• No results found

The impact of board gender diversity on firm performance in the Netherlands

N/A
N/A
Protected

Academic year: 2021

Share "The impact of board gender diversity on firm performance in the Netherlands"

Copied!
47
0
0

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

Hele tekst

(1)

MSc Business Economics

Track: Finance

Master Thesis

The impact of board gender diversity on firm performance

in the Netherlands

Name:

Marco Scholman

Student number: 10384324

Supervisor:

Prof. Dr. A. W. A. Boot

Date:

07-07-2016

(2)

Statement of Originality

This document is written by Student Marco Scholman who declares to take full responsibility for the contents of this document. I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

ABSTRACT

This empirical study investigates the effect of board gender diversity on firm financial performance. A sample of Dutch listed and non-listed firms for the period 2009-2014 is used to examine this effect. The system general method of moments (SYS-GMM) estimator is applied to address the econometric problems of reverse causality and constant unobserved heterogeneity. Firm performance is measured as the return on assets (ROA). In addition, the return on equity (ROE) is used to check whether the results are robust. Based on the evidence in this study, board gender diversity does not significantly influence firm performance when the entire sample is used. This is consistent with the social psychological and contingency theory. The sample is also divided into a subsample of listed and a subsample of non-listed firms to test whether the effect of board gender diversity on firm performance is the same for these two types of firms. It is hypothesized that board gender diversity more positively influences firm performance for the latter subsample. The results indicate that this is indeed the case. Increasing board gender diversity improves firm performance for non-listed firms, whereas this effect is insignificant for listed firms. This may indicate that listed firms are more influenced by the societal pressure to increase board gender diversity, which could harm firm performance. However, this result depends on the performance measure used and should therefore be interpreted with caution.

(4)

TABLE OF CONTENTS

1. Introduction 5

2. Theoretical background 7

2.1 The resource dependence theory 7

2.2 The human capital theory 8

2.3 The agency theory 9

2.4 The social psychological theory 10

2.5 The contingency theory 11

2.6 The business case 11

2.6 Concluding remarks theories 12

3. Empirical evidence board gender diversity and firm performance 14

3.1 International empirical evidence 14

3.2 Dutch empirical evidence 16

3.3 Key insights empirical evidence 16

4. Hypotheses 19

5. Data 20

5.1 Data sample 20

5.2 Variables 21

5.2.1 Dependent variable of interest 21

5.2.2 Independent variable of interest 22

5.2.3 Control variables 22

5.3 Descriptive statistics 24

6. Methodology 29

6.1 SYS-GMM estimator to address econometric problems 29

7. Main results 32

7.1 Two-step SYS-GMM board gender diversity and ROA entire sample 32 7.2 Two-step SYS-GMM board gender diversity and ROA listed and unlisted firms 34

8. Robustness checks 36

8.1 Two-step SYS-GMM board gender diversity and ROE entire sample 36 8.2 Two-step SYS-GMM board gender diversity and ROE listed and unlisted firms 37

9. Conclusion 39

10. References 42

(5)

1. Introduction  

There is an increasing pressure around the world to appoint women as directors in order to increase gender diversity on corporate boards. Some countries have already taken certain measures to establish this (Adams & Ferreira, 2009). For example, in Norway a quota applies for a boardroom consisting of at least 40% women. In the Netherlands there only is a target quota of 30% (European Commission, 2012a). However, according to the European

Commission the improvement in board gender diversity is too slow. Therefore, they want to implement a European legislation for the largest listed firms to have 40% of the board-member positions occupied by women in 2020.

Two claims are primarily used for justifying an increase in gender diversity. Firstly, groups that were historically excluded from board positions will have an equal opportunity when diversity is provided. Nowadays, there is a strong incentive for ensuring that

opportunities are available to all labour market participants who are able to fulfil the job (Packel & Rhode, 2014). Based on this claim, there may exits a negative relation between board gender diversity and firm performance since increasing this diversity is driven by the societal pressure for greater gender equality (Ahern & Dittmar, 2012). Secondly, it is argued that the presence of female directors improves firm financial performance if they create a more effective decision-making process by adding a different perception. The idea that there exists a positive relation between board gender diversity and firm performance is also the main motivation of the European Commission for implementing the above-mentioned

legislation and is referred to as the “business case” (European Commission, 2012b). This case does not imply that women are substitutes for male directors. Instead, it indicates that women may possess unique attributes that could be beneficial for board performance and in turn firm performance.

Despite the business case claim for a positive relation between board gender diversity and firm performance, empirical evidence regarding the link is mixed. While some studies do indeed find a positive relation, others report a negative or insignificant relation. The different results could be due to the use of a different sample period, performance measure and country as indicated by Campbell and Vera (2008). This is in line with a contingency theory

perspective. The theory suggests that certain board diversity aspects may only be desirable for some firms, under certain circumstances and under certain time periods (Fiedler, 1967;

Lawrence & Lorsch, 1967). However, investigating the relation between board gender diversity and firm performance is still important, since governments around the world are

(6)

forcing firms to increase this diversity. Therefore, this paper will investigate whether the business case can be justified on the Dutch market. The main research question can be stated as follows:

Is increasing board gender diversity associated with a better firm performance in the Netherlands?

An empirical investigation will be conducted to answer this question. To my

knowledge, all studies regarding this subject have only focussed on public quoted companies. This research therefore distinguishes itself by using a sample of non-listed Dutch firms in addition to a sample of listed firms. The total sample size is equal to 120 and the sample period is from 2009 until 2014. In first instance the whole sample will be investigated. After that, the sample will be split up in listed and non-listed companies in order to conclude if there exists a different effect. A balanced panel regression is used to measure the effect of board gender diversity on the firm performance. Different specifications of the gender

diversity measure will be applied in performing this regression. The return on assets (ROA) is used as a performance measure. To check if the results are robust the return on equity (ROE) will also be used as a performance measure.

Regressions regarding board composition are exposed to endogeneity. According to Hermalin and Weisbach (2003) it is more likely that better performing firms have a more gender-diversified workforce than worse performing firms. This might result in more boardroom positions occupied by women within better performing firms. In other words, reverse causation might be a problem. Another econometric problem that might arise with performing a regression is constant unobserved heterogeneity. To solve for these two problems the system general methods of moments (SYS-GMM) estimator will be used.

The remainder of this thesis will be organized as follows. Section 2 will describe the different theories regarding the relation between board (gender) diversity and firm

performance. Empirical evidence with respect to this relationship will be discussed in the subsequent section. Section 4 describes the formulation of the hypotheses. The next two sections will respectively discuss the data and the methodology that is used for this

investigation. The main results will be presented in seventh section. A different specification for the regression is used in the subsequent section to conclude if the main results found are robust. A conclusion is formulated in the last section.

