• No results found

When can women make a difference?

N/A
N/A
Protected

Academic year: 2021

Share "When can women make a difference?"

Copied!
47
0
0

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

Hele tekst

(1)

When can women make a difference?

Abstract: The aim of this thesis is to find out whether there is a correlation between the share of female board members and the performance of the firm. In addition to this, there will be analyzed to what extent this performance will be dependent on the industry. In order to answer this, three hypotheses are composed: (H1) The company results do not depend on the share of women in the board of directors, (H2) For all companies, women are most likely to be in boards of companies in the services industry and (H3) There is a correlation between the percentage of women on the board and the results across different industries. To test these hypotheses, several models and a sample with 1122 U.S. firms are used. The following conclusions are drawn from this: (1) There is no effect from the share of women on the board on the performance of the firm, (2) There cannot be concluded that there are significantly more women on board of the service industry and (3) The percentage of women on the board does not yield different results across different industries.

Following this, the conclusion of this thesis is that there is no correlation between the share of female board members and the performance of the firm and there is also no correlation

between the fraction of females and the difference in industries.

Keywords: Board Diversity, Firm Performance, Industry

Name: Sarah Fathi

Student number: 10060618

Date: September 5th 2014

Bachelor thesis: Economie en Bedrijfskunde

Specialisatie Financiering en Organisatie Supervisor: Ms. L. Rosendahl Huber

(2)

2 Table of Contents 1. INTRODUCTION ... 4 2. LITERATURE REVIEW ... 6 2.1INTRODUCTION HYPOTHESIS 1 ... 6 2.2TRADE-OFF ... 6

2.3FOUR FUNCTIONS OF THE BOARD ... 6

2.3.1 Resource Dependence Theory ... 7

2.3.2 Human Capital Theory ... 8

2.3.3 Agency Theory ... 9

2.3.4 Social Psychological Theory ... 9

2.3.5 Hypothesis 1 ... 10

2.4INTRODUCTION HYPOTHESES 2 AND 3... 11

2.5DIVERSITY AND INDUSTRY ... 11

3. EMPIRICAL EVIDENCE ... 13

3.1INTRODUCTION ... 13

3.2RESOURCE DEPENDENCE THEORY AND HUMAN CAPITAL THEORY ... 13

3.3AGENCY THEORY ... 15

3.4SOCIAL PSYCHOLOGICAL THEORY ... 16

3.5DIVERSITY AND INDUSTRY ... 17

3.5.1 Conclusion ... 20

4. DATA & RESEARCH METHODOLOGY ... 21

4.1INTRODUCTION ... 21

4.2DATA COLLECTION ... 21

4.3SAMPLE AND VARIABLES ... 22

4.3.1 Variables Descriptions... 22

4.3.1.1 Dependent Variables ... 23

4.3.1.2 Independent Variables ... 24

4.3.1.3 Descriptive Statistics and Correlations ... 24

4.4MODELS ... 27

5. METHOD AND RESULTS ... 29

5.1INTRODUCTION ... 29

5.2MAIN RESULTS ... 29

6. CONCLUSION AND DISCUSSION ... 36

7. METHODOLOGICAL DIFFICULTIES ... 39

REFERENCES ... 41

(3)

3

Tables and Figures

TABLE 1 ... 18

TABLE 2 ... 19

TABLE 3 ... 20

TABLE 4: REDUCTION FULL SAMPLE TO FINAL SAMPLE ... 22

TABLE 5: OVERVIEW OF VARIABLES ... 23

TABLE 6: DESCRIPTIVE STATISTICS ... 25

TABLE 7: REGRESSION MODEL 1 H1 ... 29

TABLE 8: REGRESSION MODEL 2 H2 ... 31

TABLE 9: FIRST 4 “SIMPLE” REGRESSIONS H3 ... 32

TABLE 10: FIRST 4 “COMPLETE” REGRESSIONS H3 ... 34

FIGURE 1: THE DISTRIBUTION OF FIRMS ACROSS THE INDUSTRIES ... 26

(4)

4

1. Introduction

Even though 57.7% of the working population in the U.S. consists of women (U.S. department of labor, 2012), there are still very few women in executive positions. In the Fortune 500 (list of 500 largest companies in the U.S., based on revenue) there are only 23 female CEO’s (CNN, 2013).

With 29.6%, there are slightly less highly educated (bachelor’s degree or higher) women, than the 30.3% men that are highly educated (U.S. Census Bureau, 2010). This ratio is not the same as the CEO percentage of 4.6% according to the Fortune 500. Neither comes it close to the 16.9% women on boards in the U.S. (Catalyst, 2014).

There are many studies on whether gender diversity among the board of directors affects the firm performance, the results differ from a positive effect (Adams and Ferreira, 2009) to no effect at all (Carter et al., 2010). Why are there so few women in executive positions? Differences in ability only can account for part of the low ratio (16.9%) and often it is said that discrimination and preferences explain the rest. According to Niederle and Vestlund (2007), there are two other reasons for this difference. First, in competitions where gender is mixed, men perform, on average, better than women. Second, men and women are not equally competitive in the same situation. Men are two times as likely to get in the tournament, even though there were no differences in gender performance.

This difference is caused by 2 factors; men are more overconfident than women, and men differ in their performance preferences in competitions from women (Niederle et al., 2007).

The aim of this thesis is to find out if, for the sample used in this study, it makes a difference whether there are women on the board. To find out if the share of women on a board affect company performance, I will compare the outcomes of different firms in order to answer the first sub research question:

Is there a positive correlation between the share of female board members and company performance?

(5)

5 Additionally, I will divide my sample into different industry divisions to see if the results are different across industries. I would like to examine whether there is a correlation between the performance of the firm and its industry.

My second sub research question therefore is:

To what extent will this performance be dependent on the industry?

The structure of the rest of this thesis is as follows. Section 2 consists a review of the theory from existing literature, which leads to the hypotheses, that will be tested later on in this thesis. Section 3 gives an overview of the empirical evidence of the existing literature. In section 4 the research methodology is covered, which includes the data collection, an

explanation of the variables used, the descriptive statistics and the models that are used to test the hypotheses. The results are discussed in section 5. Section 6 covers the conclusion and the discussion, followed by the methodological difficulties in section 7.

(6)

6

2. Literature Review

2.1 Introduction Hypothesis 1

This section starts with part 2.2, in which two articles state that whether diversity has a positive effect on firm performance, depends on the trade-off between benefits and costs, which are different for all circumstances. However, parts 2.3.1 up till 2.3.4 discuss four theories (partly based on the functions of the board discussed in part 2.3) that actually try to provide an answer to the question whether firm performance and board diversity are

associated with each other in general. Following these theories, a hypothesis is formed in 2.3.5.

2.2 Trade-off

Men and women function differently, but does this imply specific benefits or costs for the company and to what extent does a diverse board of directors influence the performance of the firm? According to Hamilton et al. (2003) there is a trade-off between the benefits of having a diverse board which includes the potentially related pool of diverse skills and knowledge, and the costs of communication and coordination. Which of the two is the most important depends on the circumstances (Hoogendoorn et al., 2013). In the existing literature there is no theory that forecasts whether there is a relationship between firm performance and the percentage of females in the board of directors.

