Gender Diversity in Corporate Boards and Corporate Risk Taking:
A European Union Approach
Rico G. Teikotte (S2581167)
Student MSc International Financial Management, Faculty of Economics and Business, University of Groningen, the Netherlands
Article information Abstract
Article history: Draft version 21-05-2015 Received feedback 27-05-2015 Final version 19-06-2015 JEL classification: G32 J16 O52 Keywords: Risk Gender Director Diversity European Union Word count: Excluding appendix: 14,591 Including appendix: 17,829
This study extends previous research on gender diversity by investigating the effect gender diverse corporate boards have on corporate risk taking in a European Union context. By making use of a sample of 462 companies spread over 20 Member States of the European Union for the period 1999-2013, this study revealed that an increasing share of female corporate board directors results in a decrease of a firm's corporate risk taking. This result is in line with findings in psychological, sociological, and biological research explaining that women are more risk averse than men. In addition, weak evidence is found that a masculine culture strengthens the relationship between gender diverse corporate boards and corporate risk taking. Lastly, no evidence is found that the severity of a country's gender quota moderates the relationship between gender diverse corporate boards and corporate risk taking.
Journal of International
Financial Management
1 Introduction
Events as the past corporate governance scandals, the Sarbanes-Oxley act, and the financial crisis reinforced the attention on corporate board composition (Terjesen, Sealy & Singh, 2009; Dalton & Dalton, 2010). An issue in today's board composition relates to the worldwide underrepresentation of women (Szydło, 2015), causing an underutilization of female talent (European Commission, 2012), and raises concerns and attention from practitioners, politics and scholars (Terjesen, Aguilera & Lorenz, 2014). Of the United States' Fortune 500 firms' corporate boards in 2013, only 16.9% of all board seats, including all positions of board leadership, were represented by women, which is a minor increase compared to 16.6% in 2012 (Catalyst, 2013). In Australia, only 12% of the board of directors' chairs are held by women on March 31 2014 (Workplace Gender Equality Agency, 2014). On October 2014, regarding the largest publicly listed firms of the 28 European Union Member States (EU-28), on average 20.2% of the corporate board members consisted of women (European Commission, 2015). This might feel contradicting considering the fact that 60% of the university graduates in the European Union (EU) are female (European Commission, 2015). Several voluntarily and legislative measures are accepted by the EU as a whole and by its individual Member States to foster female board members. These measures resulted in a rapid increase of the share of female directors in EU firms' corporate boards (European Commission, 2015) and therefore, this study will focus on this continent.
To date, gender diversity on corporate boards has experienced lots of attention. However, current literature especially focused on the relationship between gender diversity and firm performance (Perrault, 2014). Empirical evidence regarding this relationship found contradictory results (Choudhury, 2014). By looking directly at the relationship between board diversity and firm performance, scholars might have overlooked other important firm performance determinants. Therefore, scholars proposed to study other, more proximate relationships to examine the effect gender diversity has on firms (Johnson, Schnatterly & Hill, 2013; Choudhury, 2014; Ferreira, 2014; Hillman, 2014; Perrault, 2014).
This paper aims to respond to this research call by examining the effect an increasing share of female corporate board members has on corporate risk taking. Corporate risk taking will be studied as managerial risk choices and risk taking play a key role regarding decision making and is an important determinant for firm growth, performance and survival (Boubakri, Mansi & Saffar, 2013).
relevant hypotheses (see Section 4) do not explicitly focus on executive or non-executive directors. Both type of directors have the ability to impact a firm's corporate risk taking and both positions are underrepresented by women in the EU (European Commission, 2015).
The risk preference of men and women is broadly examined e.g. in the field of psychology and finance. These studies found that women are more risk averse than men (e.g. Eckel & Grossman, 2008). Therefore, a women's risk preference may influence corporate risk taking as the share of female corporate board members increases. Based on the preceding, the following research question applies: Does an increasing share of female corporate board
members affect a firm's corporate risk taking? Two sub-questions are formulated based on the
European Union approach of this paper. These sub-questions are discussed next.
The fist sub-question is related to a country's culture. By doing so, this paper scrutinizes whether country differences in the EU have an effect on the relationship between an increasing share of female corporate board members and a firm's corporate risk taking. Hofstede distinguishes five cultural dimensions whereby each cultural dimension specifies different characteristics of the behavior of people (Hofstede, 1998). In a masculine country men are seen as assertive and tough, whereas women are characterized as modest and tender (Hofstede, 1998). Therefore, a country's culture may influence the effect an increasing share of female board members may have on corporate risk taking as the level of masculinity influences an individual's behavior. Therefore, the second research question is: Is the impact
of an increasing share of female corporate board members on corporate risk taking affected by a EU Member States' masculinity level?
on corporate risk taking affected by a EU Member States' corporate board gender diversity stimulation approach?
Based on a sample of 462 firms spread over 20 EU Member States, this study found that an increasing share of female corporate board directors will lead to lower corporate risk taking. Further, weak evidence is found that a masculine culture strengthens the relationship between an increasing share of female corporate board directors and corporate risk taking. Lastly, no support is found that a country's gender diversity stimulation approach moderates the relationship between an increasing share of female corporate board directors and corporate risk taking.
This study provides the following contributions: Firstly, this study contributes to the existing literature that already focused on the relationship between gender diversity and corporate risk taking e.g. Adams & Funk (2012) and Mihet (2012) because, to the extent of the author's knowledge, no study exists that researched the effect gender diverse corporate boards have on a firm's corporate risk taking in a multi-country context. Therefore, this study contributes to the current literature as it is possible in this study to research the EU context as a whole and to examine the influence country specific characteristics, e.g. legal and cultural, may have on the relationship between the share of female corporate board directors and corporate risk taking. Secondly, this study provides contributions in the field of corporate risk taking as this study increases our understanding whether women are, compared to men, more risk averse in a business environment. Lastly, this study provides valuable insights for politics as a lot of debates around the subject of gender diversity in corporate boards have been conducted and still continues. The results found in this study may contribute to these debates as they shed light on the effect gender diversity has on corporate risk taking in a EU context as a whole and between individual EU Member States.
2 Gender diversity in the European Union
In order to accelerate the share of women in the EU-28 firm's corporate boards, the European Commission adopted a directive on November 14th 2012 that sets a quantitative objective that a firm's non-executive board members should at least consist for 40% of both genders by the year 2020 (European Commission, 2014). The directive applies to all large firms listed on a stock exchange of a EU Member State, whereas small and medium-sized firms are exempted (European Commission, 2014). However, this directive should not be seen as a 'rigid
quantitative quota obligation that would result in sanctions if it is not reached.' (European
Commission, 2014, p.1).
