• No results found

Corporate Governance and Corporate Social Responsibility: The Effect of Gender Diversity and Board Independence on CSR Decoupling

N/A
N/A
Protected

Academic year: 2021

Share "Corporate Governance and Corporate Social Responsibility: The Effect of Gender Diversity and Board Independence on CSR Decoupling"

Copied!
51
0
0

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

Hele tekst

(1)

Corporate Governance and Corporate Social

Responsibility: The Effect of Gender Diversity

and Board Independence on CSR Decoupling

MSc Accountancy

University of Groningen

Faculty of Economics and Business

June 2019

Jeroen S. Bosgra

S2729164

Grote Beerstraat 206-C25

9742 SJ Groningen

06 23 109 503

j.s.bosgra.1@student.rug.nl

Supervisor

dr. N. Hussain

Word count: 10.935

(2)

Abstract

The last two decades have witnessed a substantial growth of research related to CSR decoupling, which can be defined as the implementation gap between actual and communicated CSR practices. In this article, the relation between board composition and CSR decoupling is empirically examined. Drawing on institutional theory and the literature on corporate governance, I hypothesize that both gender diversity and board independence reduce the degree to which companies engage in CSR decoupling. Using data from a global sample of 18.351 firm-year observations for 4.870 listed companies between 2009 and 2017, I first find that gender diversity has a significant positive effect on CSR decoupling, which is contrary to my expectations. Secondly, I find that board independence has a significant negative effect on CSR decoupling, which is in line with my hypothesis. Results of the first additional analysis show that the positive effect of gender diversity is caused by its effect on the social pillar of CSR, and to a lesser extent the governance pillar of CSR. Furthermore, these results show that the negative effect of board independence can be fully attributed to the governance pillar of CSR. Results of the second additional analysis show that distinguishing between overstated and understated decoupling can provide fruitful insights.

Keywords: Corporate Social Responsibility, Decoupling, Corporate Governance, Gender

(3)

Table of content

1. Introduction ... 3 2. Theoretical background ... 7 2.1 Institutional theory ... 7 2.2 CSR Decoupling ... 9 2.3 Hypotheses development ... 13 2.3.1 Gender Diversity ... 13 2.3.2 Board Independence ... 15 3. Research method ... 16 3.1 Sample ... 16 3.2 Variable measurement ... 18 3.2.1 CSR decoupling ... 18 3.2.2 Gender diversity ... 20 3.2.3 Board independence ... 21 3.3 Control variables ... 21

3.4 Empirical model and estimation procedure ... 22

4. Results ... 23

4.1 Descriptive statistics and correlation analysis ... 23

4.2 Regression analysis ... 26

4.3 Robustness tests ... 28

4.4 Additional analyses ... 29

4.4.1 Decomposing CSR decoupling ... 29

4.4.2 Distinguishing between overstated and understated decoupling ... 32

5. Discussion and conclusion ... 35

5.1 Answer to the research question ... 36

5.2 Theoretical and managerial implications ... 37

5.3 Limitations and further research ... 38

References ... 39

Appendix A: ESG performance categories ... 47

(4)

1. Introduction

It was on December 15th 2018 that 15-year-old Greta Thunberg managed to silence a room filled with world leaders with her speech at the United Nations climate summit in Poland (Connect4Climate, 2018). The Swedish girl accused world leaders of decoupling their environmental talk from reality, by only speaking of green eternal economic growth without actually making an impact on environmental sustainability. Since 1995, the annual United Nations climate summit has resulted in a lot of talks and agreements but the world is still failing to actually make an impact on climate change (Plumer & Popovich, 2018).

On another level, companies have a very large impact on social and environmental issues. For instance, they are responsible for the biggest part of global emissions (Griffin & Heede, 2017). While corporate social responsibility (CSR) is a hot topic and environmental and social transparency tend to increase, the question is whether communicated CSR practices are in line with reality on the work floor. If this communication does not correctly reflect actual CSR efforts, it can be referred to as CSR decoupling. CSR decoupling can be defined as the situation where “communication about the implementation of CSR-related organizational practices does not correspond with the actual implementation of these practices” (Wickert, Scherer, & Spence, 2016, p. 29). Boiral (2013) gave a good example of CSR decoupling in practice by analyzing 23 sustainability reports of which the scope of disclosure was rated with A or A+ levels by the Global Reporting Initiative1. Still, he found that 90 per cent of the significant negative CSR news events regarding the firms that published these reports were not reported. The goal of this study is to identify factors that decrease the extent to which firms engage in CSR decoupling.

Companies may engage in CSR decoupling in two different ways (Wickert, Scherer, & Spence, 2016). The most popular conception is that of the implementation gap, where companies put much effort in communicating CSR policies and practices and less effort in implementing them into their core activities. I refer to this type of decoupling as overstated CSR decoupling. Organizational communication about CSR activities mostly induces favorable consumer responses (Marquina Feldman & Vasquez-Parraga, 2013). It would be detrimental if consumers would keep rewarding companies that adopt formal CSR policies which are decoupled from organizational practices, as this would motivate these companies to ‘not walk the CSR talk’

1 The Global Reporting Initiative is an international organization that develops sustainability reporting standards

which firms can voluntarily apply. It is generally regarded as the strictest guideline of CSR (Brown, De Jong, & Levy, 2009).

(5)

(Wickert, Scherer, & Spence, 2016). Since it is beneficial for the environment and society that companies act as socially responsible as possible, it is important to limit overstated CSR decoupling as much as possible.

The other way in which companies may engage in CSR decoupling concerns the communication gap, where companies put an emphasis on implementing CSR policies and practices but do not very actively communicate them to the public. I refer to this type of decoupling as understated CSR decoupling. An increased CSR disclosure is supposed to have a positive impact on firm value (Plumlee, Brown, Hayes, & Marshall, 2015), so this can be seen as a missed opportunity. Therefore, it is in the interest of shareholders to limit this side of CSR decoupling as well. While most people would believe that companies on average exaggerate their CSR performance, Hawn and Ioannou (2016) found that most companies undertake more internal than external CSR actions. To conclude, both sides of CSR decoupling should be limited so that CSR communications and implementation are aligned and interests of both society and shareholders are served.

Differing stakeholder expectations are generally seen as the main driver of CSR decoupling (Crilly, Zollo, & Hansen, 2012). Responding to different expectations of socially responsible processes and outcomes can be done in an organizationally integrated way or in a decoupled fashion (Weaver, Trevino, & Cochran, 1999). Strategic choices such as these are affected by corporate governance characteristics (Hitt & Tyler, 1991). I examine two of these characteristics, namely gender diversity and board independence, because of their constraining effects on earnings management. Earnings management is related to CSR decoupling, as both concepts pertain to situations where fundamental (CSR or financial) performance does not correspond with what is reported. Several articles have provided empirical evidence that a higher number of female board members reduces earnings management (Arun, Almahrog, & Aribi, 2015; Gavious, Segev, & Yosef, 2012; Lakhal, Aguir, & Lakhal, 2015; Srinidhi, Gul, & Tsui, 2011), because women are more ethical, demonstrate greater risk-aversion and improve board monitoring effectiveness. These characteristics may also cause a more gender-diverse board to reduce a firm’s CSR decoupling. Furthermore, two meta-analyses have shown that the increased monitoring effectiveness of board independence constrains earnings management (Lin & Hwang, 2010; Garcia-Meca & Sanchez-Ballesta, 2009), which may also apply to constraining CSR decoupling. Both board independence and gender diversity are therefore interesting components of corporate governance to investigate in relation to CSR decoupling.

