Audit committee financial expertise and earnings
management: The moderating effect of gender
Student name: Ingmar C. Wallenburg Student number: 10003428
Date of final version: June 14, 2015 Word count: 16.195
MSc Accountancy & Control, variant Accountancy Amsterdam Business School
Faculty of Economics and Business, University of Amsterdam Supervisor: dr. B. Qin
! "!
Statement of Originality
This document is written by student Ingmar Wallenburg 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.
! #!
Abstract
This paper examines whether gender specifics have an effect on the relationship between audit committee financial expertise and earnings management. Current literature does not agree on whether having financial experts on the audit committee leads to lower levels of earnings management. As current literature apparently doesn’t take a certain element into consideration that might explain this, I examine these conflicting views by testing the joint effect of gender and audit committee financial expertise on earnings management to find out whether adding gender as a moderator has a negative effect on this relationship. The results indicate that having financial experts on the audit committee does not lead to lower levels of earnings management, as consistent with the current conflicting literature about this matter. In addition, I find that there is no evidence that the joint effect of gender diversity and audit committee financial expertise will cause female financial experts to be better at constraining earnings management than their male counterparts. Furthermore, the results of the additional analysis show that the interaction of gender diversity and audit committee financial expertise does not work better for the chair position in the audit committee.
! $!
Table of contents !
! 1. Introduction ... 5 1.1 Background...5 1.2 Research question ...61.3 Motivation and contribution ...6
1.4 Structure ...9
2. Literature review... 10
2.1 Theoretical background ...10
2.1.1 Agency theory...10
2.2 Earnings management ...11
2.2.1 What is earnings management? ...11
2.2.2 Types of earnings management ...12
2.3 Audit committee and financial expert...13
2.3.1 Audit committee ...13
2.3.2. Financial expert ...14
2.4 Financial experts on the audit committee and earnings management...15
2.5 Gender-based differences in psychology and business ...17
3. Research methodology ... 21
3.1 Sample and data collection...21
3.2 Estimating earnings management ...22
3.3 Regression models ...25 3.4 Control variables...27 4. Empirical results ... 29 4.1 Descriptive statistics...29 4.2 Multicollinearity test...31 4.3 Main analysis ...33 4.4 Additional analysis...38 5. Conclusion... 44 References ... 46 Appendix ... 50 ! !
! %!
1 Introduction
1.1 Background
Following the corporate and accounting scandals that occurred at the start of the 21st century, U.S. Congress enacted the bill of The Sarbanes-Oxley Act of 2002 (SOX). The objective of SOX is to strengthen the corporate governance mechanisms in order to protect the
shareholders and the public as a whole from misconduct. In particular, SOX increased the audit committees’ responsibilities and authority (SOX section 2, 2002). The oversight role of the audit committee was broadened and more specified. The audit committees were now responsible for overseeing the financial reporting process and the auditor activity more comprehensively in order to ensure the reliability of financial reporting (Buchalter and Yokomoto, 2003).
According to prior literature, certain audit committee characteristics can restrain the use of earnings management (e.g. Xie, Davidson III and DaDalt, 2003; Ghosh, Marra and Moon, 2010). Ghosh et al. (2010) find that firms with larger audit committees are less likely to use earnings management. This could mean that a higher number of members on the audit committee increases the total knowledge of the committee, which would enable them to monitor the financial reporting process more effectively. Moreover, they find that earnings management does not decrease when there is a financial expert on the audit committee. Xie et al. (2003) examined the role of the board of directors, the audit committee, and the executive committee in preventing earnings management. Their findings show a negative relationship between several audit committee characteristics and earnings management. They describe that audit committees that have more independent outside members or members with corporate or banking investment experience are associated with lower levels of earnings management. Xie et al. (2003) also find that there is a negative relationship between the frequency of meetings of the audit committee and earnings management.
One of the requirements set by SOX (section 407, 2002) regarding audit committees is that public companies have to disclose whether a financial expert is included on the audit committee. This thesis is in particular interested in that aspect of the audit committee, and focuses on the relationship between financial experts on the audit committee and earnings management.
There could be factors that have a moderating effect on the relationship between financial experts on the audit committee and earnings management. A potential variable that could have a moderating effect is gender (e.g. Thiruvadi and Huang, 2011; Vähämaa, 2014).
! &!
Several studies find that firms with female CFOs (who can be seen as financial experts) are associated with less earnings management than male CFOs (e.g. Barua, Davidson, Rama and Thiruvadi, 2010; Vähämaa, 2014). In addition, some studies have examined the association between female audit committee members and earnings management. Thiruvadi and Huang (2011) and Ittonen, Miettinen and Vähämaa (2010) find that the presence of female directors on the audit committee leads to lower levels of earnings management. The presence of female audit committee members enhances the effectiveness and monitoring activities of the audit committee, which reduces the risk of material misstatements. Thus, gender has a moderating effect in these studies.
1.2 Research question
This study is an attempt to give the reader more insight into the relationship of gender,
financial expertise on the audit committee and earnings management. In addition, the purpose of this study is to examine the relationship between financial experts on the audit committee and earnings management, and whether gender will have a moderating effect on this
relationship. Based on the information that is provided in the background section the following research question can be formulated:
“Do gender specifics have an effect on the relationship between financial experts on the audit committee and earnings management?”
1.3 Motivation and contribution
As I’ve described in the background section, prior literature states that certain audit
committee characteristics are negatively associated with earnings management. However, the literature is contradictory when it comes to the relationship between financial expertise on the audit committee and earnings management. There are various studies that suggest there is a negative relationship between financial experts on the audit committee and earnings
management (e.g. Xie et al., 2003; Bedard, Chtourou and Courteau, 2004; Abbott, Parker and Peters, 2004; Lin and Hwang, 2010). Bedard et al. (2004) examined, among other things, the relationship between the expertise of the audit committee (financial, governance, and firm specific) and aggressive earnings management. They find that there is indeed a negative relationship between the presence of a financial expert on the audit committee and aggressive earnings management.
! '!
However, there are also studies that don’t find a negative relationship between
financial experts on the audit committee and earnings management (e.g. Lin, Li and Yang, 2006; Ghosh et al., 2010; Sharma and Kuang, 2014). Lin et al. (2006) examined i.a. the relationship between the financial expertise of the audit committee and the occurrence of earnings restatements. They find that there is no significantly negative relationship between audit committee financial expertise and the occurrence of earnings restatements. This result suggests that having financial experts on audit committees has no impact on the quality of financial reporting. Badolato, Donelson and Ege (2014) describe that this might be because of the new requirements of SOX (2002). These new requirements caused a high demand for financial experts on the audit committee. This means that firms can’t always get the financial experts on their audit committee they want and may have to accept lower status financial experts instead. These lower status experts may not have the same ability in decreasing material misstatements.
