The impact of gender diverse boards on firm financial performance in the Netherlands

(1)

The Impact of Gender Diverse Boards on Firm

Financial Performance in The Netherlands

Faculty of Economics and Business (FEB)

BSc Economics & Business - Specialisation Economics & Finance

Abstract

The past decades, gender diversity on corporate boards has been a frequently debated topic across the world. Despite almost a century of equal rights in The Netherlands, women are still extremely underrepresented on corporate boards. This paper adds to the worldwide debate by discussing whether more women in boardrooms would be economically beneficial in the Netherlands. Empirical data on 83 listed Dutch firms is observed from 2008 to 2014. After obtaining data of sufficient quality and controlling for industry, multicollinearity and fixed effects, results suggest that the represenation of female board members is positively associated with firm financial performance, using Tobin’s Q, RoE and RoA as performance measures.

Keywords: Gender diversity, firm financial performance, gender quota, board of directors

JEL classifications: G30, G38, J20, J16

Date:

29-6-2016

Name:

Felix Korse

Student Number:

10583084

Period:

Semester 2, 2015/2016

Specialization:

Economics and Finance

(2)

Statement of originality

This document is written by student Felix Korse who declares to take full responsibility for the contents of this document.

I declare that the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

Important notes

Nations across the world might differ in terms of terminology. In the Netherlands, most corporations still use a two-tier board model, meaning the executive board and the supervisory board are two separately functioning boards. However, internationally it is common to use a one-tier model in which directors, both executive (inside) and non-executive (outside) directors work together within the same board of directors (Lückerath-Rovers, 2011). In order to avoid confusion, this paper applies the international terminology and refers to both executive and non-executive directors when using the term “board”. A lot of the literature that will be discussed is from abroad. Boards in The Netherlands can differ in terms of economic, legal and cultural environments. This might cause differences in the effects of the explanatory variables of firm performance and their relationships. Of course, this paper adapts its analysis to the Dutch environment

(4)

1 Introduction

For the past few years there has been a worldwide discussion on the lack of females in corporate boards (Carter et al., 2010). Gender diversity creates equal opportunity for men and women, which is important from a political and social perspective. However, in many papers across the world mixed evidence has been found regarding the relationship between board gender diversity and financial performance (Nielsen & Huse, 2010).

So far, numerous countries have implemented a gender quota that requires a higher

percentage of women in corporate boards (Terjesen & Sealy, 2016). In Europe, Norway took the lead in 2002, requiring that in five years a minimum of 40% of the board had to be of the female gender (Rhode & Packel, 2011). Quickly after, other countries followed among which Spain, Iceland, France and Germany setting a feasible quota for the near future. There are ongoing debates on passing a similar law in other countries such as the U.K, Belgium, Italy and also in The Netherlands.

At the end of 2007, 5% of all board directors in The Netherlands were female (Lückerath-Rovers, 2011). The Dutch Corporate Governance Monitoring Committee (DCGMC) reacted by recognizing the importance of board diversity. Nevertheless, the DCGMC did not introduce any targets or requirements for the next year. The DCGMC did, however, suggest that Dutch corporations should “aim for a diverse composition in terms of such factors as gender and age” (Hawkins, 2014, pp. 162). The following years, the average share of female directors slowly increased to 15.5% in 2014 (Lückerath-Rovers, 2015).

On November 27th, 2014, Dutch minister of emancipation Jet Bussemaker decided it was time for change. She threatened to introduce a gender quota on January 1st 2016 in case the top 200 Dutch companies would not show significant effort to increase the female percentage on boards. However, despite extreme lack of effort by these firms, she postponed the introduction of the gender-quota. On November 16th 2015, Bussemaker stated that the top 200 firms would get four more years to increase the percentage of females on corporate boards to approximately 30% (Abels, 2015).

Especially for The Netherlands, there are few studies (e.g. Lückerath-Rovers, 2011; Marinova et al.; 2015; Molenkamp, 2015) that have empirically examined the effect of gender diverse boards on firm financial performance. These studies mainly examined small time frames of just one year, which does not make results generalizable for other financial periods (Molenkamp, 2015). Therefore, it would be interesting to examine what conclusions could be drawn with respect to the effect of diversity from 2008 to 2014 (Lückerath-Rovers, 2008). The analysis carried out in this paper includes more recent data, new perspectives and appropriate research methods.

(6)

This paper focuses on the following research question: What is the effect of the percentage of females on corporate boards on the financial performance of Dutch firms?

2 Theoretical overview

In this chapter the function and importance of a corporate board is explained in paragraph 2.1, making use of the agency theory and resource dependency theory. Also expectations with respect to the research question are formed based on these theories. Additionally, prior worldwide literature and thereafter prior Dutch literature about the impact of gender diverse boards on firm financial performance will be reviewed in paragraph 2.2. The main limitations of these papers are discussed in paragraph 2.3. Finally, hypotheses are formed in paragraph 2.4.

2.1 Theoretical background

A vast majority of academic articles about what effect the percentage of females on a board has on firm performance does not develop a theoretical framework (Terjesen et al., 2009). Nevertheless, two theories provide useful perspectives on the potential effect of a gender diverse board on firm performance. Namely, the agency theory (Jensen & Meckling, 1976) and the resource dependency theory (Pfeffer & Salancik, 1978).

The most dominant theoretical perspective applied in corporate governance studies is agency theory (Daily et al., 2003). It is also the most often used theoretical framework by researchers in economics and finance in order to examine relations between board characteristics and firm value (Carter et al., 2003). Jensen and Meckling (1976) define an agency relationship as a contract under which a person (principal) engages another person (agent) to perform some service on their behalf. They add that as both parties in this theoretical framework are assumed to be rational and to be aiming to maximize their own benefits, conflicts of interests between the principal and the agent arise. It is also described by Jensen and Meckling (1976) that agency problems can be limited by the principal in case he can monitor the actions of the other party. In an agency framework a board of directors plays a very critical role. It controls and monitors managers, which contributes to the reduction of the conflicts of interest between the stock holders (principals) and managers (agents) (Fama & Jensen, 1983). According to Carter et al. (2003), the main task of the board of directors is to resolve agency problems, making use of their ability to set compensation and to replace managers that do not provide (enough) value for the stock holders.

Adams & Ferreira (2009) suggested that female directors have better monitoring skills. These findings are supported by empirical evidence of Dang et al. (2014), which suggests that female board

(7)

directors are willing to put in more effort into tasks. Terjesen et al. (2009) showed that females are well-prepared for meetings and ask more questions than males. Fancoeur et al. (2008) add that a gender diverse board would provide new perspectives that could be useful when dealing with problematic issues. These characteristics are proven to be useful in reducing informational bias, decision-making and planning (Westphal & Milton, 2010), which in turn reduces moral hazard and adverse selection problems, two common agency problems (Linder & Foss, 2013).

Furthermore, Carter et al. (2003) described that, from an agency perspective, independence of boards is critical for boards to act in best interest of shareholders. As previously mentioned, a board exists of executive and non-executive directors. The non-executive directors are board members who do not have important relationships with stakeholders and are more independent than executive directors, as Nguyen & Nielsen (2010) described. Norwegian empirical evidence showed that an increase in the share of female directors corresponds to an increase in the share of non-executive directors (Nygaard, 2011). This suggests that female board members are more likely to become non-executive directors, which results in a more independent board. The latter is supported by Terjesen et al. (2015) by showing that gender diversity on boards increases the level of board independence.

