documented that the implementation of AS2 (in 2004 for accelerated filers and in 2005 for non-accelerated filers) increases audit fees significantly (Raghunandan & Rama, 2006). However, when AS2 was replaced by AS5 in 2007, audit fees decreased (Krishnan et al., 2011). Thus, we add the following control variables: AS2, AS5 and ACCER, where AS2 equals 1 for accelerated filers with fiscal year endings between November 15, 2004 and November 15, 2007 and for non-accelerated filers with fiscal year endings between July 15, 2005 and November 15, 2007, AS5 equals 1for fiscal years ending on or after November 15, 2007, and ACCER is set to 1 for accelerated filers. Finally, we control for year fixed effects and industry fixed effects for which we use the two-digit SIC code. All the variables are winsorized at the 5 and 95 percentiles.¹² The descriptions of all variables used in the empirical analyses are included in Appendix 1.

Since the PCAOB inspects different audit firms at different points in time, we exploit this staggered setting to estimate audit fee changes within small audit firms using only variation in the timing of the inspections. This mitigates concerns that the observed effect might come from contemporaneous effects or other changes in the audit market that are unrelated to the publication of the PCAOB inspection reports (i.e., parallel trends).

To test hypotheses H2a, H2b, and H2c, we change model 1 as follows: (2) $\begin{array}{rl}LA{F}_{it}& ={\alpha }_0+{\alpha }_{1}NON\mathrm{\_}QCD+{\alpha }_{2}POST+{\alpha }_{3}NON\mathrm{\_}QCD\ast POST+{\alpha }_{4}QCD\mathrm{\_}ND\\ & +{\alpha }_{5}QCD\mathrm{\_}ND\ast POST+{\alpha }_{6}LOGASSET{S}_{it}+{\alpha }_{7}LEVERAG{E}_{it}+{\alpha }_{8}INVER{E}_{it}\\ & +{\alpha }_{9}RO{A}_{it}+{\alpha }_{10}LOS{S}_{it}+{\alpha }_{11}FOREIG{N}_{it}+{\alpha }_{12}BUS{Y}_{it}+{\alpha }_{13}OPINIO{N}_{it}\\ & +{\alpha }_{14}LOGSE{G}_{it}+{\alpha }_{15}SHOR{T}_{it}+{\alpha }_{16}LOGAVG\mathrm{\_}ASSET+{\alpha }_{17}LOGTOTAL\mathrm{\_}FEE\\ & +{\alpha }_{18}DEFPERCENT\left(+{\alpha }_{19}DEFQC1/DEFQC2\right)+{\alpha }_{20}LOGLAG+{\alpha }_{21}ISP\\ & +{\alpha }_{22}DISAGREE+{\alpha }_{23}AS2+{\alpha }_{24}AS5+{\alpha }_{25}ACCER\\ & +\text{Year fixed effects}+\text{ Industry fixed effects}+\epsilon \end{array}$(2) The variables of interest are: NON_QCD, QCD_ND and POST. NON_QCD and QCD_ND are equal to 1 for clients of audit firms without any identified quality control deficiencies and clients of audit firms that remediated identified quality control deficiencies, respectively. Clients of auditors with disclosed quality control deficiencies are used as the reference group. For the second and third inspection rounds, we also control for the previous round inspection report findings, in particular whether or not the previous inspection report had any Part I or Part II deficiencies (DEFQC1/DEFQC2). All other variables are the same as in Model 1. Audit firms are allowed to provide their responses to the PCAOB findings which are publicly available and included in a response letter at the end of the inspection report. In some cases, audit firms state that they disagree with the PCAOB findings. It would seem very difficult, if not impossible, for audit firms disagreeing with the PCAOB findings, to charge higher audit fees because of loss of reputation in combination with a low likelihood that they will increase effort on individual engagements. As a result, we add an additional control variable DISAGREE, which equals 1 for clients of audit firms that disagree with the PCAOB findings in their response letters.

In order to compare changes in audit fees across the three inspection rounds, we construct a constant sample containing only clients that do not switch audit firm from at least one year before the publication of the first inspection report to one year after the publication of the third-round inspection report to test our third hypothesis. To conduct our empirical tests, we first define three dummy variables as our variables of interest for three inspection rounds: Period1, Period2 and Period3. Period1 (Period2/Period3) takes a value of 1 if the observation belongs to the period after the publication date of the first (second/third) round inspection reports and before the publication of the second (third/fourth) round inspection reports, zero otherwise. We also excluded the auditor-client combinations that do not have at least one observation in each period of time. Then we rerun our analysis using Model 1 replacing the variable NON_QCD with Period1, Period2 and Period3. Formally, the empirical model looks as follows: (3) $\begin{array}{rl}LA{F}_{it}& ={\alpha }_0+{\alpha }_{1}Period1+{\alpha }_{2}Period2+{\alpha }_{3}Period3+{\alpha }_{4}LOGASSET{S}_{it}+{\alpha }_{5}LEVERAG{E}_{it}\\ & +{\alpha }_{6}INVER{E}_{it}+{\alpha }_{7}RO{A}_{it}+{\alpha }_{8}LOS{S}_{it}+{\alpha }_{9}FOREIG{N}_{it}+{\alpha }_{10}BUS{Y}_{it}+{\alpha }_{11}OPINIO{N}_{it}\\ & +{\alpha }_{12}LOGSE{G}_{it}+{\alpha }_{13}SHOR{T}_{it}+{\alpha }_{14}LOGAVG\mathrm{\_}ASSET+{\alpha }_{15}LOGTOTAL\mathrm{\_}FEE\\ & +{\alpha }_{16}LOGLAG+{\alpha }_{17}ISP+{\alpha }_{18}AS2+{\alpha }_{19}AS5+{\alpha }_{20}ACCER\\ & +\text{Year fixed effects}+\text{Industry fixed effects}+\epsilon \end{array}$(3)

5. Empirical Results

5.1. Descriptive Statistics

Table 2, Panel A describes the inspection outcomes for both the audit firms and the client companies in the pre-inspection sample used for testing our first hypothesis. Out of the 417 audit firms, 323 (77%) have Part II deficiencies. For the 1811 clients included in the sample, 1503 (83%) have Part II deficiencies.

Table 2. Report related statistics.

Panel A: Audit firms and client companies for H1
	Audit firms		Client companies
	N	Percentage	N	Percentage
QCD	323	77.46	1503	82.99
NON_QCD	94	22.54	308	17.01
Total	417	100.00	1811	100.00
Panel B: Audit firms with Part II findings for H2
	First-round		Second-round		Third-round
	N	Percentage	N	Percentage	N	Percentage
QCD_D	45	13.93	28	10.81	18	10.34
QCD_ND	201	62.23	123	47.49	102	58.62
NON_QCD	77	23.84	108	41.70	54	31.03
Total	323	100.00	259	100.00	174	100.00
Panel C: Client companies with Part II findings for H2
	First-round		Second-round		Third-round
	N	Percentage	N	Percentage	N	Percentage
QCD_D	110	9.73	86	8.93	67	8.72
QCD_ND	841	74.36	547	56.80	531	69.14
NON_QCD	180	15.92	330	34.27	170	22.14
Total	1131	100.00	963	100.00	768	100.00

Table 2, Panels B and C relate to the samples used to test our second set of hypotheses. Panel B provides an overview of Part II inspection outcomes for the audit firms. The total sample includes 323 first-round inspection reports, 259 second-round inspection reports and 174 third-round inspection reports. As can be seen, 45 (14%) of the first-round inspected firms, 28 (11%) of the second-round inspected firms and 18 (10%) of the third-round inspected firms have disclosed quality control deficiencies (QCD_D). The number of audit firms with remediated quality control deficiencies is 201 (62%), 123 (47%) and 102 (58%) for the first three rounds of inspections, respectively. The number of audit firms without identified quality control deficiencies is 77 (24%) in the first inspection round, 108 (42%) in the second inspection round, and 54 (31%) in the third inspection round.

