The moderating effect of board characteristics on the CEO stock option and risk-taking relation.

(1)

The moderating effect of board characteristics on the

CEO stock option and risk-taking relation.

Master thesis, Controlling

Rijksuniversiteit Groningen, Faculty of Economics and Business

Date: 25/06/2018, Groningen

E. Davelaar

Student number: S3267911

Email e.davelaar@student.rug.nl

Word count: 11724

Thesis supervisor: Dr. R.C. Trapp

(2)

Acknowledgements

I would like to express my very great appreciation to my thesis supervisor Dr. R.C. Trapp for his help and efforts in the process of writing my thesis. Dr. R.C. Trapp was always available for questions and also provided feedback, even on short notice. The feedback and suggestions provided truly helped me to write and improve my thesis and for this I am very grateful. I also wish to thank my parents and girlfriend for their support and encouragement throughout my studies.

(4)

Abstract

Incentive alignment is a common method to align the interests of shareholders and CEOs. Especially stock option-pay is a common way of rewarding CEOs however, risk-taking literature has remained inconclusive about the effects of CEO stock options on risk-taking. The aim of this study was to increase the understanding of the CEO stock option and risk-taking relationship by combining the two literature streams of CEO stock options and board characteristics. The study focused on the effect of CEO stock options and risk-taking and how this relationship is moderated by various board characteristics. It uses a panel regression with a sample of 2,338 observations from S&P 500 companies. The findings suggest that the CEO stock options and risk-taking relationship is moderated negatively by the board characteristics gender diversity and the director’s stock options while the number of years until the director’s retirement moderate the relationship positively. This indicate that some board characteristics strengthen while other characteristics mitigate the relationship between CEO stock option pay and risk-taking. These results are of interests to researchers, boards of directors, regulators and shareholders.

(5)

1. Introduction

According to agency theory there is a conflict between the interests of CEOs and shareholders. The assumption is that CEOs pursue their own interests instead of the interests of stakeholders if there is no monitoring or no incentives present (Shleifer, & Vishny, 1996). This conflict between interests is related to the divergent risk preferences of shareholders and CEOs (Jensen, & Meckling, 1976).

Shareholders are generally interested in maximizing returns, and these large gains require large risks to be taken (Sanders, & Hambrick, 2007). Shareholders have the ability to reduce their exposure to risks by holding a diversified portfolio of stocks (Fama & Jensen, 1983). This diversified risk portfolio makes shareholders risk-neutral (Sanders, & Hambrick, 2007). The CEO’s wealth, however, is largely dependent on the success of the company and their own job security, making them more risk-averse (Fama & Jensen, 1983). CEOs might also be reluctant to pursue a high risk profile because if things go wrong it might hurt their reputation and future career prospects (Antia, Pantzalis, & Park, 2010). Therefore, it is assumed that CEOs are more risk-averse than shareholders and that compensation structures can be used to align these two risk interests (Sanders, 2001;Misangyi, & Acharya, 2014). Long-term incentives, such as stock compensation, are used to achieve risk alignment between shareholders and the CEO (Devers, McNamara, Wiseman, & Arrfelt, 2008). While stock compensation incentivizes CEOs to take more risks, CEO-shareholder alignment is not fully resolved by stock compensation (Devers et al., 2008). The evidence provided within the literature about how CEO stock compensation affects firm risk-taking remains inconclusive (Baixauli-Soler, Belda-Ruiz, & Sanchez-Marin, 2015). For instance, some studies have found that CEO stock compensation increases risk-taking (Sanders & Hambrick, 2007; Sanders, 2001; Devers et al., 2008). Some scholars even argue that CEO stock compensation pay may result in excessive risk-taking (Sanders, 2001). In contrast, other scholars contend that stock compensation decreases risk-taking (Larraza-Kintana, Wiseman, Gomez-Mejia, & Welbourne, 2007; Sawers, Wright, & Zamora, 2011). In short, existing literature remains inconclusive about the effect of CEO stock compensation on risk-taking.

A general trend related to corporate governance research may help to solve the ambiguity in risk-taking literature. Scholars argue that the ambiguity in corporate governance research might be explained by the focus on the independent effects of mechanisms (Misangyi, & Acharya, 2014; Tihanyi, Graffin, & George, 2014; Oh, Chang, & Kim, 2016). Therefore, research that considers corporate governance mechanisms as interactive has recently gained increased attention (Tihanyi et al., 2014). Companies use several corporate governance mechanisms at the same time to mitigate the agency conflict, yet most corporate governance research to date has taken an isolated perspective. Understanding how corporate

(6)

governance mechanisms interact with CEO compensation and risk-taking may decrease ambiguity in this field.

Scholars suggest that corporate governance mechanisms complement (substitute) for each other, which indicates that the presence of one corporate governance mechanism amplifies (decrease) the marginal effect of another corporate governance mechanism and therefore interact (Misangyi, & Acharya, 2014; Tihanyi et al, 2014; Oh, et al., 2016). Interactive effects might also be present between CEO compensation and risk-taking. Several authors argue that board characteristics affect risk-taking (Bargeron, Lehn, & Zutter, 2010; Bradley, & Chen, 2015; Huang, & Wang, 2015), and these studies have focused on how such characteristics directly influence risk-taking. As such, these studies did not consider the potential interaction between mechanisms; how board characteristics and CEO compensation could be interactively related in risk-taking has not yet been considered.

Such an interactive effect can be explained in the following way. The economic incentive provided by stock compensation may drive CEOs to take more risks. Especially stock option pay can increase the risk-taking of a previously risk-averse CEO (Coles, Daniel, & Naveen, 2006; Sanders, 2001). Since stock options only offer rewards if a certain threshold is met and the reward could be unlimited depending on the actual stock price on expiration date, the CEO is incentivized to adopt more risky decisions to increase the stock price. However, monitoring by the board of directors could indirectly influence this risk-taking, while the board does not have the ability to directly impact risk-taking because it is not involved in day-to-day decision-making, it can influence the CEO’s risk-taking through strategic decision-making. Consequently, if the board has a high risk-aversion, monitoring could decrease the marginal effect of the CEO incentive alignment and risk-taking. In comparison, a low risk-averse board may increase the risk-taking of the CEO. For example, when the CEO is incentivized through stock options to pursue a certain risk profile, the board of directors may persuade the CEO towards an even higher risk profile, thereby increasing the marginal effect of CEO stock options on risk-taking. In short, the board of directors can indirectly increase or decrease the risk-taking of the CEO through strategic decision-making, and therefore these mechanisms might complement or substitute each other.

This study therefore adds to prior literature because it is concerned with how board characteristics indirectly influence the CEO stock option and risk-taking relationship and therefore considers how these mechanisms may interact. I argue that different board characteristics that determine the board of directors’ risk-aversion can therefore influence the incentive alignment and risk relationship in an indirect way. This study is concerned with the most important board characteristics related to risk-aversion. Therefore, my research question is:

“Do different board characteristics moderate the CEO stock option and CEO risk-taking relationship?”

