Gender diversity quota : comply, explain or disregard?

(1)

24 Amsterdam Business School

Gender diversity quota:

comply, explain or disregard?

Name: Roos de Haas

Student number: 10247629

Thesis supervisor: dr. J.J.F. van Raak Date: 19 June 2016

Word count: 12366

MSc Accountancy & Control, specialization Accountancy Faculty of Economics and Business, University of Amsterdam

(2)

2 Statement of Originality

This document is written by student Roos de Haas who declares to take full responsibility for the contents of this document.

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

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

(3)

3

Abstract

A solution for increasing the number of women on boards of companies, is to introduce gender quota. The purpose of this study is to investigate determinants of complying with and explaining non-compliance with the board gender quota in the Netherlands. Data was hand collected from publicly available annual reports and supplemented with data from database Orbis. Empirical evidence from 370 large Dutch companies is presented. Descriptive analysis shows that 75% of the companies in this sample disregarded Dutch gender quota law, therefore the effectiveness of a comply-or-explain type of regulation for this matter, is called into question. Through cross-tabulation analyses and logistic regressions, a significant positive relation was found between the existence of a nomination committee and female representation within the committee and gender quota compliance on supervisory boards, even after controlling for firm characteristics. The existence of an audit committee and the presence of at least one woman in this committee have a strong positive effect on explaining non-compliance with board gender quota. These results holds after controlling for firm characteristics and industry fixed effects. No support was found for associations between explaining non-compliance and being audited by a female auditor and by a Big Four firm. This is probably due to the relatively small sample size. This study contributes to the gender diversity research and research on the effectiveness of comply-or-explain regulation. Moreover, this study provides practical implications for regulators. Mandating the existence of a nomination committee including at least one woman can improve the level of compliance with board gender quota. In the case of non-compliance, mandatory existence of an audit committee with at least one woman might improve the explanations provided by non-complying companies.

(4)

4

List of tables and figures

Table 1 Sample selection ... 20

Table 2 Descriptive statistics for outcome variables ... 25

Table 3 Descriptive statistics for explanatory variables ... 25

Figure 1 Auditing firms and auditors’ gender ... 26

Table 4 Correlations ... 27

Table 5 Contingency table board compliance ... 28

Table 6 Contingency table supervisory board compliance ... 29

Table 7 Contingency table compliance on both boards ... 30

Table 8 Contingency table explaining non-compliance ... 31

Table 9 Logistic regressions for Hypotheses 1 and 2, incl. firm characteristics and two-digit SIC industry control variables ... 33

Table 10 Logistic regressions for Hypotheses 1 and 2, incl. firm characteristics and one-digit SIC industry control variables ... 34

Table 11 Logistic regressions for Hypotheses 1 and 2, incl. firm characteristics and excl. SIC industry control variables ... 35

(7)

7

1. Introduction

"As long as outmoded ways of thinking prevent women from making a meaningful contribution to society, progress will be slow. As long as the nation refuses to acknowledge the equal role of more than half of itself, it is doomed to failure."

– Nelson Mandela on South Africa’s Women’s Day, 1996

1.1 Introduction

President Mandela committed parliament to gender quality and women’s emancipation from the opening of South Africa’s first democratic parliament in 1994 (Myakayaka-Manzini, 2002). Over one third of women were appointed in his cabinet. Nowadays, over 40% of Ministers and Deputy Ministers in South Africa are female.

In companies, equal gender representation is not yet achieved. Gender diversity throughout organizations and especially on the boards of companies, is a current topic both in the media and in academic research (Brunzell & Liljeblom, 2014; fd.nl, 2016). Empirical studies have examined the effect of gender diversity on firm performance (e.g. Ahern & Dittmar, 2012; D. A. Carter, Simkins, & Simpson, 2003; Rose, 2007) and on board effectiveness (Adams & Ferreira, 2009). Recently, national governments have passed laws to increase the representation of women. In the Netherlands, gender quota law concerning the board of companies applies the comply-or-explain principle. The majority of Dutch companies is far from reaching the gender quota (Pouwels & Henderikse, 2015).

1.2 Background information

Prior research has examined the theories underlying the call for more female representation in top leadership positions and especially in corporate boards. Gender differences and group effectiveness theories outline how women may contribute to the effectiveness of the board. Elements of the leadership style that is more common to women than to men were found to have a positive effect on leadership effectiveness (Eagly, Johannesen-Schmidt, & Van Engen, 2003). However, these differences might not hold at higher managerial levels, as women then tend to adopt leadership styles similar to men (Powell, 1990). Group effectiveness theories help depict why groups with varying gender diversity operate differently. Women seem to have a

(8)

8 predominantly positive influence on board effectiveness (Adams & Ferreira, 2009). However, evidence on board chairman’s perceptions shows a difference in attitude towards male and female board members, in favour of men (Brunzell & Liljeblom, 2014).

Also, the relationship between board gender diversity and firm value has been studied. Overall, these studies provide mixed evidence, hence the effect on firm performance remains ambiguous. Carter et al. (2003) were the first to empirically show the impact of board diversity on firm value. They found that the fraction of women or minorities on the board is significantly positively related to the firm value. No significant link between female board representation and firm performance was found by Rose (2007). He argues that a process of socialisation will occur that might result in unconventional board members – i.e. women – adopting the behaviour of the conventional board members, as that might be the only way to qualify for the board position.

1.3 Research focus

In recent years, national governments have passed laws concerning the representation of women that entail either voluntary or mandatory compliance. For example, a new Norwegian law mandates a representation of at least 40% women on boards. Wang & Kelan (2013) found that the gender quota in Norway not only increased gender equality within the board, but also positively affected the representation of women in other leadership positions.

In the Netherlands, no mandatory benchmark is present with respect to female board representation. However, if less than 30% of the board members and supervisory board members is male or female, companies have to justify this by providing an explanation in the annual report (Raad voor de Jaarverslaggeving, 2013). So, Dutch gender quota law applies the comply-or-explain principle that is common in some corporate governance codes in Europe (Andres & Theissen, 2008). Not one of the listed companies in the Netherlands satisfied the 30% target in 2015 (Lückerath-Rovers, 2015). Surprisingly, only 10% of the companies in the Netherlands fulfilled all reporting requirements concerning this lack of female representation (Van Arkel, 2015).

As the relevance of board gender equality is shown both in previous literature in the past decade and frequently in media throughout the developed world, it is remarkable that the majority of companies in the Netherlands does not obey current gender quota law. The case setting in the Netherlands allows for studying both the determinants of complying with the board gender quota and the determinants of explaining non-compliance. This leads to the following research

(9)

9 question: which factors affect companies’ decisions to comply with, explain or disregard board gender quotas?

1.4 Results

Data was hand collected from publicly available annual reports. Cross-tabulation analyses and logistic regressions were used to analyse determinants of complying with the board gender quota and explaining non-compliance with this quota. In general, the comply-or-explain principle that was applied in Dutch board gender law, appears to have failed in this matter: 12,2% of the 37 companies in this sample complied and from the remaining companies only 14,8% explained reasons for non-compliance. The results indicate that the existence of a nomination committee and female representation within this committee significantly positively influence compliance with the board gender quota for supervisory boards. This result holds after controlling for firm characteristics, but not when including industry fixed effects. With regard to explaining non-compliance with the board gender quota, the existence of an audit committee and the presence of at least one woman in the committee have a strong positive effect, even after controlling for firm characteristics and industry classifications. No support was found for the associations between explaining non-compliance and being audited by a female auditor and by a Big Four firm.

