The impact of gender diversity, board independence,
audit committee independence and CSR committee
existence on CSR decoupling
Master thesis, MSc Accountancy
University of Groningen, Faculty of Economics and Business
June 24, 2019 Thierry Piets S2545462 Grote Beerstraat 206 M7 9742 SJ Groningen tel.: +31615393663 e-mail: t.r.v.piets@student.rug.nl Supervisor: Dr. N. Hussain Co-assessor: Dr. S. Girdhar Word count: 10,575
Abstract
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 of the board, board independence, audit committee independence and CSR committee existence) on the phenomenon of CSR decoupling. CSR decoupling is damaging to stakeholders as it increases information asymmetry and can be deceiving in the case of overstating CSR performance. This study fills a gap in the literature on determinants of CSR decoupling. Multiple linear regression was performed while controlling for firm, industry, country and year level differences. The study finds that a majority of firms tend to understate their CSR efforts in their disclosure. Furthermore, gender diversity, board independence and CSR committee existence were found to be positively related to CSR decoupling, contrary to the author’s expectations. However, regarding environmental CSR and in cases of low CSR performance or CSR overstating, the appointment of independent board members can help lower the level of CSR decoupling. Similarly, CSR committees and audit committee independence are found to lower CSR decoupling for firms with high CSR performance. Audit committee independence is also found to have a small negative effect on environmental CSR decoupling. These findings can be helpful for firms, regulators and stakeholders interested in lower levels of information asymmetry on CSR matters.
Table of contents
1. Introduction ... 2
2. Theoretical framework ... 4
2.1 Literature review: CSR decoupling and effective corporate governance ... 4
2.2 Hypothesis 1: gender diversity in boards of directors ... 7
2.3 Hypothesis 2: board independence ... 8
2.4 Hypothesis 3: audit committee independence ... 10
2.5 Hypothesis 4: CSR committee existence ... 11
3. Research methodology ... 12 3.1 Data collection ... 12 3.2 Variables ... 13 3.2.1 CSR decoupling... 13 3.2.2 Independent variables ... 13 3.2.3 Control variables ... 14 3.3 Data analysis ... 16 4. Results ... 16 4.1 Descriptive statistics ... 16 4.2 Correlation analysis ... 17 4.3 Regression analysis ... 18 4.4 Additional tests ... 20
4.4.1 Over- and understating CSR ... 20
4.4.2 CSR disclosure and CSR performance scores ... 21
4.4.3 High CSR performance and low CSR performance ... 22
4.4.4 Environmental, social and governance decoupling ... 22
5. Discussion and conclusion ... 24
5.1 Findings ... 24
5.2 Theoretical and managerial implications ... 28
5.3 Limitations and future research opportunities ... 30
6. References... 32
1. Introduction
Firms are increasingly reporting on their corporate social responsibility (CSR) efforts in recent decades. An interesting phenomenon occurs however, as research indicates that there is often a distinction between what companies say they do and what they actually do (Graafland & Smid, 2016; Boiral, 2013; Cho, Laine, Roberts, & Rodrigue, 2015). This phenomenon can be referred to as CSR decoupling. CSR decoupling is problematic because it means transparency on organizations’ operations regarding CSR is decreased. Historically, the literature has focused on CSR decoupling as a strategic decision to deceive stakeholders (Crilly, Zollo, & Hansen, 2012). This type of CSR decoupling is problematic because stakeholders demand reliable information, while information of symbolic nature cannot be fully relied upon and therefore has a decreased usefulness.
However, research by Pope and Wæraas (2016) suggests that the overstating of CSR activities is probably much less common than is often perceived by the general public and academics. This idea is confirmed in a large study by Hawn and Ioannou (2016), which found that most firms perform more internal actions regarding CSR than external actions. As a result of understating internal CSR actions, firms’ market-value is harmed (Hawn & Ioannou, 2016). Furthermore, CSR understating means that stakeholders are unaware of positive CSR activities that firms truly undertake. Regardless of the direction of CSR decoupling, it can be damaging to the firms and their stakeholders as a form of information asymmetry.
To the best of my knowledge at the time of writing, there is a lack of research on determinants of CSR decoupling, factors which influence CSR decoupling and could be addressed to reduce CSR decoupling. Corporate governance can be described as “the system by which companies are directed and controlled” (Cadbury, 2000, p.8) and “the sum total of all formal procedures according to which a firm’s decision are made” (De Graaf & Stoelhorst, 2009, p.286). Therefore, if organizations’ decisions are dependent on their corporate governance mechanisms, it logically follows that these decisions can be influenced by changing aspects of the corporate governance system. Harjoto and Jo (2011) argue that effective corporate governance enforces organizations to use CSR as a means to resolve conflicts among stakeholders. They found evidence that both internal and external monitoring are positively related to CSR engagement.
In this research, I investigate four different corporate governance mechanisms. Regarding the board of directors, if female directors are found to behave more ethically and are more aligned with CSR matters than male directors, it is likely that they would also want to
decouple CSR less. As boards of directors are traditionally dominated by male directors, I expect that more gender diverse boards will decouple CSR to a lesser degree, based on resource dependence theory. Moreover, I expect that independent directors will behave more ethically towards stakeholders compared to dependent directors, in line with stakeholder-agency theory. It is more difficult for an executive director to objectively monitor the performance and disclosure on CSR of the firm, as they would often be monitoring their own actions. This results in a larger risk of unethical behavior. Independent directors do not face this problem and are therefore expected to be better monitors.
Similarly, I expect that an audit committee which is more independent from management will also be a better monitor regarding CSR. An important duty of the audit committee is to safeguard the disclosure of the firm and improve its transparency. Independent committee members will have increased incentives to achieve this duty, as they have less ties to the management and are less likely to be influenced by the management. This improved monitoring would result in firms with higher levels of audit committee independence to show lower levels of CSR decoupling. Finally, the existence of a committee dedicated to CSR could also reduce the levels of CSR decoupling. The existence of such a committee places emphasis on the importance of CSR within the board. I expect that a CSR committees can be an effective tool to hold management accountable for their actions and advise them on CSR matters. Both disclosure of and performance of CSR efforts could thereby improve, resulting in lower levels of CSR decoupling.
The research question of this study is then formulated as follows: To what extent are
gender diversity in boards of directors, board independence, audit committee independence and CSR committee existence related to CSR decoupling?
I combine data on firms’ CSR disclosure levels, CSR performance levels, corporate governance and financial performance from different databases to create a worldwide sample of 17,115 firm years from 40 countries and spanning from 2009 to 2017. Interestingly, I find that gender diversity of the board, board independence and CSR committee existence are mostly significantly positively related to CSR decoupling, contrary to my expectations. For the environmental dimension and for the firms which overstate their CSR or were low CSR performers, board independence is negatively related to CSR decoupling. Audit committee independence and the existence of a CSR committee are negatively related to CSR decoupling for firms with high CSR performance. Audit committee independence is also negatively related to CSR decoupling on the environmental dimension. The findings indicate that
stakeholder-agency theory and resource dependence theory do not fully explain the relationships of the tested corporate governance mechanisms in most situations.
