• No results found

The effect of Board Structure on the relationship between Gender Diversity and Corporate Risk-Taking

N/A
N/A
Protected

Academic year: 2021

Share "The effect of Board Structure on the relationship between Gender Diversity and Corporate Risk-Taking"

Copied!
48
0
0

Bezig met laden.... (Bekijk nu de volledige tekst)

Hele tekst

(1)

1

The effect of Board Structure on the relationship between Gender

Diversity and Corporate Risk-Taking

By Femke Buijs

s2162113 Submitted to

Supervisor: Prof. M. Ararat Co-Assessor: Dr. V. Purice

MSc. International Financial Management Faculty of Economic and Business

University of Groningen June 13th, 2016 A B S T R A C T

This paper examines the effect of board structure on the relationship between gender diversity and risk-taking. This is done based on a sample of 172 companies from the United Kingdom, Germany, the Netherlands, France and Switzerland, where companies have either a one-tier board structure, a two-tier board structure or are allowed to choose. A one-tier board structure, where a unitary board consisting of executives and non-executives collectively manages the company, and a two-tier board structure, where decision-management and decision-control are separated, differ substantially in their way of practice. It is therefore argued that, depending on the type of board structure, the relationship between gender diversity and risk-taking differs. After employing numerous robustness tests, I find significance evidence that board structure is an important moderator in the link between female representation and corporate risk-taking. However, although a combined sample shows significant findings, the results do not support the proposition that gender diversity is associated with a change in risk-taking in board structure specific samples. Nevertheless, as governments increasingly mandate corporate gender diversity policies, my study sheds additional light on the importance of considering individual firm characteristics by highlighting the effect of board structures.

(2)

2 1. INTRODUCTION

How an unprecedented crisis can lead to world fame is shown by the Icelandic business women Halla Tómasdóttir and Kristín Pétursdóttir, who founded Audur Capital; a financial service company known for being the only company to preserve the funds of its clients and stayed healthy after the Icelandic banks collapsed in October 2008 (Schiffers, 2015). What is surprising is that Audur Capital was run by eight women at that time.

One of the most important decision makers within a corporation is the board of directors, as they are responsible for overall firm decisions such as mergers and acquisitions, capital structures and the selection of top executives (Adams and Ferreira, 2009; Ferreira and Kirchmaier, 2013; Jensen and Meckling, 1976). Problems arising from ‘groupthink’ within the board of directors have become clear during the financial crisis of 2008 and have strengthened the constructive debate about having sufficient gender diversity in the boardroom. The main question in this debate started after recent corporate scandals such as in Lehman Brothers and Worldcom: would risk have been better managed if more women were running large companies around the world (Adams and Ragunathan, 2013)? This resulted in multiple countries introducing a Gender Balance Law (GBL), in which Norway was the first to mandate a gender quota of at least 40 percent (Bøhren and Staubo, 2016). But even though the number of women in leadership positions has increased in the past few years, still only 22.7 percent of the board members of large publicly listed companies in Europe are represented by women (European Commission, 2016).

(3)

3

explanation for the variation in findings is that the generally accepted risk-aversion among women is not applicable to women working in boards; by climbing the corporate ladder women adapt themselves to a male-dominated culture of risk-seeking board members (Adams and Funk, 2012).

Another explanation is that the influence women have in strategy implementation and the monitoring process is affected by the type of board structure. Two common types of board structures are a one-tier and a two-tier board structure. In a one-tier board structure, where a unitary board of directors comprised of executive and non-executive directors is collectively responsible for the long-term success of the company, independence is often a source of debate (Maassen and Van Den Bosch, 1999). For example, non-executive directors may not be fully able to control and supervise executive directors, while collectively managing the company (Sheridan and Kendall, 1992). A more independent approach is a two-tier board structure, where the supervising and management role are split by establishing two boards, a supervisory and a management board (Rouyer, 2013). However, supervisory boards may not always be as independent as they are supposed to be, and the large distance between non-executives and non-executives may hamper information flows, deterring the ability to monitor (Pirson and Turnbull, 2011).

(4)

4

The final results suggest that the relationship between gender diversity and corporate risk-taking varies depending on the type of board structure. Additionally, I find evidence that gender diversity, when controlling for board structures, is associated with a change in risk-taking when a combined sample is used. However, these results become inconclusive when the relationship is tested based on board structure specific subsamples. This remains the same when financial companies are added and controlled for. I further test whether critical mass theory provides additional insight, but no conclusive results are found. Nevertheless, the main finding that board structures directly affect the gender diversity-risk relationship shows that the presumable convergence in board structures (Block and Gerstner, 2016) is not visible in this research. This could be of great use for future corporate governance research, and provides useful guidance in current policy debates. Although governments have a crucial role in stimulating equal opportunities for all men and women, extensive regulation like gender quotas may not give the desired results when little consideration is given to individual firm characteristics like board structures.

The remainder of this paper is organised as follows. In the next section relevant theories and empirical evidence related to gender diversity, risk-taking and board structures are reviewed. Section 3 discusses the sample selection, together with the methodology used and the summary statistics. The results of the empirical analyses are presented in Section 4, while Section 5 explains the results in detail, before the implications are examined. Finally, Section 6 explains the limitations inherent in this study and discusses suggestions for future research.

2. LITERATURE REVIEW

2.1 Board composition and corporate governance

(5)

5

Thereafter, relevant theory based on psychological and economical research concerning gender-based differences is reviewed. Finally, empirical evidence and relevant facts with regard to board structures are analysed and the main hypothesis is formulated.

2.1.1 Agency Theory

Argued by agency theory, a separation between ownership and management can result in consequences for the alignment between the interest of the firm’s owners (principals) and the interest of managers (agents) (Jensen and Meckling, 1976), which brings forth agency costs (Fama, 1980; Fama and Jensen, 1983). In order to deal with this misalignment of interests, firms set up several corporate governance mechanisms to induce managers to take appropriate risks in line with the company’s risk policy, such as monitoring by shareholders or return-rewarding remuneration (Coles, Daniel and Naveen, 2006; Leland, 1998). One such corporate governance mechanism that is seen as a crucial instrument to impact risk-taking is the board of directors (Sila et al., 2016). A way in which boards can maintain the boundaries of risk-taking is through monitoring, as it can mitigate the incentive for managers to incorporate their personal preferences into the decision-making process, thereby protecting shareholder interest (Adams and Ferreira, 2009; Fama and Jensen, 1983; Jensen and Meckling, 1976).

A critical aspect for boards to function as effective monitors is board independence, and boards should be comprised of a sufficient number of independent outsiders as these outsiders have the incentive to strengthen their reputation as strong monitors and have less to lose by challenging managers than dependent insiders (Bøhren and Staubo, 2016; Carter, Simkins and Simpson, 2003). As women are generally more independent, providing greater oversight and are better at monitoring managers’ actions (Adams and Ferreira, 2009; Hillman et al., 2007), a higher proportion of female directors may develop a boards’ monitoring ability and can in turn affect risk-taking. However, Adams and Ferreira (2009) also show that a higher proportion of women is only beneficial if additional board monitoring is not counterproductive in well-governed firms. Therefore, gender diversity in boards would only be valuable if additional independence and monitoring would lead to an optimal risk level.

Nevertheless, due to their relative independence, board monitoring and in turn risk-taking is expected to be affected in firms with a higher proportion of female directors.

2.1.2 Resource Dependence Theory

(6)

6

(Hillman, Withers and Collins, 2009). The basis of resource dependence theory states that corporate directors bring resources to the firm, which include capital and expertise, diverse communication channels, creation of legitimacy in the external environment, and connections with vital stakeholders that can link the firm to its environment (Pfeffer and Salanic, 1978). The extension of the resource dependence theory by Hillman, Cannella and Paetzold (2000) includes that different types of directors bring different resources to the firm. Hence, a more gender diversified board will have access to unique information sets, strengthening decision-making and firm outcomes. For example, woman may attribute to helping to correct informational biases in problem solving and strategy formulation, by bringing in new perspectives on complex issues (Francoeur, Labelle, and Sinclair-Desgagne, 2008). Moreover, different and sometimes conflicting perspectives can generate a capacity for corporate board members to critically examine each other’s point of view (Ray, 2005). However, Gul, Srinidhi and Ng (2011) argue that this could also hinder board effectiveness by increasing dissension and conflict. For example, Adams and Ferreira (2004) show that boards tend to be more homogeneous in riskier, fast-paced and complex environments, as social homogeneity creates trust which is of crucial importance to maintain effective performance. On the other hand, Kravitz (2003) finds that gender diversity is beneficial if tasks are creative and complex, as it can restrict board effectiveness in simple tasks.