Table 2, Panel C provides the report-related statistics of the client companies in the sample. For the first inspection round, there are 1131 client companies included in total. In this sample, 110 (10%) of client companies have audit firms with disclosed quality control deficiencies (QCD_D), 841 (74%) are audited by firms with remediated quality control deficiencies (QCD_ND), and 180 (16%) use audit firms without identified quality control deficiencies. Out of the 963 client companies for the second inspection round, 86 (9%) have auditors with disclosed quality control deficiencies, 547 (57%) have auditors with remediated quality control deficiencies, and 330 (34%) have auditors without identified quality control deficiencies. Turning to the third inspection round, the sample includes 768 clients, of which 67 (9%) are audited by audit firms with disclosed quality control deficiencies, 531 (69%) by audit firms with remediated quality control deficiencies, and 170 (22%) by audit firms with no quality control deficiencies identified. Overall, these descriptives show that the percentage of audit firms with no quality control deficiencies identified was much higher after the second round of inspections than in the first round, but the percentage dropped again after the third-round inspection. However, the percentage of audit firms with disclosed quality control deficiencies decrease after each inspection round, suggesting an improvement in audit quality in the small audit firm market segment.

Table 3 presents descriptive statistics for our samples. For reasons of illustration and brevity, we describe only the descriptives included in Panel A relating to the pre-inspection sample used for testing H1. Inflation-adjusted audit fees paid by the clients range from $7899 to $415,048, with a mean of $93,278. The average client in the pre-inspection period has inflation-adjusted total assets of slightly more than $26.5 million, with inventory and receivables representing 26% of that amount. Average leverage is 2.138 and the mean return on assets is −1.7.¹³ The average client has 1.5 business segments. Overall, 66% of the observations are loss-making, 37% receive going-concern opinions and 8.6% report foreign income. Furthermore, 64% of the audits are conducted during the busy season and 21% are first-year clients. On average, 37% of the inspected engagements are identified as deficient in Part I of the inspection reports. The average number of days between the financial year end and the publication of the audited financial report is 80 days and 19% of the clients are audited by an industry specialist auditor. 39% of the companies in our sample are subject to AS2, 17% are subject to AS5 and 14% are accelerated filers.¹⁴

Table 3. Descriptive statistics.

Variables	N	Mean	Median	SD	Min	Max
Panel A: Pre-inspection sample for H1
AUDITFEE	4970	94,982	61,242	93,278	7899	415,048
ASSETS	4970	26,469,531	8,209,567	44,032,791	46,629	212,600,000
LEVERAGE	4970	2.138	0.521	4.631	0.052	20.3
INVERE	4970	0.255	0.194	0.238	0	0.728
ROA	4970	−1.688	−0.164	3.858	−15.879	0.236
LOSS	4970	0.664	1.000	0.472	0.000	1.000
OPINION	4970	0.365	0.000	0.481	0.000	1.000
FOREIGN	4970	0.086	0.000	0.28	0.000	1.000
BUSY	4970	0.641	1.000	0.48	0.000	1.000
SEGMENT	4970	1.538	1.000	1.218	0.000	11.000
SHORT	4970	0.213	0.000	0.409	0.000	1.000
TOTAL_FEE	4970	2,576,437	1,167,861	3,234,481	72,588	12,756,152
AVG_ASSET	4970	66,340,256	20,536,954	120,600,000	816,670	558,500,000
DEFPERCENT	4970	0.367	0.333	0.316	0.000	1.000
LAG	4970	79.999	81.000	26.615	34.000	145.000
ISP	4970	0.192	0.000	0.394	0.000	1.000
AS2	4970	0.393	0.000	0.488	0.000	1.000
AS5	4970	0.170	0.000	0.375	0.000	1.000
ACCER	4970	0.144	0.000	0.351	0.000	1.000
Panel B: First-round inspection pooled sample for H2a
AUDITFEE	4963	110,195	75,430	98,305	7773	415,048
ASSETS	4963	32,844,502	11,624,229	48,624,281	45,022	212,600,000
LEVERAGE	4963	1.767	0.429	4.18	0.052	20.3
INVERE	4963	0.253	0.193	0.232	0.000	0.728
ROA	4963	−1.289	−0.099	3.369	−15.879	0.236
LOSS	4963	0.623	1.000	0.485	0.000	1.000
OPINION	4963	0.311	0.000	0.463	0.000	1.000
FOREIGN	4963	0.115	0.000	0.319	0.000	1.000
BUSY	4963	0.643	1.000	0.479	0.000	1.000
SEGMENT	4963	1.611	1.000	1.304	0.000	11.000
SHORT	4963	0.131	0.000	0.338	0.000	1.000
TOTAL_FEE	4963	2,886,264	1,415,553	3,431,730	71,437	12,756,152
AVG_ASSET	4963	81,440,115	28,835,524	128,100,000	792,103	543,900,000
DEFPERCENT	4963	0.345	0.333	0.317	0.000	1.000
LAG	4963	79.934	82.000	21.008	39.000	125.000
ISP	4963	0.206	0.000	0.405	0.000	1.000
AS2	4963	0.363	0.000	0.481	0.000	1.000
AS5	4963	0.430	0.000	0.495	0.000	1.000
ACCER	4963	0.175	0.000	0.380	0.000	1.000
DISAGREE	4963	0.221	0.000	0.415	0.000	1.000
Panel C: Second-round inspection pooled sample for H2b
AUDITFEE	3886	120,465	87,642	101,101	7764	401,420
ASSETS	3886	36,970,154	13,769,553	51,968,638	48,525	204,300,000
LEVERAGE	3886	1.975	0.416	4.531	0.051	19.34
INVERE	3886	0.250	0.187	0.231	0.000	0.728
ROA	3886	−1.429	−0.102	3.551	−15.107	0.235
LOSS	3886	0.623	1.000	0.485	0.000	1.000
OPINION	3886	0.301	0.000	0.459	0.000	1.000
FOREIGN	3886	0.144	0.000	0.351	0.000	1.000
BUSY	3886	0.622	1.000	0.485	0.000	1.000
SEGMENT	3886	1.600	1.000	1.272	0.000	10.000
SHORT	3886	0.069	0.000	0.253	0.000	1.000
TOTAL_FEE	3886	2,932,524	1,503,597	3,326,964	78,179	13,040,797
AVG_ASSET	3886	96,774,148	40,688,928	136,900,000	785,834	533,300,000
DEFQC1	3886	0.830	1.000	0.376	0.000	1.000
DEFPERCENT	3886	0.208	0.000	0.287	0.000	1.000
LAG	3886	85.124	86.000	15.732	56.000	120.000
ISP	3886	0.197	0.000	0.398	0.000	1.000
AS2	3886	0.108	0.000	0.310	0.000	1.000
AS5	3886	0.890	0.000	0.313	0.000	1.000
ACCER	3886	0.183	0.000	0.387	0.000	1.000
DISAGREE	3886	0.111	0.000	0.314	0.000	1.000
Panel D: Third-round inspection pooled sample for H2c
AUDITFEE	3103	122,744	95,291	96,633	8551	370,578
ASSETS	3103	36,833,887	13,902,620	51,791,864	53,636	188,500,000
LEVERAGE	3103	1.837	0.410	4.191	0.049	18.486
INVERE	3103	0.248	0.187	0.23	0.000	0.719
ROA	3103	−1.361	−0.124	3.309	−14.501	0.238
LOSS	3103	0.638	1.000	0.481	0.000	1.000
OPINION	3103	0.296	0.000	0.457	0.000	1.000
FOREIGN	3103	0.159	0.000	0.366	0.000	1.000
BUSY	3103	0.637	1.000	0.481	0.000	1.000
SEGMENT	3103	1.584	1.000	1.282	0.000	10.000
SHORT	3103	0.071	0.000	0.257	0.000	1.000
TOTAL_FEE	3103	3,602,471	1,678,091	3,766,492	88,553	12,428,597
AVG_ASSET	3103	112,100,000	54,025,056	144,800,000	954,116	515,300,000
DEFQC2	3103	0.614	1.000	0.487	0.000	1.000
DEFPERCENT	3103	0.292	0.222	0.299	0.000	1.000
LAG	3103	83.878	87.000	13.714	58.000	107.000
ISP	3103	0.206	0.000	0.404	0.000	1.000
ACCER	3103	0.198	0.000	0.399	0.000	1.000
DISAGREE	3103	0.097	0.000	0.296	0.000	1.000
Panel E: Constant sample for H3
AUDITFEE	2588	137,623	96,367	116,483	17,818	521,000
ASSETS	2588	51,597,156	19,484,618	75,400,106	209,880	341,600,000
LEVERAGE	2588	1.063	0.343	2.159	0.050	9.339
INVERE	2588	0.263	0.209	0.227	0.000	0.718
ROA	2588	−0.556	−0.011	1.465	−6.316	0.238
LOSS	2588	0.517	1.000	0.5	0.000	1.000
OPINION	2588	0.195	0.000	0.396	0.000	1.000
FOREIGN	2588	0.185	0.000	0.389	0.000	1.000
BUSY	2588	0.603	1.000	0.489	0.000	1.000
SEGMENT	2588	1.617	1.000	1.287	0.000	10.000
AVG_ASSET	2588	130,700,000	58,360,208	175,800,000	1,912,836	702,900,000
TOTAL_FEE	2588	3,719,105	1,810,955	3,980,116	81,762	13,548,659
LAG	2588	77.810	79.000	15.805	45.000	105.000
AS2	2588	0.191	0.000	0.393	0.000	1.000
AS5	2588	0.720	1.000	0.449	0.000	1.000
ACCER	2588	0.251	0.000	0.434	0.000	1.000
ISP	2588	0.207	0.000	0.405	0.000	1.000

