Does increased access to non-accounting professionals enhance financial statement audit quality?

ALBERT L. NAGY, John Carroll University

Boler School of Business 1 John Carroll Blvd.

University Heights, Ohio 44118

MATTHEW G. SHERWOOD,* University of Massachusetts, Amherst

Isenberg School of Management 121 Presidents Dr,

Amherst, MA 01003

ALEKSANDRA B. ZIMMERMAN, Northern Illinois University

Barsema Hall 740 Garden Road DeKalb, IL 60115

April 2018

*The authors appreciate helpful comments on earlier drafts of this manuscript from anonymous reviewers, Kenneth Bills, Donald Dailey, Jeanette Franzel (PCAOB), and Robert Whited, along with the conference participants/reviewers of the 2017 Audit Midyear Meeting, 2017 British Finance and Accounting (BAFA) Audit Special Interest Group (SIG) Conference, and 2017 International Symposium on Auditing Research (ISAR). We are also grateful to our graduate assistants for their help in data collection.

Corresponding Author: Matthew G. Sherwood, Phone: 1-413-545-7639, Email: msherwood@isenberg.umass.edu

Data Availability: Data on office on CPA composition was hand-collected from the annual Book of Lists publications of U.S. city business journals. Data on other variables is publicly available as indicated in the paper. Data may be provided upon request.

Does increased access to non-accounting professionals enhance financial statement audit quality?

Abstract: In this study, we examine whether the personnel composition of Big 4 firm offices, in terms of non-CPAs, is associated with office-level audit quality and audit independence. The non-CPA personnel of accounting firm offices primarily consist of three groups: assurance and tax staff aspiring to be CPAs, consulting/advisory professionals, and administrative personnel. Recent regulatory and other developments in accounting and auditing suggest that in the past decade, auditors have increasingly come to rely on different types of non-audit experts/specialists when auditing increasingly complex client engagements. Such specialists include actuaries, appraisers, valuation specialists, engineers, lawyers, environmental specialists, forensic accountants, IT specialists, and data analytics professionals, among others. We hypothesize that a greater presence of non-CPAs at the office-level is associated with higher audit quality, as greater access to non-CPAs, particularly non-audit specialists, is likely to facilitate the additional use of these professionals among audit engagements, which should translate to higher audit quality. The results suggest that the support from the non-CPA personnel at the office-level enhances the quality of the audits performed by the office. Further, we do not find that a greater non-CPA presence reduces office-level audit engagement independence. This study contributes to the literature by providing evidence that office audit quality is not just a function of CPAs but also the availability of non-CPAs to support the audit engagement teams.

Keywords: non-CPAs, audit quality, human resource capital, audit firm offices, use of specialists, audit pricing

JEL: M41 - M42

1

Does increased access to non-accounting professionals enhance financial statement audit quality?

1. Introduction

This study examines whether audit outcomes (e.g., audit quality and auditor

independence) are associated with office-level human capital composition, within large public

accounting firms, and whether the association between audit outcomes and office-level human

capital composition differs for high versus low levels of client complexity. This topic is

important because large accounting firms exhaust significant resources in recruiting, hiring and

retaining qualified personnel to provide an array of services to their clients (AICPA 2010).1

However, not all office-level human capital resources focus on engagements within the

assurance practice. In many cases, a significant proportion of the office’s resources focus on

other lines of service (e.g., tax and advisory) as well as perform administration and support roles.

As a result, office-level differences in the composition of professional employees within and

between large accounting firms are likely. Using a sample of hand-collected office-level data for

Big 4 audit offices in the United States, we examine the impact of professional diversification

(i.e., human resource composition), which we define as the office’s ratio of non-CPAs to total

office employees, on audit quality and auditor independence. We find a positive and significant

association between office-level professional diversification and audit quality. Further, we find

no evidence that greater professional diversification reduces auditor independence.

To varying degrees, the Big 4 accounting firms have been diversifying their human

resource mix by growing their non-accounting focused lines of service (Harris 2014, 2015, 2016;

1 For example, PwC lists audit and assurance, tax, deals, risk assurance, cybersecurity and privacy, forensics, government services, private company services, and people and organization, among others as services provided by their firm (see http://www.pwc.com/us/en/services.html#0).

2

Doty 2015). As a result, at the global level, the Big 4’s non-audit services revenue increased by

27 percent from 2009 to 2013 (Harris 2014). Within the United States, the combined consulting

revenues of the Big 4 exceeded $19 billion in 2015 alone, with their combined market share

being 65 percent bigger than the next-largest firm type (Source Global Research 2016).2 At the

same time, revenues from audit engagements have been relatively flat (Ettredge et al. 2014). As

a result, it seems likely that this shift in office-level human resource composition has increased

the professional diversity within audit offices. Further, given the flat audit revenues during a

time of increasing non-accounting revenues, it seems reasonable to expect that the level of CPA-

aspiring junior staff and the ratio of non-CPA administrative employees to total employees has

remained relatively constant.3

In addition to enhancing accounting firms’ revenues, the growth of the Big 4’s

consulting/advisory service lines might also provide greater access to and use of in-house

expertise to support the audit function. Accounting firms, including the Big 4, commonly use

specialists without an accounting background, to support their audit engagements.4 Auditors

often call upon specialists for a variety of reasons, including, addressing complex or subjective

accounting matters (e.g. fair value measurements or intangible assets), when a client has a

complex IT system, to provide guidance on the use of emerging technologies, and when other

abnormal or unusual circumstances are present (Bauer and Estep 2014; Griffith 2016; Jenkins,

2 June 9, 2016 Press Release. See also https://blogs.wsj.com/cfo/2016/06/09/u-s-spending-on-management-consulting-climbs-to-54-7-billion-in-2015/ and http://economia.icaew.com/news/march-2016/big-four-consulting-revenues-rocket 3 If an office grows its consulting wing and hires more non-CPA professionals, one would expect the administrative staff would increase in raw numbers but not the ratio of admins to total employees. Also, given the flat audit revenues, one would expect the level of CPA-aspiring to remain constant since CPA-aspiring staff pertain mainly to assurance work. 4 For example, when discussing the role of specialists on audit engagements, Boritz, Kochetova-Kozloski, Robinson, Wong (2017, 59) state, “There is an increasing trend to use specialists with little or no accounting or auditing experience in such areas as IT, tax, valuations.”

3

Negangard, and Oler 2016; PCAOB 2017). The recent resurgence of consulting and advisory

services by the large accounting firms, coupled with the common practice of auditors engaging

specialists employed within the firm to complete audit engagements, has likely elevated the

presence and prominence of non-CPAs within the offices of public accounting firms. As a

result, we assume that the level of non-CPAs in an audit firm office is a valid proxy for the

availability of non-audit experts and specialists in an office to support the audit teams. Our

study, therefore, provides an initial exploration into the current and important issue regarding the

role of professional diversity within an audit office and explores the idea that using specialists

can influence audits, by examining the effect non-CPA specialists have on audit outcomes.

To measure the level of non-CPAs within an audit office, we hand-collect data on the

number of CPAs and the total number of employees within Big 4 offices in the United States,

from the annual Book of Lists (top/largest accounting firms) publications.5 In this study, we

introduce a novel measure for assessing audit office human resource composition: the ratio of

non-CPAs to total office employees. The ratio of non-CPAs to total office employees is an

indication of the degree to which local or in-house non-assurance specialists are available to

provide expertise to an audit engagement.6 We expect local non-CPA specialists to provide

expertise for local audit engagements and to enhance audit quality. Therefore, our hypotheses

predict a positive relationship between higher percentages of non-CPAs in the office and audit

5 We collected data from both the American Business Journals and Crain’s, which publish the majority of business journals. Keune, Mayhew and Schmidt (2016) also use data from the Book of Lists in their examination of the effect of non-Big 4 local market leadership on audit competition. 6 We recognize that non-CPAs typically fall into one of the following categories: client-facing employees aspiring to be a CPA, client-facing employees not aspiring to be a CPA, and non-client-facing administrative employees. A noted limitation of our source data is the inability to distinguish between the various non-CPA employee categories. However, under the presumption that ratios within the audit-staffing model and the administrative support model are consistent between audit offices, one can infer that non-CPA specialists represent the primary source of variance within an office’s non-CPA population.

4

quality. Of the non-CPA employee categories, we argue that the same-office specialists

supporting the audit function have the most significant impact on audit outcomes.7 Our

additional analyses provide support for this argument. We call for future researchers to determine

the effects, if any, of the use of non-certified audit staff to complete audit work and to

disentangle the effects of these non-CPA groups on audit quality.

We analyze approximately 5,000 U.S. company-year observations audited by Big 4 firms

from 2009 to 2014 and find that Big 4 offices with higher percentages of non-CPAs

(NONCPA%) are associated with higher audit quality, as measured by a lower rate of

restatements and lower performance-adjusted discretionary accruals. In addition, we do not find

a reduction in auditor independence for audit offices with a higher NONCPA%. These findings

are important because regulators have expressed concerns that increasing the non-assurance

focus of an auditor could potentially lead to reductions in auditor independence. Taken together,

our results imply that audits provided by offices with higher NONCPA% are of higher quality but

no less independent than those provided by offices with lower compositions of non-CPAs.

Furthermore, we perform additional analyses to provide some insight on our premise

regarding the effect of using specialists by considering the effects of client accounting

complexity on our results. We predict that the benefits provided by in-house specialists are most

pronounced on audits of high complexity clients, as specialists are more likely to help audit the

complex accounts and provide input regarding complex accounting systems. The results from the

7 We use the terms “local specialist,” “same-office specialist,” “in-house specialist,” and “employed specialist” interchangeably throughout this paper. The terms all represent the concept that the non-CPA specialist is housed in the same audit firm offices as the audit engagement team.

