Corporate governance and corporate social
responsibility: A meta-analytical review
P.H.K. Kaman
s1507850
MSc Accountancy & Controlling
ABSTRACT
The relationship between corporate governance (CG) and corporate social responsibility (CSR) has been studied extensively. However, the exact nature of the relationship between CG
mechanisms and CSR remains unclear. In order to fill this gap, I used a sample, consisting of 64 studies, 136 outcomes and 139.102 observations, to systematically analyze the existing research on the relationship between four board level CG mechanisms: board independence, board size, CEO duality and women on board and two CSR outcomes: CSR disclosure and CSR
performance. I find that board size and women on board are significantly related with both CSR disclosure and CSR performance. Board independence is only significantly related with CSR disclosure, while CEO duality on the other hand isn’t significantly related with both CSR outcomes. Additionally, a sub-group analysis reveals that the European and UK based studies show a significant relationship between board independence and CSR disclosure and CSR performance.
Keywords:
Board independence, board size, CEO duality, corporate governance, corporate social responsibility, disclosure, performance, women on board.
Supervisor : N. Hussain, PhD
Co-assessor : A. Rehman Abassi
Date : June 25, 2018
Inhoud
1 Introduction ... 1
2 Theory ... 3
2.2 Theoretical background and review of literature ... 3
2.2 Hypothesis development ... 5
2.2.1 Board independence and CSR ... 10
2.2.2 Women on board and CSR ... 11
2.2.3 Board size and CSR ... 12
2.2.4 CEO duality and CSR ... 13
3 Methodology ... 14
3.1 Data collection and sample ... 14
3.2 Meta-analysis procedure ... 16
4 Results ... 18
4.1. Effect size analysis... 18
4.2. Sub-group analysis ... 21
5 Discussion and conclusion ... 26
5.1 Limitations and Future Research Directions ... 28
1
Introduction
Starting with Enron in 2004 and more recently with Volkswagen, there have been many scandals related with corporate social responsibility (CSR), causing a change in the way corporations behave. Changing the focus from profit-maximization to enhancing firm value by investing in CSR as a way to enhance profitability (Liang & Renneboog, 2017). These ethical, environmental and social related corporate scandals demanded corporations to demonstrate ethical and social responsible behavior and therefore provided a ‘hot-bed’ for CSR (Arvidsson, 2010, p 341). Thus increasing the scientific debate on the relationship between CSR and corporate governance (CG) (Jain & Jamali, 2016; Walls et al., 2012).
Even though the relationship between CG and CSR has been studied extensively (Coffey & Wang, 1998; David et al, 2007; Hussain et al, 2018; Jain & Jamali, 2016; Johnson &
Greening, 1999; Oh et al, 2016; Walls et al., 2012), there is still room to investigate this
relationship further as the exact nature of the relationship remains unclear (Byron & Post, 2016) and a clear theoretical framework lacks still (Hussain et al, 2018; Walls et al. 2012). A more detailed exploration would be required to fill this gap, which this study tries to accomplish by compiling existing research about the relationship between CG mechanisms and CSR.
CG mechanisms can help resolve agency problems that may occur as a result of investing in CSR (Fama & Jensen, 1983). Yet despite this, the results of various studies on the relationship between CSR and board level CG mechanisms are fragmented (Bear et al, 2010; Johnson & Greening, 1999; Walls et al, 2012; Williams, 2003). Both Bear et al (2010) and Williams (2003) researched the effects of the CG mechanism women on board on CSR performance. However both studies showed different results, with Williams (2003) noting a negative association and Bear et al (2010) noting a positive association. Similar differences occur when comparing the results of the research by Johnson & Greening (1999) and Walls et al (2012). Both Walls et al (2012) and Johnson & Greening (1999) researched the effects of the CG mechanism board independence on CSR performance. Walls et al (2012) documented a negative association, whereas Johnson & Greening (1999) documented a positive association. While these studies focused on the relationship between CG mechanisms and CSR performance, the studies on the relationship between CG mechanisms and CSR disclosure show similar fragmentations. For example, when comparing the results of the research by Abdullah et al. (2011) and Lim & Chow
2 (2007). Both researched the effects of the board independence CG mechanism and CSR disclosure. However, the study by Abdullah et al. (2011) showed a negative association, while the study by Lim & Chow (2007) showed a positive association. More inconsistencies are shown while making a comparison between CSR disclosure and CSR performance (Abdullah et al, 2011; Bear et al, 2010; Johnson & Greening, 1999; Lim & Chow, 2007; Walls et al, 2012; Williams, 2003)). As such it remains unclear which CG mechanisms are positively and which are negatively associated with CSR. Which, in turn, limits the development of theories.
This fragmentation is my main motivation for writing this paper. This paper has several main objectives. First of all, in order to better understand the relationship between the various CG mechanisms and CSR outcomes, this paper aims to obtain correlations of research conducted in the relationship between the following four CG mechanisms: board independence, women on board, board size and CEO duality, and two CSR outcomes: CSR performance and CSR
disclosure. Secondly, to quantify prior research in this relationship. In order to achieve these objectives, this study analyses the existing research on the relationship between the CG mechanisms and CSR outcomes by conducting a meta-analyses of 64 studies.
This paper contributes to existing literature by analyzing the relationship between CG and CSR. Prior research, to the best of my knowledge, considered fewer CG mechanisms and didn’t take both outcomes at the same time into consideration. As such the scope of these studies is limited (Byron & Post, 2016; García-Meca & Sánchez-Ballesta, 2010; Jain & Jamali, 2016) and this study will therefore enable us to better understand the relationship between CG and CSR. As such the research question for this study will be:
“How are board independence, women on board, board size and CEO duality associated with CSR disclosure and CSR performance?”
The remainder of the paper is divided into several sections. In the next section the theoretical perspectives describing the relationship between CG and CSR as well as the hypothesis development is being outlined. In the section after that the data collection, the sample and the methodology behind the meta-analysis are described. The fourth section provides a description the results of the study and the final section of the paper will be the discussion and the
2
Theory
2.2 Theoretical background and review of literature
Both CSR and CG have changed over time. CSR has been around for a long time. The earliest notions of CSR being in the late 1800s. Over time what was required of corporations changed and as such CSR has been slowly changing with it (Waddock, 2008). More recently due to the rise of the internet and the focus on sustainability and the environment, corporations have begun investing more in CSR. Which is further illustrated by fortune 500 corporations spending more than 15 billion us dollars on CSR activities. As such the debate on CSR isn’t new. This debate started in the ‘50s with Bowen emphasizing the social responsibilities of corporations (Bowen,
1953; Lee, 2008). Since then, starting with Carroll’s three-dimensional model (Carrol, 1979) and
then later with stakeholder theory (Freeman & Reed, 1983), scientific contributions have helped CSR evolve (Lee, 2008).
CG has also been changing. Traditionally CG can be considered as an agency
perspective, which focuses on maximizing shareholder value and the financial performance of the corporation (Schleifer & Vischny, 1997). CG can therefore be seen as a set of rules to
structure the balance of power, responsibilities and roles among a corporation’s management, its board of directors and its shareholders (Ryan et al., 2010). Starting in the 70’s several waves and more recently corporate scandals, started shifting the focus of CG from resolving agency
problems to social responsible behavior (Aras & Crowther, 2008; Elkington, 2006), which in turn gave rise to the notion that a corporation has to satisfy the needs of all of its stakeholders and not just its shareholders (Freeman & Reed, 1983). As a result CG mechanisms are assumed to affect both a corporation’s financial and non-financial outcomes (Haniffa & Cooke, 2005; Hussain et al. 2018). The relationship between CSR and CG is further exemplified by the concept of good CG. Good CG has at its core four principles These principles being
“accountability”, “fairness”, “responsibility” and “transparency” (Aras & Crowther, 2008) and are related with a corporation’s CSR.