One would expect a negative relationship between financial expertise on the audit committee and earnings management. This is because directors with financial expertise have better knowledge and are also better at monitoring internal controls and audit activities. As a result, directors with financial expertise are better at detecting and preventing misstatements (Abbott et al., 2004; Zhang, Zhou and Zhou, 2007). Therefore, this paper hypothesizes that financial experts on the audit committee is negatively associated with earnings management. Multiple studies suggest that women are more conservative, risk averse and ethical than men and that this affects the decision-making (e.g. Byrnes, Miller and Schafer 1999; Heminway, 2007; Eckel and Grossman, 2008; Croson and Gneezy, 2009).!Croson and
Gneezy (2009) find that women are less confident than men about their performance in a task, and men are more likely to participate in risky situations than women and that women are more risk averse and conservative than men in risky situations. Beckmann and Menkhoff (2008) find that also among financial experts, women are more conservative, risk averse and less confident than men. These gender-based differences may have important implications for the quality of financial reporting and effectiveness of the audit committee (Ittonen et al., 2010; Barua et al., 2010).
As I mentioned before in the background, gender has in multiple studies a moderating effect. In various settings, female CFOs and female audit committee members lead to less earnings management in firms. This has been attributed by the authors of these studies to the fact that women are more conservative, risk averse and ethical than men, which affects the decisions females make in their position (e.g. Ittonen et al., 2010; Barua et al., 2010;
! (!
Thiruvadi and Huang, 2011; Vähämaa, 2014). For example, Thiruvadi and Huang (2011, p.486) mention: “[...] gender differences affect conservatism, managerial opportunism and risk preference of the management. Based on the consistent findings of prior studies, we hypothesize that the presence of female director in the audit committee strengthens the external governance of the firm and its auditor, thereby leading to reduced earnings management.” In addition, Peni and Vähämaa (2010) state that because women are more trustworthy than men, they are less likely to engage in earnings management (p. 632). Furthermore, Huse and Solberg (2006) find that women on corporate boards are in general better prepared than men for board meetings. This might be because they are less confident in their task than men. They also found that men lack preparation, which reduces board member’s independence and supports managerial dominance. As women do prepare well they are not as dependent on reports made by management. Moreover, women are more likely to be in compliance with rules and regulations than men, also in business settings (Baldry, 1987; Wenzel, 2002). Barua et al. (2010, p.27) state that: “females are more cautious and more likely to be compliant with accounting regulations.”
Lastly, Singh, Terjesen and Vinnicombe (2008) and Adams, Gray and Nowland (2010) find that new female directors on corporate boards generally have a higher level of accounting related education than new male directors. They also describe that a higher level of education has a positive effect on the development of independent thinking. Adams et al. (2010) further argue that female directors are capable of thinking more independently than men, which improves the effectiveness of the monitoring role they have. Hence, because female directors have higher educational qualifications than their male counterparts they are more effective in the oversight and monitoring activities they perform.
Based on the information from all studies I provided in the second part of this
paragraph, I think that financial expertise on the audit committee works better in combination with female characteristics. This in turn should result in lower levels of earnings management. Hence, gender would have a moderating effect. This is because females are more likely to be compliant with rules and accounting regulations than men. When financial experts on the audit committee are stricter in compliance of rules and accounting regulations, this should enhance the financial expertise of the audit committee, as the oversight and monitoring activities will be performed more thoroughly. Secondly, female directors on corporate boards are more likely to have a higher level of accounting related education than male directors. This would directly enhance audit committee financial expertise as the financial expert has more knowledge regarding the financial reporting process, accounting policies and the audit
! )!
process. The enhanced financial expertise that female financial experts provide will reduce the risk of misconduct like earnings management. I think that the joint effect of these female characteristics will be that female financial experts will make it more difficult for
management to engage in earnings management than their male counterparts. Therefore, this paper hypothesizes that the joint effect of gender diversity and audit committee financial expertise will cause female financial experts to be better at constraining earnings management than their male counterparts.
As I stated earlier, there are conflicting studies regarding whether having a financial expert on the audit committee is negatively associated with earnings management. This means there is a possible variable that has been left out in the previous studies that has an effect. Current literature only looks at the relationship between financial experts on the audit committee and earnings management. My contribution to the current literature would be to study what the joint effect is of a financial expert on the audit committee and gender
(moderator) on earnings management. This is the first paper to do so. The implication of this study may be that organisations will pay attention to whether they will hire a female financial expert on the audit committee in order to increase its financial expertise and constrain
earnings management.
1.4 Structure
The remainder of this paper will be organized as follows. In section two the literature review and hypotheses development will be discussed. Section three will discuss the research methodology and sample selection. The empirical results are presented in the fourth section. To finish, the last section will present the conclusions and limitations.
! *+!
2 Literature Review
I will first explain the underlying theory regarding this thesis and after that I will explain earnings management itself and the different forms of earnings management. Next, I will tell how the audit committee is defined by the Sarbanes-Oxley Act of 2002, what the role is of the audit committee and its responsibilities. Lastly, I will describe how the Securities and
Exchange Commission defines a financial expert.
2.1 Theoretical background
2.1.1 Agency theory
The agency theory is a much-discussed theory throughout the academic literature and is a relevant theory for this thesis. The agency theory is directed towards the analysis of the principal-agent relationship. The principal-agent relationship involves a party (the principal) that delegates work to a different party (the agent) (Eisenhardt, 1989). An example of a principal-agent relationship is the shareholder (the principal) of a company and the manager (the agent) who has to act on the principal’s behalf.
The structure of this relationship can cause agency problems. Agency problems can arise because the manager might have different interests than the shareholder, which can conflict with each other. Secondly, because of information asymmetry, it’s impossible for the shareholder to completely verify the actions of the manager in order to ensure whether the manager has indeed behaved accordingly. Asymmetric information in this context means that the manager has more information about his/her actions and skills than the shareholder does. As a result, the manager can act in his or her own self-interest. This is called ‘moral hazard’. As I just briefly mentioned, the shareholder has less information regarding the skills of the manager. The manager can, ex ante, claim to have certain skills and abilities. This can give rise to ‘adverse selection’, as the shareholder isn’t able to verify beforehand whether the manager actually possesses these skills or abilities (Picard, 1987; Eisenhardt, 1989; Scott, 2011).
