Women in the boardroom: do they make a difference? Exploring the effects of board gender diversity on corporate social (ir) responsibility and the contingencies upon which these relationships rest

(1)

Women in the boardroom: do they make a difference?

Exploring the effects of board gender diversity on corporate social (ir)

responsibility and the contingencies upon which these relationships

rest

By

Milou Nipshagen

S2550504

M.L.Nipshagen@student.rug.nl

Thesis supervisor: Dr. Rieneke Slager

Co-Assessor: Dr. Sarah Castaldi

University of Groningen

Faculty of Economics & Business

MSc International Business and Management

Nettelbosje 2

9747 AE Groningen

June 13

th

_{, 2019}

(2)

2

Abstract

This thesis attempts to make essential contributions to the research fields of corporate social responsibility and corporate social irresponsibility. Firstly, the relationship between gender diverse boards and irresponsible behaviour of firms is studied since very little research has focused on CSR its antithesis, and even less attention has been given to antecedents of CSIR. Secondly, although the relationship between board gender diversity and CSR is well established, this research proposes that we should re-think the best ways to operationalize CSR and to explain the mechanisms underlying the relationship. Accordingly, it is argued that when researching the effect of board gender diversity on CSR, the latter concept should be operationalized as a firm’s stakeholder orientation and the gender role theory should be used to predict the effect of a gender diverse board. Following this, it is hypothesized that board gender diversity, measured by the Blau index, positively (negatively) contributes to the stakeholder orientation (CSIR performance) of firms. Besides these main effects, this study also aims to discover to what extent CEO duality and board independence moderate these relationships. Findings of an OLS regression supported the notion that a gender diverse board positively influences the stakeholder orientation of firms. Contrary to expectations, results show that CEO duality strengthens this relationship, while board independence did not show to have a moderating effect. Moreover, no significant effect of board gender diversity on CSIR was found. Accordingly, the findings of this study challenge existing theories and provide promising avenues for future research on antecedents of CSIR, as well as environmental factors influencing the effect female executives have on CS(I)R-related outcomes.

Keywords: Corporate Social Irresponsibility, Corporate Social Responsibility, Corporate

(3)

3

Acknowledgments

(4)

4

1. INTRODUCTION

‘’Achieving gender equality requires the engagement of women and men, boys and girls, and is everyone’s responsibility’’ (Ban-Ki Moon at the Gender Equality Roundtable, 2011).

Since the financial crisis of 2008, scandals involving corporate misconduct such as inhumane working conditions, corruption, environmental pollution and child labour have gained increased attention in the academic world, politics and society. Consequently, these wrongdoings made the topic of corporate social responsibility one of the most emphasized concerns of the international business literature in the 21st century (Pearce and Manz, 2011; Matten and Moon, 2005; Pava and Krausz, 1996). However, although scholars put much attention to CSR, far less attention has been given to its antithesis: Corporate social irresponsibility.

Corporate social irresponsibility (hereafter CSIR) can be seen as an immoral decision made by a company with the goal of maximizing or creating shareholder value at the expense of other stakeholders (Armstrong, 1977). More specifically, these decisions encompass acts such as bribes, violating human rights or causing environmental damage. Although Armstrong already introduced the concept of CSIR in 1977, little academic attention has been given to the subject over the past thirty years (Riera and Iborra, 2017). This is a worrying trend, since CSIR is just as important as its counterpart. The latter is stressed by the argumentation that irresponsible practices can have major negative consequences for stakeholders and for firms themselves (McMahon 1999; Muller and Kräussl, 2011; Lange and Washburn, 2012; Kotchen and Moon, 2012; Herzig and Moon, 2013). Connected to the aforementioned, there is evidence to suggest that CSIR is negatively related to financial performance, as opposed to CSR (Riera and Iborra, 2017). Hence, it is essential for both society and businesses that MNEs refrain from behaving socially irresponsible.

(8)

8

establishing the rationale of why gender diversity contributes to CSR as well as operationalizing the construct as the stakeholder orientation of firms, which differs from previous studies mostly operationalizing CSR as corporate social performance. Hereupon, the research questions addressed in this paper are:

To what extent does gender diversity in the board of directors influence the socially (ir)responsible behaviour of MNEs? And to what extent is this relationship moderated by

CEO duality and board independence?

Following the gender role theory (Eagly, 1987), beliefs about gender roles are important determinants of how executives behave. Accordingly, female directors are expected to be more socially oriented and interested in increasing the quality of relationships with communities (Boulouta, 2013). Furthermore, since male leaders are more often narcissistic and in power for their personal benefit, rather than in benefit of the organization (Pearce and Manz, 2011), male dominated boards are assumed to increase the CSIR performance of firms. Assuming that the former prevents firms from behaving irresponsibly, and the latter contributes to unethical behaviour, it is argued that adding women to the board of directors will increase the stakeholder orientation and will prevent socially irresponsible practices of the firm. Additionally, this study hypothesizes that under certain circumstances, the relationships between board gender diversity and corporate social (ir)responsibility are altered. Along these lines, it is proposed that CEO duality positively moderates the relationship between BGD and CSIR and weakens the relationship between BGD and the stakeholder orientation of a firm. Furthermore, it is suggested that as board independence increases, the relationship between BGD and the stakeholder orientation of a firm will become stronger and the effect of BGD on socially irresponsible practices will become weaker.

(9)

9

follows this stance in the debate by empirically researching both CSR and CSIR, and by seeing them as two separate and thus self-contained constructs.

Besides these theoretical contributions, the present research also provides new insights to practice. This is since it is specifically important for MNEs and managers to learn more about CSIR and its antecedents. Also, because this work will generate fresh insights into the circumstances under which gender diversity can have an even stronger (or weaker) effect on corporate social (ir)responsibility, the developed knowledge can help executives to be more aware of how to increase (decrease) the stakeholder orientation (irresponsible behaviour) of business and accordingly increase (decrease) its associated positive (negative) consequences. Correspondingly, this thesis should provide professionals with a better understanding of the mechanisms at play to help them overcome irresponsible practices and support responsible behaviour.

To research the main- and moderating effects, the remaining part of this paper proceeds as follows: The first section illustrates the main theories and concepts related to CSR, stakeholder orientation, CSIR and board gender diversity. The second section is concerned with outlining the explanatory mechanisms of the relationships discussed in this paper, and a conceptual model is presented based on the hypotheses. In the third section, the methodology used for this study is discussed. Following this, in the fourth section the results of all analyses are presented. These results are discussed in the fifth section, followed by a conclusion in the last section.

2. THEORY

2.1 Literature Review

2.1.1 Corporate social responsibility

(10)

10

various interrelated concepts such as corporate social performance (CSP), corporate citizenship and corporate responsiveness. Nonetheless, most scholars operationalized CSR in terms of CSP (Waddock and Graves, 1997; Zhang, 2012; Johnson and Greening, 1999; Shaukat and Trojanowski, 2016). Although these studies show that CSP has been proven a good operationalization of the CSR construct, a firm’s stakeholder orientation is the most suitable for this research, since it is more specific than the broad CSP construct and is mostly in line with the main ideas of the gender role theory.

