The influence of board diversity and board size on corporate social responsibility reporting : evidence from Dutch listed firm

The influence of board diversity

and board size on corporate social

responsibility reporting

evidence from Dutch listed firm

Name: Daniëlle van den Heuvel Student number: 10507701 Supervisor: Patrick Stastra Thesis field: Organization

Bsc Business and Economics, specialisation Finance and Organization Faculty of Economics and Business, University of Amsterdam

II

Statement of Originality

This document is written by Student Daniëlle van den Heuvel 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.

III

Abstract

Over the last few decades literature on the influence of board diversity has expanded (Adams et al., 2010). Additionally, recognition of the importance of directors’ characteristics beyond mere independence has increased (Nielsen and Nielsen, 2012). In light of the current debate on board composition (e.g. gender quota) and new regulations on mandatory reporting standards by the European Commission and Dutch Government, this thesis investigates whether the composition of the board has a significant impact on a company’s CSR reporting content and quality. The focus lies on relation-oriented attributes of board diversity (e.g. board members’ age, gender and nationality) and on board size. The CSR reporting is valued using the annual benchmark report of the Dutch Ministry of Economic Affairs and Climate Policy on CSR disclosure content and quality. Data is gathered for the years 2014-2017 on 46 companies. The random-effects and population-averaged estimators both show that gender and board diversity have a significant effect on the CSR reporting score. The estimators of age and nationality diversity are both insignificant. Since the regressions indicate a significant effect, there is evidence to support the main objective of this thesis. The diversity of a board does positively affect CSR reporting. This positive influence indicates diversifying the board can help firms overcome stakeholder issues. However, there were some limitations. First, even though the sample size was large enough it needs to be further expanded before the results can be generalized. Additionally, the use of firms from one single country could subjected to by country-based regulations. Furthermore, the operationalization of the nationality variable could be more in-depth, which could possibly be essential due to continuing globalization. Finally, other relation-oriented attributes, such as religion and language difference could be further researched.

IV

Table of Contents

1. Introduction ... 1 2. Literature Review ... 3 2.1 CSR activities ... 3 2.2 CSR disclosure ... 4 2.2.1 Stakeholder theory ... 4 2.2.2 Legitimacy theory ... 5 2.3 Board diversity ... 5 2.3.1 Age Diversity ... 6 2.3.2 Gender Diversity ... 7 2.3.3 Nationality Diversity ... 8 2.3.4 Board size ... 10

3. Data and methodology ... 11

3.1 Data ... 11 3.2 Conceptual model ... 12 3.2 Dependent Variable ... 12 3.3 Independent Variables ... 12 3.3.1 Age ... 13 3.3.2 Gender ... 13 3.3.3 Nationality... 13 3.3.4 Board Size ... 14 3.4 Control Variables ... 14 3.4.1 Firm size ... 14 3.4.2 Firm Age ... 14 3.4.3 Industry ... 15 3.5 Descriptive statistics ... 15 4. Research Method ... 18

4.1 Panel data models ... 18

4.2 Hypotheses ... 18

4.3 Statistical Method... 19

4.3.1 Hausman test ... 19

4.3.2 Estimator test... 19

V

5.1 Correlation ... 20

5.2 Population-averaged estimator results ... 21

5.3 Random-effects estimator results ... 22

6. Conclusion and discussion ... 25

Bibliography ... 27

1

1. Introduction

Over the last few decades literature on the influence of board diversity has expanded (Adams et al., 2010). Additionally, recognition of the importance of directors’ characteristics beyond mere independence has increased (Nielsen and Nielsen, 2012). Traditionally, research on board diversity focused on the on task-related dimensions (e.g. tenure, education and functional background) rather than on relation-oriented attributes (e.g. age, gender and nationality). However, due to the growth of transnational corporations, these attributes also require substantial consideration. Adam and Ferreira (2009) stress that another source for increased board diversification can be found in corporate governance. Corporate governance involves balancing the interest of the company’s stakeholders within a system of regulations by which a firm is controlled. Stakeholders from political, social and business perspective have increased the demand for a demographically diversified board (Homan, 2017). An example of external pressure can be found in the discussion on implementing gender quota laws for executive and supervisory boards. In 2003, Norway was the first country to introduce a gender quota of 40 per cent for both men and women in company boards. Since 2013, the Netherlands has a target quota of 30 per cent for both genders. However, to date, this target quota is generally not met due to a missing legal foundation (Dekker, 2015). Research on board diversity has been approached from multiple perspectives. From its impact on corporate governance (Adam and Ferreira, 2009) to organizational innovation and firm performance (Chen et al., 2015), risk-taking (Gonzales and Hagendorf, 2016) and corporate social responsibility (Harjato et al. 2015). Literature on Dutch companies is scarce but displays a positive relationship between board diversity and firm performance (Homan, 2017).

Recently, the European Commission announced it will implement the United Nations Guiding Principles for Business and Human Rights, which consist of guidelines on corporate responsibility. This is the responsibility of enterprises for their impact on society. As of January 2018, organizations of public interest in the Netherlands are obligated to report on the environment, social aspects, human rights, fraud/corruption and diversity. In academic literature, this non-financial reporting is commonly referred to as corporate social responsibility (CSR) reporting or disclosure. Non-financial disclosure finds its origin within corporate governance and can be linked to many economic theoretical frameworks (e.g. agency, stakeholder and, legitimacy theory). As mentioned corporate non-financial subjects have gained sufficient attention over the last decades. Guidelines and reports on the quality of corporate non-financial disclosures started emerging in the 1990s but the alignment of the actual activities to its reporting often falls short (Lindblom, 2015).

2

In light of the current debate on board composition and new regulations on mandatory reporting standards, this thesis will investigate whether the composition of the board has a significant impact on a company’s CSR reporting content and quality. There has been prior research on this topic. However, the result of this research is mixed (Barako and Brown, 2008; Khan, 2010; Fernández-Feijoo et al., 2012). The objective of this thesis is to identify the main board member characteristics that may have a significant influence on the amount and quality of voluntary social and environmental disclosure made by listed companies in the Netherlands. This specific influence has not been researched before in the Netherlands.

This thesis focusses on the relation-oriented attributes of board diversity (e.g. board members’ age, gender and nationality). In addition, the possible influence of the board size will be studied. The CSR reporting will be valued using the annual benchmark report of the Dutch Ministry of Economic Affairs and Climate Policy on CSR disclosure content and quality. Data is gathered for the years 2014-2017 on 46 companies. The panel data thus contains 184 firm-year observations. The random-effects and population-averaged estimators both show that gender and board diversity have a significant effect on the CSR reporting score. The estimators of age and nationality diversity are both insignificant. This thesis validates the recent gender quota regulations beyond mere firm performance effects. However, even though the sample size was large enough it needs to be further expanded before the results can be generalized. Another limitation is the use of firms from one single country which could be subjected to country-based regulations. Additionally, the operationalization of the nationality variable could be more in-depth.

