THE IMPACT OF
INSTITUTIONAL
OWNERSHIP ON CSR
DECOUPLING
Master thesis, MSc Accountancy and Controlling
June 29, 2019 Word count: 10686 Nawid Rassolyrad Student number: S2365812 Address: Damsterdiep 57 9711 SH, Groningen Phone number: +31 6 179 101 62 E-mail: n.rassolyrad@student.rug.nl1
Abstract
Purpose · The aim of this research paper is to investigate whether shareholders influence
CSR decoupling within organizations. I assume that a higher degree in institutional ownership will lead to less occurrence of CSR decoupling within firms. More recently, researchers have focused on studying the outcomes of CSR decoupling, yet limited research has been done with regards to the determinants of CSR decoupling.
Design/Methodology/Approach · To receive an answer on this question, the data of the
firms from the Bloomberg ESG index between 2009 and 2017 have been used to perform a fixed effects panel data regression-model. CSR decoupling is measured as the difference in scores between CSR disclosure and CSR performance. The institutional ownership percentage is lagged by 1 year, as well as all control variables used in this research.
Findings · The empirical evidence provides us with the information that institutional
ownership does not negatively influence CSR decoupling, thereby stating that our assumed relationship is not supported by the data.
Value · This paper contributes to existing literature by providing the absence of a
significant relationship. The estimates however indicate a positive relationship, which could be an interesting starting point for further research.
Keywords: CSR decoupling · Institutional ownership · Stakeholder-Agency theory ·
2
Table of Contents
1
Introduction
3
1.1 CSR
3
1.2 CSR decoupling
4
1.3 Concerns for the institutional owners
6
1.4 The discrepancy and the influence of the owner
6
2
Literature and background
7
2.1 Stakeholder-agency theory
8
2.2 Literature review
9
3
Methodology
14
3.1 Sample and data sources
14
3.2 Variables and measures
15
4
Results
22
4.1 Statistical tests
22
4.2 Institutional ownership and CSR decoupling
23
4.3 Additional analysis
27
4.4 Multicollinearity
29
5
Discussion and conclusion
31
5.1 Conclusion
31
5.2 Implications
33
3
1
Introduction
The introduction of this thesis will commence with defining what Corporate Social
Responsibility is and why it is of importance for both organizations and their environment. The second paragraph will introduce the concept of Corporate Social Responsibility Decoupling and its impact as described by several researchers. The introduction will end with explaining the concerns for the institutional owners of corporations.
1.1 CSR
In the past two decades,Corporate Social Responsibility (CSR) has progressively grown into an important concept for organizations and their stakeholders. As demonstrated by the research of Holt and Barkemeyer (2012), these last two decades also demarcated the starting point of
coverage of media on concepts related to CSR. These concepts of CSR include the environmental protection, community development, corporate governance practices, employee relations,
diversity practices, human rights, and product quality of a corporation resulting from its business practices (Dhaliwal et al., 2011).
The rapid increase for the interest of CSR has several reasons for both organizations and their stakeholders that support its importance. The first reason why CSR is important for a firm, is introduced by the seminal paper of Carroll (1979) by designating that CSR is an ethical
obligation of a firm towards society. Moreover, Carroll (1979: p. 500) indicates that firms must accept social responsibility, since “society has expectations of business over and above legal requirements.”
The second motivation for why CSR is of importance for companies is based on the principle of reputation. In prior studies, CSR activities have been shown to be able to function as a form of reputation-building (Fombrun & Shanley, 1990; Kim et al., 2012). Additionally, firms may seek to improve their reputation through activities that are deemed socially acceptable, or it may at least maintain its reputation by withholding from socially unacceptable activities (Kim et al., 2012). The increase or maintenance of reputation can subsequently increase credibility and trust
4
of their stakeholders, since CSR reporting shows a commitment to transparency (Dhaliwal et al., 2011).
The third reason why CSR is important to firms and their managers relates to the role of
monitoring. CSR reports often cover metrics such as monetary values of penalties, sanctions for non-compliance with laws and regulations, polluting emissions, and salary ratios between men and women by employee categories. Initially, these metrics appear to be strictly CSR-related. However, these metrics address in core the risks associated with the business practices (Christensen, 2015). Reporting of these metrics by measuring and analyzing them aids both managers and stakeholders in how these risks can be addressed and prevented. This form of dealing with the risks of business can in turn aid firms to keep ahead of possible scandals that could damage its core business and reputation (Christensen, 2015).
1.2 CSR decoupling
However, managers of companies have progressively received pressure from several stakeholders to behave socially responsible in conducting business (Campbell, 2007). In addition, amidst the demands of different stakeholders, there exist multiple competing and conflicting interests in the preferred direction of the firm (Jensen, 2002). To cope with these increasing pressures and to mitigate the conflicts, firms have created an approach of building an organizational façade while turning a blind eye to the state of practice inside the entity (Bromley & Powell, 2012; Harrison & Freeman, 1990; Meyer & Rowan; 1977:359). This phenomenon is often in literature described as CSR decoupling and many definitions are used to describe this process. Of these definitions, I identified two unique forms that best describe the process of CSR decoupling.
Several papers describe the CSR decoupling process as a firm’s symbolic strategy to overstate their CSR performance in their CSR reporting disclosures so that external stakeholders perceive an artificially heightened overall CSR (e.g. Delmas & Burbano, 2011; Tashman et al., 2019). Firms perform this practice by adopting CSR policies, but decouple these policies from their activities (Crilly et al., 2012). Put differently: these companies engage in external actions to a greater extent than that the firm could justify through their internal actions (Hawn & Ioannou, 2016). This first form is known as policy-practice decoupling (Bromley & Powell, 2012).
5
The second form of CSR decoupling is different and in some ways opposite in its process.
Internally, firms thoroughly implement the adopted CSR policies, but these implemented policies reveal no or few external actions (Bromley & Powell, 2012; Hawn & Ioannou, 2016). The lack of display is because of the missing or weak linkage between a firm’s implemented policies and the core activities and tasks of the organization (Bromley & Powell, 2012). Therefore, the internal policies are used as a means to reach a specific outcome, but these intended outcomes have little to no relation to the policies and thus are decoupled from the internal actions (Bromley & Powell, 2012).
These practices of CSR decoupling have reasons for their presence. As already stated, firms increasingly receive pressure from stakeholders for improvement of responsible practices
(Campbell, 2007; Waddock, 2008). Through CSR decoupling firms and managers try to cope and mitigate additional pressure, which eventually results in perceived legitimacy even though the actions are illegitimate (Crilly et al., 2012). Stakeholders are not able to distinguish the actions to be substantial or ceremonial, which results in a wrong perception of a firm’s CSR actions (Schons & Steinmeier, 2016).