A firm’s stakeholder orientation can be defined as ‘'the degree to which a firm's management decides to focus its attention on stakeholders and integrate their interests and knowledge in its decision-making’’ (Bettinazzi and Zollo, 2017: 2465). Correspondingly, a high stakeholder orientation holds that a firm its behaviours and actions are aimed at improving or maintaining a high-quality relationship with their stakeholders and are focused on incorporating their interests into the decisions of the firm. As reported by Bettinazzi and Zollo (2017), stakeholder orientation has much in common with the stakeholder theory, which has developed into a dominant paradigm in CSR research. Subsequently, in this thesis a stakeholder-based view of the firm is adopted. The stakeholder theory of Freeman (1984), assumes that a business its social responsibility is determined by the extent to which it takes into account the interests of its stakeholders. Additionally, the theory emphasizes that firms adopting CSR practices must go beyond the scope of business as usual since managers and boards need to pursue optimal results for all stakeholders, instead of only maximizing results for their shareholders. Likewise, Donaldson and Preston (1995) identified three main aspects of the stakeholder theory, a normative, descriptive and instrumental aspect. Often, the normative and instrumental aspects are converged, so that the normative aspect supports the instrumental aspect. For example, Jones (1995) argued that firms conducting business ethically will achieve a competitive advantage, as they can develop lasting and productive relationships with their stakeholders. In line with this, the present study will follow the instrumental rationale of the stakeholder theory, assuming that normative aspects are grounded in the instrumental perspectives and that a firm’s stakeholder orientation positively contributes to the performance of a firm.

(11)

11

(2001), who argued that when a firm shows responsible behaviour, it has a responsible image and in turn gains legitimacy and support from stakeholders. Additionally, connecting instrumental arguments to stakeholder orientation, scholars propose that a high stakeholder orientation results in a dynamic capability for a firm (Sharma and Henriques, 2005) and an improved reputation and legitimacy (King, 2008), which leads to competitive advantages. Hence, firms can benefit from behaving socially responsible and having a high stakeholder orientation, and this instrumental reasoning made that the topic has gained recognition over the past few decades.

Taken together, in all reviewed studies in this section two important issues emerges. The most evident issue pertains to the complexity in defining CSR. To overcome this complexity and make sure there is a fit with the aim of this research, it is suggested to operationalize CSR as a firm's stakeholder orientation. However, there is also a less apparent issue that emerges. All mentioned studies clearly indicate that there is much knowledge on CSR its effects, conceptualization and operationalization, and its antecedents. Yet, as rightly noted by Lin-Hi and Muller (2013: 1929) ‘’the current debate about the social responsibility of

corporations has a serious blind spot: research on the notion of corporate social irresponsibility’’.

2.1.2 Corporate social irresponsibility

(12)

12

Schlegelmilch, 2013; Riera and Iborra, 2017). These debates have been summarized in appendix A and will be discussed more thoroughly in the following sections.

2.1.3 Debates in CSIR literature Biased versus unbiased observers

In his preliminary work, Armstrong defined a socially irresponsible decision as ‘'a decision which generally involves the gain of one party at the expense of the total system, and in which a vast majority of unbiased observers would agree that this is at the expense of the total system'' (Armstrong, 1977: 185). In line with this definition, the first discussion that arose concerning CSIR was that of who is the one to judge whether an action is socially irresponsible. Only a few authors focus on unbiased observers, which are actors that have no direct interests in the firm, while most scholars concentrate on biased observers, which are actors with a direct interest, such as stakeholders. Accordingly, Herzig and Moon (2013) and Armstrong and Green (2013) are one of the few to agree with the notion of unbiased observers. They see CSIR as a failure to act socially responsible in light of the expectations of a society as a whole, and that the society is the group to assess if a firm fails to act socially responsible.

Contrary to these scholars, others emphasize that stakeholders are the ones assessing the irresponsible behaviour of firms (Williams and Zinkin, 2008; Wagner, Bicen and Hall, 2008; Lange and Washburn, 2012; Antoinetti and Maklan, 2016). These authors argue, in line with the stakeholder theory, that harmful behaviour can have significant impacts on stakeholders. Interestingly, Pearce and Manz (2011) claim that CSIR should be seen in light of the perceptions of both biased and unbiased observers, and thus define CSIR as ‘’unethical behaviour of executives that shows disregard for the welfare of others, at the expense of employees, shareholders and other organization stakeholders, and even society at large’’ (Pearce and Manz, 2011: 563). Since this definition emphasizes the importance of firm executives and both biased and unbiased observers, this definition will be followed in the remainder of this thesis.

Deliberate versus accidental irresponsible behaviour

(13)

13

external to the executives or the firm. In contrast, the majority of scholars perceive CSIR as deliberate or even strategic behaviour (Surroca, Tribó and Zahra, 2013; Shea and Hawn, 2018; Lange and Washburn, 2012; Pearce and Manz, 2011; Strike, Gao and Bansal, 2006; Keig, Brouthers and Marshall, 2015). In line with Campbell’s (2007) characterization of CSR, Shea and Hawn (2018) define CSIR as knowingly and intentionally doing something that harms your stakeholders. Correspondingly, Strike et al. (2006) claim that irresponsible practices are intentional since managers pursue their personal gains and do not always take into account incurred negative consequences to their stakeholders. Likewise, Keig et al. (2015) and Surroca et al. (2013) see CSIR in light of an intentional business strategy that firms might follow to overcome incurred costs of CSR or comply to local norms in host countries. Thus, CSIR can be a deliberate strategy or happens accidentally. Since in this thesis, it is proposed that executives can be drivers of CSIR, the conceptualization of CSIR as a deliberate act is followed.

Self-contained versus continuous CSIR

Although the former and latter discrepancies are important, the most critical discussion in CSIR literature revolves around its nexus to CSR (Murphy and Schlegelmilch, 2013; Riera and Iborra, 2017). Some scholars conceptualize CSIR as being an extreme on the continuum of CSR (Jones, Bowned and Tech, 2009; Windsor, 2013; Perks et al., 2013), while most academics argue that the CSR and CSIR are two separate self-contained concepts (Strike et al., 2006; Muller and Kräussl, 2011; Lange and Washburn, 2012; Herzig and Moon, 2013; Keig et al., 2015; Shea and Hawn, 2018; Walker, Zhang and Ni, 2019).

(14)

14

mutually exclusive. In addition, this means that a firm cannot be socially responsible and irresponsible at the same time.