Besides the overall directive imposed by the European Commission, the EU-28 developed voluntary initiatives and legislative quotas to increase the share of female board members. For instance, Belgium (33%)1, France (40%), and Italy (33%) installed a gender quota, which in case of non-compliance results in sanctions (Catalyst, 2014). Belgium set different compliance dates based on the firm's characteristics. Belgian state owned enterprises, large publicly listed firms, and small publicly listed firms have to comply between 2011-2012, 2017-2018 and 2019-2020 respectively (Catalyst, 2014). French firms have to comply on January 2017 (Catalyst, 2014) and Italian firms by 2015 (European Commission, 2013). An example of a less rigid measure can be found in The Netherlands where a 'comply or explain' mechanism is applied whereby firms with a board of directors not consisting for at least 30% of both men and women have to explain each year why this target is not met (Catalyst, 2014). This Dutch measure will cease in 2016 (European Commission, 2013). A recent development is the accepted gender quota in Germany on March 6 2015, that goes into effect by 2016, whereby the board of directors of the largest firms have to be composed of at least 30% of both genders, and forces medium sized firms to hand in plans to strive to gender diversity in their corporate boards (Aljazeera America, 2015, Smale & Miller, 2015).
Significant differences regarding the gender composition of corporate boards in the EU were present between October 2010 and October 2014. According to the European Commission (2015), firms in France (+20%)2, Italy (+19.6%), Belgium (+11.9%), Germany (+11.8%), the United Kingdom (+10.8%), and Slovenia (+10.1%) faced the largest increase of female director members in their corporate boards. Countries that installed legislation, are considering to install legislation, or had an intensive public debate on this topic present the most significant improvements (European Commission, 2015). According to the European
1 The quota percentages are shown in brackets behind the country. 2
Commission (2015), on October 2014, the average share of female directors in corporate boards of the largest publicly listed firms in the EU-28 is 20.2%, whereas on October 2013, on average 17.8% of the corporate board members consisted of female directors.
Still, large variations are present regarding the share of female corporate board members between the large publicly listed firms in the EU-28. On October 2014, countries as France (32.4%)3, Latvia (31.7%) and Finland (29.2%) have the highest representation of women in their largest publicly listed firm's corporate boards, and countries as Malta (2.7%), Czech Republic (3.5%) and Estonia (7.1%) the lowest (European Commission, 2015).
3 The importance of gender diversity and findings of previous studies
The capabilities of gender diverse boards differ from homogeneous boards because in gender diverse boards, the values of both genders have to be taken in to consideration. Female directors can fulfill the more tender and feminine role, and men the more assertive and masculine role (Hofstede, 1998). In accordance with this, Schwartz & Rubel (2005) found that men and women place a different level of importance on different values. They found that men attribute more importance to self-enhancement values (e.g. power), whereas women place more importance on self-transcendence values (e.g. benevolence). Consistent with this finding of Schwartz & Rubel (2005), Eagly, Johannesen-Schmidt & van Engelen, (2003) argued that in comparison with a men's leadership style, a women's leadership style is more transformational and transactional oriented, characterized by nurturing, inspiring, and rewarding employees. However, above described findings are based on general population groups and therefore these finding may not generalize to corporate board members (Adams & Funk, 2012).
Based on the preceding, it can be argued that gender diversity matters as the behavior of both genders differ from each other. These behavioral differences are important as the board of directors have to interact between the firm and its environment, whereby gender diverse boards gain access to information that could not be obtained by homogeneous boards (Hofstede, 1998). Besides an improved connection of the firm with its environment, gender diverse boards perform a better monitoring function (Adams & Ferreira, 2009). These two observations are important because they belong to the prime functions of the board of directors (Carter, D'Souza, Simkins & Simpson, 2010). Therefore, gender diverse boards are essential in order for a firm to be successful (Hofstede 1998).
However, researchers that focused on the relationship between board diversity and firm performance found mixed results. One literature stream argues a positive relationship between board diversity and firm performance due to, for instance, an increase in creativity, innovativeness, and problem solving (Erhardt, Werbel & Shrader, 2003; Campbell & Vera, 2010; Ntim, 2015), whereas other scholars argue a negative relationship between board diversity and firm performance because female board directors are more hesitant to fire employees, well governed firms may become over-monitored, and when mandatory gender quotas are in place shareholders are restricted to choose their desired board (Adams & Ferreira, 2009; Ahern & Dittmar, 2012; Matsa & Miller, 2013). These conflicting findings indicate that confusion exists about the influence gender diverse boards may have on firm performance (Ferreira, 2014).
Another research stream focused on more proximate outcomes of gender diversity on corporate boards, such as the effect of gender diverse corporate boards have on a firm's Corporate Social Responsibility efforts (Bear, Rahman & Post, 2010), on a board's legitimacy and trustworthiness (Perrault, 2014), on a firm's reputation (Brammer, Millington & Pavelin, 2009), and on a board's attendance record and monitoring function (Adams & Ferreira, 2009). This paper will contribute to this research stream by investigating the effect gender diverse corporate boards may have on corporate risk taking.
4 Literature review and hypotheses development
4.1 Gender diversity and corporate risk taking
Men and women differ significantly regarding their risk taking behavior, where women are perceived to be more risk averse than men. This observation is e.g. found in the field of psychology (Byrnes, Miller & Schafer, 1999) and in the field of economics and finance (Jianakoplos & Bernasek, 1998; Eckel & Grossman, 2008). This sub-section will provide a literature overview with possible explanations why gender risk taking differences exist by discussing several theories provided by psychological, sociological and biological research.
4.1.1 Psychological
The first psychological theory that will be discussed is the self-efficacy theory. Self-efficacy may play an important role regarding risk taking preferences as self-efficacy is defined by Bandura (1986, p. 391) as "people's judgments of their capabilities to organize and execute
the skills one has but with judgments of what one can do with whatever skills one possesses".
When an individual believes that it can handle certain tasks it will undertake these tasks, whereas individuals tend to avoid tasks when they believe they don't possess the right capabilities to perform them (Bandura, 1986, 2001). Executives with a high level of self-efficacy, i.e. they believe that they and their firm have the right capabilities to perform certain tasks, are more willing to engage in risky activities than executives with a low level of self-efficacy (Krueger & Dickson, 1994). Gender differences exist regarding self-self-efficacy where is found that men have a higher level of self-efficacy than women (Rosenthal, Guest & Peccei, 1996) and therefore men are more willing to take on risky activities (Forlani, 2013)
Secondly, sensation seeking may explain the risk taking differences between men and women. Sensation seeking is defined by Zuckerman (1994, p. 27) as follows "Sensation
seeking is a trait defined by the seeking of varied, novel, complex, and intense sensations and experiences, and the willingness to take physical, social, legal and financial risks for the sake of such experience". Individuals scoring high on sensation seeking are willing to bear risk for
the sensations and experiences it entails (Zuckerman 1994). Zuckerman (2007) states that sensation seeking and risk taking behavior are related to each other in every risk taking area, including financial risk taking. Several studies concluded that men score higher on sensation seeking than women, which implies that men are more willing to take on risky activities for the sake of the experience (Zuckerman, 1994, 2007).
