Sophie van der Zwet
10108610
Board diversity and earnings
management
Master Thesis
Faculty of Economics & Business
Amsterdam Business School
MSc Accountancy & Control – Accountancy track
Academic year 2014-2015
Supervisor: Dr. P. Kroos
Final version 21-06-2015
1
AbstractIn recent years, the attention to board diversity increased. Board diversity relates to the composition of the board by combining attributes, characteristics and expertise of individual board members (Burton, 1991). The firm’s board of directors is the most important internal control device to control and monitor management in order to deter management from opportunistic behavior (Rose, 2007). The board has both an advisory and an oversight function.
An important issue in this context is the use of earnings management. Earnings management influences the financial data available to the stakeholders, it implies that the reported earnings deviates from the underlying economic reality. According to Carter et al (2003) board diversity is critical for the monitoring function.
This study examined the relation between board diversity and earnings management. The research question is: what is the impact of board diversity on earnings management. This study examined three aspects of board diversity, namely gender diversity, age diversity and ethnic diversity. The sample consists of U.S. firms in the period 2008 – 2013. Database research is used to test three hypotheses. One hypothesis is about gender diversity, the second about age diversity and the last about ethnic diversity. A negative beta for all three variables is expected.
After performing different OLS and robust regression analyses, for all independent variables of interest significant coefficients are found, however the results are mixed. For the first proxy of board diversity, gender diversity, only significant results are found with the Modified Jones Model without year- and industry dummies. That is, as expected, a negative relationship between the percentage of women and earnings management. For the second proxy (age diversity) all regressions found the same results, namely a significant positive coefficient. Indicating that the more age diverse a board is, the higher the likelihood of earnings management. For ethnicity there were only found significant positive results in the OLS and robust regressions without year- and industry dummies. So, while the results of the directions are mixed, it can be concluded that board diversity impacts earnings
management.
Statement of Originality
This document is written by student Sophie van der Zwet 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.
2
Table of Contents
1.
Introduction ... 3
1.1
Background ... 3
1.2
Research question ... 4
1.3
Relevance ... 4
1.4
Thesis structure ... 5
2.
Literature review and hypotheses ... 6
2.1.
Board of directors and board diversity ... 6
2.2.
Earnings management ... 8
2.2.1. What is earnings management? ... 8
2.2.2. Incentives and forms of earnings management ... 10
2.3. Effect of board diversity on earnings management ... 11
2.3.1. Gender ... 11
2.3.2. Age ... 13
2.3.3. Ethnicity ... 15
3.
Research design ... 17
3.1 Sample selection ... 17
3.2. Empirical model ... 18
3.3. Variables specification ... 18
3.3.1. Dependent variable (EM) ... 18
3.3.2. Independent variables ... 19
3.3.3. Control variables ... 20
4.
Results ... 22
4.1. Descriptive statistics ... 22
4.2. Main results ... 24
5.
Conclusion ... 29
References ... 31
Appendix ... 33
3
1. Introduction
1.1 Background
Since several large accounting scandals at the beginning of the 21th century, such as WorldCom and Enron, the integrity of the financial reports and corporate governance has become more important. Corporate governance has become a well-discussed and controversial topic in both the popular and business press (Larcker & Tayan, 2011).
Investors want to trust the financial reports and require that the financial statements give a true and fair view. This means that the financial statements are free from material misstatements and faithfully represent the actual financial performance and position of the entity (Gray & Manson, 2011).
Corporate governance is based on the idea that when there is separation between the ownership of a company and its management, self-interested managers have the opportunity to take actions that benefit themselves, with shareholders and stakeholders bearing the cost of these actions. The interest and goals are different among the principal (the stakeholders) and the agent (the management) and it is difficult for the principal to monitor the actions of the agent (Eisenhardt, 1989). This is referred as the agency problem. To reduce the agency costs, some type of control or monitoring system is put in place in the organization. That system of checks and balances
incorporated in a collection of control mechanisms that an organization adopts to prevent or discourage self-interested executives from engaging in activities detrimental to the welfare of shareholders and stakeholders, referred to as corporate governance (Larcker & Tayan, 2011; Gray & Manson, 2011; Eisenhardt, 1989).
The firm’s board of directors is the most important internal control device to control and monitor management in order to deter management from opportunistic behavior (Rose, 2007). The board has both an advisory and an oversight function. The advisory role of the board means that the board offers assistance to the management regarding the strategic and operational direction of the company. The oversight function is exercised through monitoring management, thereby ensuring that the board represent and protect the interest of the company’s shareholders (Larcker & Tayan, 2011).
An important issue in this context is the use of earnings management. Earnings management influences the financial data available to the stakeholders, it implies that the reported earnings deviates from the underlying economic reality. Graham et al. (2005) conduct a survey among
4
financial executives and documents that personal incentives such as bonus schemes and personal reputation and prestige, company reputation and avoidance of debt covenant violation are important drivers of earnings management. They found that managers want to meet or beatearnings benchmarks to build credibility with the capital market, to maintain or increase stock price, to improve the external reputation of the management team and to convey future growth
prospects. The financial executives acknowledge that a failure to hit earnings benchmarks creates uncertainty about a firm’s prospects (Graham et al., 2005; Scott, 2011).
In recent years, the attention to board diversity increased. Board diversity relates to the composition of the board by combining attributes, characteristics and expertise of individual board members (Burton, 1991). Coffey and Wang (1998) defines board diversity as the differences
between the members/directors of a board. These differences include observable differences as age, gender or skin-color, as well as less observable differences such as ethnicity, experience and creed.
According to Erhardt et al. (2003) diversity leads to a greater knowledge base, creativity and innovation, and therefore becomes a competitive advantage. According to Carter et al (2003) board diversity is critical for the monitoring function. Directors with different backgrounds and
demographics have different views than more homogeneous boards and therefore a diverse board has a much broader view which enhances the effectiveness of the board. More diversity will also increase the independence of the board, because the different board members do not share their background, school relationships, kinships or regionalism (Choi & Min, 2012). This study investigates the impact of board diversity on the use of earnings management.
1.2 Research question
On the basis of the aforementioned, my research question can be stated as follows:
What is the impact of board diversity on earnings management?
This thesis will examine the impact of different board of director’s variables on the use earnings management. It will specifically look into the impact of three established proxies of diversity on earnings management, i.e., gender diversity, age diversity and ethnicity.
1.3 Relevance
This study makes several contributions. Firstly, it contributes to the literature on earnings management. There has been an extensive stream of literature that has focused on the
5
whether board diversity could be helpful in constraining earnings management by means ofimproved monitoring by the board of directors. So, my study contributes by specifically looking at the potential impact of board diversity on earnings management.
