The moderating role of CEO power on the relation between board diversity and board monitoring quality

(1)

0

MSc Accountancy & Control, track Control

Faculty of Economic and Business, University of Amsterdam

Master thesis:

The moderating role of CEO power on

the relation between board diversity

and board monitoring quality

Master thesis

Marloes Vierkant (10675299) 20th_{of August, 2014}

First supervisor: dr. Bo Qin

Second supervisor: dr. Alexandros Sikalidis

(2)

1

Board diversity ... 6 Social diversity ... 7 Occupational diversity ... 8 Quality of monitoring ... 8 CEO compensation ... 8 CEO turnover ... 9 CEO power ... 10 Research methodology ... 11 Regression models ... 13 Dependent variables ... 15 Independent variables ... 16 Interaction terms ... 16 Control variables ... 16 Results ... 18 Descriptive statistics ... 18 Main results ... 21 CEO compensation ... 22 CEO turnover ... 25 CEO Power ... 29 Robustness test ... 29 Conclusion ... 32

Limitations and future research ... 34

(3)

2

Abstract

the research objective of this paper is to provide insight in the value of board diversity. Because the motives for considering board diversity may lead to different board performance, board diversity is split in social and occupational diversity. The relationship between both types of diversity and the quality of monitoring by the board and the effect of CEO power on this relation is tested by using and Ordinary Least Squares and a Logistic regression.

I predict that board diversity, as social and occupational diversity, has an effect on board monitoring. Board monitoring is measured as the CEO compensation setting and the replacement of inefficient CEOs. CEO power is added as a moderator variable for the relation between board diversity and monitoring quality. I document that (1) occupational diversity enhances performance induced turnover, (2) CEO power weakens the effect of social diversity on performance induced turnover and (3) board diversity has no influence on the pay for performance sensitivity.

Keywords: Corporate governance, Board of Directors, board diversity, monitoring quality, performance induced turnover, pay for performance sensitivity

(4)

3

Introduction

Background

The objective of this study is to provide empirical evidence on the relation between social and occupational board diversity and the quality of monitoring by the board of directors. My motivation for this study is the growing call for greater diversity in the board of directors due to corporate governance policies. Corporate governance policies can ask for diversity quotas in gender, ethnicity and age. The U.S. SEC (Securities and Exchange Commission) (2009), for instance, requires disclosure on whether nominating committees consider board diversity in selecting new directors. Requiring this disclosure gives investors more insight in corporate culture and governance practices. This insight could be useful for investors to see how the board considers diversity, so they can make more informed voting and investment decisions.

Board diversity is not only pushed by the SEC, many firms see board of director diversity also as an important issue. For example, Sun Oil’s CEO, Robert Campbell states: “Often what a woman or minority person can bring to the board is some perspective a company has not had before—adding some modern-day reality to the deliberation process. Those perspectives are of great value, and often missing from an all-white, male gathering. They can also be inspiration to the company’s diverse workforce” (Campbell, 1996).

There have been various studies on board diversity (Adams and Ferreira, 2009; Anderson et al, 2011; Carter et al, 2009; Van der Walt et al, 2006; Westphal et al 1995), for instance the presence of women in boards. Research showed that women in boards have higher attendance in meetings, and men in boards with women are less likely to experience attendance problems for meetings than in homogeneous boards. The same research also showed that women are more like to join monitoring committees, which suggest that gender diversity in boards enhances monitoring (Adams and Ferreira, 2009). Research showed that diversity in boards influences behavior in boards, but it also showed that diversity can change firm outcome. Carter et al (2003) showed that board diversity has a significant positive influence on firm value, with a higher ROA and Tobin’s Q (a common measure for firm performance). Using a sample of Fortune 1000 firms, they showed that there is a positive relationship between the percentage of women and minorities on corporate boards and their firm value. Van der Walt et al (2006) also showed in a study of publicly listed companies in New Zealand that there is a significant positive correlation between a high level of board diversity and the firm’s profit. A few studies have shown that different kinds of diversity can cause different outcomes. Therefore diversity is split in social diversity, which consists of readily detectible attributes, and

(5)

4

occupational diversity, which focuses on the differences in board and professional experience (Adams and Ferreira, 2009; Anderson et al, 2011; Milliken and Martins, 1996; Westphal et al, 1995). The research of Anderson et al (2011), for instance, found that shareholders place different values on social diversity and occupational diversity.

Although a lot of research is rather positive on board diversity (Adams and Ferreira, 2009; Carter et al, 2003; Van der Walt en al, 2006; Anderson et al 2011), empirical evidence does not show that diversity always leads to a better performance of the board. Diversity can also backfire on company’s boards by friction that is caused by the different backgrounds of directors. One way to explain friction as a result of diversity is the existence of faultlines (Lau et al, 1998; Thatcher, 2003), where firm performance fails due to contradicting sub-groups within a board. Diversity affects the way groups behave, therefore the composition of board of directors seem to affect overall board behavior, firm performance and the design of compensation structures. There is extensive literature available on how different categories of diversity in boards affect the board organizational behavior and firm outcome (Adams and Ferreira, 2009; Anderson et al, 2011; Milliken and Martins, 1996; Westphal et al, 1995), but literature on the effect on the quality of board monitoring is limited. Another important part that we have seen in previous literature is that the positive and negative effects of board diversity are very dependent on the research settings, therefore it is important to include specific measures in board diversity research. Westphal et al (1995) and Anderson et al (2011) examined the relation between CEO power and board diversity. Westphal et al (1995) found that when CEOs are demographically similar to the board of directors, the directors tend to give more generous CEO compensation contracts. Therefore it is not surprising that in this same study, Westphal et al (1995) also found that when CEO power is high, new directors are more likely to be demographically similar to the CEO. This indicates that powerful CEOs prefer a board that is demographically similar to them to pursue their own interests. In line with these findings, Anderson et al (2011) found a negative relationship between CEO power and board diversity. Although Anderson et al (2011) found that CEOs tend to mitigate board diversity, they also showed that shareholders place great value on the combination of powerful CEOs and diverse boards, because when CEO power is high, there is a positive relationship between board diversity and firm performance. To conclude from previous research, we can say there is mixed evidence on the effects of board diversity on board and firm performance, and there is limited research on the role of CEO power on this relationship.

This paper aims to contribute to the existing literature in three ways. First, this study is one of the first that links CEO power with social and occupational diversity. Secondly this study will contribute

(6)

5

to the literature by focusing on board monitoring, as most board diversity studies focus on firm performance. Third, this paper contributes to the field of corporate governance on board diversity by explaining the effects on monitoring quality of boards. Although this paper does not provide support for its hypotheses, I believe that this paper can be used as a basis for further research. The structure of this paper is as follows. First of all I will give a review of the relevant literature and state my hypotheses. Then, in section three, the methodology section, I will describe the research method, empirical design and the sample characteristics. In the fourth part I will discuss the main results of the paper and in the last part I will conclude with a summary of the main results, the limitations of the paper and suggestions for further research.

Literature review & hypotheses development

Institutional theory posits that organizational environments “are characterized by the rules and requirements to which individual organizations must conform if they are to receive support and legitimacy” (Scott and Meyer, 1993: p149). Organizations act in a way they think is ‘legitimate’. For board diversity this means that organizations do not choose diverse boards because they think it increases board performance, but because they want to receive support from society and stakeholders. Karen J. Curtin, former executive vice president of the Bank of America, supports this problem on board diversity: “There is real debate between those who think we should be more diverse because it is the right thing to do and those who think we should be more diverse because it actually enhances shareholder value. Unless we get the second point across, and people believe it, we’re only going to have tokenism” (Brancato and Patterson, 1999).

An important theory in explaining the effectiveness of board diversity is the agency theory. Agency theory describes the relation between the principle (i.e. shareholders) and the agents (i.e. board of directors) and is concerned with resolving conflicts and tries to align the interests of the principle and the agent to maximize firm value (Jensen and Meckling, 1976). Schleifer and Vishny (1997) state that, concerning agency theory, CEOs need independent oversight. This independent oversight can be created by the diversity of boards, because the dynamics of these boards have a positive effect on the monitoring function of the board, and can therefore minimize agency costs.