(6)

Gender diversity has been extensively researched in the past in relation to subjects like firm value (Campbell & Mínguez-Vera, 2008), firm financial performance (Low, Roberts, & Whiting, 2015), CSR performance (Harjoto, Laksmana, & Lee, 2015), CSR disclosure (Rao & Tilt, 2016) and risk-taking (Perryman, Fernando, & Tripathy, 2016). Past research has shown that women on boards of directors possess certain characteristics that may positively influence companies (Kim & Starks, 2016). For instance, female directors have fewer attendance problems and are more likely than male directors to hold a doctoral degree and have expert backgrounds outside of business (Hillman, Cannella Jr, & Harris, 2002; Adams & Ferreira, 2009). There is also reason to believe that female directors might influence CSR decoupling. Overstated CSR decoupling namely comes with a ‘decoupling risk’, and women tend to be more risk-averse than men (Graafland & Smid, 2019; Croson & Gneezy, 2009). Moreover, female directors positively influence the monitoring effectiveness of the board, which contributes to the integrity of a company’s (CSR) statements and ensures that shareholders’ interests are served (Peasnell, Pope, & Young, 2005).

Board independence has also received a lot of attention from researchers (Chen, Cheng, & Wang, 2015; Liu, Miletkov, Wei, & Yang, 2015). Boards with a higher proportion of independent directors are more able to effectively monitor management, because they are not involved in the daily operations (De Villiers, Naiker, & Van Staden, 2011) and do not hold an official position in the company (Donnelly & Mulcahy, 2008). The increased effectiveness of monitoring by the board may impact CSR decoupling, since it contributes to a more skeptical and critical view of management’s decisions and initiatives (Baysinger & Hoskisson, 1990; McNulty & Pettigrew, 1999). This could entail an independent board criticizing CSR initiatives that are decoupled from practice. Moreover, an independent board would make sure that actual CSR practices are communicated externally, since this would positively influence firm value and therefore be of shareholders’ best interest. Furthermore, independent directors are more concerned with corporate social responsibility and positively influence a firm’s corporate social performance (Ibrahim, Howard, & Angelidis, 2003; Huang, 2010). Overall, research indicates that a higher proportion of independent board members could impact both sides of CSR decoupling and consequently reduce the gap between practice and policy.

Therefore, I investigate the impact of these two board characteristics on CSR decoupling. The research question guiding this research will thus be:

(7)

What is the effect of board gender diversity and board independence on the extent to which a company engages in CSR decoupling?

I use data on 18.351 firm-year observations for 4.870 listed companies between 2009 and 2017 from all around the world to empirically test my hypotheses. Results from my random-effects regressions show that gender diversity has a significant positive effect on CSR decoupling, which is contrary to my expectations. Second, I find that board independence has a significant negative effect on CSR decoupling, which is in line with my hypothesis. Results of the first additional analysis show that the positive effect of gender diversity is caused by its effect on the social pillar of CSR, and to a lesser extent the governance pillar of CSR. Furthermore, these results show that the negative effect of board independence can be fully attributed to the governance pillar of CSR. Results of the second additional analysis show that gender diversity has a negative effect on understated CSR decoupling and a positive effect on overstated CSR decoupling. They further show that the negative effect of board independence on CSR decoupling is fully caused by its effect on overstated CSR decoupling, as board independence does not have a significant effect on understated CSR decoupling.

Research on corporate social responsibility has been popular among academics for a long time, with concepts being investigated like corporate social performance (CSP), CSR disclosure and CSR engagement (Carroll & Shabana, 2010). While the literature on CSR decoupling is still limited, the number of journal articles mentioning this subject has increased significantly since the turn of the century. Some research has been conducted to get a better understanding of CSR decoupling in practice and types of CSR decoupling (Graafland & Smid, 2019; Bromley & Powell, 2012; Hawn & Ioannou, 2016; Boiral, 2013). Furthermore, some research has been done regarding the determinants of CSR decoupling, such as customer characteristics, pressures from civil society organizations, and governmental monitoring (Christmann & Taylor, 2006; Ählström, 2010; Marquis & Qian, 2013).

A firm’s environmental and social behavior is the responsibility of the board of directors. While corporate governance characteristics have been extensively researched in relation to CSR disclosure (Said, Hj Zainuddin, & Haron, 2009; Chan, Watson, & Woodliff, 2014), there is a gap in the literature when it comes to the effect of corporate governance characteristics on CSR decoupling. I draw on institutional theory and the literatures on corporate governance and CSR to propose two limiting factors of CSR decoupling, namely board gender diversity and board

(8)

independence. These two factors have already received attention for their positive influences on, amongst others, CSR disclosure and CSR performance (Hafsi & Turgut, 2013; Liao, Luo, & Tang, 2015; Bear, Rahman, & Post, 2010). This research contributes to existing literature by examining the effect of these two corporate governance characteristics on CSR decoupling. Furthermore, this research adds to existing literature by showing the importance of decomposing CSR decoupling into the environmental, social and governance pillars of which it exists and of differentiating between understated and overstated decoupling.

By showing the beneficial effects of board independence, this research can be used by standard setters and regulatory bodies to further argue for the promoting of board independence to reduce CSR decoupling. Furthermore, the results in this research add to the discussion whether an increased representation of women in corporate boards achieves the beneficial effects it intends to. Moreover, the article helps raise awareness of overstated CSR decoupling among different stakeholders of companies. If more stakeholders are aware of the possibility that companies decouple their actual CSR practices from communicated CSR policies, they could decide to more explicitly assess these companies’ reported CSR performance. This would decrease the success rate of companies trying to mislead their stakeholders regarding their CSR practices (Brooke Elliott, Jackson, Peecher, & White, 2013).

In the remainder of this article, I will first provide a theoretical background and introduce the hypotheses. Thereafter, I will explain the research method and present the results of the study. The article ends with the conclusions of the research and a discussion of the results.

2. Theoretical background

In order to gain a better understanding of CSR decoupling, this chapter provides a theoretical background by introducing institutional theory and discussing the concept of CSR decoupling. Afterwards, the literature on gender diversity and board independence are discussed and I explain how they may influence CSR decoupling.

2.1 Institutional theory

I build on international CSR research applying an institutional lens (e.g. Tashman, Marano, & Kostova, 2019; Graafland & Smid, 2019; Crilly, Zollo, & Hansen, 2012) in studying the concept of CSR decoupling. Institutional theory views companies as operating within a “social framework of norms, values, and taken-for-granted assumptions about what constitutes

(9)

appropriate or acceptable economic behavior” (Carpenter & Feroz, 2001, p. 565). Furthermore, companies are viewed as operating within an ‘organizational field’, which can be defined as “those companies that, in the aggregate, constitute a recognized area of institutional life” (DiMaggio & Powell, 1983, p. 148). Institutional theory states that companies conform within an organizational field because doing so rewards them with increased legitimacy, resources, and survival capabilities (Scott, 1987).

According to Fernando and Lawrence (2014), there are two dimensions in institutional theory: isomorphism and decoupling. Isomorphism is the constraining process that forces units in a population to resemble other units facing the same environmental conditions (DiMaggio & Powell, 1983). Isomorphism explains through which processes companies can converge within an organizational field. The other dimension of institutional theory is decoupling, which relates to the separation of the external image of a company and the actual organizational practices (Fernando & Lawrence, 2014). In relating decoupling to CSR, Deegan (2009) states that companies may use corporate reports to construct an image of social and environmental responsibility when the actual managerial motivation is one of maximization of profitability or shareholder value.

