Essays on auditor competencies

(1)

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

2018

SIMON 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.

(2)

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

(3)

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.

(4)

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

(5)

List of Tables

Table 2.1 Descriptive Statistics ... 30

Table 2.2 Correlation Matrices ... 34

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

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

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

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

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

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

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

Table 3.3 Audit fee model ... 67

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

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

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

Table 3.7 Analysis using unexplained audit fees ... 73

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

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

Table 4.3 Earnings quality model ... 97

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

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

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

(8)

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

(9)

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

(10)

Preface

This dissertation is original, unpublished, independent work by the author, Nattavut (Simon) Suwanyangyuan.

(11)

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).

(12)

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

(13)

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

(14)

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.

(15)

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

(16)

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.

(17)

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.

(18)

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.

(19)

1.3 Figures

Figure 1.1 The significant mergers among the largest audit firms Source: GAO (2008), page 9

(20)

Figure 1.2 Audit quality framework

Source: DeFond and Zhang (2014), page 280

Client Demand Auditor Supply

Incentives Incentives

e.g. agency costs,

regulation Audit 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

(21)

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.

(22)

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

(23)

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.

(24)

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.

(25)

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.

(26)

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).

(27)

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.

(28)

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

(29)

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

𝐋𝐍𝐖𝐎𝐑𝐃𝐒_𝐭= α₀+ α₁𝐁𝐈𝐆𝐍_𝐭+ α₂𝐃𝐄𝐋𝐓𝐀_𝐑𝐎𝐀_𝒕+ α₃𝐃𝐄𝐋𝐓𝐀_𝐑𝐄𝐕_𝐭+ α₄𝐌𝐀_𝐢𝐭 + α₅𝐅𝐘_𝐑𝐄𝐓_𝐭+ α₆𝐒𝐃_𝐑𝐄𝐓𝐔𝐑𝐍_𝐭+ α₇𝐒𝐏𝐈_𝐃𝐌_𝐭+ α₈𝐂𝐀𝐏_𝐋𝐄𝐀𝐒𝐄_𝐭 + α₉𝐎𝐏_𝐋𝐄𝐀𝐒𝐄_𝐭+ α₁₀𝐑𝐃_𝐭+ α₁₁𝐈𝐍𝐓𝐀𝐍𝐆_𝐭+ α₁₂𝐒𝐈𝐙𝐄_𝐭+ 𝛼₁₃𝐋𝐍𝐀𝐆𝐄_𝐭 + α₁₄𝐌𝐓𝐁_𝐭+ α₁₅𝐅𝐂𝐅_𝐭+ α₁₆𝐃𝐄𝐑𝐈𝐕𝐀𝐓𝐈𝐕𝐄_𝐭 + α₁₇𝐋𝐍𝐁𝐔𝐒𝐒𝐄𝐆_𝐭 + α₁₈𝐋𝐍𝐆𝐄𝐎𝐒𝐄𝐆_𝐭+ 𝛼₁₉𝐒𝐃_𝐎𝐈𝐀𝐃𝐏_𝐭 + α₂₀𝐃𝐄𝐋𝐀𝐖𝐀𝐑𝐄_𝐭+ α₂₁𝐈𝐏𝐎_𝐭+ α₂₂𝐒𝐄𝐎_𝐭+ 𝛼₂₃𝐋𝐍𝐍𝐌𝐂𝐎𝐔𝐍𝐓_𝐭 + 𝐘𝐞𝐚𝐫 𝐚𝐧𝐝 𝐈𝐧𝐝𝐮𝐬𝐭𝐫𝐲 𝐟𝐢𝐱𝐞𝐝 𝐞𝐟𝐟𝐞𝐜𝐭𝐬 + ε_𝐭

(30)

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

𝐁𝐈𝐆𝟒_𝐭 = α₀+ α₁𝐋𝐍𝐀𝐒𝐒𝐄𝐓_t+ α₂𝐀𝐓𝐔𝐑𝐍_𝐭+ α₃𝐂𝐔𝐑𝐑𝐄𝐍𝐓_𝐭+ α₄𝐋𝐄𝐕𝐄𝐑𝐀𝐆𝐄_𝐭 + α₅𝐑𝐎𝐀_𝐭+ Σ𝐂𝐨𝐧𝐭𝐫𝐨𝐥 𝐯𝐚𝐫𝐢𝐚𝐛𝐥𝐞𝐬

+ 𝐘𝐞𝐚𝐫 𝐚𝐧𝐝 𝐈𝐧𝐝𝐮𝐬𝐭𝐫𝐲 𝐟𝐢𝐱𝐞𝐝 𝐞𝐟𝐟𝐞𝐜𝐭𝐬 + ε_𝐭

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.