Although the agency theory does not provide a precise prediction about the potential effect board specific characteristics have on firm performance (Hermalin & Weisbach, 2000), it does provide a theoretical basis from it becomes possible to hypothesize about the effects a gender diverse board might have on firm financial performance. Thus, from an agency theory perspective, an expected outcome to our research can be formed. Given that female board members have better monitoring capabilities and are more independent, they possess characteristics that are more likely to be beneficial to firm financial performance. Therefore, the following is expected:

Expectation from Agency theory: Gender diverse boards enhance firm financial performance.

The resource dependency theory is another popular theoretical framework used by investigators examining what effect gender diversity has on firm performance. Pfeffer and Salancik (1978) describe that this theory assumes that firms are transparent organisations that have an interdependent relationship with external entities. According to Daily et al. (2003), the provision of linkages with those entities enhances, not only organizational functioning, but also firm performance. Daily et al. (2003) also describe that board members are seen as the “boundary spanners” of the organization and its environment. Pfeffer and Salancik (1978) propose four advantages due to external linkages of a board: 1) directors provide information, expertise and knowledge, 2) directors serve as

communication pathways between firms, 3) directors provide legitimacy, meaning right, credibility and acceptance, in the environment in which the firm operates, 4) they are supported by other firms.

(8)

Potential effects gender diverse boards have on firm financial performance will be discussed from a resource dependency theory perspective in the following paragraph.

Female directors bring various beneficial capabilities to boards. For example, female board members provide different knowledge, talents and skills than the members of opposite gender (Terjesen et al., 2009). Hillman et al. (2007) state that female representation on boards results in linking firms to different parties than males do, for example to different clients, potential employees, suppliers and investors. Additionally, Hilman et al. (2007) also concluded that female directors have a better influence on powerful business parties and provide more expertise than male board members. These findings are supported by Brammer et al. (2007) who found a positive effect of female board representation on firm reputation. Furthermore, Dang et al. (2014) concluded that increasing the ratio of females on board has a positive effect on legitimacy, which is one of the previously described benefits that derive from external linkages of boards according to the resource dependency theory. The mentioned points are all likely to be beneficial to a better firm financial performance.

Thus, taking a resource dependency perspective in attempting to analyse the relationship between female board representation and firm financial performance, suggests that a higher percentage of females on boards has a positive effect on firm performance.

Expectation from resource dependency theory: Gender diverse boards enhance firm financial performance

Before formulating our final hypotheses, it is useful to review prior literature about the influence of gender diverse boards on firm financial performance.

2.2 Literature review

As stated before not much research on the effects of gender diversity on firm financial performance has been done in The Netherlands. However, prior worldwide studies show that examination of the impact of a gender diverse board on a firm’s overall performance can result in mixed conclusions (Rhode & Packel, 2011). Although most studies reported positive correlations between a gender diverse board and firm financial performance, other studies found a negative relationship between both variables and some found no significant relationship at all. The following three sections review prior worldwide research that has been done so far. Thereafter, relevant Dutch market research will be discussed. The following sections also reflect on main methodological issues of articles. The most common ways of measuring financial performance used in prior research are Tobin’s Q, Return on Equity (RoE) and Return on Assets (RoA). It should be pointed out that the studies discussed cover

(9)

different samples, measures of performance, time spans and regions. A brief overview of the findings of prior research papers covered in the following four sections is given in table 1.

(10)

2.2.2 Studies that found negative effects of gender diverse boards on firm performance

A limited number of academic papers report to have found a negative relationship between female board representation and financial performance.

In a follow-up study by Adams and Ferreira, they examined a sample of US companies ranging from 1996 to 2003 and uncovered negative relationships between gender diversity and both RoA and the natural logarithm of Tobin’s Q (Adams and Fereirra, 2009). In their conclusion, the writers add that the positive relationship between gender diversity and firm financial performance found in their previous study was not robust to any method of addressing endogeneity of gender diversity. On the contrary, the follow-up study took the problem of endogeneity more into account and therefore they conclude that the relation appears to be negative (Adams & Fereirra, 2009). However, Adams and Fereirra (2009) also argue that their findings might be due to a strong shareholder protection, together with the additional monitoring ability of the board due to extra females, which might lead to over-monitoring. Too much monitoring could lead to a negative effect on firm financial performance. Adams and Fereirra (2009) suggest that in case of a weak shareholder environment, the effect might be positive. Nonetheless, other literature found that gender diversity on boards has a negative effect on firm performance, despite a weak protection by investors (Okike, 2007; Ujunwa et al., 2012). Ahern and Ditmarr (2012), examining Norwegian firms, showed that after the introduction of a gender quota, an increase in females correspond with minor financial losses using Tobin’s Q as a measure of performance. Ahern and Ditmarr (2012) add that these losses are most likely due to lack of experience. In a different study it was shown that more gender equality in boards led to more focussed decision-making and that it is more likely for interpersonal problems to be solved due to a more sensitive approach towards others (Nielsen & Huse, 2010). In the same study it was shown that it can enhance a firm's public image by contributing to equal opportunity. These aspects could lead to higher financial firm value and performance in the long-run according to Lückerath-Rovers (2011).

2.2.3 Studies that found no effect of gender diverse boards on firm performance

A number of empirical studies that found no significant relationship between both variables are covered below.

Carter et al. (2010) found no significant effect of gender diversity on Tobin’s Q. Also Heussein and Kiwia (2009) concluded there was no clear relationship between female representation on boards and firm performance examining a sample of 250 U.S. firms. Results showed that gender diversity negatively affects Tobin’s Q find and it has an insignificant effect on RoA. This last finding is supported by Rose (2007) who, from 1998 to 2001, examined all Danish firms listed on the

(11)

directors did not have a significant effect on the performance measures RoA, RoE and RoI. Likewise, another study by Miller and Triana (2009), examining a sample of 236 firms listed in the Fortune 500, also did not find evidence on gender diversity on boards significantly affecting RoI and RoS. Farrell and Hersch (2005) presented empirical proof that firms with a higher financial performance also generally tend to have more gender diverse boards of directors. However, Farrell and Hersch (2005), also state that no evidence could be found that gender diversity improves firm performance itself.

2.2.4 The Dutch case: effect of gender diverse boards on firm financial performance

As earlier mentioned, thus far not much relevant research has been done on the Dutch market. In the next section two often cited papers and one bachelor thesis are covered.