In addition, there has been some literature on the consequences of board diversity. Several studies (for example, Erhardt et al. 2003), examined the effect of board diversity on firm
performance. This thesis contributes to the board diversity literature by examining the potential impact of board diversity on earnings management. Furthermore, most of the time the researchers examines the effect on firm performance or earnings management by only looking at one dimension of board diversity, i.e., the impact of gender diversity (Carter et al., 2003; Srinidhi et al., 2011; Peni & Vähämaa, 2010). Gender diversity on the board has become a much-discussed topic nowadays, especially the discussion about more females on the board of directors. This study extends prior research by looking to a broader view of board diversity and examining more than one variable of board diversity that could impact earnings management.
Furthermore, prior literature on earnings management has focused on the constraining role of corporate governance by looking at common proxies such as the separation of CEO and chairman positions and the independence of outside directors. This study therefore also contributes to the corporate governance literature by examining whether the monitoring role of the board can be reinforced through a greater diversity of board members.
Lastly, this thesis also features societal relevance as it contributes to the ongoing discussion about more diversity on the board of directors and its possible positive effect on performance and reporting quality. This could be useful for firms when they are composing their boards, as well as for investors, because they can use it in their investment decision to estimate the likelihood for earnings management. For audit/consultancy firms this study could also be useful, because they can take in account when the likelihood for earnings management is higher.
1.4 Thesis structure
The remaining part of this paper is structured as follows: in the second chapter the prior literature of earnings management and board diversity will be reviewed, followed by the hypotheses
development. Chapter three discusses the research methodology, including a description of the sample, empirical model and variable measurements. After that, the results will be discussed in chapter four. The last chapter contains the discussion and conclusion.
6
2. Literature review and hypotheses
2.1.
Board of directors and board diversity
A board of directors is a governing body of elected or appointed members who jointly oversee the activities of an organization. The United States has approximately 50,000 directors of large private and publicly traded corporations and the average director stays on a corporate board for seven years (Larcker & Tayan, 2011). The directors are elected by the stockholders of the firm. They are elected as representatives of the stockholders to establish corporate management related policies and to make decisions on shareholders’ behalf. They should act in the interest of both the management and the shareholders. This follows from the separation of ownership and control where executive
managers are employed to run the firm on a daily basis while the shareholders are able to diversify their risk across a range of firms. However, this creates agency problems because both parties have different interest and goals and it is difficult for the shareholders to monitor the actions of the management (Eisenhardt, 1989). The firm’s board is by far the most important internal control device seeking to control and monitor management in order to deter management from opportunistic behavior (Rose, 2007).
The board have several responsibilities. According to the document ‘Principles of Corporate Governance, the Organization for Economic Cooperation and Development (OECD) the responsibilities of the board consist of:
The corporate governance framework should ensure the strategic guidance of the company, the effective monitoring of management by the board, and the board’s accountability to the company and the shareholders.
This means the board have both an advisory and an oversight function (Larcker & Tayan, 2011). Linck et al. (2008) classified the board’s activities also into these two major functions. The advisory role of the board means that the board offers assistance to the management regarding the strategic and operational direction of the company (Larcker & Tayan, 2011). According to Linck et al. (2008), the board’s advising function involves helping management make good decisions about firm strategy and actions. The oversight function is exercised through monitoring management, thereby ensuring that the board represent and protect the interest of the company’s shareholders. As mentioned above, the agency theory can be used to explain the monitoring role of the board of directors.
7
According to Linck et al. (2008), the monitoring role requires directors to scrutinizemanagement to guard against harmful behavior, ranging from shirking to fraud. Linck et al. (2008) found that a firm’s optimal board structure is a function of the costs and benefits of monitoring and advising given the firm’s characteristics. Raheja (2005) conducted a research about the interaction of insiders and outsiders on a corporate board. She found that compared to insiders, outsiders provide more independent monitoring, but are less informed about the firm’s constraints and opportunities. Thus, as the benefits (costs) of monitoring increase, boards will do more (less) monitoring leading to more (fewer) outsiders.
So, the responsibilities of directors are different from those of management. The directors only have a role to advise on the corporate strategy but not to develop the strategy, which is the responsibility of the management of the company. The board have to ensure that the financial statements gives a true and fair view, but they do not prepare the statements (Larcker & Tayan, 2011).
Diversity
Coffey and Wang (1998) defines board diversity as the differences between the members/directors of a board. Sometimes these differences are observable, as age, gender or skin-color. Sometimes these difference are less observable, such as ethnicity, experience and creed. Burton (1991) describes board diversity as the composition of the board by combining attributes, characteristics and expertise of individual board members that contribute to board processes and decision making in a positive way.
Companies might seek diverse directors when they believe diversity of personal perspective contributes to board deliberation or decision making. However, such groups tend to have low representation in the senior ranks of corporations, relative to the general populations. For example only 3.8 percent of the CEO’s of Fortune 500 companies are ethnic minorities (African-American, Latino or Asian), and only 2.4 percent are female (Larcker & Tayan, 2011). However, research by Heidrick and Struggles (2010) found that companies have made significant effort toward recruiting diverse board members. They found that 85 percent of the companies have at least one female director on the board and 78 percent have an ethnic minority director. According to the National Association of Corporate Directors and The Center for Board Leadership (NACD), more than 75 percent believe that ethnic and gender diversity is a critical factor in board recruitment (Larcker & Tayan, 2011).
Diversity leads to a greater knowledge base, creativity and innovation, and therefore becomes a competitive advantage (Erhardt et al., 2003). According to Man & Wongs (2013), the
8
composition of the board is very important because it can restrict opportunistic behavior ofmanagers and therefore lower earnings management. In addition, Rose (2007) reports that more diverse board of directors are able to better reflect the interests of stakeholders, shareholders and society in general.
Erhardt et al. (2003), conducted a study which examines the relationship between diversity on board of directors and a firm’s financial performance. The sample consists of 127 large US companies. They measured financial performance data with return on asset and return on investment. Their analyses indicate that board diversity is positively associated with these financial indicators of firm
performance. Another research which examined the relation between board diversity and firm value is the study of Carter et al. (2003). They defined board diversity as the percentage of women,
Africans, Americans, Asians. Hispanics and other minorities on the board of directors. Based on the data from Fortune 100, they found a significant positive relation between firm value and diversity on the board of directors.
2.2.