However, the majority of scholars oppose the aforementioned view and conceptualize CSIR as a self-contained concept. This holds that CSIR and CSR are perceived as two independent concepts. Additionally, from the self-contained perspective, both constructs are perceived as mutually exclusive and can thus, independent of each other, decrease or increase at the same time. Surroca et al. (2013) empirically researched this and found that MNEs can be socially responsible in e.g. their home country and, simultaneously, behave irresponsibly in a host country. Moreover, Strike et al. (2006) their findings also support the self-contained perspective, since they found that although firms might operate in line with high CSR standards, they can also show irresponsible behaviour. Furthermore, Kang, Germann and Krewal (2016) go even further and argue that a positive causal relationship exists between the concepts, meaning that more corporate social irresponsible behaviour might lead to higher levels of socially responsible behaviour.

Hence, as the concept of CSIR only recently has gained in prominence, there is a strong need to improve its conceptualization. Nonetheless, in this thesis, social irresponsibility is proposed to be deliberate behaviour, which is assessed by (un)biased observers, and should be perceived as a self-contained concept. However, besides the significance of these debates, Riera and Iborra (2017) argue it is even more important for future research to focus on antecedents of CSIR.

2.1.4 Antecedents of corporate social irresponsibility

(15)

15

Along one of the few scholars studying board-level effects and CSIR belong Biggerstaff, Cicero and Puckett (2015) and Pearce and Manz (2011). The former authors found empirical evidence that the upper echelons of a firm play an essential part in creating an unethical culture, which subsequently results in unethical conducts of a firm. Furthermore, Pearce and Manz (2011) researched the influence of leadership on CSIR and concluded that managers with a strong need for personalized power are more often narcissistic, which might drive irresponsible behaviour. Similarly, Ormiston and Wong (2013) maintain that executives with high moral identity are less inclined to behave irresponsibly than those with low moral identities. Overall, these studies highlight the need for more in-depth research on the antecedents of CSIR, as there is some knowledge on antecedents at the institutional level, but almost no knowledge on drivers of CSIR on the organizational level. More specifically, research is needed on management systems and board compositions that could drive CSIR.

2.1.5 Corporate governance, board gender diversity and CS(I)R

In their significant literature review, Aguinis and Glavas (2012) point out that on the organizational level, most research on the predictors of CSR has focused on firm motives, mission and values, firm ownership and corporate governance. Corporate governance (hereafter CG) can be defined as ‘‘the system by which companies are directed and controlled’’ (Jo, Harjoto, 2011: 55). Considering the instrumental reasoning that CSR is an important tool for gaining legitimacy, improving a firm’s reputation and improving and maintaining relationships with stakeholders, it can be claimed that CG and CSR are inherently connected. At the basis of this connection is the importance of good stakeholder management and thus a high stakeholder orientation. More specifically, it has been suggested that the strongest aspect of CG in the management of stakeholders is the board of directors (Rao and Tilt, 2016). Adding to this proposition, recent research suggests that diverse boards in particular are critical for good stakeholder management and a well-developed CSR strategy (Galbreath, 2011; Rao and Tilt, 2016; Walls, Berrone, Phan, 2012; Giannarakis, Konteos & Sariannidis, 2014; Post et al., 2011; Bear et al., 2010; Siciliano, 1996). Accordingly, in this thesis the relationships between board gender diversity and corporate social (ir)responsibility are empirically researched.

(16)

16

of the proposed relationships, it does not clearly explain why female board members would be advantageous for socially responsible behaviour of firms.

Accordingly, it is argued that, although far less used, the gender role theory provides us with a better explanation to the relationships. According to Grosser and Moon (2016), this feminist theory appears most fruitful in addressing gender issues in the management of firms and their influence on firm-level outcomes. The theory proposes that persons are socially identified as males or females and that this identification holds that different roles are expected from these individuals within a social structure (Eagly, 1987). Accordingly, the approach holds that women and men face pressures to behave in line with these expectations about the social role they ought to employ. Thus, connecting this to organizational practices, this holds that stereotypes and beliefs about gender roles are grounded in the labour position of men and women, since this position reflects ‘’a biosocial interaction between male and female physical attributes and the social structure” (Eagly, 2009: 644). Put differently, the theory proposes that ideas and beliefs about gender roles are important determinants of how executives behave and act in their firm. Particularly, the ideas and beliefs about gender roles in the social structure of a board or management team can be seen as follows. Women are expected to be more socially oriented, while men ought to be more agentic (Boulouta, 2013). This implies that women should have a great concern for others and should be interested in increasing the quality of relationships with communities, both in- and outside of the firm. In the workplace, this holds that women and men in the same positions face different expectations in the way they behave and decide since they need to comply with the different gender roles. Gutek and Morasch (1982) empirically researched this theory and indeed found that stereotypes and beliefs about gender roles put a lot of pressure on women to act in line with the expectation to be community/socially oriented and caring. Considering that relationships with communities and a social orientation are important in both the support of a high stakeholder orientation and the prevention of CSIR, the gender role theory will be central in the hypothesis development of this research.

2.2 Hypothesis Development

2.1.1 Board gender diversity and a firm’s stakeholder orientation

(17)

17

beneficial for the stakeholder orientation of a firm and are more so than men. However, some also argue that men can be useful in managing stakeholder relationships due to their ‘old boys network’ (Miller & Triana, 2009) and their relatively high management experience in large organizations (Hillman, Canella & Harris, 2002). Analysing these views, in this section the explanatory mechanisms underlying the relationship between board gender diversity and a firm’s stakeholder orientation are discussed.

Although some scholars point out male executives can be valuable for the stakeholder orientation of a firm, Williams (2003) maintains that since male board members are more focussed on financial performance and short-term gains than their female counterparts, they are less inclined to answer to stakeholder’s demands. Also, since gender role theory posits that men are more agentic and are thus more focused on executive and financial decisions in the firm, one can assume that a board with predominantly male executives will not be valuable to the stakeholder orientation of firms.

Adding women to a BoD, on the other hand, is argued to be beneficial for a firm’s stakeholder orientation. Since women have different prior experiences than their male colleagues, often outside of the ‘’old boys networks’’ and large organizations, they can offer different perspectives (Hillman et al., 2002; Singh et al., 2008; Byron and Post, 2016). Scholars argue that with these different perspectives, women can influence board-level decisions/outcomes, including which stakeholder demands to answer to (Rao and Tilt, 2016). Likewise, Galbreadth (2011) argues that since women are expected to have higher relational abilities, they are also acting in line with this and are thus better at engaging with multiple stakeholders and are consequently better able to respond to their needs. Interestingly, this reasoning of Galbreadth (2011) is conforming to the gender role theory, since this theory also proposes that women are expected to act in line with their female gender roles, which holds that they should be community oriented.