This study makes several contributions to the literature on CSR decoupling. First of all, while there has been research showing CSR decoupling exists (Boiral, 2013; Christmann & Taylor, 2006; Graafland & Smid, 2016) and studies on the effects of the disclosure gap (Cho et al., 2015; Ogunfowora, Stackhouse, & Oh, 2018), this research helps to fill the gap in the literature on determinants of CSR decoupling. Secondly, I gain additional insight into the phenomenon of CSR decoupling and its relations with corporate governance by splitting the sample into over- and understating firms, as well as in firms which are high CSR performers and which are low CSR performers. Furthermore, this study uses a large and generalizable worldwide sample which does not focus on a single industry or country.
On a practical level, this study contributes in future decision making of firms, regulators, shareholders and other stakeholders. For example, the appointment of independent directors can help to decrease information asymmetry on environmental CSR efforts, or for CSR efforts overall for firms with low CSR performance or which tend to overstate their CSR efforts.
In the remainder of this paper, I will first discuss the literature on CSR decoupling, lay out the theoretical framework and build the hypotheses of the study. The research methodology is outlined in the third chapter. Chapter four will present the results and lastly, chapter five discusses the findings, implications and limitations of the study.
2. Theoretical framework
2.1 Literature review: CSR decoupling and effective corporate governance
Before building the hypotheses, I will first discuss the literature surrounding CSR decoupling and its components. A non-exhaustive overview of the literature is provided in table 1. A first observation from the literature is that CSR efforts and disclosure have become more ubiquitous in the past decennia.Similarly, research on CSR decoupling has gained traction in recent years. Studies by Boiral (2013) and Graafland & Smid (2016) are examples of studies which find that CSR decoupling does occur in practice. Other academics have investigated why CSR decoupling occurs. They find that firms use CSR for signaling purposes (Jamali, Lund-Thomsen, & Khara, 2017; Hyatt & Berente, 2017), to maintain legitimacy (Peters & Romi, 2016) or as a response to stakeholder pressure (Cho et al. 2015; Crilly et al., 2012). Furthermore, Wickert, Scherer and Spence (2016) have developed a framework based on contingency theory to argue that as firm size increases, the difficulty and costs to coordinate “walking” the “talk”
increase, while organizational costs to “talk” CSR decrease. This creates incentives for larger firms to focus mainly on the talking of CSR.
On the other hand, Hawn and Ioannou (2016) found that the majority of firms perform more internal actions than external actions regarding CSR, indicating that most firms do not overstate their CSR, rather they understate their CSR efforts. This finding is in line with the theory of Pope and Wæraas (2016), which suggests that it is much less common for firms to overstate their CSR efforts than is commonly perceived by both the general public and academics. CSR understating is harmful as well: it is a form of information asymmetry and leads to lower firm value (Hawn & Ioannou, 2016).
While little research has been done to study the effects of corporate governance on CSR decoupling, it becomes clear from the literature review that corporate governance does affect the components of CSR decoupling: CSR performance and CSR disclosure. Often returning studied corporate governance characteristics are the gender diversity and independence of the board of directors. Both are typically found to have positive effects on disclosure or performance of CSR. In recent years, there have also been studies which link audit committee characteristics and existence to CSR disclosure (Al-Shaer & Zaman, 2018; Appuhami & Tashakor, 2017). Furthermore, it is apparent that the increased attention from firms towards CSR corresponds with an increasing use of dedicated CSR committees (Spitzeck, 2009). These are found to be positively correlated with CSR disclosure (Peters & Romi, 2016; Liao, Luo, & Tang, 2015).
The literature review shows furthermore that agency and stakeholder theory are often used to explain the relations between corporate governance and CSR (see table 1). Effective corporate governance is necessary for genuine and sustainable CSR orientations (Jamali, Safieddine, & Rabbath, 2008). It can provide firms with the resources and incentives needed to align agents of the firms to the wants and needs of their stakeholders. The board of directors has a role to monitor and give advice to the management. The resources available to a board of directors as well as its level of independence from management impact its ability to monitor management effectively. Audit committees exist to monitor the financial and non-financial reporting of the organization, which includes CSR disclosure. CSR committees are instigated to specifically monitor the CSR activities of an organization. Stakeholder-agency theory addresses the roles these committees play in guiding management to the right path. I will expand upon these ideas in the hypothesis building.
Table 1 literature overview
Study Independent variable(s) (results in parenthesis) or contribution
Dependent variable
Applied theory Country
Graafland & Smid (2016)
Study finds that decoupling between policies and implementation does occur, but it is rarely completely decoupled. CSR decoupling Framework by Delmas & Burbano 24 countries worldwide Jamali et al. (2017)
Study finds that SMEs in India couple CSR regarding child labor, but decouple in relation to humanitarian and labor right issues, using CSR for signaling purposes.
CSR decoupling Institutional theory
India
Wickert et al. (2016)
Theoretical framework which argues that large firms are more likely to decouple CSR than small firms.
CSR decoupling Contingency and agency theory
-
Boiral (2013) Study finds that 90 percent of significant negative events were not reported, indicating CSR decoupling.
CSR decoupling Society of spectacle Global Hyatt & Berente (2017)
Study finds that internal, normative stakeholder pressures often drive substantive commitments to CSR, whereas external, normative pressures primarily drive symbolic commitments to CSR.
CSR decoupling Stakeholder and signaling theory
Global
Cho et al. (2015)
Contradictory societal and institutional pressures (+) CSR decoupling Organized hypocrisy
USA
Crilly et al. (2012)
Information asymmetry (+), competing stakeholder expectations (+/-)
CSR decoupling Stakeholder and Institutional theory
Global
Peters & Romi (2014)
Environmental committee existence (+), CSO existence (+), committee size (+/-), committee meetings (+), committee expertise (+), Voluntary disclosure Legitimacy theory US Hawn & Ioannou (2016)
CSR decoupling (-) Firm value Neo-institutional
theory
33 countries worldwide
Byron & Post (2016)
Board gender diversity (+) CSR performance Upper echelons theory
Global
Al-Shaer & Zaman (2016)
Board gender diversity (+) Sustainability disclosure Stakeholder and agency theory UK Ben-Amar et al. (2017)
Board gender diversity (+) Voluntary
disclosure Resource dependence theory Canada Liao et al. (2015)
Gender diversity (+), board independence (+), environmental committee (+) Voluntary disclosure Upper echelons theory USA Hoang et al. (2018)
Diversity-of-boards (+), diversity-in-boards (0) CSR disclosure Resource dependence and agency theory Vietnam García-Sánchez et al. (2017)
Gender diversity (+), financial expertise (+) Accounting quality Resource dependence theory 9 countries worldwide
Dunn & Sainty (2009)
Board independence (+) CSR performance Agency theory Canada
Janggu et al. (2014)
Board size (+), board ownership (0), board professionalism (+), board designation (+), board independence (0)
CSR disclosure Agency theory Malaysia
Dienes & Velte (2016)
Gender diversity (+), former manager presence (0), frequency of meetings (0), supervisory board size (0)
CSR disclosure Stakeholder and stakeholder-agent theory
Germany
Arena et al. (2015)
Positive language use (-), board stakeholder orientation (+) Environmental performance Stakeholder and resource dependence theory USA Al-Shaer & Zaman (2018)
Audit committee independence (+), AC expertise (+), AC meetings (+), sustainability committee existence (+)
CSR disclosure Resource dependence theory UK Appuhami & Tashakor (2017)
AC size (+), AC meetings (+), AC independence (+), AC independent chair (0), AC expertise (0), AC gender diversity (+) CSR disclosure Resource dependence and stakeholder theory Australia Ienciu et al. (2012)
CSR committee existence (+), board independence (+), board size (0)
Environmental disclosure
Agency theory Global
2.2 Hypothesis 1: gender diversity in boards of directors
The possible effect of gender diversity in boards of directors on CSR decoupling can be described by resource dependence theory. According to resource dependence theory, the board of directors is a resource for firms, which consists of the expertise, advice, reputation and information networks of the directors (Hillman & Dalziel, 2003), as well as their values (Byron & Post, 2016). To achieve the best possible results, all available resources need to be leveraged as best as possible. Different directors bring different resources to the table, and a higher level of diversity in the board of directors would be expected to bring more resources available. In that sense, it would be illogical for an organization to not consider half of the workforce for board positions (Daily & Dalton, 2003). Indeed, there is evidence that the exclusion of a portion of society due to gender, rather than talent, leads to sub-optimal boards (Cassell, 2000).
