The relationship between women on the board and the risk
profile of an organization
Name: Eva Verstraete
Student number: 11420413
Thesis supervisor: dr Réka Felleg
Date: June 23, 2018
Word count: 13.478
MSc Accountancy & Control, specialization Accountancy and Control
2
Statement of Originality
This document is written by student Eva Verstraete who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
3
Abstract
I examine whether gender diversity of the board of directors is related to an organization’s risk profile. As women tend to be more risk-averse, it is possible that female board representation lowers an organization’s risk profile. Previous studies find mixed results, which might be explained by the critical mass theory. This theory suggests that a certain relative proportion or absolute number of female directors should be on the board before they can have an influence on the board. Therefore, I argue that the proportion of female directors might have a negative effect on an organization’s risk profile and that the impact of women is stronger above a certain threshold. Using data of non-financial firms listed on the S&P 1500 from 2007 to 2016 I find that the proportion of women is negatively related to an organizations risk profile, measured as the annualized stock volatility. This supports the notion that women are more risk-averse and thereby affect the functioning of the board of directors as the proportion of female directors increases. Contrary to the critical mass theory, I fail to find evidence that the impact of women directors differs at or above a certain threshold. However, I find that the proportion of female directors differs among industries, which might indicate that failure to find support for the critical mass theory is caused by self-selection.
Keywords: board gender diversity, female directors, critical mass, corporate governance,
4 Contents 1 Introduction ... 5 2 Literature review ... 9 2.1 Agency theory ... 9 2.2 Risk... 10 2.3 Board composition ... 10 2.4 Gender diversity ... 12
2.5 Critical mass theory ... 14
2.6 Hypotheses development... 16 3 Methodology ... 18 3.1 Sample ... 18 3.2 Variables... 19 3.3 Empirical design ... 21 4 Results ... 23 4.1 Descriptive statistics ... 23 4.2 Hypothesis tests ... 25 5 Robustness tests ... 29 5.1 Specification of regression ... 29
5.2 Operationalization of dependent variable ... 30
5.3 Threshold of female directors ... 33
5.4 Explanation of results ... 34
5.5 Discussion of results... 35
6 Conclusion ... 36
References ... 39
5
1 Introduction
Social and ethical responsibility used to be the driver for gender diversity of corporate boards (Isidro & Sobral, 2015). Recently, however, there is an increased focus on gender diversity by means of improving corporate governance (Isidro & Sobral 2015; Terjesen et al., 2009). In general, diversity enhances a robust wealth of training and education, relevant background and experience and different perspectives, which improves the quality of decision-making (e.g., Mazur and Bialostocka, 2010). Since women are perceived to be more caring, emphatic and selfless, their perspectives differ from the perspectives of men (Santilli & Hudson, 1992; Eagly & Wood, 1991). More gender-diverse boardrooms would therefore not only enhance gender equality but also result in more different perspectives within the boardroom, which is valuable for the organization (Schwartz-Ziv & Weisbach, 2013; Mazur & Bialostocka, 2010).
Internationally, companies are under pressure to compose more gender-diverse boards in order to enhance gender parity. Organizations such as Catalyst and McKinsey are reporting statistics regarding the number of women on boards (Schwartz-Ziv, 2017). Government commissions also examine women’s representation on boards, as in the Davies Review and Hampton-Alexander Review (Hampton & Alexander, 2017; Davies, 2011). Legislators and media use these reports to pressure firms to increase the number of female board directors. Several countries have implemented quotas to mandate publicly listed companies to have a set minimum percentage of women on their board (Terjesen & Sealy, 2016). Additionally, there are international initiatives encouraging firms to increase the gender diversity of their board, for example, campaigns such as 2020 Women on Boards and the 30% Club (Willows & Van der Linde, 2016).
Given the increased focus on gender diversity, there is an increase in studies examining the effect of a more gender-diverse board. These studies argue that an increase in female directors affects the organization as men and women differ. According to these studies, women tend to be less competitive, more risk-averse and more long-term oriented in a business environment (Niederle & Vesterlund, 2007; Byrnes et al., 1999; Ely, 1994). Whether these differences have an impact on the organization is examined by several studies, which relate gender diversity of the board to corporate governance and firm performance. Regarding corporate governance performance, more female directors are associated with more effort to monitor, more efficient decision-making process and a more social and ethical compliance (Isidro & Sobral, 2015; Adams & Ferreira, 2009). However, the results of studies considering
6
the effect of gender diversity of the board on firm performance are mixed (e.g., Campbell & Mínguez-Vera, 2008; Carter et al., 2003).
The debate on the relationship between gender diversity and firm value is inconclusive as yet. Considering financial performance measures, some studies find a positive relationship between the number of female directors and measures such as Tobin’s Q and return ratios (Carter et al., 2003; Erhardt et al., 2003). While others conclude no relationship or negative relation between gender diversity and financial performance measures (Ilaboya & Ashafoke, 2017; He & Huang, 2011; Adams & Ferreira, 2009; Campbell & Mínguez-Vera, 2008). These financial performance measures are not risk-adjusted, although two firms with identical cash flows might have different risk profiles. Thus, female board representation might not affect firm performance but could have an effect on an organization’s risk profile. Studies that attempted to find whether there is a relation between the proportion of women on the board and the risk profile of an organization are also not conclusive. Sila et al. (2016) do not find sufficient evidence to conclude a relation, while Jizi and Nehme (2017) conclude that female directors reduce an organization’s stock volatility and therefore lower an organization’s risk profile.
Since the representation of women on boards is still increasing, I suggest that the mixed results can be explained by the critical mass theory of Kanter (1977). According to the critical mass theory, one female director or a minority of female directors is not sufficient to have influence since they are considered as ‘tokens’. These tokens are symbolic representatives who do not affect firm performance until a certain critical mass is exceeded. Following Kanter (1977), the crucial point to exceed would be thirty percent, therefore the impact of women directors and corporate governance performance and firm performance might differ above and below this threshold. However, previous studies generally suggest no differential impact. I argue that those studies find no or a weak link between gender diversity of the board and corporate governance performance or firm performance because the average proportion of women on boards is still rather low. Recent studies that examine the critical mass theory support this argument. Regarding corporate governance, there is a positive relation after exceeding the critical mass, as a board with 20% to 40% women is more active (Schwartz-Ziv, 2017). As regards firm performance, the proportion of women on the board improves return on assets and net interest margin stronger at or above the threshold compared to below the threshold (Kramaric & Miletic, 2017). Torchia et al. (2011) focus on the number of female directors rather than the ratio or presence of women on the board. This study aims to expand
7
the critical mass theory by using literature regarding group dynamics. According to this literature, three is a pivotal number in-group dynamics. Seeing as when there are three or more people within a group with a certain characteristic, stereotyping and biasing decreases (e.g., Asch, 1955). The results of Torchia et al. (2011) show that three or more women on the board have a positive and significant relationship with corporate governance. This indicates that the number of women on the board might be more important than the ratio or presence of female directors.