5

additional analysis indicate that the NONCPA% effect on audit quality is greater for complex

clients, which lends support to our hypothesis regarding the effect of using specialists.

In addition to providing insight into general determinants of audit outcomes, our study’s

findings contribute to auditing practice and extend the research stream on the relationship

between the use of specialists and audit quality. First, from a practice standpoint, this study

should help inform the PCAOB, accounting firms, and investors on the benefits of an audit office

having non-CPA personnel to support the audit function. While regulators in the past have been

wary of the growing non-accounting practices of the large public accounting firms, this study’s

results suggest that the availability of non-CPAs is beneficial to audit quality. Second, our study

is the first to empirically examine the effect of non-CPAs on audit quality at the office level.

Thus, we are extending the office-level audit outcomes literature stream. Third, we provide

empirical evidence on the relationship between audit office composition and auditor

independence, showing that greater access to non-CPAs is associated with higher audit quality

but not reduced auditor independence. Finally, we add to the largely behavioral research on the

use of specialists by examining the effect of availability of in-house specialists on audit quality at

the office level, thereby answering the call by Boritz et al. (2017) for research into this issue. As

extant research in this area consists of experimental and survey studies, our study is the first

large-scale archival examination of the availability of in-house expertise on audit quality.

2. Literature review and hypotheses development

While prior researchers consider the impact of advisory services on audit quality at the

firm level (Lisic, Myers, Pawlewicz and Seidel 2016) and at the engagement level (non-audit

services [NAS] such as tax provided to audit clients; e.g., Bell, Causholli, and Knechel 2015),

they have not considered the impact at the office level. Large international accounting firms in

6

the U.S. have multiple levels of operations and leadership. First, the firms are operated globally

as a network of individually held firms under the leadership of global CEOs and Boards of

Directors. Second, the firms operate on a national level with U.S. CEOs and heads of the various

service lines. Third, U.S. firms group offices geographically into regions, with several offices

comprising a region, and each region having a managing partner. The fourth level is that of the

individual office, which has a local managing partner and leadership team. Many scholars argue

that analyses in audit research should focus on the office level rather than the national level,

because individual practice offices constitute the majority of client-level audit decisions

(Wallman, 1996; Francis, Stokes and Anderson 1999; Reynolds and Francis 2000; Craswell,

Stokes and Laughton 2002; DeFond and Francis 2005). Research findings also support the

argument and provide evidence that audit quality differs within accounting firms based on audit

office characteristics such as size and location (Choi, Kim, Kim and Zang 2010; Francis, Michas,

and Yu 2013). Based on the organizational structure of the Big 4 and prior office-level audit

research, we conclude that the examination of a non-CPA composition effect on audit quality is

most appropriate at the office level.

At the AICPA 2013 Conference, former chief accountant of the Securities and Exchange

Commission (SEC) Paul Beswick stated he was concerned that audit firms’ expanding into

businesses having little relevance to the accountant's primary competencies does little to promote

audit quality and has the potential to distract a firm's leadership and other personnel from

providing appropriate attention to their audit practice. Such expansion runs the risk of damaging

the accountant’s (and audit firm’s) reputation. However, not all are convinced that performing

an array of services has a negative influence on audit quality. Representatives of the large

accounting firms argue that the in-house expertise of non-audit service professionals

7

significantly supports the audit work, particularly for the complex audits of today (Whitehouse,

2014). Both auditors and specialists expect the use of specialists to continue to rise due to the

expansion of complex accounting situations such as disclosure of fair value estimates of assets

and liabilities (PCAOB 2014, 2015, 2017).

The PCAOB recently issued a proposal to expand guidance on the use of work performed

by specialists in the audit (PCAOB 2017). The Board justified the proposal, in part, because the

use of specialists’ work continues to increase in both frequency and significance. This increase

is due to accounting estimates becoming more prevalent and more significant as business

transactions increase in complexity and financial reporting frameworks continue to evolve and

require greater use of estimates (PCAOB 2017; Pyzoha, Taylor and Wu 2016; Barac, Gammie,

Howieson, and van Staden 2016). The participants of a current qualitative study report that over

50% of audits require the use of at least one type of specialist (Boritz et al. 2017), while the

participants of another report that over 60% of audits require the use of valuation specialists

(Griffith 2016).

Furthermore, the existing research shows that the inadequate authoritative guidance

regarding specialists results in firms developing and relying on their own internal guidance on

the nature, timing, and extent of the use of specialists (Glover, Taylor and Wu 2017; Griffith

2016). As a result, despite being a key component in most of today’s audits, the availability and

use of specialists are highly subjective and variable within auditing firms. However, we do not

anticipate such variation for the other two groups of non-CPAs among the audit firms’ offices.

That is, we expect the number of unlicensed staff and office support staff to be proportionally

similar across offices due to firm- (rather than office-) level guidance on hiring assurance and tax

staff and office support. This expectation is based on our inquiries with the accounting firms

8

regarding their staffing policies.8 Consequently, the key role of in-house specialists in the audit

function, coupled with the variation of the availability of this in-house expertise among offices,

suggests that, of the three identified non-CPA groups, the consulting/advisory professional group

has the greatest impact on audit engagement quality at the office level.

The growing literature on auditors’ use of specialists mainly consists of interview-based,

survey-based, and experimental studies. Boritz et al. (2017) performed an interview-based study

of 40 practitioners that indicates both auditors and specialists are dissatisfied with the current

situation, including disagreement on how firm policies are applied and how they influence the

use of specialists. Bauer and Estep (2014, 2016) also performed an interview-based study of 33

financial and IT auditors, and posit that the involvement of IT auditors in audits is a relatively

subjective process that results in significant variation of the extent IT to which auditors are used.

Glover et al. (2017) surveyed a group of audit partners to gain further insights in areas pertaining

to fair-value measurements that have not been explored in previous literature, including the

auditors’ use of valuation specialists. Griffith (2016) interviewed a group of auditors and

valuation specialists to investigate how auditors use valuation specialists in audits of fair values.

Building on this interview/survey-based research, Cannon and Bedard (2017) provided an

engagement-level analysis challenging fair value measurements using data on the various audit

phases, including the auditors’ decision to use valuation specialists. They use data from a

sample of 115 fair value measurement audits and find that inherent risk and control risk are

associated with the decision to use a valuation specialist and that auditors typically use a

specialist when the client uses a specialist.

8 Former office managing partners note that unlicensed staff and office support would be proportionally similar across offices and regions within the same firm as such staffing levels are monitored by regional and firm level administrators. However, we do not have empirical evidence to support this anecdotal evidence.

9

Of the survey-based research that examines the role a specialist plays on audit

engagements, Griffith (2016) is of particular interest to our study. She uses Giddens’s (1990,

1991) theories of trust in expert systems to analyze the interviews of 28 auditors and 14 valuation

specialists to explain why auditors use specialists and how the auditors’ approach to specialists

might systematically affect the way auditors incorporate specialists’ work into their judgments.

She concludes that it is logical for auditors to rely on an expert system (i.e., specialists) to

provide the specific knowledge needed to complete audits.

However, Griffith also notes that, because auditing standards require auditors to take

responsibility for the work performed by specialists (PCAOB 2009, 2014, 2015, 2017), the level

of trust between the audit team and the specialist is a critical element in the willingness of the

audit team to engage and rely on a specialist. The level of trust in expert systems varies as

laypeople interact with and evaluate the system experts (Giddens 1990; Knights, Noble,

Vurdubakis and Wilmott 2001)—in this case, the specialists. Audit firms may promote rapport

between specialists and auditors by having them work in the same physical location. The goal is

to increase the sense that auditors and specialists are all members of the same team, or by

combining some elements of training for auditors and specialists to increase familiarity and

crossover knowledge (Griffith 2016). More frequent interactions between auditors and

specialists should allow for greater opportunity for trust to develop, which could promote a

greater acceptance of specialists’ work, as well as critical evaluation and integration of their

work by the auditors (Knights et al. 2001; Griffith 2016).

As mentioned above, in 2017 the PCAOB proposed changes to auditing standards related

to auditors’ use of specialists and auditing estimates. In the proposal, the PCAOB (2017,

Background) notes that “if a specialist's work is not properly overseen or evaluated by the

10

auditor, there may be a heightened risk that the auditor's work will not be sufficient to detect a

material misstatement in accounting estimates.” Based on suggestions from the PCAOB, greater

audit quality would result (or fewer audit deficiencies would be found) from an auditor’s more

appropriate use, evaluation, and integration of a specialist’s audit over complex estimates

(PCAOB 2015, 2017). Consequently, we predict that audit teams with greater access to

specialists housed in the same office (proxied by NONCPA%) will facilitate greater trust

between the audit teams and the local specialists, which should result in higher audit quality for

the engagements they perform. This leads to our first hypothesis.

H1: A positive association exists between accounting firm office-level human resource composition and audit quality.

Our second hypothesis examines the association between the non-CPA office

composition and auditor independence, which we operationalize as the likelihood to issue a

going concern modified audit opinion. The effect of the NONCPA% on auditor independence is

unclear as it presents a set of competing priorities. On the one hand, increasing the number of

specialists on an audit engagement might allow for an increase in knowledge and expertise,

which should enhance audit quality. Since specialists are not core members of the engagement

team, the specialists could be less likely to develop close relationships with the client, and thus

have fewer incentives to engage in or allow inappropriate behaviors. Further, recall that Griffith

(2016) argues that frequent interaction between auditors and specialists leads to greater trust

between the auditors and the specialist and vice versa. This suggests that increased specialist

usage should allow the audit engagement team to communicate the independence requirements

to the specialists, and allow the engagement team to better trust that the specialists will uphold

the necessary level of independence.

