Master’s Thesis Econometrics Female board membership and firm performance

Master’s Thesis Econometrics

Female board membership and firm

performance

Does gender diversity affect Tobin’s Q and ROA?

Author:

I.G.G. Snijders

Student number:

s2561247

Master’s Thesis Econometrics, Operations Research and Actuarial Studies

Female board membership and firm

performance

Does gender diversity affect Tobin’s Q and ROA?

I.G.G. Snijders

June 27, 2019

Abstract

A substantial amount of research has been done investigating the effect of female board membership on firm performance. In this paper, we examine whether the fraction of female board members affects Tobin’s Q and return on assets (ROA). Data on 183 listed firms in the Netherlands is collected to determine this effect. This paper contributes to the existing literature in being the first to relate the fraction of female board members to firm performance for all listed firm in The Netherlands. We show that there is no statistically significant effect on firm performance when considering both firms with an one and a two-tier board structure. Next, we show that there is a statistically significant effect on both performance measures when solely considering firms with a two-tier board structure. This effect is negative when considering Tobin’s Q and positive when considering ROA. Finally, the data are tested on endogeneity and an instrumental variable is used. The fraction of female board members is instrumented by the fraction of male board members connected to female directors through other board memberships. We find no significant effect on the performance measures when considering one and two-tier board structures and when considering separate management and supervisory boards.

1

Introduction

Gender diversity in the boardroom has been a topic of discussion for the last decade. However, there are big differences between the various points of view. While some advocate that an increase in boardroom diversity enhances corporate performance, others believe that a higher share of female board members leads to less experienced boards and a decrease in profits.

Recently, the Bureau for Employers Activities of the International Labour Organization (ILO) reported that women in leadership positions bring better firm performance. Survey-ing almost 13,000 firms in 70 countries, the bureau found that gender diversity initiatives in the management increased profits on average with 10 to 15%. Furthermore, the bureau used data from 186 countries from 1991 to 2017 and discovered that, on a national level, an increase in female employment had a positive effect on GDP growth. Moreover, the bureau found that the positive effects of gender diversity start to grow when 30% of the management and leadership positions are held by women. The bureau also identified some essential determinants of preventing women to hold such positions. Women often have responsibilities towards their household and family which prevent them from working the same amount of hours as men. Additionally, firms generally need to improve their policies regarding, for example, flexible working hours and paternity leave.

Furthermore, as gender diversity in leadership positions has been talked about a lot, several countries, including Spain, Iceland, France and Norway have implemented a quota concerning women in the boardroom. Numerous researchers have examined the effects of such a quota and the relationship between an increase of female board members and firm performance, resulting in different outcomes. A statutory target of 30% of the board members of listed firms being female was implemented in the Netherlands on January 1st 2013. The target aims to improve the balance between men and women in board positions. Henderikse & Pouwels (2017) showed that before the implementation of the statutory target the percentage of women on the management board was 7.4%. The percentage of women on the supervisory board was 9.8% in 2012. Only 9.3% and 14.7% of the Dutch listed companies reached the target of 30% in 2012 for the management board and the supervisory board, respectively.

The committee monitoring the target argues that the amount of women in the board of a firm is positively correlated with the operating results. Furthermore, the committee claims the target supports equal opportunities for men and women and will cause boards to be a better reflection of the market. However, these arguments remain assumptions as the committee does not include proof to support their beliefs. Furthermore, the committee imposes firms to either reach the target or explain why they couldn’t. However, there is no penalty if neither one is done. This lack of sanctions could explain the small share of firms reaching the target.

The effect of gender diversity on firm performance has been investigated by many researchers. The studies conducted are largely based on U.S. data and present different outcomes, some found positive effects, while others found negative relations or even no relation at all. This study adds to the existing literature in being the first study to relate the fraction of female board members to firm performance for all listed firms in The Netherlands solely.

merely on listed firms. We use the results to determine whether implementing the statutory target has influenced the performance of firms. To answer the question we investigate the relation between the fraction of female board members and firm performance by means of regression analysis. Considering the report from the ILO and other researches, we expect to find a positive effect between the fraction female board members and firm performance. The structure of the paper is as follows: first, the existing literature is reviewed in Section 2. The data collection procedure is described in Section 3, which also contains descriptive statistics and univariate analysis of the data. Section 4 discusses the models used and shows the empirical results of our study. The conclusion and discussion are presented in Section 5 and 6, respectively. The Appendix shows the tables and graphs discussed.

2

Literature review

In this section the existing literature is reviewed. There is a separate subsection discussing

the research of L¨uckerath-Rovers (2013), who used data on 99 Dutch listed firms and found

better performance when firms had women on their board. Furthermore, we will explain our choice of dependent variable and the endogeneity problem.

2.1

Existing literature

Numerous researchers have examined the relation between gender diversity in board posi-tions and firm performance. Some find positive effects, others found no or even negative effects. This is in line with the results of Joecks et al. (2013). They used data from 151 firms in Germany to determine whether there is a relation between gender diversity on the board and firm performance measured by return on equity (ROE). They included a quadratic term in their regression analysis and found that the relationship between the two follows a U-shape pattern: gender diversity has a negative effect on firm performance, but after 30% of female board members the firm performance increases. This result may ex-plain why some studies found positive and others found negative correlations. The different relations found will be discussed below.

2.1.1 Positive effects

Campbell & M´ınguez-Vera (2008) researched the effect of gender diversity on the board of directors on firm performance in Spain. Using panel data analysis, they did find a positive effect of gender diversity on firm value. They used Tobin’s Q as a measure for firm performance. Different variables are used as a proxy for gender diversity. First, they used a dummy indicating that there is at least one woman on the board. Furthermore, they use the fraction of women and two different indices taking both the gender categories and the evenness of the distribution of board members among the categories into account. Moreover, they used board size, leverage, firm size and return on assets (ROA) as control variables. They found a significant negative effect of firm size, measured by the natural

logarithm of total assets, on firm performance. Furthermore, they reported R2 _values

between 0.163 (when using fraction of women) and 0.099 when using a dummy.

frequently used and consistent with other research in the field. They used data from 127 US companies and correlation and regression analysis to obtain their results. Next to the positive effect of board diversity, they found that production firms report a better return

on investment compared to service firms. They reported an R2 _{value of 0.17.}

A positive relationship between the fraction of female directors on the board and firm performance was also found by Terjesen et al. (2016). They measured firm performance by Tobin’s Q and ROA. They used data from 47 countries, including the Netherlands. They found significant negative effects for board size and firm size.

Christiansen et al. (2016) examined the link between gender diversity in senior positions and ROA for two million companies in Europe. They could obtain a data set this big by including non-listed firms. Firms were excluded from the sample when having less than two board members, in order to be able to examine gender diversity on the boards. The researchers discovered that the positive effect they found was more evident in sectors where the proportion of female workers is greater, like the services sector.

