essays on auditor competencies

Essays on Auditor Competencies

by

Nattavut Suwanyangyuan

M.S., University of Rochester, 2011

B.Acc. (Hons.), Chulalongkorn University, 2008

Thesis Submitted in Partial Fulfillment of the

Requirements for the Degree of

Doctor of Philosophy

in the

Segal Graduate School

Beedie School of Business

© Nattavut Suwanyangyuan 2018SIMON FRASER UNIVERSITY

Summer 2018

Copyright in this work rests with the author. Please ensure that any reproduction or re-

use is done in accordance with the relevant national copyright legislation.

ii

Approval

Name: Nattavut Suwanyangyuan

Degree: Doctor of Philosophy

Title: Essays on Auditor Competencies

Examining Committee: Chair: Andrew Gemino Professor, Beedie School of Business

Karel Hrazdil Senior Supervisor Associate Professor

Dennis Chung Supervisor Professor

Kim Trottier Supervisor Associate Professor

Robert Grauer Internal Examiner Professor Emeritus

Dan A. Simunic External Examiner Professor, Sauder School of Business University of British Columbia

Date Defended/Approved: June 20th, 2018

iii

Abstract

This dissertation consists of three essays that present new evidence on auditor

competencies to deliver high audit quality using various auditor attributes, including auditor

size, audit firm competencies and industry specialization.

In the first essay, I provide new evidence on the influential role of external auditors in

enhancing the informativeness of 10-K reports. Specifically, I find that the client’s choice

of Big 4 auditors contributes to cross-sectional variations in 10-K disclosure volume. I also

find that the benefit of enhanced disclosures provided by Big 4 auditors (PwC, EY, KPMG

and Deloitte) is more pronounced for audit clients with poorer accrual quality and those

with higher information asymmetry. Additionally, I introduce the portion of 10-K length

unexplained by operating complexity and observable clients’ characteristics as an

alternative proxy for audit efforts. I employ this measure because abnormally long

disclosures induce external auditors to reduce the risk of material misstatement through

additional audit effort, as evidenced by higher audit fees and the increased likelihood of

going-concern opinions.

In the second essay, I provide new evidence on audit pricing differences within the Big 4

audit firms in the U.S. market. I estimate an audit fee model and consistently show that

the positive coefficient for PwC is significantly larger than those of the other Big 4 audit

firms. This result indicates that PwC earns above-average audit fee premiums relative to

the other Big 4 audit firms. Because the industry expertise research stream argues that

an audit firm with greater competency will be able to differentiate itself from its competitors

in terms of within-industry market share and charge an audit fee premium for its services,

I reveal that PwC has maintained its leadership position as the market share leader across

most industries in the U.S. market. More importantly, I find that the evidence of an industry

specialization premium is consistently observed for the group of PwC specialists, but not

for the group of other (non-PwC) specialists.

In the third essay, I provide further evidence on audit quality differences at the inter-audit

firm level. Unlike other studies that implicitly assume a homogeneous level of audit quality

within the Big 4 firms, I reveal that the existing differences among the Big 4 firms lead to

cross-sectional variations in audit quality as measured by earnings quality and going-

concern audit opinions. Specifically, I find that the negative relationship between PwC and

accrual quality appears to be larger for PwC clients than for those of EY clients. I also find

that EY clients are less likely to receive a going-concern audit opinion in both the full

sample and a subsample of severely financially distressed firms. Consistent with the

evidence of an industry specialization premium in the second essay, I find that the

association between industry expertise and higher audit quality is consistently observed

for the group of PwC specialists, but not for the group of other specialists. Considered

together, these results reinforce the importance of individual audit firm competencies,

particularly the PwC effect, and suggest that not all Big 4 audit firms are the same.

iv

Keywords: Auditor competencies; big 4 auditors; 10-K disclosure volume; audit fees;

auditor industry specialization; audit quality.

v

Table of Contents

Approval ............................................................................................................................ ii

Abstract ............................................................................................................................. iii

Table of Contents .............................................................................................................. v

List of Tables .................................................................................................................... vii

List of Figures.................................................................................................................. viii

List of Acronyms ............................................................................................................... ix

Preface .............................................................................................................................. x

Chapter 1. Introduction ................................................................................................ 1

1.1 Audit quality overview............................................................................................. 3

1.2 Audit fees and its determinants .............................................................................. 7

1.3 Figures ................................................................................................................... 9

Chapter 2. Auditor choice and 10-K disclosure volume ......................................... 11

2.1 Introduction .......................................................................................................... 11

2.2 Literature Review and Hypotheses Development ................................................ 13

2.2.1 Literature review ............................................................................................ 13

2.2.2 Hypotheses development .............................................................................. 16

2.3 Research Design .................................................................................................. 19

2.4 Sample Selection and Descriptive Statistics ........................................................ 23

2.4.1 Sample selection ........................................................................................... 23

2.4.2 Descriptive statistics ...................................................................................... 23

2.5 Empirical Results.................................................................................................. 24

2.6 Additional Analyses .............................................................................................. 27

2.7 Conclusion ........................................................................................................... 27

2.8 Figure and tables.................................................................................................. 29

Chapter 3. The difference among the Big 4 firms: Further evidence from audit pricing .................................................................................................................. 43

3.1 Introduction .......................................................................................................... 43




3.3 Research Design .................................................................................................. 51






3.6.1 Analysis using unexplained audit fees .......................................................... 56

vi

3.6.2 Analysis using industry specialization at the MSA city level ......................... 59

3.7 Conclusion ........................................................................................................... 59

3.8 Figures and tables ................................................................................................ 61

Chapter 4. The difference among the Big 4 firms: Further evidence from audit quality ................................................................................................................... 77

4.1 Introduction .......................................................................................................... 77




4.3 Research Design .................................................................................................. 83






4.6.1 Analysis using signed discretionary accruals ................................................ 89

4.6.2 Analysis using industry specialization at the MSA city level ......................... 90

4.7 Conclusion ........................................................................................................... 90

4.8 Figure and tables.................................................................................................. 92

Chapter 5. Conclusion ............................................................................................. 103

References ................................................................................................................... 104

Appendix A. Supplemental Analysis for Chapter 2 ................................................ 111

Appendix B. Supplemental Analysis for Chapter 3 ................................................ 117

Appendix C. Supplemental Analysis for Chapter 4 ................................................ 122

vii

List of Tables

Table 2.1 Descriptive Statistics ............................................................................... 30

Table 2.2 Correlation Matrices ................................................................................ 34

Table 2.3 Auditor choice and 10-K disclosure volume ............................................ 35

Table 2.4 Incremental effect of Big 4 auditors on 10-K disclosure volume .............. 36


Table 2.6 Residual disclosures and audit fees (Level specification) ....................... 39

Table 2.7 Residual disclosures and audit fees (Change specification) ................... 40

Table 2.8 Residual disclosures and going-concern opinions .................................. 41

Table 2.9 Auditor choice and 10-K disclosure volume (by each fiscal year) ........... 42


Table 3.2 Rankings of each major audit firm ........................................................... 65

Table 3.3 Audit fee model ........................................................................................ 67

Table 3.4 Audit fee model (Size partition) ............................................................... 68

Table 3.5 Assignments of industry specialists across major audit firms .................. 69

Table 3.6 Audit fee model: Estimation of industry specialist premium .................... 71

Table 3.7 Analysis using unexplained audit fees ..................................................... 73

Table 3.8 Analysis using industry specialization at the MSA city level .................... 76


Table 4.2 Analysis using accrual-based audit quality proxies ................................. 95

Table 4.3 Earnings quality model ............................................................................ 97

Table 4.4 Earnings quality model: Estimation of industry specialist effect .............. 98

Table 4.5 Auditors’ going-concern opinion model ................................................. 100

Table 4.6 Differences in earnings quality among Big 4 audit firms using signed discretionary accruals ............................................................................ 102

viii

List of Figures

Figure 1.1 The significant mergers among the largest audit firms .............................. 9

Figure 1.2 Audit quality framework ........................................................................... 10

Figure 2.1 Residual disclosure of the 10-K reports ................................................... 29

Figure 3.1 The Vault’s Annual Accounting survey .................................................... 61

Figure 3.2 The World’s 10 Most Powerful Brands .................................................... 62

Figure 4.1 Assignments of industry specialization across major audit firms ............. 92

ix

List of Acronyms

10-K An annual report required by the U.S. Securities and Exchange

Commission (SEC)

AICPA American Institute of Certified Public Accountants

AIMR Association for Investment Management and research

BIG 4 The four largest firms in the accounting and consulting industry:

PricewaterhouseCoopers, Ernst & Young, Deloitte, and KPMG.

CAQ The Center for Audit Quality

DV Dependent variable

ERC Earnings response coefficient

EY Ernst & Young (formally shortens its brand name to EY)

GAAP Generally accepted accounting principles

GAAS Generally accepted auditing standards

GAO Government Accountability Office

GC Going-concern

IAASB The International Auditing and Assurance Standards Board

KPMG Klynveld Peat Marwick Goerdeler (formally shortens its brand

name to KPMG)

MD&A Management discussion and analysis

MSA Metropolitan statistical area

PCAOB Public Company Accounting Oversight Board

PSM Propensity-score matching

PwC PricewaterhouseCoopers (formally shortens its brand name to

PwC but legally remains PricewaterhouseCoopers)

SEC The U.S. Securities and Exchange Commission

SFU Simon Fraser University

SPEC Auditor industry specialization measure

UAF A measure of unexplained audit fees

x

Preface

This dissertation is original, unpublished, independent work by the author, Nattavut

(Simon) Suwanyangyuan.

1

Chapter 1. Introduction

This dissertation is a collection of three essays with a primary focus on auditor

competencies to deliver high audit quality.

Because public companies are required to have their financial statements audited

by independent public audit firms, independent auditors play a critical role by acting in the

public interest and contributing to the credibility of financial statements on which they

report beyond management’s own assertions. However, less is known about whether the

influence of external auditors significantly contributes to variations in disclosure practices

in 10-K reports. While the auditor report states that the external auditor is solely

responsible for expressing an audit opinion on the financial information provided by

management, some argue that the auditors’ responsibility is not limited to an assurance

of GAAP compliance on a pass-fail basis (e.g., DeFond et al. 2017b).

In the first essay (Chapter 2), I highlight the extent to which the client’s choice of

external auditor contributes to variations in 10-K disclosure volume. Specifically, I find that

the choice of Big 4 auditors1 is significantly associated with increased 10-K length. This

relation appears to be stronger for audit clients with either poorer earnings quality or higher

information asymmetry, thus supporting the influential role of auditors in assisting their

clients in the form of improved disclosure quality. More importantly, the portion of 10-K

length unexplained by operating complexity and observable clients’ characteristics is

significantly associated with higher audit fees and an increased likelihood of going-

concern (hereafter GC) opinions. I argue that abnormally long disclosures induce external

auditors to reduce the risk of material misstatement through additional audit efforts; hence,

they charge fee premiums as compensation for the risk premium (Simunic and Stein 1996)

1 After the significant mergers of the 1980s and 1990s, the “Big 4” refers to the four largest audit

firms in the world: PricewaterhouseCoopers (PwC), Ernst & Young (EY), Deloitte and Klynveld Peat

Marwick Goerdeler (KPMG).

2

or issue more GC opinions to reduce their exposure to litigation risk (Kaplan and Williams

2013).

In addition to the general effect of auditor size, some researchers have begun to

examine the cross-sectional audit quality variation within Big 4 auditors. While auditor

industry specialization is often argued to capture auditor competencies to supply higher

quality audits, the lack of a consistent definition and existing measurement errors are

considered major challenges inherent to this stream of literature. For example, Audousset-

Coulier et al. (2015) conduct a comprehensive test using different combinations of

measurement variables and approaches to identify industry specialists and conclude that

the use of different auditor industry specialization proxies leads to significantly different

results regarding the impact of industry expertise in audit fee and earnings quality models.

Thus, I reinforce the importance of audit firms’ competencies and focus on subtle

variations in audit pricing and audit quality at the inter-audit firm level using the following

two auditor characteristics: (1) the audit firm’s competencies and (2) auditor industry

specialization.

In chapter 3, instead of treating all the Big 4 auditors as a homogeneous set of

audit firms with regard to their brand reputation, the results of audit pricing regressions

show that the estimated positive coefficient of PwC is significantly higher than those of the

other Big 4 firms, indicating that PwC tends to earn above-average fee premiums relative

to the other Big 4 firms. Similarly, in chapter 4, the results of audit quality regressions show

that the existing differences among the Big 4 firms, primarily regarding the PwC effect,

lead to variations in audit quality as measured by either accrual quality or the likelihood of

issuing GC opinions. Together, these results reinforce the importance of audit firm

competencies and suggest that not all Big 4 audit firms are the same.

In response to a severe measurement issue in auditor industry specialization

research (e.g., Audousset-Coulier et al. 2015; Cahan et al. 2011; Knechel et al. 2007; Li

et al. 2010), I find that PwC and Ernst & Young (hereafter EY) account for a significant

within-industry market share across almost all industries and hence are more likely to be

designated industry specialists according to the market share approach. However, I find

that the relationships between (1) industry expertise and audit fees and (2) industry

expertise and audit quality appear to be consistent only for PwC specialists but are weaker

or show no relation for other (non-PwC) specialists. This indicates that the evidence of

3

auditor industry specialization is exaggerated by the confounding effect of PwC audits and

subsequently leads to the inconsistencies found in empirical archival audit research.

This dissertation, similar to many other papers in this genre, confronts a major

challenge in addressing the self-selection issue and identifying factors that drive the

supply of audit quality. First, I introduce the use of propensity-score matching (hereafter

PSM) to address the identification concerns related to functional form misspecification by

controlling for differences in client characteristics between Big 4 and non-Big 4 groups

while estimating auditor treatment effects. I take this approach because large companies

primarily consider Big 4 firms as their external auditors (e.g., GAO 2008), as evidenced by

the association between Big 4 choice and observable client characteristics, such as client

size, firm performance and financial leverage. Second, I rely on observable input

measures of audit quality, including auditor size, auditor industry specialization and audit

firm competencies. I acknowledge that this reliance on discrete input-based audit quality

measures implicitly assumes a homogenous level of audit quality within each group and

fails to capture subtle variations in audit quality.

1.1 Audit quality overview

“The [audit quality] indicators are meant to be a tool. As such, they have

inherent limitations that have to be recognized if they are to be effective.

They do not lead directly to formulas for determining the quality of a

particular audit or whether an auditor has met its obligations.”

PCAOB Concept Release on Audit Quality Indicators (July 2015)

The concept of audit quality is fundamental in auditing research; however, there is

no consensus on a definition of audit quality or on relevant indicators to assess audit

4

quality, since the term “audit quality” can be viewed from several perspectives2. One of

the most cited definitions of audit quality is that by DeAngelo (1981, page 186), which

states that “the quality of audit services is defined to be the market-assessed joint

probability that a given auditor will both discover a breach in the client’s accounting system

and report the breach”. DeFond and Zhang (2014) also view audit quality as a function of

client demand and auditor supply, which are jointly affected by regulatory intervention, as

depicted in Figure 1.2. The authors provide a comprehensive definition of audit quality,

with higher audit quality providing greater assurance of high financial reporting quality,

conditioned on the firm’s reporting system and innate characteristics.

[Insert Figure 1.2 here]

While public companies are required under federal securities laws to have their

financial statements audited by an independent public accountant to ensure compliance

with regulations, the auditing literature provides compelling evidence that auditing adds

value by providing reasonable assurance that financial statements comply with GAAP and

faithfully reflect the client’s underlying economics. For example, GC opinions are

considered direct and useful communications from the auditor to help financial statement

users predict bankruptcy (e.g., Chen and Church 1996; Menon and Williams 2010). The

stock market’s reaction to auditor changes is also consistent with the notion that auditor

switches convey useful information to the market (e.g., Boone and Raman 2001; Chang

et al. 2010; Khalil et al. 2011). More importantly, incentives to minimize agency conflicts

(e.g., Jensen and Meckling 1976) affect companies’ demand for high-quality audit services

through the client’s choice of auditor characteristics, such as Big 4 firms or auditor industry

specialization (e.g., DeAngelo 1981; DeFond and Zhang 2014; Knechel et al. 2007).

The above is consistent with the survey results of the GAO (2008), which indicate

that the ability to handle complex company operations, technical capabilities and industry

expertise are considered the major reasons why large public companies primarily choose

2 The GAO (2004) provided a more detailed definition of a quality audit as an audit conducted in

accordance with generally accepted auditing standards (GAAS) to provide reasonable assurance

that the audited financial statements and related disclosures are (1) presented in conformity with

GAAP and (2) are not materially misstated whether due to errors or fraud. In addition, the IAASB

(2014) identified a number of elements that create an environment that maximizes the likelihood

that quality audits will be performed on a consistent basis, including: (1) inputs, (2) process, (3)

outputs, (4) key interactions within the financial reporting supply chain, and (5) contextual factors.

5

Big 4 firms as their external auditors. Since large auditors are considered to have stronger

competencies to deliver high-quality audits and to have fewer incentives to behave

opportunistically, auditor size, as measured by Big 4 membership, is widely used in the

auditing literature. This measure has been shown to be associated with almost all other

audit quality proxies, including a lower incidence of accounting fraud (e.g., Lennox and

Pittman 2010), a lower incidence of accounting restatements (e.g., Eshleman and Guo

2014), lower discretionary accruals (e.g., Becker et al. 1998; Francis et al. 1999b), higher

audit fees (e.g., Craswell et al. 1995; Hay et al. 2006), increased ERCs (e.g., Teoh and

Wong 1993), improved analyst earnings forecasts (e.g., Behn et al. 2008), and a lower

cost of debt and equity (e.g., Khurana and Raman 2004).

Contrary to popular belief regarding the auditor’s responsibility to express an audit

opinion without influencing the overall disclosure strategy of his or her audit clients, I

examine whether there is a link between the client’s choice of Big 4 auditors and disclosure

quality in the form of improved informativeness of disclosures, as this remains an

unexplored empirical question in the literature (Chapter 2).

Next, while the use of Big 4 membership implicitly assumes a homogenous level

of audit quality within Big 4 audit firms, it fails to capture subtle variations in audit quality.

The research on auditor competencies then attempts to investigate the audit quality

differentiation that occurs on the inter-audit firm level. For example, Craswell et al. (1995)

argue that the demand for quality-differentiated audits drives audit firm investments in the

development of both brand name and industry-specialized expertise and hence results in

higher audit fees. Additionally, from practitioners’ perspective, PwC (2012) states as

follows:

“Large global accounting networks have emerged in response to the

demands of multinational companies which require their auditors to have a

similar global reach and consistent auditing expertise around the world.

Over many years, those networks have invested substantially in developing

necessary tools and skills to meet the market’s demands for high quality

audits across the world.” (page 1)

Although the concept of auditor industry specialization has been extensively

examined in the auditing literature, there is no consensus on how to empirically measure

6

specialization. Because the level of specialization of audit firms is unobservable, several

indirect proxies have been introduced to capture the complexity of the auditor industry

specialization concept, including market share-based measures (e.g., Craswell et al.

1995; Zeff and Fossum 1967), the portfolio proportion of clients (e.g., Kwon 1996), and

the use of weighted market shares (e.g., Neal and Riley 2004). Regardless of the

approach used to define industry specialization, all these proxies define an industry

specialist as “an audit firm” that has differentiated itself from its competitors in the audit

market. This then raises the empirical question whether there exists a possible

confounding effect of an individual audit firm’s competencies with the effect of industry

specialization on audit pricing and on audit quality because in addition to the evidence on

auditor size and auditor industry specialization, the individual audit firm’s competencies

have long been argued to play an important role in delivering high audit quality.

The classic study of Simunic (1980) provides the first evidence that there is a

significant coefficient for Price Waterhouse3 (hereafter PW), indicating that there is price

competition with a differentiated product for PW in the audit market. This is consistent with

Craswell et al. (1995), who argue that the demand for quality-differentiated audits

motivates auditors to significantly invest in brand name reputation and industry

specialization for higher quality audits and hence results in higher audit fees. Ferguson

and Scott (2014) also find evidence that the Big 4 fee premium in the Australian market

during 2002 – 2004 is largely driven by a robust PwC brand premium, indicating that

individual brand name reputation is considered the basis for within-Big 4 product

differentiation.

Given the increased interest but limited evidence on individual audit firms’

competencies, I attempt to investigate audit pricing differences (Chapter 3) and audit

quality differences (Chapter 4) within Big 4 firms in the U.S. market during the period from

2004 – 2014. Additionally, as both PwC and EY are more likely to be designated industry

specialists across almost all industries, I further argue that the individual differences

detected within Big 4 firms can be used to explain the inconsistent results regarding the

effects of industry specialization on audit pricing and audit quality.

3 PricewaterhouseCoopers (later PwC) was later formed in 1998 from a merger between Price

Waterhouse and Coopers & Lybrand, as illustrated in Figure 1.1.

7

1.2 Audit fees and its determinants

A large body of audit fee research has examined the determinants of audit fees

over the past four decades. Simunic (1980) developed a positive model of the process by

which audit fees are determined and focused on various client characteristics that are

associated with audit fees, including client size, client risk and client complexity. Because

the evidence in audit fee research became more widespread in recent decades, I follow a

summary of the empirical evidence (Hay 2013; Hay et al. 2006) to identify observable

client attributes, auditor attributes and engagement attributes that have been shown to be

significantly associated with audit fees and have been widely used in the literature,

including total assets (LNASSET), current ratio (CURRENT), the ratio of inventory and

receivables to total assets (INVREC), financial leverage (LEVERAGE), return on assets

(ROA), the existence of foreign operations (INTL), mergers and acquisitions (MA), the

incurrence of special items (SPI_DM), the number of business segments (LNBUSSEG),

the incurrence of a loss (LOSS), market-to-book ratio (MTB), busy season of audit

engagement (BUSY), audit tenure (TENURE), initial public offerings (IPO), seasonal

equity offerings (SEO), audit opinion (OPINION), high litigation industry (HIGHLIT), and

the presence of Big 4 audit firms (BIG4).

Prior studies once introduced “the number of audit reports to be issued” as another

determinant of audit fees based on the argument that audit fees increase when the

reporting requirements are more complex and higher overall audit quality is demanded

(e.g., Palmrose 1986). Instead of using the number of audit reports, which seems to

capture reporting complexity with relatively large measurement error, or the now-

discontinued AIMR scores4, as evidenced in Dunn and Mayhew (2004), I follow Cazier

and Pfeiffer (2015) and seek to investigate whether the portion of 10-K disclosure volume,

which is unexplained by observable client characteristics and operating complexity,

contributes to variations in audit fees (Chapter 2).

Additionally, in Chapter 3, instead of treating all Big 4 firms as a homogeneous set

of audit firms with regard to their brand reputation, I break down the indicator variable of

4 Disclosure informativeness is argued to be the main driver of AIMR scores (Lang and Lundholm

1993) based on the assumption that financial analysts determine the disclosure quality of firms on

the basis of the adequacy of disclosures and the firm’s effectiveness in communicating with

investors.

8

Big 4 firms and jointly introduce indicator variables for each of the Big 4 firms to examine

the differential effects of each audit firm in the audit pricing model.

9

1.3 Figures

Figure 1.1 The significant mergers among the largest audit firms

Source: GAO (2008), page 9

10

Figure 1.2 Audit quality framework

Source: DeFond and Zhang (2014), page 280

Client Demand Auditor Supply

Incentives Incentives

e.g. agency costs,

regulationAudit Quality

e.g. reputation, litigation,

regulation

Competencies Competencies

e.g. audit committee,

internal audit function

e.g. inputs to the audit

process, expertise

Regulatory

Intervention

11

Chapter 2. Auditor choice and 10-K disclosure volume

2.1 Introduction

While the auditor's report clearly states that the auditor’s responsibility is to express

an opinion on the financial statements, there is a controversy over whether the role of the

external auditor is limited to mere GAAP compliance or rather extends to a responsibility

to assure a fair presentation to the capital markets (e.g., DeFond et al. 2017b). This debate

occurs because while management is solely responsible for the preparation and fair

presentation of financial statements that are free from material misstatement, the auditor

must plan and perform the audit to obtain appropriate and sufficient audit evidence. This

includes assessments of the accounting principles used and the significant accounting

estimates made by management as well as an evaluation of the overall financial statement

presentation. Thus, an empirical question is raised regarding whether variations in audit

quality significantly lead to variations in disclosure practices in 10-K reports.

Contrary to popular belief on the role of the auditor, the auditing standards [AU

Section 5505] explicitly require auditors to read the annual report and consider whether

such unaudited information (e.g., MD&A), or the manner of its presentation, is materially

inconsistent with information, or the manner of its presentation, appearing in the financial

statements (AICPA 1997). This is consistent with a detailed summary of observations from

roundtable discussions6 on the evolving role of the auditor, stating that

“... Less sophisticated investors may not be aware that auditors currently

provide some value by reading other information provided outside of the

5 The current version of AU Section 500 [Other Information in Documents Containing Audited

Financial Statements] is AS 2710, which is effective on or after December 31, 2016.

6 The roundtable discussions on the evolving role of the auditor were sponsored by the Center of

Audit Quality in 2011 and included the full range of financial reporting stakeholders – CEOs, CFOs,

board and audit committee members, investors, auditors, former regulators, attorneys and

academics.

12

audited financial statements for consistency with the audited financial

statements …” (CAQ 2011, page 7)

In this study, I focus on determining whether the client’s choice of Big 4 auditors

contributes to cross-sectional variations in 10-K disclosure volume. This approach is

consistent with Dunn and Mayhew (2004), who argue that a client’s choice of industry

specialists is associated with the client’s intention to provide enhanced disclosure.

However, instead of using the now-discontinued AIMR scores, I use the length of 10-K

reports (e.g., Li 2008; Loughran and Mcdonald 2014) over the eleven-year period from

2004 through 2014 and reveal that the choice of Big 4 auditors is positively associated

with 10-K disclosure volume in both the full sample and the propensity-score matching

(hereafter PSM) sample (e.g., Lawrence et al. 2011), thus supporting the notion that the

audit quality differentiation between Big 4 and non-Big 4 auditors contributes to variations

in 10-K disclosure volume.

Since there is strong evidence that Big 4 auditors are of higher quality than non-

Big 4 auditors and are associated with either lower level of discretionary accruals (e.g.,

Becker et al. 1998; Francis et al. 1999b) or a reduction in information asymmetry (e.g.,

Jensen and Meckling 1976; Watts and Zimmerman 1983), I expect the association

between Big 4 auditor choice and 10-K disclosure volume to be stronger in situations

where financial reporting users potentially need more information to understand the effect

of material transactions and/or events on the information conveyed in the financial

disclosures. As expected, the benefit of enhanced disclosures provided by Big 4 auditors

is more pronounced for audit clients with poorer accrual quality and for those with higher

information asymmetry, thus supporting the notion that the choice of Big 4 auditors signals

a client’s intention to provide enhanced disclosure quality (e.g., Hutton et al. 2003;

Baginski et al. 2004; Mercer 2004; D'Souza et al. 2010).

Next, I provide new evidence that the portion of 10-K length unexplained by

operating complexity and observable client characteristics induces higher audit effort in

performing audit services and is associated with higher audit fees and an increased

likelihood of GC opinions. In other words, abnormally long disclosures induce external

auditors to lower the risk of material misstatement through additional audit effort; hence,

they charge audit fee premiums to compensate for the risk premium (Simunic and Stein

13

1996) and issue more GC opinions to reduce their exposure to litigation risk (Kaplan and

Williams 2013).

My study makes several contributions to the literature. First, this study contributes

to the broader question of how auditor firm size is associated with the quality of a firm’s

disclosure by providing evidence of a Big 4 superiority effect through enhanced disclosure

practices in 10-K reports. This suggests that the auditor’s responsibility is not limited to an

assurance of GAAP compliance but also extends to a faithful representation of the firm’s

underlying economics, as evidenced by increased 10-K length. Second, this study fills a

gap in the literature regarding the textual analysis of corporate disclosures, as prior studies

have mainly focused on the managerial discretion in firms’ disclosure practices, by

highlighting the extent to which the client’s choice of Big 4 auditors contributes to variations

in disclosure practices in 10-K reports. Given the regulatory concerns regarding corporate

disclosure and the trend toward more detailed disclosure, this study provides useful

insights into the evolving role of external auditors in the reporting process and should be

of interest to both the SEC and PCAOB. Finally, because an abnormally long disclosure

induces external auditors to provide additional audit effort, I argue that audit effort can be

inferred from a discretionary component of 10-K disclosure volume.

The remainder of this paper is organized as follows. In section 2.2, I review the

relevant literature and develop the hypotheses. I present my research design in section

2.3. I report sample characteristics and descriptive statistics in section 2.4. Sections 2.5

and 2.6 present empirical results and additional analyses for robustness checks,

respectively. Finally, section 2.7 concludes the study.

2.2 Literature Review and Hypotheses Development

2.2.1 Literature review

This study builds on and contributes to two areas of research: (1) research on audit

quality and (2) research on corporate reporting and disclosure.

14

Research on audit quality

While the conceptual nature and definition of audit quality have been discussed for

several decades, there is no consensus on a definition of audit quality or on relevant

indicators to measure audit quality. This is because different perspectives of audit quality

infer different proxies for it (e.g., PCAOB 2015). One of the most cited definitions of audit

quality is that by DeAngelo (1981, page 186), which states that “the quality of audit

services is defined to be the market-assessed joint probability that a given auditor will both

discover a breach in the client’s accounting system and report the breach”. More

importantly, the author argues that large auditors are expected to have stronger incentives

and competencies to supply high audit quality, which has motivated much of the auditing

literature to use auditor size as a proxy for audit quality.

While public companies are required under federal securities laws to have their

financial statements audited by an independent public accountant to ensure compliance

with regulations, research on Big 4 audit quality provides ample evidence that Big 4

auditors deliver higher audit quality, as captured by various output-based audit quality

proxies, including a lower incidence of accounting fraud (e.g., Lennox and Pittman 2010),

a lower incidence of accounting restatements (e.g., Eshleman and Guo 2014), lower

discretionary accruals (e.g., Becker et al. 1998; Francis et al. 1999b), higher audit fees

(e.g., Craswell et al. 1995; Hay et al. 2006), increased ERCs (e.g., Teoh and Wong 1993),

improved analyst earnings forecasts (e.g., Behn et al. 2008), and a lower cost of debt and

equity (e.g., Khurana and Raman 2004).

In another extension, the auditing literature provides compelling evidence that the

client’s choice of auditor characteristics potentially signals client incentives to demand high

audit quality, as evidenced by the stock market’s reaction to auditor switches (e.g., Boone

and Raman 2001; Chang et al. 2010; Khalil et al. 2011; Knechel et al. 2007) and enhanced

disclosure quality (Dunn and Mayhew 2004). This is consistent with the survey results of

the GAO (2008), which indicate that the ability to handle complex company operations,

technical capabilities and industry expertise are considered major reasons why large

public companies primarily choose Big 4 audit firms as their external auditors.

