The effect of audit committee equity-based compensation on financial reporting quality : evidence from going-concern opinions

(1)

The effect of audit committee equity-based compensation on

financial reporting quality: Evidence from going-concern opinions

Name: Jorick M. Hofstra Student number: 10648909

Thesis supervisor: prof. dr. V.R. O’Connell Date: 22-06-2017

Word count: 19,982

MSc Accountancy & Control, specialization Accountancy Faculty of Economics and Business, University of Amsterdam

(2)

Statement of Originality

This document is written by student Jorick M. Hofstra who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

Abstract

Despite their role as independent supervisors of their firms’ financial reporting processes, US audit committee members are commonly compensated in the form of stocks or stock options. This study examines the effect of such equity-based audit committee compensation on financial reporting quality in terms of going-concern reporting. Using logistic regression analysis in addition to a simultaneous equations approach to mitigate endogeneity bias, I find that audit committee equity compensation decreases the likelihood that a going-concern opinion is issued, and that this effect is primarily driven by stock compensation. The results therefore suggest that, while equity-based compensation is likely effective in aligning the interests of audit committees and shareholders, this does not necessarily result in improved financial reporting quality.

(4)

Tables

Main text:

Table 2-1 Relation between ACM equity compensation and financial reporting quality ... 7

Table 2-2 Effect of GCO issuance on stock returns ... 10

Table 2-3 Effect of GCO issuance on bankruptcy probability ... 11

Table 2-4 Auditors’ belief in the self-fulfilling prophecy effect ... 12

Table 3-1 Sample selection ... 16

Table 4-1 Descriptive statistics and univariate analysis ... 25

Table 4-2 Logistic regression analysis ... 28

Table 4-3 Tests of instrument validity and regressor endogeneity ... 31

Table 4-4 Simultaneous equations analysis ... 33

Table 5-1 Summary of results ... 38

Appendices: Table A-1 Sample selection ... 50

Table A-2 Final merged sample description ... 51

Table A-3 Differences in compensation of audit committee members and non-executives ... 51

Table B-1 Datasets used ... 52

Table B-2 Merging ... 53

Table B-3 Final merged sample description ... 54

Table C-1 Correlation matrix of independent variables ... 55

Table C-2 Variance Inflation Factors (VIFs) of independent variables ... 56

Table D-1 Probit regression analysis ... 57

Table D-2 Two-stage minimum chi-square estimation... 58

(6)

1

1. Introduction

The recent financial crises, as well as the high number of corporate and accounting scandals since the beginning of this century, have increased public attention on the auditing profession and stressed the importance of high-quality financial reporting (DeFond & Francis, 2005). The audit committee, as a monitor of the financial reporting process, plays an important part in ensuring the quality of financial reporting (FASB, 2010). One of the major elements of the US Sarbanes-Oxley Act of 2002 (SOX), which was enacted in part to increase financial reporting quality as a response to large-scale audit failure, is a strengthening of the audit committee by increasing its responsibilities, setting stricter standards for its members, and expanding the committee’s authority.

However, SOX (2002) did not establish any rules or limits when it comes to audit committee member equity compensation, even though this comprises a significant part of total pay (52.7% as of 2006, in the sample of Lynch and Williams [2012]). According to SOX (2002), an audit committee member who possesses stock or stock options in the company in which he serves on the audit committee is not in breach of the independence requirement. This in contrast to Europe, for example, where the EU’s audit legislation explicitly states that audit committee members are not allowed to have any equity interest in the company in which they serve on the committee (Directive 2014/56/EU, 2014). In

addition, even though both auditors and audit committees serve to ensure the quality of financial reporting, the freedom to receive equity compensation does not apply to US

auditors; auditors cannot hold any equity in the firm they’re auditing, as doing so may impair their independence and objectivity (SOX, 2002).

Given that US legislation does not prohibit audit committee equity compensation, several researchers have examined the effect of such compensation on the quality of financial reporting. Different proxies for financial reporting quality have been used in these studies, such as the degree of earnings management (e.g., Bédard, Chtourou, & Courteau, 2004; Ghosh, Marra, & Moon, 2010; Lynch & Williams, 2012) or the amount of restatements (e.g., Archambeault, DeZoort, & Hermanson, 2008; Lin, Li, & Yang, 2006). This thesis aims to provide more evidence on the effect of different kinds of audit committee equity

compensation by examining the effect of such compensation on firms using a proxy for financial reporting quality that, to the best of my knowledge, has not been utilized before in this context: the auditor’s propensity to issue a going-concern opinion (GCO).

(7)

2

compensation on reporting quality is that, in certain cases, equity-based compensation can actually result in an increase in the quality of financial reporting (Bierstaker, Cohen, DeZoort, & Hermanson, 2012; Lynch & Williams, 2012; MacGregor, 2012; Sengupta & Zhang, 2015). This study aims to determine whether the positive effect of equity compensation on financial reporting quality (if it exists) still holds in the case of issuing a GCO. After all, previous research has shown that issuing a GCO may increase the chance of bankruptcy (Louwers, Messina, & Richard, 1999; Vanstraelen, 2003) and decrease stock returns (Jones, 1996; Menon & Williams, 2010), which means that equity compensation may actually provide an incentive not to issue a GCO, even if one is warranted. As such, the research question of this thesis is: what is the effect of audit committee equity compensation on financial reporting quality in terms of going-concern reporting?

To answer this question, I examine the relation between GCO issuance and equity-based compensation of non-executive directors for a sample of 20,466 US firm-years over the period of 2006 to 2015. In addition to traditional logistic regression, I also perform a

simultaneous equations analysis based on probit maximum-likelihood estimation in order to mitigate potential endogeneity bias. Lastly, I test the robustness of my results by performing a number of sensitivity checks.

I find that the proportion of equity-based compensation in directors’ compensation packages is significantly negatively related to GCO issuance, consistent with the hypothesis that audit committee members are less likely to support the issuance of a GCO the more equity compensation they receive. Additionally, I find that that this negative relation is driven mostly by compensation in the form of stock rather than options. This is consistent with the hypothesis that, because they generally face a greater decline in personal wealth when a GCO is issued, audit committee members compensated with stock are less likely to support GCO issuance than those rewarded with options. While the results support the notion that equity-based compensation increases audit committee members’ alignment of interests with

investors, this study also shows that this does not necessarily translate into improved financial reporting quality with respect to GCOs.

The remainder of this thesis is structured as follows. In section 2, the literature is reviewed to provide background and develop the hypotheses. Section 3 describes the sample selection procedure, and details the models and procedures used in the analyses. The results follow in section 4. In section 5, I discuss the conclusions, contribution and limitations of this study, and suggest possible avenues for future research.

(8)

3

2. Literature review and hypothesis development

In this section, I first explain what the audit committee is, what its responsibilities are, and what the US regulation around equity compensation is. I then define financial reporting quality, and review the literature on how audit committee equity compensation can affect it. Next, I explain why issuance of a going-concern opinion can act as a proxy for reporting quality, and examine the consequences of issuing a going-concern opinion according to previous research. I then use agency theory to explain how equity compensation may discourage issuance of a going-concern opinion by increasing the misalignment of interests between the principal (all existing and potential providers of capital) and the agent (the audit committee). Lastly, I develop my hypotheses based on all of the above.