Furthermore, Sabatier (2015) observed a positive effect of gender diversity on economic performance in France. He used three performance measures, namely return on equity (ROE), ROA and Tobin’s Q. In addition to the positive effect of gender diversity, he found

a positive effect of firm size as well. He reported an R2 _{value of 0.523.}

Nguyen & Faff (2007) investigated whether female board representation in Vietnam affects firm outcomes. They found that the positive effect of having women on the board is more pronounced when having a greater share of female directors. Furthermore, they found a positive effect of board size and a negative effect for the natural logarithm of total assets.

2.1.2 No significant effect

Rose (2007) investigated if board diversity, measured by gender, educational background and ethical background, has an effect on firm performance measured by Tobin’s Q. A sample of listed Danish firms was used. He did not find a statistically significant effect between female board representation and performance. His regressions reported an F-statistic of 2.05, which could imply that the model has small predictive capability.

Furthermore, Marinova et al. (2016) examined the effect of gender diversity on the board of directors on firm performance as measured by Tobin’s Q. They used a sample of 184 Dutch and Danish listed firms. Moreover, they used firm size, age, industry, the share of independent directors and the board size as explanatory variables, but did not find any significant result. They did not find a significant effect on firm performance for gender diversity either. They reported F-statistics between 1.19 and 2.09, which are again small values implying that the models have small predictive capability.

Gallucci et al. (2015) showed that the presence of women on company board does not affect firm performance in Italy. However, this positive relationship becomes significant when they consider the moderating role of the female presence in ownership. Return on sales (ROS) was used as the dependent variable. As control variables they used firm size, firm age and governance structure among others, they found an significant effect of governance structure only and reported a quite low F-statistic of 6.09.

Furthermore, Chapple & Humphrey (2014) found no correlation between having women on a board and ROA of firms in Australia. Furthermore, they show that there is no significant difference in returns of firms with one versus more women on the board.

Eckbo et al. (2016) used a difference-in-difference approach and found no effect when considering forced gender-balancing of corporate boards in Norway and the effect on Tobin’s Q.

2.1.3 Negative effects

Ahern & Dittmar (2012) found a significant drop in stock prices at the announcement of the gender quota in Norway. They showed that the quota lead to younger, less experienced boards, increases in leverage and acquisitions and deterioration in operating performance of firms. Furthermore, they found lower values of Tobin’s Q over the following years, which is consistent with their assumption that firms choose a board to maximize firm value. This would indicate that they already had an optimal board composition, which was violated by increasing the female board participation to the 30% that was imposed by the gender quota. The difference in findings between Eckbo et al. (2016) and Ahern & Dittmar (2012) could be due to the first having a much larger data-set.

Comparing affected and unaffected firms when determining the impact of gender quotas on corporate policy decision in Norway, Matsa & Miller (2013) determined that affected companies undertook fewer workforce reductions, leading to increased labour costs and employment levels and reduced short-term profits.

2.2

L¨

uckerath-Rovers (2013)

L¨uckerath-Rovers (2013) used data on 99 Dutch listed firms in the period of 2005-2007 and

investigated the difference in financial performance, measured by ROE, ROS and return on invested capital (ROIC) between firms with and without female board members. She aimed to improve the methodology of the researches of Desvaux et al. (2007) and Catalyst (2004). Neither of the studies reported if their results where statistically significant.

First, she used the Catalyst (2004) method, examining the differences in ROE, ROS and ROIC for firms with and without female directors. She reported statistically significant results for ROE and ROIC: both performance measures where higher when firms did have female directors.

She reported F-statistics of 7.89 when using the fraction of female board members and an slightly higher F-statistic of 9.01 when using the dummy variable indicating the presence of women on the board.

Her research differs from our study on several levels. Firstly, she used only listed firms with a statutory domicile in the Netherlands. This resulted in firms with a two-tier board structure only. Furthermore, our study uses other performance measures and the control variables are different. Where her study included firm size, board size and a dummy indicating the firm being in the financial sector, with firm size measured as the natural logarithm of total assets, our study measures firm size differently and includes several additional control variables, like dummies for a two-tier board structure and a male CEO. Moreover, she mentions the endogeneity problem, but does not consider it in her regression.

2.3

Dependent variable

As shown above, the dependent variable varies between studies. Most studies use perfor-mance measures as Tobin’s Q, ROE and ROA. This study will follow the existing literature in choosing the dependent variable. As Tobin’s Q and ROA are most widely used, we use these as performance measures.

Tobin’s Q is originally devised by James Tobin of Yale University. For this research, Tobin’s Q as defined by Chung & Pruitt (1994) will be used. They developed a simple formula for approximating the Tobin’s Q which only requires basic financial and accounting information. Furthermore, they empirically tested the usefulness of their simple formula. Their ’approximated q’ is defined as follows

Q = MVE + PS + DEBT

TA ,

where MVE represents the market value of the firm defined as the product of a firm’s share price and the number of common stock shares outstanding, PS is the liquidating value of the firm’s outstanding preferred stock, DEBT is defined as the value of the firm’s short-term liabilities plus the book value of the firm’s long-term debt, and TA is the book value of the total assets of the firm.

As a second measure of firm performance ROA is used. ROA measures the profitability of the firm relative to its total assets. We define ROA as follows:

ROA = NI

TA,

where NI represents net income and TA are the book value of the total assets of the firm.

2.4

Endogeneity problem

When observing the relationship between female board participation and firm performance, one can envision a endogeneity problem. This problem consists of reverse causality. Reverse causality may be an issue since higher female board representation could increase firm performance, while simultaneously firms performing better could be more likely to take the ’risk’ of hiring an extra female for a board position.

influences equity risk. They also displayed that unobserved between-firms heterogeneous factors drive the findings of a negative relationship between risk and gender diversity on the board of directors.

Furthermore, Campbell & M´ınguez-Vera (2008) discovered that gender diversity has a positive effect on firm value in Spain, but found no significant effect for the opposite causal relationship.

However, as it is likely that there exists reverse causality between the financial per-formance of a firm and the fraction of female board members, as explained before, an instrumental variable approach is conducted in this study. The instrument should be cor-related with the endogenous variable, that is, the fraction of female directors. However, it is essential that it is not correlated with the error term of the regression. Otherwise, it would suffer from the same problem as our endogenous variable.

Terjesen et al. (2016) assumed endogeneity and used the lagged exogenous variables as instruments, that is, the lagged values of board size, number of employees and working women index. They found that these lagged variables formed a valid set of instruments and found a positive relation between the fraction of female board members and firm performance.