1.5 Contribution

Prior studies have examined theories underlying the demand for more female representation in top leadership positions and especially in corporate boards. Also, the effects on including more female board directors have been researched. However, the determinants that affect whether companies choose to appoint more women to their boards have not been studied. The Dutch comply-or-explain law allows for studying determinants that affect this decision. Thereby, this study contributes to both gender diversity and comply-or-explain research. Also, this study provides a practical implication for regulators. Whereas the comply-or-explain principle appears to be not effective in improving a balanced distribution of board seats, both the existence of nomination committee and audit committee and the inclusion of women, might positively emphasize board gender diversity. Thus, there appear to be solutions to improve board gender diversity.

(10)

10

1.6 Paper structure

This study is structured as follows. In the next section, relevant literature for understanding the research question is discussed. Also, hypotheses are formulated with regard to determinants of complying or explaining the board gender quota. In the third section, data collection and research methods are explained. Next, descriptive statistics and empirical results for the cross-tabulation analyses and logistic regressions are presented. Finally, concluding remarks are offered.

(11)

11

2. Literature review

In this section, a theoretical background will be shaped in order to better understand the research question. First, the possible effects of including more women on boards are discussed. Gender differences and group effectiveness theories outline how women may contribute to the effectiveness of the board. Also, the effects of increased board gender diversity on firm value will be discussed. Next, current board gender quotas will be summarized. Specifically, the situation in the Netherlands is emphasized as this comprises the case setting examined in this research. Finally, some possible determinants for companies’ decisions to abide by the law and either comply with or explain non-compliance with the gender quota will be outlined. Hypotheses will be formulated for each possible determinant.

2.1 Effects of female board representation

First, gender differences and group effectiveness theories are discussed to improve our understanding of how women may contribute to board effectiveness (Nielsen and Huse (2010). Then, empirical evidence on the effect that a change in board representation has on firm value will be outlined.

2.1.1 Gender differences and group effectiveness theories

Gender studies consider both biological differences (nature) and environmental differences (nurture) to describe stereotypical feminine or masculine behaviour (Breesch & Branson, 2009). Group effectiveness theories analyse whether group compositions – in this case: group gender compositions – can explain any difference in group effectiveness (Gladstein, 1984).

Considering a combination of gender differences theories and group effectiveness theories can help clarify the issue of whether and how women make a difference to board effectiveness (Nielsen & Huse, 2010). Also, gender representation can be compared to the perception on specific board tasks. In combination with group dynamic theories, this will provide information on how gender diverse groups are perceived to work (Brunzell & Liljeblom, 2014). Although literature suggests that men and women overall do not differ in effectiveness, in some situations some gender related differences in behaviour and skills exist (Nielsen & Huse, 2010; Yukl, 2002).

(12)

12 Women and men are assumed to differ in their leadership behaviour (Nielsen & Huse, 2010). Characteristics that are more often attributed to men are: being assertive, ambitious, aggressive, competitive, independent and self-confident (Eagly et al., 2003). Female leaders tend to be more cooperative and collaborative and less hierarchical than male leaders. Elements of the leadership style that were more common to women than men had a positive relationship with leadership effectiveness, whereas a similar result was not found for leadership styles more common to men. However, according to social science research these gender differences are weakened in higher managerial positions (Nielsen & Huse, 2010). Feminine stereotypes are assumed to be rejected once women arrive at management positions, as they then adopt leadership styles similar to men (Powell, 1990).

So, although gender differences in leadership appear to be recognized, these differences might be minor at a higher managerial level. Rather than affecting overall board effectiveness, gender differences might affect certain board tasks. Nielsen and Huse (2010) distinguish operational control tasks and strategic control tasks. The operational control tasks relate to supervising decisions with regard to the accounting and financial situation of the firm. These tasks are routine and assume quantitative knowledge and skills. On the other hand, strategic control tasks are complex and creative, requiring more analytical skills. Managerial decisions about organizational policies on topics like safety, health and environment need to be monitored. Women on boards tend to be especially valued for their ability to bring strategic input (Bilimoria, 2000). Also, female directors are more sensitive to others’ perspectives, more focused on enhancing others’ self-respect and better able to resolve interpersonal conflicts (Bilimoria, 2000; Eagly et al., 2003; Nielsen & Huse, 2010; Terjesen, Sealy, & Singh, 2009). Consequently, it is theorized that women may be more effective in performing control tasks for issues with a strategic nature. Indeed, Nielsen and Hansen (2010) provide evidence that a higher fraction of female directors has a positive direct relationship with board strategic control. Therefore, the authors suggest that to improve board effectiveness, it is necessary to specify the nature of the board tasks and consider these when appointing women.

Empirical evidence regarding the effect of the gender composition of boards is provided by Adams and Ferreira (2009). First, boards with a high gender diversity have better board attendance rates. Women appear to have fewer attendance problems than men. Moreover, when there are more women on the board, the attendance behaviour of men also improves. Second, gender-diverse boards more strictly monitor CEOs. In firms with weak governance, gender diversity leads to a positive effect on performance. However, when governance is strong, higher gender diversity on the board could lead to overmonitoring of the firm. Overall, Adams and

(13)

13 Ferreira (2009) find a positive relationship between the gender composition of the board and board effectiveness. Women on boards seem to have a similar influence to the independent directors described in governance theory. However, this research could suffer from problems due to omitted factors, such as the overall level of corporate governance (Brunzell & Liljeblom, 2014). An alternative is to study perceptions on board work and link these to gender differences. Based on group dynamic theories, this will provide information about how groups with different gender compositions are perceived to operate. Brunzell and Liljeblom (2014) provide evidence on perspectives from within a board, by studying the board’s chairman’s perception on female board representation. In this way, the general attitude of chairmen toward gender differences can be uncovered. The authors find that most chairmen believe the board’s performance is better when gender diversity is low, thus favouring male board members. These findings validate group effectiveness theories as they show differences in performance of heterogeneous groups.

Overall, this section has attempted to provide an overview of prior female board representation research. Gender differences have been described, although these differences might not hold at higher managerial levels. Group effectiveness theories help depict why groups with varying gender diversity operate differently. Although women seem to have a predominantly positive influence on board effectiveness, evidence on perceptions shows a difference in chairman’s attitude toward male and female board members, in favour of men.

2.1.2 Effect on firm financial value

From a traditional economic viewpoint, prior studies often focused on the direct relationship between female directors and firm financial performance (Brunzell & Liljeblom, 2014; Nielsen & Huse, 2010). Shrader, Blackburn and Iles (1997) give a justification for assuming this relationship. They assume that firms with more female managers are progressive and more competitive as they better reflect the composition of society. Also, these firms have presumably better recruited skilled managers as they draw from a bigger available talent pool. According to Fields and Keys (2003), diversity brings firms heterogeneity of ideas, experiences and innovations, which predominantly causes the positive effect on firm performance.

Carter et al. (2003) are the first to empirically examine the association between board diversity and firm financial value, making use of the agency theory. Board diversity is defined as the fraction of women or ethnical minorities in the board. Agency theory explains the relationship between a principal (e.g. the shareholders) and an agent (e.g. the managers). The costs of solving disagreements and aligning interests within and across groups are considered in agency theory

(14)

14 (Terjesen et al., 2009). According to Fama and Jensen (1983), the board is an important instrument for controlling and monitoring managers. Diverse group dynamics are likely to have a positive impact on the ability to control managers (Erhardt, Werbel, & Shrader, 2003). So, diversity of the board could be used as a tool to minimize agency problems. Erhardt et al. (2003) find results that stress the potential for enhanced representation and perspective resulting from diversity.