Earnings management is a good example of an agency problem. Managers (agents) try to maximize their own interests at the cost of the shareholders (principals) by engaging in earnings management. The inaccurate financial information provided by the managers to the shareholders can cause the shareholders to make non-optimal investment and operational decisions, which can be seen as agency costs. Managers do this in order to serve their own interests (Davidson III, Jiraporn, Sang Kim and Nemec, 2004). This is an example of moral
! **!
hazard (earnings management) that is caused by information asymmetry between the managers and shareholders (Richardson, 2000)
Audit committees can be used to decrease information asymmetry by monitoring managers. As there is more complete information about the activities of managers they are more likely to act in the interests of shareholders. This decreases the likelihood of agency problems such as earnings management (Eisenhardt, 1989; Richardson, 2000; Beasley, Carcello, Hermanson and Neal, 2009). In order for audit committees to be effective, directors need to understand the financial, accounting and audit issues confronting the firm’s managers. A financial expert who has more expertise regarding these matters will be better able to provide effective oversight and reduce information asymmetry regarding accounting and auditing matters than other audit committee members who have a different background (Abbot et al., 2004). This way, financial experts on the audit committee enable audit
committees to adequately address and detect material misstatements and therefore prevent or decrease agency costs such as earnings management. Hence, because of financial experts, managers are more likely to act in the interests of shareholders (Abbot et al., 2004; Zhang, Zhou and Zhou, 2007).
In addition, gender could be an important factor to further decrease information asymmetry and agency costs. This is because female directors on corporate boards have higher educational qualifications than male directors and prepare better for board meetings, which makes female directors less dependent on reports made by management (Huse and Solberg, 2006; Singh et al., 2008; Adams et al., 2010). Based on this, female financial experts could be more effective in their monitoring and oversight activities than their male
counterparts and therefore provide more information about managers, which lessens
information asymmetry, decreases agency costs such as earnings management and aligns the interests of managers and shareholders.
2.2 Earnings management
2.2.1 What is earnings management?
In the previous section, the agency theory was described that is important to this thesis. In this section the theory of earnings management will be explained. The academic literature
provides several definitions for earnings management. I will state two definitions that are often used to define earnings management. Schipper (1989, p. 92) defines earnings management as follows:
! *"!
“[…] a purposeful intervention in the external financial reporting process, with the intent of obtaining some private gain (as opposed to, say, merely facilitating the neutral operation of the process) […]”
A more extensive definition is provided by Healy and Wahlen (1999, p. 368). They state the following:
“Earnings management occurs when managers use judgment in financial reporting and in structuring transactions to alter financial reports to either mislead some stakeholders about the underlying economic performance of the company or to influence contractual outcomes that depend on reported accounting numbers.”
2.2.2 Types of earnings management
There are two types of earnings management: accrual-based earnings management and real earnings management.
Accrual-based earnings management
Accrual-based earnings management is achieved by changing the accounting methods or estimates used when presenting a given transaction in the financial statements. It involves managers making accounting choices with the intention to deliberately bias accounting information by means of discretionary accruals. Examples of discretionary accruals are: changing the depreciation method for fixed assets, the estimate for provision for doubtful accounts and a firm may record excessive liabilities for product guarantees, contingencies and rebates. This can bias reported earnings in a particular direction without changing the
underlying transactions (Zang, 2011; Scott, 2011). There are also non-discretionary accruals, which are caused by real economic activities. For example, in times of economic prosperity a firm is likely to have higher sales. This will lead to a change in (non-discretionary) accruals such as accounts payable and accounts receivable without managers engaging in earnings management. Therefore, non-discretionary accruals are not part of accrual-based earnings management (Islam, Ali and Ahmad, 2011).
! *#!
Real earnings management
Real earnings management involves managers making real economic decisions that deviate from the normal operational practices with the intention to deliberately influence accounting information. Managers try to mislead stakeholders by making them believe certain goals have been achieved by normal operations, which isn’t the case. Unlike accrual-based earnings management, real earnings management can affect current and future cash flows directly (Roychowdhury, 2006). Examples of real earnings management can be influencing sales through discounts and cutting desirable research and development investments to boost current-period earnings. Because of this, real earnings management can be harmful to a firm's long-term profitability and competitive advantages. Managers engage in real earnings
management because it’s harder to detect than accrual-based earnings management. With regard to real economic decisions, there is no benchmark that can determine what the best decision would be in a specific situation. Therefore, even if bad decisions are made, the managers can’t be held accountable. In contrast, accrual-based earnings management is subject to monitoring and scrutiny by auditors, courts and regulators who have accounting standards as a benchmark. (Lo, 2008; Kim and Sohn, 2013).
2.3 Audit Committee and financial expert
2.3.1 Audit committee
The Sarbanes-Oxley Act of 2002 (SOX) increased audit committees’ responsibilities and authority. It raised membership requirements and committee composition to include more independent directors. Companies were also required to disclose whether or not a financial expert is on the audit committee (SOX section 2, 2002).
The Sarbanes-Oxley Act (SOX section 2, 2002) defines an audit committee as:
(A) 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; and
(B) if no such committee exists with respect to an issuer, the entire board of directors of the issuer.
! *$!
In the early stages the role and responsibilities of the audit committee were still limited compared to now. The SEC recommended in 1972 that publicly held firms should establish audit committees in order to strengthen the protection to investors further. Back then the audit committee was primarily responsible for the oversight of the financial reporting process, receipt of audit results and the selection of the independent auditor. However, later on regulatory reforms increased the audit committees’ responsibilities and authority. Most
notably, the report of the Blue Ribbon Committee (BRC) and the Sarbanes-Oxley Act of 2002 (SOX). The BRC (1999) provided recommendations for improving the effectiveness of the oversight role the audit committee has in order to improve the financial reporting. As I already have stated above, SOX increased audit committees’ responsibilities and authority (SOX section 2, 2002). It is because of these regulatory reforms that several institutions like the SEC stepped up and proposed new regulations and rules to strengthen audit committees. Audit committees would now have a more prominent role within companies and the oversight role of the audit committee was broadened and more specified. New responsibilities are for example, overseeing the auditor activity and process and overseeing the financial reporting process more thoroughly, including the review of both the scope and the results of work performed by the internal and external auditors, the review of critical accounting choices and the review and discussion of all financial reports (BRC, 1999; Vera-Munez, 2005).
2.3.2 Financial expert
Section 407 of SOX (2002) requires public companies to disclose whether a financial expert is included on the audit committee. If this is the case, the company has to disclose the name of that expert and if the expert is independent of management. When a company does not have at least one financial expert on its audit committee the company must disclose why it does not have such an expert. Even though SOX made the proposition of section 407, it was not responsible for the precise definition of the term “financial expert”. The Securities and Exchange Commission (2003) had the responsibility to define “financial expert”. With the release of the final rule by the SEC, the SEC defined a financial expert as having:
1. An understanding of generally accepted accounting principles and financial statements; 2. The ability to assess the general application of such principles in connection with the accounting for estimates, accruals and reserves;
! *%!
3. Experience preparing, auditing, analyzing or evaluating financial statements that present a breadth and level of complexity of accounting issues that are generally comparable to the breadth and complexity of issues that can reasonably be expected to be raised by the registrant's financial statements, or experience actively supervising one or more persons engaged in such activities;
4. An understanding of internal controls and procedures for financial reporting; and 5. An understanding of audit committee functions.