Besides providing resources, female representation in boards also creates an important link to female customers and the female labor force, as they can provide legitimacy in the view of the current and potential workforce which can create a greater human capital talent pool for firms (Carter, D’Souza, Simkins and Simpson, 2010; Hillman et al., 2007; Simpson, Carter and D’Souza, 2010). However, this may lead to firms appointing a female director simply because of stimulating the firm’s image and reputation (tokenism), omitting their potential contributions (Kesner, 1988). Furthermore, because of the contrast between tokens and the majority of the board, tokens are vulnerable for social and professional isolation (Kanter, 1977), hampering them to influence, for example, risk-taking.

2.1.3 Upper Echelons Theory

(7)

7

values reflecting past experience, knowledge, and personality. Cognitive frameworks of directors influence the way directors seek and interpret information, meaning they can shape decisions and the decision-making process (Post and Byron, 2015). Furthermore, several theorists have argued that complex decision-making is largely influenced by behavioural factors of the decision maker and, in general, the more complex the decisions to make, the more applicable this theory is (Cyert and March, 1963). Since cognitive frameworks of men and women differ due to different ethical reasoning, risk-aversion and values, this might explain the possible influence of gender diversity on complex strategic decisions. Moreover, according to Post and Byron (2015), gender differences on risk-aversion reflected in cognitive frameworks of corporate directors may lead to an increased incentive to monitor for female directors, as women tend to be more risk-averse than men, thereby avoiding legal, reputational and ethical risks of not doing so.

Based on agency theory and resource dependence theory, it can be seen that board composition and group dynamics matter, and that both affect the strategic decision-making process in several ways. According to agency theory, adding female directors increases board monitoring and resource dependence theory explains the qualities and perspectives women bring to the board. As dissimilar cognitive frameworks have an effect on how decision-making is accomplished according to the upper echelons theory, it is necessary to look if, and to what extent, cognitive frames of men and women differ.

2.2 Board gender diversity, board structure and hypothesis formulation 2.2.1 Gender-based differences

(8)

8

For example, Adams and Funk (2012) report that in Sweden female directors might even be more risk-seeking than men, because their risk-aversion has vanished once they have adapted to a male-dominated culture and have broken through the glass ceiling. It is reasonable to assume that female directors possess more ‘male’ characteristics that have helped them to climb the corporate ladder in the first place, making them different from the average female population. These diverse conclusions translate into the inconsistency in findings of the limited studies that investigated the relation between gender diversity and risk-taking. For example, Wilson and Altanlar (2011) find a negative relationship between female board representation and insolvency risk, Berger, Kick and Schaeck (2014) report that an increase in the proportion of female bank directors has a positive effect on portfolio risk, and Sila et al. (2016) find no evidence that the proportion of female directors has an effect on equity risk. The lack of conclusive evidence is not only visible in research on the gender diversity-risk relationship, also in the extensively researched relationship between female representation and firm performance no consensus has been reached so far. For example, some studies report that board diversity leads to better corporate performance, while others find no such relationship (e.g. Adams and Ferreira, 2009; Carter et al., 2003; Dwyer, Richard and Chadwick, 2003).

This calls for additional research on the relationship between corporate gender diversity and firm outcomes. Many researches state that the relationship is influenced by unobservable firm factors and reverse causality, as board characteristics are not exogenous variables but are endogenously chosen by firms to suit their internal and external environment (Adams and Ferreira, 2007). These two sources of endogeneity are seen as biasing estimates of analyses on how gender affects firm risk-taking. Unfortunately, measures used to control for these endogeneity issues have not resulted in any difference in the inconsistency of findings (Adams and Ferreira, 2009; Carter et al., 2010; Sila et al., 2016), and research outcomes remain to a high degree contextual. The aim of this study is to position the relationship in a different context by investigating whether country specific board structures can create reasoning behind the deviation in findings on gender diversity and firm outcomes.

2.2.2 Board structure and hypothesis formulation

(9)

9

success of a company (Rouyer, 2013). The executives are predominantly responsible for setting the company’s strategy and the associated risk appetite in achieving its strategic objectives, whereas the non-executive directors monitor this process and constructively challenge executives. In order to carry out their monitoring duty, non-executive director independence is of great importance (Block and Gerstner, 2016). Unfortunately, many researchers and reformers argue that unitary boards are not independent enough. For example, according to Sheridan and Kendall (1992), it is unsound practice if one group of directors controls and supervises another group of directors, while collectively managing a company.

A more independent approach to board structures is a two-tier board structure (Maassen and Van Den Bosch, 1999), as it separates management and control by establishing two separate boards (Rouyer, 2013), increasing the distance and in turn independence as there is no overlap of membership between the two boards (Fama and Jensen, 1983; Maassen and Van Den Bosch, 1999), which may lead to stronger and more effective monitoring by the supervisory board. Besides a higher degree of independence, the supervisory board in a two-tier board structure has also more power in the election process of management than non-executives in a unitary board, as the former is fully responsible in electing, reappointing and dismissing management, whereas in a unitary board the executives are elected by shareholders. Subsequently, the supervisory board possesses an extra indirect control over prospective risk management.

(10)

10

In addition, the degree of independence may itself be challenged. Besides monitoring, the supervisory board also advices management and is involved in decisions that are regarded as fundamental to the company. Consequently, a clear division between decision-making and decision-control may be diminished, resulting in a questionable independence of the supervisory board in its monitoring role.

In conclusion, a priori the influence of board structure on risk-taking remains ambiguous; it is therefore unclear what effect board structures could have on the gender diversity-risk relationship. However, as both board structures differ greatly in their way of practice, size, composition, norms and duties (Block and Gerstner, 2016), it is highly likely they affect the way and to what extent female directors can influence risk-taking. Therefore, based on the aforementioned theories and empirical evidence, the following hypothesis is formulated: Hypothesis 1: The type of board structure has an effect on the relationship between gender diversity and risk-taking

3. METHODOLOGY

3.1 Data and variable definitions 3.1.1 Sample selection

For this research, a data sample of German, English, Dutch, French and Swiss firms is used which is derived at the 2nd and 23rd of May, covering a period of 2002 to 2014. These firms are either listed on the Financial Times Stock Exchange 100 (FTSE 100), the Deutscher Aktienindex (DAX), the Amsterdam Exchange (AEX), Amsterdam Midcap Index (AMX), the Cotation Assistée en Continu (CAC 40) or the Swiss Market Index (SMI). As the FTSE 100 companies have to comply with the one-tier board structure following the English corporate governance code, and companies that are listed on the DAX have to comply with the German corporate governance code that stipulates a two-tier board structure, this sample provides a great opportunity to compare both board structures. Although Dutch companies can nowadays choose between both board structures, all sample firms have a two-tier board structure; the Netherlands will thus be regarded as a two-tier country. Furthermore, stock exchanges in France and Switzerland are added, where companies can decide on which board structure to obtain. In this way, the effects of other country specific characteristics are to a great extent reduced.

(11)

11

have been operating for many years. The DAX is a stock index that trades on the Frankfurt Exchange and represents 30 of the largest and most liquid German companies. The AEX and AMX, who are respectively the blue-chip and mid-cap indexes, represent the 50 largest stocks in the Netherlands. The CAC 40 depicts a capitalisation weighted measure of the 40 most important companies in France. Finally, the SMI is the blue-chip stock market index in Switzerland, and represents the 20 largest companies in Switzerland in terms of market capitalisation. Financial data and data on board characteristics are retrieved from the Thomson Reuters’ Asset 4 Database (ESG) and the Thomson Reuters’ Worldscope Database. Data concerning board structure is gathered manually if it was not available in the Thomson Reuters’ Database. Observations with less than one year of data on gender diversity are removed, which included 620 firm observations. Furthermore, I remove 47 financial firms, as these firms are subjected to specific regulations and have to comply with different accounting standards (Boubakri, Mansi and Saffar, 2013; Jackling and Johl, 2009). Accordingly, firms with SIC codes between 6000 and 6999 are excluded from the sample (Gul et al., 2011). Finally, 4 companies are taken out of the sample as two (Unilever and Royal Dutch Shell) are listed on both the AEX and the FTSE 100, and two companies have a mixed board structure. A balanced sample is not required. Consequently, the final sample consists of 1917 unique firm observations from 172 companies, of which 105 have a one-tier board structure and 67 have a two-tier board structure.