All continuous variables have been winsorized at the 5th and 95th percentile.

Table 4 reports the correlations between variables. There is little concern for multicollinearity.¹⁵ For reasons of illustration and brevity, we only present the correlation matrix of the descriptives included in Table 3 relating to the pre-inspection sample used for testing H1. There is a significant positive correlation between audit fees (LAF) and clean Part II inspection findings (NON_QCD), which already suggests that audit firm without identified quality control deficiencies have higher audit fees. Furthermore, audit fees (LAF) are also lower for audit firms with a higher percentage of deficient engagement (DEFPERCENT).

Table 4. Pairwise correlation pre-inspection sample for H1.

Variables	1	2	3	4	5	6	7	8	9	10	11	12	13	14
LAF(1)	1.00
NON_QCD(2)	0.05*	1.00
LOGASSETS(3)	0.66*	0.05*	1.00
LEVERAGE(4)	−0.29*	−0.02	−0.62*	1.00
INVERE(5)	0.14*	0.02	0.15*	−0.10*	1.00
ROA(6)	0.30*	0.03*	0.65*	−0.78*	0.17*	1.00
LOSS(7)	−0.16*	−0.06*	−0.40*	0.23*	−0.22*	−0.33*	1.00
FOREIGN(8)	0.27*	0.02	0.29*	−0.11*	0.13*	0.13*	−0.19*	1.00
BUSY(9)	0.07*	0.04*	0.01	0.04*	−0.06*	−0.03	0.03	0.02	1.00
OPINION(10)	−0.26*	−0.06*	−0.51*	0.36*	−0.15*	−0.39*	0.39*	−0.16*	0.01	1.00
LOGSEG(11)	0.22*	0.04*	0.26*	−0.12*	0.09*	0.15*	−0.12*	0.10*	0.01	−0.13*	1.00
SHORT(12)	0.03*	0.01	0.02	−0.02	0.01	−0.01	0.05*	0.05*	0.11*	0.02	0.02	1.00
LOGAVG_ASSET(13)	0.52*	0.05*	0.51*	−0.25*	0.10*	0.27*	−0.23*	0.18*	0.00	−0.31*	0.13*	0.00	1.00
LOGTOTAL_FEE(14)	0.48*	−0.17*	0.33*	−0.14*	0.01	0.12*	−0.07*	0.16*	−0.01	−0.16*	0.04*	0.01	0.60*	1.00
DEFPERCENT(15)	−0.27*	−0.43*	−0.21*	0.11*	−0.05*	−0.13*	0.08*	−0.06*	0.00	0.16*	−0.07*	−0.02	−0.35*	−0.29*
LOGLAG(16)	−0.04*	−0.04*	−0.16*	0.19*	−0.04*	−0.19*	0.17*	−0.04*	0.06*	0.22*	−0.02	0.05*	−0.14*	−0.05*
ISP(17)	0.13*	0.04*	0.10*	−0.04*	−0.04*	0.04*	0.01	0.04*	0.01	−0.04*	−0.01	0.03*	−0.07*	−0.07*
AS2(18)	0.21*	−0.05*	0.15*	−0.06*	−0.02	0.07*	−0.04*	0.04*	0.00	−0.06*	0.02	−0.01	0.21*	0.27*
AS5(19)	−0.01	−0.04*	0.02	0.04*	−0.03*	−0.03	−0.04*	0.08*	0.04*	0.03*	−0.01	0.02	0.02	0.01
ACCER(20)	0.28*	0.02	0.34*	−0.14*	−0.02	0.14*	−0.17*	0.16*	0.04*	−0.19*	0.08*	0.00	0.21*	0.18*
DISAGREE(21)	−0.09*	−0.20*	−0.03*	0.03	−0.04*	−0.04*	0.00	−0.02	−0.02	0.03*	0.01	−0.02	−0.02	0.04*
	15	16	17	18	19	20	21
DEFPERCENT(15)	1.00
LOGLAG(16)	0.09*	1.00
ISP(17)	−0.03*	−0.04*	1.00
AS2(18)	−0.07*	0.02	−0.02	1.00
AS5(19)	0.08*	0.18*	0.00	−0.36*	1.00
ACCER(20)	−0.10*	−0.11*	0.10*	0.21*	0.01	1.00
DISAGREE(21)	0.15*	0.00	−0.01	0.01	−0.02	0.01	1.00

* Shows significance at the 0.05 level.