Lastly, differences in core values between both genders may explain the difference in risk taking between men and women. Schwartz & Rubel (2005) found that women place more importance on the security and tradition values compared to men. The security and tradition values can be described as "Safety, harmony, and stability of society, of relationships, and of
self" and "Respect, commitment, and acceptance of the customs and ideas that traditional culture or religion provide the self", respectively (Schwartz & Rubel, 2005, p.1011). It is
expected that the gender difference regarding the importance of the security value may give rise to gender risk taking differences because it is argued that people tend to act in accordance with their values (Verplanken & Holland, 2002; Bardi & Schwartz, 2003). Women may try to avoid risky activities in order to satisfy their security value, whereas men place less emphasize on the security value and are therefore more willing to engage in risky activities.
4.1.2 Sociological
2002). Men and women had traditionally different social roles, where women took care for their offspring and men had the role to hunt (Wood & Eagly, 2002). These differences in social roles were created as it was crucial for the offspring's survival to enjoy the care of its mother which resulted in a higher risk averseness of women (Campbell, 1999). It is also argued that men take more risk than women because men historically faced more competition and threat than women in order to receive the right to mate, and in order to increase their social status (Wilson & Daly, 1985). The risk attitude based on ancient social roles may have an influence in today's risk preference as well, whereby women are perceived to be more risk averse than men.
Secondly, gender stereotypes may provide an explanation of gender risk taking differences as well. Gender stereotypes are formed to be able to create an expectation of an individual's behavior (Diekman, Eagly & Kulesa, 2002). Role theorists identified the expectations of the society regarding the behavior of men and women, whereby risk taking and the expectation of men to remain calm when dealing with dangerous events are considered to be a part of the male role (Becker & Eagly, 2004). This expectation that men are risk takers can actually result in risk taking behavior of this group as people tend to behave in accordance with their gender stereotype when ambiguous or confusing events occur, and to enhance their social approval and self esteem (Eagly, 2009). Therefore, the stereotype that men are risk takers may actually result in risk taking behavior of men, and therefore men may be more willing to bear risks compared with women.
Lastly, Grossman & Wood (1993) argue that women experience emotions more intense than men do which may be caused by social role differences between men and women. Differences in daily activities whereby women act more often than men as caretakers, and gender stereotypes that women are more emotionally intense are social role difference examples (Grossman & Wood, 1993). It is argued that women experience negative outcomes with a worse feeling than men do, which might negatively impact their risk taking willingness (Croson & Gneezy, 2009), as risk can be defined as: "The possibility of suffering harm or
loss; danger" (Zuckerman, 2007, p. 52). This definition indicates that the possibility of harm,
loss, or danger may increase when the level of risk increases, which is more intensely experienced by women and, therefore, making women more risk averse.
4.1.3 Biological
Kuhlman (2000), a positive relationship exists between dopamine and risk taking, meaning that when the level of dopamine increases, the persons' risk appetite increases with it. Conversely, they argue that a negative relationship exists between serotonin and risk taking. Evidence is found that men have a more reactive dopaminergic system compared with women, which explains why men take more risks than women (Zuckerman & Kuhlman, 2000).
Besides the level of dopamine and serotonin, the level of an individual's testosterone may also be a determinant of a person's risk appetite as men with high levels of testosterone behave less social, are more willing to take on risk and experience less stable sexual relationships (Archer, 2006). It is found that men have significantly higher levels of testosterone compared with women, which makes men more willing to accept risk (Sapienza, Zingales & Maestripieri, 2009).
Based on evidence found in the field of psychology (self-efficacy, sensation seeking, and values), sociology (ancient social roles, stereotypes, and emotions) and biology (dopamine, serotonin and testosterone) that women are more risk averse than men, the following hypothesis holds:
Hypothesis I: An increasing share of female corporate board directors will result in a
decrease in corporate risk taking.
An important comment has to be made regarding the just discussed theories. The discussed theories provided by psychological, sociological and biological research do not specifically focus on corporate board members. Therefore, it is possible that specific population groups e.g. corporate board members may deviate from the results found in these theories (Adams & Funk, 2012).
4.2 Gender diversity, corporate risk taking, and masculine/feminine culture
Cultural differences may have an impact on the relationship between the share of female corporate board members and corporate risk taking. A country's culture can be described as "The collective programming of the mind that distinguishes one group or category of people
from another", (Hofstede & McCrae, 2004, p.58) and implies that culture is a collective rather
Figure I: Masculinity scores by gender (Hofstede, 1998, p. 86). Hofstede's culture classification is often used to describe the cultural distance between countries as his dimensions are simple and straightforward (de Mooij & Hofstede, 2010)4. Hofstede distinguishes the following dimensions: power distance, uncertainty avoidance, individualism/collectivism, masculinity/femininity, and long term orientation (Hofstede & McCrae, 2004). Especially the masculinity/femininity dimension is expected to be of special importance for this study. Men living in a country that is classified as masculine are characterized as assertive, tough, and focused on material success, whereas women are classified as modest, tender, and concerned with the quality of life (Hofstede, 1998). In feminine countries both men and women are characterized as modest, tender, and concerned with the quality of life (Hofstede, 1998). Thus, the masculinity/feminine dimension predicts a constant disparity between men and women, except for feminine countries. Figure I depicts this relationship.
It can be argued that the difference in gender risk taking willingness is the largest in very masculine countries and the smallest in very feminine countries because Hofstede (1998) states that values of men and women differ the most in very masculine countries. It is therefore expected that the decrease in corporate risk taking accelerates when the share of female corporate board directors increases in masculine countries because of the large deviation in gender risk taking preference. Therefore, hypothesis II holds:
Hypothesis II: The negative effect an increasing share of female corporate board
directors have on corporate risk taking increases as the level of a country's masculinity increases.
4.3 Gender diversity, corporate risk taking, and legislation
Norway introduced an obligatory gender quota of 40% when on average only 9% of the Norwegian firms' corporate boards consisted out of female directors at that time (Ahern & Dittmar, 2012). The Norwegian female talent pool was not large enough to absorb the demand for qualified female directors resulting in boards with less experience, increases in a firm's
4
See for an extensive discussion about Hofstede's five dimensions: Kirkman, Lowe & Gibson (2006) Tough
Men
Women
Tender Tough
50
Country Masculinity Index
leverage, increases in a firm's acquisition activities, and deterioration of a firm's operating performance (Ahern & Dittmar, 2012).
The same situation may arise in a EU Member State. An obligatory gender quota may force firms to recruit female board directors because they have to but not because these women are necessarily suited for the job (Matsa & Miller, 2013). A well performing male director may be replaced by a female director, or female directors are added to a firm's board to conform to the gender quota without having the needed experience and knowledge for the function (Matsa & Miller, 2013). It could be the case that these female directors face difficulties in assessing a project's riskiness because they lack the knowledge to do so. The result may be that men do not face resistance of female directors in accepting a high risk project. Therefore, the preference of male board directors to engage in risky projects is not tempered by the female directors' preference of risk averseness because these women do not possess the qualifications (e.g. knowledge, experience) to do so.