In this paper, I examine whether gender diversity of the board is related to an organization’s risk profile. Specifically, I test whether there exists a critical percentage or number of female board members necessary to affect an organization’s risk profile. Studying this is important and adds to the extant literature on gender diversity of boards and firm performance for two reasons. Firstly, it is likely that the future composition of boards changes given the pressure on organizations to enhance the proportion of female directors. Whereas women are more risk-averse and less competitive, it is of value to investors to know how the change of board composition affects an organization’s risk profile. Additionally, in light of considering a gender quota, it is in the interest of legislators to be aware of the consequences for organizations. Secondly, my research differs from existing literature because I do not only test for a relationship between gender diversity and an organization’s risk profile, but examine also the existence of a relative or absolute threshold, based on the critical mass theory. To my knowledge, this is the first study examining this.
To examine the relationship between the proportion of female directors and the risk profile of an organization and whether there is a certain threshold above which the impact of female directors becomes pronounced, I perform an analysis on the annualized stock volatility of 1,543 non-financial firms in the S&P 1500 over 2007 to 2016. I run several OLS regressions on the data. The results suggest that there is a negative relationship between the proportion of female directors and the risk profile of an organization. However, I fail to find evidence for the existence of a certain relative or absolute threshold above which the impact of female board representation on an organization’s risk profile becomes more pronounced. Given that I find that the proportion of women differs among industries, I suggest that failure to support the relative and absolute threshold is a consequence of a different impact of female directors in different industries. These findings provide a unique contribution to the existing literature on board gender diversity. Furthermore, the findings of my study are relevant to investors, legislators, managers and directors. Since there is little research on the impact of women
8
directors on an organization’s risk profile and the existence of a certain threshold, my results may help to comprehend the consequences of increasing the proportion of female directors. Since my findings identify a negative relationship between the proportion of women directors and an organization’s risk profile, for investors this might imply that the risk level of their portfolio decreases when a female director is appointed. This conclusion supports the notion that women are more risk averse and thus more female directors affect the functioning of the board of directors with respect to the firm’s risk acceptance. Additionally, I find that a minority of female directors can affect an organization’s risk profile. This may imply for legislators that currently pressuring organizations to increase the proportion of female directors already affects an organization’s risk profile.
The remainder of this paper is structured as follows. The next section provides the extant literature and the hypotheses development. Thereafter the research method is described in the third section. The results of the regressions are described in the fourth section, followed by the tests conducted to examine the robustness of my results in section five. Finally, the conclusion is drawn in the last section.
9
2 Literature review
2.1 Agency theory
Separation of control and ownership of a firm is an essential development for modern economic theories. The extant literature explores the effects of this development, which is interesting as the interests of the controlling and owning party might differ. The literature is extended by the inclusion of the agency problem, which is the problem that might occur when the two parties have different objectives and responsibilities (Jensen & Meckling, 1976).
From an agency perspective, the agent should act on behalf of the principal (Fama, 1980). Regarding cooperation within the organization, the managers control the organization and thus are the agent, and the shareholders represent the principal as the owner of the organization. However, managers might act differently from the interest of shareholders because of two type of frictions. The first type of friction is the existence of asymmetric information that refers to the problem that the agent is more and better informed compared to the principal. The second type refers to imperfectly aligned incentives, according to the agency theory a consequence of both parties being utility maximizers (Holmström, 1979; Jensen & Meckling, 1976). Because of these frictions between the agent and principal, the agency problems moral hazard and adverse selection can occur. Moral hazard refers to actions and effort of the manager detrimental to the principal. Managers might perform these type of actions as the consequences are not at their expense and as the shareholders cannot observe all actions performed by the manager and its effects on the organization. Adverse selection concerns information hidden by the managers from shareholders, as a result, of which the shareholders cannot exactly assess the performance of managers that might result in poor investment decisions (Holmström, 1979; Pauly, 1974).
The principal might incur agency costs, to limit agency problems. Agency costs are the sum of monitoring and contracting costs and the residual loss. The monitoring costs are intended by the principal, to control and monitor the agent’s actions whereas contracting costs aim to align the incentives of the agent with those of the principal (Fama, 1980; Jensen & Meckling, 1976). Finally, the residual loss is the cost that occurs, as the principal accepts some deviation in the actions of the manager from the ideal as otherwise, the monitoring and contracting costs would be prohibitive (Jensen & Meckling, 1976).
10
2.2 Risk
Return and risk are fundamental concepts in financial theories. Assuming an efficient capital market, increasing return relates to carrying more risk, therefore return and risk are positively related (Bernstein, 1996). According to the Modern Portfolio Theory, investors can choose a tradeoff between risk and expected return as the investor can reduce its risk level by diversification (Markowitz, 1952). High risk of individual investments in a well-diversified portfolio might generate higher returns. Therefore, it is likely that the investor engages in more risky investments, to generate an optimal portfolio. On firm level, the investor is therefore considered as risk-seeking (Bernstein, 1996).
In the interest of the investors, the organization has to engage in risky projects with a positive net present value. As managers of the organization select the projects to engage in, it is appropriate to assume that the managers consider risk and return. However, the agency theory argues that managers are more risk-averse compared to investors who are risk-seeking on firm level, as their future employment income depends on the continuation of the organization (Berk et al., 2010; Holmström, 1999; Agrawal & Mandelker, 1987). Additionally, as contracting costs, the organization might issue common stock to the manager to increase its motives to improve firm performance. However, incentives regarding risk-taking behavior of managers might diminish even further as the wealth of managers is undiversified (Chava & Purnanandam, 2010; Fama, 1980). To encourage managers to engage in risky projects, and thereby align the incentives of shareholders and managers, the board of directors controls and monitors risk engagements of the managers (Cheng, 2008).
2.3 Board composition
According to the agency theory, the important function of the board of directors is monitoring firm management on behalf of the shareholders. Furthermore, the board should provide information, advice and counsel, create external legitimacy for the organization and gain access to external commitments or support (Pfeffer & Salancik, 1978). These activities performed by the directors are considered important drivers for firm performance (He & Huang, 2011). To understand the impact of the board on the organization, previous studies examined the influence of board composition on firm outcomes. Extant studies examined several board characteristics including board size, board independence and board directors’ diversity.
11
Jensen (1993) finds that big boards are less effective, therefore, board size might have a negative impact on the organization. Studies testing this view, find a negative relationship between board size and firm performance (Cheng, 2008; Pathan et al., 2007; Hermalin & Weisbach, 2003). While larger boards may have expertise that is more diverse, lower intergroup connection and more conflicts result in a less effective board (Goodstein et al., 1994). Next to firm performance, the organization’s risk profile is negatively affected by board size. Pathan (2009) finds a negative relationship between stock volatility and board size. More directors require more compromises, therefore, decisions become less risky. Overall, bigger boards are thus not favorable for shareholders’ wealth as they are risk-seeking on firm level.