Another reason for CSR decoupling that is argued by prior research is that management is well-intentioned, but actions are uncoordinated (Crilly et al., 2012). Due to diverse and possibly conflicting pressures of external stakeholders, management lacks an adequate response to
implement those policies that truly positively affect the firm’s core activities (Crilly et al., 2012). CSR decoupling, however, has the potential to produce adverse outcomes (e.g. Du, 2015; Jamali et al., 2017; Walker & Wan, 2012). To illustrate the potential hazardous consequences of CSR decoupling, prior studies have argued why the occurrence of CSR decoupling results in negative consequences for both the company and its stakeholders. The research of Bromley and Powell (2012) discussed three negative outcomes of CSR decoupling for organizations. Firms become internally more complex, are less stable, and displace their resources from core activities as a result of symbolical responding and adapting to a growing number of external stakeholders (Bromley & Powell, 2012). Research has also substantiated that decoupling affects legitimacy perceptions of external parties on firms in such a way that legitimacy is granted, while the
6
gestures made rest on solely symbolic and ceremonial practices (MacLean & Behnam, 2010; Westphal & Zajac, 1994). This study will focus on one specific external party, which is the institutional owner of the firm.
1.3 Concerns for the institutional owners
The above described practice of CSR decoupling thus can have severe consequences for firms. Summarizing, studies have indicated that both firms and management have the ability to improve their reputation through deceit and disseminating disinformation so as to present an
environmentally responsible public image (e.g. Laufer, 2003; Maclean & Behnam, 2010; Ramus & Montiel, 2005). CSR decoupling can be of high-risk for the institutional shareholders. As briefly brought up, Schons & Steinmeier (2016) found strong evidence that low-proximity stakeholders, i.e. stakeholders with a high physical distance between them and the firm, are unable to distinguish symbolic and substantive actions of firms.
This may have serious consequences for institutional investors, because this limited ability indicates that firms actually can benefit from building an organizational façade and solely disclose symbolic actions regarding their CSR. Moreover, investors are unable to perceive the true nature behind these symbolic actions, making it for firms easier to appear to fulfill stakeholders’ needs and maintain legitimacy without making drastic changes in their business processes (DiMaggio & Powell, 1983; Meyer & Rowan, 1977). So in this way, institutional owners can be hindered to fulfill their task as an effective governing body of management. Westphal and Zajac (1998: p. 128) even claim that “top managers can satisfy external demands for increased accountability to shareholders while avoiding unwanted compensation risk and loss of autonomy […] by bolstering such actions with socially legitimate language.”
1.4 The discrepancy and the influence of the owner
As the previous paragraph concisely indicates, there exists a discrepancy between the actual essence of reporting on CSR as a positive, value-adding practice (e.g. Di Giuli & Kostovetsky, 2014; Dhaliwal et al., 2011) and the symbolic usage of CSR reports to gain or maintain
7
Where preceding papers mostly focused on firms that act solely on a symbolic level, this study seeks to focus rather on the discrepancy within a firm that exists between substantive actions and symbolic gestures. Hence, CSR decoupling differs in that it regards both the actual performance of CSR and the disclosure of CSR. CSR decoupling can thus be perceived as the difference between the quality of the CSR disclosure and the actual CSR performance. In this way, this research is able to view CSR decoupling as the difference between the scope of the firm’s CSR reporting and the salience of the impacts that it reports (Tashman et al., 2019).
This paper will therefore analyze this effect with the following research question as guidance:
To what extent do institutional investors affect CSR decoupling?
To answer this question, I have developed five hypotheses to give a comprehensive view on the influences of institutional investors on CSR decoupling. These hypotheses will be stated in the next section, which will be logically derived from the existing literature.
The remainder of this research proceeds as follows. The next section develops the literature and background, and will describe the theory that will be used as the fundament of our expectations. These expectations will be stated through the usage of hypotheses. The ‘method’ section will elaborate on the specific manner in which the research is conducted, including a table of all the variables that have been used in this research (Table 2). Thereafter, the ‘results’ section has the goal to inclusively present all results found in this research. Finally, the main question will be answered, the limitations of the study will be specified, and the research will be complemented with theoretical and managerial implications in the ‘discussion and conclusion’ section.
2
Literature and background
To create a comprehensive view of the potential outcomes attributed CSR decoupling, existing literature regarding malpractices categorized as CSR decoupling and greenwashing was reviewed (table 1). I chose to make use of the stakeholder-agency theory introduced by Hill and Jones (1992) that incorporates both the stakeholder theory and agency theory. This chapter will
8
commence with a brief description of the main theory, which will be followed by the review of literature that will establish the foundation for the hypotheses, which function as the concluding part of this chapter.
2.1 Stakeholder-agency theory
The main theory of this thesis is an article of Hill and Jones (1992) that aids us in understanding firm’s strategic behavior through choices made by managers (agents), who are appointed as the decision makers in favor of the owners (principals). These managers, while being monitored by the owner(s) to indeed act in their interest, need to take stakeholder demands into the strategy’s account. “Managers are the only group of stakeholders who enter into a contractual relationship with all other stakeholders” (Hill & Jones, 1992: p. 134). Hill and Jones (1992) indicate with this statement that managers are not only the agent for the stockholders, but also function as the agent of all other stakeholders that have implicit or explicit contracts with the firm. The authors depict that the firm is a nexus of contracts and add that, however, not all stakeholders can be deemed as principals in the sense that they hire agents to perform some services on their behalf.
Nonetheless, Hill and Jones (1992) contend that similarities exist between the principal-agent relationship of agency theory and the stakeholder-manager relationship of the stakeholder theory. These two relationships are defined briefly to provide the reader the necessary knowledge to comprehend the sophisticated similarities.
The principal-agent relationship of agency theory is defined as “a contract under which one or more persons (the principal(s)) engage another person (the agent) to perform some service on their behalf which involves delegating some decision-making authority to the agent” (Jensen & Meckling, 1976: p. 308). Agency theory further posits that the interests of the agent and the owner diverge (Scott, 2014). These conflicts of interest are also called agency problems. In order to reach alignment between the interests of both principal and agent, contracts need to be made with appropriate incentives for the agent and the incurring of costs to monitor activities of the agent (Jensen & Meckling, 1976). The careful reader will note that this theory primarily is applied to the stockholder-manager relationship within firms, but Jensen and Meckling view that this relationship is just one of the nexus of contracts with numerous stakeholders that form a company (Hill & Jones, 1992).
9
The stakeholder-management relationship of stakeholder theory has its focus more on the impact of a firm. As Freeman (2010: p. 244) argues with his seminal book ‘Strategic Management: A Stakeholder Approach’: “Executives must seek to balance the relatively narrow interests of their individual firms with the broader concept of public interests”. The stakeholder theory indicates that not only the interests of shareholders need to be considered by the firm, but also the needs of other parties affected by the operations of the firm such as employees, state communities,
competitors, consumers and others (Freeman, 1984). Moreover, March and Simon (1958) state that these stakeholders all supply a corporation with critical resources, but in exchange expect that their interests will be fulfilled by the firm (Hill & Jones, 1992). Hence, to achieve these relationships executives are appointed as the spokesman of corporations for their stakeholders and as the builders of alliances among their internal and external stakeholders (Freeman, 1984). The main theory of my argument derives from a combination of both theories discussed. Both relationships contain implicit or explicit contracts that are meant to align the interest between both parties in the two theories. Therefore, these indications result in the notion that the principal-agent theory is a specific subset of the more broad and general set of stakeholder-agent relationships that create a modern organization (Hill and Jones, 1992).