This expectation differs from the previous developed hypotheses in the sense that in this case not the differences in risk taking behavior between both genders is examined, but the aim is to examine whether mandatory gender quotas installed by countries will lead to less qualified women in corporate boards, resulting in higher corporate risk taking compared with countries that installed non-mandatory gender quotas. However, when corporate risk taking differences are found between countries that installed mandatory gender quotas and countries that installed non-mandatory gender quotas, it is not possible in this setting to determine the cause of these differences e.g. due to gender differences in knowledge level or experience level. Hypothesis III is stated as follows:
Hypothesis III: The negative effect an increasing share of female corporate board
directors have on corporate risk taking decreases when a country installs mandatory gender quotas.
Figure II: Conceptual model, describing the proposed relationships
It is expected that (1) an increasing share of a firm's female corporate board directors will decrease a firm's corporate risk taking, (2) that a masculine country culture strengthens the first expectation, and (3) that legally enforced gender quotas will weaken the first expectation.
5 Methodology, variable description, sample selection, and descriptive statistics
5.1 Methodology
An empirical model is constructed for each hypothesis. These models are tested through an Ordinary Least Squares (OLS) cross section regression (John et al., 2008) for the year 2013. The following empirical models are constructed:
Hypothesis I: (1) Hypothesis II: (2) Hypothesis III: (3)
represents the constant, is the error term of the model, i specifies a firm, and c specifies a country. An overview of the other variables used in the three empirical models can be found in Table 8 in the Appendix. A comprehensive description of the variables used in Models (1), (2), and (3) can be found in Section 5.2.
Hypothesis II and III are structured as a conditional hypothesis and therefore interaction terms are needed to test these hypotheses (Brambor, Clark & Golder, 2005). Brambor et al. (2005) are followed in order to structure these regression model correctly. Brambor et al. (2005) state that the interaction term and the components of the interaction term have to be included in the regression model. Model (2) and (3) satisfy these requirements because the components of the interaction terms , and and are included separately in Model (2) and (3) respectively.
5.2 Dependent variable, independent variables, and control variables
John et al. (2008) is followed in computing the proxy for corporate risk taking by calculating the volatility of the firm's Return On Assets (ROA)5. This dependent variable is presented in the empirical models as . A firm's ROA is a ratio computed as the earnings before interest, taxes, depreciation, and amortization (EBITDA) divided by the firm's total assets (John et al., 2008). The ROA is calculated per year per firm for the period 1999-2013 whereby at least 5 years total assets and EBITDA data per firm is required (John et al., 2008). Subsequently, the standard deviation of the firm's ROA over the period 1999-2013 is computed (John et al., 2008). This results in one standard deviation of the ROA per firm which is a proxy for a firm's operational risk taking, because when a firm invests in riskier projects and has riskier operations it is expected that the firm's return on investments has a higher volatility (John et al., 2008).
The independent variable is represented in the empirical models as which is the share of female corporate board directors per firm per 2013. The share of female corporate board directors includes both executive directors and non-executive directors because the theories discussed in Section 4.1 do not explicitly focus on executive or non-executive directors. Both type of directors have the ability to impact a firm's corporate risk taking and both positions are underrepresented by women (European Commission, 2015). Therefore, both director types are included in the variable. Hypothesis II and III make use of additional independent variables. Regarding Hypothesis II, represents the masculinity dimension. Regarding Hypothesis III, are dummy variables that represent the first two country classification groups. is one when country c is present in country classification group one and zero otherwise. is one when country c is
present in country classification group one and zero otherwise. is not included in the model to prevent perfect correlation (Brooks, 2008). Section 6.3 will provide the rationale behind the construction of these country classification groups.
The following control variables are included based on Faccio et al. (2011), Boubakri et al. (2013), and Bruno & Shin (2014) who have found that these variables have an impact on corporate risk taking. The abbreviation per control variable that is used in the empirical models are presented in brackets. (1) The natural logarithm of a firm's total assets ( ) is used as control variable because small firms are typically perceived to behave more risky. The total assets are comprised of both current and fixed assets. (2) The ratio of debt to assets ( ) is included to control for the financial risk of a firm. It is calculated by dividing a firm's debt over its assets. Debt includes both the long and short term debt obligations, and assets includes both the current and fixed assets. (3) The market-to-book ratio ( ) is included to control for a firm's growth opportunities because more growth opportunities may result in more corporate risk taking. (4) A firm's ROA ( ) is included as control variable to control for the quality of the management. A well performing management team is capable to generate stable returns over their assets. Poor management teams may have more trouble with generating stable returns while this is not attributable to risk taking decisions but to the quality of management. (5) GDP growth ( ) is used to control for a country's overall growth. It is important to include this measure as lower country-level growth could be related to a higher firm's earnings volatility because a firm's risk willingness in a low country-level growth environment may increase to maintain the firm's profit levels. (6) Industry ( ) and country dummies ( ) are included to control
for industry and country effects that might influence corporate risk taking. Currency differences between countries are not an issue because of the use of ratios.
5.3 Sample selection
from the sample (Bruno & Shin, 2014). A sample of 532 firms remained after this correction. An additional data correction was necessary because some directors were entered multiple times in a particular board composition year. Not deleting these directors would result in a biased percentage of female corporate board directors. Accounting data is retrieved from DataStream by making use of the firm's ISIN code provided by BoardEx. DataStream had not accounting data available for all the firms introduced by BoardEx. A sample of 462 companies were available after matching the firms provided by DataStream with the firms provided by BoardEx. Delisted firms are retained in the final sample although Norway experienced that firms delisted because of the introduction of gender quotas (Ahern & Dittmar, 2012). The author is not aware that this also occurred in one of the countries studied in this paper, and therefore the delisted firms are maintained in the used sample. The final sample consists of unbalanced data including 462 firms spread over 20 EU Member States for the period 1999-2013.
5.4 Descriptive statistics
A comprehensive overview regarding the sample composition can be found in table 1. The data is retrieved from 20 countries, including in total 462 firms. On average in 2013, 2% of the corporate board members are female executive directors and 12% of the corporate board members are female non-executive directors. The percentage of female directors in the used sample is not comparable to the claim of the European Commission that the board of directors in the EU-28 consist for 17.8% of female directors at November 2013 (European Commission, 2015). This discrepancy can be explained by the fact that the European Commission only takes large publicly listed firms into account, whereas the used sample also includes medium and small sized firms.