Institutional theory states that the reason firms decouple is the tension between internal and external pressures (Meyer & Rowan, 1977). Firms have to gain social legitimacy from its stakeholders, while also striving for internal efficiency in response to practical considerations. When stakeholders have competing expectations, this adds to the degree of decoupling by firms (Crilly, Zollo, & Hansen, 2012). Overstated CSR decoupling is expected to be more likely to occur in organizational fields that are ambiguous, uncertain and lack mechanisms to monitor compliance (Greenwood & Hinnings, 1996).

According to Deegan (2009), institutional theory links organizational practices, including CSR practices, to the social framework of norms and values of the organizational field in which a company operates. Within organizational fields, legitimated structures and practices will exist based on shared norms and values. These legitimated structures and practices will spread to companies through isomorphic processes, such as coercion, imitation or normative pressures (Fernando & Lawrence, 2014). This leads to various companies being ‘forced’ to adopt CSR practices to maintain legitimacy (Jamali, Lund-Thomsen, & Khara, 2017). Because of the low

(10)

costs of CSR communication, firms may then resort to overstated CSR decoupling in order to minimize the costs of CSR implementation while enjoying the fruits of CSR disclosure.

2.2 CSR Decoupling

An overview of my literature review on CSR decoupling is presented in table 1. While the concept of CSR decoupling is becoming more popular among academics, the literature on the subject is still limited. I use the article of Wickert, Scherer and Spence (2016) to distinguish two types of decoupling. The first one is overstated CSR decoupling. The common conception of people is that companies are guilty of creating an implementation gap (Kanter, 2009). That is, they focus on communicating formal CSR policies while putting less effort in actually implementing CSR practices. In this situation, the CSR efforts of a company can be seen as being highly symbolic with the purpose of improving reputation. Research has shown that consumers respond favorably to reported CSR efforts (Deng & Xu, 2017). Because of the relatively low cost of CSR communication, this could motivate companies to focus on symbolic rather than substantive CSR efforts. This could be harmful because it is in the best interest of the environment and society that companies act as socially responsible as possible in a substantive manner.

The second type of decoupling is understated CSR decoupling and relates to the communication gap (Wickert, Scherer, & Spence, 2016). Hawn and Ioannou (2016) found that, contrary to popular belief, this type of decoupling takes place at most firms. Understated CSR decoupling pertains to companies that focus more on actual CSR practices while not actively communicating their CSR efforts externally. In this situation, external audiences could well assume that a company is not actively engaging in CSR if that company does not convey and communicate its CSR efforts to stakeholders. This is detrimental to firms and their shareholders because the company incurs costs for which the full benefits, in terms of higher financial performance (Wang, Dou, & Jia, 2016), are not obtained. To sum up, overstated CSR decoupling means that the external CSR communication exceeds the internal CSR performance and understated CSR decoupling means the opposite. Both types of decoupling concern a misalignment between CSR communications and CSR performance and it is desirable to limit this misalignment.

Based on the current CSR literature, differing stakeholder expectations are seen as the main driver of overstated CSR decoupling (Crilly, Zollo, & Hansen, 2012). On the one hand there

(11)

TABLE 1

Prior research related to CSR decoupling

Study Relevant contribution Data source Theory applied Country

Ählström (2010) Demands of CSOs may cause firms to decouple practice from the responsible business discourse.

Interviews Discourse theory; Institutional theory

Sweden

Boiral (2013) Firms’ sustainability disclosure does not necessarily correspond with their sustainability performance.

Content analysis on sustainability reports Voluntary disclosure theory; legitimacy theory Global Bromley and Powell (2012)

Decoupling can be divided in a gap between policy and practice and a gap between means and ends.

Literature review Institutional theory n/a

Brooke Elliott et al. (2013)

Firms’ disclosed CSR performance influences investors’ beliefs about firm value. Experiment using graduate business students Affect-as-information theory n/a Cho, Laine, Roberts, and Rodrigue (2015)

Contradictory societal and institutional pressures force organizations to engage in decoupling. Content analysis on websites and reports Signaling theory; legitimacy theory US Christmann and Taylor (2006)

Firms may select their level of decoupling based on customer characteristics.

Surveys No explicit theory applied.

China

Crilly, Zollo, and Hansen (2012)

Firms facing pressures to decouple, may do this either intentional or emergent.

Interviews Institutional theory Global

Delmas and Burbano (2011)

Drivers of greenwashing include consumer demand, firm characteristics, incentive

structure, and uncertain regulatory environment.

Literature review Institutional theory n/a

Du (2015) Greenwashing is significantly negatively associated with abnormal returns around the exposure of greenwashing. Content analysis of newspapers; CSMAR database No explicit theory applied China Dyck, Lins, Roth, and Wagner (2018)

Institutional ownership is positively related to CSP.

ASSET4 No explicit theory applied

Global

Fiss and Zajac (2004)

The presence of more powerful and more committed key actors among a firm’s owners reduces the likelihood of decoupling.

Content analysis on annual reports No explicit theory applied Germany Graafland and Smid (2019)

The quality of CSR reporting and locating the responsibility for CSR at the board level reduce decoupling.

(12)

TABLE 1 (continued)

Study Relevant contribution Data source Theory applied Country Hawn and

Ioannou (2016)

A wider gap between internal and external CSR actions is negatively associated with market value.

ASSET4 Institutional theory Global

Jamali, Lund-Thomsen and Khara (2017)

Local SMEs selectively decouple their CSR initiatives from the global CSR agenda.

Interviews Institutional theory India

Lyon and Maxwell (2011)

The probability that a firm engages in CSR decoupling, is affected by the industry it operates in.

Economic analysis No explicit theory applied n/a Marano, Tashman and Kostova (2017)

EM-MNEs do more CSR reporting when they come from less institutionally developed countries.

WIR; Compustat; Osiris

Institutional theory Several developing countries

Marquis and Qian (2013)

The risk of governmental monitoring is negatively

associated with CSR decoupling.

Content analysis on annual and CSR reports

Institutional theory China

Parguel, BenoitMoreau & Larceneux (2011)

Sustainability ratings could act to discourage greenwashing and encourage virtuous firms to persist in their CSR practices.

Survey and experiment

Attribution theory France

Schons and Steinmeier (2016)

CSR decoupling has a negative effect on financial performance if directed toward high-proximity stakeholders.

ASSET4 Institutional theory Global

Wagner, Lutz and Weitz (2009)

Firms whose CSR statements follow its observed CSR behavior receive lower levels of perceived hypocrisy than firms whose CSR statements precede conflicting observed behavior.

Experiment Inoculation theory n/a

Walker & Wan (2012)

Greenwashing has a significant negative effect on financial performance. Corporate websites; Compustat Institutional theory; signaling theory Canada Weaver, Trevino and Cochran (1999)

External pressures for social performance increase decoupling. Survey and archival data Institutional theory US Weber, Davis and Lounsbury (2009)

Practices and policies adopted through coercive processes are associated with more ceremonial adoption.

WDI database; COMTRADE database

Institutional theory Several developing countries

Wickert, Scherer and Sence (2016)

Firm size is positively associated with CSR decoupling.

Literature review Several theories applied

(13)

are internal stakeholders such as shareholders and creditors who want a company to deliver a positive financial performance. On the other hand, there are external influences persuading companies to improve their environmental and social performance. These external influences include legislation, public policy, public opinion and social activism (Bromley & Powell, 2012). Because of the difficulty of coping with these different influences, firms may resort to decoupling. According to Bromley & Powell (2012), “decoupling allows an organization to adopt multiple, even conflicting, policies in response to external pressures, without unduly disrupting daily operations by trying to implement inconsistent strategies” (p. 7). In this sense, companies avoid implementation costs while benefiting from the adoption of multiple policies, thus playing both sides.

