Master Thesis
The impact of director busyness on the relationship
between specific corporate governance
characteristics and earnings management
by
Bryan Jurcka 10684913
22 June 2015
Amsterdam Business School
Faculty of Economics and Business, University of Amsterdam MSc Accountancy & Control, variant Accountancy
Supervisor: Prof. dr. V. O'Connell, B.Comm, MBS, PhD (LSE), ACMA Word count: 13.428
2
Statement of Originality
This document is written by student Bryan Jurcka 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
In the early years of the 21st
century, a number of large corporate and accounting scandals took place. These scandals led to discussions about the corporate governance of large firms. One of those discussions was about the negative effects of non-executive directors who serve on multiple boards. In many studies, these directors are called busy directors and serve on three or more outside boards. In this thesis, I will discuss the impact of these busy directors on the relationship between some specific corporate governance characteristics and earnings management. These characteristics are board size, board independence, audit committee independence and the presence of a financial expert on the audit committee. Prior literature already investigated the relationship between the aforementioned characteristics and earnings management. Work on the potential impact of director busyness on these relationships has not been published by known and recognized journals, to the best of my knowledge. In this thesis, I found evidence that director busyness has a significantly positive impact on the relationship between board size and earnings management. Further, I could not find any evidence that shows that director busyness has an impact on the relationship between board independence and earnings management, between audit committee independence and earnings management and between the presence of a financial expert on the audit committee and earnings management. I also ran several sensitivity analyses. The results from these analyses point out that adding return on assets as a control variable increases the power of the whole model and also, in most cases, increases the magnitude of the relations between board busyness, audit committee busyness and financial expert busyness and earnings management. However, replacing director busyness, as a percentage of total directors, by the average number of outside boards a director serves on, does not give any different results for the most important variables.
4
Table of Contents
1. Introduction ... 5 2. Theory ... 8 2.1 Agency theory ... 8 2.2 Earnings management ... 8 2.3 Board of directors ... 9 2.4 Director busyness ... 112.5 Literature review and hypotheses ... 12
3. Research methodology ... 15 3.1 Sample selection ... 15 3.2 Earnings management ... 15 3.3 Director busyness ... 16 3.4 Control variables ... 17 3.5 Empirical model ... 17 4. Results ... 19 4.1 Descriptive statistics ... 19 4.2 Multivariate analysis ... 25 4.3 Sensitivity analysis ... 29 5. Conclusion ... 32 References ... 34
5
1. Introduction
After large corporate and accounting scandals at the beginning of the 21st
century, there was a lot of discussion about corporate governance. Ultimately, in 2002, the Sarbanes-Oxley Act (hereafter: SOX) was introduced in the U.S. to “protect investors by improving the accuracy and reliability of corporate disclosures made pursuant to the securities laws, and for other purposes” (Sarbanes-Oxley Act of 2002, p. 745). It was introduced to preclude new scandals by preventing fraudulent accounting and management misbehavior (Zhang, 2007), like scandals from Worldcom in 2001, Enron in 2001 and Tyco in 2002. Before the introduction of SOX, the federal regime only consisted of disclosure requirements. However, after the introduction, substantive corporate governance mandates were added (Romano, 2004).
Baysinger and Butler (1985, p. 101) refer to Fama and Jensen (1983) and Williamson (1983 and 1984) and state that “recent advances in economic theory suggest that the board of directors is an important part of the governance structure of large business corporations”. Farber (2005) found that fraud firms have poor governance in the year prior to fraud detection, including: fewer outside board members, fewer audit committee meetings and fewer financial experts on the audit committee. Nowadays, public debate has focused on the board of directors in large publicly traded corporations and the effectiveness of their monitoring (Fich and Shivdasani, 2006). A good example is Bank of America, listed at the New York Stock Exchange and the Tokyo Stock Exchange. The bank appointed CEO Brian Moynihan as chairman of the board at October 1, 2014. An activist investor group of the bank threated with a shareholder resolution because of this appointment, which resulted in a dual role of Moynihan. The goal of the shareholder resolution was to suggest appointing an independent board chairman. Eventually, to avoid a vote on the resolution, Bank of America agreed to produce a report on its corporate culture and business practices.
Next to the prior discussion about the dual role of the CEO, another element of the debate is about whether non-executive directors1
who serve on multiple boards have to reduce the number of boards they serve on (Fich and Shivdasani, 2006). They state that investors and policy advocates utter that sitting on multiple boards can result in “overstretched directors that may not be effective monitors on any board” (p. 721). The Cadbury Report (1992) already emphasized the role of non-executive directors in corporate governance. It stated that non-executive directors should bring an independent judgment regarding issues of strategy, performance, resources, including key appointments, and standards of conduct. It further says that the number of non-executive directors serving on a board have to be such that their opinion has a significant weight in the board’s decisions.
1
Prior literature uses the term non-executive director, as well as independent director and outside director. They all mean the same, as explained in paragraph 2.4. Therefore, in this thesis, I will use these terms
6 The Council of Institutional Investors (1998) states that non-executive directors should not serve on three or more boards, which has been followed by the paper of Ferris, Jagannathan and Pritchard (2003). Core, Holthausen and Larcker (1999) state that a busy non-executive director can be described as a non-executive director who serves on three or more other boards. For retired non-executive directors, they mention a maximum of six or more boards. In the research of Fich and Shivdasani (2006), they also use a cutoff of three or more outside boards a non-executive director has to serve on, to qualify him or her as busy. With this choice, they acknowledge consistency of this particular cutoff with prior work by Core et al. (1999) and Ferris et al. (2003).
A good example of a busy non-executive director is that of Peter Hartman. Next to vice-chairman of the board of directors of Air France-KLM, he is also chairman of the board of governors of Connekt, vice-chairmen of the Advisory Council for Aeronautics Research in Europe. Furthermore, he is a member of the board of directors of Koninklijke Ten Cate NV, Constellium N.V., Fokker Technologies Group B.V. and Texel International Airport.2
One can argue that a non-executive director who serves on multiple boards does not have enough time and maybe even enough energy to effectively monitor and advise the executive directors. The conclusion can then be drawn that there is a positive relationship between a busy non-executive director and earnings management. Ferris et al. (2003) created the busyness hypothesis, which states that serving on multiple boards, overcommits an individual. As a result, these individuals shirk their responsibilities as directors. Ferris et al. (2003) say for example that overcommitted directors might serve less frequently on important board committees such as the audit or the compensation committees. They also say that reduced monitoring by busy directors might lead to agency costs, such as increased litigation exposure for the firm.