We will use the instrument that is most widely used in the existing literature. This instrument is defined by Adams & Ferreira (2009) and denotes the fraction of male board members who have female connection through other board memberships. The researchers support their idea by explaining that the lack of connections of female board members is a main cause of the low fraction of female board members. They argue that, the greater the fraction of male directors with female connections is, the greater the gender diversity on the board should be.

The study conducted by Medland (2004) supports this. She argued that the most important obstacle to having female board members is that the informal social network between directors is dominated by men. Hence, the more connected male directors are to women, the more women should be in leadership positions. However, it is hard to measure informal social connections, but networks occurring because directors sit on multiple boards can be examined. Therefore, we will use the fraction of male directors who have female connections through other board memberships as our instrumental variable for the fraction of female board members.

3

Data

In this section we will discuss the explanatory variables. Next, the data collection procedure

is described. Finally, we will show some descriptive statistics and perform univariate

analysis by means of the correlation matrix, crosstables and boxplots.

3.1

Explanatory variables

As explanatory variables we include some firm-level characteristics conform the existing literature, such as firm age, firm size and board size.

Our main variable of interest is the fraction of female board members. This is defined by FF. Based on the existing literature we could either find a positive, a negative or even

no effect of the fraction of women on the board on our performance measures. L¨

Netherlands. Though our sample and research method differs from the one she conducted, we expect to find similar results, namely a positive effect between the proportion of female board members and firm performance.

Moreover, dummies indicating a male CEO and a two-tier board are included. In the Netherlands it is common to have a two-tier board structure. However, foreign firms are likely to not be familiar with this board structure. Therefore, it is possible to have a one-tier board structure, having the executive and supervisory directors in one board: the board of directors. As the number of female CEOs is very small, we do not expect a significant relation between CEO and our performance measures. Furthermore, we expect the two-tier board structure to be positively related to firm performance as a two-tier structure implies that the responsibilities are well-organized.

DIR is defined as the number of directors on the board of directors, that is both the management board and the supervisory board, and hence measures the size of the board. We expect board size to have a negative relation with firm performance as a large board could imply inefficiency.

In addition to DIR we use the size of the firm (dsize) as an explanatory variable. We use the number of employees and the total assets of a firm to divide the firms into two groups, small and medium sized enterprises (SME) and large enterprises, following the criteria of the Dutch Enterprise Agency. To be defined as a SME a firms must have fewer

than 250 employees and either an annual turnover of e40M of less or the total number

of assets is equal to or less than e20M. We expect firm size to have a positive effect on

Tobin’s Q as larger firms, due to higher market power, tend to be more productive. Hence, firm size is also expected to have a positive effect on ROA.

Firm age (AGE) is defined as the year, which can be 2012 or 2017 in our sample, minus the year of incorporation of the firm. We expect a negative significant effect of age on both dependent variables. Loderer & Waelchli (2010) showed that firms tend to be less profitable when they grow older. They argued that older firms have a weakened ability to compete.

The time dummy (time) has value one for 2017 and zero for 2012. We do expect a positive effect on firm performance as the economic situation of the Netherlands in 2017 was substantially better compared to 2012.

Additionally, we will replace our dummy indicating a two-tier board structure and the total number of directors with two variables explaining the number of directors on the management board (MB) and the supervisory board (SB) separately. This will decrease our number of observations as the firms with an one-tier structure will be left out of our regression. We expect both variables to have a negative effect on firm performance as expected for the total number of directors.

The variables and their definitions can be found in Table 1. Section 3.3 provides some descriptive statistics on these variables which will be elaborated in Section 3.4.

3.2

Data collection

Netherlands when starting this study. This resulted in 366 observations.

As our study is the first to investigate the relation between firm performance and the fraction of female board members for all listed firm in the Netherlands in 2012 and 2017, there does not exist a complete data-set. Hence, the data is collected using different sources. The financial data is collected from the online Bureau Van Dijk Orbis database. The firm level data contains statistics of firms listed in The Netherlands, that is, firms listed on the AEX, AMX and AScX. This includes foreign firms, as the quota includes firms with a foreign parent firm. The statistics contain financial data such as total assets, debt, operating revenue, but also include the number of employees. As this data-set is not complete for all firms, missing data is hand-collected from annual reports of the firms. This concerns roughly a third of the firms in our data-set, that is approximately, 120 observations. However, not all firms have all of their financial data online.

In addition to the financial data, data from Mijntje L¨uckerath’s Female Board Index is

used. These data contain statistics on the total number of directors on the board of Dutch firms and the number of female directors on the board. As this index does not include all listed firms in The Netherlands, the data-set remains incomplete. It is completed by hand-selecting the missing observations using annual reports. Per missing firm, that is, for approximately 150 observations, the total number of directors on the management board and supervisory board is collected from the annual reports and by researching background information the number of female directors is noted. Furthermore, in order to have a consistent instrumental variable, the annual reports of all 183 firms are inspected for both years. Then for each of the 2112 male directors in our data-set their female connections through other board memberships are examined by looking into their working experience and then inspecting the annual reports of the firms mentioned. As many directors serve on various boards and those boards often consist mainly of male directors, the data collection was very time-consuming. The inspection resulted in 535 male directors in our data-set having female connections through other board memberships.

3.3

Descriptive statistics

Table 2 shows some descriptive statistics for the total sample and the subsamples consid-ering the years 2012 and 2017. More generally, the gender quota of 30% is reached by a quarter of our sample. Furthermore, 66% of the firms has at least one female board member. In 2012 we have a third of the companies meeting the target of 30% of female board members and 63% of the firms having one or more women on their board. In 2017 a considerably lower percentage of firms reached 30% of women in their board, namely 20%. The fraction of firms with at least one female board member is slightly higher, namely 68%.

3.3.1 Whole sample

average of 3.11 directors, whereas the supervisory board is a bit larger with 5.15 directors on average. When examining the connections with other female board members, we see that 24% of the male directors has female connections though other board memberships.

3.3.2 Subsample 2012

The 2012 subsample consists of more large firms, namely 84% of the firms is considered a large firm. Furthermore, the average Tobin’s Q of the firms in the subsample is 1.24, indicating that the stock is overvalued. The number of directors on the board of directors is 7.97 on average, existing for 9% of female members. Moreover, 76% of the firms has a two-tier board structure and 97% is directed by a male CEO. The management board consists on average of 3.01 directors and the supervisory board of 5.12 directors. On average, 24% of the male directors has female connections through other board memberships.