2.1. Audit committee

SOX defines the audit committee as “a committee (or equivalent body) established by and amongst the board of directors of an issuer for the purpose of overseeing the accounting and financial reporting processes of the issuer and audits of the financial statements of the issuer.” (SOX, Sec. 2). In addition, an audit committee is obligated to contain at least one financial expert (SOX, Sec. 407), establish complaint procedures for complaints from the issuer or its employees (SOX, Sec. 301), determine the audit fee (SOX, Sec. 301), and ensure each member is a member of the board but otherwise independent (SOX, Sec. 301).

Audit committee members (ACMs) are considered independent according to SOX when they do not accept any consulting, advisory, or other compensatory fee from the firm on which they serve on the audit committee, other than in their capacity as a member of the committee, and when they are not an affiliated person of the firm or any of its subsidiaries (SOX, Sec. 301). It contains no further specification or limitations on the compensation that ACMs are allowed to receive in their capacity as members.

As such, ACMs can receive equity compensation in the company they’re monitoring and still be considered independent according to SOX. “Equity compensation” is pay in a form other than cash that represents ownership in the firm, such as shares, stock options, or restricted stock (Engel, Hayes, & Wang, 2010; Archambeault et al., 2008). Audit committee equity compensation appears to be a common phenomenon: while Engel et al. (2010) note that equity-based audit committee compensation decreased in 2002 as a result of SOX being enacted, the proportion of ACMs’ pay consisting of equity increased in the years afterwards. In addition, Lynch & Williams (2012) show that equity compensation in 2006 accounted for, on average, 52.7% of total audit committee compensation in their sample of 1,488 firms.

(9)

4

Lastly, SOX also outlines the responsibilities of audit committees: audit committees are to be directly responsible for the appointment, compensation, and oversight of the work of the auditor, including the resolution of disagreements between the auditor and management (SOX, Sec. 301). Furthermore, the audit committee must be aided in its tasks by the auditor’s reports to the audit committee, which must contain all critical accounting policies and

practices to be used, all alternative treatments of financial information and its ramifications, and other material written communications between the auditor and management (SOX, Sec. 204). In short, SOX clearly prescribes that the audit committee should play an important role in the financial reporting process.

2.2. Financial reporting quality

2.2.1. Definition of financial reporting quality

Both the IASB (2010) and the US-based FASB (2010) provide definitions of financial reporting quality in their Conceptual Frameworks. Due to the Framework convergence project between the IASB and the FASB, there is a large amount of overlap between the Frameworks and the definitions of reporting quality therein. Since the focus of this thesis is on US firms and audit committees, I define financial reporting quality using the FASB framework. According to the FASB (2010) Conceptual Framework,

“The objective of general purpose financial reporting is to provide financial

information about the reporting entity that is useful to existing and potential investors, lenders and other creditors in making decisions about providing resources to the entity” (p. 1).

High quality financial reporting, therefore, means providing a high quality of such financial information. The Framework further outlines that financial information can be considered high quality, or “useful”, when it is relevant and faithfully represents what it purports to represent (FASB, 2010, p. 16). This usefulness is enhanced when the financial information is comparable, verifiable, timely and understandable (FASB, 2010, p. 16).

2.2.2. Effect of audit committee equity compensation

Since reporting quality (as defined above) is not directly measurable, there are a number of proxies that can be used to measure the effect that audit committee equity compensation can have on the quality of financial reporting. A short overview of the findings of several previous studies on this subject follows below.

(10)

5 Earnings management as a proxy:

Earnings management can be defined as the “purposeful intervention in the external financial reporting process, with the intent of obtaining some private gain” (Schipper, 1989, p. 92). As increased earnings management can generally be expected to lead to decreased financial reporting quality, it can therefore be used as a proxy for (lowered) reporting quality (Biddle, Hilary, & Verdi, 2009).

Theorizing that larger equity holdings will align the interests of ACMs with that of management, Ghosh et al. (2010) measure the correlation between the percentage of

committee members’ equity holdings with the degree of earnings management proxied for by discretionary accruals. However, they find no significant results. Using a similar

discretionary accrual model to measure the degree of earnings management, Bédard et al. (2004) instead focus on the effect of stock option compensation. Controlling for a number of audit committee characteristics, they find that value of exercisable stock options held by ACMs has a significant positive relation with the degree of earnings management.

Another study utilizing a discretionary accruals model to measure earnings

management is that of Lynch and Williams (2012). They examine different types of audit committee equity compensation, separately measuring the effect of stock and option

compensation. Like Ghosh et al. (2010), Lynch and Williams (2012) find that the proportion of all equity compensation to total compensation is not significantly related to the degree of earnings management. However, Lynch and Williams do find significant results when separately measuring the effects of stock and option compensation. Specifically, they find that option compensation is positively related with earnings management, while stock compensation has a negative relation with the management of earnings, suggesting that compensation in the form of options may reduce reporting quality while stock compensation can actually improve it. Lynch and Williams argue that this is because compensation in the form of stock increases the alignment of interests with shareholders, while option

compensation can incentivize opportunistic behavior (e.g., earnings management) due to the limited downside risk inherent in stock options.

Restatements as a proxy:

Some studies have investigated audit committee equity compensation by examining its effect on the occurrence of restatements. A restatement is a correction of a (usually highly) material misstatement in a financial report after the financial report has been issued (Eilifsen &

(11)

6

quality because they indicate that financial reports were not faithfully represented at the time they were issued.

Lin et al. (2006) examine the effect of the percentage of outstanding common shares held by ACMs in their logistic regression model of the likelihood of restatement. In line with what would be expected based on the results of Ghosh et al. (2010), Lin et al. (2006) find no significant relation between restatement likelihood and committee members’ total stock holdings. Archambeault et al. (2008), on the other hand, investigate the effect of audit committee option compensation on the probability of restatement. Consistent with the

findings of Lynch and Williams (2012), Archambeault et al. (2008) find that the proportion of option compensation to total compensation is correlated with an increased restatement

likelihood, corresponding to a decrease in financial reporting quality. Other proxies:

Another proxy for financial reporting quality that has been used in the context of equity compensation is the quality of disclosures. Sengupta and Zhang (2015) measure disclosure quality as the likelihood of disclosure, the frequency, and the accuracy of firms’ earnings forecasts. They find that audit committee equity compensation, measured as the proportion of equity-based compensation to total compensation, is positively associated with all measures of disclosure quality. Furthermore, MacGregor (2012) investigates the relation between equity holdings by ACMs and the risk of reporting problems. He finds that equity holdings reduce this risk when there are observable risk factors, such as weak internal controls, high CEO equity incentives, or high-risk assets, providing further evidence that equity

compensation may increase financial reporting quality.

Rather than using archival research methods, Bierstaker et al. (2012) examine the effect of equity compensation on ACMs’ judgment in an experimental setting. With a participant group consisting of highly experienced public company ACMs, they find that committee members receiving vested option compensation are more likely to support the auditor in an accounting disagreement with management than members who do not receive such compensation.

