Amsterdam Business School
Faculty of Economics and Business, University of Amsterdam
Audit Committee characteristics and
earnings quality -the status since 2008-
Samson, R. (2015)Rico Samson 10644814
MSc Accountancy and Control, track Accountancy Supervisor: dhr. dr. A. (Alexandros) Sikalidis Word count: 12453
2 Statement of Originality
This document is written by Student Rico Samson, 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
Motivated with the enactment of the Sarbanes-Oxley Act in 2002 by the Securities and Exchange Commission to enhance the audit committee effectiveness and thus the quality of the earnings, I examine if audit committee characteristics (independence, size, meetings, financial expertise and ownership) are related to earnings quality for the period of 2008 till 2013. Earnings quality is measured through earnings management.
Using one measure of discretionary accruals and three of real activities manipulation, my results suggest that audit committee size is positively related to earnings quality. Indicating that large audit committees are more productive in managing the financial reporting process and are better in restricting firms to reduce their discretionary expenditures. Also, I find that audit committees with more financial experts are positively related to earnings quality. The increase of financial experts in the audit committee leads to more available specialized knowledge and hence better at executing their overseeing duties. In contrast, no evidence is found with respect to independence, meetings and ownership.
This research contributes to the literature on corporate governance, audit committee effectiveness and earnings quality.
Keywords: audit committee characteristics, corporate governance, earnings quality, earnings management, discretionary accruals, real activities manipulation
4
Acknowledgments
With this acknowledgement, I would like to take a moment to express my thankfulness to my thesis supervisor Dr. Alexandros Sikalidis from the University of Amsterdam for the guidance throughout the thesis period. Especially, I want to acknowledge Alexandros for his patience and everlasting positive attitude which helped me to finalize this thesis. Furthermore, I would like to thank my supervisor Ronald Bakker from Deloitte for repeatedly making time for me even if he had a full agenda and the all-time willingness to review my thesis. Ronald showed he has an exceptional critical view which I will take with me along my further professional career. My gratitude also goes to my fellow student colleague Ari Iancovici. You were always there for thesis related questions but even more helpful were the good laughter after long days of writing. Finally, a big thanks to my parents for all your support, I dedicate this thesis to you.
5
Table of contents
1. Introduction ... 6
2. Literature review and hypothesis development ... 10
2.1 Corporate governance and the role of the audit committee ... 10
2.2 Audit committee characteristics ... 10
2.2.1 Audit committee independence ... 11
2.2.2 Audit committee size ... 11
2.2.3 Audit committee meetings ... 12
2.2.4 Audit committee financial expertise ... 12
2.2.5 Audit committee ownership ... 14
2.3 Earnings quality ... 14
2.3.1 Earnings management ... 14
2.3.1.1 Discretionary accruals ... 15
2.3.1.2 Real activities manipulation ... 15
3. Research design ... 18
3.1 Research method ... 18
3.2 Empirically testing of discretionary accruals ... 18
3.3 Empirically testing of real activities management ... 19
3.4 Regression models ... 21
3.4.1 Relation between discretionary accruals and audit committee characteristics ... 21
3.4.2 Relation between real activities manipulation and audit committee characteristics ... 22
3.5 Control variables ... 24
3.6 Sample and data collection ... 25
4. Results ...27
4.1 Descriptive statistics ... 27
4.2 Correlation matrix ... 28
4.3 Relation between earnings management and audit committee characteristics ... 30
4.3.1 Analyzing the R-squared in the relation between discretionary accruals and audit committee characteristics ... 30
4.3.2 Analyzing the R-squared in the relation between real activities manipulation and audit committee characteristics ... 30
4.3.3 Effect of audit committee characteristics on discretionary accruals ... 31
4.3.4 Effect of audit committee characteristics on real activities manipulation ... 32
4.3.5 Analyzing hypotheses ... 34
5. Conclusion ...35
Appendix A. Variables description ... 36
Appendix B. Regression of the total accruals ... 37
6
1. Introduction
In middle of the 1990s and the 2000s, Arthur Levitt (chairman of the SEC) cited in his “Numbers Game”:
“I fear that we are witnessing an erosion in the quality of earnings, and therefore, the quality of financial reporting.” (Levitt, 1998)
Consequently, the SEC responded by enhancing the disclosure and accounting rules together with the overseeing function for the audit committee. All with the purpose to enhance the quality of the audit committees. The ultimate objective is to make the aforementioned committee more qualified, committed, independent and tough-minded (Levitt, 1998; and Duncan, 2001).
In this study, I examine the effect of multiple audit committee characteristics (independence, size, meetings, financial expertise and ownership) on the quality of earnings for the years 2008 till 2013. Earnings quality is measures through earnings management using one proxy for discretionary accruals and three for real activities manipulation.
Throughout the time numerous studies examine the impact of corporate governance on financial reporting in the US, Canada, Europe and Asia (e.g., Osma and Noguer, 2007; Klein, 2002; Shen and Chih, 2007; Peasnell, 2005; Park and Shin, 2004), they mainly use data from the 1980s and 1990s, when the regulatory environment was noticeable different (Ghosh, 2010). Namely, no requirement was in place to appoint financial experts on the audit committee. Also the audit committee was not accountable for selecting and dismissing the auditor (Cohen et al., 2004). Given the enactment of the Sarbanes-Oxley Act of 2002 to intensify the board’s oversight role, I focus on whether these reforms have effect on the quality of earnings. Based on recent research from Ghosh et al. (2010) (research on U.S. listed firms) and Al-Matari et al. (2014) (research in Oman), which focuses on the years 1998-2005 and 2011-2012, my data sample consists of the years 2008 till 2013. The most recent data is used to assess whether regulation in the United States has increased the earnings quality by means of the level of earnings management.
My study can potentially contribute to the late dispute on controls of corporate governance for audit committees and the movement towards increasing similarity of corporate governance across countries (Yoshikawa and Rasheed, 2009; Xia, 2009; Lannoo and Khachaturyan, 2004). Based on current research (e.g. Ghosh et al., 2010; Xie et al., 2003; Cohen et al., 2004; and Klein, 2002), I specifically analyze the amount of independent members that forms the audit committee (independence), the total amount of members in the committee (size), the frequency of the audit committee meetings (meetings), the amount of financial experts in the committee
7 (financial expertise) and the ownership by means of shares owned by the member of the audit committee (ownership).
First, audit committee independence, Xie et al. (2003) state that in cases of specific instances of agency problems, independent outside directors are better in protecting shareholders. The audit committee is accountable with overseeing management to safeguard the interests of shareholders, hence one would assume that more independent outside directors are negatively related to earnings management. However, Xie et al. (2003) and Ghosh et al. (2010) find that earnings management does not vary with the percentage of outside directors on the audit committee. These results are contrary to the results from Klein (2002), which state that independent members are effective in guarding the financial reporting and hence it is more probable that engaging in earnings management will be constrained. Nevertheless, according to Xie et al. (2003) earnings management appears less usual when boards include more independent outside directors in combination with more corporate experienced (financially) directors. SOX demand that the audit committee include wholly independent outside directors since dependent inside directors have little incentives to counter managerial reporting discretion (SEC, 2003). If independent outside directors under the new regulations are better in discouraging earnings management, I expect that earnings quality will increase and hence a positive relation is assumed between independence and earnings quality.