3.3.3 Subsample 2017

In 2017 the listed companies have an average Tobin’s Q of 1.94, which is considerably higher than in 2012. However, a smaller part of the subsample is considered to be a large enterprise (78%). Furthermore, an average board consists of 8.01 members, having a fraction of 15% being female. This fraction is substantially higher (by 6%) than in the subsample five years before. In addition, 74% of the firms has a two-tier board structure and again 97% has a male CEO. The management board consists of 3.18 directors, on average. The supervisory board consists on average of 5.17 directors. Moreover, a quarter of the male board members has female connections through other board memberships.

3.4

Univariate analysis

In addition to the descriptive statistics above, univariate analysis is conducted. First, the correlation matrix is analysed. Next, we examine some crosstables in order to get a first impression of our data. Lastly, we present boxplots to five a good visual representation of the differences between the two years in our data-set.

3.4.1 Correlation matrix

Focusing on ROA, one can see that there is a positive relation between FF and ROA, which is also statistically significant at the 1%-level. This indicates that firms with a higher fraction of female board members have a higher ROA. Furthermore, dsize is statistically significant and positively correlated to ROA, supporting our expectation that larger firms perform better. However, we see that AGE is positively correlated with ROA as well, which is opposite to our expectations.

Moreover, we see that firms having a male CEO, that is CEO equals one, have a smaller fraction of female board members. DIR is positively and statistically significantly correlated with the fraction of female board members, indicating that a bigger board has a higher fraction of female board members. Furthermore, AGE and dsize are positively correlated and statistically significantly different from zero, indicating that older and larger firms have a higher fraction of female board members. Additionally, we see that FF and our time dummy are positively correlated and that this correlation is statistically significantly different from zero at the 1%-level. This implies that firms in our 2017 sample have a higher proportion of women on their board of directors. The number of directors on the management board and supervisory board are both separately positively and statistically significantly correlated to the fraction of female board members, indicating that a big-ger management board and supervisory board both have higher fraction of female board members. Finally, we want a strong correlation between the fraction of female board mem-bers and our instrument, the fraction of male directors having female connections through other board memberships. We find a positive and statistically significant correlation, in-dicating that the fraction of male directors having female connections could be used as an instrument.

3.4.2 Crosstables

To get a first impression of the effect of the different explanatory variables on the dependent variables, we examine some crosstables. These tables show the one-on-one effect of the variable on the dependent variable. The results are shown in Tables 4 till 11.

We categorized the fraction of female board members into three categories. The dif-ferences between the mean of the groups are shown in Table 4. It becomes clear that a higher fraction of women on the board of directors, that is a higher category, results in a higher Tobin’s Q and ROA. That is, firms with more female directors on the board are more profitable compared to firms having less female board members.

Table 5 shows how having an one or two-tier board structure affects the measures of firm performance. For both Tobin’s Q and ROA a two-tier board structure seems to be less beneficial, resulting in a lower Tobin’s Q and ROA.

The number of directors is categorized into three groups. The results are shown in Table 6. Tobin’s Q is decreasing with each group, indicating that having more directors on the board would reduce Tobin’s Q. However, the categorized variable has an opposite effect when considering the ROA of a firm: having a bigger board of directors resulting in a higher ROA.

When considering the size dummy dsize, it becomes clear that large enterprises have a higher Tobin’s Q as is shown in Table 7. Furthermore, ROA is also higher, with small and medium sized enterprises having a negative mean.

a lower Tobin’s Q. Considering ROA, one can see that the mean differs greatly. Fairly new firms have a negative mean ROA. The ROA then becomes positive and increases with age, indicating the more years of incorporation of a firm the higher ROA. Again, the standard deviations corresponding to the means regarding the ROA for the different groups are much higher, indicating that ROA in the age groups differ greatly.

One can see, in Table 9, that the mean of Tobin’s Q in 2017 is higher compared to 2012, indicating that the firms in this subsample are more profitable. However, the ROA in 2017 seems to be lower and even negative, which is opposite to 2012, where the return is substantially positive. However, the standard deviation is considerably higher when examining the ROA.

Table 10 shows the categorized variable corresponding the number of directors in the management board. One can see that Tobin’s Q is highest when having the less directors on the management board. When considering ROA, the opposite is true.

Finally, Table 11 presents the crosstable of the categorized variable corresponding to the number of directors on the supervisory board and our performance measures. Both performance measures are highest for the last category, which corresponds to the highest number of directors on the supervisory board.

3.4.3 Boxplots

To give a good visual representation of the differences between the subsamples, boxplots are created. Figures 1 till 12 show the results. When considering Tobin’s Q, only values below 5 are shown, as to make the picture more clear. For ROA, we used values between minus 10 and 10.

First, some boxplots are made regarding the fraction of female board members. First, we categorized the variable into three groups. The first plot, shown in Figure 1, displays the differences in firm performance measured by Tobin’s Q. Though the medians are quite similar, one can see that the upper quartiles differ greatly. The second and third group, which include firms with a higher fraction of female directors, contain more profitable firms compared to the first group. Figure 2 shows the same groups but now considering the ROA of the firms. One can see that the differences between the groups are small, the boxplots are very much alike. However, the differences within the groups appear bigger, indicating that the ROA between firms in one group differs.

The boxplots in Figure 3 and Figure 4 show the relation between the performance measure and the dummy indicating whether a company has a two-tier board structure (dummy equals 1) or a one-tier structure (dummy equals 0). The difference in average Tobin’s Q between the distinct structures appears to be small: the medians coincide, though the difference within the groups seems higher when firms are having an one-tier board structure. Considering the ROA, one can see that the box corresponding to a two-tier structure is higher, indicating that the firms in this group have a higher ROA than the firm that have implemented an one-tier structure.

The effect of the size dummy on our performance measures can be examined in Figure 7 and 8. When considering Tobin’s Q, it is noticeable that the boxes are quite similar, indicating that the firms have an approximately similar Tobin’s Q. Examining the boxplots regarding ROA, one can immediately see that the small and medium sized enterprises are widely spread considering their ROA. Furthermore, the medians of both groups are close. The boxplot concerning the categorized AGE variable are displayed in Figures 9 and 10. Inspecting the plots shows a similar pattern for both measures, it seems that the second and last category score lowest.

Figures 11 and 12 show the boxplots for Tobin’s Q and ROA against the time dummy, respectively. One can see that for both performance measures, though more clear for Tobin’s Q, the 2017 subsample scores higher, indicating that the firms in this subsample are more profitable. The difference between the two subsamples is more prominent for Tobin’s Q.

4

Models and estimation

In this section we will first discuss the models used. Next, the endogeneity problem is addressed and the empirical results are shown.