(18)

18 Hypothesis 1: Board gender diversity has a positive effect on the stakeholder orientation of a firm.

2.1.2 Board gender diversity and self-contained CSIR

Hypothesizing about the relationship between board gender diversity and CSIR, assuming the latter concept to be self-contained, is not so clear-cut. As already mentioned, there is very little academic literature on antecedents of social irresponsible behaviour. However, as one of the few scholars Pearce and Manz (2011) studied the influence of leadership (self and shared leadership) on CSIR. They found that leaders with a strong need for personalized power are more often narcissistic and in power for their personal benefit, rather than in benefit of the organization. This, in turn, might lead to increased CSIR practices. Interestingly, Magee and Langly (2008) empirically researched this and found that managers with personalized power motivation were more often associated with antisocial decisions. Connected to the latter, they found that men use their power motivation more for antisocial ends than women do. From this, one can argue that male-dominated boards are more inclined to show socially irresponsible behaviour. Additionally, in accordance with the gender role theory, this study expects that female executives are more sensitive to avoid irresponsible practices than male executives do, because they are envisioned to pay attention to communities and should be socially oriented. Slote (2007) researched this idea of gender roles and found that female managers indeed are more inclined to prevent unethical business practices. Drawing on this conclusion, Boulouta (2010) argues that women, needing to act according to the stereotypes about their gender roles, are sensitive towards societal issues and thus are less inclined to support unethical business practices. Thus, since men are associated with more personalized power motivation and anti-social behaviour, predominant male boards are assumed to be more inclined to promote irresponsible behaviours. Correspondingly, grounded in gender role theory, it is proposed that adding women to boards increases the likelihood that socially irresponsible practices are put to hold. Hence, this study hypothesizes that board gender diversity negatively influences CSIR:

(19)

19 2.1.3 The moderating roles of CEO duality and Board independence

It is proposed that, under certain circumstances, the relationships between board gender diversity and CS(I)R change. Accordingly, this study proposes that CEO duality and board independence are board-level contingencies that alter the effects of board gender diversity on a firm’s stakeholder orientation and socially irresponsible practices.

CEO duality

Besides gender, it is argued that CEO duality is a significant determinant for firm-level decisions, including those concerning the stakeholder orientation and irresponsible practices of a firm. CEO duality can be defined as the presence of a CEO who also holds a chair position (Shaukat, Qiu and Trojanowski, 2016). It is argued that the gender role theory explains why this board-level factor moderates the two main relationships of this research. A significant aspect of the theory revolves around the dynamics surrounding women in their professional life, especially if they are in leadership positions. Accordingly, Agars (2004) claims that if women in these positions face serious hierarchical power, they may adopt more masculine behaviour, as this pushes them more in a masculine managerial role. Wiggenhorn, Pissaris and Gleason (2014) their findings are in line with this. However, they do not discuss their results based on the latter rationale, but propose that this is since a powerful CEO enforces entrenchment behaviour as the dual structure provides the CEO with unquestionable authority. This unlimited authority could subsequently lead to the CEO being less likely to care about stakeholders such as employees, overruling the board’s decision-making. Hence, CEOs occupying a dual position could veto decisions of the gender diverse board concerning the stakeholder orientation of the firm in a negative sense. Thus, this study argues that CEO duality will undermine the positive effect of board gender diversity on the stakeholder orientation of the firm (Hypothesis 3a). Additionally, it proposes that CEOs occupying a dual position could also overrule decisions concerning the prevention of harming stakeholders. Consequently, this research hypothesizes that CEO duality weakens the negative relationship of board gender diversity and corporate socially irresponsible behaviour (Hypothesis 3b).

(20)

20 Hypothesis 3b: CEO duality moderates the relationship between board gender diversity and CSIR of a firm such that the negative relationship between board gender diversity and socially irresponsible behaviour weakens if the CEO holds a dual position.

Board independence

The concept of board independence is most often included as a diversity indicator or an independent effect on CSR (Post et al.; Zhang 2012; Ibrahim and Angelidis 1995; Johnson and Greening, 1999; Lorenzo Sanchez & Gallego-Alvarez, 2009). Nonetheless, this thesis suggests that under a high number of independent board members, the relationships between BGD and corporate social (ir) responsibility will be different. Board independence is defined as the degree in which board members are not dependent on the organization (Ibrahim, Howard and Angelidis, 2003). According to Ibrahim et al. (2003) and Johnson and Greening (1999), independent board members provide a board with more resources and legitimacy. Correspondingly, Ibrahim et al. (2003) argue that executives of independent boards can make use of these resources and the increased legitimacy to be more sensitive to society’s needs. Besides increased sensitivity, Berghe and Baelden (2005) found that the impacts of board independence on the functioning of board members increased the qualitative decision-making and effectiveness of the board, due to increased monitoring. In line with the gender role theory, this would mean that an independent board strengthens the qualitative decision making that is connected to the female gender role which, as argued, increases the stakeholder orientation of firms and decreases harm towards stakeholders.

(21)

21 Hypothesis 4a: board independence moderates the relationship between board gender diversity and the stakeholder orientation of a firm such that the positive relationship between board gender diversity and stakeholder orientation is strengthened by board independence.

Hypothesis 4b: board independence moderates the relationship between board gender diversity and CSIR of a firm such that the negative relationship between board gender diversity and socially irresponsible behaviour is strengthened by board independence.

2.2 Conceptual Model

Considering the hypotheses, the following conceptual model is constructed.

FIGURE 1: Conceptual Model

3. METHODS

3.1 Data Collection

(22)

22

The data retrieved from these sources stems from different years, since several authors argue that corporate governance mechanisms change quite slowly and influence firm-level outcomes in time. (e.g. Bear et al., 2010; Ben-Amar, Chang and McIlkenny, 2017; Jo and Harjoto, 2012). Consequently, it is examined how a firm's stakeholder orientation and CSIR in period t are influenced by the firm its board gender diversity in t-2. Accordingly, this should also take away issues of endogeneity and reverse causality. This is since one should be cautious of the fact that firms engaging more in social responsibility might e.g. also attract more females in leadership functions (Bear et al., 2010). Following the suggestion of the latter authors to lag the independent variables for more than one year, data of the dependent variables is collected in 2016, and data for the independent and control variables is collected in 2014 for the main analyses. Nonetheless, to correctly control for all effects, robustness tests are performed.

3.2 Sample

The sample for this study is based on the largest MNEs that are headquartered in the U.S., as these MNEs show most engagement to CSR practices. Furthermore, U.S. headquartered MNEs nowadays seem to adopt a more explicit commitment to CSR compared to European headquartered MNEs, which makes the data more reliable (Boulouta, 2013). The sample of firms in this study consists of 373 S&P 500 firms whose social performance was rated by Asset4. Although there are many views on the amount of needed observations for an optimal sample size, Tabachnic and Fidell (2007) propose that for each variable, there should be at least 10 observations. Since this study includes 13 variables, at least 130 observations are needed. In this study, the 500 observations were limited to 373 observations because 24 firms did not make the S&P 500 list in all analysed years (2014 and 2016 for the main analysis, 2015 for the robustness test). In addition, observations of 103 firms were excluded due to missing data. Additionally, the sample of S&P 500 firms was chosen since it accounts for a great portion of the US economy and as these companies differ in their sizes and industrial sectors. In this sample, all industries were included, since this safeguards that there are enough observations for the study.