2.2 Literature review
Where agency theory solely falls short in explaining why and/or how social targets should be included in strategic goals (Walls et al., 2012) and where Freeman’s stakeholder concept as a theory is largely implicit (Jones, 1992), this combined concept constructs a sound paradigm that explains specific aspects of the strategic behavior of the firm (Flak & Rose, 2005). To indicate, the stakeholder-agency theory improves the way of describing and explaining the alignment of incentives and the institutional structures that have evolved, which serve the function of
monitoring and enforcing the terms in the implicit and explicit contracts between managers and stakeholders (Flak & Rose, 2005; Hill & Jones, 1992).
Where agency theory and stakeholder theory have differences in values when it comes to managers spending corporate funds on anything that does not aid in adding profit (Munilla & Miles, 2005), stakeholder-agency theory approaches this by postulating that management needs
10
to find the approach that maximizes their utility function between the divergent claims of stakeholders and stockholders (Hill & Jones, 1992). In the case of CSR, research shows
inconsistent results on whether it contributes to firms financial performance (e.g. Dhaliwal et al., 2011; Di Giuli & Kostovetsky, 2014), but several stakeholders do value firms with adequate levels of CSR performance and disclosure (Christensen, 2016; Dowling & Pfeffer, 1975; Gray et al., 1995; Guthrie & Parker, 1989). So the manager, who has the control over the decision-making apparatus of the firm and is at the nexus of contracts between the firm and stakeholders, needs to navigate between creating value for the shareholders as the call to engage in CSR, which is not proven to be of financial value.
As previously discussed, one strategy used by companies to respond to stakeholders’ expectations is that of decoupling policy from practice (Crilly et al., 2012). Through usage of symbolic
actions, firms are able to construct and maintain the legitimacy of their stakeholders, including shareholders (Dowling & Pfeffer, 1975; Elsbach, 1994; Neu et al., 1998; Schons & Steinmeier, 2016). An important assertion, however, is that the stakeholder-agency theory argues that the lack of proper monitoring is specifically the case for diffused stakeholders (Hill & Jones, 1992). With this assertion, the theory implies that possibly one stakeholder or entity with sufficient funds and power could function as a proper monitor of the firm and its managers.Moreover, the
management of firms may even “be put on the defensive by increases in stakeholder power” as a result of the increase in the amount of stock held by institutions (Hill & Jones, 1992: p. 151). Thus, following the stakeholder-agency theory of Hill and Jones (1992), the conclusion can be made that the theory fundamentally would discourage CSR decoupling. Based on the theory and the arguments presented above, I hypothesize the following relationship:
Hypothesis 1: Institutional investors negatively influence the effect of CSR decoupling within firms.
Hill and Jones thus advocate in their research that one entity with sufficient power could function more appropriate as a monitor of management. One form of institutions that could possess this monitoring ability, is a governmental institution. Although in the literature review (table 1) the study of Luo et al. (2014) showed that conflicting governmental demands in China cause firms to engage in CSR decoupling, governments and their pressure have been found to enhance CSR
11
performance of firms (Chaudhary et al., 2011; Beddewela & Fairbrass, 2015). Beddewela and Fairbrass (2015) even argue that having a good relationship with governmental institutions is of major importance for the viability of a corporation. This argument is in line with the assertion of Hill and Jones (1992) that, if governments have power over firms in terms of the need of firms to have a solid relationship, they could act as a proper monitor. Therefore, I hypothesize the second relationship as follows:
Hypothesis 2: Governmental institutions negatively influence the effect of CSR decoupling within firms.
The next specific form of institutions that I argue to have an influence on CSR decoupling of firms, is the pension fund. The study of Hong and Kacperczyk (2009) found that pension funds are a type of institutions that are exposed to social norms and public scrutiny. The effect of this exposure, is that pension funds invest more in firms that have superior CSR performance (Hong & Kacperczyk, 2009). In addition, the authors provide evidence that increases in the
shareholdings of pension funds are related to consequential improvements in CSR (Cahan et al., 2017). This means that the assertion of the stakeholder-agency theory that less diffused
stakeholders or a more concentrated stakeholder, results in better monitoring (Hill and Jones, 1992). If the monitoring indeed improves, I argue that this will be displayed in less occurrence of CSR decoupling within firms. Hence, I hypothesize the following relationship:
Hypothesis 3: Pension funds negatively influence the effect of CSR decoupling within firms.
This section will describe two hypotheses between two differing types of institutions and their influence on CSR decoupling. Table 1 shows the research of Walker and Wan (2012), which revealed that CSR decoupling has a negative effect on a firm’s financial performance. In addition, table 1 showed as well that the paper of Hawn and Ioannou (2016) found that CSR decoupling has a negative impact on a firm’s market value and thus it’s owners. A significant assertion made in the theory of Hill and Jones (1992) is that, due to the fact of an intense rise of financial
12
exert their influence over management actions. Next to these financial institutions, other institutions have evolved that try to exploit the profit opportunities of these companies (Hill & Jones, 1992). These institutions heavily rely on CSR reports, because it can increase the transparency of the firm in terms of risk management and long-term performance of the firm (Christensen et al., 2015; Dhaliwal et al., 2011). Based on the theory and mentioned papers, I therefore hypothesize the following two associations:
Hypothesis 4: Investment corporations negatively influence the effect of CSR decoupling within firms.
Hypothesis 5: Other institutions, not characterized as one of the other categories, negatively influence the effect of CSR decoupling.
13 Study Relevant contribution
Data source of the relevant
contribution Theory Applied
Samp le firms Country Cover age years Crilly et al.
(2012) The effects of conflicting stakeholder expectations on CSR decoupling
Innovest, e-Capital Partners,
Vigeo No specific theory 17 Multiple countries 1 year Hawn and
Ioannou
(2016) The negative effect of CSR decoupling on a firm's market value ASSET4 Neo-institutional theory 1492 Multiple countries years 6 Schons and
Steinmeier
(2016) The vulnerability of low-proximity stakeholders on greenwashing ASSET4
Neo-institutional theory and
stakeholder theory 220 Multiple countries years 9
Du (2015)
The negative linkage between
greenwashing and cumulative abnormal
returns Manual coding data No specific theory 14 China years 2 Tashman et
al. (2019)
Pervasiveness of of institutional voids in a firm's home country as a driver of CSR
decoupling Morgan Stanley Capital Index Neo-institutional theory 93 Multiple countries years 8 Graafland
and Smid
(2018) The quality of CSR reports as a determinant of CSR decoupling Sustainalytics Institutional theory 397 Multiple countries 1 year Walker and
Wan (2012) The negative impact of greenwashing on financial performance Manual coding data No specific theory 103 Canada 1 year Ramus and
Montiel
(2005) Evidence of differences in greenwashing within industries CEP Database + survey Institutional theory 188 Multiple countries years 6 MacLean
and Behnem
(2010) Gaining legitimacy through decoupling compliance programs Interviews
Legitimacy theory and
Neo-institutional theory 1 US 1 year Laufer
(2003) Managing reputation of corporations through greenwashing Theoretical study Legitimacy theory n/a n/a n/a Jain (2017)
Enhancing social license to operate through signalling a symbolic
multi-stakeholder image Manual coding data Signaling theory 30 US, Germany
and India years 4 Delmas and
Burbano
(2011) The drivers behind greenwashing Theoretical study Institutional theory n/a n/a n/a Marquis et
al. (2016) The negative effect of highly globalized firms on greenwashing Trucost Institutional theory 4750 Multiple countries years 4 Jamali et al.