In a study by Lückerath-Rovers (2011) the relationship between gender diversity and firm value for Dutch companies was examined, extending two well-known research methods: 1) the Catalyst method (Catalyst, 2007) and 2) the McKinsey method (McKinsey & Company, 2007). Both methods use the Return on Equity (RoE) as a performance measure. Lückerath-Rovers (2011) examined 99 firms listed in the Dutch Female Board Indices of 2005-2007. These yearly presented indeces list the percentage of females on boards for 99 companies noted on the Euronext

Amsterdam. Lückerath-Rovers (2011) concludes that Dutch firms with female directors generally tend to outperform firms without female directors. Why she chose not to use a research method based on Tobin’s Q instead of RoE was not made entirely clear, although she does name studies that used RoE and found a positive relationship and a Danish study using Tobin’s Q in which there was not found any correlation (i.e. Rose, 2007). This led Lückerath-Rovers (2011) to belief using RoE as a measure of performance would therefore be an obvious one. However, she does not refer to any prior studies that did not find a significant or negative relationship while also using RoE as dependent variable (e.g. Ahern & Dittmar, 2012; Zahra & Stanton, 1988). Moreover, unlike the Danish study Lückerath-Rovers (2011) referred to, multiple studies mentioned in section 2.2.1 did find positive relationships between gender diversity and firm performance using Tobin’s Q as performance measure (e.g. Adams, Fereirra, 2003; Campbell & Minguez-Vera, 2008; Carter et al.,2003; Deszõ & Ross, 2007; Nguyen & Faff, 2007;). Lückerath-Rovers (2011) states that the biggest shortcoming in her study was that the significant and positive relationship found did not suggest causality in any way and therefore it was premature to use it as an argument to appoint more women. Accordingly, she recommends future research to take factors that cause endogeneity problems into account. Marinova et al. (2015) used Tobin’s Q to examine a sample of 186 listed firms observed in 2007, of which 102 companies were Dutch and 84 Danish. Marinova et al. (2015) explained Denmark and the Netherlands were examined because of the similarities in terms of gender equality and corporate governance. Marinova et al. (2015) concluded not to have found a significant relationship

(12)

between gender diversity on boards and the Tobin’s Q. They advise future studies to refer to board characteristics, such as education and independence. They also recommend to use accounting-based measures, such as RoA or RoE and to focus on panel data instead of examining just a single year: “If gender-related performance effects can be traced over several years, the quality and implications of the analysis will improve considerably, as dynamic factors will also be captured in the relationship” (Marinova et al., 2015, pp. 10).

Molenkamp (2015) had the initial intention to examine the effect of diversity of education and nationality on firm financial performance. Nevertheless, his empirical research showed no proof that financial performance was affected by these variables. However, he did find a significant positive effect of gender diversity on firm performance measured as Tobin’s Q and RoA.

Table 1 provides a list of findings on the influence of gender diversity on boards on firm financial performance, as discussed above.

Table 1: Former findings Author(s),

year

region Time span GD measure FFP measure Effect, coefficient

(std.) or result Carter, et al., 2003 USA 1997 Percentage of females in board Tobin’s Q Positive, 9.43* (4.5) Carter, et al., 2010 USA 1998-2008 Number of females in board RoA Positive, 0.57* (0.12)

Deszõ & Ross, 2012 USA 1992-2006 Percentage of females X innovation intensity Tobin’s Q, RoE and RoA

Positive, 0.33* (0.11); 0.16* (0.02); 0.26* (0.08) Nguyen and Faff, 2007 Australia 2001-2011 Percentage of females Tobin’s Q Positive, 2.24* (0.55) Adams and Ferreira, 2003 cross-sectional sample 1998 Percentage of females in board Tobin’s Q Positive, 3.15* (0.002) Erhardt et al. 2003

USA 1993 & 1998 Percentage of

females in board

RoI and RoA Positive 0.32* (0.09); 0.25* (0.06) Campbell and Minguez-Vera, 2008 Spain 1995-2005 Percentage of females by the Blau index5 Tobin’s Q Positive, 16.62* (1.93) Schwartz-Ziv, 2013 Israel 2000-2007 Percentage of females in board Positive, .18 (.10)**

(13)

Rohner & Dougan, 2012 cross-sectional sample 2005-2011 Presence of at least one woman in board Positive 1 Lui et al., 2014 China 1999 to 2011 Percentage of females in board

RoA and RoS Positive, 0.07* (0,02); 0.55* (0.16) Adams and Ferreira, 2009 USA 1996-2003 Percentage of women in board Ln (Tobin’s Q) and RoA Negative, -5.92* (2.68); -231.41* (111.79] Ahern and Ditmarr, 2012 Norway 2001-2009 Percentage of women in board Tobin’s Q Negative, −1.94* (0.59) Carter et al., 2010 USA 1998-2008 Number of females in board Tobin’s Q None, -0.16 (-0.74) Heussein and Kiwia, 2009 USA 2000-2006 Dummy variable: GNDPAR2 and GNDINFit3 RoA and Tobin’s Q None, 1.26 (0 .92); Negative -1.42* (0.657)

Rose, 2007 Denmark 1998-2001 Percentage of

women in supervisory board 4 Tobin’s Q None, 0.29 (1.01) Schrader et al., 1997 USA 1992 and 1993 Percentage of women in board

RoA, RoE and RoI None, 0.02 (0.08); None -0.06 (0.081); None 0.09 (0.07) Miller and Triana, 2009 USA 2002 Percentage of females by the Blau index5

RoI and RoS standardized to one PM6

None 0.15 (0.31)

Farrell and Hersch, 2005

USA 1990-2000 RoA Likelihood of

adding a female Positive 1.93 (0.11) Lückerath-Rovers, 2011 The Netherlands 2005-2007 Dummy variable for presence of females RoE Positive 7 Marinova et al., 2015 The Netherlands & Denmark 2007 Percentage of females in board Tobin’s Q None 9.03 (9.38) Molenkamp, 2015 The Netherlands 2013 Percentage of females by the Blau index5

Buy and hold return

Positive 34.59* (9.66)

Note: Standard errors are given in parentheses. * represents significance at a 5% level and ** shows significance at a 10%

level. Positive= positively significant relationship found, Negative= negatively significant relationship found, None= insignificant positive or negative relationship found.

1

no coefficients were provided in this study. Result study: RoE was 4 percentage points higher than the average RoE of companies with no female board representation.

(14)

2

GNDPAR = an index assigned for female-male ratio. The binary code 1 is assigned where the ratio of female is greater than 25% and 0 is where it less. GNDINFit. = A binary code where 1 represents the influence of female directors in the board room where their number exceeds that of their male counterparts and 0 represents females being minorities.

3

GNDINFit =A binary code where 1 represents the influence of female directors in the board room where their number exceeds that of their male counterparts and 0 represents where females are a minority.

4

The supervisory board consists of non-executive directors and mainly deals with supervising.

the policy of the management board, ratifying managerial decisions, providing advice, as well as adopting the company’s annual accounts (Marinova et al., 2015).

5

Blau’s index= 1-∑pi 2 where p is the proportion of group members in a given category i (i=female in this case). 6

PM=Performance Measure.

Catalyst method: companies with female directors had a mean RoE of 23.3%. Companie without famela directors have an average RoE of just 11.1%. This was shown to be a significant difference of 110% with a t-value of t = 4.0. McKinsey & Company method: for boards with females, the RoE is 56% higher than the average for the overall sample. A T-test was not applicable, however it was shown that the prensence of females significantly correlated with ROE (P<0.01).

2.3 Main limitations of prior literature

After having summed up the prior empirical research relevant to the examination of the relationship between gender diversity and firm financial performance one might still be curious to why

researchers come to different conclusions. Researchers generally provide three explinations: Firstly, critical mass. According to Kramer et al. (2006), critical mass is the minimum size a subgroup required to have an effect on the overall group.