15

Research on corporate reporting and disclosure

A large body of research on corporate reporting and disclosure has focused on the

benefits of increased disclosures and has argued that an increased volume of firm

disclosures is associated with reduced information asymmetry, lower cost of immediacy,

higher trading activity and an overall improvement in the efficiency of information price

discovery (e.g., Diamond and Verrecchia 1991; Botosan 1997; Leuz and Verrecchia 2000;

Graham et al. 2005; Balakrishnan et al. 2014). However, there is a line of research on the

textual analysis of corporate disclosures that raises significant concerns regarding the

relevance of information in financial disclosures based on the presumption that “longer

and less readable documents are more deterring and require higher costs of information-

processing” (Li 2008, page 222). For example, Nelson and Pritchard (2007) find that

companies that are subject to more shareholder litigation use more readable language in

their reports and avoid boilerplate warnings, while You and Zhang (2009) document that

investors underreact to the information provided in 10-K filings, with a more pronounced

effect for companies that file more complex and less readable 10-K reports. Lawrence

(2013) also finds that individual investors are more likely to invest in firms that provide

clear and concise disclosures relative to other firms. Together, these findings suggest that

detailed and lengthy disclosures potentially create a risk of information overload and make

it more difficult for the intended users to identify the information that is most relevant.

10-K disclosure volume, as measured by the number of words in a 10-K filing, was

first introduced to capture annual report readability (e.g., Li 2008; Loughran and Mcdonald

2014). In a subsequent study by Loughran and Mcdonald (2016), the authors argue that

it is not possible to disentangle the complexity of a firm’s business from the readability of

its annual reports, and they recommend that researchers focus on a broader concept of

information complexity. Cazier and Pfeiffer (2015) use a small sample of 10-Ks and

partition the disclosure volume of 10-K reports into three major components: (1) firms’

operating complexity, (2) disclosure redundancy and (3) residual disclosure. The authors

argue that while the disclosure volume of 10-K reports is largely driven by operating

complexity and disclosure redundancies, a substantial amount of disclosure volume is

attributable to a discretionary reporting choice by management; hence, they call for future

research to investigate the factors that drive idiosyncratic disclosure, which is not

explained by either operating complexity or disclosure redundancies.

16

Prior studies further indicate that the choice of external auditor has a significant

influence on client disclosure quality. For example, Dunn and Mayhew (2004) show that

industry specialists provide value-added services, including disclosure advice, to their

audit clients in the form of improved disclosure quality. This is because when determining

whether financial statements are fairly presented or not, an auditor has to consider

whether the “information presented in the financial statements, including accounting

policies, is relevant, reliable, comparable, and understandable”, and whether the “financial

statements provide sufficient disclosures to enable users to understand the effect of

material transactions and events on the information conveyed in the financial statements”

(PCAOB 2005, page 8). Thus, the active role of the external auditor in the client’s

accounting and disclosure choices affects the content of the client’s financial statements

because the auditor must ensure that the financial statements are appropriate (Gibbins et

al. 2001).

2.2.2 Hypotheses development

The argument leading to my first hypothesis concerns the broader view of auditors’

responsibilities raised by DeFond et al. (2017b). Although both Big 4 and non-Big 4

auditors are held to the same regulatory and professional standards, financial statement

users expect high-quality auditors to consider more than technical GAAP compliance

when determining whether financial statements are fairly presented. Thus, whether and to

what extent the choice of Big 4 auditor affects the disclosure volume of 10-K reports is an

empirical issue, which is the focus of this study.

Because theory suggests that Big 4 auditors have greater incentives to maintain

high levels of audit quality and that Big 4 auditors are sought because of their incentives

and competencies to enhance the credibility of financial reporting7, the choice of Big 4

auditors should therefore help audit clients improve the informativeness of their

disclosures. Instead of using the now-discontinued AIMR scores, as evidenced in Dunn

7 Mercer (2004) suggests that the credibility of financial reporting is influenced by various factors,

including the degree of external assurance and the characteristics of the disclosure (e.g. precision,

venue, horizon and amount of supporting information). For example, supplementary disclosures

are used as a firm strategy to enhance the credibility of earnings forecasts by increasing the content

and ex post verifiability of them (e.g., Hutton et al. 2003).

17

and Mayhew (2004), I use 10-K disclosure volume to capture the informativeness8 of client

disclosures.

The existing literature often uses text-based analyses to estimate various proxies

for readability and complexity, such as 10-K document length and file size, based on the

general consensus that firms with annual reports that are less complex and easier to read

have more persistent positive earnings, experience smaller underreactions to earnings

news and attract more individual investors. Part of the evidence can be attributed to

Bloomfield (2002) incomplete revelation hypothesis, which states that “statistics that are

more costly to extract from public data are less completely revealed in market prices”.

Thus, because less readable and more complex 10-K reports likely provide managers with

more opportunities to withhold bad news from the market, this line of reasoning

hypothesizes a negative association between the choice of Big 4 auditors and 10-K

disclosure volume, which would indicate that the clients of Big 4 auditors benefit from clear

and concise corporate disclosures.

However, as Bloomfield (2008) later notes, firms could simply require longer and

more detailed explanations to support certain complex structural transactions and events,

and they respond to changes in their information environment by voluntarily increasing

both the quantity and frequency of their filings relative to what is mandated by market

regulators. If more detailed annual reports reflect new value-relevant information and are

indicative of higher reporting quality, this line of reasoning hypothesizes a positive

association between the choice of Big 4 auditors and 10-K disclosure volume, which would

indicate that the clients of Big 4 auditors benefit from improved disclosure quality through

longer and more detailed disclosures.

In sum, whether and to what extent the choice of Big 4 auditors affects the

disclosure volume of 10-K reports is an empirical issue, which is the focus of this study.

The first hypothesis is then formulated, in null form, as follows:

8 Disclosure informativeness has been argued to be the main driver of AIMR scores (Lang and

Lundholm 1993) based on the assumption that financial analysts determine the disclosure quality

of firms based on the adequacy of disclosure and the firm’s effectiveness in communicating with

investors.

18

H1: The choice of Big 4 auditors is not associated with 10-K disclosure volume.

Next, I investigate to what extent the choice of Big 4 auditors impacts 10-K

disclosure volume in the two following situations where financial reporting users potentially

need more information to understand the effects of material transactions and events on

the information conveyed in the financial disclosure.

First, while earlier research provides evidence that the clients of Big 4 auditors

report lower discretionary accruals on average than those of non-Big 4 auditors (e.g.,

Becker et al. 1998; Francis et al. 1999b), I hypothesize that the influence of Big 4 auditors

on 10-K disclosure volume will be more pronounced for clients with poorer accrual quality,

as measured by the magnitude of discretionary accruals, thus supporting these clients’

attempts to increase the credibility of their financial reports. Second, because companies

seek to shape their information environment by voluntarily disclosing more information to

reduce information asymmetries (e.g., Balakrishnan et al. 2014), I use the effective bid-

ask spread to proxy for information asymmetry and hypothesize that the influence of Big

4 auditors on 10-K disclosure volume will be more pronounced for clients with higher levels

of information asymmetry, thus supporting these clients’ attempts to decrease the

information asymmetries. Taken together, the second hypothesis is then formulated as

follows:

H2a: The association between the magnitude of discretionary accruals and 10-

K disclosure volume increases with the presence of Big 4 auditors.

H2b: The association between the effective bid-ask spread and 10-K disclosure

volume increases with the presence of Big 4 auditors.

Finally, because auditors are responsible for examining firms’ financial reporting

and expressing an opinion on its fairness, I expect that more detailed financial reporting

requires higher costs for information processing (e.g., Bloomfield 2002; Li 2008) together

with more effort in performing the audit services.

Given that the auditor’s effort level is not observable, researchers often use audit

hours to capture audit effort (e.g., Caramanis and Lennox 2008; Palmrose 1986).

However, because data availability is a major limitation, audit efforts can also be inferred

from a variety of observable auditor responses, such as audit fees and audit opinion. This

is partly because auditors can reduce the risk of material misstatement by increasing their

19

effort to avoid legal liability due to potential audit errors (e.g., McCracken 2002). Hence,

they can charge higher audit fees (Simunic and Stein 1996) and/or increase GC opinions

(Kaplan and Williams 2013). Therefore, if the influence of auditors contributes to variations

in 10-K disclosure volume, I predict that the residual disclosure of 10-K reports will be

associated with either higher audit fees or a higher propensity to issue modified audit

opinions. In other words, incremental audit effort, as measured by the amount by which

actual 10-K disclosure volume exceed predicted 10-K disclosure volume, would reflect a

greater audit effort in providing assurance services to audit clients. The third hypothesis

is then formulated, in null form, as follows:

H3a: Higher residual disclosure is not associated with higher audit fees.

H3b: Higher residual disclosure is not associated with an increased likelihood to

issue GC opinions.

2.3 Research Design

To test the first hypothesis, I include an indicator variable for Big 4 audit firms to

examine the differential effect of Big 4 auditor choice on 10-K disclosure volume. Based

on existing disclosure studies (e.g., Li 2008; Cazier and Pfeiffer 2015), I estimate the

following empirical model:

Equation 1: The disclosure model

𝐋𝐍𝐖𝐎𝐑𝐃𝐒𝐭 = α0 + α1𝐁𝐈𝐆𝐍𝐭 + α2𝐃𝐄𝐋𝐓𝐀_𝐑𝐎𝐀𝒕 + α3𝐃𝐄𝐋𝐓𝐀_𝐑𝐄𝐕𝐭 + α4𝐌𝐀𝐢𝐭

+ α5𝐅𝐘_𝐑𝐄𝐓𝐭 + α6𝐒𝐃_𝐑𝐄𝐓𝐔𝐑𝐍𝐭 + α7𝐒𝐏𝐈_𝐃𝐌𝐭 + α8𝐂𝐀𝐏_𝐋𝐄𝐀𝐒𝐄𝐭

+ α9𝐎𝐏_𝐋𝐄𝐀𝐒𝐄𝐭 + α10𝐑𝐃𝐭 + α11𝐈𝐍𝐓𝐀𝐍𝐆𝐭+ α12𝐒𝐈𝐙𝐄𝐭 + 𝛼13𝐋𝐍𝐀𝐆𝐄𝐭

+ α14𝐌𝐓𝐁𝐭 + α15𝐅𝐂𝐅𝐭 + α16𝐃𝐄𝐑𝐈𝐕𝐀𝐓𝐈𝐕𝐄𝐭

+ α17𝐋𝐍𝐁𝐔𝐒𝐒𝐄𝐆𝐭 + α18𝐋𝐍𝐆𝐄𝐎𝐒𝐄𝐆𝐭 + 𝛼19𝐒𝐃_𝐎𝐈𝐀𝐃𝐏𝐭

+ α20𝐃𝐄𝐋𝐀𝐖𝐀𝐑𝐄𝐭 + α21𝐈𝐏𝐎𝐭+ α22𝐒𝐄𝐎𝐭 + 𝛼23𝐋𝐍𝐍𝐌𝐂𝐎𝐔𝐍𝐓𝐭

+ 𝐘𝐞𝐚𝐫 𝐚𝐧𝐝 𝐈𝐧𝐝𝐮𝐬𝐭𝐫𝐲 𝐟𝐢𝐱𝐞𝐝 𝐞𝐟𝐟𝐞𝐜𝐭𝐬 + ε𝐭

See Table 2.1 for variable definitions and descriptive statistics.

20

To address the identification concerns9 related to functional form misspecification

(e.g., Boone et al. 2010; Lawrence et al. 2011), I use a propensity-score matching model10

to control for differences in client characteristics between Big 4 and non-Big 4 auditors

while estimating auditor treatment effects. Specifically, I estimate the following logistic

regression (Equation 2) and obtain the probability of hiring a Big 4 auditor based on a

broad range of observable client characteristics, including asset size, asset turnover,

current ratio, financial leverage and firm performance, together with the control variables

used in the disclosure model.

Equation 2: The auditor selection model

𝐁𝐈𝐆𝟒𝐭 = α0 + α1𝐋𝐍𝐀𝐒𝐒𝐄𝐓t + α2𝐀𝐓𝐔𝐑𝐍𝐭 + α3𝐂𝐔𝐑𝐑𝐄𝐍𝐓𝐭 + α4𝐋𝐄𝐕𝐄𝐑𝐀𝐆𝐄𝐭

+ α5𝐑𝐎𝐀𝐭 + Σ𝐂𝐨𝐧𝐭𝐫𝐨𝐥 𝐯𝐚𝐫𝐢𝐚𝐛𝐥𝐞𝐬

+ 𝐘𝐞𝐚𝐫 𝐚𝐧𝐝 𝐈𝐧𝐝𝐮𝐬𝐭𝐫𝐲 𝐟𝐢𝐱𝐞𝐝 𝐞𝐟𝐟𝐞𝐜𝐭𝐬 + ε𝐭


After obtaining the fitted values from Equation 2, I match, without replacement,

each client of a Big 4 auditor with a client of a non-Big 4 auditor that has the closest fitted

value in the same fiscal year and corresponding two-digit SIC code industry within a

maximum distance of 0.03 between the two propensity scores. This procedure creates a

pseudo-random sample in which one group of firms (the treatment group) is audited by

Big 4 audit firms, while the other group (the control group) is not audited by Big 4 audit

firms. Since the variation in the client characteristics is minimized through the propensity-

score matching procedure, the remaining differences in means between the treatment and

control groups are justified to be considered as the treatment effect.

9 Lawrence et al. (2011) show that propensity-score matching on client characteristics eliminates

the Big 4 effects and conclude that differences in quality between Big 4 and non-Big 4 audit firms

largely reflect observable client characteristics, primarily client size.

10 While Shipman et al. (2017) state that the PSM results can be highly sensitive to design choices,

such as caliper distance and whether matching was performed with or without replacement,

DeFond et al. (2017a) find that the results of Lawrence et al. (2011) arise in only a minority of

design choice. In addition, they find evidence supporting the Big 4 effects for most of the audit

quality measures in a majority of the matched samples.

21

To test the second set of hypotheses, I introduce the following two variables to

investigate the incremental effect of Big 4 auditor choice on 10-K disclosure volume: (1)

the magnitude of discretionary accruals and (2) the effective bid-ask spread.

First, with regard to the measurement of opportunistic behavior, I estimate normal

levels of accruals based on the modified Jones model11 (Dechow et al. 1995), which

defines the accrual process as a function of growth in credit sales and investment in PPE,

controlling for firm performance (Kothari et al. 2005). I then decompose total accruals into

discretionary and non-discretionary components, with a larger magnitude of discretionary

accruals (ADA_MJR) indicating more aggressive opportunistic behavior.

Second, following the same approach as Hendershott et al. (2011), I measure the

effective spread (EFFSPRD) as the difference between the bid-ask midpoint and the

actual transaction price divided by the bid-ask midpoint. Specifically, I calculate a volume-

weighted average over the 12-month period, with a larger effective spread indicating less

stock liquidity and hence more information asymmetry.

Because I expect that the benefit of enhanced disclosures provided by Big 4

auditors will be more pronounced for audit clients with poorer accrual quality and those

with higher information asymmetry, I partition the sample into two subsamples using the

median of ADA_MJR and EFFSPRD to examine the differential effects of Big 4 auditor

choice together with its incremental effect through an interaction between BIG4 and either

ADA_MJR or EFFSPRD in the disclosure model.

11 The following accrual model is estimated with a minimum of 20 observations in each industry-

year cluster:

TAt = α0 + α1 (1

ATt−1) + α2 (

(∆REVt − ∆RECt)

ATt−1) + α3 (

PPEt

ATt−1) + α4ROAt + εt

where TAt = the difference between net income and operating cash flow in year t, scaled by lagged

assets; ATt−1 = lagged total assets; ∆REVt − ∆RECt = the change in total revenues less the change

in total receivables in year t from year t-1; PPEt = the gross book value of property, plant and

equipment at the end of year t; and ROAt = income before extraordinary items in year t, scaled by

total assets.

22

To test the last set of hypotheses, I first estimate the disclosure model by fiscal

year12 and obtain the residual (RES_WRD) from this model as the portion of 10-K

disclosure volume unexplained by observable client characteristics and operating

complexity (e.g., Cazier and Pfeiffer 2015; Li 2008).

Building on prior studies, I then investigate whether abnormally long disclosures

trigger a variety of auditor responses through additional audit effort, as evidenced by either

higher audit fees or an increased likelihood of GC opinions. Specifically, I estimate the

following two models with the inclusion of control variables based on audit fee studies

(e.g., Simunic 1980; Hay et al. 2006; Hay 2013) and GC opinion studies (e.g., DeFond et

al. 2002; Reynolds and Francis 2000).

Equation 3: The audit fee model

𝐋𝐍𝐀𝐅𝐄𝐄𝐒𝐭 = α0 + α1𝐋𝐍𝐀𝐒𝐒𝐄𝐓𝐭 + α2𝐂𝐔𝐑𝐑𝐄𝐍𝐓𝐭 + α3𝐈𝐍𝐕𝐑𝐄𝐂𝐭 + α4𝐋𝐄𝐕𝐄𝐑𝐀𝐆𝐄𝐭

+ α5𝐑𝐎𝐀𝐭 + α6𝐈𝐍𝐓𝐋𝐭 + α7𝐌𝐀𝐭 + α8𝐒𝐏𝐈_𝐃𝐌𝐭 + α9𝐋𝐍𝐁𝐔𝐒𝐒𝐄𝐆𝐭

+ α10𝐋𝐎𝐒𝐒𝐭 + α11𝐌𝐓𝐁𝐭 + α12𝐁𝐔𝐒𝐘𝐭 + α13𝐓𝐄𝐍𝐔𝐑𝐄𝐭 + α14𝐈𝐏𝐎𝐭

+ α15𝐒𝐄𝐎𝐭 + α16𝐎𝐏𝐈𝐍𝐈𝐎𝐍𝐭 + α17𝐇𝐈𝐆𝐇𝐋𝐈𝐓𝐭 + α18𝐁𝐈𝐆𝟒t

+ α19𝐑𝐄𝐒_𝐖𝐑𝐃𝐭 + 𝐘𝐞𝐚𝐫 𝐚𝐧𝐝 𝐈𝐧𝐝𝐮𝐬𝐭𝐫𝐲 𝐅𝐢𝐱𝐞𝐝 𝐄𝐟𝐟𝐞𝐜𝐭𝐬 + ε𝐭


Equation 4: The going-concern opinions model

𝐆𝐂𝐭 = 𝛂𝟎 + 𝛂𝟏𝐋𝐍𝐀𝐒𝐒𝐄𝐓𝐭 + 𝛂𝟐𝐌𝐓𝐁𝐭 + 𝛂𝟑𝐋𝐄𝐕𝐄𝐑𝐀𝐆𝐄𝐭 + 𝛂𝟒𝐂𝐇𝐋𝐄𝐕𝐭+ 𝛂𝟓𝐂𝐅𝐎𝐭

+ 𝛂𝟔𝐀𝐋𝐓𝐌𝐀𝐍𝐭 + 𝛂𝟕𝐏𝐋𝐎𝐒𝐒𝐭 + 𝛂𝟖𝐇𝐈𝐆𝐇𝐋𝐈𝐓𝐭 + 𝛂𝟗𝐓𝐄𝐍𝐔𝐑𝐄𝐭

+ 𝛂𝟏𝟎𝐑𝐎𝐀𝐭 + 𝛂𝟏𝟏𝐒𝐃_𝐎𝐈𝐀𝐃𝐏𝐭 + 𝛂𝟏𝟐𝐁𝐈𝐆𝟒𝐭 + 𝛂𝟏𝟑𝐑𝐄𝐒_𝐖𝐑𝐃𝐭

+ 𝐘𝐞𝐚𝐫 𝐚𝐧𝐝 𝐈𝐧𝐝𝐮𝐬𝐭𝐫𝐲 𝐟𝐢𝐱𝐞𝐝 𝐞𝐟𝐟𝐞𝐜𝐭𝐬 + 𝛆𝐢𝐭


12 Estimating (Equation 1) by year allows the intercept and coefficients to vary by year, thereby

controlling for any year-specific events (Hribar et al. 2014).

23

2.4 Sample Selection and Descriptive Statistics

2.4.1 Sample selection

I obtain the available datasets from Loughran and Mcdonald (2014) and focus on

the textual characteristics of 10-K annual reports available on EDGAR during the 2000 to

201413 period. These datasets contain various complexity and readability measures,

including the word counts of the 10-K reports based on words appearing in the Loughran-

McDonald Master Dictionary.

To address my research questions, I merge the datasets with Compustat

fundamental annual files and CRSP monthly stock files to obtain the necessary financial

statement data for all firm-years from 2000 to 2014. I exclude all observations related to

financial (between SIC 6000 and 6999) and utility (between SIC 4900 and 4949) firms. I

delete firms with total assets of less than $1 million and negative book value of equity as

well as firms that have less than 2,000 words in their 10-K reports. I also require that firms

have a stock price of at least $1 or a total market capitalization greater than or equal to

$200 million. After imposing all the necessary requirements to the estimate disclosure

model, I obtain a sample of 43,575 firm-year observations, in which 13,818 (31.7%) and

29,757 (68.3%) reflect non-Big 4 and Big 4 accounting clients, respectively. Using

Equation (2) to calculate the propensity scores and imposing a caliper distance of 3

percent, I obtain a propensity-score matched sample of 13,152 firm-years, of which 6,576

are Big 4 clients and 6,576 are non-Big 4 clients. Finally, I winsorize observations that fall

in the top and bottom 1 percent of the distribution for each non-discrete variable to mitigate

potential problems of outliers in both samples.

2.4.2 Descriptive statistics

Table 2.1 reports the descriptive statistics for all variables used in the disclosure

model (Equation 1) during the 2004 to 2014 period.

Panel A reports the mean summary statistics for the full sample of Big 4 and non-

Big 4 auditors together with their differences in means. Overall, the descriptive results

13 The updated dataset for the fiscal year 2014 was graciously provided to me by Bill McDonald.

24

illustrate that clients of Big 4 auditors are relatively larger in size, more profitable and more

leveraged than those of non-Big 4 auditors. I also document that the mean LNWORDS of

Big 4 and non-Big 4 clients are 10.80 and 10.45, which translates into means of 49,026

and 34,493 words, respectively, indicating that clients of large audit firms tend to provide

more detailed disclosures than those of small audit firms. In Panel B, the PSM sample

based on the auditor selection model results in a total sample of 13,152 observations with

relatively similar client characteristics in which one group of firms is audited by Big 4 and

the other group is audited by non-Big 4 auditors. While the PSM model appears effective

in forming a balanced sample of Big 4 and non-Big 4 auditors, I consistently find that the

average 10-K disclosure volume is still relatively larger for clients of Big 4 auditors (10.61)

than those of non-Big 4 auditors (10.57).

Table 2.2 reports the Pearson (the upper half) and the Spearman (the lower half)

correlation coefficients among the key variables used in this study. First, the high

correlation between LNWORDS and SIZE (𝑟𝑝 = 𝑟𝑠 = 0.46) is consistent with prior studies,

suggesting that a significant portion of 10-K length is attributable to operating complexity.

I also find that BIG4 is positively correlated with LNWORDS (𝑟𝑝 = 𝑟𝑠 = 0.32) and

RES_WRD (𝑟𝑝 = 0.04; 𝑟𝑠 = 0.05) at less than 1% levels, indicating that the influence of

Big 4 auditors potentially contributes to the variation in 10-K disclosure volume. As

expected, the significant correlations between RES_WRD and both LNAFEES (𝑟𝑝 =

0.10; 𝑟𝑠 = 0.11) and GC (𝑟𝑝 = 𝑟𝑠 = 0.03) indicate that abnormally long disclosures are

associated with higher audit fees and the increased likelihood of going-concern opinions

at less than 1% levels.

2.5 Empirical Results

Table 2.3 reports the regression results14 of estimating the disclosure model with

the inclusion of BIG4 on both the full sample (Column 1) and the PSM sample (Column

2). While all explanatory variable coefficients are significant and have directional effects

consistent with those documented in previous studies, I consistently find that the estimated

14 In untabulated results, there is no sign of a severe multicollinearity problem based on the VIF

factors of each independent variable in the disclosure model.

25

coefficient of BIG4 is positive and significant (Coef. = 0.07 with t-statistic = 7.6415 for the

full sample; Coef. = 0.03 with t-statistic = 3.35 for the PSM sample), indicating that the

variation in 10-K reports between Big 4 and non-Big 4 auditors persists with the PSM

sample. This result suggests that the clients of Big 4 auditors benefit from improved

disclosure quality through longer and more detailed 10-K reports.

To examine the incremental effect of Big 4 auditors on 10-K length in the situation

where financial reporting users potentially need more information to understand the effects

of material transactions or events reported in the financial disclosure, I estimate the

disclosure model (Equation 1) with the inclusion of either ADA_MJR or EFFSPRD and its

interaction with BIG4 in Table 2.4. I then partition the full sample into subsamples with low

and high values of ADA_MJR in Columns (1) and (2) and subsamples with low and high

values of EFFSPRD in Columns (3) and (4), respectively.

As expected, I find that the coefficient of BIG4 is positive and significant (Coef. =

0.06 with t-statistic = 3.4 for the full sample; Coef. = 0.04 with t-statistic = 1.74 for the PSM

sample) in the subsample of firms with better accrual quality. Similarly, in the subsample

of firms with poorer accrual quality, I find that the incremental effect of Big 4 auditors, as

captured by the estimated coefficient of ADA_MJR*BIG4 (Coef. = 0.34 with t-statistic =

3.20 for the full sample; Coef. = 0.30 with t-statistic = 1.88 for the PSM sample), is relatively

larger than those reported in the first column. Alternatively, while I find marginal results or

no relation in the subsample of firms with lower levels of information asymmetry, the

estimated coefficient of BIG4 is positive and significant in the subsample of firms with

higher levels of information asymmetry (Coef. = 0.07 with t-statistic = 6.19 for the full

sample; Coef. = 0.05 with t-statistic = 3.84 for the PSM sample).

In sum, these results provide evidence supporting an auditor influence toward

increasing the informativeness of client disclosures, particularly when audit clients report

higher levels of discretionary accruals or experience higher levels of information

asymmetry.

15 In untabulated results, I find that the inferences are unchanged when I implement the two-way

clustering approach proposed by Petersen (2009), which is considered to be a conservative

approach to control for time and firm effects in panel datasets.

26

Next, I estimate the disclosure model by year and obtain residual disclosures as

the portion of 10-K disclosure volume unexplained by observable client characteristics and

operating complexity. RES_WRD and RES_SG are then defined as residuals from

estimating the disclosure model using the word count (LNWORDS) and the gross file size

of the complete 10-K submission text file16 (LNSIZEG) as the dependent variable,

respectively. Specifically, I investigate whether 10-K disclosure volume varies with the

auditor’s influence and induces higher audit effort through charging a fee premium or

issuing more GC opinions to high litigation risk clients. The descriptive statistics for all the

variables used in both models are reported in Table 2.5.

However, as illustrated in Panel A of 2.8 Figure and tables

Figure 2.1, while the mean RES_WRD values of firms that use Big 4 (non-Big 4)

auditors are consistently positive (negative) and do not fluctuate around zero throughout

the sample period, the mean RES_SG values of both Big 4 and non-Big 4 auditors do

appear to fluctuate around zero, with no specific pattern observed. This is partly because

the gross 10-K file size includes other components, such as HTML tags, XBRL codes and

encoded images, which are not relevant to the scope of the auditor’s report and

responsibilities. Therefore, I only use RES_WRD, and not RES_SG, as a discretionary

component in 10-K disclosure volume to capture audit effort.

To prevent the model’s residuals from being correlated with client size, I estimate

the audit fee model (Equation 3) partitioned by asset size quintiles and report the

regression results in Table 2.6. Consistent with my prediction, I find that the estimated

coefficients of RES_WRD are positive and significant at less than 1% levels across all

quintile groups, indicating that audit clients with abnormally long disclosures, on average,

pay higher audit fees.

Additionally, I use the change specification to examine the relation between

residual disclosures and the level of audit fees. The change variables in the model

(denoted by Δ) are then measured as the current year value less the prior year value of

the variables used in the audit fee model. As expected, I find that the year-to-year change

16 Instead of the word count, Loughran and Mcdonald (2014) argue that the gross file size of the

complete 10-K submission text file can be used as a simple proxy that does not require document

parsing, facilitates replication and is correlated with alternative readability constructs.

27

in the residual disclosures (Δ RES_WRD) is positively associated with the year-to-year

change in the level of audit fees (Coef. = 0.03 with t-statistic = 5.82). This result suggests

that firms with an unexpected increase in 10-K disclosure volume pay higher audit fees on

average than in the immediately preceding year. Additionally, I partition the full sample

into subsamples of non-Big 4 clients (Column 2) and Big 4 clients (Column 3), and the

results consistently show that the estimated coefficients on Δ RES_WRD are positive and

significant across both subsamples (Coef. = 0.03 with t-statistic = 1.89 for the subsample

of non-Big 4 auditors; Coef. = 0.03 with t-statistic = 5.57 for the subsample of Big 4

auditors).

Finally, I report the estimation results of the GC opinions model (Equation 4) in

Table 2.8, which consistently show that the estimated coefficients of RES_WRD are

positive and significant at less than 1% levels in both the full sample (Column 1) and the

subsample of severely financially distressed firms (Column 2). This result indicates that

the likelihood of issuing GC opinions is significantly associated with the unexplained

portion of 10-K disclosure volume, which potentially captures the level of audit effort to

identify circumstances that warrant a going-concern report.

2.6 Additional Analyses

To control for any year-specific events, such as the subprime financial crisis or the

implementation of the Dodd-Frank Act, I estimate the disclosure model by year to allow

the intercept and coefficients to vary by year. Table 2.9 reports the mean results of year

regressions for the disclosure model together with t-statistics calculated based on the

standard errors of the coefficient estimates. As expected, most variables (untabulated

results) are significant and in the predicted direction. Specifically, I consistently show that

the estimated coefficients of BIG4 are positively associated with 10-K length as measured

by the word count of the 10-K submission text file across the sample period, supporting

the argument that 10-K disclosure volume varies with auditor characteristics as captured

by auditor firm size.

2.7 Conclusion

In this study, I extend the auditor size literature by investigating whether the choice

of Big 4 auditors contributes to cross-sectional variations in 10-K disclosure volume. In

28

addition to the ample evidence that Big 4 auditors deliver higher quality, I document that

the perceived high audit quality of Big 4 auditors helps audit clients improve the

informativeness of their disclosures as measured by 10-K disclosure volume. I further

show that this relation is more pronounced in situations where the users of financial reports

potentially need more relevant information to understand the information conveyed in the

10-K reports. Together, these results indirectly address the controversial issue regarding

auditors’ responsibilities raised in the study by DeFond et al. (2017b) and support the

broader view of auditors’ responsibilities, which argues that the role of the auditor is not

limited to mere GAAP compliance.

Finally, because the auditors are responsible for examining firms’ financial

reporting and expressing an opinion on its fairness, I show that an abnormally high level

of disclosure volume is positively associated with higher audit fees and an increased

likelihood of GC opinions, thus indicating that a significant discretionary component of 10-

K disclosure volume induces external auditors to increase their audit effort.

2.8 Figure and tables

Figure 2.1 Residual disclosure of the 10-K reports

These graphs depict the mean residual disclosure of firms that use Big 4 auditors (BIG4 = 1) and non-Big 4 auditors (BIG4 = 0) by fiscal year.