In the next section, literature and theories related to board diversity and CSR reporting will be introduced and discussed. They will be used to form the hypotheses tested in this study. Section 3 elucidates the methodology of this research and encloses information on the used data. In section 4 the research method will be clarified. It includes the regression models and descriptive statistics. Section 5 shows the results. Additionally, all potential statistical problems will be addressed. Lastly, section 6 holds the conclusion and discussion.

3

2. Literature Review

This section will provide more detail on CSR and CSR reporting. Thereupon, relevant theories and empirical evidence that explain how board member diversity and board size can influence CSR activities and disclosure are examined. Based on this theory four hypotheses will be formulated.

2.1 CSR activities

In the last century, interest in the relationship between large business organizations and society has risen. Hetherington (1969) was the first to publish on the subject of corporate social responsibility stating that recent events in the United States indicated towards a new and different dimension of corporate responsibility. Since the 1970s awareness and concern about environmental issues has increased amongst environmentalist, governments and society in general (Monteiro and Aibar-Guzman, 2010). Companies have been regarded as major contributors to global warming and environmental pollution, and thus have been faced by increasing pressure to reduce their impact on the natural environment (Dunlap and Scarce, 1991). Over the last three decades, the United Nations have hosted multiple Conferences on Environment and Development, where the Declaration on Environment, consisting of principles to guide countries into future sustainable development, has been produced and expanded. In order to meet the goals set in this agreement governments impose regulations on companies based in their countries such as the new reporting regulation in the Netherlands. Likewise, in compliance with these recent developments company stakeholders have been increasingly demanding companies to accept their social responsibility and environmental impact (Hackston and Milne, 1996).

Literature on the relationship between CSR and board characteristics mostly relies on two theoretical views: agency theory and resource dependency theory (Jensen and Meckling, 1976; Pfeffer and Salinick, 1978). Agency theory explains the relationship between principals (stakeholders) and agents (corporate executives). This theory posits that efficient management will align corporate interest with long-term goals of all company stakeholders and thus promote CSR (Chang et al., 2012). Htay et al. (2012) suggest that managers can exploit information asymmetries if they desire to act in a manner contrary to the interests of stakeholders. Information disclosure reduces information asymmetry and is thus an integral part of corporate governance. The resource dependency theory holds that an organization’s external resources influence its behaviours and that a company’s management plays a vital role in acquiring these resources and information. In the context of CSR, a comprehensive understanding of stakeholders’ interests and demands relies on the managers’ qualities and is needed in order to efficiently implement CSR (Chang et al., 2017).

4

2.2 CSR disclosure

In order to comply to stakeholder’s demands companies have shifted to voluntarily providing information on environmental and social aspects in their annual reports (Monteiro and Aibar -Guzman, 2010). According to Cowen (1987), this voluntary disclosure on non-financial company information started developing in the 1980s when pressure from government regulations and environmental organizations increased and companies started being held accountable towards other stakeholders than just their shareholders. However, CSR disclosure does not necessarily guarantee quality CSR. Adams (2004) introduces the reporting-performance portrayal gap as a measure of the incompleteness of the reporting released by companies between 1993 and 1999. Guidelines developed by the Global Reporting Initiative (GRI) and other organizations did not seem to affect the quality of reporting. In 1993 KPMG started releasing a report on worldwide CSR disclosure. In their tenth edition, published in 2017, they reviewed CSR reporting from 4900 companies in 49 countries. As of 2004 the Dutch Ministry releases an annual report on CSR disclosure in Dutch companies. The rise of such reports, in addition to new legislation, indicates the increasing importance of CSR and CSR reporting for companies, government and, the public. Literature on CSR disclosure is built upon two theoretical frameworks: the stakeholder theory and the legitimacy theory.

2.2.1 Stakeholder theory

In his stakeholder concept, Freeman (1984, p.46) defines “a stakeholder as any group or individual who can affect or is affected by the achievement of the organization’s objectives”. A major objective of a firm is to balance the conflicting demands of the various stakeholders. In the stakeholder theory, the behaviour of various stakeholders is considered to be a constraint on the managers’ strategy. Freeman categorizes his concept into a corporate planning and business policy model, and a corporate social responsibility model of stakeholder management. This first model focusses on the approval of strategic decisions by relevant stakeholders. The second model adds to the first by allowing for changes in the social demand of non-traditional power groups (i.e. external influences).

Ullman (1985) expanded Freeman’s theory by developing a contingency framework for predicting levels of CSR activities and reporting. His framework presents a model linking stakeholder theory to CSR activities by using CSR disclosure. He states that if CSR activities can be used as an effective management strategy for dealing with critical stakeholders, a positive relationship between stakeholder power, CSR performance and CSR reporting is expected. However, CSR results are likely to be secondary to meeting the economic demands of stakeholders. Companies are thus more likely to disclose on CSR matters when their

5

(financial) performance is favourable (Mills and Gardner, 1984). Roberts (1992) tested Ullman’s theory and found significant effects of multiple stakeholder characteristics on CSR disclosure.

2.2.2 Legitimacy theory

The link between a company and society can be viewed as a social contract (Gray et al., 1996). CSR disclosure is considered to represent an important tool by which management can influence external perceptions about an organization. The legitimacy theory states that organizations continuously try to ensure that their operations fall within the bounds and norms of their respective societies. Within this theory, a firm would voluntarily report on activities if the management perceived that they were expected by the community the firm operates in. Firms are incentivized to use disclosure to enhance their legitimacy. If the company fails to meet up to the expectations of society (i.e. the social contract is breached), society will revoke the organization’s contract to continue its operations (e.g. consumers stop buying its products). This social contract is difficult to define because it can be explicit or implicit and is not permanent. Managers can thus have different perceptions about these various terms (O’Donovan, 2002). The legitimacy gap represents the difference between the expectations of how an organization should act and how the organization does act (Lindblom, 1994). Campbell et al. (2003) studied the extent to which voluntary disclosure represents an attempt to close the legitimacy gap but find inconclusive results. Deegan et al. (2002) state that CSR disclosure is more about reducing exposure than about providing meaningful and transparent information on social and environmental aspects of the firm. Chauvey and Giordano-Spring (2015) endorse this statement in their study of CSR disclosure amount and quality between 2004-2010 in France. They find a significant increase in CSR disclosure space and some evidence of increased quality. However, Deegan’s assumptions associated with the legitimacy theory still hold. As quality characteristics for CSR information, they identified: relevance, comparability, verifiability, clarity and, neutrality (Chauvey and Giordano-Spring, 2015).