2.3 Four Functions of the Board

However Carter et al. (2010) state that it is possible to get an answer that holds in general and does not depend on circumstances as mentioned in the previous part. The starting point for getting this answer begins with the most important functions of the board of directors. These include controlling and monitoring managers, providing counsel and information to managers, monitoring performance with appropriate laws and function as link between the firm and the external environment (for example other firms) (Monks et al., 2004). These functions are the

(7)

7 basis for several theories of different fields that will be discussed in the next parts of this section. None of this theories will predict the relationship between board diversity and firm performance directly by itself, but the combination of them will give a theoretical basis for the first hypothesis.

2.3.1 Resource Dependence Theory

The first theory is the resource dependence theory, which states that if there are different types (meaning various ages, ethics and both genders) of directors in a board it will provide different profitable resources to the firm. Pfeffer and Salancik (1978) claim that boards help to link the firm to other the external environment (a function as mentioned before). They state that this external link has four benefits: (1) supply of resources, for example expertise and information, (2) creation of communication channels which matter for the firm, (3) supply of obligations of support from other firms or important organizations in the external

environment, (4) creation of legality in the external environment for the firm (Pfeffer and Salancik, 1978). Cannella et al. (2000) extend these benefits into a classification system for various types of directors who provide different resources to a firm: business experts, community influentials, insiders and support specialists. This makes that if there are more various types of directors, so the board is more diverse, it will accommodate more valuable resources, which should lead to improved firm performance (Cannella et al., 2000).

However, the research of this thesis is about the diversity of gender and up and until now the literature was about types of men and functions of the board. Yet, directors of ethnic minorities and female directors also contribute by bringing in different resources and benefits. Cannella et al. (2002) go further on this and find that African-American female directors are less likely to be business experts, compared to African-American male directors. Again both are less likely to be experts compared to Caucasian female directors and Caucasian male directors are most likely to be business experts. In de resource dependence theory ethnicity and gender are two different components because both groups differ in human capital and backgrounds, this leads to an ability of addressing external links. Diverse directors potentially improve the information from the board meant for the managers (Cannella et al., 2002). Through the difference in gender and ethnicity it is likely that the information provided by the board is unique and helps the managers making better decisions. More diverse directors means more access to important subdivisions in the outside environment. This results in the

(8)

8 fact, that firms with a more diverse board are able to bring in more talent. Other than that a more diverse board gives a good and important signal to the product market as well as the labor market.

So concluding, the resource dependency theory states that the board is a link to external environment, which has its benefits, that can be extended into a classification of different types of directors, who all provide different resources for the firm. However, in addition to this, women and ethnic minorities accommodate also various resources and benefits to the board. Therefore a diverse board, with not only different “types”, but also diversity in gender, will have a positive effect on the information for managers, who will make, in their turn, better decisions that possibly cause the firm to perform better than with a less diverse board.

2.3.2 Human Capital Theory

The second theory is the human capital theory, which states that skills, education and experience can be used in favor of an organization (Becker, 1964) and this theory also includes the given that the difference in gender makes that directors have unique human capital (Terjesen et al., 2009). The question that comes up, given that women have human capital that is unique, is that they miss the “right” kind of human capital in order to be a board member (Terjesen et al., 2009). Terjesen et al. (2009) suggest that, following evidence on specific female human capital, women are as suitable as men for being a board member by meeting some important qualities, such as education level, but it is also less likely that women have experience with it. Following this, Cannella et al. (2002) and Peterson and Philpot (2007) state that female board members take on different roles on the boards and this is probably due to this unique human capital. Additionally, there is also an influence of the contingency theory, which means that there is not simply one correct way of leading a company and making decisions, instead the correct decision depends on the internal and external circumstances. This is also relevant here, because human capital may be useful in a certain organization in a specific point of time, but might not be useful in other circumstances (Lawrence et al., 1967).

So concluding, the human capital theory predicts that the board performance is affected by diversity of the board due to diverse and unique human capital. From a financial performance point of view, the effect could be positive or negative.

(9)

9

2.3.3 Agency Theory

The agency theory is the third theory and it is a presumption that explains the relationship between the principal, in this case the board, and the agent, here the managers. There are two problems that the theory addresses. First, the board might not have the same goals as the managers and the board is unable to oversee what the managers are doing. Second, the board and the managers might have a different attitude towards risk. Therefore, the function of controlling and monitoring managers by the board is an essential concept of this theory (Jensen et al., 1976). Carter et al. (2003) agree to this and suggest that a more diverse board is better at monitoring managers, because independency is increased by the diversity of the board. This independence is linked to monitoring, because people with a different gender or ethnic background might ask different questions than the ones directors with more “standard” backgrounds would ask. The board independence is important for the board to operate in the best possible way for the shareholders, therefore it may possibly lead to better performances of the firm (Carter et al., 2003). However Carter et al., (2003) go further and state that the agency theory doesn’t obtain an obvious prediction of the relationship between financial performance and the diversity of the board.

In the aggregate, the agency theory doesn’t accommodate the same strong support for the financial gains of a more diverse board as the resource dependency theory does, but it doesn’t preclude the possible benefits of board diversity on the performance.

2.3.4 Social Psychological Theory

The last theory is the social psychological theory, which means that demographic differences reduce the probability of social bonds between groups and that social hurdles lower the chance that viewpoints of minorities will affect the decisions of a group. Through this, the degree to which the directors of a minority (for example women) can influence the board decisions in a successful way. The concept of social psychology of minorities is deduced from the social impact theory (Westphal and Milton, 2000). This is a theory which forecasts that during group decisions, individuals who are in the majority, are more likely to have a disproportionate amount of influence. So it is possible that in a more diverse board the

directors of the minority don’t influence the decisions, because of the internal dynamics in the group. On the other hand, Westphal and Milton, 2000 also suggests that group members who

(10)

10 are part of a minority may stimulate different thinking while making decisions. However, Campbell and Minguez-Vera (2008) criticized this and concluded that a more diverse board leads to a greater diversity of options and critical thinking, which makes the taking of decisions more time consuming and less effective.

In the end it seems that in a diverse board, the minority (for example a share of women) has no effect on the board decisions, which could lead to the fact that this minority has no effect on the firm performance and therefore diversity of the board might not have an influence on the firm performance. But, the theory and established facts on the dynamics of groups, suggest that the effect of the diversity of the board on firm performance can be both positive and negative.

2.3.5 Hypothesis 1

Following these four theories where the resource dependency theory states that a diverse board, with diversity in gender, will have a positive effect on the information for managers. Who will make, in their turn, better decisions that possibly cause the firm to perform better than with a less diverse board. The human capital theory predicts that the board performance is affected by diversity of the board due to diverse and unique human capital. From a financial performance point of view, the effect could be positive or negative. The agency theory

doesn’t accommodate the same strong support for the financial gains of a more diverse board as the resource dependency theory does, but it doesn’t preclude the possible benefits of board diversity on the firm performance. And the social psychological theory concludes that theory and established facts on the dynamics of groups, suggest that the effect of the diversity of the board on firm performance can be both positive and negative.