Panel A The mean residuals from estimating the disclosure model using the word count (LNWORDS) as the dependent variable

Panel B The mean residuals from estimating the disclosure model using the gross file size of the complete 10-K submission text file (LNSIZEG) as the dependent variable

-0.030

-0.020 -0.019-0.021

-0.036 -0.035

-0.022 -0.024

-0.030 -0.030-0.025

0.008 0.008 0.0080.011

0.017 0.016

0.010 0.011 0.0120.014

0.011

-0.040

-0.030

-0.020

-0.010

0.000

0.010

0.020

2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014

Variable: RES_WRD

(1) BIG4 = 0 (2) BIG4 = 1

-0.082

-0.034

-0.015

0.007

-0.007

0.038

0.051

-0.017 -0.013 -0.017-0.025

0.026

0.0130.006

-0.005

0.003

-0.021-0.029

0.009 0.007 0.011 0.014

-0.100

-0.080

-0.060

-0.040

-0.020

0.000

0.020

0.040

0.060

2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014

Variable: RES_SG

(1) BIG4 = 0 (2) BIG4 = 1

Table 2.1 Descriptive Statistics

These tables report the summary statistics of variables used in the full sample and the propensity-

score matched samples from the fiscal year 2004 to 2014. All continuous variables are winsorized

at the 1st and 99th percentiles. *, **, *** indicate significance at the 0.10, 0.05 and 0.01 levels,

respectively, using two-tailed t-tests of differences in means.

Variable Definitions

a. Disclosure attributes:

LNWORDS = the natural logarithm of the number of word count in the 10-K complete

submission text file; LNSIZEG = the natural logarithm of the gross file size of the 10-K complete

submission text file.

b. Client Characteristics:

LNASSET = the natural logarithm of total assets (in millions) at the end of fiscal year; ATURN

= the ratio of sales to lagged total assets; CURRENT = the ratio of current assets to current

liabilities; LEVERAGE = the sum of short-term and long-term debt in year t, divided by total

assets; ROA = income before extraordinary items, scaled by average total assets.

c. Determinants of disclosure attributes:

DELTA_ROA = the annual change in ROA; DELTA_REV = the annual percentage change in

sales; MA = indicator variable equals to one if an audit’s client is engaged in a merger or

acquisition during the year, and zero otherwise; FY_RET = Raw annual return over the 12-

month fiscal period; SD_RETURN = the standard deviation of the monthly stock returns in the

prior fiscal year; SPI_DM = indicator variable equals to one if an audit’s client has any special

item during the year, and zero otherwise; CAP_LEASE = indicator variable equals to one if the

company reports a capital lease on its balance sheet, and zero otherwise; OP_LEASE =

indicator variable equals to one if the value of operating lease payments due in one year is

greater than 1 percent of total assets, and zero otherwise; RD = the amount of research and

development expense, scaled by lagged assets; INTANG = the unamortized value of

purchased intangible assets, scaled by lagged assets; SIZE = the natural logarithm of the firm’s

market value at the end of fiscal year; AGE = the natural logarithm of the number of years since

a firm’s first appearance in the COMPUSTAT annual files; MTB = the firm’s market value

divided by its book value; SPECIAL = the amount of special items, scaled by total assets; FCF

= the average operating cash flows scaled by total assets over the current and prior years;

DERIVATIVE = indicator variable equals to 1 if the company reports any current or

accumulated gains or losses on derivative transactions, and zero otherwise; LNBUSSEG = the

natural logarithm of one plus the number of business segments; LNGEOSEG = the natural

logarithm of one plus the number of geographic segments; SD_OIADP = the standard deviation

of the operating earnings in the last five fiscal years; DELAWARE = indicator variable equals

to one if an audit’s client is incorporated in Delaware, and zero otherwise; IPO = indicator

variable equals to one if an audit’s client is engaged in an initial public offering during the year,

and zero otherwise; SEO = indicator variable equals to one if an audit’s client is engaged in

any seasoned equity offering during the year, and zero otherwise; NMCOUNT = the natural

logarithm of number of non-missing items in Compustat annual files.

31

Table 2.1 – Continued

d. Other tested variables:

BIG4 = indicator variable equals to one if the firm’s auditor is a member of the Big 4 audit firms

(PwC, EY, KPMG and Deloitte) and zero otherwise; ADA_MJR = the absolute value of

discretionary accruals based on the modified Jones model (Dechow et al. 1995) controlling for

firm’s financial performance (Kothari et al. 2005); EFFSPRD = the difference between the bid-

ask midpoint and the actual transaction price divided by the bid-ask midpoint, following the

same approach as in Hendershott et al. (2011).


Panel A – The full sample

Non-Big 4 firms Big 4 firms Differences

in Means (N = 13,818) (N = 29,757)

Variable Mean Std. Dev. Mean Std. Dev. Diff. t-statistic

Disclosure Attributes LNWORDS 10.45 0.47 10.80 0.46 -0.35 -72.13 ***

LNSIZEG 14.78 1.25 15.20 1.23 -0.42 -31.83 ***

Client Characteristics LNASSET 5.08 1.61 7.16 1.62 -2.08 -113.33 ***

ATURN 0.89 0.94 0.96 0.93 -0.07 -7.17 ***

CURRENT 3.35 5.18 3.22 5.24 0.13 2.49 **

LEVERAGE 0.15 0.20 0.22 0.19 -0.06 -32.98 ***

ROA -0.02 0.19 0.02 0.18 -0.04 -20.23 ***

Determinants of Disclosure Attributes DELTA_ROA -0.31 3.38 -0.24 3.61 -0.06 -1.75 *

DELTA_REV 0.17 0.47 0.13 0.51 0.04 7.42 ***

MA 0.11 0.36 0.22 0.37 -0.11 -29.58 ***

FY_RET -0.13 0.77 -0.11 0.75 -0.03 -3.48 ***

SD_RETURN 0.13 0.07 0.11 0.07 0.02 26.01 ***

SPI_DM 0.46 0.50 0.68 0.50 -0.23 -45.05 ***

CAP_LEASE 0.43 0.50 0.53 0.50 -0.10 -19.04 ***

OP_LEASE 0.37 0.49 0.41 0.49 -0.04 -8.56 ***

RD 0.04 0.11 0.04 0.10 0.00 -3.69 ***

INTANG 0.12 0.23 0.18 0.22 -0.06 -27.79 ***

SIZE 4.46 1.28 6.96 1.29 -2.50 -161.30 ***

AGE 2.47 0.82 2.67 0.80 -0.20 -22.88 ***

MTB 1.78 1.58 1.85 1.61 -0.07 -4.29 ***

FCF 0.02 0.14 0.07 0.15 -0.05 -32.84 ***

DERIVATIVE 0.11 0.35 0.28 0.35 -0.17 -47.23 ***

LNBUSSEG 0.70 0.54 0.96 0.54 -0.26 -44.55 ***

LNGEOSEG 0.59 0.68 0.88 0.67 -0.30 -43.75 ***

SD_OIADP 0.12 0.25 0.08 0.28 0.04 15.76 ***

DELAWARE 0.41 0.50 0.61 0.50 -0.20 -39.30 ***

IPO 0.00 0.04 0.00 0.03 0.00 3.15 ***

SEO 0.07 0.28 0.09 0.28 -0.03 -10.48 ***

NMCOUNT 5.62 0.20 5.74 0.22 -0.11 -49.80 ***

Other Tested Variables ADA_MJR 0.07 0.07 0.05 0.06 0.03 30.42 ***

EFFSPRD 1.15 0.82 0.29 0.79 0.86 72.22 ***


Panel B – The Propensity-Score Matched Sample

Non-Big 4 firms Big 4 firms Differences

in Means (N = 6,576) (N = 6,576)

Variable Mean Std. Dev. Mean Std. Dev. Diff. t-statistic

Disclosure Attributes LNWORDS 10.57 0.47 10.61 0.46 -0.04 -5.51 ***

LNSIZEG 14.83 1.25 14.85 1.24 -0.02 -0.96 Client Characteristics LNASSET 5.57 1.61 5.58 1.63 -0.02 -0.64 ATURN 0.97 0.94 0.95 0.92 0.02 1.31 CURRENT 3.64 5.24 3.59 5.19 0.05 0.56 LEVERAGE 0.17 0.20 0.17 0.19 0.00 0.03 ROA -0.01 0.19 -0.02 0.18 0.00 1.01 Determinants of Disclosure Attributes DELTA_ROA -0.29 3.38 -0.31 3.61 0.01 0.21 DELTA_REV 0.17 0.47 0.17 0.51 0.00 0.04 MA 0.15 0.36 0.16 0.37 -0.01 -0.94 FY_RET -0.14 0.77 -0.15 0.75 0.01 0.69 SD_RETURN 0.12 0.07 0.12 0.07 0.00 -0.04 SPI_DM 0.55 0.50 0.55 0.50 0.00 -0.16 CAP_LEASE 0.50 0.50 0.51 0.50 -0.01 -0.66 OP_LEASE 0.42 0.49 0.42 0.49 -0.01 -0.60 RD 0.05 0.11 0.05 0.10 0.00 0.03 INTANG 0.14 0.23 0.14 0.22 0.00 -0.89 SIZE 5.25 1.28 5.24 1.29 0.01 0.47 AGE 2.50 0.82 2.49 0.80 0.01 0.81 MTB 1.88 1.58 1.85 1.61 0.03 1.01 FCF 0.04 0.14 0.03 0.15 0.00 1.31 DERIVATIVE 0.15 0.35 0.14 0.35 0.00 0.27 LNBUSSEG 0.82 0.54 0.82 0.54 0.00 0.47 LNGEOSEG 0.73 0.68 0.73 0.67 0.00 0.18 SD_OIADP 0.11 0.25 0.11 0.28 0.00 -0.03 DELAWARE 0.53 0.50 0.53 0.50 0.00 -0.05 IPO 0.00 0.04 0.00 0.03 0.00 0.77 SEO 0.09 0.28 0.08 0.28 0.00 0.59 NMCOUNT 5.68 0.20 5.68 0.22 0.00 0.36 Other Tested Variables ADA_MJR 0.06 0.07 0.06 0.06 0.01 5.29 ***

EFFSPRD 0.72 0.87 0.69 0.79 0.03 1.77 *

34

Table 2.2 Correlation Matrices

This table reports the Pearson (the upper diagonal) and the Spearman (the lower diagonal)

correlation coefficients with all bold values are significant at 0.01 levels (two-tailed p-values).

Variable definitions:

RES_WRD = Residual from estimating equation (1) with the number of words count in the 10-K

complete submission text file as the dependent variable. Other variables are defined as in Table

2.1.

[1] [2] [3] [4] [5] [6] [7] [8]

[1] LNWORDS 1.00 0.74 0.32 0.46 -0.08 -0.26 0.54 0.00 <.0001 <.0001 <.0001 <.0001 <.0001 <.0001 0.7552 43,575 43,575 43,575 34,573 40,262 43,341 43,516

[2] RES_WRD 0.70 1.00 0.05 0.00 0.02 -0.03 0.11 0.03 <.0001 <.0001 0.4956 0.0014 <.0001 <.0001 <.0001 43,575 43,575 43,575 34,573 40,262 43,341 43,516

[3] BIG4 0.32 0.04 1.00 0.58 -0.19 -0.44 0.63 -0.08 <.0001 <.0001 <.0001 <.0001 <.0001 <.0001 <.0001 43,575 43,575 43,575 34,573 40,262 43,341 43,516

[4] SIZE 0.46 0.00 0.59 1.00 -0.22 -0.60 0.81 -0.14 <.0001 0.3573 <.0001 <.0001 <.0001 <.0001 <.0001 43,575 43,575 43,575 34,573 40,262 43,341 43,516

[5] ADA_MJR -0.09 0.02 -0.17 -0.22 1.00 0.13 -0.22 0.12 <.0001 0.0001 <.0001 <.0001 <.0001 <.0001 <.0001 34,573 34,573 34,573 34,573 32,126 34,376 34,524

[6] EFFSPRD -0.37 0.00 -0.54 -0.89 0.19 1.00 -0.49 0.09 <.0001 0.634 <.0001 <.0001 <.0001 <.0001 <.0001 40,262 40,262 40,262 40,262 32,126 40,064 40,213

[7] LNAFEES 0.53 0.10 0.64 0.80 -0.20 -0.72 1.00 -0.08 <.0001 <.0001 <.0001 <.0001 <.0001 <.0001 <.0001 43,341 43,341 43,341 43,341 34,376 40,064 43,309

[8] GC 0.00 0.03 -0.08 -0.14 0.09 0.11 -0.08 1.00 0.7974 <.0001 <.0001 <.0001 <.0001 <.0001 <.0001

43,516 43,516 43,516 43,516 34,524 40,213 43,309

35

Table 2.3 Auditor choice and 10-K disclosure volume

This table reports the regression results of estimating the disclosure model (Equation 1) on both

the full sample (Column 1) and the PSM sample (Column 2). T-statistic is determined by clustered

standard errors at firm level. *, **, *** denote significance at the 0.10, 0.05, and 0.01 levels,

respectively.

(1) (2)

Full Sample PSM Sample

DV = LNWORDS DV = LNWORDS

Variables Coef. t-stat. Coef. t-stat.

Intercept 9.80 *** 67.38 10.20 *** 47.63

BIG4 0.07 *** 7.64 0.04 *** 4.09

DELTA_ROA 0.00 ** -2.47 0.00 0.17

DELTA_REV 0.01 1.23 0.00 0.36

MA 0.03 *** 5.21 0.02 * 1.82

FY_RET -0.04 *** -14.69 -0.05 *** -9.46

SD_RETURN 0.96 *** 18.57 1.00 *** 13.49

SPI_DM 0.10 *** 18.99 0.10 *** 12.58

CAP_LEASE 0.06 *** 7.27 0.06 *** 4.35

OP_LEASE 0.02 * 1.92 0.01 0.98

RD 0.32 *** 6.89 0.20 *** 2.94

INTANG -0.03 -1.58 0.01 0.23

SIZE 0.12 *** 43.50 0.12 *** 25.63

AGE -0.07 *** -17.48 -0.08 *** -13.10

MTB -0.05 *** -19.69 -0.05 *** -10.86

LEVERAGE 0.31 *** 14.36 0.37 *** 11.62

FCF -0.42 *** -14.81 -0.40 *** -10.09

DERIVATIVE 0.04 *** 5.79 0.05 *** 3.89

LNBUSSEG 0.05 *** 6.25 0.07 *** 5.13

LNGEOSEG 0.00 0.37 0.01 1.32

SD_OIADP 0.03 ** 2.30 0.03 ** 2.00

DELAWARE 0.04 *** 4.34 0.04 *** 3.71

IPO 0.10 * 1.65 0.13 1.10

SEO 0.04 *** 4.35 0.02 1.55

NMCOUNT 0.03 1.11 -0.03 -0.91

Fixed Effects Yes Yes

Observations 43,575 13,152

Adjusted R-squared 42.9% 37.7%

36

Table 2.4 Incremental effect of Big 4 auditors on 10-K disclosure volume

These tables report the benefit of enhanced disclosures provided by Big 4 auditors for audit clients with poorer accrual quality and those with higher

information asymmetry. The full sample (Panel A) and the PSM sample (Panel B) are partitioned into subsamples with low and high values of

ADA_MJR in columns (1) and (2) and subsamples with low and high values of EFFSPRD in columns (3) and (4), respectively. T-statistic is determined

by clustered standard errors at firm level. *, **, *** denote significance at the 0.10, 0.05, and 0.01 levels, respectively.

(1) (2) (3) (4)

Low ADA_MJR High ADA_MJR Low EFFSPRD High EFFSPRD

DV = LNWORDS DV = LNWORDS DV = LNWORDS DV = LNWORDS

Variables Coef. t-stat. Coef. t-stat. Coef. t-stat. Coef. t-stat.

Panel A – The full Sample

BIG4 0.06 *** 3.44 0.02 1.52 -0.03 -0.88 0.07 *** 6.19

ADA_MJR -0.30 -0.55 -0.05 -0.66 ADA_MJR*BIG4 0.28 0.44 0.34 *** 3.20 EFFSPRD -0.36 ** -2.13 -0.02 *** -4.64

EFFSPRD*BIG4 0.29 * 1.72 0.00 -0.29

Control Variables Yes Yes Yes Yes

Observations 17,103 17,470 20,128 20,134

Adjusted R-squared 43.1% 43.9% 31.8% 38.5%

Panel B – The PSM Sample

BIG4 0.04 * 1.74 -0.01 -0.39 0.01 0.20 0.05 *** 3.84

ADA_MJR -0.33 -0.49 -0.06 -0.58

ADA_MJR*BIG4 0.19 0.20 0.30 * 1.88

EFFSPRD 0.01 0.07 -0.03 *** -3.98

EFFSPRD*BIG4 0.00 -0.02 -0.01 -0.75

Control Variables Yes Yes Yes Yes

Observations 4,573 5,695 2,873 9,309


37


This table reports the summary statistics of variables used for the audit fee model (Panel A) and

the GC opinions model (Panel B) from the fiscal year 2004 to 2014. All continuous variables are

winsorized at the 1st and 99th percentiles.

Panel A – Variable definitions for audit fee model:

LNAFEES = the natural logarithm of audit fees; BIG4 = indicator variable equals to one if the firm’s

auditor is a member of the Big 4 audit firms (PwC, EY, KPMG and Deloitte) and zero otherwise;

LNASSET = the natural logarithm of total assets (in millions); CURRENT = the ratio of current

assets to current liabilities; INVREC = the ratio of total inventory and receivables to total assets;

LEVERAGE = the sum of short-term and long-term debt, divided by total assets; ROA = income

before extraordinary items, scaled by average total assets; INTL = indicator variable equals to one

if an audit’s client has international operations, and zero otherwise; MA = indicator variable equal

to one if an audit’s client is engaged in a merger or acquisition (as reported by SDC Platinum)

during the year, and zero otherwise; SPI_DM = indicator variable equal to one if an audit’s client

has a special item during the year, and zero otherwise; LNBUSSEG = the natural logarithm of one

plus the number of business segments; LOSS = indicator variable equals to one if income before

extraordinary items is negative in the current period, and zero otherwise; MTB = the firm’s market

value divided by its book value; BUSY = indicator variable equals to one if an audit’s client has a

year-end fall on December 31, and zero otherwise; TENURE = the number of years the company

has been audited by the same audit firm; IPO = indicator variable equals to one if an audit’s client

is engaged in an initial public offering during the year, and zero otherwise; SEO = indicator variable

equal to one if an audit’s client is engaged in a seasoned equity offering (as reported by SDC

Platinum) during the year, and zero otherwise; OPINION = indicator variable equal to one if an

audit’s client receives a modified audit opinion and zero otherwise, where a modified opinion is

defined as anything except a standard unqualified audit opinion coded as one by COMPUSTAT;

HIGHLIT = indicator variable equal to one for high litigation risk industries as defined in Francis et

al. (1994), and zero otherwise.

Panel B – Variable definitions for GC opinions model:

GC = indicator variable equals to one if a firm receives a going-concern report in a fiscal period,

and zero otherwise; BIG4 = indicator variable equal to one if the firm’s auditor is a member of the

Big 4 (PwC, EY, KPMG and Deloitte), and zero otherwise; LNASSET = the natural logarithm of

total assets (in millions); MTB = the firm’s market value divided by its book value; LEVERAGE =

the sum of short-term and long-term debt, divided by average total assets; CH_LEV = the difference

between current and prior year leverage; CFO = the operating cash flow, scaled by lagged total

assets; ALTMAN = the Altman z-score; PLOSS = indicator variable equals to one if income before

extraordinary items is negative in the prior period, and zero otherwise; HIGHLIT = indicator variable

equal to one for high litigation risk industries as defined in Francis et al. (1994), and zero otherwise;

TENURE = the number of years that an audit client has been audited by the same audit firm; ROA

= income before extraordinary items, scaled by average total assets; SD_OIADP = the standard

deviation of income before extraordinary items over the last five years;

38


Variables N Mean Std. Dev. Median Q1 Q3

Panel A – Variables used in the audit fee model

LNAFEES 30,918 13.68 1.26 13.71 12.85 14.51

BIG4 30,918 0.72 0.45 1.00 0.00 1.00

LNASSET 30,918 6.12 2.04 6.08 4.65 7.51

CURRENT 30,918 3.09 3.32 2.15 1.43 3.47

INVREC 30,918 0.26 0.19 0.23 0.11 0.37

LEVERAGE 30,918 0.18 0.19 0.14 0.00 0.30

ROA 30,918 0.00 0.18 0.04 -0.02 0.08

INTL 30,918 0.08 0.28 0.00 0.00 0.00

MA 30,918 0.20 0.40 0.00 0.00 0.00

SPI_DM 30,918 0.67 0.47 1.00 0.00 1.00

LNBUSSEG 30,918 1.00 0.51 0.69 0.69 1.39

LOSS 30,918 0.31 0.46 0.00 0.00 1.00

MTB 30,918 2.03 1.48 1.55 1.16 2.31

BUSY 30,918 0.68 0.47 1.00 0.00 1.00

TENURE 30,918 9.37 8.62 7.00 3.00 13.00

IPO 30,918 0.00 0.00 0.00 0.00 0.00

SEO 30,918 0.07 0.26 0.00 0.00 0.00

OPINION 30,918 0.33 0.47 0.00 0.00 1.00

HIGHLIT 30,918 0.36 0.48 0.00 0.00 1.00

Panel B – Variables used in the GC opinions model

GC 42,889 0.02 0.12 0.00 0.00 0.00

BIG4 42,889 0.69 0.46 1.00 0.00 1.00

LNASSET 42,889 6.49 2.09 6.51 5.05 7.89

MTB 42,889 1.84 1.43 1.34 1.03 2.03

LEVERAGE 42,889 0.20 0.19 0.15 0.02 0.32

CH_LEV 42,889 0.00 0.07 0.00 -0.02 0.02

CFO 42,889 0.06 0.16 0.07 0.01 0.13

ALTMAN 42,889 3.88 6.21 2.52 0.85 4.80

PLOSS 42,889 0.26 0.44 0.00 0.00 1.00

HIGHLIT 42,889 0.26 0.44 0.00 0.00 1.00

TENURE 42,889 8.53 8.14 6.00 3.00 12.00

ROA 42,889 0.00 0.16 0.03 0.00 0.07

SD_OIADP 42,889 0.09 0.23 0.04 0.02 0.08

39

Table 2.6 Residual disclosures and audit fees (Level specification)

This table presents regression results of the audit fee model (Equation 3) by each asset quintile group. *, **, *** denote significance at the 0.10, 0.05,

and 0.01 levels, respectively, using two-tailed tests. T-statistic is determined by clustered standard errors at firm level.

(1) (2) (3) (4) (5)

Quintile 1 (Small) Quintile 2 Quintile 3 Quintile 4 Quintile 5 (Large)

DV = LNAFEES DV = LNAFEES DV = LNAFEES DV = LNAFEES DV = LNAFEES

Variables Coef. t-stat. Coef. t-stat. Coef. t-stat. Coef. t-stat. Coef. t-stat.

Intercept 10.30 *** 94.37 10.02 *** 63.90 10.51 *** 48.07 10.79 *** 39.64 9.41 *** 35.55

RES_WRD 0.23 *** 9.84 0.22 *** 9.01 0.19 *** 7.34 0.20 *** 7.61 0.14 *** 5.19

BIG4 0.52 *** 19.66 0.44 *** 15.62 0.32 *** 9.50 0.18 *** 3.73 0.19 1.64

LNASSET 0.51 *** 36.31 0.51 *** 22.03 0.48 *** 17.54 0.44 *** 15.10 0.58 *** 33.56

CURRENT -0.02 *** -8.41 -0.01 *** -4.71 -0.02 *** -4.47 -0.03 *** -5.97 -0.05 *** -4.11

INVREC 0.27 *** 4.66 0.37 *** 4.00 0.65 *** 6.76 0.88 *** 6.58 1.25 *** 7.94

LEVERAGE -0.26 *** -4.20 -0.26 *** -3.73 0.00 0.00 -0.08 -1.00 -0.25 *** -2.59

ROA -0.31 *** -6.35 -0.35 *** -5.31 -0.21 ** -2.39 -0.06 -0.51 -0.53 *** -3.19

INTL 0.14 ** 2.35 0.10 1.32 0.01 0.09 0.14 ** 2.32 0.12 *** 2.83

MA 0.00 -0.17 0.05 *** 2.69 0.00 0.00 0.01 0.61 0.02 1.23

SPI_DM 0.12 *** 7.68 0.16 *** 9.18 0.15 *** 8.83 0.23 *** 8.28 0.18 *** 6.76

LNBUSSEG 0.05 * 1.91 0.03 0.93 0.05 * 1.88 0.04 * 1.68 0.04 * 1.78

LOSS 0.11 *** 5.12 0.11 *** 4.63 0.06 *** 2.69 0.12 *** 3.99 0.10 *** 2.86

MTB 0.01 ** 2.37 0.04 *** 4.19 0.00 -0.66 -0.02 ** -2.24 -0.02 -1.24

BUSY 0.03 1.43 0.07 ** 2.34 0.02 0.54 0.03 0.79 0.02 0.46

TENURE -0.01 *** -3.26 0.00 -0.65 0.00 0.07 0.00 0.21 0.00 0.19

IPO 0.00 0.00 0.00 0.00 0.00 SEO 0.02 0.76 0.06 ** 2.42 0.02 1.00 0.06 * 1.93 0.01 0.24

OPINION 0.08 *** 4.28 0.14 *** 6.55 0.13 *** 6.88 0.10 *** 4.90 0.09 *** 5.10

HIGHLIT 0.13 *** 3.39 0.06 1.15 0.18 *** 3.78 0.02 0.40 -0.04 -0.80

Fixed Effects Yes Yes Yes Yes Yes

Observations 6,429 6,180 6,168 5,678 6,463

Adj. R-squared 69.8% 55.5% 46.5% 44.4% 71.6%

40

Table 2.7 Residual disclosures and audit fees (Change specification)

This table presents regression results of the audit fee model (Equation 3) using the change

specification. The dependent variable is a year-to-year change in the level of audit fees (Δ

LNAFEES). *, **, *** denote significance at the 0.10, 0.05, and 0.01 levels, respectively, using two-

tailed tests. T-statistic is determined by clustered standard errors at firm level.

(1)

Full Sample

(2)

Non-Big 4 firms

(3)

Big 4 firms

DV = Δ LNAFEES DV = Δ LNAFEES DV = Δ LNAFEES

Variables Coef. t-stat. Coef. t-stat. Coef. t-stat.

Intercept 0.05 1.60 0.02 0.27 0.06 * 1.91

Δ RES_WRD 0.03 *** 5.82 0.03 * 1.89 0.03 *** 5.57

Δ LNASSETB 0.32 *** 27.48 0.31 *** 14.38 0.33 *** 23.94

Δ CURRENT -0.01 *** -5.05 0.00 * -1.70 -0.01 *** -5.67

Δ INVREC 0.12 *** 3.00 0.10 1.55 0.12 ** 2.32

Δ LEVERAGE 0.05 * 1.70 -0.01 -0.24 0.08 ** 2.45

Δ ROA -0.23 *** -10.66 -0.28 *** -7.10 -0.22 *** -8.16

Δ LNBUSSEG 0.01 1.17 0.02 0.96 0.00 0.33

Δ MTB 0.00 -0.82 -0.01 ** -2.08 0.00 0.99

Δ MA (0 to 1) 0.02 *** 3.96 0.02 * 1.74 0.02 *** 3.07

Δ MA (1 to 0) -0.01 -1.23 -0.01 -0.38 -0.01 -1.47

Δ SPI_DM (0 to 1) 0.03 *** 4.93 0.04 *** 3.41 0.03 *** 4.37

Δ SPI_DM (1 to 0) -0.02 *** -4.20 -0.02 * -1.74 -0.02 *** -3.50

Δ OPINION (0 to 1) 0.04 *** 4.47 0.05 *** 3.03 0.03 *** 3.60

Δ OPINION (1 to 0) 0.00 -0.12 -0.03 ** -2.44 0.00 0.03

Δ LOSS (0 to 1) 0.03 *** 4.16 0.00 -0.01 0.05 *** 4.85

Δ LOSS (1 to 0) 0.00 -0.02 -0.01 -0.77 0.00 0.44

Δ IPO (1 to 0) -0.08 -1.15 -0.18 * -1.95 0.03 0.35

Δ SEO (0 to 1) 0.03 *** 2.94 0.03 1.35 0.03 *** 2.66

Δ SEO (1 to 0) -0.05 *** -6.03 -0.04 * -1.85 -0.05 *** -5.51

Δ BIG4 (0 to 1) 0.30 *** 6.34

Δ BIG4 (1 to 0) -0.36 *** -16.15

Fixed Effects Yes Yes Yes

Observations 24,894 6,708 18,186

Adjusted R-Squared 16.5% 11.5% 15.2%

41

Table 2.8 Residual disclosures and going-concern opinions

This table presents logistic regression results of the going-concern opinions model (Equation 4). *, **, ***

denote significance at the 0.10, 0.05, and 0.01 levels, respectively, using two-tailed tests. P-values are

based on Z-statistics, which are clusted by client firm.

(1)

Full Sample

DV = GC

(2)

Sample of Severely

Financially Distressed

firms

DV = GC

Variables Coef. p-value Coef. p-value

Intercept -4.52 0.89 -3.15 0.94

RES_WRD 1.09 *** <.0001 0.99 *** <.0001

BIG4 -0.10 0.40 -0.30 ** 0.03

LNASSET -0.49 *** <.0001 -0.34 *** <.0001

MTB -0.13 *** <.0001 -0.09 *** 0.00

LEVERAGE 2.56 *** <.0001 2.39 *** <.0001

CH_LEV -0.45 0.28 -0.59 0.22

CFO -1.00 *** 0.00 -0.48 0.21

ALTMAN -0.04 *** <.0001 -0.04 *** <.0001

PLOSS 1.13 *** <.0001 0.36 * 0.07

HIGHLIT 0.15 0.45 0.15 0.49

TENURE 0.00 0.97 -0.01 0.59

ROA -3.54 *** <.0001 -3.45 *** <.0001

SD_OIADP 0.31 *** 0.00 0.27 *** 0.01



Pseudo R-Squared 37.1% 29.7%

42

Table 2.9 Auditor choice and 10-K disclosure volume (by each fiscal year)

This table reports the mean results of year regressions for the disclosure model together with t-statistics calculated based on the standard

errors of the coefficient estimates. *, **, *** indicate significance at the 0.10, 0.05 and 0.01 levels, respectively.