5.2. Univariate Results

Table 5 provides the univariate analyses for all samples based on t-tests. The results in Panel A suggest that the average fees are higher for clients of audit firms without any Part II deficiencies, providing some initial evidence to support H1. From Panel B it appears that audit fees increase significantly after the first inspection, irrespective of whether or not quality control deficiencies were identified, remediated or disclosed. However, after the second inspection, audit fees significantly increase only for clients of auditors without any quality control deficiencies. After the third inspection round, significant audit fee decreases are observed for the group of clients whose auditors have disclosed quality control deficiencies. Overall, these findings already partly confirm our expectations. Panel C shows the results for the constant sample. The results indicate that while audit fee increases are significant after the first and second inspection round, a decrease is observed after the third inspection, which is partly consistent with H3. In the next section, we use multivariate analyses to test our hypotheses by incorporating the control variables that influence audit fees.

Table 5. Univariate ananlysis.

Panel A: Pre-inspection sample for H1
Log_audit_fee	QC	NON_QCD	Diff
	11.005	11.140	0.135***
	N = 4344	N = 626
Panel B: Pre-inspection sample for H2a, H2b and H2c
	First inspection			Second inspection			Third inspection
	Pre	Post	Diff	Pre	Post	Diff	Pre	Post	Diff
QCD_D	10.451	10.565	0.114*	10.358	10.426	0.068	11.286	11.072	−0.213*
	N = 243	N = 219		N = 179	N = 139		N = 170	N = 103
QCD_ND	11.203	11.352	0.149***	11.344	11.390	0.046	11.399	11.399	0.000
	N = 2131	N = 1567		N = 1133	N = 1067		N = 1133	N = 973
NON_QCD	11.211	11.468	0.257***	11.369	11.431	0.062*	11.263	11.306	0.043
	N = 387	N = 424		N = 604	N = 775		N = 3778	N = 398
Panel C: Constant sample for H3 testing
Pre-inspection(1)	Period1(2)	Period2(3)	Period3(4)	(2)–(1)	(3)–(2)	(4)–(3)
11.303	11.528	11.621	11.518	0.225***	0.093**	−0.103**
N = 714	N = 608	N = 743	N = 523

*** p < 0.01, ** p < 0.05, * p < 0.1

5.3. Regression Results

Table 6 shows the regression results of the benchmark model to test our first hypothesis. The regression contains 4970 observations and yields an R² of 60%. NON_QCD has a significant positive coefficient (0.127, p < 0.01) indicating that audit fees were higher for companies without any quality control deficiencies in the pre-inspection period, while controlling for inspection findings at the engagement level. The results show that the coefficient on DEFPERCENT is significantly negative (−0.104, p < 0.01), implying that audit fees were on average around 10% lower for companies with a higher percentage of deficient engagements.¹⁶ Collectively, these results provide support for the first hypothesis and suggest that both Part I and Part II of the PCAOB inspection reports were priced in the market before the publication of the reports.

Table 6. Pre-inspection audit fee analysis for H1.

Log_audit_fee	Log_audit_fee
NON_QCD	0.127***
	(3.804)
LOGASSETS	0.306***
	(39.562)
LEVERAGE	0.012***
	(3.431)
INVERE	0.262***
	(5.944)
ROA	−0.027***
	(−6.248)
LOSS	0.174***
	(7.763)
FOREIGN	0.183***
	(5.095)
BUSY	0.100***
	(5.077)
OPINION	0.160***
	(6.696)
LOGSEG	0.049***
	(6.673)
SHORT	−0.025
	(−1.087)
LOGAVG_ASSET	0.051***
	(6.467)
LOGTOTAL_FEE	0.152***
	(17.021)
DEFPERCENT	−0.104***
	(−2.870)
LOGLAG	0.129***
	(4.636)
ISP	0.243***
	(10.151)
AS2	0.060
	(1.568)
AS5	0.125*
	(1.790)
ACCER	0.063**
	(2.193)
Constant	2.378***
	(9.961)
Observations	4970
R-squared	0.601
Fixed effects	Year/industry
R2 adjusted	0.595

Robust t-statistics in parentheses.

*** p < 0.01, ** p < 0.05, * p < 0.1.

Table 7 provides the regression results for hypotheses H2a, H2b and H2c. The coefficients on POST are insignificant for all three inspection rounds. The coefficients on POST represent the changes in audit fees from pre- to post inspections for clients of auditors with disclosed quality control deficiencies. Hence, we do not find support for H2c, as audit fees do not appear to change significantly for clients of audit firms with disclosed quality control deficiencies. The coefficients on NON_QCD and QCD_ND are both significantly positive for the first inspection round. This suggests that audit fees are higher before the first inspection for clients of audit firms without and with remediated quality control deficiencies compared to clients of audit firms with disclosed quality control deficiencies. These findings further support our first hypothesis that audit fees are different before the inspection, conditional on the inspection outcomes. We use an F-test to test the significance of the total effect by adding up the coefficients NON_QCD/QCD_ND with the coefficients on the interaction terms with POST. The results (Total effect NON_QCD) suggest that the total effect of NON_QCD on audit fees is significantly positive for both the first (0.086, p < 0.10) and third inspection (0.069, p < 0.10) round. In other words, for clients of audit firms with a clean Part II inspection report, audit fees increase on average around 8% and 7% after the publication of the first and third inspection reports. These findings support H2a. The results for the total effect of QCD_ND (Total effect QCD_ND) to test H2b show that audit fees increase only after the first inspection round (0.052, p < 0.05) by around 5% for clients of audit firms that remediated their quality control deficiencies, providing limited support for H2b. We also use an F-test to compare the magnitude of fee change between audit firms with clean Part II inspection reports and audit firms with remediated Part II deficiencies (NON_QCD-QCD_ND). The results suggest that the magnitude of fee increase is significantly larger for firms with clean audit firms after the third inspections, which is consistent with the notion that firms with remediated quality control deficiencies suffer from reputation damage compared to clean audit firms. We further find that the audit fees of clients of audit firms disagreeing with PCAOB inspection findings are lower after the first and second inspection round consistent with the notion that it affects audit firm reputation. We note that the coefficients on SHORT are significantly negative across all three inspection rounds, which suggests that audit fees are lower for new clients.¹⁷

Table 7. Audit fee analysis for H2.