So all theories, except the resource dependency theory can’t conclude whether a more diverse board has a positive or a negative effect on the firm performance Therefore the first

hypothesis is:

H1: The company results do not depend on the share of women in the board of directors.

(11)

11

2.4 Introduction Hypotheses 2 and 3

The nature of the industry is likely to have an effect on the amount of women in the board (Cannella et al., 2007). In the next part, part 2.5, the effect of the industry on the diversity of board will be discussed, this leads to the second hypothesis. This effect will be extended by taking along the firm performance, which will lead to the third hypothesis.

2.5 Diversity and Industry

In 2009 there were 15.2 % women among boards in the U.S., 5 years later, in 2014, it is 16.9%. So the percentage women is increasing, but still not uniform represented across firms. Hyland and Marcellino (2002), conclude for example that the fraction of women on the board of directors correlates with the board size and the industry of the firm. But the industry of the firm might also affect to which extent the board is diversified (Cheng et al., 2007). This happens for instance, with firms in the finance industry, which still have boards that are “too big, too old and too male” (Engen, 2002). It is unlikely that this will change soon, because a study of Gilpatrick (2002) shows that most boards consist of directors that are middle aged and up and who have been directors in other firms, but in the same industry. Which might be caused by the fact that directors, with communal experience in an industry, probably have similar beliefs about opportunities and threats which influence the decision making process and leads to a lower probability of diversity of the boards in the finance industry than in other industries (Spender, 1989). In contrast to this, the services industry where firms are more likely to take in female directors on the board than firms in other industries (Harrigan, 1981). Additionally, having a larger number of women working in an industry tend to increase representation of women on a board of directors (Cannella et al., 2007). Most of the highly educated women (i.e. with a bachelor degree or higher) work in the services industry (Bureau of Labor Statistics, 2011). Because it is most likely that the women on the boards are highly educated (Carter et al., 2003), this means that given the previous two findings, there would be a higher possibility that there are more women on the board of the services industry than in other industries. Therefore the second hypothesis is:

For all companies, women are most likely to be in boards of companies in the services industry.

(12)

12 Given the assumption of Cannella et al. (2007) that having a larger number of women

working in an industry tend to increase representation of women on a board of directors. It would be interesting to see if there is a pattern in some industries, while there isn’t one in others. And also to see whether there will be a correlation between the results and the pattern. Following this, the third hypothesis is:

There is a correlation between the percentage of women on the board and the results across different industries.

(13)

13

3. Empirical Evidence

3.1 Introduction

In this section the empirical evidence of the literature from section 2 will be discussed. Parts 3.2 up to 3.4 cover theories, with in part 3.2 the resource dependency and the human capital theory together because the evidence is similar to one another. Part 3.3 will cover the empirical evidence of the agency theory and in part 3.4 the evidence of the social

psychological theory will be discussed. Part 3.5 discusses the evidence on the literature of the link between diversity of the board and the industry of the firm, as well as the firm

performance.

3.2 Resource Dependence Theory and Human Capital Theory

The resource dependence theory states that if there are different types (meaning various ages, ethics and both genders) of directors in a board it will provide different profitable resources to the firm. According to the literature, this theory concludes that a diverse board, with diversity in gender, will have a positive effect on the information for managers, who will make, in their turn, better decisions that possibly cause the firm to perform better than with a less diverse board. The second theory is the human capital theory, which states that skills, education and experience can be used in favor of an organization (Becker, 1964) and this theory also includes the given that the difference in gender makes that directors have unique human capital (Terjesen et al., 2009). This theory predicts, based on the literature, that the board performance is affected by diversity of the board due to diverse and unique human capital. From a financial performance point of view, the effect could be positive or negative. Following this, the empirical evidence of these theories will be discussed.

So these two theories don’t specifically forecast a clear relation between the diversity of the board and the performance of the firm. Additionally, based on these theories, the type of diversity should be meaningful. There can be predicted that women and ethnic minorities have different effects on the board and therefore on the firm performance, due to the different human capital and external environment (Carter et al., 2010). This is for example found by Brammer et al. (2007), who conclude that, by analyzing the diversity of gender and ethnics of

(14)

14 a U.K. companies sample, “board diversity is influenced by a firm’s external business

environment and particularly an imperative to reflect corresponding diversity among its customers”. Brammer et al. (2007) contrive a meaningful cross sector variation in the diversity of gender across industries. In addition to this, human capital has another effect, in the literature part was stated that Cannella et al. (2002) and Peterson and Philpot (2007) concluded that “female board members take on different roles on the boards and this is

probably due to this unique human capital”. The evidence generated by Cannella et al. (2002), Peterson and Philpot (2007) and Peterson et al. (2007) support the opinion that female

directors and also directors of an ethnic minority possibly have diverse roles or functions on the board. They come to this conclusion by doing various analyses. Cannella et al. (2002), analyze a sample of the directors of the Fortune 1000 from 1993 and divide them into

different groups of the females (7.53%) and African-American (1.45%) directors. They don’t have an interest in the cross sectional analysis, but rather in the choices the directors made during their careers as well as the previous director functions they fulfilled. Two of the hypotheses are “A greater percentage of female and African-American directors will hold advanced degrees than will white male directors.” and “Female and African-American directors will more likely serve on multiple boards than will white male directors.” (Cannella et al., 2002). By doing a Chi-square analysis across the groups, they come to the conclusion that indeed a higher rate of women and African-Americans have an higher education than white males. They also find that women and African-Americans, once they were for the first time on a board, are more likely to join a second board. Those outcomes are in themselves promising, because if women are higher educated and on more boards at the same time, it means that there might be a higher chance of a higher share of women amongst boards in general. However, this does not give an answer to the first sub research question. Combining this result of Cannella et al. (2002) with the research of Peterson and Philpot (2007), who in their turn examine the firms from the Fortune 500 firms and do analyze this sample with a logistic regression model. The model controls for the characteristics of the directors and the firm, the resource dependence roles of the directors and the interaction between the

characteristics of the director and the gender. They find that male directors are more likely to be on executive boards, but that female directors are more likely to be on so called “public affairs committees”, which are less important (Peterson and Philpot, 2007).

These two studies lead to the conclusion that on the one hand, women are possibly more likely to be on boards in general, because they have a higher education and are on more boards at the same time. This means that if the fact, that diverse directors potentially improve

(15)

15 the information from the board meant for the managers (Cannella et al., 2002) as stated in the literature part, is true, then it is more likely that this leads to more improved information for the managers and therefore, better decisions and maybe better results for the firm. However, on the other hand, there is concluded that women are more likely to be on boards that are less “important”, this might have to do with the difference in human capital between men and women.

The conclusion of this part and taking along the first sub research question is, it can be predicted that women and ethnic minorities have different effects on the board and therefore on the firm performance, due to the different human capital and external environment. And that a more diverse board might improve the performance due to improved information for managers. However, it cannot be concluded that there is a explicit positive relation between the share of female board members and company performance.