From another angle, one can argue that non-executive directors, who serve on multiple boards, bring several advantages. Because they serve on multiple boards, they can have a lot of experience from those other firms. Next to that, they can have a broader network, which can work in favor of a particular company. They can provide advantages in the form of access to resources, suppliers and customers. In this way, one can argue that there is a negative relationship between a board with busy non-executive directors and earnings management, which means that the more outside boards a director serves on, the lower earnings management will be.
In this thesis, I am going to examine the impact of director busyness on the relationship between specific corporate governance characteristics and earnings management. More specifically, I am going to investigate whether busyness of non-executive directors has a moderating effect on the relationship between board size and earnings management and on
7 the relationship between board independence and earnings management. Furthermore, I am going to investigate whether busyness of non-executive directors serving on an audit committee has a moderating effect on the relationship between audit committee independence and earnings management. Finally, I am going to investigate whether, when an audit committee has a financial expert, the busyness of this financial expert has a moderating effect on earnings management.
In my opinion, adding busyness as a moderating variable will give new insights into this subject. Prior literature investigated the relationship between board busyness and firm performance (Andres, Van den Bongard and Lehmann, 2013; Cashman, Gillan and Jun, 2012; Fich and Shivdasani, 2006; Kiel and Nicholson, 2006) between board busyness and stock prices (Falato, Kadyrzhanova and Lel, 2014), between board busyness and market-to-book ratio (Fich and Shivdasani, 2006) or between board busyness and key strategic decisions (Harris and Shimizu, 2004). However, research regarding the relationship between busyness of non-executive directors and earnings management are very scarce. Sarkar, Sarkar and Sen (2008) did investigate the relationship between board busyness and earnings management. However, they used board busyness as an independent variable instead of a moderating variable.
In this thesis, I found that director busyness does not have an impact on the relationship between board independence and earnings management, between audit committee independence and earnings management and between the presence of a financial expert on the audit committee and earnings management. However, I do find that board busyness does have an impact on the relationship between board size and earnings management. At first, the relationship between board size and earnings management is weak and also not significant. After adding board busyness as an independent variable, this variable itself is not significantly associated with earnings management but the interaction with board size leads to a positive relationship with earnings management, significant at the 1% level. One can conclude from these findings that when a board becomes larger and at the same time consists of busy directors, earnings management will also increase. My sensitivity analysis shows that adding return on assets as a control variable results in stronger relations between earnings management and board busyness, audit committee busyness and financial expert busyness.
This thesis proceeds as follows. Section 2 will discuss theory consisting of the agency problem, earnings management, the board of directors and director busyness, and prior literature and the resulting hypotheses. Section 3 addresses my sample selection, how to measure earnings management and busyness of non-executive directors, the control variables I will use and the models I developed to answer the hypotheses. The descriptive statistics, multivariate analysis and sensitivity analysis will be presented in section 4 and I will conclude with the conclusion, discussion and limitations of this thesis, which will be discussed in section 5.
8
2. Theory
This section will begin in paragraph 2.1 with discussing theory regarding the agency problem because a board of directors has an important monitoring role, described in agency theory. Paragraph 2.2 will explain what earnings management is, the reasons behind earnings management and signals to detect earnings management. Then, paragraph 2.3 and 2.4 continue with theory regarding the board of directors and director busyness. This section will conclude with relevant prior literature, together with the associated hypothesis of this thesis in paragraph 2.5. The aim here is to derive hypotheses based on prior literature.
2.1 Agency theory
Jensen and Meckling (1976) describe an agency relationship as a contract under which one or more persons (the principal(s)) engage another person (the agent) to perform some service on their behalf, which involves delegating some decision-making authority to the agent. Here, the principal(s) is (are) the shareholder(s) of a certain firm. The agent can be seen as the management of that particular firm. Further, Jensen and Meckling (1976) mention that it can be that the agent will not act in the best interests of the principal. This can be the case because the agent will maximize his own welfare instead of the principal’s welfare. Berhold (1971) states that the principal wants to motivate the agent to make an appropriate decision, because the principal’s results depend on the agent’s decisions. Jensen and Meckling (1976, p. 308) mention in their research that “in some situations it will pay the agent to expend resources (bonding costs) to guarantee that he will not take certain actions which would harm the principal or to ensure that the principal will be compensated if he does take such actions” (p. 308). This monitoring is the job of the board of directors. They monitor the managers of the firm. At the end, the firm benefits from these monitoring activities when the benefits outweigh the costs. A benefit from a well-functioning monitoring body is that it constrains management from self-interested behavior. Earnings management should be constrained to a minimum when the board of directors functions well.
2.2 Earnings management
Dechow and Dichev (2002) describe the role of accruals as to shift or adjust recognition of cash flows over time so that earnings capture firm performance. They say that there is a need for estimates and that these estimates can be subject to errors. First, they say, there are unintentional errors. This is the case when estimates were wrong, for example due to the complexity of the estimation. Then, there are also intentional errors, which are estimates that are biased. Dechow and Dichev (2002) define this last type of errors as earnings management. Regarding earnings management, there are also two types. First, there is earnings management, which are accounting choices that managers make with the goal of deliberately biasing accounting information. And there is real earnings management, which
9 are actual decisions that managers make with the purpose of deliberately influencing accounting information. In this thesis, I will just focus on the first type of earnings management.
Dichev, Graham, Harvey and Rajgopal (2013) interviewed CFOs of public companies and two standard setters and asked their opinions about the reasons behind earnings management. The top three are as follows: (1) to influence stock price, (2) because there is outside pressure to hit earnings benchmarks and (3) because there is inside pressure to hit earnings benchmarks. Next to the reasons behind earnings management, Dichev et al. (2013) also provide a list of signals that can be used to detect earnings management. The top three are as follows: (1) GAAP earnings do not correlate with cash flow from operations; weak cash flows; earnings and cash flow from operations move in different direction for 6-8 quarters; earnings strength with deteriorating cash flow, (2) deviations from industry (or economy, peers’) norms/experience (cash cycle, volatility, average profitability, revenue growth, audit fees, growth of investments, asset impairment, A/P, level of disclosure) and (3) consistently meet or beat earnings targets (guidance, analyst forecasts).
2.3 Board of directors
Shivdasani and Yermack (1999) state that the board of directors serves as the central mechanism for monitoring the managers of a public firm. The board of directors consists of directors.They are chosen by the shareholders of the firm. These directors have a fiduciary responsibility to protect shareholders’ interest. Next to this, the directors are charged with selecting, compensating, evaluating and dismissing top management. Fama and Jensen (1983) argue that a board of directors also has the duty to assist and monitor management in their effort to maximize shareholder wealth. Peasnell, Pope and Young (2000) say that boards include non-executive directors to facilitate effective monitoring. Baysinger and Butler (1985) mention in their research that according to many reformers, the board of directors of all major U.S. firms should be independent. They explain an independent board as a board with a majority of non-executive directors.