On the contrary significant negative effects of women on corporate boards were also found. For example, Ahern and Dittmar (2012) examined the change in stock price after a new law in Norway mandated female board representation. A significant decrease in stock price for both companies with and without female directors followed immediately after the announcement of the gender quota. The decrease in stock price was significantly greater for companies that not yet had any female director. Moreover, a large decline in firm value continued over the subsequent years. Because of the increase in female directors, on average directors became younger and less experienced. More experienced board members are supposedly better at monitoring managers and maximizing shareholder wealth. However, the authors underline their inability to investigate a direct relationship between changes in the board and firm value. Adams and Ferreira (2009) find that the often cited positive relationship between board gender diversity and firm value does not hold when considering the endogeneity problems that arise because of reverse causality or differences in unobservable characteristics across firms. They too conclude that the relationship between gender diversity and firm value is complex. Moreover, they believe motivations other than enhancements in corporate governance and firm value should underpin a government’s decision to introduce mandatory gender quotas.

Rose (2007) finds yet another result when testing the relationship between board gender diversity and firm performance. In this study, diversity does not seem to influence firm financial performance. According to the author, a possible reason for this is a process of socialisation through which unconventional board members (here: women) adopt the behaviour of the existing board members. This idea is tangent to the approach chosen by Torchia, Calabrò and Huse (2011), who consider both critical mass theory and tokenism theories. They suggest that women can contribute to firm innovation when the number of women on board increases from one or two (a few tokens) to the critical mass of at least three women.

Overall, these studies provide mixed evidence with regard to a relationship between board gender diversity and firm value, as a direct relationship is difficult to establish.

(15)

15

2.2 Board gender quotas

Several countries have taken steps to increase the number of women on corporate boards (see e.g. Ahern & Dittmar, 2012; Rose, 2007; Wang & Kelan, 2013). Norway is leading in introducing a gender quota for female board chairs and female CEOs (Nielsen & Huse, 2010). From 2003, a new law requires that 40% of Norwegian public-limited firms are women (Ahern & Dittmar, 2012). As voluntary compliance failed, the law became mandatory from 2006. This legislation led to a worldwide debate on whether gender quotas should be implemented in other countries (Wang & Kelan, 2013). Also, the Norwegian setting allowed for empirical research with regard to firm value, as discussed in Section 2.1.2. Wang and Kelan (2013) find empirical evidence that the Norwegian gender quota did not only lead to more female board chairs and female CEOs, but also increased gender equality in other top leadership positions.

In the Netherlands, the Management and Supervision Act contains a rule concerning the balanced distribution of seats between men and women both on the board of directors and on the board of supervisory directors (Pouwels & Henderikse, 2015; Wijzigingswet Burgerlijk Wetboek Boek 2, 2011). From 1 January 2013, large companies1_{need to reserve at least 30% of board seats for}

women and at least 30% for men. If a company does not comply with this rule, additional information on the seats distribution should be disclosed in its annual report. This disclosure should include: why the balanced distribution was not achieved, how the company tried to achieve this and how the company intends to achieve this in the future. Legally, this is considered soft law as it relies on the principle of comply-or-explain (Rose, 2007).

An independent Talent to the Top Monitoring Commission was installed to monitor the progress of balanced distribution of men and women (Pouwels & Henderikse, 2015). The Commission published a report in September 2015 on the progress over the years 2012-2015. The report consists of a quantitative, descriptive part and a qualitative study. Most noteworthy, the report concludes that the majority of companies is far from reaching the gender quota.

2.3 Possible determinants of abiding by gender quota law

In this section, some determinants that could influence the decision of companies to comply with the gender quota will be discussed. As was noted in the previous section, in the Netherlands the

1_{The requirements apply to public limited companies (NVs) and private limited companies (BVs) that} qualify as “large legal entity” according to the accounting and reporting rules. Two of the following three criteria on two successive balance sheet dates need to be satisfied: value of assets amounts to more than €17.5 million; net turnover is more than €35 million; average number of employees is 250 or more.

(16)

16 gender quota law relies on the comply-or-explain principle. So, if less than 30% of the board members and supervisory board members is male or female, companies have to justify this deficiency in the annual report (Raad voor de Jaarverslaggeving, 2013; Wijzigingswet Burgerlijk Wetboek Boek 2, 2011). In 2015, not one of the listed companies in the Netherlands satisfied the 30% target (Lückerath-Rovers, 2015) Moreover, Pouwels & Henderikse (2015) find that in their sample consisting of 800 companies, not more than 8% of the companies fulfilled all reporting requirements. Therefore, it is interesting to examine both which factors might influence whether companies choose to comply with the gender quota and, if they do not comply, which factors influence whether companies abide by the law by then explaining why they did not comply with the quota. First, determinants of complying with gender quotas are hypothesized. Second, hypotheses regarding factors that influence whether companies explain their non-compliance are developed.

2.3.1 Complying with the gender quota

Board committees, such as corporate governance, remuneration and audit committees, specialize in narrowly defined tasks (Adams & Ferreira, 2009). The nomination committee considers the composition of the board. Candidates are reviewed and recommended as directors. Empirical evidence found that the presence of a nomination committee has a positive influence on gender balance in boards (Brunzell & Liljeblom, 2014). Moreover, a better gender balance exists when female representation in nomination committees exists as opposed to entirely male nomination committees. For instance, in their study Brunzell and Liljeblom (2014) find that all the companies in their sample with at least one female member in their nomination committee, also had at least one female board member. While on the other hand, the companies without female board members did not have any women on their nomination committee. This leads to the following hypotheses.

Hypothesis 1a: The existence of a nomination committee has a positive influence on complying with the gender quota.

Hypothesis 1b: Female representation in a nomination committee – as opposed to an entirely male nomination committee – has a positive influence on complying with the gender quota.

(17)

17

2.3.2 Explaining non-compliance with the gender quota

The audit committee recommends the appointment of independent auditors and discusses internal financial control matters. Historically, audit committees are viewed as a monitoring mechanisms that should enhance the audit attest function of financial reporting (Bradbury, 1990). Empirical evidence shows that the existence of an audit committee is positively associated with voluntary disclosure (Barako, Hancock, & Izan, 2006; Ho & Wong, 2001).

Hypothesis 2a: The existence of an audit committee has a positive influence on explaining the non-compliance with the gender quota.

Hypothesis 2b: Female representation in an audit committee – as opposed to an entirely male audit committee – has a positive influence on companies explaining non-compliance with the gender quota.

As the rule in the Netherlands includes an explain-in-the-annual-report-part, the auditing profession is part of the Dutch solution to establish a more balanced board composition. It is the duty of the auditor to point out when this information – which is legally required – is missing2_{. As}

outlined in the beginning of this section, the majority of companies in the Netherlands do not adhere to the board gender quota law. The Royal Netherlands Institute of Chartered Accountants criticized auditors and strongly emphasized that the gender quota disclosure needs to be considered when performing the audit (Van Arkel, 2015). If the information is not in the final annual report, the auditor should mention the shortcomings in the audit report. So, the auditor has influence on whether a company discloses the non-compliance with the gender quota in the annual report, as is required by law.

In section 2.1.1 gender differences theories have been discussed. Theories about gender differences in general also apply to auditor gender differences. However, gender differences specifically in an audit context have also been researched.