Second, with respect to size, the audit committee helps a firm to improve its performance and therefore improve the quality of earnings and restrain earnings management. Al-Matari et al. (2014) discuss two theories with regard to the size of an audit committee namely, the agency theory and the resource dependence theory. The agency theory argues that a firm will demonstrate poor performance when the size of the audit committee gets bigger (Al-Matari et al., 2014). Numerous authors examined this and found a negative association between audit committee size and firm performance (Bozec, 2005; Al-Matari et al., 2012; Hsu and Petchsakulwong, 2010; Mollah and Talukdar, 2007). On the contrary, the resource dependence theory states that when the size of the audit committee gets bigger, its performance is enhanced. Less members in the audit committee lacks in the field of skills and knowledge and this makes the committee ineffective. Forthwith, an audit committee with the perfect amount of directors enables them to make use of their full experience and expertise in advantage of the stakeholders (Pfeffer, 1987; and Pearce and Zahra, 1992). To demonstrate, Reddy et al. (2010), Bauer et al. (2010), Premuroso and Bhattacharya (2007) and Swamy (2011) find a positive relation between audit committee size and firm performance in countries such as New Zealand, US, India. According to Beasley and Salterio (2001) in the case of more audit committee members, firms are more likely to enlarge the amount of the committee
8 with more independent outside directors apart from the mandated minimum requirements. Hence, regarding the conflicting theories, the way of the relationship between audit committee size and earnings quality is twofold.
Third, the activity of an audit committee may contribute in restraining earnings management. Audit committees have the responsibility to be proactive and adopt a critical attitude towards the financial reporting (matters) (Turner, 2001). Hence, more frequent meetings reasonably predict higher quality of reporting and therefore, the proactive and critical attitude suggests lower level of earnings management for companies with more frequent audit committee meetings (Ghosh et al., 2010). On the other hand, the frequency of meetings can also indicate as a reactive measure rather than a proactive (Vafeas, 1999). Jensen (1993) state that audit committees often increase the frequency of their meetings in the presence of problems. Similar to audit committee size, I also predict that the relationship between earnings quality and audit committee activity is ambiguous.
Fourth, in order to monitor the financial reporting process, which entails in-depth knowledge of technical rules and/or accounting standards, members with refined financial background are likely to be successful in constraining earnings management (Ghosh et al., 2010). This is supported by Xie et al. (2003) namely, they state that a member with a financial background may be more familiar with the ways that earnings can be managed and may better understand implications of earnings management. SOX requires that a firm discloses whether the audit committee includes a financial expert and if not, why (SEC, 2003). If members with a better financial background perform better with regards to monitoring the financial reporting, hence, more financial experts in the audit committee less occurrence of earnings management is expected. I expect that the relation between financial expertise and earnings quality is positive.
At last, Ghosh et al. (2010) measures the ability of the audit committee using two different criteria’s namely, the tenure of an audit committee member and stock owned by audit committee members. With respect to the research of Ghosh et al. (2010), I will focus solely on the share ownership by audit committee members. Jensen (1993) finds that members with more share ownership results in better performance regarding their duties in accordance to shareholders’ interest. The probability of managerial fraud in the financials declines with the amount of stock owned by outside audit committee members (Beasley, 1996). Hence, I expect a positive relation between ownership and earnings quality.
In order to yield evidence for the aforementioned assumptions, I provide a relation between the audit committee characteristics and earnings quality. Earnings quality is determined by the level of earnings management. Earnings management is measures through one proxy for
9 discretionary accruals (based on the modified Jones Model from Dechow et al. 1995) and three for real activities manipulation (based on prior research from Kim et al. (2012) and Roychowdhury (2006). Audit committee characteristics are specified in five parts; independence, size, meetings, financial expertise and ownership. To rule out any alternative causes, I control for certain aspects related to audit committee characteristics.
The sample firms consist of 5939 which is a set containing the S&P400, 500 and 600 classifications. This set represents the years 2008 till 2013. Audit committee characteristics information is gathered from ISS (formerly RiskMetrics) and information on discretionary accruals and real activities manipulation is derived from COMPUSTAT, similarly for the control variables.
The findings of my research provide significant relations between size and earnings quality and financial expertise and earnings quality. The first relation is measured based on the level of the abnormal value of the discretionary expenses and the second relation is measured based in the level of discretionary accruals. Looking at the other characteristics, no evidence is found to support the hypotheses.
The remainder of the paper is structured as follows. Section 2 elaborates the relevant literature and elaborates the relationship between audit committee characteristics and earnings quality by developing the hypotheses. Section 3 describes the research design and portrays the data sample and the sample collection. Section 4 represents the descriptive statistics, correlation matrix and empirical results drawn from the regression analysis. The conclusion is presented in Section 5.
10
2. Literature review and hypothesis development
2.1 Corporate governance and the role of the audit committee
More than a decade ago, major financial reporting scandals occurred, such as Enron and also the collapse of Arthur Anderson took place. Due to this, corporate governance issues came to the forefront of major reform efforts aimed at enhancing the quality of the financial reporting process (SEC, 2003). In particular, requirements to improve corporate controls and reducing conflicts of interest, with respect to audit committees. According to the SEC (2003), an audit committee is defined 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; and if no such committee exists with respect to an issuer, the entire board of directors of the issuer.”
In theory, the need for corporate governance rests on the idea that when separation exists between the ownership of a company and its management, self-interested executives have the opportunity to take actions that benefit themselves. In this case shareholders and stakeholders are responsible for the costs of these actions (Larcker and Tayan, 2011). Because of the separation of corporate management and ownership, supervisory boards are in place to protect the shareholders’ interest. This is due to managers that may not always act in the best interest of shareholders (Jensen and Meckling, 1976). Meaning that the board of directors’ and supervisory board are responsible to oversee management activities. Because the multiple responsibilities of both functions within corporate governance, duties have to be delegated. Hence, the role of the audit committee includes providing oversight of (1) the accuracy and integrity of the firm’s financial statements, (2) compliance with legal and regulatory requirements, (3) the appointment of external auditors, (4) the effectiveness of the firm’s internal audit functions, (5) the company’s accounting process, principles and reporting, and (6) the company’s audits. The board of directors can influence the financial reporting process by selecting ‘high-quality’ audit committees (Dobija, 2015; Cohen et al., 2004; DeZoort et al., 2002).
2.2 Audit committee characteristics
The presence of an audit committee indicates greater monitoring and diligence on the part of board members (Vafeas, 1999). Therefore, the board of directors’ frequently uses standing committees, such as the audit committee, to enhance its effectiveness. According to Ghosh et al. (2010), audit committee effectiveness is determined by means of the following attributes: (1) independence, (2) size, (3) meetings, (4) financial expertise, and (5) ownership.
11 2.2.1 Audit committee independence
Klein (2002) states that audit committees, which are independent, minister as exceptional monitors of the financial reporting process. Hence, independent members are more effective in restricting earning manipulation. The late amendments SEC (2003) oblige that the audit committee solely includes independent members, because non-independent members have less incentives to counter managerial reporting discretion. For this reason, earnings quality is assumed to increase due to more effective audit committee members under the new regulations (Ghosh et al., 2010). Hence, a positive relation between audit committee independence and earnings quality is assumed. However, it could be possible that the results of Klein (2002) may not hold for the post-SOX years. Because between 1999 and 2003, the major exchanges in the US required that audit committees be ‘mostly’ independent and the SEC (2003) mandates that audit committees consist of independent directors only (Ghosh et al., 2010). Therefore, the probability is present that there is no shift in audit committee independence between the pre-SOX and post-SOX period.
H1: Ceteris paribus, the presence of solely independent audit committee members is positively related to earnings quality.