The theoretical perspectives that are often used to explain the relationship between CG and CSR are stakeholder theory (Freeman & Reed, 1983) and agency theory (Jensen &
Meckling, 1976). Agency theory suggests that there is an agent (the management) and a principal (shareholders) (Jensen & Meckling, 1976). Thus, agency theory suggests that ownership and
4 control of the corporation is separated. Agency problems can occur as a result of investing in CSR. CG mechanisms can help resolve these agency problems (Fama & Jensen, 1983).
However, shareholders aren’t the only group a corporation is responsible for (Freeman & Reed, 1983), defining all these groups as stakeholders. The word stakeholder, in turn, refers to those that have a stake in the activities conducted by the corporation and first appeared in 1963, where it was mentioned in a memorandum at the Stanford Research Institute (now SRI International, Inc.). These stakeholders can be divided into two groups. Those within the organizational structure of the corporation and those outside of it. Stakeholders within the organization are the employees, the managers and the owners. Outside of the organization the stakeholders are the customers, the government, the suppliers, the lenders, the society and the shareowners. In order to better satisfy the needs of all these stakeholders CG structures need to be designed in such a way that all stakeholders can benefit both non-financially as well as financially (Windsor, 2006). Other than engaging in activities a corporation should also disclose information about these activities to its stakeholders (Clarkson, 1995). This disclosure of social, environmental and economic information can aid in the legitimization of a corporation’s attempts of stakeholder management and can also change the expectations and perceptions of various social groups (Adams & McNicholas, 2007). As such a corporation could choose to voluntarily disclose information beyond what is considered mandatory (Boesso & Kumar, 2007). Unlike stakeholder theory, agency theory doesn’t explore the relationship between a corporation and its
stakeholders. The stakeholder-agency theory lifts this boundary, by proposing a paradigm that encompasses all the contractual relationships, both explicitly as well as implicitly, between all stakeholders (Hill & Jones, 1992). As such benefits to all stakeholders are supposed to be guaranteed under this paradigm.
Other theories that can explain the relationship between CG and CSR are legitimacy theory and resource-dependence theory. Legitimacy theory is based around the notion that in order to be successful, corporations must act within socially accepted boundaries (O’Donovan, 2002). As such in order for a corporation to gain legitimacy it needs to gain acceptance by its environment (Kostova & Zaheer, 1999). While there are some differences between stakeholder theory and legitimacy theory, there is also some overlap between these two theories. In order for corporations to gain legitimacy, corporations need to act in accordance with what is accepted as social responsible behavior by society and by doing so a corporation needs satisfy the needs of
stakeholders other than its shareholders (Freeman & Reed, 1983).
The resource dependence theory is based around the notion that a corporation is
dependent on its environment (Pfeffer & Salancik, 1978). Focus of this theory is on improving a corporation’s environmental performance depending on the board size and the composition of the board (De Villiers et al. 2011). The theory proposes that external dependencies can be managed by a corporation’s board (Pfeffer & Salancik, 1978) and depending on what types of resources a director brings to the board that director can be qualified as a business expert, a support specialist or a community influential (Hillman et al., 2000). Basically, directors are seen as a source of knowledge due to their experience, expertise and a skills. Which allows an expert director to have access to information as well as various other resources (Hillman & Dalziel, 2003).
All the mentioned theories can at least in part explain the relationship between CG and CSR and as such I will use all them while developing my hypothesizes. As such agency theory, stakeholder theory, stakeholder-agency theory, legitimacy theory and resource dependence theory will form the foundation of the theoretical framework of this study.
2.2 Hypothesis development
In literature CG is often described through the variables board independence and board composition (Hussain et al, 2018). These variables, in turn, are defined through a number of parameters. As can be seen in table 1, the ones that are used most often are the percentage of independent directors, the percentage of women on board, the board size and CEO duality. While other variables are available, they lack the sample size in order to be worthy of interest.
Table 1 also shows that the most used theories to explain the relationship between CG and CSR are agency and stakeholder theory. Despite this other theories like legitimacy and resource dependence theory have been used. Also shown in table 1 is that each study has been linked to either CSR disclosure or CSR performance. Given this, I will develop hypotheses in the next subsection for the CG mechanisms board independence, women on board, board size and CEO duality. Additionally an overview of the theories and assumptions associated with the CG mechanisms and CSR is presented in table 2.
Table 1 Sample studies (N=65)
Studies Theory Applied Governance variable(s) Outcome
variable
Sample firms
Country Coverage year
Abdullah et al. (2011) RDT Board independence Disclosure 100 Malaysia 1
Akhtaruddin et al. (2009) Agency Board independence, Board size, Board audit committee Disclosure 105 Malaysia 1
Allegrini & Greco (2013) Agency Board independence, Board size, CEO duality, Board meetings Disclosure 177 Italy 1
Bear et al. (2010) Agency, Signaling, RDT Women on Board, CEO duality Performance 51 Global 1
Ben Amar et al. (2015) RDT Board independence, Women on Board Disclosure 100 Canada 7
Benomran et al. (2015) Legitimacy Board independence, Board size, CEO duality Disclosure 42 Libya 6
Berrone & Gomez-Mejia (2009) Agency, Institutional,
Environmental management
Board independence, , CEO duality, Director ownership, CG index Performance 469 US 7
Boesso & Kumar (2007)1 Stakeholder Board independence Disclosure 36, 36 Italy, US 1
Bukair and Rahman (2015) Legitimacy Board independence, Board size, CEO duality Disclosure 53 5 middle east 1
Chang et al. (2012) Institutional Board size Performance 200 Korea 1
Chang et al. (2017) Agency, RDT Board independence, Board size. Director ownership, Education of
directors, Age of directors
Performance 293 Korea 3
Cordeiro et al. (2017) n neo-institutional theory Board independence, Board size, Cross directorship Performance 500 India 4
Cuadrado-Ballesteros et al. (2015) Stakeholder-agency Board independence, Women on Board, Board size, Board
meetings, Foreign directors
Disclosure 575 Global 7
Cucari et al. (2018) Stakeholder, RDT Board independence, Women on Board, Board CSR committee,
Age of directors
Disclosure 54 Italy 4
Dardour & Husser (2016) Stakeholder-agency Board independence, Board size, CEO duality Disclosure 89 France 5
Dunn & Sainty (2009) Agency Board independence Performance 104 Canada 4
El Ghoul et al. 2016 Agency, Expropriation
hypothesis, Reputation/ horizon effect
Board size Performance 335 East Asia 10
Eng & Mak (2003) Theory of the firm Board independence, Director ownership Disclosure 158 Singapore 1
Fernandez-Gago et al. (2016) Stakeholder, Agency,
Stewardship theory
Board independence Performance 35 Spain 6
Fodio & Oba (2012) RDT, Slack resources Women on Board Disclosure 16 Nigeria 3
Frias‐Aceituno et al. (2013) Stakeholder Board independence, Women on Board, Board size, Board
meetings, Foreign directors
Disclosure 568 Global 3
García-Sánchez et al. (2015) N/A Board independence, Women on Board, Board size, Board
meetings
Table 1 continued
Studies Theory Applied Governance variable(s) Outcome
variable
Sample firms
Country Coverage year
Glass et al. (2016) N/A Women on Board, CEO duality, Age of directors Performance 500 US 10
Godos-Diez et al. (2018) Agency, Stakeholder, RDT,
Signaling
Board size, CEO duality, Board meetings, Board CSR committee Performance 81 Spain 5
Gul & Leung (2004) Agency, Stewardship Board independence, CEO duality, Director ownership, Board