There are two main sets of legal rules on the supervision of corporate management: one-tier boards and two-tier boards (Jungmann, 2006). A one-tier board consists of executive directors and non-executive directors and a two-tier board consists of a management board and a supervisory board. Because this research is focused on North American firms, the focus will be on one-tier boards. Theory about two-tier boards lies outside the scope of this research and therefore will not be discussed.
The directors of a one-tier board can be divided into three categories (Adams, Hermalin and Weisbach, 2010). First, there are inside directors. These are directors who are a full-time employee of the firm. Second, there are outside directors. These directors are also called independent directors or non-executive directors. A distinction with inside directors is
10 that an outside director’s primary employment is with a different organization than the firm on whose board he serves. Third and last, there are gray directors. These are outside directors, but there are doubts whether they are really independent. Adams et al. (2010) state that outside directors mostly have a background that will enable them to be valuable for a board of directors. It can also be that they represent an important constituency.
The New York Stock Exchange states in its Listed Company Manual that a majority of independent directors will increase the quality of the board oversight and lessen the possibility of damaging conflicts of interest. Further it states that a director can be qualified as an independent director when “the board of directors affirmatively determines that the director has no material relationship with the listed company (either directly or as a partner, shareholder or officer of an organization that has a relationship with the company)” (section 303A.00).
In the Corporate Governance Requirements of the NASDAQ Stock Market is explained that independent directors act on behalf of investors to maximize shareholder value and guard against conflicts of interest. Further, it states that a board that consists of a majority of independent directors empowers these directors to carry out more effectively these responsibilities. An independent director is described by the NASDAQ Stock Market as “a person other than an Executive Officer or employee of the Company or any other individual having a relationship which, in the opinion of the Company's board of directors, would interfere with the exercise of independent judgment in carrying out the responsibilities of a director”.
Adams et al. (2010) argue that many firms have bankers on their board. They are added to a board because in this way they can monitor the firm for the bank they work for. Next to this, they can also be added to a board because of their financial expertise. Labor is another group that also often serves on a board of directors. Adams et al. (2010) say that labor wants to be represented on the board so that it can influence management to take actions which are favorable to workers. Another group, which is often represented on a board, are people who have political connections. Also venture capitalists can serve on a board of directors. Many firms are founded with the help of funds from venture capitalists. A condition to these funds, usually, is a degree of control to these venture capitalists. Finally, Adams et al. (2010) mention CEOs from another firm, as one of the most common occupations of outside directors. They state that CEOs have management skills and an understanding of the issues facing top management. Many of these people have full-time jobs or serve on many boards. Adams et al. (2010) argue that there is a concern that such busy directors will not be able to devote sufficient effort to any board. Therefore, the effectiveness of these busy directors has become an issue for discussion.
SOX requires public companies in the U.S. to have an audit committee, which comprises of independent members of the board of directors. The act describes an audit committee as “a committee (or equivalent body) established by and amongst the board of directors of an issuer
11 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” (Sarbanes-Oxley Act of 2002, p. 747). Klein (2002) uses three definitions of independence. First, she uses the interpretation of independence as the percentage of independent directors. Second, she refers to the guidelines of the NASDAQ and NYSE. Those two consider an audit committee independent if all members are independent directors. The third interpretation she uses for independence is whether a majority of the members are independent. Each issuer should disclose whether or not the audit committee is comprised of at least one member who is a financial expert. To define a financial expert, one should consider whether a person has, through education and experience, an understanding of generally accepted accounting principles and financial statements, experience in the preparation or auditing of financial statements, experience with internal accounting controls and an understanding of audit committee functions.
2.4 Director busyness
Those busy directors, mentioned by Adams et al. (2010), are directors who serve on multiple outside boards. Most researchers agree about the definition of a busy director. The Council of Institutional Investors published in 1998 their Corporate Governance Policies. Here, they stated that “absent unusual, specified circumstances, directors with full-time jobs should not serve on more than two other boards. Currently serving CEOs should not serve as a director of more than one other company, and then only if the CEO’s own company is in the top half of its peer group. No other director should serve on more than five for-profit company boards” (p. 8). Core et al. (1999) say that busy outside directors are outside directors who serve on three or more other outside boards. For retired outside directors, they mention a cutoff of six or more outside boards. Also Ferris et al. (2003) use this cutoff of three or more outside boards, just like Fich and Shivdasani (2006), Cashman et al. (2012) and Field et al (2013). Cashman et al. (2012) used independent variables from prior literature like Fich and Shivdasani (2006) and Ferris et al. (2003) to proxy for busyness. Next to that, they tried to find other proxies that can give a better view of the demands placed on a director’s time. The first dimension they focused on is the complexity of the firm. Here, they looked at firm size, number of business segments and R&D intensity. The second dimension stands for additional time commitments that the individual may face. For this dimension, the authors looked at whether the director is serving on another S&P 1500 firm, whether the director serves on other outside board committees and whether the director serves on the outside board of a firm that operates in a different industry. Concerning the aforementioned alternative proxies, they state that these are just as informative as the regular proxies used by prior literature and that they are highly correlated, suggesting that they add little in explaining firm performance. Therefore, they advise future researchers to stick with the proxies from prior literature. The
12 way in which I calculate busyness for all four hypotheses in this research will be explained in paragraph 3.3.
2.5 Literature review and hypotheses
In this paragraph, I will discuss literature where the relationship is investigated between particular board characteristics and earnings management. These particular board characteristics are board size, board independence, audit committee independence and the presence of a financial expert on a board of directors. In this thesis, I will add board busyness, audit committee busyness and financial expert busyness as a moderating variable and I will compare the results of this thesis with the literature that I will discuss subsequently.