Earnings management
2.2.1. What is earnings management?
Separation of ownership and control leads to the situation where executive managers that run the firm on a daily basis are better informed than the owners of the firm. However, for the existence of well-functioning capital markets, investors have to be able to make informed decisions. This information asymmetry makes it difficult for outside capital providers to assess the profitability of the firm’s investment opportunities. Therefore, publicly held firms are obliged to disclose a minimum level of information to corporate outsiders to alleviate the problems associated with information asymmetry. Financial reporting can add value if they enable financial statements to effectively portray differences in firms’ economic positions and performance in a timely and credible manner (Beyer et al., 2010). Financial reporting plays an important role in the before mentioned agency problem. In an agency relationship, one party acts on behalf of another. The interests and goals are different among the principal and agent and it is difficult for the principal to monitor the actions of the agent (Eisenhardt, 1989). The mandated disclosure should to some extent address this moral hazard problem, because on the basis of the information owners are better able to infer whether the managers has taken actions which are in the owners best interest, instead of the managers best interest. Likewise, the disclosure can also address the adverse selection problem
9
because now owners are better able to evaluate the abilities of the incumbent manager and to evaluate the current viability and future prospects of the firm. On basis of this information, investors could decide to dismiss the current manager or to increase or decrease the level of ownership in the respective firm (Beyer et al., 2010). On top of the mandatory information to be disclosed to the capital markets, firms can also decide to voluntary disclose additional information (e.g.,management forecasts). An important theory in this context is the signaling theory. This theory addresses the aforementioned problems of information asymmetry in capital markets. The signaling theory says that one party, the management, send signals about the organization and their
performance to another party, the financial market. This could lead to creditable signals for stakeholders (Morris, 1987).
The disclosed information is for an important part based on accrual accounting. A role of accounting accruals is to provide a measure of short-term performance that more closely reflect expected cash flows than do realized cash flows. This will lead to a more informative measure of economic performance. Dechow (1993) investigated circumstances under which accruals are predicted to improve earnings’ ability to measure firm performance, as reflected in stock returns. She, indeed, found that accruals come up with a more informative measure about the economic performance. However, accrual accounting also implies that estimates and discretion is required. This can be used by corporate managers to achieve some personal gain. So, this means that managers can use their knowledge about the company and the business and its opportunities to select reporting methods, estimates, and disclosures that match the firms’ business economics, potentially increasing the value of accounting as a form of communication. However, managers’ use of judgements could also create opportunities for “earnings management”, in which managers choose reporting methods and estimates that do not accurately reflect their firms’ underlying economic performance (Healy & Wahlen, 1999).
According to Healy and Wahlen (1999) the most used definition of earnings management is:
“Earnings management occurs when managers use judgment in financial reporting and in structuring transactions to alter financial reports to either mislead some stakeholders about the underlying economic performance of the company or to influence contractual outcomes that depend on reported accounting numbers.”
The auditor also plays an important role in this context. The audit provides independent assurance to the users of the financial statements that the financial statements presented a “true and fair” view of the company’s underlying financial performance and position. Prior research suggested that
10
higher quality auditors reduce the level of accrual earnings management (Chi et al., 2011). For example, Becker et al. (1998) examined the effect of audit quality on earnings management. The result of their research indicates that higher audit quality is associated with less accounting flexibility and this implies less earnings management.2.2.2. Incentives and forms of earnings management
There are different incentives for earnings management. First, earnings are managed to meet or beat earnings benchmarks (Graham et al., 2005). Investors often use the forecasts of analysts to device whether to invest or not. Managers will therefore have an incentive to manage earnings upwards in order to meet or beat analysts’ forecasts, thereby enjoying a share price increase. Graham et al. (2005) conduct a survey among financial executives and found that managers want to meet or beat earnings benchmarks to build credibility with the capital market, to maintain or increase stock price, to improve the external reputation of the management team and to convey future growth prospects. Failure to hit earnings benchmarks creates uncertainty about a firm’s prospects.
Sometimes earnings are managed to minimize income. This happens when there is no benefit for profits in excess of a certain level. When a managers’ bonus have an upper limit, income will be minimized up to the upper limit. A company in financial trouble can manage earnings downward in a particular year to make an even bigger loss (it has little to lose at this point) and thereby already incorporating future losses, which results in lower expenses in the future. This earnings management technique is known as big bath accounting. Income can also be minimized for tax purposes (Scott, 2011).
Earnings are also managed because of managers’ personal reasons, for example to increase their compensation and reputation. When the company use bonuses in a managers’ compensation package, the managers could make discretionary accrual decisions and use accounting judgement to maximize their short-term bonuses. This income maximization form of earnings management is named the bonus plan hypothesis (Scott, 2011).
Another incentive to manage earnings is to avoid the violation of debt contracts. This is named the debt covenant hypothesis. Debt covenants are made with the intention to restrict managers from making investment and financing decisions that reduce the value of debtholder claims. This covenants are frequently written in terms of accounting numbers. Violating debt covenants are costly, therefore managers that are close to violating debt covenants make
11
accounting choices that reduce the likelihood of default (Watts & Zimmerman, 1986). DeFond and Jiambalvo (1994) found that firms in the year prior to the violation, have abnormal total and working capital accruals that are significantly positive. This result shows that firms use earnings management to avoid debt contract violation.There are two types of earnings management, namely accrual-based earnings management and real earnings management. Accrual-based earnings management refers to accounting choices that managers make with the intention of managers to deliberately bias accounting information and real earnings management refers to actuals decisions that managers make with the intention of manager to deliberately influence accounting information. Accruals do not affect cash flows and are largely subject to the discretion of managers. Real earnings management decisions are decisions that adversely affect firm value in the long run (e.g., lowering investments to boost current profit numbers), because it has negative consequences on future cash flows (Chi et al., 2011). This thesis will focus on accrual earnings management.
2.3. Effect of board diversity on earnings management
In this subchapter the different proxies for board diversity are explained. In this study there are three variables used to examine the impact of board diversity on the use of earnings management, i.e., gender, age, and ethnicity.
2.3.1. Gender
In recent years there has been increasing attention to gender diversity, it has become a well-discussed topic in both the popular and business press. In 2008 there were in the United States, 16 percent women of all directors on the board and 89 percent of the boards had at least one woman director in their board. In 2003 there were only 13 percent woman, this means that the share of female directors on the board is rising (Larcker & Tayan, 2011). However, women are still significantly under-represented on boards of directors relative to the overall population.
In recent years, several countries attempts to increase the female representation. For example, in Norway they implemented a law requiring that approximately 40 percent of the directors of publicly traded companies has to be female. Other countries in Europe have or are considering similar laws. France recently passed similar legislation. In the Netherlands since 2013 it is legally established that efforts should be made to have at least 30 percent female directors. Spain enacted a 40 percent female requirement starting in 2015 (Larcker & Tayan, 2011). In the United
12
States they did not yet enact similar legislations on female quota, but the voluntary femalerepresentation on corporate boards is also rising (Srinidhi et al., 2011; Larcker & Tayan, 2011). Ahern and Dittmar (2010) researched the potential negative impact of an arbitrary female quota using Norwegian data. Female quota could be a risk when companies, in order to meet the (mandatory) arbitrary quotas, recruit underqualified directors. This is caused by recruitment efforts that are not focused on getting the best person for the job, but getting the best candidate that also satisfies a demographic profile. This referred to as tokenism (Larcker & Tayan, 2011). Ahern and Dittmar (2010) found that this law led to considerable changes in board compositions, not only in gender, but also in age, education and experience, which led to a significant decrease in firm value. This was not only a result of more female, but also to the inexperience of new directors. But, also for non-quota settings, such as in the U.S., a greater pressure to increase gender diversity on the board could translate into selecting less qualified directors, and subsequently lower firm value.