With respect to board independence, the outcomes of studies testing the relation with firm outcomes are mixed. Independent directors are outside directors, not affiliated with the firm through significant business or family relationships with the organization (Adams & Ferreira, 2009). According to Fama and Jensen (1983), independent directors should increase shareholders’ wealth. In accordance with this view, various studies find a positive relation between board independence and firm performance in terms of corporate governance and financial performance (Pathan et al., 2007; MacAvoy & Millstein, 1999; Rosenstein & Wyatt, 1990). However, other studies find no or a negative relationship between board independence and firm performance (Ramdani & Van Witteloostuijn, 2010; Hermalin & Weisbach, 2003). Concerning the organization’s risk profile, Sila et al. (2016) argue that there is a positive relationship with board independence, as independent directors are more shareholders focused. In the interest of shareholders diversified portfolio, independent directors would enhance risk-taking by the organization.
The diversity of directors might increase the variety of perspectives within the board, which enhances the monitoring function of the board and thereby improves corporate governance and performance of the organization (Mazur and Bialostocka, 2010; Carter et al., 2003). Additionally, Hillman et al. (2000) hypothesized that diversity of the board increases the ability to connect with stakeholders, which improves firm performance. Diversity of the board refers to the diversity of characteristics of directors such as ethnicity, gender, age, industry experience, education and cultural background. The results of studies testing the relation between board diversity and firm performance are mixed. Some of the studies support the view of Carter et al. (2003) and Hillman et al. (2000), others fail to indicate a relationship between board diversity and firm performance (Ferreira, 2010). The relationship between diversity of directors and the risk profile of an organization is not examined yet.
12
2.4 Gender diversity
From a social and ethical perspective, because of equal opportunities and representation, ideally, both men and women are represented on the board (Isidro & Sobral, 2015). Organizations and government commissions, who report on statistics regarding women’s representation on boards, promote this perspective. With these reports, legislators and the media try to pressure firms to compose more gender-balanced firms through these reports (Schwartz-Ziv, 2017). Furthermore, firms can be required to have a certain percentage of women on their board as several countries implemented board gender quotas (Terjesen & Sealy, 2016). Currently, in the top 1000 U.S. companies based on revenue, women hold 19.8% of board seats. As organizations are pressured to increase the number of female directors, it is likely that the percentage of board seats held by women will increase (2020 Women on Boards, 2018).
From an agency perspective, more gender-diverse boards might result in perspectives that are more diverse and thereby improve corporate governance and firm performance (Carter et al., 2003). The difference in perspectives of men and women are driven by the differences in characteristics of men and women. In general, women are perceived to be more caring, emphatic and selfless (Santilli & Hudson, 1992; Eagly & Wood, 1991). In a professional setting, women of the same ability as men tend to be less competitive, more risk-averse and more long-term oriented (Niederle & Vesterlund, 2007; Byrnes et al., 1999; Ely, 1994). Because of these differences, compared to men, women are less extreme and more consistent investors. As a director, women also appear to behave differently from men.
Since female directors behave differently, it appears that gender diversity of the board has an impact on corporate governance. Adams and Ferreira (2009), examine the relationship between gender diversity and monitoring, with CEO turnover as a proxy for monitoring intensity. They find that a board with more female directors is likely to assign more effort in monitoring. This is evident from the fact that the CEO is more likely to be held responsible for performance if the number of female directors increases, as the CEO’s turnover is more dependent on the CEO’s performance. Moreover, Isidro and Sobral (2015) conclude that women improve monitoring because of improved decision-making and an increase in board independence. Additionally, they find evidence that female directors improve social and ethical compliance. Thus overall, gender diversity of the board has a significant impact on corporate governance (Adams & Ferreira, 2009).
13
Given the impact on corporate governance, the effect of gender diversity on firm performance is also extensively studied. However, in contrast to the significant relationship regarding corporate governance, the results of the relation between board gender diversity and firm performance are ambiguous (Adams & Ferreira, 2009). Studies testing for a relationship operationalize firm performance in multiple ways including Tobin’s Q, shareholder value creation and return ratios. The outcomes of these studies are mixed. Several studies provide a positive relationship between gender diversity and firm performance (Carter et al., 2003; Erhardt et al., 2003). However, other studies find evidence of a negative relationship (Ilaboya & Ashafoke, 2017; Isidro & Sobral, 2015; He & Huang, 2011; Campbell & Mínguez-Vera, 2008). Additionally, there are studies concluding that there is no relationship between board gender diversity and firm performance (Isidro & Sobral, 2015). According to Adams and Ferreira (2009), the mixed results indicate that increased monitoring can improve as well as reduce shareholder value. Additionally, in an environment where the application of female directors is mandated by a quota, more female directors on the board diminish firm performance (Ahern & Dittmar, 2012). This might also indicate over-monitoring or contra selection of female directors as the firms are forced to appoint more women (Adams & Ferreira, 2009). Ahern and Dittmar (2012) propose that insufficient selection occurs because the quota is recently adopted, this is a short-term effect that decreases over time.
Finally, as the attitude of men and women towards risk aversion differs, it is also possible that gender diversity has an impact on an organization’s risk profile. Since women are more risk averse, an increase in the number of female directors is expected to lower the risk acceptance of the board. As the decisions of the board might become less risky and monitoring risk-taking behavior of management becomes tighter, this might decrease the organization’s risk profile. Jizi and Nehme (2017), who find a negative relationship between female board representation and stock volatility as a proxy of the organization’s risk profile, support this view. In contrast, Sila et al. (2016) find no evidence that the risk profile of the organization is affected by gender diversity of the board. Concluding, there is no consensus on whether board gender diversity is related to an organization’s risk profile.
14
2.5 Critical mass theory
The impact of female directors on the organization might be limited since female directors are still considered as tokens. Tokenism refers to a few individuals that are seen as symbolic representatives of a specific social category, female directors are thus rather seen as representatives of women rather than being acknowledge and singled out because of their accomplishments. This stereotyping of female directors limits their ability to exert influence on the board (Kanter, 1977).
Kanter, (1977) studied group interaction of diverse groups in a political context. She argues that groups can be categorized into four different types, according to the composition of the group regarding a specific characteristic. The categories she identifies are uniform, skewed, titled and balanced groups. Within a uniform group, all members have the same characteristic. Regarding gender, a uniform group exists of only men or women. A skewed group consists of members with and without the specific characteristic, however, there is one dominant that controls the group. With respect to gender, there is at least one female but less than 20% is female in a male-dominated skewed group and vice versa, in case of a female-dominated skewed group. Titled groups are less extreme, in this type of group the minority have potential allies and may thereby control the group while it is supposed that in general the dominant still controls the group. The ratio of the minority is 20-40%, regarding gender a male-dominated group thus consists of 20-40% women. Finally, within a balanced group, the minority and majority become potential subgroups, other factors than the specific characteristic are important when it comes to control within the group. With respect to gender, between 40-60% of the group is a woman in a balanced board (Kanter, 1977). Concluding, if the proportion of the minority group increases, the potential of the group to exert influence increases as well.