Xie, Davidson III and DaDalt (2003) examined the role of the board of directors, the audit committee and the executive committee in preventing earnings management. They say that a smaller board may be less encumbered with bureaucratic problems, which makes them more functional than a larger board. Next to that, they may provide better financial reporting oversight. On the other hand, a larger board can have a broader range of experience and may be more likely to have independent directors with corporate or financial experience. Given these conflicting arguments, Xie et al. (2003) do not have directional expectations regarding the relationship between board size and earnings management. Ultimately, they find a negative and significant relationship between board size and earnings management, which means that larger boards are associated with lower levels of discretionary current accruals. Karamanou and Vafeas (2005) state that adding more people to the board enhances its knowledge, but, larger boards are less flexible and more inefficient than smaller boards. Rahman and Ali (2006) show a significantly positive relationship between discretionary accruals and board size. This means that the larger the board, the higher the discretionary accruals are. The aforementioned researches all investigated the relationship between earnings management and, among others, board size. I will add board busyness as an interacting variable to investigate whether busyness has a moderating impact on this relationship. Like mentioned in section 1, director busyness can lead to a negative impact because of a lack of time or energy to serve on more outside boards. However, it can also lead to a positive impact because serving on multiple outside boards results in experienced directors with a broad network. Because board busyness can either have a negative or a positive impact, hypothesis 1 is formulated as a two-tailed hypothesis. Therefore, a significantly positive relationship as well as a significantly negative relationship both will be support for this hypothesis. Hypothesis 1 is formulated as follows:
H1: The presence of one or more busy non-executive directors on a board of directors impacts the relationship between board size and earnings management.
13 Sarkar, Sarkar and Sen (2008) find that board independence, measured by majority of non-executive directors or by the percentage of non-non-executive directors of a board, have no significant relation with discretionary accruals. Liu and Lu (2007) notice a significantly negative relationship between the percentage of non-executive directors on a board and earnings management. Park and Shin (2004) find no evidence in Canada for an association between the degree of accrual manipulation and the proportion of non-executive directors on the board of directors. As a possible reason, they mention that these non-executive directors in Canada may lack financial sophistication and/or access to relevant information to detect and correct earnings management. The results of Rahman and Ali (2006) show that they could not find a significant relationship between the percentage of non-executive directors on a board and earnings management. Ghosh, Marra and Moon (2010) observe a decline in discretionary accruals when boards become more independent. Peasnell, Pope and Young (2005) find that firms with a higher percentage of non-executive directors are associated with less income-increasing earnings management when pre-managed earnings are below zero or below last year’s earnings. Here, they conclude that non-executive directors play an important monitoring role. On the other hand, they find no evidence that non-executive directors constrain income-decreasing earnings management. These two findings show that monitoring by non-executive directors is asymmetric, with the emphasis on constraining income-increasing earnings management. Davidson, Goodwin-Stewart and Kent (2005) find support for their hypothesis that earnings management is negatively associated with board independence. The aforementioned researches all investigated the relationship between earnings management and board independence. For this thesis, I will add board busyness as an interacting variable to investigate whether board busyness has a moderating impact on this relationship. Like mentioned earlier in this paragraph, board busyness can either have a negative or a positive impact. Therefore, hypothesis 2 is also formulated as a two-tailed hypothesis. A significantly positive relationship as well as a significantly negative relationship both will be support for this hypothesis. Hypothesis 2 is formulated as follows: H2: The presence of one or more busy independent directors on a board of directors
impacts the relationship between board independence and earnings management. Bédard, Chtourou and Courteau (2004) find that an audit committee has to comprise of one hundred percent non-executive directors for efficient monitoring. Klein (2002) says that firms that move from a majority-independent to a minority-independent audit committee experience larger increases in adjusted abnormal accruals. She concludes that a majority outside membership may be a critical threshold for deriving a meaningful relation between director independence and the absolute value of the adjusted abnormal accruals. Xie et al. (2003) find that there is a negative relationship between the percentage of non-executive directors on the audit committee and earnings management. Klein (2002) finds higher
14 abnormal accruals for firms with audit committees comprised of less than a majority of non-executive directors. Next to that, she finds a negative association between abnormal accruals and the percentage of non-executive directors serving on the audit committee. Rahman and Ali (2006) do not find sufficient evidence that there is a negative relationship between discretionary accruals and the percentage of non-executive directors, which is also the conclusion of Piot and Janin (2007). Yang and Krishnan (2005) examined the association between audit committee characteristics and measures of quarterly earnings management. They found that the experience of an audit committee member, measured by the amount of other boards they are serving on as an independent director, is significantly associated with lower quarterly discretionary accruals. All these researches investigated the relationship between earnings management and audit committee independence. I will add board busyness as an interacting variable. Now I can investigate whether director busyness has an moderating impact on the relationship between earnings management and audit committee independence. Because director busyness can either have a negative or a positive impact, also hypothesis 3 is formulated as a two-tailed hypothesis. Therefore, a significantly positive relationship as well as a significantly negative relationship both will be support for this hypothesis. Hypothesis 3 is formulated as follows:
H3: The relationship between the percentage of independent directors serving on an audit committee and earnings management will be impacted by the presence of one or more busy independent directors on the audit committee.
Bédard et al. (2004) hypothesized that firms that have a financial expert on their audit committee are less likely to engage in aggressive earnings management. For their sample of firms from COMPUSTAT for 1996, they find evidence that supports their hypothesis. They state that financial expertise seems to decrease the likelihood of earnings management, positive as well as negative. They argue that the experience of non-executive directors allows them to develop their monitoring capabilities. Carcello, Hollingsworth, Klein and Neal (2006) find in their research a significantly negative relationship between the presence of at least one audit committee financial expert and abnormal accruals. Further, they divided their finding into accounting financial expert and non-accounting financial expert. They only find a significantly negative association between an accounting financial expert and earnings management. Bédard et al. (2004) and Carcello et al. (2006) both investigated the relationship between earnings management and the presence of a financial expert on the audit committee. However, I would like to investigate what impact a busy financial expert on the audit committee has on the aforementioned relationship. Also here, I have no directional expectations. Therefore, hypothesis 4 is formulated in a two-tailed way. Support for this hypothesis can either be a significantly negative relationship as well as a significantly positive relationship. Hypothesis 4 is formulated as follows:
15 H4: The relationship between the fact that the audit committee contains a financial expert and earnings management will be impacted when on or more of these experts can be qualified as a busy director.
3. Research methodology
3.1 Sample selection
To test the aforementioned hypotheses in paragraph 2.5, I initially started with a sample of firms from COMPUSTAT for the period 2008 through 2013. This led to a sample of 10.088 firm year observations, which can be seen at table 1. Next, I collected data about director characteristics from ISS by importing CUSIP numbers I retrieved from the data I already collected from COMPUSTAT.
TABLE 1
Sample N
Number of firm-year observations for the sample period 2008 through 2013 10.088 Less: Number of firm-year observations with missing data (1.205)
regarding director characteristics
Less: Number of firm-year observations with missing data (3.228) regarding firm financial information
Less: Firm-year observations from financial/insurance companies (SIC 6000-6411) (178)
Number of firm-year observations in the final sample 5.477
Number of unique firms 1.232
Then, I imported both collected data sets into Stata and merged them to get one large data set. Because of missing data, I had to remove 1.205 firm-year observations from my sample regarding director characteristics and 3.228 firm-year observations regarding firm financial information. Finally, I removed 178 firm-year observations from financial institutions and insurance companies. These are firms with an SIC code between 6000 and 6411. In the end, removing the aforementioned year observations leads to a final sample of 5.477 firm-year observations, containing 1.232 unique firms.