There are a number of ways in which female directors can be valuable resources in the board of directors. Firstly, the values of female directors tend to be more aligned with corporate social performance and women tend to care more about social performance issues (Byron & Post, 2016). Women are also found to be more diligent, committed and involved in general, while being less oriented towards self-interest (Liao et al., 2015). Moreover, female directors do less often have a background in business and show more interest in philanthropic and
community service activities. Female directors are also more likely to hold advanced degrees, which could lead to them being more concerned with CSR issues. (Byron & Post, 2016) Furthermore, researchers argue that women have a lower tendency to portray unethical business behavior and are more socially oriented than men (Kassinis, Panayiotou, Dimou, & Katsifaraki, 2016). Other researchers argue that women, compared to men, play a larger role in establishing positive values towards carbon emissions reduction, with the greatest value on ‘going green’ being put by highly educated females. (De Silva & Pownall, 2014).
The above mentioned characteristics of women support the idea that women in boards have a positive influence on CSR performance and disclosure. They correspond with two of the four responsibilities of CSR as described by Carroll (1999), namely: the ethical responsibility to do what is right and fair, and the philanthropic responsibility to provide resources for social, educational, cultural or recreational purposes.
Supporting the theoretical underpinnings, Byron and Post (2016) found that gender diversity is significantly positively related to CSR performance, through a meta-analysis on 26,710 firm years. Research by Williams (2003) also found that firms with a higher level of gender diversity have higher levels of charitable giving. Moreover, gender diversity was found to be positively related to CSR disclosure (Liao et al., 2015; Al-Shaer & Zaman, 2016). Abbott, Parker and Presley (2012) found that the presence of female board members is negatively related to the likelihood of a financial restatement, indicating that these firms have a higher quality of financial reporting through improved corporate governance.
Based on the findings of previous research and drawing on resource dependence theory, a gender diverse board could be better equipped to monitor management regarding both CSR activities and reporting. I expect gender diversity in boards of directors to have a negative relationship with CSR decoupling. The first hypothesis is then defined as follows:
H1: There is a negative relationship between gender diversity in boards of directors and CSR
decoupling.
2.3 Hypothesis 2: board independence
Stakeholder-agency theory (Hill & Jones, 1992) can be used to explain the role of independent and dependent directors of the board on CSR decoupling. CSR decoupling leads to a form of information asymmetry between management and stakeholders, as what management reports to its stakeholders is different than how it factually acts. As outsiders, stakeholders have difficulty assessing whether the management of a firm acts in their interest. Management can have several reasons to decouple CSR disclosure from its factual
performance. Stakeholders will want to reduce the information asymmetry by gathering information. The cost of doing so may however be prohibitive for individual stakeholders (Hill & Jones, 1992). Corporate governance mechanisms can be utilized to incentivize and monitor management. The board of directors is one such mechanism, with both a monitoring and an advisory role to the management of the firm. It commonly consists of both independent and dependent members. Independent directors, also referred to as outside directors, are those who are not employees of the firm and are not otherwise affiliated with it. They are also not directly involved in the day-to-day operations of the organization (De Villiers, Naiker, & Van Staden, 2011). This means that they have less ties to the management of the firm. As a result, one would expect independent directors to be able to perform their monitoring and advisory roles more objectively.
Indeed, it is commonly accepted that this is the case (Liao et al., 2015). Independent directors are less aligned with management than dependent directors and are less likely to collude with internal directors, because they are less incentivized to do so (Carter, Simkins & Simpson, 2003). Furthermore, as the careers of independent directors do not depend as much on the CEOs of the firms they monitor compared to those of dependent directors, CEOs hold less power over independent directors (Core, Holthausen, & Larcker, 1999). It is argued that because of their reduced dependence on management and their diverse backgrounds, independent directors tend to have a strong stakeholder orientation (Coffey & Wang, 1998). They are also more likely to take a broader view on organizational performance, one that does not primarily focus on financial measures and benefits (Liao et al., 2015). It is this broader and more stakeholder-oriented view, combined with an increased independence from management, that makes the appointment of independent directors an effective monitoring mechanism that can restrict potential opportunistic behavior of agents.
These theoretical arguments are supported by evidence. A higher level of independence in the board of directors is found to be positively related to environmental performance (De Villiers et al., 2011).Furthermore, a higher level of independent directors is thought to improve the reporting system of the firm and found to lead to higher levels of environmental transparency (Michelon & Parbonetti, 2012). I expect that, through their decreased ties to executive management, independent directors can be more objective, more effectively hold management accountable, and are less likely to deceive stakeholders. This should lead to less information asymmetry and CSR decoupling. Thus, I expect to find a negative relationship between the level of independence of the board of directors and CSR decoupling. The second hypothesis is defined as follows:
H2: There is a negative relationship between board independence and CSR decoupling.
2.4 Hypothesis 3: audit committee independence
Stakeholder-agency theory can also explain the role of audit committees on CSR decoupling. An audit committee is a committee under the board of directors which is tasked to monitor financial and non-financial reporting. Similar to the board of directors, it is thus a corporate governance mechanism utilized by stakeholders to monitor the management of a firm. Audit committees are expected to minimize information asymmetry between management and stakeholders (Karamanou & Vafeas, 2005). An audit committee does not have much influence in steering the organization by itself, but it can be seen as “the ultimate monitor” on the reporting activities of an organization and it protects and preserves the interests and equity of shareholders (Zhang, Zhou, & Zhou, 2007). Audit committees are found to improve the quality of sustainability reporting and its credibility, through their expertise and oversight (Al-Shaer & Zaman, 2018). Trotman and Trotman (2015) found evidence that audit committee members took a monitoring role on reporting processes of sustainability reports.
The effectiveness of an audit committee is expected to be increased as it becomes more independent, as less ties to management means the audit committee can be more objective in monitoring and holding management accountable. Thereby it becomes a better governance mechanism in the stakeholder-agent problem. Indeed, audit committees consisting of more independent directors are positively related with the quality of sustainability disclosure (Al-Shaer & Zaman, 2018), in line with stakeholder-agency theory.