2.2.2 Audit committee size
Jensen (1993) claims that a streamlined committee is better in overseeing management. Including the contrasting views on a streamlined committee, one could argue that a smaller amount of members in the committee is more efficient in overseeing the financial reporting process and thus restricting the likelihood for managers to engage in earnings manipulation. Also, a smaller streamlined committee is argued to be more efficient in decision-making discussion and have lower incremental cost of poorer communication (Jensen, 1993). Otherwise, one could also argue that large audit committees are better in overseeing the financial reporting process. Beasley and Salterio (2001) concludes that when the size of the audit committee enlarges, it is more probable that firms comprise independent members beside the required minimum amount. This will result in more effectiveness of the audit committee. Likewise, problems concerning audit committees are perceived to be complex and technical. Thus, financial expertise in financial reporting is required to assess the severity of these problems (Bull and Sharp, 1989; McDaniel et al., 2002).
Due to the contrasting perspectives and together with the regulatory initiatives which impose no maximum audit committee size, the direction of the relation between audit committee size and earnings quality is inclusive, however a positive relation is expected between size and
12 earnings quality based on the result of Beasley and Salterio (2001) and the new amendment (SEC, 2003).
H2: Ceteris paribus, firms with more audit committee members (size) are positively related to earnings quality.
2.2.3 Audit committee meetings
Audit committee meetings are twofold. On the one hand, more activity indicates the level of caution and scrutiny practiced by the members. Naturally, audit committee members possess a proactive attitude and ask intrusive questions about the financial reporting process (Turner, 2001). Consequently, when the frequency of audit committee meetings has a proactive nature, suggested is that earnings quality is higher for firms with more periodically meetings. On the other hand, an increase in meetings can also signify the existence of issues. Audit committees are more likely to schedule more meetings in case of significant financial reporting and auditing topics. In this case the frequency of meetings is perceived as a reactive measure (Vafeas, 1999). Hence, the frequency of meetings serves both as a reactive and a proactive measure. For this reason, the direction of the relationship between earnings persistence and audit committee activity is ambiguous. But for this research, I expect a positive relation between audit committee meetings and earnings quality based on aforementioned findings from Turner (2001).
H3: Ceteris paribus, firms with frequent audit committee meetings are positively related to earnings quality.
2.2.4 Audit committee financial expertise
In order to perform efficient monitoring on the financial reporting process, members must consist of relevant financial expertise in order to restrict earnings management. Since the implementation of SOX, it is mandatory that a firm disclose whether financial experts are included in the audit committee. If there are zero financial experts, the firm has to disclose why (SEC, 2003). Concerning related research, Bedard et al. (2004) and Carcello et al. (2006) state that when at least one member has financial expertise, this relates to a lower probability of aggressive earnings management. Correspondingly, Puat Nelson and Devi (2013) finds that firms with accounting affiliated directors and managerial experiences are negatively associated to the amount of earnings management. Another finding is that accounting affiliated audit committees are related to higher financial reporting quality, and accounting expertise contributes to better overseeing tasks
13 performed by the audit committee (Defond et al., 2005; Krishnan and Visvanathan, 2009). In addition to the expertise of audit committee members, Puat Nelson and Devi (2013) distinguish four types:
i. The financial expert, is a person with professional accounting certification, at least a postgraduate qualification, and experience in a senior managerial position.
ii. The accounting expert, is a person with professional accounting certification, and experience in a senior managerial position, the difference between the financial expert and the accounting expert here, is that the accounting expert excludes any postgraduate qualification.
iii. The non-accounting professional expert, is a person that is professionally affiliated with any professional body or organization in any field such as Engineers, Architects, Lawyers and Mariners, and has at least a postgraduate qualification, and also experience in a senior managerial position.
iv. A non-accounting expert, is an AC member with postgraduate qualification and senior managerial experience.
Puat Nelson and Devi (2013) expects that members with prior background in auditing and managerial positions such as CFO and/or CEO will perform better in restricting earnings management (Dechow et al., 1996; Beasley et al., 1999). Also, Hertz and Schults (1999) find that past involvement in accounting and auditing will increase the accuracy of the audit committee members and hence will increase the audit committee effectiveness.
However, looking at quarterly discretionary accruals, Yang and Krishnan (2005) did not find a powerful relation between financial experts and audit committee effectiveness. Given these points, audit committee members with more financial expertise have more specialized knowledge and are therefore better at performing their overseeing financial reporting duties (Ghosh et al., 2010). Thus, a positive relation is expected between audit committee financial expertise and earnings quality.
H4: Ceteris paribus, firms with more financial experts in their audit committee are positively related to earnings quality.
14 2.2.5 Audit committee ownership
To measure the audit committee ability, shares owned by audit committee members and tenure of a typical audit committee member is used (Ghosh et al., 2010). I will solely focus on the ownership of shares for this research. Audit committee members with larger share ownership are more likely to perform their duties in accordance with shareholders’ interest (Jensen, 1993). Hence, I assume a positive relation between ownership and earnings quality.
H5: Ceteris paribus, firms with stock ownership audit committee members are positively related to earnings quality.
2.3 Earnings quality
The definition of earnings quality is central in financial and accounting studies. However, according to Dichev et al. (2013), multiple definitions and measurements are widely used in academic research. A few of those measures used by researchers are accruals, earnings persistence and real activities manipulation. In order to measure earnings quality, Dechow et al. (2010) state that quality is unforeseen on the context and it is depending on the performance of the firm. Dechow et al. (2010, p. 344) define earnings quality as following:
“Higher quality earnings gives more information about the features of a firm’s financial performance which could be of interest to a decision made by e.g. a manager.” In addition to this definition, earnings quality is a function of the elemental economic performance related to the decision made and the capability of the accounting system to calculate the accompanying performance (Dechow et al., 2010; Schipper and Vincent, 2003).
I will define earnings quality by means of earnings management where earnings management is divided into one proxy for discretionary accruals and three proxies for real activities manipulation.
2.3.1 Earnings management
Earnings management is extensively used in the academic research and hence numerous definition are used. According to Healy and Wahlen (1999), earnings management is recognized when managers adopt judgment in financial reporting and structure financial activities in such a way to either delude stakeholders about the underlying economic conditions of the firm, or to control contractual outcomes which are dependent on financial figures. Another definition by Schipper (1989) puts the focus on the private gain that can be achieved through earnings management. When looking at the findings of Watts and Zimmerman (1986) regarding the positive
15 accounting theory, which describes the firm as a nexus of contracts. This gives the managers a chance to exercise opportunistic behavior where accounting is treated as device to control for contractual outcomes. On the one hand managers demand the ability to pick accounting methods that will represent the true underlying performance and tries to lower the contracting costs as much as possible. Hence, it is not possible to set up contracts for explicit direction for the use of accounting standards. Due to these aspects, managers can apply earnings management (Watts and Zimmerman, 1986). For managers there is room for opportunistic behavior by means of engaging in discretionary accruals and real activities manipulation.
2.3.1.1 Discretionary accruals
One way to measure the level of earnings management can be achieved by assessing the discretionary accruals. Accounting figures can be separated in accrual items and cash items, where accruals items reflect future income and expenses. According to Ball and Brown (1968), a combination of both items expose reveal more exactly information about the firm performance. Managers are in the position to control manipulate the accrual items above cash items due to more subjectivity of the manager on the accrual items (Dechow and Dichev, 2002). Consequently, managers are challenged to provide estimates on the accounting figures of the firm in order to reflect its true performance. Hence, this subjectivity provides the change to engage in manipulation activities (Dechow and Dichev, 2002).