Audit quality is defined as the probability that the auditor discovers and reports any material misstatement in the financial statements (DeAngelo, 1981). Application of the selectivity hypothesis suggests that female auditors will discover more material misstatements, as they process more available information while men tend to be selective in processing information (Meyers-Levy, 1989). The findings of Breesch and Branson (2009), suggest that female auditors

(18)

18 indeed discover more potential misstatements than male auditors, although they analyse the misstatements less accurately. Also, female auditors are more risk-averse than male auditors. According to the study of Pouwels and Henderikse (2015), auditors in the Netherlands were only slightly familiar with the new law. Also, auditors considered the matter of low priority. Not all companies in the study considered the lack of gender diversity a problem. Some stated that it has become “a female thing”. This insinuates that women might consider gender diversity a bigger problem than men do. If so, female auditors might more often point out to companies the issue of gender equality.

As previous literature shows that female auditors will discover more potential misstatements, companies they audit might more likely obey the law and explain their gender quota non-compliance. This leads to the following hypothesis.

Hypothesis 3: Female auditors have a positive influence on companies explaining the non-compliance with the gender quota in their annual reports.

Auditing firms, and especially the Big Four, are important in professionalization and regulatory processes (Cooper & Robson, 2006). When auditing firms are more central to issues such as female board representation the public interest and politicians might also get more involved. In other words, the Big Four play a crucial role in the development of codes of best practice (Edgley, Sharma, & Anderson-Gough, 2015). The auditing profession itself has traditionally been dominated by men (Kornberger, Carter, & Ross-Smith, 2010). However, recently auditing firms, and especially the Big Four, have promoted their commitment to gender equality. Therefore, it is expected that when a company is audited by a Big Four firm, this will have a positive influence on the provision of an explanation as to why companies fail to comply with the quota.

Hypothesis 4: Big Four engagementhas a positive influence on explaining the non-compliance with the gender quota.

(19)

19

3. Methodology

This chapter focuses on the research method that has been used to address the research question: which factors affect companies’ decisions to comply with, explain or disregard board gender quotas? First, the method to obtain sample data will be described. Next, the measures for the variables of interest and control variables will be defined. Finally, the methods used to test the hypotheses will be outlined.

3.1. Data and sample selection

In the first place, the company data is obtained from Orbis3_{, which is used as the starting point for}

the sample selection. All companies that are subject to the revision in the Dutch Management and Supervision Act, concerning a balanced distribution of board seats, are selected. These are Dutch public limited companies (NVs) and private limited companies (BVs) that qualify as large legal entities. This implies satisfying two out of three criteria on two successive balance sheet dates: a total asset value of more than €17.5 million; a net turnover of more than €35 million and/or an average number of employees of 250 or more. For each criterion, a list of companies was obtained from database Orbis. Next, the data were matched and filtered to select only those companies that satisfy at least two of the criteria. This resulted in a list of 3,795 companies that met the criteria in 2014. As the revision in the Dutch Management and Supervision Act came into effect on 1 January 2013, companies will have been able to change the composition of their boards to a larger extent in 2014 compared to 2013. Indeed, the level of compliance was higher in the year 2014 as opposed to 2013 (Pouwels & Henderikse, 2015). Therefore, this study will focus on 2014 only. These companies were sorted based on their total asset value. The top 500 companies were selected for further analysis. In order to test the hypotheses, information about the composition of the board of directors and, if present, the board of supervisory directors; explanations in case of non-compliance with the gender quota; the existence and composition of board committees; the signing auditing firm and the gender of the signing auditor needed to be collected. This information was hand collected from publicly available annual reports4_{. For 104 companies, the}

annual report of 2014 was not accessible online. Although these annual reports had been filed with the Dutch Chamber of Commerce, the reports are either not accessible online or only

3_{Orbis is a database published by Bureau van Dijk. It contains information on over 200 million companies} worldwide, with a focus on private company information.

4_{The annual reports were obtained from company.info, a database that contains qualitative information} about all companies in the Netherlands.

(20)

20 available at the company’s office which is permitted according to article 394.4 from the Dutch Civil Code Book 2. Furthermore, 26 companies could not achieve a balanced distribution of board seats as their board of directors only consisted of one seat and they did not have a board of supervisory directors. Interestingly, in all of those companies the sole director was male. These companies have been eliminated from the sample. This leads to a final sample of 370 companies. Table 1 summarizes the formation of the final sample.

TABLE 1SAMPLE SELECTION

Number of observations Companies that meet the “large legal entity” criteria in the Netherlands in 2014 3,795

Selection of 500 companies based on total asset value 500 Annual report not available in 2014 104 Balanced distribution of board seats not possible as there is only one

(supervisory) director 26

Final sample 370

3.2. Variables

In this research, four outcome variables are of interest. The first variable measures whether a company complies with the board gender quota for its board. The second variable measures whether a company complies with the gender quota for its supervisory board, if present. In the Netherlands, companies are allowed to choose between one-tier and two-tier board models according to the revised Management and Supervision act (Peij, Bezemer, & Maassen, 2012). One-tier board models consist of both executive and non-executive directors. The latter supervise or control the first, whilst in the same board. In a two-tier board model these roles are clearly separated with a management board and a supervisory (non-executive) board (Maassen & Van Den Bosch, 1999). The Dutch gender board quota applies explicitly to both boards. Therefore, the third variable measures whether compliance is achieved on all existent boards. In a one-tier board model, this variable concerns the one existing board only. In that case, the value of this variable equals the value of the (single) board compliance variable. However, in a two-tier board model this variable represents compliance for both boards. Compliance is also assumed when the management board consists of only one director and the supervisory board achieved gender equality. For the sake of simplicity, this variable is referred to as “compliance on both boards”, even though in some companies only one board is installed. The fourth variable measures

(21)

21 whether – in case of non-compliance with the both board gender quota – companies explain their reasons for not complying with the desired gender composition of boards.

As discussed in the previous section, the information necessary to test the hypotheses was hand collected from annual reports. Information about the (supervisory) board composition was always presented in the annual reports. In the case of non-compliance with the board gender quota, an explanation is mandated according to Dutch law. Companies have to explain in the annual report (1) why a gender-balanced distribution of seats was not achieved, (2) how the company tried to achieve this and (3) how the company intends to achieve this in the future (Pouwels & Henderikse, 2015). The annual reports were thoroughly scrutinized in order to determine whether satisfactory explanations on all three parts were provided.

The explanatory variables are the hypothesized determinants of compliance and disclosure in case of non-compliance. Through analysing the annual reports, it was also determined whether a company had a nomination committee and/or an audit committee and, if so, whether any woman sat on the committee. Also, information about the signing auditing firm and signing auditor was stored. After collecting these data, the gender of the signing auditor was determined5_.

3.3. Method

All outcome variables and the explanatory variables that are of interest for testing the hypotheses are binary. The outcome variables regarding compliance with the (supervisory) board gender quota are set to 1 in the case of compliance. With regard to the committees, the independent variables take the value 1 if the particular committee is present and when female representation exists. Female representation exists when the committee consists of at least one female member. So, no balanced distribution of seats is required – contrary to the variables on board composition. This is consistent with the hypotheses developed in section 2.3. Also, a variable regarding the gender of the auditor will be added. This variable equals 1 in the case of a female auditor. A Big Four dummy will be added in order to test Hypothesis 4.