All variables used in this research and their descriptions can be found in Table A.1 in the Appendix. The abbreviations of the control variables described in Table A.1 will be stated within brackets in the variable descriptions below.

3.1.2 Dependent variable

(12)

12 3.1.3 Independent variable

In this research, the independent variable is gender diversity in unitary boards and supervisory boards. Based on previous literature, three different measures are used. First of all, many studies have used the proportion of women (%_Women) on the board of directors as a measure of gender diversity (Adams and Ferreira, 2009; Liu, Wei and Xie, 2014; Sila et al., 2016; Francoeur et al., 2008). Accordingly, I use this as the main measure of gender diversity, which is the ratio of female directors with respect to the board of directors. Unfortunately, the proportion of women may not always be an appropriate metric to measure gender diversity, as variations can stem from two sources: a change in the number of women and an alteration in the overall board size. Therefore, alternative measures are used in this research, namely the number of women (Nr_Women) and a dummy variable (Women_Dummy) that equals 1 if a board has at least one female director and 0 otherwise (Bøhren and Staubo, 2016; Gul et al., 2011; Liu et al., 2014). The latter can also be used to examine critical mass theory, which stipulates that a critical mass needs to be reached before the influence of women in the decision-making process materialises (Kanter, 1977).

3.1.4 Control variables

(13)

13

board meetings (Board_Meetings) during each year. Finally, a dummy variable that controls for board structures (BoardStructure_Dummy) is included in combined sample analyses. This dummy equals 0 for one-tier boards and 1 for two-tier boards, and may also serve as a corporate governance proxy for countries (Ferreira and Kirchmaier, 2013).

(14)

14 3.2 Regression model

To test whether the relationship between female representation and risk-taking is affected by board structure, a pooled ordinary least squares (OLS) regression model is used. A common concern that arises with a pooled OLS regression model is the potential for an endogeneity problem as pointed out by Hermalin and Weisbach (1988).

3.2.1 Endogeneity

Establishing a causal relationship between gender diversity and risk-taking in one-tier and two-tier board structures is challenging. There is a general consensus in literature that board characteristics are not exogenous random variables, but endogenously chosen to fit a firms’ characteristics and its information and operating environment (Adams and Ferreira, 2007; Fama and Jensen, 1983). Endogeneity may lead to inconsistent and biased estimates that would make a reliable interpretation of the outcomes virtually impossible (Wintoki, Linck and Netter, 2012). To accurately test whether the gender diversity-firm risk-taking relationship is affected by board structure, at least two alternative sources of endogeneity have to be considered which are particularly likely to bias my estimates: omitted unobservable firm factors and reverse causality.

Omitted unobserved factors

(15)

15

Under each of these explanations, a significant relationship between female board representation and firm risk-taking could be observed while no such relationship may exist. To address this endogeneity issue to a certain extent, a panel regression with either fixed effects or random effects could be used (Adams & Ferreira, 2009; Faccio et al., 2015; Liu et al. 2014). Fixed effects estimations potentially mitigate the bias arising from omitted unobserved firm characteristics (Wintoki et al., 2012), and can be split into firm fixed effects (𝛼𝑖) and time fixed effects (λ𝑡). The former is included to help eliminate constant omitted variables bias and the latter controls for ‘economy-wide yearly fluctuations’ (Liu et al., 2014: 8).

An alternative to the fixed effects model is the random effects model. Identical to the fixed effects approach, the random effects approach assigns an intercept to each entity that is constant over time, and assumes the relationships between the dependent and independent variables to be equal both cross-sectionally and temporarily. The main difference arises from the added random variable assigned to each intercept which varies cross-sectionally but is constant over time (Brooks, 2008). The major drawback of the random effects model is its lack of validity when individual unobserved variables are correlated with the explanatory variables, which is often the case. Nevertheless, I perform the Hausman test to test whether the fixed effects approach or the random effects approach is preferable. The null hypothesis is that the unobserved omitted firm specific effects are uncorrelated with the explanatory variables, which would make the random effects model appropriate. The alternative hypothesis is that the random effects model is biased and thus an inappropriate estimator. Using the Hausman test, the returned p-value is 0.000, which is smaller than the smallest significance level of 1 percent. The null hypothesis is therefore rejected; hence time and cross-sectional fixed effects (FE) will be used to address the first endogeneity problem. Unfortunately, using a fixed effects estimator is insufficient to deal with the second source of endogeneity, reverse causality, as it assumes that the independent variable (female board representation) is completely independent of past values of the measured variable (risk-taking) (Wintoki et al., 2012).

Reverse causality

(16)

16

reported risk-aversion (Bernasek and Shwiff, 2001; Farrel and Hersch, 2005; Sunden and Surette, 1998). This may result in a strong, but biased negative relationship between female board directors and firm risk-taking. To address this endogeneity problem, I use a one-year lagged board gender diversity variable and one-year lagged board independence and board size variable, as board characteristics need time to influence risk-taking (Liu et al., 2014).

3.2.2 FE OLS regression model

As stated above, I use a pooled OLS regression including both cross-sectional and time fixed effects, and one-year lagged variables for gender diversity, board independence and board size. To test the significance of board structures, the interaction term 𝐵𝑜𝑎𝑟𝑑𝑆𝑡𝑟𝑢𝑐𝑡𝑢𝑟𝑒_𝐷𝑢𝑚𝑚𝑦𝑖,𝑡 ∗ 𝐺𝑒𝑛𝑑𝑒𝑟_𝐷𝑖𝑣𝑒𝑟𝑠𝑖𝑡𝑦𝑖,𝑡−1 is included in the regression (Brambor, Clark and Golder, 2005). Furthermore, according to Brambor et al., (2005), the constitutive terms of the interaction model should be present in the model. The regression model including all countries combined is therefore formulated as follows:

𝐴𝑙𝑙_𝑅𝑖𝑠𝑘𝑖,𝑡 = α0+ 𝛽1𝐺𝑒𝑛𝑑𝑒𝑟_𝐷𝑖𝑣𝑒𝑟𝑠𝑖𝑡𝑦𝑖,𝑡−1+ 𝛽2𝐵𝑜𝑎𝑟𝑑𝑆𝑡𝑟𝑢𝑐𝑡𝑢𝑟𝑒_𝐷𝑢𝑚𝑚𝑦𝑖,𝑡+ 𝛽3𝐵𝑜𝑎𝑟𝑑𝑆𝑡𝑟𝑢𝑐𝑡𝑢𝑟𝑒_𝐷𝑢𝑚𝑚𝑦𝑖,𝑡∗ 𝐺𝑒𝑛𝑑𝑒𝑟_𝐷𝑖𝑣𝑒𝑟𝑠𝑖𝑡𝑦𝑖,𝑡−1+ 𝛽4%_𝐼𝑛𝑑𝑒𝑝𝑒𝑛𝑑𝑒𝑛𝑐𝑒𝑖,𝑡−1+ 𝛽5𝐵𝑜𝑎𝑟𝑑_𝑇𝑒𝑛𝑢𝑟𝑒𝑖,𝑡 + 𝛽6𝐵𝑜𝑎𝑟𝑑_𝑆𝑖𝑧𝑒𝑖,𝑡−1+ 𝛽7𝐵𝑜𝑎𝑟𝑑_𝑀𝑒𝑒𝑡𝑖𝑛𝑔𝑠𝑖,𝑡+ 𝛽8𝑀𝑇𝐵𝑖,𝑡+

𝛽9𝐶𝐴𝑃𝐸𝑋𝑖,𝑡+ 𝛽10𝐺𝑟𝑜𝑤𝑡ℎ𝑖,𝑡+ 𝛽11𝐹𝑖𝑟𝑚_𝑆𝑖𝑧𝑒𝑖,𝑡+ 𝛽12𝜎𝑅𝑂𝐴𝑖,𝑡+λ𝑡+ ε𝑖,𝑡 (1)

(17)

17

fixed effects term and ε𝑖,𝑡 is the error term. Cross-sectional fixed effects (α𝑖) are excluded from this regression as the board structure dummy creates a near singular matrix which counteracts the firm fixed effects model.