3.3 Measurement of the Variables

3.3.1 Dependent variables

(23)

23

should be made is that the way they measure the scores differ. This has implications for the findings of studies, depending on which database is used (Garcia-Sanchez et al., 2014). Hereupon, Bettinazi and Zollo (2017) argue that the indicators of KLD are very grainy and include items that do not pertain to stakeholder orientations. Correspondingly, thesis uses a different database for collecting data: the Thomson Reuters Asset4 database. This database is best since it has specific assessments of companies their stakeholder orientations and offers 400 different ESG metrics. In addition, the methodology of their data gathering differs from that of other ESG databases since it eradicates potential self-selection concerns, which decreases chances that the estimation of a coefficient is biased (Bettinazi and Zollo, 2017). This holds that Asset4 its ESG metrics are measured based on reported public domain data such as annual reports, websites of nongovernmental organizations and firm surveys. This results in 10 different categories of the ESG topics (table 1).

TABLE 1:

Overview of ESG and Controversy Topics

ESG topics Category Number of variables

underlying category

Number of variables used in analysis

Environmental Resource use 19 -

Emissions 22 - Social Workforce 29 16 Human rights 8 4 Community 14 8 Product responsibility 12 4 Governance Management 34 3 CSR strategy 12 - Shareholders 8 - Controversy topics Environmental controversies Resource use 2 2 Emissions 1 1

Social Controversies Workforce 13 13

Human rights 4 4

Community 6 6

Product responsibility

12 12

Governance Controversies Management 2 2

Shareholders 6 6

(24)

24

Conforming to the approach of Bettinazi and Zollo (2017), this research measures a firm’s stakeholder orientation by aggregating the variables of the ESG categories that correspond to an orientation of firms towards employees, customers and local communities1. For this, only category score variables are used since Asset4 calculates them by equally weighting and z-scoring all data points of the variables, and comparing them against all firms in the sample. Accordingly, these variables show the relative orientation of firms towards these three stakeholders, resulting in interval variables with scores between zero and 100, where zero is a very low orientation and 100 is a very high orientation (appendix B).

An MNE its orientation towards employees is measured as the sum of four macro variables Employment Quality, Workforce Diversity and Opportunity, Workforce Health and Safety and Workforce Training and Development. Since all these macro-items are normalized scores with a positive value ranging from 0 to 100, all variables can be aggregated into one interval employee orientation variable. Consumer orientation is assessed by the Customer/Product Responsibility variable. This variable reflects a company's management commitment and effectiveness in generating sustainable and long-term revenue growth for its customers and has a positive value of zero to 100. Like the other two orientation variables, the consumer orientation variable will thus also be of an interval measurement level with positive values ranging from zero to 100 (Appendix B). The orientation towards communities is proposed to be an item of the Society/Community and the Society/Human Rights variables. These items have a positive value, ranging from zero to 100.

In order to decide if aggregating these items is reliable, reliability analyses were performed (appendix C). Since these results show that merging the items together is reliable, sum variables were computed. Furthermore, in line with the methodology of Bettinazi and Zollo (2017), the community, consumers and employee variables are aggregated into one overall stakeholder orientation variable with a value from zero to 100. To test if this is reliable, both a reliability analysis and a factor analysis were performed (appendix C). These show that a sum variable could be computed of all underlying items, except for the Society/Human Rights item which was thus excluded from the sum variable.

Of equal importance in this research is MNEs their irresponsible behaviour in their stakeholder orientation. Until now, the majority of scholars empirically researching CSIR have

1_{Although the stakeholder orientation of a firm also pertains to an orientation towards suppliers, the latter group}

(25)

25

used the KLD concerns measures (Strike et al., 2006; Tang, Qian, Chen and Shen, 2015; Kang et al., 2016; Ormiston and Wong, 2013). Interestingly, the Thomson Reuters Asset4 database also provides a similar measure for irresponsible practices: controversy scores. A controversy score is measured by Asset4 as follows: if, during the fiscal year, a scandal occurs in one of the controversy categories (e.g. client loyalty, employment quality, health and safety and human rights, table 1), the firm is given a score of 1. Then, the firm is benchmarked against other firms in its sample. So if, for example, a firm has two controversies in the category of client loyalty in 2016 and 3 of the firms in its sample do not have any (0) controversies in this category, the overall score for the latter firms in this category will be (1+(3/2)) / 4 = 0.63/63% (formula 1). Additionally, for the former firm with two controversies, their score will be 25: (0 + (2/2)) / 4 = 0.25/25% (formula 2). Thus, the polarity of the controversy scores is negative, where a score of zero is a high level of controversies and 100 is a very low level of controversies, as compared to other firms in the sample.

(26)

26 3.3.2 Independent variable

Board gender diversity is operationalized in terms of heterogeneity of gender in a board.

Accordingly, the Blau (1977) index of heterogeneity is used to capture the construct. The index measures the distribution of board members their gender (formula 3).

𝐻 = 1 − ∑_𝑖𝑘 = 𝐼𝑃𝑖2 (3)

Where I is the number of categories (2 for gender diversity) and pi is the proportion of the group (female or male board members) in each category. The index can thus take values from 0, in which there is only one gender in the board and no heterogeneity, to 0.5 when there is an equal number of male and female board members. The index is used since it meets criteria that are argued to be important in constructing a good measure of diversity. Ben-Amar et al. (2017: 376) propose that ‘'it has a zero point to represent complete homogeneity, larger numbers indicate greater diversity, the index does not assume negative values and the index is not unbounded''. Additionally, to test the robustness of both models, the percentage of women on the board of directors is used as operationalization of board gender diversity. This is since, according to Schwartz-Ziv (2017), one should be taking into account the idea that for a significant effect on firm-level outcomes, a critical mass of women in the board should be reached. Along these lines, the scholar also proposes that this critical mass is best measured as the percentage of women in the BoD (Schwartz-Ziv, 2017).

3.3.3 Moderating variables

(27)

27 3.3.4 Control variables

While the focus here is on one antecedent of CSR and CSIR, both concepts are determined by many factors, and one must control for these factors accordingly. For example, studies suggest that larger boards put more emphasis on CSR practices, while smaller boards tend to show an opposite effect and thus put less emphasis on CSR (Post et al., 2011). Hence, board size, measured as the number of directors, needs to be taken into account and should be controlled for when researching (ir)responsible behaviour (Ibrahim, Howard and Angelidis, 2003; Johnson and Greening, 1999; Webb, 2004). Board member’s experience in the BoD is also controlled for since Ibrahim et al. (2003) argue that a more experienced board could overrule effects of e.g. diversity in boards, influencing potential outcome effects of diverse boards. Thus, this study takes the average years of experience of board members in the current board into account. Moreover, the extent to which a CEO is dependent upon the performance of a firm is also a factor influencing the responsibility practices of firms (Dyck, Lins, Roth and Wagner, 2019). This is since dependent CEOs can e.g. block decisions of a board on CSR related issues, when this can have negative consequences for the performance of the firm. In line with this, CEO

dependence upon the performance of the organization, measured as the presence of CEO whose

compensation is linked to total shareholder return, is an antecedent that is controlled for.