Drawing on the available research and stakeholder-agency theory, I expect more independent audit committees to limit the level of CSR decoupling because of their increased ability to reduce information asymmetry. If an audit committee is more independent, I expect it to be more objective and to be better capable of monitoring and holding management accountable. Because of the increased objectivity, I expect independent audit committee members to be less willing to deceive stakeholders, as they have less incentives to do so. Therefore, I expect to find a negative relationship between the level of independence of an audit committee and CSR decoupling. The third hypothesis is then formulated as follows:
H3: There is a negative relationship between the independence of an audit committee and CSR
2.5 Hypothesis 4: CSR committee existence
A CSR committee is a dedicated committee on the level of the board of directors, which has oversight on the CSR activities of a company. Similar to the case of the audit committee, stakeholder-agency theory can explain the possible role of the CSR committee on CSR decoupling. CSR committees could lead to reduced information asymmetry regarding CSR by placing emphasis on the importance of CSR on a board level.
A study by Lock and Seele (2016) found that 80.5 percent of 41 firms from 2013 with the best performance in economic, social and environmental terms had a committee or governance board solely committed to CSR. Similarly, Eccles, Ioannou and Serafeim (2014) found that firms were more likely to have a separate committee responsible for CSR if they had a sustainability strategy. Spitzeck (2009) noted that CSR was increasingly “infused and embedded” in the corporate governance structures of UK-based firms, as CSR committees had become more common from 2002 to 2005.
Although research on the effects of CSR committees specifically is limited, Peters and Romi (2014) found that environmental committees are positively related with disclosure on greenhouse gas emissions. Liao et al. (2015) argue that environmental committees may enhance employees’ awareness around the environmental impact of their jobs, and indeed they found that the existence of environmental committees is positively related to environmental disclosure.
I expect that, through their monitoring role dedicated to CSR activities and CSR reporting, CSR committees can decrease the information asymmetry that exists between management and stakeholders. Thus, I expect a negative relationship to exist between the existence of CSR committees and CSR decoupling and therefore, I formulate the following hypothesis:
H4: There is a negative relationship between the existence of a CSR committee and CSR
decoupling.
Figure 1: conceptual model
3. Research methodology
3.1 Data collectionThis study was performed using quantitative data gathered from databases containing large quantities of data on publicly traded companies. The Bloomberg Professional database was used to gather CSR disclosure data, to measure how firms scored on CSR disclosure. The CSR performance data was taken from Thomson Reuter’s ASSET4 Database, which independently tracks firms’ performance on environmental, social and corporate governance (ESG) metrics. The ESG score from this database was used to measure firms’ CSR performance.
Data for the independent variables on the board of directors, audit committee and sustainability committee were taken from the BoardEx database of Wharton Research Database Services (WRDS). This database allows access to data of the existence and composition of boards and committees, including personal information such as education, gender, age and independence. This data was then translated into the independent variables of board independence, gender diversity of the board, audit committee independence and sustainability committee existence per firm year.
Finally, Thomson Reuter’s ASSET4 Database was used to gather financial control variable data. All gathered data was subsequently merged into one database file using ISIN and year variables. While originally data from 2002 to 2017 was planned to be used, this was not possible in practice due to missing data in the Bloomberg Professional database from before
CSR
decoupling
Board gender
diversity
Board
independence
Audit
committee
independence
CSR
committee
existence
2009. The initial Bloomberg sample contained 87,325 firm years, while the ASSET4 database contained 115,312 firm years. After merging the datasets and removing observations with missing data for all used variables except research & development costs, the final sample consisted of 17,115 firm years from 4,637 firms, from the years 2009 to 2017 and from 40 countries worldwide. Firms from the financial sector were excluded as they can be subject to different disclosure requirements (Gul & Leung, 2004) and their high levels of leverage are considered to be different in meaning than for non-financial firms (Fama & French, 1992), which could skew the results. Furthermore, observations from countries with less than 25 firm years were removed in order to avoid representativeness issues. Appendix I shows how the observations are distributed between years, industries and countries.
3.2 Variables
3.2.1 CSR decoupling The dependent variable, CSR decoupling, is operationalized by taking the absolute value of the difference between the CSR performance score and the CSR disclosure score for each firm year. The performance and disclosure scores both range from 0 to 100. It is important to note that in this calculation, following Hawn and Ioannou (2016), the CSR performance score was taken from one year prior to the CSR disclosure score. The underlying logic is that disclosure, such as a sustainability or integrated report, usually refers to the prior period. Hawn and Ioannou (2016) also argue that it will take at least a year before internal actions are implemented, legitimized and can make an impact. The performance score should therefore be lagged. The dependent variable can thus be defined as follows:
𝐶𝑆𝑅𝐷𝑒𝑐𝑜𝑢𝑝𝑙𝑖𝑛𝑔𝑖𝑡 = |𝐶𝑆𝑅𝐷𝑖𝑠𝑐𝑙𝑜𝑠𝑢𝑟𝑒𝑆𝑐𝑜𝑟𝑒𝑖𝑡− 𝐶𝑆𝑅𝑃𝑒𝑟𝑓𝑜𝑟𝑚𝑎𝑛𝑐𝑒𝑆𝑐𝑜𝑟𝑒𝑖𝑡−1|
Firm is noted by i and year is noted by t. The resulting score indicates the level of CSR decoupling. By taking the absolute value of the difference, I have a dependent variable which is 0 in case of no decoupling (when the performance score is exactly the same as the disclosure score) and 100 in case of the highest possible level of decoupling between performance and disclosure.
3.2.2 Dependent variables The first independent variable, gender diversity within the board of directors, is measured as the percentage of female directors in a board. A relative measure provides more information on the gender diversity of boards than the absolute number of female directors on a board. Board independence, the second independent variable, is measured as the
percentage of board members who are fully independent. Similarly, audit committee independence is measured as the percentage of audit committee members who are fully independent. Lastly, CSR committee existence is noted as 0 or 1; 0 indicating no CSR committee and 1 indicating the existence of a CSR committee. A CSR committee is defined as a committee solely dedicated to CSR matters. This data was collected using the BoardEx database from WRDS, containing the names of board level committees. These were then filtered to find all CSR committees, using search terms such as “CSR”, “Sustainability”, “Environmental”, “Social”, “Corporate governance”.
3.2.3 Control variables Based on review of prior studies, I control for the following firm level accounting variables: firm size, leverage, profitability, sales growth and research and development intensity. Firm size is measured as the log of total assets (Al-Tuwaijri, Christensen, & Hughes, 2004). According to Liao et al. (2015), firm size acts as a proxy for social pressure which firms experience to act on CSR matters, suggesting a negative coefficient. However, a positive coefficient is also be a possibility as Wickert et al. (2016) argue that as firm size increases, the cost to “talk” CSR decreases while the cost to “walk” CSR increases. Return on assets (ROA) is used as a proxy for profitability, following Liao et al. (2015). A positive coefficient is expected as Graafland & Smid (2016) argue that more profitable companies will find it easier to dedicate resources to implement CSR policies. Leverage is measured using the debt-to-asset ratio, calculated as total debt divided by total assets. This control variable is included as it can influence how firms deploy their assets with regards to CSR (Marano et al., 2017). Sales growth is controlled for as it is found to be positively related to CSR performance (Lev, Petrovits, & Radhakrishnan, 2010). It is measured as the percentage growth of net revenues compared to the previous year. Following Tashman, Marano and Kostova (2019), I control for research and development (R&D) intensity, as more innovative firms may be more capable of improving their CSR performance (McWilliams & Siegel, 2000). R&D intensity is measured as R&D costs divided by net revenue. In case of missing R&D data, I follow Marano et al. (2017) and use the industry mean per firm year.