4.1

Models

The following model will be used to determine the relationship between female board representation and firm performance:

PMit = α + β1FFit+ β2twoit+ β3CEOit+ θ1DIRit

+ θ2dsizeit+ θ3AGEit+ δtimeit+ εit,

(1) where i denotes firm i and t can be 2012 or 2017. Moreover, PM is defined as either Tobin’s Q (Q) or ROA (ROA). FF represents the quantifying variable for female board members. This is defined as the number of female board members divided by the number of total directors on the board. Furthermore, two indicates if the firm has a two-tier board structure, CEO specifies having a male CEO. Moreover, DIR is the variable for board size and dsize indicates that the firm is considered to be a large enterprise. AGE represents the years since the incorporation of the firm and time is a dummy indicating the year 2017.

To examine the effect of the number of directors on the management board and the supervisory board separately, we will estimate the following model as well:

PMit= α + β1FFit+ β2CEOit+ θ1MBit+ θ2SBit

+ θ3dsizeit+ θ3AGEit+ δtimeit+ εit,

(2) where i again denotes firm i and t can be 2012 or 2017. Furthermore, MB and SB are defined as the number of directors on the management board and the supervisory board, respectively. The other variables are defined as in (1).

4.1.1 Endogeneity problem

will use a two-stage least square regression to examine whether the problem exists within our data-set. Depending on this outcome, an instrumental variable approach is conducted, where FF will be instrumented by the fraction of male directors having female connections though other board memberships.

4.2

Empirical results

Before we use the models described in Section 4.1, we test our data on multicollinearity and heteroskedasticity. Table 3 shows the correlations coefficients of the variables used. The absolute values of the correlations coefficients range from 0.0017 to 0.4130. Gujarati & Porter (1999) suggest that in order to eliminate a multicollinearity problem correlation coefficients should be below 0.80. This is the case with all our coefficients, hence we can eliminate the multicollinearity problem. The potential heteroskedasticity problem is dealt with by using robust standard errors clustered at firm-level.

4.2.1 Total board of directors

To examine the effects of the fraction of female board members on firm performance, we first estimate (1) with Tobin’s Q as the dependent variable. The first column of Table 12 shows the results of this estimation. One can see that the coefficient corresponding to the fraction of female board members is positive, indicating that a higher proportion of female board members results in a firm performing better. This is conform our expectations. Furthermore, having a two-tier board structure decreases the Tobin’s Q and having a male CEO increases the profitability of a firm, which is opposite to what we expected. The number of directors on the board of directors has a negative coefficient which implies that having more directors decreases the Tobin’s Q of a firm. Moreover, we expected the firm size to have a positive correlation with Tobin’s Q, as larger firms tend to be more productive, due to having higher market power. Conform our prediction, the coefficient of our size dummy is positive. Furthermore, the coefficient corresponding to AGE is negative, which implies that older firms have a lower Tobin’s Q. This supports our expectations based on Loderer & Waelchli (2010) who showed that older firms tend to be less profitable due to their weakened ability to compete. However, all above mentioned coefficients are statistically insignificant, causing the results to be unable to interpret.

Nevertheless, our time dummy time is statistically significantly different from zero and can thus be interpreted. The positive coefficient of 0.695 shows that the Tobin’s Q of a firm is 0.695 higher in 2017 than the performance measure in 2012. We did expect a positive effect on firm performance as the economic situation in the Netherlands was considerably better in 2017 compared to 2012.

Concluding, we can say that, when considering the total board of directors and Tobin’s Q as a performance measure, the fraction of female board members does not significantly influences firm performance. Furthermore, our time dummy does have a significant effect on Tobin’s Q. This positive effect implies that firms performed better in 2017 compared to 2012. With the quota of 30% female board members implemented in 2013, one would expect this effect to be partly due to the quota. However, as the economic situation in the Netherlands was considerably better in 2017 compared to 2012, we cannot conclude this from our results.

estimated results are shown in the second column of Table 12. The coefficient of the frac-tion of female board members is positive, indicating a positive effect of the fracfrac-tion on ROA. However, this coefficient is not statistically significantly different from zero and can thus not be interpreted. Furthermore, we again find statistically insignificant coefficients for a two-tier board structure, having a male CEO, the number of directors and the time dummy. Therefore, these results cannot be interpreted. Nevertheless, as the coefficients of dsize and AGE ate positive and statistically significantly different from zero, we can conclude that both large and older firms report a higher ROA.

Concluding, we can say that, regarding the total board of directors and ROA, the fraction of female board members does not have a statistically significant effect on firm performance. Moreover, the size dummy and the variable indicating the age of a firm have a positive and significant effect on ROA, implying that large firms as well as older firms have higher returns.

4.2.2 Management board and supervisory board

In this section we consider the models without the variable indicating a two-tier board structure and the number of directors on the board of directors. Instead, we use two variables indicating the number of directors in the management board and the supervi-sory board separately. As not all firms have a two-tier board structure, the number of observations in these model will be lower.

First, we estimate (2) with Q as the dependent variable. The estimated results are shown in the first column of Table 13. The fraction of female board members has a negative coefficient, indicating that a higher fraction of female board members decreases the Tobin’s Q of a firm. This result is statistically significantly different from zero at the 5%-level. The effect is opposite our expectations of Tobin’s Q increasing with the fraction of female board members. Furthermore, the coefficients corresponding to the number of directors on the management board and supervisory board are both not statistically significantly different from zero, as are the coefficients corresponding to CEO, dsize and AGE. However, our time dummy does has a significant effect on Tobin’s Q, indicating that firms have higher Tobin’s Q in 2017 compared to 2012. This effect is similar to the effect found when considering all firms.

Concluding, we found significant effects for the proportion of female board members (negative) and the time dummy (positive). This implies that, when considering firms with a two-tier board structure and Tobin’s Q as a performance measure, increasing the fraction of women on the board negatively affects firm performance. Furthermore, firms are again considered to perform better in 2017 compared to 2012.

zero at the 10%-level, we can conclude that an additional year since the incorporation of a firm increases the ROA.

Concluding, the fraction of female board members has a positive and statistically sig-nificant effect on ROA of firms with a separate management board and supervisory board. However, this effect is only statistically significant at the 10%-level. Furthermore, firms considered to be large firms by the Dutch Enterprise Agency have a higher return compared to small and medium sized firms. Older firms have a higher return as well.

4.2.3 L¨uckerath-Rovers (2013)

In this subsection, the results found before are compared to the results found by L¨

uckerath-Rovers. She found a significant positive relation between the fraction of female board members and firm performance measured by ROE. In contrast, we found no statistically significant relation when using using the total board of directors. Furthermore, she found a positive and statistically significant effect of firm size on her performance measure. In our study, this effect was only statistically significant when considering the ROA, not with Tobin’s Q. The difference in statistically significant effects may be explained by the fact

that L¨uckerath-Rovers only used firms with two-tier board in her sample.