3.2 Earnings management
Earnings management is the dependent variable in this research. It will be measured by discretionary accruals, scaled by total assets at t-1. Dechow, Sloan and Sweeney (1995) say that the Jones Model implicitly assumes that revenues are non-discretionary. They developed
16 the Modified Jones Model to eliminate the conjectured tendency of the Jones Model to measure discretionary accruals with error when discretion is exercised over revenues. To determine discretionary accruals I use model (1), which is in line with the Modified Jones Model. All variables are winsorized to the 1st
and 99th
percentiles of their distributions to avoid results due to extreme values. It is not necessary to winsorize 1/Ai,t-1 because this variable is already constrained by a minimum of zero.
TAi,t = α1 + β2 (1/Ai,t-1) + β3 (ΔSALEi,t – ΔRECi,t) + β4 PPEi,t + β5 ROAi,t + εi,t (1)
In the above model, TAi,t stands for total accruals in year t scaled by total assets at t-1 for firm i, calculated by subtracting cash flow from operations from income before extraordinary items and then divide this result by total assets at t-1. Ai,t-1 are the total assets at t-1 for firm i; ΔSALEi,t are the sales in year t less sales in year t-1 scaled by total assets at t-1 for firm i; ΔRECi,t are total receivables in year t less total receivables in year t-1 scaled by total assets at t-1 for firm i; PPEi,t is the gross property, plant and equipment in year t scaled by total assets at t-1 for firm i; and ROAi,t stands for return on assets in year t for firm i, calculated by dividing income before extraordinary items by total assets at t-1.
The error term in model (1) stands for the part of total accruals that cannot be explained by the independent variables. Since total accruals is the sum of non-discretionary accruals and discretionary accruals, this error term can be interpreted as the discretionary accruals scaled by total assets at t-1 for firm i and will function as the dependent variable in this research.
3.3 Director busyness
In this research, I will use director busyness as a moderating variable. Director busyness can be defined as the total number of outside boards an independent director is serving on. Prior literature, such as Core et al. (1999), Ferris et al. (2003), Fich and Shivdasani (2006), Field et al. (2013) and Cashman et al. (2012) use a cutoff of three other outside boards an independent director is serving on to qualify this director as a busy director. In this research I will use this same cutoff.
For the first and second hypothesis, I will calculate board busyness by dividing the number of busy independent directors on a board by total independent directors on this board. For the third hypothesis, I will calculate audit committee busyness by dividing the number of busy independent directors on the audit committee by total independent directors on the audit committee. At last, for the fourth hypothesis, I will calculate financial expert busyness by dividing the number of busy financial experts serving on an audit committee by total financial experts on the audit committee. All these moderating variables will be continuous, with ranging possibilities from 0% to 100%.
17
3.4 Control variables
Next to the aforementioned variables that will be used in this research, I also selected control variables to measure the relative impact of the independent variables. I selected total assets, market value and revenues as a control variable to proxy for firm size. I also added leverage as a control variable, because leverage may be associated with discretionary accruals. Becker, DeFond, Jiambalvo and Subramanyam (2010) argue that managers of highly leveraged firms have incentives to make income-increasing discretionary accruals to avoid debt covenant violation. From this finding, one can expect that there is a positive relationship between leverage and discretionary accruals.
3.5 Empirical model
To test the hypotheses, formulated in paragraph 2.5, I developed the following models. They are all four built up in the same way. First, discretionary accruals are noted as the dependent variable, which is a proxy for earnings management. Then, the intercept is noted. In the models, the intercept is noted as α0. Subsequently, the independent variables are noted. These independent variables are the corporate governance characteristics mentioned in paragraph 2.5 (β1), director busyness (β2) and the interaction between these two (β3). For model (2), the corporate governance characteristic is board size. For model (3) it is board independence, for model (4) it is audit committee independence and for model (5) it is financial expert presence. In every model, β2 stands for director busyness. However, in every model this variable is calculated differently. In model (2) and (3), β2 stands for board busyness. In model (4), β2 stands for audit committee busyness. In model (5), β2 stands for financial expert busyness. The calculation of these busyness variables will be explained after every model. Finally, every model has the control variables assets, revenues, market value and leverage. To test the first hypothesis, I will use model (2):
DAi,t = α0 + β1 BSZi,t + β2 BDi,t + β3 (BSZi,t * BDi,t) + β4 Ln_Assetsi,t +
β5 Ln_Revi,t +β6 Ln_MKVi,t + β7 Levi,t + εi,t (2)
In model (2), DAi,t stands for the discretionary accruals in year t scaled by total assets in year t-1 for firm i; BSZi,t is board size in year t for firm i; BDi,t is the percentage of busy independent directors on the board of directors in year t for firm i, calculated as total number of busy independent directors on the board divided by total number of independent directors on the board; Ln_Assetsi,t is the natural logarithm of the assets in year t for firm i; Ln_Revi,t is the natural logarithm of revenues in year t for firm i; Ln_MKVi,t is the natural logarithm of the market value in year t for firm i; and Levi,t is the leverage in year t for firm i, calculated as total debt divided by total assets. A significant β3 in model (2) would offer support for hypothesis 1. A negative coefficient will mean that the higher board size, the lower earnings
18 management, when the board consists of one or more busy directors. A positive coefficient will mean that the higher board size, the higher earnings management, when the board consists of one or more busy directors.
To test the second hypothesis, I will use model (3):
DAi,t = α0 + β1 INDi,t + β2 BDi,t + β3 (INDi,t * BDi,t) + β4 Ln_Assetsi,t +
β5 Ln_Revi,t +β6 Ln_MKVi,t + β7 Levi,t + εi,t (3) In model (3), INDi,t stands for board independence and is the percentage of independent directors on the board of directors in year t for firm i, calculated as total number of independent directors on the board divided by total number of directors on the board. A significant β3 in model (3) would offer support for hypothesis 2. A negative coefficient will mean that the higher board independence, the lower earnings management, when the board consists of one or more busy directors. A positive coefficient will mean that the higher board independence, the higher earnings management, when the board consists of one or more busy directors.