In addition, self-categorization theory implies that people identify themselves with a particular social group, which can be within other groups. People within these social groups have the tendency to follow the norms that distinguish their group from other social groups. Tajfel and Turner (1986) state that people within groups evaluate each other more positive than people who do not belong to

(7)

6

these groups, this effect is called: in-group favoritism. The existence of subgroups within a board can therefore backfire the positive influence of board diversity. This implies however, that if boards are strongly diverse, they can overcome this problem because there are no groups big enough to form a subgroup. Another counterargument on the effect of board diversity is recognized in different studies on faultlines (Lau et al, 1998; Thatcher, 2003). Faultlines are hypothetical dividing lines that can split a group into subgroups. These group lines are mostly based on demographic characteristics like age, gender and ethnicity, but nondemographic attributes can also lead to forming subgroups. Lau et al (1998) state that when groups do not divide into faultlines in the beginning of group forming, and group members get to know each other, demographic attributes become less salient and nondemographic attributes may become key attributes for group faultline structures. According to Thatcher (2003), groups with no faultlines (very diverse groups) or strong faultlines (groups with two strong subgroups) have higher conflict levels, lower group morale and lower performance than groups with medium faultlines. Furthermore, they found that groups with moderate diversity can have strong faultlines and therefore have more relationship and process conflict.

The board of directors

The board of directors play an important role in corporate governance. All publicly listed companies are required to have such a board. Van den Berghe and Levrau (2004) state that the role of directors is dual. The first role of the board of directors includes the need for insight in the organizations’ strategy, and therefore be entrepreneurial. We can also call this the advisory function of the board. The second role of the board of directors is monitoring and control. Boards have to install a system and set up rules to minimize mismanagement and corruption. After setting up this system and rules they have to strictly monitor the outcomes (Larcker and Tayan, 2011). Since this paper investigates the quality of supervision of the board so I will only focus on the monitoring role of the board. Characteristics of monitoring mechanisms at public firms are board composition and size, director compensation, equity ownership structure, and the level of disciplinary takeover activity. The use of these monitoring mechanisms influence, for instance, CEO replacement decisions (Huson et al 2001).

Board diversity

Diversity in boards in two different categories: social diversity and occupational diversity(Anderson, 2011; Milliken and Martins, 1996; Westphal, 1995). Anderson (2011) states that social diversity includes measures for directors’ age, gender and ethnicity and occupational diversity measures differences in directors‘ education, work experience and board experience. They state that both types of diversity may be relevant and beneficial for the decision making process of the boards. Their research showed that both categories of diversity in boards have a positive influence on firm

(8)

7

performance. The results show that occupational diversity in boards has a stronger impact on firm performance than social diversity. This can be explained by the fact that social diversity is more observable by outsiders and is more relevant to a political view, since corporate governance policies ask for diversity quotas. Occupational diversity is less susceptible to political policies and plays a bigger role in board advising and monitoring, so board effectiveness increases due to better advising and monitoring. As a conclusion of their results Anderson et al (2011) suggests that shareholders place larger valuations on occupational diversity than on social diversity.

Milliken and Martins (1996) showed the same distinction earlier, in their research they showed that previous researchers also made a distinction between different types of diversity. The distinction that is made is the distinction between diversity that is ’observable with readily detectible attributes‘ and the diversity that is ‘less visible with underlying attributes’. We can compare the first distinction with social diversity and the second distinction with occupational diversity.

I expect that different categories of diversity affect board performance in a different way because the background of diversity in boards is different. Attributes of social diversity are readily detectable, and can therefore easily be subject to diversity quota´s. In this light, social diversity may arise from an outside push, because of political and corporate governance policies. Attributes of occupational diversity, on the other hand, are less visible. This implies that boards with occupational diversity take diversity into consideration because they believe that diverse boards are better boards, and they look further than just the observable diversity attributes. Because boards of social and occupational diversity consider board diversity for different reasons, this may affect board performance.

Social diversity

Social diversity includes measures for directors’ age, gender and ethnicity (Anderson et al, 2011). Milliken and Martins (1996) use a similar distinction in diversity where the type of diversity that is observable has measures like ethnic background, age or gender. According to institutional theory, as stated before by Scott and Meyer (1993), organizations will act in a way they think is legitimate to gain support from shareholders, society, the media and the government. Firms who try to be legitimate are likely to pay more attention to director demographics if there are demands from stakeholders (Ferreira, 2012). Milliken and Martins (1996) state that one reason for making a distinction between observable and nonobservable diversity is that the observable type will evoke certain responses. When diversity in groups is visible, people are likely to respond in a way that is due to biases, prejudices, or stereotypes.

The main benefit of social diversity is that it potentially brings a broader range of ideas to the boardroom (Anderson et al, 2011). Regarding age diversity, older directors lend greater stability and

(9)

8

experiential wisdom to the board, while younger directors bring greater energy to the board and have less problems bearing risk. Where the research of Anderson et al (2011) is mostly positive on age diversity, Milliken and Martins (1996) found some negative effects of age diversity. They found that firms with greater age diversity seem to have higher turnover and their directors are absent more frequently. Adams and Ferreira (2009) showed that gender diversity has positive effects on board performance, because boards with more women on it have fewer attendance problems and that women have higher focus on monitoring. Like gender and age diversity, ethnic diversity has similar benefits to board performance but has no other identifiable benefits.

Occupational diversity

Anderson et al (2011) measure occupational diversity as differences in directors‘ education, work experience and board experience. Milliken and Martins’ (1996) measures of the less observable diversity type is based on measures like education, technical abilities, functional background, tenure in the organization, or socioeconomic background, personality characteristics, or values. This type of diversity relates to the differences in personality and values of a group. Although the differences in this type of diversity are harder to observe than social diversity, these differences can create major differences on individual perspectives on key issues, and is therefore an important type of diversity. Professional diversity has a substantial impact on the directors’ identification, their perception of problems and their solutions to these issues. Directors with accounting/finance expertise, for instance, may be more sensitive to financial issues and can easily communicate with directors with a similar professional background (Anderson et al, 2011).

Quality of monitoring

CEO compensation

Although most studies on diversity focus on firm performance, less studies investigate the monitoring performance of the board, like the effectiveness of CEO compensation setting. Westphal et al (1995) state that there is a strong positive relation between demographic similarity of the board and the CEOs total compensation. A reason for this is that demographic similarity enhances the boards’ perception of the value of CEOs leadership. Results also showed that there is a negative relationship between increased CEO-board similarity and subsequent change in compensation contingency. In addition to these results, Adams and Ferreira (2009) showed that the proportion of the compensation that is contingent with firm performance compared to overall director pay is higher in firms with relatively more women on boards. So in general the studies of Westphal et al (1995) and Adams and Ferreira (2009) showed that in increase in similarity between the CEO and the board of directors is associated with more generous compensation contracts, since the proportion of

(10)

9

compensation contingent on firm performance decreases when similarity between the CEO and the board of directors increases. These results are also in line with in-group favoritism as Tajfel and Turner (1986) stated in their research.

Westphal et al (1995) argue that similarity between the CEO and the board of directors can have significant influence on the decision making process, due to in-group favoritism. CEOs are more likely to influence the compensation-setting process in boards with directors demographically similar to them, since CEOs have more power in boards with directors who are demographically similar to each other. Previous studies ( Westphal et al, 1995; Anderson et al, 2011) showed that CEOs use their power to influence the compensation-setting process by trying to attract directors who are demographically similar to them because they are less critical and set higher CEO compensation. The results from the research of Westphal et al (1995) show that if CEO and the board of directors are demographically similar to each other, CEO compensation is higher.. This means that if board diversity is higher, less directors are similar to the CEO and there is less in-group favoritism, and therefore is pay performance sensitivity is higher. Although various researchers (Anderson et al, 2011; Tajfel and Turner, 1986; Westphal et al, 1995) implicate that diversity has a positive effect on pay performance sensitivity of the firm, counterarguments also exist. As stated before, faultlines cause subgroup conflicts and board performance therefore suffers (Lau et al, 1998; Thatcher, 2003). According to the work of Anderson et al (2011) I expect that social diversity has a weaker influence on the pay-performance sensitivity of the firm than occupational diversity. Therefore I hypothesize:

H1a: Social board diversity is related to pay-performance sensitivity of the CEO of the firm.

