Board Composition & Corporate Social Responsibility Performance:

(1)

Master Thesis

Board Composition & Corporate Social

Responsibility Performance:

I n vestigating t h e Moderating Effect o f Organisational Ag e

by

Sjoerd Kwakkel

Student number: S3016811 E-mail: s.kwakkel@student.rug.nl

June 7, 2017

Supervisor: Drs. J. (Hans) van Polen (j.van.polen@rug.nl)

Co-assessor: Mr. P.J. (Paulo) Marques Morgado (p.j.marques.morgado@rug.nl) Word count: 14.459

MSc International Business & Management 2016 – 2017 University of Groningen

Faculty of Economics and Business PO Box 800

(2)

ABSTRACT

This thesis fills the gap in the literature by investigating the effect of board composition on the CSR performance of the firm, whilst combining it with the moderating demographic factor of organisational age. The moderating effect of organisational age on the relation between board composition and CSR performance has been studied by building on the Upper Echelons Theory. A sample of 165 European based firms is taken from the Standard & Poor’s Euro 2015 list, to test the effects. A (moderated) regression analysis method has been applied to test the hypotheses. Enough evidence has been found for the hypotheses suggesting a positive relation between board composition (in terms of board size and gender diversity) and CSR performance. No significant moderating effect of organisational age could be found for the relation between board size and CSR performance. At last, contrary to the expectation, a significant negative relation of organisational age has been found to affect gender diversity and CSR performance. This leads to the conclusion that organisational age negatively moderates the positive relation between board composition (in terms of gender diversity) and CSR performance. These findings add to both literature as well as the business practice in identifying the effect of board composition on CSR performance, with organisational age operating as a moderator variable.

Key words: Board Composition, Board size, Corporate Social Responsibility (CSR), Performance,

(3)

ACKNOWLEDGEMENTS

I would like to use this opportunity to express my sincere gratitude and appreciation towards a number of individuals who have helped and supported me during this thesis and the entire MSc program. First, I would like to thank my supervisor and co-assessor, Drs Hans van Polen and Mr. Paulo Morgado, for their support and useful feedback during this process. Second, I would like to express my gratitude to my girlfriend, Kelly, and my entire family for their support and believe in the successful completion of this thesis, and therefore International Business & Management program. Third, I would like to thank everyone who has contributed in any viable way, by providing tips, reviewing the report or supporting me during the process. I could not have completed this without the support of all of you. Lastly, but most important, I would like to thank the Lord, in which I believe everything has been made possible. “For the Lord gives wisdom; from His mouth come knowledge and understanding” (Holy Bible, 1982: Proverbs 2:6).

(4)

List of Abbreviations

Abbreviations

CEO – Chief Executive Officer

CSR – Corporate Social Responsibility

ESG – Environmental, Social, and Governance HR – Human Resources

HRM – Human Resource Management KLD – Kinder Lydenberg Domini database KPI – Key Performance Indicator

(5)

List of Figures and Tables

Figures

Figure 1 – Conceptual Model [page 21]

Figure 2 – Q-Q Plot for the Dependent Variable (CSR Performance) [page 57]

Figure 3 – Scatterplot of the Normal Residuals against the Unstandardized Residuals of

the Dependent Variable (CSR Performance) [page 58]

Tables

Table 1 – Overview of Used Variables [page 27]

Table 2 – Descriptive Statistics of the Sample Size [page 30] Table 3 – Descriptive Statistics of the Industry [page 31] Table 4 – Overview of the Regression Results [page 35] Table 5 – Analysis Overview [page 37]

Table 6 – Descriptive Statistics of the Country of Origin [page 56]

(6)

INTRODUCTION

Back in 1917, Henry Ford might have been one of the first to put Corporate Social Responsibility (CSR) into practice when he defended his decision to reinvest Ford Motor Company’s earnings on plant extension. He stated the purpose of the company as follows: “To

do as much as possible for everybody concerned, to make money and use it, give employment, and send out the car where people can use it ... and incidentally to make money” (Lee, 2008:

54). Henry Ford already showed a form of Social Responsibility back then by taking responsibility beyond the direct shareholders.

For several decades, CSR has increasingly received tremendous attention in both the business practice as well as in literature. The widespread attention of scholars from different fields resulted in the application of multiple theories (like the resource based view, stakeholder theory and transaction costs economics) and independent variables (mostly corporate size and financial performance) to the field of CSR, in order to expand the understanding of CSR within the firm (Caroll & Shabana, 2010; Cowen, Ferreri, & Parker, 1987; Frynas & Yamahaki, 2016; McWilliams, Siegel, & Wright, 2006; Moratis, 2016; Wang, 2015). Despite the widespread attention, several subjects within the field still requires (additional) examination. Studying the relation between board composition, organisational factors (like age) and CSR performance is one of the proposed subjects for future research. This subject only received limited attention (Elsayed & Paton, 2009; Rao & Tilt, 2016; Withisuphakorn & Jiraporn, 2016), even though its proven effect on the strategic decision process (Elbanna & Child, 2007; Wang, Gao, Hodgkinson, Rousseau, & Flood, 2014).

(8)

The need to study the effect between board composition and CSR becomes even more evident when having a closer look at the literature, and their calls for future research to increase the empirical evidence and understanding. Even back in 1972, Pfeffer already indicated that future research on additional dimensions of the board are needed in order to further identify the understanding of the “... organizational orientation to external

environmental factors” (Pfeffer, 1972: 226). O’Neill et al. (1989) also call for future research,

suggesting that background characteristics determine the level of corporate social responsiveness orientation, so additional research is needed to further identify this effect. Additional calls for research are based on the involvement of organisational factors like firm size, industry, and ownership, amongst organisational age, and their influence on the managers’ decision making process (Wang et al., 2014). Kor (2006) specifically calls for scholars to expand the field of study of board demographics on other organisational outcomes. More recently, some suggest that, as a result of the increasing importance of CSR to a firm’s strategy, future research of the influence of board composition on the firm’s CSR performance is vital (Rao & Tilt, 2016). In studying the effect of board composition on the CSR performance of the firm, and by adding the organisational factor of age, this study answers the long-time calls from previous scholars, and contributes to the predictability of the strategic decisions based on the board’s background characteristics (O’Neill et al., 1989).

(9)

moderating effect of organisational age on the relation between board composition and CSR performance, the following research question has been defined.

What is the moderating effect of organisational age on the relation between board composition and the firm’s CSR performance?

This study is considered to be important since the board of directors is both responsible and accountable for the firm’s actions with regards to the wide range of stakeholders, but the understanding of this is still scarce (Rao & Tilt, 2016). Further identifying the effects of board composition contributes to a better understanding of the boards ’ role and the effect of organisational age within this (following the study of Elbanna & Child, 2007). This study further clarifies the effect of human- and organisational factors in relation to the MNC’s CSR performance, hereby contributing to the academic literature. Subsequently, this increased understanding could lead to specific HR actions (specialised training or aimed hiring of board members and policy adaptions) on the business level, or to specific policy actions (e.g. stimulation of diversified boards), which stimulates the envisioned CSR outcome, therefore contributing to the business practice.