Hawn and Ioannou (2016) provide some reasons why companies might engage in understated CSR decoupling. They say that companies may insufficiently communicate their CSR efforts to external audiences because of a policy of secrecy or because they believe that more experimentation with CSR is needed until their CSR efforts have the right scope and nature. Furthermore, they say that companies may not fully reveal their CSR actions because they might not want to attract additional stakeholder pressure or attention by social movements, or because they “might be more risk-averse in terms of how investment analysts may perceive such actions” (Hawn & Ioannou, 2016, p. 2583).

The concept of CSR decoupling is related to greenwashing, but it is important that a clear distinction is made. Whereas CSR decoupling refers to the gap between communicated CSR policies and actual CSR practices, greenwashing is defined as the situation where a firm has poor environmental performance and positive communication about environmental performance (Delmas & Burbano, 2011). Where CSR decoupling concerns an implementation or communication gap, greenwashing is concerned with a performance gap.

Some research has been done regarding the determinants of CSR decoupling. Christmann and Taylor (2006) showed that customer characteristics, such as customer preferences, customer monitoring and sanctions by customers may decrease the degree of CSR decoupling by firms. Ählström (2010) provided significant evidence that pressures from civil society organizations (CSOs) increase the likelihood of CSR decoupling by firms, which is of course contrary to the intentions of these CSOs. Other determinants for which significant statistical evidence has been found, include the likelihood of governmental monitoring (Marquis & Qian, 2013), firm size

(14)

(Wickert, Scherer, & Spence, 2016) and negative media attention for ethical failures (Weaver, Trevino, & Cochran, 1999). However, to the best of my knowledge the effect of board characteristics on CSR decoupling has not yet been researched.

2.3 Hypotheses development 2.3.1 Gender Diversity

A substantial and growing literature links board gender diversity to observable firm outcomes (Joecks, Pull, & Vetter, 2013). In this regard, gender diversity is defined as the percentage of women on the board of directors (Campbell & Mínguez-Vera, 2008). Previous studies have shown that gender diversity has positive effects on both CSR disclosure (Rao & Tilt, 2016) and CSR performance (Harjoto, Laksmana, & Lee, 2015), so one might think that it has a negative effect on CSR decoupling. While the results are partially mixed, the overall trend in the literature is that gender diversity has beneficial influences. This has led some countries to introduce gender quotas for corporate boards. However, while board diversity is increasing, women still remain underrepresented in the boardroom and progress is slow (Catalyst, Quick Take: Women on Corporate Boards, December 2018).

Women have certain characteristics that could be of influence in the boardroom. One of them is that women have a lower risk appetite than men (Croson & Gneezy, 2009). Arch (1993) explains this phenomenon by stating that men see a risky situation as a challenge that calls for participation, while women see it as a threat that encourages avoidance. While this was mostly focused on risk related to subjects like investment management and strategic planning, it could also apply to the risk related to overstated decoupling. Firms face a ‘decoupling risk’ if their CSR impacts are not in line with their promises as formulated in their formal policy’s. They run the risk of “harming their reputation even more than companies with similar (or worse) social impacts that did not raise expectations by such policies” (Graafland & Smid, 2019, p. 260). The lower risk appetite of women suggests that female board members would try to reduce the decoupling risk, which entails reducing the gap between a firm’s policies and practices.

Another positive influence of female board members is that they tend to increase the monitoring effectiveness of the board. Female directors have a wider range of professional experience and have a higher chance of having an advanced degree, compared to male directors, which enhances the board’s expertise (Hillman, Cannella Jr, & Harris, 2002). These added qualities, in turn, enable the board to more effectively monitor management (Hillman & Dalziel, 2003).

(15)

Adams and Ferreira (2009) provided significant statistic evidence for this. They found that female directors have fewer attendance problems than male directors. Furthermore, they found that a more gender diverse board provides their directors with more pay-performance incentives and increases the number of board meetings. If the board is more effective in undertaking its monitoring role, it will be more likely to demand explanations for management’s decisions (McNulty & Pettigrew, 1999; Baysinger & Hoskisson, 1990). This means that if management tries to engage in overstated CSR decoupling, an effectively monitoring board will criticize the emerging gap between policies and practices. The board will then demand a change, due to the potential harm that CSR decoupling can cause (Wagner, Lutz, & Weitz, 2009).

Furthermore, an effectively monitoring board will be better able to protect shareholders’ interests (John & Senbet, 1998). Such a board will make sure that companies that put effort in substantive CSR actions communicate this externally, since this has a positive impact on financial performance (Wang, Dou, & Jia, 2016). Otherwise, the company would incur the costs of CSR efforts but not receive the related benefits from external audiences. In other words, I argue that a higher presence of female board members will increase the monitoring effectiveness of the board and thus increase the degree to which shareholders’ interests are protected, therefore reducing understated CSR decoupling.

Thus, the effect women have on the monitoring effectiveness of boards, together with their lower risk appetite, suggests that a higher representation of women on boards may reduce CSR decoupling. In line with previous studies (Campbell & Mínguez-Vera, 2008; Adams & Ferreira, 2009), I measure the effect of gender diversity in two ways. The first measure is the percentage of female board members and the second measure is the mere presence of women on the board. This leads to the following hypotheses:

Hypothesis 1a: The percentage of female members on the board of

directors is negatively associated with the degree to which a company engages in CSR decoupling.

Hypothesis 1b: Having at least one female member on the board of

directors is negatively associated with the degree to which a company engages in CSR decoupling.

(16)

2.3.2 Board Independence

Board characteristics have been extensively researched in the past in relation to CSR. Of these characteristics, board independence has been closely examined in the existing literature (e.g. Liao, Luo, & Tang, 2015; Tulung & Ramdani, 2018; Bhagat & Black, 2001). When talking about board independence, an independent director can be defined as a “person other than an officer or employee of the company or its subsidiaries or any other individual having a relationship which, in the opinion of the board of directors, would interfere with the exercise of independent judgement in carrying out the responsibilities of a director”2. In the present, most companies understand the important roles played by independent directors since the collapse of large firms like WorldCom and Enron. If a board consists for a large part of directors that are independent of management, it is better able to perform its independent oversight function (Klein, 2002).

By more effectively monitoring management, board independence may lead to the board being better able to protect shareholders’ interests (John & Senbet, 1998). For instance, Peasnell, Pope and Young (2005) found that board independence reduces income-increasing earnings management practices by management. This type of earnings management makes it appear as if the company performed well while losses are actually deferred to a later period, which is not in line with shareholders’ interests. Other researchers found that the proportion of outside directors on a board is associated with a better corporate reputation (Musteen, Datta, & Kemmerer, 2010) and an improved investment efficiency (Liu, Miletkov, Wei, & Yang, 2015). However, research on board independence has provided mixed results which is mainly due to cultural differences. According to international research, a higher proportion of non-executive directors has a negative effect on firm performance in New Zealand (Fauzi & Locke, 2012), has no effect on firm performance in Malaysia and Bangladesh (Johl, Kaur, & Cooper, 2015; Rashid, De Zoysa, Lodh, & Rudkin, 2010) and has a positive effect on firm performance in Ireland and China (Liu, Miletkov, Wei, & Yang, 2015; O'connell & Cramier, 2010). Nevertheless, nearly all researchers agree that board independence improves the monitoring efficiency of the board.