H1b: Occupational board diversity is related to pay-performance sensitivity of the CEO of the firm.

CEO turnover

One of the most important decisions to be made by the board in the realm of monitoring is efficient CEO dismissal. Decisions on CEO dismissal have long-term implications for several management decisions like investment, operating and financial decisions. If the board fails to remove the CEO when performance is poor, the effectiveness of this internal monitoring mechanism is lacking (Huson et al, 2001).

Recent literature on the quality of internal monitoring is mainly focused on the likelihood of top management turnover as a consequence of bad financial performance or managerial ownership. Although there is literature on the influence of board diversity on CEO turnover, results are unclear. Westphal et al (2005) find that if board members are more demographically similar to each other, they are more likely to take actions after problems have occurred. This means that a board with low demographic diversity is probably more likely to replace the CEO when performance is bad than

(11)

10

boards that are more demographically diverse. Adams and Ferreira (2009) found that the opposite is true. In their research they showed that CEO turnover is more sensitive to stock return performance when firms have relatively more women on boards, which means that boards with higher gender diversity are more likely to hold CEOs accountable for poor firm performance. Similar to the results of the study of Anderson et al (2011) I expect that social diversity has a weaker influence on performance induced CEO turnover than occupational diversity. Therefore I predict the following hypothesis:

H2a: Social board diversity is related to performance induced CEO turnover. H2b: Occupational board diversity is related to performance induced CEO turnover.

CEO power

Prior research on board diversity and firm performance has resulted in ambiguous results, Combs (2007) explains this due to the missing moderating factor CEO Power. Research has showed that CEO power influences board diversity, and the relationship between board diversity and firm performance (Anderson et al, 2011; Westphal et al, 1995; Combs, 2007). Anderson et al (2011) and Westphal et al (1995) both showed that powerful CEOs tend to limit board diversity because they know board members will be more sympathetic to them.

Combs (2007) states in his research on board composition that when CEOs lack power no extra layer of control is needed. He also showed that when CEO power is measured with CEO ownership and duality, CEO power moderates the board composition-firm performance link. Anderson et al (2011) also investigated the relation between board composition and firm performance with the moderator CEO Power. Their results showed that CEO power strengthens the positive relationship between board diversity and firm performance. This implies that in the presence of powerful CEOs, shareholders place greater value on a diverse boards than non-diverse boards. Based on previous literature, my third hypothesis is:

H3a: CEO power influences the effect of social diversity on the pay-performance sensitivity of the CEO of the firm

H3b: CEO power influences the effect of occupational diversity on the pay-performance sensitivity of the CEO of the firm

H3c: CEO power influences the effect of social diversity on the performance induced CEO turnover. H3d: CEO power influences the effect of occupational diversity on the performance induced CEO

(12)

11

A summary of the theoretical model is provided below in Figure 1:

Figure 1

Research methodology

This research examines the relation between diversity in the board of directors and the quality of monitoring the CEO. This is measured by the efficiency of CEO compensation and the efficiency of replacing inefficient CEOs. Data of the background of directors is required; the compensation of the CEO; performance induced CEO turnover; CEO Power and some control variables. I will identify the composition of board of directors, compensation data, data on CEO turnover, CEO power and data regarding my control variables by using the databases of Wharton Research Data Services (WRDS).

Table 1 - Summary of sample selection process.

Sample # observations

# of director observations 2011/2012 27706

Less: directors with missing diversity data 117

Total: director observations 27589

# firm year observations RiskMetrics 2956

# firm year observations CompuStat 22603

# firm year observations Execucomp 3680

Total: firm year observations after merging 2630

Less: firm year observations with insufficient financial data 426

(13)

12

I will measure the differences in directors’ age, gender and ethnicity to identify social diversity. Diversity of directors’ age is measured as the differences of director age across the entire board. I calculated the mean age of directors in the board, and then calculated the differences compared to that mean age divided by the board size. I divided this by the higher outcome to create a score between 0 and 1. Gender diversity is measured as the number of female directors on the board divided by the total board size. If there are two women in a board that consists of eight directors, the gender diversity score is 2/8=0.25. Ethnic diversity is calculated as the number of African-American, Asian, Hispanic, Indian, Middle-Eastern and Native American directors serving on the board divided by the total board size. This means that if there is one Asian, one Indian and four Caucasian directors on one board, the ethnic diversity score of that firm is 2/6=0.33. Because all separate diversity measures lie between 0 and 1, I added the diversity measures to each other. For the total social diversity score the total scores of age, gender and ethnic diversity are divided by the number of variables available for this measure, so three.

In this study I measure occupational diversity by professional diversity and board experience diversity. Professional diversity is measured by the number of employment categories in a board and the number of financial experts in the board. Professional background is the employment category, which can be academic, accountant, attorney/counsel, consultant, executive, financial services, medical, other, prof Director, real estate services and retired. The professional background of directors is the amount of different background directors have divided by the total number of backgrounds available.. So a board with only directors with an accounting background have a score of 1/11=0,091 where boards with directors with background in consulting, financial services and medicine have a score of 3/11=0,273. A second measure for professional diversity is the number of financial experts on the board. This is measured by the total of financial experts on the board, divided by total board size. This means that with four directors who are financial experts, on board with ten directors in total, the diversity score regarding to the number of financial experts is 4/10=0.25.

Board experience is measured by director tenure and the number of other boards directors serve on. Director tenure is measured as the coefficient of variation on of the time directors are employed in the firm. I calculated the mean tenure for directors in a firm, and then calculated the differences in tenure based on that mean director tenure score. The last measure of occupational diversity is the difference in the number of other corporate boards that directors serve on. Because all separate diversity measures lie between 0 and 1. After that the total occupational diversity score is divided by four because there are four separate diversity measures.

(14)

13

To measure CEO power, I use, according to Anderson et al (2011), a factor score of CEO tenure (CEO_Tenure), CEOs equity holding (Perc_Total_Shares) and free cash flow (operating cash flow minus capital expenditures) divided by total assets (Free_Cash_ROA).

Regression models

First, I will estimate the determinants of diversity in boards. I estimate these determinants to see which variables influence board diversity. I use two different models for social and occupational diversity. The following empirical models are used in an Ordinary Least Squares regression:

Div_Social= β0 + β1CEO_Power + β2Board_Size + β3Indep_Boards + β4Firm_Size + β5Growth

β6Leverage + β7ROA + β8Earnings_Per_Share + ∑ βi DummyYeari + ∑ βj DummyIndustry j (1)

Div_Occupational= β0 + β1CEO_Power + β2Board_Size + β3Indep_Boards + β4Firm_Size + β5Growth

β6Leverage + β7ROA + β8Earnings_Per_Share + ∑ βi DummyYeari + ∑ βj DummyIndustry j (2) The first determinant for board diversity is CEO power, since research showed powerful CEOS tend to mitigate board diversity (Anderson et al, 2011; Westphal et al, 1995). According to Lehn et al (2009) larger boards inherently are more diverse, so include incorporate board size (Board_Size) as a possible determinant for board diversity. According to Anderson et al (2011) the third determinant is Indep_Boards. This is calculated as the number of independent board members divided by the total board size. Randøy et al (2006) found that larger firms have more diverse boards because they are

more sensitive to social and political concerns, therefore I include firm size (Firm_Size). Lehn et al (2009) find growth opportunities as a determinant board composition, so I include Growth as a variable. Growth is measured as the market to book ratio, which is calculated by dividing the market value of the firm by the book value of the firm. I included Leverage because Pearce et al (1992) state that the leverage level of the firm influence board size and composition. Leverage is the debt to equity ratio, which is calculated as the total liabilities divided by total assets. The last variables that may influence board diversity are return on assets (ROA) and earnings per share (Earnings_Per_Share). The control variables DummyYear and DummyIndustry are held constant to test the relative impact of the independent variables.