The audit committee is responsible for overseeing the financial reporting and audit process. The oversight role of the audit committee needs to ensure the reliability of financial reporting in order to protect the shareholders and the public as a whole from misconduct like earnings management (SOX section 2, 2002; Buchalter and Yokomoto, 2003). However, in order for the audit committee to be able to monitor the financial reporting process effectively, directors need to understand the financial, accounting and audit issues of the firm. A financial expert who has more expertise will be able to provide effective oversight regarding accounting and auditing matters. This way, financial experts on the audit committee should be able to address and detect material misstatements and therefore prevent misconduct such as earnings
management (Abbot et al., 2004; Zhang, Zhou and Zhou, 2007).
However, literature is conflicting whether having a financial expert on the audit committee is indeed negatively associated with earnings management. Some studies find a negative relationship, and some don’t (e.g. Xie et al., 2003; Bedard et al., 2004; Lin et al., 2006; Badolato et al., 2014). In order to fill the gap in the current literature, this paper will research whether there is indeed a negative association between having financial experts on the audit committee and earnings management. I will contribute to the current literature by adding a moderator (gender), and find out its effects on this relationship.
2.4 Financial experts on the audit committee and earnings management
Prior literature has extensively researched the relationship between audit committee characteristics and earnings management. Several studies have investigated the samecharacteristics and found similar results. Based on the studies conducted by Xie et al. (2003), Abbott et al. (2004), Bedard et al. (2004) and Sharma and Kuang (2014) audit committee independence has found to be negatively associated with earnings management. In addition,
! *&!
Ghosh et al. (2010) and Goh (2009) find that larger audit committees are more effective in limiting earnings management.
However, prior literature is conflicting when it comes to the relationship between financial expertise on the audit committee and earnings management. Does having a financial expert on the audit committee constrain earnings management? Some studies argue that audit committees with a financial expert on the board doesn’t lead to lower levels of earnings management (e.g. Lin et al., 2006; Qin, 2007; Sharma and Kuang, 2014; Badolato et al., 2014). But there are also studies that say otherwise, and conclude that a financial expert on the audit committee does lead to lower levels of earnings management (e.g. Xie et al., 2003; Abbott et al., 2004; Bedard et al., 2004).
I will now discuss some studies to give some more depth regarding the subject matter. Abbott et al. (2004) examine if financial expertise on the audit committee is negatively
associated with the occurrence of financial restatements. Among other things, the regression results show that having at least one financial expert on the audit committee is significantly, negatively associated with the occurrence of financial restatements (p-value < 0.05). This result suggests that having financial experts on the audit committee increases the financial reporting quality. Another research that shows there is a negative relationship was conducted by Bedard et al. (2004). They examined whether the quality of financial information is influenced by the audit committee’s expertise. In particular, if financial expertise had any influence on the levels of earnings management as measured by the amount of discretionary accruals. They find that the presence of at least one financial expert on the audit committee reduces the likelihood of earnings management in a firm.
Prior studies are conflicting regarding this topic. Sharma and Kuang (2014) studied the association between audit committee characteristics (e.g. financial expertise) and earnings management in New Zealand. They find that independent directors with financial expertise on the audit committee are negatively associated with earnings management. However, there is no significant negative relationship between non-independent directors possessing financial expertise and earnings management. Hence, this suggests that financial expertise alone is not sufficient to reduce earnings management. This is also shown by Badolato, Donelson and Ege (2014). They investigated the joint effects of financial expertise on an audit committee and the status of the audit committee on earnings management. The results show that audit committees with a high status and financial expertise are negatively related with earnings management. While audit committees with only financial expertise and no relative high status are not negatively associated with earnings management. In addition, Qin (2007)
! *'!
demonstrates that only financial experts with accounting expertise are associated with a higher quality of reported earnings.
Audit committee’s with financial experts have better knowledge and are better at monitoring internal controls and audit activities. Because of this, the effectiveness of internal controls is increased in preventing and detecting material misstatements. Also, audit committee’s with financial expertise are better able in understanding audit issues, risks and the procedures as a whole. This way, financial experts on the audit committee enable audit committees to
adequately address and detect material misstatements (Abbott et al., 2004; Zhang et al., 2007). Consequently, I expect that the presence of financial experts on the audit committee will be negatively associated with earnings management. This leads to the first hypothesis of this thesis:
H1: Financial experts on the audit committee are negatively associated with earnings management.
2.5 Gender-based differences in psychology and business
Whether females are more conservative, risk averse and ethical than men is an important (economic) issue. If females are indeed more conservative, risk averse and ethical than men, this could affect their decision making in all aspects. A lot of studies in sociology, psychology and experimental economics have examined these gender-based differences. A lot of which conclude that women are indeed more conservative, risk averse and ethical than men (e.g. Byrnes, Miller and Schafer 1999; Fehr-Duda, Gennaro, and Schubert, 2006; Heminway, 2007; Eckel and Grossman, 2008). Some studies that examined this issue found out that gender is related to asset allocation. The portfolios managed by women are less risky than that by men (e.g. Jianakoplos and Bernasek, 1998).
However, the problem with empirical studies is that there is too much variation in the methods researchers use. This makes it hard to compare results and judge whether the
majority of the literature agrees whether females are indeed more risk averse and conservative than men. Charness and Gneezy (2012) solved this problem by gathering data from existing studies that vary in a lot of ways but are all based on the same investment game. To their own surprise they find a very clear and consistent answer: women are more risk averse than men. Beckmann and Menkhoff (2008) find that the financial expertise of fund managers doesn’t
! *(!
overrule gender-based differences, which means that also among financial experts, women are more conservative, risk averse and less confident than men. Pierce and Sweeney (2010) find that females are more ethical than males regarding the likelihood to engage in auditor dysfunctional behavior. Females have less intention to engage in auditor dysfunctional behavior compared to males. Also, females have a higher ethical judgment and perceive a higher ethical intensity than males. They also demonstrate that males perceive greater
pressure to engage in unethical behavior than females to achieve success and profitability for the firms. The authors state that this is in line with other studies that say that women are more likely to obey rules than men. In addition, Peni and Vähämaa (2010) state that because
women are more trustworthy than men, they are less likely to engage in earnings management (p. 632).
Important, but unanswered questions remain. Why are women more conservative, risk averse and ethical than men and how do gender-based differences affect the decisions of women in the business world? Croson and Gneezy (2009) find that there are three different aspects: emotions, the interpretation of risky situations and overconfidence. Croson and Gneezy (2009) mention that, regarding emotions, women experience emotions more strongly than men. Specifically, women are more nervous and feel more fear than men when a
negative outcome is expected. On the contrary, men tend to feel angry in identical situations, which makes them evaluate a given situation as less risky than women would. Because women are more likely to be afraid of losing than men, this translates into women
overweighting the probability of losing. Secondly, men are more likely to interpret a risky situation as something that could be a challenge, which stimulates men and increases the likelihood men participate in risky situations. Whereas women are less likely to participate as they interpret risky situations as threats and respond with avoidance.!Lastly, men are more overconfident than women in uncertain situations. They say that women are less confident than men and that, in general, men tend to be more overconfident than women about their performance in a task (Croson and Gneezy, 2009).