Fama and Jensen (1983) argue that independent directors are more effective in monitoring management because of their incentive to maintain their reputational capital. They want to gain

(17)

and retain a reputation of being good at what they do, which is being effective monitors of management. By monitoring management more effectively, independent directors are more likely to demand explanations for management’s decisions and criticize misguided initiatives (Baysinger & Hoskisson, 1990; McNulty & Pettigrew, 1999). Independent board members are thus more skeptic and critical in performing the monitoring role of the board. I argue that a more effectively monitoring board will also be more critical regarding initiated CSR policies and initiatives, in the sense that it will be more likely that decoupled CSR communications are criticized and opposed so as to reduce overstated CSR decoupling.

Furthermore, previous research has shown that independent directors positively influence a firm’s corporate social performance (Huang, 2010; Strandberg, 2005). These studies make clear that independent directors generally have an intrinsic motivation to enhance a firm’s CSR. If a firm commits to certain CSR initiatives and implements them, an independent board will make sure that it is disclosed since this increases firm value and thus is in shareholders’ best interests (Plumlee, Brown, Hayes, & Marshall, 2015; Wang, Dou, & Jia, 2016). I thus also argue that an independent board will reduce understated CSR decoupling. Therefore, I develop the following hypothesis:

Hypothesis 2: The percentage of independent members on the board of

directors is negatively associated with the degree to which a company engages in CSR decoupling.

3. Research method

In this chapter the research method of this study is explained. First, I explain how I produced the sample that is used in this research. Second, I discuss the dependent variable of the research, namely CSR decoupling. Third, I elaborate on the independent variables and the controls that are used. Finally, I present my empirical model and discuss the estimation procedure that is used.

3.1 Sample

Most public companies are being evaluated on their environmental, social and governance (ESG) performance by various rating agencies. In this research, I use ESG data provided by Bloomberg since it more closely reflects on the level of disclosure of firms compared to other

(18)

providers of ESG ratings3. Bloomberg collects ESG disclosure data for more than 10.000 publicly-listed companies all over the world. The initial sample of this research consists of all the companies of which Bloomberg provides an ESG disclosure score and ranges from 2009 to 2017. The sample begins at 2009 because this is the first year for which I can retrieve ESG disclosure data from Bloomberg. 2017 was the most recent year for which Bloomberg provided ESG disclosure scores for most companies. This results in an initial sample of 87.314 firm-year observations from 10.951 companies.

For the measurement of CSR decoupling, I also need data on companies’ ESG performance, which I retrieve from Asset4. Due to missing data, this reduces the sample to 27.950 firm-years from 5.398 companies. Furthermore, I collect data on gender diversity and board independence from Asset4. After merging the data, this results in an unbalanced panel dataset of 18.351 firm-year observations for 4.870 listed companies between 2009 and 2017. Table 2 contains an

3 Based on Siew, Balatbat, & Carmichael (2016, p. 5)

TABLE 2

Geographical representation of sample firms

Region Number of firm-years Percent

Anglo-Saxon outside Europe

United States, Canada, Australia and New

Zealand

7.536 41,1

Asia 5.044 27,5

Other 1.920 10,5

Anglo-Saxon within Europe

Great Britain and Ireland

1.507 8,2

Other Western Europe

Germany, Austria, Switzerland, France, Belgium, The Netherlands and Luxembourg

1.396 7,6

Scandinavia

Norway, Sweden, Denmark and Finland

530 2,9

Mediterranean Europe

Greece, Italy, Spain, Portugal

418 2,3

(19)

overview of the geographical representation of my final sample. As this table shows, most companies in my sample are located in Anglo-Saxon countries outside Europe or in Asia.

3.2 Variable measurement 3.2.1 CSR decoupling

This research makes use of available data from online databases. Table 3 contains a list of all the variables used in my research model. The first step of my research was to collect data about CSR decoupling. As mentioned before, the definition of CSR decoupling is the situation where “communication about the implementation of CSR-related organizational practices does not correspond with the actual implementation of these practices” (Wickert, Scherer, & Spence, 2016, p. 29).

In line with Tashman, Marano, and Kostova (2019), I measure the degree of CSR decoupling as the absolute value of the difference between the standardized value of a firm’s lagged CSR performance score and actual CSR disclosure score. I take the absolute value of this difference because I look at the degree of misalignment. I standardized these values for accuracy purposes, as they were collected from two different ESG rating agencies. In line with Hawn and Ioannou (2016), I use the one year lagged scores of CSR performance. A higher score of CSR decoupling means that a firm has a greater misalignment between its intensity of CSR disclosure and its actual CSR performance. CSR performance can be defined as the quality of a company’s policies and programs that relate to the company’s societal relationships and the observable outcomes associated with it (Wood, 1991). CSR disclosure can be defined as “the extent to which firms report on a comprehensive set of CSR issues” (Marano, Tashman, & Kostova, 2017, pp. 390-391).

For the measurement of CSR disclosure, I make use of ESG data collected by Bloomberg. This database uses annual reports, sustainability reports, press releases and third-party research to collect data on several hundreds of metrics that cover all aspects of ESG4. It does not measure performance, but transparency. All these metrics are combined to create three different scores based on different types of disclosure: evironmental disclosure, social disclosure, and governance disclosure. Bloomberg also combines these metrics to one overall ESG disclosure score, which I use to measure CSR disclosure. In combining them, the metrics are weighted

4 From the Bloomberg ESG infographic, developed by Framework LLC. Retrieved from:

(20)

TABLE 3 Variable measurement

Variable Proxy Measurement

Dependent variable

CSRdecoupling The degree of

misalignment between the company’s CSR performance and CSR disclosure

The absolute value of the difference between the standardized values of a company’s one year lagged CSR performance score and actual CSR disclosure score.

Independent variables Explanatory variables

PercWOB Percentage of women on the board of directors

The percentage of women on the board of directors.

PresenceWOB Presence of women on the board of directors

This measure will be 1 if a company has one or more female directors on the board of directors. Otherwise, the measure will be 0.

BoardInd Percentage of independent board members

The percentage of independent members on the board of directors.

Control variables

FirmSize The size of the company The natural logarithm of the total assets at the end of the fiscal year, measured in US dollars.

RoA Return on assets Earnings before interest and taxes divided by average total assets, lagged by one year.

SalesGrowth Sales growth The percentage change in sales relative to the year before, lagged by one year.

RnDintens R&D intensity The ratio of R&D expenditures to sales.

Leverage Leverage Total debt divided by total shareholders’ equity. CapIntens Capital intensity The ratio of assets to sales.

GovHoldings Governmental holdings The percentage of holdings of 5% or more held by a government institution.

BoardSize Board size The total number of board members at the end of the fiscal year.

BMattendance Board meeting attendance

The average overall attendance percentage of board meetings.

BoardMeetings Number of board meetings

(21)

according to their importance in general and their importance in the industry that a company operates in. The ESG disclosure score can have a value between 0,1 and 100.

I make use of ESG data of Asset4 to measure CSR performance. As a part of Thomson Reuters, Asset4 is one of the biggest ESG rating agencies and collects data in various ESG categories. A total of 178 metrics are used to produce ESG performance scores in ten different categories. These are weighted on their relative importance in general and in their industry and combined to one overall ESG performance score. The score is expressed as a percentage, ranging from 0% to 100%. These ten categories can also be combined to the same three types of scores as the ESG disclosure scores of Bloomberg. To get a better idea of what constitutes the ESG performance score, refer to Appendix A for an overview of the ten categories of which it is comprised. I use the overall ESG performance score to measure CSR performance.