11

On the other hand, having a greater NONCPA% presence might cause the assurance

practice as a whole to perceive additional pressure with regards to revenue generation. If the

assurance practice believes that losing an audit client could result in the loss of office-level

influence of resources, because of the strength of the other lines of service, the assurance practice

might become more acquiescent to their client’s desires and reduce their independence. Even if

the audit engagement increases the use of the specialists to increase audit quality, if the perceived

benefits of reducing independence outweigh the perceived costs, engagement teams might be

willing to reduce their independence standards. In sum, we are unable to predict whether offices

with high compositions of non-CPAs will have less auditor independence than those with low

compositions of non-CPAs because the auditors might or might not perceive the power/influence

of the non-CPAs as a valid threat. This leads to our second hypothesis, stated as follows:

H2: There is a no relation between an accounting firm office’s non-CPA human resource composition and auditor independence.

3. Research Design

3.1 Data and sample

To determine the professional diversity within offices’ human resource capacity, we

hand-collect office personnel composition data of the top (largest) accounting firms in major

U.S. cities for the years 2009-2014. The lists are included within a special annual publication

entitled the “Book of Lists” that is published by either American City Business Journals or

Crain’s Business Journals. The information included in the list of top accounting firms is not

identical for each MSA.9 However, each list reports the accounting firm’s rank in the metro area

9 While the Book of Lists classifies the publications by “city,” Metropolitan Statistical Area (MSA) is a more accurate description. The MSA is a geographic unit comprising one or more counties, including the counties containing the core urban area, and adjacent counties with a high degree of social and economic integration (see http://www.census.gov/population/metro/). For example, neither the U.S. Census Bureau defined MSAs nor the

12

based on size, while the majority also include the number of CPAs employed and totally number

of employees by office. We require the Book of Lists to report the number of CPAs and the

number of total accounting office employees for at least three years to be included in the

sample.10 Appendix B lists, in alphabetical order, the MSAs and the average NONCPA%

across all of the Big 4 firm offices in that MSA by year of available data. 11

Table 1 reports the sample attrition for each of the unique samples we employ when

estimating our regression models. We begin by merging the hand-collected MSA-accounting

firm-year dataset with data from Audit Analytics and Compustat based on auditor, year, and

MSA code. The merger of these datasets resulted in a combined dataset containing 25,405

company-year observations between 2009 and 2014. To eliminate the potential influence of

differences in firm-wide resources that might occur between Big 4 and non-Big 4 firms, we

limited our sample to the 18,048 observations audited by Big 4 firms. Following prior research,

we omit 6,890 financial institutions and utility firms due to their unique nature and a high degree

of regulation (Reichelt and Wang 2010). This results in a base sample of 11,158 company-year

observations. As shown in Panel A of Table 1, for the misstatement analysis, we reduce the

sample by 5,538 observations missing necessary data elements, resulting in a sample of 5,620

Book of Lists distinguish between Kansas City, Kansas and Kansas City, Missouri, despite the fact that they are separate cities. We use the terms city, MSA, and metro area interchangeably throughout this study. 10 As noted in the prior footnote Keune, Mayhew and Schmidt (2016) also use data from the Book of Lists in their examination of the effect of non-Big 4 local market leadership on audit competition. Their sample size is slightly larger than our sample size . This is because we use the Book of Lists data in a slightly different manner, and thus face data constraints that they do not. Specifically, we require the Book of Lists data to report both the number of CPAs and the total number of employees in order to compute our NONCPA%, where as they require only the rank of the firm office in relation to the other local offices. Thus, given that not all Book of Lists publications include both the number of CPAs and the total number of employees our MSA sample size is smaller than that used by Keune et al. (2016). 11 While most of the MSA in our sample contain offices of each Big 4 accounting firm, the presence of each Big 4 firm within an MSA is not a requirement for inclusion in our sample. In a sensitivity test, we limit the sample to only MSAs with offices of each Big 4 accounting firm and find results consistent with those reported in Section 4.

13

observations.12 Panel B of Table 1 shows the sample selection for the discretionary accruals

analysis. A total of 5,578 observations are omitted due to missing control variables, resulting in

a sample of 5,580 observations. Panel C shows the Going Concern audit opinion sample. We

limit the sample to financially distressed firms, of which there are 1,651 which have all of the

data elements required by our estimation model.

INSERT TABLE 1

3.2 Variables and models

Using the Book of Lists data for each audit firm office, we introduce NONCPA% as a

novel measure to proxy for the composition of non-CPAs for an accounting firm’s office. The

variable represents the professional diversity within an accounting firm office. We estimate it

using equation (1), as follows:

NONCPA% = (Number of Employees –Number of CPAs) / Number of Employees. (1).

The numerator in equation 1 represents the number of non-CPAs at the accounting-firm

MSA level for which data are available via the Book of Lists. 13 The denominator of equation (1)

represents the total number of employees in the accounting firm with the corresponding MSA.

12 2014 is used as the sample cutoff for examining material misstatements which results in a restatement of previously issued financial statements to allow sufficient time for detection and correction of the material misstatement. This is in line with prior research (Francis et al. 2013). 13 In a few major cities (e.g., New York City, Los Angeles), accounting firms may have multiple physical office locations. However, the Book of Lists reports personnel figures for all of an accounting firm’s offices within a city in total (i.e., the MSA level, as described in the prior footnote). We map auditor offices (per the audit report) into MSAs, which allows us to match audit office data with Book of Lists data according to MSA. Thus, the numerator represents an accounting firm’s number of employees within an MSA, minus its number of employees within the MSA. Based on discussions with the Big 4 audit firms, due to the geographical proximity of auditor offices within an MSA there is a consistent tone across offices within an MSA. In sensitivity analyses, we limit our sample to include only the primary signing office within each MSA and find substantially similar results.

14

The larger the NONCPA%, the greater the composition of non-CPA personnel for the office,

suggesting greater professional diversity within the accounting firm office.14

Based on DeFond and Zhang (2014), who recommend using multiple proxies when

testing for effects on audit quality, we utilize multiple audit quality models to test our first

hypothesis (H1). In our main analyses, we measure material misstatements that result in

subsequent restatements and performance-adjusted abnormal discretionary accruals.

We rely on prior literature and employ the following accounting restatement logistic

regression model to test Hypothesis 1:

Prob(MISSTATE) = b0 + b1NONCPA% + b2SIZE + b3ROA + b4LEVERAGE + b5LOSS + b6BM + b7ISSUE + b8RESTRUCT + b9MERGER + b10SEG + b11LIT + b12FOREIGN + b13ICMW + b14AUDTEN + b15SI + b16CAPITAL + b17INDSPE + b18OFFICE_SIZE + Industry Dummies + Year Dummies + MSA Dummies + e (2) The dependent variable, MISSTATE, equals 1 when a firm’s annual financial report

contains a material misstatement of audited data that is restated in a subsequent period, and 0

otherwise. The material misstatements were restated within three years following fiscal year’s

end. The coefficient of NONCPA% shows the intercept shift, capturing the change in likelihood

of a material misstatement as the percentage of non-CPAs within an audit office increases. A

negative coefficient for NONCPA% indicates a decrease in the likelihood of a misstatement

occurring for audits performed by offices with a greater proportion of non-CPAs.

Additional control variables for equation (2) are drawn from the prior literature on

misstatements (Burns and Kedia 2006; Chen and Li 2013; Lobo and Zhao 2013). These

variables encompass various proxies for firm performance and audit risk, both of which affect

14 As discussed in the Introduction and Literature Review and Hypotheses Development sections, the non-CPAs within an office consist of CPA-aspiring professionals, permeant non-CPA professionals, and administrative/support professionals.

15

the likelihood of a financial statement misstatement. Client size is measured as the natural log of

the clients’ total assets (SIZE). Firm performance and profitability controls include return on

assets (ROA), the debt-to-total assets ratio (LEVERAGE), the capital level (CAPITAL), net losses

(LOSS), and the amount of special item income reported (SI). Firm complexity control variables

include restructuring charges (RESTRUCT), mergers and acquisitions (MERGER), highly

litigious industries (LIT), foreign transactions (FOREIGN), the number of segments (SEG), and

the strength of the firm’s internal control environment, measured as the existence of material

weaknesses in internal controls over financial reporting (ICMW). The book-to-market equity

ratio (BM) and the issuance of new debt or equity (ISSUE) control for financing activities.

Characteristics of the firm’s auditors such as industry specialization at the national and city level

(INDSPE), auditor tenure (AUDTEN), and the size of the audit office performing the auditing

(OFFICE_SIZE) are included as additional control variables. Definitions of the control variables

can be found in Appendix A.

As our second measure of audit quality, we examine the relation between NONCPA%

and discretionary accruals. We rely on prior literature (Reichelt and Wang 2010) to develop our

estimation model and futher test Hypothesis 1 using the following OLS regression model:

DA= b0+ b1 NONCPA% + b2 SIZE + b3 LEVERAGE + b4 LIT + b5 SCALEDOANCF + b6 SALEGR + b7 GC + b8 FOREIGN + b9SEG + b10 ICMW + b11 AUDTEN + b12OFFICE_SIZE + Industry Dummies + Year Dummies + MSA Dummies + e (3)

The dependent variable (DA) is the absolute value of performance-adjusted discretionary

accruals. A positive coefficient associated with NONCPA% will indicate that client

discretionary accruals are higher for offices where non-CPAs make up a larger percentage of the

number of employees. The higher the discretionary accruals, the greater the unconstrained

earnings management in the company, and therefore, the lower the audit quality. We follow

16

prior studies on the use of discretionary accruals to obtain the additional control variables for

equation (3) (Ashbaugh, LaFond, Mayhew 2003; Hribar and Nichols 2007; Lim and Tan 2008;

Wang and Zhou 2012). These control variables include company size (SIZE), leverage ratio

(LEVERAGE), litigiousness of the client’s industry (LIT), standard deviation of operating cash

flows (SCALEDOANCF), sales growth (SALESGR), issuance of going concern modified audit

opinion (GC), presence of foreign operations (FOREIGN), natural logarithm of the client’s total

number of segments (SEG), presence of an internal control material weakness (ICMW), auditor’s

tenure with the client (AUDTEN), and size of the audit firm office performing the audit

(OFFICE_SIZE). See Appendix A for variable definitions.