When considering the management board and supervisory board separately, we do find a statistically significant relation between the fraction of female board members and both our performance measures as well. However, the effect is negative when considering Tobin’s

Q. This is in contrast to the positive effect that L¨uckerath-Rovers found. We do find a

statistically significant and positive effect when considering ROA. The positive effects found

by both L¨uckerath-Rovers and by us are both only statistically significantly different from

zero at the 10%-level, which could imply that the result may not be completely reliable. The statistically significant positive effect of firm size on ROA is similar to the results

found by L¨uckerath-Rovers.

4.3

Instrumental variable approach

As explained before, our model could suffer from endogeneity. This could be fixed by using an instrumental variable approach. We will first determine whether our model indeed suffers from endogeneity. If this is true, we will use an instrumental variable approach to estimate our models again.

4.3.1 Test for endogeneity

We will test for endogeneity by regression our expected endogenous variable, that is, frac-tion of female board members, on our exogenous variables and our instrument. We use the fraction of male board members who are connected to women through other boards as our instrument. Next, we estimate the residuals of this regression and then estimate the following model:

PMit = α + δˆeit+ βxit+ εit, (3)

where where PM is our performance measure and can be Q or ROA, xit is a column vector

of exogenous variables and ˆeit are the residuals from the model in 3. This test was first

models, we find that δ is not statistically significantly different from zero, indicating that we cannot reject the null hypothesis of exogeneity, implying that the fraction of female board members could be exogenous. However, we will conduct an instrumental variable approach to ensure that we rule out possible endogeneity.

4.3.2 Total board of directors

The results from the first and second stage of the regressions are shown in the second and first column of Table 14, respectively.

In the first stage regression, one can see that all coefficients are statistically significantly different from zero at the conventional significance levels.

Next, we estimate the second stage regression, where the fraction of female board mem-bers is instrumented by the fraction of male directors having female connections through other boards. Examining the results, it becomes clear that none the coefficients are statis-tically significantly different from zero.

Furthermore, we estimate the first and second stage regressions regarding ROA. The results are shown in, respectively, the fourth and third column of Table 14.

Examining the results from the regression of ROA on the other variables, it becomes clear that only the size dummy is a variable of interest. Size is positively associated with return on assets, supporting our assumption that larger firms tend to be more profitable due to higher market power.

Concluding, we can say that when using an instrumental variable approach only the size of a firm has a statistically significant effect on ROA. When considering Tobin’s Q, no statistically significant effects where found.

4.3.3 Management board and supervisory board

The results from the first and second stage of the regressions are shown in the second and first column of Table 15.

In the first stage regression, one can see that the coefficients corresponding to the number of directors on the management board and the size dummy are statistically in-significantly different from zero at the conventional significance levels.

Next, we estimate the second stage regression, where the fraction of female board mem-bers is instrumented by the fraction of male directors having female connections through other boards. Examining the results, it becomes clear that our time dummy is the only variable of interest as this is the only variable statistically significantly different from zero at a conventional significant level. The positive effect is conform our expectation that firms have a higher Tobin’s Q in 2017 compared to 2012.

Furthermore, we estimate the first and second stage regressions regarding the return on assets. The results are shown in, respectively, the fourth and third column of Table 15. Again the coefficients corresponding to the number of directors on the management board and the size dummy are statistically insignificant.

Examining the results from the second stage, we see that only the coefficient correspond-ing to the size dummy is statistically significantly different from zero at the conventional significance levels, indicating that larger firms are more profitable. This finding is conform our predictions made in Section 3.1.

when using an instrumental variable approach. The effects of the time dummy on Tobin’s Q and of the size dummy on ROA remain statistically significant and positive when using an instrumental variable.

Overall, we found that the fraction of female board members on the board of directors is not statistically significantly different from zero in most of our models. This indicates that gender diversity does not have an effect on the performance of firms. However, the effect is statistically significant when considering the number of directors on the manage-ment board and supervisory board separately, instead of the total number of directors at once. Nevertheless, this effect becomes statistically insignificant when conducting an instrumental variable approach.

5

Conclusion

To explain the effect of the fraction female board members on firm performance for Dutch listed firms for 2012 and 2017 the following is examined. First, an univariate analysis method is used to observe the possible effects. Examining the correlation matrix a signif-icant positive relation was found between both Tobin’s Q and ROA and the fraction of female board members. Furthermore, multicollinearity was ruled out.

Next, multivariate analysis was adopted to determine the effects. First, the total board of directors was considered, this included firms with a one and a two-tier board structure. We found no statistically significant effect between the fraction of female board members

and our performance measures. We did find that large firms reported a higher ROA

than small and medium sized firms. Furthermore, firms had a higher Tobin’s Q in 2017 compared to 2012. With the quota of 30% female board embers implemented in 2013, one could expect this effect to be partly due to the quota. However, as the economic situation in the Netherlands was considerably better in 2017 compared to 2012, we cannot conclude that this effect was due to the implementation of the quota.

Additionally, only firms with a two-tier board structure where considered. This resulted a negative and statistically significant effect of the fraction of female board members on Tobin’s Q. This was opposite our expectations based on existing research. The fraction of female board members had a positive and statistically significant effect on ROA, however, this effect was only statistically significantly different from zero at the 10%-level. Again, firms reported a better Tobin’s Q in 2017 and large firms had a higher ROA.

We tested our data on endogeneity and found no evidence that the endogeneity prob-lem is present in our data-set. However, to ensure that we rule out the small possibility of endogeneity, an instrumental variable approach was conducted. This approach did not significantly affect our results when considering both firms with a one and two-tier board structure. The effect of the fraction of female board members remained statistically in-significant.

When considering only firms with a two-tier board structure, the statistically significant effect of the fraction of female board members found before disappeared when using an instrumental variable approach. The effects of the time dummy on Tobin’s Q and of the size dummy on ROA remained statistically significant and positive when using an instrumental variable.

using an instrumental variable. However, since we do not have clear signs of endogeneity in our data-set, the instrumental variable approach is not entirely necessary. Furthermore, we do find a statistically significant and positive effect of our time dummy on Tobin’s Q. Nevertheless, as the economic situation in the Netherlands was considerably better in 2017 compared to 2012, it cannot be concluded whether this effect is partly due to the implementation of the statutory target in the Netherlands.

6

Discussion

This paper contributes to the existing literature in being the first study to link the frac-tion of female directors to firm performance for listed firms in the Netherlands. Existing literature has focused on other European and non-European firms or examined the effect

for the Netherlands among other countries. L¨uckerath-Rovers (2013) did investigate the

effect for the Netherlands, but only included firms with two-tier boards and did not test endogeneity in her data-set. Previous research on board diversity and firm performance has shown different results, including both positive and negative effect, but no significant effects as well. This paper shows that using an instrumental variable approach causes the effect found for firms with two-tier board structures to become insignificant.