Finally, I control for year, industry and country effects with the use of dummy variables. Industries are divided into 8 categories based on their SIC codes, as shown in Appendix I. Table 2 summarizes the description of variables.
Table 2 Variable description
Variable name Proxy Measurement
Dependent variables
CSRDecoupling CSR decoupling Absolute value of
CSRDisclosureScore – lagged CSRPerformanceScore
CSRPerformanceScore CSR performance Score from 0 to 100
CSRDisclosureScore CSR disclosure Score from 0 to 100
Independent variables
GenDiv Gender diversity within
boards of directors
Number of female board members divided by total number of board members
BoardInd Board independence Number of independent board
members divided by total number of board members
AudInd Audit committee
Independence
Number of independent audit committee members divided by total number of audit committee members CSRComm CSR committee existence 1 if CSR committee exists, 0 if no CSR committee exists Control variables
FirmSize Firm size Log of total assets
Leverage Leverage Total debt divided by total assets
Profit Profitability ROA: net income divided by total
assets
SalesGrow Sales growth The percentage of growth of net
revenue compared to the previous year
RDint R&D intensity R&D costs divided by net revenue
Year Year Dummy variables for each year
Industry Industry Dummy variables for each industry category
Country Country Dummy variables for each
country
3.3 Data analysis
The hypothesized relation, including control variables, is formulated as follows:
𝐶𝑆𝑅𝐷𝑒𝑐𝑜𝑢𝑝𝑙𝑖𝑛𝑔𝑖𝑡 = 𝛽0 + 𝛽1 𝐺𝑒𝑛𝐷𝑖𝑣𝑖𝑡 + 𝛽2 𝐵𝑜𝑎𝑟𝑑𝐼𝑛𝑑𝑖𝑡 + 𝛽3 𝐴𝑢𝑑𝐼𝑛𝑑𝑖𝑡+
𝛽4 𝐶𝑆𝑅𝐶𝑜𝑚𝑚𝑖𝑡+ 𝛽5 𝐹𝑖𝑟𝑚𝑆𝑖𝑧𝑒𝑖𝑡+ 𝛽6 𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒𝑖𝑡 + 𝛽7 𝑃𝑟𝑜𝑓𝑖𝑡𝑖𝑡+ 𝛽8 𝑆𝑎𝑙𝑒𝑠𝑔𝑟𝑜𝑤𝑖𝑡+ 𝛽9 𝑅𝐷𝑖𝑛𝑡𝑖𝑡+ 𝛽10 𝑌𝑒𝑎𝑟𝑖𝑡+ 𝛽11 𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦𝑖𝑡+ 𝛽12 𝐶𝑜𝑢𝑛𝑡𝑟𝑦𝑖𝑡+ Ɛ
Ɛ denotes the error term and β represents the correlation coefficients. To estimate the above formulated model, multiple linear regression analysis was performed in Stata 15. This regression was performed using robust standard errors, to control for heteroscedasticity. Heteroscedasticity can occur when using panel data. Untabulated results show that using robust standard errors had very little impact on the significance of the results compared to not doing so, meaning that the results are not influenced by this choice.
4. Results
In this chapter, I will present the results of the study. The descriptive statistics and the correlation matrix are displayed first, before I provide the results of the first six models testing the controls and hypotheses. Finally, I perform several additional tests to gain more insight into the phenomenon of CSR decoupling.
4.1 Descriptive statistics
Table 3 provides the descriptive statistics of this study. To correct for the influence of extreme outliers, the accounting variables (FirmSize, Leverage, Profit, SalesGrow and RDint) were first winsorized in Stata 15 at the 1st and 99th percentile.
Interestingly, the mean CSRDecoupling score is -24.29, indicating that on average, firms scored better on their CSR performance than on their disclosure. Female board membership is 13.96% on average, with a minimum of 0% and a maximum of 71.43%. Board
independence is 53.31% on average, with percentages ranging from 1.34% to 95.28%. Audit committee independence is higher at 85.33% on average, ranging from 0 to 100%.
Table 3 Descriptive statistics
Variables N Mean SD Min Max
CSRDecoupling (absolute difference) 17115 31.88 22.65 0 91.34
CSRDecoupling (relative difference) 17115 -24.29 30.64 -91.34 79.80
CSRDisclosureScore 17115 27.53 15.18 0.83 85.54 CSRPerformanceScore 17115 51.82 30.89 2.42 97.43 EnvironmentalDecoupling (absolute) 17115 42.34 31.15 0 95.26 SocialDecoupling (absolute) 17115 36.60 25.76 0 97.62 GovernanceDecoupling (absolute) 17115 31.15 20.44 0 97.62 GenDiv 17115 13.96 12.02 0 71.43 BoardInd 17115 53.31 29.87 1.34 95.28 AudInd 17115 85.33 25.95 0 100 CSRComm 17115 0.53 0.50 0 1 FirmSize 17115 16.56 2.79 11.55 24.31 Leverage 17115 24.71 18.83 0 82.31 Profit 17115 4.78 8.66 -40.37 31.15 SalesGrow 17115 8.03 24.82 -51.57 147.58 RDint 17115 12.21 40.46 0 303.79 4.2 Correlation analysis
Next, Pearson’s correlation analysis was performed to test the correlation between variables and to look for possible multicollinearity issues. The results of the analysis are presented in table 4. The table shows there is a high correlation coefficient of 0.63 (p<0.05) between board independence and audit committee independence. This could be explained because audit committee members are also part of the board of directors; the audit committee is a committee of the board of directors. Nonetheless, a commonly used threshold for multicollinearity is a coefficient of 0.6, which is in this case exceeded, indicating multicollinearity could be an issue. However, research by Dormann et al. (2016) shows collinearity becomes a serious threat with correlation coefficients higher than 0.7, which is not the case in this study. Another way to test for multicollinearity is to measure the Variance Inflation Factors when performing the regression analysis. The highest VIF was found to be 4.09 (as shown in table 5), which is below the common threshold of 10. This means that
multicollinearity does not pose a threat to the results of my analysis and the conclusions based upon it.
Table 4 Pearson’s correlation matrix
Variables (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) CSRDecoupling (1) 1.00 GenDiv (2) 0.14* 1.00 BoardInd (3) 0.07* 0.24* 1.00 AudInd (4) 0.02* 0.07* 0.63* 1.00* CSRComm (5) 0.21* 0.13* -0.03* -0.04* 1.00 FirmSize (6) -0.07* -0.11* -0.20* -0.10* 0.13* 1.00 Leverage (7) -0.00 0.01 0.00 0.01 0.01 0.06* 1.00 Profit (8) 0.02* 0.03* 0.03* 0.03* 0.01 -0.01 -0.08* 1.00 SalesGrow (9) -0.01 -0.03* 0.01 0.02* -0.05* -0.07* -0.02* 0.09* 1.00* RDint (10) 0.02* -0.01 0.03* 0.02* -0.05* -0.29* -0.03* -0.15* 0.07* 1.00* *p<0.05 (two-tailed). 4.3 Regression analysis
Initially, six multiple linear regression models were run in Stata 15 to test the hypotheses of this study. The results are outlined in table 5. In the first model, only the relation between the control variables and CSR decoupling is tested. The second to fifth models test H1, H2, H3 and H4 respectively, in addition to the controls. In model 6, all variables are tested at once.