Besides a FE OLS regression including all countries in the sample, two separate regressions are performed using only companies with a one-tier board structure and companies with a two-tier board structure, to make a comparison between the effects of both board structures available. The regression model including only companies with a one-tier board structures is as follows:

𝑂𝑛𝑒_𝑅𝑖𝑠𝑘𝑖,𝑡 =

α0 + 𝛽1𝐺𝑒𝑛𝑑𝑒𝑟_𝐷𝑖𝑣𝑒𝑟𝑠𝑖𝑡𝑦𝑖,𝑡−1+ 𝛽2%_𝐼𝑛𝑑𝑒𝑝𝑒𝑛𝑑𝑒𝑛𝑐𝑒𝑖,𝑡−1+ 𝛽3𝐵𝑜𝑎𝑟𝑑_𝑇𝑒𝑛𝑢𝑟𝑒𝑖,𝑡+ 𝛽4𝐵𝑜𝑎𝑟𝑑_𝑆𝑖𝑧𝑒𝑖,𝑡−1+ 𝛽5𝐵𝑜𝑎𝑟𝑑_𝑀𝑒𝑒𝑡𝑖𝑛𝑔𝑠𝑖,𝑡+ 𝛽6𝑀𝑇𝐵𝑖,𝑡+ 𝛽7𝐶𝐴𝑃𝐸𝑋𝑖,𝑡+ 𝛽8𝐺𝑟𝑜𝑤𝑡ℎ𝑖,𝑡+ 𝛽9𝐹𝑖𝑟𝑚_𝑆𝑖𝑧𝑒𝑖,𝑡+ 𝛽10𝜎𝑅𝑂𝐴𝑖,𝑡 +λ𝑡+ α𝑖 + ε𝑖,𝑡 (2) The regression model including only companies with a two-tier board structures is as follows: 𝑇𝑤𝑜_𝑅𝑖𝑠𝑘𝑖,𝑡 = α0 + 𝛽1𝐺𝑒𝑛𝑑𝑒𝑟_𝐷𝑖𝑣𝑒𝑟𝑠𝑖𝑡𝑦𝑖,𝑡−1 + 𝛽2%_𝐼𝑛𝑑𝑒𝑝𝑒𝑛𝑑𝑒𝑛𝑐𝑒𝑖,𝑡−1+ 𝛽3𝐵𝑜𝑎𝑟𝑑_𝑇𝑒𝑛𝑢𝑟𝑒𝑖,𝑡+ 𝛽4𝐵𝑜𝑎𝑟𝑑_𝑆𝑖𝑧𝑒𝑖,𝑡−1+ 𝛽5𝐵𝑜𝑎𝑟𝑑_𝑀𝑒𝑒𝑡𝑖𝑛𝑔𝑠𝑖,𝑡+ 𝛽6𝑀𝑇𝐵𝑖,𝑡+ 𝛽7𝐶𝐴𝑃𝐸𝑋𝑖,𝑡+ 𝛽8𝐺𝑟𝑜𝑤𝑡ℎ𝑖,𝑡+ 𝛽9𝐹𝑖𝑟𝑚_𝑆𝑖𝑧𝑒𝑖,𝑡+ 𝛽10𝜎𝑅𝑂𝐴𝑖,𝑡𝑡+ α𝑖 + ε𝑖,𝑡 (3) 3.3 Descriptive statistics 3.3.1 Descriptive statistics

Table 1 represents sample summary statistics based on 172 firms listed on the FTSE 100, DAX, AEX, AMX, CAC 40 and SMI within the period 2002-2014. All data variables, except for the MTB ratio, are winsorised at the 0.01 and 0.99 level to curtail the effect extreme outliers might have on outcomes. The MTB ratio is winsorised at the 0.05 and 0.95 level due to its high extreme values.

(18)

18

average proportion of women on boards is 12.3 percent, in which the maximum (41.2 percent) is still below half of the board members. The average number of females is 1.52 and the mean dummy is 78.9 percent, meaning that 78.9 percent of the firms in the sample selection have at least one woman in their board of directors. Figure A.1 in the Appendix shows the trend of women directors by year and by board structure. Figure A.1A illustrates that the percentage of female directors gradually rises from about 6 percent in 2002 to about 23 percent for one-tier boards and 18 percent for two-tier boards. A similar trend is visible in Figure A.1B, which exhibits a gradual growth in the percentage of firms that have at least one female director. Surprisingly, the proportion of women on the board of directors and the percentage of firms that have at least one female board member is higher for one-tier boards than for two-tier boars, further showing the difference between board structures. An equal observation can be derived from Table A.2 in the Appendix, which shows are higher mean for one-tier boards with regard to both the proportion of women and the amount of women in boards. Only the dummy variable shows a higher mean for two-tier boards, where on average 80.3 percent of the boards appointed at least one female director.

(19)

19 Table 1

Summary statistics

This table presents the descriptive statistics of all board and firm variables. The data sample consists of 1917 firm observations of 172 companies over a period of 2002-2014. The data is retrieved from the Thomson Reuters’ Asset 4 Database (ESG) and the Thomson Reuters’ Worldscope Database. Variable definitions are presented in Table A.1. All data variables, except for the MTB ratio, are winsorised at the 0.01 and 0.99 level. The MTB ratio is winsorised at the 0.05 and 0.95 level.

Variable Mean Min Median Max Standard

deviation N Leverage 0.407 0.000 0.393 1.000 0.215 1896 %_Women 0.123 0.000 0.100 0.412 0.101 1917 Nr_Women 1.524 0.000 1.000 7.000 1.447 1902 Women_Dummy 0.789 0.000 1.000 1.000 0.408 1917 %_Independent 0.590 0.000 0.583 1.000 0.235 1467 Board_Tenure 6.024 0.000 5.770 14.564 2.623 1760 Board_Size_log 1.049 0.699 1.041 1.342 0.149 1905 Board_Meetings 7.860 4.000 8.000 17.000 2.632 1833 MTB 1.708 0.940 1.471 3.605 0.735 1896 CAPEX 0.046 0.000 0.039 0.177 0.034 1908 Growth_log 0.004 -0.024 0.002 0.057 0.010 1816 Firm_Size_log 7.090 5.706 7.096 8.274 0.588 1908 σROA 0.035 0.004 0.025 0.213 0.034 1434

3.3.2 Correlation among variables

(20)

20 Table 2

Correlation matrix

This table presents the correlation matrix among the independent and dependent variables used in this research. The data sample consists of 1917 firm observations of 172 companies over a period of 2002-2014. The data is retrieved from the Thomson Reuters’ Asset 4 Database (ESG) and the Thomson Reuters’ Worldscope Database. Variable definitions are presented in Table A.1. All data variables, except for the MTB ratio, are winsorised at the 0.01 and 0.99 level. The MTB ratio is winsorised at the 0.05 and 0.95 level.

(21)

21 4. EMPIRICAL RESULTS

4.1 Pooled and FE OLS regression with lagged board characteristics

I first analyse the effect of gender diversity on risk-taking for all companies combined, I then include the board structure dummy and the interaction term to investigate whether the relationship between gender diversity and risk-taking differs with the type of board structure. Table 3 presents the results of the OLS regression (1) with and without time and cross-sectional fixed effects, three lags of gender diversity measures and two lags for other board characteristics, namely board independence (%_Independent) and board size (Board_Size). Columns (1) and (4) present results with the percentage of female directors, columns (2) and (5) with the number of female directors, and columns (3) and (6) with a women dummy as the independent variable defined in Table A.1.

The overall results suggest that no significant relationship exists between gender diversity and risk-taking. That is, by increasing the percentage or number of women in corporate boards, or by appointing at least one woman, would not lead to a deviation in risk-taking. With regard to other board characteristics, Table 3 shows that board tenure is significant and has a negative relationship with firm risk-taking for all proxies, which is consistent with the idea that a longer tenure entrenches board members and lead to lower risk-taking (Berger et al., 1997). Furthermore, growth and investment opportunities seem to explain a great deal in the relationship between gender diversity and risk. The MTB ratio is negative and significant for all proxies. In addition, CAPEX has a positive impact on risk-taking in the pooled OLS model, consistent with the idea that stronger investment and growth opportunities result in greater risk-taking (Guay, 1999). However, the significant relationship disappears when fixed effects are added, suggesting that the positive relationship comes from unobservable firm heterogeneity. The opposite happens for the growth coefficient, which becomes significantly positive after adding fixed effects.