According to the minority theory, below a certain threshold the minority has no potential to exert influence, however, there is a positive relationship between the size of the minority group and its potential to control the group. Thereby Kanter (1977), suggests that after passing the critical mass, a certain threshold, the minority group are no more tokens but has an effect within the group. Based on the different categories and the impact of the minority within the group, Kanter (1977) argues that minority should comprise at least thirty percent to have an effect. This critical mass theory gained support from several scholars who apply the theory to a legislative and political context (e.g., Childs & Krook, 2008).
15
The critical mass theory is barely applied to a business or board of directors’ context yet (Terjesen et al., 2009). However, the critical mass theory might explain why previous studies testing for a direct association between board gender diversity and several firm outcomes are not conclusive. When examining the relationship between gender diversity and corporate governance, Schwartz-Ziv (2017) finds a board with 20-40% women directors is more active compared to boards with fewer women. In addition, recent scholars find that exceeding the critical mass significantly improves firm performance (Kramaric & Miletic, 2017; Joecks et al., 2013). Therefore, the critical mass theory is also supported within the context of the board of directors. Moreover, Joecks et al. (2013) find the relation between gender diversity and firm performance is U-shaped. This indicates that there is also a critical mass for male directors within the board, necessary to achieve optimal firm performance. Thus, the interaction of perspectives of man and woman optimizes firm performance. Among other things, the perspective of women regarding monitoring causes in a female-dominated board over-monitoring, while a male-dominated board proceeds to less over-monitoring, both are detrimental to firm performance (Joecks et al., 2013).
Rather than testing for a threshold of thirty percent, other scholars test for an absolute critical mass of women directors within the board (e.g., Torchia et al., 2011). According to previous literature, the number three is pivotal within group dynamics (e.g., Asch, 1955). Especially when it comes to gender. One female director would attract attention, as gender is a key feature noticed. This results in biased perceptions of the other directors, which limit the impact of the solo female director. A second female director decreases stereotyping, but still, these female directors are seen as representatives of all women whereby their impact is limited. A third female director leads to normalization, the gender of directors becomes less noticed as being a man or woman is less extreme. As stereotyping and bias are eliminated, three female directors are supposed to have an impact (Erkut et al., 2008; Asch, 1995). Building on this literature Joecks et al. (2013) and Torchia et al. (2011) examine the relationship between three or more women on the board and firm performance. These studies conclude that three or more women on the board significantly improve firm performance. As the presence of three or more women on the board is sufficient to have potential allies within the minority group, this absolute threshold might be more important than the relative threshold.
16
2.6 Hypotheses development
The agency theory argues that managers might not always act in the best interest of shareholders because of information asymmetry and imperfectly aligned incentives. As the risk aversion of managers and shareholders differs, the board of directors is put in place as a governance mechanism to control and monitor the behavior of managers. Prior studies show that the characteristics of the board of directors, such as size, can have a particularly large impact on the risk profile of the organization (e.g., Pathan, 2009). Gender diversity of the board might be a feature of the board composition affecting the risk profile of the organization, as women tend to be more risk-averse. This is interesting as currently, organizations are pressured to increase the number of female directors on the board (Schwartz-Ziv, 2017; Terjesen & Sealy, 2016). I argue that if female directors enhance corporate governance and if they are more risk averse, an organization with more female directors is expected to have a lower risk profile. Accordingly, my first hypothesis is stated as follows:
Hypothesis 1: There is a negative relationship between the proportion of women directors and the organization’s risk profile.
Testing this hypothesis either results in confirmation of the study of Jizi and Nehme (2017) who find a negative relationship or confirm Sila et al. (2016) who conclude no association between the proportion of women directors and the organization’s risk profile. The difference in results might be a feature of their dataset, as the studied period and country differ. Sila et al. (2016) examine a much broader period, namely 1996 to 2010, whereas Jizi and Nehme (2017) examine only 2008 to 2013. Furthermore, Sila et al. (2016) focus on U.S. firms while Jizi and Nehme (2017) focus on firms in the UK. These countries differ with respect to board gender diversity, as the U.S. used to be an early pioneer of diversity. The female directors of the sample of Sila et al. (2016) might be considered as tokens, whereas the increased proportion of women on boards of the last years explains support for the relationship with the risk profile of an organization by Jizi and Nehme (2017). Since Sila et al. (2016) reject this hypothesis, it might be that the impact of female directors on an organization’s risk profile is only conspicuous after exceeding a certain threshold. A rejection of this hypothesis might imply a misspecification, my second and third hypothesis try to resolve this issue.
17
The critical mass theory suggests that a minority group should comprise at least thirty percent to have an influence (Kanter, 1977). This is in accordance with previous studies applying the critical mass theory in a board of directors’ context (Kramaric & Miletic, 2017; Schwartz-Ziv, 2017; Joecks et al., 2013). Therefore, I argue that the effect of the risk-averse attitude of women on an organization’s risk profile becomes pronounced once the proportion of women on a board exceeds thirty percent. Correspondingly, my second hypothesis is formulated as follows:
Hypothesis 2: The negative impact of female board directors on an organization’s risk profile is stronger after exceeding the relative threshold of thirty percent female directors on the board.
To expand the critical mass theory, Torchia et al. (2011) test whether the number of female directors is critical to have an effect on firm innovation. They find that having three women directors is the minimum to contribute to firm innovation. This finding is in accordance with previous theories regarding group dynamics, which argue that three is a pivotal number (Erkut et al., 2008; Asch, 1995). In addition, Joecks et al. (2013), who find that three or more female directors increase firm performance, as compared to boards with less female directors, confirm this absolute threshold of three female directors. I argue that if women directors enhance corporate governance and if they are more risk averse this only affects the organization’s risk profile when there are at least three female directors. In accordance, I state my third hypothesis as follows:
Hypothesis 3: The negative impact of female board directors on an organization’s risk profile is stronger after exceeding the absolute threshold of three female directors on the board.