A shortcoming of this research is the fact that just two years of data are used. Further research could expand the data-set by including more years of information. Furthermore, instead of using this study as a final answer in the discussion whether gender diversity on boards has an effect on firm performance, this research could be a starting point. Additional research could be done using more variables, for example concerning board characteristics.

References

Adams, R. B. & Ferreira, D. (2009), ‘Women in the boardroom and their impact on governance and performance’, Journal of Financial Economics 94(2), 291–309.

Ahern, K. R. & Dittmar, A. K. (2012), ‘The changing of the boards: The impact on firm valuation of mandated female board representation’, Quarterly Journal of Economics 127(1), 137–197.

Campbell, K. & M´ınguez-Vera, A. (2008), ‘Gender diversity in the boardroom and firm financial performance’, Journal of Business Ethics 83(3), 435–451.

Catalyst (2004), The bottom line: Connecting corporate performance and gender diversity, Catalyst, New York.

Chapple, L. & Humphrey, J. E. (2014), ‘Does board gender diversity have a finan-cial impact? Evidence using stock portfolio performance’, Journal of Business Ethics 122(4), 709–723.

Christiansen, L., Lin, H., Pereira, J., Topalova, P. B. & Turk-Ariss, R. (2016), Gender diversity in senior positions and firm performance: Evidence from Europe, IMF Working Paper, No 16/50.

Desvaux, G., Devillard-Hoellinger, S. & Baumgarten, P. (2007), Women matter: Gender diversity, a corporate performance driver, McKinsey.

Durbin, J. (1954), ‘Errors in variables’, Revue de l’Institut International de Statistique / Review of the International Statistical Institute 22(1/3), 23–32.

Eckbo, B. E., Nygaard, K. & Thorburn, K. S. (2016), Does gender-balancing the board reduce firm value?, CEPR Discussion Paper No. DP11176.

Erhardt, N. L., Werbel, J. D. & Shrader, C. B. (2003), ‘Board of director diversity and firm financial performance’, Corporate Governance: An International Review 11(2), 102–111. Gallucci, C., D’Amato, A. & Santulli, R. (2015), ‘Women on board of directors and firm perfomance: The moderating role of female ownership. Empirical evidence from the italian wine industry’, Journal of Financial Management, Markets and Institutions 3(2), 225–244.

Gujarati, D. N. & Porter, D. C. (1999), Essentials of econometrics, Vol. 2, Irwin/McGraw-Hill Singapore.

Hausman, J. A. (1978), ‘Specification tests in econometrics’, Econometrica: Journal of the Econometric Society 46(6), 1251–1271.

Henderikse, W. & Pouwels, B. (2017), Een beetje beter, maar nog lang niet voldoende. Bedrijvenmonitor Topvrouwen 2017, Zeist: Van Doorne-Huiskes en Partners.

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), 61–72.

Julizaerma, M. K. & Sori, Z. M. (2012), ‘Gender diversity in the boardroom and firm per-formance of Malaysian public listed companies’, Procedia-Social and Behavioral Sciences 65, 1077–1085.

Loderer, C. F. & Waelchli, U. (2010), Firm age and performance, Working Paper (Univer-sity of Bern, Switzerland).

L¨uckerath-Rovers, M. (2013), ‘Women on boards and firm performance’, Journal of

Man-agement & Governance 17(2), 491–509.

Marinova, J., Plantenga, J. & Remery, C. (2016), ‘Gender diversity and firm performance: Evidence from Dutch and Danish boardrooms’, The International Journal of Human Resource Management 27(15), 1777–1790.

Matsa, D. A. & Miller, A. R. (2013), ‘A female style in corporate leadership? Evidence from quotas’, American Economic Journal: Applied Economics 5(3), 136–69.

Medland, D. (2004), ‘Small steps for womankind’, Corporate Board Member Europe, Winter .

Rose, C. (2007), ‘Does female board representation influence firm performance? The Dan-ish evidence’, Corporate Governance: An International Review 15(2), 404–413.

Sabatier, M. (2015), ‘A womens boom in the boardroom: Effects on performance?’, Applied Economics 47(26), 2717–2727.

Sila, V., Gonzalez, A. & Hagendorff, J. (2016), ‘Women on board: Does boardroom gender diversity affect firm risk?’, Journal of Corporate Finance 36, 26–53.

Terjesen, S., Couto, E. B. & Francisco, P. M. (2016), ‘Does the presence of independent and female directors impact firm performance? A multi-country study of board diversity’, Journal of Management & Governance 20(3), 447–483.

Appendix

Tables

Table 1: Variables and their description

Variable Abbreviation Definition

Tobin’s Q Q The market value of equity plus liquidating

value firm’s outstanding preferred stock plus value short term liabilities and book value firm’s long term debt divided by the book value of the total assets

Return on assets ROA The net income of a firm divided by the book

value of the total assets

Female fraction FF Percentage of women on board of directors,

that is both management board and super-visory board

Two two Dummy variable indicating the firm has a

two tier board

CEO CEO Dummy variable indicating a male CEO

Board size DIR Total number of directors on board of

direc-tors, that is on both management board and supervisory board

Dummy size dsize Dummy variable for the size of the firm

fol-lowing the criteria of the Netherlands Cham-ber of Commerce

Age AGE Number of years since incorporation firm

Time time Dummy variable indicating the year 2017

Management board MB Number of directors on management board

Supervisory board SB Number of directors on supervisory board

Men on other boards FMOB Fraction of male board members who have

Table 2: Descriptive statistics

Variable Mean N Mean (2012) N (2012) Mean (2017) N (2017)

Table 4: Female fraction and perfor-mance measures FF (categorized) Q ROA 1 1.423 −3.445 (2.027) (18.816) 2 1.471 2.782 (1.571) (13.500) 3 2.087 4.854 (4.221) (10.006)

Table 5: Two-tier board structure and performance measures two Q ROA 0 2.455 1.096 (5.086) (15.022) 1 1.423 0.265 (1.787) (16.157)

Table 6: Board size and performance

measures

DIR (categorized) Q ROA

0 1.998 −1.204 (4.665) (18.720) 1 1.615 1.023 (2.190) (13.481) 2 1.413 1.224 (1.232) (15.613)

Table 7: Firm size and performance mea-sures dsize Q ROA 0 1.198 −8.235 (1.274) (21.350) 1 1.801 3.293 (3.154) (12.588)

Table 8: Years since incorporation and performance measures AGE (categorized) multicolumn1cQ ROA 0 1.764 −5.240 (2.367) (25.806) 1 1.700 0.495 (2.320) (10.820) 2 1.915 2.004 (4.101) (12.736) 3 1.262 4.767 (1.406) (9.574)