(2017) An overview of dynamics used by firms that perform CSR decoupling Survey + interviews Institutional theory 37
Pakistan, India and
China years 5 Parguel et al.
(2011) The impact of sustainability ratings on greenwashing Web survey Attribution theory n/a France 1 year De Vries et
al. (2015)
The negative effect of communicating environmental motives on greenwashing
suspicion Experimental research method No specific theory n/a Netherlands 1 year Luo et al.
(2014)
The negative influence of conflicting governmental demands on CSR
decoupling Rankins CSR ratings No specific theory 584 China years 3 Christmann
(2004)
The effect of stakeholder pressures on global standardization of environmental
policies Survey Resource dependency theory 98 US 1 year Christmann
and Taylor
(2006) CSR decoupling through obtaining symbolic standard certifications KLD Database + survey No specific theory 172 China 1 year Marano et al.
(2017) Decoupling CSR reports as a means to overcome barriers of legitimacy
GRI's Sustainability Disclosure
Database Institutional theory 157 Multiple countries years 8
14
3
Methodology
This chapter begins with the description of the data sources used and reports the total sample size for this study. Next, the measurement of all the variables that are used will be explained and finally the descriptive tables are presented, which will as well be explained in chapter four.
3.1 Sample and data sources
The hypotheses that in the previous chapter derived from the stakeholder-agency theory were tested on the firms listed in the BESGPRO Index. The sample consists of 5444 firms listed within the BESGPRO Index of Bloomberg, which were measured for 9 years between 2009 and 2017. Bloomberg developed this index together with a leading provider of environmental, social and governance data, MSCI ESG Research, presenting the ESG data weighted-scores on both firm’s CSR disclosure and CSR performance quality (Bloomberg, n.d.). Bloomberg and MSCI ESG Research evaluated firms’ CSR disclosure and performance in four standardized ranges: Overall CSR disclosure and performance, Environmental disclosure and performance, Social disclosure and performance, and Governance disclosure and performance. This index provides us with two possibilities, which will be discussed in this paragraph shortly and will be argued in more depth in the next paragraph. The first possibility given, is that we can differentiate between three forms of CSR decoupling: CSR decoupling as the absolute difference between the CSR performance score and CSR disclosure score, the actual difference between CSR performance score and CSR disclosure score, and the difference when comparing the CSR disclosure score with the CSR performance score. Secondly, this index provides us with the possibility to analyze and differentiate between the four ranges of the scores and particularly compare the difference between them over several years to determine possible influences or effects, making the index members a suitable population for this research.
Next to the Bloomberg database as a data source, I relied on the Compustat Global (Standard and Poor’s, 2019) database to deliver the figures on the six control variables in this research. All six control variables, Company size, Financial return, Leverage, Tangibility, Tobin’s q, Capital
intensity, and Board size were collected through the global data source and merged with the
15
The third and final data source used in this research was the Asset4 database of Refinitiv (formerly Thomson Reuters). Similar to the former data source, Refinitiv provides global data regarding firm ownership. One of the strong advantages of Refinitiv is the categorization within the ownership data, providing the ability to collect and measure the data for each category
separately, as well as a measuring the effect of all categories composed as a whole. Together with their depth of historical data, this data source finalizes our dataset for the years from 2009 to 2017. The final sample after merging all data sources consists of 37,690 firm-year observations and covers 5417 firms from 106 countries during the period 2009-2017.
3.2 Variables and measures
This second paragraph will elaborate more on our variables and the way of measurement. Firstly, the dependent variable is discussed. Afterwards, the independent variables are explained in more detail. Thirdly, the control variables used in this research are conferred. This chapter will end with a table of all used variables and measures (table 2), a table with the descriptive statistics of this research (table 3), and three tables that will be discussed in chapter four (table 4, table 5, and table 6).
Dependent variable
As mentioned briefly in the previous paragraph, the Bloomberg data source provides the possibility to measure the difference between a firms’ CSR disclosure and performance standardized scores. In this research, I differentiate between three different forms of CSR
decoupling, which will be discussed in this paragraph. Firms receive a score from 0-100 on both their CSR performance and CSR disclosure efforts on the next four dimensions: overall,
environmental, social, and governance. These four scores are added up with an equal weight and eventually divided by four for both CSR performance and CSR disclosure.
The first way of measuring CSR decoupling is known as Absolute CSR decoupling. This is done by subtracting the CSR disclosure score from the CSR performance score and notating the non-negative value. By taking the absolute value, the study of Bromley and Powell (2012) is used, which argues that CSR decoupling consists not only of policy-practice decoupling (i.e. routinely violating or unimplementation of rules), but also of means-end decoupling (i.e. opaque linkage between formal policies and the intended outcome). The higher the score of CSR decoupling, the
16
higher is the existence of misalignment between a firms’ actual performance regarding CSR and the quality of CSR disclosure in either way. Validation through this manner of measuring the dependent variable, is proven by the study of Tashman et al. (2019) who consequently built upon the study of Marquis et al. (2016).
The second method used in this paper is by measuring the positive gap between the CSR performance score and CSR disclosure score, called Positive CSR decoupling. This is measured by subtracting the CSR disclosure score from the CSR performance score. By doing so, the gap is included of firms that implement the CSR policies thoroughly, but lack a well-defined
relationship to their core tasks, which is called means-ends decoupling (Bromley & Powell, 2012). Note that, although defined as Positive CSR decoupling, this is not considered to have any positive effects for firms, since this form does not aid firms in achieving their intended outcomes (Bromley & Powell, 2012). The purpose of this is to view the effects of our independent variables on solely this form of CSR decoupling.
The third differentiation made, is the subtraction of the CSR performance score from the CSR disclosure score. Through this measure, this study captures the effect of institutional ownership on firms that do not implement those rules that firms have disclosed in their CSR reports, which is otherwise known as policy-practice decoupling in the study of Bromley and Powell (2012). Using these three forms of CSR decoupling, it will give us a thorough understanding of the effects of our independent variables, which will be elaborated on in the next part, on the complete CSR decoupling measure. To reduce the unfavorable effects of outliers, the logarithm is taken of
Absolute CSR decoupling, Positive CSR decoupling, and Negative CSR decoupling. Also, all
three variables are winsorized at the first and ninety-ninth percentiles. Independent variables
Prior studies have mostly dealt with the impact of CSR on shareholders as one homogenous entity (e.g. Albuquerque et al., 2017; Godfrey et al., 2009; Masulis & Reza, 2015). In this research, I examine the impact of a specific type of shareholder: the institutional investor. The reasoning behind the choice of institutional ownership is because of the fact that these type of shareholders own and vote on the bulk of the world’s equity capital (Dyck et al., 2018). To
17
illustrate the presence of this group in the equity market, figure 1 depicts the percentages owned by institutional investors in several U.S. indices. According to a study done by P&I Online in 2017, institutions own about 80% of the equity market capital in the United States.