MNE size, measured as the number of employees, also needs to be considered as a

control variable. Scholars suggest that larger MNEs have more financial resources and face higher stakeholder pressures for CSR, which is why they are connected to more CSR practices than smaller MNEs (Laudal, 2011). Financial performance, measured as the profit margin of a firm, should also be controlled for. High financial performance of MNEs is linked to more CSR as firms with higher financial performance have more resources to spend on CSR activities (Brammer & Millington, 2004). In contrast, Brammer and Millington (2004) argue that

financial risk, measured as the liquidity ratio of firms, hampers CSR initiatives of firms. Thus,

both financial risk and financial performance are firm-level variables that are controlled for. Last, several studies propose that the Industry in which MNEs operate has an influence on CSR-related outcomes (Ben-Amar et al., 2017; Post et al., 2011; Boulouta, 2013). This is since industries differ in resources and stakeholder pressures for CSR. Thus, industry groups were identified by using ICB industries, from which a dummy variable was created with a manufacturing industry (0) or a service industry (1)2_.

(28)

28

3.4 Empirical Data Analysis

In order to empirically research the hypotheses, two regressions (stepwise ordinary least squares and logistic regressions) are performed using the IBM SPSS Statistics software. As proposed by Lattin, Caroll and Green (2003), independent variables were standardized so that multicollinearity is reduced. For the relationships between the independent variable, moderators and stakeholder orientation, an OLS regression is proposed to be the most suitable regression analysis. This is since the relationship is tested between a continuous independent variable, two moderators and the continuous dependent variables of a stakeholder orientation score (0 to 100). Additionally, to test the moderating effects, an Andrew F. Hayes process regression was performed (Aguinis and Gottfredson, 2010).However, before conducting these regressions, assumptions for a parametric test should be met (Lattin et al., 2003). Correspondingly, variables were tested on their distribution by a Shapiro-Wilk test and P-P plot, linearity was checked by plotting the studentized residuals and dependent variables, homoscedasticity was tested using a Breusch-Pagan and Koenker test for the equality of variances and multicollinearity was checked for by looking at correlations and the variance inflation factor (VIF) values (Appendix D). Since all preliminary requirements are met for the stakeholder orientation model, an OLS regression is performed.

Although the stakeholder orientation model meets all requirements for a parametric test, the first test of normality of the social irresponsibility model already shows that it does not adhere to this first assumption (Appendix D). The CSIR performance variable, while continuous, appears to be divided into two groups. Accordingly, it should be seen as a dichotomous variable with two groups, those with high CSIR performance (scores from 0 to 35) and those with low CSIR performance (35 to 100). Because of this, an OLS regression cannot be used for analysing the relationships between the independent variables and CSIR performance, and a logistic regression should be used as a method of analysis instead (Tabacknick and Fidell, 2007). To perform this regression, some assumptions should still be met. Correspondingly, multicollinearity among the independent variables was still tested. As presented in appendix D, all assumptions for the logistic regression were met.

3.4.1 Descriptive statistics

(29)

29

that the firms analysed can be placed more on the high-end of CSIR performance continuum, since a lower score holds that firms have had more controversies as compared to other firms. Furthermore, the average gender diversity in boards indicates that they are a bit diverse, since the average index is 0.29 (SD = 0.11), while the average percentage of women in the boards is 18.58% (SD = 9.27). On average, 84.22% of the directors are independent (SD = 9.23) and for 73% of the firms, CEO duality is in place (SD = 0.44). In the sample, firms have an average net profit margin of 11.64%. Since the firm size, measured as the number of employees, varies considerably, natural logarithms were used. In addition, boards consist on average of 11 executives and 71% of the CEOs are dependent upon the performance of the organization (SD = 0.40). Furthermore, as the table in Appendix E illustrates, the majority of firms in the sample are in consumer discretionary industry (15.9%), while there are only few firms in the real estate industry (2.6%). Additionally, the 2015 sample used for the robustness test also has a sample size of 373 firms. Interestingly, in 2015 the average Blau index score is 0.30 (SD = 0.10), which is slightly higher than the year before. In contrast, the average percentage (82.31%) of independent board members is slightly lower in the 2015 sample, while CEO duality does not differ compared to 2015.

TABLE 2: Descriptive Statistics

2014

Variables Min Max Mean SD

Dependent variables

Stakeholder orientation 12.06 95.26 68,53 17.33

CSIR performance 0.00 68.00 38.58 24.92

Independent variables

BGD (Blau index) 0.00 0.50 0.29 0.11

BGD (% women on the BoD) 0.00 55.60 18.58 9.27 Moderators

Board independence 38.00 95.00 84.22 9,23

CEO duality 0 1 0.73 0.44

Control variables

ICB industry type 0 1 0.30 0.45

Financial performance -15.05 56.72 11.64 8.68

Firm size (log) 1.91 5.73 4.71 4.88

Financial risk 0.21 10.81 1.91 1.27

Board size 4 18 10.95 2.14

Board experience 4 20 9.55 2.99

CEO dependence 0 1 0.71 0.40

(30)

30

The correlation table (table 3) depicts that board gender diversity significantly correlates with both stakeholder orientation and CSIR performance r (0.241) = 0.000, p < 0.001 and r (-0.138) = 0.006, p < 0.01. Board gender diversity operationalized as the percentage of women in the BoD also significantly correlates with both dependent variables, for stakeholder orientation,

r (0.220) = 0.000, p < 0.001 and for CSIR performance r (-0.122) = 0.016, p <0.05. Moreover,

board independence does not significantly correlate with CSIR performance, r (-0.074) = 0.148,

p > 0.05 but does so with stakeholder orientation, r (0.276) = 0.000, p < 0.01. Also, it correlates

significantly with both board gender diversity and the percentage of women in the BoD, r (0.137) = 0.000, p < 0.01 and r (0.124) = 0.015, p < 0.05. CEO duality does not correlate to any of the two dependent variables r (0.006) = 0.903, p > 0.05 and r (0.061) = 0.230, p > 0.05, nor does it correlate with the percentage of women in the BoD. Additionally, the table shows that, although not significantly, board experience correlates considerably high with CEO duality

r (-0.501). To make sure there is no multicollinearity of these predictors, the VIF-value was

(31)

31

TABLE 3: Correlation Matrix

Note: N=373; independent variables are measured in 2014, dependent variables are measured in 2016, **p≤0.01, *p≤0.05