Log_audit_fee	First inspection	Second inspection	Third inspection
POST	−0.017	0.047	0.015
	(−0.311)	(0.580)	(0.217)
NON_QCD	0.186***	0.198***	−0.152***
	(3.520)	(3.284)	(−2.997)
1.POST#1.NON_QCD	0.103	0.002	0.054
	(1.558)	(0.021)	(0.719)
QCD_ND	0.126***	0.207***	−0.048
	(3.101)	(3.790)	(−1.206)
1.POST#1. QCD_ND	0.069	−0.023	−0.003
	(1.227)	(−0.284)	(−0.044)
DEFQC1		−0.140***
		(−5.575)
DEFQC2			−0.088***
			(−4.641)
LOGASSETS	0.320***	0.339***	0.330***
	(42.491)	(40.841)	(38.457)
LEVERAGE	0.023***	0.026***	0.024***
	(6.211)	(7.048)	(5.693)
INVERE	0.245***	0.248***	0.303***
	(5.937)	(5.523)	(6.392)
ROA	−0.021***	−0.019***	−0.019***
	(−4.452)	(−3.873)	(−3.168)
LOSS	0.148***	0.139***	0.068***
	(7.770)	(6.728)	(3.121)
FOREIGN	0.220***	0.208***	0.180***
	(8.201)	(7.833)	(6.892)
BUSY	0.074***	0.040**	0.023
	(4.180)	(2.176)	(1.175)
OPINION	0.174***	0.176***	0.130***
	(7.361)	(6.817)	(4.856)
LOGSEG	0.035***	0.029***	0.005
	(5.355)	(3.763)	(0.790)
SHORT	−0.098***	−0.109***	−0.196***
	(−3.763)	(−2.832)	(−4.588)
LOGAVG_ASSET	0.045***	0.044***	0.029***
	(6.414)	(5.313)	(3.455)
LOGTOTAL_FEE	0.141***	0.132***	0.146***
	(16.818)	(13.435)	(16.230)
DEFPERCENT	0.013	−0.257***	−0.204***
	(0.384)	(−5.349)	(−5.010)
LOGLAG	0.144***	0.059	−0.046
	(4.462)	(1.106)	(−0.698)
ISP	0.201***	0.118***	0.127***
	(9.503)	(4.886)	(5.303)
AS2	−0.020	−0.117
	(−0.496)	(−0.701)
AS5	0.048	−0.085
	(0.827)	(−0.494)
ACCER	0.108***	0.098***	0.045*
	(4.670)	(3.868)	(1.831)
DISAGREE	−0.146***	−0.138***	0.039
	(−7.340)	(−4.293)	(1.150)
Constant	2.190***	2.544***	3.253***
	(8.723)	(7.103)	(5.994)
Observations	4963	3886	3103
R-squared	0.667	0.721	0.745
Fixed effects	Year/industry	Year/industry	Year/industry
R2 adjusted	0.662	0.715	0.739
F-test Total effect NON_QCD	0.086*	0.049	0.069*
F-test Total effect QCD_ND	0.052**	0.025	0.012
F-test NON_QCD-QCD_ND	0.034	0.025	0.057***

Robust t-statistics in parentheses.

*** p < 0.01, ** p < 0.05, * p < 0.1.

Table 8 presents the OLS regression results for testing H3. The model explains 76% of the variance. All the control variables have the expected signs. The coefficients on Period1 (0.117, p < 0.01), Period2 (0.240, p < 0.01) and Period3 (0.256, p < 0.01) are all significantly positive. Economically, the coefficients suggest that audit fees increase on average around 12%, 24% and 26% respectively for clients who do not switch audit firms after the first, second and third round inspections, compared to audit fees charged before the first-round inspections. However, the F-test indicates that there is no significant difference between the coefficient on Period2 and Period3 (0.016, p > 0.1). Overall, the results suggest that for companies that did not switch auditors, audit fees increased after the first and second inspection round. However, after the third inspection, audit fees did not change significantly compared to the period after the second inspection, suggesting that there is a saturation point after which there is no further incremental increase in audit fees for audit firms that have been subject to multiple inspection rounds. This would be in line with H3.

Table 8. Audit fee analysis with constant sample for H3.

Log_audit_fee	Log_audit_fee
Period1	0.117***
	(3.145)
Period2	0.240***
	(5.014)
Period3	0.256***
	(4.261)
LOGASSETS	0.354***
	(40.381)
LEVERAGE	0.029***
	(4.054)
INVERE	0.317***
	(6.710)
ROA	−0.056***
	(−4.841)
LOSS	0.105***
	(5.089)
FOREIGN	0.287***
	(11.318)
BUSY	0.016
	(0.894)
OPINION	0.105***
	(3.280)
LOGSEG	0.008
	(1.030)
LOGAVG_ASSET	0.003
	(0.370)
LOGTOTAL_FEE	0.145***
	(16.622)
LOGLAG	0.102**
	(2.214)
ISP	0.082***
	(3.375)
AS2	0.007
	(0.104)
AS5	0.076
	(0.828)
ACCER	0.042*
	(1.829)
Constant	2.411***
	(9.453)
Observations	2588
R-squared	0.771
Fixed effects	Year/industry
R2 adjusted	0.764
F-test Period2-Period1	0.123***
F-test Period3-Period2	0.016

Robust t-statistics in parentheses, *** p < 0.01, ** p < 0.05, * p < 0.1.

6. Additional Analyses

6.1. Personnel Adjustments

To further examine how audit firms respond to inspection outcomes, we investigate the personnel adjustments of audit firms. We examine changes in human resources (the number of CPAs) using the Form2 Data published on the PCAOB website. Starting in 2010, all PCAOB-registered audit firms are required to submit this form, which covers a 12-month period from April 1 to March 31. These reports contain information on the number of CPAs working for the firms. We regress the natural logarithm of the number of CPAs on the indicator variables based on the Part II inspection results and two audit firm level control variables LOGTOTAL_FEE and LOGAVG_ASSET, as defined in our main analysis: (4) $\begin{array}{rl}LOGCPAS& ={\alpha }_0+{\alpha }_{1}NON\mathrm{\_}QCD+{\alpha }_{2}POST+{\alpha }_{3}NON\mathrm{\_}QCD\ast POST\\ & +{\alpha }_{4}QCD\mathrm{\_}ND+{\alpha }_{5}QCD\mathrm{\_}ND\ast POST+{\alpha }_{6}LOGTOTAL\mathrm{\_}FEE\\ & +{\alpha }_{7}LOGAVG\mathrm{\_}ASSET+\epsilon \end{array}$(4) The results presented in Table 9 show that the number of CPAs increased after the second inspection round for audit firms that successfully remediated their quality control deficiencies. For audit firms with disclosed quality control deficiencies and audit firms without any quality control deficiencies, the number of CPAs does not change significantly. Collectively, we find some support for the theory that audit firms that remediated their quality control deficiencies increased their audit effort after the publication of the inspection reports, at least through personnel adjustments.

Table 9. Number of CPAs analysis.

Log_NCPA	First inspection	Second inspection	Third inspection
POST	0.382	0.346	0.043
	(1.057)	(1.456)	(0.160)
NON_QCD	1.488***	0.959***	1.156***
	(4.525)	(4.803)	(6.818)
1.POST#1.NON_QCD	−0.561	−0.384	−0.223
	(−1.047)	(−1.478)	(−0.797)
QCD_ND	0.589**	0.683***	0.790***
	(2.342)	(3.661)	(5.076)
1.POST#1. QCD_ND	−0.283	−0.0299	−0.035
	(−0.746)	(−0.118)	(−0.130)
LOGSUMFEE	0.339***	0.449***	0.490***
	(4.440)	(12.98)	(16.77)
LOGAVGASSET	0.212***	0.213***	0.147***
	(6.495)	(10.84)	(7.730)
Constant	−6.263***	−7.246***	−6.678***
	(−6.525)	(−14.29)	(−14.72)
Observations	220	862	917
R-squared	0.433	0.426	0.421
Fixed effects	Year	Year	Year
R2 adjusted	0.400	0.418	0.413
F-test Total effect NON_QCD	−0.179	−0.038	−0.180
F-test Total effect QCD_ND	0.099	0.316*	0.008

Robust t-statistics in parentheses.