(7)

I draw on the complementary view to study how stock options are interactively influenced by board characteristics such as the gender diversity of board members, the board’s independence, the directors’ stock option pay and the retirement age of directors, and how these may interact with risk-taking. This study focuses on gender diversity as scholars argue that women tend to be more risk-averse in comparison to their male counterparts (Adams, & Ferreira, 2009; Chen, Crossland, & Huang, 2016). Another important characteristic that has received considerable attention is board independence; independent boards may have less knowledge about what is actually going on in the firm. Therefore, due to regulatory compliance it could be argued that these outside directors are less willing to take risks (Pathan, 2009). The time to retirement of board members could also impact the board’s risk-aversion. According to (Fama and Jensen, 1983), directors are motivated to perform their role of monitoring by the reputation gained from the market of directors. However, this may not hold true for directors close to retirement age, as these individuals could be less concerned with their reputation and future board memberships. The effect of providing stock options to directors is also considered in this research, since stock pay for directors might also influence the boards’ level of risk-aversion.

The study uses a S&P 500 sample of the years 2007–2017. The results of my study indicate that CEO stock options increase risk-taking. Furthermore, consistent with my expectations I find that gender diversity mitigates the effect of CEO stock options on risk-taking. The number of years until directors’ retirement strengthens the effect of CEO stock options and risk-taking. Contrary to my predictions, my findings indicate that providing directors with stock options decreases the effect of CEO stock option pay on risk-taking, which indicates that these mechanisms might be substitutes. There is no evidence for a moderating effect of director independence found within this study.

This study contributes to prior corporate governance literature by considering stock options with a moderating effect of board characteristics on risk-taking and makes the following contribution. As previously mentioned, existing literature remains ambiguous about how incentive alignment affects risk-taking (Baixauli-Soler et al., 2015). This study may advance our understanding of the relationship between CEO stock option pay and risk-taking by considering corporate governance as interactive mechanism, as advocated by recent research (Tihanyi et al., 2014; Misangyi, & Acharya, 2014; Oh et al., 2016). Considering that these mechanisms may interact, could help to develop a better understanding of how the incentive alignment and risk-taking relationship is interrelated, and could therefore explain the ambiguity in current risk-taking literature. Consequently, this study contributes to the literature on risk-taking and the discussion of gender quota on boards (Adams and Ferreira, 2009), director compensation and board composition in general by illuminating the effects of these characteristics. This study may provide insights about how these characteristics might moderate the relationship of stock options and risk-taking.

(8)

From a practical perspective this research is important when designing corporate governance systems because board characteristics could have a decreasing or increasing effect on the incentivized risk relationship. Therefore, considering these board characteristics when designing a corporate governance system could be critical. Scholars have argued that understanding these board characteristics is a crucial step towards better understanding how boards function (Adams, Licht & Sagiv, 2011). For instance, when stock options are used to increase risk-taking it is important to consider the effects of specific board characteristics as well, as this might increase the resulting risk-taking and could even lead to excessive risk-taking. In contrast, other board characteristics could decrease risk-taking as a result of the board’s own risk-aversion, decreasing the intended effect of stock options. Designing a good corporate governance system could benefit shareholders, and understanding which board characteristics could influence the incentive alignment and risk-taking relationship could be crucial in designing effective corporate governance. Consequently, this study could benefit shareholders since prior research has found that good corporate governance leads to higher stock returns (Gompers, Ishii, & Metrick, 2003).

The remainder of this paper is organized as follows. Section 2 provides the theoretical background of risk-taking literature regarding executive compensation and board characteristics, at the end of this section my hypotheses are formed. In the section 3 the method is addressed in which in-depth information is provided about the sample, variables and the data collection. The results of this paper are provided in section 4. Within section 5 the discussion and conclusions can be found, which provide theoretical and practical implications, the limitations of the study and recommendations for future research.

2. Literature review and hypothesis development

2.1.1 Executive compensation and risk-taking

As privately-owned companies have a great impact on the overall economy (Fauver, Hung, Li, & Taboada, 2017), understanding the investment policy of CEOs is important. According to agency theory, the risk-aversion of shareholders and CEOs differs (Jensen, & Meckling, 1976). Incentive alignment is one of the most important tools in corporate governance to mitigate this divergent risk interest and as such has received considerable attention (Dalton, Hitt, Certo, & Dalton, 2007). One of the most important consequences regarding the difference in risk-aversion is that a CEO might pursue a suboptimal investment policy from the shareholders’ perspective. This problem arises because the CEO and shareholders have different risk preferences. As described in the introduction, shareholders are generally interested in maximizing returns, and such large gains require risk to be taken (Sanders, & Hambrick, 2007). The divergent risk preferences comes from the shareholders’ ability to diversify risks. Shareholders can easily divide their risk by splitting their investments between different stocks, which

(9)

makes them less dependent on a single company. Due to this diversification shareholders are assumed to be risk-neutral (Sanders, & Hambrick, 2007). As CEOs do not share this risk-neutral position, since they do not have the ability to diversify their risk, they are assumed to be more risk-averse. The CEO’s wealth is determined particularly by the company and the chances of losing his or her job (Fama & Jensen, 1983).Furthermore, the CEO is concerned with personal reputation and future managerial career prospects, and therefore pursues lower-risk policies that have lower chances of failure (Antia, et al., 2010).

According to Misangyi and Acharya (2014), compensation structure can be used to align the risk interests of CEOs and shareholders. Scholars have long argued that stock-based incentives could be granted to CEOs to align risk interests (Jensen, & Murphy, 1990) because such incentives encourage the CEO to focus on value creation, which is in the shareholders’ interests. According to Ndofor, Wesley and Priem (2015), incentive alignments consist of stock-based incentives. These stock rewards could increase wealth-maximizing behavior and could therefore lead to a higher level of profitability for the firm.

An important contribution to this discussion was made by the study of Sanders (2001), which drew a distinction between stock and stock options. The study demonstrated that stock and stock option pay could lead to opposite outcomes. Sanders found that stock pay leads to lower acquisitions, which reflects decreased risk-taking. Stock option pay, however, leads to more acquisitions, which reflects increased risk-taking. This is the result of the different incentives that stock and stock options provide. Stock pay directly influences a CEO’s wealth because a decline in share price will have a direct effect on his or her portfolio. This could lead to an increase in risk-aversion because the CEO’s wealth is already highly dependent on the company. Stock option pay, however, incentivizes the CEO to take more risks, because stock option pay contains a limited downside risk and unlimited upside potential (Sanders, 2001). Therefore, stock option pay could be used to increase risk-taking of a risk-averse CEO as opposed to stock pay. Sanders’ study makes an important contribution because prior to this research both stock and stock option pay where assumed to align risk preferences. Following Sanders’s rationale this study is focused on stock option pay.