H2 tests whether accounting firm offices with greater professional diversity reduce their

independence when delivering audits. We proxy for auditor independence using the issuance of

a going concern modified opinion to financially distressed clients.15 We rely on prior research

and employ the following binary going concern audit opinion model to test Hypothesis 2:

Prob(GOING_CONCERN=1)=b0+b1NONCPA%+b2SIZE+b3ROA+b4LEVERAGE+b5BM +b6ISSUE+b7MERGER+b9 NEGEQUITY+b10SALESGR+b13ICMW+b11INVAT +b12EXTFINDMD+b8AUDTEN+b14 INDSPE+b15OFFICE_SIZE +Industry Dummies+ Year Dummies+ MSA Dummies+ e (4)

A negative coefficient for NONCPA% indicates decreasing auditor independence for audit

clients of offices having relatively high composition of non-CPAs.

The control variables are based upon several recent studies on determinants of issuing a

going concern audit opinion (Lobo and Zhao 2013; Numan and Willekens 2011; Krishnan,

Krishnan, and Song 2011), and encompass various proxies for firm stability, audit effort and

audit risk. We expect that auditors will be less likely to issue a going concern opinion when

15 We provide results in Table 6 based on two common going concern measures: 1) New Going Concern audit opinions and 2) any Going Concern audit opinion.

17

clients are larger (SIZE), have higher profitability (ROA), have less leverage (LEVERAGE), and

have higher book-to-market equity ratio (BM). Further, we expect the likelihood of issuing a

Going Concern to increase when clients engage in mergers and acquisitions (MERGER), issue

new debt or equity (ISSUE), experiencing negative sales growth (SALEGR), or negative equiy

(NEGEQUITY), have a greater need for external financing, and convert inventory less quickly

(INVAT). We control for auditors who are industry specialists at the national and city level

(INDSPE), auditor tenure (AUDTEN), and the size of the auditor’s office (OFFICE_SIZE).

Lastly, we control for the internal control environment: material weaknesses noted in the internal

controls over financial reporting audit opinion (ICMW). We expect a higher likelihood of going

concern issuance for clients audited by industry specialist auditors and clients with an internal

controls material weakness. Control variables are defined in Appendix A.

4. Results

4.1 Descriptive statistics

Table 2 reports descriptive statistics (mean, median, and standard deviation) for the

sample of firm-year observations used in the estimation of each of the models presented in

Section 3. The average NONCPA% of the samples is consistently between 0.60 and 0.63;

however, as can be seen in Appendix B, the averages by city and year vary widely. The control

variables common to multiple models are consistent across the samples, with SIZE hovering

around 20.8, while the median number of publicly traded audit clients is 10.0. Table 3 reports

the Pearson correlation coefficients between the primary test variable NONCPA% and the control

variables used in the models. The results indicate that NONCPA% is negatively correlated with

both measures of audit quality, restatements, and discretionary accruals, a finding consistent with

the higher audit quality expectation as hypothesized in H1. The results also show an

18

insignificant correlation between the NONCPA% variable and the likelihood of issuing a Going

Concern audit opinion, which is consistent with Hypothesis 2.

INSERT TABLE 2

INSERT TABLE 3

4.2Hypothesis testing

Test of Hypothesis 1: Audit Quality—Restatements

Table 4 reports the logistic estimation of equation (2) using a sample of 5,620 firm-year

observations. Coefficients, z-statistics, and significance levels are shown along with model

statistics. The model is significant with an area under the ROC curve of 71%, indicating

acceptable model fit. The negative (–1.23) and significant (p < 0.5) coefficient on NONCPA%

indicates that offices with relatively more non-CPAs are associated with lower restatement rates.

This finding suggests that offices with higher compositions of non-CPAs compared to CPAs are

associated with higher audit quality, as measured by a lower propensity for audited client

financial statements to be materially misstated and subsequently restated. Consistent with prior

literature, ROA is negatively associated with restatements, and SI and ICMW are positively

related to restatements. The NONCPA% result supports H1 and suggests that high non-CPA-

composition offices perform higher quality audits due to the support provided to the audit

function from in-house specialists, administrative staff, and/or non-CPA assurance and tax staff.

INSERT TABLE 4

Test of Hypothesis 1: Audit Quality–Discretionary Accruals

Table 5 reports the results of estimating equation (3) for the absolute value of

performance-adjusted discretionary accruals using ordinary least squares regression analysis.

The model is statistically significant and explains 19 percent of the variation in performance-

19

adjusted discretionary accruals. The coefficient on NONCPA% is negative (−0.017) and

significant (p < 0.05), suggesting offices with high non-CPA compositions are associated with

higher quality audits, as their clients report lower levels of discretionary accruals. Consistent

with prior literature, SIZE, LEVERAGE, SCALEDOANCF, SEGMENTS, and AUDTEN are

negatively related to discretionary accruals, and LIT, SALEGR, and GC are positively related to

discretionary accruals. Results for the other control variables are insignificant.

The results of the discretionary accruals model provide further evidence to support H1,

that office composition in terms of non-CPAs positively affects engagement-level audit quality.

The negative association between NONCPA% and discretionary accruals indicates that offices

with relatively high levels of non-CPAs generate higher quality audits because they are more

effective at constraining client earnings management through accruals. In sum, the results of

both audit quality models suggest that the availability of specialists, administrative personnel,

and non-CPA assurance and tax staff within an office improves audit quality. These results

support the theory that auditors develop trust with same-office specialists, which in turn creates

ontological security for the auditors. This trust appears to improve the coordination and critical

evaluation of work between specialists and auditors, which translates to higher audit quality.

INSERT TABLE 5

Test of Hypothesis 2: Auditor Independence

Table 6 shows the results of estimating equation (4) using the issuance of a going concern

audit opinion as the dependent variable. Column (1) reports the results when the dependent

variable is the issuance of a first time going concern (NEWGC), while column two reports the

likelihood of issuing any going concern modified audit opinion. In both cases, the coefficient on

NONCPA% is negative and insignificant. This result suggests that the likelihood an audit client

20

of an accounting firm office with a greater proportional presence of non-CPAs is no more or less

likely to receiving either a new or a recurring going concern audit opinion than a client of an

office with less of a proportional non-CPA presence. In other words, the degree to which the

professionals focusing on traditional accounting services comprise an office’s work force does

not influence auditors’ independence. This finding is consistent with the null prediction

presented by H2. Consistent with prior literature, larger firms and firms with greater book-to-

market ratios are less likely to receive either type of going concern modified audit opinion.

Whereas, firms with negative equity, greater need of external financing, and material weaknesses

within their internal controls over financial reporting are more likely to receive a going concern

modified audit opinion. In summary, taken together with our audit quality results, the auditor

independence test suggest that offices with higher non-CPAs to CPAs compositions deliver

higher quality audits without a reduction in independence. The additional analyses discussed

below lends further support for this conclusion.

INSERT TABLE 6

5. Additional analyses and limitations

5.1 Client Complexity

In this paper, we argue that audit offices with a greater composition of non-CPAs

produce higher quality audits, primarily due to the same office specialists available for audit

support. By their nature, audit engagements of increased complexity should represent an increase

in the likelihood of expert services needed to complete the audits. Thus, we expect to find a more

profound effect of the office non-CPA composition on audit quality for audit engagements of

more complex clients.

21

Client complexity varies both between and within audit firms based on the characteristics

of the client base at the MSA level. Further, the local client base is subject to change on an

annual basis due to clients switching auditors from one year to the next. Therefore, we compute

a measure of client complexity for each accounting firm, by MSA, on an annual basis. For

example, for our sample, we compute the median value of the Pension and Retirement Expense

on an annual basis for each Big 4 office at the MSA level. We use the annual median value of

the Pension and Retirement Expense for each Audit Firm−MSA combination to bifurcate the

client sample into clients of high or low complexity, based on clients having pension and

retirement expenses, which are above or below the annual median for the local client base. This

results in a unique division of the sample by location and audit firm, on an annual basis. This

division accounts for the influence of both macroeconomic factors that might influence all firms

as well as local microeconomic factors that might influence only a few MSAs.

We employ a variety of proxies of firm complexity likely to be associated with the use of

outside specialists such as valuation and actuarial specialists, including the annual pension and

retirement expense, goodwill, soft assets, and investments. Consistent with our assumption,

upon estimating equations (2) and (3) on the high and low complexity samples, we find, for the

sub-sample of complex audit clients, a larger NONCPA% results in a greater reduction in the

likelihood of a material misstatement of audited financial data, as well as in the discretionary

accruals level, than it does for the non-complex sub-sample. Table 7 reports the model

coefficients on NONCPA% in each of the additional analyses using the above complexity proxies

(we exclude the control variables for the sake of brevity). As shown in Table 7, the main audit

quality results reported above are driven by higher complexity clients who most likely require

the use of specialists; in all instances, the coefficient on NONCPA% is statistically significant (p

22

< 0.05) in the appropriate higher audit quality direction for the higher complexity clients, but not

significant for the lower complexity clients. Moreover, we find (untabulated data) that when the

division between high and low complexity is based on measures that would not be expected to be

correlated with the use of specialists, such as accounts receivable and fixed assets, the likelihood

of misstatement and the level of discretionary accruals are not significantly associated with the

NONCPA%, thus providing counterfactual evidence of the specialist effect. These additional

analyses further strengthen the inferences drawn from our main analyses and support our

previous argument that the results are driven by the non-CPA consulting/advisory professional

group rather than the non-certified audit and tax staff or the administrative personnel groups.