Table 9: Time dummy and performance measures time Q ROA 0 1.240 1.830 (1.700) (14.939) 1 1.937 −0.014 (3.355) (16.089)

Table 10: Size management board and performance measures MB (categorized) Q ROA 0 1.568 2.173 (3.052) (14.055) 1 1.302 0.044 (1.029) (18.553) 2 1.481 1.036 (1.768) (14.802)

Table 12: Regression with total board of directors Variable Q ROA FF 0.5638 8.5857 (1.6473) (7.5062) two −0.4742 −2.3729 (0.6273) (2.4796) CEO 0.7863 −1.4393 (0.5408) (3.4258) DIR −0.1149 −0.0904 (0.0916) (0.3200) dsize 0.8322 10.5318 ∗ ∗∗ (0.5603) (3.5673) AGE −0.0033 0.0418 (0.0025) (0.0186) time 0.6947 ∗ ∗∗ −0.4963 (0.2206) (1.4879) Constant 1.2444 −7.5407∗ (0.6751) (4.5570)

Notes: Clustered robust standard errors in parentheses. ***p < 0.01, **p < 0.05, *p < 0.10

Table 13: Regression with management board and su-pervisory board Variable Q ROA FF −2.1576 ∗ ∗ 16.3293∗ (1.0656) (9.6833) CEO −0.0367 −0.4609 (0.0.3745) (4.2164) MB −0.0411 0.0929 (0.0459) (0.2906) SB 0.1057 −0.2349 (0.0874) (0.6912) dsize −0.6325 12.5241 ∗ ∗∗ (0.9263) (4.0768) AGE −0.0026 0.0362∗ (0.0030) (0.0212) time 0.6946 ∗ ∗∗ −1.1046 (0.1659) (1.8545) Constant 1.5816 ∗ ∗ −12.7236 ∗ ∗ (0.7002) (6.0275)

Table 14: Instrumental variables regression with total board of directors Variable Q FF ROA FF FF −12.9242 45.2762 (12.3755) (31.1294) two −1.2236 −0.0432∗ −1.3205 −0.0266 (1.3380) (0.0221) (2.5356) (0.0188) CEO −1.6684 −0.1629 ∗ ∗∗ 4.1453 −0.1611 ∗ ∗∗ (2.0251) (0.0450) (6.1564) (0.0422) DIR −0.0541 0.0043∗ −0.4069 0.0042 ∗ ∗ (0.0759) (0.0024) (0.3268) (0.0020) dsize 1.7166 0.0486 ∗ ∗ 9.8091 ∗ ∗∗ 0.0456 ∗ ∗ (1.1734) (0.0202) (3.6864) (0.0177) AGE 0.0022 0.0004 ∗ ∗∗ 0.0201 0.0004 ∗ ∗∗ (0.0058) (0.0.0001) (0.0197) (0.0001) time 1.6300 0.0647 ∗ ∗∗ −3.3803 0.0637 ∗ ∗∗ (1.0006) (0.0111) (2.5160) (0.0103) FMOB 0.0944 ∗ ∗ 0.1519 ∗ ∗∗ (0.0387) (0.0377) Constant 4.0090 0.1722 ∗ ∗∗ −22.9218∗ 0.1432 ∗ ∗∗ (3.1949) (0.0563) (7.0981) (0.0483) F-statistic 1.56 51.99 3.45 76.25

Notes: Clustered robust standard errors in parentheses. ***p < 0.01, **p < 0.05, *p < 0.10

Table 15: Instrumental variables regression with management board and supervisory board

Variable Q FF ROA FF FF −12.2372 90.6684 (12.3724) (75.3820) CEO −1.3752 −0.1458 ∗ ∗∗ 12.6873 −0.1759 ∗ ∗∗ (1.7502) (0.0472) (13.8848) (0.0420) MB −0.0960 −0.0026 0.1845 −0.0013 (0.0823) (0.0032) (0.3370) (0.0031) SB 0.2147 0.0155 ∗ ∗∗ −1.1246 0.0118 ∗ ∗∗ (0.2422) (0.0040) (1.1143) (0.0035) dsize 0.8094 0.0193 10.6624 ∗ ∗ 0.0160 (0.5155) (0.0216) (4.8393) (0.0.0224) AGE 0.0021 0.0004 ∗ ∗∗ 0.0007 0.0005 ∗ ∗∗ (0.0062) (0.0.0001) (0.0456) (0.0002) time 1.5576∗ 0.0612 ∗ ∗∗ −7.3154 0.0789 ∗ ∗∗ (0.8246) (0.0123) (6.0638) (0.0113) FMOB 0.0749∗ 0.0841∗ (0.0387) (0.0377) Constant 1.8349 0.1010∗ −23.3506∗ 0.1357 ∗ ∗∗ (1.3991) (0.0563) (12.8885) (0.0480) F-statistic 1.80 53.63 2.49 43.78

Figures

Fraction of female boardmembers(categorical)

Figure 1: Boxplot Tobin’s Q against fraction of female board members (categorized)

-1 0 -5 0 5 10 R e tu rn O n Asse ts 1 2 3

Fraction of female boardmembers (categorical)

Figure 2: Boxplot ROA against fraction of female board members (categorized)

0 1 2 3 4 T o b in 's Q 0 1 Two-tier board

Figure 3: Boxplot Tobin’s Q against dummy indicating two-tier board

-1 0 -5 0 5 10 R e tu rn O n Asse ts 0 1 Two-tier board

0 1 2 3 4 T o b in 's Q 0 1 2

Number of directors (categorical)

Figure 5: Boxplot Tobin’s Q against number of directors (categorized) -1 0 -5 0 5 10 R e tu rn O n Asse ts 0 1 2

Number of directors (categorical)

Figure 6: Boxplot ROA against number of directors (categorized) 0 1 2 3 4 T o b in 's Q 0 1 Size

Figure 7: Boxplot Tobin’s Q against size dummy -1 0 -5 0 5 10 R e tu rn O n Asse ts 0 1 Size

Figure 8: Boxplot ROA against size dummy

0 1 2 3 4 T o b in 's Q 0 1 2 3 Age (categorical)

Figure 9: Boxplot Tobin’s Q against age (categorized) -1 0 -5 0 5 10 R e tu rn O n Asse ts 0 1 2 3 Age (categorical)

0 1 2 3 4 T o b in 's Q 0 1 Time

Figure 11: Boxplot Tobin’s Q against time dummy -1 0 -5 0 5 10 R e tu rn O n Asse ts 0 1 Time

Figure 12: Boxplot ROA against time