LITERATURE REVIEW

In this section, the literature will be reviewed for the different concepts used. Firstly, as for the dependent variable; the history of the CSR literature and the measurement of performance will be reviewed. Subsequently, the main theory applied to this study, the Upper Echelons Theory, will be explained in setting the guidelines of the independent variable. As what follows, the independent variable, board composition, will be discussed. In this section, the two chosen variables, which form the board composition, board size and gender diversity, will be revised, followed by the moderating effect of organisational age. This section will end with the formulation of the hypotheses and the conceptual model.

CSR performance

(10)

Davis, 1960; Grether, 1969). Later on, Carroll (1974) recapitulates on the subject and suggests guidelines for further discussion. As literature develops throughout the years, broader concepts are defined and studied. Sethi (1979), for example, developed a conceptual framework for firms to analyse the firm’s social issues and their response patterns. Buehler & Shetty (1975) and Jones (1980) added to this field of study by providing practical guidelines for firms to apply CSR. In reviewing the CSR debate, Walton (1982) identified the CSR performance rhetoric and the clash amongst definitions as a result of the widespread attention. In his 1983 paper, still in search for what CSR exactly should entail, Carroll provides an answer by stating that “CSR is comprised of four parts: economic, legal, ethical and

voluntary or philanthropic” (1983: 604), a concept which is later defined as the Pyramid of

CSR, and often referred to within the CSR context (Carroll, 1991; Crowther & Aras, 2008; Schwartz & Carroll, 2003; Wood, 2010).

As drawn here, the subject of CSR is characterized by a broad development and the application of several variables throughout the years. Some of the variables which have been studied in relation to CSR are, amongst others, financial performance (Balabanis, Phillips, & Lyall, 1998; Cochran & Wood, 1984; Q. Wang, Dou, & Jia, 2016; Zahra, Oviatt, & Minyard, 1993), governance modes (Bondy, Matten, & Moon, 2008; Calabrese, Costa, Menichini, & Rosati, 2013; Filatotchev & Nakajima, 2014; Jo & Harjoto, 2012), firm strategy (Areal, McInthosh, & Sheppy, 2016; Marshall & Brown, 2003; Martínez-Ferrero, Rodríguez-Ariza, & García-Sánchez, 2016), and human resources (Fenwick & Bierema, 2008; Jamali, El Dirani, & Harwood, 2015; Yi et al., 2015). All these scholars studied the role of CSR in different situations, thereby contributing to the overall literature and understanding of CSR within firms.

(11)

Due to the popularity of CSR, many scholars came up with different definitions, or made small adjustments to existing definitions (Clarkson, 1995; Crowther & Aras, 2008; Sarkar & Searcy, 2016). As a result of this diversity, scholars started debating their results and indicators of CSR performance in defending their studies, which led to flawed discussions (Schwartz & Carroll, 2003). In order to prevent any uncertainties about the definition of CSR in this report, the definition of CSR as given by McWilliams & Siegel is used (2001). This definition is believed to best encompass the broad field of CSR and is used by several other scholars as well (Chin, Hambrick, & Trevino, 2013; Filatotchev & Nakajima, 2014; Jo & Harjoto, 2012; Waldman et al., 2006). CSR engages in ‘‘actions that appear to further some social good, beyond the interests

of the firm and that which is required by law’’ (McWilliams & Siegel, 2001: 117).

Even though defining CSR gives a good indication about the concept of this study, it still does not say anything about how to measure the CSR performance. Previous scholars have measured CSR in numerous ways. Wood states that several performance measures have been developed, which include measurements such as “corporate image, responsiveness to social

demands ... ; community relations and charitable giving, absence of law-breaking behaviour; ... executives good intentions, particularly in handling a crisis; or anything else that suits the user’s purposes.” (Wood, 1991: 67). Additionally, many scholars tried to identify the

performance of CSR by using a financial indicator (Griffin & Mahon, 1997; Jo & Harjoto, 2012; Zahra et al., 1993). The diversity within the performance measures indicates the diversity within the definition of CSR performance itself. However, several scholars have suggested that the use of solely financial indicators, in identifying the effect of CSR, is not ideal (Jones & Wicks, 1999; McWilliams et al., 2006). McWilliams et al. (2006) state that CSR simply cannot be measured solely by financial indicators since this is a firm level indicator, whereas CSR activities mainly occur at the plant and/or product level, making it a restrictive indicator for CSR performance.

(12)

Even back in 1974, Carroll already identified that CSR knows several responsibilities like external and social responsibilities (Carroll, 1974). All these scholars indicate that the performance indicator of CSR must surpass the restrictive financial measure and should focus on social and governmental measurements, too.

Later on, it is noticed that more scholars use the social and environmental indicators as mentioned by Wood and Carroll, and the governance indicator as mentioned by Clarkson (Deckop et al., 2006; Marshall & Brown, 2003; Shaukat, Qiu, & Trojanowski, 2016). It is therefore believed that a good CSR performance indicator surpasses the financial measures and focuses on environmental, social and governance measures to indicate the CSR performance of the firm.

The Upper Echelons Theory

Before discussing the independent variable of board composition, the main theory (applicable to the independent variable) will be discussed. This study is based upon the Upper Echelon Theory, as described by Hambrick & Mason (1984), to explain the relation between board diversity and the firm’s CSR performance. The first notion of this theory was introduced by Hambrick & Mason, back in 1984. They introduced the concept that the top executives’ background partially predicts the organisational outcomes, in their paper ‘Upper Echelon: The organisation as a reflection of its top managers’. Based on bounded rationality, they suggest that top executives differently value, and act upon, the changing environment of the firm, thereby influencing the activities and performance of the firm (Hambrick, 2007).

More specifically, the cognitive base and values are identified as the important characteristics of the manager forming the background. This also includes the 1) “knowledge

or assumptions about future events”, 2) “knowledge of alternatives”, and 3) “knowledge of consequences attached to alternatives” (Hambrick & Mason, 1984: 195). The manager’s

(13)

O’Neill, Saunders & McCarthy’s (1989) study is one of the first to look at the background of board members and their effect on Corporate Social Responsiveness Orientation. Even though the Upper Echelons theory has been linked to CSR at an earlier stage by O’Neill, Saunders, & McCarthy (1989), it still took quite some time for the scholars to follow this lead. Swanson (1999) indicated that the personal values of a manager influence the strategy for CSR performance. Additional suggestions, provided by other scholars, indicate that the CEO perceptions and values are of influence in the environmental corporate social performance (Agle et al., 1999) and that CEO payment structure influences corporate social performance (Deckop et al., 2006).

Later on, Yi et al. (2015) studied the effect of the CEO’s hubris (the level of extreme self-confidence and pride) on CSR, in combination with the firm’s access to resources. An initial negative effect has been found which is weakened by a high external dependence on resources. Other studies focused on more specific subjects like entrepreneurial behaviour, Small- and Medium sized Enterprises (Hooi, Ahmad, Amran, & Rahman, 2016) and charitable donations (Wang et al., 2014). Nevertheless, studies focusing on board composition (based upon the Upper Echelons Theory) in relation with CSR, still remains scarce despite the fact that these are needed to further increase the board’s individual understanding of CSR within the organisation (Rao & Tilt, 2016; Waldman et al., 2006). Since it is the objective of this study to identify the board members’ characteristics in explaining their behaviour towards CSR, the Upper Echelons Theory has been applied.