As mentioned in the introduction, at present research remains inconclusive regarding how CEO stock option pay affects taking. Some authors find that CEO stock option pay may result in more risk-taking (Sanders & Hambrick, 2007; Sanders, 2001; Devers et al., 2007). For instance, Sanders and Hambrick (2007) argue that the effect of CEO stock option pay might be explained by the CEO’s incentive to invest in more risky projects to raise share prices in the interest of personal gain. Other scholars argue that CEO stock option pay decreases risk-taking (Larraza-Kintana, et al., 2007; Sawers, et al., 2011). The behavioral agency model can be used to explain this effect. According to Larraza-Kintana et al. (2007), stock options are valued by CEOs at their “intrinsic value” and therefore the value

(10)

of options may become perceived as part of the total wealth of the CEO, just like stocks. This is because up to 80% of stock options end in a positive pay out (Murphy, 2002). Therefore, as CEOs expect that these stock options have value and as they are perceived as part of the CEO’s own wealth, it increases their aversion. Consequently, as the value of stock options increases the CEO becomes more risk-averse. Therefore, when CEOs are provided with stock options they will only be incentivized to take more risks when stock options are not in a profit state yet. As a result, the effect on risk-taking is not static but changes over time. Another line of reasoning for either a positive or negative effect of stock options is that the effect of CEO stock option pay may be moderated by other factors. For example, Baixauli-Soler et al. (2015) argue that there might be top management characteristics, such as gender diversity, which moderate the relationship between stock option pay and risk-taking. Which indicates that the CEO stock option and risk-taking relationship might be more complex. Both top management and board characteristics have received considerable attention in studies of risk-taking.

2.1.2 Board characteristics and risk-taking

The board of directors is responsible for monitoring the decisions of a CEO (Fama, & Jensen, 1983). The assumption is that without external monitoring or incentive structures, CEOs will pursue their self-interest instead of the self-interest of stakeholders (Shleifer, & Vishny, 1996). Despite the positive effects of incentive alignments there are some scholars who argue that in some cases, incentive alignment results in unwanted incentives, for instance excessive risk-taking (Sanders, 2001). Therefore, incentive alignment should be complemented by monitoring. Agency problems arise because of the information asymmetry between shareholders and executives. According to Ndofor et al. (2015), information asymmetry could lead to CEOs deviating from shareholders’ preferred risk-taking. Because of the lack of information on the part of the shareholders, the CEO can engage in self-serving behavior, such as deviating from the risk preferences of shareholders, with relative ease (Richardson, 2000; Ndofor, et al. 2015). According to Hillman and Dalziel (2003), the most used assumption of agency theory is that effective monitoring can reduce agency costs because monitoring by the board of directors could decrease the possibility for the CEO to deviate from the risk interests of shareholders.

The board has several opportunities to perform their role as monitors. For instance, the board of directors is involved in deciding on strategic issues for companies (Adams et al., 2010). The board participates in the process of formulating, evaluating and approving firm strategies (He, & Huang, 2011). Secondly, the board of directors can appoint a CEO, fire a CEO, evaluate the performance of a CEO and determine compensation (Bednar, 2012). Therefore, some authors claim that the board of directors is responsible for making sure the company creates value for their shareholders (Adams, 2016). This might explain why boards and their characteristics have received so much attention in studies of corporate governance. For example, Adams et al. (2011) argued that understanding board characteristics is a crucial step towards understanding these boards’ function.

(11)

Board characteristics are also considered important in taking. Prior literature has focused on risk-taking and board compensation (Adams, & Ferreira, 2008; Dah & Frye, 2017), the duality function of the CEO (Adams et al., 2010), board independence (Huang, & Wang, 2015; Bargeron et al., 2010; Adams et al., 2010) and board size (Wang, 2012; Ferrero-Ferrero, Fernández-Izquierdo, & Muñoz-Torres, 2012). Since a board’s risk-aversion could be dependent on these different characteristics, understanding how these board characteristics are related to risk-taking is important. The aforementioned studies focused on the isolated effects of these board characteristics on risk-taking. Other board characteristics may also play an important role. For example, Adams et al. (2010) recognized that the incentives of individual board members might differ.

2.1.3 Interactive effect of board characteristics and CEO stock option pay

Most research conducted concerning CEO compensation and risk-taking has not considered additional mechanisms and thus assumes independent relationships between mechanisms. This type of research does not consider that additional mechanisms might interactively influence risk-taking. Additional mechanisms may increase or decrease the relationship between CEO stock option pay and risk-taking. I argue that the interactive effect of other mechanisms might explain why the CEO compensation and risk-taking literature remains inconclusive.

Recently, scholars have suggested that corporate governance mechanisms should be considered as a bundle rather than in isolation. Multiple scholars argue that corporate governance mechanisms may complement each other (Hoskisson, Castleton, & Withers, 2009; Schepker & Oh, 2013; Oh et al., 2016), suggesting that the presence of one corporate governance mechanism amplifies the marginal effect of another corporate governance mechanism on risk-taking. Other scholars (Rediker, & Seth, 1995; Ward, Brown, & Rodriquez, 2009; Oh et al., 2016) posit that corporate governance may have a substitutive effect, meaning that the presence of one mechanism could also decrease the marginal effect of another mechanism because multiple corporate governance mechanisms might interactively influence the organizational outcome (Oh et al., 2016). These mechanisms interactively influence each other and can be seen as synergetic. The effect of the bundle of mechanisms is thus more than just the sum of independent mechanisms (Misangyi, & Acharya, 2014). Considering that different corporate governance mechanisms interactively influence each other is important because in practice, corporate governance mechanisms work as a bundle rather than in isolation.

Drawing on the complementary versus substitute framework, mechanisms may also act as substitutes or complements in risk-taking. This is one of the most salient issues in previous risk-taking research because it focuses on the assumption that mechanisms work in isolation. Therefore, studies that use this framework do not consider the interactive effects between mechanisms. For instance, providing the CEO with stock options does not mean the CEO can pursue any risk policy he or she is incentivized to follow. Monitoring by the board could influence a CEO’s policy choices as this monitoring may set a limit.

(12)

As previously mentioned the board has several tools to influence the risk-taking of the CEO in an indirect way by approving or rejecting a CEO’s plan, how the board reacts to this plan might depend on their level of risk-aversion. Monitoring by the board of directors may interact with the incentive of stock option pay. Providing the CEO with stock options might have an impact on risk-taking, which can be mitigated or strengthened depending on the board’s risk preference, which in turn might be determined by specific characteristics. Consequently, considering this interactive effect of board characteristics on risk-taking could be fruitful for further research on risk-taking. Understanding how board characteristics moderate the relationship between CEO stock options and risk-taking could be crucial as it could mitigate two problems. First, if a CEO is provided with stock options, the resulting incentive is to pursue more risky policies. Through monitoring, the board of directors is assumed to counterbalance any deviation from the risk preference of shareholders. However, if the board allows an overly high risk profile, the incentive of stock options for the CEO could lead to excessive risk-taking. Secondly, the intended incentive provided by stock options (i.e. for the CEO to pursue a higher risk profile) could be mitigated by an overly risk-averse board. Both cases might ultimately lead to deviating from the risk preferences of shareholders. Therefore, understanding how these board characteristics interactively influence risk-taking is important. Furthermore, from a theoretical perspective, such an understanding might help to explain the ambiguity in the current CEO stock option and risk-taking literature.