Figure 1
Institutional ownership is measured in this study as the percentage of shares owned by
institutional investors. The Refinitiv data source provides us with these discrete values for each company in the sample, through the assortment of several ownership groups: Government holdings, investment corporations holdings, pension funds holdings, and percentage owned by other holdings. After accumulation of these percentages, we can observe the percentage owned by all institutional investors. To effectively assess all hypotheses, which are presented in chapter two, the percentages owned by Governmental institutions, Investment corporations, Pension
funds, and Other institutions are also separately measured with regards to their relationship with
the dependent variable. In this research Institutional ownership, Governmental institutions,
18
institutional investors are only able to influence decision making of firms when they truly own shares of the corporation.
Control variables
To control for the unreliability of results, we make use of several sorts of firm-level control variables as mentioned in paragraph 3.1. The first control variable Firm size is measured through taking the logarithm of the average of the total assets and total revenues owned by the
corporation. Multiple studies have shown that indeed firm size has a significant relationship with CSR performance (e.g. Dyck et al., 2018; Graafland & Smid, 2018; Lepoutre & Heene, 2006). In particular, the smaller the firm the less the firm is able to implement CSR programs in a fitting manner (Lepoutre & Heene, 2006).
The second firm-level control variable used in this thesis is Financial return. Previously executed research of Delmas & Burbano (2011) has shown that the profitability of an organization, has influences on its overall strategy. Even more so, Delmas and Burbano (2011) argued that firms with higher profitability, are positively related to the practice of decoupling. The reason behind this statement is that they contended that the firms with higher profits are better able to endure and resist bottom line shocks from reputational damage when being alleged of greenwashing. To measure this control variable, we make use of the measurement introduced by Griffin and Mahon (1997) by taking the average of the combination of three measures for financial performance: the return on assets, the return on equity and the return on capital.
The third and fourth firm-level control variable, Leverage and Tangibility, are suggested by the study of Hong et al. (2012) to measure credit constraints of a firm, which can prohibit extra liquidity of a firm. Firms with financial constraints, including credit constraints, have been found to dismiss any CSR activities (Campbell, 2007; Chan et al., 2017). In line with the study of Dyck et al. (2018), I measure Leverage as the ratio of the total debt to the total assets and Tangibility as the property, plant and equipment to the total assets.
Drawing on previous literature, I include the fifth control variable Tobin’s q, which is measured as the market capitalization of equity plus total debt divided by total assets (Dyck et al., 2018).
Tobin’s q is found by literature as a proxy for firm value and is as well found to be positively
19
The sixth firm-level control variable used is Capital intensity. In their study, Russo and Fouts (1997) identified that this control variable is able to affect how firms arrange assets to improve their CSR performance. This variable is measured as the ratio of assets to sales, following the research done by Tashman et al. (2019).
The seventh and final firm-level control variable utilized is Board size. The paper of de Villiers et al. (2011) showed that firms with larger boards are more likely to enhance environmental
performance. Thus, to control for this effect, the Board size variable is included in our research. This variable is measured as the number of board members in the firm (de Villiers et al., 2011). To reduce possible biases and increase robustness of the estimates, usage of lagged controls by one year is applied for the independent variables and all of the control variables. In addition, the control variables have been winsorized at the first and ninety-ninth percentiles to reduce the unintentional effects of outliers.
20
Table 2 Variables and measurement of data
Variable Name Variable type Value Measurement
CSR Decoupling Dependent Continuous CSR disclosure score minus CSR performance score Institutional investor Independent Discrete Percentage of shares owned by
institutional investors Governmental
institutions
Independent Discrete Percentage of shares owned by governments
Investment corporations Independent Discrete Percentage of shares owned by investment corporations Pension funds Independent Discrete Percentage of shares owned by
pension funds
Other institutions Independent Discrete Percentage of shares owned by other institutions
Size of firm Control Continuous The logarithm of the average of total revenues & total assets of the firm
Financial return Control Continuous Combination of return on assets, return on equity & return on capital
Leverage Control Continuous The total debt divided by the total assets
Tangibility Control Continuous Property, plant and equipment divided by the total assets Tobin’s Q Control Continuous Market capitalization of equity
plus total debt divided by total assets
Capital intensity Control Continuous Ratio of the current assets to the sales
Board size Control Continuous The number of members within
21
Table 3 Descriptive statistics
Variable Name Observations (N) Mean (µ) Standard deviation (σ) Min Max
CSR Decoupling 37,690 108.17 70.17892 0 359.61 Institutional investor 37,690 10.79% 14.73% 0% 100% Governmental institutions 37,690 1.93% 9.70% 0% 100% Investment corporations 37,690 7,23% 9.66% 0% 100% Pension funds 37,690 1,32% 2.30% 0% 90% Other institutions 37,690 0,31% 7.59% 0% 100% Size of firm 37,690 6.92 1.22 2.51 11.56 Financial return 37,690 4.56 127.34 -21777.57 3472.43 Leverage 37,690 0.25 0.21 0 7.95 Tangibility 37,690 0.29 0.27 -0.012 1.91 Tobin’s Q 37,690 1.43 1.30 0.091 7.79 Capital intensity 37,690 84.10 13346.61 0.00 2589834 Board size 37,690 6.75 5.48 0 20
Table 4 Breusch-Pagan test for heteroscedasticity
Breusch-Pagan / Cook-Weisberg test for heteroscedasticity Ho: Constant variance
Variables: fitted values of CSR Decoupling chi2(1) = 3.75
22
Table 5 Breusch-Pagan LM test for random effects vs. OLS
Variable Var sd = sqrt(Var)
CSR decoupling 0.8126 0.9014 E 0.6286 0.7929 U 0.1827 0.4275 Test: Var(u) = 0 Chibar2(01) 5384.06 Prob > chibar2 0.0000
Table 6 Hausman test for fixed effects vs. random effects model
Coefficient(fixed) Coefficient(random) Difference Sqrt Institutional investor 0.0011 0.0007 0.0004 0.0005 Governmental institution -0.0000 0.0009 -0.0009 0.0018 Investment corporation 0.0004 0.0001 0.0003 0.0004 Pension fund 0.0051 0.0015 0.0036 0.0026 Other institutions 0.0039 0.0011 0.0028 0.0015 Size of firm 0.0014 0.0526 -0.0512 0.0300 Financial return 0.0002 0.0019 -0.0017 0.0003 Leverage -0.0521 -0.0348 -0.0173 0.0466 Tangibility -0.1061 0.0801 -0.1860 0.0623 Tobin’s q -0.0264 -0.0167 -0.0097 0.0055 Capital intensity 0.0000 0.0000 0.0000 0.0000 Board size 0.0014 0.0063 -0.0048 0.0003
H0: Difference in coefficients is not systematic Chi2(12)= 265.22
Prob > chi2= 0.0000
4
Results
This chapter will start the first paragraph with the statistical foundations for the choice of the kind of regression. In the sequential paragraph, the results of the statistical analysis will be discussed for all hypotheses and in the final paragraph of this chapter the executed additional analysis will be explained and presented.