Board composition

(14)

Scholars kept developing the literature throughout the years by adding more depth into their studies and by adding factors and variables. Other board composition studies focused, for example, on nationality, director’s age, educational background, board independence, and CEO duality (Bear, Rahman, & Post, 2010; Müller, 2014; Nas & Kalaycioglu, 2016; Post, Rahman, & Rubow, 2011; Rao & Tilt, 2016; Romano & Guerrini, 2014; G. Wang et al., 2016). Also several dependent variables have been studied in combination with this board composition like export performance, environmental performance, reputation, and financial performance (Bear et al., 2010; Dalton, Daily, Ellstrand, & Johnson, 1998; Kesner & Dalton, 1985; Müller, 2014; Nas & Kalaycioglu, 2016; Post et al., 2011).

Many of these studies rely on the Agency Theory in explaining the effect of the board on the different outcomes of the firm, and often find significant relations (Bear et al., 2010; Nas & Kalaycioglu, 2016; Sur, Lvina, & Magnan, 2013). However, the agency theory is not found to be explanatory in all cases (Dalton et al., 1998; Deutsch, 2005). The Upper Echelons Theory is believed to provide more explanatory power to the individual characteristics playing a role in the decision making process of managers, compared to the agency theory (Kor, 2006). The main reason for this can be found in the Agency Theory which focuses on the conflict of interests (individual accomplishments against corporate accomplishments) between the individual manager and other stakeholders (Fama & French, 1993; Jensen & Meckling, 1976). This theory provides explanations for managers to execute self-interested actions, something that is not aimed to understand in the Upper Echelons Theory, and this study. Within the Upper Echelons Theory, the characteristics, values and beliefs of a management team which drive a certain corporate (strategic) decision are sought after in order to explain certain behaviour (Hambrick, 2007).

(15)

of inside-outside orientation, average age, education and financial stake) and corporate social responsiveness orientation (O’Neill et al., 1989), and board composition (in terms of percentage of outside board directors, gender and average age) and environmental corporate social responsibility (Post et al., 2011), all finding positive correlations, except for one (the manager’s financial stake in the case of O’Neill et al.). These results lead to suggest that a variety of board indicators have a positive relation with the CSR performance of the firm. No further literature studying the effect of board composition on environmental issues could be found.

Board size

One of the variables, which form the board composition in this study, is the size of the board. Previous scholars already gave this variable attention in their studies, finding varied results. Many of the scholars studying the effect of board size on the financial performance of the firm found a negative relation (Peters & Bagshaw, 2014; Xie & Fukumoto, 2013). Where some scholars found no relation (Horváth & Spirollari, 2012), others found a positive relation (Kim, Cha, Cichy, Kim, & Tkach, 2012; Topal & Dogan, 2014), and a few even suggest an optimal size of the board (Garg, 2007; Jensen, 1993).

One of the studies to suggest a specific size of the board is Jensen (1993). These findings are supported by the suggestion that boards bigger than 7-9 individuals are less efficient and easier to control by the CEO. Garg (2007) adds to this argument by stating that an agency problem occurs for bigger boards, resulting in a more symbolic function instead of an efficient function of the management team. Nevertheless, these studies specifically focus on the financial performance of the firm. It is argued that the top executive compensation is sensitive to the size of the firm, where its profitability is not. Thus, the operating costs are increased whereas it does not directly affect firm value, thereby decreasing (putting pressure on) the firm’s operating performance (Nguyen, Rahman, Tong, & Zhao, 2015). Additional studies regarding the size of the board have focused on the export performance (Nas & Kalaycioglu, 2016), corporate failure (Chaganti et al., 1985), firm value (Nguyen et al., 2015), and CSR (Post et al., 2011), and provide less useful information for this study.

(16)

& Muhammad, 2015; Majeed, Aziz, & Saleem, 2015; Yusoff, Dalila, Jamal, & Darus, 2016). They all find that a bigger board increases the disclosure of CSR related activities for the firm. Whilst this does not imply that the actual performance will also increase, it does provide evidence for the positive effect the managers can have on CSR related issues within the firm.

When looking specifically at the relation between board size and CSR performance, Post et al. (2011) show a positive relation between board size and the environmental corporate social responsibility score of the firm. However, the significance was lost in their study when other board composition variables were added. Only de Villiers, Naiker, & van Staden (2011) are found to have directly studied the effect of board size on CSR, and to have indicated a positive relation. They argue that larger boards are more likely to increase CSR experienced managers (or even experts) to deal with specific issues, like environmental performance, which subsequently influences the environmental performance of the firm.

Aggarwal & Nanda (2004) studied the relation between the board size on both the managerial incentives as well as on the social objectives. They found that the board size is negatively related to the managerial incentives of the firm, implying that there are fewer incentives to increase shareholder value when the board is big. On the other hand, they found a positive relation between board size and social objectives, which include environmental issues, thereby indicating that a bigger board positively influences the social objectives of the firm. This study gives a plausible explanation for the diversified results on board size in the literature by explaining that the managerial incentives are negatively related whereas the social objectives are positively related to the number of top executives on the board.

With the results of Aggarwal & Nanda, and the additional studies, which indicate a positive relation between board size and Corporate Social Responsibility, keeping in mind, a positive effect is expected within this study. The argumentation of both Aggarwal & Nanda as de Villiers et al. is followed, suggesting that a bigger board can both hire more expertise in dealing with CSR related issues, as well as the fact that these CSR related issues are often a part of the manager’s individual objectives within the board (Aggarwal & Nanda, 2004; de Villiers et al., 2011). This argumentation therefore leads to the following hypothesis.

(17)

Gender diversity

The second variable used in this study to indicate board composition is gender diversity. In the current modern society, subjects like multiculturalism and gender diversity are becoming increasingly important (Rao & Tilt, 2016). Both scholars, firms and policymakers are studying the effects of gender diversity to the firm, and how this effect is influenced (García-Meca, García-Sánchez, & Martínez-Ferrero, 2015; Rao & Tilt, 2016). Since boards traditionally consists out of males, often referred to as ‘the old-boys network’, the urgency to investigate the effect of gender diversity within the board is considered relatively high (Ujunwa, 2012), and therefore explicitly indicated as a direction for future research (Gong, 2006; Rao & Tilt, 2016). Ruigrok et al. explains this as “in order to successfully manage diversity on corporate

boards, it is vital to explore how ... gender diversity interacts with more traditional forms of board diversity” (2007: 547).

In general, it is suggested that an increase in diversity (in terms of gender, presence of minorities, nationalities, backgrounds, etc.) expands the information within the decision process, thereby improving the organisational performance or reputation (Bear et al., 2010; Carter et al., 2003; García-Meca et al., 2015; Ruigrok et al., 2007; Williams & O’Reilly, 1998). These suggestions originate from the field of social studies , which have studied the proposed effect of female directors on firms, based on psychology. Jaffee & Hyde (2000), for example, found in their study that women are more likely to use care reasoning, in comparison to men. Another study indicates that women, compared to men, are more likely to recognize unethical actions within the firm (Khazanchi, 1995).