However, prior literature also suggests that increased gender diversity can be beneficial as it may increase the performance of companies and add value. For example, women might evaluate information and consider risk and reward differently than men, this will lead to enhanced decision making. Furthermore, women on board can lead to more independence, because it reduces social similarities that lead to premature consensus (Carter, 2003; Adams & Ferreira, 2009). Adams et al. (2010) also argued that female directors exhibit more independent thinking and improve the
monitoring process. Female board representation might also lead to better board dynamics, because women might exhibit higher levels of trustworthiness and cooperation than men (Larcker & Tayan, 2011). In addition, the research of Adams and Ferreira (2009) showed that female directors exhibit greater diligence in monitoring. Furthermore, Adams et al. (2010) found that investors value the addition of female directors to the board. Together, better monitoring and lower information asymmetry leads to better earnings quality (Srinidhi et al., 2011).
Most of the studies found evidence that supports the idea that female representation can improve governance quality. For example Peni and Vähämaa (2010) examined the association between earnings management and the gender of firm’s executives. They focus on the gender of the CFO and CEO. They used a sample of S&P 500 firms with 1955 firm-year observations. They used discretionary earnings to measure earnings management. The result of this research was that firms with a female CFO are associated with income decreasing discretionary accruals, this implies that female CFO’s are using more conservative earnings management strategies. This is in line with other studies which also found that women are more conservative and less risk-taking, which will result in lower earnings management (Rose, 2007; Adams & Ferreira, 2009; Srinidhi, 2011).
13
In the research of Srinidhi et al. (2011) they examined whether U.S. corporations withgender-diverse boards exhibit higher earnings quality, as a tangible consequence of the higher level of monitoring. According to Srinidhi et al. (2011) and Clarke (2005) gender-diverse boards have more informed deliberations and discuss tougher issues that are often considered unpalatable by all-male boards. Female board participation also promotes more effective board commination to investors. Taken this together, they expect that this more informed board discussions and better
communication improve the board’s monitoring ability (Srinidhi et al, 2011). They measured earnings quality by discretionary accruals. In conjunction with other studies, such as Adams and Ferreira (2009) and Rose (2007) they found that female board participation increases earnings quality by improving the oversight function of the board. The overall implication of their research was that when greater oversight is desired and better earnings quality is demanded by investors, inclusion of female directors on the board is a plausible way to achieve these objectives.
Based on the above discussed literature, the first hypothesis will be:
H1: There is a negative relationship between gender diversity and the use of earnings management.
2.3.2. Age
The age range of directors is also an important indicator of board diversity. Traditionally, most board members are middle to retirement aged members. For example, the average age of directors in the U.S. is 61 years. Approximately 75 percent of boards have a mandatory retirement age, which is usually 70 years or older (Larcker & Tayan, 2011). Most boards require some spread in age and age diversity is becoming more important (Li et al., 2011). Directors who differ in age have different qualities, insights, perspectives of the world and are a better representation of the society. Mclntyre et al. (2007) researched the effect of age diversity on firm performance. They found that age
diversity is positively related with firm performance.
Houle (1990) distinguished boards into three groups. He found that the youngest board members are more innovative and has the energy and drive to succeed and they are prepared for their management position in order to ensure the future of the company, while the eldest part of the board has the necessary experience and connections, whereas the middle-aged part focuses on their responsibility in the company and society. The eldest group can provide experience, wisdom and a strong network to the youngest group members. Furthermore, Houle (1990) argues that boards can be more efficient when there is more age-diversity. The members are from different generations and this will provide a set of unique resources to a board. By an age-diverse board
14
composition, they can benefit from each other’s strengths (Eulerich et al., 2014). Because of the different generations, for example not sharing the same childhood, independence could also increase, leading to better monitoring quality (Larcker & Tayan, 2011).Davidson et al. (2007) found that firms with an older CEO, who is near retirement, have larger discretionary accruals in the year prior turnover. They used 1017 CEO turnovers of companies listed on the S&P 1500 in their sample. To measure discretionary accruals, they used the Modified-Jones model. They found that older CEOs nearing retirement age are associated with greater income increasing earnings in the year’s prior retirement. According to Davidson et al. (2007) this is caused by the time horizon of the executives; older executives near retirement may be more concerned with current or short-term performance, rather than long-run performance. Another incentive for earnings management by an older CEO is because he or she do care less about his further career perspectives, because he is soon going to retire. So, he is willing to take more risk to maximize his compensation, because there are less consequences of dismissal. Reitenga and Tearney (2003) also found an increase in discretionary accruals in the years prior to the CEO’s retirement. This can be countered by more age diversity, namely by combining older board members with younger board members, who can prevent that the board only focuses on the short term by using earnings
management strategies.
In summary, an age-diverse board ensures that the different perspectives from the different age groups have an inhibiting effect on the likelihood that one perspective is becoming too
important and will take the overhand. For example the short term perspective of elder board members, caused by their limited employment horizon, found by the above discussed studies of Davidson et al. (2007) and Reitenga and Tearny (2003). This effect is mitigated with a more age-diverse board for two reasons. Firstly, then there are less old (near retirement) board members and, on the other side, when the board is more age-diverse, there are also other age groups which more focuses on the long-term, which will result in a mitigating effect. As mentioned before, Houle (1999) found that board members in the middle-aged part focuses on their responsibility in the company and society and the long-term, this perspectives could mitigate the incentives of other/elder board members to manage earnings upwards in order to achieve short-term goals. Thus in this way they could benefit from each other strength’s and younger directors can prevent the older (near
retirement aged directors) possible use of earnings management strategies, while learning from the experience and wisdom of the older members. So, I will expect that a more age-diverse board will lower the likelihood of earnings management.
15
Based on the abovementioned literature, this will lead to the following hypothesis:H2: There is a negative relationship between age diversity and the use of earnings management.
2.3.3. Ethnicity
Ethnicity can be defined as a group of people who regard themselves to be different from others. The ethnic groups are united by common traditional, cultural, linguistic, ritualistic, behavioral and religious traits. According to the National Association of Corporate Directors and The Center for Board Leadership (NACD), more than 75 percent believe that ethnic diversity is a critical factor in board recruitment (Larcker & Tayan, 2011). However, now only 3.8 percent of the CEO’s of Fortune 500 companies are ethnic minorities (African-American, Latino or Asian). Only a quarter of the directors in the U.S. have international work experience and only 7 percent are of foreign birth (Larcker & Tayan, 2011).
According to Rahman and Ali (2006), cultural and ethnic factors are important as the traditions of a nation are instilled in its people and might help explain why things are as they are. Which could also affect the likelihood of earnings management.