Conclusions:

A summary of the findings of the studies discussed is presented in Table 2-1. Most (3 out of 4) studies that examine the effect of stock options find that options decrease reporting quality, possibly because the limited downward risk inherent in this type of compensation

(12)

7

holdings, the relation with reporting quality is unclear, and appears to be either insignificant or positive depending on the proxy used. Possibly because it increases the alignment of interests between ACMs and shareholders, equity-based compensation has been found to increase reporting quality by increasing disclosure quality, increasing the audit committee’s risk responsiveness, and increasing the likelihood that ACMs support the auditor in disputes with management. However, equity compensation has not been found to have an effect on earnings management or restatement likelihood. As for stock compensation, the only researchers to specifically examine the effect of such compensation on reporting quality are Lynch and Williams (2012), who find that this effect is positive. Lastly, it should be noted that they find that the effects of option and stock compensation on reporting quality are opposite to each other, which may explain why several studies found total equity-based compensation to have an insignificant effect.

Table 2-1: Relation between ACM equity compensation and financial reporting quality

Study Proxy used Subject of research Findings Relation with

reporting quality Ghosh et al. (2010) Earnings management Total equity holdings

ACMs’ equity holdings are not significantly related to earnings management Not significant Bédard et al. (2004) Earnings management Option compensation

There is a significant positive relation between ACM option compensation and earnings management.

Negative Lynch & Williams (2012) Earnings management

Option, stock, and total equity-based compensation

Option compensation has a positive relation, stock compensation has a negative relation, and total equity-based compensation has no significant relation with earnings management.

Options: Negative Stock: Positive Total: Not significant Lin et al. (2006)

Restatements Total equity holdings

ACMs’ equity holdings are not significantly related to restatement likelihood. Not significant Archambeault et al. (2008) Restatements Option compensation

There is a significant positive relation between ACM option compensation and the likelihood of restatement.

Negative Sengupta & Zhang (2015) Disclosure quality Total equity-based compensation

There is a significant positive relation between ACM total equity-based compensation and disclosure quality.

Positive MacGregor (2012) Risk of reporting problems Total equity holdings

ACMs’ equity holdings are significantly negatively related to the risk of reporting problems. Positive Bierstaker et al. (2012) Propensity to support auditor Option compensation

There is a significant positive relation between vested option compensation and the ACM’s propensity to support the auditor in disagreements with management.

(13)

8

2.3. The going-concern opinion

2.3.1. Going-concern opinions as a proxy for financial reporting quality

According to US accounting standards, auditors have the responsibility of assessing whether there is substantial doubt about the audited firm’s ability to continue as a going-concern for a reasonable period of time, i.e., a period of time not exceeding one year past the date the audit was performed (AICPA, 2015, pp. 573-574). If the auditor were to conclude that such

substantial doubt exists, the auditor should include an emphasis-of-matter paragraph in the auditor's report to reflect this; the so-called “going-concern opinion” (AICPA, 2015, p. 576).

The value of any financial information produced by the financial reporting process is contingent on the assumption that the reporting entity is a going concern, an assumption that is commonly referred to as the going-concern basis of accounting (FASB, 2014, p. 6). If it is determined that a firm’s liquidation is imminent, the firm should instead be accounted for under the liquidation basis of accounting, in accordance with ASC 205-30 of US GAAP (FASB, 2014, p. 6). Therefore, accurately reporting on whether or not the going-concern assumption is justified is a crucial part of high-quality financial reporting.

Given the importance of the going-concern assumption, the auditor’s propensity to issue a going-concern opinion (after controlling for other factors) has often been used as a proxy for financial reporting quality, as auditors are likely to be significantly pressured by management not to issue such an opinion even if one is warranted (Bronson, Carcello, Hollingsworth, & Neal, 2009; Bruynseels & Cardinaels, 2013; Carcello & Neal, 2000; Carey & Simnett, 2006; DeFond, Raghunandan, & Subramanyam, 2002; Knechel & Vanstraelen, 2007). A key audit committee task is to mitigate managerial pressure on the auditor, and aid the auditor in any disputes with management (Bronson et al., 2009; Carcello & Neal, 2000; SOX, 2002). Given the important role audit committees play in the financial reporting

process (as described in 2.1.), audit committees can therefore significantly influence the GCO decision.

2.3.2. Effects of issuance of a going-concern opinion

Next, I review the literature to assess the effects of GCO issuance, focusing on the possible effects on stock returns and on the probability of bankruptcy.

GCO issuance and stock returns

Earlier studies on the subject suggest GCOs may have little effect on stock returns. Dodd, Dopuch, Holthausen and Leftwich (1984) examine the effect on stock returns of several

(14)

9

different types of audit reports, and find no significant abnormal stock returns of GCO firms in the period surrounding GCO issuance. Elliot (1982) also examines abnormal stock returns surrounding the announcement of different types of audit reports, with a relatively small sample of 16 GCO firms. He finds significantly negative cumulative returns in the 45 weeks prior to GCO issuance, but no significant abnormal returns afterwards, suggesting that investors learn of the negative firm conditions that lead to a GCO several weeks before it is issued.

However, most of the more recent research suggests that GCO issuance does affect stock returns. Jones (1996) theorizes that the difference (in terms of stock returns) between receiving a GCO and not receiving one is larger for firms that are financially distressed, as issuing a clean opinion likely has a stronger reassuring effect on investors if the firm in which they have invested is distressed. In line with these expectations, Jones (1996) finds that abnormal returns were negative for distressed firms receiving a GCO, while they were

positive for distressed firms receiving a clean opinion. Ogneva and Subramanyam (2007) also examine abnormal stock returns after GCO issuance, using a sample of US and Australian firms that is significantly larger (n = 1,250) than the samples used in prior research. Similar to Jones (1996), they find significantly negative abnormal returns after GCO issuance. However, Ogneva and Subramanyam (2007) also find that abnormal returns are no longer significant when they control for expected earnings, suggesting that the magnitude of

negative stock returns depends on the degree to which the GCO was expected. Lastly, Menon and Williams (2010) also find that GCO issuance has a negative effect on stock returns, using a large sample (n = 1,194) of US firms. In addition, they show that the magnitude of the negative effect may be driven by the type of shareholder, as institutional shareholders seem to react more strongly to GCO issuance than other investors.

A summary of the findings of the studies discussed is presented in Table 2-2. In conclusion, it seems that issuing a GCO generally decreases stock returns of US firms, but that the magnitude of decrease depends on a number of factors, such as the degree to which the GCO was expected and the level of institutional ownership.

GCO issuance and bankruptcy

Several studies have suggested that issuing a GCO may increase the chance of bankruptcy due to the self-fulfilling prophecy effect. The “self-fulfilling prophecy effect” is the idea that issuing a GCO can actually cause a firm to go bankrupt, or at least increase the chance of bankruptcy, because issuing a GCO can have several negative consequences for the firm such

(15)

10 Table 2-2: Effect of GCO issuance on stock returns

Study Country Findings Relation between

GCO issuance and stock returns Dodd et al.

(1984)

US No significant change in abnormal stock returns surrounding GCO issuance.

Not significant

Elliot (1982)

US Increase in abnormal stock returns preceding GCO, but not following GCO .

Not significant

Jones (1996)

US Abnormal returns were significantly negative for distressed firms receiving a GCO.