3.3 Agency Theory

The agency theory is a presumption that explains the relationship between the principal, in this case the board, and the agent, here the managers. The conclusion from the literature part was that the agency theory doesn’t accommodate the same strong support for the financial gains of a more diverse board as the resource dependency theory does, but it doesn’t preclude the possible benefits of board diversity. Next, the empirical evidence of this theory will be discussed.

The agency theory presents the possibility that if directors are more diverse, they are better management monitors. According to the theory, the kind of relation between the diversity of the board and the performance of the firm is not quite clear. Adams and Ferreira (2009) form the basis of this theory with their article and discover that boards with more gender diversity dedicate more effort to the monitoring of managers. But they also state that there is a negative relation between the rate of female board members and the Tobin’s Q, which follows from an U.S. firms analysis. So in the end, more and persistent monitors might be positive as well as negative in terms of effect on the results (Adams and Ferreira, 2009). The article of Carter et al. (2003) is also a fundamental to the agency theory, as was shown in the literature part of this thesis and Carter et al. (2003) come up with an analysis of U.S. firms from which can be concluded that there is a positive relation between the gender diversity of the board and the Tobin’s Q, which is the opposite of the result of Adams and Ferreira (2009).

(16)

16 In conclusion, boards with more gender diversity dedicate more effort to the

monitoring of managers. But there is also a negative relation between the rate of female board members and the Tobin’s Q. So in the end, more and persistent monitors might have a

positive as well as negative effect on the results. However, Carter et al. (2003) did find a positive correlation between the diversity of the board and the performance of the firm. Therefore recalling the first sub research question, it can be concluded that there might be a positive relation between the share of female board members and company performance.

3.4 Social Psychological Theory

The last theory is the social psychological theory, which means that demographic differences reduce the probability of social bonds between groups and that social hurdles lower the chance that viewpoints of minorities will affect the decisions of a group. The conclusion drawn from the discussed literature was that in a diverse board, the minority (for example a share of women) has no effect on the board decisions, which could lead to the fact that this minority has no effect on the firm performance and therefore diversity of the board might not have an influence on the firm performance. However, the theory and established facts on the dynamics of groups, suggest that the effect of the diversity of the board on firm performance can be both positive and negative. Following this conclusion, the empirical evidence of this theory will be discussed next.

The social psychological theory indicates that diverse directors might not have an hold on the decisions which the board makes and this is caused by the internal dynamics in the group (in this case the board). Also, more diversity of the board might encourage innovative and creative thoughts, but the making of decisions can be slower and might be the cause of conflict between the directors (Carter et al., 2010). Westphal and Milton (2000) did a study that had a different outcome. They got data by means of a survey amongst directors of middle and large companies in the U.S.. In their study their dependent variable is the influence of board members on the decisions that the board makes and they make use of an OLS multiple regression analysis. They estimate the effects of being in a minority to education, race, functional and industry backgrounds and the gender. The outcome is that if directors who are in a minority and have been in a minority position before, can raise their power to exert influence on the focus of the board. Whilst if the directors were previously majority directors and are now minority directors, they are less likely to have any influence. The overall result of the study of Westphal and Milton (2000) is that demographic minorities can reduce the

(17)

17 possibility of having almost no influence on the board. This is possible if they have

experienced being part of the minority directors before or if they have a social connection with other board members (of the majority) and in that way form a majority. Another

fundamental article for this theory is from Campbell and Minguez-Vera (2008). They examine the relation between the diversity of gender and the performance of the firm by using a

sample of Spanish companies. They conclude that the gender diversity of the board has a positive effect on the value of the firm, which is measured by the Tobin’s Q.

So the conclusion of this empirical part of the theory and its main articles is, that more diversity of the board might encourage innovative and creative thoughts, but the making of decisions can be slower and might be the cause of conflict between the directors.

Furthermore, demographic minorities can reduce the possibility of having almost no influence on the board. This might mean that even though the decision process is slower and more difficult, if the minority board members (for example a share of women) can get through to the majority (so create social bonds or overcome social hurdles), then the minority members might give the board more creative and innovative ideas that possibly help improve the decisions eventually made and therefore might result in better results by the performance. However, this seems unlikely to happen because, as stated in the literature, the internal dynamics are presumably always an element of group decisions. The only study that had a positive conclusion about the relation between gender diversity of the board and the value of the firm was for Spanish firms and this thesis focuses mainly on U.S. firms. Therefore the answer of the first sub research question based on this theory is that it cannot be concluded that there is a positive relation between the share of female board members and company performance.

3.5 Diversity and Industry

Hyland and Marcellino (2002), conclude in the literature part that the fraction of women on the board of directors correlates with the board size and the industry of the firm. Now the empirical evidence will be discussed.

Hyland and Marcellino (2002) use data from the top 100 public companies in the U.S. and one of their hypothesis are formulated as “the overall percentage of board seats filled by women will be higher in the industries of finance/insurance/real estate and services than in other industries”. They estimate this through OLS regression, where the number of female

(18)

18 directors is the dependent variable. Table 1 shows that this hypothesis is only partly

supported, because even though the effect for services is positive, it is not significant (Hyland and Marcellino, 2002). Overall they conclude that the fraction of women on the board of directors correlates with its size and the industry of the firm.

Table 1

Source: table from Hyland and Marcellino (2002, p. 29)

According to the literature part, the industry of the firm might affect to which extent the board is diversified (Cheng et al., 2007). Following next will be the empirical evidence on which this is based.

Cheng et al. (2007) use a sample of the top 100 firms in Australia and they want to find out if there is a correlation between type of industry and the independence and diversity of the board of directors. They distinguish 4 types of industry, “consumer services/products”, “financials”, “materials and industrials” and “others”. Hyland and Marcellino (2002), did linked the type of industry with the diversity of the board, but Cheng et al. (2007) state that there is no significant relation between gender diversity and the type of industry. As can be seen in table 2, they do find a relation between the type of industry and the age and also between the industry and the independence of directors (Cheng et al. 2007).

(19)

19 Table 2

Source: table from Cheng et al. (2007, p. 203)

In the literature was stated that having a larger number of women working in an industry tend to increase representation of women on a board of directors (Cannella et al., 2007). The empirical evidence for this will follow next.

For the research whether having more female employees working in an industry is positively associated with the performance of the firm, Cannella et al., (2007) used data of the Bureau of Labor Statistics to find out how many women there were working in each industry division (2 SIC digits). They find that women are mostly working in the health and financial services and are underrepresented in construction and utilities. For the analyses they use a population averaged model and “generalized estimating equations (GEE)” which adjusts for correlation between the fixed effects and independent variables (this correlation was

discovered because the sample data failed on the Hausman test).

First they run multiple logistic models, which means that they separate boards with one woman from boards with more women in order to look for robustness. But the results were significantly the same, so therefore they only use the logistic regression. Table 3 shows the results of the analyses, instead of coefficients there is made use of odd ratios. So an odd ratio of 1.00 means that there is no effect and a ratio larger than 1.00 means that if the independent variable increases, the likelihood of the dependent variable also increases.