Model 1 shows us that FirmSize is significantly positively related to the dependent variable CSRDecoupling (β=0.34, p<0.01). A marginally positive relationship is found between RDint and CSRDecoupling (β=0.01, p<0.10). Leverage, Profit and SalesGrow show no significant relationship to the dependent variable in this model. The explanatory power of this model testing only control variables was 2.1%.
Model 2 tests the relationship between GenDiv and CSRDecoupling and finds they are significantly positively related (β=0.25, p<0.01). Model 3 finds a significantly positive
relationship between BoardInd and CSRDecoupling (β=0.03, p<0.01). Model 4 finds no significant relationship between audit committee independence and CSRDecoupling. In model 5, a significantly positive relationship is found between CSRComm and CSRDecoupling (β=10.63, p<0.01). This model also shows that CSRComm is the best predictor of CSRDecoupling tested so far, as the R2 value of this model is 0.072.
Finally, the complete model is tested in model 6. This model had the largest degree of explanatory power with an R2 value of 0.081. The outcome of the analysis provides no evidence in support of my hypotheses. AudInd does not seem to be significantly related to CSRDecoupling. Meanwhile, GenDiv (β=0.18, p<0.01), BoardInd (β=0.02, p<0.01) and CSRComm (β=10.01, p<0.01) all showed a significant but positive relationship to CSRDecoupling.
Regarding the control variables, FirmSize was negatively related to CSR decoupling (β=-0.21, p<0.10) after including all other test variables, supporting the argument of Liao et al. (2015) that firm size acts as a proxy for the social pressure which larger firms experience. RDint was significantly positively related to CSR decoupling (β=0.01, p<0.01). Other control variables (Leverage, Profit, SalesGrow) showed no significant relationship to CSR decoupling.
Table 5 Multiple linear regression models
Models (1) (2) (3) (4) (5) (6) GenDiv 0.25 *** 0.18 *** BoardInd 0.03 *** 0.02 *** AudInd 0.00 -0.01 CSRComm 10.63 *** 10.01 *** FirmSize 0.34 *** 0.22 * 0.31 ** 0.34 *** -0.14 -0.21 * Leverage -0.00 -0.00 -0.00 -0.00 -0.01 -0.00 Profit 0.04 0.03 0.03 0.04 0.03 0.02 SalesGrow -0.01 -0.01 -0.01 -0.01 -0.01 -0.00 RDint 0.01 * 0.01 ** 0.01 0.01 * 0.01 * 0.01 ** Constant 24.01 *** 22.73 *** 23.40 *** 23.75 *** 26.44 *** 25.58 *** Industry/year/country
dummies included Yes Yes Yes Yes Yes Yes
R2 0.021 0.037 0.023 0.021 0.072 0.081
Highest VIF 4.01 4.02 4.02 4.01 4.08 4.09
F-value 8.24 *** 13.04 *** 8.64 *** 8.13 *** 24.02 *** 25.96 ***
4.4 Additional tests
In order to test the robustness of the results and to gain additional insight into the phenomenon of CSR decoupling and the effect that gender diversity, board independence, audit committee independence and CSR committee existence have on it, I perform several additional tests. In the first additional test, I divide the sample into two groups: firms which overstate their CSR performance, and those that understate. In the second test, I use the full sample again to test the relationship of the independent variables on CSR performance and CSR disclosure separately. In the third additional test, I divide the sample into one group of firms with high CSR performance and one with low CSR performance, and test the effects of the independent variables on these groups separately. Finally, I perform a dimensional analysis by testing the relationships of the independent variables on environmental, social and governance decoupling. All additional tests are performed using multiple linear regression.
4.4.1 Over- and understating CSR To see whether there are differences in the relationships of the independent variables and CSRDecoupling between over- and understating firms, I divide the sample into two groups. The first sample consists of only those firms which score better on their CSR disclosure than on their CSR performance (CSRDisclosureScoreit –
CSRPerformanceScoreit-1 > 0). These firms thus overstate their CSR. The second sample
consists of only those firms which understate their CSR; these firms score better on their CSR performance than on their disclosure (CSRDisclosureScoreit – CSRPerformanceScoreit-1 < 0). The dependent variable is still measured as an absolute value in both models. In line with the study of Hawn and Ioannou (2016), understating CSR was more common than overstating CSR. The sample of overstating firms consists of 4,135 firm years, whereas the sample of understating firms consists of 12,983 firm years. The results of the additional test are provided in table 6.
The overstating model shows that firms with higher board independence levels show lower levels of CSRDecoupling, as BoardInd is found to be significantly negatively related to CSRDecoupling (β=-0.03, p<0.01) for the overstating firms. While GenDiv and CSRComm show negative relationships in this sample as well, these relationships are not statistically significant.
For the firms that understate their CSR performance, which is the majority of firms studied, GenDiv (β=0.20, p<0.01) and CSRComm (β=10.12, p<0.01) are significantly positively related to CSRDecoupling, in line with the findings of models 2, 5 and 6. However, the level of board independence is not significantly related to CSRDecoupling for this sample.
Table 6 Linear regression models of additional tests 1-3
Test Over- & understating CSR CSR Disclosure & performance CSR performance level Model Overstating Understating Disclosure Performance High Low
GenDiv -0.01 0.20 *** 0.03 *** 0.27 *** 0.04 *** 0.01 BoardInd -0.03 *** 0.01 -0.00 0.07 *** 0.02 ** -0.03 *** AudInd 0.00 -0.00 0.01 * -0.01 * -0.02 ** -0.00 CSRComm -0.18 10.12 *** 3.79 *** 18.27 *** -2.79 *** 1.37 *** FirmSize 2.94 *** -0.87 *** 4.01 *** 2.16 *** -4.38 *** 3.41 *** Leverage 0.00 0.00 -0.00 -0.01 -0.01 -0.01 Profit 0.04 ** 0.01 0.09 *** 0.10 *** -0.14 *** 0.01 SalesGrow -0.02 *** 0.01 -0.03 *** -0.03 *** 0.04 *** -0.02 *** RDint -0.01 *** 0.02 ** -0.01 *** 0.01 0.02 *** -0.01 ** Constant -34.56 *** 40.73 *** -37.92 *** 4.45 125.91 *** -45.72 *** Industry/year/country
dummies included Yes Yes Yes Yes Yes Yes
R2 0.293 0.092 0.493 0.164 0.514 0.441
Highest VIF 4.58 3.94 4.09 4.09 3.15 4.98
F-value 29.07 *** 21.88 *** 356.22 *** 55.63 *** 114.07 *** 53.53 ***
N 4135 12983 17115 17115 4496 4224
*p<0.10, **p<0.05, ***p<0.01 (two-tailed).
4.4.2 CSR disclosure and CSR performance scores I previously explained that I expected to find negative relationships between the independent variables and CSR decoupling, as they have been found to be significantly positively related to either CSR disclosure or CSR performance, or both. To gain more insight into why the tested independent variables are not negatively related to CSRDecoupling, I test their impact on CSR performance and CSR disclosure separately. In the disclosure model, I test the relationships between the independent variables and the dependent variable CSRDisclosureScore, while the performance model is used to test the relationship between the independent variables and the dependent variable CSRPerformanceScore. The results are provided in table 6.