Negative Ogneva & Subramanyam (2007) US & Australia

Abnormal returns were significantly negative following a GCO, and dependent on the degree to which it was expected.

Negative

Menon & Williams (2010)

US Abnormal returns were significantly negative following a GCO, and at least partially driven by the level of institutional ownership.

Negative

Notes: “Sample size” refers to the number of firm-years examined in which a GCO was issued.

as decreased stock returns (as discussed above) or increased difficulty in restructuring debt or raising capital (Carey, Geiger, & O’Connell, 2008; Citron & Taffler, 2001).

Consistent with the self-fulfilling prophecy effect, Louwers et al. (1999) find that GCO issuance leads to an increase in the chance of bankruptcy in their sample of US firms, this chance being highest in the year immediately following GCO issuance. However, they find that the risk of bankruptcy is significantly lessened afterwards. Also in line with the self-fulfilling prophecy effect are the results of Vanstraelen’s (2003) study in the Belgian setting, as she finds that issuance of both first-time and repeated GCOs significantly increase the likelihood of bankruptcy. Using a sample of private rather than public firms, Gaeremynck and Willekens (2003) find similar results, providing further evidence that GCO issuance is positively related with bankruptcy risk.

However, other non-US studies show that issuing a going-concern opinion may not increase the chance of bankruptcy. Using a sample of Australian firms that received a first-time GCO matched with financially distressed non-GCO firms, Carey et al. (2008) find no evidence that issuance of a GCO increases the chance of bankruptcy for Australian firms. Citron and Taffler (2001) also use a sample of matched, financially distressed firms in their study of the effect of GCOs in the UK. They, too, find that the firms in their sample that received a GCO are no more likely to go bankrupt than those who received clean opinions.

(16)

11

true, auditors tend to believe in the self-fulfilling prophecy effect. Applying a game-theoretic analysis, Matsumura, Subramanyam and Tucker (1997) find that auditors are less likely to issue a GCO the higher the risk of dismissal and the stronger the self-fulfilling prophecy effect. Building on this analysis, Tucker, Matsumura and Subramanyam (2003) perform an experimental test based on an adjusted version of Matsumura et al.’s (1997) game-theoretic model of GCO judgment. Tucker et al. (2003) find that the self-fulfilling prophecy effect affects the decision making of both auditors and clients, causing a decrease in the amount of GCOs issued with the magnitude of decrease being negatively correlated to the forecast accuracy of audit evidence. Further evidence on the self-fulfilling prophecy effect is provided by Guiral, Ruiz and Rodgers (2011), who conduct an experimental study with participants that include senior managers and partners from an international accounting firm. They find that believing in the self-fulfilling prophecy effect increases participants’ tendency to accept mitigating evidence, while lowering the tendency to acknowledge contrary evidence.

A summary of the findings of the studies discussed is presented in Tables 2-3 and 2-4. In conclusion, evidence on the self-fulfilling prophecy effect (i.e. the effect that issuing a GCO increases the chance of bankruptcy) is somewhat mixed, although it seems to have at least some effect in the US. Despite the mixed evidence, auditors generally seem to believe in the self-fulfilling prophecy effect, decreasing their tendency to issue a GCO.

Table 2-3: Effect of GCO issuance on bankruptcy probability

Study Country Type of

research

Findings Relation between

GCO issuance and the probability of bankruptcy Louwers et al.

(1999)

US Archival Risk of bankruptcy significantly increased following GCO, especially in the first year.

Positive

Vanstraelen (2003)

Belgium Archival Both first-time and subsequent GCOs significantly increase the likelihood of bankruptcy.

Positive

Gaeremynck & Willekens (2003)

Belgium Archival Private firms’ risk of bankruptcy significantly increased following GCO.

Positive

Carey et al. (2008)

Australia Archival GCO firms no more likely to go bankrupt than matched non-GCO firms.

Not significant

Citron & Taffler (2001)

UK Archival GCO firms no more likely to go bankrupt than matched non-GCO firms.

(17)

12 Table 2-4: Auditors’ belief in the self-fulfilling prophecy effect

Study Country Type of

research

Findings Evidence for a

belief in self-fulfilling prophecy effect? Matsumura et al. (1997) n/a Game-theoretic analysis

Under certain circumstances, a belief in the self-fulfilling prophecy effect reduces the likelihood of GCO issuance.

Yes Tucker et al. (2003) US Game-theoretic analysis & Experimental

Belief in the self-fulfilling prophecy effect decreases the amount of GCOs issued, with the decrease being negatively correlated to the forecast accuracy of audit evidence.

Yes

Guiral et al. (2011)

n/a* Experimental Belief in the self-fulfilling prophecy effect affects participants’ interpretations of audit evidence.

Yes

Notes: * Guiral et al. (2011) do not disclose participant nationality, mentioning only that the participants are employees of an international accounting firm.

2.4. Agency theory and equity compensation

Agency theory aims to describe the relationship between a principal and an agent acting on the principal’s behalf (Eisenhardt, 1989). Jensen and Meckling (1976) define such a

relationship as “a contract under which one or more persons (the principal(s)) engage another person (the agent) to perform some service on their behalf which involves delegating some decision-making authority to the agent.” (p. 308). The main task of the audit committee is to ensure the quality of financial reporting (SOX, Sec. 2), which means providing high-quality financial information to all existing and potential providers of capital (FASB, 2010, p. 1).As such, the employment of an ACM can be described as a principal-agent relationship: through legislation, the existing and potential providers of capital (the principals) engage the ACM (the agent) to perform the service of monitoring the financial reporting process on their behalf.

Given that the agent and the principals have different utility functions, this may lead to unwanted behavior of the agent from the perspective of the principals. If both the

principals and the agent seek to maximize their utility (i.e., assuming self-interest), and one assumes effort to be costly, it is likely that the agent will not always act in the best interest of the principals (Jensen & Meckling, 1976; Watts & Zimmerman, 1978). This study focuses on a type of incentive-based compensation that is commonly used to align the interests of the agent with the principals: equity-based compensation (Morck, Shleifer, & Vishny, 1988).

(18)

13

Equity compensation gives ACMs a direct financial interest in increasing the value of the firm, and as such, is expected to better align the interests of the audit committee with that of shareholders (Jensen & Meckling, 1976). The findings of several studies discussed in section 2.2 (i.e., Bierstaker et al., 2012; Lynch & Williams, 2012; MacGregor, 2012; Sengupta & Zhang, 2015) seem to be consistent with this expectation, as they find that equity

compensation can benefit shareholders by, for example, decreasing earnings management or improving disclosure quality.

Audit committee equity compensation, therefore, aligns the audit committee’s

interests with that of shareholders. However, the principals of the audit committee are not just shareholders, but all existing and potential providers of capital (FASB, 2010, p. 1). In

situations where the interests of shareholders deviate from that of the other (potential) capital providers, audit committee equity compensation could therefore increase the misalignment of interests between agent and principals.

2.5. Hypotheses

Agency theory predicts that equity-based compensation will help align the interests of audit committee members with that of shareholders (Jensen and Meckling, 1976; Watts &

Zimmerman, 1978). In addition, several studies have shown that such equity-based

compensation can improve financial reporting quality (e.g., Bierstaker et al., 2012; Lynch & Williams, 2012; MacGregor, 2012; Sengupta & Zhang, 2015).