The result of this regression is that if more women work in a certain industry, that there are more women represent the board in the same industry.

(20)

20 Table 3

Source: table from Cannella et al. (2007, p. 948)

3.5.1 Conclusion

The results of this part of the empirical evidence are that if more women work in a certain industry, that there are more women represent the board in the same industry. On top of that, the fraction of women on the board of directors correlates with the size of the board and the industry of the firm. Opposite to this, there is also found that there is no significant relation between gender diversity and the type of industry, but there is a relation between the type of industry and the age and also between the industry and the independence of directors. This was however a study with Australian firms, while this thesis is about U.S. firms, therefore this negative result is taken into account, but because the positive result came from a study with U.S data, the conclusion of this part will be that there might be a correlation between the diversity of the board and the type of industry.

(21)

21

4. Data & Research Methodology

4.1 Introduction

In this section the process from data collection to the model composition is discussed. First, in part 4.2, I’ll start with the method was used to collect my data, which databases were inquired and which companies were chosen and in what way it leads to the final sample. Second, part 4.3 describes how many observations are in the final sample and why some of them were removed from the sample. This includes the discussion of the variables that I use and the descriptive statistics. Last, in part 4.4 I’ll explain the three models that I use to test the hypotheses.

4.2 Data Collection

To answer the first of my two research sub questions “Is there a positive correlation between the share of female board members and company performance?” Data on the gender of the CEO’s and board members of U.S. companies and the results of those companies across 2004 is needed. The reason for using U.S. companies is because, in contrast to other countries, this data is most easily accessible. I used the data of 2004, because the database contains more data from this year than years before or after, which makes that my sample is as large as possible. For the available U.S. companies WRDS’ database “Compustat Monthly Updates - Fundamentals Annual” was used to collect data on firm performance and other financial details. I collect my data about the gender of the CEO’s and board members from those companies from “Compustat Executive Compensation - Annual Compensation”.

For the second research sub question “To what extent will this performance be dependent on the industry?” I need, next to the same data on firm performance and other financial and gender information, also data on the industries of those U.S. companies. Therefore I use the SIC codes, which I also collect from the WRDS’ database “Compustat Monthly Updates - Fundamentals Annual”.

Part of the variables I need are just available in either one of the databases, including the sales, the fact that a CEO was male or female, the age and gender per board member and the industry SIC codes. However the majority of the data is manually merged, this applies to

(22)

22 the Tobin’s Q, the board size, the fraction female directors on a board and the average age per board. I will discuss all this variables later in detail and also how I computed them.

4.3 Sample and Variables

For the two databases I’ve got 2 different outputs, which are not the same because the

Financial sheet (from “Compustat Monthly Updates - Fundamentals Annual”) shows data per company and the Gender sheet (from “Compustat Executive Compensation - Annual

Compensation”) shows data per board member. Therefore the two sheets were combined manually into one which includes all data available of both sheets. This leads to a sample that exists of 3352 observations. The sample is reduced to 1165 observations due to missing data on the board size and therefore the number of females, fraction of females and the average age of the board. Then the number of observations becomes 1127 due to negative or zero data on the book value, which is caused by a negative or missing book value per share. Finally the sample is further reduced to 1122 observations due to negative or missing logsales. Table 4 shows this reduction.

Table 4: Reduction Full sample to final sample

Corrections Observations

Full Sample 3352

Missing Board Size -2187

Negative or Zero Bookvalue -38 Negative or Missing Log(sales) -5

Final Sample 1122

4.3.1 Variables Descriptions

To answer my research questions, I use three hypotheses; (H1) The company results do not depend on the share of women in the board of directors. (H2) For all companies, women are most likely to be in boards of companies in the services industry and (H3) There is a

correlation between the percentage of women on the board and the results across different industries. However, because I use almost the same variables for every hypothesis I will first

(23)

23 give an overview (Table 5) of all the variables I use and then I will explain them beginning with the dependent variables and then the independent variables.

Table 5: Overview of variables

Variable Name Hypothesis 1 Hypothesis 2 Hypothesis 3

Tobin´s Q + * + Fraction Female * + * Board Size * * LogSales * * Female CEO * Average Age * Agriculture 1 * * Mining * * Construction * * Manufacturing * * Transport 2 * * Wholesale 3 * * Retail Trade * * Finance 4 * * Services * * Administration 5 * * Dependent: +, Independent: *

1 = Agriculture, Forestry, Fishing, 2 = Transportation & Public Utilities, 3 = Wholesale Trade, 4 = Finance, Insurance, Real Estate,

5 = Public Administration

4.3.1.1 Dependent Variables

For firm performance I use the Tobin’s Q, because several studies from the literature review and empirical evidence part also use this measure (Adams and Ferreira, 2009; Campbell and Minguez, 2008; Carter et al., 2003). The Tobin’s Q is defined as “the ratio of the firm’s market value to its book value” (Adams and Ferreira, 2009). The market value is available in the WRDS database, but the book value is only per share, therefore I calculate the total book value by multiplying it with the total shares.

The fraction of females is calculated by dividing the number females on the board by the total board size. The number of females was available in the WRDS’s database, but the board size was not and so this was manually merged with the number of board members per company in order to get the fraction females per company.

(24)

24

4.3.1.2 Independent Variables

The board size consists of the total number of board members in a company and this is not available in the WRDS database and I had to count the number of board members of each company in my sample. This variable was used in this thesis because Hyland and Marcellino (2002), Adams and Ferreira (2009) and Cheng et al. (2007) use the board size to make sure the effects they find are due to gender diversity and not board size.

The logsales is simply the logarithm of the sales, which is available in the WRDS database. The logarithm of the sales is used as a proxy for firm size according to inter alia Adams and Ferreira (2009) and Carter et al. (2003).

Another independent variable that I use is whether a company has a female CEO, this is also a variable that is available in the database. I transferred the male/females output from the database to a dummy with a 0 if the CEO is male and a 1 if the CEO is female. This control variable is used in several studies I used in the literature and empirical evidence parts of this thesis (Carter et al., 2010 and Adams and Ferreira, 2009)

The average age is also calculated by hand, because the age of the individual directors was available but had to be averaged and transferred to match the correct financial results of the company from the other output. According to the study of Carter et al. (2003) this variable is an important control variable for a proper regression.

To indentify the different industries the WRDS database uses 4 digit SIC codes 1000 up to and including 9997. For the purpose of this study this was reduced to only 1 digit SIC codes (which is specified in Appendix A) and turned into dummies. Cheng et al. (2007) states that the industry of an firm influences the board diversity and therefore should be included in the model.

4.3.1.3 Descriptive Statistics and Correlations

Table 3 shows the descriptive statistics of the variables I use for testing the hypotheses. It shows for example that the average percentage females on the board is 6.36%, but there is a company with 83.33% women. The average board consists of almost 6 people, but the smallest board had only 2 members. This variable has quit a large standard deviation, which indicates that the data are not close to the mean. In my sample 1.69% of the companies has a female CEO and the average of the average age of the board is 61.16 years.