Pearson Zero-Order Correlations

Variables 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 1. Stakeholder orientation 1,000 2. CSIR performance -0.072 1,000 3. BGD (Blau index) 0.241** -0.138** 1,000 4. BGD (% of women in BoD) 0.220** 0.190** 0.964** 1,000 5. Board independence 0.276** -0.074 0.137** 0.124* 1,000 6. CEO duality 0.006 0.061 0.054 0.064 -0.115* 1,000

7. ICB industry type -0.115* -0.067 -0.085 -0.078 0.225** -0.083 1,000

(32)

32 3.4.2. Robustness tests

Tabachnick and Fidell (2007) argue that, to ensure the outcomes of the regression analyses have satisfactory validity, robustness tests should be conducted. Correspondingly, both regressions are repeated and the consistency of the results are compared. A first robustness test implies a replacement of 2014 data with the 2015 data of the independent variables. Additionally, an adjustment is made by replacing board gender diversity (the Blau index) by the critical mass measure of the percentage of women in the board of directors. Also, although not including board-level diversity, one of the items (Workforce diversity and Opportunity) of the aggregated employee-orientation variable includes aspects such as equal opportunity for women and men in the workforce of the firm. Accordingly, to make sure this is not a cause of endogeneity, a third robustness test is run, excluding this item in the stakeholder orientation variable. In the next section, the regression results of the main models as well as the robustness test results are presented.

4. RESULTS

In this section, results of the stepwise OLS and logistic regressions are presented and analysed. To test the hypotheses concerning stakeholder orientation, a stepwise OLS regression was conducted. In addition, stepwise logistic regressions were performed to test the hypotheses on the CSIR performance of firms. All independent variables were standardized and measured in t-2 (2014), while the dependent variable was measured in t (2016). Since all assumptions for a parametric test were met for the stakeholder orientation model (appendix D), table 4 shows the outcomes of an ordinary least squares regression. In the table, the unstandardized coefficients, the standard errors and significance level are depicted. Furthermore, the explanatory power of each model is included by presenting the R-squares, adjusted R-squares, constant and F-statistics of each model. Moreover, the table illustrating the outcomes of the logistic regression also shows the unstandardized coefficients, the standard errors and significance level. However, other than in the OLS regression, the Cox and Snell and Nagelkerke R-squares are included.

Model 1 of table 4 shows that the control variables explain about one sixth of the variance in the outcome variable stakeholder orientation, since the R2_{is 0.158 and the adjusted}

R2_{is 0.141, with a very high significance (p = 0.000). Interestingly, when comparing the first}

(33)

33

(34)

34

TABLE 4:

Regression Analysis Stakeholder Orientation Model

Note: N=373; independent variables are standardized and form 2014, ***p≤0.01. **p≤0.05. *p≤0.1

Variables Model 1 Model 2 Model 3 Model 4 Model 5 Model 6

Independent variable

Board gender diversity 3.042 (0.847)*** 3.047 (0.850)** 3,019 (0,876)*** 2.635 (0.836) *** 2.591 (0.838)***

Moderating variables

CEO duality 0.157 (1.832) 0.261 (1.906)

Board independence 3.388 (00845) *** 3.355 (0.846)***

Interactions

Board gender diversity X CEO duality

4.041 (1.891)**

Board gender diversity X board independence -0.693 (0.746) Control variables Financial performance -1.660 (0.839)** -1.860 (0.828)** -1.860 (0.829) ** -1.811 (0.825)** -1.700 (0.812) ** -1.692 (0.812)** Firm size Financial risk 2.312 (0.877)*** 1.529 (0.972) 1.787 (0.875)** 1.532 (0.953) 1.792 (0.878)** 1.498 (0.902) 1.598 (0.878)* 1.387 (0.911) 1.573 (0.859)* 1.211 (0.872) 1.623 (0.861)* 1.198 (0.801) CEO dependence 7.411 (1.866)*** 7.155 (1.838)*** 7.155 (1.840)*** 2.980 (0.824)*** 6.357 (1.812) *** 2.797 (0.811)*** Board size 3.623 (0.860)*** 2.988 (0.864)*** 2.987 (0.865)*** 3.055 (0.816)*** 2.884 (0.847)*** 2.814 (0.851)*** Board experience -0.425 (0.892) -0.253 (0.875) -0.420 (0.890) -0.211 (0.872) -0.415 (0.812) -0.212 (0.875)

Industry dummies Yes Yes Yes Yes Yes Yes

Constant 58.540 (2.228)*** 58.572 (2.192)*** 58.697 (2.637)*** 63.380 (1.89)*** 31.425 (7.106)*** 65.430 (1.877)***

R-square 0.158 0.177 0.185 0.241 0.219 0.221

Adjusted R-square 0.141 0.160 0.169 0.194 0.204 0.204

(35)

35

TABLE 5:

Logistic Regression Analysis: CSIR Performance Model

Note: N=373: independent variables are standardized and from 2014. ***p≤0.01. **p≤0.05. *p≤0.1

Variables Model 1 Model 2 Model 3 Model 4 Model 5 Model 6

Independent variable

Board gender diversity -0.034 (0.125) -0.050 (0.126) -0.161 (0.234) -0.020 (0.126) -0.021 (0.126)

Moderating variables

CEO duality -0.408 (0.258) -0.419 (0.259)

Board independence -0.145 (0.127) -0.143 (0.126)

Interactions

Board gender diversity X CEO duality

-0.156 (0.276)

Board gender diversity X board independence -0.052 (0.119) Control variables Financial performance -0.337 (0.121)** -0.335 (0.121) *** -0.340 (0.122) *** -0.340 (0.122)*** -0.377 (0.121) *** -0.339 (0.122)*** Firm size -0.752 (0.162)*** -0.744 (0.164)** -0.763 (0.165)** -0.776 (0.167)*** -0.737 (0.164)*** -0.734 (0.129)*** Financial risk -0.051 (0.140) -0.049 (0.124) -0.041 (0.122) -0.034 (0.140) -0.028 (0.124) -0.027 (0.122) CEO dependence 0.326 (0.290) 0.306 (0.291) 0.249 (0.188) 0.326 (0.290) 0.306 (0.291) 0.249 (0.188) Board size -0.345 (0.125)*** -0.338 (0.127)** -0.336 (0.128)** -0.338 (0.128)*** -0.335 (0.128)*** -0.343 (0129)*** Board experience -0.110 (0.127) -0.110 (0.124) -0.102 (1.122) -0.110 (0.127) -0.110 (0.124) -0.102 (1.122)

Industry dummies Yes Yes Yes Yes Yes Yes

Constant 0.625 (0.217)*** 0.622 (0.127)*** 0.723 (0.228)*** 0.206 (0.230) 0,655 (0.220)*** 0.509 (0.141)***

Cox and Snell R-square 0.136 0.136 0.142 0.142 0.139 0.139

(36)

36

4.1. Regression Results

Table 4 presents the results of the analysis of relations between the predictors and the outcome variable stakeholder orientation. Firstly, in order to determine whether the control variables have significant explanatory power, the first model presents the regression run with the controls. In the second model, the first hypothesis is tested that there is a positive effect of BGD on the stakeholder orientation of a firm. In order to analyse whether the stakeholder orientation is higher for firms with gender diverse boards, an ordinary least squares regression was performed of board gender diversity on stakeholder orientation (model 2, table 4). This test was significant,

p = 0.00, B = 3.042, p < 0.001. Since the unstandardized coefficient is positive and significant,

there is enough evidence to infer that board gender diversity has a positive effect on the stakeholder orientation of a firm, which supports hypothesis 1.