To measure the efficiency of CEO compensation I measure pay for performance. Based on the theoretical model in figure 1, the following models are used to test the first hypothesis:

(15)

14

LnCEO_Total_Compensation= β0 + β1Firm_Performance + β2Div_Social +

β3Firm_Performance*Div_Social β4CEO_Power + β5CEO_Power*Div_Social + β6CEO_Tenure +

β7CEO_Age + β8Board_Size + β9Indep_Boards + β10Growth + β11Firm_Size + β12Dummy_Gender_CEO

+ β13CEO_Duality + β14Leverage + ∑ βi DummyYeari + ∑ βj DummyIndustry j (3) LnCEO_Total_Compensation= β0 + β1Firm_Performance+ β2 Div_Occupational +

β3Firm_Performance*Div_Occupational + β4CEO_Power + β5CEO_Power*Div_Occupational +

β6CEO_Tenure + β7CEO_Age + β8Board_Size + β9Indep_Boards + β10Growth + β11Firm_Size +

β12Dummy_Gender_CEO + β13CEO_Duality + β14Leverage + ∑ βi DummyYeari + ∑ βj DummyIndustry j

(4)

Firm_Performance = firm performance measures. This are ROA and Earnings_Per_Share Div_Social = The social diversity score of the firm

Div_Occupational = The occupational diversity score of the firm

LnCEO_Total_Compensation = the natural logarithm of the total compensation of the CEO

The models as seen above are used in an Ordinary Least Squares (OLS) regression.

If there is pay for performance, I expect β1 to be positive. A significant β3 would suggest that board diversity has an effect on the quality of supervision. If board diversity enhances pay performance sensitivity of the CEO, I expect β3 to be positive. When board diversity enhances pay performance sensitivity of the CEO, it means board efficacy increases. A significant negative β3 means that board diversity reduces board efficacy. If β5 is positive, it means that in the presence of powerful CEOs, the compensation of the CEO increases when boards are diverse.

To examine the efficiency of replacing inefficient CEOs, I use the following empirical models:

Logit[Pr (Y t=1)] = β0 + β1Firm_Performance+ β2Div_Social + β3Firm_Performance*Div_Social +

β4CEO_Power β5CEO_Power*Div_Social + β6CEO_Tenure + β7CEO_Age + β8Board_Size +

β9Indep_Boards + β10Growth + β11Firm_Size + ∑ βi DummyYeari + ∑ βj DummyIndustry j (5) Logit[Pr (Y t=1)] = β0 + β1Firm_Performance + β2Div_Occupational +

β3Firm_Performance*Div_Occupational + β4CEO_Power + β5CEO_Power*Div_Occupational +

β6CEO_Tenure + β7CEO_Age + β8Board_Size + β9Indep_Boards + β10Growth + β11Firm_Size + ∑ βi

DummyYeari + ∑ βj DummyIndustry j (6)

Firm_Performance = firm performance measures. This are ROA and Earnings_Per_Share Div_Social = The social diversity score of the firm

(16)

15

Div_Occupational = The occupational diversity score of the firm Y = CEO replacement in year t.

If there is performance-induced turnover, I expect β1 to be negative in this logistic regression. A

significant β3 would suggest that board diversity has an effect on the quality of supervision. If board diversity strengthens performance induced turnover, which means it increases board efficacy, I expect β3 to be negative. A positive β3 means that board diversity mitigates board efficacy. If β5 is negative, it means that in the presence of powerful CEOs, CEO turnover decreases when boards are diverse.

To measure if CEO power enhances of weakens the relation between social and occupational diversity and the quality of monitoring I use the following models:

LnCEO_Total_Compensation= β0 + β1Firm_Performance + β2Div_Social + β3CEO_Power + β4

Firm_Performance*Div_Social + β5 Firm_Performance*CEO_Power + β6 Div_Social*CEO_Power + β7 Firm_Performance*Div_Social*CEO_Power + β8CEO_Tenure + β9CEO_Age + β10Board_Size + β11Indep_Boards + β12Growth + β13Firm_Size + β14Dummy_Gender_CEO + β15CEO_Duality +

β16Leverage + ∑ βi DummyYeari + ∑ βj DummyIndustry j (7)

Logit[Pr (Y t=1)] = β0 + β1Firm_Performance + β2Div_Social + β3CEO_Power + β4

Firm_Performance*Div_Social + β5 Firm_Performance*CEO_Power + β6 Div_Social*CEO_Power + β7 Firm_Performance*Div_Social*CEO_Power + β8CEO_Tenure + β9CEO_Age + β10Board_Size + β11Indep_Boards + β12Growth + β13Firm_Size + β14Dummy_Gender_CEO + β15CEO_Duality +

β16Leverage + ∑ βi DummyYeari + ∑ βj DummyIndustryj (8)

Dependent variables

The total compensation of a CEO (CEO_Total_Compensation) is the total compensation as reported to the SEC. This is the sum of salary, bonus and other compensation as recorded to the SEC.

The turnover rate is calculated as a dummy variable (CEO_Turnover_Dummy) if the CEO was replaced in that specific year. This means a variable of 1 if the CEO changed in that year, and a variable of 0 if the CEO remained the same in that selected year.

My hypotheses suggests that CEO power influences the effect of diversity on board monitoring quality so I will use CEO power as a moderator. According to Anderson et al (2011), CEO power is measured as a factor score of CEO tenure (CEO_Tenure), CEOs equity holding (Perc_Total_Shares) and free cash flow (operating cash flow minus capital expenditures) divided by total assets (Free_Cash_ROA).

(17)

16

Independent variables

The diversity scores for firms are divided in two categories, social diversity and occupational diversity. I measure social diversity by age diversity, gender diversity and ethnic diversity. Occupational diversity is measured by the differences in professional background and board experience of directors within their board. I added the separate diversity scores to measure social and occupational diversity, because all values lie between 0 and 1.

The firm performance measures (Firm_Performance) consist of the return on assets (ROA) and the earnings per share (Earnings_Per_Share). The firms’ ROA is the net income divided by the book value of total assets. Earnings_Per_Share is calculated as the firms’ profit divided by the total number of shares outstanding.

Interaction terms

β3 is an interaction term, which consists of two firm performance measures. These measures are return on assets (ROA) and earnings per share (Earnings_Per_Share). The variables of the firm performances are multiplied by the diversity score of the firm, where each firm has two different diversity scores, the social diversity score (Div_Social) and the occupational diversity score (Div_Occupational). The interaction terms of this paper are ROA*Div_Social, ROA*Div_Occupational, Earnings_Per_Share*Div_Social and Earnings_Per_Share*Div_Occupational.

Control variables

To test the relative impact of the dependent variables I include several control variables. I included control variables of which I think also have influence on CEO compensation and CEO turnover so I have to keep them constant. The variables that are included as control variables are described below.

Most control variables are applicable to both CEO compensation and CEO turnover. First I control for the tenure of the CEO (CEO_Tenure) because when a director has been a CEO for a longer timeframe the amount of total compensation is probably higher. It is also logical that if tenure is higher, CEOs are more likely to retire. Another control variable is CEO_Age. CEOs will naturally have higher compensation because they have more experience and might work longer for the same firm. Also, when the CEOs age is higher, it is more likely that they retire because they are closer to the retirement age. The third control variable is CEO_Power. According to Finkelstein et al (1989) CEO power influences CEO compensation. Weisbach et al (1988) state that strong CEOs are less likely to be replaced than weak CEOs, so I also include CEO_Power). According to Mikkelson (1997) I include Board_Size and Indep_Boards as a control variable for CEO turnover. Greater director independence

(18)

17

from senior management potentially improves the monitoring quality of the board and is therefore also a control variable for CEO compensation. The board independence is measured by the number of outside directors divided by board size. In line with Core (1999) I expect that larger firms with greater growth opportunities and more complex operations will demand higher-quality managers with higher wages, so I include Growth and Firm_Size. Puffer et al (1991) state that growth and firm size also influence the turnover rate, so both Growth and Firm_Size are also included for CEO_Turnover_Dummy.