Consequently, it could be argued that these board characteristics could interactively affect risk-taking, as the board sets the norm for the risk-taking of the CEO and therefore these characteristics could either have a complimentary or a substitutive effect. However, to the best of my knowledge, there has no research to date that has focused on how board characteristics interact in CEO stock options and risk-taking. Therefore, this study aims to fill this gap in research and is concerned with how the most important board characteristics affect the CEO stock option and risk-taking relationship.

2.2 Hypothesis development

2.2.1 CEO stock options and risk-taking

Some scholars argue that paying executives with stock options is crucial for compensation packages (Murphy, 2013). A stock option gives its owner the right to buy a share of stock at a pre-specified, fixed price in a pre-specified period (Hall, & Murphy, 2003). Since the owner of a stock option has the opportunity but not the obligation to exercise this right, the minimum value of an option is zero. However, the maximum value of the option is unlimited, as this is dependent on the difference between the strike price and current share price. Consequently, stock options encourage the CEO to take more risks, which could better align the interests of shareholders and CEOs. According to Sanders (2001), the effect of stock options can be explained by prospect theory. According to prospect theory, people tend to respond to a loss more strongly than to a profit (Tversky, & Kahneman, 1991). If the CEO is provided with stock options, he or she has nothing to lose and can only gain (Sanders, 2001). Therefore, providing

(13)

stock options could decrease the CEO’s risk-aversion. Consequently, considering their unlimited upside and limited downside potential, stock options could be used when attempting to increase the risk-taking of a risk-averse CEO. This leads to the following hypothesis:

H1: CEO stock options are positively related to CEO risk-taking.

2.2.2 Board independence

Boards of directors are faced with litigation risks (Bargeron et al., 2010), which may influence a board’s risk-aversion. This may be especially true for independent directors. This can be explained by considering the difference between independent directors, also called “outside directors,” and non-independent or “inside directors.” Outside directors are directors who did not previously work in the company and therefore are considered to be independent (Hillman, & Dalziel, 2003). Inside directors currently hold or previously held a position within the company before joining the board. It can be expected that inside directors have considerably more information and experience related to the firm than outside directors.

The information and experience of inside directors may allow for better judgement regarding the decisions and (strategic) proposals that the CEO makes. It is likely that outside directors will lack this information and experience (Bargeron et al., 2010; Adams et al., 2010). Due to the liability concerns that a board faces, outside directors may be more risk-averse than inside directors. This can be explained by the greater information asymmetry for outside directors compared to inside directors (Pathan, 2009). Outside directors have to make a judgement that involves a higher level of uncertainty than inside directors. Considering the fact that directors are faced with litigation risks, this could lead to a higher aversion of an outside director compared to an inside director. Because of this increased risk-aversion a more independent board may be more reluctant to approve risky projects. Therefore, I argue that while CEOs are encouraged to take more risks when provided with stock options, a more independent board may mitigate this effect. Therefore, the positive relationship of CEO stock options on CEO risk-taking may be mitigated in cases where board independence is high. This leads to the following hypothesis:

H2: The positive relationship between CEO stock options and CEO risk-taking is weaker when board independence is high.

2.2.3 Gender diversity of board members

There is growing attention directed towards the effects of a gender-diverse board (Jizi, & Nehme, 2017). In fact, gender diversity has recently been one of the most frequently considered board characteristics for both policy makers (Adams, 2016) and researchers (Baixauli-Soler, et al., 2015). In 2007, only 14.8% of the board positions for S&P 500 companies were held by women (Catalyst, 2007). In 2016, this percentage rose to 21.2% (Catalyst, 2016). Some countries have even adopted quotas for female board

(14)

memberships. Considering the assumption that women and men behave differently with regard to risk-taking, it is important to understand the effect on risk-taking that this increasingly female board membership has had.

Female board membership may affect risk-taking as it might influence the boards’ risk-aversion. It is assumed that females tend to be more risk-averse, and scholars have indeed found evidence that this is the case (Sapienza, Zingales, & Maestripieri, 2009; Faccio, Marchica & Mura, 2016; Chen, et al., 2016; Loukil, & Yousfi, 2016). Some scholars argue that this increased risk-aversion can be explained by psychological differences between men and women (Adams, 2016). For example, Niederle and Vesterlund (2007) argue that women do not have the same confidence as their male counterparts. Furthermore, Andersen, Ertac, Gneezy, List and Maximiano (2013) argue that women are not as competitive as men. Because they are more competitive, men might take more risks in order to outperform their competitors in attempting to become the “winner.” Despite empirical findings, whether and how a gender-diverse board decreases risk-taking remains inconclusive. Some scholars argue that this ambiguity could result from the fact that the women on boards are not “ordinary” women in society (Adams, 2016). However, as of today there have been no empirical results provided to support this claim.

The ambiguity in the results of studies related to gender diversity might also be explained by the focus on isolated effects of gender diversity. It could be argued that gender diversity may influence the effect of stock options on risk-taking in a similar manner as board independence. The influence of gender diversity on risk-taking could be dependent on other mechanisms. This can be explained as follows. When stock options are lacking, the CEO is presumed to be more risk-averse and the board of directors can only influence risk-taking indirectly by setting the risk norm. Therefore, the effect of gender diversity may not be observable when the CEO is also risk-averse (i.e. when no stock options are granted to the CEO). This is because in this case, both the CEO’s and the board’s risk-aversion might be aligned. However, providing CEOs with stock options could result in a divergence in risk-aversion between the board and the CEO. This may especially hold true for a gender-diverse board because female board members are presumed to be more risk-averse. Consequently, I argue that CEOs are encouraged to take more risks when provided with stock options, but that this effect may be mitigated as gender diversity on the board increases. Therefore, the positive relationship of CEO stock options on CEO risk-taking may be decreased in cases where gender diversity is high, leading to the following hypothesis:

H3: The positive relationship between CEO stock options and CEO risk-taking is weaker when a board’s gender diversity is high.

2.2.4 Incentive alignment for board members

The directors themselves can be considered agents acting on behalf of the shareholders as principals (Adams et al., 2010). Therefore, the board could be incentivized to perform the role of actually

(15)

monitoring the CEO according to the expectations of the shareholders. Incentive alignment for a board of directors is therefore also commonly used to align the interests of the directors and shareholders. For instance, a study conducted by Bryan and Klein (2004) concluded that almost 70% of the compensation that directors receive in Fortune 500 companies includes incentive pay. Other scholars have also found that directors seem to respond to financial incentives to attend board meetings, which in turn could affect monitoring (Adams & Ferreira, 2008).

However, it might be important to consider that providing directors with stock options may influence their risk-aversion in the same way it does for CEOs. As previously argued, this is because of the characteristics of stock options, i.e. a limited downside potential and an unlimited upside potential (Sanders, 2001). As a consequence, a director’s risk-aversion may be decreased when he or she is provided with stock options, which in turn might influence strategic decision-making and directors’ authorizations to decrease or increase CEO risk-taking. The board could, for example, approve more risky projects when they are incentivized with stock options because they are encouraged through their own economic benefit to do so therefore, allowing CEO’s, which are provided with stock options to execute more risky projects. Ultimately, this might even lead to excessive risk-taking because both mechanisms are focused on increasing risk-taking rather than counterbalancing one another. Therefore, the positive relationship between CEO stock options and CEO risk-taking may be strengthened in cases where stock options for directors are high. This leads to the following hypothesis:

H4: The positive relationship between CEO stock options and CEO risk-taking is stronger when board stock option pay is high.