*** p < 0.01, ** p < 0.05, * p < 0.1.

6.2. Change in the Number of Clients

In addition to audit fee changes, we also investigate whether there is a change in the number of audit firms’ clients following inspection outcomes. Besides reporting the number of CPAs they employ, audit firms also report the number of public clients in the annually published Form 2. We extract the data and use the natural logarithm of the number of public clients as the dependent variable instead of the natural logarithm of the number of CPAs, and rerun Model 4. As shown in Table 10, we find a weak decrease in the number of clients for audit firms without any quality control deficiencies. In combination with the results of our main analysis that audit fee increases are driven by audit firms without quality control deficiencies, these findings would suggest that certain clients in the small audit firm market seem to prefer lower audit fees to better audit quality.

Table 10. Number of clients analysis.

Log_NCLIENT	First inspection	Second inspection	Third inspection
POST	−0.017	−0.108	−0.018
	(−0.061)	(−0.734)	(−0.121)
NON_QCD	−0.550**	−0.582***	−0.410***
	(−2.054)	(−4.071)	(−4.016)
1.POST#1.NON_QCD	0.099	0.097	−0.132
	(0.278)	(0.594)	(−0.815)
QCD_ND	−0.319	(−1.688)	−0.343***
	(−1.354)	(3.661)	(−3.835)
1.POST#1. QCD_ND	0.272	−0.032	0.008
	(0.957)	(−0.204)	(0.052)
LOGSUMFEE	0.821***	0.765***	0.781***
	(18.40)	(38.68)	(47.34)
LOGAVGASSET	−0.231***	−0.132***	0.147***
	(−10.09)	(10.84)	(7.730)
Constant	−4.305***	−5.329***	−0.126***
	(−7.541)	(−18.62)	(−11.09)
Observations	220	862	917
R-squared	0.685	0.697	0.744
Fixed effects	Year	Year	Year
R2 adjusted	0.400	0.693	0.740
F-test Total effect NON_QCD	0.082	−0.011	−0.150*
F-test Total effect QCD_ND	0.255	−0.140	−0.010

Robust t-statistics in parentheses.

*** p < 0.01, ** p < 0.05, * p < 0.1.

6.3. Total Audit fee Change

As our results suggest that audit firms without quality control deficiencies appear to lose clients while charging higher audit fees to their remaining clients, we investigate how the total audit fees collected by audit firms from their public clients change after the publication of inspection reports, conditional on Part II inspection findings. Using data from Audit Analytics, we calculate the natural logarithm of the total audit fees collected by each audit firm per year. We use t-tests to compare the change in total fees from pre- to post-inspection periods, based on the outcome of the inspection report. The results presented in Table 11 suggest that total audit fees increase significantly both for auditors with remediated quality control deficiencies and for auditors without quality control deficiencies after the first round of inspection, while audit fees do not change significantly for auditors with disclosed quality control deficiencies. After the second inspection, total audit fees drop significantly for audit firms with disclosed quality control deficiencies, while total audit fees do not change significantly for either audit firms with remediated quality control deficiencies or audit firms without quality control deficiencies. After the third inspection, total audit fees do not change significantly for auditors without quality control deficiencies, and they drop significantly for the two other types of audit firms. In summary, the results give some indication that audit firms without quality control deficiencies are better off than the other two groups of audit firms.

Table 11. Total audit fees analysis.

Log_SUMFEE	Pre	Post	Diff	p-value
Panel A: First inspection
QCD_D	12.18	12.34	0.16	0.126
	N = 220	N = 120
QCD_ND	12.82	13.04	0.22***	0.003
	N = 1016	N = 573
NON_QCD	12.47	12.77	0.3***	0.002
	N = 401	N = 291
Panel B: Second inspection
QCD_D	12.54	12.09	−0.45**	0.013
	N = 78	N = 58
QCD_ND	13.49	13.33	−0.16	0.102
	N = 312	N = 255
NON_QCD	12.86	12.85	−0.01	0.476
	N = 340	N = 391
Panel C: Third inspection
QCD_D	12.71	12.29	−0.42*	0.098
	N = 61	N = 28
QCD_ND	13.48	13.09	−0.39***	0.003
	N = 286	N = 182
NON_QCD	12.94	12.9	−0.04	0.412
	N = 126	N = 120

*** p < 0.01, ** p < 0.05, * p < 0.1.

6.4. Financial Risk of new Clients Post-Inspection

In our main analysis, we restrict our sample to clients that do not switch audit firms after the publication of the inspection report. To further corroborate our findings on how disclosure of inspection outcomes affects audit firms’ behavior, we investigate the financial risks of new clients added to the audit firm portfolio after each round of inspection in the initial year of engagement. We identify the new clients from Audit Analytics and match them with the financial information from Compustat. Using t-tests, we compare whether ROA, LOSS and LEVERAGE for the new clients are different across audit firms with different inspection outcomes. The results are presented in Table 12.¹⁸ In general, compared to audit firms with disclosed quality control deficiencies, the results show that financial risks are lower for the new clients of audit firms without or remediated quality control deficiencies, implying that these audit firms became more selective in their client acceptance decisions, which is a key feature of a well-designed internal quality control system.

Table 12. Financial risks of new clients post inspections.

	QCD_D(1)	QCD_ND(2)	NON_QCD(3)	Diff(1)–(2)	Diff(1)–(3)	Diff(2)–(3)
Panel A: First inspection
ROA	−4.770	−2.681	−1.863	−2.089***	−2.907***	−0.818*
	N = 68	N = 646	N = 150
LEVERAGE	4.507	2.477	2.622	2.030***	1.885***	−0.145
	N = 68	N = 646	N = 150
LOSS	0.736	0.700	0.507	0.036	0.229***	0.193***
	N = 68	N = 646	N = 150
Panel B: Second inspection
ROA	−7.615	−3.843	−4.200	−3.772***	−3.415***	−0.357
	N = 48	N = 271	N = 239
LEVERAGE	8.110	2.967	3.732	5.143***	4.378***	−0.765
	N = 48	N = 271	N = 239
LOSS	0.792	0.720	0.770	0.072	0.022	−0.050*
	N = 48	N = 271	N = 239
Panel C: Third inspection
ROA	−5.369	−4.225	−3.576	−1.144	−1.793	−0.649
	N = 31	N = 199	N = 57
LEVERAGE	4.885	2.661	2.638	2.224**	2.247*	0.023
	N = 31	N = 199	N = 57
LOSS	0.839	0.809	0.789	0.030	0.050	0.020
	N = 31	N = 199	N = 57

*** p < 0.01, ** p < 0.05, * p < 0.1.

6.5. Sensitivity Analyses

To rule out the possibility that very small client companies might make different pricing decisions, we exclude clients with assets of less than one million US dollars. The untabulated results show that our main results hold when excluding these small client companies. Furthermore, instead of using robust standard errors, we cluster the standard errors by audit firms to control for the heteroscedasticity within each audit firm in all our regressions. The findings remain unchanged.