As common, men earn more than women so it is naturally that their compensation is higher than that of female CEOs, therefore I have to keep the variable Dummy_Gender_CEO constant. Core et al (1999) found in their results that duality of the CEO influences CEO compensation, therefore I included CEO_Duality as a control variable. The last firm specific control variable I use is leverage (Leverage). When a firm has a higher leverage, which means that they have more debt than equity, it is more likely to give a higher compensation to the CEO than firms with lower leverage. Leverage is measures as the total liabilities divided by total assets

Finally, I control for the years and the industries. I use a dummy variable (DummyYear) for the years 2011 and 2012. I also control for industry by including dummy variables for each two-digit SIC codes in the sample.

The variable definitions are summarized in table 2.

Table 2 variable definitions

Variable Name Description

Div_Social A measure for social diversity, this is a measure of the differences in directors age, gender and ethnicity within a board

Div_Occupational A measure for occupational diversity, this is a measure of the differences in directors employment category, % financial experts, nr of other boards directors serve on and directors tenure within a board

LnCEO_Total_Compensation Natural logarithm of the total compensation of CEOs as directed to the SEC

CEO_Turnover_Dummy A dummy variable for CEO turnover, where 1=CEO changed and 0=CEO remained the same

ROA Return on assets (net income/book value of total assets)

Earnings_Per_Share Firms profit/number of shares

Earnings_Per_Share*Diversity The earnings per share multiplied by the social/occupational diversity score

ROA*Diversity The return on assets multiplied by the social/occupational diversity score

(19)

18

CEO_Power A factor score of CEO_Tenure, Perc_Total_Shares and Free_Cash_ROA

Firm_Performance*Diversity

*CEO_Power Firm performance multiplied by social/occupational diversity, multiplied by CEO power

CEO_Tenure Number of years CEO has been in the current role

CEO_Age The age of the CEO

Indep_Boards The number of independent board members/board size

Board_Size The total board size of the firm

Growth Book value/market price

Firm_Size Natural logarithm of the market value of the firm

Dummy_Gender_CEO A dummy variable where 1=man and 0=woman

CEO_Duality A dummy variable for CEOs other functions. 1=CEO also chairman and 0=CEO is not also chairman

Leverage Total liabilities/total assets

Results

Descriptive statistics

The sample consists of firms that are included in the Standards & Poor’s (S&P) 400, 500 and 600. The data comes from the years 2011 and 2012 since that are the only years where subsequent data is available on board diversity because requirement from the SEC on disclosures regarding board diversity started from 2009. Any data before 2011 was not sufficient regarding board diversity. I focus on S&P firms because they have to provide disclosure concerned to nominating new directors and board directors.

The summary statistics contains a sample of 2206 firm-year observations, as seen in Table 2. As described before I use two measures of diversity, Div_Social and Div_Occupational. I made a diversity score for each firm, the scores for social diversity lie between .01 and .49 and the scores for occupational diversity lie between .10 and .53. All the proxies for diversity are measured in percentages of the lowest and highest scores. All the diversity scores lie between 0 and 1, so I added the separate diversity scores to each other. Because social diversity is measured by three proxies (Div_Age, Div_Gender and Div_Ethnicity) and occupational diversity by four (Div_Employment, Div_Financial, Div_Tenure and Div_Nr_Boards) I divided the score by the number of proxies.

Because CEO_Total_Compensation and Firm_Size have a wide spread, I have dropped the bottom and top 1 percent of these variables. The mean of CEO total compensation

(CEO_Total_Compensation) is $6237,15. Because the difference between the highest and the lowest CEO total compensation does not say much, I made a natural logarithm variable for that (LnCEO_Total_Compensation).

(20)

Table 3 Descriptive statistics

Variable Mean sd min max p1 p25 p50 p75 p99

Div_Social 0.151 0.069 0.006 0.492 0.17 0.104 0.141 0.19 0.361 Div_Occupational 0.225 0.056 0.203 0.529 0.126 0.185 0.22 0.25 0.405 CEO_Total_Compensation 6,237.145 5,208.541 450.575 30,460.190 587.588 2,547.316 4,723.05 8,433.33 26,455.110 CEO_Turnover 0.207 0.405 0.000 1.000 0.000 0.000 0.000 0.000 1.000 ROA 0.055 0.083 −1.703 0.783 −0.144 0.018 0.051 0.089 0.272 Earnings_Per_Share 1.970 2.877 −29.40 35.120 −4.720 0.730 1.700 2.920 9.080 CEO_Power 0.640 0.320 −1.084 1.689 0.052 0.374 0.627 0.923 1.296 CEO_Tenure 6.931 3.718 0.000 13.000 0.000 4.000 7.000 11.000 13.000 CEO_Age 56.24 7.109 33.000 96.000 41.000 52.000 56.000 61.000 75.000 Board_Size 9.275 2.204 4.000 20.000 5.000 8.000 9.000 11.000 16.000 Indep_Boards 0.811 0.099 0.250 1.000 0.555 0.750 0.857 0.889 0.923 Growth 0.573 0.411 −2.288 5.144 −0.022 0.313 0.499 0.767 1.861 Firm_Size 69.299 151.299 1.323 1,620.013 1.711 8.115 19.592 57.512 867.969 Dummy_Gender_CEO 0.963 0.188 0.00 1.000 0.000 1.000 1.000 1.000 1.000 CEO_Duality 0.822 0.383 0.00 1.000 0.000 1.000 1.000 1.000 1.000 Leverage 0.544 0.230 0.00 1.820 0.083 0.381 0.542 0.705 1.004

CEO_Total_Compensation is in thousand American dollars

(21)

Table 4 Correlation matrix Div _So cia l Div_O ccu pa tio na l CEO _T ota l_ Co mp en sa tio n CEO _T ur no ve r ROA Earn in gs _P er _S ha re CEO_ Po w er CEO _T en ur e CEO_ Ag e Bo ar d_ Si ze Inde p_ Bo ar ds Gro w th Firm _S iz e Dum m y_ G en de r_ CE O CE O _D ua lit y Le ver ag e Div_Social 1 Div_Occupational 0.038 1 CEO_Total_Compensation 0.139 −0.0925 1 CEO_Turnover −0.0416 −0.0491 0.053 1 ROA 0.034 −0.0081 0.087 −0.0326 1 Earnings_Per_Share 0.040 −0.0115 0.226 −0.0426 0.501 1 CEO_Power −0.0292 0.032 −0.033 0.081 0.299 0.134 1 CEO_Tenure −0.0612 0.034 −0.043 0.102 0.031 0.031 0.931 1 CEO_Age −0.13 0.136 0.065 0.176 −0.0454 0.002 0.263 0.278 1 Board_Size 0.185 0.109 0.357 −0.0357 −0.0513 0.131 −0.1588 −0.1303 0.063 1 Indep_Boards −0.0607 −0.1574 0.194 0.002 0.012 0.087 −0.107 −0.0904 −0.0722 0.171 1 Growth −0.0583 0.094 −0.1426 0.014 −0.2867 −0.1746 −0.1505 −0.0523 0.070 0.058 0.022 1 Firm_Size 0.141 −0.0405 0.626 0.026 0.121 0.201 −0.0503 −0.0765 0.039 0.361 0.150 −0.1502 1 Dummy_Gender_CEO −0.1249 0.014 −0.0135 0.017 0.001 0.010 0.066 0.077 0.043 0.003 −0.0126 0.002 −0.0285 1 CEO_Duality −0.006 −0.0039 −0.0061 −0.1448 −0.0148 0.033 0.046 0.063 −0.0378 0.020 0.114 0.032 −0.0018 0.010 1 Leverage 0.062 0.020 0.204 −0.0355 −0.2963 −0.0292 −0.1693 −0.0853 0.068 0.421 0.146 −0.0172 0.103 −0.0305 0.049 1

(22)

21

The correlation matrix is shown in Table 3. As seen in the table no variables show correlation with social diversity (Div_Social) and occupational diversity (Div_Occupational). Board_Size and Firm_Size show a positive correlation with the dependent variable CEO_Total_Compensation.