Table 2 presents the descriptive statistics of the cross sectional data. The variables RISK1, FEMR, LNTA, D/A, ROA, and GDPGR are winsorised at the 0.01 and 0.99 level to reduce the effect extreme values might have on the outcome. The MTB variable is winsorised at the 0.05 and 0.95 level because of the high level of extreme values. The firms' maximum standard deviation of the ROA is 1.86 and the minimum is 0.009. The female board representation is on average 12.25%6. The average masculinity score is 45.68 with a maximum of 88 and a minimum of 5. The maximum and minimum value of the GRPD1, GRPD2, and GRPD3 variables are one and zero respectively because these variables are dummy variables that only take one or zero as value. The average firm size is 14.68, and the
6
average debt to assets ratio is 28.01%. The market to book ratio is on average 2.05, and the average return on assets is 9.97%. The GDP growth is on average -0.24% with a maximum of 1.99%.
Table 1: Sample composition
This table gives an insight in the composition of the final sample. The total number of firms per country is indicated, even as the percentage of female board directors (% Females) as of 2013, and their function: executive director (Exc. Dir.) or non-executive director (Stnd. Dir.).
Table 2: Descriptive statistics, cross sectional data
This table presents the descriptive statistics of the total sample. RISK1 is the standard deviation of a firm's ROA for the period 1999-2013, where the period is annual. FEMR represents the percentage females on a firm's corporate board, including both executive and non-executive directors. D/A stands for the debt-to-asset ratio whereby a firm's debt is divided over a firm's total assets. MTB is the market-to-book ratio. LNTA stands for the natural logarithm of a firm's total assets, and GDPGR represents the GDP growth per country per year. MAS represents the masculinity scores per country, and GRPD1, GRPD2, and GRPD3 are the three country classification variables to classify firms according to the severity of their gender diversity measures where GRPD1 includes the countries with the most rigid gender diversity measures and GRPD3 includes the countries with the least severe gender diversity measures.RISK1, FEMR, LNTA, D/A, ROA, and GDPGR are winsorised at the 0.01 and 0.99 level. MTB is winsorised at the 0.05 and 0.95 level.
Variable Mean Median Maximum Minimum Std. Dev. Observations
RISK1 0.1103 0.0557 1.8613 0.0092 0.2249 462 FEMR 0.1225 0.1111 0.4109 0.0000 0.1095 412 MAS 45.6834 50.0000 88.0000 5.0000 24.4269 458 GRPD1 0.1602 0.0000 1.0000 0.0000 0.3672 462 GRPD2 0.5455 1.0000 1.0000 0.0000 0.4985 462 GRPD3 0.2944 0.0000 1.0000 0.0000 0.4563 462 LNTA 14.6791 14.6676 19.3690 8.6073 2.1022 398 D/A 0.2801 0.2528 1.1893 0.0000 0.2181 398 MTB 2.0475 1.5150 7.2680 0.0000 1.8591 462 ROA 0.0997 0.1035 0.5380 -0.4335 0.1162 386 GDPGR -0.0024 0.0011 0.0199 -0.0332 0.0136 462
Country Nr. Firms Females Exc. Dir. Stnd. Dir.
Table 3: Pearson correlation matrix, cross sectional data, Hypothesis I
This table presents the correlation matrix and includes the independent variable (RISK1), the dependent variable (FEMR), and the control variables (LNTA, D/A, MTB, ROA, and GDPGR). The correlation matrix is based on the winsorised data.
** indicates that the correlation is significant at the 0.01 level. * indicates that the correlation is significant at the 0.05 level.
1 2 3 4 5 6 7 1 RISK1 1 2 FEMR -0.12 * 1 3 LNTA -0.35 ** 0.29 ** 1 4 D/A 0.01 -0.01 0.14 ** 1 5 MTB 0.14 ** 0.11 * -0.02 -0.18 ** 1 6 ROA -0.30 ** 0.10 * 0.26 ** -0.11 * 0.22 ** 1 7 GDPGR 0.06 0.27 ** 0.19 ** -0.19 ** 0.14 ** 0.06 1 6 Hypothesis Testing
Before testing the hypotheses, a White heteroskedasticity test is performed to test whether the winsorised data is biased by heteroskedasticity. The heteroskedasticity test is shown in table 12 in the Appendix. The results indicate that heteroskedasticity is present in the winsorised cross sectional data and therefore the error terms are estimated using White's Modified Standard Error Estimates (WMSEE) (Brooks, 2008).
6.1 Hypothesis I
coefficient become statistically significant at the 0.05 significance level. Model (4) complements Model (2) by adding country dummies. This adjustment results in changes compared to Model (2) in several ways. First, the D/A coefficient becomes statistically significant (P<0.1) and positive. This contradicts existing literature stating that firms with more debt face more scrutiny from their debt holders which decreases their corporate risk taking (John et al., 2008). Second, the GDPGR coefficient becomes negative which is in line with the literature stating that a lower GDP growth results in higher corporate risk taking (Bruno & Shin, 2014). Third, the GDP growth coefficient becomes statistically significant at the 0.1 significance level. Model (5) includes all variables, country dummies, and industry dummies. The FEMR coefficient is negative and statistically significant (P<0.01), the LNTA coefficient is negative and statistically significant (P<0.01), the MTB coefficient is positive and statistically significant (P<0.05), and the ROA coefficient is negative and statistically significant (P<0.05). The D/A, and GDPGR coefficients are both insignificant. Throughout all models, the FEMR coefficient is negative and statistically significant, ranging between a 0.01 and 0.05 significance level. Therefore, Hypothesis I is supported.
6.2 Hypothesis II
Hypothesis II is tested by using a masculinity variable (MAS) that represents the masculinity score of the EU Member States present in the sample. The masculinity scores are retrieved from http://geert-hofstede.com/.
6.3 Hypothesis III
The method of Abiad, Detragiache & Tressel (2008) is followed in transforming a country's rules and regulations in testable variables. Abiad et al. (2008) constructed several classification groups per country to research financial reforms. Hypothesis III will be tested in a similar way by classifying countries based on the severity of the installed gender stimulation measures. Countries are divided into three groups, described as follows7:
Group 1: Countries that installed sanctions, e.g. fines or dismissal of current board members.
Group 2: Countries that did not install sanctions but require firms to, for instance, report their actions in striving to a gender diverse board in an official report, e.g. annual report or corporate governance report.
Group 3: Countries that did not install sanctions, and do not require firms to report their actions in striving to a gender diverse board in an official report e.g. annual report or corporate governance report.