(7)

2. Theoretical background

Two types of arguments are usually used to justify the increase of board gender diversity: (1) ethical arguments and (2) economic arguments (Walt & Ingley, 2003). According to the former board diversity is a desirable end in itself. It is argued that excluding groups on the basis of gender, race or other demographic characteristics is unjustifiable (Singh et al., 2001). Previous ethical studies indicate that the underrepresentation of certain groups calls for ethical considerations in order to ensure that all qualified market participants for a certain job have equal opportunities. These studies suggest that with an increase in board diversity a more equitable outcome for society is achieved (Carver, 2002;Keasey et al., 1997) and that it could be economically beneficial if assuming that board performance is influenced by the board’s composition, which in turn partially influences firm performance (Dalton et al., 1999;Kiel & Nicholson, 2003). Economic arguments have also arisen from the fact that systematically excluding certain market participants that are able to fulfil the job could harm firm financial performance (Burke, 2000;Carver, 2002).

There are different theories with underlying motivations for what the economic benefits of board gender diversity are. It is difficult to test these motivations since there does not exist a comprehensive framework in order to measure the variables behind them (Fields & Keys, 2003). The different theories will all be presented below and include: the resource dependence theory, human capital theory, agency theory, social psychological theory and the contingency theory. In addition, the business case view will be discussed, which is referred to as “commercial reasons” by Walt & Ingley (2003). It is important to bear in mind that the theories and the business case view focus on the diversity as a whole, and not specifically on gender diversity. However, it is possible to apply these theories on gender diversity.

2.1 The resource dependence theory

According to Pfeffer and Salancik (1978) the board plays a key role in establishing a link between the company and other external organizations in order to address environmental dependencies. Therefore, the board can be seen as an essential strategic resource for a certain company. Pfeffer and Salancik (1978) provide four possible benefits resulting from the external linkages: (1) obtaining valuable resources such as information and expertise, (2) obtaining new channels of communication with business relations that are important to the

(8)

firm, (3) support or commitments from important external organizations and (4) the establishment of legitimacy for the company in the external environment.

Hillman et al. (2000) have expanded the resource dependency theory and have translated these four benefits into different types of directors who could provide the firm various resources. These are insiders, business experts, support specialists, and community influentials. Insiders are board directors who are or were actively involved in the firm in the past. They provide expertise regarding the firm itself. Business directors provide expertise in decision-making, problem solving and competition for large companies. They can be

compared with CEOs or senior managers within large, for-profit companies. Support specialists provide expertise in banking, law, marketing, and access to important resources including legal support. Lastly, community influentials provide the firm resources regarding non-business perspectives on problems, issues and ideas. Furthermore, they deliver expertise about influential groups within the community. Influentials can be compared with community or social leaders such as politicians (Hillman et al., 2000).

Hillman et al. (2000) argue that having different types of directors will lead to different available resources that will be beneficial to the firm. In essence, they indicate that more diversity within corporate boards will create more valuable resources. This ultimately promotes firm performance.

2.2 The human capital theory

The human capital theory has some complementary concepts regarding the role of board diversity derived from the resource dependence theory. According to Terjesen et al. (2009) the human capital theory is derived from a study performed by Becker (1964). This theory tries to explain what the role of a person’s capability is in terms of experience, education, and skills that can be used to positively influence the company. Moreover, gender diversity among directors ensures unique human capital (Terjesen et al., 2009).

Given that women do have unique human capital, the question is if women lack the “right” human capital for a director position? Empirical evidence provided by Terjesen et al. (2009) shows that women have equal qualifications as men in terms of qualities such as the education level. However, an equal amount of experience as business experts is less likely for women. It appears to be that female directors perform different roles on the board relative to male directors. This is possibly caused by their unique human capital (Hillman et al., 2002). Men and female directors could therefore complement each other. As a result, the human

(9)

capital theory suggests that board diversity influences board performance and in turn firm performance due to diverse and unique capital. Although, the human capital theory does not specifically predict a positive effect of board diversity on corporate performance it is highly suggestive of such an effect (Carter et al., 2010).

2.3 The agency theory

The agency theory, provided by Jensen and Meckling (1979), is often used as a theoretical framework for explaining the relation between board characteristics and firm performance (Carter et al., 2003). The theory explains that there exists a conflict of interest between managers and shareholders, and that increasing the board independency could reduce this conflict. This in turn will positively influence firm performance. As a result, the theory proposes that the board must play a key role in controlling and monitoring the managers. Therefore, the influence of board diversity on board independency should be investigated. Or differently formulated, does increasing the diversity within corporate boards promote the independency of the board (Carter et al., 2003)?

One argument in favour of this question is that more diversity could lead to more different ways of thinking within boardrooms in comparison with homogeneous boards. As a result, this may create a more activist board because outside directors with non-traditional characteristics could be seen as the ultimate outsider (Carter et al., 2003). However, Carter et al. (2003) also point out that a different perspective may not necessarily create more effective monitoring. It may be the case that diverse board members are marginalized. Given that the diverse director only has a marginal influence, it is expected that this director does not have an incentive to build reputation as an expert monitor. Moreover, according to Hermalin and Weisbach (2000) the principal agent theory indeed gives many insights, but it is lacking of explaining board-specific phenomena.

Empirical evidence provided by Adams and Ferreira (2009) suggested that the independency of female directors is greater than for male directors. However, they mention that independence is not specifically a female trait. It could also be possible that other

unobserved factors are the driven force behind the observed independence of female directors. For example, the difference in social and business networks of male and female board

(10)

Assuming that female directors are indeed more independent than male directors, Adams and Ferreira (2007) argue that independence comes with costs and benefits. Therefore, one could not expect that increasing the presence of women on the board and thus the board independence will promote firm performance. Previous literature has also shown that the effect of board independence on firm performance is inconclusive (Adams et al., 2010). Therefore, based on the discussion above, it can be derived that the agency theory does not provide a clear link between board diversity and firm performance.

2.4 The social psychological theory

According to Westphal and Milton (2000) diversity reduces the social cohesion between groups. In addition, it will be less likely that the perspectives of minorities (such as women) have a significant influence on group decisions if social barriers are present. This social psychological concept is derived from the social impact theory (Westphal & Milton, 2000). According to this theory, individuals with a majority status are able to more heavily influence group decisions. Evidence has shown that these individuals often resist the influence of minority group members (e.g. female directors). As a result, diverse directors will not have or only have a little impact on the decisions within boardrooms.

Westphal and Milton (2000) have also argued that minorities may improve the decision-making process due to stimulating a different way of thinking. This in turn could positively influence firm performance. However, Campbell and Mínguez (2008) indicate that board diversity results in a less efficient decision-making process due to a greater possibility of the occurrence of a conflict between board members. Empirical evidence suggests that diversity may stimulate both, divergent thinking and group conflict (Williams & O’Reilly, 1998). Forbes and Milliken (1999) reviewed the evidence form a number of studies and concluded that the effectiveness of the board probably significantly depends on social psychological related processes. They further argue that it is likely that each demographic board characteristic has many complex and opposing effects on the processes that influence the performance of the board and in turn firm performance.

In short, the social psychological gives an inconclusive answer for what the expected effect of diversity in the boardroom on firm performance is since it may exerts both, positive and negative effects.