According to Larcker and Tayan (2011) companies might seek directors of diverse ethnic origin when they believe diversity of personal perspective contributes to board deliberation or decision making. It could improve decision making by ensuring that the board has the full array of knowledge in terms of market dynamics. According to Nielsen and Nielsen (2013) it is especially important for international firms to have different ethnicities and nationalities in the board, because this provides a wider range of culturally sensitive knowledge and experience and better reflect the society. Another incentive to increase ethnic diversity is that a diverse board composition more closely reflects the diversity of the broader U.S. population (Larcker & Tayan, 2011).
Erhardt et al. (2003) and Carter et al. (2003) examined the relationship between diversity on boards of directors with firm financial performance. They define board diversity, amongst other, as the percentage of Africans, Americans, Asians, Hispanics and other minorities on the board of directors. They both found a significant positive relation between firm value and ethnic board diversity.
Furthermore, agency theory supports the idea that to increase the independence of the board, boards should include outside directors. These directors are needed to monitor and control (Larcker & Tayan, 2011; Jensen & Meckling, 1976). More ethnic diversity will increase the
16
school relationships, kinships or regionalism. Furthermore, Oxelheim and Randoy (2003) argued that foreign board members can accomplish active monitoring because they are perceived as ‘outsiders’. This will increase the likelihood that the board acts on the behalf of the shareholders rather than on behalf of the managers or other board members. This increases the monitoring quality of the board, which will decrease the likelihood of earnings management (Choi & Min, 2012).Han et al. (2010) examined whether the degree to which managers exercise earnings discretion relates to their value system (i.e., culture). The found an association between culture and earnings management and they found that this association varies with the strength of investor protection. Furthermore they found that the culture values of individualism and uncertainty avoidance plays an important role in explaining earnings discretion.
I expect that more ethnic diversity on the board leads to improved board independence due to a smaller overlap in background, relationships etc. among board members, as well as a greater diversity in values (e.g., individualism vs. collectivism). This may result in better monitoring, which decreases the likelihood of using earnings management. This lead to the following hypothesis:
17
3. Research design
3.1 Sample selection
In order to examine the impact of board diversity on earnings management I will perform a quantitative archival study (database research). The data is collected from Wharton Research Data Services (WRDS), from the “ISS” (formerly RiskMetrics) database and from the “Compustat
Fundamentals Annual” database. The “ISS” database (formerly RiskMetrics) is used for the governance data, which I needed for my independent variables, this includes data about: gender, age and ethnicity. The initial “ISS” sample consists of all U.S. firms in the period 2008 – 2013. This period is used because this was the maximum time period with available data in the “ISS” database. This initial sample yielded 5760 observations.
In addition, the data required to measure my dependent variable (earnings management, measured by discretionary accruals) is extracted from the “Compustat” database.
Following prior research, firms from the financial services industry (i.e. banks and insurance companies) have been excluded (SIC codes 6000 - 6999) since their firm characteristics and
regulations are different, which are difficult to compare. Specifically, accruals of these firms are likely to be different from accruals of firms in other industries. Following Roychowdhury (2006) the sample will also be restricted by eliminating firms in utility industries (SIC codes 4900-4999). Firms with insufficient financial data or governance data are also excluded. After merging “ISS” with “Compustat” and dropping the missing values (now and later in the regression process) and the aforementioned SIC codes, it leaves us with a final sample of 3886 firm-year observations from 826 unique firms.
Table 1: Sample selection
Number of observations
Sample merged ISS & Compustat matched
5760 -Financial services (SIC
6000-6999) deleted
466 - Utility firms (SIC 4900-4999) 299
- Missing values 1109
18
3.2. Empirical model
The following empirical model is used to conduct a regression analyses to measure the effect of board diversity on earnings management:
EM = β
0+ β
1GEN + β
2AGE + B
3ETH + β
4LNSIZE + β
5LEV + β
6ROA + β
7BIG4dum
+ β
8LOSSdum + ε
EM proxies for accrual-based earnings management. GEN, AGE and ETH denote the degree of gender diversity, age diversity and ethnic diversity, respectively. The other variables are control variables. In the next paragraph the variables will be defined in more detail. In table 2, at the end of this chapter, you can find a summary of the description and measurement of the dependent, independent and control variables.
The coefficients of β1, β2 and β3 are needed to test the impact of respectively gender diversity, age diversity and ethnic diversity on earnings management. On the basis of my hypotheses I expect β1, β2 and β3 <0.
3.3. Variables specification
3.3.1. Dependent variable (EM)
The dependent variable in this study is earnings management. This is measured by the use of discretionary accruals. Discretionary accruals are those accruals over which managers can exercise some control through accounting discretion and estimates, which can be used to engage in earnings management to achieve personal gain (Healy & Wahlen, 1999; Scott, 2011).
There are different models to calculate discretionary accruals. Dechow et al. (1995)
compared and discussed different models and concluded that the Modified-Jones model is the best model for detecting earnings management1. Therefore the Modified-Jones model will be used in this
study.
1 Dechow et al. (1995) compared and discussed five models to measure accruals, namely the Healy model
(1985), DeAngelo model (1986), The Industry Model (1991), Jones Model (1991) and the Modified Jones Model (1995).
19
To measure Total Accruals (TAit), data is retrieved from Compustat and it is calculated by deducting cash flow from operating activities for firm i in year t (in Compustat: OANCF) from income before extraordinary items (in Compustat: IBit), as done for example in the research of Cahan et al (2011): TAit = IBit – OANCFitTo measure the discretionary accruals, the following formula of Modified Jones Model is used:
Where,
TAit = Total accruals for firm i in year t (Compustat: IBit –OANCFit) Ait-1 = Total assets in the previous year, lagged assets (Compustat: ATit)
Δ REVit= Change in revenue of firm i between years t and t-1. (Compustat: SALEt – SALEit-1 ) Δ RECit= Change in receivables of firm i between years t and t-1 (Compustat: RECTt – RECT t-1) PPEit = The level of property, plant and equipment of firm i in year t (Compusat: PPEGTt) εit = Error term, proxy for the level of unsigned discretionary accruals of firm i in year t.
After calculating the total accruals and performing the OLS regression for every two digit SIC group the coefficients α1, α2 and α3 are estimated. The estimation error (εt) is used as proxy for the level of firm-specific discretionary accruals as part of the total accruals. I will use the unsigned values of discretionary accruals, because this study looks to the amount of earnings management, rather than the direction of earnings management.
3.3.2. Independent variables
In this study there are three explanatory variables used to measure the effect of board diversity on earnings management. The first one is gender (GEN), gender is measured as the percentage of women in the board. This is in line with the research of Carter et al. (2003). It will be calculated as the number of woman board directors divided by the total number of board members. This data is found in the “ISS” database.