(25)

25 Looking at the descriptive statistics of other studies, Adams and Ferreira (2009) use a sample of 8,253 firms, which is significantly larger than the sample used for this thesis. This was also the largest sample that I could find in the literature I used and therefore it might come closest to the actual population. For the fraction of females on the board they find a mean of 8.5% but there are boards in their sample that consist for 50% of women. Their average board size is 9.38 and this varies from one or more boards with 3 members to a board with 39 members. The percentage of firms with a female CEO is 3.7% and the average age of the board members is 58.9.

Comparing this larger sample to my own sample, it becomes clear that my sample is different for all of the variables mentioned above. However for most of the values the

difference is not large. The most striking difference is that Adams and Ferreira (2009) have a sample with a board consisting of 39 members and have a higher average fraction of females, but for my sample the maximum of this is higher.

Table 6: Descriptive Statistics

Obs Mean SD Min Max

Tobin's Q 1122 3.913023 25.1135500 0 828.0551 Fraction Female 1122 0.063642 0.1099012 0 0.833333 Board Size 1122 5.945633 1.1666200 2 12 LogSales 1122 3.283764 0.6745112 0.8539414 5.421586 Female CEO 1122 0.016934 0.1290082 0 1 Average Age 1122 61.163970 4.2606670 45.2 82.5

The following figure shows the distribution of firms across the industries. As can be seen below, there are more than 500 firms in my sample that operate in the manufacturing industry and very few firms that belong to the public administration and agriculture, forestry, fishing industries.

(26)

26 Figure 1: The distribution of firms across the industries

To get an idea of the relationship between the dependent and the independent variables I used in this study, I look at the pairwise correlations between each of these variables. The

correlation matrix is shown in Appendix B.

The results in the table of Appendix B show that there is no correlation of any kind between the Tobin’s Q and the variables I use for hypotheses 1 and 3 (where Tobin’s Q is the dependent variable). For the second hypothesis however I use the fraction female as a

dependent variable and as can be seen in Appendix B, there are several significant

correlations. First there are positive correlations between the fraction of females and the size of the board (P<0.05) and the dummy female CEO (P<0.01). Which means that the amount of women on the board increases if the board size increases, more women on the board also lead to more female CEO’s. There is negative correlation as well, between the fraction of females and the average age (P<0.01). This implies that if there are more women on the board, the average age drops. These are no causal relations and it is therefore impossible to say if for example more women on the board lead to a more female CEO’s, or that more female CEO’s lead to more women on the board.

Next to this, the independent variables also correlate with each other, for example the board size and the logsales show a positive correlation (P<0.01), meaning that if the board size increases, the logsales also does this. There are also negative correlations, for example between the board size and the average age (P<0.05), which says that if the size of the board increases, the average age will decline.

0 100 200 300 400 500 600 Agriculture, Forestry, Fishing

Mining Construction Manufacturing Transportation & Public Utilities Wholesale Trade Retail Trade Finance, Insurance, Real Estate Services Public Administration

(27)

27 The table also shows that there is no danger for multicollinearity since the highest correlation coefficient is 0.631 and this is well below the 0.9

4.4 Models

For my study I use two sub research questions: “Is there a positive correlation between the share of female board members and company performance?” and “To what extent will this performance be dependent on the industry?”. To answer this questions I will use three hypotheses, for which I use different models. I will discuss the different models on the basis of the hypotheses.

H1: The company results do not depend on the share of women in the board of directors. To test this relationship between firm performance and the share of women on the board, I use the Tobin’s Q as the dependent variable. This is my measure for the performance in the first model. The fraction of females, the board size, logsales, female CEO and average age are the independent variables in model 1.

Tobin’s Q = β0 + β1Fraction Females + β2Board Size + β3LogSales + β4Female CEO +

β5Average Age + ε

H2: For all companies, women are most likely to be in boards of companies in the services

industry.

I test this relation between the amount of women on the board and the industry of the company by taking the variable female fraction as a dependent variable and the Tobin’s Q, board size, logsales, and the 10 industries as independent variables.

Fraction Females = β0 + β1Tobin’s Q + β2Board Size + β3Sales + β4Female CEO +

β5Average Age + β6IND1 + β7IND2 + … + β15IND10 + ε

H3: There is a correlation between the percentage of women on the board and the results

across different industries.

I want to know if the percentage of women on the board yield different results across different industries. To test this hypothesis, I use several models, as I will show below. I use interaction

(28)

28 terms (for example, FF1 for Fraction Female * IND1), to see if there is a relation between the share of women on the board and the result of the company per industry. This term basically makes it possible to calculate the average effect of the percentage women on the board per industry. The different regressions I will perform consist of 10 “simple” regressions, one for each interaction term and associated industry and 10 “complete” regressions with each interaction term and all the industries (all but one, because otherwise there will be perfect multicollinearity and the regression will fail). Finally I will do one regression with all interaction terms (all but one) and all industries (again all but one).

“Simple” models:

Tobin’s Qi = β0 + β1Fraction females + β2INDi + β3FFi + εi

with i = {1,…,10} “Complete” models:

Tobin’s Qi = β0 + β1Fraction Females + β2IND1 + β3IND2 + β4IND3 + … + β11FFi +

+β12Board Size + β13LogSales + β14Female CEO + β15Average Age + εi

with i = {1,…,10}

After that I do one last regression with all the interaction terms and all industries to see if the significant results will hold:

Tobin’s Q = β0 + β1Fraction Females + β2IND1 + β3IND2 + β4IND3 + … + β11FF1 +

… + β20FF10 + β21Board Size + β22LogSales + β23Female CEO +

(29)

29

5. Method and Results

5.1 Introduction

In this section I present in part 5.2 the estimation method I use to test my hypotheses. After this I present the results per hypothesis also in part 5.2 and I interpret these results. I check whether the independent variables are significant and I will discuss the values of the R2 and adjusted R2 and explain the difference between these two R2.

5.2 Main Results

I do several OLS regressions to estimate my models. By using OLS I assume that the assumption which have to hold in order to get valid hypothesis tests and unbiased and

consistent estimators. After I have done the regressions I will briefly discuss the problems that can occur by threats of internal validity and whether that might be the case in my regression.

H1: The company results do not depend on the share of women in the board of directors.

Table 7: Regression Model 1 H1

The values between the brackets are the standard errors Independent and Control

Variables Dependent Variable Tobin’s Q Fraction Female -2.224 (7.291) Board Size -0.486 (0.660) LogSales 0.200 (1.140) Female CEO 0.051 (6.185) Average Age -0.046 (0.178) Constant 9.119 (11.942) Prob > F 0.9826 R2 0.0006 Adjusted R2 -0.0038 Number of Observations 1122

(30)

30 To know whether this hypothesis is rejected or not, I check the p-value of the model. This value needs to be smaller than 0.05 (5% significance) in order to reject the hypothesis. In this case the value is 0.9826 and therefore the model doesn’t explain the dependent variable Tobin’s Q. Also the variable Fraction Female, which is the most important independent variable, has a negative value. Therefore it looks like it has a negative effect on the firm performance, but is not significant and I do not reject the first hypothesis, so the company results do not depend on the share of women in the board of directors. It can also be seen that none of the regression coefficients show an effect significantly different from zero.