2.3.1.2 Real activities manipulation
Perhaps a more comprehensive measure to detect earnings management is real activities management. To give a definition on real activities management I refer to Roychowdhury (2006), namely:
“Real activities manipulation is defined as management actions that deviate from normal business practices, undertaken with the primary objective of meeting certain earnings thresholds.” (Roychowdhury, 2006)
To support this citation, Roychowdhury (2006) state that some real activities manipulation techniques like discounting prices and decreasing discretionary expenditures are considered common for specific economic conditions. Though, significant frequent use of these techniques above the anticipated threshold, with regard to meet or beat a goal, managers are perceived to engage in real activities manipulation. By means of extensive prior research empirical confirmation was provided that managers participate in earnings management research (Healy, 1985; Defond and Jiambalvo, 1994). The evidence of these papers mainly focusses on accruals, where real activities manipulation establishes a more extensive empirical method by incorporating the
16 following measures; cash flow from operations, production costs and discretionary expenses (Roychowdhury, 2006).
Supporting Roychowdhury (2006), Graham et al. (2005) concludes that managers value the certainty to meet earning goals and are prepared to engage in real activities manipulation to complete these goals. The consequence of managers engaging in real activities manipulation is a decrease in firm value. Activities undertaken in the present period can cause an adverse reaction on the cash flows in future periods (Roychowdhury, 2006). An example for this is when managers apply significant price discounts in order to boost the sales and meet their short-term earnings goal. This can cause a future drop in purchasers because they hope to see these discounts in following periods (Roychowdhury, 2006).
When comparing accrual manipulation to real activities manipulation, Graham et al. (2005) and Bruns and Merchant (1990) find that financial executives impose more eagerness to manipulate earnings by means of real activities rather than accruals. This finding could be based on two grounds. In the first place, regulators and/or auditors can detect accrual manipulation more easily than for example significant price discounts. In the second place, in some situations when the part between the amount of earnings that are not manipulated and the threshold exceeds the part which is possible to manipulate through accruals, it is not feasible to manipulate real activities at year-end (Roychowdhury, 2006).
Based on the extensive research, Roychowdhury (2006) concentrates on these three methods for manipulation; sales manipulation, reduction of discretionary expenditures and overproduction.
First, sales manipulation can be divided in the effort managers take to temporarily boost the sales to meet short-term goals through 1) adopt credit terms that are more compliant or 2) provide price discounts. The first attempt is achieved when, for example firms offer 0% financing at the end of the year. This method results in a lower cash flow over the whole financing period. The second attempt done by managers in order to boost sales is to offer temporary price discounts so that sales shift from next year to current year. Once the firm adjust the prices back to the original prices, the increased sales amount presumably vanish. Consequently, the boost in sales seems to reflect positive margins. However, as a result of smaller margins due to the price discounts, the production costs are significant higher when this is compared to sales (Roychowdhury, 2006).
Second, reducing discretionary expenditures, like maintenance, advertising and research and development is done in order to decline expenses and hence, increase earnings. Normally these expenditures are expensed in the same period as they are aroused. These expenditures do not instantly cause profits. When the amount of reduced discretionary expenditures is high, managers
17 are likely to engage in manipulation activities.
Third, overproduction is a method used by managers to lower the fixed costs because the overhead costs are distributed among more products. This will initially cause an increase in earnings, only if the reduced overhead costs are bigger than possible increases in marginal costs (Roychowdhury, 2006). Eventually, when this method is applied cost of goods sold show a lower amount and the operating margins illustrate more positive figures. However, overproduction comes with additional storage- and production costs which are not assigned to the period of the sales. Under these circumstances, addition marginal costs regarding overproduction leads to more annual production costs with respect to the sales (Roychowdhury, 2006).
18
3. Research design
3.1 Research method
In order to give answer to the research question I will assess five hypotheses with regard to the relation between audit committee characteristics and earnings quality. To do so, I will conduct four regression analysis. I will assess earnings quality based on the level of earnings management. The examination for earnings management will be done through one proxy for discretionary accruals and three proxies for real activities manipulation.
The audit committee characteristics are based on prior research from Gosh et al. (2010) and Bedard et al. (2004). According to the Dechow et al. (1995) paper which evaluates the ability of models to detect earnings management, they conclude that the modified Jones model provides the most powerful tests of earnings management. The modified Jones model is based on the initial Jones (1991) model. Hence, regarding dictionary accruals (DA), I will use the modified Jones model from (Dechow et al. (1995). Consequently, the second measure for assessing the level of earnings management is real activities manipulation. Real activities manipulation is measured by means of three models, following Kim et al. (2012) and Roychowdhury (2006), namely 1) the abnormal cash flow from operations (AB_CFO), 2) the abnormal production costs (AB_PROD) and 3) the abnormal discretionary expenses (AB_DISEXP).
3.2 Empirically testing of discretionary accruals
As mentioned before, a measure for the earnings quality is earnings management and can be obtained by the estimation of the discretionary accruals. Dechow et al. (1995) find that the modified Jones Model, provides the most powerful tests of earnings management.
In comparison with the Jones Model (1991), the modification is designed to eliminate the conjectured tendency to measure discretionary accruals with error when discretion is exercised over revenues. Under the modified version, nondiscretionary accruals are estimated during the event period (i.e., during periods in which earnings management is hypothesized) as:
𝑁𝐷𝐴𝑡 = 𝛼1(1/𝐴𝑡−1) + 𝛼2(∆𝑅𝐸𝑉𝑡− ∆𝑅𝐸𝐶𝑡) + 𝛼3(𝑃𝑃𝐸𝑡) (1)
The total accruals contains two parts, nondiscretionary – and discretionary accruals. The total accruals are measured as:
19 Based on preceding formula it is possible to calculate the discretionary accruals by means of the following calculation:
𝐷𝐴𝑡 = 𝑇𝐴𝑡− 𝑁𝐷𝐴𝑡 (3)
Where,
𝑁𝐷𝐴𝑡 Estimated nondiscretionary accruals at year 𝒕 𝑇𝐴𝑡 Estimated total accruals at year 𝑡 =
𝐷𝐴𝑡 Estimated discretionary accruals at year 𝑡
𝐴𝑡−1 Total assets at 𝑡 -1
∆𝑅𝐸𝑉𝑡 Revenues in year 𝜏 less net receivables in year 𝜏 -1 (scaled by total assets
at t-1)
∆𝑅𝐸𝐶𝑡 Net receivables in year 𝜏 less net receivables in year 𝑡 -1 (scaled by total
assets at 𝑡 -1)
𝑃𝑃𝐸𝑡 Gross property plant and equipment in year 𝜏 (scaled by total assets at 𝑡 -1)
𝛼1, 𝛼2, 𝛼3 Firm-specific parameters
3.3 Empirically testing of real activities management
In order to empirically test real activities manipulation, I rely on prior preceding research from Kim et al. (2012), Roychowdhury (2006) and Dechow et al. (1995). After reviewing aforementioned studies, I use three measures to identify real activities manipulation, namely 1) the abnormal levels of operating cash flows (AB_CFO), 2) the abnormal production costs (AB_PROD) and 3) the abnormal discretionary expenses (AB_DISEXP). Based on Dechow et al. (1995), the following models are predicted for the mentioned three measures.
As mentioned in section 2.3.2., the probability for low operating activities net cash flow is low due to sales manipulation. To assess the first measure which is the level of operating cash flow (AB_CFO), I use the model of Roychowdhury (2006).
𝐶𝐹𝑂𝑡 𝐴𝑡−1 = 𝛼0+ 𝛼1 ( 1 𝐴𝑡−1) + 𝛽1(𝑆𝑡− 𝐴𝑡−1) + 𝛽2( ∆𝑆𝑡 𝐴𝑡−1) + 𝜀𝜏 (4) Where,
20 CFOt Cash flow from operations in year t
A Total assets
S Total net sales
∆𝐒 𝑆𝑡− 𝑆𝑡−1, change in net sales
For every firm-year, abnormal cash flow from operations is the residual estimated from the industry-year model and firm-year’s sales and lagged assets (Kim et al., 2012).