20
Second, the variable age is used (AGE). This study looks into diversity, so I will use agediversity. This will be measured by the standard deviation of the age of the board members of a respective firm (Murray, 1989).
Finally, ethnicity (ETH) is used as independent variable of interest2. Ethnicity variable is
found in the “ISS” database. The percentage of foreigners in the board is used, a foreign ethnicity is defined as another ethnicity than the Caucasian race. Most of the directors of U.S. firms which I found in the “ISS” database are of the Caucasian ethnicity. This will be measured by percentage non-Caucasian directors on the board. This is calculated by counting all other ethnicities than non-Caucasian and this number dividing by the total number of the board members. This is in line with the research of Rose (2007).
3.3.3. Control variables
In this research, I used several control variables to control for other factors which otherwise bias the regression coefficients of interest.
Firstly, I included control variable to control for the impact of company size (LNSIZE) on earnings management. I controlled for this effect because prior studies have shown that larger firms tend to have lower levels of accruals than smaller firms (e.g. Francis & Wang, 2008). Watts and Zimmerman (1990) found that large firms have a negative relationship with earnings management. Erhardt et al. (2003), Adam and Ferreira (2009) and Peni and Vähämaa (2010) argued that larger firms have lower information asymmetry and monitoring by auditors is high. Firm size will be measured by the natural logarithm of total assets.
Secondly, leverage (LEV) is also a commonly used control variable (e.g. Francis & Wang, 2008). This variable is also included as control variable in my study, because a high leveraged firm increases the likelihood of earnings management by avoiding violation of a debt covenant and it increases the likelihood of bankruptcy. Both of this will lead to increased likelihood earnings management (DeFond & Jiambalvo, 1994). Leverage will be measured as total liabilities divided by total assets.
Another control variable which is used is return on assets (ROA). This is calculated by net income divided by assets. This is a commonly used control variable, it is an important benchmark to measure the performance of a company and it gives a benchmark of how efficient the board used
2 This study focuses on ethnicity, because this variable is available in the databases of U.S. firms. Nationality is
not available and the only way to use this is by hand-collecting this data, this is very time-consuming when using U.S. firms, so for this reason this study examines the potential impact of ethnicity on earnings management.
21
the active to generate earnings. With this control variable we control for differences in performance (Dechow et al., 1995; Kothari et al., 2005).As fourth control variable a Big 4 dummy variable is included to control for audit quality. Research found that there is a relation between the likelihood of earnings management and being audited by a Big 4 or non-Big 4 audit firm. Michas (2011) found for example that total and
discretionary accruals are both lower at Big 4 client companies in comparison with non-Big 4 companies. Lower discretionary accruals are a signal of higher audit quality because auditors will better monitor to constrain the opportunities for earnings management by the managers (Becker, 1998).
Another control variable which is used is ‘LOSS’, this is a dummy variable for firms with prior year losses. According to Francis and Wang (2008) there is a positive relation between a prior year loss and income increasing accruals. Companies with a loss in prior year tend to use more
discretionary accruals to have better financial numbers and to prevent losses in this year.
Lastly dummies for year- and industry effects will be used to improve the robustness of the findings. Year dummies (DUM_YEAR) are included because years could be related with the growth of the economy. The industry dummies (DUM_INDUS) are included because different industries have different risks and economic conditions.
Table 2: Summary of variable definitions
Variables Description
Dependent
EM Accrual-based earnings management, measured by Modified Jones Model Independent
GEN Gender, measured by percentage of female directors on the board
AGESD Age diversity, measured by the standard deviation of the age of the board members ETH Ethnicity, measured by percentage non-Caucasian directors on the board
Control
LNSIZE Firm size, measured by natural log of total assets, LnAssets LEV Leverage, measured by total liabilities divided by total assets ROA Return on assets, net income divided by assets
BIG4dum Firms with an auditor from Big 4 audit firms, dummy variable equal to 1 for those firms audited by a Big 4 audit firm, zero otherwise
LOSSdum Dummy variable for firms with prior-year losses equal to 1 when there was a loss in previous year, zero otherwise
DUM_YEAR Year dummies
22
4. Results
4.1. Descriptive statistics
In table 3 you can find a descriptive summary of the variables used in the empirical models. For discretionary accruals, as aforementioned, the absolute value is used as a proxy for EM. The table shows that the mean is 0.034. This value is in line with prior research.
The mean of the percentage of woman on the board is 12.3 percent. As mentioned before, there is increasing attention for females in the board of directors. So, I expect an increasing percentage. I also compute the percentage women on the board over the sample years (see appendix table 1) that shows an increase from 11.6 percent in 2008 to 13.6 percent in 2013. The standard deviation of the age is 7.52. And on average 22.3 percent of the board members is from another ethnicity than Caucasian.
Furthermore the table shows that 93 percent of my sample firms is audited by a Big 4 audit firm. This was expected because the sample consists of listed U.S firms which have most of the time Big 4 auditors. The mean of the leverage is 50.2 percent, which indicates that the firms are equally financed with equity and debt. Furthermoe, about 15 percent of the firms reports a loss.
Table 3: descriptive statistics of the variables
Variables N Mean p25 p50 p75 SD EM 3886 .034 .014 .029 .054 .024 GEN 3886 .123 .000 .111 .200 .102 AGESD 3886 7.520 5.975 7.274 8.899 2.266 ETH 3886 .223 .000 .111 .300 .274 LNSIZE 3886 7.814 6.657 7.644 8.794 1.537 LEV 3886 .502 .350 .500 .631 .218 ROA 3886 .051 .025 .058 .095 .096 BIG4dum 3886 .930 1.000 1.000 1.000 .255 LOSSdum 3886 .148 .000 .000 .000 .356
In table 4 you can find the sample composition across industries, based on two-digit SIC codes. The largest part of the sample, almost 55 percent, consists of the “Manufacturing industry” (SIC codes: 20– 39). Inside this industry, the “Chemicals & allied products” (SIC: 28), the “Industrial &
23
equipment” (SIC: 36) and the “Instruments & miscellaneous manufacturing” (SIC: 38-39) are the largest (not tabulated). Furthermore, the Services industry makes up a large part of the sample, approximately 17 percent. The largest part of this industry group is the “Business services” (SIC: 73). The “Retail Trade industry” (SIC codes: 52-59) also makes up a considerably part of the sample. The composition in this study is quite consistent with the distribution of firms over industries in prior research. As aforementioned, to address for potential differences between industries, in my multivariate analyses I control for industry effects.Table 4: Sample composition by industry
2SIC Industry Observations Percentage
Mining (SIC 10-14) 220 5.66%
Construction (SIC 15-17) 90 2.32%
Manufacturing (SIC 20-39) 2089 53.76%
Transportation, Communication & Public Utilitites (SIC 40-49) 227 5.84%
Wholesale Trade (SIC 50-51) 182 4.68%
Retail Trade (SIC 52-59) 416 10.71%
Finance, Insurance, Real Estate (SIC 60-67) dropped 0 -
Services (SIC 70-89) 662 17.04%
Total 3886 100.00%
Table 5 reports a Pearson correlation matrix with correlation coefficients between the dependent, independent and control variables. The correlations between accrual-based earnings management and my three proxies for board diversity are all significant at a 5 percent significance level. As expected, a negative correlation is found between gender and EM. The other two proxies are positive related with earnings management. Moreover, it seems that firms with greater gender diversity have lower age and ethnic diversity. Furthermore, as expected, bigger firms (measured by LNSIZE) are negatively correlated with earnings management (p<0.01) and have greater gender diversity, but lower age and ethnic diversity. As also expected, a significant positive correlation coefficient is found for LOSS with EM (p<0.01) indicating that when a firm have a loss in prior year, the likelihood of earnings management is higher. Lastly, being audited by a big 4 auditor is
correlated with less earnings management (p<0.01).