Secondly, the problem of time refers to an examined time period that is too small for results to be valid and reliable for future periods (Lückerath-Rovers, 2011; Marinova et al., 2015;

Molenkamp, 2015).

Thirdly, endogeneity problems due to unobserved factors or reverse causality are often not fully addressed (Adams & Ferreira, 2009; Farell & Hersch, 2005). The findings mentioned above, in most cases, do not necessarily imply causal relationships. One previously mentioned study however did claim to have found a causal relationship after finding a positive effect of board diversity on firm performance (Campbell and Minguez-Vera, 2008). The writers claim that there was no problem of reverse causality in their examination. This conclusion was drawn after employing a two-stage least squares (2SLS) test to assess whether female board representation really influences firm

performance or whether well performing corporations are just more likely to have a gender diverse board (Campbell and Minguez-Vera, 2008).

This paper attempts to circumvent these three problems as much as possible. Critical mass is difficult to deal with as the average female board representation of a sample of 83 Dutch listed companies was, although increasing over the years, just 17.0% in 2014 (Lückerath-Rover, 2015). Additionally, in the same year 39% of the companies examined did not have a single woman on their

(15)

board. Moreover, in 2008, a majority of 69 % did not show any interest in female directors. In order to prevent time-related problems, a broader time-span of 7 years is covered. Additionally, the possible existence of the endogeneity problem will be taken into account in paragraph 3.2 by discussing the possibility of applying 2SLS (two-stage-least-squares) regression, Arellano and bond regression or a fixed effects panel data method.

2.4 Hypotheses

This research paper examines the effect of gender diverse boards on firm financial performance in The Netherlands. Based on the empirical results a conclusion can be drawn on whether female representation on boards have a positive, negative or no effect on firm financial performance. In this paragraph, the hypotheses are given by making use of the expectations from the agency and

resource dependency theories and our literature review from the previous chapters.

Before covering prior relevant literature, the Agency theory and Resource Dependency Theory were discussed and expectations were formed. Based on both theories female representation is expected to have a positive effect on firm financial performance. However, before hypothesizing about a potential outcome, prior relevant studies were reviewed. Looking at Table 1, it comes to notice that most worldwide studies found that female representation on boards is able to drive firm financial performance. Although little Dutch market research has been done up to now, two papers found positive and significant relationships. Combining the literature review with the two

expectations it is likely to find that gender diversity on boards has a positive effect on firm financial performance. Correspondingly, primarily:

(H1): Gender diverse boards of directors will positively affect firm financial performance of The

Netherlands

Looking at table 1, it comes to notice that most researchers made use of either a market-based financial performance measure or accounting-based performance measures. The by prior researchers most common used market-based performance measure is Tobin’s Q, as for the accounting-based performance measure RoE and RoA were most frequently used. Moreover,

Marinova et al. (2015) recommend future research to use accounting-based measures next to Tobin’s Q. Taking the above into consideration, hypothesis 1 can be further specified into:

(H1a): Gender diverse boards of directors will positively affect a market-based financial performance

(16)

(H1b): Gender diverse boards of directors will positively affect accounting-based financial performance

of Dutch firms

As earlier stated, directors provide knowledge, information and expertise, according to the resource dependency theory. Additionally, Marinova et al. (2015) recommended future researchers to take board characteristics like education into account. Singh et al., (2008) showed that female board members are more likely to be well educated and to hold advanced degrees, e.g. a master degree in business administration. In a Canadian study it was found that about 90% of the females on board were university graduates, which suggests that female board members are more intelligent or determined than males (Burke, 1997). Furthermore, Bantel & Jackson (1989) found that a high level of education enhances innovation. Therefore, a higher level of board intelligence can be expected to enhance firm financial performance through enlarging the effect of gender diversity on firm financial performance. So, secondarily:

(H2): the relationship between gender diversity of boards and firm financial performance of Dutch

firms is positively moderated by the level of education of boards. Which can, similarly, be further specified to:

(H2a): the relationship between gender diversity of boards and market-based firm financial

performance of Dutch firms is positively moderated by the level of education of boards. (H2b): the relationship between gender diversity of boards and accounting-based firm financial

performance of Dutch firms is positively moderated by the level of education of boards.

As described above, the agency theory stresses the importance of presence of independent board members, also known as outside directors. Also, as earlier stated, it was shown by Nygaard (2011) that an increase in the amount of women corresponds to a higher level of board

independence. Additionally, Daily et al. (2003) describe that from also from a resource dependency perspective it follows that independence can contribute to higher firm performance, since this characteristic provides access to resources needed. Combining these points suggests that the degree of board independence might moderate the effect gender diversity has on firm performance. So,

(H3): The effect of gender diversity of boards and firm financial performance of Dutch firms is

positively moderated by the level of board independence. Which is similarly split up into:

(H3a): The relationship between gender diversity of boards and market-based firm financial

(17)

(H3b): The relationship between gender diversity of boards and accounting-based firm financial

performance of Dutch firms is positively moderated by the level of independence of boards.

3 Methodology

This chapter explains the research method applied to measure the impact gender diverse boards have on firm financial performance. Firstly, in section 3.1 the dependent, independent and control variables are discussed. Secondly, in section 3.2 possible solutions to the problem of endogeneity are discussed. Furthermore, in section 3.3 the data and sample used are covered. Finally, section 3.4 provides a summary of the statistical analyses performed on the data.

3.1 Variables

3.1.1 Dependent variables

Firm financial performance will be measured using both market-based and accounting-based performance measures. The market-based measure that will be used is an approximation of Tobin’s Q. As can be seen in table 1, this performance measure is frequently used in prior research. Tobin’s Q is measured by dividing the total market value of a company by their replacement cost, i.e. the sum of the market value of stock and the book value of debt divided by the book value of assets (Tobin, 1969). Or in formula form:

𝑇𝑜𝑏𝑖𝑛’𝑠 𝑄 = 𝑀𝑉 (𝑠𝑡𝑜𝑐𝑘)+𝐵𝑉(𝑑𝑒𝑏𝑡)

𝐵𝑉(𝑎𝑠𝑠𝑒𝑡𝑠) Equation (1)

Tobin’s Q provides a clear yardstick for firm financial performance. A value of Tobin’s Q above 1, indicates that a firm’s stock is more expensive than replacement cost of its assets, which suggests that the stock is overvalued. Therefore, these companies are expected by investors to create more value by making use of available resources effectively (Campbell & Minguez-Vera, 2007). Conversely, firms with a value between 0 and 1 indicate undervaluation of stock and therefore are associated with making poor use of available resources.