4.1 Statistical tests
As briefly mentioned above, statistical tests need to be executed to determine which statistical analysis seems fit. To perform these statistical tests of significance, the data is firstly examined
23
for the presence of heteroscedasticity in the attained data. Therefore, the Breusch-Pagan test for heteroscedasticity was conducted between the dependent and independent variables. Table 4 shows the result that the p-value of this test is under the margin of 0.05, meaning that H0 of constant variance is rejected. Hence, our sample is in first instance under the influence of
heteroscedasticity, which needs a correction to validate the results obtained through the statistical tests of significance.
Another test performed before estimation, is the Breusch-Pagan LaGrange Multiplier test presented in table 5. This test is used to determine whether we need to perform a pooled OLS regression or a random effects regression-model. The test provides us with a p-value of 0.0000, indicating high significance and thereby contributing us with the knowledge that the usage of an ordinary least squares regression is not suitable for the dataset.
To determine which test in fact is needed when regarding the options for a fixed effects model and a random effects model, we make use of the Hausman test. The results of the Hausman test are displayed in table 6 and indicate with a significant p-value for the chi-score that the variables differ on a systematic basis. This result consequently provides the information that we need to perform a fixed effects model. Furthermore, to correct for the existence of heteroscedasticity I include the robust standard errors estimator in the regression (Hoechle, n.d.).
4.2 Institutional ownership and CSR decoupling
The main goal of our tests is to examine the relationship between institutional ownership and CSR decoupling. To estimate this relationship, this research makes use of the fixed effects regression model for two reasons next to the above provided test in tables 5 and 6. Primarily, since the Hausman test provided this research with the fact that time-invariant unobserved firm characteristics exist, the application of a firm fixed effects model is suited to control for that (Gormley & Matsa, 2014). Secondarily, since the panel data consist of a relatively short time series but available for a high amount of cross-sectional units, the choice for a fixed effects approach aids the research by allowing differences in individual effects (Holtz-Eakin et al., 1988).
As shown in chapter three, the dependent variable CSR decoupling is measured in three ways:
Absolute CSR decoupling, Positive CSR decoupling, and Negative CSR decoupling. Also shown
24
institutional investors: Governmental institutions, Investment corporations, Pension funds and
Other institutions. To test the hypotheses related to the four groups of institutional investors and
the one variable that accumulates all institutional groups, this research makes use of the following regression models:
Model 1
Log(Absolute CSR decouplingit)= α+ β1Institutional investorit-1+ β2Size of firmit-1+
β3Financial returnit-1+β4Leverageit-1+ β5Tangibilityit-1+β6Tobin's qit-1+β7Capital intensityit-1+ β8Board size+ ϵi
Model 2
Log(Positive CSR decouplingit)= α+ β1Institutional investorit-1+ β2Size of firmit-1+
β3Financial returnit-1+β4Leverageit-1+ β5Tangibilityit-1+β6Tobin's qit-1+β7Capital intensityit-1+ β8Board size+ ϵi
Model 3
Log(Negative CSR decouplingit)= α+ β1Institutional investorit-1+ β2Size of firmit-1+
β3Financial returnit-1+β4Leverageit-1+ β5Tangibilityit-1+β6Tobin's qit-1+β7Capital intensityit-1+ β8Board size+ ϵi
where the dependent variable is, as explained in chapter three, the log of the absolute difference between the CSR disclosure score and the CSR performance score. For clarity purposes
β1Institutional investorit-1 is used in the models, but all four subgroups of institutional investors have been as well tested separately, as shown in table 7. Furthermore, in this model the α stands for the constant for the entire sample, where β1 is our lagged independent variable, together with β2 to β7 as our control variables. The εi term stands for the random error component, which measures the difference between the average CSR decoupling difference at year i and the average
CSR decoupling difference in the entire sample.
Table 7 contains the result for the main test of our hypotheses, which forecasted that firms with relatively more Institutional ownership(H1), Governmental institutions(H2), Investment
corporations(H3), Pension funds(H4), and Other institutions(H5) cause firms to engage less in CSR decoupling.
The first result in table 7 indicates that a positive, non-significant relation exists between (lagged) institutional investor and all three forms of CSR decoupling of a firm (Model 1: β = 0.0007, p =
25
0.073; Model 2: β = 0.0012, p = 0.056; Model 3: β = 0.0005, p = 0.355). causing that our first hypothesis is rejected. To illustrate, the coefficient indicates that a one standard deviation increase in Institutional ownership is related to a 0.0007 standard deviation increase in Absolute
CSR decoupling. The p-value indicates a strong and positive relation in the first model, but not
significant at the 5% level and is not supported through the other forms of CSR decoupling. The second result in table 7 yields interesting figures. In both the relationship between
Governmental institutions and Absolute CSR decoupling(Model 1) as well as between
Governmental institutions and Positive CSR decoupling(Model 2) there have been found positive,
significant relationships (β = 0.0014, p = 0.048; β = 0.0026, p = 0.001). Between Governmental
institutions and Negative CSR decoupling a positive, insignificant relationship has been found (β
= 0.0005, p = 0.355). Although some significant relations have been found, the second hypothesis in this thesis is rejected. The reason for this is that the hypothesis assumed a negative
relationship, where a positive relationship has been found.
The third hypothesis suggested a negative association between pension funds and CSR
decoupling. When observing the statistical analysis in table 7, all values indicate an insignificant relationship between Pension funds and the three forms of CSR decoupling (Model 1: β = 0.0016, p = 0.532; Model 2: -0.0019, p = 0.507; Model 3: β = 0.0055, p = 0.176). These results show a positive correlation between Pension funds and Absolute CSR decoupling and Negative CSR
decoupling, but a negative one between Pension funds and Positive CSR decoupling. All are,
however, insignificantly associated. As a result, also the third hypothesis is rejected.
With regards to the linkage between Investment corporations and three forms of CSR decoupling, the results indicate a negative, highly insignificant linkage (Model 1: β = -0.0000, p = 0.987; Model 2: β = -0.0001, p = 0.894; Model 3: β = -0.0001, p=0.869). Therefore, the third hypothesis is rejected.
The fifth and last hypothesis puts forward a negative relation between Other institutions and the three forms of CSR decoupling. The results, however, indicate a positive relationship, with a significant relation between Other institutions and Negative CSR decoupling ( β = 0.0022, p = 0.04). The other two models show strongly insignificant association (Model 1: β = 0.0012, p = 0.178; Model 2: β = 0.0004, p = 0.789), which form insufficient grounds to conclude an overall
26
significant effect. Because of positive nature of the relationship, the last hypothesis is as well rejected.