However, reporting in line with existing shareholder interests may not always improve reporting quality. High-quality financial reporting entails providing high-quality financial information about a firm that is useful to all the firm’s existing and potential providers of resources; not just the current shareholders (FASB CF, 2015, p. 1). While issuing a going-concern opinion may provide valuable information to potential providers of capital, it can be considered to be against the interests of existing shareholders given the resulting potential decrease in stock returns (Jones, 1996; Menon & Williams, 2010; Ogneva & Subramanyam, 2007) and possible increase in the risk of bankruptcy (Gaeremynck & Willekens, 2003; Louwers et al., 1999; Vanstraelen, 2003). In addition, previous research indicates that the auditor’s GCO decision is influenced by a belief in the self-fulfilling prophecy effect (Guiral et al., 2011; Matsumura et al., 1997; Tucker et al., 2003). It is difficult to say to which degree this belief transfers to audit committee members, but if it does, it could further decrease the chance a going-concern opinion is actually issued. Given the negative expected consequences of GCO issuance for shareholders, the effect of equity

(19)

14

compensation on financial reporting quality may therefore not be positive when reporting quality is proxied for by the auditor’s propensity to issue a going-concern opinion.

Management is likely to oppose issuance of a going-concern opinion due to the negative consequences of such an opinion for both management and shareholders. As a result, auditors are likely to be significantly pressured by management not to issue a going-concern opinion, even if one is warranted (Bronson et al., 2009; Bruynseels & Cardinaels, 2013; Carcello & Neal, 2000; Carey & Simnett, 2006; DeFond et al., 2002; Knechel & Vanstraelen, 2007). One of the key tasks of the audit committee is ensuring auditor

independence by mitigating such pressure from management (Bronson et al., 2009; Carcello & Neal, 2000; SOX, 2002).

As I expect that equity-based audit committee compensation increases the

committee’s alignment with existing shareholders’ interests, and issuing a GCO is against their interests, I expect that equity compensation will decrease the committee’s efforts to mitigate managerial pressure on the auditor not to issue a GCO. Stated differently, I expect that audit committee equity compensation decreases the likelihood that a going-concern opinion is issued. Therefore, the first hypothesis is:

H1: The occurrence of audit committee equity compensation is negatively associated

with the likelihood of GCO issuance.

Extending H1, I also expect that the alignment of interests between audit committee members and existing shareholders increases when the equity-based compensation comprises a greater part of committee members’ total compensation package. As such, the second hypothesis is:

H2: The proportion of equity compensation to total audit committee compensation is

negatively associated with the likelihood of GCO issuance.

Stock option awards provide a convex payoff schedule whereas stock grants provide a linear payoff schedule, which means that stock option compensation provides a more limited decrease in wealth than stock compensation when the share price falls (Bryan, Hwang, Klein & Lilien, 2000; Lynch & Williams, 2012; Rajgopal & Shevlin, 2002). As such, an audit committee member’s decrease in personal wealth following a share price drop as a result of GCO issuance may be lower when he or she is compensated with options than when he or she is compensated with shares. I therefore expect that option compensation decreases the

likelihood of GCO issuance less than stock compensation. The third hypothesis is:

H3: Audit committee option compensation is more positively (or less negatively)

(20)

15

3. Research methodology

3.1. Sample

I examine the effect of audit committee equity compensation on the auditor’s propensity to issue a going-concern opinion by analyzing archival data on the composition of director compensation and firm characteristics of a number of US companies in the time period post-SOX. As the current regulation on audit committees became effective as of April 2003 (SOX, 2002, SEC 301), I start with an initial sample that covers the period of fiscal year 2004 to the most recent fiscal year for which data is available: 2016.

I retrieved the data on firms’ director compensation from the Capital IQ – People Intelligence database, as it is the largest database available to me that has detailed

information on the composition of the compensation packages of board members. It should be noted that the Capital IQ database does not provide information on directors’ committee membership; it only (reliably) distinguishes between executive directors and non-executive directors. However, it seems reasonable to assume that the compensation packages of audit committee members (ACMs) are similar to that of non-executive directors who are not on the audit committee (NACMs), given that ACMs are selected from the pool of all non-executive directors (SOX, 2002). As such, I do not expect there to be much difference in the

compensation structures of both groups.

To investigate the validity of this assumption, I analyzed the differences in compensation between ACMs and NACMs by combining the Execucomp – Director Compensation database with the ISS – Directors database, the latter of which contains the necessary information on audit committee membership1. I found that there is little difference in the compensation packages of ACMs and NACMs: the value of total compensation for NACMs is only 0.4% lower than for ACMs, and there is no statistical difference in the value of total equity compensation at all (see Appendix A for further detail on this analysis). I therefore consider the compensation of all non-executives to be a reasonable proxy for ACM compensation.

Table 3-1 describes my sample selection procedure. The initial sample, retrieved from the Capital IQ database, contained 2,894,322 observations of the employees of 41,500 firms

1_{I did not use ISS for my main analysis because it contains an insufficient number of GCO}

years. Merging ISS with my datasets of other firm characteristics resulted in a sample of 9,000 firm-year observations, but a GCO was issued in only 5 of them.

(21)

16 Table 3-1: Sample selection

Observations

Initial Capital IQ sample No gvkey

Removal of non-directors Non-US firms

Negative cash or equity compensation No compensation data available

Less than 2 non-executives per firm-year

2,894,219 (426,374) (1,522,065) (666,483) (113) (12,312) (30,412) Capital IQ sample, director observations

Removal of duplicate firm-year observations

236,460 (197,690)

Capital IQ sample, firm-years 38,770

Lost by merging Compustat – Fundamentals Annual Lost by merging AuditAnalytics

Lost by merging Compustat – Security Monthly Lost by merging Compustat – Historical Segments Removal of observations in 2004, 2005, 2016 (15,764) (1,561) (484) (158) (320) Final sample 20,466

over the period 2004 to 2016. I removed observations lacking a gvkey-identifier (426,374), observations of persons not on the board of directors (1,522,065), firms outside the US (666,483), observations with incorrect (113) or missing (12,312) compensation data, and firms with less than 2 non-executives on the board (30,414) leaving 236,460 director-firm-year observations. Calculating the average director compensation package for each firm and removing the observations of individuals resulted in a final Capital IQ sample of 38,770 firm-year observations.

To obtain the necessary information on firm characteristics, auditor opinions, stock returns, and firm segments, I merged my Capital IQ sample with datasets from Compustat - Fundamentals Annual, AuditAnalytics, Compustat - Security Monthly, and Compustat - Historical Segments, respectively (see Table B-1 and B-2 in Appendix B for more detail on these datasets). Lastly, I excluded observations from 2004, 2005, and 2016 from my sample (320), as the only data available for these years is the data of a small number of big firms. My final sample consists of 20,466 firm-year observations over the period 2006 to 2015, covering

(22)

17

3,473 unique firms and 14,606 different non-executive directors (see Table B-3 in Appendix B).