composition
Disclosure 385 Hong Kong 1
Hafsi and Turgut, (2013) Agency Board independence, Women on Board, Board size, CEO duality,
Director ownership, Experience of directors, Age of directors
Performance 95 USA 1
Halme & Huse (1997) Agency Board size Disclosure 140 4 European 1
Haque (2017) Agency, RBV, Busyness
hypothesis
Board independence, Women on Board, Board size, CEO duality, Board CSR committee, Experience of directors
Performance 256 UK 13
Harjoto et al. (2015) Stakeholder Women on Board, Cross directorship, Experience of directors, Age
of directors
Performance 1489 US 13
Huafang & Jianguo (2007) Agency, Signaling, Board independence, CEO duality Disclosure 559 China 1
Ionel-Alin et al. (2012) Agency Board independence, Board size, Board CSR committee Disclosure 54 Global 1
Isa & Muhammad (2015) Agency Women on Board, Board size, Director ownership, Board
composition
Disclosure 6 Nigeria 10
Janggu et al. (2014) Agency Board independence, Board size, Director ownership, Cross
directorship, Foreign directors
Disclosure 100 Malaysia 1
Jia & Zhang (2013) Critical mass theory Women on Board, Board size Performance 519 China 3
Jia & Zhang (2013) Agency Board size Performance 475 China 1
Johnson & Greening (1999) Agency Board independence Performance 252 US 1
Kassinis et al. (2002) Stakeholder, Agency Board size, Cross directorship, Experience of directors Performance 209 US 5
Kaymak et al. (2017) Stakeholder, Agency Board independence, Board size, CEO duality Disclosure 80 International 1
Khan (2010) Legitimacy Women on Board, Board composition, Foreign directors Disclosure 30 Bangladesh 1
Kilic et al. (2015) Stakeholder, Legitimacy Board independence, Women on Board, Board size Disclosure 25 Turkey 5
Krishnamurti et al (2018) N/A Board independence, Performance 167 Global 1
Lee et al. (2016) Stakeholder Board independence, Board size Performance 215 Korea 6
Liao et al. (2015) Stakeholder; Agency;
Legitimacy
Board independence, Women on Board, Board size, CEO duality, Director ownership, Board meetings, Board CSR committee, Board composition
Table 1 continued
Studies Theory Applied Governance variable(s) Outcome
variable
Sample firms
Country Coverage year
Lim & Chow (2007) Agency Board independence, Board size Disclosure 181 Australia 1
Lone et al. (2016) Agency Board independence, Women on Board, Board size Disclosure 50 Pakistan 5
Majeed et al. (2015) Legitimacy, Social and
political cost theory
Board independence, Women on Board, Board size, Foreign directors
Disclosure 100 Pakistan 5
McGuinness et al. (2017) CMT, Social network ties,
Team dynamics
Board independence, Women on Board, Board size, CEO duality Performance 635 China 4
Michelon & Parbonetti (2012) Stakeholder Board independence, Board size, CEO duality, Board CSR
committee, Community inferentiality board
Disclosure 114 Global 1
Naseem et al. (2017) legitimacy theory Board independence, Women on Board, Board size, Board
meetings
Disclosure 179 Pakistan 6
Oh et al. (2015) Grounded theory Board independence, Board size, Director ownership Performance 654 US 1
Oh et al. (2016) Upper Echelon Board independence, Board size, CEO duality Performance 223 US 6
Oh et al. (2017) N/A Board independence, Board size Performance 1559 US 7
Ortiz-de-Mandojana et al. (2014) Institutional Board independence, CEO duality, CEO duality, Board CSR
committee
Performance 210 North
America, EU 1
Prado-Lorenzo & Garcia-Sanchez (2010) Agency, Stakeholder Board independence, Women on Board, Board size, CEO duality,
Board meetings
Disclosure 283 Global 1
Rao et al. (2012) Agency Board independence, Women on Board, Board size Disclosure 96 Australia 1
Rodríguez-Ariza et al. (2016) SEW Women on Board, Board size, Board meetings Performance 550 Global 6
Saha & Akter (2013) Agency Board independence, Board size, Board audit committee Disclosure 40 Bangladesh 1
Setó-Pamies (2015) Agency, RDT Women on Board Performance 94 Global 1
Surroca and Tribo (2008) Stakeholder, Ent.Hyp., Board independence, CEO duality Performance 358 Global 4
Tauringana &Chithambo (2015) Stakeholder-agency Board size, Director ownership, Board composition Disclosure 215 UK 4
Uyar et al. (2013) Agency, Signaling Board independence, Board size Disclosure 131 Turkey 1
Villaro-Peramato et al. (2016) Agency Board independence Performance 1490 Global 8
Walls et al. (2012) Agency, Stakeholder Board independence, Women on Board, Board size, CEO duality,
Board CSR committee
Performance 500 US 9
Williams (2003) N/A Women on Board Performance 185 US 4
Table 1 Overview of assumptions and theories
CG mechanisms Agency Theory Stakeholder and Legitimacy Theory2 Resource dependence theory
Board Independence
Board independence can help resolve agency problems (Fama & Jensen, 1983, Hussain et al, 2018; Jensen & Meckling, 1976) as outsiders tend to have a broader view of the performance of a corporation (Wang & Coffey, 1992).
Outside directors could ensure that a corporation acts on behalf of all of its stakeholders (Haniffa & Cooke, 2005) and are therefore more likely to act in accordance with what is accepted as social responsible behavior and help confer legitimacy to a corporation (Freeman & Reed, 1983).
Outside directors could be seen as a valuable source of resources (Hillman & Dalziel, 2003) and could positively affect CSR performance (De Villiers et al. 2011).
Women on board
Women and minority directors are less profit-driven and are more responsive to the welfare of broad variety of stakeholders and CSR (Wang & Coffey, 1992).
The presence of women on board could positively influence CSR due to the role women tend play in their professional environment (Betz et al., 1989; Frias-Aceituno et al., 2013) as women tend to play the role of wife and mother.
Female directors could be considered a resource based on the assumption that board diversity brings different experience, expertise and a skills (Amran et al., 2014). Women directors tend to be more susceptible to philanthropic behavior and to CSR in general than their male counterparts (Hussain et al, 2018; Williams, 2003).
Board size The quality of the monitoring is reduced as the board size increases (De Andres et al. 2005).
Large boards increase the likelihood investments in CSR are made as large boards often include members that are oriented towards pro-social activities (Kock et al., 2012).
A large board size provides the corporation with more directors and could increase the social capital of the corporation (Pfeffer & Salancik, 1978). Additionally, a large board could increase the collective information and lead to increased performance (Dalton et al, 1999; Guest, 2009).
CEO duality
CEO duality blurs the line between management and control (Fama and Jensen, 1983) and compromises the board’s ability to monitor properly (Rechner and Dalton, 1991).
CEO duality could positively affect CSR as it could increase stakeholder representation (Finkelstein & Hambrick, 1990)
CEO duality represents high stakeholder orientation (Finkelstein & Hambrick, 1990) and high stakeholder orientation is positively associated with CSR (Arena et al., 2015; Mallin et al., 2013).
2 Both theories suggest that in order for corporations to gain legitimacy, corporations need to act in accordance with what is accepted as social responsible behavior by society and by doing so
10
2.2.1 Board independence and CSR
As stated before, agency problems can arise as a result of investments in CSR (Fama & Jensen, 1983). Board independence can help resolve these agency problems as the actions of an agent can be monitored more effectively by an independent board (Hussain et al, 2018). In addition, outsiders tend to have a broader view of the performance of a corporation (Wang & Coffey, 1992), whereas insiders tend to be more preoccupied with short-term economic gains. Therefore, independent boards can help reduce agency problems (Fama & Jensen, 1983, Jensen &
Meckling, 1976). Stakeholder theory also suggests a positive association between board
independence and CSR performance (Hussain et al, 2018) as outside directors are assumed to be less subjected to internal pressures. Outside directors can be seen as a way to ensure that
corporations not only act on behalf of its shareholders, but on behalf of all of its stakeholders (Haniffa & Cooke, 2005). Outside directors will therefore be more likely to act in accordance with what is accepted as social responsible behavior and help confer legitimacy to a corporation (Freeman & Reed, 1983).