A cross-tabulation analysis will be carried out. Through this analysis, associations between particular variables can be determined. However, the direction of a relationship cannot be analysed. Therefore, a logistic regression will be run. Also, this allows for including control variables and checking for interdependencies between independent variables.

5_{The gender of the signing auditor was determined by accessing the corporate website of the auditing} firm and, where necessary, the LinkedIn profile of the auditor in question.

(22)

22 For all logistic regressions, control variables with regard to firm characteristics are included. These relate to the return on assets and equity, leverage, total assets value, revenue per employee and whether the company is listed on the stock exchange. Prior evidence shows that the industry of a company may influence the level of board diversity. For instance, healthcare and technology-related service industries are more likely to appoint female directors (Harrigan, 1981). On the other hand, the banking and finance sector have “too big, too old, and too male” boards (Engen, 2002; Kang, Cheng, & Gray, 2007). Therefore, an industry control variable is included. The Standard Industry Classification (SIC) codes are used. Consistent with prior research, industry fixed effects based on two-digit SIC codes are included.

For each explanatory variable of interest, a distinct regression will be run. This results from the differences in sample size. That is, female representation in a committee can only be determined in a company that has installed the particular committee.

The first equations test Hypothesis 1 (both sub-hypotheses a and b), concerning compliance with board gender quota. These equations are used for logistic regressions regarding the first three outcome variables: board compliance, supervisory board compliance and compliance on both boards. The probability of board compliance is hypothesized to be a function of the following variables. 𝐶𝑜𝑚𝑝𝑙𝑦 = 𝛽0+ 𝛽1𝑁𝑜𝑚𝐶𝑜𝑚 + 𝛽2𝑅𝑂𝐴 + 𝛽3𝑅𝑂𝐸 + 𝛽4𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒 + 𝛽5𝑙𝑜𝑔(𝐴𝑠𝑠𝑒𝑡𝑠) + 𝛽6𝑅𝑒𝑣𝑒𝑛𝑢𝑒 𝑝𝑒𝑟 𝐸𝑚𝑝𝑙𝑜𝑦𝑒𝑒 + 𝛽7𝐿𝑖𝑠𝑡𝑒𝑑 + 𝛽8−56 𝑆𝐼𝐶 𝑑𝑢𝑚𝑚𝑖𝑒𝑠 (1a) 𝐶𝑜𝑚𝑝𝑙𝑦 = 𝛽0+ 𝛽1𝐹𝑒𝑚𝑁𝑜𝑚 + 𝛽2𝑅𝑂𝐴 + 𝛽3𝑅𝑂𝐸 + 𝛽4𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒 + 𝛽5𝑙𝑜𝑔(𝐴𝑠𝑠𝑒𝑡𝑠) + 𝛽6𝑅𝑒𝑣𝑒𝑛𝑢𝑒 𝑝𝑒𝑟 𝐸𝑚𝑝𝑙𝑜𝑦𝑒𝑒 + 𝛽7𝐿𝑖𝑠𝑡𝑒𝑑 + 𝛽8−56 𝑆𝐼𝐶 𝑑𝑢𝑚𝑚𝑖𝑒𝑠 (1b) Where:

NomCom = existence of a nomination committee

FemNom = female representation in the nomination committee

ROA = Return on Assets

ROE = Return on Equity

Leverage = book value of long-term debt divided by book-value of total assets Log (Assets) = natural logarithm of total assets

Revenue per Employee = total revenue divided by number of employees

Listed = dummy variable that equals 1 when company is listed on a stock exchange

(23)

23 Hypotheses 2-4 refer to explaining non-compliance. Again, variables controlling for firm characteristics and industry classification are included. This results in the following equations. 𝐸𝑥𝑝𝑙𝑎𝑖𝑛 = 𝛽0+ 𝛽1𝐴𝑢𝑑𝐶𝑜𝑚 + 𝛽2𝑅𝑂𝐴 + 𝛽3𝑅𝑂𝐸 + 𝛽4𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒 + 𝛽5𝑙𝑜𝑔(𝐴𝑠𝑠𝑒𝑡𝑠) + 𝛽6𝑅𝑒𝑣𝑒𝑛𝑢𝑒 𝑝𝑒𝑟 𝐸𝑚𝑝𝑙𝑜𝑦𝑒𝑒 + 𝛽7𝐿𝑖𝑠𝑡𝑒𝑑 + 𝛽8−56 𝑆𝐼𝐶 𝑑𝑢𝑚𝑚𝑖𝑒𝑠 (2a) 𝐸𝑥𝑝𝑙𝑎𝑖𝑛 = 𝛽0+ 𝛽1𝐹𝑒𝑚𝐴𝑢𝑑 + 𝛽2𝑅𝑂𝐴 + 𝛽3𝑅𝑂𝐸 + 𝛽4𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒 + 𝛽5𝑙𝑜𝑔(𝐴𝑠𝑠𝑒𝑡𝑠) + 𝛽6𝑅𝑒𝑣𝑒𝑛𝑢𝑒 𝑝𝑒𝑟 𝐸𝑚𝑝𝑙𝑜𝑦𝑒𝑒 + 𝛽7𝐿𝑖𝑠𝑡𝑒𝑑 + 𝛽8−56 𝑆𝐼𝐶 𝑑𝑢𝑚𝑚𝑖𝑒𝑠 (2b) 𝐸𝑥𝑝𝑙𝑎𝑖𝑛 = 𝛽0+ 𝛽1𝐹𝑒𝑚𝐴𝑢𝑑𝑖𝑡𝑜𝑟 + 𝛽2𝑅𝑂𝐴 + 𝛽3𝑅𝑂𝐸 + 𝛽4𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒 + 𝛽5𝑙𝑜𝑔(𝐴𝑠𝑠𝑒𝑡𝑠) + 𝛽6𝑅𝑒𝑣𝑒𝑛𝑢𝑒 𝑝𝑒𝑟 𝐸𝑚𝑝𝑙𝑜𝑦𝑒𝑒 + 𝛽7𝐿𝑖𝑠𝑡𝑒𝑑 + 𝛽8−56 𝑆𝐼𝐶 𝑑𝑢𝑚𝑚𝑖𝑒𝑠 (3) 𝐸𝑥𝑝𝑙𝑎𝑖𝑛 = 𝛽0+ 𝛽1𝐵𝑖𝑔𝐹𝑜𝑢𝑟 + 𝛽2𝑅𝑂𝐴 + 𝛽3𝑅𝑂𝐸 + 𝛽4𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒 + 𝛽5𝑙𝑜𝑔(𝐴𝑠𝑠𝑒𝑡𝑠) + 𝛽6𝑅𝑒𝑣𝑒𝑛𝑢𝑒 𝑝𝑒𝑟 𝐸𝑚𝑝𝑙𝑜𝑦𝑒𝑒 + 𝛽7𝐿𝑖𝑠𝑡𝑒𝑑 + 𝛽8−56𝑆𝐼𝐶 𝑑𝑢𝑚𝑚𝑖𝑒𝑠 (4) Where:

AudCom = existence of an audit committee

FemAud = female representation in the audit committee FemAuditor = signing auditor is female

BigFour = signing audit firm is either Deloitte, EY, KPMG or PwC

(24)

24

4. Results

In this section, the results of the statistical analyses are presented. First, descriptive statistics about the outcome variables and the explanatory variables are described. Second, a univariate analysis on the relationship between these variables is presented. Last, a multivariate analysis including multiple logistic regressions will be shown.