Model 3 (table 4) shows that the moderating variable of CEO duality does not have a direct effect on the outcome variable since B = -0.157, p = 0.932. In order to analyse hypothesis 3a and thus whether CEO duality negatively moderates the positive relationship between board gender diversity and the stakeholder orientation of a firm, a process regression of Andrew F. Hayes was performed (model 4, table 4). The overall model shows p < 0.01 and R2 = 0.241. As the table shows, the interaction effect is significant p = 0.033, and B= 4.041. This means, that while there is enough evidence to infer that CEO duality moderates the relationship between BGD and the stakeholder orientation of a firm, hypothesis 3a is rejected. This is since the opposite is found of what was hypothesized, and the effects of board gender diversity become stronger under CEO duality, while the effect is not influenced by no CEO duality (figure 2).

FIGURE 2:

(37)

37

Model 5 (table 4) illustrates that board independence does have a direct effect on the outcome variable (B = 3.388, p = 0.00). However, in order to analyse the hypothesis (4a) that board independence strengthens the positive relationship between board gender diversity and the stakeholder orientation of a firm, a process regression of Andrew F. Hayes was performed (model 6, table 4). The overall model shows p < 0.01 and R2 = 0.221. The interaction effect is insignificant p = 0.367 and B = -0.693. Thus, there is not enough evidence to infer that board independence positively moderates the relationship between BGD and the stakeholder orientation of a firm, and hypothesis 4a is rejected.

Besides a hypothesized relationship between board gender diversity and the stakeholder orientation of a firm, in hypothesis 2 it was proposed that BGD has a negative relationship to the CSIR performance of a firm. In order to analyse whether the CSIR performance is lower for firms with gender diverse boards, a logistic regression was performed of board gender diversity on the binary variable of CSIR performance (Model 2, table 5). This test was insignificant, B = -0.034 p = 0.782, p > 0.05. Accordingly, there is not enough evidence to infer that board gender diversity has a negative effect on the CSIR performance of a firm, which rejects hypothesis 2. Model 3 (table 4) of the CSIR performance analysis shows that CEO duality does not have a significant direct effect on the CSIR performance of a firm (B = -0.408, p = 0.341). Model 4 (table 5) shows the results of a process regression of Andrew F. Hayes, to test hypothesis 3b. The overall model shows that B = -0.156, p = 0.542 and Nagelkerke R-Square = 0.194. Thus, there is not enough evidence to accept hypothesis 3b that CEO duality positively moderates the relationship between BGD and CSIR performance. Last, to test the hypothesis (4b) that board independence strengthens the negative relationship between BGD and CSIR performance, a

process regression of Andrew F. Hayes was performed. The overall model shows that

B = -0.052, p = 0.682 and Nagelkerke R-Square = 0.190. Thus, hypothesis 4b is rejected.

4.2. Robustness Tests

(38)

38

effect of CEO duality in t-1 (model 4) shows that this effect is insignificant (B = 2.002, p = 0.170), indicating that there is not enough evidence to infer that CEO duality positively moderates the relationship between BGD and a firm’s stakeholder orientation, as findings of the main analysis did show. As with the main findings, model 5 and 6 illustrate that board independence significantly influences the stakeholder orientation of a firm, but does not moderate the relationship between BGD as measured in t-1 and the stakeholder orientation of an MNE (B = 3.695, p = 0.000; B= -0.789, p = 0.411), thus also rejecting hypothesis 4a.

The second model in Appendix H confirms the rejection of hypothesis 2, since BGD

measured in 2015 shows no significant relationship to the CSIR performance of firms (B = -0.142, p = 0.243). Also, in accordance with the previous analyses, model 4 shows that

when changing the time lag to 2015, there is still not enough evidence to infer that CEO duality moderates a relationship between BGD and CSIR performance, thus rejecting hypothesis 3b. However, what model 3 does show is that in t-1, CEO duality has a marginally significant effect on the CSIR performance of a firm (B = -0.432, p = 0.092). This is different from the model in t-2 (table 5) since this model shows B = -0.408 and p = 0.162. In addition, model 5 and 6 of the CSIR performance model (Appendix H) also validate previous findings that board independence does not moderate the relationship between BGD and the CSIR performance of firms, thus rejecting hypothesis 4b (B = -0.012, p = 0.711).

(39)

39

Furthermore, model 2 (Appendix J) confirms the rejection of hypothesis 2, since the percentage of women in the BoD shows no significant relationship to the CSIR performance of firms (B = -0.033, p = 0.290). Thus, these findings also show that, as with all other analyses, there is not enough evidence to infer that women in boards will have a negative effect on the CSIR performance of a firm. Also, conforming to the previous analyses, model 4 (Appendix J) shows that when changing the Blau index into the percentage of women in the BoD, there is still not enough evidence to infer that CEO duality moderates a relationship between BGD and CSIR performance, thus rejecting hypothesis 3b (B = -0.093, p = 0.743). Along the same lines, model 6 of the CSIR performance model (Appendix J) also validates the finding that board independence does not moderate the relationship between BGD and the CSIR performance of firms, thus rejecting hypothesis 4b (B = -0.082, p = 0.142).

Last, the robustness test in Appendix K illustrates that excluding the Workforce Diversity and Opportunity item in the Stakeholder orientation variable does not imply major changes in the findings of the main model. Besides minor changes in the R-squares and adjusted

R-squares (from 0.177 and 0.160 in the main model to 0.119 and 0.107) and in the coefficients,

all previously found results were supported in this model. This supports the idea that since the Workforce Diversity and Opportunity item measures e.g. gender equality, this measure does not apply to the BoD, thus not impacting e.g. causes of endogeneity. Hence, all robustness tests confirm most of the outcomes of the main analyses, except for the moderating effect of CEO duality, when changing the time lag from t-2 to t-1. These findings will be considered and discussed in the subsequent section.

5. DISCUSSION

5.1 Discussion and Implications

The present study was designed to examine the influence of gender diversity in the boardroom on the socially (ir)responsible behaviour of 373 US headquartered S&P 500 firms. Analysing these relationships and the contingencies upon which they rest, this thesis expands our knowledge and contributes to theory and practice in significant ways. Correspondingly, in this section, the results of all analyses are discussed, as well as their contribution to theory and practice.

Women in the boardroom: do they make a difference? Exploring the effects of board gender diversity on corporate social (ir) responsibility and the contingencies upon which these relationships rest