To test the third hypothesis, I will use model (4):
DAi,t = α0 + β1 IND_AUCi,t + β2 BD_AUCi,t + β3 (IND_AUCi,t * BD_AUCi,t)
+ β4 Ln_Assetsi,t + β5 Ln_Revi,t +β6 Ln_MKVi,t + β7 Levi,t + εi,t (4) In model (4), IND_AUCi,t stands for audit committee independence and is the percentage of independent directors on the audit committee in year t for firm i, calculated as total number of independent directors on the audit committee divided by total number of directors on the audit committee; and BD_AUCi,t, which stands for audit committee busyness and is the percentage of busy independent directors on the audit committee in year t for firm i, calculated as total number of busy independent directors on the audit committee divided by total number of independent directors on the audit committee. A significant β3 in model (4) would offer support for hypothesis 3. A negative coefficient will mean that the higher audit committee independence, the lower earnings management, when the audit committee consists of one or more busy directors. A positive coefficient will mean that the higher audit committee independence, the higher earnings management, when the audit committee consists of one or more busy directors.
19 To test the fourth and last hypothesis, I will use model (5):
DAi,t = α0 + β1 AUC_FEXPi,t + β2 BD_FEXPi,t + β3 (AUC_FEXPi,t * BD_FEXPi,t) + β4 Ln_Assetsi,t + β5 Ln_Revi,t +β6 Ln_MKVi,t + β7 Levi,t + εi,t (5)
In model (5), AUC_FEXPi,t is a dummy variable and measures whether there is a financial expert on the audit committee in year t for firm i; and BD_FEXPi,t, which stands for financial expert busyness and is the percentage of busy financial experts on the audit committee in year t for firm i, calculated as total number of busy financial experts on the audit committee divided by total number of financial experts on the audit committee. A significant β3 in model (5) would offer support for hypothesis 4. A negative coefficient will mean that the presence of at least one financial expert on the audit committee is related to lower earnings management, when the financial expert(s) is (are) busy. A positive coefficient will mean that the presence of at least one financial expert on the audit committee is related to higher earnings management, when the financial expert(s) is (are) busy.
4. Results
4.1 Descriptive statistics
Table 2 provides the descriptive statistics for the variables, which are used in the models. Panel A shows the accruals, panel B shows the regular independent variables, panel C shows the interacting independent variables and panel D shows the control variables.
In panel A, discretionary accruals are reported. They are scaled by total assets at t-1 and can therefore be interpreted as a percentage of the balance sheet total. In this research, I use discretionary accruals as a proxy for earnings management and therefore this variable is the dependent variable in this research. It has a mean of 0,001, meaning that, overall, management made income-increasing accrual decisions. However, these decisions have a low impact on the balance sheet, since the extent of the discretionary accruals is just 0,1% of the balance sheet total.
In panel B, the regular independent variables are reported. Board size varies from minimum five to maximum fifteen board members and comprises of employees and insiders, independent directors and linked directors. Board size has a mean of 9,146 members, which is much lower than the 12,48 and 13 reported by Xie et al. (2003), using a sample of S&P firms for the years 1992, 1994 and 1996, and Core et al. (1999), using publicly traded U.S. firms over an unknown period. On the other hand, the board size mean of the sample of this thesis is much larger than those of Field et al. (2013) and Kiel and Nicholson (2006). They show a mean of 6,8 for a sample of venture-backed firms incorporated in the U.S. that went
20 public from 1996 through 2008 and 5,7 for a sample of 1.326 firms listed on the Australian Securities Exchange as at 30 June 2003.
When two or more predicting variables strongly correlate with each other, this has an effect on the estimates for a regression model (Chen, Ender, Mitchell and Wells, 2003). These estimates involve the coefficients and the standard errors of the concerning variables and can be impacted upwards or downwards. Because interactions contain predicting variables that are also noted separately in a model, there is a real chance that these variables will cause collinearity. Because of this real chance of collinearity, I centered all predicting variables. For every observation, I subtracted the mean of the particular variable from the individual observation. Because of this, all predicting variables are centered around their mean. These are the variables reported in panel B and panel C from table 2. However, to keep the descriptive statistics understandable, panel B and C report the original values instead of their centered values. The analysis regarding collinearity for my sample will be further explained in paragraph 4.2.
Panel B shows that 79,4% of the directors on the board are independent. Compared to the sample of Klein (2002), who reports for her sample that 58,4% of the directors on the board are independent, this is a relatively high percentage. A higher percentage is provided by Park and Shin (2004). For their sample of Canadian firms from 1991 to 1997, they show that 67,3% of the board members are outsiders. Quite similar with Park and Shin (2004), Cashman et al. (2012) report a board independence of 69%.
The audit committee has to comprise fully of independent directors according to SOX. From my selected sample, 99,2% of the audit committee members are independent. This is very well in line with SOX. Furthermore, SOX requires firms to disclose whether or not the audit committee contains at least one financial expert (Sarbanes-Oxley Act of 2002, p. 790). Who can be qualified as a financial expert is explained in paragraph 2.3. From table 2 can be read that 98,9% of the firms have a financial expert on their audit committee. Also at this point regarding the presence of a financial expert on the audit committee, the firms from the sample operate in line with SOX.
Panel C reports the interacting independent variables. These variables interact with the regular independent variables, which can be seen in the models in paragraph 3.5. This panel reports a board busyness of 8,7%. This means that 8,7% of the independent directors on the firm’s boards of the sample can be qualified as a busy director. Fich and Shivdasani (2006) show a much higher percentage of 52,26% of busy independent directors in their sample. Andres et al. (2013) show a similar percentage as Fich and Shivdasani (2006) for their sample of 133 German firms for the period 2003 through 2006, in that 52,44% of the directors are busy. Cashman et al. (2012) show a percentage of busy directors of 16%, which is much closer to my sample than that of Fich and Shivdasani (2006) and Andres et al. (2013).
21
TABLE 2
Descriptive statistics
Panel A: Accruals
Variable Mean Std. dev. Median Minimum Maximum
Discretionary accrualsa 0,001 0,052 0,004 -0,323 0,201
Panel B: Regular independent variables
Mean Std. dev. Median Minimum Maximum
Board size 9,152 2,076 9 5 15
Board independence 0,794 0,105 0,820 0,500 0,920
Audit committee independence 0,992 0,045 1,000 0,670 1,000
Financial expert on audit 0,989 0,104 1,000 0,000 1,000
committee
Panel C: Interacting independent variables
Mean Std. dev. Median Minimum Maximum
Board busyness 0,087 0,114 0,000 0,000 0,500
Audit committee busyness 0,083 0,144 0,000 0,000 0,600
Financial expert busyness 0,093 0,232 0,000 0,000 1,000
Panel D: Control variables
Mean Std. dev. Median Minimum Maximum
Assetsb 21,648 1,530 21,502 18,689 25,487 Revenuesb 21,516 1,482 21,396 18,357 25,339 Market valueb 21,662 1,518 21,531 18,528 25,728 Leveragec 0,503 0,202 0,511 0,095 1,031 a
Scaled by total assets at t-1 b
Values represent the natural logarithm c
Calculated as debt divided by assets
Finally, Core et al. (1999) report a director busyness of 45,17%. All these percentages are much higher than the 8,7% of my sample. A possible explanation can be that the samples of Fich and Shivdasani (2006) and Core et al. (1999) consist of firm-year observations between 1982 and 1995. In this period, the discussion about director busyness was not that popular as it is nowadays. It could be that director busyness was normal those days and therefore the percentages for director busyness are so much higher. Panel C further reports that 8,3% of the
22 audit committee members are busy directors and 9,3% of the financial exerts serving on the audit committee can be qualified as busy. Like already mentioned before, the part about audit committee busyness and financial expert busyness is not investigated yet. Descriptive statistics from prior literature is therefore not available to compare with the descriptives of my own sample.