Also two market-based performance measures are included in this paper, RoE and RoA. Again, as can be derived from table 1, a considerable number of prior studies used these measures as well. A high RoE implies that much of the corporation’s probability is generated by efficiently use of money that shareholders have invested. The formula used to determine RoE is as follows:

(18)

𝑅𝑒𝑡𝑢𝑟𝑛 𝑜𝑛 𝐸𝑞𝑢𝑖𝑡𝑦 (𝑅𝑜𝐸) = _{𝑇𝑜𝑡𝑎𝑙 𝐴𝑠𝑠𝑒𝑡𝑠}𝑁𝑒𝑡 𝐼𝑛𝑐𝑜𝑚𝑒*100% Equation (2) A high RoA indicates how profitable a company is with respect to its total assets, i.e. how effectively a company is turning its funding into earnings. The formula used to calculate RoA is:

𝑅𝑒𝑡𝑢𝑟𝑛 𝑜𝑛 𝐴𝑠𝑠𝑒𝑡𝑠 (𝑅𝑜𝐴) = _{𝑇𝑜𝑡𝑎𝑙 𝐴𝑠𝑠𝑒𝑡𝑠}𝑁𝑒𝑡 𝐼𝑛𝑐𝑜𝑚𝑒*100% Equation (3) The above types of performance measures differ in a few respects. Firstly, RoE and RoA look back in time while Tobin’s Q is forward-looking (Chen et al. 2008). Secondly it should be taken into

consideration that net income involves future events to a limited extent, examples that matter for net income are also depreciation and amortization. However, the Q-measure is influenced by a large set of ambiguous factors, for example predictions of investors and risk (Chen et al., 2008; Campbell & Minguez-Vera, 2008).

3.1.2 Independent and control variables

In order to reliably measure the linear relationships between gender diversity, education and independence on firm financial performance measures (Tobin’s Q, RoE and RoA), 3 control variables are also included in our model. By including these variables in the model, their effects on the

dependent variables are taken into account and therefore the chance of obtaining reliable statistical results increases. Below it is described how the independent and control variables are measured. Firstly, the main variable of interest included in our regression is BOARD GENDER DIVERSITY, which is measured as the fraction of female directors on the board. In order to test hypothesis 2, the variable EDUCATION is included. Education is measured using an approach designed by Engelen et al. (2012). It is measured as the fraction of board directors holding a master degree. Hypothesis 3 is tested by including the variable INDEPENDENCE. The Dutch Corporate Governance Committee (2003) advises all supervisory directors (non-executives), except for one, to be independent. A Dutch study by Van Ees et al. (2007) showed that supervisory directors of the majority of companies in Holland can be considered to be independent. As a result, following Marinova et al. (2015), independence is measured as the number of supervisory directors divided by the total amount of directors (non-executives and (non-executives together).

Secondly, control variables included are FIRM SIZE, LEVERAGE and ln(SALES). Firm size is measured by the natural logarithm of the total assets of the firm, leverage as the ratio debt to total asset. Sales growth is measured as the natural logarithm of net sales each year. The log

transformation of total assets and sales helps to reduce skewness of these variables (Dezsö, C. & Ross, 2012). Additionally, interaction terms between gender diversity and education and gender diversity and independence will be included in the analyses.

(19)

3.2 Preventing endogeneity

As mentioned earlier in this paper a serious problem in examining the effect of board gender diversity on firm performance is the possibility of obtaining biased results due to the problem of endogeneity. Most prior literature examining panel data and taking endogeneity problems into account argue that often these problems are caused by either reverse causality or fixed effects (Adams & Ferreria, 2009). Reverse causality means that either a more gender diverse board can lead to higher firm financial performance, or financially successful firms are more likely to hire female board members. In that case, OLS coefficients obtained would be biased and inconsistent and therefore not imply causal relationships (Marinova et al., 2015). Endogeneity can also arise from fixed effects due to omitted time-invariant firm caracteristics, such as corporate culture (Adams & Ferreira, 2008). Reverse causaltity could be prevented by applying the two-stage-least-square method (2SLS) or by including lagged dependent variables. Endogeneity which arises from fixed effects can be reduced by using the fixed effects regression method (FERM). These three methods and to what extent they can be applied in this study will be discussed below.

The first way of reducing the possible reverse causal relationship between gender diversity and firm financial performance is the 2SLS method. Making use of this method makes it possible to estimate the exogenous effect of initially endogenous independent variables on the dependent variable. The problem of endogeneity is taken into account by conducting a regression that exists of two stages:

1st stage

Diversitŷ = δ₀ + δ₁Z₁+ … + δ_mZ_mi + δ_𝑚+1W_1i+ … + δ_𝑚+𝑟W_ri+ 𝑢_𝑖 Equation (4) 2nd stage

Firm Financial performance = α0+ β1

·

Diversitŷ + β2

·

W1 + … + βi

·

Wi + 𝑣𝑖

Equation (5)

When looking at the possibly endogenous variable gender diversity, in order to apply 2SLS at least one instrument Z is needed in the 1st stage which fulfils the following conditions: 1) the instruments used have an effect on the fraction of female board members (instrumental relevance) and 2) the instrument used should not have a significant effect on firm financial performance (instrumental exogeneity). If these conditions are not met, 2SLS yields no or inconsistent results and the 2SLS estimator will not be normally distributed (Stock & Watson, 2007).

Adams and Fereira (2008) discuss that it is very difficult to find instruments that meet both of the aforementioned conditions, as most factors correlated with the percentage of female board members are often other governance characteristics that should be included in performance

(20)

regressions. Examples include firm size and board independence. Likewise Marinova et al. (2015) argue it is very difficult to find valid instruments. Nevertheless, they did manage to find two instruments that met the above criteria to a certain extent, namely the share of women in the industry and the share of women squared. However, both instruments were shown to be weakly correlated with gender diversity when testing for instrumental relevance. After the second stage gender diversity showed to have an insignificant effect on firm performance, while in the OLS-regression gender diversity was found to have a significant positive effect. This OLS-OLS-regression shows a possible result of conducting 2SLS with weak instruments.

Given the above implications, this study does not apply 2SLS in order to reduce the endogeneity problems arising from reverse causality.

There is an alternative that can be used to reduce endogeneity caused by reverse causal relationships between gender diversity on boards and firm financial performance, namely by adding lagged dependent variables to models. These models take into account that gender diversity

correlates with past financial performance, which results in dynamic panel data models. However, lagged dependent variables will very likely result in problems with autocorrelation in regressions (Angrist & Pischke, 2009). Furthermore, although lagged dependent variables often acquire large and significant coefficients, the coefficients of other variables collapse to relatively small and insignificant values and sometimes even take on wrong values (Achen, 2000). In order to prevent these problems, researchers could apply the Arellano and Bond method (Dezsö & Ross, 2012). However, dynamic panel data models require advanced skills and complex estimation techniques (dr. J.C.M. van Ophem, personal communication, july 23, 2016; Angrist & Pischke, 2009).

Accordingly, this study does not conduct an Arellano and Bond test and hence, the obtained results in chapter 4 are not robust to any method of addressing endogeneity due to reverse causality between gender diversity and firm financial performance.

Nevertheless, a substantial part of the potential endogeneity between gender diversity and firm performance could also be a result of unobserved firm and time fixed variables (Adams Ferreira, 2009). A way of reducing this source of endogeneity is using a method called the FERM, which controls for some type of omitted variables without actually observing them. Contrary to the 2SLS and Arellano and Bond methods, the FERM can be used in this study as its application is relatively straightforward. Moreover, the FERM is a very appropriate method to apply since this study examines panel data or longitudinal data, i.e. the same relationships of the same companies are observed in multiple periods from 2008 to 2014.