The second measure relates to the abnormal production costs (AB_PROD). Concluding from the literature review, production costs are formulated as the total of cost of goods sold and mutation in inventory during the fiscal year. Also, the expenses are designated as a linear function of contemporaneous sales (Kim et al., 2012; Roychowdhury, 2006). Therefore in order to measure the production costs, I first estimate the models for cost of goods sold and inventory. The model, according to Roychowdhury (2006), for cost of goods sold is:
𝐶𝑂𝐺𝑆𝑡 𝐴𝑡−1 = 𝛼0+ 𝛼1( 1 𝐴𝑡−1) + 𝛽 ( 𝑆𝑡 𝐴𝑡−1) + 𝜀𝑡 (5) The model, according to Roychowdhury (2006), for inventory growth is designed as:
∆𝐼𝑁𝑉𝑡 𝐴𝑡−1 = 𝛼0+ 𝛼1( 1 𝐴𝑡−1) + 𝛽 ( ∆𝑆𝑡 𝐴𝑡−1) + 𝛽2( ∆𝑆𝑡−1 𝐴𝑡−1) + 𝜀𝑡 (6) Where,
∆𝐼𝑁𝑉𝑡 Inventory change in period t
The production costs are defined, following Kim et al. (2012) through the following equation: 𝑃𝑅𝑂𝐷𝑡= 𝐶𝑂𝐺𝑆𝑡+ ∆𝐼𝑁𝑉𝑡 (7) Hence, with the use of equation (5) and (6), the normal production costs are estimated (Kim et al., 2012) as follows: 𝑃𝑅𝑂𝐷𝑡 𝐴𝑡−1 = 𝛼0+ 𝛼1( 1 𝐴𝑡−1) + 𝛽1( 𝑆𝑡 𝐴𝑡−1) + 𝛽2( ∆𝑆𝑡−1 𝐴𝑡−1) + 𝛽3( ∆𝑆𝑡−1 𝐴𝑡−1) + 𝜀𝑡 (8) Where,
21 The production costs are defined through the sum of cost of goods sold and the mutation in inventories.
The third measure is to capture the abnormal discretionary expenses (AB_DISEXP). Based on Kim et al. (2012), the following equation is used:
𝐷𝐼𝑆𝐸𝑋𝑃𝑡 𝐴𝑡−1 = 𝛼0 + 𝛼1( 1 𝐴𝑡−1) + 𝛽 ( 𝑆𝑡−1 𝐴𝑡−1) + 𝜀𝑡 (9) Where,
𝐷𝐼𝑆𝐸𝑋𝑃𝑡 Discretionary expenses in period t
The normal level of discretionary expenses is defined as the sum of selling, general and administrative expenses (XSGA), research and development expenses (XRD) and advertisement expenses (XAD).
To summarize, when firms engage in real activities manipulation 1) abnormal operating cash flows (AB_CFO) will decrease, 2) abnormal production costs (AB_PROD) will increase and 3) abnormal discretionary expenses (AB_DISEXP) will decrease (Roychowdhury, 2006).
3.4 Regression models
I will examine the relation between earnings management and audit committee characteristics in this section. To do so, I will formulate regression models based on prior literature and analyze the results. In total I will use four regression frameworks, the first one, in section 3.4.1, focuses on discretionary accruals and the last three, in section 3.4.2, aims on the relation between real activities manipulation and audit committee characteristics.
3.4.1 Relation between discretionary accruals and audit committee characteristics
I examine the relationship between earnings management, following the model from Ghosh et al. 2010) by means of discretionary accruals, and audit committee characteristics using the following regression model:
𝐷𝐴𝑡 = 𝛽0+ 𝐵1𝐼𝑛𝑑𝑒𝑝𝑒𝑛𝑑𝑒𝑛𝑐𝑒𝐴𝐶+ 𝛽2𝑆𝑖𝑧𝑒𝐴𝐶+ 𝛽3𝑀𝑒𝑒𝑡𝑖𝑛𝑔𝑠𝐴𝐶 + 𝛽4𝐸𝑥𝑝𝑒𝑟𝑡𝐴𝐶
+ 𝛽5𝑂𝑤𝑛𝑒𝑟𝑠ℎ𝑖𝑝𝐴𝐶 + 𝛽6𝑆𝑖𝑧𝑒 + 𝛽7𝑅𝑂𝐴 + 𝛽8𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒 + 𝛽9𝐿𝑜𝑠𝑠 + 𝛽10𝐵𝑖𝑔4 + 𝜀 (10)
22 Dependent variable Definition
𝐷𝐴𝜏 Discretionary accruals (DA = TA – NDA)
Independent variables Definition
𝐼𝑛𝑑𝑒𝑝𝑒𝑛𝑑𝑒𝑛𝑐𝑒𝐴𝐶 The amount of independent members in the audit committee 𝑆𝑖𝑧𝑒𝐴𝐶 The number of member in the audit committee
𝑀𝑒𝑒𝑡𝑖𝑛𝑔𝑠𝐴𝐶 The number of meetings held by an audit committee in a given fiscal-year
𝐸𝑥𝑝𝑒𝑟𝑡𝐴𝐶 The amount of financial experts in the audit committee
𝑂𝑤𝑛𝑒𝑟𝑠ℎ𝑖𝑝𝐴𝐶 The amount of shares owned by the member of the audit committee
Control variables Definition
𝑆𝑖𝑧𝑒 Logarithmic transformation of the fiscal year-end value of total assets
𝑅𝑒𝑡𝑢𝑟𝑛 𝑜𝑛 𝐴𝑠𝑠𝑒𝑡 (𝑅𝑂𝐴) Ratio: Net income / total assets 𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒 Ratio: Total liabilities / total assets
𝐿𝑜𝑠𝑠 Dummy variable: 1 = Net income is a loss, 0 = Net income is a profit
𝐵𝑖𝑔4 Dummy variable: 1 = Audited by one of the Big4 (KPMG, Deloitte, EY or PWC), 0 = Not audited by the Big4.
Considering that independent outside members are better in protecting shareholders and hence, are better in discouraging earnings, 𝐵1is likely to be negative. Assuming that when the size of the
audit committee is related to (less) enhanced performance, 𝛽2 is expected to be negative (positive).
Since the proactive and reactive nature of audit committee meetings, 𝛽3 is likely to be positive
(negative) reliant on the nature of the frequency of meetings. Audit committee members with a financial background may be more familiar to manage earnings and may better understand implications of earnings manipulation, hence 𝛽4 is expected to be negative. Lastly, 𝛽5 is assumed
to be negative since financial importance increases incentives for audit committee members to monitor management.
3.4.2 Relation between real activities manipulation and audit committee characteristics In order to capture the relation between earnings management, by means of real activities manipulation, and audit committee characteristics, I use the model from Kim et al. (2012) to
23 measure the three manipulation methods.