Panel D reports the control variables. All of them are transformed to their natural logarithm because of their positive skewness. This transformation reduced their skewness from between 6 and 10 to near zero for all variables. The means of 21,648 for assets, 21,516 for revenues and 21,662 for market value correspond with an original value of approximately 2.521 million for assets, 2.209 million for revenues and 2.557 million for market value. Finally, panel D reports leverage, calculated as debt divided by assets, of 50,3%. This means that the firms from the sample have, on average, assets that are twice as much as debt.
Table 3 reports the correlations between the variables. Within parentheses, the significance is reported. The correlations can have values between -1 and 1, where -1 means a total negative correlation and +1 means a total positive correlation. When there is no correlation between the variables, the correlation coefficient is zero. If there is a correlation higher (lower) than 0,7 (-0,7), the two variables are classified as highly correlated. This cutoff is widely used by researchers. When two independent variables in a multiple regression model are highly correlated with each other, this phenomenon is also called collinearity. Correlations higher (lower) than 0,7 (-0,7) are made bold in table 3. From this table it is apparent that there are strong significant correlations at panel A through D between (4) assets and (5) revenues, between (4) assets and (6) market value, between (5) revenues and (6) market value. One explanation can be that this is the case because they all can be seen as an indicator of the size of the firm. Usually, large firms have relatively high assets, revenues and market value. Next to these correlations, there is also a strong correlation in panel D between (1) financial expert on audit committee and (3) financial expert on audit committee * financial expert busyness. Because these two variables are both independent variables in model (5), this high correlation can be classified as collinearity. Further explanation regarding this issue can be found in paragraph 4.2. There, one can clearly see the effect of this collinearity on the outcome of regression model (5).
To further test for collinearity, I ranked the variance inflation factors (VIF) and tolerance (1/VIF) for all variables (untabulated). When a variable has a VIF of ten or greater, this indicates a need for further investigation. The tolerance, calculated by 1/VIF, has a cutoff of 0,1. When the tolerance is 0,1 or lower, there is a possibility of collinearity. In all four models, only the control variable assets has a VIF greater than 10 and a tolerance lower than 0,1. This is the case because I centered all independent variables, like explained earlier in this paragraph. The VIF value and tolerance of the control variable assets hardly deviate from the aforementioned cutoffs. Therefore, I will not further investigate this variable with regards to collinearity.
TABLE 3
Pearson's r correlations
Panel A: hypothesis 1 Panel B: hypothesis 2
(1) (2) (3) (4) (5) (6) (7) (1) (2) (3) (4) (5) (6) (7) (1) 1,0000 (1) 1,0000 ( - ) ( - ) (2) 0,1756 1,0000 (2) 0,1023 1,0000 (0,0000) ( - ) (0,0000) ( - ) (3) 0,0333 0,0121 1,0000 (3) -0,0406 -0,0871 1,0000 (0,0138) (0,3707) ( - ) (0,0026) (0,0000) ( - ) (4) 0,6076 0,2171 0,0458 1,0000 (4) 0,2676 0,2171 -0,0212 1,0000 (0,0000) (0,0000) (0,0007) ( - ) (0,0000) (0,0000) (0,1167) ( - ) (5) 0,5971 0,2225 0,0521 0,9120 1,0000 (5) 0,2203 0,1902 -0,0156 0,9120 1,0000 (0,0000) (0,0000) (0,0001) (0,0000) ( - ) (0,0000) (0,0000) (0,2496) (0,0000) ( - ) (6) 0,5251 0,1902 0,0643 0,8860 0,8178 1,0000 (6) 0,2272 0,2225 0,0057 0,8860 0,8178 1,0000 (0,0000) (0,0000) (0,0000) (0,0000) (0,0000) ( - ) (0,0000) (0,0000) (0,6751) (0,0000) (0,0000) ( - ) (7) 0,3759 0,1615 -0,0151 0,4632 0,4659 0,2493 1,0000 (7) 0,2369 0,1615 -0,0495 0,4632 0,4659 0,2493 1,0000 (0,0000) (0,0000) (0,2653) (0,0000) (0,0000) (0,0000) ( - ) (0,0000) (0,0000) (0,0002) (0,0000) (0,0000) (0,0000) ( - )
Where (1) stands for board size; (2) for board busyness; (3) for board size * board busyness; (4) for assets; (5) for revenues; (6) for market value; and (7) for leverage.
Where (1) stands for board independence; (2) for board busyness; (3) for board independence * board busyness; (4) for assets; (5) for revenues; (6) for market value; and (7) for leverage.
24
TABLE 3
Pearson's r correlations
Panel C: hypothesis 3 Panel D: hypothesis 4
(1) (2) (3) (4) (5) (6) (7) (1) (2) (3) (4) (5) (6) (7) (1) 1,0000 (1) 1,0000 ( - ) ( - ) (2) 0,0421 1,0000 (2) 0,0422 1,0000 (0,0018) ( - ) (0,0018) ( - ) (3) -0,3113 0,0592 1,0000 (3) -0,9658 0,2183 1,0000 (0,0000) (0,0000) ( - ) (0,0000) (0,0000) ( - ) (4) 0,0540 0,1604 -0,0301 1,0000 (4) 0,0494 0,1176 -0,0177 1,0000 (0,0001) (0,0000) (0,0260) ( - ) (0,0003) (0,0000) (0,1903) ( - ) (5) 0,0502 0,1730 -0,0200 0,9120 1,0000 (5) 0,0465 0,1267 -0,0126 0,9120 1,0000 (0,0002) (0,0000) (0,1398) (0,0000) ( - ) (0,0006) (0,0000) (0,3521) (0,0000) ( - ) (6) 0,0420 0,1585 -0,0262 0,8860 0,8178 1,0000 (6) 0,0462 0,1129 -0,0158 0,8860 0,8177 1,0000 (0,0019) (0,0000) (0,0528) (0,0000) (0,0000) ( - ) (0,0006) (0,0000) (0,2416) (0,0000) (0,0000) ( - ) (7) 0,0438 0,0919 -0,0182 0,4632 0,4659 0,2493 1,0000 (7) 0,0424 0,0495 -0,0285 0,4632 0,4658 0,2493 1,0000 (0,0012) (0,0000) (0,1792) (0,0000) (0,0000) (0,0000) ( - ) (0,0017) (0,0002) (0,0348) (0,0000) (0,0000) (0,0000) ( - )
Where (1) stands for audit committee independence; (2) for audit committee busyness; (3) for audit committee independence * audit committee busyness; (4) for assets; (5) for revenues; (6) for market value; and (7) for leverage.