In most cases the correlation coefficients are below 0.50 and above -0.50, which indicates that there are no strong correlations between the variables and thus no severe multicollinearity in the data. The only exception which coefficient is lower than -0.5 is the correlation between LOSS and ROA, which is -0.675 at a significance level of 1 percent. This is caused by the fact that in both variables net income plays a role.
24
Table 5: Pearson and Spearman correlation matrixVariables EM GEN AGESD ETH LNSIZE LEV LOSSdum BIG4dum ROA
EM 1.000 GEN -0.065 1.000 AGESD 0.074 -0.140 1.000 ETH 0.064 -0.078 -0.004 1.000 LNSIZE -0.126 0.284 -0.176 -0.092 1.000 LEV -0.007 0.278 -0.141 -0.028 0.395 1.000 LOSSdum 0.222 -0.035 0.037 0.061 -0.153 0.046 1.000 BIG4dum -0.042 0.190 -0.083 -0.026 0.271 0.200 -0.033 1.000 ROA -0.202 0.045 -0.014 -0.052 0.101 -0.115 -0.675 0.006 1.000 Bold = Significant at 5% level (p<0.05)
4.2. Main results
Table 6 reports the results for my multivariate analyses. Both OLS regressions and robust regressions are performed, subsequently excluding and including year- and industry dummies as control
variable. The variables are winzorized to control for outliers, which may affect the results of the regressions. In table 6 the left three columns report the OLS results where I only include the control variables, the control variables and the year effects and lastly the control variables and year- and industry effects respectively. The right three columns do the same for my robust regression results. These results are less sensitive to outliers as it eliminates those observations that have a very strong impact on the final regression results as well as downwards weights some observations in producing the final regression output.
With respect to my independent variables of interest, all of my proxies for board diversity (GEN, AGESD and ETH) are significant when I exclude the year- and industry effects, both in the OLS and in the robust regression. With reference to the research question, this means that board diversity indeed has an impact on the use of earnings management. More specifically, gender diversity is negatively and significantly associated with accrual-based earnings management (p<0.1) in both the OLS and robust regression. This is in line with my first hypothesis. However, in the regressions with year dummies or both year and industry dummies, no significant GEN coefficients are found for both the OLS and robust regressions. So, with this regressions this will lead to the rejection of my first hypothesis.
25
With respect to my second proxy for board diversity, the standard deviation of age (AGESD), a significant positive coefficient is found in all regression models (both the OLS regression and the robust regression, with and without year and industry dummies). This means that more age diversity in the board is associated with more accrual-based earnings management. However, this is not in line with my expectation in the second hypothesis, where I expected a negative relationship between age diversity and earnings management. So, my second hypothesis is rejected. A potential explanation for the found positive coefficient is that more age diversity results in a bigger gap between the elder members and younger members and their perspectives and manner of thinking and working could differ, leading to a less efficient decision-making process.For my third proxy of board diversity, i.e. ethnicity, I find mixed results. That is, only significant results are found in the OLS and robust regressions without year and industry control dummies (at p<0.01). However, a positive association between age diversity and earnings
management is found, which was not expected in my third hypothesis. This could be caused through the fact that it is more difficult to come to an agreement, which will lead to efficient decision-making, with different ethnic board members with different backgrounds, cultures and norms and values. In all other regressions models, the coefficient on ethnic diversity is not significant. So, H3 can be rejected.
With regard to the control variables, three significant control variables are found. Firstly, the results in both the OLS and robust regressions shows a negative association between LNSIZE and accrual-based earnings management (significance level p<0.01). Suggesting that a bigger firm is associated with less earnings management. Furthermore, a significant negative association, at p<0.01 in all regressions, is found for ROA and earnings management, meaning the higher the ROA, the less likely earnings management. These findings are confirmed when I look at the dummy for loss-making firms. That is, a significant negative coefficient is found for the loss dummy (LOSSdum), in both the OLS and robust regression, with and without year and industry controls. This is consistent with the intuition that when there was a prior year loss, the likelihood of earnings management is higher.
Finally, the regression models as a whole are all significant (for example, F=35.63; p<0.01 and F=36.22; p<0.01 for the OLS and robust regressions without year and industry effects).
26
Table 7: OLS and robust regression results, with EM based on Modified Jones ModelOLS Robust GEN -.007* (0.090) -.006 (0.133) -.003 (0.534) -.007* (0.099) -.006 (0.166) -.002 (0.618) AGESD .001*** (0.001) .001*** (0.001) .000*** (0.008) .001*** (0.002) .001*** (0.001) .000** (0.011) ETH .004*** (0.007) .001 (0.555) .000 (0.943) .004*** (0.006) .001 (0.630) .000 (0.864) LNSIZE -.001*** (0.000) -.001*** (0.000) -.001*** (0.000) -.001*** (0.000) -.001*** (0.000) -.001*** (0.000) LEV .003 (0.121) .002 (0.202) .002 (0.435) .003 (0.112) .003 (0.207) .002 (0.408) ROA -.0222*** (0.000) -.020*** (0.000) -.020*** (0.000) -.024*** (0.000) -0.021*** (0.000) -.021*** (0.000) BIG4dum -.001 (0.570) -.001 (0.434) -.001 (0.334) -.001 (0.467) -.002 (0.321) -.002 (0.271) LOSSdum .009*** (0.000) .009*** (0.000) .009*** (0.000) .010*** (0.000) .010*** (0.000) .009*** (0.000) Year dummies Industry dummies No No Yes No Yes Yes No No Yes No Yes Yes Obs. F-value 3886 35.63 3886 26.28 3886 7.91 3886 36.22 3886 26.93 3886 7.81 *,**,*** corresponds with 10%, 5% and 1% significance level (two-tailed).