(21)

3.3 Building the model

When dealing with endogeneity and panel data, the so-called fixed effect regression is a method used to control for 1) firm-fixed factors, omitted variables varying across companies but do not change over time (such as cultural norms) and 2) time-fixed factors, omitted variables that are constant across entities but evolve over time (such as mandatory corporate governance rules or laws). This model controls for and partials out effects of these firm-fixed and time-fixed

characteristics, which makes it possible to measure the net effect of the variables of interest on firm financial performance. Below, it is firstly shown how entity-fixed factors are taken into account in the performed regression. Thereafter, the same is done with respect to time-fixed factors.

By studying changes in the dependent variable over time, it becomes possible to eliminate the effect of omitted variables that differ across the examined companies but are constant over time (Stock & Watson, 2015). Not taking these so-called entity fixed effects into account could give rise to possible endogeneity problems in the examination of the effect gender diversity has on firm

performance (Hermalin &Weisbach, 2003; Carter et al., 2003). Consider the regression model below: Y_it = β₀ + β₁ · X_it + β₂· Z_i + u_it Equation (6)

For t = 1, … , T and i = 1, … , N. X1,it is the value of the regressor for company i in time in period t.

Where Ziis an unobserved variable that varies from one entity to another but does not change over time (for example, Zirepresents cultural attitudes toward corporate governance). The coefficient of interest is β1, which represents the effect X has on Y, holding constant the unobserved company specific characteristics Zi. Because Zivaries across companies but is constant over time, the regression in equation (5) can be interpreted as having N different intercepts, one for each company. Equation (5) can be rewritten as:

Yit = β1 · Xit + αi+ uit Equation (7)

For t = 1, … , T and i = 1, … , N.

Xit is the value of the regressor for company i in time in period t.

With αi= β0 + β2· Zi.

Equation (6) is the fixed effects regression model in which α1,…, αnare treated as unknown intercepts to be estimated, one for each company. Because intercept αi can be thought of as the effect of being a company i, the terms α1,…, αnare known as entity fixed effects. The variation in firm fixed effects comes from omitted variables that, like Ziin Equation (6), vary across firms but not over time.

(22)

This single regression model can be extended. The effect of multiple determinants of Y (firm performance measures) that are correlated with X (Gender diversity) and also change over time are taken into account by including them in the regression, to avoid omitted variable bias. Doing so, equation (7) is extended as follows:

Y_it= β₁ · X_1,it+ … + β_k· X_k,it+ α_i+ u_it Equation (8) For t = 1, … , T and i = 1, … , N.

X1,it is the value of the first regressor for company i in time in period t, X2,it the value

of the second regressor etc. and α1,…, αn are

the company-specific intercepts. with αi = β0+ β2· Zi

When equation (8) model is further simplified mathematically, with the use of Stata it becomes possible to compute fixed effect estimators by applying OLS. For the sake of simplicity, the mathematical procedures taken are shown for the single regressor version of the fixed effects model. The first step is to obtain the time-average equation of the equation (7) by only looking at the

average variables of the regressor:

Step 1) Y̅it= β1 · X̅it + αi + u̅it Equation (9)

For t = 1, … , T and i = 1, … , N.

X̅it is the average value of the regressor for company i in time in period t and α1,…, αn

are the company-specific intercepts. with αi = β0+ β2· Zi

Note that: α̅i = (1_T) ∗ T ∗ αi= αi The second step is obtaining the fixed effect equation by simply subtracting equation (9) from equation (8), which leaves:

Step 3) Yit− Y̅it= β1(Xit− X̅it) + (uit − u̅it) + (αi− αi) Equation (10) For t = 1, … , T and i = 1, … , N. X_it (X̅it) is the (average) value of the regressor for company i in time in period t and α1,…, αnare the company-specific intercepts.

(23)

As can be seen, the fixed effect and intercept are removed from the equation. Equation (10) can be further simplified to the so-called entity-demeaned formula:

Step 4) Ỹit= β1 · X̃it + ũit Equation (11)

For t = 1, … , T and i = 1, … , N.

X̃_1,it is the demeaned value of the regressor for company i in time in period t.

Alternatively, the fixed regression mode can be expressed using binary variables to denote firm-specific states. To develop the fixed effects regression model using binary variables, it is not possible to include all n binary variables plus a common intercept, for if this would be done the regressors will be perfect multicollinear. Therefore, D1i is omitted for the first company. The resulting equations is:

Yit = β0 + β1 · Xit+ γ2· D2i+ … + γn· Dni+ uit Equation (12)

For t = 1, … , T and i = 1, … , N

Let D2i equal 1 when i = 1 and otherwise zero, etc.

To derive the relationship between coefficients in (12) to intercepts in equation (7), the regression lines can be compared for each entity. In equation (7), for the first firm the regression equals Yit = β0 + β1 · Xit, so would be α1= β0. However, for the second and the other

companies the regression equals Yit = β0 + β1 · Xit+ 𝛾i so, so α𝑖 = β0+ 𝛾𝑖 for i > 2. As can be seen in step 3), by subtracting the time means the αi-term is completely

eliminated, which means it can be assumed that there will no more correlation with the error term. So, not only the intercept but also the firm fixed effect β2·Zi is eliminated from the regression which results in reduced correlation between the error term and the variable of interest Xit. Similarly, the firm fixed effects are no longer included in the error term in equation (12) since firm-specific dummies were included in the model. In both formulas, the coefficient on Xit is the same from one firm to another. The firm specific intercepts in equation (9) and the binary regressors in equation (12) have a common source: unobserved variable Zi,which changes across firms, stays constant over time. This proves that both ways of writing down the FERM are equivalent.

As stated, biases could also arise from so-called time-fixed variables, omitted variables that are constant across entities but evolve over time (e.g. laws). The time fixed model with a single regressor can be written as follows:

Y_it = β₁ · X_it+ λ_t+ u_it Equation (13)

For t = 1, … , T and i = 1, … , N.

Xit is the value of the regressor for company i in time in period t and λ1, … , λt are

(24)

Similar to what was done in Equation (12), binary indicators can be generated to denote time fixed effects. Adjusting the above equation results in:

Y_it= β₀+ β₁· X_it+ Ϭ₂B2_t+ … + Ϭ_TBT_𝑡+ u_it Equation (14) For t = 1, … , T and i = 1, … , N.

X_it is the value of the regressor for company i in time in period t and λ1, … , λt are

time fixed effects. B2t= 1 if t = 1 and B2_t= 0 otherwise etc.

Endogeneity caused by variables that is constant across entities but evolves over time has now been eliminated from the equation.

In this study, both firm and time fixed variables included in the models. Combining Equation (12) with Equation (14) results in the following equation:

Y_it= β₀+ β₁X_it+ ϒ₂D2_i + … + ϒ_nD2n_i + Ϭ₂B2_t+ … + Ϭ_TBT_t + u_it Equation (15) For t = 1, … , T and i = 1, … , N

X_1,it is the value of the first regressor for company i in time in period t, X2,it the value

of the second regressor etc.

Let D2i equal 1 when i = 1 and otherwise zero, etc.

B2t= 1 if t = 1 and B2t= 0 otherwise etc. By having eliminated both sources of endogeneity from the model, simple OLS-regression would result in an estimation of β1, however with wrong standard errors. These standard

errors can be calculated correctly by making use of the xtreg-command in Stata. So, if a substantial part of the potential endogeneity between gender diversity and firm performance would be caused by both firm and time-fixed variables, the chance of obtaining an unbiased and consistent estimator has significantly increased.