3.2.2 Gender diversity

Board diversity can be defined as the variation among board members, which can be derived from multiple sources such as managerial background, personalities, education, gender and age (Coffey & Wang, 1998). I use one aspect of board diversity, namely gender diversity. Gender diversity can be defined as the representation of gender differences on boards of directors (Erhardt, Werbel, & Shrader, 2003). I collect the data on gender diversity from Bloomberg. I measure gender diversity as the percentage of female members on the board of directors. If a company has a supervisory board and a management board (two-tier board), the measure will be calculated based on the supervisory board.

TABLE 3 (continued)

Variable Proxy Measurement

Control variables (continued)

IndDummy The industry of the company

The primary industry in which a company is active, ranging from 1 (Agriculture) to 8 (Services). The industry agriculture has been omitted.

YearDummy Year Nine dummy variables are used to control for year fixed effects. The year 2009 has been omitted. RegionDummy The region in which the

company is incorporated

Six dummy variables are used to control for the effect of regional differences. The region ‘Other’ has been omitted.

(22)

3.2.3 Board independence

Board independence can be defined as the percentage of independent board members on the board of directors (Osma, 2008). An independent director can be defined as “a person other than an officer or employee of the company or any other individual having a relationship, which, in the opinion of the company's board of directors, would interfere with the exercise of independent judgment in carrying out the responsibilities of a director” (NASDAQ, 2006). I collect the data on board independence from Asset4. I measure board independence as the percentage of independent board members. Consistent with the measurement of gender diversity, if a company has a supervisory board and a management board (two-tier board), the measure of board independence will be calculated based on the supervisory board.

3.3 Control variables

Delmas and Burbano (2011) argued that firm characteristics may influence the overall strategies available to a company. I therefore control for firm size, since smaller firms are less able to explicitly implement CSR programs due to a lack of experience (Lepoutre & Heene, 2006) and tend to less actively communicate CSR than larger firms (Wickert, Scherer, & Spence, 2016). I measure firm size using the natural logarithm of total assets in US dollars. I also control for

return on assets because a positive (negative) financial performance could enable (constrain)

firms to invest in costly CSR implementation programs (Waddock & Graves, 1997). I measure return on assets with a one year lag as earnings before interest and taxes divided by average total assets. In line with Hussain, Rigoni and Orij (2018) I control for sales growth, since a growth in sales may provide firms with opportunities to invest in CSR initiatives. I measure sales growth with a one year lag as the percentage change in sales relative to the year before.

I control for R&D intensity as the ratio of R&D expenditures to sales, because innovative firms could be better able to improve their CSR activities (McWilliams & Siegel, 2000). Furthermore, I control for leverage as the ratio of debt to equity, as firms with more debt have fewer resources to engage in substantive CSR practices (Marano, Tashman, & Kostova, 2017). I control for

capital intensity as the ratio of assets to sales, because it may influence the way in which firms

deploy their assets to improve CSR performance (Russo & Fouts, 1997). I also control for

governmental holdings because ownership by governmental institutions may affect firms’

propensity to engage in CSR decoupling (Marquis & Qian, 2013). I measure governmental holdings as the percentage of holdings of 5% or more that are held by a government institution.

(23)

I control for board size as the total number of board members, because larger boards may provide more expertise and knowledge (Dalton, Daily, Johnson, & Ellstrand, 1999) related to CSR. Moreover, I control for board meeting attendance as the average percentage of board members attending board meetings in a given year. A higher attendance suggests that board members are more serious and active in monitoring management and thus contribute to board effectiveness (Mishra & Mohanty, 2014). I control for the number of board meetings because a higher number of board meetings suggests that a higher level of monitoring activity is delivered (Laksmana, 2008). I also control for the industry in which a firm operates. The level of decoupling is expected to be more likely in organizational fields that are ambiguous and lack mechanisms to monitor compliance (Greenwood & Hinnings, 1996). Furthermore, incentives to translate CSR policies into implementation programs differ between industries (Brammer & Pavelin, 2006). I use eight dummy variables to control for industry, based on the industry groups of the Standard Industrial Classification. I control for year fixed effects by using nine different dummy variables. Delmas and Burbano (2011) argued that the regulatory context of firms will affect greenwashing. If the regulation of corporate social performance is weak, firms will more easily get away with decoupling. I therefore control for this effect with six regional dummy

variables. I use the categorization as proposed by Moon et al. (2012), which is based on

different types of capitalism and can be viewed in table 2.

I collected the region and industry data from Bloomberg. The data on firm size, return on assets, sales growth, R&D intensity, leverage and capital intensity were collected from Worldscope and the data on governmental holdings is collected from Datastream. The data on board size, board meeting attendance and number of board meetings were collected from Asset4. To avoid eliminating firm-years with missing data on one or two control variables, I replaced missing data on firm size, return on assets, sales growth, leverage and capital intensity with the means of respective control variables. This concerned an insignificant maximum of 2,1% of all firm-years in my sample. I also replaced missing data on R&D intensity and board meeting attendance with their respective means, which concerned respectively 60% and 25% of the firm-years. Reperforming the regressions without the firm-years for which the missing data on board meeting attendance was replaced provided qualitatively similar results.

3.4 Empirical model and estimation procedure

My empirical model to analyze the association between gender diversity and board independence on the one hand and the degree of CSR decoupling on the other is as follows:

(24)

𝐶𝑆𝑅𝑑𝑒𝑐𝑜𝑢𝑝𝑙𝑖𝑛𝑔𝑖,𝑗 = 𝛽0+ 𝛽1𝑃𝑒𝑟𝑐𝑊𝑂𝐵𝑖,𝑗+ 𝛽2𝐵𝑜𝑎𝑟𝑑𝐼𝑛𝑑𝑖,𝑗+ 𝛽3𝐹𝑖𝑟𝑚𝑆𝑖𝑧𝑒𝑖,𝑗 + 𝛽4𝑅𝑜𝐴𝑖,𝑗+ 𝛽5𝑆𝑎𝑙𝑒𝑠𝐺𝑟𝑜𝑤𝑡ℎ𝑖,𝑗+ 𝛽6𝑅𝑛𝐷𝑖𝑛𝑡𝑒𝑛𝑠𝑖,𝑗 + 𝛽7𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒𝑖,𝑗+ 𝛽8𝐶𝑎𝑝𝐼𝑛𝑡𝑒𝑛𝑠𝑖,𝑗+ 𝛽9𝐺𝑜𝑣𝐻𝑜𝑙𝑑𝑖𝑛𝑔𝑠𝑖,𝑗 (1) + 𝛽10𝐵𝑜𝑎𝑟𝑑𝑆𝑖𝑧𝑒𝑖,𝑗 + 𝛽11𝐵𝑀𝑎𝑡𝑡𝑒𝑛𝑑𝑎𝑛𝑐𝑒𝑖,𝑗 + 𝛽12𝐵𝑜𝑎𝑟𝑑𝑀𝑒𝑒𝑡𝑖𝑛𝑔𝑠𝑖,𝑗+ 𝛽13𝐼𝑛𝑑𝐷𝑢𝑚𝑚𝑦𝑖,𝑗 + 𝛽14𝑌𝑒𝑎𝑟𝐷𝑢𝑚𝑚𝑦𝑖,𝑗+ 𝛽15𝑅𝑒𝑔𝑖𝑜𝑛𝐷𝑢𝑚𝑚𝑦𝑖,𝑗+ 𝜀𝑖,𝑗

where: 𝐶𝑆𝑅𝑑𝑒𝑐𝑜𝑢𝑝𝑙𝑖𝑛𝑔𝑖,𝑗 = the level of CSR decoupling of firm i for year j; 𝜀𝑖,𝑗 is the error term; and all other variables are defined in Table 3. PercWOB can be replaced by PresenceWOB to arrive at my second empirical model.