2.3 Board diversity

Both task-related and relation-oriented attributes are part of an individuals’ human capital: an individual’s knowledge, habits, social and personal attributes embodied in the ability to perform labour (Becker, 1962). Hambrick and Mason (1984) state that human capital is a key determinant of organizational outcomes in their upper echelons theory. Research on diversity has mostly been focused on task-related attributes which are often associated with positive consequences. When regarding relationship-oriented attributes it is stated that limited information, bias perception and filtering influence managers’ objectivity during decision-making processes (Homan, 2017). According to Carter et al. (2010), varying characteristics

6

help to ease the understanding of complex environments. Horwitz and Horwitz (2007) add that heterogeneous (i.e. diverse) teams display a broader range of knowledge and information. However, Pfeffer (1985) states that social similarity is a viable asset is for communication and coherence. Social dissimilarities between colleagues could thus lead to dissatisfaction and ineffectiveness. Li and Wu (2015) add that diversity could lead to increased group conflict and reduce workforce collaboration.

Bear et al. (2010) studied how board member diversity affects CSR and how, in turn, CSR influences corporate reputation. Significant results were found on the relation between gender composition of the board and CSR. Additionally, CSR positively impacted corporate reputation, mediating the relationship between female board members and corporate reputation. Harjoto et al. (2015) studied the positive effects of gender, tenure and expertise as driving factors of firms’ CSR activities. Dienes and Velte (2015) researched the influence of multiple board member characteristics on CSR reporting in Germany, which has a similar two-tier system in corporate boards. They found a significant relationship between women on boards and CSR reporting intensity. This thesis studies the influence of relations-orientated personal attributes (i.e. age, gender and nationality) and board size. In the next paragraphs, these characteristics will be further discussed and the corresponding hypotheses will be established.

2.3.1 Age Diversity

Research on human capabilities shows that people of different age categories show different types of knowledge and cognitive abilities (Bugg et al. 2006; Horn and Cattell, 1967). Young individuals often have more knowledge of recent technologies and adapt to new situations easier due to their greater mental capacity (i.e. learning speed and efficiency). They have a higher ability to solve new problems, use logic in new situations and, identify patterns. In addition, they display better access to their long-term memory and are more able to use their skills knowledge, and experiences. However, with age individuals acquire skills that are assumed to improve with experience and learning (Shock, 1962; Welford, 1977). Additionally, older individuals have a higher working morale, awareness of quality and exhibit more know -how (Grund and Westergaard-Nielsen, 2008). In context of the stakeholder theory, older manager’s human capital could make them able to align different stakeholder’s interest more effectively. In addition, these qualities could also mean they have access to more viable resources, which is in line with the resource dependency theory.

Age diversity can be categorized into two age-related differences: differences in knowledge and values (Homan, 2017). Age-related knowledge is gathered with time and when it is shared, combined and integrated within groups it could improve the quality of

decision-7

making, creativity, efficiency, problem-solving and productivity. Value differences stem from historical experience and life-stage influences. Board composition can be either homogeneous or heterogeneous in regard to these age-related knowledge and values. Literature on promoting age diversity within boards is mixed. Lawrence and Lorsch (1986) state that the frequency of communication within a board is positively associated with homogeneity in age. Similar values ensure goal coherence and communication (Murray, 1989). In contrast, a heterogeneous board can ensure a more efficient division of labour due to capitalization of human capital (Houle, 1990). Lazear (1999) argues that the human capital of young and old workers is usually complementary to each other. Findings of empirical studies on the effect of board age diversity on firm performance are inconsistent (Mahadeo et al., 2012; Kilduff et al., 2000; Zimmerman, 2008; Murrray, 1989).

There has not been any study on the influence of age diversity on CSR disclosure. In light of CSR, Diamantopoulos et al. (2003) argue that younger individuals are more concerned about the environment and generally possess more knowledge on environmental issues. However, Ruegger and King (1992) state that older individuals tend to have more moral reasoning and thus, should have more concern for CSR. In addition, Gardyn (2003) argues that older individuals appear to behave more in line with their environmental consciousness. Ben-Amar and McIlkenny (2015) found a positive relation between board effectiveness and the voluntary disclosure of climate change business impacts. Since previous literature does not indisputably answers the question if heterogeneous boards are more effective then homogeneous ones, nor does it irrefutable indicate what the effect of average age should be, this study will assume age has no effect on CSR reporting.

Hypothesis 1: Age diversity within board members has no effect on the content and quality of the firm’s CSR reporting.

2.3.2 Gender Diversity

Various countries are currently debating introducing some sort of female quota for managing and supervisory boards (i.e. Norway, Germany, the Netherlands). Literature on the effectiveness of such quota and the added value of women on boards is mixed when considering firm performance. This debate is fuelled by the so-called glass ceiling; vertical segregation leading to an absence of women in upper-level positions of companies. Barreto, Ryan and Smitt (2009) argue gender stereotyping to be one of the causes of this invisible barrier. Betz, O’Connell and Shepard (1989) use the structural and socialization approach to study the relationship between gender and ethics. They found men to be more concerned with money and advancement; women were more interested in relationships and helping others. Men were also twice as likely to engage in unethical behaviour in a business situation. These

8

findings are in line with theory on the different moral orientations of men and women (Gilligan, 1982). Women conceptualize moral question as problems of care (e.g. compassion and empathy), while men view them as problems of rights. These moral differences are closely linked to views on CSR activities and reporting importance.

Arguments for increasing gender diversity are either ethical of business case-based (Robinson & Dechant, 1997). The ethical argument is based on equality reasons. Business case argument state that an enhancement of diversity will lead to a better firm performance, in line with the upper echelons and human capital theory. Women value their responsibilities as manager more, leading to more effective corporate governance (Miller and Triana, 2009). However, Tharenou et al. (1994) suggest evidence of female human capital shows they are equally qualified as man in terms of several important qualities, but traditionally have made fewer investments in education and business experience, reflected in lower pay and less promotion. Singh et al. (2008) refute this claim since they found women in 100 UK firms to be more likely to possess an MBA degree and have internal experience. Still, they are less likely to have COO/CEO experience. These findings are in line with recent developments in the Netherlands, in 2008-2009 twenty per cent more women attended a (research) university then men (Stoffelen, 2010). According to Oakley (2000) women are not rewarded with the same organizational opportunities (i.e. training or development programs), nor promotion or a higher salary by the male manager. This is commonly denoted as the main issue of the glass ceiling problem.