INSERT TABLE 7

5.2 Audit Efficiency

Our second additional analysis examines the association between office composition and

audit efficiency, which we operationalize as lower audit fees. Audit teams face competing

priorities when it comes to the use of specialists. On the one hand, specialists provide valuable

knowledge and expertise, which should enhance audit quality. On the other hand, the per-hour

cost of a specialist is generally higher than that of a standard audit team member of equivalent

rank within the firm. The balance between engagement quality and engagement economics

might lead some audit teams to shy away from fully engaging specialists in the audit, or at a

minimum the audit team might try to limit the number of hours allotted for specialists. We

therefore estimate the below regressions model as an exploratory analysis examining the

correlation between audit fees and the office-level ratio of non-CPAs to total employees.

FEES_AUDIT=b0 + b1NONCPA% +Ccontrols + Industry Dummies + Year Dummies + MSA Dummies + e (5)

23

A negative coefficient for NONCPA% indicates decreasing fees paid by audit clients of offices

having relatively high compositions of non-CPAs. The control variables (listed in Table 8 below)

are based upon several recent studies on determinants of audit fees (Lobo and Zhao 2013;

Numan and Willekens 2011; Krishnan, Krishnan, and Song 2011), and encompass various

proxies for audit effort and audit risk.

Table 8 shows the results of estimating equation (5) using natural log of audit fees as the

dependent variable. The coefficient on NONCPA% is negative (−0.185) and significant (p <

.05). This result suggests that offices with relatively high levels of non-CPAs are on average

associated with lower audit fees. Consistent with prior literature, audit fees are higher for bigger

clients with greater sales, higher leverage, liquidity issues, and more complexity (more segments,

special items, foreign operations, internal control issues, and clients audited by industry

specialist auditors). Taken together, our audit quality and audit fee results suggest that offices

with higher non-CPAs to CPAs compositions deliver higher quality audits at a lower price. This

increase in audit efficiency is potentially due to greater availability of in-house specialists.16

5.3 Limitations

This study is subject to several limitations. First, due to the limitations of our data, we are

not able to fully discern the underlying drivers of the positive association between the level of

non-CPAs in an audit firm office and the office’s audit quality. While we present theoretical

arguments and support for our assumptions that the variation in the availability and use of same-

office (in-house) specialists to assist audit teams is likely the reason offices with more non-CPAs

deliver higher quality audits, it is also possible that the differences in the levels of and utilization

16 To ensure that our audit fee results are not driven by a higher demand/supply and/or pricing of audit vis-à-vis non-audit services in higher non-CPA- composition offices, we examine whether NONCPA% is significantly related to ratio of audit-fees to non-audit-fees, and find it to be insignificant. Therefore, clients of higher NONCPA% offices are not supplementing their audit fee by purchasing more non-audit services from their auditors.

24

of not-yet-CPA-certified audit staff leads to higher quality audits. Former office managing

partners note that unlicensed staff and office support would be proportionally similar across

offices and regions within the same firm, as such staffing levels are determined and monitored by

regional and firm-level administrators. However, we do not have empirical evidence to support

this anecdotal evidence. The employee mix at an individual office could be correlated with the

ratio of non-CPAs to total employees, and with audit quality, for example, if an audit relies more

or less heavily on lower level staff as part of the audit engagement team. We call for future

research to further disentangle the individual effects of the non-CPA office groups on audit

quality.

Another limitation of our measure is that not all non-audit specialists are non-CPAs.

Particularly, in some areas such as forensics, specialists may be more likely to have a CPA

background. If some specialists are CPAs, then our measure underestimates the availability of

CPAs in an office, and this would bias against findings results. Nonetheless, we acknowledge

that this is a limitation of our measurements. Whether in-house specialists and consultants are

one and the same remains a question that continues to be debated in the literature and is a fruitful

avenue for future research (Bauer Estep 2014, 2016; Jenkins et al. 2016; Boritz et al. 2017).

Further, our measurements do not account for the allocation of audit work across the CPAs

within the audit office. The influence of audit work allocation across CPAs is an area which

would benefit from future research.

Finally, our analysis provides evidence of an association between the non-CPA

proportions of audit firm offices only, rather than a causal effect. Future empirical research or

archival research using natural settings of audits including/using specialists would be best suited

for causal inferences of the effect of specialists on firm audit quality (as called for in Boritz et al.

25

(2017). Our paper provides some initial broad archival empirical evidence that the availability of

specialists is associated with higher audit quality, yet more exploration remains to be done in this

area.

6. Conclusion

This study examines whether the audit office human resource composition in terms of

non-CPAs influences engagement-level audit quality. The three identified non-CPA groups of

an office are CPA-aspiring professional assurance and tax staff, consulting/advisory

professionals, and administrative support personnel. While all these non-CPA groups might

support the audit function of the office either directly and indirectly, we argue that, the same

office (in-house or employed) specialists from the consulting/advisory group have the greatest

effect on audit quality. The increased accounting complexity of business transactions, along

with the rising fair value measurement requirements in financial statements, and complexity of

IT systems, has caused auditors to call on specialists more frequently to complete their audits

(PCAOB 2015; IFIAR 2015). Furthermore, inadequate authoritative guidance on the use of

specialists in audits has resulted in firms developing and relying on their own internal guidance

on the nature, timing, and extent of use of specialists (Glover et al. 2017; Griffith 2016). The

non-standardization of use of specialists, as expected, results in significant variation of this

practice among the firms and audit firm offices. We do not anticipate such variation and impact

on audit quality from the other non-CPA groups.

We introduce a novel measure for assessing auditors’ access to non-accounting subject-

matter specialists, which we calculate using hand-collected, data. The self-reported data, allow

for the comparison of the number of non-CPAs to total employees, at the office-level. The

results suggest that the non-CPA percentage of an office is positively related to output-based

26

audit quality measures (restatements and discretionary accruals). Furthermore, we find no

evidence that a greater non-CPA presence reduces auditor independence. In an additional

analysis, we find a negative relation between non-CPA office percentage and audit fees,

suggesting that greater non-CPA percentage within an office has a positive association with audit

efficiency.

Our results have implications for the regulation of auditing. Based on our findings, a

strong presence of non-CPA personnel in an audit office is associated with higher engagement

audit quality. Additional analyses provides some support that the key driver of these results is

the in-house or employed specialist services provided by the consulting/ advisory group of the

audit firm office. Our results inform audit regulators on the role of specialists in the delivery of

audits by public accounting firms. The availability of non-CPA professionals from the

consulting and advisory lines of audit firm offices, rather than simply posing problems for audit

quality from an auditor independence standpoint, appear to be associated with higher audit

quality.

27

REFERENCES

AICPA. 2010. Positions in public accounting. Available at http://www.aicpa.org/CAREER/ CAREERPATHS/PUBLICACCOUNTING/Pages/default.aspx

Ashbaugh, H., R. LaFond, and B. W. Mayhew. 2003. Do nonaudit services compromise auditor independence? Further evidence. The Accounting Review 78 (3): 611-639.

Barac, K., E. Gammie, B. Howieson, and M. Van Staden. 2016. The capability and competency requirements of auditors in today's complex global business environment. Edinburgh: ICAS and FRC. Available at https://www.icas.com/__data/assets/pdf_file/0003/ 239457/The-capability-and-competency-requirements-of-auditors-in-todays-complex-global-business-environment.pdf

Bauer, T., and C. Estep. 2014. The IT auditor function on financial statement and integrated audits: Description of practice and avenues for future research (November 30, 2014). Available at SSRN: https://ssrn.com/abstract=2579193

Bauer, T., and C. Estep. 2016. One team or two? Investigating relationship quality between auditors and IT specialists: Implications for audit team identity and the audit process (December 14, 2016). Available at SSRN: https://ssrn.com/abstract=2579198

Bell, T. B., M. Causholli, and W. R. Knechel. 2015. Audit firm tenure, non‐audit services, and internal assessments of audit quality. Journal of Accounting Research 53 (3): 461-509.

Beswick, P. 2013. Remarks at the AICPA 2013 Conference on Current SEC and PCAOB developments in Washington, DC (December 9). Available at: https://www.sec.gov/ News/Speech/Detail/Speech/1370540488257

Boritz, J. E., N. Kochetova-Kozloski, L. A. Robinson, and C. Wong. 2017. Auditors’ and specialists’ views about the use of specialists during an audit. Working Paper, University of Waterloo. (March 31, 2017) Available at: https://ssrn.com/abstract=2534506

Burns, N., S. Kedia. 2006. The impact of performance-based compensation on misreporting. Journal of Financial Economics 79 (1): 35-67.

Cannon, N. and J. C. Bedard. 2017. Auditing challenging fair value measurements: Evidence from the field. The Accounting Review 92 (4): 81-114. doi: http://dx.doi.org/10.2308/accr-51569.

Chen, F., and Y. Li. 2013. Voluntary adoption of more stringent governance policy on audit committees. The Accounting Review 88 (6): 1939-1969.

Choi, J. H., C. Kim, J. B. Kim, and Y. Zang. 2010. Audit office size, audit quality, and audit pricing. Auditing: A Journal of Practice & Theory 29 (1): 73-97.

Craswell, A., D. Stokes, and J. Laughton. 2002. Auditor independence and fee dependence. Journal of Accounting and Economics 33 (2): 253-275.

DeFond, M., and J. Zhang. 2014. A review of archival auditing research. Journal of Accounting and Economics 58: 275-326.

DeFond, M. L., and J. Francis. 2005. Audit research after Sarbanes-Oxley. Auditing: A Journal of Practice and Theory 24 (Supplement): 5-30.