I test my model using random-effects regressions, as this is the common technique used to analyse panel data sets with many cross sections and few time periods (Tashman, Marano, & Kostova, 2019). My data consists of 4.870 firms (cross sections) of which 64% has data available on CSR decoupling for only 4 years or less. Moreover, using a random-effects model is preferable for generalizing the findings beyond the sample of the study (Maddala, 1992). Lastly, some of my variables do not vary over time, such as the industry and region dummy variables. This makes a fixed-effects regression inappropriate as this would have forced me to drop these variables from my model. Prior to the analyses, I winsorized the variables RoA,

SalesGrowth and CapIntens at the top and bottom 1 percent and I winsorized the variables Leverage and RnDintens at the top and bottom 5 percent to minimize the impact of extreme

values.

4. Results

The results of the statistical analyses are presented in this chapter. First, some descriptive statistics and the correlations for all the variables in my model are provided. Thereafter, the results of the regression analysis and the robustness tests are discussed. This chapter ends with the results of my additional analyses.

4.1 Descriptive statistics and correlation analysis

Table 4 provides the summary statistics for the full sample that I used to test my empirical model. It shows that firms in my sample on average engage more in overstated CSR decoupling than in understated CSR decoupling (mean = ,46). Furthermore, overstated CSR decoupling occurs in more extreme forms than understated CSR decoupling (6,17 versus -2,78). Table 4

(25)

suggests that women are underrepresented in boards of directors, with an average of 11 percent of the board members being female and 36 percent of the boards having no female members at all. The average board of directors of the firms in my sample consists almost half of independent directors (mean = 49,11). Table 4 further shows that an average firm in my sample has a return on assets of 4,7 percent; in 12 percent of the firm-year observations, the firm experienced a loss in the previous year. The average size (in terms of the book value of assets) of the firms in my sample is $2,57 billion (unreported in Table 4) and the average sales growth is 13%. Table 4 also shows that the average ratio of R&D expenditures to sales is 7,82 and that the average debt to equity ratio is 89,21. Furthermore, the average assets to sales ratio is 5,53 and a low number of significant holdings are held by government institutions (mean = 2,36; only 7 percent of the firm-year observations involve such governmental holdings. Finally, the average board consists of just over ten members, the average attendance percentage of board meetings is 89,24% and the average number of board meetings that are held is between nine and ten.

TABLE 4 Summary statistics

Mean SD Min Max

CSRdecoupling 1,10 ,87 ,01 6,17 CSRdecoupling_non_abs ,46 1,32 -2,78 6,17 PercWOB 11,46 11,17 0 83,31 PresenceWOB ,64 ,48 0 1 BoardInd 49,11 30,63 1,34 95,28 FirmSize 7,27 1,30 1,83 12,01 RoA 4,73 9,51 -46,61 33,07 SalesGrowth ,13 ,44 -,64 2,72 RnDintens 7,82 4,74 0 19,44 Leverage 89,21 100,12 0 381,37 CapIntens 5,53 9,75 ,32 63,75 GovHoldings 2,36 10,55 0 100 BoardSize 10,11 3,51 1 30 BMattendance 89,24 9,00 34 100 BoardMeetings 9,50 4,89 1 125 n = 18.351 observations for 4.870 firms. To provide better insight into firms’ CSR decoupling, I included the summary statistics of the non-absolute value of CSR decoupling.

(26)

TABLE 5

Pearson’s correlation matrix

1 2 3 4 5 6 7 8 9 10 11 12 13 14 1 1 2 ,004 1 3 -,004 ,804*** 1 4 -,107*** ,131*** ,140*** 1 5 ,178*** -,088*** -,089*** -,221*** 1 6 -,007 ,044*** ,042*** ,011* ,083*** 1 7 -,031*** -,074*** -,082*** -,005 -,099*** -,011** 1 8 -,036*** ,044*** ,047*** ,071*** -,175*** -,223*** ,103*** 1 9 ,036*** ,041*** ,064*** -,010* ,231*** -,075*** -,041*** -,024*** 1 10 -,020*** -,021*** -,008* -,008 ,099*** -,259*** ,072*** ,244*** ,173*** 1 11 ,031*** ,023*** ,005 -,046*** ,179*** ,013*** -,021*** -,030*** ,066*** ,045*** 1 12 ,066*** ,026*** ,077*** -,135*** ,166*** ,011* -,025*** -,046*** ,080*** ,045*** ,053*** 1 13 ,037*** ,011* -,020*** -,271*** ,055*** ,015** ,016*** -,037*** -,008 ,005 ,034*** -,077*** 1 14 ,060*** -,020*** -,053*** -,055*** ,180*** -,101*** -,007 -,029*** ,106*** ,082*** ,119*** ,009 ,046*** 1 1. CSRdecoupling; 2. PercWOB; 3. PresenceWOB; 4. BoardInd; 5. FirmSize; 6. RoA; 7. SalesGrowth; 8. RnDintens; 9. Leverage; 10. CapIntens; 11. GovHoldings; 12. BoardSize; 13. BMattendance; 14. BoardMeetings; ***, **, * coefficients are significant at the levels of respectively 1%, 5% and 10%. J.S . Bos gra M S c. T he si s – A & C 25

(27)

Table 5 provides the correlations for my main variables. The correlation patterns indicate no multi-collinearity, the only correlated variables are the two different measures of gender diversity (PercWOB and PresenceWOB), but they will not be included simultaneously in the regression analysis. Furthermore, the variance inflation factors (not reported) indicate no multi-collinearity issues. The highest VIF remains well below a value of 10, which has been suggested as the maximum level of VIF (O'Brien, 2007).

4.2 Regression analysis

Table 6 presents the main results of the random-effects regressions with CSR decoupling as the dependent variable. All regressions include dummy variables that control for region, industry and year fixed effects. I start my empirical analysis by running a regression which only tests the effects of the control variables on my dependent variable. The results of this regression are presented in column (1) of table 6 and show that only four control variables have a significant effect on CSR decoupling. FirmSize has a significant positive effect on CSR decoupling (β = 9,80; p-value < ,01) and this effect holds for the other models. This means that larger firms generally have a larger gap between their CSR performance and their CSR disclosure. The results also show that RoA (β = -2,60; p-value < ,01) and RnDinten (β = -2,97; p-value < ,01) have a significant negative effect on CSR decoupling, which holds for the other models. Furthermore, BoardSize has a significant positive effect on CSR decoupling (β = 1,67; p-value < ,1), but this effect is weak and does not hold for the models in which BoardInd is included. All other control variables do not have a significant effect in any of my empirical models.