In accordance with the recent views on CSR there have been various studies investigating the effect of female board members on CSR and CSR reporting, which all found a positive relationship (Ferrero-Ferrero et al. (2012), Bear et al. (2010), Dienes and Velte (2016), and Setó-Pamies (2015)). According to Dienes and Vielte (2015), the relationship between female board members and CSR reporting intensity is likely to be connected to the effect women have on the improvement and performance of CSR activities. And since a company’s sustainability performance is represented in its disclosure, a higher proportion of women on the supervisory board could contribute to enhancing its reporting intensity. In line with previous empirical research, a positive relationship between the CSR disclosure score and women on corporate boards is expected.

Hypothesis 2: Gender diversity within board members has a positive effect on the content and quality of the firm’s CSR reporting.

2.3.3 Nationality Diversity

Nationality is increasingly becoming a more important dimension of board diversity in European countries (Ruigrok et al., 2005). Globalization and increasing worldwide operations

9

require firms to acquire knowledge of international markets and regulatory regimes. Ruigrok et al. (2005) studied the composition of nationality in top management teams and found a significant positive link between firm internationalization and internationalization of the management teams. They regard this increase as a response to internal and external pressure arising from internationalization processes. According to Estélyi and Nisar (2016), board member nationality could make an important contribution to determining stakeholder’s interest. Individuals from dissimilar backgrounds are likely to bring different perspectives due to different life experiences. They add that foreign board members could influence the quality of decision-making because they have different interpretations of a situation. Nationality diversity would thus reduce individual biases and prejudices within groups. In addition, internationalization of the board could enhance the company’s international network and satisfy foreign stakeholders as they might feel their interest are better represented by a foreign manager, this complies with both the resource dependency and the stakeholder’s theory (Oxelheim et al., 2013).

Carter et al. (2010) provide empirical evidence that ethical diversity on corporate board positively influence performance. Hillman et al (2002) found demographical characteristics of directors to have an influence on strategic decision-making because of the differences in human capital. Peterson et al (2007) found both women and African-American directors to assume different roles on the board relative to male and/or Caucasian directors. They suggested this could also be related to their unique human capital. Ruigrok, et al. (2007) investigated the educational background of foreign managers at Swiss public listed companies and found significant evidence that foreign managers are more likely to have a business education background (e.g. MBA or compatible master’s degree) than their domestic colleagues. They suggest a person’s chances for worldwide corporate success depend on the level and nature of their education. In contrast, DuFrene et al (2009) bring forward the issue of possible cross-cultural communication problems among board members with a different cultural background.

As with literature on the relationship between board members’ age and CSR disclosure, literature on board nationality diversity and CSR disclosure is limited. Haniffa and Cooke (2005) studied the influence of the ethnic background of managers and shareholders on CSR reporting for Malaysian firms but found a significant relationship between domestic managers and CSR disclosure. In contrast, research has shown a positive relationship between foreign managers and CSR disclosure in Bangladesh and Kenya (Kahn, 2010; Barako and Brown, 2008). Due to the inconsistency of previous research, the absence of an effect of board nationality on CSR reporting is expected.

10

Hypothesis 3: Nationality diversity within board members has no effect on the content and quality of the firm’s CSR reporting.

2.3.4 Board size

Both theoretical and empirical literature on the influence of board size on CSR reporting is mixed. While Htay et al., (2012) find a significant negative relationship in their study on CSR reporting, Rao et al. (2012) and Jizi et al. (2013) find a significant ly positive one. Moreover, multiple studies on CSR disclosure and board size have insignificant results (Said et al., 2009; Cheng and Courtenay, 2006; Amran et al. 2013). According to Jensen (1993), less CSR reporting could result from the higher agency costs large board require. Kassinis and Vafeas (2002) relate their negative relationship between CSR and board size to less coherency and thus less ability to make decisions. In contrast, Dalton et al. (1999) consider the presence of more human capital within larger boards to positively influence CSR reporting. Larger boards with frequent meetings often display more efficient monitoring, less manipulation of financial results (Xiao et al., 2005) and , encounter fewer problems regarding asymmetric information (Kanagaretnam et al., 2007). According to Pferrer (1973), Osemeke (2012) and Frias-Aceituno et al. (2012) board size is positively related to both CSR and CSR reporting. However, both Karamanou and Vafeas (2005) and Fuente et al. (2017) found insignificant results for relationship between board size and CSR reporting. Again, due to the mixed literature on the relationship between CSR reporting and board size the absence of a relationship is expected.

Hypothesis 4: Board size has no effect on the content and quality of the firm’s CSR reporting.

Board size is generally measured as the total amount of executive board members. However, since Dutch regulation prescribes a two-tier system, consisting of a managing and supervisory board this method cannot be applied. Additionally, the managing board can be divided into executive and non-executive members. The executive board members are commonly referred to as the executive board. Lückerath-Rovers (2014) examined both the executive board and the supervisory board in her study on board composition in the Netherlands. Since the definition of the board is inconsistent in previous literature on CSR reporting with Homan (2017) studying only the managing board and Dienes and Velte (2017) solely investigating the supervisory board, the combination of the two could provide some new insights. Moreover, the managing board generally has more influence on the CSR activities while the supervisory board has more influence on the repor ting of those activities. Since one cannot report on activities that do not exist and the CSR reporting score evaluates both, including the executive and supervisory board is the preferred option.

11

3. Data and methodology

This section describes the research method used in this study. First, the data sample will be discussed. Secondly, the conceptual model will be introduced. Consequently, the dependent, independent and control variables will be discussed and linked to previous literature. Lastly, the descriptive statistics will be given.

3.1 Data

The methodology on which the dependent variable (CSR reporting score) in this study is based is new to literature. In the benchmark reports by the Dutch Ministry, the content and quality of CSR reporting in annual reportsfrom more than 400 large companies is valued. The report is concurrent with the measure of quality CSR identification by Chauvey and Giordano-Spring (2015). The score in these reports is based on companies’ annual reports and their self-assessments. Both documents are analysed by consultants of Ernst and Young, and any irregularities are pointed out and discussed. Based on these documents the scores will be determined, these are again reviewed by EY. The quality of the assessment is monitored by a panel of twelve experts. They also assess the highest scoring companies and evaluate the quality-based scores. The criteria and corresponding scores can be found in appendix A.I. In 2014 the benchmark report adopted international developments such as the GRI guidelines and since 2018 regulation on CSR reporting in the Netherlands has changed. Due to these developments, the decision is made to solely use the reports of 2014-2017.

The company selection in this study is based on the available information. The data on board composition, board member’s characteristics and total assets were computed manually using the annual reports (in line with e.g. Adams et al., 1995, 1998; Gray et al., 1996; Khan, 2010). The company age is collected using the historical information on each companies’ websites. Companies’ CSR reporting is valued using the annual benchmark report. In 2017 there were 85 Dutch NV’s (ltd.’s) listed on the Euronext Amsterdam, in this study 46 companies were used. The sole use of listed firms is to rule out any distortions created by different regulations, these firms all have the same obligation to publish extensive annual statements needed for this study. The decrease in the number of companies has multiple reasons. Fifteen companies were dropped because they weren’t listed on the Euronext for the four consecutive years. Another twenty-two due to absence in one or more of the benchmark reports. One company was discarded because of major developments within the firm and another because of a missing annual report for 2017. More detailed information on the company selection can be found in appendix A.II and A.III.