(11)

2.5 The contingency theory

Carter et al. (2010) argued that board gender diversity could positively or negatively influence firm performance as derived from the contingency theory. It could even be possible that there is no influence at all. According to the contingency theory the optimal course of action should be consistent with the firm’s external and internal circumstances. Certain board diversity aspects (such as board gender diversity) may only be desirable for some firms, under certain circumstances and under certain time periods (Fiedler, 1967; Lawrence & Lorsch, 1967). In line with this, Aguilera et al. (2008) argued that firm characteristics (e.g. firm size, firm age and the growth phases) might have an influence on the effectiveness of changing certain board characteristics (e.g. increasing board gender diversity). For example, they indicate that board monitoring may be of less importance than the knowledge and resource contributions of board directors in the first stages of the firm. However, as the firm matures and grows there is a greater need for external resources. An expansion of current board member networks and contacts may be established by increasing board gender diversity and helps to establish a better link between the company and other external organizations (Hillman et al., 2000). Thus the greater need for external resources may make it more likely that a firm increases board gender diversity (Aguilera et al., 2008).

2.6 The business case

According to the business case there exists a positive relation between board diversity and firm performance. It is argued that diversity does have an effect on the short- and long-term performance based on several propositions. However, these propositions are not derived from any theoretical framework. Nevertheless, Robinson and Dechant (1997) provided some empirical evidence together with intuitive examples to support each proposition.

The first proposition is that marketplace understanding will be improved with board diversity. The diversity of the firm will be more properly matched with the potentially diverse costumers and suppliers, since demographic projections show that diversity within the

marketplace is increasing. This will cause a better market penetration. The second proposition states that board diversity positively influences creativity and innovation. According to Robinson and Dechant (1997) attitudes, cognitive functioning and beliefs do not follow a random distribution in the population. Instead, it appears to be that they vary systematically with demographic variables (e.g. gender, age and race). Thirdly, as already mentioned in

(12)

subsection 2.4, diverse board members may promote an effective decision-making process. In the short run, the process may become less effective due to a greater likelihood of a conflict, whereby reaching consensus will take more time. However, diversity creates more divergent perspectives. This gives decision makers more alternatives to choose from and a more careful evaluation of the consequences for these alternatives (Robinson & Dechant, 1997). The fourth proposition indicates that board diversity generates more effective corporate leadership. Heterogeneity within the boardroom avoids that there is a too narrow perspective, while diversity stimulates a broader view. This results in a board that will be more able to deal with the environmental complexities (Robinson & Dechant, 1997). Lastly, diversity increases the chance that more effective global relationships will be established. Firms operating in an international environment are especially sensitive to cultural differences. Heterogeneous boards could reduce this sensitivity (Robinson & Dechant, 1997). In contrast with the previous propositions, there are fewer linkages to gender diversity with this proposition.

2.6 Concluding remarks theories

Section 2 provided five different theories and a non-theoretical perspective (the business case) regarding the effect of board (gender) diversity on firm performance. Table 1 summarizes this effect based on these (non-) theoretical perspectives. The effect is positive according to the resource dependency theory, human capital theory and the business case. The social

psychological theory argues that board diversity does not influence firm performance since there are offsetting effects. According to the contingency theory, board diversity may only be the best possible action under certain external and internal circumstances. The agency theory does not predict a clear link for the relation between board diversity and firm performance. In the next section, empirical evidence with respect to this relation will be discussed.

(13)

Table 1: The expected effect of board diversity on firm performance

Theories Effect of diversity on firm performance Explanation of effect Resource dependency theory

Positive Appointing different types of board directors will create more valuable resources, which increases firm performance.

Human capital theory

Positive Male and female directors perform different roles on the board due to their unique human capital. It is highly suggestive that this diverse and unique capital will positively affect firm financial performance. Agency theory No clear link The effect of board diversity on board independency is not clear.

Moreover, board independence comes with costs and benefits. Empirical evidence regarding the effect of independence on firm performance is also inconclusive. Social psychological theory No significant effect

Diverse board members do not have or only have a little influence on board decisions. Therefore, this argument suggests that board diversity has an insignificant effect on firm performance. Moreover, it is argued that board diversity may have both, a positive and negative effect on firm performance. Positive because diversity stimulates divergent thinking and negative because diversity increases the possibility of the occurrence of a conflict between board members.

Contingency theory

No significant effect

Diverse board members may have a positive, negative or insignificant effect on firm performance. The optimal course of action to undertake depends on the external and internal circumstances.

Business case1 Positive Positive effect is based on several propositions, which are not derived from a theoretical framework. Empirical evidence and intuitive examples are given to support each proposition.

                                                                                                               

1  As already indicated, the business case is a non-theoretical perspective. However it is important to discuss

(14)

3. Empirical evidence board gender diversity and firm performance

Empirical evidence regarding the effect of board gender diversity on corporate performance is mixed: some studies provide a positive effect, while others find a negative or insignificant link. An overview of previous studies is displayed in table 2. Some of these studies will be discussed below. Different methods have been used to measure the effect of gender diversity on firm performance. In most studies the dependent variable is a firm financial performance measure (e.g. Tobin’s Q, ROA), and the independent variable is a measure for the gender diversity. Gender diversity is usually measured by the percentage of women on corporate boards or a dummy variable that equals one if a certain amount of women is present in the boardroom.

3.1 International empirical evidence

Carter et al. (2003) were one of the first researches that empirically tested if there is a direct link between board gender diversity and firm performance. They have taken a sample of public US firms, used the Tobin’s Q as performance measure and the percentage of female directors as measure for board gender diversity. A two stage least squares (2SLS) regression was conducted to control for the reverse causality problem. They find that greater board gender diversity is significantly associated with greater firm performance. One main

limitation of the research by Carter et al. (2003) is that they only have used one year (1997) as sample period. Therefore, controlling for constant unobserved heterogeneity (a problem related to omitted variable bias) was not possible. In contrast, Adams and Ferreira (2009) used a dataset of public US firms over a longer period (1998-2002), which made it possible to control for this heterogeneity. They have used the same measures for firm performance and board gender diversity. In addition, they have also performed a 2SLS regression to address reverse causality and find the opposite result of Carter et al. (2003). This may indicate that the positive relation between board gender diversity and firm performance is driven by omitted time- and firm-invariant factors.

Evidence from European studies is also inconclusive. Campbell and Vera (2008) analysed a sample of listed firms in Spain for the period 1995 to 2000. They have controlled for both, constant unobserved heterogeneity and reverse causality. The latter is controlled for with performing an instrumental variable regression. They conclude that a gender-diversified board does have a positive impact on firm performance (as measured by the Tobin’s Q).

(15)

In contrast, Bøhren and Strøm (2007) used a panel data sample of Norwegian listed firms and concluded that board gender diversity is negatively associated with corporate performance. They have used the same measures for board gender diversity and firm financial performance as Campbell and Vera (2008). The use of a panel dataset made it possible to control for constant unobserved heterogeneity. The generalized method of moments (GMM) estimation technique was applied instead of a 2SLS regression to address the reverse causality problem. It is important to bear in mind that the results found in one country may not be generalized. Different countries are exposed to different factors that might have an influence on board (gender) diversity. One should thus always be cautious when comparing the results of studies conducted in different countries, as is the case with the two studies described above (Carter et al., 2010).