2.2.5 Time until retirement of board members

According to Fama and Jensen (1983), the wider market for a board of directors functions as an important incentive. Within this market, future board memberships are dependent on the reputation an individual gains in previous monitoring roles (Adams et al., 2010; Levit & Malenko, 2016). This is an important incentive because as previously mentioned, agents are considered to be self-interested. This market for board of directors might incentivize self-interested board members to perform the role of monitoring the CEO adequately, which also involves monitoring of any deviation from the shareholders’ risk interests, which could potentially harm the reputation of the board of directors.

However, the incentive for future board memberships may be of lower importance for directors who are close to retirement age, as a considerable share of these board members might be less interested in future board membership. Consequently, it could be argued that directors who are close to retirement age are potentially willing to take more risks. Additionally, these directors may have been working for a considerable number of years and therefore may also be less economically dependent on future board memberships. Furthermore, directors who are close to retirement age may not expect to be present in the future when a risky decision actually impacts the firm, which might in turn decrease their

(16)

risk-aversion. This decreased risk-aversion could foster possibilities for the CEO to take risks when he is provided with stock options. Therefore, the positive relationship between CEO stock options and CEO risk-taking may be strengthened in cases where the time before retirement of the board decreases. This leads to the following hypothesis:

H5: The positive relationship between CEO stock options and CEO risk-taking is stronger when the board is closer to retirement age.

Figure 1 presents a conceptual model of the aforementioned hypotheses. Figure 1. Conceptual Model

3. Method

3.1 Sample

The archival research approach was used to analyze the moderating effect of board characteristics on the relationship between CEO stock options and risk-taking. This method allows for the use of a broad sample and statistical generalization. In order to test the proposed hypotheses I used S&P 500 companies because U.S. companies have a shareholders view (Bodnar, Jong, & Macrae, 2003), in which value maximizing, and therefore considerable risk-taking, is deemed important.

To increase the relevance of the study the most recently available data has been used. Furthermore, as some years might be biased due to specific time-related events, a time period of 11 years was chosen. The dataset spans the years 2007-2017. This dataset consists of companies included in the S&P 500 in at least one of the years during this period. The database started with 4,730 firm years after including the dependent and independent variables the database resulted in 3,413 observation and after merging and removing missing values of the control variables it resulted in a panel data set of 2,338firm years.

CEO stock options Risk-taking

H1

₊

H2: Director independence H3: Director gender diversity H4: Director stock options H5: Director years to retirement

H2: - H3: - H4: + H5: +

(17)

16

3.2 Independent variables

The variable OPTIONCEO is the value of stock options, derived from BoardEx. This variable contains the value of stock options awarded to the CEO during the fiscal year (Sanders, 2001). In most studies, the valuation of CEO stock options is derived by using one of the two common valuation methods. The methods for valuing stock options include the Black Scholes valuation method (Hou, Priem, & Goranova, 2017; Rajgopal, & Shevlin, 2002; Hall, & Murphy, 2000) and the present value method (Sanders, 2001). The Security Exchange Commission (SEC) approves both methods. Furthermore, these methods are also highly correlated (Sanders, 2001). I used the Black Scholes model to calculate the value of options. As mentioned before these methods are highly correlated and therefore, I do not expect that this choice results in any bias. The value of the stock options were calculated using a 10-year U.S. T-bill, which has a risk-free rate of 6.5%. The stock options were calculated as a percentage of the total salary of the CEO because the effect of stock pay may be dependent on other payments received (Sanders, 2001). The total salary consists of cash bonuses, long-term incentives and the base salary.

3.3 Dependent variables

There are many different specifications of the concept of risk-taking, which serves as a dependent variable in many studies. Capital expenditure is used in this study as an indication of risk-taking (Hayes, Lemmon, & Qiu, 2012; Sanders, & Hambrick, 2007; Bargeron, et al., 2010) because a CEO could choose a less risky alternative like distributing dividends or holding cash instead of investing. CAPEX was calculated as the total capital expenditure during a year divided by the total property, plant and equipment (Christensen, Dhaliwal, Boivie, & Graffin, 2015) both where derived from Compustat. For robustness purposes, a second risk-taking proxy, stock return volatility, was also used. Risky decisions could either result in a bigger gain or a bigger loss (Sanders, & Hambrick, 2007), reflected in either a higher or a lower stock price. These bigger win or loss in turn increases the volatility of the share. Therefore, increased risk-taking could be reflected in stock return volatility (Guay, 1999). STOCKRETURN was calculated using the 12-month average of the standard deviation of value-weighted returns (including dividends) per month, and was derived from CRSP. Also for robustness purposes, standard deviation of Return on Assets (ROA) (Phatan, 2009) was used as a proxy for risk-taking. ROA was calculated using the standard deviation of ROA for the company during the current and the following year and was derived from Compustat.1

3.4 Moderating variables

Board independence is measured by DIRINDEP, which was calculated as the percentage of independent board members divided by the total number of members on the board. In doing this, I drew on the relative rather than the absolute value of independence. This is because it could be argued that the more

1_{Acquisitions are also a commonly used to measure risk-taking. For example, a study conducted by Sanders}

(2001) focused on acquisitions. However, acquisitions are not always the type of risk-taking that shareholders want because they have a high chance of failure and therefore destroying value (Sanders & Hambrick, 2007).

(18)

highly a certain characteristic is represented on the board the higher the chances of actually observing its effects. Gender diversity of the board is measured as DIRECTORGEN. This variable consists of the gender diversity ratio of the full board. The third moderating variable that measures directors’ stock options is OPTIONDIR. This variable was calculated as the value of stock options according to the Black Scholes valuation method (Hou, et al., 2017; Rajgopal, & Shevlin, 2002; Hall, & Murphy, 2000). The value of the stock options was calculated using a 10-year U.S. T-bill, which has a risk-free rate of 6.5%. Stock options were measured as the absolute value of stock options instead of based on relative values to total pay because the study of Adams and Ferreira (2008) concluded that directors seemed to respond to a relatively low financial incentive (of only $1,000) to attend board meetings. The fourth moderating variable, a director’s years until retirement, is YEARTORETIR, which was calculated using an assumed retirement age of 70 (Adams and Ferreira, 2008). The average age of directors was subtracted from 70 to calculate the average years until retirement for directors. All data for the moderating variables was derived from BoardEx.