The structure of the data violates the assumption of independence across observations necessary for OLS regressions (Chang et al., 2017). The use of longitudinal panel data is in

12

line with other similar studies (Carter et al., 2010; Chang et al., 2017; Homan, 2017; Huijsmans, 2017). It allows for both cross-sectional and time-series components. The final dataset consists of panel data of 184 firm-year observations.

3.2 Conceptual model

This study focuses on the influence of board diversity and board size on CSR reporting. Board diversity is a broad subject, lacking a widely accepted definition. This thesis constituted board diversity by using proxy measures for the age, gender and nationality of individual board members. The goal is to test whether these factors influence the quality of CSR reporting by the company. The basic regression model is as follows:

CSR reporting score =

β

0

+ β

1

age + β

2

gender + β

3

nationality + β

4

board size + β

5

firm size

+ β

6

firm age + industry dummies + ε

3.2 Dependent Variable

The dependent variable in this study will be the score on CSR reporting from annual benchmark reports. As mentioned, the use of this particular score as a (dependent) variable is new to literature. Most research on CSR reporting uses a binary coding technique (Barako and Brown, 2008; Htay et al. 2012; Gamerschag et al., 2011). Some added the length in term of words per reporting subject as an additional variable (Haniffe and Cooke, 2005; Khan et al., 2010). These measures are all on a quantitative basis. Both Karamanou and Vafeas (2005) and Fuente et al. (2017) used the Global Reporting Initiative guidelin es and manually generate binary scores based on the alignment between the disclosure and these guidelines. This method provides somewhat more detail on the quality of the CSR reporting. However, the report of the Dutch Ministry is based on similar guidelin es and is well-throughout in the way it is constructed as illustrated in the data section. The scores provided in this report reflect on both content and quality related topics. This report is thus the first to overcome the performance-reporting portrayal gap introduced by Adams (2004). In the report a maximum of 200 points can be received. All scores are divided by 200 to find values between 0 and 1.

3.3 Independent Variables

The independent variables have been widely discussed in the literature review. This section discusses the chosen operationalisation of all variables, backed by theoretical evidence. An overview of all variables can be found in appendix table A.IV.

13

3.3.1 Age

The independent variable age can be approached in two ways: as the average age of all board members or as the variation among board members (i.e. an indicator of heterogeneity). Both Post et al. (2011) and Hafsi and Turgut (2012) used the average age when researching the influence of board diversity on CSR activities and found results contradicting with their theory. Grund and Westergaard-Nielsen (2008) used average age and standard deviation of age and found both to be effective. However, when searching for the influence of age using the standard deviation provides more information. The influence of young or old managers could potentially be more interesting than the influence of the median aged managers. In addition, average age could be an inconclusive measure since a homogeneous and heterogeneous board could possibly have the same average age while the composition and thus its influence could significantly differ. In line with literature on firm performance, this study will use the standard deviation of age as a measure of heterogeneity within the board (Kilduff et al., 2000; Grund and Westergaard-Nielsen, 2008).

3.3.2 Gender

In regards of CSR reporting gender diversity has been approached in literature in multiple ways; as the proportion of female board members (Barako and Brown, 2008; Khan, 2010), as a count of female representatives (Bear et al., 2010), or by using a threshold for a minimal number of female representatives (Fernández-Feijoo et al., 2014). This last method is based on a study by Konrad et al. (2008) stating that women feel more comfortable being themselves in the presence of at least two other female colleagues and will thus feel freer to raise issues and are more active. Since smallest board size in this data set is five (and three women would implicate only two men) this last method is not applicable to this study. In addition, the use of a threshold has solely been used by Fernandez-Feijoo (2014) and his colleagues and has not been found in any other literature. The use of a proportion of women as a proxy for gender diversity is the most common method in previous research and will thus be used in this thesis.

3.3.3 Nationality

Ruigrok et al. (2007), Kahn (2010) and Barako and Brown (2008) used the absolute number of foreign board members as an indicator of diversity. However, when regarding diversity as heterogeneity within a board this method is not satisfying. The Blau (1997) heterogeneity index is another dominant measure in calculating nationality diversity in literature on board composition (Nielsen and Nielsen, 2012; Homan, 2017; Huijsmans, 2017 and, Estélyi and Nisar, 2016). This indicator illustrates the level of diversity between individuals by measuring the proportion of a board that belongs to a specific category (nationality in this case) and will be used in this study. It is calculated as:

14 Blau indicator = 1 - ∑pi2

Where pi denotes the proportion of board members who belong to a specific category (i)

divided by the total size of the board. The higher the indicator the higher the diversity within the board.

3.3.4 Board Size

Board size is often used as a control variable in literature on board composition due to its possible influence on firm performance (Homan, 2017). However, since the influence of its influence on CSR reporting is mixed (Htay et al., 2012; Cheng, 2006; Jizi et al, 2013), and this relationship has not been studied in the Netherlands, it is used as an independent variable in this study. As mentioned in section 2.3.4 it will be calculated as the total amount of executive and supervisory board members.

3.4 Control Variables

In all previous literature on related subjects, multiple control variables are added to the regression in order to eliminate their effects on CSR reporting

.

3.4.1 Firm size

According to Cowen et al. (1987), the association between large companies and CSR reporting is due to the fact that they are likely to undertake more activities and have a larger impact on society. In addition, they are subjected to more scrutiny by political and social stakeholders. Therefore, they are under greater pressure to report on their CSR activities in order to legitimize their business. These assumptions are in line with the stakeholder and legitimacy theory. This variable is introduced in the regression as the natural logarithm of the total assets of the company (Trotman and Bradley, 1981; Dienes and Velte, 2016; Haniffa and Cooke, 2005; Huijsmans, 2017; Chang et al., 2012; 2017).

3.4.2 Firm Age

A firm’s age corresponds to its life phase (Shabana et al., 2018). Different phases relate to different quantities of growth and development. Younger firms are less motivated to voluntarily disclose non-financial information due to the cost and complexity related to gathering such data. In contrast, older firms have a reputation and brand name to maintain. In addition, they are more likely to have sufficient capital to finance the additional reporting (Hossain and Reaz, 2007). Both Chang et al. (2017) and Moore (2001) added firm age to their regression on voluntary disclosure and found a significant effect.