(18)

Shifting the focus to the field of CSR, it is indicated that women are more likely to identify unethical behaviour and are more likely to plead for social responsiveness, thereby suggesting that female directors will have a positive influence on the firm’s CSR performance (Post et al., 2011; Yusoff et al., 2016). Post et al. (2011) argue that this difference between males and females is a result of differences in moral reasoning. Bear et al. (2010) follows a similar line of reasoning in their study where they suggest that female directors might sensitize the board in their CSR initiatives. This reasoning seems uniformly applied throughout the literature in different fields, as Shaukat, Qiu & Trojanowski perfectly comprise: “both theory and empirical

evidence suggest that women may bring a number of competencies and stakeholder-related values to a corporate board” (2016: 574).

The empirical studies, indeed, provide enough evidence to suggest that a higher level of gender diversity (more female board members) leads to an increased CSR performance. The number of female directors is found to have a positive influence on the CSR strength ratings (Bear et al., 2010), three or more female directors lead to higher Kinder Lydenberg Domini (KLD) CSR score (Post et al., 2011), and gender diversity is highly correlated to the Environmental CSR scores of the firm (Shaukat et al., 2016). The KLD score is one of the few widely used indicators for CSR performance within the CSR literature (Shaukat et al., 2016).

By following the positive results of previous scholars, this study expects a positive effect between gender diversity in relation to the CSR performance of the firm, which leads to the following hypothesis. As a concept of gender diversity, the study of Bear et al. is followed. This study conceptualizes gender diversity as “the percentage of woman (relative to board size)” (Bear et al., 2010: 217). This implies that the most diversified board will consist out of a perfectly equal distribution between males and females. This will be measured by means of the Blau’s index, which is further explained in the research methodology section.

Hypothesis 1b: A more equal level of gender diversity of the board is positively related to the CSR performance of the firm.

Organisational age

(19)

and indicates that firm growth and its variability, and the probability of firm failure decreases as the organisation ages (Evans, 1987). This field of study is extended by looking at the impact of organisational age on the innovation output of firms. It is indicated that the organisational age has a positive effect on the innovation output, thereby again indicating the influence of the organisational demographics on its output (Hansen, 1992).

Additionally, another study, which focused on the effect of organisational age on productivity and profitability, finds evidence that in small firms the productivity is high whilst the profitability is low, whereas the opposite is true for larger firms (Majumdar, 1997). In developing economies, the relation between organisational age and firm growth is also found to be positive (Shanmugam & Bhaduri, 2002). Other studies find support for the earlier findings, in that age holds a negative relation with firm growth, but a positive relation with productivity growth (Huergo & Jaumandreu, 2004) and survivability (Yasuda, 2005), indicating the highly context-dependency of age on the firm.

(20)

The influence of organisational factors (like board size and organisational age) on individual managers and CSR performance is very scarce in the literature. Only one study could be found that investigates this relation. Wang et al. (2014) are, to the best of knowledge, the first to combine the organisational factors like firm size, age, industry, and ownership in their influence on the manager’s decision-making process in the CSR related activity of a charitable donation. This study builds upon Elbanna & Child who indicate that organisational structure influences strategic decision making (Elbanna & Child, 2007). Despite being alone in studying this effect, Wang et al. found positive results between firm age, the decision-making process of managers and the CSR activity.

As the studies mentioned above already indicate, not much has been done to combine the organisational factors with the human factors at the firm, while focussing on an CSR related outcome. This despite the expected increased understanding of multilevel variables on the CSR performance (O’Neill et al., 1989; Rao & Tilt, 2016). Nevertheless, there is believed to be enough evidence (in combination with Withisuphakorn & Jiraporn and Elsayed & Paton) to base the following hypotheses upon. This implies that the organisational age is expected to positively influence the relation as indicated in H1a and H1b, as specified next.

Hypothesis 2a: The positive relation between the board size and the CSR performance is positively moderated by the organisational age of the firm.

Hypothesis 2b: The positive relation between a more equal level of gender diversity within the board and the CSR performance is positively moderated by the organisational age of the firm.

(21)

Figure 1 Conceptual Model

RESEARCH METHODOLOGY

In this section, the methodological outline of this study is given. The two main databases for the data collection, the ASSET4 database and Orbis database, have been reviewed. Subsequently, the size of the sample is given as a result of the selection process. In the second part of this section, the measures will be discussed. This section will be closing with the data analysis method in which the used (moderated) regression method is explained.

Data collection

(22)

of European countries, instead of only a few European countries. Therefore, by selecting this list, this study complies to the suggestions of Post et al. (2011) and Lee (2008) to focus on a broad range of firms and in a globalised context, respectively.

The main objective of this study is to indicate the moderating effect of organi sational age within the relation between board composition and CSR performance. In order to measure this, a new dataset has been created. This dataset combines the secondary data of both the Thomson Reuters ASSET4 database and the Orbis database (as provided by Bureau van Dijk). First, the Standard & Poor’s EURO list has been taken as a starting point from the Thomson Reuters ASSET4 database. Subsequently, the required variables (dependent, independent & control variables), except for the moderator variable, could be found in the Thomson Reuters ASSET4 database. With the ‘board size’ and ‘women on board’ variables of the Thomson Reuters ASSET4 database, the percentage of ‘men on board’ could be calculated. This variable is additionally needed to calculate Blau’s Index of Heterogeneity (further explained in the ‘Measures’ section). Subsequently, Blau’s index has been calculated by using the Excel program.

For the moderator variable, the Orbis database has been used to add the data to the dataset. Since the Orbis database only shows the year of incorporation, the organisational age (as measured from 2015) has been calculated by using the Excel program, as well. Additionally, the full company name and ISIN code have been subtracted from the Orbis database, too. To prevent any flaws within the dataset, this ISIN code (unique company code) has been used when combining the variables from both databases.

(23)

Subsequently, the KPI’s are further organised into 15 categories, which collectively form the three pillars (environmental, social and governance), and thereby the ESG score of the firm (Cheng et al., 2014). The environmental pillar would typically include information on water recycled, carbon emissions, energy used, and spills and pollution controversies whereas the social pillar would typically include information on injury rate, training hours, donations, employee turnover, and health and safety controversies, lastly, the corporate governance pillar, would typically include the compensation policy, shareholder rights, board function, and vision and strategy (Cheng et al., 2014; Ioannou & Serafeim, 2012; Shaukat et al., 2016). Since

“Thomson Reuters ASSET4 ... specializes in providing objective, relevant, auditable and systematic ESG information and investment analysis tools to professional investors”, it has

been chosen as the main database for the variables in this study (Ioannou & Serafeim, 2012: 844). Shaukat et al. (2016) even reason that the Thomson Reuters ASSET4 database is preferred over the KLD database because of their exclusive and comprehensive information regarding the performance indicators of the firm.

After checking for any inconsistencies within the Thomson Reuters ASSET4 database, 13 firms have been deleted because of missing variables. Within the Orbis database, 8 additional firms turned out not to be present, making it impossible to retrieve the organisational age data. In combining both databases, all firms with the missing data turned out to be unique, except for one. As a result, 20 firms have been deleted from the sample because of their missing data, leaving the total sample size to 165 firms.