I continue my empirical analysis by investigating the effect of my independent variables on CSR decoupling. I use two different measures for gender diversity, the percentage of women on the board (PercWOB) and the presence of women on the board (PresenceWOB). I run random effects regressions with these two measures separately, of which the results are depicted in Table 6, column (2) and (3). These results show that PercWOB has a significant positive effect on CSR decoupling (β = 2,17; p-value < ,05), which is contrary to my expectations. Table 6, column (5) presents the results for the full empirical model and shows a significant positive effect of PercWOB on CSR decoupling as well (β = 2,35; p-value < ,05). I hypothesized that the percentage of women on the board would have a negative effect on CSR decoupling, so hypothesis 1a is rejected. Furthermore, the results in Table 6, column (3) show that

(28)

TABLE 6

Results of the random-effects regression: CSRdecoupling

(1) (2) (3) (4) (5) (6) FirmSize 9,80*** (,011) 9,59*** (,011) 9,64*** (,011) 9,83*** (,011) 9,61*** (,011) 9,66*** (,011) RoA -2,60*** (,001) -2,79*** (,001) -2,77*** (,001) -2,47** (,001) -2,66*** (,001) -2,64*** (,001) SalesGrowth ,13 (,013) ,28 (,013) ,26 (,013) ,06 (,013) ,21 (,013) ,19 (,013) RnDintens -2,97*** (,002) -2,97*** (,002) -2,97*** (,002) -2,98*** (,002) -2,97*** (,002) -2,97*** (,002) Leverage -,27 (,001) -,57 (,001) -,58 (,001) -,24 (,001) -,55 (,001) -,56 (,001) CapIntens -,43 (,001) -,35 (,001) -,36 (,001) -,38 (,001) -,29 (,001) -,30 (,001) GovHoldings -,63 (,001) -,71 (,001) -,66 (,001) -,56 (,001) -,63 (,001) -,59 (,001) BoardSize 1,67* (,002) 1,71* (,002) 1,68* (,002) 1,25 (,002) 1,25 (,002) 1,21 (,002) BMattendance ,92 (,001) ,68 (,001) ,71 (,001) ,26 (,001) -,04 (,001) ,01 (,001) BoardMeetings 1,10 (,001) 1,16 (,001) 1,18 (,001) 1,02 (,001) 1,08 (,001) 1,10 (,001) PercWOB (H1a) 2,17** (,001) 2,35** (,001) PresenceWOB (H1b) 1,44 (,017) 1,64 (,017) BoardInd (H2) -3,17*** (,001) -3,46*** (,001) -3,43*** (,001)

IndustryDummies Included Included Included Included Included Included

YearDummies Included Included Included Included Included Included

RegionDummies Included Included Included Included Included Included

R-squared ,073 ,075 ,075 ,074 ,076 ,076 Wald 𝜒2 813,65 *** 809,35 *** 806,30 *** 823,54 *** 821,40 *** 818,17 *** n 18.732 18.376 18.376 18.707 18.351 18.351

The values in parentheses are the standard errors. ***, **, * coefficients are significant at the levels of respectively 1%, 5% and 10%.

(29)

significant). This is consistent with the results of the full empirical model using this measure for gender diversity, which are presented in table 6, column (6). It also shows an insignificant positive effect of PresenceWOB on CSR decoupling (β = 1,64; not significant). As I hypothesized that the presence of women on the board would have a negative effect on CSR decoupling, hypothesis 1b is also rejected. Overall, the results weakly suggest that female board members increase firms’ CSR decoupling rather than decrease it.

To test my second hypothesis, I run a regression that tests the effect of BoardInd on CSR decoupling. The results are presented in table 6, column (4) and show that BoardInd has a significant negative effect on CSR decoupling (β = -3,17; p-value < ,01). This is consistent with the results of the regressions of my full model. These results are presented in table 6, column (5) and (6) and show a significant negative effect of BoardInd on CSR decoupling as well (β = -3,46 and -3,43; p-value < ,01). I hypothesized that board independence would reduce the degree to which firms engage in CSR decoupling, so my results are in line with hypothesis 2.

4.3 Robustness tests

I conduct several robustness tests to corroborate my results. For these tests I use both of my complete models, which only differ in the measurement of gender diversity. Results of the robustness tests are reported in Appendix B, panel A and B. To address the possibility of heteroskedasticity, I run a Breusch-Pagan test for both models. The results show that I must reject the null hypothesis that the residuals in my models are constant. I thus cannot be sure that the error structures in my models are not heteroskedastic. Therefore, I also run random-effects regressions of my models with robust standard errors which take heteroskedasticity into account (White, 1980). The results from these regressions confirm the main results.

Furthermore, I also run random-effects regressions using clustered robust standard errors using two different clustering methods. First, I used firms’ country of corporation as cluster variable and excluded the region dummy variables. Second, I generated a new variable using the first two digits of firms’ SIC code and used this variable as cluster variable to run the regression, while excluding the industry dummy variables. The results of these regressions remain largely similar to the results from my main regression analysis. A notable change is that the negative effect of CapIntens on CSR decoupling is very significant if I cluster by the first two digits of firms’ SIC codes. Lastly, I also test both of my full models using a linear regression with robust standard errors clustered by firm. These results are also qualitatively similar to the main

(30)

regressions. The only noteworthy difference is that the positive effects of BoardSize and

BoardMeetings on CSR decoupling are significant in the linear regression. Overall, the

robustness tests confirm the main results.

4.4 Additional analyses

4.4.1 Decomposing CSR decoupling

To get a further understanding of the results of my main analysis, I decompose my dependent variable into the three underlying decoupling scores. As mentioned in the methods section, I used ratings from two different rating agencies to calculate the CSR decoupling variable. Both rating agencies also provide ratings on the three different pillars on which respectively the total CSR performance and total CSR disclosure scores are based. These are the environmental, social and governance pillars. I use the same method to generate the three decoupling variables as I used for generating the total CSR decoupling score5. Table 7 shows the summary statistics for the decomposed decoupling variables. All three types of decoupling have similar statistics, with only the governance decoupling score having a noteworthy lower minimum value than the other two decoupling scores.

Appendix A shows the different categories which constitute the three pillar scores for performance, as provided by Asset4. The categories which constitute the disclosure scores as generated by Bloomberg are similar, including: carbon emissions, waste disposal, diversity,

5 To generate the respective decoupling scores, I use the absolute value of the difference between the one year

lagged performance score and the disclosure score.

TABLE 7

Additional analysis A: Summary statistics

n Mean SD Min Max

ENVdecoupling 12.318 ,977 ,792 0 5,295 ENVdecoupling_non_abs 12.318 ,144 1,249 -2,597 5,295 SOCdecoupling 15.235 ,996 ,811 0 5,384 SOCdecoupling_non_abs 15.235 ,265 1,257 -2,763 5,384 GOVdecoupling 18.351 1,084 ,833 0 5,941 GOVdecoupling_non_abs 18.351 ,455 1,289 -5,258 5,941

To provide better insight into firms’ decoupling, I included the summary statistics of the non-absolute values of the different types of decoupling.

Referenties

GERELATEERDE DOCUMENTEN

In addition, some even contest the view that CSR decoupling is a bad thing (Christensen et al., 2013). Subsequently, looking at the composition of the CSR committee, results show

Using a sample of 17,115 firm years from 40 countries for the time period of 2009 to 2017, this study investigates the role of four corporate governance mechanisms (gender diversity

The combination of board independence and board gender diversity is only not significant to environmental decoupling (-0,0159), while showing significant negative correlations

Corporate social responsibility, real activities earnings management, and corporate governance: evidence from Korea.. Real and accrual‐based earnings management in

Women on Boards : Do They Affect Sustainability Reporting ?, Corporate Social Responsibility and Environmental Management, 21: 351-364. Strategic Management: A

In addition to this, the size and gender diversity of the audit committee only have a negative effect on CSR decoupling as a whole while the age and tenure of audit committee

(2018) Independence (-) Interdependence governance mechanisms (-) CSR gap (-) Independence (+), board diversity (+) CSR reporting Decoupling practices Market value

Consistent with prior studies (e.g. Bear et al., 2010; Byron &amp; Post, 2016; Webb, 2004) and with public debates about female representation on corporate boards, gender diversity