15

3.4.3 Industry

From a theoretical perspective, industry effects should play a significant role in voluntarily non-financial reporting. Companies within a certain industry could produce a bandwagon effect on the level of disclosure adopted by companies in the same industry (Cooke, 1989). Disclosure quality and quantity are formed by the stakeholder’s demand which is related to a firm’s industry. Hence, disclosure levels between different industries are likely to differ reflecting their unique characteristics (Wallace et al., 1994). However, empirically this relationship has not been confirmed. Haniffe and Cooke (2005), Adams et al. (1998) and Brown and Deegan (1998) all used industry dummies and did not found significant influences for any of them. Branco and Rodrigues (2008) adapted two categories (more or less sensitive) and both were insignificant. Only Gamerschlag et al. (2011) found company and energy supplying industries to incorporate more CSR information, while service-related industries published less information in their CSR disclosure. Although the empirical results are mostly insignificant, based on the theoretical fundament industry dummies are added to the regression model. The industries’ division is taken from the benchmark report.

3.5 Descriptive statistics

Table I illustrates the descriptive statistics. In table II the panel data summary can be found. The mean of gender is 0.178 indicating the average board comprises of 17.8 per cent female members. This is in line with results found by Lückerath-Rovers (2014). The index for

nationality is relatively low, this is due to the fact that there are eleven companies in the data

set that solely have Dutch board members. There are no boards that scored a one on the

nationality index indicating none of the boards is completely heterogeneous. Firm size is

measured as the natural logarithm of total assets to reduce the effects of outliers. The mean of 7.597 indicates average total assets of €1,992 million. Firm age has some deviation between the mean of 66.435 and the median of 56. This is due to the distribution of the data (many younger and older firms, less middle-aged firms), there are no outliers. Theoretically, the CSR

index ranges from zero to one, however, none of the companies had a perfect score explaining

why the maximum is 0.995. For firm size the range is between 5 and 15, which translates to a range of €49 to €60,475 million.

Table II illustrates the summary of the panel data. It distinguishes the between (across individual) and the within (over time) variation of the variables. As shown firm age is a time-invariant variable (within variation is zero). For all other variables, the value of the between variation is larger than the within variation. If this is the case, the within estimation may lead to efficiency loss. However, since this study uses the between estimation (i.e. random-effects regression, see section 4.3.1) this is not an issue.

16

Table I

Descriptive Statistics

(in)dependent variables

mean

median

std. dev.

min

max

CSR index

0.629

0.668

0.266

0.0600

0.995

age

7.581

7.538

1.918

3.640

15.13

gender

0.178

0.167

0.121

0

0.500

nationality

0.346

0.278

0.311

0

0.861

board size

8.663

8

2.325

5

15

firm size

7.597

7.605

1.643

3.892

11.01

firm age

66.435

56

37.937

6

148

N = 184 for all variables

17

Table II

Panel Data Summary

variable

mean

std. dev.

min.

max

obs.

CSR index

overall

0.629

0.266

0.60

0.995

N = 184

between

0.255

0.14

0.969

n = 46

within

0.084

0.344

1.005

T = 4

age

overall

7.581

1.918

3.640

15.134

N = 184

between

1.767

4.575

13.743

n = 46

within

0.779

5.782

10.249

T = 4

gender

overall

0.178

0.121

0

0.500

N = 184

between

0.110

0

0.418

n = 46

within

0.051

0.136

0.328

T = 4

nationality

overall

0.346

0.311

0

0.861

N = 184

between

0.309

0

0.837

n = 46

within

0.055

0.127

0.652

T = 4

board size

overall

8.663

2.325

5

15

N = 184

between

2.277

5

13.25

n = 46

within

0.553

7.413

10.413

T = 4

firm size

overall

7.597

1.643

3.891

11.007

N = 184

between

1.645

4.179

10.898

n = 46

within

0.194

5.858

8.181

T = 4

firm age

overall

78.304

59.392

6

280

N = 184

between

59.885

6

280

n = 46

within

0

78.304

78.304

T = 4

18

4. Research Method

This section sets forth the regression models used in this study. Additionally, statistical hypotheses will be formulated. Lastly, the econometric tests will be introduced and evaluated.

4.1 Panel data models

To test the individual hypothesis stated in section 3 the four regressions stated underneath were created. All equations used in this study are based on the conceptual model demonstrated in section 3.1. In some cases, they are either expanded (with dummy variables) or reduced (in case of testing a single independent variable) depending on the aim of the regression. This is in line with similar research by Chang et al. (2017), Dienes and Vielte (2017), Huijsmans (2017) and Homan (2017).

(1) CSR indexit = β0 + β1ageit + β2firm sizeit + β3firm ageit + λt + εit

(2) CSR indexit = β0 + β1genderit + β2firm sizeit + β3firm ageit + λt + εit

(3) CSR indexit = β0 + β1nationalityit + β2firm sizeit + β3firm ageit + λt + εit

(4) CSR indexit = β0 + β1board sizeit + β2firm sizeit + β3firm ageit + λt + εit

(4) CSR indexit = β0 + β1ageit + β2genderit + β3nationalityit + β4firm sizeit

+ β5firm ageit + λt + εit

Where subscript i represents each company and subscript t represents the year. λt represents

industry dummies to control for industry-effects. And εit reflects the error term. The fifth regression all variables will be studies together, due to correlation (see section 5.7) board size is not added to this equation.

4.2 Hypotheses

The following hypotheses correspond to the hypotheses formulated in section 2.

(1) H0: β1 = 0 H1: β1 ≠ 0

(2) H0: β1 = 0 H1: β1 ≠ 0

(3) H0: β1 = 0 H1: β1 ≠ 0

(4) H0: β1 = 0 H1: β1 ≠ 0

19

4.3 Statistical Method

In panel data regressions three types of regressions are possible: between estimator (random -effects), within estimator (fixed-effects) and pooled OLS or population-averaged estimators (Cameron and Trivedi, 2010, pp.235-269). In order to decide between a random or fixed effects model, a Hausman test can be performed. It tests if the unobserved omitted firm-specific effects are uncorrelated with the independent variables. If the null cannot be rejected the use of a random effects test is preferred. This is the case because the random effects model assumes variation across the dependent variables to be random. It assumes that the company’s error term is not correlated with the independent variables and thus allows for time-invariant variables to act as explanatory variables. While the fixed effects model explores the influences of the independent variables on the dependent variable within each company that vary over time. This company effect is assumed to be a fixed over time and thus corrected for by the model. These time-invariant variables are unique to each company and should not correlate with any other company characteristics. In case of correlation between companies’ error term, the fixed effect model cannot be used. With pooled OLS and population-averaged (PA) estimators yit is simply regressed on xit using both between and within variation in the

data. These estimators are consistent if the random-effects model holds and are inconsistent if the fixed-effects model is appropriate. PA estimators can lead to more efficient estimators than pooled OLS. In the PA-measure individual effects are assumed to be random and are averaged out. In short-panel data (short time period) this mitigated the existence of non-independent observations.