Main results

To estimate the determinants of social and occupational diversity I used the Ordinary Least Squares Model (OLS). Table 5 and present this test. Both social and occupational diversity are included in the table. Industry and year dummies are included in the regression but not shown.

Table 5 Ordinary Leas Squares model analysis

Dependent variable

Independent variables Div_Social Div_Occupational

constant 0.157*** 0.281*** (7.77) (16.77) CEO_Power −0.009* 0.007* (−1.93) (1.76) Board_Size 0.005*** 0.004*** (6.49) (5.82) Indep_Boards −0.052*** −0.094*** (−3.55) (−7.65) Firm_Size 0.000*** −0.000** (3.73) (−2.37) Growth −0.005 0.008** (−1.36) (2.50) Leverage −0.001 −0.013* (−0.15) (−1.89) ROA −0.008 0.014 (−0.41) (0.79) Earnings_Per_Share 0.001 −0.000 (0.90) (−0.45)

Year and Industry dummies Yes Yes

Observations 2204 2204

***. Correlation is significant at the 0.01 level (two−tailed) **. Correlation is significant at the 0.05 level (two−tailed) *. Correlation is significant at the 0.10 level (two−tailed)

As seen in table 5 Board_Size has a strong positive significant relationship with the dependent variables Div_Social and Div_Occupational. Indep_Boards also has a strong significant relationship with Div_Social and Div_Occupational but this relationship is negative. Firm_Size has a strong significant relationship with Div_Social at the 1 percent level and a relationship with Div_Occupational at the 5 percent level, but where Firm_Size has a positive relationship with

(23)

22

Div_Social, there is a negative relationship for Firm_Size and Div_Occupational. Growth has a negative relationship with Div_Occupational at the 5 percent significance level, and Leverage has a negative relation with Div_Occupational at the 10 percent level. I do not have enough evidence to assume relationships between the other independent variables and the dependent variables Div_Social and Div_Occupational.

CEO compensation

To test the influence of social and occupational diversity on CEO compensation, measured as pay for performance sensitivity I have used an OLS model. I use LnCEO_Total_Compensation as dependent variable in this test. To see the influence of board diversity I use the first firm performance measure return on assets (ROA), secondly I use the test with the performance measure earnings per share (Earnings_Per_Share). I analyze the two firm performance measures with both social diversity (Div_Social) and occupational diversity (Div_Occupational).

Table 6 depicts the results of the regression models with the dependent variable LnCEO_Total_Compensation and the firm performance measure return on assets (ROA). Column 1 and 4 show the results of the baseline models. Column 2 and 5 show the results of the model including the interaction term between firm performance and diversity, and column 3 and 6 show the results regarding the moderator CEO power.

As seen in the table ROA shows a strong significant positive relation with LnCEO_Total_Compensation at the 1 percent significance level for both social and occupational diversity. Div_Social has a weak positive significant relationship with the dependent variable, and for Div_Occupational this relation is strongly negative. CEO_Power has a strong negative relation with LnCEO_Total_Compensation at the 1 percent level. The interaction terms regarding diversity and firm performance, ROA*Div_Social and ROA*Div_Occupational do not show a significant relation with the dependent variable. The interaction terms regarding CEO power, ROA*Div_Social*CEO_Power and ROA*Div_Occupational*CEO_Power, are also not significant.

The control variables CEO_Tenure, Board_Size, Indep_Boards, Firm_Size and Leverage show for both social and occupational diversity a strong positive relationship with LnCEO_Total_Compensation. CEO_Age showed for both Div_Social and Div_Occupational a positive relationship with the dependent variable at the 10 percent level. Growth showed a weak negative significant relation at the 5 percent level. There is not enough evidence to conclude a relationship between the other variables Dummy_Gender_CEO and CEO_Duality and the dependent variable.

(24)

23

Table 6 Ordinary Least Squares Model Analysis

Dependent variable: LnCEO_Total_Compensation

(1) (2) (3) (4) (5) (6) Div_Social Div_Occupational constant 5.895*** 5.882*** 6.044*** 6.214*** 6.233*** 6.236*** (26.27) (26.15) (25.81) (27.74) (27.64) (23.42) ROA 1.554*** 1.843*** 1.872*** 1.548*** 0.962 0.009 (5.56) (4.04) (3.60) (5.54) (1.13) (0.01) Diversity 0.494** 0.594** −0.540 −0.896*** −1.047*** −1.080 (2.22) (2.33) (−1.10) (−3.39) (−3.11) (−1.59) CEO_Power −0.721*** −0.742*** −1.014*** −0.703*** −0.699*** −0.698** (−3.44) (−3.51) (−4.27) (−3.36) (−3.34) (−2.29) ROA*Diversity −1.912 −0.801 2.665 7.206 (−0.80) (−0.22) (0.73) (0.93) ROA*CEO_Power 0.430 1.016 (0.85) (0.56) CEO_Power*Diversity 1.936*** 0.004 (2.74) (0.00) ROA*Diversity*CEO_Power −4.457 −5.242 (−1.01) (−0.62) CEO_Tenure 0.062*** 0.063*** 0.063*** 0.060*** 0.060*** 0.060*** (3.63) (3.70) (3.66) (3.55) (3.53) (3.55) CEO_Age 0.004* 0.004 0.003 0.004* 0.004* 0.004* (1.66) (1.63) (1.49) (1.75) (1.79) (1.82) Board_Size 0.068*** 0.068*** 0.067*** 0.074*** 0.075*** 0.074*** (8.41) (8.42) (8.28) (9.20) (9.23) (9.17) Indep_Boards 1.127*** 1.125*** 1.168*** 1.018*** 1.021*** 1.021*** (7.41) (7.40) (7.64) (6.65) (6.66) (6.65) Growth −0.084** −0.088** −0.082** −0.080** −0.078** −0.083** (−2.12) (−2.20) (−1.96) (−2.02) (−1.96) (−2.01) Firm_Size 0.002*** 0.002*** 0.002*** 0.002*** 0.002*** 0.002*** (18.64) (18.66) (18.71) (18.71) (18.69) (18.69) Dummy_Gender_CEO −0.024 −0.024 −0.035 −0.041 −0.041 −0.041 (−0.31) (−0.31) (−0.46) (−0.54) (−0.54) (−0.53) CEO_Duality −0.016 −0.017 −0.016 −0.014 −0.014 −0.014 (−0.42) (−0.44) (−0.43) (−0.38) (−0.38) (−0.38) Leverage 0.927*** 0.930*** 0.930*** 0.914*** 0.916*** 0.918*** (11.18) (11.21) (11.17) (11.03) (11.05) (11.05)

Year and Industry dummies Yes Yes Yes Yes Yes Yes

Observations 2204 2204 2204 2204 2204 2204

(25)

24

(1) (2) (3) (4) (5) (6) Div_Social Div_Occupational constant 5.970*** 5.977*** 6.133*** 6.275*** 6.244*** 6.143*** (26.77) (26.75) (26.39) (28.16) (27.87) (23.24) Earnings_Per_Share 0.036*** 0.032** 0.049*** 0.036*** 0.068*** 0.133** (6.56) (3.24) (2.65) (6.58) (2.77) (2.36) Diversity 0.452** 0.388 −0.436 −0.879*** −0.613* −0.141 (2.04) (1.55) (−0.84) (−3.33) (−1.85) (−0.21) CEO_Power −0.159 −0.153 −0.382** −0.145 −0.147 −0.004 (−1.05) (−1.01) (−2.12) (−0.96) (−0.98) (−0.02) Earnings_Per_Share 0.027 −0.098 −0.139 −0.410* *Diversity (0.54) (−0.85) (−1.33) (−1.70) Earnings_Per_Share −0.029 −0.092 *CEO_Power (−1.08) (−1.29) CEO_Power*Diversity 1.147 −0.738 (1.49) (−0.77) Earnings_Per_Share 0.246 0.387 *Diversity*CEO_Power (1.15) (1.25) CEO_Tenure 0.016 0.015 0.019 0.015 0.015 0.018 (1.31) (1.27) (1.49) (1.22) (1.24) (1.42) CEO_Age 0.003 0.003 0.003 0.003 0.003 0.003 (1.39) (1.43) (1.26) (1.50) (1.37) (1.41) Board_Size 0.067*** 0.067*** 0.066*** 0.073*** 0.072*** 0.072*** (8.31) (8.32) (8.11) (9.05) (8.85) (8.82) Indep_Boards 1.133*** 1.130*** 1.158*** 1.027*** 1.021*** 1.006*** (7.48) (7.46) (7.59) (6.74) (6.69) (6.57) Growth −0.093** −0.094** −0.091** −0.089** −0.089** −0.093** (−2.38) (−2.40) (−2.32) (−2.27) (−2.29) (−2.36) Firm_Size 0.002*** 0.002*** 0.002*** 0.002*** 0.002*** 0.002*** (18.10) (18.08) (18.09) (18.15) (18.15) (18.12) Dummy_Gender_CEO −0.029 −0.030 −0.041 −0.044 −0.043 −0.045 (−0.37) (−0.39) (−0.53) (−0.57) (−0.56) (−0.58) CEO_Duality −0.021 −0.021 −0.0203 −0.020 −0.019 −0.021 (−0.58) (−0.57) (−0.54) (−0.54) (−0.52) (−0.57) Leverage 0.855*** 0.852*** 0.842*** 0.842*** 0.846*** 0.847*** (10.67) (10.61) (10.45) (10.52) (10.56) (10.54)