Measures

The measures of the selected variables will be presented here, starting with the dependent variable (CSR performance) and followed by the independent (board size and gender diversity), moderator (organisational age), and control variables (firm size, industry and firm performance).

Dependent variable

(24)

within the literature review, the identification of a CSR performance indicator should surpass the financial measure. It is believed that financial performance measures are too limited in indicating the effect of CSR (Clarkson, 1995; Marshall & Brown, 2003; Wood, 1991). Therefore, this study uses a broader and more comprehensive indicator for CSR performance than several individual (and restricted) indicators like CSR investment (Yang, Colvin, & Wong, 2013) and CSR disclosure (Yusoff et al., 2016). This study specifically follows the more recent notions by Deckop et al. (2006) and Shaukat et al. (2016) who suggest that the CSR performance indicators should include environmental, social and governance measures.

Within the Thomson Reuters ASSET4 database, several CSR indicators are provided. In order to select the most suitable indicator, the given indicators are reviewed. As a basic thought, the indices as provided by Deckop et al. (2006), Marshall & Brown (2003) and Shaukat et al. (2016), as presented in the literature review, have been followed. These indicate that a CSR performance measure should enclose the ESG indices. Eventually, as most appropriate CSR performance indicator, the ‘Integration/Vision and Strategies’ variable has been selected within the Thomson Reuters ASSET4 database since this variable is found to encompass the sought-for CSR ESG indicators (Shaukat et al., 2016), and is therefore considered to be the most appropriate to indicate CSR performance in this study.

Independent variable

The independent variable, board composition, is defined by the size of the board and the top executive’s gender in this study. These, and similar like, variables have been chosen more often since they offer the advantage of being observable and, subsequently, can be collected fairly easy, as well (Carter et al., 2003; Rao & Tilt, 2016). The size of the board has been selected as a part of the board composition variable because of the impact boards have on the performance of the firm (Mazutis, 2013; O’Neill et al., 1989). Within the Thomson Reuters ASSET4 database, the size of the board is given as a numerical number. Since this gives the best indication of the number of top executives, which are active within the board, this indicator will be used.

(25)

between males and females is examined by use of the Blau Index of heterogeneity. Within the Thomson Reuters ASSET4 database, the percentage of women on the board is given, which is used in calculating Blau’s Index for the sample.

Blau’s index of heterogeneity (Dwyer et al., 2003; Gong, 2006) has been used because it offers an objective measurement method. This method is common within the organisational demography and group diversity research (Gong, 2006; Solanas, Selvam, Navarro, & Leiva, 2012), and indicates the level of heterogeneity within a variable. Within this study, this entails the diversity of gender (males and females) within the board. Blau’s index is calculated by the following formula:

𝐵𝑙𝑎𝑢 𝐼𝑛𝑑𝑒𝑥 = (1 − ∑𝑛_{𝑖 =1}𝑝_𝑖2)

Where 𝑝_𝑖 is the percentage of board members in each category (in this case the share of women), and 𝑛 is the total number of board members. The values of the Blau index for gender diversity can range from 0 as a minimum to 0,5 as a maximum. 0 indicates a complete homogeneous board, consisting out of men only, whereas 0,5 indicates a perfectly heterogeneous board, consisting out of an evenly distribution between men and women (Campbell & Mínguez-vera, 2008). Since Blau’s index consists out of two variables within this study, the data has been normalised prior to analysis, following Solanas et al. (2012). As can be expected, the normalised Blau’s index shows a range from 0 to 1, following the same line of reasoning.

Moderator variable

As moderator variable, the organisational factor of age has been used. The literature review already indicates the straightforwardness of measuring this variable. Naldi & Davidsson define organisational age as “the number of years the firm has been in existence” (2014: 694). The same definition has been used in this study, which is also in compliance with several other studies which measure organisational age (Matemilola, Bany-Ariffin, Nassir, & Azman-Saini, 2017; Naldi & Davidsson, 2014; Rafiq, Salim, & Smyth, 2016).

(26)

Control variables

To control other predictors, which could directly influence the relation, several control variables have been selected. As for the control variables, the firm size, industry type and firm performance will be used to control for a more accurate result within the analysis.

Firm size is known to have several indicators throughout the literature (Dwyer et al., 2003;

Orlitzky, 2001). For this study, the proxy as provided by, among many others, Rafiq et al. (2016) is used (Mitchell, Boyle, Nicholas, Maitland, & Zhao, 2016). Rafiq et al. measure firm size by the number of employees, an approach that has also been adopted in this study. Using the Thomson Reuters ASSET4 database, the number of employees per firm (both fulltime and part-time) is given and therefore selected as a useful proxy for firm size.

Industry type is the second control variable selected for this study. The industry type is

proven to have a direct effect on the firm (Garg, 2007; Kaczmarek, 2009). More specifically, it directly influences the firms growth (Naldi & Davidsson, 2014) and environmental performance (Post et al., 2011). As industry indicator, the Standard Industrial Classification (SIC) code has been selected. This code divides the firms based on their operations and markets them into specific segments. The four digits, as given by the Thomson Reuters ASSET4 database, indicate the industry the firm is active in. This code differentiates nine segments (Pierce & Schott, 2012):

- Agriculture, Forestry and Fisheries (SIC 0100-0999) - Mining (SIC 1000-1499)

- Construction (SIC 1500-1999) - Manufacturing (SIC 2000-3999)

- Transportation, Communication and Utilities (SIC 4000-4999) - Wholesale and Retail (SIC 5000-5999)

- Finance, Insurance and Real estate (SIC 6000-6799) - Service Industry (SIC 7000-8999)

- Public Administration (SIC 9100-9799)

(27)

year (Mitchell et al., 2016; Rafiq et al., 2016). As a main currency, the EURO has been selected since all firms are from Europe. Using the Thomson Reuters ASSET4 database, the total revenue over 2015 per firm is given, and is therefore selected as a useful proxy for firm performance. To conclude, an overview of the used variables and their proxies can be found in table 1.

Table 1

Overview of Used Variables

Variables Proxy

Independent variable

Board size Number of executives within the board of the firm

Gender diversity Division between males and females within the board, as indicated by the normalised Blau’s index of heterogeneity

Dependent variable

Corporate Social Responsibility performance

CSR performance indicator as given by the ASSET4 database within the category ‘Integration/Vision and Strategies’

Moderator variable

Organisational age Total operating years since incorporation

Control variables

Firm size Number of employees per thousands

Industry type Industry group as indicated by the ASSET4 database

Firm performance Total revenue of the firm in millions €

Data analysis

(28)

been tested. The analyses have been done by using the statistical computer program of IBM, SPSS Statistics 24th version.

To test the hypotheses, models have been used to indicate the relation between the previous indicated variables. The subsequent models are based on the following regression formula.