Where (1) stands for financial expert on audit committee; (2) for financial expert busyness; (3) for financial expert on audit committee * financial expert busyness; (4) for assets; (5) for revenues; (6) for market value; and (7) for leverage.
4.2 Multivariate analysis
Table 4 reports the regression results for hypothesis 1. The independent variables are board size, board busyness and their interaction board size * board busyness. The control variables are assets, revenues, market value and leverage. These control variables are the same for all four models. Panel A reports a very small coefficient and t-value for board size. Therefore, this variable is not significant, meaning that there is no significant relationship between the size of a board of directors and earnings management. However, these results cannot be compared with prior literature because of the presence of the interaction variable in this regression. Now, the term for board size picks up the effect of board size when board busyness is zero. When I drop the interaction term in the model, running the regression resulted in a similar outcome of t = 0,09, p = .93 (untabulated). This finding is not in line with research of Xie et al. (2003) who found a negative and significant relationship between board size and earnings management, nor is it in line with Rahman and Ali (2006) who found a significant positive relationship between board size and discretionary accruals. Board busyness has also these low values and is not significant, meaning that there is no significant relationship between the percentage of busy directors on a board of directors and earnings management. However, the interaction term board size * board busyness has a coefficient of 0,008, a t-value of 2,84 and is significant at the 0,01 level. This means that there is a positive and significant relationship between board size and earnings management, when the particular board contains busy directors.
TABLE 4
Regression results hypothesis 1
Variable Coefficient Std. error t-value p-value
Intercept -0,043 0,013 -3,19 0,001 ***
Board size 0,000 0,000 0,06 0,948
Board busyness -0,001 0,006 -0,20 0,842
Board size * board busyness (H1) 0,008 0,003 2,84 0,005 ***
Assets 0,005 0,001 3,05 0,002 *** Revenues -0,000 0,001 -0,16 0,875 Market value -0,002 0,001 -1,73 0,084 * Leverage -0,019 0,004 -4,49 0,000 ***
* Significant at the 0,1 level
** Significant at the 0,05 level
*** Significant at the 0,01 level
26 The coefficient of 0,008 indicates that the effect of board size on earnings management is 0,008 higher for every single increase in board size, when this board consists solely of busy directors. So, when a board increases by one director, ceteris paribus, earnings management increases by 0,008. From these results I can conclude that a higher percentage of busy non-executive directors on a board of directors has a significant and strengthening impact on the relationship between board size and earnings management. Based on these results, hypothesis 1 should be supported.
Table 5 reports the regression results for hypothesis 2. The independent variables are board independence, board busyness and board independence * board busyness. All independent variables of hypothesis 2 are not significant. From these results can be concluded that hypothesis 2 cannot be supported. There is no significant relationship between board independence and earnings management. Like mentioned before, these results cannot be compared with prior literature because of the presence of the interaction term in this regression. Therefore, I dropped the interaction term in the model to compare the results with prior literature. Running the regression resulted in a similar outcome of t = -1,42, p = .16 (untabulated). This is in accordance with research of Sarkar et al. (2008), Park and Shin (2004) and Rahman and Ali (2006) who all did not find any significant relationship between board independence and earnings management. Though, my results are not in line with Liu and Lu (2007), Ghosh et al. (2010), Peasnell et al. (2005) and Davidson et al. (2005). They all found a negative and significant relationship between board independence and earnings management, meaning that the more independent a board becomes, the more earnings management decreases. The conclusion from this regression can be drawn that a higher percentage of busy non-executive directors on a board of directors does not impact the relationship between board independence and earnings management. Hypothesis 2 can therefore not be supported.
Table 6 reports the regression results for hypothesis 3. Here, independent variables are audit committee independence, audit committee busyness and audit committee independence * audit committee busyness. Audit committee independence is negatively significant at the 0,1 level. This finding means that there is a negative and significant relationship between the percentage of independent directors serving on an audit committee and earnings management. The more independent an audit committee is, the more earnings management decreases. Again, these results cannot be compared with prior literature because of the presence of the interaction term in this regression. I dropped the interaction term in the model. Running the regression resulted in t = -1,64, p = .10 (untabulated). This relationship, which I found, is similar to that of Xie et al. (2003) and Klein (2002), who also found a significantly negative relationship between audit committee independence and earnings management. In this research, there is no evidence for a significant relationship between audit committee busyness and earnings management. There is also no evidence for a significant relationship between
27 audit committee independence with the presence of busy independent directors and earnings management.
TABLE 5
Regression results hypothesis 2
Variable Coefficient Std. error t-value p-value
Intercept -0,048 0,011 -4,23 0,000 ***
Board independence -0,010 0,007 -1,42 0,157
Board busyness -0,001 0,006 -0,12 0,906
Board independence * board busyness (H2) 0,007 0,055 0,13 0,894
Assets 0,005 0,001 3,15 0,002 *** Revenues -0,000 0,001 -0,18 0,860 Market value -0,002 0,001 -1,60 0,110 Leverage -0,019 0,011 -4,34 0,000 ***
* Significant at the 0,1 level
** Significant at the 0,05 level
*** Significant at the 0,01 level
From these results, I can conclude that hypothesis 3 cannot be supported. A higher percentage of busy directors serving on an audit committee have no impact on the relationship between audit committee independence and earnings management.
Finally, table 7 reports the regression results for hypothesis 4. Independent variables in this regression are the presence of a financial expert on the board of directors, financial expert busyness and financial expert presence * financial expert busyness. All independent variables are not significantly associated with earnings management. Bédard et al. (2004) and Carcello et al. (2006) found both a negative and significant relationship between the presence of a financial expert serving on an audit committee and earnings management. I could not find this relationship, after dropping the interaction variable out of the regression (t = -0,85, p = .39).