For abnormal cash flow from operations:
𝐴𝐵_𝐶𝐹𝑂𝑡 = 𝛼0+ 𝛼1𝐼𝑛𝑑𝑒𝑝𝑒𝑛𝑑𝑒𝑛𝑐𝑒𝐴𝐶+ 𝛼2𝑆𝑖𝑧𝑒𝐴𝐶+ 𝛼3𝑀𝑒𝑒𝑡𝑖𝑛𝑔𝑠𝐴𝐶+ 𝛼4𝐸𝑥𝑝𝑒𝑟𝑡𝐴𝐶
+ 𝛼5𝑂𝑤𝑛𝑒𝑟𝑠ℎ𝑖𝑝𝐴𝐶+ 𝛼6𝐷𝐴𝑡+ 𝛼8𝑆𝑖𝑧𝑒 + 𝛼9𝑅𝑂𝐴 + 𝛼10𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒 + 𝛼11𝐿𝑜𝑠𝑠 + 𝛼12 𝐵𝑖𝑔4 + 𝜀𝑡 (11) For abnormal production costs:
𝐴𝐵_𝑃𝑅𝑂𝐷𝑡 = 𝛼0+ 𝛼1𝐼𝑛𝑑𝑒𝑝𝑒𝑛𝑑𝑒𝑛𝑐𝑒𝐴𝐶+ 𝛼2𝑆𝑖𝑧𝑒𝐴𝐶+ 𝛼3𝑀𝑒𝑒𝑡𝑖𝑛𝑔𝑠𝐴𝐶+ 𝛼4𝐸𝑥𝑝𝑒𝑟𝑡𝐴𝐶 + 𝛼5𝑂𝑤𝑛𝑒𝑟𝑠ℎ𝑖𝑝𝐴𝐶+ 𝛼6𝐷𝐴𝑡+ 𝛼8𝑆𝑖𝑧𝑒 + 𝛼9𝑅𝑂𝐴 + 𝛼10𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒 + 𝛼11𝐿𝑜𝑠𝑠
+ 𝛼12 𝐵𝑖𝑔4 + 𝜀𝑡 (12) For abnormal discretionary expenses:
𝐴𝐵_𝐷𝐼𝑆𝐸𝑋𝑃𝑡 = 𝛼0+ 𝛼1𝐼𝑛𝑑𝑒𝑝𝑒𝑛𝑑𝑒𝑛𝑐𝑒𝐴𝐶+ 𝛼2𝑆𝑖𝑧𝑒𝐴𝐶+ 𝛼3𝑀𝑒𝑒𝑡𝑖𝑛𝑔𝑠𝐴𝐶 + 𝛼4𝐸𝑥𝑝𝑒𝑟𝑡𝐴𝐶 + 𝛼5𝑂𝑤𝑛𝑒𝑟𝑠ℎ𝑖𝑝𝐴𝐶+ 𝛼6𝐷𝐴𝑡+ 𝛼8𝑆𝑖𝑧𝑒 + 𝛼9𝑅𝑂𝐴 + 𝛼10𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒 + 𝛼11𝐿𝑜𝑠𝑠 + 𝛼12 𝐵𝑖𝑔4 + 𝜀𝑡 (13)
Where,
Dependent variables Definition
𝐴𝐵_𝐶𝐹𝑂𝑡 The level of abnormal cash flows from operations, measured as deviations from the predicted values from the corresponding industry-year regression
𝐴𝐵_𝑃𝑅𝑂𝐷𝑡 The level of abnormal production costs, measured as deviations from the predicted values from the corresponding industry-year regression
𝐴𝐵_𝐷𝐼𝑆𝐸𝑋𝑃 The level of abnormal discretionary expenses, measured as deviations from the predicted values from the corresponding industry-year regression
* The remaining variables are already mentioned at regression (10)
According to Kim et al. (2012), in order to control earnings firms are likely to engage in a mixture of real activities manipulation and discretionary accruals. In which the firm will engage depends on which is less costly for the firm (Zang, 2011; Cohen et al. 2008). Supported by Zang (2011), the compromise between real activities manipulation and discretionary accruals is dependent on the corresponding costs. To control for aforementioned compromise, I include DA, a proxy for
24 discretionary accrual earnings management, as a control variable in equation (11), (12) and (13) (Cohen et al., 2008).
3.5 Control variables
In order to rule out any alternative causes within this research, I include the following control variables, namely (a) size, (b) ROA, (c) leverage, (d) loss and (e) big4.
First, (a) size (logarithmic transformation of the fiscal year-end value of total assets) is included. Large firms have the propensity to be associated with more prestigious accounting firms than small firms (Shu and Chiang, 2014). Since prestigious accounting firms are the, large, Big4 firms. Becker et al. (1998) find that large auditors are more successful in detecting (aggressive) earnings management. Furthermore, Shu and Chiang (2014) state that large firms are more likely than small firms to have a steady flow of earnings. This implies that the reported earnings are good forecasters and thus, inherently less earnings management. Coupled with, large firms generally have less asymmetric information and disclose more timely accounting information (LaFond and Watts, 2008). Based on the previous articles, I would expect a negative relation between earnings management and firm size.
According to Warfield et al. (1995) large firms are more often to utilize the width in accounting to lower political costs. Based on this research I would assume a positive relation between earnings management and firm size.
Second, the (b) ROA ratio (net income divided by total assets) is enclosed to control for possible shifts in firm performance. This is relevant because Baxter and Cotter (2009) state that an increase in earnings quality is related with a shift in firm performance. Thus, I would expect a negative relation between earnings management and the ROA ratio.
Third, I included (c) leverage (total liabilities divided by total assets) due to the fact that firms with a high level of leverage tend to experience more debt- and shareholder conflicts according to Watts (2003). This is supported by DeFond and Jiambalvo (1994) that firms with debt influences the incentives for management to manipulate financial statements. Therefore, I would assume a positive relation between earnings management and leverage.
Fourth, I included a dummy variable for (d) loss (where 1 means there is a negative net income and 0 when there is a positive net income). DeFond and Subramanyam (1998) state that firms in financial distress have bigger incentives to exercise their discretion to manage their earnings. When a firm is in a loss position, they more likely to the big bath (Kirschenheiter and Melumad, 2002). So I would assume a positive relation between earnings management and loss.
At last, a dummy variable for (e) Big4 (where 1 means firms that are audited by one of the Big4 audit firms: KPMG, Deloitte, EY or PWC and 0 means firms that are not audited by one of
25 the Big4. As mentioned under control variable (a), large auditors are more successful in restraining aggressive earnings management. (Becker et al. 1998). Further, earnings tend to be more conservative for firms that are audited by the Big4 and the level of total – and abnormal accruals are lower for these firms (Michas, 2011). Given these arguments, I would expect a negative relation between earnings management and Big4.
3.6 Sample and data collection
The initial sample consist of US listed firms classified under the S&P 400, S&P 500 and S&P 600 with information available on corporate governance, which particularly concerns audit committee characteristics, from ISS (formerly RiskMetrics) during the period 2008 - 2013. Where the S&P 400 records 400 mid-cap stocks, the S&P 500 represents more than 80% of the stock market (large-cap) and the S&P 600 records 600 small-cap stocks. This research focusses on the implementation of the SOX regulations regarding audit committees, which applies solely in the United States, therefore the sample composes of US listed firms. Especially, I obtain data on audit committee characteristics (independence, size, activity, expertise and ownership) from ISS. The data input for the modified Jones Model and real activities manipulation is derived from COMPUSTAT, similarly for the control variables (size, ROA, leverage, loss and Big4). The motivation to conduct research for the years 2008 till 2013 because this covers the latest data available from Wharton Research Data Services (WRDS).
Table 1
Observation collection: period 2008 - 2013
1 Observations on Audit Committee characteristics 83.341
2 (Non Audit Committee members) (49.234)
3 (Board affiliation non-independent audit committee members) (300)
4 (Observations with missing financial values) (2.952)
5 (Observations with missing information from COMPUSTAT) (9.951)
6 (Observations with missing information on RAM_DISEXP) (14.965)
Final sample 5.939
The sample distribution per industry is presented in Table 2.1. The Manufacturing industry is the most heavily represented with 55%. Furthermore, S&P 600 is the largest observation classification group in the sample with 38% followed by the S&P 500 with 38% (S&P 600 has 5 observations more).