3.5 Control variables

There are several control variables included in the model. First, a control variable was included to control for the size of the company, as company size has been associated with corporate risk-taking (Miller and Wiseman, 2002). COMPANYSIZE was measured by the logarithm of total assets (Ferrero-Ferrero, et al., 2012). Scholars have argued that CEO power leads to increased risk-taking (Lewellyn, & Muller-Kahle, 2012). To control for the effects of CEO power, the control variable DUALITYCEO was included because CEOs who hold a dual position, in which the CEO is also chairmen of the board, may influence the board’s decision-making power. Scholars have found that CEO duality decreases the influence the board has in decision-making (Dah & Frye, 2017; Krause, Semadeni, & Cannella, 2014). Therefore, when CEOs perform a dual role, the specific effects related to board characteristics might be influenced. DUALITYCEO was included as a dummy variable, with 1 used for non-dual CEOs and 0 used for duality. BOARDSIZE was also included, as larger boards are associated with lower risk-taking (Nakano, & Nguyen, 2012). As previously argued, gender may influence risk-aversion of board members, but this might also be the case for CEOs as well. Prior studies have acknowledged that a CEO’s gender is associated with risk-taking (Khan, & Vieito, 2013). Therefore, the control variable GENDERCEO was included to control for CEO gender-related effects. GENDERCEO is a dummy variable where 1 represents a male CEO and 0 represents a female CEO. Since scholars argue that risk-taking is increased when loss situations occur (Larraza-Kintana, et al., 2007) a control variable was included for profitability, as profitability may impact actual risk-taking. PROFITABILITY was measured as a natural logarithm of ROA. Furthermore, year dummies were included as some year-related events might influence the CAPEX expenses. One such specific year event could be the financial crisis, which is covered in the sample period. Industry dummies were also included, as some industries might highly influence the CAPEX expenditures for companies. The control variables SIZE and

(19)

PROFITABILITY where derived from Compustat while GENDERCEO, DUALITYCEO, BOARDSIZE, INDUSTRYDUMMY and YEARDUMMY where derived from Boardex. See Table 1 for a complete overview of the variables and their calculations.

Table 1. Variable measurement

Variable name Type Proxy Measurement Source

OPTIONCEO Independent variable

CEO stock options

Fair value of stock options provided during a year using the Black Scholes valuation method divided by total compensation

BoardEx

STOCKRETURN Dependent variable Risk-taking Standard deviation of stock returns including dividends over 12-month period

CRSP

CAPEX Dependent variable Risk-taking Capital expenditure divided by property plant and equipment

Compustat ROA Dependent variable Risk-taking Standard deviation of net income

divided by total assets

Compustat

DIRINDEP Moderating variable Board independence

Independent directors according to BoardEx, divided by total directors

BoardEx

DIRECTORGEN Moderating variable Board gender diversity

Gender ratio during a year; both executive and non-executive directors

BoardEx

OPTIONDIR Moderating variable Board stock option pay

Fair value of stock options provided during a year; both executive and non-executive directors

BoardEx

YEARTORETIR Moderating variable Board retirement age

The retirement age of board of directors is assumed to be 70 years; calculated as the current age subtracted from 70

BoardEx

SIZE Control variable Size of the company

Natural logarithm of total assets Compustat PROFITABILITY Control variable Profitability of

the company

Natural logarithm of ROA Compustat GENDERCEO Control variable CEO gender Gender of the CEO: 1 = male 0 =

female

BoardEx DUALITYCEO Control variable CEO duality Dummy variable if CEO held a dual

position during the fiscal year; 0 = no duality 1 = duality

BoardEx

BOARDSIZE Control variable The total size of the board

The total size of the board

BoardEx INDUSTRYDUMMY Control variable Industry

effects

The industry based on the two digit

SIC code BoardEx

YEARDUMMY Control variable Year effects The year effects based on the fiscal

(20)

4. Results

4.1 Descriptive statistics and correlation matrix

Table 2 shows the descriptive statistics of the variables. As can be derived from the table, on average 22% of the total salary of the CEOs in the sample consists of stock options. However, some CEOs have salaries consisting of up to 87% stock options.

Table 2. Descriptive statistics

Obs. Mean Std. Dev. Min Max

Independent variable OPTIONCEOᵃ 2,338 0.22 0.20 0.00 0.87 Dependent variables STOCKRETURNᵃ 2,338 0.04 0.01 0.02 0.06 CAPEXᵃ 2,338 0.10 0.06 0.02 0.32 ROAᵃ 2,281 0.02 0.03 0.00 0.16 Moderating variables DIRINDEP 2,338 0.87 0.07 0.56 0.94 DIRECTORGEN 2,338 0.90 0.06 0.54 1 OPTIONDIR 2,338 50,809 81,307 0.00 1,201,783 YEARTORETIR 2,338 13 5.72 -9 35 Control variables COMPANYSIZEᵇ 2,338 4.16 0.53 3.12 5.93 GENDERCEOᶜ 2,338 0.98 0.14 0.00 1.00 DUALITYCEOᶜ 2,338 0.41 0.49 0.00 1.00 BOARDSIZEᵃ 2,338 10.98 2.05 7 17 PROFITABILITYᵇ 2,338 2.28 0.70 -0.48 3.51

The table contains the independent, dependent, moderating and controls variables. ᵃ Variables have been winsorized to remove extreme outliers.

ᵇ Natural logarithms. ᶜ Dummy variables.

Table 3 shows the Pearson correlation matrix. The highest correlation is between PROFITABILITY and COMPANYSIZE (-0.53). This is surprising and might be explained by PROFITABILITY being measured as ROA. The control variable ROA has total assets as a denominator and therefore is negatively related to the control variable COMPANYSIZE because it also uses total assets. The second highest correlation is between BOARDSIZE and COMPANYSIZE (0.49). Furthermore, STOCKRETURN (0.04) and CAPEX (0.12) seem to be significantly related to OPTIONCEO, which provides initial support for H1. Variance inflation factor analysis revealed that the highest factor is 1.83 except for the industry dummies, which indicate a high multicollinearity for this control variable.

(21)

20 However, the industry dummies are still deemed important because industry might highly influence the amount of CAPEX expenditures and for this reason were included in the model.2

(22)

Table 3. Pearson correlation matrix (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) 1. OPTIONCEO 1 2. STOCKRETURN 0.04* 1 3. CAPEX 0.12** -.05* 1 4. ROA -0.00 -0.02 .06** 1 5. YEARTORETIR 0.06** .07** .11** 0.02 1 6. DIRINDEP -0.02 -0.03 -.19** -.05* 0.08** 1 7. OPTIONDIR 0.31** -0.13** .18** 0.01 -0.07** -0.05* 1 8. DIRECTORGEN 0.06** 0.08** .05* 0.03 0.01 0.10** -0.08** 1 9. COMPANYSIZE -0.15** -0.07** -.20** -0.12** -0.14** 0.25** 0.10** -0.12** 1 10. GENDERCEO -0.00 -0.07* 0.02 0.01 0.01 0.08** -0.14** 0.15** -.05** 1 11. DUALITYCEO 0.00 0.04* .05* 0.01 -0.12** -0.05* 0.01 0.07** 0.06** -0.03 1 12. BOARDSIZE -0.14** 0.01 -.17** -0.07** -0.09** 0.22** 0.01 -0.18** 0.49** -0.05* -0.04* 1 13. PROFITABILITY 0.14** -0.03 .11** 0.10** 0.04 -0.11** 0.14** -0.04 -0.53** -0.02 -0.11** -0.17** 1 ** Correlation is significant at the 0.01 level (2-tailed).