28

Doty, J. 2015. Keynote Address to the Baruch College 14th Annual Financial Reporting Conference, New York City (May 7). Available at: https://pcaobus.org/News/ Speech/Pages/05072015_Baruch.aspx

Francis, J.R., and M. Yu. 2009. Big 4 office size and audit quality. The Accounting Review 84 (5): 1521-1552.

Francis, J.R., P. N. Michas, and M.D. Yu. 2013. Office size of Big 4 auditors and client restatements. Contemporary Accounting Research 30 (4): 1626-1661.

Francis, J.R., D. Stokes, and D. Anderson. 1999. City markets as a unit of analysis in audit research and the re-examination of Big 6 market shares. Abacus 35 (2): 185-206.

Giddens, A. 1990. The Consequences of Modernity. Stanford, CA: Stanford University Press. Giddens, A. 1991. Modernity and Self-Identity. Stanford, CA: Stanford University Press. Glover, S. M., M. H. Taylor, and Y. Wu. 2017. Current practices and challenges in auditing fair

value measurements and complex estimates: Implications for auditing standards and the academy. Auditing: A Journal of Practice & Theory 36 (1): 63-84. doi: http://dx.doi.org/ L10.2308/ajpt-51514

Griffith, E. E. 2016. Auditors, specialists, and professional jurisdiction in audits of fair values. Working Paper, University of Wisconsin−Madison.

Harris, S. 2014. The rise of advisory services in audit firms. Speech to the Practicing Law Institute 12th Annual Directors’ Institute on Corporate Governance, New York City (November 24). https://pcaobus.org/News/Speech/Pages/11242014_Harris.aspx

Harris, S. 2015. Issues for the academic community to consider. Speech at the PCAOB/Annual Meeting, Washington, DC (April 16). Available at: https://pcaobus.org/News/ Speech/Pages/04162015_Harris.aspx

Harris, S. 2016. Issues for the academic community to consider. Speech at the PCAOB/Annual Meeting, Washington, DC (April 15). Available at: https://pcaobus.org/News/Speech/Pages/Steve-Harris-speech-PCAOB-AAA-conference-2016.aspx

Hribar, P. and C. Nichols. 2007. The use of unsigned earnings quality measures in tests of earnings management. Journal of Accounting Research 45 (5): 1017-1053.

International Forum of Independent Audit Regulators (IFIAR). 2016. Report on 2015 Survey of Inspection findings. Available at https://ifiar.org/IFIAR/media/Documents/ General/About%20Us/IFIAR-2015-Survey-of-Inspection-Findings.pdf

International Forum of Independent Audit Regulators (IFIAR). 2015. Report on 2014 Survey of Inspection findings. Available at: http://www.h3c.org/textes/Etude_IFIAR_2014-EN.pdf

Jenkins, J. G., E. M. Negangard, and M. Oler. 2016. Understanding firms’ usage of forensic specialists in the financial statement audit. Working Paper, Virginia Polytechnic and State University.

Keune, M.B., B.W, Mayhew, and J. J. Schmidt. 2015. Non-Big 4 local market leadership and its effect on competition. The Accounting Review 91(3): 907-931.

Knights, D., F. Noble, T. Vurdubakis, and H. Willmott. 2001. Chasing shadows: Control, virtuality and the production of trust. Organization Studies 22 (2): 311-336.

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

29

Lim, C.Y. and H.T. Tan. 2008. Non‐audit service fees and audit quality: The impact of auditor specialization. Journal of Accounting Research 46 (1): 199-246.

Lisic, L. L., L. A. Myers, R. J. Pawlewicz, and T. A. Seidel. 2016. Do accounting firm consulting revenues affect audit quality? Evidence from the pre- and post-SOX era (December 2016). Available at http://ssrn.com/abstract=2460102

Lobo, G. J., and Y. Zhao. 2013. Relation between audit effort and financial report misstatements: Evidence from quarterly and annual restatements. The Accounting Review 88 (4): 1385-1412.

Numan, W., and M. Willekens. 2011. An empirical test of spatial competition in the audit market. Journal of Accounting and Economics 54 (1-2): 450-465.

PCAOB. 2009. Auditing fair value measurements and using the work of a specialist. Standing Advisory Group Meeting. October 14-15. Washington, DC: PCAOB.

Public Company Accounting Oversight Board (PCAOB). 2014. Staff consultation paper, auditing accounting estimates and fair value measurements – preliminary overview of comments. Presented by Barbara Vanich, Associate Chief Auditor, to Staff Advisory Group Meeting on November 21. Available at http://pcaobus.org/News/Events/ Documents/11202014_SAG/11212014_Fair_Value.pdf

Public Company Accounting Oversight Board (PCAOB). 2015. The auditors’ use of the work of specialists. Staff Consultation Paper No. 2015-01, Washington, DC: PCAOB. Available at http://pcaobus.org/Standards/Documents/SCP-2015-01_The_Auditor's_ Use_of_the_Work_of_Specialists.pdf

Public Company Accounting Oversight Board (PCAOB). 2017. Proposed amendments to auditing standards for auditor's use of the work of specialists. PCAOB Release No. 2017-003, June 1, 2017. PCAOB Rulemaking Docket Matter No. 044. Washington, D.C.: PCAOB. Available at: https://pcaobus.org/Rulemaking/Docket044/2017-003-specialists-proposed-rule.pdf and https://pcaobus.org/News/Releases/Pages/specialists-proposal-6-1-17.aspx

Pyzoha, J. S., M. H. Taylor, and Y.-J. Wu. 2016. The effects of tone-at-the-top messaging and specialists on auditors’ judgments during complex audit tasks (April 2016). Available at SSRN: https://ssrn.com/abstract=2888084.

Reichelt, K. J., and D. Wang. 2010. National and office‐specific measures of auditor industry expertise and effects on audit quality. Journal of Accounting Research 48 (3): 647-686.

Reynolds, J. and J.R. Francis. 2000. Does size matter? The influence of large clients on office-level auditor reporting decisions. Journal of Accounting and Economics 30 (3): 375-400.

Wallman, S. 1996. The future of accounting, Part III: Reliability and auditor independence. Accounting Horizons 10 (4): 76-97.

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

Whitehouse, T. 2014. Regulators looking anew at audit firm consulting (December). Available at www.ComplianceWeek.com

30

APPENDIX A Variable definitions

Variable Name Variable Definition

Dependent Variables

DA = Absolute value of performance adjusted discretionary current accruals. In order to calculate, we partition the entire population of Compustat firms, excluding financial sector firms, by two-digit SIC code, and industries with fewer than 15 firms are deleted. We estimate parameters for normal accruals for each two-digit SIC industry by year using the following equation: NAt = β0 + β1 (1/TAt-1) + β2 (∆Revt) + β3 (ROAt-1) + εt, where: NAt = current accruals, reflected by net income before extraordinary items plus depreciation and amortization minus operating cash flows scaled by total assets at the beginning of year. TAt-1 = total assets at the beginning of the fiscal year t. ∆Revt = net sales in year t less net sales in year t-1 scaled by the beginning of the year total assets. ROAt-1 = income before extraordinary items scaled by total assets in year t-1. All variables are winsorized at the 1st and 99th percentiles. The parameters estimated from the above equation are used to calculate expected current accruals (ENA): ENAt = b0+ b1 (1/TAt-1)

+ b2 (∆Revt - ∆ARt) +b3 (ROAt-1), where ∆ARt = accounts receivable in year t less accounts receivable in year t-1, scaled by the beginning of year total assets. The discretionary current accruals (DA) are calculated as follows: DAt = NAt – ENAt.

FEES_AUDIT = The natural logarithm of the annual audit fee as reported in Audit Analytics (which pulls this data from the client’s Proxy Statement).

MISSTATE = Indicator variable set to 1 if the audited financial statements issued for the fiscal-year ending were subsequently restated due to a material misstatement, 0 otherwise.

Test Variable

NONCPA% = Number for non-CPAs per accounting firm at the MSA level divided by the total number of employees per firm at the MSA level.

Control Variables

ARAT = The firm's receivables (RECT) divided by its total assets (AT).

ATURN = Sales (SALE) divide by total assets (AT) at the beginning of the year (lagged total assets).

AUDTEN = Natural logarithm of number of years the client firm has engaged the opining auditor.

31

BM = The firm's book-to-market ratio defined as its book value (CEQ) divided by market value of equity (CSHO*PRCC_F).

BUSY = 1 if client fiscal year end is between December 1 and March 31, 0 otherwise.

CAPITAL = Net plant, property, and equipment (PPENT) divided by total assets (AT).

CURR = Current assets (ACT) divided by current liabilities (LCT).

EXTFINDMD = 1 if free cash flow (defined as operating cash flow (OANCF) – the average capital expenditures from the prior three years (CAPX) is less than -0.5, 0 otherwise.

FEES_NAS = The natural logarithm of the annual non-audit fees paid by the client to the auditor as reported in Audit Analytics (taken from the client’s Proxy Statement).

FOREIGN = 1 if a client has foreign transactions (FCA not equal to zero), 0 otherwise.

GC = Indicator variable set to one if the auditor issues a going concern audit opinion, 0 otherwise.

ICMW = 1 if the firm has an internal controls material weakness related to the fiscal year end, 0 otherwise.

INDSPE = 1 if the auditor is a national industry specialist and a MSA-level industry specialist, both based on 30% market share of audit fees in the industry and year, 0 otherwise.

INVAT = The firm’s inventory (INV) divided by its total assets (AT).

ISSUE = 1 if a client issues new debt (DLTIS) or equity (SSTK), 0 otherwise.

LEVERAGE = Long-term debt (DLTT) + short-term debt reported in current liabilities (DLC)/total assets (AT).

LIT = 1 if a client operates in the following industries: biotechnology (2833–2836 and 8731–8734), computers (3570–3577 and 7370–7374), electronics (3600–3674), and retail (5200–5961) industries (based on Francis et al., 1994), 0 otherwise.