The implementation of the binding quota in Norway, which requires companies to have 40 percent of the board member positions occupied by women, is often seen as the “Holy Grail” from econometric view. Ahern and Dittmar (2012) and Matsa and Miller (2013) use the binding quota to perform a natural experiment. The quota allows them to treat the increase in board gender diversity as an exogeneous event. The conclusion from both papers is the same: implementation of the binding board gender quota influenced performance negatively. According to Ferreira (2015) this conclusion seems plausible. It is likely that forcing firms to comply with a certain regulation, which they did not met before comes with costs. The firms may have been forced to hire insufficiently qualified directors to comply with the quota. Therefore, the benefits from an increase in board gender diversity would only be fully realised in the long run, when there is an increase in the supply of qualified directors. Although Ferreira believes that the conclusion is plausible, he is concerned about the

methodology used to reach the conclusion. He poses five difficulties when using the binding quota of Norway as natural experiment to investigate the relation between board gender diversity and firm performance. For example, one difficulty is related to the timing. There is a lot of freedom in defining the exact date of the quota shock (Ferreira, 2015). Ahern and Dittmar (2012)use 2003 as the shock date whereas Matsa and Miller (2013) use 2006 as the event date.

(16)

3.2 Dutch empirical evidence

Relatively little studies regarding gender diversity and firm performance have focused on the Dutch market. Lückerath-Rovers (2011) and Marinova et al. (2016) conducted such studies. Their results and methodologies will be discussed below.

Firstly, Lückerath-Rovers (2011) used a sample consisting of 99 Dutch listed firms on the Amsterdam Euronext stock exchange over the period 2005-2007. The main result was that firm performance (as measured by the ROE) is significantly higher when the board consists of both, men and women. This result may support the idea that gender diversified boards have better connections with the relevant shareholders at all levels of the company, which in turn improves the company’s reputation. This can be derived from the resource dependency theory. However, it should be noted that the result is insignificant when the return on sales (ROS) or the return on invested capital (ROIC) are used as performance measures. Moreover the main limitation of this study is that it does not take into account the problem of reverse causality. It could be the case that board gender diversity is caused by firm performance.

In contrast, Marinova et al. (2016) do take account of this problem by applying an instrumental variable regression. The percentage of women per industry is used as an instrumental variable. However, it should be mentioned that the used instrument is weak. Using a sample of 102 Dutch listed firms on the Euronext Amsterdam in 2007 they do not find a significant relation between board gender diversity and firm performance. This study is limited by the fact that it only uses one year (2007) for the empirical analyses. Therefore, the econometric problem of constant unobservable heterogeneity cannot be controlled. This was also the case with the research by Carter et al. (2003), mentioned in the previous subsection.

3.3 Key insights empirical evidence

As can be concluded from the previous two subsections and table 2, the empirical evidence regarding the effect of board gender diversity and firm performance is inconclusive. For example, Carter et al. (2003) have found a positive relation between board gender diversity and firm performance whereas Adams and Ferreira (2009) have found a negative result. They both used a sample of US firms, the Tobin’s Q as measure for firm performance and the percentage of women on the board as measure for board gender diversity. As already mentioned in subsection 3.1, one main difference is that Adams and Ferreira (2009) used a greater sample period, which made it possible to control for constant unobserved

(17)

heterogeneity. European studies (including the ones conducted on the Netherlands) did also not find a conclusive result regarding the relation between board gender diversity and firm performance. Campbell and Vera (2008) conducted a study with respect to this relation for Spanish firms and found a positive effect after controlling for constant unobserved

heterogeneity and reverse causality. Bøhren and Strøm (2007) also controlled for these econometric problems and found the opposite result of Campbell and Vera (2008).

One of the main differences was the technique they have used in both studies to control for the problem of reverse causality. Bøhren and Strøm (2007) applied the GMM technique, whereas Campbell and Vera (2008) have used a 2SLS regression to address this problem.

In short, based on the evidence above there is no support for the introduction of a board gender quota, neither is there support for a case against such a quota. This thesis will add evidence to the limited studies regarding the direct relation between board gender diversity and corporate performance. The econometric problems of reverse causality and constant unobserved heterogeneity will be addressed by using the system GMM (SYS-GMM) estimator.The methodology, including the way the econometric problems will be solved are discussed in more detail in section 6. The main contribution of this thesis is the use of a sample of non-listed Dutch firms in addition to a sample of Dutch listed firms.

(18)

Table 2: Empirical results regarding the effect of gender diversity on firm performance

Author(s),

Year investigation period Sample, country, Measure for gender diversity Performance measure Main result

Shrader et al.

1997 200 public firms, US, 1992 Percentage of women ROA, ROE, ROI, ROS Negative effect Carter et al.,

2003 638 public firms, US, 1997 dummy (one if women on Percentage of women, board ≥ 1)

Tobin’s Q Positive effect

Erhardt et al. 2003

112 public firms, US, 1998

Percentage of women ROA, ROI Positive effect

Randoy et al.,

2006 Scandinavia, 2005 459 public firms, Percentage of women market value ROA, stock Insignificant result Bøhren &

Strøm, 2007 Norway, 1989-2002 >129 public firms, Percentage of women Tobin’s Q Negative effect Rose, 2007 >100 public firms,

Denmark, 1998-2002

Percentage of women Tobin’s Q Insignificant result

Campbell & Vera, 2008

68 public firms, Spain, 1995-2000

Percentage of women, dummy (one if women on board ≥ 1), Blau-,

Shannon-index

Tobin’s Q Insignificant result for dummy, positive result for

other gender diversity measures Adams & Ferreira, 2009 1939 public firms, US, 1996-2003 Percentage of women, dummy (one if women on

board ≥ 1)

Tobin’s Q, ROA Negative effect

Miller &

Triana, 2009 326 public firms, US, 2003 Blau-index ROI, ROS Insignificant result Bohren &

Strom, 2010 Norway, 1989-2002 203 public firms, Percentage of women Tobin’s Q, ROA, ROS Negative effect Campbell &

Vera, 2010

68 public firms, Spain, 1995-2000

Percentage of women, dummy (one if women on

board ≥ 1)

Tobin’s Q Positive effect

He & Huang,

2011 530 public firms, US, 2001-2007 Blau-index ROA Negative effect Lindstaedt et

al., 2011 160 public firms, Germany, 2002-2010

Percentage of women ROA, ROE, price-to-book

value

Positive effect for firms with a high percentage of female

employees Lükerath-Rovers, 2011 99 public firms, Netherlands, 2005-2007

Percentage of women ROE, ROS, ROIC

Positive effect for ROE, insignificant result for ROS

and ROIC Ahern &

Dittmar, 2012 Norway, 2001-2009 248 public firms, Percentage of women Tobin’s Q Negative effect Mahadeo et

al., 2012 371 directors of 39 firms, Mauritius, 2007

Percentage of women ROA Positive effect

Smith et al., 2012

350 public firms, UK, 1996-2011

Percentage of women ROA, ROE, TSR Insignificant result Matsa &

Miller, 2013

104 public firms, Norway, 1999-2009

Percentage of women ROA Negative effect

Vafaei et al.,

2015 ±137 public firms, Australia, 2005-2011

Percentage of women, dummy (one if women on

board ≥ 1)

Tobin’s Q, ROA,

ROE, CFO/TA Positive effect

Marinova et al., 2016

102 public firms, Netherlands, 2007

Percentage of women, dummy (one if women on

board ≥ 1)

Tobin’s Q Insignificant result

Definitions of the abbreviations: return on assets (ROA); return on equity (ROE); return on investment (ROI); return on sales (ROS); return on invested capital (ROIC); total shareholder return (TSR); cash flow from operations (CFO); total assets (TA).