18
3 Methodology
3.1 Sample
To test my hypotheses, I focus on U.S. public-listed companies of the S&P 1500. The data regarding board characteristics and risk profile of the S&P 1500 firms is highly available. Additionally, these firms are comparable to the previous studies (Jizi & Nehme, 2017; Sila et al., 2016). I use the Institutional Shareholder Services (ISS) database as a starting point to develop my sample. This database provides board characteristics of S&P 1500 firms only, hence limits my sample to these firms. Furthermore, the maximum time interval available in this database is a ten-year period from 2007 to 2016, therefore I examine this period. The initial sample consists of 14,281 firm-year observations and is an unbalanced panel. According to prior literature, financial service firms are subject to different types of business risk compared to other firms (Jizi & Nehme, 2017; Sila et al., 2016). Furthermore, the accounting standards for financial service firms deviate from those of other firms. Therefore, I exclude financial service firms from my sample. To obtain data regarding firms’ risk profile, I use the Center for Research in Security Prices (CRSP) database. As the risk profile cannot be determined without return figures, I drop observations without this data. Additionally, I use the Compustat and ExecuComp database to obtain data regarding control variables. Observations with missing values for firm characteristics, firm financial performance and CEO characteristics, are dropped. Consequently, my final sample consists of 9,821 firm-year observations of 1,543 unique companies. The sample selection procedures are summarized in Table 1.
19
3.2 Variables
Following prior studies examining the relation between board gender diversity and the organization’s risk profile (Jizi & Nehme, 2017; Sila et al., 2016), I measure the dependent variable, the organization’s risk profile, in terms of annualized stock volatility (VOL). Following prior literature, I use the daily holding period returns without dividends of the firms of my sample to calculate the stock volatility of the observed years (Sila et al., 2016). The annualized stock volatility is the square of the daily variance multiplied by the number of trading days of the year concerned. The formula of the annualized stock volatility is as follows:
𝑉𝑂𝐿𝑖,𝑡 = (∑(𝑥 − 𝑥̅) 2
𝑛 − 1 )
2
∗ 𝑛 (1)
where, x represents the daily returns without dividends and n the number of observations, which in this case is equal to the number of trading days of the year concerned. The risk profile of each organization is determined for each year separately.
Prior literature suggests that the Blau Index is the most appropriate measure of gender diversity of the board (Joecks et al., 2013; He & Huang; 2011; Campbell & Mínguez-Vera, 2008). The Blau Index is a quantitative measure that considers both the variety and distribution of the categorical variable, thereby it gives a ratio to the diversity of the board (Blau, 1977). The Blau index is calculated as:
H = 1 − ∑ 𝑠𝑐2 𝑘
𝑐=1
(2)
in which k reflects the number of categories, thus in the case of gender k equals two since there are two types of gender, either male or female. The sc stands for the proportion of board
members with characteristic, c, in this case male or female. By this formula, the Blau Index provides the same ratio for a board with 20% female directors and 80% male directors as for an opposite board with 80% female directors and 20% male directors (Joecks et al., 2013). However, contrary to the studies proposing the Blau Index, I expect that female-dominated boards behave differently from male-dominated boards. As I am especially interested in whether more women on the board have an impact on an organization’s risk profile, and not whether more diversity of the board affects an organization’s risk profile, the Blau Index is inappropriate in the context of my hypotheses. Therefore, I use the percentage of female directors on the board (PW) as the measure for the main independent variable of interest. The percentage is determined for each firm-year observation by dividing the number of women on
20
the board by the total number of directors. Furthermore, to test my second and third hypotheses, whether women have a stronger impact on an organization’s risk profile after exceeding a relative or absolute threshold, I create two dummy variables to divide my sample. In accordance with prior literature, I divide the sample into two groups considering the percentage of women to test for a relative threshold (Joecks et al., 2013). In particular, one group includes organizations with less than thirty percent female directors and the other involves organizations that have at least thirty percent female directors. Organizations with less female directors than the relative threshold are considered as the reference group, the first dummy variable is created to represent the other group (DRT). The value of the dummy is ‘1’ if the organization has at least thirty percent women directors, otherwise the value is ‘0’. Also following prior literature, I divide the sample into two groups based on the number of female board directors to test for an absolute threshold (Schwartz-Ziv 2017; Toricha et al., 2011). The first group represents boards with less than three female directors, the second group represents boards with three or more female directors. Organizations with less female directors than the absolute threshold are considered as the reference group, the second dummy variable is created to represent the other group (DAT). The value of the dummy is ‘1’ if the organization has at least three female directors, otherwise the value is ‘0’.
As control variables, I added a number of features of the firm that have an effect on an organization’s risk profile. I follow Sila et al. (2016) by controlling for firm size, measured as the log value of total assets (FS). The natural log is taken to control for outliers, and thereby improve the normal distribution of the variable. Prior research suggests that firm size affects both gender diversity of the board and the organization’s risk profile: according to Adams and Ferreira (2009), the proportion of women is positively affected by firm size. With respect to the organization’s risk profile, firm size has a negative impact, as bigger firms have more opportunities to diversify their investments, which results in less stock volatility (Baek et al., 2004). Second, I control for financial firm performance, as it affects the opportunity for investment and growth, which relates positively to taking risk (Guay, 1999). As a proxy for financial firm performance, I include the variables return on assets (ROA) and Tobin’s Q (TQ). The ROA is a profitability ratio, defined as the ratio of net income to total assets. Tobin’s Q represents firm value and is calculated as the market value of stocks and book value of debt divided by the book value of total assets (Campbell & Mínguez-Vera, 2008). Third, I include features of the CEO in terms of gender (CEOG) and tenure (CEOT). The risk attitude of the CEO is associated with these features and therefore these features might affect the origination’s risk profile. It is argued that longer tenure results in less risk-taking by the CEO (Sila et al.,
21
2016). CEO tenure is measured as the log of the years since he/she became CEO. Furthermore, it is argued that women are more averse, therefore a female CEO might result in less risk-taking. To control for the CEO’s gender I include a dummy, the value is ‘1’ if the CEO of the organization is a woman, otherwise, the value is ‘0’. Finally, I add the following board characteristics: board size (BS), independence of the board (BI) and CEO duality (CD), to control for other board characteristics than gender diversity of the board. Board size is negatively related to the organization’s risk profile, as a large board results in more compromises and therefore the decisions become less risky. On the other hand, a more independent board is positively related to an organization’s risk profile since more independent directors increase shareholder-focus, whereby risk-taking increases (Sila et al., 2016). Board size is measured in terms of the number of directors on an organization’s board. Board independence is calculated as the percentage of independent board members. Finally, CEO duality is represented by a dummy variable, the value is ‘1’ if the CEO is also the chairman of the board, otherwise, the value is ‘0’. The CEO duality is associated with softer governance, therefore I argue that CEO duality might have a negative impact on an organization’s risk profile (Adams & Ferreira, 2009).
3.3 Empirical design
The first hypothesis aims to determine if the proportion of female directors is negatively related to the risk profile of an organization. For testing this hypothesis, I estimate an ordinary least square (OLS) regression Model as presented below. The model was derived from Sila et al. (2016) and Jizi and Nehme (2017).