Year and Industry

dummies Yes Yes Yes Yes Yes Yes

Observations 2204 2204 2204 2204 2204 2204

(26)

25

In table 7 the results of the model with the dependent variable LnCEO_Total_Compensation and the firm performance measure Earnings_Per_Share are shown. Columns 1 and 4 show the results for the model without the interaction terms, column 2 and 5 show the results regarding diversity and firm performance, and column 3 and 6 show the results regarding CEO power. Earnings_Per_Share shows a strong positive relation with the dependent variable at the 1 percent level for Div_Social and Div_Occupational.

Div_Social shows a significant positive and Div_Occupational shows a strong significant negative relation with LnCEO_Total_Compensation. None of the interaction terms show a significant relation with the dependent variable LnCEO_Total_Compensation. The control variables Board_Size, Indep_Boards, Firm_Size, Dummy_Gender_CEO, CEO_Duality and Leverage show a strong positive significant relation with the dependent variable at the 1 percent level. Growth shows a significant negative relation with the dependent variable. The results of the control variables do not show significant differences between social and occupational diversity.

Because I do not have enough evidence to provide evidence for a relation with interaction terms ROA*Div_Social and ROA*Div_Occupational and the dependent variable LnCEO_Total_Compensation I reject hypothesis 1a and 1b

Because I had to reject hypothesis 1a and 1b, there is not enough support for the first hypothesis, therefore I reject hypothesis 1. However I cannot accept the first hypothesis, the results of the OLS models show that there are significant relationships with several control variables and the dependent variable LnCEO_Total_Compensation.

CEO turnover

To test the influence of the interaction terms of social and occupational diversity (Div_Social and Div_Occupational) and the firm performance measures return on assets (ROA) and earnings per share (Earnings_Per_Share) on CEO turnover (CEO_Turnover_Dummy), I use a Logistic Regression model because the dependent variable CEO_Turnover_Dummy is a dummy variable. The CEO changes 456 times in this dataset and the CEO remains the same person 1748 times of our total of 2204 firm-year observations.

Table 8 shows the results of the regression models with the dependent variable LnCEO_Total_Compensation and the firm performance measure return on assets (ROA). Column 1 and 4 show the results of the baseline models, column 2 and 4 show the results of the model with the interaction term regarding diversity and firm performance, and column 3 and 6 show the results on the model with the interaction term with firm performance, diversity and CEO power.

(27)

26

As seen in the table ROA has no significant relation with CEO_Turnover_Dummy in the baseline model. Div_Social has no significant relation, and Div_Occupational has a significant strong negative relation with CEO_Turnover_Dummy. The interaction term ROA*Div_Occupational shows a weak negative relation with the dependent variable CEO_Turnover_Dummy at the 5 percent significance level. Earnings_Per_Share*Div_Social*CEO_Power shows a weak positive significant relation with CEO_Turnover_Dummy.

CEO_Tenure and CEO_Age have a significant positive relation with the dependent variable at respectively the 1 and 5 percent level. Firm_Size shows a weak positive relation at the 10 percent level for Div_Social and Div_Occupational. There is not enough evidence to conclude a relation between the control variables CEO_Power, Board_Size, Indep_Boards, Growth and the dependent variable CEO_Turnover_Dummy. There are no significant differences between Div_Social and Div_Occupational regarding the influence of control variables on the dependent variable.

Table 9 depicts the results on the firm performance measure Earnings_Per_Share and the dependent variable CEO_Turnover_Dummy. Where Div_Social shows no significant relation, Div_Occupational shows a negative relation with the dependent variable at the 1 percent level. CEO_Power shows a negative relation with the dependent variable, for Div_Social at the 5 percent level and for Div_Occupational at the 10 percent significance level.

The interaction term Earnings_Per_Share*Div_Occupational shows a negative relation with the dependent variable CEO_Turnover_Dummy at the 5 percent significance level. Earnings_Per_Share*Div_Social*CEO_Power shows a strong positive relationship with CEO_Turnover_Dummy.

CEO_Tenure and CEO_Age show a positive relation with CEO_Turnover_Dummy for both Div_Social and Div_Occupational at the 1 percent level. Firm_Size shows for Div_Social a significant positive relation, and for Div_Occupational a weak positive relation with the dependent variable CEO_Turnover_Dummy. There is not enough evidence to conclude a relation between the control variables Board_Size, Indep_Boards and Growth and the dependent variable CEO_Turnover_Dummy.

I conclude that there is no relation between ROA*Div_Social and Earnings_Per_Share*Div_Social and the dependent variable so I reject hypothesis 2a. Both ROA*Div_Occupational and

(28)

27

Table 8 Logistic Regression model analysis

Dependent variable: CEO_Turnover_Dummy

(1) (2) (3) (4) (5) (6) Div_Social Div_Occupational constant −6.025 *** −5.905 *** −6.256 *** −5.494 *** −5.758 *** −6.507 *** (−7.21) (−7.01) (−7.10) (−6.50) (−6.75) (−6.32) ROA −0.476 −2.736 0.435 −0.429 7.033** 15.64** (−0.48) (−1.46) (0.16) (−0.43) (2.01) (2.20) Diversity −0.964 −1.743* 2.109 −3.500*** −1.697 1.695 (−1.11) (−1.67) (1.03) (−3.23) (−1.26) (0.61) CEO_Power −1.170 −1.074 −0.563 −1.136 −1.225 0.009 (−1.49) (−1.36) (−0.62) (−1.45) (−1.57) (0.01) ROA*Diversity 15.05 −23.66 −34.05** −73.97** (1.45) (−1.23) (−2.22) (−2.28) ROA*CEO_Power −10.03*** −12.70 (−2.97) (−1.50) CEO_Power*Diversity −8.043*** −5.429 (−2.69) (−1.38) ROA*Diversity*CEO_Power 85.420*** 58.830 (3.58) (1.49) CEO_Tenure 0.142** 0.135** 0.178*** 0.140** 0.148** 0.147** (2.25) (2.11) (2.73) (2.22) (2.36) (2.31) CEO_Age 0.063*** 0.063*** 0.065*** 0.068*** 0.067*** 0.067*** (7.40) (7.43) (7.57) (7.92) (7.81) (7.81) Board_Size −0.027 −0.028 −0.024 −0.026 −0.029 −0.029 (−0.88) (−0.91) (−0.76) (−0.84) (−0.95) (−0.95) Indep_Boards 0.834 0.820 0.574 0.572 0.510 0.475 (1.41) (1.38) (0.96) (0.96) (0.85) (0.79) Growth 0.087 0.106 −0.015 0.117 0.106 0.097 (0.60) (0.71) (−0.09) (0.81) (0.74) (0.63) Firm_Size 0.001* 0.001* 0.001* 0.001* 0.001* 0.001* (1.92) (1.80) (1.80) (1.67) (1.69) (1.69)