Mean Coef. t-stat. 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 Std. Err

BIG4 0.07 3.89 *** 0.06 0.05 0.05 0.06 0.10 0.09 0.06 0.06 0.07 0.08 0.07 0.02

Control Variables Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes

Fixed effects Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes

0.060.05

0.05

0.06

0.100.09

0.06

0.060.07

0.08

0.07

0.00

0.02

0.04

0.06

0.08

0.10

0.12

2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014

BIG

4 C

oe

ffic

ien

t

Fiscal Year

43

Chapter 3. The difference among the Big 4 firms: Further evidence from audit pricing

3.1 Introduction

While the Big 4 firms are recognized to be of superior quality (e.g., DeAngelo 1981)

and have been treated as a homogeneous set of firms with regard to brand name

reputation, audit firm characteristics were previously found to be associated with either an

audit fee premium or discount. For example, in the classic study by Simunic (1980), an

audit fee premium was observed for the clients of PW, indicating that there is price

competition in the audit market with a differentiated product for PW. In addition, Peat

Marwick17 (later KPMG) was considered a price cutter and earned below average top-tier

premiums (e.g., Bernstein 1978; Moizer 1997). However, results in the 1990s and early

2000s did not show a significant premium for any audit firm, and research on auditor

industry specialization subsequently examined the assumption that the development of

industry-specialized expertise facilitates the delivery of high-quality audits, thereby

increasing auditor reputation and earning such auditors a significant fee premium.

In this study, I focus on audit pricing differences within the Big 4 firms, which are

the dominant suppliers of audit services to large corporations in the U.S. market, during

the period 2004–2014. In response to a call for more studies on the PwC premium by Hay

(2013), I raise the empirical question whether, in addition to the general Big 4 audit fee

premium, individual differences can be detected within the Big 4 firms in the U.S. market.

I argue that if Big 4 firms are equally treated as a homogeneous set of high-quality firms

and earn significant fee premiums because of their brand name reputation as a group,

there should be no significant audit pricing differences among them. However, I find that

17 KPMG was later formed in 1987 from a merger between Klynveld Main Goerdeler (KMG) and

Peat Marwick International.

44

not only has PwC maintained its leadership position18 as the market share leader in the

U.S. market, the firm also earns an above-average Big 4 premium over the other Big 4

firms. I also find that KPMG has been identified as an industry specialist in the fewest

number of industries and earns a below-average Big 4 premium in the audit fee model.

According to the GAO (2008), large companies primarily choose Big 4 audit firms

as their external auditors due to their technical capabilities and industry-specialized

expertise. Thus, each audit firm has to compete intensively with other competitors to meet

the demand for high audit quality through the development of brand name reputation and

industry specialization (e.g., Craswell et al. 1995). For example, PwC (2012) states:

“The PwC Network has invested hundreds of millions of dollars in the

development and roll-out of a new proprietary global audit software tool that

will help further improve quality on even our most complex multinational

clients; trains over 60,000 audit professionals annually at a significant cost;

and maintains global systems to provide controls over audit quality and

compliance with many independence requirements - both those imposed

externally and those required by PwC internal policy.” (page 3)

Therefore, industry-specialized expertise is likely to enable differentiation of the

audit services within the Big 4 firms, improve audit quality, and result in higher audit fees.

Although the concept of auditor industry specialization has been extensively

examined in the auditing literature, there is no consensus on how to empirically measure

specialization. Because the level of specialization of audit firms is unobservable, several

indirect proxies have been introduced to capture the complexity of the auditor industry

specialization concept, including market share-based measures (e.g., Craswell et al.

1995; Zeff and Fossum 1967), the portfolio proportion of clients (e.g., Kwon 1996), and

the use of weighted market shares (e.g., Neal and Riley 2004). Audousset-Coulier et al.

(2015) recently reveal that a variation of these approaches yields inconsistent

18 Panel A of Table 3.2 shows that PwC earned 36% of the total audit fees in fiscal year 2004,

followed by EY (24%), KPMG (19%), Deloitte (17%) and other audit firms (4%), The overall

rankings and market shares of audit firms for other fiscal years are qualitatively similar to those in

fiscal year 2004.

45

classifications of audit firms as industry specialists and subsequently leads to inconsistent

inferences drawn from audit pricing and earnings quality models.

Regardless of the approach used to define industry specialization, all these proxies

define an industry specialist as “an audit firm” that has differentiated itself from its

competitors in the audit market. This then raises the empirical question whether there

exists a possible confounding effect of an individual audit firm’s competencies with the

effect of industry specialization on audit pricing because in addition to the evidence on

auditor size and auditor industry specialization, the individual audit firm’s competencies

have long been argued to play an important role in delivering high audit quality.

More importantly, the industry specialization designation results reveal that when

the within-industry market share approach is used, PwC is more likely to be designated

an industry specialist in the highest number of industries (e.g., Audousset-Coulier et al.

2015; Cahan et al. 2011; Knechel et al. 2007; Li et al. 2010), followed by EY, Deloitte and

KPMG. This then raises questions regarding the confounding effect of an individual audit

firm on the estimated effect of auditor industry specialization in the audit fee model. I test

whether a lack of proper control over the PwC fee premium has a significant effect on the

inferences drawn from the audit pricing model by examining the differential effect of the

industry specialization premium on the audit fees charged by PwC specialists and other

specialists. In particular, the industry specialization premium is consistently observed for

the group of PwC specialists but not for the group of other (non-PwC) specialists.

In additional analyses, I estimate the audit fee model partitioned by asset size

quintiles to prevent the model’s residuals from being correlated with client size and

consistently find that the evidence of a PwC fee premium and its confounding effect is

robust to client size subsample regressions. Sensitivity tests using two alternative

measures of industry specialization at the metropolitan statistical area (hereafter MSA)

city level (e.g., Ferguson et al. 2003; Francis et al. 2005b) and a measure of unexplained

audit fees (hereafter UAF) further confirm that individual differences in audit fees can be

detected within the Big 4 audit firms.

My study makes several contributions to the audit fee literature. First, while Big 4

firms tend to be homogeneously identified through the Big 4 grouping, I document that

PwC earns an above-average Big 4 fee premium, indicating that there are pricing

46

differences at the inter-audit firm level. Furthermore, the evidence of a PwC fee premium

suggests that individual brand name reputation also plays an important role in the audit

market and partially explains the existing differences within Big 4 firms. Second, I provide

an explanation for the inconsistent results regarding the impact of industry specialization

on audit fees by revealing the differential effect of the industry specialization premium on

the audit fees charged by PwC specialists and other specialists. In particular, I document

that the positive relation between audit fees and industry specialization is exaggerated by

the confounding effect of the PwC fee premium. Finally, I emphasize the relevance of

distinguishing among individual Big 4 firms and further argue that treating all Big 4 firms

as a homogeneous set of firms potentially creates an important omitted variable in certain

audit fee studies (e.g., Hribar et al. 2014).

The remainder of this study is organized as follows. In section 2, I review the

related literature and develop the hypotheses. In section 3, I present my research design,

and I report the sample selection and descriptive statistics in section 4. Sections 5 and 6

present empirical results and additional tests, respectively. Finally, section 7 concludes

the study.



While the mainstream audit literature focuses on the use of Big 4 membership19 to

proxy for high audit quality, DeFond and Zhang (2014) argue that larger audit firms not

only have greater incentives (e.g., reputation loss from discovery of a lower quality audit)

to supply high-quality audits but also have greater competencies (e.g., audit inputs,

industry expertise) to deliver high-quality audits. Thus, the research on auditor

competencies has begun to tease out the effects of competency on audit quality using

19 The indicator for Big 4 has been shown to be associated with almost all other audit quality proxies,

including a lower incidence of accounting fraud (e.g., Lennox and Pittman 2010), a lower incidence

of accounting restatements (e.g., Eshleman and Guo 2014), lower discretionary accruals (e.g.,

Becker et al. 1998; Francis et al. 1999b), higher audit fees (e.g., Craswell et al. 1995; Hay et al.

2006), increased ERCs (e.g., Teoh and Wong 1993), improved analyst earnings forecasts (e.g.,

Behn et al. 2008), and a lower cost of debt and equity (e.g., Khurana and Raman 2004).

47

several auditor characteristics, including auditor industry specialization and audit firm

competencies.

Evidence from auditor industry specialization

The concept of auditor industry specialization was introduced in the early literature

to examine quality variation at the inter-audit firm level. The supply of quality-differentiated

audits can be motivated by the agency and contracting theory (e.g., Jensen and Meckling

1976; Watts and Zimmerman 1983) as service differentials that explain both brand name

and industry specialization reputation as a function of increasing agency costs. Craswell

et al. (1995) argue that quality-differentiated audits based on industry expertise are

economically demanded and drive audit firm investments in the development of industry

specialization through technology and expertise. More importantly, because the

development of such expertise is argued to be costly, industry specialization is presumably

associated with higher audit fees (e.g., Ferguson et al. 2003; Francis et al. 2005b; Carson

2009) and higher quality audit services (e.g., Balsam et al. 2003; Romanus et al. 2008;

Lim and Tan 2010).

The classification of industry expertise is discussed in the early study by Zeff and

Fossum (1967) as well as Palmrose (1986) based on the market share of each audit firm

in each industrial category. The market leader is then designated an industry specialist

based on the assumption that industry specialists are expected to provide higher audit

quality, attract more clients and ultimately have larger market shares. Because information

on audit fees was not publicly available before the 2000s, the choice of variables used for

the calculation of auditor market share has not been consistent in the literature. For

example, some researchers use either client size or the number of audit clients as an

alternative proxy to calculate the market share of each audit firm.

An alternative approach to identify industry specialists is the client portfolio-based

approach (e.g., Kwon 1996; Neal and Riley 2004), which is based on the assumption that

industries in which a given audit firm holds the largest portfolio share reflect industries to

which the audit firm has allocated higher-than-average resources and developed industry-

specific knowledge. This approach captures the aggregate distribution of audit services

across various industries for each audit firm and defines industries in which an audit firm

is considered an industry specialist as those that constitute its three largest portfolio

48

shares. Neal and Riley (2004) also introduce the weighted market share cut-off, which

captures the complementary effects of the market share and portfolio approaches.

However, while the impact of industry expertise on either audit pricing or audit

quality has been extensively examined in the literature, the results of studies on industry

specialization assignments suggest that PwC has managed to differentiate itself from its

competitors with regard to its within-industry market share position. For example, Knechel

et al. (2007) document that the identification of industry specialists is highly stable, and

PwC dominates in 23 out of 40 industries, followed by EY (14 industries), Deloitte (10

industries) and KPMG (5 industries). Li et al. (2010) also document that PwC has the

highest number of joint national and city industry leaders in each of the six years in their

sample. Cahan et al. (2011) use the official websites of Big 4 firms to develop a list of key

industries and consistently document that the self-proclaimed areas of industry experts

are stable and consistent with the number of industries in which the audit firms qualify as

industry specialists based on the market share approach.

Evidence from audit firm competencies

The classic study by Simunic (1980) provided early evidence that there is a

significantly positive coefficient of PW in the audit pricing model, suggesting that an

indicator of PW captures some unique characteristics to its audit clients and results in a

significant fee premium relative to other competitors. Moizer (1997) shows that in addition

to the evidence of a top-tier audit fee premium, PW and Peat Marwick earn audit fee

premiums and discounts in the U.S. market, respectively. Another dimension of the quality-

differentiated characteristic of large audit firms documented by Dunn and Mayhew (2004)

is that Coopers and Lybrand LLP (later PwC) and a group of PwC partners provided

additional services beyond the standard audit through the implementation of enhanced

financial disclosure practices.

While the results in the 1990s and early 2000s did not show a significant premium

for any audit firm, Hay et al. (2006) and Hay (2013) argue that there is more evidence that

PwC received a significant fee premium in international settings (e.g., Firth and Lau 2004;

Pong and Burnett 2006). Ferguson and Scott (2014) also document that the Australian

market over the period 2002–2004 was dominated by the three largest audit firms – PwC,

49

EY, and KPMG – and that the evidence of an audit fee premium among these firms was

driven by a robust PwC fee premium.


The nature of the competition among Big 4 audit firms has been a concern for

regulators, particularly when the audit market is highly concentrated among the four

largest firms. The survey results of the GAO (2008) reveal that the current level of

concentration does not adversely affect choice, audit prices and audit quality. Additionally,

PwC (2012) argues that there is already intense competition among the large audit firms

to develop their own networks and intensively invest in industry expertise, methodologies

and talented people to meet the market’s demands for high-quality audits and effectively

serve the capital markets at competitive prices. This is consistent with the statements

explicitly stated on the Big 4 firms’ official websites:

“At PwC, we’re doing just that—investing in leading-edge technology,

significant process improvements, and leadership and performance

development for our people.”

“With a common, consistent strategy and structure, we [Ernst & Young]

serve our global and local clients with the same intensive focus on quality.”

“Applying Lean methodologies to the financial statement audit in order to

help enhance audit quality and increase value is a uniquely KPMG

innovation.”

“Deloitte’s Assurance services deliver advanced-level insights across all

industries – this uniquely positions us to understand the challenges and

opportunities facing your business.”

Because there is substantial evidence of a PwC fee premium together with its

leadership position in terms of either aggregated audit revenues or aggregated client asset

size in the U.S. market, I consider the possibility that there will be individual differences in

audit pricing among these Big 4 audit firms. As reported in 3.8 Figures and tables

50

Figure 3.1, the identification of a PwC brand reputation is supported by the results

of Vault’s annual accounting survey, which has consistently ranked PwC as the most

prestigious audit firm20 from the year 2000 to 2016. Additionally, since 2007, the Brand

Finance Index21 has ranked PwC among both the top 100 global brands and the world’s

10 most powerful brands based on Brand Finance’s Brand Strength Index (BSI), with an

estimated brand value of $18.5 billion in 2017, as illustrated in Figure 3.2.

Thus, I hypothesize that the individual firm reputational effect, particularly the PwC

effect, will be associated with above-average fee levels, thereby supporting the early

evidence of a PW effect provided by Simunic (1980). The first hypothesis is therefore

stated as follows:

H1: PwC earns a significant audit fee premium compared to other Big 4 firms.

Next, according to the breakdown of industry specialists by audit firm and fiscal

year, while Big 4 firms are designated industry specialists, I notice that PwC is designated

an industry specialist most often across industries and years, followed by EY, Deloitte and

KPMG (e.g., Cahan et al. 2011; Knechel et al. 2007; Li et al. 2010). The prevalence of

PwC across most auditor industry specialization measures raises an interesting question

regarding whether the positive association between audit fees and industry specialization

is confounded by the PwC fee premium.

Specifically, I argue that the confounding effect will result in a higher fee premium

charged by PwC specialists and a lower fee premium (or no relation) charged by other

(non-PwC) specialists. In other words, the evidence of industry expertise merely reflects

the individual firm reputational effect, primarily the PwC premium effect, in the audit fee

model. The second hypothesis is therefore stated as follows:

20 Vault’s annual accounting survey is conducted by asking thousands of accounting professionals

at the top audit firms to rate their competitor firms in terms of prestige and to provide a few words

describing their perception of those competitor firms.

21 Founded in 1996, Brand Finance is the world’s leading independent branded business valuation

and strategy consultancy. The assessment measures a range of metrics, including brand

awareness, satisfaction and recommendations, financial performance and internal investment,

market share and revenues. It also examines corporate responsibility, governance, and the views

of internal and external stakeholders. Organizations are then judged relative to their competitors.

51

H2a: PwC specialists receive higher audit fees than non-specialists.

H2b: Other (non-PwC) specialists receive higher audit fees than non-specialists.

H2c: PwC specialists receive higher audit fees than other (non-PwC) specialists.

3.3 Research Design

To test the first hypothesis, I investigate whether there are cross-sectional

differences in audit pricing detected within the Big 4 firms by breaking down the indicator

variable of Big 4 firms into four individual firm indicator variables (PwC, EY, KPMG and

DELT). Based on existing audit fee studies (e.g., Simunic 1980; Hay et al. 2006; Hay

2013), I estimate the following empirical model:

𝐋𝐍𝐀𝐅𝐄𝐄𝐒𝐭 = α0 + α1𝐋𝐍𝐀𝐒𝐒𝐄𝐓𝐭 + α2𝐂𝐔𝐑𝐑𝐄𝐍𝐓𝐭 + α3𝐈𝐍𝐕𝐑𝐄𝐂𝐭 + α4𝐋𝐄𝐕𝐄𝐑𝐀𝐆𝐄𝐭

+ α5𝐑𝐎𝐀𝐭 + α6𝐈𝐍𝐓𝐋𝐭 + α7𝐌𝐀𝐭 + α8𝐒𝐏𝐈_𝐃𝐌𝐭 + α9𝐋𝐍𝐁𝐔𝐒𝐒𝐄𝐆𝐭

+ α10𝐋𝐎𝐒𝐒𝐭 + α11𝐌𝐓𝐁𝐭 + α12𝐁𝐔𝐒𝐘𝐭 + α13𝐓𝐄𝐍𝐔𝐑𝐄𝐭 + α14𝐈𝐏𝐎𝐭

+ α15𝐒𝐄𝐎𝐭 + α16𝐎𝐏𝐈𝐍𝐈𝐎𝐍𝐭 + α17𝐇𝐈𝐆𝐇𝐋𝐈𝐓𝐭 + α18𝐏𝐰𝐂𝐭 + α19𝐄𝐘𝐭

+ α20𝐊𝐏𝐌𝐆𝐭 + α21𝐃𝐄𝐋𝐓𝐭 + 𝐘𝐞𝐚𝐫 𝐚𝐧𝐝 𝐈𝐧𝐝𝐮𝐬𝐭𝐫𝐲 𝐅𝐢𝐱𝐞𝐝 𝐄𝐟𝐟𝐞𝐜𝐭𝐬 + ε𝐭


H1 predicts that the differences between the estimated coefficients of PwC and the

other Big 4 firms will be positive and significant. In other words, I expect α18 to be

significantly larger than α19, α20 and α21. Additionally, to focus on the pricing differences

within the Big 4 firms, I exclude companies that are audited by non-Big 4 firms. I then

exclude an indicator variable of PwC to avoid perfect multicollinearity in the regression

model and re-run the audit fee model using the sample of Big 4 firms. H1 also predicts

that the coefficients of EY, KPMG and DELT will be negative and significant.

To test the second hypothesis, I investigate whether the evidence of auditor

industry specialization (SPEC) largely reflects the effect of the PwC fee premium.

First, I use audit fees to calculate the auditor market share within an industry and

take the within-industry market share approach, in which an audit firm is considered an

industry specialist if the individual audit firm holds a significant portion of the market share

in that industry. Specifically, I adopt two widely used assignment approaches to designate

industry specialists – dominant market share (AF_DOM) and market share cut-off at 30

52

percent (AF_30). The dominant market share approach includes the market leader in each

year and industry (based on two-digit SIC code) together with the firm with the second

highest market share in industries where there is not a major difference (less than 10

percent) between the market shares of the leader and the second-highest firm. For the

market share cut-off approach, I require the audit firm to earn audit fee revenue from an

industry that amounts to more than 30 percent22 of the total audit fees charged to all firms

in that particular industry. Finally, because the choice of variables used for the calculation

of market share is not consistent in the literature, I also include AT_DOM and AT_30 as

alternative proxies of industry specialization using client size to measure the market share

within each industry.

Second, I use two interaction terms to examine the differential effect of industry

specialization on the audit fees charged by PwC specialists and other (non-PwC)

specialists. SPEC – Others is the interaction between SPEC and an indicator variable that

equals one if the incumbent firm is not PwC and zero otherwise. SPEC – PwC is an

interaction term between SPEC and PwC. I predict that if industry specialization results in

higher audit fees, the estimated coefficients of both SPEC – PwC (H2a) and SPEC –

Others (H2b) will be positive and significant. Otherwise, the difference between the

estimated coefficients of SPEC – PwC and SPEC – Others in the audit fee model (H2c)

would be suggestive of either a PwC fee premium or a confounding effect of the PwC

premium on the industry specialization premium effect.



My sample includes all U.S.-listed firms with available datasets from the Audit

Analytics database from 2004 through 2014. I then merge these data with the Compustat

fundamental annual files to obtain necessary financial data for all firm-year observations.

I exclude all observations related to firms in the financial (between SIC 6000 and 6999)

and utility (between SIC 4900 and 4949) industries because the audit fee model for these

22 As is common in the industry specialization literature, I assume that each industry is

evenly split among the Big 4 audit firms (1/4 = 25%) and that industry specialists are those that

have a market share that is 20 percent greater than the calculated average of 25 percent.

53

firms is different from other industries (e.g., Fields et al. 2004; Hay et al. 2006). I also

exclude firms with missing values and firms with total assets of less than $1 million. These

sample selection procedures yield a final sample of 39,985 firm-year observations,

including 14,147 for clients of non-Big 4 firms (35.4 percent) and 25,838 for clients of Big

4 firms (64.6 percent). Finally, I winsorize all observations that fall in the top and bottom 1

percent of the distribution for each non-discrete variable to mitigate potential problems of

outliers.


Table 3.1 reports the descriptive statistics for all variables used in the audit fee

model during the 2004 to 2014 period.

To provide detailed insight into the rankings of each major audit firm across the

sample period, I report the number of audit clients, the aggregated client assets, and the

aggregated audit revenues for each of the Big 4 firms and non-Big 4 firms for the fiscal

years 2004, 2009 and 2014 in Panel A of Table 3.2. Consistent with Kwon (1996), I find

that EY led the rankings for the number of clients, with 22%, 21% and 21% of the overall

number of audits performed in 2004, 2009 and 2014, respectively. However, when the

aggregated client assets and audit revenues are considered as the rankings criteria, I find

that PwC placed first throughout almost all sample periods. This indicates that the average

size of PwC’s clients is relatively larger than the average size of the clients of the other

Big 4 audit firms. Panel B provides detailed information on the sample composition by

client size quintile for the fiscal years 2004, 2009 and 2014. The results consistently show

that the PwC concentration ratio23 increases with the size of audit clients.

23 The PwC concentration ratio is the percentage of total number of audit clients, the aggregated

client assets, and the aggregated audit revenues, earned by PwC.

54


Table 3.3 reports the regression results24 used to test Hypothesis 1. Instead of

treating Big 4 firms as a homogeneous set of firms, I break down the indicator variable of

Big 4 firms into four individual firm indicator variables. As reported in Column (1), the

estimated coefficients on PwC (0.53), EY (0.45), KPMG (0.41) and DELT (0.45) are

positive and significant at less than 1% levels. I also find that the differences between the

coefficients of PwC and the other Big 4 firms are positive and significant at less than 1%

levels, indicating that the magnitude of the PwC coefficient is significantly larger than those

of the other Big 4 firms. Additionally, I find that the estimated coefficient of KPMG is

significantly smaller than those of the other Big 4 firms, supporting the early evidence of

Bernstein (1978), who argued that Peat Marwick was a price cutter in the U.S. market.

Column (2) reports the regression results for the sample of Big 4 firms, and the

results consistently show that the estimated coefficients on EY (-0.08), KPMG (-0.13) and

DELT (-0.08) are negative and significant at less than 1% levels. Together, these results

indicate that on average, PwC earns a significant audit fee premium relative to the other

Big 4 firms. Similarly, when I estimate the regression model while excluding KPMG in

Column (3), the estimated coefficients on PwC (0.13), EY (0.04) and DELT (0.05) are

positive and significant, indicating that a below average premium is observed for KPMG.

Given that client size is the factor that is most closely related to auditor selection

(e.g., Lawrence et al. 2011), I divide each year and industry sample into quintiles based

on total client assets and report the results using the same format as in Table 3.3.

As reported in Panel A of Table 3.4, while the differences between the coefficients

on PwC and either EY or DELT exhibit weaker results across almost all asset size

quintiles, I find that the magnitudes of the PwC coefficients are significantly larger than

those of KPMG across all asset size quintiles. Alternatively, the estimated results obtained

from the sample of Big 4 firms in Panel B are qualitatively similar to those reported in Panel

24 The variance inflation factor of each independent variable is estimated to detect severe

multicollinearity problems in the audit pricing model. In untabulated results, there is no sign of

severe multicollinearity problems.

55

A. Collectively, these results suggest that there is clear evidence of variations in audit

pricing across the Big 4 audit firms.

To examine a possible confounding effect of an individual audit firm’s

competencies with the effect of industry specialization on audit pricing, I first identify

industry specialists in each industry according to the four different assignment methods

and measurement variables: the dominant market share (AF_DOM and AT_DOM) and

the market share cut-off at 30 percent (AF_30 and AT_30). As I have argued, the presence

of a PwC fee premium might be a potential confounder that correlates both audit fees

(dependent variable) and auditor industry specialization25 (independent variable) in the

audit fee model. Panel A of Table 3.5 reveals that the popular measures of industry

specialization are highly concentrated in Big 4 audit firms, particularly for PwC and EY,

throughout the sample period. Again, I find that KPMG is identified as an industry specialist

in the fewest industries across four industry specialist measures.

Next, I report both industry specialists’ and PwC specialists’ concentration ratios

for different client size categories in Panel B of Table 3.5. The results show that larger

audit clients are more likely to choose an audit firm with industry-specialized expertise.

This result occurs because client size potentially captures various aspects of a firm’s

operational and business environment. I also find that the PwC specialists’ concentration

ratio is relatively high across all size categories and sample periods, suggesting that the

relationship between audit fees and auditor industry specialization may be exaggerated

by the confounding effect of the PwC fee premium.

The regression results used to test Hypothesis 2 are separately reported in Panel

A of Table 3.6 for each measure of auditor industry specialization: AF_DOM (Column 1),

AF_30 (Column 2), AT_DOM (Column 3) and AT_30 (Column 4). Regardless of the

measures used to identify industry specialists, I find that the estimated coefficients on

SPEC - PwC are positive and significant at less than 1% levels. On the other hand, the

estimated coefficients on SPEC - Others are positive and significant at 5% levels only in

Columns (1) and (2). More importantly, I find that the magnitude of the SPEC - PwC

25 In untabulated results, I find that the Spearman correlation coefficients between an indicator

variable of PWC and various industry specialization measures range from 0.35 to 0.45 and are

significant at less than 1% levels.

56

coefficient is significantly larger than those of SPEC – Others across various measures of

industry specialists in the audit fee model, thus supporting the evidence of a PwC fee

premium in the audit market.

Additionally, I report the regression results by asset size quintile using the same

format as in Panel A, and the results consistently show that the estimated coefficients on

SPEC - PwC are positive and significant across various measures of industry specialists

and asset size quintiles. On the other hand, I find marginal results or even a negative

relationship (fee discount) in some instances regarding the impact of other (non-PwC)

specialists on audit fees, which raises more doubts about the validity of the industry

specialization fee premium. Once again, the differences between the estimated

coefficients of SPEC – PwC and SPEC – Others support the evidence of a PwC fee

premium in the audit market.


I perform several sensitivity tests to determine the sensitivity of my main results

regarding a PwC fee premium to different specifications.

3.6.1 Analysis using unexplained audit fees

I follow the methodology introduced by Hribar et al. (2014) and obtain a measure

of unexplained audit fees (UAF) by regressing the natural logarithm of audit fees on a set

of explanatory variables, including an indicator variable of Big 4 firms, in the audit fee

model by year and asset size decile.

Panel A of

57

Table 3.7 presents the mean UAF values separately reported by each fiscal year

end across all major groups of audit firms. While the mean UAF values of companies

audited by other (non-PwC) audit firms are either negative or nonsignificant across the

sample period, I document that the mean UAF values of companies audited by PwC are

positive and significantly different from zero, thus supporting the existing variation in audit

pricing at the inter-audit firm level.

In Panel B, I report univariate tests of the differences in means of unexplained audit

fees using an indicator variable of PwC together with an indicator variable for the dominant

market leader26 (AF_DOM) to partition the full sample into four subsamples: (1) companies

that are audited by PwC specialists, (2) companies that are audited by specialists but not

by PwC, (3) companies that are audited by PwC but not by specialists and (4) companies

that are audited by neither PwC nor specialists. Next, I compare the mean UAF values

between each pair of subsamples with a presumption that a strong positive UAF infers an

ability of audit firms to earn significant fee premiums.

Consistent with Huang et al. (2007), I find that the mean UAF values for the group

of industry specialists (Subsample A) are significantly larger than those for the group of

non-industry specialists (Subsample B) only in some periods, indicating that an industry-

specific premium is sensitive to the time period. More interestingly, when I divide the

sample of industry specialists into clients of PwC specialists (Subsample A1) and other

specialists (Subsample A2), the mean UAF values for the group of PwC specialists are

significantly larger than those of other specialists across almost all fiscal years. On the

other hand, I find no statistically significant differences in mean UAF values between

clients of PwC specialists (Subsample A1) and clients of PwC non-specialists (Subsample

B1), thus supporting the evidence of a PwC fee premium, but not an industry-specific fee

premium.

Finally, I create a matched-pair sample based on the asset size (LNASSET) within

a maximum distance of 3 percent between two auditor groups: PwC and other (non-PwC)

Big 4 firms27. I require matched-pair observations to possess common attributes, including

26 In untabulated results, I find that the inferences are unchanged when I use alternative measures

of industry specialization, including AT_DOM, AF_30 and AT_30.

27 I take this approach because I mainly focus on the existing variation within Big 4 audit firms.

58

complexity of operations (INTL), financial performance (LOSS), firm-specific event (MA)

in the same fiscal year and two-digit SIC codes.

Next, I compare the mean UAF values between each pair of subsamples and test

whether the PwC fee premium can be attributed to a specific client’s characteristics. Panel

C of

59

Table 3.7 reports the results for each matched sample using t-tests and Wilcoxon

signed ranked tests. Overall, I document that the mean UAF values are significantly larger

for PwC clients at less than a 1% level and that the mean UAF values are significantly

larger for the clients of PwC specialists, regardless of the measures used to identify

industry specialists, at less than 1% levels.

In sum, by focusing on the variations in audit pricing at the inter-audit firm level,

these results indicate that not all Big 4 audit firms are the same.

3.6.2 Analysis using industry specialization at the MSA city level

Because another stream of research focuses on auditor industry specialization at

the MSA city level (e.g., Ferguson et al. 2003; Francis et al. 2005b), I repeat my main tests

for Hypothesis 2 using alternative measures of industry specialization at the MSA city

level. Specifically, I define cities using the U.S. Census Bureau’s definition of MSA and

identify metropolitan areas based on the 5-digit zip codes available in Compustat for each

firm-year observation. I then classify an audit firm as an industry specialist using the

following two measures at the MSA city level: CAF_LARGE, which represents an audit

firm that has the largest market share in each two-digit SIC code and fiscal year, and

CAF_30, which represents an audit firm that has a market share greater than or equal to

30 percent.

Consistent with the main analysis, I examine the differential effects of the industry

specialization premium on the audit fees charged by PwC specialists and other specialists

using two interaction terms: “SPEC – Others” and “SPEC – PwC”. Again, as reported in

Table 3.8, while the estimated coefficients on both SPEC – Others and SPEC – PwC are

positive and significant at less than 1% levels, I find that the magnitudes of the SPEC -

PwC coefficients are significantly larger than those of SPEC – Others, indicating that the

individual PwC reputation results in higher audit fees.

3.7 Conclusion

This paper addresses the concept of individual auditor reputation by investigating

audit pricing differences within Big 4 audit firms during the period 2004–2014. In addition

to the general Big 4 premium, I argue that the individual auditor reputational effect also

60

plays an important role in the U.S. market and enables PwC to earn an above-average

fee premium relative to the other Big 4 firms.

Furthermore, I argue that the positive relationship between audit fees and auditor

industry specialization is exaggerated by the confounding effect of the PwC fee premium

in the audit fee model. This occurs because the popular measures used to identify industry

specialists are more likely to designate PwC as an industry specialist relative to the other

Big 4 firms. I also find that the estimated coefficients for the group of PwC specialists are

consistently observed and significantly larger than those for the group of other (non-PwC)

specialists across various measures of auditor industry specialization.