In line with the prior stated expectations, GenDiv (β=0.03, p<0.01) and CSRComm (β=3.79, p<0.01) are found to be significantly and positively related to the levels of CSR disclosure. GenDiv (β=0.27, p<0.01) and CSRComm (β=18.27, p<0.01) also show significantly positive relationships to CSRPerformanceScore. The coefficients of these variables are larger for the performance model than for the disclosure model, suggesting that gender diversity of
the board of directors and the existence of a CSR committee have more impact on CSR performance than on CSR disclosure. It is possible that this is where the significantly positive relations to CSRDecoupling in prior models come from. Furthermore, BoardInd shows no significant relation to CSR disclosure, in line with previous research by Janggu, Darus, Zain and Sawani (2014), and shows a significantly positive relationship to CSR performance (β=0.07, p<0.01). AudInd is marginally significantly and positively related to CSR disclosure (β=0.01, p<0.10) and marginally significantly and negatively related to CSR performance (β=-0.01, p<0.10).
4.4.3 High CSR performance and low CSR performance In the third additional test, I create two subsamples. The first model of this test, high, measures the effect of the independent variables on CSRDecoupling for firms which show high levels of CSR Performance. These are firms who score one standard deviation or more higher than the mean on the CSRPerformanceScore. The second model, low, is used to test the effect of the independent variables on CSRDecoupling for firms which show low levels of CSR performance. The cut off is again one standard deviation from the mean level of CSRPerformanceScore. The results are shown in table 6.
The high performance model shows that for firms with high CSR performance scores, AudInd (β=-0.02, p<0.05) and CSRComm (β=-2.79, p<0.01) are negatively related to CSRDecoupling, contrary to models 4, 5 and 6 and in line with hypotheses 3 and 4, respectively. GenDiv (β=0.04, p<0.01) and BoardInd (β=0.02, p<0.01) remain significantly positively related to CSRDecoupling.
For firms with low CSR performance scores, GenDiv and AudInd are found to not be significantly related to CSRDecoupling. However, BoardInd is significantly negatively related to CSRDecoupling (β=-0.03, p<0.01), meaning that hypothesis 2 does hold for firms with low CSR performance. Furthermore, CSRComm is significantly positively related to CSRDecoupling (β=1.37, p<0.01), in line with models 5 and 6.
4.4.4 Environmental, social and governance decoupling As both Bloomberg Professional and Thomson Reuter’s ASSET4 Database provide not only aggregate CSR scores, but also scores on the three dimensions of CSR (environmental, social and governance CSR), it is possible to analyze the effects of the tested governance mechanisms on decoupling per dimension. By doing so, I can gain insight into how the tested variables impact each dimension, and to what extent. Table 7 provides the results of three multiple linear regression models, one
model for each dimension. Similarly to the main analysis, the absolute value of the difference between the disclosure and performance scores is used.
Regarding environmental decoupling, the results show a significantly positive relation for GenDiv (β=0.18, p<0.01), a significantly negative relation for BoardInd (β=-0.03, p<0.01), a marginally significantly negative relation for AudInd (β=-0.02, p<0.10) and a significantly positive relation for CSRComm (β=17.75, p<0.01).
For social decoupling, I find a significantly positive relation for GenDiv β=(0.15, p<0.01), as well as for CSRComm (β=12.04, p<0.01). BoardInd and AudInd show no statistically significant relation to social decoupling.
Finally, for governance decoupling, no statistically significant relation is found for GenDiv. BoardInd however, is significantly positively related (β=0.04, p<0.01) to governance decoupling, AudInd is marginally significantly related (β=0.01, p<0.10) and CSRComm is again significantly positively related (β=3.73, p<0.01).
By comparing the R2 values and F-statistics of the different models, it is noticeable that the tested variables have the most impact on environmental decoupling, followed by social decoupling. The R2 value of 0.031 for the governance decoupling model means that governance decoupling is the least impacted dimension by the tested variables. Similarly, the regression coefficient of CSRComm has the highest value for the environmental decoupling model (β=17.75) and the lowest for the governance decoupling model (β=3.73). It seems then, that the independent variables mostly impact CSR decoupling on the environmental dimension.
Table 7 Environmental, social and governance decoupling
Models Environmental Social Governance
GenDiv 0.18 *** 0.15 *** 0.02 BoardInd -0.03 *** -0.01 0.04 *** AudInd -0.02 * -0.02 0.01 * CSRComm 17.75 *** 12.04 *** 3.73 *** FirmSize 2.03 *** 1.33 *** 0.53 *** Leverage -0.01 0.00 0.01 Profit 0.06 ** 0.08 *** 0.04 ** SalesGrow -0.02 *** -0.01 -0.01 RDint 0.02 *** 0.01 ** 0.01 Constant 6.31 7.63 * 11.95 *** Industry/year/country
dummies included Yes Yes Yes
R2 0.156 0.108 0.031
Highest VIF 4.09 4.09 4.09
F-value 51.77 *** 33.69 *** 9.16 ***
N 17115 17115 17115
*p<0.10, **p<0.05, ***p<0.01 (two-tailed).
5. Discussion and conclusion
In this chapter, I will first discuss the findings of the study. Subsequently, I will discuss its theoretical and practical implications. Finally, I will reflect on the limitations of the study and address future research opportunities.
5.1 Findings
This study set out to find the impact that certain corporate governance mechanisms have on the phenomenon of CSR decoupling. The researched variables are: gender diversity of the board of directors, independence of the board of directors, independence of the audit committee, and existence of a dedicated CSR committee. The findings are based on an analysis of 17,115 firm years, from 40 countries worldwide and from the time period of 2009 to 2017.
Before discussing the individual hypotheses, it is interesting to point out that the average score on the independent variable CSRDecoupling was found to have a negative value, meaning that on average, firms scored better on their CSR performance than on their CSR disclosure.
While the opposite also occurred in the dataset, this finding means that many firms do not overstate CSR, as is commonly perceived by both academics and the general public (Pope & Wæraas, 2016). Rather, firms are more likely to understate their CSR efforts. Hawn and Ioannou (2016) found similar results in their study on internal and external CSR. They provide several reasons for this observation: (1) firms might not want to report activities until they are more confident about them, (2) firms might want to avoid additional pressure from stakeholders, and (3) firms might be risk-averse regarding the perceptions of investment analysts. Another explanation for this type of CSR decoupling is that it stems from an overall policy of secrecy.
Regarding the first hypothesis, I expected to find a negative relationship between gender diversity of the boards of directors and the level of CSR decoupling. The results showed a significant positive relationship however, indicating that firms with more gender diversity in their boards of directors tend to decouple their CSR disclosure performance more. This relationship holds for both firms which understate CSR and for those which overstate CSR, as well as for the subsample of firms with high CSR performance. The dimensional analysis showed positive relationships on environmental and social CSR decoupling, and no significant relationship to governance decoupling. For firms with low CSR performance, gender diversity is not significantly related to CSR decoupling either. Hypothesis 1 is therefore not supported by the evidence.
Based on resource dependence theory, I argued that higher levels of gender diversity would be beneficial to firms through the values and backgrounds which women bring to the board. Indeed, I found that higher levels of gender diversity in the boards lead to higher disclosure and performance scores. Yet it also leads to higher levels of CSR decoupling. This can possibly be explained by the finding that the CSR performance scores are much more affected by the level of gender diversity than the CSR disclosure scores.