4.1. Descriptive statistics

Table 2 presents descriptive statistics for the outcome variables regarding complying with and explaining board gender quota. As all outcome variables are binary, the value is either zero or one. The first row shows that approximately 15% of 359 companies complied with the gender quota for its board. In a one-tier board model, this variable represents the only board a company has. Eleven companies could not achieve a balanced distribution of seats, as there was only one director. However, these companies adopted the two-tier board model and thus did have a supervisory board. Otherwise, the companies would have been excluded from the sample, as described in section 3.1. The majority of the companies in the sample implemented a supervisory board. One in five supervisory boards consisted at least for 30% of female directors and at least for 30% of male directors. Of the 370 firms in the sample, 45 firms (or 12,2%) complied fully to the board gender quota, i.e. both in their management and supervisory board, if present.

Following from the revision in Dutch law, the remaining 88% of the firms in the sample were required to provide an explanation as to why they did not comply with the board gender quota. The final row in Table 2 shows that only 14,8% of non-complying firms actually provided an adequateexplanation. So, 85,2% of the 325 companies did not provide an explanation. This means that concerning the entire sample, 277 out of 370 companies did not comply with the law which requires to provide an explanation in case of non-compliance. Consequently, three-quarters of the firms in this sample did not obey the law.

Thus, it appears that the comply-or-explain principle is not effective with regard to gender quota. Furthermore, hand collection of data from annual reports showed that some companies only refer to the Dutch gender quota with regard to the supervisory board, thereby omitting the gender quota for the management board (e.g. in the annual report of Gasterra B.V.6_{). This could indicate}

unawareness of companies regarding the exact requirements of the revised Dutch law.

(25)

25 TABLE 2DESCRIPTIVE STATISTICS FOR OUTCOME VARIABLES

Variables Observations Mean Standard deviation

Board compliance 359 0.148 0.355

Supervisory board compliance 205 0.195 0.397

Compliance on both boards 370 0.122 0.327

Explain non-compliance 325 0.148 0.355

In Table 3 the descriptive statistics of the independent variables of interest are listed. Again, all variables are binary and thus have a minimum value of 0 and a maximum value of 1. Of all the firms in the sample, a little over 20% installed a nomination committee. More companies implemented an audit committee: 32,4%. In half of the nomination committees, at least one female member was present. This occurred less often in audit committees, where approximately one third of the committees contained female members. Of all 370 audits, only 7% were signed off by a female auditor. Furthermore, 97% of all audits were performed by a Big Four audit firm.

TABLE 3DESCRIPTIVE STATISTICS FOR EXPLANATORY VARIABLES

Variables Observations Mean Standard deviation

Nomination committee 370 0.238 0.426

Female representation in nomination committee

89 0.517 0.503

Audit committee 370 0.324 0.469

Female representation in audit committee

120 0.308 0.464

Female auditor 370 0.073 0.260

Big Four firm 370 0.970 0.170

Information about the auditors’ gender, specified by audit firm, is further disclosed in Figure 1. The auditor gender is expressed as a percentage of all audits in the sample. The figure shows that PwC has the highest portion of audit engagements signed off by a female auditor in this sample, followed by EY. Deloitte, KPMG and the non-Big Four firms barely have any audits signed off by a woman. The figure also shows which company conducted most of the audits in the sample. PwC performed one third of all 370 audits. A quarter of the companies was audited by EY.

(26)

26 FIGURE 1AUDITING FIRMS AND AUDITORS’ GENDER

For all variables, Pearson’s correlation coefficients and accompanying p-values were retrieved for analysis. However, for the sake of conciseness, Table 4 presents only the correlations between the most interesting variables. The variable for compliance on both boards correlates very highly with board compliance (0.875, significant at 0.001 level). This makes sense, as the first variable is derived from the latter. Also, the variables regarding the existence of a nomination committee and an audit committee are correlated (0.793, significant at 0.001 level). This suggests that often firms who implement a nomination committee also implement an audit committee.

(27)

27 TABLE 4CORRELATIONS

* p < .05 ** p < .01 *** p < .001

7_{RPE = Revenue per Employee}

com p ly boar d com pl ys up com pl yb ot h expl ain n om com fe m n om audc om fe m aud fe m audi tor b igf ou r roa roe leve ra ge loga ss et s R PE 6 lis te d complyboard 1 complysup 0.282*** 1 complyboth 0.875*** 0.550*** 1 explain 0.077 -0.014 -0.023 1 nomcom 0.008 0.192** -0.014 0.380*** 1 femnom 0.149 0.378*** 0.199 0.085 0.115 1 audcom -0.019 0.145* -0.046 0.299*** 0.793*** -0.102 1 femaud 0.218* 0.402*** 0.319*** 0.168 0.169 0.287** 0.061 1 femauditor 0.109* 0.087 0.150** -0.067 -0.035 0.038 -0.039 0.065 1 bigfour -0.063 0.035 -0.081 0.066 0.060 0.112 0.053 0.087 0.049 1 roa 0.057 -0.006 0.021 -0.010 0.025 -0.043 -0.014 0.026 -0.004 0.024 1 roe 0.032 -0.020 -0.003 0.071 -0.039 -0.098 -0.046 0.013 0.007 -0.006 0.451*** 1 leverage -0.053 0.054 -0.048 -0.013 -0.064 0.026 -0.062 -0.028 0.051 -0.010 -0.260*** -0.007 1 logassets 0.058 0.194** 0.053 0.069 0.268*** 0.202 0.256*** 0.249** -0.013 0.096 0.043 -0.069 0.021 1 RPE7 _-0.024 _-0.086 _-0.017 _-0.040 _-0.069 _-0.101 _-0.004 _0.107 _0.007 _-0.030 _0.036 _-0.004 _0.033 _0.124* ₁ listed -0.039 0.030 -0.076 0.264*** 0.624*** -0.077 0.595*** 0.113 -0.097 0.078 0.023 -0.022 -0.077 0.289*** -0.061 1

(28)

28

4.2. Univariate analysis

In order to test the hypotheses, a cross-tabulation analysis is performed. The chi-square statistic (χ2_{) is used}

to analyse the joint frequency distributions. This determines whether the variables are either statistically associated or independent. The univariate analysis is subdivided into two sections: the first covers the compliance outcome variables and the second section the outcome variable regarding explaining non-compliance.

4.2.1 Complying with gender quota

Table 5, Table 6 and Table 7 below present the contingency tables for the dependent variables board compliance, supervisory board compliance and compliance on both boards. The percentages of (non-)compliance are shown for each independent variable. In parentheses, the actual joint frequencies and the expected frequencies are separated by a slash. The probability associated with the chi-square statistic indicates the extent to which it would be incorrect to assert a relationship between the variables in the population from which the sample is drawn. Thus, the lower the probability associated with the chi-square statistic, the stronger the relationship between the variables.

For the (single) board compliance, the cross-tabulation analysis shows that the existence of a nomination committee is not significantly related to board compliance. Female representation is not significantly associated with board compliance either.

TABLE 5CONTINGENCY TABLE BOARD COMPLIANCE

Board compliance χ2 _Pr 0 1 Total Nomination committee 0 85.40 (234/233.5) 14.60 (40/40.5) 100 (274) 0.025 0.875 1 84.71 (72/72.5) 15.29 (13/12.5) 100 (85) Total 85.24 (306) 14.76 (53) 100 (359) Female representation 0 90.24 (37/34.7) 9.76 (4/6.3) 100 (41) 1.875 0.171 1 79.55 (35/37.3) 20.45 (9/6.7) 100 (44) Total 84.71 (72) 15.29 (13) 100 (85)

The percentages of (non-)compliance are shown for each independent variable. In parentheses, the actual joint frequencies and the expected frequencies are separated by a slash.