𝑌 = 𝛽₀+ 𝛽₁𝑥₁+. . . +𝛽_𝑛𝑥_𝑛+ 𝜀 Where,

Y represents the dependent variable, or the outcome variable, and is defined by the

CSR performance indicator

𝜷_𝟎 represents the constant variable 𝜷_𝒏𝒙_𝒏 represents the variables, in which,

𝛽₁𝑥₁ (𝐵𝑜𝑎𝑟𝑑 𝑠𝑖𝑧𝑒) represents the independent variable of board size 𝛽₂𝑥₂ (𝐺𝑒𝑛𝑑𝑒𝑟 𝑑𝑖𝑣𝑒𝑟𝑠𝑖𝑡𝑦) represents the independent variable of gender diversity

𝛽₃𝑥₃ (𝑂𝑟𝑔𝑎𝑛𝑖𝑠𝑎𝑡𝑖𝑜𝑛𝑎𝑙 𝑎𝑔𝑒) represents the moderator variable of organisational age

𝛽₄𝑥₄ (𝐹𝑖𝑟𝑚 𝑠𝑖𝑧𝑒) represents the control variable of firm size

𝛽₅𝑥₅ (𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦 𝑡𝑦𝑝𝑒) represents the control variable of industry type 𝛽₆𝑥₆ (𝐹𝑖𝑟𝑚 𝑝𝑒𝑟𝑓𝑜𝑟𝑚𝑎𝑛𝑐𝑒) represents the control variable of firm performance

𝛽₇𝑥₇ (𝐵𝑜𝑎𝑟𝑑 𝑠𝑖𝑧𝑒 ∗ 𝑂𝑟𝑔𝑎𝑛𝑖𝑠𝑎𝑡𝑖𝑜𝑛𝑎𝑙 𝑎𝑔𝑒) represents the moderating effect of organisational age on board size

𝛽₈𝑥₈ (𝐺𝑒𝑛𝑑𝑒𝑟 𝑑𝑖𝑣𝑒𝑟𝑠𝑖𝑡𝑦 ∗ 𝑂𝑟𝑔𝑎𝑛𝑖𝑠𝑎𝑡𝑖𝑜𝑛𝑎𝑙 𝑎𝑔𝑒) represents the moderating effect of organisational age on gender diversity

(29)

Within the first model, the effect of the control variables on the dependent variable (CSR performance) has been executed. This has been done to check the impact of the control variables on the CSR performance indicator. This is enclosed within the following first model:

(1) 𝑌 = 𝛽0+ 𝛽4𝑥4 (𝐹𝑖𝑟𝑚 𝑠𝑖𝑧𝑒) + 𝛽5𝑥5 (𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦 𝑡𝑦𝑝𝑒) +

𝛽₆𝑥₆ (𝐹𝑖𝑟𝑚 𝑝𝑒𝑟𝑓𝑜𝑟𝑚𝑎𝑛𝑐𝑒) + 𝜀

Within the second model, the first independent variable of board composition, board size, has been added. The regression between the two independent variables has been done separately to avoid the problem of multicollinearity. Within this model, the suggested relation between the board size and the firm’s CSR performance is tested as indicated within the following model:

(2) 𝑌 = 𝛽0+ 𝛽1𝑥1 (𝐵𝑜𝑎𝑟𝑑 𝑠𝑖𝑧𝑒) + 𝛽4𝑥4 (𝐹𝑖𝑟𝑚 𝑠𝑖𝑧𝑒) + 𝛽5𝑥5 (𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦 𝑡𝑦𝑝𝑒) +

𝛽₆𝑥₆ (𝐹𝑖𝑟𝑚 𝑝𝑒𝑟𝑓𝑜𝑟𝑚𝑎𝑛𝑐𝑒) + 𝜀

Within the third model, the second independent variable of board composition, gender diversity, has been added. Within this model, the suggested relation between the gender diversity and the firm’s CSR performance is tested as indicated within the following model:

(3) 𝑌 = 𝛽0+ 𝛽2𝑥2 (𝐺𝑒𝑛𝑑𝑒𝑟 𝑑𝑖𝑣𝑒𝑟𝑠𝑖𝑡𝑦) + 𝛽4𝑥4 (𝐹𝑖𝑟𝑚 𝑠𝑖𝑧𝑒) +

𝛽₅𝑥₅ (𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦 𝑡𝑦𝑝𝑒) + 𝛽₆𝑥₆ (𝐹𝑖𝑟𝑚 𝑝𝑒𝑟𝑓𝑜𝑟𝑚𝑎𝑛𝑐𝑒) + 𝜀

The fourth model is the first to test the moderating effect of organisational age. Model 4 tests the suggested moderated effect of organisational age on the relation between the first independent variable of board composition (board size) and the firm’s CSR performance. The cross-product term has been used for the moderation effect (𝛽₇𝑥₇), as indicated within the following model:

(4) 𝑌 = 𝛽0+ 𝛽1𝑥1 (𝐵𝑜𝑎𝑟𝑑 𝑠𝑖𝑧𝑒) + 𝛽3𝑥3 (𝑂𝑟𝑔𝑎𝑛𝑖𝑠𝑎𝑡𝑖𝑜𝑛𝑎𝑙 𝑎𝑔𝑒) + 𝛽7𝑥7 (𝐵𝑜𝑎𝑟𝑑 𝑠𝑖𝑧𝑒 ∗

𝑂𝑟𝑔𝑎𝑛𝑖𝑠𝑎𝑡𝑖𝑜𝑛𝑎𝑙 𝑎𝑔𝑒) + 𝛽₄𝑥₄ (𝐹𝑖𝑟𝑚 𝑠𝑖𝑧𝑒) + 𝛽₅𝑥₅ (𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦 𝑡𝑦𝑝𝑒) + 𝛽₆𝑥₆ (𝐹𝑖𝑟𝑚 𝑝𝑒𝑟𝑓𝑜𝑟𝑚𝑎𝑛𝑐𝑒) + 𝜀

(30)

(5) 𝑌 = 𝛽₀+ 𝛽₂𝑥₂ (𝐺𝑒𝑛𝑑𝑒𝑟 𝑑𝑖𝑣𝑒𝑟𝑠𝑖𝑡𝑦) + 𝛽₃𝑥₃ (𝑂𝑟𝑔𝑎𝑛𝑖𝑠𝑎𝑡𝑖𝑜𝑛𝑎𝑙 𝑎𝑔𝑒) + 𝛽₈𝑥₈ (𝐺𝑒𝑛𝑑𝑒𝑟 𝑑𝑖𝑣𝑒𝑟𝑠𝑖𝑡𝑦 ∗ 𝑂𝑟𝑔𝑎𝑛𝑖𝑠𝑎𝑡𝑖𝑜𝑛𝑎𝑙 𝑎𝑔𝑒) + 𝛽₄𝑥₄ (𝐹𝑖𝑟𝑚 𝑠𝑖𝑧𝑒) + 𝛽₅𝑥₅ (𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦 𝑡𝑦𝑝𝑒) + 𝛽₆𝑥₆ (𝐹𝑖𝑟𝑚 𝑝𝑒𝑟𝑓𝑜𝑟𝑚𝑎𝑛𝑐𝑒) + 𝜀

EMPIRICAL RESULTS AND ANALYSIS

Within this section, the data is tested and analysed. At first, an overview of the statistics is presented. Subsequently, the preliminary analysis tests the assumptions needed before running the (moderated) regression analysis. This section will be clos ed by testing the hypothesis and interpreting the results.