Prior studies on the relationship between board independence and CSR disclosure show mixed results. On one hand there have been studies (Akhtaruddin et al. 2009; Cucari et al., 2017; Ionel-Alin et al., 2012; Lim & Chow, 2007; Lone et al, 2016; Rao et al., 2012; Uyar et al. 2013) reporting a positive association. On the other hand there have been studies reporting a negative association (Eng & Mak, 2003; Jangu et al, 2014, Majeed et al, 2015). Agency theory suggests that independent directors can improve the monitoring quality of the board (Hussain et al, 2018; Jensen & Meckling, 1976) and therefore increase the quality and quantity of the disclosure (Fama & Jensen, 1983). The existing fragmentation calls for further investigation and based upon the existing theories and the prior studies, I hypothesize that:
Hypothesis 1a Board independence is positively associated with CSR disclosure
The relationship between board independence and CSR performance shows a similar
fragmentation. On one hand there have been studies reporting a positive association (Dunn & Sainty, 2009; Fernández-Gago et al, 2016; Haque, 2017; Johnson & Greening, 1999). On the other hand there have been studies reporting a negative association (Ortiz‐de‐Mandojana, 2016; Surroca & Tribo, 2008). In addition, based upon the resource dependence theory, outside directors could be seen as a valuable source of resources (Hillman & Dalziel, 2003). In turn,
these resources could positively affect CSR performance (De Villiers et al. 2011). This
fragmentation calls for further investigation and based upon the existing theories, I hypothesize that:
Hypothesis 1b Board independence is positively associated with CSR Performance 2.2.2 Women on board and CSR
Prior studies argued that diversity in a board can improve the chance that different ideas and perspectives will be considered when making decisions (Liao et al, 2015; Post et al., 2011). As such, gender is often debated upon when discussing board composition characteristics (Liao et al, 2015). Additionally, it is argued that the presence of women on the board could positively influence CSR is due to the role women tend play in their professional environment (Betz et al., 1989; Frias-Aceituno et al., 2013) as women tend to play the role of wife and mother. These and other differences in backgrounds may cause women directors to be more susceptible to
philanthropic behavior and to CSR in general than their male counterparts (Hussain et al, 2018; Williams, 2003). As such, under the notion of resource dependence theory, the presence of women on board can be considered a resource (Bear et al, 2010).
Prior studies on the relationship between women on board and CSR disclosure show mixed results. On one hand there have been studies (Fodio et al, 2012; Isa & Muhammed, 2015; Rao et al., 2012) reporting a positive association. Additionally, both Amran et al. (2014) and Post et al (2011) argue that when there is a certain number of women on board, the amount and the quality of disclosure increases. Amran et al (2014) argues there have to be at least two or three women on board. Post et al (2011) argues, based upon critical mass theory, that there has to be at least three women on board. On the other hand Cucari et al. (2017) shows no positive association between women on board and CSR disclosure. The existing fragmentation calls for further investigation and based upon the existing theories and the prior studies, I hypothesize that:
Hypothesis 2a Women on board is positively associated with CSR disclosure
Based on prior research, the presence of women directors on the board has been argued to have a positive effect on the philanthropy of a corporation and thus in turn on CSR performance (Wang
12 and Coffey, 1992; Williams, 2003). Additionally, prior studies reported a positive association between women on board and CSR performance (Bear et al, 2010; Hafsi & Turgut, 2013). On the other hand, Galbreath (2016) reports the relationship to be inconclusive. This fragmentation calls for further investigation and based upon the above discussion, I hypothesize that:
Hypothesis 2b Women on board is positively associated with CSR Performance 2.2.3 Board size and CSR
Considering that each director can be considered a resource. A large board size could have a positive effect on CSR as a larger board would have more directors. As this would provide the corporation with more resources and therefore increase the social capital of the corporation (Pfeffer & Salancik, 1978). Additionally, a larger board could result in an increase in the capacity of a board to monitor (John & Senbet, 1998) and could increase the collective information and therefore lead to increased performance (Dalton et al, 1999; Guest, 2009). However, the potential benefits of this may be outweighed by the potential costs that are associated with larger groups (Jensen & Meckling, 1993) as from the perspective of agency theory the quality of the monitoring is reduced as the board size increases (De Andres et al. 2005). Additionally a smaller board size could improve efficiency (John & Senbet, 1998), but this smaller board size may come at the cost of a higher work load for the directors. Which, in turn, would reduce the ability of the board to monitor and control effectively.
Prior studies on the relationship between board size and CSR disclosure show mixed results. On one hand there have been studies reporting a positive association (Majeed et al, 2015; Jangu et al, 2014; Rao et al., 2012). On the other hand Prado-Lorenzo and Garcia-Sanchez (2010) reports a negative association and argues that the efficiency of CG suffers from a large board size. As such the relationship between board size and CSR disclosure is unclear. Following the arguments of Prado-Lorenzo and Garcia-Sanchez (2010) and De Andres et al. (2005) and taking the perspective of agency theory into consideration, I hypothesize the following relationship:
Prior studies report both a positive association between board size and CSR performance
(Godos-Díez et al, 2016; Oh et al, 2015) as well as a negative association (Kassinis et al., 2002). Following the arguments of Prado-Lorenzo and Garcia-Sanchez (2010) and De Andres et al. (2005) and taking that perspective of agency theory into consideration, I hypothesize the following relationship:
Hypothesis 3b Board size is negatively associated with CSR Performance 2.2.4 CEO duality and CSR
CEO duality refers to the situation when a person is both the CEO of the corporation as well as chairman of the board. This situation blurs the line between management and control (Fama and Jensen, 1983) and compromises the board’s ability to monitor properly (Rechner and Dalton, 1991) as this situation compromises the carefully designed system of checks and balances. Michelon and Parbonetti (2012) argue that CEO duality reduces the transparency and
accountability of a corporation. In addition, CEO duality can lead to agency problems as CEO duality results in the concentration of managerial power and reduced board independence (Boyd, 1994).
Prior studies reveal that the relationship between CEO duality and CSR disclosure show mixed results. On one hand there have been studies reporting no association between CEO duality and CSR disclosure (Liao et al., 2015; Michelon & Parbonetti, 2012). On the other hand there have been studies reporting a negative association (Allegrini & Greco, 2013; Gul & Leung, 2004). Based upon the results of prior studies as well as the perspective taken from the agency theory suggest, I hypothesize the following relationship:
Hypothesis 4a CEO duality is negatively associated with CSR disclosure
Mallin and Michelon (2011) report a negative association between CEO duality and CSR performance and suggests that the ability of a corporation to pursue and implement CSR is reduced due to CEO duality. As such I hypothesize that:
14
3
Methodology
In order to better understand the relationship between the CG and CSR, I conducted a meta-analysis of existing findings on this relationship. I studied the CG mechanisms: board
independence, women on board, board size and CEO duality in relationship with the two CSR outcomes: CSR performance and CSR disclosure.
3.1 Data collection and sample
A thorough search was conducted of existing literature in order to identify the primary studies. This search was conducted during September 2016 – February 2017 using varying electronic and online databases like Econlit, EBSCOhost, Google Scholar and Web of Science. The search keywords that were used were terms related to CG and CSR. In order for a thorough and
comprehensive search of literature both individual as well as combinations of the keywords were used. Some examples of the used search keywords are ‘board composition’, ‘board diversity’, ‘board independence’, ‘board size’, ‘CEO duality’, ‘corporate governance’, ‘corporate social responsibility’, 'cross directorship', ‘CSR committee’, 'CSR performance', 'director age', 'director expertise', 'director ownership', ‘environmental reporting’, 'foreign director', ‘nonfinancial disclosure’, ‘sustainability reporting’, ‘voluntary’, etc. The initial search resulted in 126
empirical studies being selected as potentially usable. After checking the studies, it was decided to only use studies with at least one of the following variables: board independence, women on board, board size or CEO duality. As these variables were the most often present in the studies and thus would allow for a better comparison. Out of these studies 27 didn’t include one of these variables and as such they were dropped for not fulfilling that criteria.
In order to conduct the meta-analysis I used the Hedges and Olkin (1985) meta-analysis (HOMA), which will be further detailed under the meta-analytic strategy. To be able to conduct this meta-analysis the partial correlation coefficient needed to be calculated, which requires a t-statistic to be either present in the study or to be able to calculate the t-t-statistic. 35 more studies, including Amran et al (2014), De Villiers et al (2011), Mall et al (2013) and Wang and Coffey (1992), either didn’t include a t-statistic or the t-statistic could not be calculated and as such they were dropped from the sample. This left me with a sample of 64 studies. One of the selected studies, Boesso and Kumar (2007), measured Italy and US separately and included a separate
pearson correlation coefficient and t-statistic for each. As such they were both used. The final sample is described in table 1.