𝑉𝑂𝐿 = ∝ + 𝛽1 𝑃𝑊 + 𝛽2 𝐹𝑆 + 𝛽3 𝑅𝑂𝐴 + 𝛽4 𝑇𝑄 + 𝛽5 𝐶𝐸𝑂𝑇 + 𝛽6 𝐶𝐸𝑂𝐺 + 𝛽7 𝐶𝐸𝑂𝐷 + 𝛽8 𝐵𝑆 + 𝛽9 𝐵𝐼 + 𝛽10∑ 𝐼𝑁𝐷𝑈𝑆𝑇𝑅𝐼𝐸𝑆 + 𝛽11∑ 𝑌𝐸𝐴𝑅𝑆 + 𝜀
(3)
In line with my first hypothesis, I expect 𝛽1, the coefficient on the proportion of women to be negative. In order to support the hypothesis and thus conclude a negative relation between women on the board and an organization’s risk profile, 𝛽1 should be significantly negative.
22
The second and third hypotheses aim to determine whether the impact of a female board director is stronger after exceeding a relative threshold of thirty percent women directors or absolute threshold of three women directors. In order to measure the incremental impact on an organization’s risk profile after exceeding the threshold, I compare the impact of female directors of firms without or with less female directors than the relative or absolute threshold with the impact of female directors of firms with the number of women female directors at the relative or absolute threshold and above. To test my second hypothesis, I run the OLS regression Model (3) for the firms of the sample below the relative threshold and those firms with a percentage of female directors equal to or above the relative threshold. The cut-off of the relative threshold is at thirty percent female directors on the board, which is represented by the dummy DAT. An OLS regression is run to enable subsequent comparison of the significance and sign of the two coefficients in the two regressions on PW, 𝛽1. My third hypothesis is tested similarly, however, I estimate the OLS regression Model (3) separately based on the absolute threshold, whereby the limit is at three female directors instead of thirty percent women directors. Then I compare the significance and sign of the coefficients in the two regressions, as with my second hypothesis. Hypothesis two is confirmed if the coefficient on PW, 𝛽1, in the regression of organizations with more than thirty percent female directors is significantly lower than in the comparable regression at the organizations with less than thirty percent or three female directors.
The OLS regressions I run to test my hypotheses are robust and clustered by firm in order to control for differences in the standard errors because of arbitrary intra-group correlation and to control for heteroscedasticity. Additionally, I include time and industry dummies in the OLS regressions to control for the biasing effect of time trends and industry characteristics at the coefficient of my dependent variable (VOL). Furthermore, to control for outliers, I winsorize
ROA and TQ at the top and bottom 1% and VOL and FS only at the top 1% as there are no
outliers at the bottom of these variables. The cutoff for outliers considered is approximately three times the standard deviation.
23
4 Results
4.1 Descriptive statistics
Table 2 provides the descriptive statistics of the variables for the full sample and subsamples based on the number of women on the board. The mean level of the proportion of women on the board of the full sample is 0.132. This is higher compared to Sila et al. (2016) and Jizi and Nehme (2017) who find a mean of 0.096 and 0.088, respectively. This also appears in the distribution of observations of each subsample. The percentage of boards without any female directors in the full sample of Sila et al. (2016) is 37%, while in my sample is 26%. On the other hand, the percentage of organizations with three or more female directors in my full sample is 8% and 4% respectively, whereas in the sample of Sila et al. (2016) are 4% and 1%. As the examined country and sample periods of this study and of Sila et al. (2016) and Jizi and Nehme (2017) differ, comparison of these means is complicated since increase of the proportion of women on the boards is stimulated differently in the U.S. and UK and thereby the proportion of female board representation increases differently over time in the different countries. In my sample, the proportion of female directors increase from an average of 10.7% in 2007 to 16.9% in 2016. Thus, the difference in the mean level of the proportion of female directors between the study of Sila et al. (2016), Jizi and Nehme (2017) and my sample is probably caused by different sample periods. The mean annualized stock volatility as the proxy of the organization’s risk profile is 0.377. Contrary to the variable proportion of female directors, this variable, annualized stock volatility, varies less over different countries and is therefore more comparable with the other studies. The mean of the annualized stock volatility of the samples of Sila et al. (2016) and Jizi and Nehme (2017) is 0.451 and 0.378, respectively. Due to the different sample periods, the mean of annualized stock volatility of Sila et al. (2016) is significantly higher compared to the mean of annualized stock volatility in my sample (p = 0.000, t = -29.0443). Finally, the mean of the dummy variables for the absolute and relative critical mass of three or more female directors and thirty percent or more female directors are 0.118 and 0.069. This indicates that a limited number of observations exceed the absolute or relative threshold.
24
Table 3 summarizes the results of the Pearson correlation analysis performed. This analysis determines the separate correlation coefficients between the dependent, independent and control variables used in the regressions. A correlation coefficient between two independent variables that exceeds 0.8 could point out multi-collinearity. Multi-collinearity might disturb the dataset as it indicates that one independent variable could be linearly predicted by another independent variable. This might result in unreliable inferences based on the dataset. The Pearson correlation matrix shows that none of the correlation coefficients exceeds 0.8. However, it could be noticed that some of the correlation coefficients are relatively high. First striking is the correlation coefficients of PW and DRT (0.624), PW and DAT (0.585) and DRT and DAT (0.707). This correlation is expected, as these variables all express a form of female representation within the board. The variables PW, DRT and DAT are examined in different regression models, therefore multi-collinearity between these variables can be neglected. Second, the correlation coefficient of BS and FS (0.608) stand out. The correlation between these variables is in accordance with Eisenberg et al. (1998), who find that bigger firms need larger boards. Lastly, the correlation coefficient of ROA and TQ (0.493) is relatively high. This correlation can be explained by the fact that both variables express the financial performance of an organization. Furthermore, the results of the Pearson correlation analysis provide some indicative evidence for the relation expected based upon my first hypothesis. The correlation coefficient between VOL and PW (-0.210) are negative, which reflects a negative impact of the proportion of women on the annual stock volatility, which is in accordance with my first hypothesis.
[Insert Table 3 about here]
Since the variables BS, FS, ROA and TQ are examined simultaneously, multi-collinearity should be tested. The variance inflation factor (VIF) test estimates if high correlation coefficients lead to multi-collinearity within the regression. A VIF score higher than 10 indicates multi-collinearity between two or more variables affecting the regression. The results of the tests are summarized in Table 4. The table shows that no variable has a VIF value higher than 1.71, therefore there is no concern of multi-collinearity in any of the regression models.