(31)

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

(32)

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

𝐋𝐍𝐀𝐅𝐄𝐄𝐒_𝐭= α₀+ α₁𝐋𝐍𝐀𝐒𝐒𝐄𝐓_𝐭+ α₂𝐂𝐔𝐑𝐑𝐄𝐍𝐓_𝐭+ α₃𝐈𝐍𝐕𝐑𝐄𝐂_𝐭+ α₄𝐋𝐄𝐕𝐄𝐑𝐀𝐆𝐄_𝐭 + α₅𝐑𝐎𝐀_𝐭+ α₆𝐈𝐍𝐓𝐋_𝐭+ α₇𝐌𝐀_𝐭+ α₈𝐒𝐏𝐈_𝐃𝐌_𝐭+ α₉𝐋𝐍𝐁𝐔𝐒𝐒𝐄𝐆_𝐭 + α₁₀𝐋𝐎𝐒𝐒_𝐭+ α₁₁𝐌𝐓𝐁_𝐭+ α₁₂𝐁𝐔𝐒𝐘_𝐭+ α₁₃𝐓𝐄𝐍𝐔𝐑𝐄_𝐭+ α₁₄𝐈𝐏𝐎_𝐭 + α₁₅𝐒𝐄𝐎_𝐭+ α₁₆𝐎𝐏𝐈𝐍𝐈𝐎𝐍_𝐭+ α₁₇𝐇𝐈𝐆𝐇𝐋𝐈𝐓_𝐭 + α₁₈𝐁𝐈𝐆𝟒_t

+ α₁₉𝐑𝐄𝐒_𝐖𝐑𝐃_𝐭+ 𝐘𝐞𝐚𝐫 𝐚𝐧𝐝 𝐈𝐧𝐝𝐮𝐬𝐭𝐫𝐲 𝐅𝐢𝐱𝐞𝐝 𝐄𝐟𝐟𝐞𝐜𝐭𝐬 + ε_𝐭

Equation 4: The going-concern opinions model

𝐆𝐂_𝐭= 𝛂_𝟎+ 𝛂_𝟏𝐋𝐍𝐀𝐒𝐒𝐄𝐓_𝐭+ 𝛂_𝟐𝐌𝐓𝐁_𝐭+ 𝛂_𝟑𝐋𝐄𝐕𝐄𝐑𝐀𝐆𝐄_𝐭+ 𝛂_𝟒𝐂𝐇_𝐋𝐄𝐕_𝐭+ 𝛂_𝟓𝐂𝐅𝐎_𝐭 + 𝛂_𝟔𝐀𝐋𝐓𝐌𝐀𝐍_𝐭+ 𝛂_𝟕𝐏𝐋𝐎𝐒𝐒_𝐭+ 𝛂_𝟖𝐇𝐈𝐆𝐇𝐋𝐈𝐓_𝐭+ 𝛂_𝟗𝐓𝐄𝐍𝐔𝐑𝐄_𝐭 + 𝛂_𝟏𝟎𝐑𝐎𝐀_𝐭+ 𝛂_𝟏𝟏𝐒𝐃_𝐎𝐈𝐀𝐃𝐏_𝐭+ 𝛂_𝟏𝟐𝐁𝐈𝐆𝟒_𝐭 + 𝛂_𝟏𝟑𝐑𝐄𝐒_𝐖𝐑𝐃_𝐭 + 𝐘𝐞𝐚𝐫 𝐚𝐧𝐝 𝐈𝐧𝐝𝐮𝐬𝐭𝐫𝐲 𝐟𝐢𝐱𝐞𝐝 𝐞𝐟𝐟𝐞𝐜𝐭𝐬 + 𝛆_𝐢𝐭

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

(33)

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

(34)

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 (𝑟_𝑝= 𝑟_𝑠= 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 (𝑟_𝑝= 𝑟_𝑠 = 0.32) and RES_WRD (𝑟_𝑝= 0.04; 𝑟_𝑠= 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 (𝑟_𝑝= 0.10; 𝑟_𝑠= 0.11) and GC (𝑟_𝑝= 𝑟_𝑠= 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.

(35)

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.

(36)

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.

Essays on auditor competencies