LOSS = 1 if the firm’s net income before extraordinary items (IB) is negative, 0 otherwise.

MERGER = 1 if a client undertook a large merger or acquisition (AQS not equal to zero), 0 otherwise.

NEGEQUITY = 1 if the long-term liabilities (LT) exceed total assets (AT) and 0 otherwise.

32

OFFICE_SIZE = The total number of publically traded audit clients the company’s audit firm auditing the company had in year t, measured at the MSA level.

RESTRUCT = 1 if aggregate restructuring charges (RCP) in year t and t–1 is negative, 0 otherwise.

ROA = The firm's return-on-asset ratio calculated as net income before extraordinary items (IB) divided by beginning of the year total assets (lagged AT).

SALEGR = Percentage increase in sales from the prior year ((sales in year t – sales in year t – 1)/sales in year t – 1)).

SCALEDOANCF = Standard deviation of cash flows from operations over the prior ten years scaled by total assets, or operating cash flow (OANCF)/assets (AT).

SEG = Logarithm of the sum of the number of business segments reported by the Compustat Segments database.

SI = Special items (SPI) divided by total assets (AT).

SIZE = The natural log of the firm’s total assets (AT) measured in millions of dollars.

Industry dummies = Industry dummy variable based on company’s two-digit SIC code

Year dummies = Indicator variable based on the fiscal year end of the audited financial statements.

MSA dummies = Industry dummy variable based on the audit firm office’s Metropolitan Statistical Area location.

33

Appendix B Metropolitan Statistical Areas (MSAs) and years examined in this study and average NONCPA% by year

MSA  2009  2010  2011  2012  2013  2014 Albany-Schenectady-Troy  64.26%  64.26%  61.97%  66.83%  67.92%   Albuquerque  52.05%  57.14%  48.39%  42.86%  42.86%  41.67% Atlanta - Sandy Springs - Marietta  77.87%  72.26%  74.80%  71.31%  73.89%  74.97% Austin - Round Rock  67.19%  69.91%  67.25%  65.88%  72.42%  66.70% Baltimore - Towson  63.63%  65.62%  65.44%  69.01%  60.39%  58.00% Birmingham - Hoover  35.60%  44.87%  31.52%  25.67%  34.83%  38.41% Boston - Cambridge - Quincy  71.17%  67.55%  62.86%  65.20%  68.10%  69.65% Buffalo - Cheektowaga - Tonawanda  47.24%  48.83%  42.26%  43.47%  35.38%  42.00% Chicago-Naperville-Joliet  58.55%  60.18%  68.68%  71.27%     Cincinnati - Middletown  59.73%  65.29%  60.88%  60.22%  61.08%  63.48% Cleveland-Elyria-Mentor  38.95%    48.23%    52.19%  55.70% Dallas-Fort Worth-Arlington  76.34%  70.14%  74.06%  68.81%  74.83%  75.02% Dayton  42.73%  59.92%  77.62%  80.56%  81.25%  84.03% Denver-Aurora  59.13%  59.44%  57.03%  57.85%  55.88%   Detroit-Warren-Livonia  29.35%  31.11%  34.66%  35.08%  35.36%   Greensboro-High Point  49.51%  42.00%  46.03%  44.75%  36.22%  35.94% Honolulu  56.81%  53.47%  47.33%  45.34%  23.29%  45.65% Houston-Baytown-Sugar Land  68.26%  67.34%  67.54%  68.12%  74.48%  69.99% Jacksonville  10.16%  38.56%  38.50%  40.82%  41.25%  32.82% Kansas City  48.03%  48.66%  53.34%  55.00%  50.52%  52.93% Louisville  46.36%  46.32%  47.63%  44.35%  48.93%  45.63% Milwaukee-Waukesha-West Allis        48.43%  47.94%  44.99% New York-Northern New Jersey-Long Island  31.65%  40.57%  40.57%  47.52%     Philadelphia-Camden-Wilmington  70.51%  69.54%  67.50%  70.18%  70.55%  71.78% Phoenix-Mesa-Scottsdale      54.13%  60.57%  65.77%  64.74% Pittsburgh  76.62%  70.85%  70.34%  71.61%  74.55%  74.79% San Antonio-New Braunfels  56.91%  45.31%  53.80%  54.53%  54.69%  57.02% San Jose-Sunnyvale-Santa Clara  75.19%    72.69%  73.05%  61.61%   St. Louis  65.95%  59.07%  57.35%  54.04%  53.73%  52.18% Tampa-St. Petersburg-Clearwater  64.08%    48.91%  52.91%  59.60%  66.83% Washington-Arlington-Alexandria  77.37%  74.39%  70.37%  71.71%  72.64%  84.78% 

Total Average  54.86%  56.51%  55.83%  56.49%  57.07%  58.94% 

34

TABLE 1 Sample attrition by analysis

Panel A: Sample for misstatement analysis

Merge of Top Accounting Firm by City/Year Data with Audit Analytics and Compustat Databases for 2009-2014, Big 4 Audited Observations Only

18,048

Less: Financial Institutions and Utilities 6,890 Sub-Total 11,158 Less: Observations Missing Control Variables for Misstatement Analysis 5,538 Total Company-Year Observations in Misstatement Sample (Table 4)

5,620

Panel B: Sample for discretionary accruals analysis Observations between fiscal years 2009 and 2014 (inclusive) 11,158 Less: Observations Missing Control Variables Required in Discretionary Accruals Analysis

5,578

Discretionary Accruals Analysis Sample (Sample for Table 5)

5,580

Panel C: Sample for Going Concern Analysis Distressed Firm Observations between fiscal years 2009 and 2014 (inclusive) 4,743 Less: Observations Missing Control Variables Required in Going Concern Analysis 3,092 Discretionary Accruals Analysis Sample (Sample of Table 6)

1,651

Note: Table 1 reports the sample determination process for each of the analyses conducted in the paper. The samples are limited to companies engaging a Big 4 audit firm and are located within one of the cities listed in Appendix B.

35

TABLE 2 Descriptive Statistics

Sample: Misstatement

(N=5,620) Disc. Accr. (N=5,580)

Going Concern (N=1,651)

Variable Mean Med. St. Dev

Mean Med. St. Dev

Mean Med. St. Dev

AUDTEN 2.48 2.48 0.90 2.49 2.48 0.89 2.26 2.30 0.87 BM 0.44 0.42 0.93 0.40 0.45 1.52 CAPITAL 0.26 0.17 0.24 DA 0.05 0.03 0.06 EXTFINDMD 0.18 0.00 0.38 FOREIGN 0.38 0.00 0.48 0.38 0.00 0.49 GC 0.02 0.00 0.14 ICMW 0.02 0.00 0.16 0.02 0.00 0.15 0.04 0.00 0.20 INDSPE 0.19 0.00 0.39 0.18 0.00 0.39 INVAT 0.09 0.03 0.14

ISSUE 0.91 1.00 0.28

LEVERAGE 0.25 0.20 0.27 0.25 0.21 0.26 0.30 0.24 0.34 LIT 0.34 0.00 0.47 0.34 0.00 0.47

LOSS 0.31 0.00 0.46

MERGER 0.10 0.00 0.31 0.08 0.00 0.27 MISSTATE 0.06 0.00 0.23

NEGEQUITY 0.11 0.00 0.32

NEWGC 0.04 0.00 0.19

NONCPA% 0.61 0.64 0.15 0.62 0.64 0.15 0.63 0.67 0.14

OFFICE_SIZE 18.13 10.00 23.87 17.22 10.00 22.22 19.11 10.00 23.26

RESTRUCT 0.47 0.00 0.50

ROA 0.02 0.08 0.36 0.04 0.08 0.22 -0.13 -0.02 0.38 SALEGR 0.10 0.05 0.51 0.13 0.00 0.83

SCALEDOAN 0.06 0.09 0.22 SEG 0.93 1.10 0.55 0.96 1.10 0.54 SI 0.02 0.01 0.06 SIZE 20.84 20.87 1.85 20.90 20.93 1.79 20.03 19.93 1.80 Note: Table 2 reports the descriptive statistics of all of the variables used in estimating the models used to test the study’s hypotheses. See Appendix A for variable defniitions.

36

12 13 14 15 16 17 18 19 20 21 22 23 12 1.000 13 -0.137 1.000 14 0.090 0.178 1.000 15 -0.003 0.051 -0.008 1.000 16 0.012 0.015 0.028 0.016 1.000 17 -0.025 -0.025 -0.045 -0.040 -0.033 1.000 18 0.016 0.103 0.127 0.060 0.027 -0.072 1.000 19 -0.006 -0.153 -0.417 0.052 -0.021 -0.044 0.025 1.000 20 -0.029 0.066 -0.002 0.066 -0.004 0.032 -0.067 -0.031 1.000 21 -0.218 -0.156 -0.344 0.045 -0.001 -0.063 0.028 0.626 -0.065 1.000 22 -0.053 0.157 0.035 0.112 0.044 -0.189 0.297 0.151 -0.031 0.184 1.000 23 0.071 0.043 0.211 0.038 0.021 0.008 0.133 -0.146 -0.015 -0.110 0.016 1.000 24 0.115 -0.284 -0.336 0.005 0.002 -0.027 0.095 0.290 -0.072 0.331 0.075 -0.136 Note: Table 3 represents the correlation among all the variables used in this study. Bolded items are significant at the 0.05 level. See Appendix A for variable definitions

TABLE 3 Pearson Correlation Coefficients for Variables 1 2 3 4 5 6 7 8 9 10 11

1 NONCPA% 1.000 2 AUDTEN -0.054 1.000 3 BM -0.032 0.020 1.000 4 CAPITAL 0.070 0.028 -0.001 1.000 5 DA -0.011 -0.151 -0.085 -0.140 1.000 6 FOREIGN 0.006 0.044 -0.044 -0.111 -0.006 1.000 7 GC 0.024 -0.045 -0.261 -0.012 0.170 -0.010 1.000 8 ICMW 0.008 -0.036 -0.050 -0.007 0.032 0.037 0.025 1.000 9 INDSPE -0.163 0.064 -0.008 0.095 -0.036 -0.040 -0.014 -0.002 1.000