(22)

22 Table 3

Pooled and FE OLS with lagged board characteristics of all companies combined: effect of gender diversity on leverage.

This table presents the results of regression (1) without board specific variables. Columns (1)-(3) exclude fixed effects, columns (4)-(6) include fixed effects. Leverage is calculated as the financial debt divided by the sum of equity and financial debt. %_Women equals the proportion of women in the board of directors. Nr_Women is the amount of women in the board of directors. Women_Dummy equals 1 if at least one woman is present in the board of directors and 0 otherwise. %_Independent is the percentage of independent board members. Board_Tenure is the average number of years each board member has been on the board of directors. Board_Size_log is the natural logarithm of the number of board directors. Board_Meetings is the number of board meetings each year. MTB is calculated as the market value of total assets divided by the book value of total assets. CAPEX is the ratio of capital expenditures with respect to total assets. Growth_log is the logarithm of the yearly growth in assets. Firm_Size_log is the logarithm of total assets. σROA is calculated as the volatility over a period of five years of the ratio of EBIT to total annual assets. The data is retrieved from the Thomson Reuters’ Asset 4 Database (ESG) and the Thomson Reuters’ Worldscope Database. Corresponding standard errors are shown in parentheses. ***, ** and * denote statistical significance at the 1%, 5% and 10% level respectively. Leverage (1) (2) (3) (4) (5) (6) Constant 0.315*** (0.098) 0.33*** (0.100) 0.305*** (0.097) 0.323 (0.238) 0.323 (0.238) 0.343 (0.237) %_Women𝑡−1 0.095 (0.070) -0.011 (0.056) Nr_Women 𝑡−1 0.006 (0.006) 0.001 (0.004) Women_Dummy𝑡−1 0.019 (0.020) -0.020 (0.013) %_Independent 𝑡−1 -0.005 (0.031) -0.004 (0.031) -0.009 (0.032) -0.027 (0.029) -0.028 (0.029) -0.029 (0.028) Board_Tenure -0.011*** (0.003) -0.012*** (0.003) -0.011*** (0.003) 0.000 (0.002) 0.000 (0.002) 0.000 (0.002) Board_Size_log𝑡−1 0.030 (0.066) 0.018 (0.070) 0.034 (0.065) -0.064 (0.059) -0.068 (0.060) -0.065 (0.058) Board_Meetings 0.005* (0.003) 0.005** (0.003) 0.005* (0.003) -0.000 (0.001) -0.000 (0.002) -0.000 (0.002) MTB -0.017* (0.009) -0.016* (0.009) -0.016* (0.009) -0.053*** (0.009) -0.053*** (0.009 -0.051*** 0.009 CAPEX 0.366** (0.175) 0.361** (0.175) 0.356** (0.173) 0.071 (0.162) 0.068 (0.162) 0.073 (0.161) Growth_log -0.763 (0.653) -0.813 (0.651) -0.796 (0.652) 0.774** (0.318) 0.771** (0.318) 0.775** (0.317) Firm_Size_log 0.017 (0.015) 0.017 (0.015) 0.018 (0.015) 0.030 (0.033) 0.031 (0.033) 0.030 (0.033) σROA -0.828*** (0.178) -0.833*** (0.178) -0.836*** (0.177) 0.875*** (0.137) 0.873*** (0.137) 0.877*** (0.136) R-squared 0.057 0.056 0.056 0.864 0.864 0.864 Adjusted R-squared 0.049 0.048 0.047 0.839 0.839 0.840 N 1134 1134 1134 1134 1134 1134

Firm fixed effects Time fixed effects

(23)

23

It is interestingly to observe that all measures of gender diversity become positively and significantly related to leverage. This indicates that by raising the percentage of women, the number of women or by appointing a women in the board of directors, an increase in the proportion of assets that is financed with debt would occur, indicating a tendency for enhanced risk appetite in the financial decision-making process (Friend and Lang, 1988; Leland, 1998). This corresponds with the findings of Adams and Funk (2012), who report that female directors might even be more seeking than men due to their vanished risk-aversion by adapting themselves to the male-dominated culture, and contradicts the conclusions of Faccio et al. (2015), who document that by appointing a female CEO a firm reduces corporate risk-taking, and firms that are run by female CEOs have less volatile earnings and lower leverage ratios. Finally, the standard deviation of the return on assets becomes negatively related to risk-taking at the 1 percent significance level for all proxies.

(24)

24 Table 4

Pooled and FE OLS with lagged board characteristics of all companies combined including board structure variable and the interaction term: effect of gender diversity on leverage.

This table presents the results of regression (1). Columns (1)-(3) exclude fixed effects, columns (4)-(6) include time fixed effects. Leverage is calculated as the financial debt divided by the sum of equity and financial debt. %_Women equals the proportion of women in the board of directors. Nr_Women is the amount of women in the board of directors. Women_Dummy equals 1 if at least one woman is present in the board of directors and 0 otherwise. BoardStructure_Dummy equals 0 in case of a one-tier board structure and 1 in case of a two-tier board structure. %_Independent is the percentage of independent board members. Board_Tenure is the average number of years each board member has been on the board of directors. Board_Size_log is the natural logarithm of the number of board directors. Board_Meetings is the number of board meetings each year. MTB is calculated as the market value of total assets divided by the book value of total assets. CAPEX is the ratio of capital expenditures with respect to total assets. Growth_log is the logarithm of the yearly growth in assets. Firm_Size_log is the logarithm of total assets. σROA is calculated as the volatility over a period of five years of the ratio of EBIT to total annual assets. The data is retrieved from the Thomson Reuters’ Asset 4 Database (ESG) and the Thomson Reuters’ Worldscope Database. Corresponding standard errors are shown in parentheses. ***, ** and * denote statistical significance at the 1%, 5% and 10% level respectively.

Leverage (1) (2) (3) (4) (5) (6) Constant 0.287*** (0.099) 0.308*** (0.101) 0.277*** (0.098) 0.265*** (0.099) 0.303*** (0.101) 0258*** (0.098) %_Women𝑡−1 0.177** (0.080) 0.247*** (0.086) Nr_Women 𝑡−1 0.015** (0.007) 0.021*** (0.007) Women_Dummy𝑡−1 0.047** (0.023) 0.056** (0.024) BoardStructure_Dummy 0.047* (0.025) 0.053** (0.022) 0.107** (0.043) 0.050** (0.024) 0.054** (0.022) 0.108** (0.043) BoardStructure_Dummy* Gender_Diversity𝑡−1 -0.315* (0.163) -0.032*** (0.011) -0.114** (0.046) -0.328** (0.163) -0.033*** (0.011) -0.118** (0.046) %_Independent 𝑡−1 -0.003 (0.031) -0.007 (0.031) 0.013 (0.033) -0.004 (0.031) -0.009 (0.031) 0.011 (0.033) Board_Tenure -0.011*** (0.003) -0.012*** (0.003) -0.010*** (0.003) -0.010*** (0.003) -0.011*** (0.003) -0.010*** (0.003) Board_Size_log𝑡−1 0.074 (0.071) 0.080 (0.075) 0.049 (0.068) 0.049 (0.071) 0.040 (0.077) 0.029 (0.069) Board_Meetings 0.006** (0.003) 0.006** (0.003) 0.006** (0.003) 0.006** (0.003) 0.006** (0.003) 0.006** (0.003) MTB -0.018* (0.009) -0.017* (0.009) -0.017* (0.009) -0.014 (0.010) -0.013 (0.010) -0.013 (0.010) CAPEX 0.388** (0.176) 0.372** (0.175) 0.382** (0.174) 0.356** (0.176) 0.344* (0.176) 0.350** (0.175) Growth_log -0.696 (0.653) -0.725 (0.650) -0.755 (0.651) -1.042 (0.672) -1.100 (0.670) -1.105 (0.671) Firm_Size_log 0.012 (0.015) 0.009 (0.015) 0.012 (0.015) 0.016 (0.015) 0.013 (0.015) 0.017 (0.015) σROA -0.814*** (0.178) -0.813*** (0.178) -0.796*** (0.178) -0.821*** (0.179) -0.819*** (0.17) -0.806*** (0.179) R-squared 0.060 0.063 0.061 0.070 0.072 0.069 Adjusted R-squared 0.050 0.053 0.051 0.053 0.055 0.052 N 1134 1134 1134 1134 1134 1134

Firm fixed effects Time fixed effects

(25)

25 4.2 Board structure specific sub-sample analyses

Contrary to the results of the FE OLS regression (1) in Table 4, the results for some of the three proxies of gender diversity become insignificant for both board structures in regressions (2) and (3) in Table 5 and 6 respectively, which could be due to a sample size issue. For the regression model regarding one-tier companies the question of unobserved sources of firm heterogeneity rises, as the results become insignificant after fixed effects are added. Overall, there is only weak evidence that women have a significant impact on risk in both board structures, as only the number of women in supervisory boards remains significant after controlling for fixed effects.