A parallel can be drawn with women versus men on corporate boards. Male directors tend to feel confident about their capabilities and don’t feel the need to prepare well before a meeting. Huse and Solberg (2006) found that women on corporate boards are in general better prepared than men for board meetings. Lack of preparation reduces board member’s
independence and supports managerial dominance. But because female directors do prepare well, they, and therefore the board, become less dependent on presentations and reports made by management, which increases the effectiveness of the board and reduces the likelihood of
! *)!
misconduct like earnings management.
Moreover, women are more likely to be in compliance with rules and regulations than men (Baldry, 1987, Wenzel, 2002). Wenzel (2002) finds that regarding tax compliance, women are more compliant than men in income reporting and deduction claims. This argument is also seen in business studies. Barua et al. (2010) examine the relationship between the gender of CFO and the quality of accruals. They find that female CFOs report lower levels of total and current discretionary accruals and have lower accrual estimation errors. They attribute these results to: “females are more cautious and more likely to be compliant with accounting regulations. This leads to the argument that female CFOs are less likely to be aggressive in making judgments related to discretionary accruals.” (p.27). In addition, Srinidhi, Gul and Tsui (2011) find that female representation on the audit committee is associated with higher quality of accruals and earnings.
I have mentioned several examples that show that gender-based differences affect decisions women make. This is something that is also seen on the audit committee. Thiruvadi and Huang (2011) find that the presence of female directors on the audit committee improves the quality of financial reporting and increases the effectiveness of corporate governance. They say that this means that the presence of female directors on the audit committee leads to lower levels of earnings management. The reason they provide why this is the case is because the effects of gender-based differences cause the presence of women on audit committees to have a positive effect in their monitoring and oversight role (p. 484). Ittonen et al. (2010) come to similar findings and state that gender-based differences improve the corporate
governance, as female directors on the audit committee help to better perceive potential biases in strategy formulation and risk assessments. This in turn can enhance the effectiveness and monitoring role of the audit committee (p.118; p.136). It appears that gender has a moderating effect in these studies as well.
Singh et al. (2008) find that new female directors on corporate boards generally have a higher level of education than new male directors. In particular, the proportion of new female directors who have a MBA degree is greater than that of their male counterparts. This is consistent with Adams et al. (2010) who find that female directors hold higher educational qualifications than men. In addition, Singh et al. (2008) describe that a higher level of education has a positive effect on the development of independent thinking. Adams et al. (2010) further argue that female directors are capable of thinking more independently than men, which improves the effectiveness of the monitoring role they have. Hence, because female directors have higher educational qualifications than their male counterparts they are
! "+!
more effective in the oversight and monitoring activities they perform.
Based on the information from all studies I provided in paragraph 2.5, I think that financial expertise on the audit committee works better in combination with female characteristics. This in turn should result in lower levels of earnings management. Hence, gender would have a moderating effect. To underpin this opinion, I’ll provide some arguments that are based on the female characteristics discussed in the abovementioned studies. Firstly, females are more likely to be compliant with rules and accounting regulations than men. When financial experts on the audit committee are stricter in compliance of rules and accounting regulations, this should enhance the financial expertise of the audit committee, as the oversight and monitoring activities will be performed more thoroughly. Secondly, female directors on corporate boards are more likely to have a MBA degree than male directors. This would directly enhance audit committee financial expertise as the financial expert has more accounting and auditing knowledge. As a result, the audit committee will be better able to detect misconduct like earnings management. In addition, female directors prepare better for board meetings than men. Financial expertise of the audit committee should be enhanced even further when the financial expert prepares better for board meetings, as she will have more knowledge than her male counterpart who is more likely to attend meetings without preparation. In sum, I expect that these three main female characteristics will lead to more effective oversight and increased knowledge regarding the financial reporting process, accounting policies and the audit process, and therefore, enhance the financial expertise of the audit committee. The enhanced financial expertise that female financial experts provide will reduce the risk of misconduct like earnings management. Consequently, I think that the interaction between ‘financial expertise’ and ‘gender’ on earnings management will cause female financial experts to be better at constraining earnings management than their male counterparts. Hence, gender would have a moderating effect. This leads to the second hypothesis of this thesis:
H2: The joint effect of gender diversity and audit committee financial expertise will cause female financial experts to be better at constraining earnings management than their male counterparts.
! "*!
3 Research methodology
3.1 Sample and data collection
In this study I will perform an archival research. The sample will be based on the U.S. market. I will gather data from ISS (formerly RiskMetrics) regarding gender and whether a director on the audit committee is a financial expert. Next, I will use Compustat to gather data about earnings management. Lastly, I will retrieve analyst forecast data from I/B/E/S. In order to conduct my research I will use the maximum available years and firms in the database. The firms in these databases roughly correspond with the S&P 1500. The period is 2007 till 2013. I exclude financial institutions with Standard Industrial Classifications (SICs) between 6000 and 6999 because these firms are heavily regulated, this could be problematic when
computing discretionary accruals for these firms (e.g. Burgstahler and Eames, 2003). I have winsorized the data at the 1st and 99th percentiles of distribution to address the potential impact of outliers.
Table 1: Sample description of main sample!
Sample selection Compustat ISS Total firm observations between 2007-2013 99,787 96,679 SIC codes 6000-6999 (38,487)
Missing values for accruals (12,012) Other missing values (4,625) Deleting duplicates for merging (5,625) Less not matched observations (31,564)
Less other than audit committee members (57,182)
Deleting duplicates for merging (29,843)
Less not matched observations (2,180)
! ""!
3.2 Estimating earnings management
Accrual-based earnings management: discretionary accruals
Following prior literature (e.g. Dechow, Sloan and Sweeney, 1995; Cohen, Dey and Lys, 2008; Zang, 2011), I use, amongst other things, discretionary accruals to proxy for accrual-based earnings management. Discretionary accruals are the difference between the actual (total) accruals and the normal level (non-discretionary) of accruals in a firm. Dechow et al. (1995) investigated several models to find out what model is best to measure the level of earnings management. I estimate the discretionary accruals by using the Cross-sectional Modified Jones Model of Dechow et al. (1995), as this is the most powerful test of earnings management (Dechow et al., 1995; Cohen et al., 2008). I exclude SIC two-digit industries with less than 10 observations each year in order to mitigate the industry-specific effects on accruals. This will lead to less bias in the model.