3.2. Logistic regression model

Given that my dependent variable (whether or not a GCO is issued) is binary, I test my hypotheses using logistic regression, utilizing the following base model:

GCO = β0 + β1Equity + β2LnTA + β3ZFC + β4Return + β5Vol + β6PriorGCO +

β7RepLag + β8Big4 + β9BoardSize + β10LnAge + ε (1)

The dependent variable, GCO, equals 1 if a going-concern opinion is issued, and 0 otherwise.

3.2.1. Primary independent variables (Equity)

The independent variables of primary interest are a number of measures of equity compensation represented in the above function by the variable Equity, as I intend to separately examine the effects of different types of equity compensation received by non-executive directors (NEDs). These four measures are (in bold):

EqInd = Indicator variable that equals 1 if NEDs’ compensation packages contain a form of equity compensation in the given firm-year, 0 otherwise.

EqTotal = Average of (value of total NED equity compensation / value of total NED compensation).

EqStock = Average of (value of NED stock compensation / value of total NED compensation).

EqOption = Average of (value of NED stock option compensation / value of total NED compensation).

The corresponding regression models are:

Model 1: GCO = β0 + β1EqInd + β2LnTA + β3ZFC + β4Return + β5Vol + β6PriorGCO +

β7RepLag + β8Big4 + β9BoardSize + β10LnAge + ε (2)

Model 2: GCO = β0 + β1EqTotal + β2LnTA + β3ZFC + β4Return + β5Vol + β6PriorGCO

+ β7RepLag + β8Big4 + β9BoardSize + β10LnAge + ε (3)

Model 3: GCO = β0 + β1EqStock + β2LnTA + β3ZFC + β4Return + β5Vol + β6PriorGCO

(23)

18 Model 4: GCO = β0 + β1EqOption + β2LnTA + β3ZFC + β4Return + β5Vol +

β6PriorGCO + β7RepLag + β8Big4 + β9BoardSize + β10LnAge + ε (5)

Following my hypotheses, I expect EqInd and EqTotal to have a negative and significant influence on GCO (H1 and H2), and I expect EqOption to have a more positive influence on GCO than EqStock (H3).

3.2.2. Control variables

The remaining independent variables are control variables (in bold): LnTA = Natural log of total assets.

Prior research finds that issuance of a GCO is less likely the larger a firm is because larger firms are less likely to default, and because audit fees (and by extension the costs of losing the client) are higher the bigger the client (e.g., Bronson et al., 2009; Bruynseels &

Cardinaels, 2013; Carcello & Neal, 2000; Knechel & Vanstraelen, 2007; Mutchler,

Hopwood, & McKeown, 1997; Wu, Hsu, & Haslam, 2016). I therefore expect the coefficient of LnTA to be negative. I use the natural log of total assets to measure firm size as it is the most commonly used firm size proxy in the aforementioned research.

ZFC = Zmijewski’s (1984) financial condition score, which indicates the degree of financial distress.

Previous studies have found that more financially distressed firms are more likely to receive going-concern opinions (e.g., Bronson et al., 2009; Bruynseels & Cardinaels, 2013; Carcello & Neal, 2000; Knechel & Vanstraelen, 2007; Mutchler et al., 1997). Following Bronson et al. (2009), Bruynseels and Cardinaels (2013), Carcello and Neal (2000) and DeFond et al. (2002), I measure financial distress using the coefficients of Zmijewski’s probit model from the sample of 40 bankrupt and 800 non-bankrupt firms. The Zmijewski score is calculated as follows:

Zmijewski score = –4.336 – 4.513X1 + 5.679X2 + 0.004X3

Where:

X1 = Net income/Total assets

X2 = Total liabilities/Total assets

X3 = Current assets/Current liabilities

As a higher score indicates more financial distress, I expect the coefficient of ZFC to be positive.

(24)

19 Return = Annualized monthly stock returns.

Vol = Annualized monthly stock volatility.

Consistent with previous GCO research (e.g., Bruynseels & Cardinaels, 2013; DeFond et al., 2002; Kaplan & Williams, 2012; Lim & Tan, 2008), I add market-based variables (Return and Vol) to further control for financial distress. In line with the aforementioned research, I expect the coefficient of Return to be negative and the coefficient of Vol to be positive.

PriorGCO = Indicator variable that equals 1 if a going-concern opinion was issued in the prior year, 0 otherwise.

Previous studies indicate that issuance of a GCO in the prior year increases the likelihood that a GCO is issued in the current year (e.g., Bronson et al., 2009; Bruynseels & Cardinaels, 2013; Carcello & Neal, 2000; DeFond et al., 2002; Mutchler, 1985; Wu et al., 2016). As such, I expect the coefficient of PriorGCO to be positive.

RepLag = Reporting lag, i.e., the number of days between fiscal year-end and the annual report filing date.

I include RepLag as a control variable because several studies have shown a positive relation between reporting lag and GCO issuance (e.g., Bruynseels & Cardinaels, 2013; DeFond et al., 2002; Kaplan & Williams, 2012; Lim & Tan, 2008). I therefore expect the coefficient of RepLag to be positive.

Big4 = Indicator variable that equals 1 if the current audit firm is one of the Big 4, 0 otherwise.

Several studies on GCO issuance take into account whether or not a firm is being audited by a Big N auditor, as Big N auditors are expected to provide a higher audit quality and, as a result, are expected to be more likely to issue a GCO (e.g., Bruynseels & Cardinaels, 2013; Carcello & Neal, 2000; DeFond et al., 2002). A number of those studies (e.g., Bruynseels & Cardinaels, 2013; DeFond et al., 2002) find a significant positive relation between the

presence of a Big N auditor and the auditor’s propensity to issue a going-concern opinion. As such, I control for the presence of a Big 4 auditor by adding the indicator variable Big4, expecting the coefficient of this variable to be positive.

BoardSize = Total number of board members.

Another control variable that has been used in GCO research is the size of the board of directors (e.g., Bruynseels & Cardinaels, 2013; Carcello & Neal, 2000; Wu et al., 2016). On the one hand, a larger board may increase the quality of financial reporting because the

(25)

20

broader array of expertise and views of its members could improve monitoring (Bédard et al., 2004; Xie,Davidson, & DaDalt, 2003). On the other hand, a smaller board may be less encumbered by bureaucratic and communicative problems, and thus improve the quality of financial reporting (Bédard et al., 2004; Xie et al., 2003). As such, I control for board size by adding the variable BoardSize, and make no prediction on the relation between BoardSize and the auditor’s propensity to issue a GCO.

LnAge = Age of the firm, measured as the natural log of the number of years since the company was founded.

Firm age has been used as a control variable in several GCO studies, as younger firms are more likely to experience financial distress or receive a GCO (Callaghan, Parkash, & Singhal, 2009; DeFond et al., 2002; Knechel & Vanstraelen, 2007). I therefore control for firm age using LnAge¸ and expect the coefficient to be negative.

3.3. Simultaneous equations model

A potential problem with logistic regression analysis is that it does not take into account the possibility of endogeneity, i.e., the possibility that the independent variables of interest are correlated with the error term (Wooldridge, 2010). It seems probable that the level of equity compensation might affect the likelihood of GCO issuance, but at the same time, the

likelihood of GCO issuance might affect equity compensation as well. For example, the board of a distressed firm may, as a reaction to its distressed state, decide to increase equity-based director compensation in an attempt to increase monitoring efforts and improve performance. In other words, Equity and GCO might be endogenously determined, as both are likely to be related to financial distress. While I attempt to control for financial condition via a number of control variables, it is unlikely that these variables are a better indicator of financial distress than actual issuance of a GCO (DeFond et al., 2002). As a result, it is possible my results will be biased.