Descriptive Statistics

To get an overview of the main variables used in this study, table 2 with the descriptive statistics is discussed first. As indicated in table 2, the size of all variables is 165 as a result of eliminating all missing values beforehand (see data collection). In this table, the industry variable has been excluded since the descriptive statistics in this table would be less meaningful for implementation. An overview of the industry dispersion within the sample size is given in table 3 and is discussed later.

Table 2

Descriptive Statistics of the Sample Size

a_{x 1.000} b_{x 1.000.000}

Variable Valid (N) Mean Std. Deviation Minimum Maximum

(31)

The table shows that the CSR performance of the sample size, on average, is quite high (Mean = 84,27), but also dispersed (Min = 11,04 and Max = 94,89), which might cause problems in the analysis since it is the dependent variable. The average board size of the sample size is nearly 13,5 (Mean = 13,48), and less dispersed (Min = 5 and Max = 30). Considering the Blau’s Index of heterogeneity, the average is given at a 0,76 (Mean = 0,7642). The given minimum and maximum could be expected because of the calculation of the normalised Blau’s Index. An average of 0,76 implies that most board of directors still are unequally distributed. But as the maximum indicates, at least one firm can be indicated with a perfectly equal distribution within the board of directors. The average firm age of the sample size is nearly 68 years (Mean = 67,84) and shows a high dispersion (Min = 3 and Max = 350). A high level of dispersion is also measured at the control variables Firm size (Min = 1.315 and Max = 610.076) and Net sales (Min = 1.069.000 and Max = 213.292.000). The average value for the firm size is 76.893 employees (Mean = 76.893,45) and the average net sales exceed the €26 million (Mean = 26.202.324,04).

Table 3

Descriptive Statistics of the Industry

Industry Frequency Percent Cumulative Percent

Agriculture, Forestry and Fisheries 0 0,00 0,00

Mining 1 0,6 0,6

Construction 6 3,6 4,2

Manufacturing 49 29,7 33,9

Transportation, Communication & Utilities 49 29,7 63,6

Wholesale & Retail 10 6,1 69,7

Finance, Insurance & Real Estate 8 4,8 74,5

Service Industry 42 25,5 100,0

Public Administration 0 0,0 100,0

Total 165 100,0

(32)

(0,00%), and Public Administration (0,00%) industries are not represented within the sample. Since the sample has been selected by means of a European based listing, the country of origin is considered, as well. Table 6 in appendix 1 shows the table. Within the sample, countries from France (25,5%) and Germany (21,8) account for almost half of the observations. Next, Spain (12,1%), the Netherlands (11,5%), Italy (9,7%), and Finland (5,5%) follow. Other countries represented in this sample, but with fewer observations, are Belgium (4,8%), Ireland (4,2%), Austria (1,8%), Luxembourg (1,8%), and Portugal (1,2%). As indicated, the sample shows a bias towards France and Germany with the respectively higher representation levels.

Preliminary analysis

Before testing the hypotheses using the regression method, several assumptions need to be fulfilled (Laerd Statistics, 2013). These assumptions are identified as linearity, reliability of measurement (also known as multicollinearity), homoscedasticity, and normality. It is important to test these assumptions before running the regression analyses because they prevent a “Type I or Type II error, [and] over- or under-estimation of significance or effect size” (Osborne & Waters, 2002: 1). To be aware of any flaws and biases within the analysis, and to examine whether the conclusions can be based upon significant and relevant data, the assumptions have been tested. In examining these assumptions, the guidelines provided by Laerd Statistics have been followed (2013).

Normality

(33)

Whilst some argue that a violation of the normality assumption does not affect the estimates in multilevel methods, and should therefore not be considered (Box, 1976; Gelman & Hill, 2007), others argue that it is important in interpreting the results (Cohen, Cohen, West, & Aiken, 2013; Osborne & Waters, 2002). In order to cope with the violation of the normality assumption, an alternative use (called the bootstrapping approach) has been applied (Leeflang, Wieringa, Bijmolt, & Pauwels, 2015).

By using the bootstrapping approach, the data of the current (random) sample are used in order to approximate the cumulative distribution function of the entire population (Driscoll, 1990). Using the bootstrapping approach, increases the normality distribution of the sample and improves the accuracy of the analysis. The bootstrap distribution is found to be as nearly continuous, even within small sample sizes (Andrews & Buchinsky, 2000), and has therefore been applied in the data analysis of the hypotheses. Subsequently, as already indicated in the previous section (descriptive statistics), the dependent variable of CSR performance shows a highly-dispersed distribution, which causes the violation of the normality assumption. By specifically conducting the bootstrapping approach for this variable, its normality can be presumed, thereby increasing the significance of the results . Therefore, for the CSR performance variable, the statistical justified stratified bootstrapping approach has been applied to assure the normality within the sample.

Linearity

Linearity indicates the relation between the independent variables (in this case the board size and gender diversity) and the dependent variable (CSR performance). Violation of this assumption could lead to data transformation or the use of a different research method. (Laerd Statistics, 2013). To identify the linearity between these variables, scatterplots have been used. The first scatterplot shows a positive linear relation between Board size and CSR performance, confirming the linearity assumption for the first independent variable.

(34)

Homoscedasticity

The homoscedasticity assumption measures the variances of the residuals. If the variance is not constant, the homoscedasticity assumption is violated affecting the confidence intervals. The homoscedasticity assumption can be examined by plotting the normal residuals against the unstandardized residuals (Laerd Statistics, 2013). Appendix 2, figure 3, shows the variance of residuals of both the independent variables in relation to the dependent variable. As the scatterplot indicates, the assumption of homoscedasticity is not violated.

Multicollinearity

The multicollinearity assumption tests whether the independent variables of the sample size are highly correlated. A high correlation between these variables could lead to problems in understanding the regression results (Laerd Statistics, 2013). To test this assumption, the Variance Inflation Factor (VIF) is used. In order to meet the assumption of multicollinearity, the VIF should not exceed a score of 3 (Kutner, Nachtsheim, Neter, & Li, 2005). As appendix 2, figure 9, indicates, the VIF score for both the variables does not exceed a score of 3, therefore confirming the assumption. This indicates that multicollinearity is not a problem in this dataset.

Hypotheses testing

As the previous section already indicated, all the assumptions have been confirmed except for the normality assumption. Because of this violation, the stratified bootstrapping approach has been applied to the models in testing the hypotheses. The dependent variable of CSR performance has been subjected to stratification due to its non-normality. The hypotheses are tested by conducting linear and moderating regression analyses with the statistical computer program of IBM, SPSS Statistics 24th version.

(35)

Table 4

Overview of the Regression Results Model Variablesa ₁ ₂ ₃ ₄ ₅ Constant (CSR) 85,547* 80,301* 68,151* 78,147* 67,222* Independent variable Board Size 0,442* 0,464* Blau’s Index 21,140* 21,110* Moderator variable Firm Age 0,019* 0,017* Moderating effects

Board Size x Firm Age 0,001

Blau’s Index x Firm Age -0,155*

Control variables Firm Sizeb _0,012* _0,006 _0,013* _0,005 _0,012* Industry -0,908* -0,920* -0,681* -0,789* -0,708* Firm Performancec _0,109* _0,101* _0,104* _0,097* _0,106* Model summary R-squared 0,055 0,063 0,107 0,066 0,118 *p < 0.05

a_{Unless otherwise noted, bootstrap results are based on 1.000 stratified bootstrap samples} b_{x 1.000}

c_{x 1.000.000}

Hypothesis 1a – The size of the board is positively related to the CSR performance of the firm

(36)

Hypothesis 1b – A more equal level of gender diversity of the board is positively related to the CSR performance of the firm

Model 3, subsequently tested the proposed effect of the second independent variable, gender diversity, by means of Blau’s index, on the CSR performance of the firm as indicated by hypothesis 1b. As model 3 indicates, a positive significant effect has been found between an equally diversified board (21,140) and the CSR performance of the firm (p < 0.05). Subsequently, the dependency level of this variable is higher than the first model as well (R -squared 0,107). This leads to enough evidence to support hypothesis 1b, as well.