Measures were taken in order to follow the assumption of independence, which underlies the validity of a meta-analysis. In some cases, the studies of interest used more than one measure of the variable disclosure or performance or used more than one proxy. For instance, Bear et al (2011), uses two dimensions of CSR performance, ‘institutional strength’ and ‘technical strength’. In these cases, the weighted average correlation was calculated in order to obtain the desirable one outcome of a relation per study.
Review of the sample revealed that both CSR disclosure and CSR performance are measured in two different ways: unstandardized score and standardized score. CSR disclosure was considered as a standardized score when it was based on Bloomberg, CCER questionnaire, CDP questionnaire, GRI, integrated reporting, TRI, etc. On the other hand CSR disclosure in cases where the author constructed used a self-constructed disclosure index, CSR disclosure was considered as a unstandardized score. Likewise, CSR performance was considered a
standardized score when it was based on ESG or KLD from ASSET4, while all other CSR performance measures were considered unstandardized.
As stated before the CG mechanisms of interest for this study are board independence, women on board, board size and CEO duality. As such we extracted from our dataset, these predictors of CSR disclosure and CSR performance. A description of the total number of predictors for both CSR disclosure and CSR performance is given in table 3.
Table 3 Descriptive statistics
CSR disclosure CSR performance Total
Board independence 29 19 48
Women on board 14 12 26
Board size 23 18 41
CEO duality 10 11 21
Total 76 60 136
As can be seen, in most cases I have more predictors for CSR disclosure than CSR performance. In addition, we have at least 10 predictors for each CG mechanism for each CSR outcome. This increases the validity of our analysis.
16
3.2 Meta-analysis procedure
The general procedure to conduct a meta-analysis is to gather and then combine the effect size estimates from the various studies (Field, 2005). Effect size estimates provide information about the strength (significance) and the direction (association) of the relationship between two
variables. In this study I used both the pearson correlation coefficient r and the partial correlation coefficient rxy,z for effect size estimates. While the pearson correlation coefficient could be
gathered from the studies, the partial correlation coefficient had to be calculated using the following formula (Greene,2008) (1):
( /( + df)) (1)
In this formula t stands for the t-statistic and df stands for the degrees of freedom. In some cases the t-statistic wasn’t listed and as such had to be calculated by dividing the coefficient by the standard error. Given that this formula doesn’t produce negative numbers, the numbers had to be converted from positive to negative if the regression coefficient was negative.
I used the correlation coefficients from the sample and the Hedges and Olkin (1985) meta-analysis procedure, otherwise known as HOMA, in order to transform the effect size estimates into Fisher’s z score. The Fisher’s z score was calculated using the following formula (2):
Z = [ln(1 + r) − ln(1 − r) = (2) In this formula r stands for either the pearson or the partial correlation coefficient. The Fisher’s z score is then used to calculate weighted average correlation coefficient using the following formula (3):
z ∑ W∑ W"!# !Z !
! "
!# (3)
In this formula Wi stands for the weight of each study and K stands for the number of studies. The weighted average correlation coefficient, %̅ , and SE(%̅ ), the standard deviation of the weighted average correlation coefficient, are then used to calculate the confidence interval at 95%, using the following formula (4):
In order to convert the weighted average correlation coefficient bac k to a correlation, the following formula was used (5):.
r̅ =ee 00− 1+ 1 =ee00− e+ e 00 (5)
Additionally both the Q-statistic as well as the I-squared were calculated. The Q-statistic was calculated to test the heterogeneity of the outcomes. A significant result of a Q-test indicates that the effect size varies across studies. The Q-statistic is calculated using the following formula (6):
Q = 3 W! 4 !#
(Z!− z !) (6)
Some authors argue that the size of the sample can influence the robustness of a Q-test (Higgins et al, 2003). For this reason I also calculated the I-squared as the I-squared is the percentage of variation across studies due to the true differences in effect sizes rather than chance. I-square was calculated using the following formula (7):
I =Q − (K − 1)Q (7)
For the statistical model I used the random effects model instead of the fixed effects model (Hedges and Olkin, 1985) as the the random effects model goes under the assumption that the studies in the sample are not similar and as such the factors that could moderate the relationship between variables are considered. Therefore, the random effects model can differentiate in sub-groups. On the other hand, the fixed effects model goes under the assumption that the observed variability can be attributed to the sampling error and that otherwise the studies in the sample are studying the same effect size. I expect that the associations between the variables are not the same and as such I used the random effects model.
18
4
Results
4.1. Effect size analysis
Tables 4 and 5 show the results of the meta-analysis and the correlations between the CG mechanisms: board independence, women on board, board size and CEO duality and the CSR outcomes: CSR disclosure and CSR performance. I used the random effects model to perform the meta-analysis.
Table 4 reveals that board independence is positively associated with CSR disclosure and that this relationship is very significant (p<0,01) for both the rxy,z as well as the r results. As such
hypothesis 1a is supported. On the other hand table 5 reveals that the association between board independence and CSR performance is not significant (p>0,05) for the rxy,z result. Therefore,
even though the r result shows a positive and significant relationship, hypothesis 1b isn’t supported.
Table 4 and table 5 show that women on board is positively associated with both CSR disclosure as well as CSR performance. The relationship between women on board and CSR disclosure is significant (p<0,05) for the rxy,z result and very significant for the r result. As such
hypothesis 2a is supported. On the other hand the relationship between women on board and CSR performance is very significant (P<0,01) for both the rxy,z as well as the r results. As such
the relationship between women on board and CSR performance is stronger than the relationship between women on board and CSR disclosure. It also means that hypothesis 2b is supported.
Tables 4 and 5 reveal that board size is positively associated with both CSR disclosure as well as CSR performance and that these relationships are very significant (p<0,01) for both the
rxy,z as well as the r results. As such hypothesis 3a and 3b are all rejected as board size is
positively associated with CSR disclosure and CSR performance and not negatively associated as hypothesized.
The results between CEO duality and CSR disclosure and CSR performance in tables 4 and 5 reveal that these relationships are not significant (p>0,05) for both the rxy,z as well as the r
Table 4 HOMA results: CG mechanisms and CSR disclosure
Partial correlation (rxy,z) Pearson correlation (r)
Variables K N ř Z-value CI 95% I2 Q-stats
ř Z-value CI 95% I2 Q-stats Board independence 29 11155 0.056 2.696** 0.015/0.096 69.535 91.908*** 0.112 3.166** 0.043/0.180 92.764 386.98*** Women on board 14 8391 0.102 2.449* 0.020/0.181 90.375 135.062*** 0.231 4.681*** 0.136/0.322 92.813 180.873*** Board size 23 9496 0.119 6.375*** 0.083/0.155 53.585 47.398** 0.221 5.682*** 0.146/0.293 87.384 174.384*** CEO duality 10 2597 -0.037 -1.152n.s. -0.101/0.026 57.837 21.346* 0.009 0.156n.s. -0.106/0.125 90.792 97.74***
Table 5 HOMA results: CG mechanisms and CSR performance
Partial correlation (rxy,z) Pearson correlation (r)
Variables K N ř Z-value CI 95% I2 Q-stats
ř Z-value CI 95% I2 Q-stats Board independence 19 31398 0.028 1.908n.s. -0.001/0.056 78.710 84.546*** 0.135 4.451*** 0.076/0.193 95.760 424.515*** Women on board 12 29047 0.058 4.425*** 0.032/0.084 70.428 37.197*** 0.079 2.892** 0.026/0.132 93.728 175.394*** Board size 18 31651 0.046 2.679** 0.012/0.080 86.933 130.102*** 0.187 3.293*** 0.076/0.292 98.939 1602.869*** CEO duality 11 15367 0.022 1.814n.s. -0.002/0.046 43.552 17.715n.s. 0.114 2.645** 0.030/0.197 95.842 240.509*** K is the number of sample studies. N is the number of observations. ř is overall effect size. C.I. is confidence interval. I2 = I-squared. p< 0.001 = ***, p<0.01 = **, p<0.05 = *, n.s. insignificant.