10 INVAT -0.008 0.097 0.048 -0.072 -0.012 0.090 -0.025 -0.005 0.031 1.000 11 ISSUE 0.046 0.082 -0.001 0.006 -0.023 0.202 0.013 0.033 -0.011 0.016 1.000 12 LEVERAGE -0.051 -0.075 -0.300 0.200 -0.023 -0.072 0.139 0.006 0.025 -0.075 0.030 13 LIT 0.043 -0.008 -0.062 -0.177 0.196 0.110 0.071 0.021 -0.060 0.087 0.191 14 LOSS 0.068 -0.154 -0.056 -0.066 0.219 0.032 0.229 0.068 -0.043 -0.035 0.141 15 MERGER 0.016 -0.010 -0.008 -0.051 0.002 0.065 -0.034 0.006 -0.031 -0.009 0.126 16 MISSTATE -0.026 0.006 0.033 0.004 0.005 -0.003 -0.004 0.058 -0.005 0.002 0.070 17 OFFICE_SIZE -0.276 -0.038 0.005 -0.185 0.078 -0.017 -0.028 -0.016 0.010 -0.052 -0.278 18 RESTRUCT -0.031 0.129 -0.075 -0.099 -0.023 0.281 0.022 0.025 -0.009 0.104 0.225 19 ROA -0.035 0.097 0.021 0.093 -0.310 0.059 -0.298 -0.024 0.016 0.089 -0.003 20 SALEGR 0.012 -0.114 -0.033 -0.049 0.177 -0.013 -0.016 0.015 -0.017 -0.042 0.041 21 SCALEDOANCF -0.025 0.107 0.053 0.154 -0.283 0.057 -0.418 -0.007 0.017 0.072 -0.005 22 SEG 0.045 0.122 -0.035 0.062 -0.059 0.314 -0.052 0.027 -0.010 0.142 0.382 23 SI 0.026 -0.065 -0.072 -0.044 0.323 0.028 0.129 0.038 -0.032 -0.011 -0.025 24 SIZE -0.119 0.250 0.063 0.187 -0.343 0.015 -0.215 -0.021 0.138 -0.068 0.117

37

TABLE 4 Likelihood of material misstatement resulting in restatement

Regression Type: Logit

Dependent Variable: MISSTATE

Variable Coef. z-stat

INTERCEPT -3.697 *** -2.94 NONCPA% -1.229 ** -1.66 SIZE 0.010 0.20 ROA -0.380 ** -2.47 LEVERAGE 0.251 1.13 LOSS -0.182 -1.09 BM 0.184 ** 1.98 RESTRUCT -0.086 -0.52 MERGER -0.062 -0.31 SEG 0.168 0.97 LIT -0.242 -0.96 FOREIGN -0.420 ** -2.29 ICMW 0.800 *** 2.63 AUDTEN -0.010 -0.11 SI 1.791 ** 2.30 CAPITAL -0.484 -0.98 INDSPE 0.104 0.55 GC -0.418 -0.73 OFFICE_SIZE 0.002 0.36 Industry, Year and MSA FE Yes

N of Obs. 5,620 Chi2 136.12 Prob > Chi2 0.000 Pseudo R-Squared 0.079 Area Under ROC 0.705 Note: Table 4 reports the results of estimating the likelihood of material misstatement between fiscal years 2009 and 2012. *, **, *** indicate statistical significance at the 0.10, 0.05 and 0.01 levels, respectively. Variables with directional prediction represent a one-tailed tests, while all other variables are in two-tailed tests. Variable definitions are provided in Appendix A.

38

Table 5 Office Non-CPA percentage and the absolute value of performance-adjusted discretionary accruals

Regression Type: OLS

Dependent Variable: Absolute value of performance-adjusted discretionary accruals

Variable Coef. t-stat INTERCEPT 0.192 *** 12.81 NONCPA% -0.017 ** -1.97 SIZE -0.006 *** -9.21 LEVERAGE -0.012 * -1.95 LIT 0.012 *** 2.67 SCALEDOANCF -0.039 *** -3.45 SALEGR 0.017 *** 4.26 GC 0.043 *** 3.63 FOREIGN -0.001 -0.32 SEG -0.003 * -1.69 ICMW 0.008 1.4 AUDTEN -0.003 ** -2.61 OFFICE_SIZE 0.001 0.74 Industry, Year and MSA FE Yes N of Obs. 5,580 F-Value 5.85 Prob >F2 0.000 Adjusted R2 0.19 Note: Table 5 reports the results of estimating the firm-level performance-adjusted discretionary accrual using firm year observations between 2009 and 2014. *, **, *** indicate statistical significance at the 0.10, 0.05 and 0.01 levels, respectively, in two-tailed tests. Variable definitions are provided in Appendix A.

39

Table 6 Likelihood of Issuing a Going Concern Audit Opinion Regression Type Logit Logit

Dependent Variable:

First Time GC

(Column One)

Any GC

(Column Two)

Variable Coef. z-stat Coef. z-stat

INTERCEPT 1.465 0.43 6.84 2.055** NONCPA% -2.592 -1.24 -1.26 -0.553 SIZE -0.479 -2.95*** -0.78 -4.69*** ROA 0.359 0.58 -0.19 -0.523 LEVERAGE 0.258 0.63 -0.05 -0.116 BM -0.341 -2.56** -0.38 -ISSUE -0.565 -0.9 -0.60 -1.154 MERGER -0.448 -0.42 -0.89 -0.907 AUDTEN -0.174 -0.84 0.07 0.287 NEGEQUITY 0.765 1.05 1.82 3.157**SALEGR 0.045 0.3 -0.16 -1.08 INTCOV 0.829 2.77*** 1.37 4.431**EXTFINDMD 1.494 2.72*** 1.60 4.36*** ICMW 1.119 1.82* 1.79 2.367** INDSPE -0.086 -0.16 -0.26 -0.527 OFFICE_SIZE -0.001 -0.12 -0.01 -0.543 Industry, Year and MSA Yes Yes N of Obs. 1,651 1,651 Chi2 325.88 277.44 Prob > Chi2 0.000 0.0006 Psuedso R-Squared .408 0.532 Area Under ROC .939 0.953 Note: Table 6 reports the results of predicting the likelihood of an auditor modifying the audit opinion for going

concern issues for companies that are in financial distress between the fiscal years of 2009 and 2014. * **, *** indicates statistical significance at the 0.10, 0.05 and 0.01 levels respectively, in two-tailed tests. Variable definitions are provided in in Appendix A.

40

Table 7 Additional analyses for office Non-CPA percentage for effects on audit quality of high versus low accounting complexity clients

Misstatement Likelihood

Model Discretionary Accruals

Model Accounting Complexity

Measure Test

Variable

Client Accounting Complexity Coef. z P>|z| Coef. t P>|t|

Pension and Retirement Exp. NONCPA% High -3.638 -2.940 0.003 -0.030 -2.517 0.012 Pension and Retirement Exp. NONCPA% Low -0.993 -0.820 0.413 -0.026 -1.290 0.198

Soft Assets NONCPA% High -3.343 -3.140 0.002 -0.035 -2.283 0.023 Soft Assets NONCPA% Low 0.447 0.360 0.721 -0.018 -0.898 0.369

Goodwill NONCPA% High -2.574 -2.480 0.013 -0.031 -2.483 0.013 Goodwill NONCPA% Low -1.203 -0.960 0.336 -0.016 -0.681 0.496

Short-Term Investments NONCPA% High -2.129 -2.020 0.043 -0.046 -2.090 0.037 Short-Term Investments NONCPA% Low -3.078 -1.430 0.152 -0.016 -0.990 0.323 Note: Table 7 reports the results of additional analyses for the effect of office non-CPA percentage on the engagement audit quality of high versus low accounting complexity engagements. The results indicate the the effect of non-CPA presence in the office is pronounced for high accounting complexity clients but not low complexity clients. The first set of results shows the estimation of the restatement model equation for high versus low complexity clients and the second set of column results shows the estimation of the discretionary accruals model equal using the same breakdown.

41

Table 8 Office Non-CPA percentage and client audit pricing Regression Type: OLS

Dependent Variable: Audit Fees

Variable Coef. t-stat

INTERCEPT 4.020 *** 18.85 NONCPA% -0.185 ** -2.01 SIZE 0.481 *** 55.55 ATURN 0.097 *** 3.48 ARAT 0.872 *** 4.79 INVAT 0.140 1.06 CURR -0.016 *** -4.11 LEVERAGE 0.101 *** 2.61 ROA -0.188 *** -3.76 LOSS 0.124 *** 6.07 GC 0.166 *** 2.93 BM -0.015 -1.48 AUDTEN 0.023 * 1.82 BUSY 0.018 0.73 SEG 0.087 *** 4.47 SI 0.289 ** 2.3 FOREIGN 0.182 *** 8.1 MERGER 0.079 *** 4.09 ISSUE 0.018 0.48 ICMW 0.321 *** 5.22 INDSPE 0.045 * 1.69 OFFICE_SIZE -0.001 -0.92 Industry, Year and MSA FE Yes N of Obs. 5,620 F-Value 80.261 Prob >F2 0.000 Adjusted R2 0.81

Note: Table 8 reports the results of estimating firm-level audit fees between the fiscal years of 2009 and 2014.

*, **, *** indicate statistical significance at the 0.10, 0.05 and 0.01 levels, respectively, in two-tailed tests. Variable definitions are provided in in Appendix A.