* Useful source for the table: Joecks, J., Pull, K., & Vetter, K. (2013). Gender diversity in the boardroom and firm performance: what exactly constitutes a ‘critical mass’?. Journal of Business Ethics, 118(1), pp. 61-72.

(19)

4. Hypotheses

The objective of this thesis is to investigate the impact of board gender diversity on firm performance. As already indicated, the resource dependency theory, human capital theory and the business case suggest that there exists a positive relation between board diversity and firm performance. The social psychological theory suggests that board gender diversity does not influence firm performance, the contingency theory suggest that diversity may only be optimal under unique circumstances, and the agency theory does not provide a clear link. In addition, the previous section showed there is no conclusive result based on the empirical evidence.

The business case is used by the European Commission as main motivation for greater gender diversity within boards (European Commission, 2012b). They even are considering the implementation of a binding board gender quota for the largest listed firms. Therefore, to extent the existing literature and add to the current European policy debate, this paper will test the following hypothesis:

H1: Increasing board gender diversity will positively influence firm performance.

The entire sample will also be divided into two subsamples: a subsample of listed and a subsample of non-listed firms. By doing this it can be examined if the influence of gender diversity on corporate boards is different for these two types of firms.

Listed firms often have a great number of (anonymous) shareholders whereas non-listed firms often have a few large shareholders. Therefore, non-listed firms are more in the spotlight with more attention from the media. Because of this, it seems plausible that board directors within listed firms face more pressure from the public opinion than non-listed firms (Klaassen & Rijken, 2010). As a result, it could be the case that listed firms experience a greater influence by the social pressure to increase board gender diversity than non-listed firms. According to Ahern and Dittmar (2012), increasing board gender diversity born out of this societal pressure could have a negative effect on firm performance. The second

hypothesis is therefore as follows:

H2: Increasing board gender diversity will more positively influence firm performance for non-listed than for listed firms.

(20)

5. Data

This section will firstly discuss how the data sample used for conducting the empirical

analysis is created. In the second subsection the variables (dependent, independent and control variables) will be explained. Lastly, descriptive statistics regarding the used sample will be discussed.

 

5.1 Data sample

A balanced panel data sample of Dutch firms is used to conduct the empirical analysis. The sample consists of 60 listed and 60 non-listed firms. The Dutch market is chosen because relatively little studies regarding the relation between board gender diversity and firm

performance have focussed on this market. In addition, data with respect to the representation of women on corporate boards for Dutch listed companies is easily assessable. Specifically, this data can be gathered from “The Dutch Female Board Index” provided by Lückerath-Rovers.2 The index is freely available and does only include Dutch NVs. This ensures that the analyses will not be influenced by foreign legal systems. The Dutch Female Board Index was first published in 2007. However, the effective sample period will be from 2009 until 2014 instead from 2007 until 2014 (2015 is excluded because annual reports for the fiscal year 2015 are not yet available for many non-listed firms). The reason for this is that lagged values of variables are used as instruments to perform the regression analysis. Further details

regarding the use of these lagged values will be discussed in the methodology section. “Company.info” is used to construct a sample of non-listed Dutch NVs. The sample consists of the 60 largest non-listed companies measured by the total revenue. Data regarding the corporate boards is retrieved from the annual reports of the non-listed firms. These annual reports are also available with the “company.info” database. Photos of the board members are used if the gender is not indicated explicitly. References indicating the gender such as Mr or Mrs are used in the absence of pictures within the annual report. External sources are applied if none of the indication factors are presented in the report. One such source is for example the website of the company. The procedure is comparable to the one used by Ahern and Dittmar (2010).

                                                                                                               

(21)

Data regarding the performance measures (ROA and ROE) and the control variables (board size, independent directors, ln(firm size), ln(firm age) and the lagged ROA and ROE) is collected from the databases “DataStream” and “AMADEUS” for the listed firms. Annual reports are used to gather this data for the non-listed firms if it is not available for these firms with the databases just mentioned.

To be part of the final sample the firm has to satisfy two more requirements. Firstly, the relevant data for a listed or non-listed company has to be available for the entire effective sample period (2009-2014) and the years 2007 and 2008 (to be able to use lagged values as instruments). Secondly, financial firms are excluded because they are exposed to specific accounting rules. The one-digit SIC 1987 codes are used to determine if a company belongs to the financial sector.

5.2 Variables

5.2.1 Dependent variable of interest

The dependent variable is a firm financial performance measure. As already indicated in the third section, several measures have been used for firm performance. There is still little agreement about which one should be used. The two types often applied are accounting-based measures (such as the ROA, ROE and ROI) and market-based measures (such as the Tobin’s Q and portfolio returns). Both measures have their strengths and weaknesses. Market-based measures, especially the Tobin’s Q, are the most frequently used measures within corporate governance studies. However, one main disadvantage is that these measures are subjective. They are strongly dependent on the market reactions that reflect the perceptions and

behaviour of investors (Fama, 1991). This study will use accounting-based measures since the sample also consists of Dutch non-listed firms and therefore market values are not observed for these firms. The main drawback of these measures is that they are heavily influenced by the applied asset-valuation methodology (Marinova et al., 2016). In first instance, the ROA will be used as a firm performance measure. For robustness, the ROE will also be used. The ROA is measured as the net income divided by the total assets and the ROE is defined as the net income over the total equity.

(22)

5.2.2 Independent variable of interest

The independent variable of interest is board gender diversity. This variable will be measured in four different ways: (1) the percentage of women in the boardroom, (2) a dummy variable that equals 1 if there is at least one woman present in the boardroom and 0 otherwise, (3) the Blau index (Blau, 1977), and (4) the Shannon index (Shannon, 1948). The Blau index is calculated as 1 − ! 𝑃!!

!!! , where 𝑃! equals the percentage of directors for each gender and

thus the number of categories (n) equals two (men and women). The index takes on a value of 0.5 when the composition of the board is equally divided between men and women. The value equals 0 when only men or women represent the board. The Shannon index is calculated as − !!!!𝑃! ∗ ln  (𝑃!). The definitions of 𝑃! and n are equivalent to the Blau index. The Shannon

index is different from the Blau index in the sense that it is more sensitive to small changes in the board gender composition (Randel, 2002). The values of this index lie between 0, when there is no gender diversity, and 0.69 when there is an equal number of women and men on the board.

According to Randel (2002) gender diversity is present in the boardroom when both men and women occupy the positions. Thus, a board consisting of only female directors is also not gender diversified. Therefore, the Blau and Shannon index are more suitable to measure diversity instead of using the first two measures described above.