The findings suggest that out of all the CG mechanisms board size has the strongest relationship and CEO duality the weakest relationship with both CSR outcomes. Cohen (1992) suggests that an effect size is considered to be large when it is 0,50. An effect size is considered to be medium when it is 0,30 and it is considered to be small when it is 0,10. Given that none of the effect sizes is larger than 0,30, none of the CG mechanisms appear to be closely associated with CSR.
4.2. Sub-group analysis
The I-square can be used to test for heterogeneity (Higgins et al, 2003). This means that I-square can be used to measure inconsistencies across studies in the meta-analysis Given that the I-square is greater than 75% in most cases, this means that there is a significant amount of heterogeneity. In order to test for the factors that could cause this fragmentation, I conducted a sub-group analysis. For this sub-group analysis I selected three potential covariates: the coverage years (COVyear), the geographical location of the countries (GEO) and the measurement of the outcome variables (CSRmeas). In order to check for differences in results, the outcome variables, CSR disclosure and CSR performance, were categorized in standardized and unstandardized. For this I used a dummy variable referred to as CSRmeas. This variable equals 0 for unstandardized measurements and 1 for standardized measurements. COVyear is a dummy variable that equals 1 if the study uses panel data and 0 otherwise. GEO is divided into 5 different categories. This variable equals 1 for African and Gulf countries, 2 for Asian countries, 3 for international, 4 for Europe and the UK and 5 for Canada and the US. Tables 6 and 7 show the results of the sub-group analysis.
In some cases the results of the sub-group analysis should be interpreted with caution as the number of studies in those cases is lower than 3. The sub-group analysis reveals that, for both the rxy,z as well as the r results, the relationship between the CG mechanisms: board
independence and women on board and CSR is stronger in the studies using unstandardized measurements. In the case of board independence, the relationship with CSR disclosure is stronger in the studies using standardized measurement.
Table 6 Sub-group analysis of CG mechanisms on CSR disclosure
Partial correlation (rxy,z) Pearson correlation (r)
Variables
sub-groups level K N ř Z-value CI 95% I
2 Q-stats ř Z-value CI 95% I2 Q-stats
BOARD.IND 29 11155 0.056 2.696** 0.015/0.096 69.535 91.908*** 0.112 3.166** 0.043/0.180 92.764 386.980*** CSRmeas 0 21 5654 0.035 1.321n.s. -0.017/0.087 69.729 66.071*** 0.077 1.539n.s. -0.021/0.173 90.861 218.837*** 1 8 5501 0.095 2.586** 0.023/0.165 70.716 23.903** 0.190 2.515* 0.042/0.330 95.736 164.161*** GEO 1 2 215 -0.051 -0.544n.s. -0.230/0.132 43.909 1.783n.s. -0.120 -0.791n.s. -0.396/0.176 75.627 4.103* 2 12 3706 0.030 0.949n.s. -0.032/0.092 79.613 53.956*** 0.082 1.404n.s. -0.032/0.194 94.193 189.428*** 3 8 5261 0.041 0.997n.s. -0.040/0.122 40.774 11.819n.s. 0.090 1.207n.s. -0.056/0.233 86.305 51.114*** 4 6 1432 0.152 3.479*** 0.067/0.234 0.000 4.470n.s. 0.230 2.884** 0.075/0.374 79.138 23.967*** 5 1 541 0.015 0.162n.s. -0.165/0.194 0.000 0.000n.s. 0.250 1.344n.s. -0.116/0.557 0.000 0.000n.s. COVyear 0 19 3014 0.045 1.548n.s. -0.012/0.101 61.457 46.701*** 0.086 1.705n.s. -0.013/0.184 77.979 81.738*** 1 10 8141 0.069 2.200* 0.008/0.129 80.047 45.106*** 0.150 2.352* 0.025/0.271 96.920 292.162*** WoB 14 8391 0.102 2.449* 0.020/0.181 90.375 135.062*** 0.231 4.681*** 0.136/0.322 92.813 180.873*** CSRmeas 0 9 3937 0.184 3.153** 0.070/0.293 89.217 74.188*** 0.324 4.988*** 0.201/0.436 92.368 104.824*** 1 5 4454 -0.005 -0.072n.s. -0.144/0.134 92.340 52.218*** 0.083 1.000n.s. -0.080/0.241 93.886 65.422*** GEO 1 2 108 0.527 4.280*** 0.307/0.693 0.000 0.052n.s. 0.384 2.641** 0.104/0.607 76.819 4.314* 2 5 2158 0.075 1.087n.s. -0.060/0.208 87.642 32.367*** 0.372 4.719*** 0.224/0.503 93.901 65.58*** 3 3 4944 0.008 0.097n.s. -0.148/0.163 88.407 17.252*** 0.097 0.997n.s. -0.094/0.281 96.072 50.921*** 4 3 640 0.014 0.157n.s. -0.158/0.185 94.010 33.388*** 0.066 0.633n.s. -0.138/0.266 0.000 1.781n.s. 5 1 541 0.196 1.420n.s. -0.075/0.44 0.000 0.000n.s. 0.230 1.375n.s. -0.099/0.514 0.000 0.000n.s. COVyear 0 4 738 0.201 2.428* 0.039/0.352 87.764 24.518*** 0.018 0.208n.s. -0.147/0.181 73.537 11.337* 1 10 7653 0.069 1.421n.s. -0.026/0.162 91.367 104.255*** 0.298 6.309*** 0.209/0.383 92.464 119.423*** Board size 23 9496 0.119 6.375*** 0.083/0.155 53.585 47.398** 0.221 5.682*** 0.146/0.293 87.384 174.384*** CSRmeas 0 18 5142 0.143 6.293*** 0.099/0.186 31.127 24.683n.s. 0.220 5.476*** 0.142/0.294 87.039 131.166*** 1 5 4354 0.069 2.062* 0.003/0.134 75.901 16.598** 0.221 3.200** 0.087/0.347 85.115 26.873*** GEO 1 3 275 0.266 3.935*** 0.136/0.386 66.640 5.995* 0.190 1.836n.s. -0.013/0.378 79.370 9.695** 2 8 2504 0.140 4.628*** 0.081/0.198 22.856 9.074n.s. 0.235 4.150*** 0.126/0.339 91.729 84.634*** 3 5 5138 0.052 1.641n.s. -0.010/0.113 75.071 16.045** 0.186 2.713** 0.052/0.313 0.000 3.210n.s. 4 7 1579 0.129 4.025*** 0.067/0.191 0.000 2.274n.s. 0.237 4.056*** 0.124/0.344 83.715 36.843*** COVyear 0 13 1829 0.090 3.053** 0.032/0.147 55.372 26.889** 0.268 5.875*** 0.181/0.351 42.147 20.742n.s. 1 10 7667 0.140 5.683*** 0.092/0.187 52.434 18.921* 0.167 3.487*** 0.074/0.257 93.731 143.571***
Table 6 continued
Partial correlation (rxy,z) Pearson correlation (r)
Variables
sub-groups level K N ř Z-value CI 95% I
2 Q-stats ř Z-value CI 95% I2 Q-stats
CEO duality 10 2597 -0.037 -1.152 n.s. -0.101/0.026 57.837 21.346* 0.009 0.156n.s. -0.106/0.125 90.792 97.74*** CSRmeas 0 6 971 -0.084 -1.853n.s. -0.172/0.005 9.231 5.509n.s. -0.068 -0.775n.s. -0.237/0.104 71.157 17.335** 1 4 1626 0.005 0.119n.s. -0.080/0.09 74.559 11.792** 0.109 1.085n.s. -0.088/0.298 95.244 63.075*** GEO 1 2 215 -0.009 -0.110n.s. -0.169/0.152 0.000 0.488n.s. 0.131 0.793n.s. -0.192/0.428 0.000 0.239n.s. 2 2 944 -0.116 -2.214* -0.216/-0.013 0.000 0.210n.s. -0.120 -0.805n.s. -0.393/0.172 0.000 0.392n.s. 3 3 477 0.042 0.705n.s. -0.074/0.156 69.812 6.625* -0.055 -0.422n.s. -0.301/0.198 48.088 3.853n.s. 4 3 961 -0.033 -0.693n.s. -0.125/0.06 39.828 3.324n.s. 0.086 0.694n.s. -0.156/0.319 96.719 60.962*** COVyear 0 8 1980 -0.049 -1.250n.s. -0.125/0.028 65.312 20.180** -0.015 -0.187n.s. -0.175/0.145 92.486 93.159*** 1 2 617 0.001 0.012n.s. -0.141/0.143 0.000 0.179n.s. 0.099 0.618n.s. -0.212/0.392 49.309 1.973n.s.