(23)

4.2 Testing of H1

H1 hypothesized a positive relationship between OPTIONCEO and CAPEX. I used a panel regression to test H1 because it increases the number of observations in the sample (Wooldridge, 2015) and therefore increases the degrees of freedom and reduces multicollinearity issues (Hsiao, 2007). The sample in the panel regression consists of 2,338 observations. In this model the dependent variable used as a proxy for risk-taking was CAPEX. To test H1 the following regression was used:

CAPEXit = α₀ + β₁ OPTIONCEOit + β₂ COMPANYSIZEit + β₃ BOARDSIZEit +

β₄ GENDERCEOit + β₅ DUALITYCEOit + β₆ PROFITABILITYit + β₇ YEARDUMMYit +

β₈ INDUSTRYDUMMYit + ᵋit

The first model in Table 4 includes the control variables. The second model includes the stock options for the CEO. As can be derived from Table 4, the control variables COMPANYSIZE, PROFTIABILITY and DUALITYCEO are significant at the p < 0.10 level, while PROFITABILITY and DUALITYCEO are significant at the p < 0.05 level. With regard to H1, there is a significant positive association (β = 0.018; p < 0.01) between CEO stock options and CAPEX. This finding is consistent with the prediction and provides support for H1, indicating that as CEO stock options increase, risk-taking also increases. Table 4. Panel regression for H1

Model 1: Panel regression (Dependent variable CAPEX)

Model 2: Panel regression (Dependent variable CAPEX)  SE  SE Intercept 0.153*** (0.032) 0.147*** (0.032) Controls COMPANYSIZE -0.008* (0.005) -0.008* (0.005) BOARDSIZE 0.000 (0.001) 0.000 (0.001) PROFITABILITY 0.013*** (0.002) 0.013*** (0.002) GENDERCEO 0.013 (0.014) 0.012 (0.014) DUALITYCEO 0.006** (0.003) 0.006** (0.003) Main effects OPTIONCEO 0.018*** (0.005)

YEARDUMMY YES YES

INDUSTRYDUMMY YES YES

n 2,338 2,338

R square 0.1551 0.1606

∆ R Square 0 0.0055

(24)

4.3 Testing of H2

H2 hypothesized a negative moderating effect of DIRINDEP on the CEO stock option and risk-taking relationship. To test this moderating effect, both DIRINDEP and OPTIONCEO were standardized. The following model was used to test H2 by running a panel regression:

CAPEXit = α₀ + β₁ OPTIONCEOit + β₂ DIRINDEPit + β₃ DIRINDEPit*OPTIONCEOit +

β₄ COMPANYSIZEit + β₅ BOARDSIZEit + β₆ PROFITABILITYit + β₇ GENDERCEOit + β₈

DUALITYCEOit + β₉ YEARDUMMYit + β₁₀ INDUSTRYDUMMYit + ᵋit

Table 5 shows the results where the first model includes the control variables, the second model includes the direct effects of OPTIONCEO and DIRINDEP and the third model includes the moderating effect of DIRINDEP. It can be concluded from Model 3 that there is no significant interaction between DIRINDEP and OPTIONCEO, and Model 3 provides no evidence to support H2. However, there is evidence for a direct effect, given that Model 2 and Model 3 indicate a significant and negative relationship between DIRINDEP and CAPEX (β = -0.003; p < 0.05).

Table 5. Panel regression for H2

Model 1: Panel regression (Dependent variable

CAPEX)

Model 3: Panel regression (Dependent variable CAPEX)  SE  SE  SE Intercept 0.153*** (0.032) 0.144*** (0.032) 0.144*** (0.032) Controls COMPANYSIZE -0.008* (0.005) -0.008* (0.004) -0.008* (0.004) BOARDSIZE 0.000 (0.001) 0.000 (0.001) 0.000 (0.001) PROFITABILITY 0.013*** (0.002) 0.013*** (0.002) 0.013*** (0.002) GENDERCEO 0.013 (0.014) 0.012 (0.014) 0.012 (0.014) DUALITYCEO 0.006** (0.003) 0.006** (0.003) 0.006** (0.003) Main effects OPTIONCEO 0.018*** (0.005) 0.018*** (0.005) DIRINDEP -0.003** (0.001) -0.003** (0.001) Two-way interaction OPTIONCEO X DIRINDEP 0.000 (0.001)

YEARDUMMY YES YES YES

INDUSTRYDUMMY YES YES YES

n 2,338 2,338 2,338

(25)

∆ R Square 0 0.014 0.0009

*** p < 0.01, ** p < 0.05, * p < 0.1

4.4 Testing of H3

The third hypothesis proposed a negative moderating effect of gender diversity on the CEO stock option and risk-taking relationship. In this model, the gender diversity ratio is 1.0 for a 100% male board and 0.5 for a maximally gender-diverse board (i.e. 50% female and 50% male). Both DIRECTORGEN and OPTIONCEO were standardized in Model 2 and Model 3 in order to test the moderating effect. The following model was adopted for testing H3:

CAPEXit = α₀ + β₁ OPTIONCEOit + β₂ DIRECTORGENit + β₃ DIRECTORGENit*OPTIONCEOit +

While Model 1 includes the control variables, the second model presents the direct effects of CEO stock options and gender diversity. The third model comprises the moderating effect of DIRECTORGEN. As can be derived from this model, there is a positive and significant interaction between OPTIONCEO and DIRECTORGEN (β = 0.005; p < 0.01), which is consistent with the prediction of H3. This finding indicates that if there are more women on the board, the effect of stock options on risk-taking is lower. Table 6. Panel regression for H3

CAPEX)

Model 3: Panel regression (Dependent variable CAPEX)  SE  SE  SE Intercept 0.153*** (0.032) 0.146*** (0.032) 0.152*** (0.032) Controls COMPANYSIZE -0.008* (0.005) -0.008* (0.005) -0.007 (0.004) BOARDSIZE 0.000 (0.001) 0.000 (0.001) 0.000 (0.001) PROFITABILITY 0.013*** (0.002) 0.013*** (0.002) 0.012*** (0.002) GENDERCEO 0.013 (0.014) 0.011 (0.014) 0.011 (0.014) DUALITYCEO 0.006** (0.003) 0.006** (0.003) 0.006** (0.003) Main effects OPTIONCEO 0.019*** (0.005) 0.015*** (0.005) DIRECTORGEN 0.001 (0.001) 0.000 (0.001) Two-way interaction

(26)

n 2,338 2,338 2,338 R square 0.1551 0.1611 0.1793 ∆ R Square 0 0.011 0.0182 *** p < 0.01, ** p < 0.05, * p < 0.1

4.5 Testing of H4

H4 hypothesized a positive moderating effect of OPTIONDIR on the relationship between OPTIONCEO and risk-taking. As before, to test the moderating effect both OPTIONDIR and OPTIONCEO were standardized in Model 2 and Model 3. The following model was used to test the hypothesis:

CAPEXit = α₀ + β₁ OPTIONCEOit + β₂ OPTIONDIRit + β₃ OPTIONDIRit*OPTIONCEOit +

Table 7 presents the results of the panel regression. The first model includes only the control variables, while the second model also incorporates the direct effects of CEO stock options and director stock options. The third model includes the moderating effect of OPTIONDIR. Model 3 shows that the moderating effect of OPTIONDIR is significant (p < 0.01), however, contrary to my prediction it is negative (β = -0.002). This indicates that as directors receive more stock options the relationship between CEO stock options and risk-taking becomes weaker. Furthermore, another important observation is that the direct effects of both OPTIONDIR (β = 0.006; p < 0.01) and OPTIONCEO (β = 0.006; p < 0.01) are positive and significant.