26 Table 2.1
Observations per industry
# Industry SICH Observations % of total
1 Manufacturing 2000 - 3999 3.298 55%
2 Transportation, Communication, Electric, Gas and Sanitary 4000 - 4999 44 1%
3 Wholesale and Retail Estate 5000 - 5999 1.590 27%
4 Financial Services 6000 - 6999 6 0%
5 Services 7000 - 8999 1.001 17%
Total observations 5.939 100%
Table 2.2
Observations per S&P classification
# S&P Classification Observations % of total
1 S&P 400 1.458 25%
2 S&P 500 2.238 38%
3 S&P 600 2.243 38%
27
4. Results
In this section I will discuss the descriptive statistics, regressions performed and the correlation matrix. In the end I will elaborate the hypotheses based on the empirical results.
4.1 Descriptive statistics
The descriptive statistics for the final sample can be found in Table 3. The mean value of DA is 0.067 which is in line with Ghosh et al. (2010) and Kim et al. (2012), namely, 0.052 and 0.005. The mean values of AB_CFO, AB_PROD, AB_DISEXP are 0.130, -0.093 and -0.067, also in line with the results from Kim et al. (2012). In fact, the results from Kim et al. (2012) are, respectively, 0.129, -0.096 and -0.059 showing almost equivalent figures. The mean values of AB_CFO and AB_PROD assume that this sample firms interfere in manipulation methods like overproduction and sales manipulation (Kim et al., 2012).
Next, I also report in the descriptive statistics for audit committee characteristics measures. The average audit committee consists of 100% independent members and is formed with, on average, 4 members. The measure IndependenceAC is in line with the SOX regulations, namely stating that all audit committee members must be independent (SOX, 2002). SizeAC shows a mean value of 3.9 which is similar to the results (3.7) of Ghosh et al. (2010). All firms had one meeting (MeetingsAC) per year and consists of two financial experts (ExpertAC) with a maximum of eight experts. All these audit committee members, on average, own 94355 shares (OwnershipAC).
Finally, the control variables included in the regression models are also displayed in Table 3. Firm Size, which is calculated by natural log of the total assets, shows a mean value of 3.369. 91.6% of my sample firms is audited by one of the Big4 accounting firms and, on average, firms have a 6.4% return on asset (ROA). Further, the mean value for Leverage is 48.9% indicating that, on average, the total liabilities are almost the half of the total assets for the firms in my sample. The mean value of Loss is 0.119, indicating that 707 firms reported a loss.
28 4.2 Correlation matrix
The Pearson correlation coefficients are shown in Table 4. The Pearson test is tool to measure the relation between the variables used in my regression models. Where a correlation coefficient of 1 displays perfect correlation, a coefficient of 0 means no correlation. At first sight, the correlation coefficients show a weak correlation. However, the correlation between ROA and Loss (-0.637) and the correlation between Size and Leverage (0.412) show moderate correlation. The correlation between ROA and Loss can be explained by means of firms with a negative net income are not performing well, hence less return on their invested assets. The correlation between Size (logarithmic of total assets) and Leverage makes sense based on the ratio of Leverage (Total liabilities divided by total assets). Furthermore, the correlation between AB_CFO (measure of sales manipulation) and AB_PROD (measure of overproduction) also presents a moderate correlation (-0.508). Indicating that when managers engage more in sales manipulation, it is most likely that the managers will engage less in overproduction. Additionally, IndependenceAC and MeetingsAC are omitted due to perfect collinearity.
Table 3 Descriptive statistics
N Mean Min Max Std. Deviation
Earnings Management
1 DA 5939 0.067 -2.231 1.163 0.165
2 AB_CFO 5939 0.130 -1.600 1.032 0.196
3 AB_PROD 5939 -0.093 -1.494 0.874 0.187
4 AB_DISEXP 5939 -0.067 -1.536 1.610 0.224
Audit Committee characteristics
5 IndependenceAC 5939 1 1 1 0 6 SizeAC 5939 3.879 1 8 0.965 7 MeetingsAC 5939 1 1 1 0 8 ExpertAC 5939 2.082 0 8 1.183 9 OwnershipAC 5939 94355 0 1.81E+07 525435 Control Variables 10 Size 5939 3.369 1.736 5.435 0.709 11 ROA 5939 0.064 -1.138 0.405 0.095 12 Leverage 5939 0.489 0.063 2.310 0.222 13 Loss 5939 0.119 0 1 0.323 14 Big4 5939 0.916 0 1 0.278
Table 4
Pearson correlation coefficients matrix Observations = 5939 1. DA AB_ 2. CFO 3.
AB_PROD AB_DISEXP 4. IndepenceAC 5. SizeAC 6. Meetings7. AC 8. Expert AC 9. Ownership AC 10.
Size ROA 11. Leverage 12. Loss 13. Big14. 4 1. DA 1 2. AB_CFO 0.289 1 3. AB_PROD -0.100 -0.508 1 4. AB_DISEXP -0.096 -0.276 -0.289 1 5. IndepenceAC . . . . . 6. SizeAC 0.040 -0.030 0.002 -0.007 . 1 7. MeetingsAC . . . . 8. ExpertAC -0.018 -0.057 0.042 -0.009 . 0.270 . 1 9. OwnershipAC -0.001 0.007 -0.002 0.006 . 0.028 . 0.060 1 10. Size 0.085 0.111 0.020 -0.169 . 0.349 . 0.280 0.035 1 11. ROA 0.186 0.218 -0.178 -0.015 . 0.077 . -0.001 -0.003 0.099 1 12. Leverage -0.014 -0.174 0.096 -0.008 . 0.249 . 0.220 -0.026 0.412 -0.109 1 13. Loss -0.146 -0.129 0.083 -0.029 . -0.115 . -0.029 0.006 -0.160 -0.637 0.048 1 14. Big4 -0.011 -0.007 -0.037 0.044 . 0.140 . 0.165 0.011 0.286 -0.026 0.188 -0.030 1 Note: The variables IndependenceAC and MeetingsAc are omitted due to perfect collinearity.
4.3 Relation between earnings management and audit committee characteristics
For testing the relation between earnings management and audit committee characteristics, four regression analyses are conducted using STATA. A regression analysis is a measure to assess the relationship between certain independent variables and one dependent variable. This analysis provides the Beta-coefficient, direction of the contribution and the significance level of the independent variables to the dependent variable while the other independent are maintained at a stable factor. Furthermore, this analysis provides insight in how much variance in the dependent variable is explained by the model of independent variables, expressed in the form of the adjusted R square.
All the aforementioned hypotheses in Section 2.2 presume a specific direction about the relation between audit committee characteristics and earnings management. Earnings management is measured by means of discretionary accruals and real activities manipulation. The results of the regression for discretionary accruals and the three regressions for real activities manipulation; 1) abnormal cash flow from operations, 2) abnormal production costs and 3) abnormal discretionary expenses are presented in Table 5. First, I will analyze the R-squared for the four measures and then I will discuss the coefficients, the direction of the coefficient and the significance level of the audit committee characteristics and control variables.
4.3.1 Analyzing the R-squared in the relation between discretionary accruals and audit committee characteristics
Looking at the R-squared of the regression model for discretionary accruals. The explanatory power of this model is not high, namely 4.23% of the variance in DA is explained by the model. Compared to Ghosh et al. (2010) (R-squared of 7,12%), the R-squared in my sample is relatively low. The contrast can be explained through the more comprehensive set of audit committee characteristics and control variables incorporated in their regression model. Also, Ghosh et al. (2010) use a sample consisting of firms from pre- and post-SOX year (1998-2005), where my research focuses on 2008 till 2013.