K is the number of sample studies. N is the number of observations. ř is overall effect size. C.I. is confidence interval. I2 = I-squared.
BOARD.IND = board independence. WoB= women on board. CSRmeas is the measurement of CSR disclosure. GEO is the country location. COVyear is the coverage year. p< 0.001 = ***, p<0.01 = **, p<0.05 = *, n.s. insignificant.
Table 7 Sub-group analysis of CG mechanisms on CSR performance
Partial correlation (rxy,z) Pearson correlation (r)
Variables
sub-groups level K N ř Z-value CI 95% I
2 Q-stats ř Z-value CI 95% I2 Q-stats
BOARD.IND 19 31398 0.028 1.908n.s. -0.001/0.056 78.710 84.546*** 0.135 4.451*** 0.076/0.193 95.760 424.515*** CSRmeas 0 10 16299 0.034 1.609n.s. -0.007/0.075 78.033 40.97*** 0.126 3.27** 0.051/0.2 95.665 207.589*** 1 9 15099 0.020 0.886n.s. -0.025/0.065 81.218 42.594*** 0.144 3.556*** 0.065/0.222 91.660 95.919*** GEO 2 4 5171 0.038 1.358n.s. -0.0170/0.093 50.509 6.062n.s. 0.193 3.085** 0.071/0.308 97.499 119.942*** 3 5 9258 -0.020 -0.708n.s. -0.0740/0.035 81.037 21.094*** 0.026 0.435n.s. -0.09/0.14 93.083 57.829*** 4 2 2460 0.181 4.054*** 0.0940/0.265 90.666 10.714** 0.331 3.614*** 0.156/0.486 0.000 0.442n.s. 5 8 14509 0.015 0.66n.s. -0.0290/0.058 58.212 16.751* 0.121 2.489* 0.026/0.214 93.524 108.086*** COVyear 0 5 1378 -0.007 -0.186n.s. -0.079/0.066 75.821 16.543** 0.082 1.283n.s. -0.043/0.204 89.638 38.603*** 1 14 30020 0.034 2.142* 0.003/0.065 80.577 66.931*** 0.151 4.33*** 0.083/0.218 96.622 384.831*** WoB 12 29047 0.058 4.425*** 0.032/0.084 70.428 37.197*** 0.079 2.892** 0.026/0.132 93.728 175.394*** CSRmeas 0 6 11567 0.074 3.696*** 0.0350/0.113 69.190 16.228** 0.141 3.778*** 0.068/0.213 94.831 96.731*** 1 6 17480 0.046 2.367* 0.0080/0.085 73.701 19.012** 0.013 0.326n.s. -0.063/0.088 87.527 40.087*** GEO 2 2 3325 0.056 1.518n.s. -0.016/0.128 0.000 0.599n.s. 0.059 0.876n.s. -0.072/0.188 76.539 4.262* 3 4 8108 0.088 2.772** 0.026/0.15 79.931 14.948** 0.185 3.357*** 0.078/0.288 95.843 72.162*** 4 1 2315 0.041 0.827n.s. -0.056/0.137 0.000 0n.s. -0.020 -0.215n.s. -0.2/0.161 0.000 0n.s. 5 5 15299 0.061 2.386* 0.011/0.111 80.220 20.222*** 0.045 0.989n.s. -0.044/0.134 90.948 44.189*** COVyear 0 3 240 0.302 4.661*** 0.179/0.417 0.000 1.066n.s. 0.148 1.811n.s. -0.012/0.3 71.082 6.916* 1 9 28807 0.046 4.629*** 0.026/0.065 59.122 19.57* 0.071 2.425* 0.014/0.127 95.204 166.812*** Board size 18 31651 0.046 2.679** 0.012/0.080 86.933 130.102*** 0.187 3.293*** 0.076/0.292 98.939 1602.869*** CSRmeas 0 11 15462 0.053 2.245* 0.007/0.099 91.354 115.665*** 0.128 2.152* 0.011/0.241 97.998 499.544*** 1 7 16189 0.038 1.297n.s. -0.019/0.095 54.520 13.193* 0.274 3.741*** 0.133/0.404 98.533 408.908*** GEO 2 8 8458 0.033 1.045n.s. -0.028/0.093 47.312 13.286n.s. 0.127 2.085* 0.008/0.242 94.568 128.867*** 3 2 7963 -0.018 -0.309n.s. -0.129/0.094 98.388 62.032*** -0.129 -1.083n.s. -0.348/0.104 99.432 176.056*** 4 2 2713 0.059 0.956n.s. -0.062/0.178 84.629 6.506* 0.416 3.639*** 0.202/0.593 88.628 8.793** 5 6 12517 0.093 2.546* 0.022/0.164 88.823 44.736*** 0.284 4.093*** 0.151/0.407 97.565 205.326*** COVyear 0 4 1404 0.078 1.762n.s. -0.009/0.163 43.189 5.281n.s. 0.253 2.052* 0.012/0.466 84.665 19.563*** 1 14 30247 0.041 2.177* 0.004/0.077 89.116 119.444*** 0.169 2.611** 0.042/0.289 99.175 1576.613***
Table 7 continued
Partial correlation (rxy,z) Pearson correlation (r)
Variables
sub-groups level K N ř Z-value CI 95% I
2 Q-stats ř Z-value CI 95% I2 Q-stats
CEO duality 11 15367 0.022 1.814n.s. -0.002/0.046 43.552 17.715n.s. 0.114 2.645** 0.030/0.197 95.842 240.509*** CSRmeas 0 4 5346 0.024 1.111n.s. -0.018/0.065 77.170 13.14** 0.061 0.809n.s. -0.087/0.206 97.331 112.419*** 1 7 10021 0.021 1.258n.s. -0.012/0.055 0.000 4.421n.s. 0.148 2.474* 0.031/0.261 95.209 125.241*** GEO 2 1 2412 -0.018 -0.5760n.s. -0.079/0.043 0.000 0n.s. 0.290 3.038** 0.106/0.455 0.000 0n.s. 3 3 709 0.032 0.7830n.s. -0.048/0.113 39.482 3.305n.s. -0.032 -0.446n.s. -0.172/0.109 83.730 12.293** 4 2 2713 0.004 0.1410n.s. -0.05/0.058 0.000 0.098n.s. 0.179 2.488* 0.038/0.313 97.980 49.513*** 5 5 9533 0.037 2.34* 0.006/0.068 49.847 7.976n.s. 0.110 2.377* 0.019/0.199 76.715 17.178** COVyear 0 3 356 0.039 0.6920n.s. -0.071/0.148 44.979 3.635n.s. 0.074 0.769n.s. -0.115/0.259 85.450 13.746** 1 8 15011 0.021 1.6420n.s. -0.004/0.047 49.969 13.991n.s. 0.124 2.602** 0.031/0.215 96.772 216.819*** K is the number of sample studies. N is the number of observations. ř is overall effect size. C.I. is confidence interval. I2 = I-squared.