(26)

26 Table 5

Pooled and FE OLS with lagged board characteristics of all companies with a one-tier board structure: effect of gender diversity on leverage.

This table presents the results of regression (2). Columns (1)-(3) exclude fixed effects, columns (4)-(6) include fixed effects. Leverage is calculated as the financial debt divided by the sum of equity and financial debt. %_Women equals the proportion of women in the board of directors. Nr_Women is the amount of women in the board of directors. Women_Dummy equals 1 if at least one woman is present in the board of directors and 0 otherwise. %_Independent is the percentage of independent board members. Board_Tenure is the average number of years each board member has been on the board of directors. Board_Size_log is the natural logarithm of the number of board directors. Board_Meetings is the number of board meetings each year. MTB is calculated as the market value of total assets divided by the book value of total assets. CAPEX is the ratio of capital expenditures with respect to total assets. Growth_log is the logarithm of the yearly growth in assets. Firm_Size_log is the logarithm of total assets. σROA is calculated as the volatility over a period of five years of the ratio of EBIT to total annual assets. The data is retrieved from the Thomson Reuters’ Asset 4 Database (ESG) and the Thomson Reuters’ Worldscope Database. Corresponding standard errors are shown in parentheses. ***, ** and * denote statistical significance at the 1%, 5% and 10% level respectively.

Leverage (1) (2) (3) (4) (5) (6) Constant 0.281** (0.115) 0.334*** (0.119) 0.284** (0.116) 0.219 (0.293) 0.216 (0.293) 0.228 (0.293) %_Women𝑡−1 0.181** (0.083) 0.018 (0.067) Nr_Women𝑡−1 0.015** (0.007) 0.005 (0.005) Women_Dummy𝑡−1 0.039 (0.024) -0.016 (0.014) %_Independent𝑡−1 -0.059 (0.051) -0.057 (0.051) -0.055 (0.050) -0.023 (0.042) -0.024 (0.042) -0.026 (0.042) Board_Tenure -0.017*** (0.003) -0.018*** (0.003) -0.017*** (0.003) 0.003 (0.003) 0.003 (0.003) 0.003 (0.003) Board_Size_log 𝑡−1 0.310*** (0.098) 0.259** (0.100) 0.288*** (0.099) -0.015 (0.071) -0.035 (0.074) -0.014 (0.071) Board_Meetings 0.003 (0.003) 0.003 (0.003) 0.003 (0.003) 0.004* (0.002) 0.003* (0.002) 0.003* (0.002) MTB -0.005 (0.011) -0.004 (0.011) -0.004 (0.011) -0.041*** (0.010) -0.040*** (0.010) -0.041*** (0.010) CAPEX 0.224 (0.219) 0.220 (0.219) 0.206 (0.219) 0.220 (0.209) 0.220 (0.208) 0.203 (0.209) Growth_log -0.754 (0.796) -0.806 (0.793) -0.815 (0.796) 0.763** (0.378) 0.752** (0.378) 0.765** (0.378) Firm_Size_log -0.013 (0.018) -0.012 (0.018) -0.012 (0.018) 0.028 (0.041) 0.030 (0.041) 0.029 (0.041) σROA -0.599*** (0.195) -0.596*** (0.195) -0.599*** (0.195) 0.666*** (0.145) 0.664*** (0.145) 0.672*** (0.145) R-squared 0.062 0.061 0.059 0.867 0.867 0.867 Adjusted R-squared 0.050 0.050 0.048 0.844 0.844 0.844 N 822 822 822 822 822 822

Firm fixed effects Time fixed effects

(27)

27 Table 6

Pooled and FE OLS with lagged board characteristics of all companies with a two-tier board structure: effect of gender diversity on leverage.

This table presents the results of regression (3). Columns (1)-(3) exclude fixed effects, columns (4)-(6) include fixed effects. Leverage is calculated as the financial debt divided by the sum of equity and financial debt. %_Women equals the proportion of women in the board of directors. Nr_Women is the amount of women in the board of directors. Women_Dummy equals 1 if at least one woman is present in the board of directors and 0 otherwise. %_Independent is the percentage of independent board members. Board_Tenure is the average number of years each board member has been on the board of directors. Board_Size_log is the natural logarithm of the number of board directors. Board_Meetings is the number of board meetings each year. MTB is calculated as the market value of total assets divided by the book value of total assets. CAPEX is the ratio of capital expenditures with respect to total assets. Growth_log is the logarithm of the yearly growth in assets. Firm_Size_log is the logarithm of total assets. σROA is calculated as the volatility over a period of five years of the ratio of EBIT to total annual assets. The data is retrieved from the Thomson Reuters’ Asset 4 Database (ESG) and the Thomson Reuters’ Worldscope Database. Corresponding standard errors are shown in parentheses. ***, ** and * denote statistical significance at the 1%, 5% and 10% level respectively.

Leverage Model (1) (2) (3) (4) (5) (6) Constant 0.053 (0.197) 0.026 (0.197) 0.067 (0.193) 0.768* (0.435) 0.727* (0.434) 0.839* (0.432) %_Women 𝑡−1 -0.177 (0.132) -0.156 (0.100) Nr_Women𝑡−1 -0.017* (0.010) -0.016** (0.008) Women_Dummy 𝑡−1 -0.107*** (0.039) -0.036 (0.030) %_Independent𝑡−1 -0.030 (0.041) -0.028 (0.040) 0.016 (0.044) -0.040 (0.040) -0.044 (0.040) -0.048 (0.039) Board_Tenure 0.000 (0.005) 0.001 (0.005) 0.000 (0.005) -0.005 (0.004) -0.006 (0.004) -0.004 (0.004) Board_Size_log𝑡−1 -0.186* (0.103) -0.131 (0.112) -0.194** (0.096) -0.231** (0.102) -0.205** (0.102) -0.226** (0.099) Board_Meetings 0.012*** (0.004) 0.012*** (0.004) 0.013*** (0.004) -0.007*** (0.003) -0.007*** (0.003) -0.007** (0.003) MTB -0.082*** (0.020) -0.082*** (0.020) -0.076*** (0.019) -0.102*** (0.020) -0.099*** (0.020) -0.089*** (0.020) CAPEX 0.597** (0.294) 0.570* (0.292) 0.656** (0.282) -0.201 (0.271) -0.186 (0.270) -0.189 (0.266) Growth_log 0.072 (1.115) 0.182 (1.114) -0.208 (1.099) 1.099* (0.609) 1.124* (0.606) 1.075* (0.607) Firm_Size_log 0.085*** (0.031) 0.082** (0.032) 0.088*** (0.031) 0.015 (0.060) 0.018 (0.060) 0.003 (0.060) σROA -0.836** (0.323) -0.859*** (0.322) -0.766** (0.315) 1.173*** (0.285) 1.17*** (0.284) 1.236*** (0.277) R-squared 0.190 0.194 0.199 0.884 0.885 0.882 Adjusted R-squared 0.163 0.168 0.173 0.849 0.851 0.847 N 312 312 312 312 312 312

Firm fixed effects Time fixed effects

(28)

28

4.3 Women in the management board in two-tier companies

The results are straightforward, evidence exists that board structure affects the relationship between gender diversity and risk. However, in this research only women in the supervisory board are used for companies with a two-tier board structure, while executive women are included in the proxies for gender diversity in companies with a one-tier board structure. The management board also influences risk-taking through the executive channel, it might give some additional insight to see how the percentage of female executives interacts with female representation in supervisory boards. I first check whether there is a correlation between women in supervisory and in management boards. The returned correlation value is 0.137, which is rather low. Hence, I can safely add a control variable for female managers and interaction term FemaleManagers𝑡−1*Gender_Diversity𝑡−1 without being concerned about multicollinearity.