4.3.2 Analyzing the R-squared in the relation between real activities manipulation and audit committee characteristics
Measuring real activities manipulation is done in threefold, and hence, three R-squared are presented in Table 5. With respect to AB_CFO (abnormal cash flow from operations), 16.44% of the variance in abnormal cash flow from operations is explained by the regression model. Regarding AB_PROD (abnormal production costs), 4.86% of the variance is explained and concerning AB_DISEXP (abnormal discretionary expenses), 5.25% of the variance is explained. Especially the model AB_CFO is relatively high compared to the other measures. Kim et al. (2012)
31 presents an R-squared of 15.1% for AB_CFO, 8.2% for AB_PROD and 3.1% for AB_DISEXP. The R-squared in my model, for AB_CFO and AB_PROD, are in line with the results in Kim et al. (2012). However, for AB_PROD, I note a difference in explained variance in the model. This is because Kim et al. (2012) included more control variables to control for their specific topic on a firm’s CSR performance and the behavior for influencing financial reporting. This includes control variables such as MB (market-to-book equity ratio) and FIRM_AGE.
4.3.3 Effect of audit committee characteristics on discretionary accruals
The results of the regression from discretionary accruals is shown in Table 5, where the coefficients and significance levels are presented. Referring to Section 2.2 where I defined all hypotheses, I note that all hypotheses assume a positive relation between the audit committee characteristics and earnings quality. Meaning that when the audit committee characteristics increases, the quality of the earnings will also increase. In order for earnings quality to increase, earnings management should reflect a low level of discretionary accruals (DA). A higher value for DA signifies a higher level of earnings management and hence, I expect that the audit committee characteristics coefficients are negative in order to accept the hypotheses.
Now I can assess my results with respect to the results of Ghosh et al. (2010). Ghosh et al. (2010) uses the absolute value of discretionary accruals, based on the Jones Model, as a measure of earnings management. They performed multiple regression which all capture different periods regarding pre- and post-SOX regulation. Their coefficients for SizeAC (-0.000) and MeetingsAC (0.001) are significant. However, I do not find significance for those characteristics. With respect to SizeAC, my coefficient (0.002) is not statistically significant and also in the wrong direction. According to Ghosh et al. (2010) this is not a surprising finding namely, there are conflicting views on the relation between audit committee size and earnings management. Regarding MeetingsAC, in my sample all firms had one audit committee meeting per fiscal year and therefore the variable is fully omitted because of perfect collinearity. Also IndependenceAC is omitted because of full collinearity. Meaning that MeetingsAC and IndependenceAC are in perfect linear relation with the four regression models. This makes sense because SOX mandates that all members of the audit committee should be independent (SEC, 2003). Hence, non-independent members in the audit committee are nog permitted for firms acting under SOX. Therefore, the sample only exist of independent audit committee members from ISS (formerly RiskMetrics). At last, I note an insignificant coefficient for OwnershipAC in the assumed direction. This is in line with Ghosh et al. (2010), namely they show a coefficient of -0.001.
Looking at my other results, ExpertAC is significant (p < 0.01) with a coefficient of -0.005. This is not in line with the results from Ghosh et al. (2010). This can be explained by the
non-32 consistent sample selection from CorporateLibrary for the post-SOX period by Ghosh et al. However, my results are in line with research from Bedard et al. (2004) and Carcello et al. (2006), namely a coefficient of -2.817 (p < 0.01) and -1.361 (p < 0.05). They find that financial experts within the audit committee are related to a lower level of aggressive earnings management. Given these points, I can state that the size of the audit committee has an effect on the level of discretionary accruals.
4.3.4 Effect of audit committee characteristics on real activities manipulation
The outcomes of the regression from real activities manipulation are also displayed in Table 5. For each measure the coefficients and significance levels are displayed. I used three proxies to measure real activities manipulation, namely 1) abnormal operating cash flows (AB_CFO), 2) abnormal production costs (AB_PROD) and 3) abnormal discretionary expenses (AB_DISEXP). When firms engage in real activities manipulation 1) AB_CFO is negative, 2) AB_PROD is positive, and 3) AB_DISEXP is negative (Roychowdhury, 2006). Hence, I expect that the audit committee characteristics coefficients are positive for AB_CFO and AB_DISEXP and negative for AB_PROD.
AB_CFO shows the following coefficients for SizeAC (-0.010), ExpertAC (-0.008) and OwnershipAC, all significant (p < 0.01) except for OwnershipAC. All coefficients do not point in the assumed direction as mentioned above. However, with respect to SizeAC, Sun and Liu (2014) also report a negative coefficient (not significant) of -0.005. Also, Kim et al. (2012) finds a negative significant coefficient for their dependent variable. Sun and Liu (2014) find a, not significant, coefficient of 0.002 for ExpertAC in their results. This can be explained by the variation in definition of an audit committee member. Sun and Liu (2014) define a financial expert when he/she is or was a 1) certified public accountant, 2) auditor, 3) principal, 4) CFO, 5) controller of 6) chief accounting officer (Sun and Liu, 2014), where my sample is based on directly accessible information on financial expertise from COMPUSTAT. Also, their information on audit committee characteristics is hand collected. Next to these significant variables, five of the control variables are significant and control the amount of AB_CFO, which in total explains 16.44% of the variance in AB_CFO. The control variables ROA and DA represents the most explanatory power for this regression model. Indicating a significant (p < 0.01) coefficient of, respectively, 0.335 and 0.293.
The coefficients which are presented for the audit committee characteristics measured by means of AB_PROD, show a significant value of 0.005 for ExpertAC (p < 0.01), an insignificant value of -0.002 for SizeAC and an insignificant value for OwnershipAC. The expected direction holds for SizeAC except that it is insignificant. Compared to Sun and Liu (2014), their results show
33 a negative coefficient for ExpertAC of -0.014, 0.023 for SizeAC and 0.023 for OwnershipAC (all are insignificant). Looking at the control variables, I see a significant coefficient for ROA of -0.378 (p < 0.01). Meaning ROA is main responsible for the level of AB_PROD in this regression model.
The final regression for real activities manipulation that I performed in this research is AB_DISEXP. The SizeAC significant (p < 0.01) coefficient is 0.010, ExpertAC coefficient is 0.002 and lastly the OwnershipAC also shows a positive coefficient. The beforehand assumed direction is positive and all coefficient meet this assumption. However, Sun and Liu (2014) finds insignificant coefficients for SizeAC (-0.021), ExpertAC (-0.010) and OwnershipAC (0.228). Only OwnershipAC shares the same direction. Furthermore, especially DA influences the level of AB_DISEXP with a significant coefficient of -0.110 (p < 0.01).
Table 5 Regression results
Discretionary accruals and real activities manipulation
DA AB_CFO AB_PROD AB_DISEXP
Variables Coefficient Coefficient Coefficient Coefficient
Audit Committee characteristics
IndependenceAC**** - - -
-SizeAC 0.002 ***-0.010 -0.002 ***0.010
MeetingsAC**** - - -
-ExpertAC ***-0.005 ***-0.008 ***0.005 0.002
OwnershipAC -3.54e-10 -4.90e-10 -6.17e-10 5.61e-09
Control Variables Size ***0.020 ***0.055 0.005 ***-0.077 ROA ***0.262 ***0.335 ***-0.378 **0.0824 Leverage -0.017 ***-0.190 ***0.064 ***0.081 Loss **-0.018 ***0.028 ***-0.030 0.003 Big4 *-0.014 -0.001 ***-0.046 ***0.073 DA ***0.293 ***-0.082 ***-0.110 R-squared 0.0410 0.1644 0.0486 0.0525 n 5939 5939 5939 5939
*, ** and *** represent the significance levels of 0.1, 0.05 and 0.01. ****: The variables IndependenceAC and MeetingsAC are omitted due to perfect collinearity.