4.3.1 Hausman test

The Hausman test is performed in order to decide between the random -effects and the fixed effects model for each regression (and to determine whether pooled OLS or PA estimators can be used). All results indicate (p > 0.05) that the random-effects model should be used (see appendix table A.VI). Additionally, the results indicate that pooled OLS and PA estimators is also appropriate for the data.

4.3.2 Estimator test

The decision between the OLS or PA estimators is based on the compassion of the size of the standard errors. With pooled OLS company clusters are added. It is essential the OLS standard errors are corrected for clustering on the company (Cameron and Trivedi, 2010, p. 250). In addition, the robust standard errors are used in the PA regression to ensure a fair comparison with the clustered standard errors of the pooled OLS regression. Details can be found in appendix table A.VI. All results indicate the use of the PA model.

20

5. Empirical results

In this section, all the results will be represented and clarified. Section 5.1 presents the correlation matrix and evaluates any potential problems. Sections 5.2 and 5.3 present the results from the population-averaged and random-effects regressions.

5.1 Correlation

To cope with non-normality total assets have been transformed by their natural logarithm, this improves their distribution in regard to skewness and kurtosis. The correlation between the unique errors and the variables in a regression is commonly referred to as the problem of endogeneity. Since the null hypothesis of the Hausman test cannot be rejected the problem of endogeneity is assumed to be absent (Huijsmans, 2017). Both the RE as PA estimator in this model are consistent but inefficient due to the within-panel correlation of the error terms (Cameron and Trivedi, 2010, pp.255-262). This problem can be solved by using robust standard errors, which are clustered on a company basis. These clustered robust standard errors eliminate any heteroscedasticity problems.

Table III

Correlation Matrix

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(1) CSR index

1.000

(2) age

-0.2486

1.000

(3) gender

0.3758

-0.1146

1.000

(4) nationality

0.1860

-0.2726

0.2337

1.000

(5) board size

0.4593

-0.0752

0.3424

0.6182

1.000

(6) firm size

0.4280

-0.0862

0.3415

0.3735

0.6801

1.000

(7) firm age

0.4970

-0.1677

0.2287

-0.1723

0.0079

0.2967

1.000

(8) banking

0.0658

-0.0241

0.1653

-0.0035

0.1863

0.0483

-0.1572

(9) construction

0.1015

0.1176

-0.2281

-0.3440

-0.1215

0.0482

0.2942

(10) media

0.0729

-0.0204

0.1945

-0.0522

-0.0091

0.0986

0.3949

(11) products

0.1548

-0.1078

0.0434

0.3734

0.0964

0.1338

0.1000

(12) real estate

-0.1012

0.0967

0.0553

-0.1978

-0.1759

0.0286

-0.0419

(13) retail

-0.0696

-0.0416

0.0189

-0.1300

-0.2559

-0.2381 -0.1334

(14) services

-0.1524

0.1750

-0.0827

-0.0903

-0.0229

-0.2135 -0.1824

(15) technics

-0.1072

-0.1889

-0.1004

0.1422

0.1203

0.0680

-0.1744

21

All variables are reviewed for multicollinearity. As demonstrated in table III the correlation between CSR index and the independent variables is relatively small. Board size, percentage of female board members, firm size and firm age all have a moderate correlation of > 0.3. The standard deviation of age has a relatively small negative correlation with the CSR index ( -0.2486). Nationality is the least correlated independent variable to dependent variable with 0.1860. Notably, the correlation between nationality and board size is large (0.6182). As a result of this large correlation, board size will not be added in the fifth regression where all independent variables are regressed together in order to avoid multicollinearity. Appendix table A.VII illustrates the correlation between the industries.

5.2 Population-averaged estimator results

Table VI reports the results from the population-averaged estimator regression. The first four regressions consist of one single independent variable and correspondent to hypotheses one to four. The last regression holds a combination of independent variables age, gender and nationality and tests hypothesis five.

From regression (1) and (3) it can be concluded that heterogeneity of the board in terms of age and nationality does not have an influence on the CSR index. This is in line with hypotheses (1) and (3). Regression (2) shows that there is a significant relationship between

gender diversity in terms of percentage female board members and the CSR index. The size

of the coefficient is 0.456 (z= 2.63, p=0.008) and reflects an increase of 9.12 point for every ten per cent increase of female directors. Again, this result is in line with the hypothesis. In regression (4) a significant influence of board size (z=2.94, p=0.003) on the CSR index is found. Even though board size is highly correlated with firm size, the effect of firm size is found to be insignificant in this regression, indicating that the board size is more valuable than firm size in regards of CSR reporting. The size of the coefficient is 0.034 indicating an increase of 7.8 points for every additional board member. This result is contradictory to the hypothesis which stated board size would not have an effect. In the (5th_{)regression independent variables}

age, gender and, nationality are combined and the results are in accordance with the individual

regressions. Both age and nationality are insignificant. Gender is significant with coefficient 0.444 (z=2.50, p=0.012). Since at least one of the coefficients of the independent variables is not zero the research question of this thesis that board diversity influences the CSR reporting score is confirmed.

When regarding the control variables. Firm age is the only one with a significant coefficient for all regression. The firm size coefficient in regressions (1), (2), (3) and (5) is significant. It is insignificant is regression four due to the high correlation with board size. The

22

industry dummies are insignificant in all regressions. This last finding is in accordance with previous empirical studies.

5.3 Random-effects estimator results

Table VII presents the results of the random-effects estimator regression. The results of this regression are similar to the population-averaged regression. In regressions (1) and (3) on independent variables age and nationality are again insignificant. The coefficients in regressions (2) and (4) for both gender (z=3.68, p=0.000) and board size (z=3.23, p=0.001) are again significant. With sizes 0.455 and 0.032 respectively they are close to the PA estimators. Regression (5) also shows similar results for age, gender and nationality. Again, only the fourth hypothesis is refuted. In the RE regression firm size is only significant in regressions (1) – (3). Firm age and the industry dummies have the same effects in both models.