The results are presented in Table A.4 in the Appendix. By adding these two variables the relationship between gender diversity in supervisory boards and risk-taking becomes insignificant. However, without controlling for fixed effects, it seems that increasing the percentage of women in the management board does increase leverage. This is similar to the results of Liu et al. (2014), who find relatively strong evidence that female executives influence firm performance. For the interaction term only weak significant evidence is found. Thus, there is no credible evidence that increasing the ratio of female managers has any effect on the relationship between gender diversity in the supervisory board and risk-taking, but there is likewise no proof to suggest otherwise due to low sample issues.

4.4 Robustness tests

4.4.1 Mean/Median splits and the Critical Mass Theory

(29)

29

median %_Women in Panel B, whereas the below mean/median coefficients are all negative. Furthermore, all coefficients are insignificant. This is in line with the conclusions from regressions (2) and (3), where no significant relationship between gender diversity and risk-taking could be established.

(30)

30 Table 7

FE OLS regression using a sub-sample analysis: classified by board structure, mean and median

This table presents the results of %_Women on risk-taking for sub-samples classified by board structure, mean and median. Panel A shows the results for firms with a one-tier board structure, and Panel B for firms with a two-tier board structure. Leverage is calculated as the financial debt divided by the sum of equity and financial debt. %_Women equals the proportion of women in the board of directors. %_Independent is the percentage of independent board members. Board_Tenure is the average number of years each board member has been on the board of directors. Board_Size_log is the natural logarithm of the number of board directors. Board_Meetings is the number of board meetings each year. MTB is calculated as the market value of total assets divided by the book value of total assets. CAPEX is the ratio of capital expenditures with respect to total assets. Growth_log is the logarithm of the yearly growth in assets. Firm_Size_log is the logarithm of total assets. σROA is calculated as the volatility over a period of five years of the ratio of EBIT to total annual assets. The data is retrieved from the Thomson Reuters’ Asset 4 Database (ESG) and the Thomson Reuters’ Worldscope Database. Results of control variables are omitted in the table to save space. Cross-sectional and time fixed effects are included in all regressions. Corresponding standard errors are shown in parentheses. ***, ** and * denote statistical significance at the 1%, 5% and 10% level respectively.

Leverage

Panel A: One-tier Above Mean Below Mean Above Median Below Median

%_Women𝑡−1 0.137 (0.085) -0.181 (0.146) 0.086 (0.080) -0.097 (0.177) Observations 433 389 471 351 R-squared 0.890 0.897 0.889 0.899

Panel B: Two-tier Above Mean Below Mean Above Median Below Median

%_Women𝑡−1 0.016 (0.128) -0.252 (0.206) -0.0244 (0.099) -0.238 (0.240) Observations 162 283 231 214 R-squared 0.911 0.904 0.910 0.919

The results are demonstrated in Table 8 below. Cross-sectional effects are omitted due to a near singular matrix. Surprisingly, only 109 of 172 companies accomplished a critical mass in their board of directors at least once in the sample period. However, having three or more female directors does not result in a significant impact on risk-taking in terms of leverage, compared to firms with two or less female directors, which do have a significant and positive %_Women coefficient.

(31)

31 𝜎𝑅𝑂𝐴𝑖,𝑡 = α0 + 𝛽1%_𝑊𝑜𝑚𝑒𝑛𝑖,𝑡−1+ 𝛽2%_𝐼𝑛𝑑𝑒𝑝𝑒𝑛𝑑𝑒𝑛𝑐𝑒𝑖,𝑡−1+ 𝛽3𝐵𝑜𝑎𝑟𝑑𝑆𝑡𝑟𝑢𝑐𝑡𝑢𝑟𝑒_𝐷𝑢𝑚𝑚𝑦𝑖,𝑡+ 𝛽4𝐵𝑜𝑎𝑟𝑑𝑆𝑡𝑟𝑢𝑐𝑡𝑢𝑟𝑒_𝐷𝑢𝑚𝑚𝑦𝑖,𝑡∗ 𝐺𝑒𝑛𝑑𝑒𝑟_𝐷𝑖𝑣𝑒𝑟𝑠𝑖𝑡𝑦𝑖,𝑡−1+ 𝛽5𝐵𝑜𝑎𝑟𝑑_𝑇𝑒𝑛𝑢𝑟𝑒𝑖,𝑡+ 𝛽6𝐵𝑜𝑎𝑟𝑑_𝑆𝑖𝑧𝑒𝑖,𝑡−1+ 𝛽7𝐵𝑜𝑎𝑟𝑑_𝑀𝑒𝑒𝑡𝑖𝑛𝑔𝑠𝑖,𝑡+ 𝛽8𝑀𝑇𝐵𝑖,𝑡+ 𝛽9𝐶𝐴𝑃𝐸𝑋𝑖,𝑡+ 𝛽10𝐺𝑟𝑜𝑤𝑡ℎ𝑖,𝑡+ 𝛽11𝐹𝑖𝑟𝑚_𝑆𝑖𝑧𝑒𝑖,𝑡+ 𝛽12𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒𝑖,𝑡+λ𝑡+ ε𝑖,𝑡 (4)

Where leverage (𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒𝑖,𝑡) is included as a control variable. The results in Table 8 show that using the standard deviation of the return on assets as a proxy for risk-taking leads again to a significant, but negative, relationship between the percentage of women and the volatility of the return on assets for companies with two or less female directors. For companies with three or more female directors, although the coefficient has turned negative, it is still not significant.

(32)

32 Table 8

FE OLS regression by using a sub-sample analysis: Critical Mass Theory

This table presents the results of regression model (1) and (4), where the sample is divided into two sub-samples classified by the number of female directors. Sub-Sample 1 consists of firm observations with zero, one or two female directors, and Sub-Sample 2 consists of firm observations with three or more female directors. Leverage is calculated as the financial debt divided by the sum of equity and financial debt. %_Women equals the proportion of women in the board of directors. %_Independent is the percentage of independent board members. Board_Tenure is the average number of years each board member has been on the board of directors. Board_Size_log is the natural logarithm of the number of board directors. Board_Meetings is the number of board meetings each year. MTB is calculated as the market value of total assets divided by the book value of total assets. CAPEX is the ratio of capital expenditures with respect to total assets. Growth_log is the logarithm of the yearly growth in assets. Firm_Size_log is the logarithm of total assets. σROA is calculated as the volatility over a period of five years of the ratio of EBIT to total annual assets. The data is retrieved from the Thomson Reuters’ Asset 4 Database (ESG) and the Thomson Reuters’ Worldscope Database. Corresponding standard errors are shown in parentheses. ***, ** and * denote statistical significance at the 1%, 5% and 10% level respectively.

Leverage σROA

Sub-Sample 1 Sub-Sample 2 Sub-Sample 1 Sub-Sample 2

Constant 0.153 (0.111) 0.329 (0.226) 0.029** (0.012) 0.085*** (0.031) %_Women 𝑡−1 0.271* (0.140) 0.218 (0.152) -0.050*** (0.016) -0.009 (0.021) BoardStructure_Dummy 0.027 (0.030) -0.016 (0.073) -0.010*** (0.003) 0.001 (0.010) BoardStructure_Dummy* %_Women 𝑡−1 -0.023 (0.301) -0.165 (0.281) 0.137*** (0.033) -0.011 (0.039) %_Independent𝑡−1 -0.012 (0.033) -0.005 (0.065) 0.018*** (0.004) 0.008 (0.009) Board_Tenure -0.009** (0.004) -0.014*** (0.004) -0.002*** (0.000) -0.001** (0.001) Board_Size_log𝑡−1 -0.019 (0.081) 0.385** (0.160) 0.025*** (0.009) -0.028 (0.022) Board_Meetings 0.005 (0.003) 0.007 (0.005) 0.001** (0.000) 0.003*** (0.001) MTB -0.022** (0.010) 0.010 (0.019) 0.004*** (0.001) 0.009*** (0.003) CAPEX 0.638*** (0.240) -0.010 (0.358) -0.014 (0.027) 0.087* (0.049) Growth_log -0.467 (0.948) -3.56** (1.785) -0.185* (0.107) -0.889*** (0.244) Firm_Size_log 0.047*** (0.017) -0.048 (0.030) -0.003 (0.002) -0.007* (0.004) σROA -1.879*** (0.305) -0.313 (0.427) Leverage 0.024*** (0.004) -0.006 (0.008) R-squared 0.126 0.108 0.167 0.261 Adjusted R-squared 0.105 0.046 0.146 0.209 N 829 305 829 305