Table 6 depicts the cross sectional regression output and presents three models. Model (1) includes the dependent variable RISK1, the FEMR variable, the two country group dummies GRPD1 and GRPD2, and the interaction terms GRPD1*FEMR and GRPD2*FEMR. The FEMR coefficient is negative but statistically insignificant. The GRPD1 coefficient is statistically significant (P<0.01) and negative. The coefficient of the interaction term GRPD1*FEMR is positive but statistically insignificant, and the coefficient of the interaction term GRPD2*FEMR is negative but also statistically insignificant. Therefore, no support is found for Hypothesis III in this model. After including the control variables in Model (2), the FEMR variable has a negative coefficient but is still statistically insignificant. GRPD1 and GRPD2 are both statistically insignificant and have a negative and positive coefficient respectively. The coefficient sign of the interaction terms remain the same and are still statistically insignificant. Model (3) extents Model (2) by including industry dummies. The regression output in Model (3) shows that also in this model the FEMR coefficient is negative but statistically insignificant. The GRPD1 coefficient is positive but insignificant, and the GRPD2 coefficient is positive and statistically significant (P<0.1). Both interaction terms have a negative and insignificant coefficient. Note that the FEMR variable is statistically
7
insignificant in all three models. This may be caused due to correlation effects. Based on the results found in all three models, it can be concluded that Hypothesis III is not supported because the firms positioned in countries classified in group 1 show no statistically significantly higher corporate risk taking when the share of female board directors increases, compared with firms positioned in countries classified in group 3. The GRPD1*FEMR interaction term even shows a negative coefficient which contradicts Hypothesis III.
6.4 Female Representation: dummy Variable
The presence of female corporate board directors is measured by using the share of female corporate board members in all previous discussed regressions. However, other authors (e.g. Adams & Ferreira, 2009) use a dummy variable to measure the presence of female corporate board members. To test whether the results of this approach are similar to the results found in Sections 6.1, 6.2, and 6.3, a dummy variable FEMRD is created that is one if the firm has female corporate board members and zero otherwise. The results of this regression can be found in table (7). This table includes three models. Model (1), (2), and (3) represent Hypothesis I, II, and III respectively.
Firstly, Model (1) of table (7) shows that the FEMRD coefficient has a negative sign and is statistically significant (P<0.05). This is in line with the findings from table 4, however in table 4 Model (5) the FEMR coefficient sign has a higher statistical significance level (P<0.01). Another deviation can be found regarding the GDPGR coefficient. In table 4 Model (5), the GDPGR coefficient is not statistically significant whereas the GDPGR coefficient in table 7 Model (1) is statistically significant at the 0.1 significance level.
Secondly, Model (2) of table 7 shows similar results compared with table 5 Model (3). However, the FEMR, MAS, and the GDPGR coefficient in table 7 Model (2) are statistically significant at the 0.1 level compared with a significance level of 0.05 in table 5 Model (3).
Thirdly, Model (3) in table 7 shows some deviations with the results found in table 6 Model (3). The GRDPD1 coefficient has changed in sign and the GRPD2 coefficient is not statistically significant anymore. Also the interaction term GRPD1*FEMR coefficient is changed in sign and the D/A coefficient is not statistically significant anymore. The GDPGR coefficient has become statistically insignificant in table 7 Model (3).
Table 4: Hypothesis I, cross section results
The table presents the regression results based on the winsorised data, whereby 5 different models are specified. Every model includes more or different variables. Model (5) represents the described empirical Model (1). 2013 is the year included in the cross sectional regression. All models have White's Modified Standard Error Estimates (WMSEE) to control for heteroskedasticity issues. The T-statistics are presented in brackets.
*** indicates that the coefficient is significant at the 0.01 level. ** indicates that the coefficient is significant at the 0.05 level. * indicates that the coefficient is significant at the 0.1 level.
Dependent variable: RISK1
(1) (2) (3) (4) (5) CONSTANT 0.1310 *** 0.4714 *** 0.4250 **** 0.6921 *** -0.0363 (6.8486) (5.6734) (6.2376) (3.5140) (-0.1112) FEMR -0.2318 *** -0.1373 ** -0.1408 ** -0.2073 ** -0.2098 *** (-2.7593) (-2.1038) (-2.1965) (-2.2189) (-2.6441) LNTA -0.0240 *** -0.0223 *** -0.0262 *** -0.0241 *** (-4.8522) (-4.9382) (-4.7764) (-4.9827) D/A -0.0005 -0.0546 0.0180 * -0.0442 (-0.0126) (-1.5690) (0.4926) (-1.2482) MTB 0.0184 * 0.0208 ** 0.0194 * 0.0214 ** (1.7972) (1.9993) (1.8836) (2.0284) ROA -0.4110 ** -0.4530 ** -0.3950 ** -0.4150 ** (-2.0289) (-2.2081) (-1.9805) (-2.0754) GDPGR 1.6557 *** 1.5093 ** -110.2979 * 210.0634 (3.0230) (2.2384) (-1.7030) (1.4553)
Industry dummies NO NO YES NO YES
Country dummies NO NO NO YES YES
WMSEE YES YES YES YES YES
Observations8 412 383 383 383 383
R2 0.0140 0.2183 0.4341 0.2660 0.4745
Adjusted R2 0.0116 0.2058 0.3679 0.2168 0.3804
8
Table 5: Hypothesis II, cross section results
The table presents the regression results based on the winsorised data, whereby 3 different models are specified. Every model includes more or different variables. Model (3) represents the described empirical Model (2). 2013 is the year included in the cross sectional regression. All models have White's Modified Standard Error Estimates (WMSEE) to control for heteroskedasticity issues. The T-statistics are presented in brackets.
*** indicates that the coefficient is significant at the 0.01 level. ** indicates that the coefficient is significant at the 0.05 level. * indicates that the coefficient is significant at the 0.1 level.
Dependent variable: RISK1
(1) (2) (3) CONSTANT 0.1950 ** 0.5718 *** 0.5645 *** (2.5743) (4.9732) (5.4192) FEMR -0.4596 -0.4552 ** -0.4222 ** (-1.5748) (-2.1611) (-2.2479) MAS -0.0013 -0.0019 ** -0.0019 ** (-1.0015) (-1.9771) (-2.3279) MAS*FEMR 0.0048 0.0065 * 0.0054 (0.9329) (1.8055) (1.6053) LNTA -0.0249 *** -0.0240 *** (-4.9126) (-5.0692) D/A 0.0025 -0.0522 (0.0693) (-1.4456) MTB 0.0182 * 0.0204 ** (1.9024) (2.1012) ROA -0.3617 ** -0.4004 ** (-1.9734) (-2.1862) GDPGR 1.7631 *** 1.5732 ** (3.3247) (2.2704)
Industry dummies NO NO YES
Country dummies NO NO NO9
WMSEE YES YES YES
Observations 408 380 380
R2 0.0207 0.2377 0.4596
Adjusted R2 0.0134 0.2213 0.3922
9
Table 6: Hypothesis III, cross section results
The table presents the regression results based on the winsorised data, whereby 3 different models are specified. Every model includes more or different variables. Model 3 represents the described empirical Model (3). 2013 is the year included in the cross sectional regression. All models have White's Modified Standard Error Estimates (WMSEE) to control for heteroskedasticity issues. The T-statistics are presented in brackets.