Hypothesis 2a – The positive relation between the board size and the CSR performance is positively moderated by the organisational age of the firm

Model 4 is the first model to test the moderating effect of organisational age on the, earlier confirmed, relation between board size and CSR performance, specified in hypothesis 2a. As indicated, the effect is found to be positive (0,001), but very small, and not significant. This leads to the rejection of hypothesis 2a as a result of not finding enough evidence to support the expected positive moderating effect of organisational age in this situation. The R-squared score is slightly higher than the second model (R-squared 0,066).

Hypothesis 2b – The positive relation between a more equal level of gender diversity within the board and the CSR performance is positively moderated by the organisational age of the firm

At last, model 5 tests the moderating effect of organisational age on the, earlier confirmed, relation between gender diversity (measured by Blau’s index) and CSR performance, specified in hypothesis 2b. As the model indicates, the relation is found to be negatively significant (-0,155), which contradicts the hypothesis. The dependency level is highest for this model (R-squared 0,118) which indicates that organisational age influences the CSR performance score of the MNC. This leads to the conclusion that, despite being significant (p < 0.05), hypothesis 2b must be rejected since a negative moderating effect has been found instead of a positive moderating effect.

(37)

section, the outcomes of the regression analysis will be interpreted and discussed in more detail, and conclusions will be drawn.

Table 5 Analysis Overview

Hypothesis Description Expectation Outcome

H1a A positive relation between board size and CSR performance

Positive relation Confirmed

H1b A positive relation between gender diversity and CSR performance

Positive relation Confirmed

H2a A positive moderated effect of organisational age on the relation between board size and CSR performance

Positive relation Rejected

H2b A positive moderated effect of organisational age on the relation between gender diversity and CSR performance

Positive relation Rejected (negative relation)

DISCUSSION

In this section, the results of the analysis are being discussed. Specifically, all hypotheses are reviewed individually, and in its entirety, to answer the research question. To recall, the research question is “What is the moderating effect of organisational age on the relation

between board composition and the firm’s CSR performance?” Next, both the academic and

practical implications of the results of this study have been given. This section will come to an end by discussing the limitations and by providing guidelines for future research.

To study this research question, the initial relation between board composition and CSR performance has been studied first. Subsequently, the moderating effect of organisational age has been added to the relation. The moderated regression analysis has been conducted to compute the results, which can be seen in table 4 of the previous section.

(38)

score (0,063) is higher compared to model 1 (R-squared 0,055) which also indicates that board size influences the CSR performance score of the MNC. These results imply that the quantity of the board does matter within the context and performance of the firm. Specifically, the size of the board has found to positively influence the firm’s CSR performance (Jizi, Salama, Dixon, & Stratling, 2014). Haji (2013) argues that the board of directors is an important aspect in the overall management of the firm, and therefore has the power to exert CSR practices. The knowledge base of the board of directors increases with every extra individual added to the board (Kassinis & Vafeas, 2002), which can inspire firms to engage into CSR related activities (de Villiers et al., 2011). Therefore, the size of the board holds a positive relation to the CSR performance of the firm, a line of reasoning which fits the regression results.

Regarding hypothesis 1b, the regression result indicates an even bigger effect of the boards’ gender diversity on the CSR performance of the firm (21,140 at the 5% significance level). While it could be expected that this regression outcome is higher compared to the size of the board, because of the limited range of the variable itself (0 to 1), the R-square score (0,107) also indicates a higher dependency level of gender diversity on CSR performance compared to model 1 (R-squared 0,055). As the results specify, a higher level of gender diversity contributes to the CSR performance of the firm. Ideally, the firm should hold an equal distribution between male and female directors within the board to increase the CSR performance of the firm. This is in line with prior studies which also indicate a positive effect between diversity and the general firm performance (Dwyer et al., 2003; Schwab, Werbel, Hofmann, & Henriques, 2016; Solakoglu, 2013).

(39)

moderating impact on the firm as the hypothesis proposes, the dependency level of organisational age is considerably lower compared to the dependency level of board size itself on CSR performance.

However, caution is needed with such a suggestion since the regression result is not significant. This non-significance does not imply that the null-hypothesis (organisational age does not have a moderating effect on board size in relation to CSR performance) should be accepted. To further study the effect of organisational age, future research is needed to further explore possible relations before conclusions can be drawn where this study can be used as a preliminary study.

While hypothesis 2a could not be confirmed due to its non-significance, the results of hypothesis 2b are even more surprising. Instead of the expected positive moderating effect of firm age on the relation between gender diversity and CSR performance, the result of the regression indicated a negative, but significant, relation (-0,155 at 5% significance level). As the R-squared (0,118) indicates for gender diversity, organisational age does have its influence on the MNC’s CSR performance, compared to the R-squared (0,107) of model 3. This implies that, in the case of gender diversity, organisational age has its influence on the MNC’s CSR performance. However, as is the case with the previous hypothesis, the dependency level for organisational age is lower, compared to the dependency level of the gender diversity alone. Considering the regression results, it indicates that this role of organisational age is negative. To be more specific, the positive relation of gender diversity on CSR performance is weakened, as the firm grows older. Following this result, a firm that wants to score high on CSR performance, can best have an equal distributed board and be very young.

(40)

The results of this study lead to a partial answer of the research question. The moderating effect of organisational age can only be identified for the relation between gender diversity and CSR performance, but not for board size. This means that organisational age only has a significant effect on one of the two studied variables of board composition, but it confirms that organisational age moderates the effect of board composition on the MNC’s CSR performance.

This leads to the conclusion that the moderating effect of organisational age on the relation between board composition and CSR performance shows a negative relation.

Against the expectation, the significant relation is found to be negative which implies that organisational age negatively influences the relation between gender diversity and the firm’s CSR performance. These results have several implications for both the literature and the business practice which are discussed in the next section.

Implications

The results of this study have several implications for both the academic literature as well as the business practice. Firstly and most important, by studying the effect of organisational age on the relation between board composition and the firm’s CSR performance, this study answers the research question. The research question indicates the gap within the literature in that both human- and organisational factors which influences the MNC’s CSR performance have not been studied together. As shown in the literature review, only a few scholars combine the organisational factor of age to the operations and outcomes of the firm (see the study of Wang et al., 2014), or call to extend this (Kor, 2006). By combining human- and organisational factors with the CSR outcome, both variables which influence the firm’s CSR performance, are jointly studied (O’Neill et al., 1989; Rao & Tilt, 2016; Yi et al., 2015).