25
Table 5 presents the comparison of means of the organization’s risk profile, VOL, in different absolute and relative categories of female board representation. This comparison of means of
VOL between different types of gender-diverse boards is tested by two-sample t-tests. The
mean of VOL is significantly lower when the number of female directors or the percentage of women on the board increases. For example, the mean of VOL of an organization with female directors is 8.24% significantly lower compared to organizations without any women on the board (t = -20.52). Seeing as the mean of VOL is sustained significant decreasing when a woman director is added to the board or when the proportion of female directors increases with ten percentage points, these results imply a negative relationship between the number or proportion of female directors and an organization’s risk profile, which is in accordance with my expectations. However, the mean of VOL is already significant increasing at boards with less female board representation than the absolute or relative threshold. For example, if the mean of VOL of a board with two female directors compared to a board with one female director is already significantly lower. This indicates that on male-dominated boards, female directors can affect an organization’s risk profile, which is contradictory to the minority theory.
[Insert Table 5 about here]
4.2 Hypothesis tests
My first hypothesis tests whether there is a negative relationship between the proportion of women on the board and an organization’s risk profile, the results of the OLS regression are summarized in Table 6. To support this hypothesis the coefficient of PW, 𝛽1, should be negative and significant. As shown in Table 6, the coefficient is -0.040 and significant at the 10% level. This result is quantitatively and qualitatively similar when tested with a fixed effect panel regression, as the results are similar to these but are not tabulated. The result implies that organizations with a higher percentage of female board representation have a lower risk profile. Consequently, I find support for my first hypothesis and confirm a negative relationship between the proportion of women on the board and an organization’s risk profile. However, the economic impact of the change in an organization’s risk profile is modest. Seeing as an average board size is 9.14 directors, replacing a male director by a female results in an increase of PW of approximately ten percent. This ten percent increase of PW results in a 0.40% lower VOL. As the mean and standard deviation of VOL are 37.7% and 17.4%, respectively, I conclude that given the magnitude and fluctuation of VOL, the decrease of 0.40% is not economically significant.
26
The results of my first hypothesis test indicate a negative relationship between the proportion of female directors and an organization’s risk profile. This conclusion is not aligned with Sila et al. (2016) who find no results, but it is in accordance with the results of Jizi and Nehme (2017). I assume that Sila et al. (2016) do not find a relationship as their examined period is more extensive. The emancipation of women is still in progress, especially when it comes to women in business. Therefore, the number of female directors and their ability to make an impact differs over time (Adams & Ferreira, 2009). Thus the broad sample period, and particularly including the early years, might cause that Sila et al. (2016) find no relationship, while my results, based on a later sample period, indicate a negative relationship between the proportion of women and an organization’s risk profile.
The regression model shows an overall adjusted R-squared of 0.588, this implies that 58.8% of the variation in the annualized stock volatility is explained by the variables of interest in the model. This is in accordance with the adjusted R-squared provided by the OLS regression of Sila et al. (2016), who find an adjusted R-squared of 0.595. Jizi and Nehme (2017) provide a slightly lower adjusted R-squared of 0.447. Regarding the control variables, I find that most of them are related to an organization’s risk profile and the sign of their impact is in accordance with my expectations. The control variables FS, BS, ROA and TQ, are negatively related to
VOL, which is in accordance with Jizi and Nehme (2017) and Sila et al. (2016). Contrary to
my expectations, I find that BI is negatively related to an organization’s risk profile. This also contradicts the findings of Jizi and Nehme (2017) and Sila et al. (2016) who find no significant relationship. Finally, I find no relationship between the control variables controlling for features of the CEO, being CEOG, CEOD and CEOT. This is in accordance with Sila et al. (2016), however Jizi and Nehme (2017) find a negative relationship between CEOD and an organization’s risk profile. Since the relation between an organization’s risk profile and the control variables is mainly in accordance with the findings of Sila et al. (2016) and Jizi and Nehme, I conclude that my sample is comparable to the previous literature.
27
Hypothesis 2 examines whether the impact of female board directors on an organization’s risk profile is stronger when the percentage of female board directors exceeds the relative threshold of thirty percent. Table 7, Columns 1 and 2, shows the change in risk profile of organizations with a proportion of female board directors below and at or above the threshold. The coefficient of PW, 𝛽1, is negative and significant of firms with a proportion of female directors below the threshold of thirty percent (𝛽1= -0.042, p = 0.007). This implies that below the threshold of thirty percent there is a negative relationship between the proportion of female directors and an organization’s risk profile. However, contrary to my expectations, the coefficient on PW, 𝛽1, is not significant for the at or above threshold firms’ regression (𝛽1= 0.046, p = 0.481). This means that there is no relationship between the proportion of women on the board and an organization’s risk profile above the threshold of thirty percent. The coefficient on PW, 𝛽1, for firms with a proportion of female directors below the threshold is not significantly different from firms with a proportion of female directors at or above the threshold (Chi2 = 0.77, p = 0.381). These results imply that the proportion of female directors do not have a stronger effect on the risk profile of an organization above the relative threshold of thirty percent female board representation. Therefore, I reject my second hypothesis.
Furthermore, this indicates that the relationship between the proportion of female directors and an organization’s risk profile, is driven by observations with a proportion of women between zero and thirty percent. The models show an adjusted R-squared of 0.585 and 0.584 for firms below and firms at or above the relative threshold, relatively. This indicates that the regression model provides sufficient explanation to determine what influences the annualized stock volatility for firms below and at or above the relative threshold. Furthermore, this indicates that the regression is comparable to previous literature (Jizi and Nehme, 2017; Sila et al.,2016).
28
Hypothesis 3 examines whether the impact of female board directors on an organization’s risk profile is stronger when there are three or more female directors, which is seen as an absolute threshold. The results of comparing the impact of women board directors on an organization’s risk profile of organizations below and at or above the threshold are shown in Table 7, Columns 1 and 2. In accordance with the results of my second hypothesis, the 𝛽1, the coefficient of PW, is negative and significant for firms with less female directors than the absolute threshold of three women on the board, however, the coefficient is not significant for organizations with three or more female directors (𝛽1= -0.047, p = 0.003; 𝛽1= 0.026 p = 0.627, respectively). The difference in the coefficients on PW, 𝛽1, is insignificant (Chi2 = 0.80, p = 0.372). This means that similar to the previous results of testing the relative threshold, that the female directors do not have a stronger impact on the risk profile of an organization above the absolute threshold of three female directors on the board. Furthermore, these results imply that the relationship between the proportion of female directors and an organization’s risk profile is driven by observations with zero to two women on the board. Accordingly, I reject my third hypothesis.
The model for firms below and at or above the absolute threshold show an adjusted R-squared of 0.580 and 0.549, respectively. This implies that the regression model provides sufficient explanation to determine what influences the annualized stock volatility for firms below and at or above the absolute threshold.
[Insert Table 8 about here]
Overall, I find no evidence that above a relative or absolute threshold the proportion of female directors has an impact on an organization’s risk profile. However, the coefficient of the proportion of women above the threshold are not significantly different from the proportion of women below the threshold, thus I conclude that there is no difference between the impact of women directors when their proportion of women directors is below or above the threshold. Still, given that women above the threshold have no significant impact on an organization’s risk profile is interesting to examine further, as this would imply that above the thresholds, the proportion of women has no effect on an organization’s risk profile. In order to test the specification of the regression, operationalization of variables and applicability of theory, I run several robustness tests. The results of the robustness tests are described in the following section.
