essays on auditor competencies
Post on 25-Oct-2021
1 Views
Preview:
TRANSCRIPT
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
3.2 Literature Review and Hypotheses Development ................................................ 46
3.2.1 Literature review ............................................................................................ 46
3.2.2 Hypotheses development .............................................................................. 49
3.3 Research Design .................................................................................................. 51
3.4 Sample Selection and Descriptive Statistics ........................................................ 52
3.4.1 Sample selection ........................................................................................... 52
3.4.2 Descriptive statistics ...................................................................................... 53
3.5 Empirical Results.................................................................................................. 54
3.6 Additional Analyses .............................................................................................. 56
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
4.2 Literature Review and Hypotheses Development ................................................ 80
4.2.1 Literature review ............................................................................................ 80
4.2.2 Hypotheses development .............................................................................. 82
4.3 Research Design .................................................................................................. 83
4.4 Sample Selection and Descriptive Statistics ........................................................ 86
4.4.1 Sample selection ........................................................................................... 86
4.4.2 Descriptive statistics ...................................................................................... 86
4.5 Empirical Results.................................................................................................. 87
4.6 Additional Analyses .............................................................................................. 89
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 2.5 Descriptive Statistics ............................................................................... 37
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 3.1 Descriptive Statistics ............................................................................... 63
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 4.1 Descriptive Statistics ............................................................................... 93
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
See Table 2.1 for variable definitions and descriptive statistics.
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
See Table 2.5 for variable definitions and descriptive statistics.
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
See Table 2.5 for variable definitions and descriptive statistics.
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).
Table 2.1 – Continued
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 ***
Table 2.1 – Continued
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
Adjusted R-squared 39.1% 38.1% 34.0% 35.8%
37
Table 2.5 Descriptive Statistics
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
Table 2.5 – Continued
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
Fixed Effects Yes Yes
Observations 42,889 5,362
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.
3.2 Literature Review and Hypotheses Development
3.2.1 Literature review
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.
3.2.2 Hypotheses development
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
See Table 3.1 for variable definitions and descriptive statistics.
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.
3.4 Sample Selection and Descriptive Statistics
3.4.1 Sample selection
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.
3.4.2 Descriptive statistics
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
3.5 Empirical Results
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.
3.6 Additional Analyses
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)”
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
63
Table 3.1 Descriptive Statistics
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
Other audit firms 1,314 37% 219.2 2% 438.8 7%
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
Other audit firms 1,244 37% 397.0 3% 459.3 7%
66
Table 3.2 - Continued
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
Variables Coef. t-stat. Coef. t-stat. Coef. t-stat.
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
Fixed Effects Yes Yes Yes
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)
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.
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
Adjusted R-squared 56.8% 41.9% 44.1% 48.8% 67.6%
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%
Other audit firms 0.0% 0.0% 0.0% 0.0%
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%
Other audit firms 0.1% 0.0% 0.0% 0.0%
70
Table 3.5 – Continued
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
Variables Coef. t-stat. Coef. t-stat. Coef. t-stat. Coef. t-stat.
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
Table 3.6 – Continued
Panel B – The full sample (Size partition)
(1) (2) (3) (4) (5)
Quintile 1 (Small) Quintile 2 Quintile 3 Quintile 4 Quintile 5 (Large)
N = 8,252 N = 7,992 N = 7,978 N = 7,485 N = 8,278
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.
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
Adjusted R-squared 63.6% 59.0% 48.2% 45.0% 68.1%
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
Adjusted R-squared 63.6% 58.9% 48.2% 45.0% 68.1%
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
Adjusted R-squared 63.5% 58.9% 48.1% 45.0% 68.0%
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
Adjusted R-squared 63.5% 58.9% 48.1% 45.0% 68.0%
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
Table 3.7 - Continued
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 *
*, **, *** denote significance at the 0.10, 0.05, and 0.01 levels, respectively, using two-tailed tests.
75
Table 3.7 - Continued
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
Variables Coef. t-stat. Coef. t-stat.
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
Fixed Effects Yes Yes
Observations 28,212 28,212
Adjusted R-squared 84.94% 84.92%
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.
4.2 Literature Review and Hypotheses Development
4.2.1 Literature review
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.
4.2.2 Hypotheses development
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
See Table 4.1 for variable definitions and descriptive statistics.
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
See Table 4.1 for variable definitions and descriptive statistics.
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).
4.4 Sample Selection and Descriptive Statistics
4.4.1 Sample selection
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.
4.4.2 Descriptive statistics
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).
4.5 Empirical Results
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 Additional Analyses
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
Table 4.1 Descriptive Statistics
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
Table 4.1 - Continued
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
*, **, *** denote significance at the 0.10, 0.05, and 0.01 levels, respectively, using two-tailed tests.
96
Table 4.2 – Continued
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 ***
*, **, *** denote significance at the 0.10, 0.05, and 0.01 levels, respectively, using two-tailed tests.
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
Variables Coef. t-stat. Coef. t-stat. Coef. t-stat. Coef. t-stat.
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
Adjusted R-squared 21.5% 47.2% 12.5% 37.6%
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
Variables Coef. t-stat. Coef. t-stat. Coef. t-stat. Coef. t-stat.
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
Fixed Effects Yes Yes Yes Yes
Observations 41,167 41,167 30,595 30,595
Adjusted R-squared 21.5% 21.5% 47.1% 47.2%
99
Table 4.4 – Continued
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
Variables Coef. t-stat. Coef. t-stat. Coef. t-stat. Coef. t-stat.
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
Fixed Effects Yes Yes Yes Yes
Observations 41,167 41,167 30,595 30,595
Adjusted R-squared 21.5% 21.5% 47.1% 47.2%
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
Fixed Effects Yes Yes Yes
Observations 40,918 40,918 40,918
Pseudo R-squared 52.8% 52.8% 52.8%
101
Table 4.5 – Continued
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
Fixed Effects Yes Yes Yes
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
Variables Coef. t-stat. Coef. t-stat.
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
Observations 43,575 13,152
Adjusted R-squared 77.9% 76.3%
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
Variables Coef. t-stat. Coef. t-stat. Coef. t-stat. Coef. t-stat.
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
Adjusted R-squared 79.1% 77.6% 80.5% 76.7%
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
Adjusted R-squared 76.3% 76.9% 78.2% 76.4%
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)
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.
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
Variables Coef. t-stat. Coef. t-stat. Coef. t-stat.
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
Observations 42,889 5,362
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
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 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
*, **, *** denote significance at the 0.10, 0.05, and 0.01 levels, respectively, using two-tailed tests.
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
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 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 *
*, **, *** denote significance at the 0.10, 0.05, and 0.01 levels, respectively, using two-tailed tests.
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
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 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
*, **, *** denote significance at the 0.10, 0.05, and 0.01 levels, respectively, using two-tailed tests.
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
Fixed Effects Yes Yes Yes Yes
Observations 40,918 40,918 8,652 8,652
Pseudo R-squared 52.7% 52.8% 44.0% 44.1%
top related