The reason why I don’t use the original Jones Model (1991) is because Dechow et al. (1995) demonstrate that the original Jones Model contains an inherent problem. The original Jones Model makes the assumption that managers have no discretion regarding revenues. Management, however, does have this discretion. So, if management uses this discretion to manage earnings, the Jones model doesn’t include the related discretionary accruals. For example, consider a situation where a company recognises revenues at year-end, even though it hasn’t received any cash yet and it’s questionable whether these revenues have been earned. As revenues are recognised earlier by using managerial discretion, discretionary accruals arise through an increase in accounts receivable. The Jones Model would exclude this part of the discretionary accruals, causing the estimate of earnings management to be less accurate (Dechow et al., 1995). Therefore, Dechow et al. (1995) adjusted the original Jones Model and made some small improvements. The change in revenues is adjusted for the change in
receivables in the event period (Dechow et. al, 1995, p.199). This adjustment will make the estimation of earnings management more accurate.
TA represents total accruals defined as: TAit = EBXIit – CFO. In order to calculate this in Compustat I took the income before extraordinary items and deducted the operating activities net cash flow and the extraordinary items and discontinued operations.
! "#!
After calculating the total accruals for each firm, each year, estimations can be made of !1, !2 and !3 by applying the OLS (Ordinary Least Squares) regression. After the OLS regression has determined the estimations for !1, !2 and !3, the calculation can be made for the non-discretionary accruals, with the residual being the non-discretionary accruals. The result is the formula following the Modified Jones Model:
(1)
Where:
i = Stands for firm i
TAit = The total accruals for the year t
Ai,t-1 = The total assets for the year t-1
"REVit = Revenue in year t less revenue in year t-1
"RECit = Net receivables in year t less net receivables in year t-1
PPEit = The gross property, plant and equipment in year t
!it = The error term in year t, which is the residual of the model (1) and is interpreted as the
discretionary accruals.
!1, !2 and !3 = coefficients in the model
Measures for accrual-based earnings management and real earnings management To increase the validity of this research I will add more measures rather than taking just discretionary accruals as a proxy for earnings management. These additional measures will capture both accrual-based earnings management as well as real earnings management. Consistent with Krishnan, Raman, Yang & Yu (2011) I will measure earnings management by looking at how firms engage in earnings management to avoid failing to meet or beat the analysts’ consensus forecasts on earnings, avoid reporting a loss and avoid reporting an earnings decline.
When firms are able to just meet or beat analysts’ earnings forecasts this can be a sign that those firms have engaged in earnings management (Matsumoto, 2002; Graham, Harvey and Rajgopal, 2005; Burgstahler and Eames, 2006). Graham et al. (2005) describe that firms are willing to sacrifice economic value in order to hit earnings targets. Firms are willing to engage in (real) earnings management by postponing or eliminating hiring, R&D, advertising,
! "$!
or even investments to meet or beat earnings targets. They even state that: “[...] many managers would reject a positive NPV project in order to meet the analyst consensus
estimate!” (p.37). In addition, Burgstahler and Eames (2006) find that firms that just meet or beat analysts’ earnings forecasts also use discretionary accruals to achieve this. Graham et al. (2005) also mention that managers prefer to meet or slightly beat the earnings targets instead of surprising the market every time with massive positive earnings surprises as this could harm the firms’ credibility or lead to ever increasing analysts’ earnings forecasts. The main motivations as to why managers want to meet or beat earnings targets so badly is because if a firm isn’t able to do so, it creates suspicion or uncertainty in the market whether there might be deep, previously unknown problems at the firm or its future prospects (Graham et al., 2005). In order to analyze this measure I will look at the analyst forecast error, which is calculated by actual EPS (earnings per share) less analysts’ consensus earnings forecast. This will be defined as an indicator variable with a value of 1 if the analysts’ earnings forecast error is between # 0 and $ 0.01, and 0 otherwise, as consistent with Krishnan et al. (2011):
Analyst forecast error = Actual EPS – Forecasted EPS (2)
The other two measures – avoid reporting a loss or a decline in earnings – have also been found to be associated with earnings management. Burgstahler and Dichev (1997) find that firms engage in earnings management (both real and accrual-based) to avoid reporting losses and earnings decreases. In particular, firms turn a decline in earnings into a small increase or avoid reporting a loss by “[...] both ‘real’ operating actions, reflected in cash from operations, and actions of a ‘bookkeeping’ nature, reflected in discretionary accruals [...]” (Burgstahler and Eames, 2006 p. 651). The benefits earnings management provides increases substantially when a decline in earnings can be managed into a small increase or when a small loss can be managed into small positive earnings. In addition to the abovementioned reasons as to why managers engage in earnings management, there is also the lower transaction costs firms face with stakeholders when they report higher earnings. What’s more, when a loss or a decline in earnings can be avoided, firms face better loan terms as there is a lower likelihood of default or delay in loan payments (Burgstahler and Dichev, 1997). In order to analyze firms that avoid reporting a loss or avoid reporting a decline in earnings I will go through the scaled earnings and scaled earnings changes, respectively. This will be defined as an indicator
variable with a value of 1 if the net income divided by the market value of equity is between # 0 and < 0.02, and 0 otherwise for the ‘avoid a loss’ measure. The ‘avoid an earnings decline’
! "%!
measure will be defined as an indicator variable with a value of 1 if the change in net income from year t-1 to t divided by the market value of equity at the end of year t-2 is # 0 and < 0.01, and 0 otherwise, as consistent with Krishnan et al. (2011):
Scaled earnings = Net Income/ Market Value of Equityt-1 (3)
Scaled earnings changes = "Net Income/Market Value of Equityt-2 (4)
3.3 Regression models
The first regression model is related to the first hypothesis. This model will try to find
whether there is a significant relationship between earnings management and audit committee financial expertise. %1 represents the association between earnings management and an audit committee having a financial expert on the audit committee. I predict for the first hypothesis that there is negative relationship between earnings management and audit committee financial expertise. Therefore, I expect %1 to be a significantly negative coefficient.
EM1 = #0 + #1ACfinexp + #2ACfemale + #3SIZE + #4GROWTH + #5LEV +
#6ROA + #7LOSS + #8BIGN + ! (5)
The interaction between audit committee financial expertise and gender has been added in the second empirical model with %3 to test the second hypothesis. %3 represents the interaction between gender and financial expertise on the audit committee and the relationship between this interaction and earnings management. I predict that the joint effect of gender diversity and audit committee financial expertise will cause female financial experts to be better at constraining earnings management than their male counterparts. Therefore, I expect %3 to be a significantly negative coefficient. The description of the control variables and the relationship of these control variables with earnings management will be explained in the following paragraph.
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
! "&!
EM2 = #0 + #1ACfemale + #2ACfinexp + #3ACfemale x ACfinexp + #4SIZE +
#5GROWTH + #6LEV + #7ROA + #8LOSS + #9BIGN + ! (6)
Table 2: Description of the variables used in the regression models
Variable Description
DA The discretionary accruals.