I attempt to mitigate such endogeneity bias by estimating GCO using a simultaneous-equations system, as such an approach generally controls for endogeneity better than single-equation logistic or probit regression (Kaplan & Williams, 2012; Larcker & Rusticus, 2010; Wooldridge, 2010). Given that my regression models have a dichotomous dependent variable (GCO) with a (presumed to be) endogenous regressor (Equity), I use maximum likelihood estimation (MLE) to fit my probit models for GCO. Given that both the independent and the endogenous dependent variable are dichotomous in Model 1, I use the MLE method for

(26)

21

bivariate probit models as described by Greene (2003, pp. 710-712) for Model 1. As the endogenous dependent variable in Model 2-4 is continuous rather than dichotomous, I estimate these models by instead using the MLE method for probit models with a continuous endogenous regressor as described by Wooldridge (2010, pp. 475-477).

I chose MLE over alternative approaches such as those proposed by Newey (1987) or Rivers and Vuong (1988) because MLE, while more computationally intensive, is more efficient than any such two-stage approaches (Heckman, 1978; Newey, 1987; Rivers & Vuong, 1988; Wooldridge, 2010). In addition, using Wooldridge’s (2010) probit model and Greene’s (2003) bivariate probit model should produce more consistent estimators than approaches involving separate 2SLS regressions of models for both the discrete (GCO) and the continuous (Equity) endogenous variable, as used by, for example, Kaplan and Williams (2012). This is because it avoids having to use a non-linear function in the first stage of 2SLS regression, which has been shown to generally lead to inconsistent estimates (Wooldridge, 2010).

The simultaneous-equations system consists of a structural model for GCO (see: Model 1a to 4a in Table 4-3, p. 32) in addition to a reduced-form model for Equity (see: Model 1b to 4b in Table 4-3, p. 32), and is specified as follows, with the instrumental variables denoted in bold:

GCO = β0 + β1Equity + β2LnTA + β3ZFC + β4Return + β5Vol + β6PriorGCO +

β7RepLag + β8Big4 + β9BoardSize + β10LnAge + ε1 (6)

Equity = β0 + β1LnTA + β2ZFC + β3Return + β4Vol + β5PriorGCO + β6RepLag +

β7Big4 + β8BoardSize + β9LnAge + β10Segments + β11CashCon + ε2 (7)

Where:

Segments = Number of business, operating, or geographic firm segments in a given firm-year.

CashCon = Three-year average of ((common and preferred dividends + cash flow used in investing activities – cash flow from operating activities)/total assets).

To effectively reduce endogeneity bias, a major challenge is finding one or more instrumental variables that are both exogenous and unique to the endogenous variable in question (Larcker & Rusticus, 2010); in other words, variables that are both exogenous to director equity-based

(27)

22

compensation and not correlated with GCO issuance. As such, I chose the number of firm segments and the degree to which a firm is cash constrained as instruments.

Segments

The first instrumental variable, Segments, equals the number of firm segments (business, operating, or geographic) a firm reported in a given year, a common proxy for firm

complexity (e.g., Lang & Stulz, 1994; Linck, Netter, & Yang, 2008; Markarian & Parbonetti, 2007). To avoid counting insignificant or inactive segments, I included only segments that reported sales amounting to at least 2% of total sales. Given that monitoring costs tend to increase with firm complexity (Bushman et al., 2004; Linck et al., 2008), prior studies have found that the proportion of directors’ compensation that is equity-based also tends to

increase with firm complexity (Bryan et al., 2000; Ertugrul & Hegde, 2008), likely in order to incentivize increased monitoring effort (Bushman et al., 2004; Linn & Park, 2005). I

therefore expect a positive relation between the number of segments and the proportion of total equity compensation in directors’ compensation packages.

In their study of the relations between the director compensation mix and a number of other factors, Bryan et al. (2000) find that the number of segments is positively related to the proportion of stock grants to total compensation, but negatively related to the proportion of stock options in the total compensation package. This is likely because (as mentioned earlier) the costs of monitoring increase with firm complexity, making it more difficult and costly to prevent excessive managerial risk-taking in complex firms. Given the limited downward risk inherent in options, option awards provide a weaker incentive for directors to mitigate such overinvestment than stock grants (Bryan et al., 2000; Hovakimian & Hovakimian, 2009; Lynch & Williams, 2012). As such, it seems probable that complex firms are more willing to use stock than options when compensating directors. Following from this, I expect Segments to be positively related to the proportion of stock grants, and negatively related to the

proportion of option awards. At the same time, no relation between firm complexity and GCO issuance has been found in previous GCO research using sets of control variables similar to mine (e.g., Goh, Krishnan, & Li, 2013; Wu et al., 2016), making Segments a suitable instrumental variable.

Cash constraints

The second instrumental variable, CashCon, reflects the degree to which a firm is cash constrained. Consistent with Core and Guay (1999) and Carter, Lynch and Tuna (2007), I measure the degree to which a firm is cash constrained as the three-year average of

(28)

23

((common and preferred dividends + cash flow used in investing activities – cash flow from operating activities)/total assets). It has frequently been suggested that firms experiencing cash constraints use equity-based compensation as a substitute for cash compensation (e.g., Carter et al., 2007; Core & Guay, 1999; Dechow, Hutton, & Sloan, 1996; Ittner, Lambert, & Larcker, 2003; Yermack, 1995). I therefore expect a positive relation between cash

constraints and the proportion of total equity compensation in directors’ compensation packages.

Regarding the director compensation mix, Bryan et al. (2000) also find that cash constraints increase the proportion of option awards issued, while decreasing the proportion of issued stock grants. An explanation for this is that cash constraints increase the degree of costly underinvestment (Hovakimian & Hovakimian, 2009; Lamont, 1997). Given that option awards disincentivize underinvestment while stock grants can incentivize it (Bryan et al., 2000), and given that it is within managers’ power to prevent (to a certain extent)

underinvestment even under cash constraints (Hovakimian & Hovakimian, 2009), it seems likely that cash constrained firms reward directors with option awards more often than with stock grants in order to incentivize them to mitigate underinvestment. I therefore expect CashCon to be negatively related to the proportion of stock grants, and positively related to the proportion of option awards. In addition, Carey and Simnett (2006) and Defond et al. (2002) find no significant relation between GCO issuance and their proxy for cash constraints when using models for GCO issuance that also include Zmijewski’s financial condition score, which seems to indicate that ZFC already effectively controls for the positive correlation that one might expect between financial distress and cash constraints. As a result, I consider CashCon to be a suitable instrumental variable2.

2_{Additionally, it should be noted that the coefficients of both Segments and CashCon are insignificant}

(29)

24

4. Results

In this section, the results of my analyses are presented. First, descriptive statistics on the sample are provided, along with a univariate analysis of the differences between the GCO and non-GCO groups. I subsequently present the results of the multivariate logistic regression analysis, followed by the results of the simultaneous-equations analysis intended to mitigate potential endogeneity bias. Lastly, a number of sensitivity checks are performed.