One concern in testing our first hypothesis is that certain client characteristics determine the clients’ choice of an audit firm with or without Part II PCAOB deficiencies, which may affect audit fees. To address this concern, we follow a propensity score matching (PSM) procedure to randomize the endogenous treatment. We first estimate a logistic regression where the dependent variable is NON_QCD. The explanatory variables include client companies’ total assets (LOGASSETS), leverage (LEVERAGE), inventory turnover (INVERE), return on asset (ROA), loss indicator of the previous year (LOSS), number of business segments (LOGSEG), going concern opinion indicator (OPINION), foreign income indicator (FOREIGN), busy season indicator (BUSY), short tenure indicator (SHORT), proxies for audit firm size (LOGAVG_ASSET and LOGTOTAL_FEE), year and industry dummies. Using the propensity scores from the logit model, we match each observation that has an audit firm with quality control deficiencies to an observation with a clean audit firm that has the closest match.¹⁹ Within each matched pair, the two firms exhibit similar probabilities to choose a clean or a deficient auditor. Next, we run t-tests to compare the audit fees between the treated sample and the matched sample. The untabulated results are consistent with our main results that audit fees are higher for clients of audit firms without identified quality control deficiencies before the first-round inspection.

Finally, we rerun our regression analysis by dropping the sample restriction used in the main analysis to include only the audit-client combinations that have at least one fiscal year end before and after the publication of the inspection report. The results remain unchanged using the full panel without this restriction.

7. Conclusion

We examine the impact of PCAOB inspections on audit fees of small audit firms over time. Since price competition is an important feature of the atomistic small audit firm market, a signal about the quality of small audit firms is expected to lead to price changing. We start our analysis by investigating whether audit fees were different prior to the commencement of inspections conditional on inspection outcomes relating to quality control deficiencies. We find that audit firms without quality control deficiencies were charging higher audit fees before the start of PCAOB inspections. This would suggest that PCAOB-identified quality control differences were already partly known and priced in the small audit firm segment market and representative of audit effort. Next, we examine whether inspection outcomes on the quality control system, arguably providing a strong signal about audit quality, causes changes in audit fees while controlling for the inspection outcomes at the engagement level. We find that audit fees increase after PCAOB inspections, but that this increase is mainly driven by audit firms without quality control deficiencies, suggesting a positive reputation effect. While we observe an increase in the number of CPAs employed by audit firms with remediated quality control deficiencies, we find that they only have limited ability to charge higher audit fees. Hence, these audit firms appear to experience difficulties in passing on to their clients the incurred additional costs to bring their quality control system up to standard due to reputation damage. Interestingly, we find that, instead of attracting more clients, audit firms without quality control deficiencies experience a decrease in the number of public clients. In combination with our finding that audit fees of these firms further increased, this would suggest that in the small audit firm market, certain clients are more concerned with obtaining a lower audit fee than an audit of higher quality. We further document that audit firms without disclosed quality control deficiencies appear to become more selective in their client acceptance decisions, as newly accepted clients have lower financial risk. For audit firms with disclosed quality control deficiencies, we do not find evidence of fee changes. We speculate that the increased fee gap compared to audit firms without quality control deficiencies might help these firms to retain their audit fee levels. Finally, we show that the magnitude of change in audit fees decreases after multiple inspection rounds. Collectively, our evidence suggests that PCAOB inspections have led to important changes in the small audit firm market segment.

Our study is subject to a number of data limitations. First, it would be desirable to have a fully balanced sample of audit clients across the period of investigation. Given limited data availability, a reasonable sample size is achievable only by including all clients with no less than one year of available data in each of the pre- and post-inspection periods. Second, the Form 2 data we used for analyzing the change in the number of CPAs and the number of public clients is available only from 2010 onward. Thus, the analysis based on these data is particularly an issue for the first inspection round for which data are limited. Third, we acknowledge that we might not fully capture audit firm size, which arguably affects the ability to manage inspection risk, with the control variables we have available on a yearly basis. Finally, we acknowledge limitations of generalizability of our findings which relate to the US institutional setting.

Overall, this study contributes to the relatively limited literature on the impact of PCAOB inspections of small audit firms on audit quality. In particular, we exploit variation in quality control inspection outcomes amongst small audit firms, and investigate how this affects audit pricing. While recent studies mainly relating to annually inspected audit firms report a number of positive economic effects of PCAOB inspections, our findings suggest that for the small audit firm market, the effects are not unequivocally positive. This seems to be driven by a lower demand for high quality in this audit market segment. This paper therefore also extends the literature on the potentially adverse effects of high fee pressure caused by competition and client-specific demand characteristics. These insights can be useful to practitioners, regulators and oversight bodies throughout the world. Future research would benefit from a further investigation of the impact of PCAOB inspections on other audit quality dimensions of small audit firms over time. Furthermore, future research could advance our insights on the impact of public oversight by contributing to the limited research on its effects in other parts of the world (e.g., Brocard et al., 2018; Carson et al., 2017; Sundgren & Svanström, 2017). For example, researchers could exploit the European setting where there is substantial variation in inspection regimes in terms of for example frequency of inspections, disclosure of inspections and enforcement of sanctions.

Acknowledgements

We thank Juha-Pekka Kallunki (Associate editor) and the anonymous reviewer for their helpful and constructive comments. We also thank NWO, the Netherlands Organisation for Scientific Research, for its financial support for this research project (NWO grant number 406-13-071). We gratefully acknowledge the valuable comments received from Limei Che, Matt Ege, Kris Hardies, Caren Schelleman, and workshop participants at ESSEC Business School, University of Antwerp, the 2017 EARNET symposium, the 2017 UF International Conference on Assurance and Governance and the 2017 International Symposium on Audit Research. Finally, we wish to thank Mona Offermanns and Robert Knechel for their involvement at an early stage of this project.

Additional information

Funding

We thank Juha-Pekka Kallunki (Associate editor) and the anonymous reviewer for their helpful and constructive comments. We also thank NWO, the Netherlands Organisation for Scientific Research, for its financial support for this research project (NWO grant number 406-13-071).

Notes

1 Other countries include Norway, The Netherlands for Big 4 firms from 2014, and the UK for annually inspected audit firms.

2 Examples of such actions include notifying the SEC, the US Justice Department, and disciplinary proceedings by the PCAOB such as censuring, suspending and barring auditors, or revoking the registration of audit firms, all of which can be accompanied by large financial penalties.

3 Audit firms with remediated quality control deficiencies are not included as a comparison group in Hermanson and Houston (2008). We understand that larger audit firms may be more likely to get an inspection report without quality control deficiencies as they might be better able to manage inspection risk compared to smaller audit firms. PCAOB inspection reports disclose variables reflecting audit firm size including number of partners, number of issuer clients, and number of total professionals to total clients in the inspection year (i.e., every 3rd year for the audit firms in our sample). Since we do not have panel data available, we are unable to control for these audit firm characteristics. As an alternative, we control for audit firm size using the average client total assets of the audit firm and the total revenue of the audit firms in the models to test our hypotheses.

4 We note that there might also be a reputation spillover effect from PCAOB inspection reports to non-audit services, which is beyond the scope of this paper.