In sum, this study raises concerns regarding whether it is appropriate to treat all

Big 4 firms as a homogenous group of firms in the audit fee model, as I argue that not all

Big 4 audit firms are the same. The evidence of a PwC fee premium further suggests that

structural, cultural and technological differences within the Big 4 firms potentially

contribute to variations in the revenues generated from audit services, particularly given

the existence of the PwC phenomenon, as previously suggested in Hay (2013).

61

3.8 Figures and tables

Figure 3.1 The Vault’s Annual Accounting survey

The survey consists of questions about life at the professional’s firm (or former

firm) and a prestige rating. As for the prestige rating, participants were asked to rate

companies other than their own on a scale of 1 to 10, with 10 being the most prestigious.

It is important to note that all participants were asked only to rate firms which they were

familiar and were not permitted to rate their own (or former employer).

The Vault’s Accounting survey regarding the most Prestigious Accounting firms

Ranking

Calendar Year

2016 2015 2014 2013 2012 2011 2010 2009 2008 2007

PwC 1st 1st 1st 1st 1st 1st 1st 2nd 1st 2nd

EY 2nd 2nd 2nd 2nd 3rd 2nd 2nd 3rd 2nd 1st

KPMG 4th 4th 4th 4th 4th 4th 4th 4th 4th 4th

Deloitte 3rd 3rd 3rd 3rd 2nd 3rd 3rd 1st 3rd 3rd

Source: http://www.vault.com

Some interesting responses from participants regarding the perception for each of

the Big 4 firms include:

“Challenging and rewarding work with high profile clients—you get to work with

Fortune 500 companies on a daily basis (PwC)”

“Runs employees ragged—insane working hours (PwC)”

“The variety of client experiences is amazing and contributes to rapid learning

and growth (Deloitte)”

“Demanding, often unpredictable hours (Deloitte)”

“Challenging, enjoyable work (Ernst & Young)”

“Not on PwC’s level but very well respected (Ernst & Young)”

“The great people and family-like culture (KPMG)”

“Known as a distant No. 4 in the Big 4 but still very prestigious (KPMG)”

http://www.vault.com/company-rankings/accounting/most-prestigious-accounting-firms

62

Figure 3.2 The World’s 10 Most Powerful Brands

The Brand Finance index is an annual

assessment of the brand value of over 500 of the

world’s best-known businesses. The assessment

measures a range of metrics including brand

awareness, satisfaction and recommendations,

financial performance and internal investment,

market share and revenues. It also examines

corporate responsibility, governance, and the views

of internal and external stakeholders. Organizations

are then judged relative to their competitors.

Overall, PwC has retained its position as the

number one professional services brand, and one

of the world’s top ten most powerful brands in the

Brand Finance Index 2017.

Source: http://www.brandfinance.com

http://www.brandfinance.com/

63


These tables report the summary statistics of variables used for the audit fee model from the fiscal

year 2004 to 2014. All continuous variables are winsorized at the 1st and 99th percentiles.

Variable definitions for audit fee model:

LNAFEES = the natural logarithm of audit fees; LNASSET = the natural logarithm of total assets

(in millions); CURRENT = the ratio of current assets to current liabilities; INVREC = the ratio of total

inventory and receivables to total assets; LEVERAGE = the sum of short-term and long-term debt,

divided by total assets; ROA = income before extraordinary items, scaled by average total assets;

INTL = indicator variable equals to one if an audit’s client has international operations, and zero

otherwise; MA = indicator variable equal to one if an audit’s client is engaged in a merger or

acquisition (as reported by SDC Platinum) during the year, and zero otherwise; SPI_DM = indicator

variable equal to one if an audit’s client has a special item during the year, and zero otherwise;

LNBUSSEG = the natural logarithm of one plus the number of business segments; LOSS =

indicator variable equals to one if income before extraordinary items is negative in the current

period, and zero otherwise; MTB = the firm’s market value divided by its book value; BUSY =

indicator variable equals to one if an audit’s client has a year-end fall on December 31, and zero

otherwise; TENURE = the number of years the company has been audited by the same audit firm;

IPO = indicator variable equals to one if an audit’s client is engaged in an initial public offering

during the year, and zero otherwise; SEO = indicator variable equal to one if an audit’s client is

engaged in a seasoned equity offering (as reported by SDC Platinum) during the year, and zero

otherwise; OPINION = indicator variable equal to one if an audit’s client receives a modified audit

opinion and zero otherwise, where a modified opinion is defined as anything except a standard

unqualified audit opinion coded as one by COMPUSTAT; HIGHLIT = indicator variable equal to

one for high litigation risk industries as defined in Francis et al. (1994), and zero otherwise; BIG4 =

indicator variable equals to one if the firm’s auditor is a member of the Big 4 audit firms (PwC, EY,

KPMG and Deloitte) and zero otherwise; PWC (EY, KPMG and DELT) = indicator variable equals

to one if the firm’s auditor is PwC (Ernst & Young, KPMG and Deloitte, respectively), and zero

otherwise.

64

Table 3.1 - Continued

Variable N Mean Std. Dev. Median Q1 Q3

LNAFEES 39,985 13.412 1.380 13.497 12.422 14.349

LNASSET 39,985 5.601 2.309 5.671 3.959 7.247

CURRENT 39,985 2.864 2.901 1.983 1.260 3.295

INVREC 39,985 0.255 0.195 0.220 0.095 0.371

LEVERAGE 39,985 0.246 0.305 0.167 0.008 0.356

ROA 39,985 -0.098 0.420 0.026 -0.092 0.076

INTL 39,985 0.074 0.262 0.000 0.000 0.000

MA 39,985 0.166 0.372 0.000 0.000 0.000

SPI_DM 39,985 0.655 0.475 1.000 0.000 1.000

LNBUSSEG 39,985 0.964 0.487 0.693 0.693 1.386

LOSS 39,985 0.397 0.489 0.000 0.000 1.000

MTB 39,985 2.311 2.221 1.618 1.168 2.482

BUSY 39,985 0.693 0.461 1.000 0.000 1.000

TENURE 39,985 8.451 8.178 6.000 3.000 11.000

IPO 39,985 0.023 0.149 0.000 0.000 0.000

SEO 39,985 0.072 0.258 0.000 0.000 0.000

OPINION 39,985 0.360 0.480 0.000 0.000 1.000

HIGHLIT 39,985 0.369 0.483 0.000 0.000 1.000

BIG4 39,985 0.646 0.478 1.000 0.000 1.000

65

Table 3.2 Rankings of each major audit firm

Panel A Rankings of each major audit firm by the number of audit clients, the aggregated client assets and the aggregated audit revenues for the fiscal year 2004, 2009 and 2014.

Name of

Major audit firms

Number of

audit clients

%

audited Rank

Aggregated

client assets

(in billions)

%

assets

audited Rank

Aggregated

audit

revenues

(in millions)

% audit

revenues

earned Rank

Fiscal Year 2004

PwC 735 19% 2nd 2,810.3 33% 1st 1,879.9 36% 1st

Ernst & Young 818 22% 1st 2,113.6 25% 2nd 1,274.1 24% 2nd

Deloitte 547 14% 4th 1,688.5 20% 4th 901.7 17% 4th

KPMG 596 16% 3rd 1,897.7 22% 3rd 1,017.6 19% 3rd

Other audit firms 1,077 29% 78.6 1% 211.1 4%

Fiscal Year 2009

PwC 530 15% 2nd 3,118.1 28% 1st 1,614.6 27% 1st

Ernst & Young 751 21% 1st 2,834.6 26% 2nd 1,523.8 26% 2nd

Deloitte 483 14% 3rd 2,370.7 22% 4th 1,255.7 21% 3rd

KPMG 441 13% 4th 2,412.0 22% 3rd 1,057.9 18% 4th


Fiscal Year 2014

PwC 501 15% 2nd 4,395.2 30% 2nd 1,966.0 29% 1st

Ernst & Young 705 21% 1st 4,507.8 31% 1st 1,921.2 28% 2nd

Deloitte 429 13% 4th 3,188.3 22% 3rd 1,346.3 20% 3rd

KPMG 478 14% 3rd 2,198.2 15% 4th 1,100.1 16% 4th


66


Panel B PwC concentration ratios by client size quintile for the fiscal year 2004, 2009 and 2014.

PwC Concentration Ratio

Size Category

of Audit clients

Number of

audit clients

Aggregated

client assets

(in billions)

Aggregated

audit

revenues

(in millions)

%

audited

%

assets audited

% audit

revenues

earned

Fiscal Year 2004

Quintile 1 (Small) 784 18 100 6% 17% 14%

Quintile 2 744 74 249 15% 18% 20%

Quintile 3 760 217 524 21% 19% 23%

Quintile 4 701 596 808 26% 21% 28%

Quintile 5 (Large) 784 7,684 3,603 30% 34% 41%

Fiscal Year 2009

Quintile 1 (Small) 729 27 141 2% 13% 8%

Quintile 2 710 125 347 9% 11% 14%

Quintile 3 705 383 615 14% 12% 16%

Quintile 4 655 874 1,014 23% 25% 26%

Quintile 5 (Large) 720 9,544 3,774 28% 30% 32%

Fiscal Year 2014

Quintile 1 (Small) 697 44 144 3% 25% 13%

Quintile 2 664 166 401 6% 9% 11%

Quintile 3 677 562 782 16% 14% 20%

Quintile 4 627 1,292 1,194 22% 25% 26%

Quintile 5 (Large) 692 12,623 4,272 29% 31% 34%

Note: The PwC concentration ratio is the percentage of total number of audit clients, the aggregated client assets, and the aggregated audit revenues, earned by PwC.

67

Table 3.3 Audit fee model

This table presents regression results of the audit fee model for the full sample (Column 1) and the sample

of Big 4 audit firms (Columns 2 and 3). *, **, *** denote significance at the 0.10, 0.05, and 0.01 levels,

respectively, using two-tailed tests. T-statistic is determined by clustered standard errors at firm level.

(1) (2) (3)

Full Sample Big 4 audit firms Big 4 audit firms

DV = LNAFEES DV = LNAFEES DV = LNAFEES


Intercept 10.14 *** 146.18 10.63 *** 94.36 10.50 *** 94.85

LNASSET 0.50 *** 110.98 0.49 *** 82.40 0.49 *** 82.40

CURRENT -0.02 *** -12.54 -0.03 *** -8.85 -0.03 *** -8.85

INVREC 0.48 *** 11.84 0.76 *** 11.95 0.76 *** 11.95

LEVERAGE -0.12 *** -6.04 -0.15 *** -4.77 -0.15 *** -4.77

ROA -0.14 *** -9.09 -0.38 *** -11.09 -0.38 *** -11.09

INTL 0.15 *** 5.72 0.20 *** 6.62 0.20 *** 6.62

MA 0.03 *** 3.48 0.04 *** 3.75 0.04 *** 3.75

SPI_DM 0.16 *** 17.44 0.17 *** 14.33 0.17 *** 14.33

LNBUSSEG 0.06 *** 4.42 0.05 *** 3.50 0.05 *** 3.50

LOSS 0.14 *** 13.13 0.10 *** 6.85 0.10 *** 6.85

MTB 0.01 *** 5.15 0.01 ** 2.53 0.01 ** 2.53

BUSY 0.05 *** 3.25 0.05 *** 2.71 0.05 *** 2.71

TENURE 0.00 -0.92 0.00 1.03 0.00 1.03

IPO 0.28 *** 12.41 0.30 *** 11.84 0.30 *** 11.84

SEO 0.04 *** 3.27 0.04 *** 3.03 0.04 *** 3.03

OPINION 0.12 *** 13.32 0.11 *** 10.54 0.11 *** 10.54

HIGHLIT 0.10 *** 4.35 0.07 ** 2.29 0.07 ** 2.29

PwC 0.53 *** 24.11 0.13 *** 5.62

EY 0.45 *** 22.91 -0.08 *** -4.34 0.04 ** 1.98

KPMG 0.41 *** 18.82 -0.13 *** -5.62 DELT 0.45 *** 20.11 -0.08 *** -3.62 0.05 ** 1.97


Observations 39,985 25,838 25,838

Adjusted R-squared 84.95% 75.61% 75.61%

Difference | PwC – EY 0.08 *** 4.10 Difference | PwC – KPMG 0.12 *** 5.42 Difference | PwC – DELT 0.08 *** 3.49 Difference | KPMG – EY -0.04 ** -1.97

Difference | KPMG – DELT -0.04 * -1.90

68

Table 3.4 Audit fee model (Size partition)

These tables present regression results of the audit fee model by each asset quintile group for the full sample (Panel A) and the sample

of Big 4 audit firms (Panel B). *, **, *** denote significance at the 0.10, 0.05, and 0.01 levels, respectively, using two-tailed tests. T-statistic

is determined by clustered standard errors at firm level.

(1) (2) (3) (4) (5)




Panel A – The full sample

PwC 0.72 *** 12.66 0.64 *** 17.76 0.49 *** 13.19 0.30 *** 6.21 0.29 *** 2.78

EY 0.66 *** 14.88 0.57 *** 18.41 0.44 *** 12.81 0.21 *** 4.60 0.18 * 1.72

KPMG 0.55 *** 11.39 0.46 *** 12.87 0.37 *** 10.33 0.20 *** 4.15 0.21 ** 2.00

DELT 0.53 *** 7.75 0.51 *** 13.28 0.43 *** 11.21 0.24 *** 4.83 0.21 ** 1.97 Observations 8,252 7,992 7,978 7,485 8,278

Adjusted R-squared 63.6% 59.1% 48.2% 45.1% 68.0%

Diff. | PwC – EY 0.06 0.94 0.07 * 1.87 0.06 1.62 0.09 *** 2.81 0.11 *** 3.37

Diff. | PwC – KPMG 0.17 ** 2.43 0.18 *** 4.4 0.12 *** 3.31 0.09 ** 2.54 0.08 ** 2.17

Diff. | PwC – DELT 0.19 ** 2.28 0.13 *** 2.97 0.06 1.58 0.06 1.45 0.08 ** 2.23

Panel B – The sample of Big 4 audit firms

EY -0.09 *** -2.78 -0.04 -1.17 -0.06 * -1.70 -0.18 *** -4.51 -0.08 ** -2.05

KPMG -0.20 *** -5.36 -0.12 *** -3.49 -0.08 ** -2.23 -0.13 *** -2.73 -0.06 -1.38

DELT -0.12 *** -2.98 -0.05 -1.40 -0.06 -1.42 -0.13 *** -2.73 -0.07 -1.51

Observations 5,427 5,150 5,151 4,649 5,461


69

Table 3.5 Assignments of industry specialists across major audit firms

Panel A Frequencies (in percentage) of industries in which each audit firm is identified as an industry specialist for the fiscal year 2004, 2009 and 2014.

Variable definitions for auditor industry specialization measures:

AF_DOM (AT_DOM) = indicator variable equals to one if the company uses an audit firm that has either the largest market share based on aggregated audit fees (or aggregated client assets for AT_DOM) or the second highest market share in industries where the difference between the market shares of the leader and the second-highest firm is less than 10 percent, and zero otherwise; AF_30 (AT_30) = indicator variable equals to one if the company uses an audit firm that has a market share based on aggregated audit fees (or aggregated client assets for AT_30) greater than or equal to 30 percent in that particular industry, and zero otherwise.

Measurement variable

Audit fees

Measurement variable

Client assets

Name of

Major audit firms

Dominance

Market shares

(AF_DOM)

Market shares

Cut-off at 30%

(AF_30)

Dominance

Market shares

(AT_DOM)

Market shares

Cut-off at 30%

(AT_30)

Fiscal Year 2004 PwC 43.4% 60.1% 63.3% 49.6%

Ernst & Young 34.6% 26.4% 23.5% 40.7%

Deloitte 13.2% 8.0% 9.0% 7.5%

KPMG 8.7% 5.6% 4.1% 2.3%

Other audit firms 0.0% 0.0% 0.0% 0.0%

Fiscal Year 2009 PwC 38.3% 35.4% 36.8% 40.2%

Ernst & Young 41.3% 36.4% 33.4% 41.6%

Deloitte 12.5% 20.4% 20.0% 13.8%

KPMG 7.9% 7.8% 9.8% 4.4%


Fiscal Year 2014 PwC 38.0% 37.2% 36.2% 41.8%

Ernst & Young 31.0% 42.8% 45.8% 42.0%

Deloitte 20.9% 11.1% 11.2% 10.8%

KPMG 9.9% 8.9% 6.8% 5.4%


70


Panel B Industry specialist and PwC specialist concentration ratios by client size quintile for the fiscal year 2004, 2009 and 2014.

Concentration Ratio

Industry Specialist

Concentration Ratio

PwC Specialist

Size Category

of Audit clients

%

audited

%

assets

audited

% audit

revenues

earned

%

audited

%

assets

audited

% audit

revenues

earned

Fiscal Year 2004

Quintile 1 (Small) 8% 21% 18% 56% 53% 62%

Quintile 2 20% 28% 28% 64% 55% 65%

Quintile 3 31% 35% 36% 56% 41% 53%

Quintile 4 37% 35% 38% 62% 49% 62%

Quintile 5 (Large) 44% 55% 55% 61% 59% 71%

Fiscal Year 2009

Quintile 1 (Small) 6% 22% 15% 30% 48% 44%

Quintile 2 20% 34% 32% 34% 21% 33%

Quintile 3 34% 38% 39% 28% 13% 25%

Quintile 4 45% 49% 51% 38% 36% 40%

Quintile 5 (Large) 56% 67% 62% 39% 37% 43%

Fiscal Year 2014

Quintile 1 (Small) 5% 24% 19% 36% 30% 45%

Quintile 2 19% 33% 30% 23% 16% 26%

Quintile 3 37% 36% 45% 34% 22% 33%

Quintile 4 43% 48% 50% 39% 39% 42%

Quintile 5 (Large) 54% 67% 63% 43% 41% 48%

Note:

1. Industry specialist concentration ratio is the percentage of total number of audit clients, the aggregated client assets, and the aggregated audit revenues, earned by industry specialists (AF_DOM = 1).

2. PwC specialist concentration ratio is the percentage of total number of audit clients, the aggregated client assets, and the aggregated audit revenues, earned by PwC specialists (AF_DOM = 1 and PwC = 1).

71

Table 3.6 Audit fee model: Estimation of industry specialist premium

These tables present the regression results of the audit fee model with the inclusion of auditor industry specialization measures. *, **, ***

denote significance at the 0.10, 0.05, and 0.01 levels, respectively, using two-tailed tests. T-statistic is determined by clustered standard

errors at firm level.

(1) (2) (3) (4) Panel A – The Full Sample SPEC = AF_DOM SPEC = AF_30 SPEC = AT_DOM SPEC = AT_30 DV = LNAFEES DV = LNAFEES DV = LNAFEES DV = LNAFEES


Intercept 10.15 *** 147.75 10.14 *** 147.29 10.14 *** 147.71 10.14 *** 147.63

LNASSET 0.50 *** 111.06 0.50 *** 111.34 0.50 *** 111.17 0.50 *** 111.12

CURRENT -0.02 *** -12.51 -0.02 *** -12.52 -0.02 *** -12.54 -0.02 *** -12.53

INVREC 0.48 *** 11.84 0.48 *** 11.75 0.48 *** 11.76 0.48 *** 11.78

LEVERAGE -0.12 *** -5.91 -0.12 *** -5.99 -0.12 *** -5.98 -0.12 *** -5.97

ROA -0.14 *** -9.11 -0.14 *** -9.02 -0.14 *** -9.10 -0.14 *** -9.07

INTL 0.15 *** 5.68 0.15 *** 5.62 0.15 *** 5.69 0.15 *** 5.68

MA 0.03 *** 3.42 0.03 *** 3.46 0.03 *** 3.45 0.03 *** 3.45

SPI_DM 0.16 *** 17.43 0.16 *** 17.40 0.16 *** 17.36 0.16 *** 17.38

LNBUSSEG 0.06 *** 4.48 0.06 *** 4.59 0.06 *** 4.55 0.06 *** 4.54

LOSS 0.14 *** 13.08 0.14 *** 13.05 0.14 *** 13.13 0.14 *** 13.13

MTB 0.01 *** 5.08 0.01 *** 5.19 0.01 *** 5.17 0.01 *** 5.15

BUSY 0.05 *** 3.26 0.05 *** 3.27 0.05 *** 3.26 0.05 *** 3.27

TENURE 0.00 -1.10 0.00 -1.07 0.00 -1.07 0.00 -1.08

IPO 0.28 *** 12.51 0.28 *** 12.55 0.28 *** 12.49 0.28 *** 12.48

SEO 0.04 *** 3.41 0.04 *** 3.38 0.04 *** 3.35 0.04 *** 3.35

OPINION 0.12 *** 13.30 0.12 *** 13.53 0.12 *** 13.41 0.12 *** 13.46

HIGHLIT 0.10 *** 4.43 0.10 *** 4.55 0.10 *** 4.46 0.10 *** 4.46

BIG4 0.43 *** 24.33 0.44 *** 25.55 0.45 *** 25.53 0.45 *** 25.63

SPEC - Others 0.03 ** 2.18 0.05 ** 2.56 -0.01 -0.72 0.00 0.08

SPEC - PwC 0.13 *** 6.75 0.14 *** 6.70 0.11 *** 5.31 0.12 *** 5.48

Observations 39,985 39,985 39,985 39,985

Adjusted R-Squared 84.96% 84.96% 84.94% 84.94%

Difference | SPEC: PwC – Others 0.10 *** 4.51 0.10 *** 3.64 0.13 *** 5.1 0.12 *** 4.53

72


Panel B – The full sample (Size partition)

(1) (2) (3) (4) (5)


N = 8,252 N = 7,992 N = 7,978 N = 7,485 N = 8,278



B1: SPEC = AF_DOM

SPEC - Others -0.07 -1.47 0.02 0.75 0.04 * 1.77 0.02 0.87 0.05 ** 1.97

SPEC - PwC 0.19 *** 2.74 0.14 *** 3.86 0.10 *** 3.01 0.09 *** 2.86 0.14 *** 4.38


Diff. | SPEC: PwC – Others 0.26 *** 3.52 0.12 *** 2.81 0.05 1.44 0.07 * 1.86 0.09 *** 2.66

B2: SPEC = AF_30

SPEC - Others -0.14 ** -2.55 0.03 0.82 0.07 ** 2.23 0.04 1.09 0.07 ** 2.56

SPEC - PwC 0.15 * 1.95 0.08 ** 1.97 0.09 ** 2.41 0.08 ** 2.30 0.15 *** 4.18


Diff. | SPEC: PwC – Others 0.29 *** 3.33 0.06 1.08 0.02 0.53 0.05 0.98 0.08 * 1.83

B3: SPEC = AT_DOM

SPEC - Others -0.08 -1.56 -0.01 -0.19 -0.02 -0.84 0.00 0.03 -0.03 -1.23

SPEC - PwC 0.16 ** 2.02 0.10 ** 2.55 0.07 * 1.95 0.09 ** 2.51 0.09 *** 2.61


Diff. | SPEC: PwC – Others 0.24 *** 2.82 0.11 ** 2.34 0.09 ** 2.2 0.09 ** 2.14 0.12 *** 3.13

B4: SPEC = AF_30

SPEC - Others -0.06 -1.08 -0.01 -0.33 0.00 0.10 0.00 0.09 -0.02 -0.57

SPEC - PwC 0.17 ** 2.06 0.10 ** 2.24 0.07 * 1.84 0.09 ** 2.55 0.10 *** 2.81


Diff. | SPEC: PwC – Others 0.23 ** 2.54 0.11 ** 2.09 0.06 1.43 0.09 ** 2.07 0.11 *** 2.83

73

Table 3.7 Analysis using unexplained audit fees

Panel A The mean unexplained audit fees (UAF) by each fiscal year across major groups of audit firms.

Non-Big 4 firms PwC EY Deloitte KPMG

N Mean N Mean N Mean N Mean N Mean Fiscal Year 2004 1,247 0.00

828 0.06

939 0.02

657 -0.08

680 -0.02

t-statistic [Ho = 0] -0.06 3.04 *** 0.96 -3.83 *** -0.85 Fiscal Year 2005 1,416 0.00

692 0.06

904 0.00

603 -0.04

568 -0.04

t-statistic [Ho = 0] -0.01 3.13 *** 0.27 -1.9 * -1.84 *

Fiscal Year 2006 1,468 0.00

605 0.08

880 -0.02

572 -0.01

518 -0.04

t-statistic [Ho = 0] 0.03 4.19 *** -1.39 -0.7 -1.76 *

Fiscal Year 2007 1,469 0.00

569 0.04

841 -0.02

529 0.02

458 -0.02

t-statistic [Ho = 0] 0.19 1.98 ** -1.45 0.85 -1.1 Fiscal Year 2008 1,364 0.00

543 0.03

789 -0.04

488 0.04

448 -0.01

t-statistic [Ho = 0] 0.06 2.03 ** -2.84 *** 1.93 * -0.61 Fiscal Year 2009 1,293 0.00

530 0.03

751 -0.02

481 0.01

444 -0.01

t-statistic [Ho = 0] 0.1 1.87 * -1.59 0.67 -0.79 Fiscal Year 2010 1,210 0.00

530 0.05

710 -0.02

459 0.00

439 -0.03

t-statistic [Ho = 0] -0.04 2.95 *** -1.31 -0.18 -1.53 Fiscal Year 2011 1,133 0.00

505 0.07

712 -0.03

425 0.00

452 -0.04

t-statistic [Ho = 0] -0.06 4.49 *** -2.1 ** 0.13 -2.07 **

Fiscal Year 2012 1,125 0.00

502 0.07

717 -0.03

406 0.01

448 -0.05

t-statistic [Ho = 0] 0.05 4.34 *** -1.94 * 0.57 -2.95 ***

Fiscal Year 2013 1,182 0.00

516 0.08

719 0.00

410 -0.02

453 -0.07

t-statistic [Ho = 0] -0.06 4.78 *** -0.11 -1.14 -4.17 ***

Fiscal Year 2014 1,240 0.00

504 0.08

705 0.01

431 -0.01

478 -0.08

t-statistic [Ho = 0] 0.03 4.6 *** 0.46 -0.32 -4.48 ***

*, **, *** denote significance at the 0.10, 0.05, and 0.01 levels, respectively, using two-tailed tests.

74


Panel B The differences in means of unexplained audit fees (UAF)

This table reports univariate tests of the differences in means of UAF using an indicator of PwC together with an indicator of the dominant

market leader (AF_DOM) to partition the full sample into four subsamples: (1) companies that are audited by PwC specialists, (2) companies

that are audited by specialists but not by PwC, (3) companies that are audited by PwC but not by specialists, and (4) Firms that are audited

by neither PwC nor specialists.

Fiscal

Year

Subsample A

AF_DOM = 1

Subsample B

AF_DOM = 0

Differences in means:

Unexplained audit fees (UAF)

[A1]

PwC = 1

[A2]

PwC = 0

[B1]

PwC = 1

[B2]

PwC = 0

[D1]

A - B

[D2]

A1 - A2

[D3]

B1 - B2

[D4]

A1 - B1

[D5]

A2 - B2

N Mean N Mean N Mean N Mean t-stat t-stat t-stat t-stat t-stat

2004 697 0.07 554 0.04 131 -0.01 2,969 -0.02 4.55 *** 0.94 0.33 1.52 2.68 ***

2005 570 0.08 725 0.02 122 -0.02 2,766 -0.02 3.83 *** 2.11 ** -0.04 1.98 ** 1.88 *

2006 464 0.09 756 0.01 141 0.04 2,682 -0.02 3.21 *** 2.91 *** 1.46 1.01 1.30

2007 416 0.05 797 0.00 153 0.01 2,500 -0.01 1.32 1.82 * 0.34 0.97 0.26

2008 405 0.04 701 -0.01 138 0.01 2,388 -0.01 1.07 2.00 ** 0.40 0.80 -0.07

2009 400 0.04 732 -0.01 130 0.01 2,237 0.00 0.92 1.93 * 0.26 0.89 -0.14

2010 438 0.06 615 -0.02 92 -0.01 2,203 -0.01 1.22 3.44 *** 0.02 1.49 -0.91

2011 392 0.08 641 -0.02 113 0.07 2,081 -0.01 1.58 4.05 *** 2.02 ** 0.29 -0.57

2012 408 0.06 640 -0.04 94 0.12 2,056 -0.01 0.29 4.13 *** 3.00 *** -1.24 -1.62

2013 408 0.10 609 0.00 108 0.01 2,155 -0.02 3.77 *** 4.29 *** 0.81 2.05 ** 0.91

2014 377 0.09 656 0.01 127 0.03 2,198 -0.02 3.95 *** 3.31 *** 1.44 1.67 * 1.81 *


75


Panel C The differences in means of unexplained audit fees (UAF) using the matched sample within Big 4 audit firms

The matched sample is created based on the size of Big 4 clients within a

maximum distance of 3 percent. I require matched-pair observations to possess

common attributes including complexity of operations (INTL), financial

performance (LOSS), firm-specific event (MA) in the same fiscal year and two-digit

SIC codes.

Variable: UAF

PwC

PwC = 1

Other Big 4 firms

PwC = 0 Student's

Signed

Rank

N Mean N Mean p-value p-value

Panel A – The subsample of Big 4 clients (BIG4 = 1)

UAF

(BIG4 = 1) 1,496 0.07 1,496 -0.01 0.00 *** 0.00 ***

Panel B – The subsample of companies that are audited by industry specialists

UAF

(BIG4 = 1; AF_DOM = 1) 1,407 0.07 1,407 0.00 0.00 *** 0.00 ***

(BIG4 = 1; AT_DOM = 1) 759 0.08 759 -0.02 0.00 *** 0.00 ***

*, **, *** indicate significance at the 0.10, 0.05 and 0.01 levels, respectively, based on two-tailed t-tests and Wilcoxon signed-rank tests of differences in means.

76

Table 3.8 Analysis using industry specialization at the MSA city level

This table presents the regression results of the audit fee model with the inclusion of alternative

measures of industry specialization at the MSA city level. *, **, *** denote significance at the 0.10,

0.05, and 0.01 levels, respectively, using two-tailed tests. T-statistic is determined by clustered

standard errors at firm level.

Variable definitions:

CAF_LARGE = indicator variable equals to one if the company uses an audit firm that has the

largest market share in that particular industry at the MSA city level, and zero otherwise; CAF_30

= indicator variable equals to one if the company uses an audit firm that has a market share based

on aggregated audit fees greater than or equal to 30 percent in that particular industry at the MSA

city level, and zero otherwise.