EM1 (Scaled Earnings) Indicator variable equal to 1 if net income divided by the market value of equity # 0 and < 0.02, and 0 otherwise (Krishnan et al., 2011).
EM2 (Scaled Earnings Changes) Indicator variable equal to 1 if the change in net income from year t-1 to t divided by the market value of equity at the end of year t-2 is # 0 and < 0.01, and 0 otherwise (Krishnan et al., 2011).
EM3 (Analyst Forecast) Indicator variable equal to 1 if analyst
forecast error # 0 and $ 0.01, and 0 otherwise (Krishnan et al., 2011).
ACfemale Fraction (%) of audit committee that is
female.
ACfinexp Fraction (%) of audit committee that is a
financial expert.
ACfemale x ACfinexp The interaction between audit committee
financial expertise and gender. This is calculated by multiplying ACfinexp with ACfemale.
SIZE The natural logarithm of total assets.
GROWTH The market-to-book ratio.
LEV LEV is the leverage of the firm, measured as
total debt divided by total assets.
ROA The return on assets is included to control for
differences in performance, measured as net income divided by total assets.
LOSS Indicator variable equal to 1 if in a particular
year a loss occurred, and 0 otherwise.
BIGN Indicator variable equal to 1 if the auditor is a
Big4 audit firm, and 0 otherwise. !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
"!The dependent variable ‘EM’ varies in four forms: DA, EM1, EM2 and EM3. !
! "'!
3.4 Control variables
!I will include several control variables to control for effects that might have an influence on earnings management. The control variables used in this research are the following. I will control for firm size (SIZE) because audit firms are less likely to intervene and adjust earnings management made by big clients (Nelson, Elliott and Tarpley, 2002). Secondly, Barton and Simko (2002) argue that large firms are more pressured to meet or beat analysts’ forecasts, which might tempt firms to engage in earnings management. Lastly, firms might also be tempted to engage in earnings management to reduce potential political costs (Warfield, Wild and Wild, 1995). Based on these studies I believe firm size is positively associated with earnings management.
The market-to-book ratio will be used as a proxy to control for firms’ growth
(GROWTH). It is argued that firms with growth potential show low levels of transparency and therefore have greater opportunities to engage in earnings management (Meek, Rao and Skousen, 2007). Consequently, I expect that growth is positively associated with earnings management. The leverage of firms (LEV) will also be included in the empirical model as a control variable. When firms are close to the violation of debt covenants, managers are more likely to shift income from future periods to the current period to avoid these violations (DeFond and Jiambalvo, 1994: Scott, 2011). Therefore, I expect a positive relationship between leverage and earnings management.
The return on assets (ROA) is included to control for differences in performance. It is measured as net income divided by total assets. I don’t have an expectation for ROA. I will also control whether firms are audited by a Big 4 audit firm (BIGN). Prior literature states that Big 4 audit firms provide higher audit quality than non-Big 4 audit firms. This is because Big 4 audit firms are less likely to be economically dependent on a particular client than non-big 4 audit firms. This makes them better able to resist client pressure on earnings management. Moreover, Big 4 audit firms possess greater in-house experience and expertise than non-Big 4 audit firms (Becker, DeFond, Jiambalvo and Subramanyam, 1998; Francis and Yu, 2009; Choi, Kim, Kim and Zhang, 2010). Based on this, I predict that Big 4 audit firms are negatively associated with earnings management.
Lastly, I will control for firms that have incurred a loss (LOSS). When firms experience periods of financial stress, reorganization or when firms have to report a loss, management might decide to use income-decreasing discretionary accruals and take a bath. They do this so management won’t have to do this in the future and “clear the decks”.
! "(!
the future. This will increase the probability of future reporting profits (Scott, 2011).
Furthermore, reporting a loss will heavily increase transaction costs with stakeholders. This creates incentives for firms to avoid losses and engage in earnings management (Burgstahler and Dichev, 1997). As a result, I expect a positive relationship between firms that incur a loss and earnings management.
Table 3: Estimated direction of the variables used in the regression models
Expected sign
Variable
DA/EM1/EM2/EM3 ACfemale - ACfinexp - ACfemale x Acfinexp - SIZE + GROWTH + LEV + ROA ? LOSS + BIGN -! ")!
4 Empirical Results
4.1 Descriptive statistics
In the first section of this chapter the descriptive statistics are described. As I have mentioned in 3.1, the sample used in this thesis is based on the U.S. market, in particular, the S&P 1500 from the period 2007-2013. Data has been gathered from ISS, Compustat and I/B/E/S. Financial institutions with SIC codes between 6000-6999 were deleted from the sample. Observations that did not have complete information regarding the calculation of
discretionary accruals or contained other missing values were also deleted. Lastly, duplicates and observations that could not be matched were deleted from the sample as well. This resulted in a final sample of 7,474 firm year observations as presented in table 1.
Table 4 shows the distribution of firm-year observations by industry. The distribution of the observations shows that the observations are evenly divided over the sample years. With 938 observations, 2007 is the year with the least observations in the sample. The most
observations are observed in the year 2012 with 1,147 observations. Lastly, there are 1560 uniquely identifiable firms in the total sample.
Table 5 presents the distribution of all industries that are present in the sample (see appendix). The sample contains firms from 52 different industries, based on the two-digit SIC code. The industry ‘automotive repair, services and parking’ with the corresponding SIC code 75 has the lowest frequency; there are 12 observations reported. Industry 73 has the highest amount of observations in the sample: ‘Business services’ has 860 reported observations.
! #+!
Table 6 shows the descriptive statistics of all the variables from the regression models that are used in the analyses. The table shows that the dependent variable ‘discretionary accruals’ (DA) has a mean of -.0248858, which is comparable to prior studies (Becker et al., 1998; Cohen et al., 2008). The means of EM1, EM2 and EM3 are, respectively, .057742, .1890897 and .0534282. Regarding the independent variables, the table shows that about 47% of an audit committee consists of financial experts (ACfinexp). An audit committee consists for roughly 13% out of female members (ACfemale). With respect to the control variables, the following descriptive statistics can be obtained. Big 4 audit firms (BIGN) audit the vast
majority of the firms. The table shows that over 88% of the sample is being audited by a Big 4 audit firm. Size has an average value of 7.44. The growth of firms (GROWTH) is determined by the market-to-book ratio, which has a mean of 2.69. Firm performance is measured as net income divided by total assets (ROA), which has a percentage of 2.8. The average ratio of leverage (LEV) is .18, which means that 18% of the firm’s assets is financed by debt. Lastly, the descriptive statistics show that 18% of the sample shows a loss as seen over the total sample period.