It should be pointed out that instead of using FERM, the random effects regression model model (RERM) could also have been used in order to conduct a panel data examination. However, the RERM assumes no correlation exists between the error term and variables of interest, i.e. no fixed effects. By investigating whether correlation exists, it could be concluded whether the FERM is indeed the appropriate model to apply. So-called Hausman tests were conducted for each regression, which in each case resulted in p-values, implying that the null-hypothesis (H0 = corr(𝑋𝑖𝑡, 𝑢𝑖𝑡)=0)

(25)

factors which confirms the importance of addressing this type of endogeneity of diversity in the performance regressions. Accordingly, the FERM is conducted for each of the six models.

The fixed effect model presented in Equation (15) is a simplified version of the model that will be used in this paper, since several explanatory variables are taken into account. This paper examines not only the effect of a gender diverse board on firm performance, but also the effects of other independent, control and interaction variables are taken into account. These other

determinants of firm performance can simply be added to our model by extending equation (15) to a fixed effects regression model with multiple regressors, which is shown below:

Yit = β0+ β1· Xit+ … + βk · Xk,it+ ϒ2D2i+ … + ϒn Dni+ Ϭ2B2t+ …

+ ϬtBnt + uit

Equation (16) For t = 1, … , T and i = 1, … , N Applying the variables to the above fixed effects regression with multiple regressors provides:

Model(1)

Tobin′_{s Q = β}

0+ β1· GENDER DIVERSITYit+ β2· EDUCATIONit

+ β₃· INDEPENDENCE_it+ β₄· FIRM SIZE_it+ β₅· LEVERAGE_it

+ β₆ · Ln(SALES)_it+ ϒ₂D2_i+ … + ϒn Dn_i+ Ϭ2B2_t+ … + Ϭ_tBn_t + εit For t = 1, … , T and i = 1, … , N

Model (2)

RoE_it= β₀+ β₁· GENDER DIVERSITY_it+ β₂· EDUCATION_it

+ β₃· INDEPENDENCE_it + β₄· FIRM SIZE_it+ β₅· LEVERAGE_it

+ β6 · Ln(SALES)it+ ϒ2D2i+ … + ϒn Dni+ Ϭ2B2t+ … + ϬtBnt + εit For t = 1, … , T and i = 1, … , N

Model (3)

RoA_it= β₀+ β₁· GENDER DIVERSITY_it+ β₂· EDUCATION_it

+ β₃· INDEPENDENCE_it+ β₄· FIRM SIZE_it+ β₅· LEVERAGE_it

+ β₆ · Ln(SALES)_it+ ϒ₂D2_i+ … + ϒn Dn_i+ Ϭ2B2_t+ … + Ϭ_tBn_t + ε_it For t = 1, … , T and i = 1, … , N

Model (1)-(3) will be all used to test hypothesis 1, (H1).

In order to test hypotheses 2 & 3, (H1) & (H2), the 3 formulas above are slightly modified in

order to examine if education and independence moderate the effect female representation has on firm financial performance.

(26)

As can be seen above the moderating effects are taken into account by adding 2 extra variables of interest called multiplication terms to model (1)-(3).

Model (4) Tobin′_{s Q}

it = β0+ β1· DIVERSITYit+ β2· EDUCATIONit + β3· INDEPENDENCEit + β₄· FIRM SIZE_it+ β₅· LEVERAGE_it+ β₆ · Ln(SALES)_it

+ β₇· DIVERSITY · EDUCATION_it+ β₈· DIVERSITY · INDEPENDENCEit + ϒ₂D2_i+ … + ϒn Dn_i+ Ϭ2B2_t+ … + Ϭ_tBn_t + ε_it

For t = 1, … , T and i = 1, … , N Model (5)

RoE_it = β₀+ β₁· DIVERSITY_it+ β₂· EDUCATION_it + β₃· INDEPENDENCE_it + β₄· FIRM SIZE_it+ β₅· LEVERAGE_it+ β₆ · Ln(SALES)_it

For t = 1, … , T and i = 1, … , N Model (6)

RoA_it = β₀+ β₁· DIVERSITY_it+ β₂· EDUCATION_it + β₃· INDEPENDENCE_it + β₄· FIRM SIZE_it+ β₅· LEVERAGE_it+ β₆ · Ln(SALES)_it

For t=1,…,T and i=1,...,N Model (4) – (6) will be all used to test hypothesis 2-3, (H2) & (H3).

Before performing a regression on the aforementioned models, a robustness test should be conducted to check for multicollinearity. Multicollinearity relates to the level of non-independence of independent variables and can pose problems for the estimation of coefficients due to inflation of variance of parameter estimates (Dormann et al., 2013). This could result in unstable and

unidentifiable estimation results. These results could lead to misinterpretation of relevant predictors in a statistical model. Especially when interaction terms are added to the model, multicollinearity is more likely to become a problem due to high correlation between the new interaction term and the two variables used to calculate it. Dormann et al. (2013) describe that the most commonly applied method to test for multicollinearity is making use of the Pearson coefficient: if two independent variable show to have a higher correlation coefficient than 0.7(|r| <0.7), one of them should be dropped from the regression. In table 5 in the appendix, the correlation matrix shows that none of

(27)

the independent variables mentioned above are this highly correlated. Thus, all of the aforementioned independent variables will be used in the regression.

The examined relationship between gender diversity and firm performance will be robust to multicollinearity and endogeneity arising from fixed effects.

3.4 Data & sample

The purpose of this paper is to examine the effect of board gender diversity on firm performance using a sample of Dutch NV companies listed on the Amsterdam Euronext . The relationship between gender diversity and firm financial performance is examined over a period from 2008 to 2014. The examination of a different timeframe than a very recent one is not an option due to the extreme lack of female board members before 2008.

Following Marinova et al. (2015), banks, soccer clubs and insurance companies are excluded due to their specific accounting methods, which could pose difficulties for the calculation of Tobin’s Q. Also listed investment funds are excluded because of the special nature and management of these companies (Lückerath-Rovers, 2011). Additionally, companies with missing data, companies that were taken over by foreign corporations and companies that changed their constituent country are excluded. Data of sufficient quality was eventually available for 83 companies. Information and data on the identity and characteristics of the directors as well as market value and accounting data is obtained from the database ORBIS, provided by Bureau van Dijk, and Thomson Reuters Eikon. These online information platforms provide general and financial data of over 200 million public and private companies worldwide. If needed, data is also obtained from annual reports of the companies

examined. In case an annual report does not provide sufficient information, it is attempted to be found through indulging in business news sites such as Bloomberg, financial magazines and social business platforms, such as LinkedIn.

3.5 Summary statistics

Table 2 shows the number of observations, minimum, maximum, mean and standard deviation values for all variables used in the models described in section 3.1. for the overall sample with a time-span of 2008-2014. The overall average percentage of females on boards in the Netherlands is to be exact 9.86%. The minimum and maximum values are 0.00% and 42.86%. Besides descriptive statistics for the overall sample, the yearly descriptive statistics for gender diversity are given in table 2. The average percentage of females in boards of directors increased over the years, to be exact

(28)

from 5.7% to 15.0%. Also, it can be derived from the yearly mean that up to 2012 at least half of the examined companies had no female board members at all.