Year and Industry

Observations 2204 2204 2204 2204 2204 2204

(29)

28

Table 9 Logistic Regression model analysis

Dependent variable: CEO_Turnover_Dummy

(1) (2) (3) (4) (5) (6) Div_Social Div_Occupational constant −6.041 *** −6.179 *** −6.467 *** −5.506 *** −5.732 *** −6.071 *** (−7.24) (−7.32) (−7.37) (−6.52) (−6.71) (−5.91) Earnings_Per_Share −0.033 0.023 0.172* −0.034 0.170 0.047 (−1.44) (0.45) (1.65) (−1.50) (1.63) (0.20) Diversity −0.978 −0.261 3.314 −3.543*** −2.026 −0.330 (−1.12) (−0.25) (1.62) (−3.26) (−1.53) (−0.12) CEO_Power −1.171** −1.186** −0.775 −1.100* −1.133** −0.489 (−2.06) (−2.12) (−1.12) (−1.95) (−2.02) (−0.46) Earnings_Per_Share −0.425 −2.769*** −0.910** −0.961 *Diversity (−1.20) (−3.42) (−1.99) (−0.92) Earnings_Per_Share −0.198* 0.071 *CEO_Power (−1.70) (0.23) CEO_Power*Diversity −6.349** −3.358 (−2.16) (−0.87) Earnings_Per_Share 3.490*** 0.583 *Diversity*CEO_Power (3.57) (0.43) CEO_Tenure 0.144*** 0.145*** 0.146*** 0.139*** 0.142*** 0.123** (3.04) (3.10) (2.94) (2.96) (3.02) (2.38) CEO_Age 0.063*** 0.063*** 0.062*** 0.068*** 0.067*** 0.067*** (7.48) (7.47) (7.30) (8.00) (7.90) (7.80) Board_Size −0.023 −0.023 −0.019 −0.021 −0.026 −0.018 (−0.75) (−0.72) (−0.59) (−0.69) (−0.84) (−0.59) Indep_Boards 0.866 0.906 0.970 0.605 0.562 0.672 (1.46) (1.53) (1.61) (1.02) (0.94) (1.12) Growth 0.069 0.060 0.090 0.097 0.093 0.116 (0.48) (0.42) (0.61) (0.68) (0.65) (0.78) Firm_Size 0.001** 0.001** 0.001** 0.001* 0.001* 0.001* (2.07) (2.18) (2.30) (1.83) (1.81) (1.78)

Year and Industry

Observations 2204 2204 2204 2204 2204 2204

(30)

29

Earnings_Per_Share*Div_Occupational do show a strongnegative significant relation with the dependent variable CEO_Turnover_Dummy, therefore I can accept hypothesis 2b. Because I rejected hypothesis 2a, but accepted hypothesis 2b, I partially accept the second hypothesis.

However I cannot fully accept the second hypothesis, the results of the OLS models show that there are significant relationships with three control variables and the dependent variable CEO_Turnover_Dummy.

CEO Power

The interaction terms regarding CEO power (ROA*Div_Social*CEO_Power, ROA*Div_Occupational*CEO_Power, Earnings_Per_Share*Div_Social*CEO_Power and Earnings_Per_Share*Div_Occupational*CEO_Power) did not show a relation with the dependent variable LnCEO_Total_Compensation, therefore I reject hypothesis 3a and 3b.

Regarding the dependent variable CEO_Turnover_Dummy the interaction terms ROA*Div_Occupational*CEO_Power and Earnings_Per_Share*Div_Occupational*CEO_Power did not show a significant relation, so I have to reject hypothesis 3c. Since the interaction terms ROA*Div_Social*CEO_Power and Earnings_Per_Share*Div_Social*CEO_Power showed a strong positive significant relation with CEO_Turnover_Dummy, I can accept hypothesis h3d.

Because I accepted hypothesis 3d, but rejected hypotheses 3a, 3b and 3c I do not have enough support for the third hypothesis, and therefore I reject hypothesis 3.

Robustness test

Because I rejected the first and third hypothesis, and only partly accepted the second hypothesis, I included an extra firm performance measure for a robustness test. The variable I am using is return on equity (ROE). ROE is calculated as the net income divided by shareholders equity. With this new variable the Ordinary Least Squares model with LnCEO_Total_Compensation and the logit regression with CEO_Turnover_Dummy were re-run. Table 10 and 11 show the results of the regressions. As seen in table 10 Div_Social shows a positive, and Div_Occupational a significant negative relation with LnCEO_Total_Compensation at respectively the 5 percent and 1 percent level. The interaction terms regarding firm performance and diversity do not show a significant relation with the dependent variable. ROE*Div_Social*CEO_Power does not show a significant relation and ROE*Div_Occupational*CEO_Power shows a negative significant relation with the dependent variable at the 5 percent level.

(31)

30

(1) (2) (3) (4) (5) (6) Div_Social Div_Occupational constant 5.968 *** 5.983 *** 6.123 *** 6.274 *** 6.274 *** 6.265 *** (26.45) (26.49) (26.06) (27.83) (27.83) (24.22) ROE 0.006 −0.029 −0.084* 0.006 0.006 −0.160 (0.87) (−1.05) (−1.77) (0.92) (0.10) (−1.55) Diversity 0.462** 0.434* −0.599 −0.865*** −0.865*** −0.865 (2.06) (1.92) (−1.25) (−3.24) (−3.21) (−1.47) CEO_Power 0.144 0.125 −0.068 0.158 0.158 0.171 (0.98) (0.85) (−0.37) (1.08) (1.08) (0.72) ROE*Diversity 0.191 0.679** 0.003 0.814* (1.30) (2.18) (0.01) (1.70) ROE*CEO_Power 0.059 0.305** (0.77) (1.99) CEO_Power*Diversity 1.588** 0.053 (2.42) (0.06) ROE*Diversity *CEO_Power −0.609 −1.504** (−1.34) (−2.09) CEO_Tenure −0.005 −0.004 −0.007 −0.007 −0.007 −0.008 (−0.49) (−0.36) (−0.57) (−0.58) (−0.58) (−0.67) CEO_Age 0.002 0.002 0.002 0.002 0.002 0.002 (1.04) (1.03) (0.89) (1.13) (1.13) (1.08) Board_Size 0.071*** 0.072*** 0.072*** 0.077*** 0.077*** 0.077*** (8.80) (8.77) (8.77) (9.54) (9.54) (9.54) Indep_Boards 1.204*** 1.200*** 1.249*** 1.100*** 1.101*** 1.105*** (7.88) (7.85) (8.14) (7.16) (7.15) (7.18) Growth −0.117*** −0.120*** −0.115*** −0.114*** −0.114*** −0.114*** (−2.96) (−3.04) (−2.90) (−2.87) (−2.87) (−2.88) Firm_Size 0.002*** 0.002*** 0.002*** 0.002*** 0.002*** 0.002*** (18.92) (18.92) (18.91) (18.98) (18.98) (18.92) Dummy_Gender_CEO −0.006 −0.006 −0.017 −0.022 −0.022 −0.021 (−0.08) (−0.08) (−0.21) (−0.29) (−0.29) (−0.27) CEO_Duality −0.011 −0.011 −0.013 −0.010 −0.010 −0.012 (−0.28) (−0.29) (−0.33) (−0.26) (−0.26) (−0.31) Leverage 0.827*** 0.826*** 0.821*** 0.814*** 0.814*** 0.823*** (10.10) (10.08) (10.02) (9.93) (9.93) (10.03)

Year and Industry

Observations 2195 2195 2195 2195 2195 2195

The moderating role of CEO power on the relation between board diversity and board monitoring quality

Master thesis:

The moderating role of CEO power on

the relation between board diversity

and board monitoring quality

Contents

Abstract

Introduction

Background

Literature review & hypotheses development

The board of directors

Board diversity

Quality of monitoring

CEO power

Research methodology

Regression models

Dependent variables

Independent variables

Interaction terms

Control variables

Results

Descriptive statistics

Main results

CEO compensation

CEO turnover

CEO Power

Robustness test