5 If audit firms have quality control deficiencies identified during the inspection reports, the PCAOB states in Part II of the inspection report: ‘Any defects in, or criticisms of, the Firm's quality control system are discussed in the non-public portion of this report and will remain non-public unless the Firm fails to address them to the Board's satisfaction within 12 months of the date of this report’. If the audit firm does not have any quality control deficiency identified, the PCAOB states in Part II of the inspection report: ‘The inspection team did not identify anything that it considered to be a quality control defect that warrants discussion in a Board inspection report.’ This distinction in wording allows us to differentiate between firms with and without quality control deficiencies.

6 As Audit Analytics contains neither all inspected audit firms nor the full set of an audit firm's clients, it is not possible to match the inspected audit firms with all their clients. Furthermore, information is incomplete for certain client observations owing to missing data or missing identifiers for matching the different databases. Hence, a number of inspection reports are excluded from the analysis.

7 The number of years before the publication of the first-round inspection varies for client-audit combinations. For example, Moore Stephens Frost PLC had its first inspection report published on 21 January 2005 (this is the time that the PCAOB published its first inspection report on small audit firms). The pre-inspection period includes all clients from 2003 and 2004. Radin Glass & Co LLP had its first inspection report published on 25 January 2007, and thus the pre-inspection period includes all clients from 2003 to 2006.

8 There are two reasons why the number of inspection reports decreases from the first to the second and third inspection round. First, new audit firms entering the market at different times after 2005 did not have their second/third inspections conducted at the end of the sample period. Second, according to Lennox and Pittman (2010), small audit firms that are unable to afford the costs of remaining in the public sector exit the market.

9 For example, for the second inspection sample, we excluded all the observations that have a fiscal year end before the publication of the first inspection and after the publication of the third-round inspection.

10 All variables with dollar values are adjusted for inflation.

11 If an audit firm only has one client or all clients are from the same industry, the auditor is not recognised as an industry specialist.

12 We winsorize our data at the 5th and 95th percentile to make sure that most of the continuous variables remain in the range of three standard deviations from the means.

13 The average ROA is highly negative even after winsorizing at the 5th percentile. This is partly driven by the financial crisis. We control for year fixed effects in our empirical analyses.

14 AS2 and AS5 are not included in the descriptive statistics and the regression analysis for the third-round inspection as only one observation in our third inspection sample is subject to AS2 and all remaining observations are subject to AS5.

15 All VIF values are below 4 (excluding the dummy variables and interaction terms).

16 As we use the natural logarithm of audit fee as our dependent variable, the economic magnitude is calculated as (1-e-0.101)*100 percent.

17 We acknowledge that small audit firms deregistering from the PCAOB in response to inspections might reduce the supply of audit services and therefore may have an effect on audit fees set by other firms. We calculated, based on data on the number of US small audit firms from PCAOB annual reports and on the number of US issuers audited by US small audit firms from Audit Analytics, that the average number of issuer clients per audit firm increases from 2005 to 2009 from 7.24 to 8.43 and then decreases to 6.58 in 2015. These data do not seem to provide support for a more concentrated small audit firm market over the entire sample period. We cannot include these yearly data in our models given that we include year fixed effects. Further, if the audit market is becoming more competitive, audit fees should arguably increase for all audit firms, which is not consistent with our findings that audit firms with disclosed quality control deficiencies decrease audit fees.

18 The descriptives in Table 12 differ from the descriptives in Table 3 as the samples used are different. For example, we include new clients added to the audit firm portfolio after publication of the first round of inspection as our sample in Table 12, Panel A, while in Table 3, Panel B, we included all clients of an audit firm who did not switch auditors for at least one year after the publication of the first round inspection reports. Another reason for the difference is that the sample size for Table 12 is much smaller compared to the sample size for Table 3. Consistent with Hermanson et al. (2007), there is high variability in our data (extreme observations remain even after winsorizing at the 5th and 95th percentile). As a result, the means can vary significantly, especially when the sample size is small.

19 We match without replacement and with a caliper of 0.05.

Appendix A1. Variable definitions.

Table

POST	Indicator equal to 1 if the fiscal year ended after the publication of the inspection reports.
QCD_D	Indicator equal to 1 for auditors with disclosed QCDs.
QCD_ND	Indicator equal to 1 for auditors with identified but not disclosed QCDs.
QCD	Indicator equal to 1 for auditors with identified QCDs.
NON_QCD	Indicator equal to 1 for auditors without any QCDs identified.
AUDITFEE	Amount of audit fees paid by the client for the audit in a particular financial year, adjusted for inflation.
LAF	The natural logarithm of AUDITFEE.
ASSETS	Total assets of the client firm, adjusted for inflation.
LOGASSETS	The natural logarithm of ASSETS.
LEVERAGE	Sum of current and long-term debt divided by total assets.
INVERE	Sum of inventory and receivables of the client divided by total assets.
ROA	Net income divided by total assets.
LOSS	Indicator equal to one when net income is below zero.
FOREIGN	Indicator equal to one when the client reports foreign income.
BUSY	Indicator equal to one when the financial year-end is in December.
OPINION	Indicator equal to one when a going concern opinion was issued.
SEGMENT	Number of business segments of the company.
LOGSEG	The natural logarithm of SEGMENT.
SHORT	Indicator equal to one for first year engagement as a new client.
AVG_ASSET	Average client total assets of the audit firm, adjusted for inflation.
LOGAVG_ASSET	The natural logarithm of AVG_ASSET.
TOTAL_FEE	Total audit fee collected by the auditor in a year, adjusted for inflation.
LOGTOTAL_FEE	The natural logarithm of TOTAL_FEE.
DEFPERCENT	The percentage of deficient engagement as indicated in Part I of the PCAOB inspection reports, calculated as the total number of engagements inspected divided by the number of deficient engagements.
LAG	Number of days between a company's fiscal year end and the date of auditor report.
LOGLAG	The natural logarithm of LAG.
ISP	Indicator equal to 1 if clients are in the two-digit SIC industry that their audit firms comprise their largest portfolio shares (measured by audit fees).
AS2	Indicator equal to 1 if clients are accelerated filers with fiscal years between 15 November 2004 and 15 November 2007 or if clients are non-accelerated filers with fiscal years ending between 15 July 2005 and 15 November 2007.
AS5	Indicator equal to 1for fiscal years ending on or after 15 November 2007.
ACCER	Indicator equal to 1 for accelerated filers.
DEFQC1	Indicator equal to 1 for auditors with Part I or Part II deficiencies identified in first inspection reports.
DEFQC2	Indicator equal to 1 for auditors with Part I or Part II deficiencies identified in second inspection reports.
DISAGREE	Indicator equal to 1 for auditors who disagree with the PCAOB findings as stated in their response letters.
Period1	Indicator equal to 1 if the observation belongs to the period after the publication date of the first and before the publication of the second-round inspection reports.
Period2	Indicator equal to 1 if the observation belongs to the period after the publication date of the second and before the publication of the third-round inspection reports.
Period3	Indicator equal to 1 if the observation belongs to the period after the publication date of the third-round inspection reports and before the publication of the fourth-round inspection reports.
LOGNCPAS	Natural logarithm of the number of CPAs of the audit firm as indicated in Form 2 data.
LOGNCLIENTS	Natural logarithm of the number of public clients of the audit firm as indicated in Form 2 data.