The firm financial performance measures are ROE, RoA and Tobin’s Q. Table 2 shows that the mean value of RoE is 8.89 with a standard deviation of 22.09. For RoA, respectively, 5.24 and 22.09 and for Tobin’s Q is 1.59 and 1.60. As shown, the accounting-based performance measures have relatively high standard deviations.

The average fraction of board directors holding a master degree is 0.62. As can also be seen, no boards could be observed with less than 24% of their member holding this type degree. A few observations even showed that certain board only contain members holding a master degree, which may indicate a policy that requires a person to have a master-degree in order to become a director.

In the board 56% of the total amount of directors are independent. A few of the companies observed showed to have no independent board members, which can be confirmed by the minimum value of 0.00%.

The firm size was measured as the natural logarithm of the net sales. The average overall measured firm size over a time-span of 2008-2014 was 12.22.

The average leverage ratio over the years was 0.56, indicating that on average more than half of the total assets of companies are financed with debt.

No large outliers were observed for the dependent, independent and control variables panel data set, therefore it was decided not to winsorize the data.

(29)

Table 2 presents the Descriptive Statistics of the Dutch sample examined, which contains 581 observations of 83 firms over the period of 2008-2014. The sample was drawn from the Amsterdam Euronext. Tobin’s Q is measured by dividing the total market value of a company by their replacement cost, RoE as the percentage of Net Income with respect to total equity and RoA as the percentage of net income with respect to total assets. GENDER DIVERSITY is measured as the fraction of female board directors, EDUCATION as the fraction of board directors holding a master degree. INDEPENDENCE is measured as the number of supervisory directors divided by the total amount of directors (non-executives and executives together), FIRM SIZE as the natural logarithm of the total assets of the firm, LEVERAGE as the amount of leverage, ln(SALES) as the natural logarithm of net sales.

Table 2: Descriptive statistics overall sample

N Minimum Maximum Mean Median Standard

deviation Tobin’s Q 581 0.02 10.35 1.59 0.86 1.60 RoE (%) 581 -130.01 89.98 14.89 16.88 22.09 RoA (%) 581 -58.99 61.67 5.24 7.33 11.34 Diversity 581 0.00 0.43 0.10 0.00 0.10 Education 581 0.24 1.00 0.62 0.66 0.18 Independence 581 0.00 0.84 0.56 0.48 0.19 Firm size 581 2.40 21.22 12.22 11.97 2.81 Leverage 581 0.02 0.87 0.56 0.55 0.21 Ln(Sales) 581 -0.85 4.80 0.22 0.19 0.56

Table 3 presents the yearly Descriptive Statistics, regarding gender diversity, of the Dutch examined sample, which contains 581 observations of 83 firms over the period of 2008-2014. Again, GENDER DIVERSITY is measured as the fraction of women on the board

Table 3: Descriptive statistics gender diversity per year

N Minimum Maximum Mean Median Standard

deviation 2008 83 0.00 0.36 0.06 0.00 0.06 2009 83 0.00 0.31 0.07 0.00 0.09 2010 83 0.00 0.33 0.08 0.00 0.10 2011 83 0.00 0.38 0.09 0.00 0.08 2012 83 0.00 0.38 0.10 0.00 0.12 2013 83 0.00 0.43 0.14 0.14 0.09 2014 83 0.00 0.39 0.16 0.17 0.19

(30)

4 Results

In this chapter, in paragraph 4.1 the regression results of models (1), (2) and (3) will be discussed. Thereafter, the empirical findings regarding models (4), (5) and (6) are covered in paragraph 4.2. The obtained effects on each of the three firm financial performances (Tobin’s Q, RoE and RoA) are separately presented in subparagraphs. Coefficient estimates for models (1)-(6) are presented in table 4.

4.1 Empirical findings regarding gender diversity, independence and education on firm

financial performance

4.1.1 Tobin’s Q

Firstly, the effect of the independent variables: a positive significant effect of gender diversity on firm financial performance was found (β= 1.837, p-value= 0.000), education has an insignificant positive impact (β= 0.103, p-value= 0.586) and the effect of board independence found was also insignificant, however negative (β= -0.099, p-value= 0.483). Regarding the control variables, for both firm size and sales growth moderate, respectively negative and positive effects were obtained (β= -0.378, p-value= 0.04; β= 0.203, p-value= 0.000). Leverage was shown to affect Tobin’s Q insignificantly and negatively (β= -0.398, p-value= 0.180).

4.1.2 RoE

Again gender diversity has a significant positive effect on firm financial performance (β=5.316, p-value= 0.023). Also education and independence were again found not to impact financial

performance significantly (β= 3.799, p-value= 0.540; β= 0.663, p-value= 0.630). Control variables firm size and sales growth both enhance firm performance significantly (β= 13.740, p-value= 0.009; β= 11.628, p-value= 0.001). Unlike in the case of Tobin’s Q, the obtained effect of firm size is positive. Using ROE as firm financial performance measure, leverage has a significantly and negatively impact (β= -34.881, p-value=0.004).

(31)

4.1.3 RoA

Unlike with Tobin’s Q and RoE, gender diversity was found to have no significant effect on RoA (β= 3.672, p-value= 0.140), which indicates that gender diversity does not influence firm financial performance. The other independent variables independence and education are again not

significant. (β= -0.980, p-value= 0.780; β= 1.462, p-value= 0.721). Estimates for the control variables firm size, leverage and sales growth are again all significant (β= 6.423, value= 0.002; β= -36.845, p-value= 0.009; β= 4.766, p-p-value= 0.000).

Given the empirical evidence that the percentage of female directors in boards has a positive and significant effect on Tobin’s Q and RoE and an insignificantly positive effect on RoA. These findings might provide support that gender diversity on boards of directors in Dutch listed firms has a positive effect on firm financial performance. However, although Durbin-Wu-Hausman tests showed that a significant level of endogeneity was caused by fixed effects, the results are not robust against all sources of endogeneity. Even though the overall level of endogeneity is reduced after applying the FERM, the affect variables have cannot be assumed to be exogenous because additional endogeneity problems could also arise because of the problem of reverse causation, which is not controlled for in this study. Therefore, conclusions drawn based on model (1)-(3) do not suggest any causality and thus it should be considered premature to use the found positive impact as a reason to appoint more women to boards.

4.2 Empirical findings regarding the effects of interactions between gender diversity, board

independence and education on firm financial performance

4.2.1 Tobin’s Q

Both of the interaction terms Gender Diversity X Education and Gender Diversity X Independence show insignificant effects on Tobin’s Q (β= 0.977, p-value= 0.4335; β= 0.253, p-value= 0.4899). The same holds for the individual coefficients of education and independence (β= 0.217, p-value= 0.796; β= -0.144, p-value= 0.836). Additionally, the effect of gender diversity on this financial performance measure is still positive and significant (β= 1.257 p-value= 0.044). Again for sales growth and firm size significant coefficients were found (β= 0.196, p-value= 0.000; β= -0.411, p-value= 0.000), while leverage was, just like in model (1), observed to have an insignificant effect (β= -0.401, p-value= 0.128). Additionally, the findings presented in table 4 suggest that, in terms of the adjusted R2, model 4 is the preferred model.

The impact of gender diverse boards on firm financial performance in the Netherlands