4.1. Descriptive statistics and univariate analysis

Table 4-1 presents descriptive statistics on the differences between the group of firms that received a GCO (n = 1,028) and the group of firms that did not (n = 19,438). Specifically, I compare the means, medians and standard deviations for the amount and composition of annual NED compensation, as well as for the control variables. I start by winsorizing all continuous variables at the 1st and 99th percentile in order to reduce the influence of outliers. Using Student’s t-tests, I then find that all the differences in means between the two groups are significant at (at least) the 5 percent level, except for the value of option awards.

However, as several of the variables examined have non-normal distributions, the Wilcoxon rank-sum test is likely to be more appropriate. Using this test, I find that the difference is significant for each variable tested.

Unsurprisingly, the average value of each type of NED compensation is lower in the GCO group, likely because firms that receive GCOs tend to be smaller than those that do not (Carcello & Neal, 2000; DeFond et al., 2002). Panel B indicates that equity compensation is quite common in both groups, as over 80% of NEDs’ compensation packages contain some form of equity compensation, and around 50% of average total compensation is equity-based. Panel B further indicates that, while GCO firms use equity compensation less often, equity compensation tends to form a slightly larger part of NED’s pay packages compared to non-GCO firms. In addition, non-GCO firms appear to be more likely to choose option awards over stock grants to compensate their directors.

Panel C presents the differences between the GCO and non-GCO group in terms of the control variables used. Consistent with my expectations and prior research (e.g.,

Bruynseels & Cardinaels, 2013; DeFond et al., 2002), the differences in ZFC, Return and Vol indicate that firms that receive a GCO are more financially distressed than those that do not. Also consistent with prior research (e.g., Bronson et al., 2009; Carcello & Neal, 2000; Wu et al., 2016), Panel C indicates that, on average, GCO firms are smaller, more likely to have received a GCO the previous year, have larger reporting lag, and are younger than non-GCO

(30)

25 Table 4-1: Descriptive statistics and univariate analysis

Panel A: Value of annual NED compensation (in thousands $)

GCO firms (n = 1,028) Non-GCO firms (n = 19,438)

Difference in means Student t-value Wilcoxon rank-sum z-value

Compensation Mean Median St. Dv. Mean Median St. Dv.

Cash 26.21 18.33 28.79 55.92 53.10 33.12 29.71 28.21*** 29.87*** Stock grants 14.87 0.00 39.06 54.26 37.50 61.58 39.38 20.29*** 24.34*** Option awards 25.81 0.72 51.10 27.98 0.00 60.23 0.22 1.13 8.68*** Other 0.22 0.00 0.97 0.29 0.00 0.97 0.07 2.38** 9.81*** Total 70.81 49.26 79.38 143.43 128.20 98.67 72.62 23.20*** 28.34***

Panel B: The NED compensation mix (in percentages)

Variable Mean Median St. Dv. Mean Median St. Dv.

EqInd 81.81 100.00 38.60 89.19 100.00 31.05 7.38 7.33*** 7.32***

EqStock 15.49 0.00 28.33 31.34 33.81 27.59 15.85 17.93*** 19.12***

EqOption 35.76 23.84 37.66 18.17 0.00 27.47 19.06 19.58*** 16.69***

EqTotal 51.30 52.86 36.93 49.53 52.26 25.84 1.77 2.09** 2.30**

(31)

26 Table 4-1: Descriptive statistics and univariate analysis (continued)

Panel C: Control variables

Variable Mean Median St. Dv. Mean Median St. Dv.

LnTA 2.55 2.36 1.85 6.41 6.41 2.12 –3.86 57.20*** 44.35*** ZFC 10.20 5.15 11.51 –1.31 –1.56 2.29 11.51 110.00*** 44.08*** Return –0.10 –0.36 0.86 0.12 0.07 0.55 –0.22 11.97*** 19.19*** Vol 0.29 0.26 0.14 0.13 0.11 0.08 0.16 60.86*** 40.00*** PriorGCO 0.67 1 0.47 0.01 0 0.12 0.66 130.01*** 96.34*** RepLag 91.35 90 22.70 65.84 60 16.56 25.51 47.11*** 39.03*** Big4 0.28 0 0.45 0.74 1 0.44 –0.46 32.70*** 31.88*** BoardSize 4.60 4 1.59 5.81 6 2.10 –1.22 18.34*** 18.73*** LnAge 2.61 2.83 1.14 3.37 3.40 1.04 –0.76 22.53*** 21.62***

Notes: *, **, *** indicate significance at the 10-percent, 5-percent, and 1-percent level, respectively (two-sided).

“NED” = Non-executive director. “GCO” = Going-concern opinion. “Other” compensation refers to reported compensation not classified as either cash, options, or stock, and consists of, for example, changes in pension plans or non-qualified deferred compensation. The other variable definitions can be found in section 3.2.

(32)

27

firms. Lastly, I also find that GCO firms are less likely to have a Big 4 auditor, and that they tend to have smaller boards. A possible cause of this finding, I suspect, is the strongly positive correlation these two variables have with firm size (see: Table C-1).

To check for potential multicollinearity problems, I also perform an analysis of the correlations between the variables used (see: Appendix C). The correlation matrix in Table C-1 shows that, except for EqStock-LnTA (correlation = 0.4C-1) and EqStock-RepLag (correlation = -0.31), correlations between the independent variables of interest and the control variables are less than 0.30. Additionally, Table C-2 presents the Variance Inflation Factors (VIFs) for the right-hand variables in each of the GCO-models used. The table shows that the VIFs are low, ranging from 1.09 to 1.22 for the independent variables of interest, and with the means ranging from 1.47 to 1.49. As such, there does not appear to be any indication of serious multicollinearity problems.

4.2. Logistic regression analysis

Table 4-2 presents the results of the logistic regression analyses of Model 1 to 4. The pseudo R2 of the models range from 62.96% to 63.04%, which is fairly high relative to comparable GCO research (e.g., Bruynseels & Cardinaels, 2013; Carcello & Neal, 2000; DeFond et al., 2002; Wu et al., 2016). A possible explanation for this difference is that I control for firms’ financial condition by not only using Zmijewski’s financial condition score (ZFC), but also by using market-based control variables (Ret and Vol), thus capturing the effect of financial condition on GCO issuance more effectively.

First, I use Model 1 to examine the effect of the presence of equity-based

compensation on the likelihood that a GCO is issued. Contrary to H1, I find that this effect is statistically insignificant (p=0.313), indicating that the fact that ACMs’ compensation

packages contain some form of equity compensation is not in itself predictive of GCO issuance.

Instead of measuring the effect of the presence of equity compensation, Model 2 is used to measure the effect that the proportion of equity compensation to total compensation has on the likelihood that a firm receives a GCO. I find that this proportion is significantly negatively related to GCO likelihood (p=0.015), supporting H2 and indicating that GCO issuance becomes less probable as ACMs are rewarded with more equity. Paired with the rejection of H1, these results seem to indicate that, while the proportion of equity

compensation is negatively related to GCO issuance, this effect becomes significant only at a sufficiently high proportion of equity-based compensation to total compensation.

The effect of audit committee equity-based compensation on financial reporting quality : evidence from going-concern opinions