Table 7 contains also the results for the seven control variables, Size of firm, Financial return,
Leverage, Tangibility, Tobin’s q, Capital intensity, and Board size. Firstly, the results for the
relationship between the (lagged) Size of firm, Financial return, and Tobin’s Q control variables and CSR decoupling indicate a highly significant relationship with regards to both Model 1 and Model 2, showing that indeed our control variable has validated our research. Also the
relationship between the (lagged) Tangibility control variable and Absolute CSR decoupling and
Negative CSR decoupling indicate a positive and significant relationship, validating that this
variable is in accordance with prior literature. Board Sizeand Capital intensity are the only two control variables that have a significant relationship with the second method, Positive CSR
decoupling. Both control variables have additionally a highly significant relationship with the
first model, with Capital intensity to have as well a highly significant relationship with the last model. Leverage, however, is the only variable that shows a negative insignificant relation with all forms of CSR decoupling (β = -0.04, p=0.40).
27
Table 7 Fixed effects-regression GLS regression
Absolute CSR decoupling (1) Positive CSR decoupling (2) Negative CSR decoupling (3) The institutional investor 0.0007 (0.073) 0.0012 (0.056) 0.0005 (0.355) Governmental institutions 0.0014 (0.048)* 0.0026 (0.001)** 0.0005 (0.646) Investment corporations -0.0000 (0.987) -0.0001 (0.894) -0.0001 (0.869) Pension funds 0.0016 (0.532) -0.0019 (0.507) 0.0055 (0.176) Other institutions 0.0012 (0.178) 0.0004 (0.789) 0.0022 (0.040)* Size of firm 0.0427 (0.000)*** 0.0046 (0.607) 0.0976 (0.000)*** Financial return 0.0016 (0.002)** 0.0009 (0.267) 0.0016 (0.010)** Leverage -0.0431 (0.222) -0.0649 (0.242) 0.0051 (0.898) Tangibility 0.0586 (0.035)* 0.0091 (0.825) 0.0948 (0.003)** Tobin’s Q -0.0209 (0.000)*** -0.0156 (0.056) -0.0233 (0.000)*** Capital intensity -0.0045 (0.000)*** -0.0039 (0.001)** -0.0056 (0.000)*** Board size 0.0065 (0.000)*** 0.0104 (0.000)*** -0.0007 (0.530) Constant 4.0768 4.3492 3.638 Probability > Chi2 0.000 0.000 0.000 ρ 0.209 0.405 0.254
Note: *Coefficient is significant at the 0.05 level; **Coefficient is significant at the 0.01 level; ***Coefficient is significant at the 0,001 level.
4.3 Additional analysis
Now that the main analysis is presented, this research moves on to the additional analysis. To check for the robustness of the regression results, four additional tests are conducted. Table 8 presents those different models and relations. As mentioned in chapter three of this research, the CSR decoupling variable consisted of four different categories in total. For reminding purposes those four dimensions were: the difference between the overall CSR performance and disclosure,
28
the difference between environmental performance and disclosure, the difference between social performance and disclosure, and the difference between the governance performance and
disclosure. To understand if the influence of institutional ownership does affect decoupling within the dimensions, the relationships are separately examined. In this way, this study clarifies if the degree of institutional ownership has any impact on the dimensions of CSR.
Table 8, however, shows that institutional investors on neither of the four dimensions have a significant impact. Moreover, the p values of the dimensions separately have less of a relationship than the main regression of the four dimensions combined. The next paragraph in this chapter shows the tests performed to check for any multicollinearity issues.
Table 8 Additional analysis
Variable (1) (2) (3) (4) Coefficient (p value) Institutional investor 0.0142 (0.646) 0.0078 (0.816) 0.0129 (0.666) -0.0062 (0.774) Size of firm 2.3215 (0.000)*** 5.4742 (0.000)*** 3.3714 (0.000)*** -0.3296 (0.223) Financial return 0.0954 (0.020)* 0.0589 (0.150) 0.0851 (0.024)* 0.0009 (0.978) Leverage 0.4021 (0.872) -1.1598 (0.667) 0.6132 (0.790) 0.2604 (0.886) Tangibility 3.4048 (0.061) 4.5887 (0.018)* 1.6851 (0.313) 4.0661 (0.003)** Tobin’s Q -1.1208 (0.004)** -1.0383 (0.013)* -0.9268 (0.010)* -0.1857 (0.508) Capital intensity -0.1738 (0.000)*** -0.1910 (0.000)*** -0.1600 (0.000)*** -0.0458 (0.188) Board size 1.2637 (0.000)*** 1.3110 (0.000)*** 0.9770 (0.000)*** 0.3550 (0.000)*** Constant 24.5147 7.7126 16.2502 33.1738 Probability>Chi2 0.000 0.000 0.000 0.027 ρ 0.524 0.539 0.449 0.206
Note: *Coefficient is significant at the 0.05 level; **Coefficient is significant at the 0.01 level; ***Coefficient is significant at the 0,001 level.
29
4.4 Multicollinearity
Multicollinearity is the phenomenon that involves several variables which independently interact with each other (Blumberg et al., 2014). Multicollinearity exists if there are high
inter-correlations between the independent variables used in the analysis. These inter-inter-correlations between variables could possibly influence the regressions in the main analysis in terms of confidentiality of the results. Therefore, the Pearson correlation coefficient test has been performed, which shows the associations between the independent variable. To assess whether multicollinearity is present, the standard of Blumberg et al. (2014) is used which assumes multicollinearity if the Pearson correlation coefficient is higher than 0.8 or lower than -0.8.
In addition, the variance inflation factor (VIF) is provided, which also reflects the presence and/or absence of multicollinearity between the independent variables. The study of Hair et al. (1995) is applied, that argues the maximum level of the 10 for the VIF.
Table 9 presents the correlation matrix of the variables used in the main statistical analysis together with the variation inflation factor of each variable separately. The table shows that both boundaries (Blumberg et al., 2014 ; Hair et al., 1995) applied, are not crossed. Thus,
30
Table 9 Multicollinearity analysis
Note: Correlations with p-values less than 0.01 are indicated in bold. The matrix comprises all numbers of observations, equals 37,690 Institutional
investor Governmental institution Investment corporation Pension fund Other institutions firm Size of Financial return Leverage Tangibility Tobin’s q Capital intensity Board size VIF
Institutional investor 1 6.81 Governmental institution 0.571 1 2.94 Investment corporation 0.394 -0.104 1 3.35 Pension fund 0.108 0.112 -0.044 1 1.10 Other institutions 0.423 -0.025 -0.081 0.004 1 2.09 Size of firm -0.073 0.159 -0.352 0.111 0.054 1 6.81 Financial return -0.034 0.004 -0.064 -0.001 0.017 0.110 1 1.38 Leverage -0.006 0.020 0.034 -0.007 0.006 0.080 -0.068 1 2.86 Tangibility -0.024 0.048 -0.075 0.091 -0.012 0.017 -0.056 0.310 1 2.36 Tobin’s q 0.021 -0.077 0.128 -0.034 -0.026 -0.274 0.195 -0.108 -0.058 1 2.19 Capital intensity 0.012 0.051 -0.030 -0.005 0.002 -0.055 -0.208 -0.042 -0.145 -0.183 1 1.39 Board size -0.027 0.049 -0.102 -0.004 0.018 0.200 0.021 0.023 0.021 -0.060 0.013 1 2.63 Mean VIF 2.99
31
5
Discussion and conclusion
This section will provide the conclusions in first place. Hereafter, the implications of the results will be described in the second paragraph and the limitations will be discussed in the last paragraph.