For this model the values of R2 and the adjusted R2 are very low, so the independent variables I use in my regression do not explain a big part of the variation in the dependent variable (Tobin’s Q). However, R2

measures the part of the dependent variable, that is explained by the model, so a value near 1 means that the independent variables can predict the dependent very well, while a value of almost 0 indicates that the independent variables are not able to predict the Tobin’s Q at all well (Stock and Watson, 2012). This does not mean that this is necessary a bad model, it is possible that the model is good even though the R2 is low and the other way around (Stock & Watson, 2012). If the results are significant, then interpretations can be made and conclusions can be drawn despite the low value of R2.

There are however some problems with R2, for example if you add another (relevant or not) independent variable, the R2 will rise or stay the same, no matter the relevance and it will never go down. Therefore a larger model appears to have a better fit (Stock & Watson, 2012). Another problem is that there is a chance of so called “overfitting” and this might cause a noise between the variables (Stock & Watson, 2012). That is why the adjusted R2 is introduced, it corrects to some extent for the number of explanatory variables that a model uses (Stock & Watson, 2012).

(31)

31 H2: For all companies, women are most likely to be in boards of companies in the services

industry.

Table 8: Regression Model 2 H2 Independent and Control Variables Dependent Variable Fraction Female Tobin's Q -0.000 (0.000) Board size 0.005*** (0.003) LogSales -0.006 (0.005) Agriculture 1 -0.053 (0.077) Mining -0.025 (0.018) Construction -0.029 (0.026) Manufacturing -0.001 (0.011) Transport 2 0.014 (0.014) Wholesale 3 0.006 (0.022) Retail Trade 0.061* (0.015) Finance 4 0.017 (0.013) Administration 5 -0.054 (0.023) Constant 0.048** (0.023) Prob > F 0.0001 R2 0.0358 Adjusted R2 0.0253 Number of Observations 1122 *P<0.01, **P<0.05, ***P<0.1

The values between the brackets are the standard errors

1 = Agriculture, Forestry, Fishing, 2 = Transportation & Public Utilities, 3 = Wholesale Trade, 4 = Finance, Insurance, Real Estate,

5 = Public Administration

In this regression only the constant, the board size and the Retail Trade industry (7th) have a significant impact on the fraction females (dependent variable) because their coefficient is different from 0. (Therefore it can be seen that the p-value of the model is 0.0001 which means that not all the coefficient are 0). So the significant part of the fraction females depends

(32)

32 on 0.048 + 0.005 * Board Size + 0.061 * IND7. Which means that if for example a company has a board size of 6 and operates in the retail trade industry (so the dummy is 1), the fraction of females is partly explained by 0.048 + 0.005 * 6 + 0.061 * 1 = 0.1388 (=13.88%). If the company then for instance expands the board by 1 member, the fraction females will rise with 0.005 (=0.5%), which is a very low rise but is consistent with the low percentage of women on boards in general.

I leave the 9th industry out, this is the benchmark and in this case the services industry, because I want to know if there is a difference between this industry and the other industries. As can be seen in the table above, the only industry dummy that is significant is the 7th industry, so I can conclude that there is a significant difference between the retail trade industry and the service industry. However, if I look at the 2nd hypothesis, I cannot conclude that in the service industry there are more female directors than in other industries (except the retail trade industry) because there are no significant differences between the service industry and the other ones. The values of the R2 and adjusted R2 are low, so there is not a large part of the variance of the dependent variable (fraction female) explained by the independent

variables.

H3: There is a correlation between the percentage of women on the board and the results

across different industries.

Table 9: First 4 “Simple” Regressions H3 Independent and Control Variables Dependent Variable Tobin’s Q FF and IND i = 1 i = 2 i = 3 i = 4 Fraction Female -2.392 (6.833) -2.127 (6.915) -2.449 (6.859) 1.300 (8.874) FFi 0 (Omitted) -8.016 (50.004) -2.596 (96.171) -7.484 (13.928) INDi -3.539 (17.794) 0.872 (3.886) -1.702 (6.135) 2.656 (1.737) Constant 4.072* (0.868) 4.019* (0.892) 4.100* (0.877) 2.816** (1.189) Prob > F 0.9241 0.9815 0.9749 0.4718 R2 0.0001 0.0002 0.0002 0.0022 Adjusted R2 -0.0016 -0.0025 -0.0025 -0.0004 Number of Observations 1122 1122 1122 1122 *P<0.01, **P<0.05

The values between the brackets are the standard errors

(33)

33 From the 10 simple regressions I do (the other 6 are in Appendix B), only the constant is significantly different from 0 and of 2 of them the interaction terms FF(1 & 10) are omitted because there are not enough observations. I wanted to test if there is a “pattern” by doing all these simple regressions per interaction term and use the other industries as benchmark. In that way a significant coefficient by the FF would prove a significant difference between the included FF and the other excluded FF’s. So unfortunately following this simple regression I can conclude that there is no difference in the average effect of the percentage females in the board per industry. I will now look at the more complete regression to see if that gives significant results.

(34)

34 Table 10: First 4 “Complete” Regressions H3

Independent and Control Variables Dependent Variable Tobin’s Q FF i = 1 i = 2 i = 3 i = 4 Fraction Female -1.297 (7.403) -1.069 (7.486) -1.284 (7.4252) 1.911 (9.498) Board Size -0.527 (0.666) -0.530 (0.667) -0.526 (0.667) -0.523 (0.666) LogSales 0.484 (1.169) 0.485 (1.169) 0.484 (1.170) 0.480 (1.169) Female CEO -0.330 (6.215) -0.393 (6.225) -0.333 (6.219) -0.351 (6.217) Average Age -0.059 (0.180) -0.060 (0.180) -0.059 (0.180) -0.054 (0.180) FFi 0 (Omitted) -10.481 (50.337) -2.418 (96.705) -7.631 (14.151) Agriculture 1 -3.779 (18.032) -3.761 (18.040) -3.7781 (18.040) -3.613 (18.040) Mining 0.368 (4.218) 0.699 (4.510) 0.368 (4.220) 0.450 (4.222) Construction -2.272 (6.162) -2.267 (6.165) -2.213 (6.592) -2.164 (6.167) Manufacturing 1.086 (2.640) 1.088 (2.641) 1.086 (2.641) 1.516 (2.759) Transport 2 -1.640 (3.236) -1.643 (3.238) -1.640 (3.238) -1.681 (3.238) Wholesale 3 -2.246 ( 5.146) -2.248 (5.148) -2.246 (5.148) -2.256 5.147) Retail Trade -0.891 (3.595) -0.903 (3.597) -0.892 (3.597) -1.079 (3.613) Finance 4 -1.992 (3.124) -1.995 (3.125) -1.992 (3.125) -2.047 (3.126) Administration 5 -2.770 ( 11.572) -2.753 (11.578) -2.769 (11.578) -2.603 (11.580) Constant 9.305 (12.100) 9.418 (12.118) 9.307 (12.106) 8.851 (12.134) Prob > F 0.9968 0.9984 0.9985 0.9976 R2 0.0034 0.0034 0.0034 0.0036 Adjusted R2 -0.0092 -0.0101 -0.0102 -0.0099 Number of Observations 1122 1122 1122 1122