Firm fixed effects Time fixed effects

(33)

33 4.4.2 Effect of the financial industry

In this section, as a robustness check for regression (1) including the board structure variable and interaction term (Table 4), a collection of financial companies is added to the sample, which are usually omitted in the main analysis in most governance studies (Sila et al., 2016). This has as a benefit of increasing the sample size, which may give additional insights. To control for financial companies, a dummy variable (Financial_Dummy) is included in the regression. Again, it is not possible to perform a cross-sectional fixed effects OLS regression as this creates a near singular matrix. The results can be found in Table 9 below. Similar to the results from regression (1), all of the proxies for gender diversity are significantly positive. Furthermore, the financial dummy is positively related to leverage at the 1 percent significance level. This indicates that, compared to non-financial companies, financial companies have higher leverage ratios. This makes sense as, in general, borrowed money is the stock in trade of the financial industry.

(34)

34 Table 9

Pooled and FE OLS with lagged board characteristics of all companies combined, including financial companies: effect of gender diversity on leverage.

This table presents the results of regression (1) including financial companies. Columns (1)-(3) exclude fixed effects, columns (4)-(6) include time fixed effects. Leverage is calculated as the financial debt divided by the sum of equity and financial debt. %_Women equals the proportion of women in the board of directors. Nr_Women is the amount of women in the board of directors. Women_Dummy equals 1 if at least one woman is present in the board of directors and 0 otherwise. BoardStructure_Dummy equals 0 in case of a one-tier board structure and 1 in case of a two-tier board structure. Financial_Dummy equals 1 in case of a financial company and 0 otherwise. %_Independent is the percentage of independent board members. Board_Tenure is the average number of years each board member has been on the board of directors. Board_Size_log is the natural logarithm of the number of board directors. Board_Meetings is the number of board meetings each year. MTB is calculated as the market value of total assets divided by the book value of total assets. CAPEX is the ratio of capital expenditures with respect to total assets. Growth_log is the logarithm of the yearly growth in assets. Firm_Size_log is the logarithm of total assets. σROA is calculated as the volatility over a period of five years of the ratio of EBIT to total annual assets. The data is retrieved from the Thomson Reuters’ Asset 4 Database (ESG) and the Thomson Reuters’ Worldscope Database. Corresponding standard errors are shown in parentheses. ***, ** and * denote statistical significance at the 1%, 5% and 10% level respectively.

Leverage (1) (2) (3) (4) (5) (6) Constant -0.139* (0.078) -0.118 (0.081) -0.155** (0.078) -0.153* (0.078) -0.110 (0.081) -0.174** (0.078) %_Women 𝑡−1 0.105 (0.074) 0.181** (0.078) Nr_Women𝑡−1 0.012* (0.006) 0.019*** (0.006) Women_Dummy 𝑡−1 0.034 (0.021) 0.048** (0.022) BoardStructure_Dummy 0.048** (0.022) 0.053*** (0.020) 0.095** (0.040) 0.051** (0.022) 0.055*** (0.020) 0.100** (0.040) BoardStructure_Dummy* Gender_Diversity𝑡−1 -0.223 (0.144) -0.022** (0.010) -0.086** (0.042) -0.228 (0.144) -0.023** (0.010) -0.093** (0.042) Financial_Dummy 0.066*** (0.023) 0.067*** (0.023) 0.068*** (0.023) 0.064*** (0.023) 0.064*** (0.023) 0.067*** (0.023) %_Independent𝑡−1 -0.039 (0.029) -0.045 (0.029) -0.026 (0.030) -0.038 (0.029) -0.044 (0.029) -0.026 (0.030) Board_Tenure -0.011*** (0.003) -0.011*** (0.003) -0.01*** (0.003) -0.010*** (0.003) -0.011*** (0.003) -0.009*** (0.003) Board_Size_log𝑡−1 -0.010 (0.062) -0.012 (0.066) -0.035 (0.061) -0.035 (0.063) -0.055 (0.067) -0.055 (0.061) Board_Meetings 0.007*** (0.002) 0.007*** (0.002) 0.007*** (0.002) 0.007*** (0.002) 0.007*** (0.002) 0.007*** (0.002) MTB -0.009 (0.009) -0.010 (0.009) -0.009 (0.009) -0.007 (0.009) -0.006 (0.009) -0.006 (0.009) CAPEX 0.198 (0.166) 0.189 (0.166) 0.196 (0.165) 0.175 (0.167) 0.167 (0.166) 0.165 (0.165) Growth_log -0.800 (0.568) -0.789 (0.566) -0.849 (0.566) -1.093* (0.583) -1.110* (0.581) -1.152** (0.582) Firm_Size_log 0.081*** (0.011) 0.079*** (0.011) 0.083*** (0.011) 0.085*** (0.011) 0.082*** (0.011) 0.086*** (0.011) σROA -0.094** (0.039) -0.094** (0.039) -0.094** (0.039) -0.089** (0.039) -0.089** (0.039) -0.092** (0.039) R-squared 0.153 0.155 0.154 0.162 0.165 0.162 Adjusted R-squared 0.146 0.147 0.146 0.150 0.152 0.150 N 1418 1418 1418 1418 1418 1418

Firm fixed effects Time fixed effects

(35)

35 5. DISCUSSION AND CONCLUSION

Recently, many firms are under legislative and public pressure to embrace the presence of women in the boardroom (Ferreira and Kirchmaier, 2013). Several countries, such as Germany, France and the Netherlands, have implemented mandatory gender quotas towards stimulating firms to appoint a greater proportion of women. As the level of female directors has increased through the years (European Commission, 2016; Faccio et al., 2015), a growing section in literature has investigated the economic consequences of female participation on corporate boards. However, the conclusions remain to a high degree inconsistent and inconclusive (Carter et al., 2010; Francoeur et al, 2008; Gul et al., 2011).

In this study the widely researched relationship is positioned in a different context. Based on agency theory, resource dependence theory, upper echelons theory and numerous relevant psychological and economic theories with regard to gender-based differences and corporate governance, it was hypothesised that the choice of having a one-tier or a two-tier board structure affects the relationship between gender diversity and risk-taking. A one-tier board structure uses a unitary board that is collectively responsible for the company’s long-term strategy and oversees this process at the same time. In a two-tier board structure, the roles of decision-management and decision-control are split by establishing two separate boards, a management and a supervisory board. Both board structures differ greatly in their way of practice, size, composition, norms and duties (Block and Gerstner, 2016), it is therefore likely that the type of board structure female directors are appointed to affects the intensity in which they can impact risk policies.

Referenties

GERELATEERDE DOCUMENTEN

bijvoorbeeld naar Duits recht – niet een regel op grond waarvan een lening door een aandeelhouder of moedervennootschap, verstrekt op een moment dat het eigen vermogen van

The surface water (groundwater) fraction was calculated by summing all self- supplied withdrawals or public supply deliveries of surface water (groundwater) within a CFS Area

In evaluating the composition and thickness of the oxide skin on the eutectic surface, the adventitious C overlayer could be ignored, as it would affect the signal from both

The combination of board independence and board gender diversity is only not significant to environmental decoupling (-0,0159), while showing significant negative correlations

The data concerning directors’ and CEOs’ skills, CEO power, board size, gender diversity, and, for some companies, other variables was manually collected from the

Van Grieken already serves for the second time on a supervisory board in which the CEO was honored as the female entrepreneur of the year in the Netherlands (Zakenvrouw van

Examining the effect of firm size in their sample of A-share- listed non-financial companies, Li and Chen (2018) identify that an increased fraction of female board members enhances

The results show overall good agreement between experimental and numerical data with average error of 7.2% for thermocouple measurements and 1% for Acoustic Gas