5.1 Conclusion
Using a comprehensive sample found in the Bloomberg ESG index that holds companies across more than 100 countries, this thesis investigated the relationship between institutional ownership and CSR decoupling as a result of the literature review in research in the field of accountancy. In line with existing literature, I adopted three manners of measurement for the dependent variable, CSR decoupling. Provided by the study of Tashman et al. (2019), CSR decoupling was taken as the absolute difference between the standardized values of the quality of CSR disclosure and the CSR performance. This means of measurement is reported to represent a better-quality approach over existing measures, where generally firms’ symbolic involvement with CSR was measured by the quality of solely a firm’s CSR disclosure score (Marquis & Qian, 2014; Tashman et al., 2019). To create a more thorough analysis, I chose to not only utilize the absolute difference, but also the positive and negative gaps that exist between CSR performance and CSR disclosure. The positive gap is the gap that exists in firms that have a better CSR performance than their CSR disclosure and vice versa for the negative gap.
Having used these different forms of CSR decoupling, the analysis of this study holds several interesting insights with regards to the examined relationships. Support for the five hypothesized relationships between the degree of several forms of institutional investors and CSR decoupling is not found, although it can be stated that some forms have shown statistical significant
relationships. Governmental institutions have shown to be positively associated with two out of three forms of CSR decoupling. In other words put: The higher the degree of governmental institutions, the higher the chance of occurrence of CSR decoupling practices within firms, and specifically the higher the chance of Positive CSR decoupling. This indicates that firms with relatively more governmental owners use more the malpractice of CSR decoupling and, in a specific sense, implement policies regarding CSR thoroughly but decouple these policies from
32
the firm’s core tasks (Bromley & Powell, 2012). Another interesting finding, although not
supported in my hypothesis, is the weakly significant, positive relation between Other institutions and Negative CSR decoupling. This result suggests that firms with relatively high institutional owners that are not characterized as one of the three other categories, decouple more their policies from practice than firms with less Other institutions. An important note to make is that this significant linkage is not supported by one of the two other forms, which implies a relatively weak foundation.
Also contrary to my expectations, is the nature of the relationship found between the variables. The majority of the results indicate that there is a positive linkage between the degree of
institutional investors in a company and the occurrence of CSR decoupling within a firm, while the hypotheses suggested a negative linkage. The only negative association was between
Investment corporations and CSR decoupling, but the results showed a remarkably insignificant
relation between them.
While bearing in mind the implications this has and will have for further research and practice, we move on to the main question in this study: to what extent do institutional investors affect CSR decoupling? Although the five hypotheses in this research were rejected, the statistics suggest a positive relation between the aggregated variable of all institutional investors and CSR decoupling (Table 7: p-value 7.3%, positive coefficient). Thus the results of the statistical
analysis express a positive, but insignificant relationship that the higher the degree of institutional ownership in a corporation, the higher the degree of CSR decoupling within an organization. Next to the aggregated variable, this research found significant results between governmental institutions and CSR decoupling. Thus, when answering the main question, it can be concluded that governmental institutions positively affect CSR decoupling. In other words, the more shareholdings of a governmental institutions, the more occurrence of CSR decoupling. The results showed furthermore that firms with more governmental shareowners, conduct specifically more in decoupling that heightens a firm’s CSR performance than that could be justified by trough their CSR disclosure.
Further, additional tests were executed to evaluate the impact of institutional ownership on the separate measurements of CSR decoupling. In comparison to the main analysis, the results
33
demonstrate that institutional ownership does not impact any dimension of CSR significantly. Although the relationship is even less strong in all of the different dimensions when comparing to the main statistical analysis, it can be concluded that these additional analyses are in line with the main results that a significant relationship lacks. Thus, separating the CSR decoupling in the four dimensions does not add any information to our research.
5.2 Implications
While recently researchers have focused their attention upon CSR decoupling, very little is
known about the determinants of CSR decoupling. This study investigates this gap in literature by looking at the relationship between several sorts of institutional ownership and the practice of CSR decoupling. Although no significant relationship has been found in this study, there are still some implications that can be taken into consideration. The first implication is associated with the nature of the relationships, where a positive association was found in the majority of the cases. Out of this positive association can be deducted that firms with a low percentage of institutional ownership do not need to be directly associated with a higher chance of CSR decoupling. In fact, the trend pointed out by this study is quite the contrary. The tendency is that firms with higher concentrations of institutional investors, have a higher chance of CSR
decoupling. Accordingly, implications for researchers is that a negative relationship needs not to be the main focal point of interest since the figures exclude a trend that can be interpreted as such.
The implications it has on practice is that firms and their board of directors need to be aware of the fact that a higher degree in institutional investors does not correlate to less chance or occurrence of CSR decoupling, while a positive correlation remains to be unknown. Moreover, firms that are owned for a large part by governmental institutions are possibly practicing more CSR decoupling.
As indicated by the results, there is a positive, significant linkage between Governmental
institutions and two forms of CSR decoupling. This opens up possibilities for future research,
34
5.3 Limitations and future research
I acknowledge that there are some potential shortcomings and limitations that I faced during this study. The first limitation is related to usage of the CSR decoupling measurement, which is described in the conclusion paragraph in this section. This research makes use of scores on CSR disclosure and performance, which are also known as aggregates. The study of Waddock (2003) indicates that the ratings given to firms are based on a combination of qualitative and quantitative data, making these ratings subject to interpretation and judgement. While Waddock (2003: p. 372) refutes this argument partially by asserting that “these same judgements are no different from the interpretations that underlie financial and accounting statements”, it does not take away the fundamental reliability concerns of figures based on interpretation. Another potential shortcoming in the usage of scores is related to the validity. As firstly pointed out by the paper of Schepers and Sethi (2003), the question is whether these CSR scores measure what they intend to measure. As the definition of CSR emerged through the years from a more discretionary responsibilities’ view to a more integral responsibilities view, it is of high importance that the indicators used for the measurement of CSR move along with this shift in definition (Waddock, 2003). Following the argumentative discourse of Entine (2003), the decisions made for certain indicators for categories as well as within the various categories that exist in the scores are at least to call arbitrary.
Because these decisions, although they were made clearly and consistently, validity could be limited.
Second, a limitation is possibly the case when regarding the external validity of the overall research. Since it is inherently impossible to include all factors which may influence the relation between CSR decoupling and institutional ownership, a relative small sample of variables is used in this study. Also, the degree of causal inference made between the variables in this study could be invalid and would affect its’ internal validity. Taken together, the possible limitations in the internal validity could affect the external validity of this research and consequently making the outcomes of this study less generalizable to and across other situations, people, stimuli and times (Aronson et al., 2007).