*** indicates that the coefficient is significant at the 0.01 level. ** indicates that the coefficient is significant at the 0.05 level. * indicates that the coefficient is significant at the 0.1 level. Dependent variable: RISK1
(1) (2) (3) CONSTANT 0.1440 *** 0.4448 *** 0.4287 *** (4.1939) (5.8503) (6.2937) FEMR -0.2261 -0.0780 -0.0840 (-1.3928) (-0.8707) (-0.7633) GRPD1 -0.0943 *** -0.0145 0.0067 (-2.6669) (-0.6129) (0.2163) GRPD2 0.0033 0.0386 0.0526 * (0.0733) (1.4221) (1.7214) GRPD1*FEMR 0.1993 0.0559 -0.0259 (1.1747) (0.4539) (-0.1604) GRPD2*FEMR -0.0509 -0.1162 -0.0998 (-0.2493) (-0.9228) (-0.6797) LNTA -0.0234 *** -0.0227 *** (-4.7597) (-4.7337) D/A -0.0013 -0.0585 * (-0.0361) (-1.6492) MTB 0.0180 * 0.0202 ** (1.7786) (1.9798) ROA -0.4046 ** -0.4417 ** (-2.0189) (-2.1973) GDPGR 1.4141 ** 1.2757 * (2.3886) (1.8067)
Industry dummies NO NO YES
Country dummies NO NO NO10
WMSEE YES YES YES
Observations 412 383 383
R2 0.0299 0.2254 0.4441
Adjusted R2 0.0180 0.2045 0.3718
Table 7: Hypothesis III, cross section results, dummy variable approach
The table presents the regression results based on the winsorised data, whereby 3 different models are specified. FEMRD is a dummy variable that measures the presence of female corporate board directors. FEMRD is one when female board directors are present in the firm's corporate board and zero otherwise. Model (1), (2), and (3) represent Hypothesis I, II, and III respectively. 2013 is the year included in the cross sectional regression. All models have White's Modified Standard Error Estimates (WMSEE) to control for heteroskedasticity issues. The T-statistics are presented in brackets.
*** indicates that the coefficient is significant at the 0.01 level. ** indicates that the coefficient is significant at the 0.05 level. * indicates that the coefficient is significant at the 0.1 level.
Dependent variable: RISK1
(1) (2) (3) CONSTANT -0.1895 0.5785 *** 0.4267 *** (-0.5789) (4.8850) (6.1383) FEMRD -0.0375 ** -0.1317 * -0.0387 (-2.2613) (-1.7677) (-1.4417) MAS -0.0025 * (-1.8998) MAS*FEMRD 0.0019 (1.5008) GRPD1 -0.0149 (-0.4231) GRPD2 0.0421 (1.0346) GRPD1*FEMR 0.0221 (0.5718) GRPD2*FEMR -0.0043 (0.0975) LNTA -0.0227 *** -0.0225 *** -0.0207 *** (-4.5302) (-4.6913) (-4.2099) D/A -0.0405 -0.0480 -0.0574 (-1.1752) (-1.3578) (-1.6391) MTB 0.0214 ** 0.0197 ** 0.0206 ** (2.0230) (2.1327) (2.0210) ROA -0.4213 ** -0.3946 ** -0.4447 ** (-2.0792) (-2.2384) (-2.2149) GDPGR 282.1385 * 1.3006 * 1.0084 (1.9244) (1.9413) (1.4830)
Industry dummies YES YES YES
Country dummies YES NO NO
WMSEE YES YES YES
Table 7 continued
R2 0.4707 0.4588 0.4445
Adjusted R2 0.3760 0.3914 0.3722
7 Robustness tests
7.1 Endogeneity
Endogeneity effects will be taken into account regarding all robustness tests because Adams & Ferreira (2009) conclude that not taking into account endogeneity will give rise to biased results. This study may also suffer from endogeneity as the dependent and independent variables may influence each other. It may be the case that more female board directors result in lower corporate risk taking but corporate risk taking itself may also influence the decision of hiring female board members. All robustness tests use a technique that is often used to control for endogeneity issues, namely the use of one year lagged independent and control variables (Adams & Ferreira, 2009; Hermalin & Weisbach, 2003).
7.2 Panel regression
significant (P<0.01) indicating that country, industry, and year dummies have to be included. However, year dummies are not included in the panel regression models because of near multicollinearity problems.
The panel regression outputs for Hypothesis I, II, and III can be found in table 15, 16, and 17 respectively in the Appendix. Differences in output regarding Hypothesis I can be found when comparing Model (5) of table 4 with Model (3)11 of table 15. The D/A coefficient is statistically significant in the panel regression output but not in the cross sectional output. Further, all coefficients that are statistically significant in the cross sectional regression output are also statistically significant in the panel regression. However, the MTB and ROA coefficient are less significant in the cross sectional regression. The GDPGR coefficient is positive in the cross sectional output and negative in the panel regression output. The results of the panel regression output do not alter the previously made conclusion regarding Hypothesis I.
Regarding Hypothesis II, when comparing Model (3) of table 16 with model (3) of table 5, the results show that the MAS coefficient is negative and statistically significant in both tables (P<0.05 and P<0.01 respectively). However, Model (3) of both tables shows a statistically insignificant interaction term, where the cross sectional model predicts a positive interaction coefficient and the panel regression a negative interaction coefficient. Although the coefficient of the interaction term in table 16 Model (3) of the panel regression is in line with Hypothesis II, the coefficient is statistically insignificant and therefore Hypothesis II is not supported. The results of the panel regression output in table 16 do not alter the previously made conclusion that no support is found for Hypothesis II.
The panel regression model constructed for Hypothesis III is different from the other panel regressions. Countries began to install gender quotas with sanctions from 2011 (see table 13 in the Appendix). The most recent year of available data was 2013. Therefore, it was only possible to include the years 2011, 2012, and 2013 in the panel regression regarding Hypothesis III. Comparing Model (3) in table 6 with Model (3) in table 17 it can be found that the FEMR coefficient is statistically significant (P<0.1) and negative in Model (3) of the panel regression, whereas the FEMR coefficient is negative and insignificant in the cross sectional regression. Further, the GRPD1 coefficient is negative and statistically significant in the panel regression (P<0.05) but positive and insignificant in the cross sectional regression. The GRPD2 variable is statistically significant (P<0.1) and positive in the cross sectional
regression but negative and insignificant in the panel regression. Further, the interaction terms have a negative coefficient in the cross sectional regression and a positive coefficient in the panel regression and are statistically insignificant in both models. Therefore, the results of the panel regression output do not alter the previously made conclusion regarding Hypothesis III.