5.2.3 Control variables

A first control variable included in the regression is the board size. Yermack (1996) finds that board size is associated with a negative influence on firm performance as measured by the ROA and Tobin’s Q. The result he found is motivated by the agency theory. Jackling and Johl (2009) found the opposite result, which is supported by Dalton et al. (2003). It is argued that the pool of knowledge increases with more board members. This leads to a more effective decision-making process, which eventually is beneficial for firm performance (Jackling and Johl, 2009). This reasoning is based on the resource dependency theory. Lipton and Lorsch (1992) argued that a board should maximally consist of ten directors (preferably eight or nine). Firms with a larger number of directors should not abruptly reduce the board size. Instead, it should be established gradually through attrition by retirement.

(23)

The second control variable is the share of independent board members. It is

recommended by the Dutch Corporate Governance Code that the members of the supervisory board, except one, have to be independent (Corporate Governance Committee, 2003).

However, according to van Ees et al. (2007) all supervisory board members are independent for most of the Dutch listed companies. In this research it is assumed that the same holds for the non-listed Dutch firms. Although the Dutch Corporate Governance Code only applies to listed firms, the majority of the non-listed firms within the used sample indicate that they comply with (most parts) of the Code. Therefore, the share of independent board members is equal to the amount of supervisory board members over the total board size. According to the agency theory increasing the number of independent directors has a positive impact on corporate performance. However, the empirical results are inconclusive (Baysinger & Butler, 1985; Jackling & Johl, 2009).

The third control variable is firm size. Firm size will be measured by the natural logarithm of total assets to reduce the skewness of the distribution.Large firms have some advantages over small firms. For example, larger firms are more able to exploit economies of scale and scope and have more diverse capabilities. These advantages have a beneficial effect on the operating efficiency, which in turn promotes firm performance (Penrose, 1959). Using a panel dataset, Lee (2009) confirms the existence of a positive relation between firm size and firm performance. On the other hand, according to Shepherd (1986) there is a correlation between size and market power. Market power creates x-inefficiencies, which negatively influences corporate performance (Leibenstein, 1976).

The fourth control variable included in the regression is firm age. This variable is calculated as the natural logarithm of the number of years the firm exists since its

incorporation date. Again, the logarithm is used to reduce the skewness of the distribution. One reason to believe that firm age positively influences firm performance is that over time firms have better knowledge about their abilities and how they can operate more efficiently. In addition, according to empirical evidence the best performing firms survive (Baker & Kennedy, 2002). Ericson and Pakes (1995) have found empirical evidence for an

improvement in performance with firm age. In contrast, Loderer and Waelchli (2010) have found the opposite result. It is argued that performance could be harmed by firm age because of organizational rigidities and inertia, which both come with age (Leonard-Barton, 1992; Hannan & Freeman, 1984).

(24)

The last control variable is the lagged value of the performance measure (ROA or ROE). This variable is included to reflect the dynamic nature of performance. Inclusion of the lagged performance is analogous to the papers of Adams and Ferreira (2009) and Carter et al. (2010). They have implemented this variable for two reasons: (1) current performance could be influenced by past performance and (2) past performance could influence the board structure, which implies reverse causality.

All the variables (except for the gender diversity variables) are winsorized at the 1% level to correct for outliers. With winsorizing these outliers are not deleted from the sample but are watered down. Using this technique alleviates the danger of biased estimators.

5.3 Descriptive statistics

The descriptive statistics for the entire sample are shown in table 3. The representation of female board directors was on average equal to 7.82% (or 0.68 expressed in numbers) for the period 2009 until 2014. It should also be mentioned that none of the firms within the sample have a board consisting of only female directors. The maximum percentage of female

directors is 55.56%. Differently formulated, if a sample firm has a female director then it will always also have a male director. Therefore, in essence board gender diversity is always present to some degree for these firms.  The financial performance measures (ROA and ROE) indicate that over the sample period the firms were on average financially successful.

However, it should be noted that even after winsorizing at the 1% level there is a wide

variation in the performance measures, especially for the ROE. The minimum ROE is equal to -169.87% and the maximum is equal to 121.36%. The minimum and maximum ROA are respectively equal to -39.06% and 30.11%. The average board size is equal to 7.42, which is similar to the optimal board size proposed by Lipton and Lorsch (1992).

(25)

Table 3: Descriptive statistics entire sample

Variable name Number of firm-year

observations Number of firms Mean deviation Standard Min Max

Firm characteristic Ln(firm size) ROA Lagged ROA ROE Lagged ROE Firm age Ln(firm age) Board characteristic Board size Independent directors

Gender diversity measures

Gender diversity (number) Gender diversity (%) Gender diversity (1/0) Blau index Shannon index 720 720 720 720 720 720 720 720 720 720 720 720 720 720 120 120 120 120 120 120 120 120 120 120 120 120 120 120 19.90 2.78% 2.97% 5.92% 7.08% 50.92 3.55 7.42 63.97% 0.68 7.82% 0.44 0.12 0.20 2.01 9.06 9.74 29.88 31.53 48.46 0.88 2.82 0.11 0.95 0.10 0.50 0.15 0.23 15.70 -39.06% -39.06% -169.87% -169.87% 2 1.39 1 33.33% 0 0% 0 0 0 24.58 30.11% 30.11% 121.36% 121.36% 248 5.80 16 88.89% 5 55.56% 1 0.5 0.69

This table shows the descriptive statistics for the entire sample for the period 2009 until 2014. Ln(firm size) is measured by the natural logarithm of the total assets. ROA is measured by the net income divided by the total assets. ROE is measured by the net income divided by the total equity. Lagged ROA and ROE are measured in the same way, only with the first lagged value. Ln(firm age) is equal to the natural logarithm of the recent year minus the firm’s year of incorporation. Board size is measured as the number of executive plus non-executive board members. The variable independent directors is equal to the number of non-executive board members divided by the variable board size. Gender diversity (number) equals the number of female directors on the board. Gender diversity (%) is calculated as the number of female directors divided by the variable board size.

Gender diversity (1/0) is a dummy variable that equals 1 if there is at least one female director on the board and

0 otherwise. The Blau index is measured by 1 − ! 𝑃!!

!!! , where 𝑃! equals the percentage of board directors for each gender and thus the number of categories (n) equals two (men and women). Lastly, the Shannon index is calculated as − !!!!𝑃!∗ ln  (𝑃!). The definitions of 𝑃! and n are equivalent to the Blau index. All variables (except for the gender diversity measures) are winsorized at the 1% level.

Table 4 describes the evolution of female representation on the board over the sample period. Panel A of the table presents the percentage of firms with zero, one, two, or more than two board positions occupied by women. In 2009 these percentages were respectively

60.83%, 29.17%, 6.67% and 3.33%. By 2014, the percentage of firms without female directors dropped to 46.67%. The percentages of the other categories increased by 2014, except for the firms with only one female director on the board. This trend is consistent with the trend presented in the Dutch Female Board Index 2014. Panel B of table 4 describes the evolution of the means of the four board gender diversity measures, which will be used for analysing the effect of gender diversity on firm performance. All these measures have

(26)

increased from 2009 until 2014. In short, based on the results in table 4 (panel A and B) it can be concluded that board gender diversity has increased during the sample period.