Additional analysis
Kothari (2005) found in his research that it is important to include also the differences in performance, based on ROA, in the Modified Jones Model. In order to control for the impact of performance on accruals, as an additional test I also performed a regression of the performance-adjusted Modified Jones Model and compare this results with the normal Modified Jones model. I repeat the regressions, OLS as well as robust regressions with including and excluding year and industry effects, but now with earnings management as the dependent variable computed on the basis of the performance adjusted Modified Jones Model. In table 8 the results of this regressions are summarized.
However, this additional analyses does not really affect my main inferences, because it does not differ very much with the main model regressions. The only difference is the first variable AGE, where a significant negative coefficient (only at 10 percent significance level) is found in the main regressions without year and industry dummies, while now no significant coefficients at all are found. Suggesting that there is no relationship between gender and earnings management. Further,
27
as also found in the above main model regressions, a significantly positively association between age diversity and accrual-based earnings management is found, at a 5 percent significance level. The third independent variable, ETH, is only significant in the OLS and robust analysis in the regressions without year and industry dummy effects. This is also the same result as found in the mainregressions.
Finally, the regression models as a whole are significant (for example, F=35.45; p<0.01 and F=34.60; p<0.01 for the regressions without year and industry effects).
Table 8: OLS and robust regression results, with EM based on Performance-adjusted Modified Jones Model OLS Robust GEN -.003 (0.539) -.002 (0.669) .001 (0.846) -.002 (0.646) -.001 (0.769) .001 (0.800) AGESD .001*** (0.000) .001*** (0.000) .000*** (0.001) .001*** (0.000) .001*** (0.000) .001*** (0.002) ETH .002* (0.068) -.001 (0.514) .002 (0.221) .003* (0.057) -.001 (0.563) -.002 (0.298) LNSIZE -.002*** (0.000) -.002*** (0.000) -.002*** (0.000) -.002*** (0.000) -.002*** (0.000) -.002*** (0.000) LEV .005** (0.016) .004** (0.042) .003 (0.169) .005** (0.026) .004* (0.064) .003 (0.211) ROA .079*** (0.000) .006*** (0.000) -.027*** (0.000) -.034*** (0.000) -0.031*** (0.000) -.032*** (0.000) BIG4dum -.001 (0.581) -.001 (0.414) -.002 (0.347) -.001 (0.429) -.002 (0.283) -.002 (0.255) LOSSdum .008*** (0.000) .008*** (0.000) .007*** (0.000) .008*** (0.000) .008*** (0.000) .007*** (0.000) Year dummies Industry dummies No No Yes No Yes Yes No No Yes No Yes Yes Obs. F-value 3886 33.45 3886 24.99 3886 7.77 3886 34.60 3886 25.99 3886 7.70
*,**,*** corresponds with 10%, 5% and 1% significance level (two-tailed).
In summary the results which were found are mixed. In the regressions without year and industry dummies based on the Modified Jones Model a significant negative gender diversity coefficient is found. However, when I based my regressions on the recommended performance-adjusted Modified Jones Model by Kothari (2005), no significant results are found for my gender proxy for board diversity. For the proxy age all results of the OLS and robust regressions (also in the additional analysis), with and without year and industry effect, are in line with each other and shows the same significant positive coefficients. For ethnicity there were only found significant results in the OLS and
28
robust regressions without year and industry dummies. This also applies to my additional analysis of the Performance-adjusted Modified Jones Model.29
5. Conclusion
This study examined the relation between board diversity and earnings management. So, the research question was: “What is the impact of board diversity on earnings management?” This study used three proxies for board diversity, namely gender diversity, age diversity and ethnic diversity.
The board has two important functions, namely both an advisory and a monitoring function (Larcker & Tayan, 2011). According to Carter et al. (2003), board diversity plays a critical in the board’s monitoring function. The firm’s board of directors is namely the most important internal control device to control and monitor management in order to deter management from
opportunistic behavior (Rose, 2007). An example for this opportunistic behavior is earnings management, which is examined in this study. Earnings management influences the financial data available to the stakeholders, it implies that the reported earnings deviates from the underlying economic reality.
To test this relationship between board diversity and the use of earnings management I performed quantitative archival study and three hypotheses were used; one for every aspect of diversity. The sample consists of all U.S. firms in the period 2008-2013. Based on the prior literature and studies, a negative relationship between board diversity and earnings management was
expected. More specifically, for all three diversity aspect a negative beta was expected.
First, three different OLS regressions including and excluding year and industry dummies for the Modified Jones Model are performed. As additional test, there is also a robust regression performed which is less sensitive to outliers. In the end the performance-adjusted Modified Jones Model is used as additional regression analysis. This model is recommended by Kothari (2005). After performing different regression analyses, I found mixed results. In the regressions without year and industry dummies based on the Modified Jones Model a significant negative gender diversity coefficient (measured by percentage of females in the board) is found. Meaning that more gender diversity is associated with less earnings management. So, this is in line with my first hypothesis, which can be accepted in this way. However, when I based my regressions on the recommended performance-adjusted Modified Jones Model by Kothari (2005) or I include year and/or industry dummies, no significant results are found for my gender proxy of board diversity. For the proxy age diversity all results of the OLS and robust regressions, with and without year and industry dummies, are in accordance with each other and shows the same significant coefficients. Same results are found in my additional analysis. The results suggest that there is a positive association between age diversity and accruals-based earnings management, suggesting that when a board is more age diverse, the likelihood of earnings management increases. For ethnicity there were only found significant results in the OLS and robust regressions without year and industry dummies. This also
30
applies to my additional analysis of the Performance-adjusted Modified Jones Model.To answer the research question whether board diversity impacts the likelihood of earnings management and in which way, it can be stated that board diversity, proxied by gender, age and ethnicity, indeed impacts the likelihood of accrual-based earnings management. However, the results found for the gender variable were only significant and negative in the Modified Jones Model regressions without year and industry dummies, this is in line with my first hypothesis. With regard to the age variable it can also be concluded that it also significantly impacts earnings management, because in all regressions, significant positive results are found. However this is not in the expected direction, as hypothesized in H2. For ethnicity only for the models without year and industry controls a significant positive coefficient is found. So, while the results are not all in the expected direction as hypothesized in my three hypotheses, significant evidence is found to conclude that board diversity indeed impacts the likelihood of earnings management.
As with all studies, this study also have some limitations with associated suggestions for future research. Firstly, I only looked to three aspects of board diversity which could impact the use of earnings management. A suggestion for future research could be to include more proxies for board diversity. It could be interesting to include more board of director aspects, such as director’s expertise, skills and educational background. This would give a more comprehensive view of board diversity. I also only looked to accruals-based earnings management. There is also another form of earnings management, namely real earnings management. It may be interesting to include also this form of earnings management to give a broader view of earnings management. Another suggestion for future research is to use a broader time interval. This study only used 2008 to 2013 in the sample. For example, it could be interesting to examine the effects of board diversity on the use of earnings management in period directly before and after the SOX and compare these periods.