(1) (2)

DV = LNAFEES DV = LNAFEES

SPEC = CAF_LARGE SPEC = CAF_30


Intercept 10.22 *** 132.78 10.22 *** 132.34

LNASSET 0.49 *** 93.74 0.49 *** 93.55

CURRENT -0.03 *** -11.86 -0.03 *** -11.85

INVREC 0.50 *** 10.15 0.50 *** 10.16

LEVERAGE -0.14 *** -5.87 -0.14 *** -5.92

ROA -0.16 *** -8.49 -0.16 *** -8.52

INTL 0.14 *** 4.75 0.14 *** 4.77

MA 0.03 *** 2.68 0.03 *** 2.67

SPI_DM 0.14 *** 13.11 0.14 *** 13.08

LNBUSSEG 0.07 *** 4.70 0.07 *** 4.62

LOSS 0.12 *** 9.66 0.12 *** 9.67

MTB 0.01 *** 3.61 0.01 *** 3.61

BUSY 0.02 1.35 0.02 1.36

TENURE 0.00 -1.25 0.00 -1.15

IPO 0.29 *** 11.35 0.29 *** 11.34

SEO 0.03 * 1.95 0.03 * 1.95

OPINION 0.11 *** 11.00 0.11 *** 11.00

HIGHLIT 0.09 *** 3.54 0.09 *** 3.62

BIG4 0.43 *** 21.61 0.43 *** 21.31

SPEC - Others 0.09 *** 5.62 0.08 *** 4.92

SPEC - PwC 0.17 *** 6.71 0.16 *** 6.40




Diff. | SPEC: PwC – Others 0.08 *** 2.89 0.08 *** 2.98

77

Chapter 4. The difference among the Big 4 firms: Further evidence from audit quality

4.1 Introduction

For over three decades, the Big 4 audit firms28 have generally been grouped as a

homogeneous set of firms regarding their brand-name reputation. Auditor size, as

measured by Big 4 membership, is then argued to be associated with greater incentives

and competencies in providing higher audit quality. In recent years, there has been a

growing stream of literature on auditor competencies that examines quality variation at the

inter-audit firm level by focusing on a variety of audit firm characteristics, including industry

specialization and individual brand name reputation. While the lack of a consistent

definition and measure of auditor industry specialization is considered a major challenge

inherent to this literature (e.g., Audousset-Coulier et al. 2015), it is interesting to observe

that PwC and EY are more likely to be designated industry specialists29 relative to other

audit firms across almost all industries (e.g., Audousset-Coulier et al. 2015; Cahan et al.

2011; Knechel et al. 2007; Li et al. 2010), which raises an empirical question whether

individual differences can be detected and associated with variations in audit quality

delivered by the Big 4 firms.

In this study, I focus on audit quality differences within the Big 4 firms, which are

the dominant suppliers of audit services to large corporations in the U.S. market, during

the period 2004–2014. Because PwC and EY have successfully differentiated themselves

from their competitors in terms of market share across most industries, I hypothesize that

an individual audit firm’s competencies, as perceived through its brand name reputation,

are likely to play an important role in providing high-quality audits and can potentially be

28 In this study, I use Big 4 as a generic term encompassing the Big 8, Big 6, Big 5, and Big 4 to

reflect the significant mergers of the 1980s and 1990s, as illustrated in Figure 1.1.

29 The existing auditing literature generally uses the within-industry market share approach based

on audit fees, sales, or size to identify an audit firm that holds a significant portion of the market

share in an industry as an industry specialist.

78

used as a basis for within-Big 4 service differentiation. Thus, I expect to observe cross-

sectional variations in audit quality at the inter-audit firm level.

Since there is no consensus on a definition of audit quality or on relevant indicators

to assess audit quality, I introduce two output-based proxies to measure the level of audit

quality delivered. First, I measure audit quality using clients’ earnings quality based on the

notion that high-quality audits help to constrain opportunistic earnings management (e.g.,

Becker et al. 1998; Francis et al. 1999a) and cause firms to report lower amounts of

estimated discretionary accruals. Alternatively, because auditor communication is a direct

and useful way for an auditor to help financial statement users predict bankruptcy, I use

the propensity to issue a qualified GC opinion as another measure of audit quality based

on the argument that higher quality auditors are more likely to use a GC opinion (e.g., Lim

and Tan 2008; Reichelt and Wang 2010; Weber and Willenborg 2003).

Overall, I consistently find that the estimated coefficients of each of the Big 4 firms

(PwC, EY, KPMG and Deloitte) are negative and significant in the earnings quality model,

which suggests that Big 4 auditors do improve earnings quality. More interestingly, while

both PwC and EY are more likely to be identified as industry specialists across almost all

industries, I find that the differences between the coefficients of PwC and EY are negative

and significant at 10% levels. Similarly, I find that the negative relation between EY and

GC opinions is relatively larger than those of the other Big 4 firms in both the full sample

and the subsamples of severely financially distressed firms, indicating that on average,

EY is less likely to issue GC audit opinions relative to the other Big 4 firms. Together,

these results reinforce the importance of an individual audit firm’s competencies,

especially for PwC and EY, in delivering high audit quality.

Next, because the existing measures of auditor industry specialization heavily

designate PwC and EY as industry specialists, a separate but related question is whether

the effect of industry specialization is confounded by individual differences within the Big

4 firms. Specifically, I investigate the differential effects of individual Big 4 firms by

introducing three interaction terms to capture the impact of industry expertise on audit

quality for each group of industry specialists: PwC, EY and others.

Consistent with prior studies, the results are mixed regarding the impact of industry

expertise on audit quality. Specifically, either marginal results or no relation is observed

79

across various settings. However, when I break down the indicator variable of industry

expertise into three groups, the negative relationship between industry specialization and

earnings quality appears to be stronger for the group of PwC specialists but weaker or

insignificant for the other two groups. Similarly, I find that the clients of PwC specialists

are more likely to receive a going-concern audit opinion, with weaker evidence for the

other two groups. Together, these results suggest that the inconsistent results regarding

the influence of industry expertise on audit quality can be partially explained by the

individual differences within the Big 4 audit firms.

Finally, sensitivity tests using either signed discretionary accruals (e.g., DeFond et

al. 2017a; Hribar and Nichols 2007) or auditor industry specialization measures at the

MSA city level (e.g., Ferguson et al. 2003; Francis et al. 2005b) further confirm that

individual differences in audit quality have been detected within the Big 4 firms.

My study makes several contributions to the auditing literature. First, while the Big

4 auditors have long been treated as a homogenous group, I document that there is

significant audit quality variation within these firms, suggesting that individual audit firms’

competencies potentially have a significant effect on improving audit quality. Specifically,

I argue that certain differentiating characteristics of PwC auditors are associated with

above-average audit quality and hence can be considered as a basis for service

differentiation within the Big 4 firms. This emphasizes the relevance of distinguishing

among individual Big 4 firms and partially explains why PwC has maintained its leadership

position in terms of total audit revenues and why it has consistently been ranked as the

most prestigious audit firm30 from the year 2000 to 2017.

Second, I reveal that the inconsistencies found in the industry specialization

literature are confounded by the individual differences within the Big 4 firms, primarily by

the effects of PwC and EY. This occurs because while both PwC and EY are dominant in

the overall audit market, only PwC appears to successfully differentiate itself from its

competitors by offering above-average audit quality.

30 Vault’s annual accounting survey is conducted by asking thousands of accounting professionals

at the top audit firms to rate their competitor firms in terms of prestige and to provide a few words

describing their perception of those competitor firms.

80

Access to proprietary data from audit firms and their audit clients would be

beneficial to address questions that pertain to within-Big 4 audit quality differences.

The remainder of this study is organized as follows. In section 2, I review the

related literature and develop relevant hypotheses. In section 3, I present my research

design, and I report the sample selection and descriptive statistics in section 4. Section 5

and 6 present empirical results and additional tests, respectively. Finally, section 7

concludes the study.



While the concept of audit quality is fundamental in auditing research, there is no

consensus on a definition of audit quality or on relevant indicators to assess audit quality,

since the term “audit quality” can be viewed from several perspectives.

One of the most cited definitions of audit quality is that by DeAngelo (1981, page

186), which states that “the quality of audit services is defined to be the market-assessed

joint probability that a given auditor will both (a) discover a breach in the client’s accounting

system, and (b) report the breach”. Thus, one way to infer audit quality is to consider input-

based measures, including auditor firm size and industry specialization. While there is

strong evidence indicating that auditor size, as measured by Big 4 membership, leads to

higher audit quality31, some researchers have begun to investigate audit quality variation

within these Big 4 firms instead of treating all these firms as a homogenous group. For

example, Simunic (1980) finds a positive and significant coefficient of PW in the audit fee

model, suggesting that there is price competition based on a differentiated product for PW

in the audit market. Dunn and Mayhew (2004) also find that Coopers and Lybrand LLP

assisted their audit clients with an annual assessment of industry-specific disclosures and

31 The indicator for Big 4 has been shown to be associated with almost all other audit quality proxies,

including a lower incidence of accounting fraud (e.g., Lennox and Pittman 2010), a lower incidence

of accounting restatements (e.g., Eshleman and Guo 2014), lower discretionary accruals (e.g.,

Becker et al. 1998; Francis et al. 1999b), higher audit fees (e.g., Craswell et al. 1995; Hay et al.

2006), increased ERCs (e.g., Teoh and Wong 1993), improved analyst earnings forecasts (e.g.,

Behn et al. 2008), and a lower cost of debt and equity (e.g., Khurana and Raman 2004).

81

that a group of PwC partners assisted their audit clients in targeted industries by

implementing enhanced disclosure practices, indicating that certain traits of individual

audit firms are likely to play a significant role in delivering high-quality audits.

Evidence from auditor industry specialization

The industry expertise research further extends the auditor size literature by

focusing on audit quality differentiation at the inter-audit firm level. Specifically, industry

specialists are expected to develop their knowledge of industry business and accounting

practices, which are specific to their clients in each industry, and to subsequently provide

higher audit quality than non-specialists (e.g., Balsam et al. 2003; Lim and Tan 2010; Low

2004; Romanus et al. 2008; Solomon et al. 1999), which indicates that industry specialists

have greater competency in delivering high-quality audits. However, because an audit

firm’s level of industry specialization is unobservable, a significant challenge in this stream

of literature is identifying industry specialists. According to Zeff and Fossum (1967) and

Palmrose (1986), within-industry market share is widely used in the literature based on a

diverse array of measurement variables, such as client asset size, the number of clients,

and audit fees32, for the calculation of auditor market share. Specifically, the market share

approach assumes that an audit firm that allocates its resources and audit effort to develop

the industry-specific knowledge to become an industry specialist will be able to

differentiate itself from its competitors in terms of market share in the audit market.

However, while the impact of industry expertise on audit quality has been

extensively examined in the literature, the results of industry specialization assignments

consistently show that PwC has managed to differentiate itself from its competitors with

regard to its within-industry market share position. For example, Knechel et al. (2007)

document that the identification of industry specialists is highly stable, with of PwC

dominating in 23 out of 40 industries, followed by EY (14 industries), Deloitte (10

industries) and KPMG (5 industries). Li et al. (2010) also document that PwC had the

highest number of joint national and city industry leaders in each of the six years in their

sample. Cahan et al. (2011) use the official websites of Big 4 firms to develop a list of key

industries and consistently document that the self-proclaimed areas of industry experts

32 Because information on audit fees was not publicly available before the 2000s, researchers

generally used either client assets or sales revenue as an indirect proxy for audit fees.

82

are stable and consistent with the number of industries in which the auditor qualifies as an

industry specialist based on the market share approach.


The nature of the competition among Big 4 firms has been a concern to regulators,

particularly with the audit market concentrated among the four largest audit firms (e.g.,

GAO 2008). While auditor size captures both auditor incentives and competencies, it is

interesting to observe that of the Big 4 firms, PwC and EY are most often identified as

industry specialists across almost all industries and sample periods (e.g., Cahan et al.

2011; Knechel et al. 2007; Li et al. 2010). Because industry-specific skills and technical

capabilities are considered a basis for within-Big 4 product differentiation, this indicates

that individual differences33 can therefore be detected within the Big 4 firms in the U.S.

market. Specifically, I hypothesize that individual differences within the Big 4 firms will be

associated with differences in audit quality. Thus, the first hypothesis is stated in null form

as follows:

H1: There are no significant differences in audit quality within the Big 4 firms.

While researchers have begun to investigate audit quality variation within the Big

4 auditors by teasing out the effects of competency on audit quality, the prevalence of

PwC and EY across most industry specialization measures raises an interesting question

whether the association between audit quality and industry specialization is confounded

by individual differences within the Big 4 firms. In other words, I argue that the evidence

of industry expertise merely reflects the individual audit firms’ competencies and

subsequently leads to inconsistent results regarding the impact of industry specialization

on audit quality in archival auditing research. The second hypothesis is therefore stated in

null form as follows:

H2: There are no significant differences in audit quality among groups of

industry specialists with respect to individual audit firms’ competencies.

33 The identification of an individual brand name reputation is further supported by the results of

Vault’s annual accounting survey, which has consistently ranked PwC as the most prestigious audit

firm, followed by EY, Deloitte and KPMG.

83

4.3 Research Design

To test the first hypothesis, I investigate whether there are individual differences in

perceived audit quality within the Big 4 firms by breaking down the indicator variable of

Big 4 into four individual firm indicator variables (PwC, EY, KPMG and DELT) in the

earnings quality model. H1 predicts that the differences in the estimated coefficient for

each Big 4 firm in the earnings quality model are suggestive of individual auditor

competencies. In addition, to focus on differences in accrual quality within the Big 4 firms,

I re-run the earnings quality regression model using the sample of Big 4 firms.

In this study, I consider two widely used accrual models to identify abnormal

accruals as a measure of earnings quality. First, following Jones (1991) and Dechow et

al. (1995), I estimate normal levels of accruals based on the modified Jones model, which

defines the accrual process as a function of growth in credit sales and investment in PPE,

while controlling for firm performance (Kothari et al. 2005). The following model34 is

estimated with a minimum of 20 observations in each industry-year cluster:

𝐓𝐀𝐭 = α0 + α1 (1

ATt−1) + α2 (

(∆REVt − ∆RECt)

ATt−1) + α3 (

PPEt

ATt−1) + α4ROAt + εt

where TAt = the difference between net income and operating cash flow in year t,

scaled by lagged assets; ATt−1 = lagged total assets; ∆REVt − ∆RECt = the change in total

revenues less the change in total receivables in year t from year t-1; PPEt = the gross book

value of property, plant and equipment at the end of year t; and ROAt = income before

extraordinary items, scaled by total assets. The absolute value of the error term

(ADA_MJR) is then used as a measure of earnings quality, and a larger unsigned

magnitude of discretionary accruals indicates poorer accrual quality. Second, I use the

standard deviation of the residuals from regressions relating current accruals to cash flows

(e.g., Dechow and Dichev 2002; Francis et al. 2005a) as an alternative measure of accrual

quality. Specifically, I define the total change in working capital (ΔWCt) as the change in

accounts receivables plus the change in inventory and the change in other assets minus

the change in accounts payable and the change in taxes payable, scaled by average total

34 Following Kothari et al. (2005), I include a constant term to mitigate model misspecification.

84

assets. The following model is estimated with a minimum of 20 observations in each

industry-year cluster:

𝚫𝐖𝐂𝐭 = α0 [1

ATt−1] + α1 [

CFOt−1

ATt−1] + α2 [

CFOt

ATt−1] + α3 [

CFOt+1

ATt−1] + α4 [

∆REVt

ATt−1] + α5 [

PPEt

ATt−1] + εi,t

where ΔWCt is the total change in working capital in year t; and CFOt is the firm’s cash flow

from operations in year t. The standard deviation of the error terms over years t-4 through

t (SD_FLOS) is then used as a measure of earnings quality, and a higher standard

deviation denotes lower accrual quality.

Building on prior earnings quality studies (e.g., Audousset-Coulier et al. 2015;

Dechow et al. 2010), I estimate the earnings quality model with the inclusion of four

individual firm indicator variables as follows:

Equation 1: The earnings quality model

𝐀𝐃𝐀_𝐌𝐉𝐑𝐭(𝐨𝐫 𝐒𝐃_𝐅𝐋𝐎𝐒𝐭)

= 𝛂𝟎 + 𝛂𝟏𝐒𝐈𝐙𝐄𝐭 + 𝛂𝟐𝐒𝐃_𝐒𝐀𝐋𝐄𝐭 + 𝛂𝟑𝐒𝐃_𝐂𝐅𝐎𝒕 + 𝛂𝟒𝐂𝐅𝐎𝐭

+ 𝛂𝟓𝐋𝐄𝐕𝐄𝐑𝐀𝐆𝐄𝐭 + 𝛂𝟔𝐋𝐎𝐒𝐒𝐭 + 𝛂𝟕𝐌𝐓𝐁𝐭 + 𝛂𝟖𝐇𝐈𝐆𝐇𝐋𝐈𝐓𝐭

+ 𝛂𝟗𝐀𝐋𝐓𝐌𝐀𝐍𝐭 + 𝛂𝟏𝟎𝐀𝐂𝐂𝐑_𝐋𝐀𝐆𝐭 + 𝛂𝟏𝟏𝐏𝐰𝐂𝐭 + 𝜶𝟏𝟐𝐄𝐘𝐭

+ 𝛂𝟏𝟑𝐊𝐏𝐌𝐆𝐭 + 𝛂𝟏𝟒𝐃𝐄𝐋𝐓𝐭

+ 𝐘𝐞𝐚𝐫 𝐚𝐧𝐝 𝐈𝐧𝐝𝐮𝐬𝐭𝐫𝐲 𝐟𝐢𝐱𝐞𝐝 𝐞𝐟𝐟𝐞𝐜𝐭𝐬 + 𝛆𝐭


Alternatively, I assess the robustness of my results by examining the association

between those auditor attributes and auditors’ going-concern opinions. Because large

companies are more likely to hire the Big 4 auditors, the clients of Big 4 auditors are less

likely to received GC opinions (e.g., DeFond and Lennox 2011; Francis and Yu 2009).

This is consistent with Carson et al. (2012), documenting that the overall frequency of GC

opinions issued in the U.S. was 36.7% for companies with small market capitalizations

(less than $75 million) but only 0.33% for those with large market capitalizations (greater

than $500 million). However, some researchers argue that industry specialists (e.g., Lim

and Tan 2008; Reichelt and Wang 2010) and larger offices of Big 4 auditors (Francis and

Yu 2009) are more likely to issue GC opinions since they have greater expertise in

detecting the circumstances that warrant a GC report.

85

Building on prior studies (e.g., DeFond et al. 2002; Reynolds and Francis 2000), I

predict the issuance of a going-concern opinion by estimating the following logistic model

to test the first hypothesis:

Equation 2: The auditors’ going-concern opinion model

𝐆𝐂𝐭 = α0 + α1𝐋𝐍𝐀𝐒𝐒𝐄𝐓𝐭 + α2𝐌𝐓𝐁𝐭 + α3𝐋𝐄𝐕𝐄𝐑𝐀𝐆𝐄𝐭 + α4𝐂𝐇_𝐋𝐄𝐕𝐭

+ α5𝐂𝐅𝐎𝐭 + α6𝐀𝐋𝐓𝐌𝐀𝐍𝐭 + α7𝐏𝐋𝐎𝐒𝐒𝐭 + α8𝐇𝐈𝐆𝐇𝐋𝐈𝐓𝐭

+ α9𝐓𝐄𝐍𝐔𝐑𝐄𝐭 + α10𝐑𝐎𝐀𝐭 + α11𝐒𝐃_𝐎𝐈𝐀𝐃𝐏𝐭 + α11𝐏𝐰𝐂𝐭

+ 𝛼12𝐄𝐘𝐭 + α13𝐊𝐏𝐌𝐆𝐭 + α14𝐃𝐄𝐋𝐓𝐭

+ 𝐘𝐞𝐚𝐫 𝐚𝐧𝐝 𝐈𝐧𝐝𝐮𝐬𝐭𝐫𝐲 𝐟𝐢𝐱𝐞𝐝 𝐞𝐟𝐟𝐞𝐜𝐭𝐬 + ε𝐢𝐭


To test the second hypothesis, I examine whether the evidence of auditor industry

specialization (SPEC) largely reflects the effect of individual audit firms, primarily PwC and

EY, in the earnings quality model. As is common in the auditor industry specialization

literature, the variable SPEC is identified based on the dominant market share approach35

with two different measurement variables: audit fees (AF_DOM) and client assets

(AT_DOM). The dominant market share approach includes the market leader in each year

and industry (based on two-digit SIC code) together with the firm with the second-highest

share in industries in which there is not a major difference (less than 10 percent) between

the leader and the second-highest firm.

Specifically, I introduce three interaction terms to examine the differential effect of

industry specialization for each group of PwC, EY and other industry specialists in the

earnings quality model. SPEC – PwC is an interaction term between SPEC and PwC.

SPEC – EY is an interaction term between SPEC and EY. SPEC – Others is an interaction

between SPEC and an indicator variable that equals one if the incumbent firm is neither

PwC nor EY and zero otherwise. I predict that if industry specialization improves audit

quality, the estimated coefficients for each group of industry specialists will be negative

and significant in the earnings quality model (Equation 1) but positive and significant in

the auditors’ going-concern opinion model (Equation 2). Otherwise, the differences among

35 In untabulated results, I find that the inferences are unchanged when I use alternative measures of industry specialization using the market share cut-off at 30 percent.

86

the estimated coefficients of industry specialization for each group of PwC, EY and other

industry specialists would be suggestive of a confounding effect of individual audit firms’

competencies.

In addition, because the decision to issue a going-concern opinion is most

noticeable for severely financially distressed firms, I re-run the GC model using the

subsample of severely financially distressed firms that report both negative net income

and negative cash flow from operations (e.g., Blay and Geiger 2013; Callaghan et al.

2009).



I utilize the available datasets from Compustat fundamental annual files and CRSP

monthly stock files to obtain the necessary financial statement data for all firm-years from

2004 to 2014 and exclude all observations related to financial services (between SIC 6000

and 6999). I delete firms with total assets of less than $1 million and then merge the

datasets with the Audit Analytics database to obtain auditor identities in order to examine

the differential effects of certain auditor attributes on audit quality. These sample selection

procedures yield a final sample of 41,167 and 40,920 firm-year observations for the

earnings quality (Equation 1) and auditors’ going-concern opinion (Equation 2) models,

respectively. Finally, I winsorize observations that fall in the top and bottom 1 percent of

the distribution for each non-discrete variable to mitigate potential problems of outliers.


Table 4.1 reports the descriptive statistics for all variables used in the discretionary

accruals (Panel A) and going-concern opinions (Panel B) analyses during the 2004 to

2014 period.

Table 4.2 reports univariate tests of the differences in means of the accrual-based

audit quality proxies between two auditor groups: Big 4 and non-Big 4 firms (Column 1),

PwC and other Big 4 firms (Column 2), EY and other Big 4 firms (Column 3) and specialists

and non-specialists (Column 4). Consistent with the results in auditing research (e.g.,

Becker et al. 1998; Francis et al. 1999a; Krishnan 2003), the detailed yearly results

87

indicate that both the magnitude of discretionary accruals (ADA_MJR) and the standard

deviation of residuals (SD_FLOS) are significantly smaller for clients of Big 4 firms at less

than 1% levels in Panels A and B, respectively. While the differences in means for the

other pairs of auditor attributes appear to be weaker and sensitive to the period, I find that

the discretionary accruals are consistently smaller in magnitude for the subsamples of

PwC auditors (Column 2) and industry specialists (Column 4) but consistently larger in

magnitude for the subsample of EY auditors (Column 3) across almost all fiscal years,

which raises questions concerning the reliability and validity of these industry specialist

measures, as previously documented in Audousset-Coulier et al. (2015).


Table 4.3 reports the regression results used to test Hypothesis 1. Instead of

treating Big 4 firms as a homogeneous set of firms, I break down the indicator variable of

Big 4 auditors into four individual firm indicator variables and consistently find that the

coefficients on each individual firm indicator variable are negative and significant at less

than 1% levels. However, I find that the differences between the estimated coefficients of

PwC and EY are negative and significant (Coef. = -0.002 with t-statistic = -1.73 and -1.82,

when the dependent variable is ADA_MJR and SD_FLOS, respectively), as reported in

Columns (1) and (2). I also find that the difference between the estimated coefficients of

PwC and KPMG is negative and significant (Coef. = -0.003 with t-statistic = -2.76, when

the dependent variable is SD_FLOS). Considered together, these results indicate that

variations across individual auditors leads to variations in audit quality.

Columns (3) and (4) present the results using the sample of Big 4 firms to focus

on quality differentiation within the Big 4 firms. The significantly positive coefficients on EY

(Coef. = 0.002 with t-statistic = 1.71 and 1.84, when the dependent variable is ADA_MJR

and SD_FLOS, respectively) and KPMG (Coef. = 0.003 with t-statistic = 3.08, when the

dependent variable is SD_FLOS) further suggest that finer partitioning of the Big 4 firms

helps to address questions of audit quality differentiation within the Big 4 firms.

As I have argued, the prevalence of PwC and EY across various industry specialist

measures might be a potential confounder that correlates both earnings quality

88

(dependent variable) and auditor industry specialization measure36 (independent variable)

in the earnings quality model. This is because PwC and EY were most often identified as

industry specialists throughout the 2004–2014 period, as graphically illustrated in 4.8

Figure and tables

Figure 4.1.

Table 4.4 presents the accrual-based regression results used for the test of H2 by

each measure of industry specialization: AF_DOM (Panel A) and AT_DOM (Panel B). In

Panel A, while the estimated coefficient on SPEC is either nonsignificant (Column 1, when

the dependent variable is ADA_MJR) or marginally significant (Column 3, when the

dependent variable is SD_FLOS), I find that the estimated coefficient on SPEC – EY is

positive and significant (Coef. = 0.002 with t-statistic = 1.78, when the dependent variable

is ADA_MJR) and that the estimated coefficient on SPEC – PwC is negative and

significant (Coef. = -0.002 with t-statistic = 2.42, when the dependent variable is

SD_FLOS).

Again, as reported in Panel B, while the estimated coefficients on SPEC are

nonsignificant, the estimated coefficients on SPEC – PwC are negative and significant

(Coef. = -0.002 with t-statistic = 1.96, when the dependent variable is ADA_MJR; Coef. =

-0.003 with t-statistic = 2.62, when the dependent variable is SD_FLOS). Thus, it is likely

that the negative relationship between earnings quality and industry specialization is

confounded by the individual audit firms’ competencies, particularly the PwC effect.

As an additional robustness check, I construct a logistic model to predict the

issuance of a going-concern opinion and report the estimation results in Table 4.5.

Because Big 4 clients are generally in better financial condition (e.g., Carson et al. 2012),

I consistently find that the coefficients on each of the Big 4 firms (Column 1) and the Big

4 auditors (Column 2) are negative and significant at less than 1% levels. However, I

reveal that the negative coefficient of PwC (EY) is relatively smaller (larger) than those of

the other Big 4 firms. In addition, while SPEC is positive but nonsignificant (Column 2), I

find that only the estimated coefficient on SPEC – PwC is positive and significant (Coef. =

36 In untabulated results, I find that the Spearman correlation coefficients between an indicator

variable of PWC and various industry specialization measures range from 0.35 to 0.45 and are

significant at less than 1% levels.

89

0.26 with p-value = 0.02), indicating that PwC specialists are more likely to issue a GC

opinion (Column 3).

In Panel B, I focus on the severely distressed subsample, and the results

consistently show that the negative impact of EY is relatively larger than those of the other

Big 4 firms. Additionally, while the estimated coefficient on SPEC is marginally significant

(Column 2), the estimated coefficients on SPEC – PwC (Coef. = 0.28 with p-value = 0.04)

and SPEC – Others (Coef. = 0.48 with p-value = 0.03) are positive and significant in the

severely distressed subsample, as reported in Column (3). Together, these regression

results indicate that the estimated effect of SPEC is confounded by the negative impact of

EY on the likelihood of issuing a GC audit opinion and that there is significant variation in

audit outcomes within the Big 4 firms.


4.6.1 Analysis using signed discretionary accruals

While the magnitude of discretionary accruals is widely used in the accounting

literature as a proxy for earnings quality, some researchers argue that the use of signed

or absolute measures of discretionary accruals may yield different results and suggest

utilizing signed accruals rather than absolute accruals (e.g., DeFond et al. 2017a; Hribar

and Nichols 2007). I report the results using DA_MJR as the dependent variable and

further partition the sample into income-increasing (positive accruals) and income-

decreasing (negative accruals) earnings management subsamples in Table 4.6.

As reported in Panel A, I find that the coefficients on each individual firm indicator

variable are positive (negative) and significant at less than 1% levels in the negative

accruals (positive accruals) subsample. I also find that the differences between the

estimated coefficients of EY and DELT (Coef. = -0.003 with t-statistic = -1.88 for the

negative accruals subsample) and PwC and DELT (Coef. = -0.002 with t-statistic = -1.71

for the positive accruals subsample) are negative and significant. In Panel B, I find that

the estimated coefficient on SPEC – Others is positive and significant (Coef. = 0.003 with

t-statistic = 1.70) in the income-increasing subsample. On the other hand, the coefficient

on SPEC – EY is negative and significant (Coef. = -0.004 with t-statistic = -2.32) in the

income-decreasing subsample.

90

In line with previous analyses, these results further suggest that variations across

individual auditors lead to variations in audit quality at the inter-firm level.

4.6.2 Analysis using industry specialization at the MSA city level

Because there is another stream of research that focuses on auditor industry

specialization at the MSA city level (e.g., Ferguson et al. 2003; Francis et al. 2005b), I

further introduce an alternative industry specialization proxy at the MSA city level to

examine the effects of industry specialization together with the PwC effect. Specifically, I

define cities using the U.S. Census Bureau definition of MSAs to identify metropolitan

areas based on the 5-digit zip codes available in Compustat for each audit client and

obtain industry specialist proxies at the MSA city level. Again, I repeat the same set of

analyses and find that the inferences are virtually unchanged when the dependent variable

is ADA_MJR, thus supporting the notion that the association between earnings quality and

industry specialization is confounded by the individual audit firms’ competencies.

4.7 Conclusion

This paper extends the auditor size literature by investigating whether audit quality

differentiation occurs at the inter-audit firm level. While a homogenous level of audit quality

is generally assumed among large audit firms, I find that there are subtle variations in audit

quality, as measured by either accruals quality or the propensity to issue a GC opinion,

within the Big 4 firms and that the association between individual audit firms and audit

quality appears to be stronger for PwC clients but weaker for EY clients. More importantly,

I reveal that the impact of industry specialization on audit quality is confounded by

individual firms’ competencies, since the existing measures of industry specialization

heavily identify either PwC or EY as an industry specialist in each industry throughout the

sample period.

This paper makes several important contributions to the audit quality literature.

First, I raise the concern of whether it is appropriate to treat all Big 4 firms as a

homogenous group and further argue that not all Big 4 audit firms are the same. Evidence

on individual auditor characteristics, primarily for PwC and EY, is consistent with individual

auditor competencies having a significant effect on improving audit quality. Second, I

provide an explanation for the inconclusive results regarding the validity of industry

91

specialization measures in archival auditing research by showing that the confounding

effect of individual differences within the Big 4 firms leads to significantly different results

regarding the impact of industry specialization on audit quality.

Finally, to the extent that external auditors are held to the same professional

standards and are subject to the same regulations, an individual audit firm’s competencies

likely play an important role in providing high audit quality.

92

4.8 Figure and tables

Figure 4.1 Assignments of industry specialization across major audit firms

Frequencies (in percentage) of industries in which major audit firms are identified as industry specialists across two assignment approaches (AF_DOM and AT_DOM) from 2004 to 2014.