Table 4: Evolution of female directors on the board by year

Panel A: Percentage of firms with female directors by year

Year N Women on board = 0 Women on board = 1 Women on board = 2 Women on board > 2 2009 2010 2011 2012 2013 2014 120 120 120 120 120 120 60.83 59.17 58.33 60.00 50.83 46.67 29.17 29.17 27.50 20.83 27.50 28.33 6.67 8.83 10.00 13.33 15.00 13.33 3.33 3.33 4.17 5.58 6.67 11.67

Panel B: Means board gender diversity measures by year

Year N Gender diversity (%) Gender diversity (1/0) Blau index Shannon index

2009 120 5.77% 0.39 0.09 0.16 2010 2011 2012 2013 2014 120 120 120 120 120 6.34% 6.84% 7.05% 9.43% 11.51% 0.41 0.42 0.43 0.49 0.53 0.10 0.11 0.12 0.15 0.17 0.17 0.18 0.19 0.23 0.27

Panel A of this table shows the percentage of firms with female directors by year. Four dummies have been created for this: (1) a dummy that equals 1 if there are zero female directors on the board and 0 otherwise, (2) a dummy that equals 1 if there is one female director on the board and 0 otherwise, (3) a dummy that equals 1 if there are two female directors on the board and 0 otherwise and (4) a dummy that equals 1 if there are more than two female directors on the board and 0 otherwise. Panel B of this table shows the evolution of the means by year of the four different gender diversity measures that are used.

The distribution of female directors for the sample period by industry is presented in table 5, where the industry classification is based on the one-digit SIC 1987 codes. The transportation and public utilities sector is the industry with the highest representation of female directors (13.39%). The industry with the lowest percentage (or number) of female directors is the construction sector (4.72%). Theory suggests that the industries with the highest representation of female employees will probably have a positive effect on the number of female board directors. In the resource dependency theory Pfeffer and Salancik (1978) suggested that a firm that attracts more female customers may have a greater incentive to have more positions occupied by women at all levels in order to more effectively satisfy these customers. According to a report by the CBS (2015) the construction sector had the

(27)

lowest percentage of female employees in 2014 (11.51%).3 This suggests that the results in table 5 may provide some evidence for the above-mentioned theory by Pfeffer & Salancik (1978).

Table 5: Distribution of female directors by industry

Industry N Mean female directors (number) Mean female directors (%)

Mining Construction Manufacturing Transportation & public utilities

Wholesale trade Retail trade Services 2 13 41 13 18 9 24 0.92 0.37 0.72 1.19 0.43 0.70 0.68 7.73 4.72 7.98 13.39 6.17 6.40 8.01

This table shows the mean number and mean percentage of female directors by industry for the sample period 2009-2014. The one-digit SIC 1987 codes are used to determine in which industry a firm belongs. Three industry classifications ((1) agriculture, forestry and fishing, (2) public administration and (3) non-classifiable) have been left out of this table since none of the firms in the sample belonged in one of these industries. In addition, the finance sector is also not displayed. This is because firms belonging to this sector were excluded from the final sample.

Table 6 shows a comparison of means between boards with at least one female director and boards without female directors for the whole sample period. In essence, it is a comparison of means between gender diversified and gender undiversified boards. This is because as already mentioned the maximum percentage of female directors on the board is equal to 55.56%.Before calculating the statistical difference a Levene test is used to see whether the variances of the two groups are equal. If not, the Welch’s test is used for calculating the statistical difference of the means. Based on the table, the following can be concluded for firms that have female directors: these firms are larger, are more mature, have a lager board size and have a greater share of independent directors. For the performance measures ROA and ROE there does not seem to be a statistical difference between firms with and without female directors. Therefore, based on table 6, firms with female directors do not perform better than firms without female directors. It should be noted that causality cannot be derived from the table. However, the comparisons in means suggest that the choices of firms to appoint female directors could be influenced by firm characteristics. Therefore, it is important to control for these firm characteristics when performing the empirical analysis.                                                                                                                

3

Using table 3.15 of the CBS report (2015) this number is calculated as “vrouwen / totaal” = 55 / 304 = 11.51%. These numbers can be found in the row called “bouwnijverheid” (Dutch word for the construction sector).

(28)

Table 6: Comparison of firms with and without female directors on the board

Firm characteristic Mean for firm-years with female directors (n = 317) (gender diversified board)

Mean for firm-years without female directors (n = 403) (gender undiversified board)

Statistical Difference Ln(firm size) ROA ROE Ln(firm age) Board size Independent directors 20.88 2.63% 6.00% 3.74 9.00 65.42% 19.13 2.90% 5.85% 3.41 6.18 62.83% 1.75*** -0.26 0.15 0.33*** 2.82*** 2.59***

This table displays the means of different firm characteristics for firms with and without female directors on the board. A Welch’s test is used to calculate the statistical difference of means between the two groups if the variances appear to be unequal. Definitions of the variables can be found in table 3. *** Indicates that the difference is significant at the 1% level.

                                                             

(29)

6. Methodology

As already mentioned, a sample of Dutch listed and non-listed firms is taken to test the two hypotheses formulated in section 4. The following model is used to examine the impact of board gender diversity on firm performance:

𝑃𝑒𝑟𝑓𝑜𝑟𝑚𝑎𝑛𝑐𝑒!" = 𝐶𝑜𝑛𝑠𝑡𝑎𝑛𝑡  +  𝛽! 𝐺𝑒𝑛𝑑𝑒𝑟  𝐷𝑖𝑣𝑒𝑟𝑠𝑖𝑡𝑦!"#+ 𝛽! 𝑋!"+ 𝛿!+ 𝜀!"

Where,

𝑃𝑒𝑟𝑓𝑜𝑟𝑚𝑎𝑛𝑐𝑒!" = Dependent variable that measures firm performance defined as

return on assets (ROA) or return on equity (ROE).

𝐺𝑒𝑛𝑑𝑒𝑟  𝐷𝑖𝑣𝑒𝑟𝑠𝑖𝑡𝑦!"# = Variable for the presence of board gender diversity. As already indicated this will be measured in four different ways: (1) the percentage of female directors (gender diversity (%)), (2) a dummy variable equal to 1 if there is at least one female director present and 0 otherwise (gender diversity (1/0)), (3) the Blau index and (4) the Shannon index.

𝑋!" = A vector of the five control variables that are included: board

size, independent directors, ln(firm size), ln(firm age) and the lagged ROA (or lagged ROE)

𝛿! = Time fixed effects, which control for variables that are constant across firms but differ over time.

𝜀!" = The error term, which is the sum of variation in firm performance that cannot be explained by the dependent or control variables.

6.1 SYS-GMM estimator to address econometric problems

In contrast to reasons for believing that gender diversity improves firm performance, there are also theoretical reasons to believe that the relation goes the other way around. According to Hillman et al. (2002) the scarcity of women with experience as directors may lead to self-selection of those women into better performing firms. In addition, better performing firms may experience more pressure to comply with a target quota for gender diversity. This could

(30)

be due to a greater need for legitimacy (Meyer & Rowan, 1977), or due to greater latitude and more available resources to take the risk of hiring a female director (Pfeffer & Salancik, 1978). These arguments may imply that the degree of board gender diversity endogenously depends on firm performance. In other words, a positive relation between board gender diversity and firm performance may be driven by reverse causality.