(29)

29 On the contrary, the existence of a nomination committee does have a strong relationship with compliance regarding supervisory boards. Moreover, the presence of at least one woman in the nomination committee appears to be significantly related to supervisory board compliance.

TABLE 6CONTINGENCY TABLE SUPERVISORY BOARD COMPLIANCE

Supervisory board compliance χ2 _Pr

0 1 Total Nomination committee 0 86.51 (109/101.4) 13.49 (17/24.6) 100 (126) 7.5452 0.006 1 70.89 (56/63.6) 29.11 (23/15.4) 100 (79) Total 80.49 (165) 19.51 (40) 100 (205) Female representation 0 89.19 (33/26.2) 10.81 (4/10.8) 100 (37) 11.297 0.001 1 54.76 (23/29.8) 45.24 (19/12.2) 100 (42) Total 70.89 (56) 29.11 (23) 100 (79)

Compliance on both boards shows relationships similar to that of single board compliance. The existence of a nomination committee is not a significant factor, whereas female representation in the particular committee is.

The chi-square test assumes that expected frequencies for each cell have a value greater than 5. Overall, the sample is large enough to prevent these expected frequencies to adopt a value lower than 5. However, for testing the association between female representation in the nomination committee and compliance on both boards, the expected frequency is low. Therefore, Fisher’s exact test is executed. This shows an exact p-value of 0.061. So, the association between having at least one woman in the nomination committee and compliance on both boards is significant at a 0.10 level.

Overall, preliminary support is found for Hypothesis 1a, but only with regard to supervisory board compliance. Also, preliminary evidence is found that provides support for Hypothesis 1b. This applies to the variables for both supervisory board compliance and compliance on both boards.

(30)

30 TABLE 7CONTINGENCY TABLE COMPLIANCE ON BOTH BOARDS

Compliance on both boards χ2 _Pr

0 1 Total Nomination committee 0 87.59 (247/247.7) 12.41 (35/34.3) 100 (282) 0.069 0.793 1 88.64 (78/77.3) 11.36 (10/10.7) 100 (88) Total 87.84 (325) 12.16 (45) 100 (370) Female representation 0 95.24 (40/37.2) 4.76 (2/4.8) 100 (42) 3.477 0.062 1 82.61 (38/40.8) 17.39 (8/5.2) 100 (46) Total 88.64 (78) 11.36 (10) 100 (88)

4.2.2 Explaining non-compliance with gender quota

The results of the cross-tabulation analysis for explaining non-compliance are shown in Table 8. The associations between this dependent variable and the independent variables for the audit committee, female representation in the audit committee, Big Four involvement and the gender of the signing auditor are tested. Again, for all explanatory variables the percentages of explaining non-compliance are shown. The numbers in parentheses are the total actual joint frequencies and expected frequencies in each category.

The relationship between explaining non-compliance and the installation of an audit committee is profoundly significant. As opposed to female representation in nomination committees, the presence of at least one woman shows a weaker relationship. However, the association is significant at a 0.10 level. In all audits performed by non-Big Four audit firms, companies did not succeed in explaining their non-compliance. However, the independent variable of Big Four involvement does no turn out to be significantly related to explaining non-compliance. Finally, the gender of the signing auditor is weakly associated with explaining board gender quota non-compliance.

Again, considering the expected frequencies in each cell, it is appropriate to compute Fisher’s exact for the variables regarding the signing auditor’s gender and Big Four engagement. For the first, the exact p-value equals 0.197. For the latter, Fisher’s exact is 0.274. So, both variables have a weak relationship with whether companies explain their non-compliance.

Thus, strong preliminary support is found for Hypothesis 2a. So, the existence of an audit committee is indeed positively associated with disclosure, which is in line with previous

(31)

31 evidence regarding voluntary disclosure (Barako et al., 2006). Some preliminary evidence that provides support for Hypothesis 2b is also found. Hypotheses 3 and 4 are not supported.

TABLE 8CONTINGENCY TABLE EXPLAINING NON-COMPLIANCE

Explaining non-compliance χ2 _Pr 0 1 Total Audit committee 0 92.66 (202/185.8) 7.34 (16/32.2) 100 (218) 29.037 0.000 1 70.09 (75/91.2) 29.91 (32/15.8) 100 (107) Total 85.23 (277) 14.77 (48) 100 (325) Female representation 0 74.68 (59/55.4) 25.32 (20/23.6) 100 (79) 3.034 0.082 1 57.14 (16/19.6) 42.86 (12/8.4) 100 (28) Total 70.09 (75) 29.91 (32) 100 (107) Female auditor 0 84.64 (259/260.8) 15.36 (47/45.2) 100 (306) 1.449 0.229 1 94.74 (18/16.2) 5.26 (1/2.8) 100 (19) Total 85.23 (277) 14.77 (48) 100 (325) Big Four 0 100.00 (8/6.8) 0.00 (0/1.2) 100 (8) 1.421 0.233 1 84.86 (269/270.2) 15.14 (48/46.8) 100 (317) Total 85.23 (277) 14.77 (48) 100 (325)

The percentages of explaining non-compliance are shown for each independent variable. In parentheses, the actual joint frequencies and the expected frequencies are separated by a slash.

4.3 Multivariate analysis

Where the univariate analysis only considers whether a dependency between variables exists, a multivariate analysis can control for interdependencies between independent variables, including control variables. In the previous section, relationships between dependent and independent variables were studied by means of cross-tabulation analyses. Some significant associations were observed. However, in this analysis no distinction was made between dependent and independent variables, i.e. dependent variables were not explicitly defined. Therefore, logistic regressions are carried out.

For each equation, developed in section 3.3, an individual regression is run. The sub-hypotheses 1a and 1b are tested for single board compliance, supervisory board compliance and compliance

(32)

32 on both boards (identified by the affix of i, ii and iii respectively). This results in a total of 10 models. The results of the equations are presented in Table 9. Each cell shows the estimated coefficient and the z-value (in parentheses). A superscript * [**] denotes significance at the 10% [5%] level. All explanatory variables of interest are presented in bold. The logistic regressions for the equations regarding compliance on both boards and female representation in nomination committees (1biii), the presence of a female auditor (3) and Big Four engagement (4) could not be run8_{. All regressions included the two-digit SIC codes as control variables. This resulted in the}

inclusion of 49 dummy variables. In order to present the results concisely, results in Table 9 are presented without the output of the SIC control variables.

The likelihood ratio chi-square test (LR chi2_{) and accompanying probability show whether the}

overall model is statistically significant. At a 10% level, only the logistic regressions regarding Hypotheses 2a and 2b are significant overall. For all other models, the inclusion of the explanatory variable of interest does not fit the data significantly better than the model with only the constant. The results show a positive and significant association between explaining non-compliance and the existence of an audit committee, thereby supporting Hypothesis 2a. The positive and highly significant coefficient on female representation in audit committees (FemAud) suggests that the presence of at least one woman in the audit committee has a positive influence on companies explaining non-compliance with board gender quota. This supports Hypothesis 2b. In both regressions, the coefficients for the leverage control variable is significant too. In the regression regarding Hypothesis 2b, the coefficients for the control variables regarding total assets value, revenue per employee and whether the company is listed turn out to be significant as well.