Perhaps, it is a conscious decision from female board members to decouple disclosure from performance. For example, CSR reporting is found to be costly and time-consuming, and it is difficult for managers to know if the reports fulfill the information needs of stakeholders (O’Dwyer, 2002). Other disincentives to disclose voluntary information according to managers are the danger of releasing information which harms the competitive position of the firm, and setting a disclosure precedence which is difficult to keep up (Beattie & Smith, 2010). Additionally, not all CSR information may be perceived as being worth reporting. To clarify this idea, Godfrey, Merrill and Hansen (2009) argue that there are two forms of CSR: activities targeted at primary stakeholders and activities targeted at secondary stakeholders. Primary
stakeholders are those who are essential to firms’ operation, while secondary stakeholders are those who can influence the primary stakeholders (Freeman, Harrison, & Wicks, 2007). CSR activities targeted at the first group did not create the same benefits for the firm as those targeted at secondary stakeholders (Godfrey et al., 2009). The first type of CSR may be perceived as more self-serving and thereby not create the moral capital which other CSR activities create. Additionally, as primary stakeholders stand closer to the firm compared to secondary stakeholders, they simply may have less need for disclosure on CSR activities targeted at them. For these reasons, it could be that CSR disclosure is decoupled from its performance. Alternatively, women on the board may not be able to affect disclosure of CSR to the level they can and do influence CSR performance, due to their role as board member.
The second hypothesis focused on the level of independence of the board of directors. I expected a negative relationship to exist between the independence of the board and CSR decoupling. On the contrary, board independence is significantly positively related to CSR decoupling. However, a significant negative relationship is found on the environmental dimension of CSR decoupling. Moreover, for those firms which overstate their CSR, board independence is also negatively related to decoupling, meaning that firms with more independent directors on their board overstate their CSR performance less. The same effect is observed for firms which have low CSR performance scores. Higher levels of board independence are negatively related to CSR decoupling for these firms. Thus, hypothesis 2 only holds true regarding environmental CSR, for firms which overstate their CSR and for firms which have low levels of CSR performance.
It can be argued that overstating is more damaging to stakeholders than understating CSR. Information asymmetry exists in both situations. However, in the case of overstating CSR, where disclosure outweighs performance, it is more likely that the disclosure is of symbolic nature, i.e. stakeholders are being lied to. Similarly, low CSR performance is more unfortunate for the stakeholders of those companies than high CSR performance. It could be that in these cases, independent directors are more willing to decrease CSR decoupling to relieve the information asymmetry between the firm and its stakeholders than dependent directors, in line with stakeholder-agency theory. In other situations, decoupling might be a conscious decision, as discussed previously.
The third hypothesis stated that there would be a negative relationship between the independence of an audit committee and CSR decoupling. The main analysis showed that audit committee independence is not significantly related to CSR decoupling. However, for firms with high CSR performance scores, audit committee independence was found to have a small,
statistically significant negative effect on CSR decoupling. The second additional test gave some explanation for these findings: it seems that audit committee independence has little effect on CSR performance and disclosure separately, as only small effects were found on both. The dimensional analysis showed however that there is a marginally significant negative effect of audit committee independence on environmental decoupling. To conclude, my findings on audit committee independence are mixed and show only weak effects of audit committee independence on CSR decoupling.
I expected that the more independent audit committees are, the more likely they are to improve CSR disclosure, based on stakeholder-agency theory. This increased level of CSR disclosure would lead to decreased CSR decoupling. Previous studies found mixed results regarding the relationship of audit committee independence on voluntary disclosure (Appuhami & Tashakor, 2017). Al-Shaer and Zaman (2018) found a positive relationship of audit committee independence on disclosure regarding sustainability matters, which can partly explain the finding of a marginally significant negative effect on environmental CSR decoupling. Li, Mangena and Pike (2012) found no significant relationship between independent audit committees and voluntary disclosure in their study and provide a possible reason for this result: it is possible that independent audit committee members may be more concerned with releasing proprietary information to competitors. Alternatively, it could be the case that audit committees, through their monitoring role, have more influence on the quality of what is reported than on the decision of what is and what is not reported. This would automatically weigh down the measurable effect that audit committee independence has on CSR disclosure levels. The use of different measurement methods of CSR disclosure could then be a possible reason why some studies do and other studies do not find a positive relationship between audit committee independence and CSR disclosure.
The fourth and final hypothesis of this study focused on the existence of a CSR committee. Specifically, I was interested to see whether the existence of such a committee could decrease the level of CSR decoupling. This hypothesis was initially not supported, as the results show that on the whole, a relatively strong positive relationship exists. This holds true for all three dimensions of CSR decoupling, although the degree of correlation is highest for environmental CSR and lowest for governance CSR. However, for those firms with high CSR performance, CSR committees are significantly negatively related to CSR decoupling. In this situation, hypothesis 4 does hold true. Another interesting finding is that no significant relation was found for firms which overstate CSR. Possibly it is the stakeholder-agency role of the CSR committee which limited managers’ incentives to decouple CSR disclosure from performance
in this case. Similarly, Lock and Seele (2016) find that accusations of CSR being a PR exercise (most in line with the situation of overstating CSR) are not reflected in the organizational structure of most sustainable companies they studied.
Still, the question remains why CSR committees are mostly positively related to CSR decoupling. While in the past, some studies have found insignificant relationships between CSR committees and CSR performance and disclosure (Rodrigue, Magnan, & Cho, 2013; Michelon & Parbonetti, 2012), this is clearly not the case in this sample as I found CSR committees are significantly positively related to both. However, the influence on CSR performance is much stronger than that on CSR disclosure, similar to the case of gender diversity of the boards. Again, this could be the result of a conscious decision to decouple CSR, or because CSR committees have more influence on performance than on disclosure. Some evidence for the latter is provided by Ricart, Rodríguez & Sánchez (2005), who found that CSR committees were mainly responsible for advising the board of directors on CSR policies, as well as evaluating and reviewing these policies. Of course, both explanations could also be simultaneously at play.
The research question posed in the introduction, to what extent are gender diversity in
boards of directors, board independence, audit committee independence and CSR committee existence related to CSR decoupling?, can now be answered. This study found statistically
significant and positive relations to CSR decoupling for gender diversity, independence of the boards of directors and the existence of a CSR committee, contrary to my expectations. Audit committee independence is only found to have small negative effects on CSR decoupling on the environmental dimension and for firms with high levels of CSR performance. Furthermore, the hypothesis that board independence is negatively related to CSR decoupling holds true for decoupling on the environmental dimension, for firms with low CSR performance and for firms which overstate their CSR performance. Lastly, CSR committees are only negatively related to CSR decoupling for firms with high CSR performance.
5.2 Theoretical and managerial implications
The main theoretical contribution of this study is the investigation into the effect of corporate governance mechanisms on the phenomenon of CSR decoupling. To the best of my knowledge, this study is the first to study the effects which gender diversity of the board, board independence, audit committee independence and CSR committee independence have on CSR decoupling. A large worldwide sample size consisting of 17,115 firm years from 2009 to 2017 and from 40 countries was used to provide a good level of generalizability of the findings. The