7.3 Hypothesis III: Country classification
Abiad et al. (2008) are followed in structuring the country classification groups. However, Hypothesis III is tested using country dummies (see Section 6.3), which is not proposed by Abiad et al. (2008). Therefore, to test whether the results in Section 6.3 are robust, Abiad et al. (2008) are followed whereby a variable is constructed that can take the value 1, 2, or 3 which is in accordance with the country classification groups 1, 2, and 3 respectively. By doing so, only one country classification variable is included in the empirical model, namely the COUNCL variable. Table 18 in the Appendix presents the panel regression output including the country classification variable as proposed by Abiad et al. (2008). The regression output shows that the interaction term is negative and statistically insignificant. This indicates that firms positioned in country classification group 1 face higher corporate risk taking levels when the share of female board members increases compared with firms positioned in country classification group 3, which is in line with Hypothesis III. However, the interaction term is statistically insignificant and therefore Hypothesis III is not supported.
7.4 FEMR dummy
Table 19 in the Appendix shows the panel regression results of Hypothesis I, II, and III whereby the presence of female corporate board members is measured with a female representation dummy (FEMRD). Hypothesis III is tested using group dummies because the test in Section 7.3 did not presented significant differences between the test performed with group dummies and the test performed with the COUNCL variable. The findings in table 19 are not different in the sense that the conclusions in Section 7.2 regarding Hypothesis I and Hypothesis III have to be changed. However, the results for Hypothesis II show that the coefficient of the interaction term MAS*FEMRD is statistically significant (P<0.01) and therefore lends support in favor of Hypothesis II. A statistically significant result in favor of Hypothesis II is not found in previous tests and therefore this finding is striking.
7.5 Sample robustness
firms have a very high percentage of female corporate board directors but this is only based on one firm. The same counts for Bulgaria. It seems like Bulgarian firms do not have any female corporate board directors, but again this is only based on one firm. To test whether countries that have a low amount of firms included in the used sample affect the results, countries with less than ten firms are removed from the sample. This is the case for the following countries: Belgium, Bulgaria, Cyprus, Denmark, Finland, Hungary, Poland, Portugal, and Slovenia. With this adjusted sample the three hypotheses are tested again. The panel regression results are shown in table 20 in the Appendix. Model (1), (2), and (3) represent Hypothesis I, II, and III respectively. All models have one year lagged independent- and control variables to control for endogeneity effects. Model (1) does not alter our previous made conclusion that evidence is found in support for Hypothesis I. Model (2) is comparable to the previous panel regression results, and also Model (3) does not show striking differences with the previous panel regression results. Therefore, countries with a low level of firms included in the sample do not alter the panel regression results presented in Section 7.2 in such a way that previously made conclusions have to be changed.
In Section 7.4 is concluded that when the presence of female corporate board members is measured through a dummy in a panel regression, Hypothesis II is accepted. Model (4) in table 20 shows the results of Hypothesis II when the presence of female corporate board members is measured through a dummy variable (FEMRD) when using the adjusted sample. Also this model, as was the case in Section 7.4, supports Hypothesis II because the coefficient of the interaction term MAS*FEMR is statistically significant (P<0.05) and negative.
8 Discussion and conclusion
This paper studied a relevant and widely discussed topic, namely gender diversity in a firm's corporate board. Current literature argues inter alia that gender diverse boards gain access to other information than homogeneous boards, and that increasing the share of female board directors result in a better monitoring function (Hofstede 1998; Adams & Ferreira, 2009).
More specifically, this research investigated whether the inclusion of more female directors in a firm's corporate board leads to lower corporate risk taking. This relationship is expected as psychological, sociological, and biological evidence suggests that women are more risk averse than men. However, a caution has to be made because these studies draw their conclusions based on certain population groups but not specifically based on corporate board members (Adams & Funk, 2012). The results in Section 6 and 7 indicate that when the share of female corporate board directors in a firm located in a EU Member State increases, the firm's corporate risk taking will decline. This result is in line with the discussed theories in Section 3 of Schwartz & Rubel and Eagly et al. (2003) that behavior differences exist between men and women. Further, this paper's result is in line with psychological, sociological, and biological findings and therefore does not lend support that female board directors have different risk preferences compared with certain population groups as is suggested by Adams & Funk (2012).
between both genders. However, the results found are not strong which may be explained by the fact that a masculine society emphasizes achievements and competitiveness, focuses on money and assets, and encourages to be strong and successful, which may stimulate a person's risk taking behavior (Mihet, 2012). These characteristics of a masculine culture may therefore influence the risk taking behavior of men and women to such extent that the value differences between men and women in a masculine society do not result in large deviations in a gender's risk taking willingness.
Lastly, this paper took a whole other approach by not examining the behavioral differences between men and women but by researching whether a country's gender diversity stimulation measure has an impact on the relationship between gender diversity and corporate risk taking. By doing so, the EU context of this paper is further utilized. It is argued that rigid gender quotas stimulate the recruitment of female corporate board directors that do not possess the right qualifications for the function (Ahern & Dittmar, 2012). This may result in an increase in a firm's corporate risk taking in comparison with countries that installed voluntarily or no gender quotas because the latter two groups have more time to select appropriate female corporate board directors who know how to assess risky projects. However, no support is found for this relationship in Section 6 and 7. An explanation for this finding might be that the EU has a large enough talent pool to absorb the demand for qualified female talent because 60% of the EU university graduates are female (European Commission, 2015), as already mentioned in Section 1. Politics may use this finding in support of a gender quota because this finding may help them to refute arguments of opponents of gender quotas stressing that the female talent pool is not large enough to absorb the demand for qualified females when an obligatory gender quota is installed.
It can be concluded that an increasing share of female corporate board directors will lead to a decline in corporate risk taking. In addition, weak support is found that this relationship is strengthened when the share of female corporate board directors increases in a masculine country. Lastly, it can be concluded that the relationship between an increasing share of female corporate board directors and corporate risk taking is not weakened when the share of female corporate board directors increases in countries that installed rigid gender quotas with sanctions.
9 Relevance for practice
This study found a strong indication that an increasing share of female corporate board members will result in lower corporate risk taking. Practical implications for this finding may differ between firms that emphasize their shareholders and firms that emphasize their stakeholders.
Shareholder oriented firms have the objective to maximize their shareholder's wealth, meaning that high risk projects are preferred in order to maximize the possible returns to their shareholder, and because shareholders are able to diversify their risk exposure (Devers, McNamara, Wiseman & Arrfelt, 2008). Shareholder oriented firms may therefore become reluctant to include female board directors because of their risk averseness which results in lower returns. Shareholder oriented firms may engage in lobby activities to prevent a gender quota in their country.
A firm that has a stakeholder view does not only focus on its shareholders, but focuses on every group that is affected by the firm's attempts to realize their objectives (Fassin, 2008). Stakeholder oriented firms may stimulate the recruitment of female board directors because stakeholders as banks, governments, and labor unions prefer conservative firm investments (John et al., 2008). Firms with bank loans, for instance, may feel pressure to reduce their corporate risk taking in order to safeguard the bank's interests in the firm (John et al., 2008). So, stakeholder oriented firms may feel pressure from their stakeholders to recruit female corporate board directors and may engage in lobby activities to stimulate gender diversity measures.