29
5 Robustness tests
5.1 Specification of regression
My first hypothesis suggests a negative relationship between the proportion of female board directors and an organization’s risk profile. Model (3), used to test the existence of the relationship, is testing for a linear relationship. However, the negative linkage between the proportion of women and an organization’s risk profile may follow a curvilinear pattern as well, such that the effect of women on an organization’s risk profile might slightly increase or decrease when the proportion of women is higher (Ostertagová, 2002). To address this possibility, I perform a robustness test considering a curvilinear relationship.
In order to test for a curvilinear relationship, I use the same sample as for the main analysis. However, I generate an additional variable the squared value of the percentage of women directors (PW2). I estimate Model (4) presented below, based on the model of Ostertagová (2002) that tests for a curvilinear relationship. I adjusted the model by replacing the independent and dependent variables in accordance with my previous Model (3).
𝑉𝑂𝐿 = ∝ + 𝛽1 𝑃𝑊 + 𝛽2 𝑃𝑊2 + 𝛽3 𝐹𝑆 + 𝛽4 𝑅𝑂𝐴 + 𝛽5 𝑇𝑄 + 𝛽6 𝐶𝐸𝑂𝑇 +
𝛽7 𝐶𝐸𝑂𝐺 + 𝛽8 𝐶𝐸𝑂𝐷 + 𝛽9 𝐵𝑆 + 𝛽10 𝐵𝐼 + 𝛽11∑ 𝐼𝑁𝐷𝑈𝑆𝑇𝑅𝐼𝐸𝑆 + 𝛽12∑ 𝑌𝐸𝐴𝑅𝑆 + 𝜀 (4)
My first hypothesis proposes a negative relationship between the proportion of women and an organization’s risk profile. To conclude a curvilinear negative relationship, I expect 𝛽1 to be significantly negative and 𝛽2 to be significant. A positive 𝛽2 indicates a concave relationship, which indicates that the impact of women increases when the proportion of women becomes higher. In contrast, a negative 𝛽2 indicates a convex relationship whereby the impact of women decreases when the proportion of women becomes higher.
The results of the regression testing for a curvilinear relationship are shown in Table 9. Contrary to my expectations, the 𝛽1 and 𝛽2 are insignificant (𝛽1= -0.057, p = 0.221; 𝛽2= 0.049, p = 0.681). These results imply that there is no curvilinear relationship between the proportion of women and an organization’s risk profile. Thus, there is no decrease or increase in the impact of women as the proportion of women on the board increases. Therefore, I conclude a linear relationship between the proportion of female directors and an organization’s risk profile.
30
5.2 Operationalization of dependent variable
When conducting the hypothesis tests, I used VOL as a measure of an organization’s risk profile, which is in accordance with Jizi and Nehme (2017) and Sila et al. (2016). However, a difference in risk-taking behavior of the organization is not necessarily reflected in an organization’s stock volatility (Chaudhuri & Koo, 2001). To validate my results, I consider the research and development (R&D) ratio as a measure of an organization’s risk-taking behavior. The R&D ratio is the ratio of R&D expenses to total assets. According to prior literature, the R&D ratio is an appropriate measure of the organization’s risk-taking behavior, which is, among other things, determined by the board of directors (Bargeron et al., 2010; Coles et al., 2006). Spending on R&D is considered risky since the payoffs are highly uncertain. Therefore, a relatively high R&D ratio indicates an organization that is more risk-taking than average, which implies more risk acceptance by the policies of the board of the directors. As women are considered more risk-averse, I expect that there is a negative relationship between the proportion of women directors and the R&D ratio. In order to test my expectation, I use the same sample as for the main analysis. The data regarding board characteristics is consistent with the main sample. The proxy of an organizations risk profile, annualized stock volatility, is replaced by the R&D ratio (RD). The R&D expenses and total assets, necessary to determine the R&D ratio, are obtained from Compustat. I assume missing R&D expenses indicate no R&D expenditure, hence I replace missing R&D expenses by zero. Consequently, my final sample for this robustness test consists of 9,821 firm-year observations, which is equal to the main sample. Also similar to the original tests, I run OLS regression (3) to test the hypotheses. In order to test for a relative or absolute threshold, I run the OLS regression separately for organizations below the relative or absolute threshold and organizations at or above the threshold. Thereafter, I compare the significance and sign of the two coefficients in the two regressions on PW, 𝛽1.
31
Table 10 shows the results of the first OLS regression, testing for a negative relationship between the proportion of female directors and an organization’s risk-taking behavior. The adjusted R-squared is 0.247, which indicates that 24.7% of the variation in the dependent variable is explained by the model. This value is lower compared to the adjusted R-squared of the model when using annualized stock volatility as the dependent variable. This implies that the variation in R&D ratio is explained by different independent variables than VOL, however, the value is high enough to set a conclusion based on this regression model. In accordance with my expectations, the results show that the coefficient of PW is negative and significant (𝛽1 = -0.020, p = 0.047). This implies a negative relationship between the proportion of female directors on the board and the R&D ratio. Therefore, an increase in the proportion of female board directors results in a decrease of the R&D ratio. This result is consistent with the original analysis using annualized stock volatility as the measure for an organizations risk profile, therefore this result strengthens the confirmation of my first hypothesis.
In accordance with my second and third hypotheses tests, I run the OLS regressions separately for firms below and at or above the relative and absolute threshold. Table 11, Columns 1 and 2, shows the change in R&D ratio with a proportion of female board directors below and at or above the relative threshold. The coefficient of PW, 𝛽1, is negative and significant of both firms with a percentage of female directors below the threshold and firms with a proportion female directors at or above the threshold (𝛽1= -0.019 p = 0.001; 𝛽1= -0.048 p = 0.024, respectively). Contrary to my previous tests regarding annualized stock volatility and the relative threshold, there is a significantly negative relationship between the proportion of women and an organization’s risk profile above the threshold. This indicates that there is a relationship between women directors on the board and the R&D ratio both below and at or above the threshold. However, the difference in the coefficients on PW, 𝛽1, is insignificant (Chi2 = 0.93, p = 0.334). While the coefficients are not significantly different, the slope on PW is steeper above the threshold compared to below. The statistically insignificant results could be caused by the difference in standard errors, as the observations below and at or above the threshold are divided over different samples, which result in a difference in standard errors. Therefore, despite the statistically insignificant results, I argue that there is a difference between the impact of female directors below and at or above the absolute threshold. Hence, I conclude that this result supports that the impact of women on an organization’s R&D ratio is stronger above a relative threshold of thirty percent.