BOARD.IND = board independence. WoB= women on board. CSRmeas is the measurement of CSR disclosure. GEO is the country location. COVyear is the coverage year. p< 0.001 = ***, p<0.01 = **, p<0.05 = *, n.s. insignificant.
26 Additionally the results reveal that the relationship between board independence and CSR is very significant for studies which collected data from European or UK corporations, while this
relationship is insignificant for studies which collected data from other locations. In the case of board size, the results also reveal that the relationship with CSR disclosure is stronger for studies which collected data from European and UK corporations. This suggests that the relationship between CG and CSR disclosure is stronger in European countries and in the UK.
5
Discussion and conclusion
The primary objective of this paper is to analyze the relationship between CG and CSR by systematically analyzing existing research on the relationship between the CG mechanisms: board independence, women on board, board size and CEO duality and the CSR outcomes: CSR disclosure and CSR performance.
The primary results indicate that all the CG mechanisms except CEO duality influence the CSR outcomes of a corporation significantly. Interestingly the relationship between women on board is more significant for CSR performance than it is for CSR disclosure. Which suggests that female directors care more about the social responsible actions of a corporation, rather than about disclosing information about these actions. The results of this study also provide further support to the findings of Harjoto et al (2015) and Haque (2017) by showing that women on board significantly affects CSR performance. Additionally board size is positively and significantly related with both CSR disclosure as well as CSR performance. This could be explained through resource dependence theory as a large board size increase the social capital of the corporation (Pfeffer & Salancik, 1978) and could lead to increased performance (Dalton et al, 1999; Guest, 2009). Additionally these results further supports the findings of Majeed et al (2015) by showing board size is positively and significantly related with CSR disclosure. Board independence is significantly related with CSR disclosure, but the relationship with CSR performance is not significant. This could suggest that an independent board is more concerned with gaining legitimacy, which could be done by means of CSR disclosure (Cho & Patten, 2007).
However given the small effect sizes, none of the CG mechanisms appear to be closely associated with CSR. As such the question remains whether there truly is a link between CG and
CSR (Walls et al, 2012) and the results therefore require further theorization (Hussain et al, 2018). Despite this, the meta-analysis helps us better understand the relationship between CG and CSR as the analysis identifies both the directions as well as the strengths of the relationships between the CG mechanisms and CSR outcomes. The study reveals both strong as well as weak relationships between CG and CSR.
While prior studies (Byron & Post, 2016; García-Meca & Sánchez-Ballesta, 2010; Jain & Jamali, 2016) also researched the relationship between CG and CSR. The scope of these studies was limited as these studies considered fewer CG mechanisms and didn’t take both CSR
disclosure as well as CSR performance at the same time into consideration. This study
contributes to the prior studies by performing a separate analysis for both CSR disclosure as well as CSR performance. The analysis shows the relationship between the CG mechanisms and the CSR outcomes. However due to the fact that not all studies provide both a partial and a pearson correlation, the overall sample size is only 64 studies.
Prior findings (Aquilera et al, 2006) suggested that geographical settings could influence the relationship between CSG and CSR. The results of the meta-analysis support this as the sub-group analysis reveals that the relationship between board independence and CSR is very significant for studies which collected data from European or UK corporations, while being insignificant otherwise. This variation in the results can be explained through institutional theory (Jain & Jamali, 2016). According to institutional theory corporations are embedded in society and as such are subject to the values, norms and beliefs that society imposes upon the
corporation (Ioannou & Serafeim, 2012). Different countries have different societies and therefore different values, norms and beliefs. These differences result in different rules
corporations need to abide by and therefore influence actions of the corporation differently. In addition to providing an academic contribution, this study will enable corporations to better understand the relationship between CG and CSR. Which will enable corporations to utilize CG mechanisms more effectively in improving either their CSR disclosure or their CSR performance.
28
5.1 Limitations and Future Research Directions
The results of prior research in CG and CSR literature has been quantitatively summarized in this study. This study has several limitations. First, environmental and social issues are handled differently by corporations and there is more than one CSR dimension (Bansal & Gao, 2008). A limited number of studies prevented me from analyzing the different CSR dimensions separately with CG mechanisms.
Secondly, this study only focusses on board level CG mechanisms and as such only covers one aspect of CG. In order to fully understand the relationship between CG and CSR all the CG aspects should be incorporated in future studies.
Thirdly, the group analysis may not be accurate in all cases. As currently some sub-groups have a low number of sample studies and therefore a low number of observations. A higher number of studies and observations could lead to different findings and as such would therefore be desirable.
Fourth, a meta-analysis can only be used to analyze the relationship between two variables and as such cannot provide deeper insight is able to analyze direct relationships between two variables and cannot offer deeper insights of different interventions or mediations in a model. Combining meta-analysis and structural equation modeling can offer the comparison between different theoretical paradigms by testing more than one relation simultaneously (Bergh et al., 2016).
Thirdly, while the study shows a positive and significant relationship between board size and CSR disclosure and CSR performance, the question remains whether this is due to a larger board size or due to board composition as independent directors and female directors are also positively linked with CSR.
References
References with * are included in the meta-analysis
Adams, C., & McNicholas, P. (2007). Making a difference: Sustainability reporting, accountability and organisational change. Accounting, Auditing & Accountability
Journal, 20(3), 382-402.
Abdullah, S. N., Mohamad, N. R., & Mokhtar, M. Z. (2011). Board independence, ownership and CSR of Malaysian Large Firms. Corporate Ownership & Control, 8(3), 417-431.* Akhtaruddin, M., Hossain, M. A., Hossain, M., & Yao, L. (2009). Corporate governance and
voluntary disclosure in corporate annual reports of Malaysian listed firms. Journal of
Applied Management Accounting Research, 7(1), 1-19.*
Allegrini, M., & Greco, G. (2013). Corporate boards, audit committees and voluntary disclosure: evidence from Italian Listed Companies. Journal of Management & Governance, 17(1), 187-216.*
Aguilera, R. V., Williams, C. A., Conley, J. M., & Rupp, D. E. (2006). Corporate governance and social responsibility: A comparative analysis of the UK and the US. Corporate
Governance: An International Review, 14(3), 147-158.
Aras, G., & Crowther, D. (2008). Governance and sustainability: An investigation into the relationship between corporate governance and corporate sustainability. Management
Decision, 46(3), 433-448.
Arvidsson, S. (2010). Communication of corporate social responsibility: A study of the views of management teams in large companies. Journal of Business Ethics, 96(3), 339-354. Bansal, P. (2005). Evolving sustainably: a longitudinal study of corporate sustainable
development. Strategic Management Journal, 26: 197-218.
Barako, D.G., Hancock, P. and Izan, H.Y. (2006). Factors Influencing Voluntary Corporate Disclosure by Kenyan Companies. Corporate Governance: An International Review, 14: 107-125.
Bear, S., Rahman, N., & Post, C. (2010). The impact of board diversity and gender composition on corporate social responsibility and firm reputation. Journal of Business Ethics, 97(2), 207-221.*
Ben-Amar, W., Chang, M., & McIlkenny, P. (2015). Board gender diversity and corporate response to sustainability initiatives: evidence from the Carbon Disclosure Project.
Journal of Business Ethics, 1-15.*
Benomran, N. A., Haat, M. H. C., Hashim, H. B., & Mohamad, N. R. B. (2015). Influence of Corporate Governance on the Extent of Corporate Social Responsibility and
Environmental Reporting. Journal of Environment and Ecology, 6(1), 48-68.* Berrone, P., & Gomez-Mejia, L. R. (2009). Environmental performance and executive
compensation: An integrated agency-institutional perspective. Academy of Management
Journal, 52(1), 103-126.*
Betz M, O’Connel L, Shepard JM. 1989. Gender differences in proclivity for unethical behavior.
Journal of Business Ethics 8: 321–324.
B., & P. (2016). Women on boards of directors and corporate social performance: A meta- Analysis. Corporate Governance: An International Review, 24(4), 428-442. Boesso, G. and Kumar, K. (2007). Drivers of corporate voluntary disclosure. Accounting,