8_{These logistic regressions exclude the independent variables of respectively female representation in} nomination committees, female auditors and Big Four involvement. An explanation for this could be that the joint frequencies of the particular dependent and independent variable are too low. This also resulted from the univariate analysis, where this problem was resolved by computing Fisher’s exact.

(33)

33 TABLE 9LOGISTIC REGRESSIONS FOR HYPOTHESES 1 AND 2, INCL. FIRM CHARACTERISTICS AND TWO -DIGIT SIC INDUSTRY CONTROL VARIABLES

1ai 1aii 1aiii 1bi 1bii 2a 2b

Constant 1.09E-08

(-0.01) 2.34E-10 (-0.01) 1.22E-08 (-0.02) 5633156 (0.00) (-1.62) 0.000 3.80E-06 (-0.01)

3.59E+18** (2.52) NomCom 0.694 (-0.47) 3.728* (1.81) (-0.75) 0.529 FemNom 3.263 (1.04) 18.857** (2.06) AudCom 4.431* (1.95) FemAud 906.894** (2.57) ROA 1.025 (0.75) (1.42) 1.117 1.065* (1.75) (0.35) 1.039 (0.29) 1.044 (-0.36) 0.985 (-0.42) 0.956 ROE 1.003 (0.76) (-0.96) 0.976 (-0.42) 0.998 (-0.42) 0.989 (0.08) 1.002 (1,12) 1.007 (-0.87) 0.986 Leverage 0.12 (-1.48) (0.32) 1.579 (-0.82) 0.305 22.208 (0.82) (0.08) 1.227 11.545* (1.68) 1384798** (2.40) Log (Assets) 1.324 (1.27) 1.645** (2.13) (1.08) 1.297 (-0.25) 0.893 (1.04) 1.566 (-0.67) 0.850 0.025** (-2.68) Revenue per Employee (-0.40) 1.000 (-1.09) 1.000 (-0.42) 1.000 (1.07) 1.000 (-0.86) 0.999 (-0.39) 1.000 1.001* (1.89) Listed 1.219 (0.19) (-1.27) 0.242 (-0.44) 0.573 6.38E-08 (0.00) 4.38E-07 (-0.01) (1.23) 3.012 1114.459** (2.23) Log likelihood -73.0691 -49.0532 -62.8867 -16.8945 -18.7412 --58.942 -13.417 Number of observations 182 112 174 34 43 188 56 LR chi2_(df) 16.79 (23) 25.61 (19) 17.46 (20) 5.51 (13) 19.28 (14) 50.45 (25) 43.50 (17) Prob > chi2 _0.8195 _0.1414 _0.6231 _0.9622 _0.1544 _0.0019 _0.0004 Pseudo R2 _0.1030 _0.2070 _0.1219 _0.1402 _0.3397 _0.2997 _0.6185

Each cell shows the estimated coefficient and the z-value (in parentheses). A superscript * [**] denotes significance at the 10% [5%] level. All explanatory variables of interest are presented in bold.

As the majority of the models is not statistically significant, the regressions are carried out again, but now using the one-digit SIC codes instead of the two-digit SIC codes. This significantly reduces the number of independent variables taken into account. This might improve the predictive power of the models, as the sample is overall not that large. Agresti (2007)argues that the sample size limits the number of variables that can be studied. Based on statistical evidence on the reliability of logistic regressions, he suggests that 10 observations are needed for every variable included. The results for the regressions with one-digit SIC control variables are shown in Table 10. Again, the estimated coefficient and the z-value (in parentheses) are shown in each cell. A superscript * [**] denotes significance at the 10% [5%] level. All explanatory variables of interest are presented in bold. Again, the logistic regressions regarding the presence of a female auditor (3) and Big Four engagement (4) could not be run. However, the relation between compliance on both boards and female representation in nomination committees can be tested now. Four of the eight models turn

(34)

34 out to be overall significant. For these models, including the explanatory variable of interest fits the data significantly better than the models without it. Regarding the hypothesized determinants of compliance with the board gender quota – the existence of a nomination committee and female representation in the committee – no statistical support was found. However, both the presence of an audit committee and the presence of at least one woman in this particular committee does turn out to significantly positively influence the probability of explaining non-compliance with board gender quota. This supports Hypotheses 2a and 2b. Again, in the regression regarding Hypothesis 2b, some coefficients for the control variables are significant. Specifically, those regarding leverage, total assets and whether the company is listed.

TABLE 10LOGISTIC REGRESSIONS FOR HYPOTHESES 1 AND 2, INCL. FIRM CHARACTERISTICS AND ONE -DIGIT SIC INDUSTRY CONTROL VARIABLES

1ai 1aii 1aiii 1bi 1bii 1biii 2a 2b

Constant 2.43E-08

(-0.02) -1.2E-10 (-0.02) 6.37E-09 (-0.02) 11.895 (0.47) 0.000 ** (-2.0) 689344.4 (1.42) (-0.01) 0.000 1.31e+11** (3.16)

NomCom 1.033 (0.05) 2.0777 (1.27) (-0.07) 0.956 FemNom 2.203 (0.94) (1.40) 3.633 24.956* (1.90) AudCom 3.627** (2.350) FemAud 29.885** (2.650) ROA 1.006 (0.21) (0.65) 1.032 (0.95) 1.025 (0.67) 1.069 (0.52) 1.063 (0.12) 1.014 (-1.39) 0.951 (-0.84) 0.940 ROE 1.003 (1.00) (-0.19) 0.997 (0.20) 1.001 (-0.39) 0.990 (0.03) 1.001 (-0.20) 0.999 (0.82) 1.003 (-0.56) 0.994 Leverage 0.204 (-1.32) (-0.29) 0.676 (-0.58) 0.493 (0.10) 0.753 (-0.73) 0.205 (0.26) 2.528 (0.92) 2.775 39.349* (1.89) Log (Assets) 1.356 (1.60) 1.860** (2.88) (1.61) 1.409 (-0.76) 0.789 1.768* (1.65) 0.267* (-1.77) (-1.24) 0.780 0.132** (-3.43) Revenue per Employee (-0.44) 1.000 (-0.94) 1.000 (-0.54) 1.000 (1.36) 1.001 (-0.73) 0.999 (-0.16) 0.999 (-0.41) 1.000 (1.20) 1.000 Listed 0.828 (-0.26) (-1.38) 0.340 (-0.98) 0.423 (-0.46) 0.565 0.096 * (-1.90) (0.56) 3.217 (1.23) 2.084 8.942** (2.48) Log likelihood -90.2488 -58.3258 -78.3530 -22.5922 -26.4565 -13.3122 -80.760 -29.910 Number of observations 258 149 267 54 60 41 241 79 LR chi2_{(df) 9.00 (15)} _27.34 (14) 9.30 (15) 6.57 (11) 20.39 (12) 10.85 (9) 45.15 (15) 41.650 Prob > chi2 _0.8778 _0.0174 _0.8616 _0.8331 _0.0600 _0.2859 _0.0001 _0.0001 Pseudo R2 _0.0476 _0.1899 _0.056 _0.1269 _0.2782 _0.2896 _0.2185 _0.4104

Each cell shows the estimated coefficient and the z-value (in parentheses). A superscript * [**] denotes significance at the 10% [5%] level. All explanatory variables of interest are presented in bold.

Additional regressions excluding the SIC control variables were also carried out. This further improves overall significance of most logistic regressions. The results are displayed in Table 11.

Gender diversity quota : comply, explain or disregard?