Panel A – Dominant market share approach based on aggregated audit fees (AF_DOM)

Panel B – Dominant market share approach based on aggregated client assets (AT_DOM)

Note:

1. AF_DOM (AT_DOM) = indicator variable equals to one if the company uses an audit firm that has either the largest market share based on aggregated audit fees (or aggregated client assets for AT_DOM) or the second highest market share in industries where the difference between the market shares of the leader and the second-highest firm is less than 10 percent, and zero otherwise

0%

10%

20%

30%

40%

50%

60%

04 05 06 07 08 09 10 11 12 13 14

Fiscal Year

PwC

EY

Others

0%

10%

20%

30%

40%

50%

04 05 06 07 08 09 10 11 12 13 14

Fiscal Year

PwC

EY

Others

93


These tables report the summary statistics of variables used for the earnings quality model (Panel

A) and the GC opinions model (Panel B) from the fiscal years 2004 to 2014. All continuous variables

are winsorized at the 1st and 99th percentiles.

Panel A – Variable definitions for earnings quality model

ADA_MJR = the unsigned magnitude of discretionary accruals based on the modified Jones model

while controlling for firm’s financial performance (Kothari et al. 2005); DA_MJR = the signed

discretionary accruals; SD_FLOS = the standard deviation of the residuals over years t-4 through

t based on the modified Dechow and Dichev model (e.g., Dechow and Dichev 2002; Francis et al.

2005a); SIZE = the natural logarithm of the firm’s market value at the end of fiscal year; SD_SALE

= the standard deviation of total sales over the last five years; SD_CFO = the standard deviation

of the operating cash flow over the last five years; CFO = the operating cash flow, scaled by lagged

total assets; LEVERAGE = the sum of short-term and long-term debt, divided by average total

assets; LOSS = indicator variable equals to one if income before extraordinary items is negative in

the current period, and zero otherwise; MTB = the firm’s market value divided by its book value;

HIGHLIT = indicator variable equals to one for high litigation risk industries as defined in Francis et

al. (1994), and zero otherwise; ALTMAN = the Altman z-score; ACCR_LAG = total lag accruals,

scaled by lagged total assets; BIG4 = indicator variable equals to one if the firm’s auditor is a

member of the Big 4 (PwC, EY, KPMG and Deloitte), and zero otherwise.

Variables N Mean Std. Dev. Median Q1 Q3 ADA_MJR 41,167 0.061 0.076 0.035 0.015 0.073

SD_FLOS 30,595 0.049 0.043 0.036 0.020 0.064

SIZE 41,167 5.846 2.329 5.971 4.183 7.503

SD_SALE 41,167 0.208 0.233 0.134 0.068 0.253

SD_CFO 41,167 0.152 0.344 0.060 0.032 0.122

CFO 41,167 0.021 0.265 0.076 0.001 0.137

LEVERAGE 41,167 0.247 0.282 0.186 0.013 0.359

LOSS 41,167 0.369 0.482 0.000 0.000 1.000

MTB 41,167 2.152 1.825 1.598 1.166 2.342

HIGHLIT 41,167 0.346 0.476 0.000 0.000 1.000

ALTMAN 41,167 2.724 8.504 2.692 1.176 4.944

ACCR_LAG 41,167 -0.045 0.188 -0.019 -0.073 0.023

BIG4 41,167 0.675 0.468 1.000 0.000 1.000

94


Panel B – Variable definitions for auditors’ GC opinions model

GC = indicator variable equals to one if a firm receives a going-concern report in a fiscal period,

and zero otherwise; LNASSET = the natural logarithm of total assets (in millions); MTB = the firm’s

market value divided by its book value; LEVERAGE = the sum of short-term and long-term debt,

divided by average total assets; CH_LEV = the difference between current and prior year leverage;

CFO = the operating cash flow, scaled by lagged total assets; ALTMAN = the Altman z-score;

PLOSS = indicator variable equals to one if income before extraordinary items is negative in the

prior period, and zero otherwise; HIGHLIT = indicator variable equal to one for high litigation risk

industries as defined in Francis et al. (1994), and zero otherwise; TENURE = the number of years

that an audit client has been audited by the same audit firm; ROA = income before extraordinary

items, scaled by average total assets; SD_OIADP = the standard deviation of income before

extraordinary items over the last five years; BIG4 = indicator variable equal to one if the firm’s

auditor is a member of the Big 4 (PwC, EY, KPMG and Deloitte), and zero otherwise.

Full sample Severely distressed firms

(N = 40,920) (N = 8,653)

Variables Mean Std. Dev. Median Mean Std. Dev. Median

GC 0.07 0.26 0.00 0.26 0.44 0.00

LNASSET 5.83 2.32 5.89 3.70 1.79 3.64

LEVERAGE 0.25 0.28 0.19 0.28 0.39 0.12

CH_LEV 0.00 0.12 0.00 0.03 0.19 0.00

CFO 0.02 0.26 0.08 -0.32 0.37 -0.17

ALTMAN 2.71 8.44 2.70 -2.24 13.80 0.04

PLOSS 0.37 0.48 0.00 0.88 0.33 1.00

HIGHLIT 0.35 0.48 0.00 0.51 0.50 1.00

TENURE 8.69 8.15 6.00 6.10 5.59 5.00

ROA -0.06 0.31 0.03 -0.46 0.46 -0.30

SD_OIADP 0.19 0.59 0.05 0.53 1.03 0.17

BIG4 0.67 0.47 1.00 0.43 0.50 0.00

95

Table 4.2 Analysis using accrual-based audit quality proxies

Panel A Differences in means of the absolute value of discretionary accruals controlling for firm’s financial performance (ADA_MJR)

Note: SPEC = Dominant market share approach based on aggregated audit fees (AF_DOM)

FISCAL

YEAR

Full Sample Sample of Big 4 firms

(1)

Difference

B4 - Non-B4 t-Stat

(2)

Difference

PwC - Others t-Stat

(3)

Difference

EY - Others t-Stat

(4)

Difference

SPEC – Others t-Stat

2004 -0.052 -15.08 *** -0.003 -1.30

0.002 0.67

-0.004 -1.92 *

2005 -0.046 -15.39 *** -0.007 -2.96 *** 0.003 1.26

-0.003 -1.57

2006 -0.053 -15.90 *** -0.006 -2.52 ** 0.008 3.06 *** -0.001 -0.29

2007 -0.049 -15.67 *** -0.004 -1.56

0.007 2.84 *** 0.004 1.72 *

2008 -0.040 -15.08 *** -0.006 -2.82 *** 0.007 3.08 *** -0.004 -1.85 *

2009 -0.040 -14.40 *** -0.007 -2.95 *** 0.008 3.22 *** -0.004 -1.67 *

2010 -0.043 -15.05 *** -0.003 -1.23

0.005 1.82 * -0.005 -2.28 **

2011 -0.039 -12.87 *** -0.002 -0.62

0.004 1.53

-0.004 -1.67 *

2012 -0.038 -13.03 *** -0.003 -1.24

0.005 2.12 ** -0.004 -1.70 *

2013 -0.043 -13.13 *** -0.005 -2.18 ** 0.007 2.50 ** -0.004 -1.85 *

2014 -0.043 -13.21 *** -0.005 -1.94 * 0.007 2.74 *** -0.001 -0.39


96


Panel B Differences in means of the standard deviation of residuals from the modified Dechow and Dichev model (SD_FLOS)

Note: SPEC = Dominant market share approach based on aggregated audit fees (AF_DOM)

FISCAL

YEAR

Full Sample Sample of Big 4 firms

(1)

Difference

B4 - Non-B4 t-Stat

(2)

Difference

PwC - Others t-Stat

(3)

Difference

EY - Others t-Stat

(4)

Difference

SPEC – Others t-Stat

2004 -0.037 -17.31 *** -0.003 -1.70 * 0.002 1.19

-0.003 -2.19 **

2005 -0.041 -20.67 *** -0.005 -3.40 *** 0.003 1.71 * -0.006 -4.03 ***

2006 -0.040 -20.72 *** -0.006 -3.63 *** 0.002 1.24

-0.005 -3.74 ***

2007 -0.037 -19.90 *** -0.004 -2.80 *** 0.005 3.24 *** -0.003 -1.94 *

2008 -0.033 -17.65 *** -0.004 -2.66 *** 0.005 3.53 *** -0.003 -2.10 **

2009 -0.036 -17.50 *** -0.006 -4.14 *** 0.007 3.95 *** -0.003 -2.04 **

2010 -0.034 -17.00 *** -0.005 -3.56 *** 0.006 3.53 *** -0.004 -2.75 ***

2011 -0.031 -15.88 *** -0.005 -3.13 *** 0.006 3.68 *** -0.006 -3.75 ***

2012 -0.031 -16.38 *** -0.004 -2.71 *** 0.006 3.61 *** -0.007 -4.57 ***

2013 -0.033 -17.18 *** -0.005 -3.40 *** 0.005 3.06 *** -0.005 -3.78 ***

2014 -0.031 -16.64 *** -0.004 -3.00 *** 0.005 3.21 *** -0.006 -4.35 ***


97

Table 4.3 Earnings quality model

This table presents regression results of the earnings quality model for the full sample (Columns 1 and 2) and the sample of Big 4 audit firms

(Columns 3 and 4). *, **, *** denote significance at the 0.10, 0.05, and 0.01 levels, respectively, using two-tailed tests. T-statistic is determined by

clustered standard errors at firm level.

(1) (2) (3) (4)

Full Sample Full Sample Big 4 audit firms Big 4 audit firms

DV = ADA_MJR DV = SD_FLOS DV = ADA_MJR DV = SD_FLOS


Intercept 0.109 *** 13.91 0.083 *** 14.76 0.086 *** 7.81 0.073 *** 11.16

SIZE -0.005 *** -17.48 -0.004 *** -18.79 -0.004 *** -14.96 -0.004 *** -15.93

SD_SALE 0.032 *** 11.36 0.053 *** 21.79 0.024 *** 7.27 0.046 *** 15.40

SD_CFO 0.016 *** 6.41 0.031 *** 8.93 0.024 *** 5.10 0.033 *** 5.30

CFO -0.034 *** -8.82 -0.013 *** -4.95 -0.014 *** -2.62 -0.011 *** -3.03

LEVERAGE -0.001 -0.57 -0.004 ** -2.24 -0.008 *** -3.35 -0.009 *** -4.58

LOSS -0.005 *** -4.96 0.004 *** 4.99 0.001 1.13 0.005 *** 6.47

MTB 0.007 *** 15.96 0.004 *** 12.42 0.007 *** 12.19 0.003 *** 7.56

HIGHLIT -0.001 -0.34 0.004 *** 2.97 -0.002 -1.09 0.003 * 1.95

ALTMAN -0.001 *** -8.36 -0.001 *** -8.43 -0.001 *** -5.25 -0.001 *** -5.28

ACCR_LAG -0.010 ** -2.47 -0.002 -0.53 -0.012 ** -1.97 -0.003 -0.72

PwC -0.013 *** -9.17 -0.009 *** -7.58

EY -0.011 *** -7.64 -0.007 *** -5.97 0.002 * 1.71 0.002 * 1.84

KPMG -0.011 *** -7.63 -0.006 *** -4.90 0.002 1.53 0.003 *** 3.08

DELT -0.012 *** -8.27 -0.007 *** -5.96 0.001 0.95 0.001 1.49

Fixed Effects Yes Yes Yes Yes

Observations 41,167 30,595 27,777 21,639


Difference | PWC - EY -0.002 * -1.73 -0.002 * -1.82 Difference | PWC - KPMG -0.002 -1.48 -0.003 *** -2.76 Difference | PWC - DELT -0.001 -0.82 -0.001 -1.59 Difference | EY - KPMG 0.000 0.12 -0.001 -1.12 Difference | EY - DELT 0.001 0.83 0.000 0.13

98

Table 4.4 Earnings quality model: Estimation of industry specialist effect

These tables present regression results of the earnings quality model with the inclusion of auditor industry specialization measures (SPEC). The

dominant market shares based on aggregated audit fees (AF_DOM) and aggregated client assets (AT_DOM) are used to identify industry specialists

at the national level in Panels A and B, respectively. *, **, *** denote significance at the 0.10, 0.05, and 0.01 levels, respectively, using two-tailed

tests. T-statistic is determined by clustered standard errors at firm level.

Panel A SPEC = Dominant market share approach based on aggregated audit fees (AF_DOM)

(1) (2) (3) (4)

DV = ADA_MJR DV = ADA_MJR DV = SD_FLOS DV = SD_FLOS

SPEC = AF_DOM SPEC = AF_DOM SPEC = AF_DOM SPEC = AF_DOM


Intercept 0.109 *** 13.95 0.109 *** 13.89 0.083 *** 14.70 0.082 *** 14.64

SIZE -0.005 *** -17.61 -0.005 *** -17.54 -0.004 *** -18.78 -0.004 *** -18.73

SD_SALE 0.032 *** 11.35 0.032 *** 11.35 0.053 *** 21.79 0.053 *** 21.78

SD_CFO 0.016 *** 6.40 0.016 *** 6.40 0.031 *** 8.94 0.031 *** 8.94

CFO -0.034 *** -8.83 -0.034 *** -8.81 -0.013 *** -4.93 -0.013 *** -4.91

LEVERAGE -0.001 -0.58 -0.001 -0.58 -0.004 ** -2.25 -0.004 ** -2.25

LOSS -0.005 *** -4.97 -0.005 *** -4.97 0.004 *** 5.03 0.004 *** 5.03

MTB 0.007 *** 15.98 0.007 *** 15.97 0.004 *** 12.44 0.004 *** 12.45

HIGHLIT -0.001 -0.34 -0.001 -0.31 0.004 *** 2.91 0.004 *** 2.94

ALTMAN -0.001 *** -8.35 -0.001 *** -8.34 -0.001 *** -8.43 -0.001 *** -8.43

ACCR_LAG -0.010 ** -2.46 -0.010 ** -2.47 -0.002 -0.55 -0.002 -0.56

BIG4 -0.012 *** -9.29 -0.012 *** -9.29 -0.007 *** -6.26 -0.007 *** -6.23

SPEC 0.001 1.13 -0.001 * -1.75 SPEC - PwC -0.001 -0.96 -0.002 ** -2.42

SPEC - EY 0.002 * 1.78 -0.001 -0.99

SPEC - Others 0.002 1.63 0.000 0.50


Observations 41,167 41,167 30,595 30,595


99


Panel B SPEC = Dominant market share approach based on aggregated client assets (AT_DOM)

(1) (2) (3) (4)

DV = ADA_MJR DV = ADA_MJR DV = SD_FLOS DV = SD_FLOS

SPEC = AT_DOM SPEC = AT_DOM SPEC = AT_DOM SPEC = AT_DOM


Intercept 0.109 *** 13.94 0.109 *** 13.91 0.083 *** 14.70 0.082 *** 14.66

SIZE -0.005 *** -17.51 -0.005 *** -17.46 -0.004 *** -18.82 -0.004 *** -18.79

SD_SALE 0.032 *** 11.35 0.032 *** 11.35 0.053 *** 21.79 0.053 *** 21.77

SD_CFO 0.016 *** 6.40 0.016 *** 6.40 0.031 *** 8.93 0.031 *** 8.93

CFO -0.034 *** -8.82 -0.033 *** -8.81 -0.013 *** -4.92 -0.013 *** -4.92

LEVERAGE -0.001 -0.59 -0.001 -0.58 -0.004 ** -2.25 -0.004 ** -2.25

LOSS -0.005 *** -4.95 -0.005 *** -4.95 0.004 *** 5.00 0.004 *** 5.01

MTB 0.007 *** 15.97 0.007 *** 15.96 0.004 *** 12.44 0.004 *** 12.45

HIGHLIT -0.001 -0.35 -0.001 -0.33 0.004 *** 2.90 0.004 *** 2.93

ALTMAN -0.001 *** -8.35 -0.001 *** -8.34 -0.001 *** -8.43 -0.001 *** -8.43

ACCR_LAG -0.010 ** -2.47 -0.010 ** -2.47 -0.002 -0.54 -0.002 -0.54

BIG4 -0.012 *** -9.04 -0.012 *** -9.05 -0.007 *** -6.45 -0.007 *** -6.42

SPEC 0.000 -0.14 -0.001 -1.46 SPEC - PwC -0.002 ** -1.96 -0.003 *** -2.62

SPEC - EY 0.002 1.24 -0.001 -0.84

SPEC - Others 0.001 0.55 0.001 0.89


Observations 41,167 41,167 30,595 30,595


100

Table 4.5 Auditors’ going-concern opinion model

These tables present logistic regression results of the GC opinions model with the inclusion of

individual audit firm (column 1) and auditor industry specialization measures (columns 2 and 3) for

the full sample (Panel A) and the sample of firms with severely financially distressed firms (Panel

B). The dominance market shares approach based on aggregated audit fees (AF_DOM) is used to

identify industry specialists at the national level. *, **, *** denote significance at the 0.10, 0.05, and

0.01 levels, respectively, using two-tailed tests. T-statistic is determined by clustered standard

errors at firm level.

Panel A – The Full Sample

(1) (2) (3)

DV = GC DV = GC DV = GC

SPEC = AF_DOM SPEC = AF_DOM SPEC = AF_DOM

Variables Coef. p-value Coef. p-value Coef. p-value

Intercept -2.53 0.92 -2.52 0.92 -2.52 0.92

LNASSET -0.46 *** <.0001 -0.46 *** <.0001 -0.46 *** <.0001

MTB -0.15 *** <.0001 -0.15 *** <.0001 -0.15 *** <.0001

LEVERAGE 1.46 *** <.0001 1.46 *** <.0001 1.47 *** <.0001

CH_LEV -0.44 *** 0.00 -0.44 *** 0.00 -0.45 *** 0.00

CFO -0.19 * 0.07 -0.18 * 0.08 -0.18 * 0.08

ALTMAN -0.04 *** <.0001 -0.04 *** <.0001 -0.04 *** <.0001

PLOSS 1.33 *** <.0001 1.33 *** <.0001 1.33 *** <.0001

HIGHLIT -0.13 0.15 -0.13 0.16 -0.14 0.14

TENURE 0.01 *** 0.00 0.01 *** 0.00 0.01 *** 0.00

ROA -1.58 *** <.0001 -1.58 *** <.0001 -1.58 *** <.0001

SD_OIADP 0.23 *** <.0001 0.23 *** <.0001 0.24 *** <.0001

PwC -0.21 ** 0.05 EY -0.51 *** <.0001 KPMG -0.35 *** 0.00 DELT -0.23 ** 0.03

BIG4 -0.38 *** <.0001 -0.38 *** <.0001

SPEC 0.08 0.35

SPEC - PwC 0.26 ** 0.02

SPEC - EY -0.10 0.45

SPEC - Others -0.01 0.96


Observations 40,918 40,918 40,918

Pseudo R-squared 52.8% 52.8% 52.8%

101


Panel B – The sample of firms with severely financially distressed firms

(1) (2) (3)

DV = GC DV = GC DV = GC

SPEC = AF_DOM SPEC = AF_DOM SPEC = AF_DOM

Variables Coef. p-value Coef. p-value Coef. p-value

Intercept -0.91 0.80 -0.90 0.81 -0.91 0.80

LNASSET -0.38 *** <.0001 -0.38 *** <.0001 -0.38 *** <.0001

MTB -0.10 *** <.0001 -0.10 *** <.0001 -0.10 *** <.0001

LEVERAGE 1.14 *** <.0001 1.14 *** <.0001 1.14 *** <.0001

CH_LEV -0.34 * 0.05 -0.34 * 0.05 -0.34 * 0.06

CFO 0.24 * 0.07 0.24 * 0.06 0.24 * 0.06

ALTMAN -0.03 *** <.0001 -0.03 *** <.0001 -0.03 *** <.0001

PLOSS 0.47 *** <.0001 0.47 *** <.0001 0.47 *** <.0001

HIGHLIT -0.18 0.11 -0.17 0.12 -0.18 0.11

TENURE 0.02 *** 0.01 0.02 *** 0.01 0.02 *** 0.01

ROA -1.46 *** <.0001 -1.47 *** <.0001 -1.47 *** <.0001

SD_OIADP 0.20 *** <.0001 0.20 *** <.0001 0.20 *** <.0001

PwC -0.15 0.25 EY -0.54 *** <.0001 KPMG -0.36 ** 0.01 DELT -0.04 0.76

BIG4 -0.39 *** <.0001 -0.38 *** <.0001

SPEC 0.19 * 0.09

SPEC - PwC 0.28 ** 0.04

SPEC - EY -0.13 0.48

SPEC - Others 0.48 ** 0.03


Observations 8,652 8,652 8,652

Pseudo R-squared 44.2% 44.0% 44.1%

102

Table 4.6 Differences in earnings quality among Big 4 audit firms using signed discretionary accruals

These tables present regression results of the earnings quality model using signed discretionary accruals (DA_MJR) with the inclusion of individual

audit firm (Panel A) and auditor industry specialization measures (Panel B). The dominant market shares approach based on aggregated audit fees

(AF_DOM) is used to identify industry specialists at the national level. *, **, *** denote significance at the 0.10, 0.05, and 0.01 levels, respectively,

using two-tailed tests. T-statistic is determined by clustered standard errors at firm level.

Panel A Panel B

DV = DA_MJR DV = DA_MJR

Negative Accruals Positive Accruals Negative Accruals Positive Accruals

Variables Coef. t-stat. Coef. t-stat. Variables Coef. t-stat. Coef. t-stat.

Intercept -0.098 *** -11.39 0.109 *** 12.36 Intercept -0.098 *** -11.41 0.108 *** 12.29

SIZE 0.005 *** 15.74 -0.004 *** -12.39 SIZE 0.005 *** 15.82 -0.004 *** -12.37

SD_SALE -0.021 *** -7.17 0.036 *** 10.03 SD_SALE -0.021 *** -7.17 0.036 *** 10.04

SD_CFO -0.017 *** -6.33 0.011 *** 3.45 SD_CFO -0.017 *** -6.33 0.011 *** 3.45

CFO -0.019 *** -4.16 -0.082 *** -16.53 CFO -0.019 *** -4.16 -0.082 *** -16.52

LEVERAGE 0.008 *** 3.41 0.007 ** 2.46 LEVERAGE 0.008 *** 3.40 0.007 ** 2.45

LOSS -0.012 *** -9.70 -0.022 *** -15.36 LOSS -0.012 *** -9.68 -0.022 *** -15.35

MTB -0.007 *** -16.49 0.004 *** 7.98 MTB -0.007 *** -16.51 0.004 *** 7.99

HIGHLIT -0.001 -0.28 -0.002 -0.77 HIGHLIT -0.001 -0.35 -0.002 -0.79

ALTMAN 0.001 *** 11.24 0.000 -0.80 ALTMAN 0.001 *** 11.23 0.000 -0.79

ACCR_LAG 0.017 *** 3.87 0.003 0.52 ACCR_LAG 0.017 *** 3.87 0.003 0.52

PwC 0.008 *** 4.48 -0.016 *** -9.12 BIG4 0.008 *** 4.96 -0.014 *** -9.01

EY 0.006 *** 3.13 -0.015 *** -8.72 SPEC - PwC 0.000 0.25 -0.002 -1.14

KPMG 0.006 *** 3.79 -0.014 *** -7.50 SPEC - EY -0.004 ** -2.32 0.000 -0.36

DELT 0.009 *** 5.01 -0.014 *** -7.60 SPEC - Others -0.001 -0.59 0.003 * 1.70

Observations 20,699 20,468 Observations 20,699 20,468

Adj. R-squared 22.2% 25.2% Adj. R-squared 22.2% 25.2%

Diff. | EY – DELT -0.003 * -1.88 Note: SPEC = AF_DOM

Diff. | PwC – DELT -0.002 * -1.71

103

Chapter 5. Conclusion

Over the years, the role of external auditors has been examined to assess its

relevancy considering changing market practices and investor information needs. While

evidence in the auditing literature shows that audits have value in providing reasonable

assurance regarding whether financial statements comply with GAAP on a pass-fail basis,

this dissertation provides new evidence on the influential role of external auditors in

enhancing the informativeness of financial disclosures.

The results in Chapter 2 indicate that in addition to managerial discretion, the

choice of Big 4 auditors is attributable to variations in 10-K disclosure volume. I also argue

that the residual disclosure of 10-K reports induces external auditors to increase their audit

effort to reduce their exposure to litigation risk, as evidenced by higher audit fees and the

increased likelihood of GC opinions. Therefore, I argue that audit effort can be inferred

from a discretionary component of 10-K disclosure volume.

In addition to the general effect of auditor size, I further argue that there is cross-

sectional variation in both audit fees (Chapter 3) and audit quality (Chapter 4) among the

Big 4 firms. Consistent with prior studies that argue that there is now more evidence of a

PwC audit fee premium, I reveal that PwC earns an above-average Big 4 premium and

that the firm delivers higher audit quality relative to the other Big 4 audit firms. This result

is consistent with evidence from research on auditor industry specialization, which

commonly designates PwC as an industry specialist due to its leadership position across

most industries in the U.S. market. Furthermore, I argue that the effects of industry

expertise on audit fees and on audit quality are exaggerated by the confounding effect of

audit firms’ competencies, primarily the effects of PwC and EY, thus raising more doubts

regarding the robustness of prior empirical results found in auditor industry specialization

research. Together, the results in Chapter 3 and 4 reinforce the importance of individual

audit firm competencies and suggest that not all Big 4 firms are the same.

104

References

[1] AICPA. 1997. Au section 550 other information in documents containing audited financial statements. New York, NY: AICPA.

[2] Audousset-Coulier, S., A. Jeny, and L. Jiang. 2015. The validity of auditor industry specialization measures. Auditing: A Journal of Practice & Theory 35 (1):139-161.

[3] Baginski, S. P., J. M. Hassell, and M. D. Kimbrough. 2004. Why do managers explain their earnings forecasts? Journal of Accounting Research 42 (1):1-29.

[4] Balakrishnan, K., M. B. Billings, B. Kelly, and A. Ljungqvist. 2014. Shaping liquidity: On the causal effects of voluntary disclosure. The Journal of Finance 69:2237-2278.

[5] Balsam, S., J. Krishnan, and J. S. Yang. 2003. Auditor industry specialization and earnings quality. Auditing: A Journal of Practice & Theory 22 (2):71-97.

[6] Becker, C. L., M. L. DeFond, J. Jiambalvo, and K. Subramanyam. 1998. The effect of audit quality on earnings management. Contemporary Accounting Research 15 (1):1-24.

[7] Behn, B. K., J. H. Choi, and T. Kang. 2008. Audit quality and properties of analyst earnings forecasts. The Accounting Review 83 (2):327-349.

[8] Bernstein, P. W. 1978. Competition comes to accounting. Fortune 98 (1):88-&.

[9] Blay, A. D., and M. A. Geiger. 2013. Auditor fees and auditor independence: Evidence from going concern reporting decisions. Contemporary Accounting Research 30 (2):579-606.

[10] Bloomfield, R. 2008. Discussion of “annual report readability, current earnings, and earnings persistence”. Journal of Accounting and Economics 45 (2):248-252.

[11] Bloomfield, R. J. 2002. The “incomplete revelation hypothesis” and financial reporting. Accounting Horizons 16 (3):233-243.

[12] Boone, J. P., I. K. Khurana, and K. Raman. 2010. Do the big 4 and the second-tier firms provide audits of similar quality? Journal of Accounting and Public Policy 29 (4):330-352.

[13] Boone, J. P., and K. K. Raman. 2001. Auditor resignations versus dismissals: An examination of the differential effects on market liquidity and trading activity. Advances in Accounting 18:47-75.

[14] Botosan, C. A. 1997. Disclosure level and the cost of equity capital. The Accounting Review 72 (3):323-349.

[15] Cahan, S. F., D. C. Jeter, and V. Naiker. 2011. Are all industry specialist auditors the same? Auditing: A Journal of Practice & Theory 30 (4):191-222.

105

[16] Callaghan, J., M. Parkash, and R. Singhal. 2009. Going-concern audit opinions and the provision of nonaudit services: Implications for auditor independence of bankrupt firms. Vol. 28.

[17] CAQ. 2011. Observations on the evolving role of the auditor: A summary of stakeholder discussions.

[18] Caramanis, C., and C. Lennox. 2008. Audit effort and earnings management. Journal of Accounting and Economics 45 (1):116-138.

[19] Carson, E. 2009. Industry specialization by global audit firm networks. The Accounting Review 84 (2):355-382.

[20] Carson, E., N. L. Fargher, M. A. Geiger, C. S. Lennox, K. Raghunandan, and M. Willekens. 2012. Audit reporting for going-concern uncertainty: A research synthesis. Auditing: A Journal of Practice & Theory 32 (sp1):353-384.

[21] Cazier, R. A., and R. J. Pfeiffer. 2015. Why are 10-k filings so long? Accounting Horizons 30 (1):1-21.

[22] Chang, H., C. S. A. Cheng, and K. J. Reichelt. 2010. Market reaction to auditor switching from big 4 to third-tier small accounting firms. Auditing: A Journal of Practice & Theory 29 (2):83-114.

[23] Chen, K. C. W., and B. K. Church. 1996. Going concern opinions and the market's reaction to bankruptcy filings. The Accounting Review 71 (1):117-128.

[24] Craswell, A. T., J. R. Francis, and S. L. Taylor. 1995. Auditor brand name reputations and industry specializations. Journal of Accounting and Economics 20 (3):297-322.

[25] D'Souza, J., K. Ramesh, and M. Shen. 2010. Disclosure of gaap line items in earnings announcements. Review of Accounting Studies 15 (1):179-219.

[26] DeAngelo, L. E. 1981. Auditor size and audit quality. Journal of Accounting and Economics 3 (3):183-199.

[27] Dechow, P., W. L. Ge, and C. Schrand. 2010. Understanding earnings quality: A review of the proxies, their determinants and their consequences. Journal of Accounting and Economics 50 (2-3):344-401.

[28] Dechow, P. M., and I. D. Dichev. 2002. The quality of accruals and earnings: The role of accrual estimation errors. The Accounting Review 77 (s-1):35-59.

[29] Dechow, P. M., R. G. Sloan, and A. P. Sweeney. 1995. Detecting earnings management. The Accounting Review 70 (2):193-225.

[30] DeFond, M., D. H. Erkens, and J. Zhang. 2017a. Do client characteristics really drive the big n audit quality effect? New evidence from propensity score matching. Management Science 63 (11):3628-3649.

106

[31] DeFond, M., C. Lennox, and J. Zhang. 2017b. Some controversies in the auditing literature. Working Paper.

[32] DeFond, M., and J. Y. Zhang. 2014. A review of archival auditing research. Journal of Accounting and Economics 58 (2-3):275-326.

[33] DeFond, M. L., and C. S. Lennox. 2011. The effect of sox on small auditor exits and audit quality. Journal of Accounting and Economics 52 (1):21-40.

[34] DeFond, M. L., K. Raghunandan, and K. R. Subramanyam. 2002. Do non-audit service fees impair auditor independence? Evidence from going concern audit opinions. Journal of Accounting Research 40 (4):1247-1274.

[35] Diamond, D. W., and R. E. Verrecchia. 1991. Disclosure, liquidity, and the cost of capital. The Journal of Finance 46 (4):1325-1359.