Table 6: Descriptive statistics of variables used in empirical models
a
Because of missing values for scaled earnings, 893 observations were deleted.
b Because of missing values for scaled earnings changes, 1498 observations were deleted. c
Because of merging a third dataset (I/B/E/S) with the final sample and missing values for analyst forecast error, 867 observations were deleted.
! #*!
4.2 Multicollinearity test
A multicollinearity test is performed to test the correlation between independent variables that are used in the regression models. In this research the Pearson correlation coefficient is used, as this is the most used coefficient in the literature. The correlation coefficient measures the strength of the linear relationship between the dependent variable and the independent
variables. The value of the Pearson correlation coefficient can vary between -1 and 1. A value of -1 indicates a perfect negative linear relationship. And a value of 1 indicates a perfect positive linear relationship. When the value is 0, there is no linear relationship. If the Pearson coefficient has a value that is either too high or too low it can affect the reliability of the model, which means that those variables should be excluded from the analyses. To determine what is “too high” or “too low” I will use a benchmark that Dancey and Reidy (2004) provide in their paper. A Pearson correlation coefficient higher than .7 or lower than -.7 influences the reliability of the model. Table 7 shows the Pearson correlation matrix. All the outcomes are within the acceptable range of -.7 and .7.
When examining table 7 further, most of the coefficients of the dependent variables go to the predicted direction, as based on prior literature. For ACfinexp, the coefficients of DA, EM1, EM2 and EM3 show a negative value, which is consistent with what I described in section 3. For the second independent variable, ACfemale, three out of four earnings management measures show the predicted sign for the coefficients: DA 0.0097), EM1 (-0.0418) and EM3 (-0.0031), with EM1 having a significant value. This means that if the fraction of female members on the audit committee increases, earnings management
decreases. Regarding the control variables, the proxies for earnings management are mostly significantly positively related with both SIZE and GROWTH. As expected, firms that have a loss or are highly leveraged are more likely to engage in earnings management.
Moreover, ACfinexp and ACfemale are both significantly positively correlated with both BIGN and SIZE. This can be interpreted that larger firms, and firms that are being audited by a Big 4 audit firm, are associated with more financial experts and females on the audit committee. ACfemale is significantly negatively correlated with LOSS, which means that a higher amount of female members in the audit committee is associated with a lower likelihood of a loss. In addition, BIGN is significantly positively correlated with SIZE, and significantly negatively correlated with LOSS. This means that larger firms are more likely to be audited by a Big 4 audit firm than smaller firms. And firms that are being audited by a Big 4 audit firm are less likely to experience a loss. Lastly, ROA is significantly negatively correlated with LOSS, which is rather intuitive. This is because ROA measures firm
Table 7: Pearson Correlation Matrix
performance. If the firm performs badly, there is a higher likelihood the firm will experience a loss.
4.3 Main analysis
In addition to the Pearson correlation matrix presented in table 7, I perform the VIF test (variance inflation factor) to assess the severity of multicollinearity between the independent variables of the empirical models related to the hypotheses. The literature provides several recommendations as to what levels of VIF are acceptable. The general rule of thumb is that if VIF has a value of 10 or above, this warrants further investigation (e.g. Kennedy, 1992; Hair, Anderson, Tatham, & Black, 1995). However, some literature provide a lower maximum value for VIF. For example, Rogerson (2001) recommends a maximum VIF value of 5. The results of the VIF tests for the first hypothesis are presented in table 8. Column DA (1.77) contains the highest VIF value of all columns, and column EM1 (1.68) has the lowest VIF value. In any case, all VIF values are well below the various maximum values stated in the literature.
Table 8 also presents the regression results of the empirical models related to hypothesis 1 in which I predict that audit committee financial expertise is negatively related to earnings management. The results in column DA show that ACfinexp is negatively related to
discretionary accruals. However, the coefficient of ACfinexp (-0.035) is not significant. The second independent variable, ACfemale, shows a negative coefficient as well (-0.063). Again, this coefficient is not significant. Hence, both independent variables show a coefficient that is in line with the expected direction. All control variables show a coefficient that is significant at the 0.01 level. The control variables BIGN (-0.154), SIZE (0.019), ROA (0.723), LEV (0.186) and LOSS (0.166) are all in line with the expected direction as described in section 3, except for GROWTH (-0.014).
The results in column EM1 show that ACfinexp is negatively related to scaled earnings. However, the coefficient of ACfinexp (-0.118) is not significant. The second independent variable, ACfemale, shows a negative coefficient as well (-0.897). This coefficient is significant at the 0.05 level, which would mean that the presence of female members on the audit committee leads to lower levels of earnings management. Thiruvadi and Huang (2011) find the same results in their study. Of all control variables, only SIZE (-0.134) is significant. SIZE is negatively associated with earnings changes, which is the opposite direction from what I expected. Habib and Bhuiyan (2011) provide a possible explanation for
! "#!
this. They say that larger firms have a higher likelihood of implementing strong internal controls than smaller firms.
The results in column EM2 show that ACfinexp is negatively related to scaled earnings changes, which is in line with the predicted direction. However, the coefficient of ACfinexp (-0.159) is not significant. The second independent variable, ACfemale, shows a positive coefficient (0.032) that is not significant. This was not expected, as the predicted sign is positive. Regarding the control variables, all variables are significant at the 0.01 level, except for BIGN (-0.275), which is significant at the 0.1 level. The control variables BIGN (-0.275),
SIZE (0.206), GROWTH (0.094), and ROA (2.003) are all positively related with scaled
earnings changes as consistent with what I described in section 3. Both LEV (-0.989) and
LOSS (-1.316) are negatively associated with scaled earnings changes. The negative association of LOSS was not expected, but is not surprising as Bédard et al. (2004) and Sharma and Kuang (2014) show a negative direction for loss.
The results in column EM3 show that no variable is significantly related with analysts’ earnings forecasts error. The coefficients of both ACfinexp (-0.221) and ACfemale (-0.052) are negative. Hence, both independent variables show a coefficient that is in line with the expected direction. Of the control variables most of the variables show the expected direction of the coefficients. More specifically, this is applicable to BIGN (-0.136), GROWTH (0.109),
ROA (0.896) and LEV (0.032).
As for the independent variables, I can conclude that both gender and audit committee financial expertise does not lead to lower levels of earnings management. For gender, the coefficients of three out of four proxies are not significant. This means that the presence of females on the audit committee does not lead to lower levels of earnings management, which is inconsistent with the findings of Thiruvadi and Huang (2011), who state that the effects of gender-based differences cause the presence of women on audit committees to have a positive effect in their monitoring and oversight role, which reduces the level of earnings
management. As described above, the coefficients of audit committee financial expertise are not significant in all four proxies of earnings management. This means that financial experts on the audit committee are not negatively associated with earnings management as consistent with the findings of e.g. Sharma and Kuang (2014) and Badolato et al. (2014). Hence, the first hypothesis of this research is rejected.