23

Table VI

Population-averaged estimator regression

(1)

(2)

(3)

(4)

(5)

variables

CSR index

CSR index CSR index CSR index CSR index

age

-0.014

-0.012

(0.011) (0.010)

gender

0.456***

0.444**

(0.167) (0.172)

nationality

0.043

0.028

(0.086) (0.083)

board size

0.034***

(0.011)

firm size

0.038**

0.030*

0.036**

0.009

0.028*

(0.017) (0.016) (0.018) (0.021) (0.017)

firm age

0.003***

0.003***

0.004***

0.004***

0.003***

(0.001) (0.001) (0.001) (0.001) (0.001)

banking

0.134

0.084

0.129

0.103

0.097

(0.150) (0.159) (0.150) (0.118) (0.158)

construction

0.008

0.016

-0.015

0.011

0.058

(0.137) (0.143) (0.144) (0.110) (0.138)

media

-0.106

-0.156

-0.129

-0.113

-0.132

(0.106) (0.108) (0.102) (0.094) (0.111)

products

0.064

0.044

0.047

0.060

0.051

(0.095) (0.093) (0.093) (0.087) (0.095)

real estate

-0.061

-0.111

-0.070

-0.002

-0.077

(0.073) (0.061) (0.079) (0.067) (0.074)

retail

0.083

0.044

0.084

0.126

0.054

(0.081) (0.077) (0.083) (0.080) (0.077)

service

0.038

0.006

0.022

0.014

0.027

(0.076) (0.072) (0.079) (0.069) (0.073)

constant

0.185

0.095

0.072

-0.022

0.189

(0.179) (0.139) (0.149) (0.131) (0.158)

observations

184

184

184

184

184

number of companies

46

46

46

46

46

random effects

Yes

Yes

Yes

Yes

Yes

Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.10

24

Table VI

Random-effects estimator regression

(1)

(2)

(3)

(4)

(5)

variables

CSR index

CSR index CSR index CSR index CSR index

age

-0.014

-0.012

(0.011) (0.011)

gender

0.455***

0.443**

(0.173) (0.177)

nationality

0.024

0.012

(0.089) (0.086)

board size

0.032***

(0.011)

firm size

0.035*

0.028*

0.035*

0.010

0.026

(0.018) (0.017) (0.019) (0.021) (0.018)

firm age

0.003***

0.003***

0.004***

0.004***

0.003***

(0.001) (0.001) (0.001) (0.001) (0.001)

banking

0.133

0.084

0.126

0.105

0.095

(0.154) (0.164) (0.157) (0.123) (0.165)

construction

0.006

0.015

-0.023

0.008

0.049

(0.141) (0.148) (0.149) (0.115) (0.143)

media

-0.108

-0.157

-0.131

-0.115

-0.135

(0.108)

(0.110)

(0.105)

(0.097)

(0.114)

products

0.063

0.043

0.049

0.059

0.052

(0.098)

(0.096)

(0.096)

(0.090)

(0.098)

real estate

-0.062

-0.111

-0.078

-0.007

-0.084

(0.076) (0.064) (0.081) (0.069) (0.077)

retail

0.078

0.040

0.077

0.121

0.047

(0.085) (0.079) (0.087) (0.082) (0.081)

service

0.035

0.004

0.017

0.013

0.022

(0.079) (0.075) (0.083) (0.071) (0.076)

constant

0.203

0.111

0.090

-0.008

0.207

(0.190) (0.146) (0.158) (0.138) (0.171)

observations

184

184

184

184

184

number of companies

46

46

46

46

46

random effects

Yes

Yes

Yes

Yes

Yes

Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.10

25

6. Conclusion and discussion

This thesis aimed to identify the main board member characteristics that may have a significant influence on the content and quality of voluntary social and environmental disclosure made by listed companies in the Netherlands. The dependent variable CSR index is based on the annual benchmark report by the Dutch Ministry on CSR reporting. The independent variables studied are board member’s age, gender and nationality and the board size. These variables were selected based on previous literature and linked to multiple theories such as the stakeholder and legitimacy theory. The theoretical review displayed the absence of research on the influence of board member’s age on CSR reporting. Likewise, literature on the effects of nationality diversity and board size was mixed. Gender diversity and CSR reporting were said to have a positive relationship. The results from the population-averaged regression and the random-effects regression are quite similar and indicate a significant effect of both gender diversification and board size on the CSR index. The influence of diversity in terms of board members’ age and nationality was insignificant. Since the regressions indicate a significant effect, there is evidence to support the main objective of this thesis. The diversity of a board does positively affect CSR reporting. This study adds to literature by exploring a new method of evaluating companies’ CSR reporting and using a new combination of independent variables. Additionally, the relationship between CSR reporting score and board composition has not been studied in the Netherlands before.

With the confirmation of the research question comes the implication for firms to diversify their boards. With the increasing external pressure from stakeholders and through new European and Dutch regulation, CSR reporting is gaining importance. Since board diversity has a positive influence on the content and quality of firms’ CSR reporting, diversifying the board can help firms overcome stakeholder issues. In this thesis, the positive relationship between female representatives on boards and the CSR reporting has been proven. This validates the recent developments of gender quota for boards on a wider scope than mere firm performance.

Since globalisation and the growth of transnational corporations require knowledge on international markets and regulatory regimes, the nationality diversification within boards could potentially be essential in order to meetup to various stakeholders’ demand. The absence of an effect of nationality in this study could possibly be explained by the difficulties operationalizing this variable. Gray et al. (1996) illustrate that some countries are geographically close and economically similar, which could lead to substantial influence on corporate behaviour. In his case, a board consisting of a French, German, Dutch, Swedish, Swiss and UK member identify as more diversified than it is in reality. Alternatively, Ruigrok et

26

al. (2007), Kahn (2010) and Barako and Brown (2008) used the absolute number of foreign board members as an indicator of diversity. However, with this measure, the board illustrated by Gray et al. (1996) would still display more diversified than it is. The board members in this study were predominantly for a Western background and thus might not have been so diversified. When searching for a solution to this problem, the aim should be on the focus on how diverse different nationalities are in reality. Possibly looking into differences per continent or creating a scale comparing all found nationalities and scoring their differences should be further researched, for this last method Hofstede’s (1984) dimensions could potentially be used.

This study has some additional limitations. First, the sample size was relatively small. Due to changes in the scoring of the benchmark reports and changing regulations the data was limited to the years 2014-2017 which is considered to be a short-panel. Moreover, 39 out of 85 companies had to be dropped from the sample due to missing data. This led to 184 firm-year observations in the panel data regressions. Even though this is large enough, N does not approach infinity, which is idealistically desired. Secondly, even though random -effects estimators allow for generalizations beyond the regression sample. The data is obviously limited to the Netherlands and could thus be subject to country-bound effects. Conducting this research across multiple countries (including country clusters or dummies) and increasing the sample size would enhance the understanding of the effect of board member characteristics and board composition on the CSR reporting quality.

In addition to further research aiming at a larger sample including multiple countries and developing a more sophisticated method for rating the nationality diversity, the choice of relation-oriented attributes could be further investigated. Diversity on basis of religion or language could enhance the issue of possible cross-cultural communication problems and thus have an influence on board activities.