[36] Dunn, K. A., and B. W. Mayhew. 2004. Audit firm industry specialization and client disclosure quality. Review of Accounting Studies 9 (1):35-58.

[37] Eshleman, J. D., and P. Guo. 2014. Do big 4 auditors provide higher audit quality after controlling for the endogenous choice of auditor? Auditing: A Journal of Practice & Theory 33 (4):197-219.

[38] Ferguson, A., J. R. Francis, and D. J. Stokes. 2003. The effects of firm-wide and office-level industry expertise on audit pricing. The Accounting Review 78 (2):429-448.

[39] Ferguson, A., and T. Scott. 2014. What if there were three? Audit pricing within the big 4 and the pricewaterhousecoopers' premium in the australian audit market. International Journal of Auditing 18 (1):57-67.

[40] Fields, L. P., D. R. Fraser, and M. S. Wilkins. 2004. An investigation of the pricing of audit services for financial institutions. Journal of Accounting and Public Policy 23 (1):53-77.

[41] Firth, M., and T. Lau. 2004. Audit pricing following mergers of accounting practices: Evidence from hong kong. Accounting and Business Research 34 (3):201-213.

[42] Francis, J., R. LaFond, P. Olsson, and K. Schipper. 2005a. The market pricing of accruals quality. Journal of Accounting and Economics 39 (2):295-327.

[43] Francis, J., D. Philbrick, and K. Schipper. 1994. Shareholder litigation and corporate disclosures. Journal of Accounting Research 32 (2):137-164.

[44] Francis, J. R., E. L. Maydew, and H. C. Sparks. 1999a. The role of big 6 auditors in the credible reporting of accruals. Auditing: A Journal of Practice & Theory 18 (2):17-34.

[45] Francis, J. R., E. L. Maydew, and H. C. Sparks. 1999b. The role of big 6 auditors in the credible reporting of accruals. Auditing: a Journal of Practice & theory 18 (2):17-34.

107

[46] Francis, J. R., K. Reichelt, and D. Wang. 2005b. The pricing of national and city-specific reputations for industry expertise in the us audit market. The Accounting Review 80 (1):113-136.

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

[48] GAO. 2004. Mandatory audit firm rotation study: GAO Washington, DC.

[49] ———. 2008. Audits of public companies: Continued concentration in audit market for large public companies does not call for immediate action: GAO Washington, DC.

[50] Gibbins, M., S. Salterio, and A. Webb. 2001. Evidence about auditor–client management negotiation concerning client’s financial reporting. Journal of Accounting Research 39 (3):535-563.

[51] Graham, J. R., C. R. Harvey, and S. Rajgopal. 2005. The economic implications of corporate financial reporting. Journal of Accounting and Economics 40 (1-3):3-73.

[52] Hay, D. 2013. Further evidence from meta‐analysis of audit fee research. International Journal of Auditing 17 (2):162-176.

[53] Hay, D. C., W. R. Knechel, and N. Wong. 2006. Audit fees: A meta‐analysis of the effect of supply and demand attributes. Contemporary Accounting Research 23 (1):141-191.

[54] Hendershott, T., C. M. Jones, and A. J. Menkveld. 2011. Does algorithmic trading improve liquidity? The Journal of Finance 66 (1):1-33.

[55] Hribar, P., T. Kravet, and R. Wilson. 2014. A new measure of accounting quality. Review of Accounting Studies 19 (1):506-538.

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

[57] Huang, H. W., L. L. Liu, K. Raghunandan, and D. V. Rama. 2007. Auditor industry specialization, client bargaining power, and audit fees: Further evidence. Auditing: A Journal of Practice & Theory 26 (1):147-158.

[58] Hutton, A. P., G. S. Miller, and D. J. Skinner. 2003. The role of supplementary statements with management earnings forecasts. Journal of Accounting Research 41 (5):867-890.

[59] IAASB. 2014. A framework for audit quality: Key elements that create an environment for audit quality. New York, NY: The International Federation of Accountants (IFAC).

108

[60] Jensen, M. C., and W. H. Meckling. 1976. Theory of the firm: Managerial behavior, agency costs and ownership structure. Journal of Financial Economics 3 (4):305-360.

[61] Jones, J. J. 1991. Earnings management during import relief investigations. Journal of Accounting Research 29 (2):193-228.

[62] Kaplan, S. E., and D. D. Williams. 2013. Do going concern audit reports protect auditors from litigation? A simultaneous equations approach. The Accounting Review 88 (1):199-232.

[63] Khalil, S. K., J. R. Cohen, and G. M. Trompeter. 2011. Auditor resignation and firm ownership structure. Accounting Horizons 25 (4):703-727.

[64] Khurana, I. K., and K. K. Raman. 2004. Litigation risk and the financial reporting credibility of big 4 versus non-big 4 audits: Evidence from anglo-american countries. The Accounting Review 79 (2):473-495.

[65] Knechel, W. R., V. Naiker, and G. Pacheco. 2007. Does auditor industry specialization matter? Evidence from market reaction to auditor switches. Auditing: A Journal of Practice & Theory 26 (1):19-45.

[66] Kothari, S. P., A. J. Leone, and C. E. Wasley. 2005. Performance matched discretionary accrual measures. Journal of Accounting and Economics 39 (1):163-197.

[67] Krishnan, G. V. 2003. Does big 6 auditor industry expertise constrain earnings management? Accounting Horizons 17:1.

[68] Kwon, S. Y. 1996. The impact of competition within the client's industry on the auditor selection decision. Auditing: A Journal of Practice & Theory 15 (1):53-70.

[69] Lang, M., and R. Lundholm. 1993. Cross-sectional determinants of analyst ratings of corporate disclosures. Journal of Accounting Research 31 (2):246-271.

[70] Lawrence, A. 2013. Individual investors and financial disclosure. Journal of Accounting and Economics 56 (1):130-147.

[71] Lawrence, A., M. Minutti-Meza, and P. Zhang. 2011. Can big 4 versus non-big 4 differences in audit-quality proxies be attributed to client characteristics? The Accounting Review 86 (1):259-286.

[72] Lennox, C., and J. A. Pittman. 2010. Big five audits and accounting fraud. Contemporary Accounting Research 27 (1):209-+.

[73] Leuz, C., and R. E. Verrecchia. 2000. The economic consequences of increased disclosure. Journal of Accounting Research 38:91-124.

[74] Li, C., Y. A. Xie, and J. A. Zhou. 2010. National level, city level auditor industry specialization and cost of debt. Accounting Horizons 24 (3):395-417.

109

[75] Li, F. 2008. Annual report readability, current earnings, and earnings persistence. Journal of Accounting and Economics 45 (2-3):221-247.

[76] Lim, C.-Y., and H.-T. Tan. 2010. Does auditor tenure improve audit quality? Moderating effects of industry specialization and fee dependence. Contemporary Accounting Research 27 (3):923-957.

[77] 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.

[78] Loughran, T., and B. Mcdonald. 2014. Measuring readability in financial disclosures. The Journal of Finance 69 (4):1643-1671.

[79] ———. 2016. Textual analysis in accounting and finance: A survey. Journal of Accounting Research 54 (4):1187-1230.

[80] Low, K. Y. 2004. The effects of industry specialization on audit risk assessments and audit-planning decisions. The Accounting Review 79 (1):201-219.

[81] McCracken, S. 2002. An extension of anile and nallebuff's conservatism and auditor-client negotiations: Implementation of auditor's asymmetric loss function. Asian Review of Accounting 10 (1):121-146.

[82] Menon, K., and D. D. Williams. 2010. Investor reaction to going concern audit reports. The Accounting Review 85 (6):2075-2105.

[83] Mercer, M. 2004. How do investors assess the credibility of management disclosures? Accounting Horizons 18 (3):185-196.

[84] Moizer, P. 1997. Auditor reputation: The international empirical evidence. International Journal of Auditing 1 (1):61-74.

[85] Neal, T. L., and R. R. Riley. 2004. Auditor industry specialist research design. Auditing: A Journal of Practice & Theory 23 (2):169-177.

[86] Nelson, K. K., and A. C. Pritchard. 2007. Litigation risk and voluntary disclosure: The use of meaningful cautionary language. Working Paper.

[87] Palmrose, Z. V. 1986. Audit fees and auditor size - further evidence. Journal of Accounting Research 24 (1):97-110.

[88] PCAOB. 2005. Review of existing standards - principles of reporting: PCAOB Washington, DC.

[89] ———. 2015. Concept release on audit quality indicators. Release no. 2015-005.: PCAOB Washington, DC.

[90] Petersen, M. A. 2009. Estimating standard errors in finance panel data sets: Comparing approaches. Review of Financial Studies 22 (1):435-480.

110

[91] Pong, C., and S. Burnett. 2006. The implications of merger for market share, audit pricing and non-audit fee income: The case of pricewaterhousecoopers. Managerial Auditing Journal 21 (1):7-22.

[92] PwC. 2012. Point of view: Competition and choice in the audit market.

[93] 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.

[94] Reynolds, J. K., 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.

[95] Romanus, R. N., J. J. Maher, and D. M. Fleming. 2008. Auditor industry specialization, auditor changes, and accounting restatements. Accounting Horizons 22 (4):389-413.

[96] Shipman, J. E., Q. T. Swanquist, and R. L. Whited. 2017. Propensity score matching in accounting research. The Accounting Review 92 (1):213-244.

[97] Simunic, D. A. 1980. The pricing of audit services - theory and evidence. Journal of Accounting Research 18 (1):161-190.

[98] Simunic, D. A., and M. T. Stein. 1996. Impact of litigation risk on audit pricing: A review of the economics and the evidence. Auditing: A Journal of Practice & Theory 15:119.

[99] Solomon, I., M. D. Shields, and O. R. Whittington. 1999. What do industry-specialist auditors know? Journal of Accounting Research 37 (1):191-208.

[100] Teoh, S. H., and T. J. Wong. 1993. Perceived auditor quality and the earnings response coefficient. The Accounting Review 68 (2):346-366.

[101] Watts, R. L., and J. L. Zimmerman. 1983. Agency problems, auditing, and the theory of the firm - some evidence. Journal of Law & Economics 26 (3):613-633.

[102] Weber, J., and M. Willenborg. 2003. Do expert informational intermediaries add value? Evidence from auditors in microcap initial public offerings. Journal of Accounting Research 41 (4):681-720.

[103] You, H. F., and X. J. Zhang. 2009. Financial reporting complexity and investor underreaction to 10-k information. Review of Accounting Studies 14 (4):559-586.

[104] Zeff, S. A., and R. L. Fossum. 1967. An analysis of large audit clients. The Accounting Review 42 (2):298-320.

111

Appendix A. Supplemental Analysis for Chapter 2

As previously suggested in Shipman et al. (2017), I present the first-stage

estimates used to calculate the propensity score in Table . The pseudo R-squared and the

area under the ROC curve for BIG4 are 58.8% and 0.91, respectively.

Instead of the word count, Loughran and Mcdonald (2014) propose that a gross

file size of 10-K complete submission text files is a better proxy for readability than

previously utilized indices (e.g., the Fog Index) because this measure is easy to replicate

without any document parsing and correlates with alternative readability constructs.

However, in their subsequent literature survey article, Loughran and Mcdonald (2016)

reinforce this definition that although file size might serve a proxy for readability, it cannot

fully separate the fundamental complexity of the firm’s business from the language

complexity of its annual report. Thus, I repeat the same set of analyses using the natural

logarithm of the gross file size of the 10-K complete submission text file (LNSIZEG) as the

dependent variable in Table and Table A..

Consistent with the evidence illustrated in Panel B of 2.8 Figure and tables

Figure 2.1, I find that the effects of Big 4 auditors are insignificantly different from

those of non-Big 4 auditors. I also repeat the same set of analyses using the residual

disclosures (RES_SG) from estimating the disclosure model with the 10-K gross file size

as the dependent variable. Overall, I consistently find that audit clients with abnormally

long disclosures, on average, pay higher audit fees and are more likely to receive a going-

concern audit opinion, as reported in Table A and Table , respectively.

112

Table A.1 First-Stage Prediction Model

This table presents the first-stage prediction used for estimating propensity scores in the auditors’

selection model. *, **, *** denote significance at the 0.10, 0.05, and 0.01 levels, respectively.

(1)

DV = BIG4

Variables Coef. t-stat.

Intercept -6.37 -0.71

LNASSET 0.60 *** 18.52

ATURN 0.09 *** 3.94

CURRENT 0.00 0.47

LEVERAGE -0.31 *** -2.96

ROA -1.05 *** -7.07

DELTA_ROA 0.00 -0.21

DELTA_REV -0.31 *** -8.98

MA -0.10 ** -2.45

FY_RET -0.16 *** -7.08

SD_RETURN -0.85 *** -3.08

SPI_DM 0.16 *** 5.11

CAP_LEASE -0.05 -1.36

OP_LEASE 0.35 *** 8.93

RD 3.36 *** 14.93

INTANG -0.50 *** -6.55

SIZE 0.62 *** 20.42

AGE -0.02 -0.84

MTB -0.06 *** -3.80

FCFLOW 0.74 *** 3.92

DERIVATIVE 0.05 1.13

LNBUSSEG -0.18 *** -4.75

LNGEOSEG 0.12 *** 4.10

SD_OIADP -0.30 *** -4.76

DELAWARE 0.45 *** 14.11

IPO -0.36 -0.92

SEO -0.25 *** -4.58

NMCOUNT 0.17 1.22

Fixed effects Yes

Observations 43,575

Pseudo R-Squared 58.8%

Area under ROC Curve 0.91

113

Table A.2 Auditor choice and 10-K disclosure volume (LNSIZEG)

This table reports the regression results of estimating the disclosure model (Equation 1) on both

the full sample (Column 1) and the PSM sample (Column 2). T-statistic is determined by clustered

standard errors at firm level. *, **, *** denote significance at the 0.10, 0.05, and 0.01 levels,

respectively.

(1) (2)

Full Sample PSM Sample

DV = LNSIZEG DV = LNSIZEG


Intercept 14.89 *** 56.64 15.63 *** 37.86

BIG4 0.03 ** 2.14 0.01 0.91

DELTA_ROA 0.00 -0.11 0.00 0.92

DELTA_REV -0.02 *** -2.94 -0.03 ** -2.06

MA 0.04 *** 4.40 0.04 ** 2.30

FY_RET -0.04 *** -9.56 -0.03 *** -3.88

SD_RETURN 0.49 *** 6.15 0.61 *** 5.09

SPI_DM 0.10 *** 12.31 0.09 *** 7.23

CAP_LEASE 0.06 *** 4.72 0.03 * 1.69

OP_LEASE 0.02 * 1.85 0.02 1.23

RD -0.05 -0.81 -0.05 -0.51

INTANG -0.06 *** -2.65 -0.01 -0.24

SIZE 0.19 *** 47.53 0.16 *** 20.49

AGE -0.01 ** -2.14 -0.03 ** -2.53

MTB -0.08 *** -16.51 -0.06 *** -8.70

LEVERAGE 0.38 *** 11.84 0.37 *** 7.23

FCF -0.46 *** -10.13 -0.41 *** -6.00

DERIVATIVE 0.09 *** 8.68 0.11 *** 5.62

LNBUSSEG 0.06 *** 4.89 0.07 *** 3.25

LNGEOSEG 0.01 1.49 0.03 * 1.92

SD_OIADP 0.13 *** 6.55 0.12 *** 4.75

DELAWARE -0.02 * -1.93 0.00 -0.08

IPO 0.03 0.34 0.07 0.71

SEO -0.04 *** -3.02 -0.05 ** -2.40

NMCOUNT 0.06 1.41 0.00 0.00



114

Table A.3 Incremental effect of Big 4 auditors on 10-K disclosure volume (LNSIZEG)

These tables report the benefit of enhanced disclosures provided by Big 4 auditors for audit clients with poorer accrual quality and those with higher

information asymmetry. The full sample (Panel A) and the PSM sample (Panel B) are partitioned into subsamples with low and high values of

ADA_MJR in columns (1) and (2) and subsamples with low and high values of EFFSPRD in columns (3) and (4), respectively. T-statistic is determined

by clustered standard errors at firm level. *, **, *** denote significance at the 0.10, 0.05, and 0.01 levels, respectively.

(1) (2) (3) (4) Low ADA_MJR High ADA_MJR Low EFFSPRD High EFFSPRD DV = LNSIZEG DV = LNSIZEG DV = LNSIZEG DV = LNSIZEG


Panel A – The full Sample

BIG4 0.07 ** 2.35 0.01 0.59 0.08 1.62 0.05 *** 2.81

ADA_MJR 1.22 1.17 0.04 0.31 ADA_MJR*BIG4 -1.05 -0.93 -0.07 -0.43 EFFSPRD -0.53 * -1.93 0.01 ** 2.32

EFFSPRD*BIG4 -0.17 -0.60 -0.01 -0.90

Control variables Yes Yes Yes Yes

Observations 17,103 17,470 20,128 20,134


Panel B – The PSM Sample

BIG4 0.05 1.35 -0.03 -1.08 0.06 1.25 0.02 0.79

ADA_MJR -0.11 -0.09 0.02 0.10

ADA_MJR*BIG4 0.37 0.24 0.12 0.49

EFFSPRD 0.09 0.27 0.00 0.42

EFFSPRD*BIG4 -0.44 -1.33 0.00 0.01

Control variables Yes Yes Yes Yes

Observations 4,573 5,695 2,873 9,309


115

Table A.4 Residual disclosures (RES_SG) and audit fees (Level and Change specifications)

Panel A - Level Specification

This table presents regression results of the audit fee model (Equation 3) by each asset quintile group. *, **, *** denote significance at the 0.10, 0.05,

and 0.01 levels, respectively, using two-tailed tests. T-statistic is determined by clustered standard errors at firm level.

(1) (2) (3) (4) (5)




RES_SG 0.09 *** 6.39 0.10 *** 5.77 0.10 *** 5.68 0.12 *** 5.82 0.11 *** 5.05

BIG4 0.52 *** 19.51 0.44 *** 15.47 0.32 *** 9.42 0.17 *** 3.42 0.20 * 1.74

Control Variables Yes Yes Yes Yes Yes

Observations 6,429 6,180 6,168 5,678 6,463

Adj. R-squared 69.3% 54.9% 46.1% 44.0% 71.6%

Panel B - Change Specification

This table presents regression results of the audit fee model (Equation 3) using the change specification. The dependent variable is a year-to-year

change in the level of audit fees (Δ LNAFEES). *, **, *** denote significance at the 0.10, 0.05, and 0.01 levels, respectively, using two-tailed tests.

T-statistic is determined by clustered standard errors at firm level.

(1)

Full Sample (2)

Non-Big 4 firms (3)

Big 4 firms

DV = Δ LNAFEES DV = Δ LNAFEES DV = Δ LNAFEES


Δ RES_SG 0.02 *** 4.74 0.01

1.52 0.02 *** 4.42

Control Variables Yes Yes Yes

Observations 24,894 6,708 18,186

Adjusted R-Squared 16.4% 11.4% 15.1%

116

Table A.5 Residual disclosures (RES_SG) and going-concern opinions

This table presents logistic regression results of the going-concern opinions model (Equation 4). *,

**, *** denote significance at the 0.10, 0.05, and 0.01 levels, respectively, using two-tailed tests. P-

values are based on Z-statistics, which are clusted by client firm.

(1)

Full Sample

DV = GC

(2)

Severely Financially

Distressed firms

DV = GC

Variables Coef. p-value Coef. p-value

Intercept -4.72 0.89 -3.33 0.93

RES_SG 0.29 *** 0.00 0.18 * 0.06

BIG4 -0.10 0.38 -0.33 ** 0.02

LNASSET -0.44 *** <.0001 -0.27 *** <.0001

MTB -0.12 *** <.0001 -0.07 ** 0.02

LEVERAGE 2.41 *** <.0001 2.15 *** <.0001

CH_LEV -0.43 0.30 -0.63 0.19

CFO -0.78 *** 0.01 -0.24 0.52

ALTMAN -0.04 *** <.0001 -0.04 *** <.0001

PLOSS 1.21 *** <.0001 0.37 * 0.06

HIGHLIT 0.18 0.36 0.18 0.42

TENURE 0.00 0.66 -0.01 0.41

ROA -3.59 *** <.0001 -3.49 *** <.0001

SD_OIADP 0.30 *** 0.00 0.26 ** 0.01


Pseudo R-Squared 37.1% 29.7%

117

Appendix B. Supplemental Analysis for Chapter 3

Instead of the dominant market share approach based on aggregated audit fees, I repeat

the same set of analyses for the differences in means of unexplained audit fees using alternative

measures of industry specialization, including AT_DOM, AF_30 and AT_30 in Panels A, B and C

of Table , respectively.

Consistent with the main analysis reported in

118

Table 3.7, I find that the differences in mean UAF values between two auditor groups:

industry specialists and non-industry specialists (Column D1), PwC specialists and PwC non-

industry specialists (Column D4) and other (non-PwC) specialists and other (non-PwC) non-

industry specialists (Column D5) appear to be weaker and sensitive to the period. On the other

hand, I find that the mean UAF values for the group of PwC specialists (Subsample A1) are

significantly larger than those for the group of PwC non- specialists (Subsample B1) across almost

all fiscal years. Similarly, I find that the mean UAF values for the group of PwC non-specialists

(Subsample B1) are significantly larger than those of other (non-PwC) non-specialists

(Subsample B2), thus supporting the evidence of a PwC fee premium, but not an industry-specific

fee premium.

119

Table B.1 Analysis using unexplained audit fees

This table reports univariate tests of the differences in means of UAF using an indicator of PwC together with an indicator of auditor industry

specialization (SPEC) to partition the full sample into four subsamples: (1) companies that are audited by PwC specialists, (2) companies

that are audited by specialists but not by PwC, (3) companies that are audited by PwC but not by specialists, and (4) Firms that are audited

by neither PwC nor specialists.

Panel A SPEC = Dominant market share approach based on aggregated client assets (AT_DOM)

Fiscal

Year

Subsample A

AT_DOM = 1

Subsample B

AT_DOM = 0



[A1]

PwC = 1

[A2]

PwC = 0

[B1]

PwC = 1

[B2]

PwC = 0

[D1]

A - B

[D2]

A1 - A2

[D3]

B1 - B2

[D4]

A1 - B1

[D5]

A2 - B2


2004 496 0.07 685 -0.01 332 0.03 2,838 -0.01 1.78 * 2.80 *** 1.50

1.09

0.00

2005 419 0.07 659 -0.01 273 0.05 2,832 -0.01 1.55

2.61 *** 1.75 * 0.58

0.07

2006 362 0.07 626 -0.03 243 0.09 2,812 -0.01 0.64

3.12 *** 3.20 *** -0.46

-0.76

2007 343 0.03 588 -0.01 226 0.04 2,709 0.00 0.48

1.52

1.29

-0.07

-0.26

2008 274 0.03 559 -0.04 269 0.04 2,530 0.00 -1.06

2.35 ** 1.26

-0.09

-1.91 *

2009 349 0.03 553 -0.03 181 0.03 2,416 0.00 -0.23

2.24 ** 0.87

0.24

-1.31

2010 338 0.05 499 -0.01 192 0.05 2,319 -0.01 0.92

2.27 ** 2.03 ** -0.15

-0.29

2011 326 0.07 599 -0.04 179 0.08 2,123 -0.01 0.11

4.06 *** 2.88 *** -0.27

-1.62

2012 315 0.06 629 -0.03 187 0.09 2,067 -0.01 -0.03

3.52 *** 3.33 *** -0.89

-1.28

2013 330 0.11 583 -0.01 186 0.04 2,181 -0.02 3.19 *** 4.43 *** 1.71 * 2.03 ** 0.63

2014 311 0.10 519 0.00 193 0.04 2,335 -0.02 3.09 *** 3.70 *** 1.92 * 1.78 * 0.81


120

Table B.1 – Continued

Panel B SPEC = Market share cut-off at 30 percent approach based on aggregated audit fees (AF_30)

Fiscal

Year

Subsample A

AF_30 = 1

Subsample B

AF_30 = 0



[A1]

PwC = 1

[A2]

PwC = 0

[B1]

PwC = 1

[B2]

PwC = 0

[D1]

A - B

[D2]

A1 - A2

[D3]

B1 - B2

[D4]

A1 - B1

[D5]

A2 - B2


2004 587 0.07 385 0.03 241 0.03 3,138 -0.02 3.47 *** 1.25

1.38

0.92

1.57

2005 421 0.07 428 0.03 271 0.05 3,063 -0.02 3.26 *** 1.14

1.98 ** 0.48

1.99 **

2006 333 0.08 451 0.01 272 0.07 2,987 -0.02 2.90 *** 2.14 ** 2.71 *** 0.46

1.24

2007 198 0.01 417 0.03 371 0.05 2,880 -0.01 1.45

-0.74

2.58 *** -1.22

1.93 *

2008 183 0.02 373 0.03 360 0.04 2,716 -0.01 1.59

-0.18

2.12 ** -0.59

1.68 *

2009 239 0.02 402 0.01 291 0.04 2,567 -0.01 1.10

0.26

1.83 * -0.58

0.92

2010 278 0.05 369 -0.01 252 0.06 2,449 -0.01 1.19

1.78 * 2.39 ** -0.27

0.16

2011 271 0.06 376 -0.02 234 0.09 2,346 -0.01 1.06

2.60 *** 3.95 *** -1.11

-0.22

2012 242 0.06 411 -0.04 260 0.08 2,285 -0.01 0.09

3.58 *** 3.42 *** -0.47

-1.39

2013 257 0.10 467 -0.01 259 0.06 2,297 -0.02 2.66 *** 3.82 *** 2.94 *** 1.14

0.66

2014 242 0.09 436 0.02 262 0.06 2,418 -0.02 3.36 *** 2.54 ** 3.16 *** 0.81

1.86 *


121

Table B.1 – Continued

Panel C SPEC = Market share cut-off at 30 percent approach based on aggregated client assets (AT_30)

Fiscal

Year

Subsample A

AT_30 = 1

Subsample B

AT_30 =0



[A1]

PwC = 1

[A2]

PwC = 0

[B1]

PwC = 1

[B2]

PwC = 0

[D1]

A - B

[D2]

A1 - A2

[D3]

B1 - B2

[D4]

A1 - B1

[D5]

A2 - B2


2004 495 0.07 547 0.01 333 0.04 2,976 -0.02 2.61 *** 1.87 * 1.87 * 0.77

1.05

2005 398 0.06 502 -0.01 294 0.05 2,989 -0.01 1.60

2.19 ** 2.12 ** 0.20

0.19

2006 273 0.06 519 -0.02 332 0.09 2,919 -0.01 0.54

2.29 ** 3.65 *** -0.92

-0.24

2007 271 0.02 484 0.00 298 0.05 2,813 -0.01 0.55

0.65

1.92 * -0.73

0.35

2008 263 0.03 499 -0.03 280 0.04 2,590 0.00 -0.44

1.96 ** 1.35

-0.03

-1.21

2009 353 0.04 513 -0.02 177 0.02 2,456 0.00 0.01

2.22 ** 0.71

0.39

-1.16

2010 342 0.05 473 -0.02 188 0.05 2,345 -0.01 0.85

2.66 *** 1.76 * 0.12

-0.67

2011 333 0.07 477 -0.03 172 0.08 2,245 -0.01 0.73

4.06 *** 2.87 *** -0.29

-1.40

2012 324 0.06 511 -0.04 178 0.09 2,185 -0.01 0.04

3.95 *** 3.13 *** -0.75

-1.74 *

2013 322 0.11 468 -0.01 194 0.04 2,296 -0.02 3.08 *** 4.62 *** 1.79 * 1.93 * 0.15

2014 315 0.10 459 0.00 189 0.04 2,395 -0.01 3.14 *** 3.83 *** 1.90 * 1.72 * 0.64


122

Appendix C. Supplemental Analysis for Chapter 4

Instead of the dominant market share approach based on aggregated audit fees

(AF_DOM), I repeat the same set of analyses for the auditors’ going-concern opinion

model using the dominant market share approach based on aggregated client assets

(AT_DOM) in Table .

Consistent with the main analyses, I find that, while SPEC is positive but

nonsignificant (Columns 1 and 3), only the estimated coefficient on SPEC – PwC is

positive and significant (Coef. = 0.24 with p-value = 0.06) for the full sample, indicating

that the estimated effect of SPEC is confounded by the positive impact of PwC on the

likelihood of issuing a GC audit opinion and that there is significant variation in audit

outcomes within the Big 4 firms. However, I find that the estimated coefficient on SPEC –

PwC is positive but nonsignificant (Coef. = 0.22 with p-value = 0.15) for the severely

distressed subsample.

123

Table C.1 Auditors’ going-concern opinion model

These tables present logistic regression results of the GC opinions model with the inclusion of auditor industry specialization measures for the full

sample (Columns 1 and 2) and the sample of firms with severely financially distressed firms (Columns 3 and 4). *, **, *** denote significance at the

0.10, 0.05, and 0.01 levels, respectively, using two-tailed tests. T-statistic is determined by clustered standard errors at firm level.

Full sample Severely Financially Distressed Firms

(1) (2) (3) (4)

DV = GC DV = GC DV = GC DV = GC

SPEC = AT_DOM SPEC = AT_DOM SPEC = AT_DOM SPEC = AT_DOM

Variables Coef. p-value Coef. p-value Coef. p-value Coef. p-value

Intercept -2.53 0.92 -2.53 0.91 -0.90 0.81 -0.90 0.81

LNASSET -0.46 *** <.0001 -0.46 *** <.0001 -0.38 *** <.0001 -0.38 *** <.0001

MTB -0.15 *** <.0001 -0.15 *** <.0001 -0.10 *** <.0001 -0.10 *** <.0001

LEVERAGE 1.46 *** <.0001 1.47 *** <.0001 1.14 *** <.0001 1.14 *** <.0001

CH_LEV -0.44 *** 0.00 -0.45 *** 0.00 -0.34 * 0.05 -0.35 * 0.05

CFO -0.18 * 0.09 -0.18 * 0.08 0.24 * 0.06 0.24 * 0.06

ALTMAN -0.04 *** <.0001 -0.04 *** <.0001 -0.03 *** <.0001 -0.03 *** <.0001

PLOSS 1.33 *** <.0001 1.33 *** <.0001 0.47 *** <.0001 0.47 *** <.0001

HIGHLIT -0.14 0.15 -0.14 0.14 -0.18 0.12 -0.18 0.10

TENURE 0.01 *** 0.00 0.01 *** 0.00 0.02 *** 0.01 0.02 *** 0.01

ROA -1.58 *** <.0001 -1.58 *** <.0001 -1.47 *** <.0001 -1.47 *** <.0001

SD_OIADP 0.24 *** <.0001 0.24 *** <.0001 0.20 *** <.0001 0.20 *** <.0001

BIG4 -0.35 *** <.0001 -0.35 *** <.0001 -0.36 *** 0.00 -0.35 *** 0.00

SPEC 0.01 0.92 0.13 0.26 SPEC - PwC 0.24 * 0.06 0.22 0.15

SPEC - EY -0.16 0.28 -0.12 0.55

SPEC - Others -0.15 0.36 0.25 0.22


Observations 40,918 40,918 8,652 8,652

Pseudo R-squared 52.7% 52.8% 44.0% 44.1%