In a lot of previous studies regarding this topic the authors have used an instrumental variable regression to solve for the reverse causality problem. However, in most cases the instruments used were weak. This will likely result in biased instrumental variable estimators. In addition, all variables regarding firm policies may be seen as endogenous variables

(Hermalin & Weisbach, 2003; Dezso & Ross, 2012). Therefore, this empirical analysis will use the general method of moments (GMM) estimator to solve for reverse causality. Two common types of this estimator are the Arellano and Bond’s (1991) difference GMM (DIFF-GMM) and the system GMM (SYS-(DIFF-GMM) estimator developed by Arellano and Bover (1995) and improved by Blundell and Bond (1997). The DIFF-GMM estimator uses the first differences of the regression model to control for constant unobserved heterogeneity and then uses lagged values of the internal endogenous variables as instruments. With the SYS-GMM estimator a system of two simultaneous equations is used: (1) an equation in levels and (2) an equation in first differences. The inclusion of the levels equation generates additional

instruments, where the own lagged first differences of the endogenous regressors in this equation are used as instruments. The equation in first differences uses the lagged levels of the endogenous regressors as instruments.

Blundell and Bond (1997) have shown that the SYS-GMM estimator solves three complications associated with the DIFF-GMM estimator. Firstly, the precision of the

estimators is improved. Secondly, the SYS-GMM estimator solves for the large finite sample bias associated with DIFF-GMM. Lastly, identifying a causal relation using the DIFF-GMM estimator may be difficult when there is little variation in the independent variable of interest. This is because time-invariant variables are differenced out with this estimator. Therefore, to address these three problems this empirical analysis uses the SYS-GMM estimator.

There are two versions of the SYS-GMM estimator: (1) the one-step SYS-GMM estimator and (2) the two-step SYS-GMM estimator. The two-step version is asymptotically more efficient than the one-step version. However, the two-step standard errors are typically downward biased (Arellano & Bond, 1998; Blundell & Bond, 1998). Therefore, the corrected standard errors proposed by Windmeijer (2005) are used with two-step SYS-GMM. This

(31)

could result in more efficient robust estimations for the two-step version instead of the one-step version, especially for SYS-GMM. As a result, this thesis will use the two-one-step version.

The inclusion of explicit dummies for time-invariant variables (such as firm or industry fixed effects) is possible when a SYS-GMM regression is performed.

Asymptotically, this would have no influence on the coefficient estimates of the regression variables because it is assumed that all instruments are uncorrelated to all the dummies for time-invariant regressors. It could be expected that removing these dummies from the error term does not alter the moments (Roodman, 2009). However, according to Roodman (2009) introducing explicit time-invariant dummies would still be a mistake. The use of lagged internal instruments would not be valid anymore with this implicit within group

transformation, especially when the number of time periods is small. Therefore, dummies for time-invariant regressors are not included when performing the SYS-GMM regression. In contrast, Roodman (2009) points out that year dummies (firm-invariant dummies) should always be included to correct for variables that are constant across firms but differ over time.

Before performing any regression with the SYS-GMM estimator, the included variables need to be divided into one of the three categories: endogenous, predetermined or exogenous variables. As already mentioned, all variables regarding firm policies may be seen as endogenous variables. Therefore, the potentially endogenous variables included in the regression are the gender diversity measures, the board size and the share of independent directors. According to Roodman (2009) standard treatment for these variables is to use the second lag of the same variable as an instrument. Firm size and firm age are treated as

predetermined variables. In most studies these firm characteristic variables are assumed to be strictly exogenous. However, there are arguments that these variables could be influenced by past values of the performance measures (Hall & Weis, 1967; Ozkan, 2001; Campbell et al., 2008). Therefore, these variables may not be strictly exogenous and are treated as

predetermined. Standard procedure for predetermined variables is to use the first lag of the same variable as an instrument (Roodman, 2009). Only the year dummies are treated as exogeneous variables.

The Hansen J statistic will be used to indicate whether the instruments are valid. The null hypothesis of this test is that the instruments are valid. In other words, the higher the p-value of the Hansen J test, the higher the validity of the instruments. The AR(2) test will be used to indicate whether the residuals in first differences are serially correlated. Under the null hypothesis there is no such correlation. A high p-value indicates that the used lags for the variables are valid. A deeper lag should be applied if this is not the case.

(32)

7. Main results

This section will discuss the results of the two-step SYS-GMM regressions for the sample period 2009-2014 to investigate the relation between board gender diversity and firm

performance. Firstly, in the upcoming subsection the whole panel dataset is used to perform a regression and test the first hypothesis: increasing board gender diversity will positively influence firm performance. In subsection 7.2 the sample is divided in a subsample of 60 listed and a subsample of 60 non-listed firms to test the second hypothesis: increasing board gender diversity will more positively influence firm performance for non-listed than for listed firms.

7.1 Two-step SYS-GMM board gender diversity and ROA entire sample

The results for the two-step SYS-GMM regressions with the entire panel data sample are shown in table 7. The first thing that should be noted is that the null hypotheses of the Hansen J and AR(2) tests cannot be rejected for all the regression specifications. This means that the lagged internal instruments used are valid and that autocorrelation of the second order does not appear to be a problem.

The results further suggest that there appears to be no significant link between board gender diversity and firm performance (as measured by ROA) for all the regressions with different specifications of gender diversity (gender diversity (%), gender diversity (1/0), Blau index and Shannon index). Formulated differently, the hypothesis that board gender diversity does have a positive influence on firm performance is not supported. This result is consistent with the result found by Smith et al. (2012). They have used a sample of UK listed firms between 1996 and 2010 and have applied a DIFF-GMM regression instead of a SYS-GMM regression.

Concerning the control variables, the natural logarithm of firm size (ln(firm size)) is positive and statistically significant at the 10% level for all the model specifications. It is consistent with the argumentthat larger firms are more able to exploit economies of scale and scope and have more diverse capabilities, which promotes operating efficiency and eventually positively influences firm performance (Penrose, 1959). The variable independent directors is also statistically significant (at the 10% level), except when the dummy variable is used as specification for gender diversity, and appears to have a negative influence on firm

Referenties

GERELATEERDE DOCUMENTEN

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

In summary, regarding the relationship between board gender diversity and firm performance, despite the mixed results, studies which assert a positive effect of the presence of

The regression equation is almost the same, besides that the average age of board members, percentage of women and foreign directors will be replaced with age diversity,

While family ties and an industry’s proportion of female employees were found to be positively associated with board gender diversity, the remaining variables of

This thesis uses an international dataset, to empirically test the relationship between board gender diversity and firm financial performance, with the

Perhaps, an increase in the amount of female board members does not affect dividend payout in countries with a civil law origin because the amount of female directors within

In short, this study believes the relationship between BGD and CFP to be positively moderated by national culture since the characteristics belonging to a high score on

But, there is no relationship between a (gender and foreign) diverse board policy and firm performance. Keywords: Agency Theory, Human Capital Theory, Resource Dependence