CAPEX)

Model 3: Panel regression (Dependent variable CAPEX)  SE  SE  SE Intercept 0.153*** (0.032) 0.159*** (0.031) 0.162*** (0.031) Controls COMPANYSIZE -0.008* (0.005) -0.011** (0.004) -0.011** (0.004) BOARDSIZE 0.000 (0.001) 0.000 (0.001) 0.000 (0.001) PROFITABILITY 0.013*** (0.002) 0.012*** (0.002) 0.012*** (0.002) GENDERCEO 0.013 (0.014) 0.014 (0.013) 0.013 (0.013) DUALITYCEO 0.006** (0.003) 0.006** (0.003) 0.006** (0.003)

(27)

Main effects

OPTIONCEO 0.015*** (0.005) 0.017*** (0.005)

OPTIONDIR 0.006*** (0.001) 0.008*** (0.001)

Two-way interaction

OPTIONCEO X OPTIONDIR -0.002*** (0.001)

4.6 Testing of H5

H5 hypothesized a positive moderating effect of the directors’ average years until retirement on the CEO stock option and risk-taking relationship. To test the moderating effect both YEARTORETIR and OPTIONCEO were standardized in Model 2 and Model 3. The model below was adopted to perform the panel regression:

CAPEXit = α₀ + β₁ OPTIONCEOit + β₂ YEARTORETIRit + β₃ YEARTORETIRit*OPTIONCEOit +

Table 8 presents the results of the panel regression. The first model includes the control variables. The second model includes the direct effects of OPTIONCEO and YEARTORETIR. The moderating effect of YEARTORETIR in incorporated in the third model. The moderating effect of YEARTORETIR is significant and positive (β = 0.002; p < 0.01). This result is consistent with the prediction and therefore provides support for H5, indicating that as the directors are closer to retirement age the relationship between CEO stock options and risk-taking becomes stronger.

(28)

CAPEX)

Model 3: Panel regression (Dependent variable CAPEX)  SE  SE  SE Intercept 0.153*** (0.032) 0.144*** (0.032) 0.141*** (0.032) Controls COMPANYSIZE -0.008* (0.005) -0.007* (0.005) -0.007 (0.005) BOARDSIZE 0.000 (0.001) 0.000 (0.001) 0.000 (0.001) PROFITABILITY 0.013*** (0.002) 0.013*** (0.002) 0.013*** (0.002) GENDERCEO 0.013 (0.014) 0.012 (0.014) 0.012 (0.014) DUALITYCEO 0.006** (0.003) 0.007** (0.003) 0.007*** (0.003) Main effects OPTIONCEO 0.018*** (0.005) 0.017*** (0.005) YEARTORETIR 0.001 (0.001) 0.001 (0.001) Two-way interaction OPTIONCEO X YEARTORETIR 0.002*** (0.001)

4.7 Robustness

In order to illuminate whether my findings are contingent on the specification of risk-taking, I conducted a robustness test. In Model 1, below, I integrated all moderating variables into a single model and conducted a panel regression. The dependent variable of Model 1 is CAPEX. As can be derived from Table 9, additional support is provided for H1, H3, H4 and H5, as the previously detected effects remain significant (p < 0.05). Furthermore, when STOCKRETURN is used as a proxy for risk-taking in Model 2 there is a significant positive relationship between OPTIONCEO and STOCKRETURN, which supports H1 (β = 0.001; p < 0.01). However, Model 2 does not provide support for the predicted moderating effect. It does, however, indicate direct effects of DIRECTORGEN (p < 0.01), OPTIONDIR (p < 0.01) and YEARTORETIR (p < 0.10). Model 3 presents a panel regression with ROA as a proxy for risk-taking. As can be understood from the model, there is no significant positive relationship between OPTIONCEO and ROA to support H1. Due to the availability of data, the number of

(29)

observations of ROA differs. However, due to the size of the sample I do not expect that this biased the results in any way.

Overall, these findings suggest that CEO stock options increase risk-taking, which is consistent with the findings of other scholars (Sanders & Hambrick, 2007; Sanders, 2001; Devers et al., 2007). However, the moderating effects of the board characteristics are only present in the model when the dependent variable is CAPEX, indicating that the relationships for some proxies are moderated by board characteristics (CAPEX) while others are not (STOCK RETURN, ROA). This implies that the measurement of risk-taking is relevant.

Table 9. Panel regression for robustness

Model 1: Panel regression (Dependent variable CAPEX) Model 2: Panel regression (Dependent variable STOCKRETURN) Model 3: Panel regression (Dependent variable ROA)  SE _ SE _ SE Intercept 0.108*** (0.024) _0.028*** (0.000) _0.033*** (0.011) Controls COMPANYSIZE -0.013*** (0.004) _{-0.003*** (0.001)} _-0.003 (0.002) BOARDSIZE 0.001 (0.001) _0.001*** (0.000) _-0.000 (0.000) PROFITABILITY 0.011*** (0.002) _-0.001* (0.000) _-0.000 (0.001) GENDERCEO 0.017 (0.014) _{-0.011*** (0.002)} _-0.002 (0.007) DUALITYCEO 0.008*** (0.003) _0.002** (0.001) _0.000 (0.002) Main effects OPTIONCEO 0.003** (0.001) _0.001*** (0.000) _-0.000 (0.001) DIRINDEP -0.006*** (0.001) _-0.000 (0.000) _-0.001* (0.001) DIRECTORGEN 0.001 (0.001) _0.001*** (0.000) _-0.000 (0.001) OPTIONDIR 0.010*** (0.001) _{-0.002*** (0.000)} _-0.000 (0.001) YEARTORETIR 0.004*** (0.001) _0.001* (0.000) _0.000 (0.001) Two-way interaction OPTIONCEO X DIRINDEP -0.001 (0.001) _0.000 (0.000) _0.001 (0.001) OPTIONCEO X DIRECTORGEN 0.005*** (0.001) _0.000 (0.000) _{-0.002*** (0.001)} OPTIONCEO X OPTIONDIR -0.002** (0.001) _0.000 (0.000) _-0.000 (0.000) OPTIONCEO X YEARTORETIR 0.002*** (0.001) _-0.000 (0.000) _0.001 (0.001)

n 2,338 2,338 2,281

The moderating effect of board characteristics on the CEO stock option and risk-taking relation.