The values between the brackets are the standard errors

1 = Agriculture, Forestry, Fishing, 2 = Transportation & Public Utilities, 3 = Wholesale Trade, 4 = Finance, Insurance, Real Estate, 5 = Public Administration

Looking at the results from the complete regression (of which the rest can be found in Appendix C), I see no significant results at all, not even the constant, therefore I cannot

(35)

35 conclude anything useful to support my hypothesis. There is no correlation between the percentage of women on the board and the results across different industries

I did do an F-test, to compare the simple model with the complete model and to see which one is statistically better. There is no proof that the complete model is better and that the

coefficients of the additional variables are significantly different from 0, as was expected looking at the results.

As can be seen in table 9 and 10, the R2 and adjusted R2 of the complete regression are bigger than ones from the simple regression. But I can conclude that the complete model is more complete than the simple model, even though the F-test failed to prove the significance.

Finally, I do a regression with all variables I used before, this regression is shown below in Appendix E.

This model was designed to check whether the significant results from the previous models/regressions would remain. Now that I don’t have any significant results I can’t compare the coefficients in an useful way.

(36)

36

6. Conclusion and Discussion

The purpose of this thesis is to study a possible difference between men and the share of women on the board and the effect on the company’s results. Part of the literature I found, before I started this thesis, concluded that there is a positive correlation between the percentage women on the board and the performance of the firm. Other articles however stated that there is no difference at all. Therefore the first part of my research question is: “Is there a positive correlation between the share of female board members and company

performance?” Additional to this it might be that there is a difference in results across different industries. Therefore the second part of my research question became: “To what extent will this performance be dependent on the industry?”

To answer these two sub research questions I use 3 hypotheses, the first one is (H1): The company results do not depend on the share of women in the board of directors. To test this hypothesis I use a model to see whether selected independent variables, most importantly share of females, have an effect on the performance measure (Tobin’s Q), which results in the fact that none of coefficients of the variables had any impact on the performance. If it was the case that the hypothesis would be rejected, then the conclusion would have been that there is a correlation (positive or negative) between the performance and the share of women. However this is not the case, so the conclusion is, that there is no effect from the share of women on the board on the performance of the firm.

The second hypothesis, (H2): for all companies, women are most likely to be in boards of companies in the services industry. This is tested with a model of which the fraction

females is the dependent variable and the 9th industry (Service) is the benchmark and is not included in this model. After doing a regression it turns out that only the board size and the 7th industry (Retail Trade) have significant coefficients. The conclusion from this regression is that if the board size rises, the fraction females on the board also rises. Also the significance of the dummy of the 7th industry indicates that there is a significant difference between the retail trade industry and the service industry. Unfortunately the other 8 industries are not significant, so there cannot be concluded that there are significantly more women on board of the service industry.

The last and third hypothesis (H3): There is a correlation between the percentage of women on the board and the results across different industries. To test this I use for every industry interaction term (10 in total), 2 different models, a simple and a complete one. It

(37)

37 turns out that neither of the models contain significant coefficients. Hence the conclusion of this last hypothesis is that there is no pattern, so the percentage of women on the board does not yield different results across different industries.

Returning to the sub research questions, the conclusion is, based on the hypotheses, that the first part “Is there a positive correlation between the share of female board members and company performance?”, turns out to be untrue for my sample. The second part “To what extent will this performance be dependent on the industry?” can also not be answered in the way I hoped before I started this research. There is no correlation between the fraction of females and the difference in industries.

Because almost none of the coefficients are significant, I wonder whether maybe I should not have done an OLS regression. Maybe the assumptions, that are made in order to justify this method, are not valid. To examine this, I have to look into the internal validity, which has 5 threats (Stock & Watson, 2012). The first threat is the chance of omitted variables (Stock & Watson, 2012), so for example I wanted to include ROA in my model, but I did not have access to that part of the database. So now ROA is part of the error term, but there is a possibility that there is a correlation between the independent variables and the error term. This makes that the first assumption for OLS might be violated. The second threat is the functional form misspecification, which means that it could be that the “real” model contains an exponential variable (Stock & Watson, 2012). So for some independent variable of my model, for example Average Age, there might be a exponential version (Average Age)2. This exponential variable is again excluded (as with omitted variable bias) and part of the error term, which then might correlate with the independent variables. However, I’ve tested this and it is not the case for my sample. The third threat is the measurement error (Stock & Watson, 2012), which is caused by the fact that the real variable I use might not be

observable, but instead the variable you can observe is something related to this variable. If this is the case, what I basically have is the related variable which consists of the real variable plus an error in it. Then again this error term within the variable will end up in the error term of the model and correlate with the independent variables of the model. This is not the case in my models, because all my variables were observable. The fourth threat is the sample

selection, this means that a sample is not randomly drawn from a population (Stock &

Watson, 2012). So for example I look at women on the board and whether those women have an effect on the results. But that sample is not actually random, because there are only those women on the board who have certain qualities, while other qualities might be also important

(38)

38 to influence the results of the firm. This leads to endogenous sample selection and I will discuss it using literature in the next paragraph. The last threat is simultaneous causality, this means that causality runs both ways (Stock & Watson, 2012). For example in my sample, there are no causal relations, but there is a correlation between more women on the board and more female CEO’s. But this could also be the other way around, that more female CEO’s lead to more women on the board. I’m not sure if this is also a problem in my sample, because I can’t prove causality, but it is likely to be so.

So the fact that I might have omitted variables, such as ROA, and the simultaneous causality, might be the causes of the large amount of insignificant results.

Referenties

GERELATEERDE DOCUMENTEN

Single human primary chondrocytes directly after isolation (P=0) and after culture expansion at normoxia and hypoxia (P=2 and P=4) and chondrocytes within human cartilage tissue at

In addition, using data of minimum and maximum values on Table 2a, both negative and positive earnings managements viz-a-viz absolute discretionary accruals as the proxy

Associations between genetic variants of the KSHV infection pathway and antipsychotic treatment response as defined by the change in log-transformed PANSS scores over 12 months

Assuming that the former prevents firms from behaving irresponsibly, and the latter contributes to unethical behaviour, it is argued that adding women to the board

Thus, this research will focus specifically on microcredit access and the effects it might have on male participants as compared to female participants, concerning outstanding

The third and final chapter will uncover whether these differences in advice are reflected in the works that were available on the English, Scottish and Dutch market and what

I improved on this prior research by drawing on several theories from psychology and business literature and its application to the corporate board domain, our

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