The impact of Financial Analysts’ Coverage on Corporate Social Responsibility Decoupling

The impact of Financial Analysts’ Coverage on Corporate

Social Responsibility Decoupling

Master Thesis Accountancy & Controlling

University of Groningen, faculty of Economics and Business June 22, 2020 RAMON BOSKER Studentnumber: 2725894 Sabangplein 9a 9715CV Groningen tel.: +31 (0)6-36398140 e-mail: R.bosker@student.rug.nl Supervisor: N. Hussain Wordcount: 9,832

ABSTRACT

This article examines the relationship between corporate social responsibility (CSR) decoupling and financial analysts. CSR decoupling can be defined as the difference between CSR disclosure and CSR performance. First, the relationship between CSR decoupling and financial analysts’ recommendations is examined. Second, this article explores the relationship between analyst coverage and CSR decoupling. The relevant data is extracted by using the Thomson Reuters Asset4, Thomson Reuters Eikon and Bloomberg databases. The sample consists of 4,947 and 23,990 firm-year observations for analyst recommendations and CSR decoupling respectively, for the period 2006-2016. The hypotheses were tested using a random effects regression based on panel data. The results show that there is no significant relationship between CSR decoupling and analyst recommendations. Furthermore, the results show that analyst coverage is positively related to CSR decoupling. This study also found that the majority of CSR decoupling engagements is due to understating CSR performance.

1

TABLE OF CONTENT

2

1. INTRODUCTION

In the last couple of decades, organizations recognize the growing importance of the engagement in corporate social responsibility (CSR). This growing importance can be seen as the result of a shift of a shareholder-perspective to a stakeholder-perspective for organizations. For example, in 2019, almost 200 CEO’s claimed that the main focus of their organization shifted from solely serving their owners (shareholders), to a more broad perspective: serving their customers, employees, suppliers and whole communities (stakeholders)1. Engaging in CSR is an important aspect of meeting stakeholder demands, and it is equally important for companies to make sure that their stakeholders are informed about their actions regarding CSR (Tata & Prasad, 2015). For a company, providing this information is a good opportunity to inform the public that they are doing the best they can to be socially responsible. However, some might use CSR information as a facade to hide actual CSR performance (Tashman, Maravo & Kostova, 2018). For example, The Guardian found some companies who misled customers with their environmental claims2. According to this article, in 2017, BMW claimed that their new car (the i3 electric) achieved the status of ‘’zero emissions’’. However, the car included an option of the implementation of a small petrol engine. Ryanair claimed that they had the lowest emissions as UK airline, but the information they used was outdated (based on data from 2011), and not all relevant airlines were included. These are small examples of the difference between the social responsibility a company claims to achieve, and the actual performance regarding social responsibility. In this study, this gap is called CSR decoupling.

CSR is a well-discussed topic in the literature. McWilliams and Siegel (2001) interpret CSR as ‘‘actions that appear to further some social good, beyond the interest of the firm and that which is required by law’’. It means that actions regarding CSR do not need to be in the best interest of the company, but in the best interest of society. With that in mind, it should indicate that regarding CSR actions, societal needs come first and organizational needs come second. Since CSR information is important to many stakeholders, companies turn to CSR reporting more often. This type of reporting contains information which indicates the degree to which a company engages in CSR. However, this type of reporting often misstates actual CSR performance, which creates a gap between what a company claims to do with CSR, and what a company actually does with CSR (Delmas & Burbano, 2011). This ‘decoupling’ of CSR

1 _{https://opportunity.businessroundtable.org/ourcommitment/}

3

information can be the result of symbolic strategy whereby CSR performance is overstated, to boost a company’s legitimacy (Tashman et al., 2018).

Even though the increase of CSR engagement by companies can be a result of moving towards a stakeholder orientation, CSR is becoming more important to shareholders as well (Lee, Palmon & Yezegel, 2016). Where investors were mostly concerned with a firm’s financial information, non-financial information like CSR activities gains importance nowadays. To obtain financial information, investors make a lot of use of financial analysis conducted by financial analysts. According to Graham and Dodd’s Security Analysis, which has also been known as a financial analyst’s bible, financial analysis is about obtaining and presenting the important facts, in a way which is most useful to investors (Cottle, Murray & Block, 1989). However, since investors are laying more emphasis on non-financial information like CSR, financial analysts are moving to a point were analyzing CSR information is just as important as the regular financial data. According to Luo, Wang, Raithel & Zheng (2015), a major part of financial analysts stated that they do not reflect positive on stock which poses CSR risks to investors. Since CSR decoupling is about information not reflecting performance, financial analysts are involved in the matter because they use CSR information for analyzing purposes. They need to carefully consider whether CSR decoupling is occurring, and if so, they will have to adjust their recommendations. In this study, I examine the effect of CSR decoupling on financial analysts’ recommendations. Furthermore, I examine the effect of analyst coverage on CSR decoupling. The research question for this study is:

Does CSR decoupling lead to more negative recommendations by financial analysts, and is analyst coverage negatively associated with CSR decoupling?

To answer the research question, data from databases Thomson Reuters Asset4, Thomson Reuters Eikon and Bloomberg is used. Asset4 provides environmental, social and governance (ESG) scores regarding company performance. Bloomberg provides ESG-scores regarding company disclosure. The CSR decoupling score is the absolute difference between the CSR disclosure and CSR performance score. Thomson Reuters Eikon is used in order to obtain data regarding analyst recommendations. The sample contains 28.518 firm year observations for CSR decoupling and 4.830 for recommendations of global companies within a time period of 2006-2016. The results showed some surprising results. According to the regressions based on this sample, CSR decoupling does not necessarily lead to more negative sell recommendations by financial analyst. Furthermore, the results show that analyst coverage is positively associated

4

with CSR decoupling, where a negative relationship was expected. Another interesting result is that the calculated CSR decoupling scores were more due to understated CSR performance, instead of overstated CSR performance. I expected that most of the engagements in CSR decoupling by companies were a result of overstating their CSR performance.

This study contributes to the literature around CSR and financial analysts in multiple ways. First, it adds to the existing literature by exploring the relationship between CSR decoupling and analyst recommendations. Second, since the literature in the field of CSR decoupling is still very limited, this paper can increase our general understanding of CSR decoupling. Third, I build on recent work by Hawn & Ioannou (2016) and Tashman et al. (2018) by using the same research methodology regarding CSR decoupling. This decreases the heterogeneity within the CSR decoupling literature. Fourth, this study adds to the agency theory literature, by using this theory in explaining the hypothesized relationships. At last, the results of this study provide insights regarding overstating and understating CSR performance. Since the latter occurred more, which is against popular believe (Hawn & Ioannou, 2016), it provides research opportunities. This study shows that the possibility of companies understating their CSR performance, also called the communication gap, should not be ignored.

This paper is organized as follows. I first discuss the notion of CSR decoupling and financial analysts in more depth in the literature review and hypothesis development. Subsequently, I describe the research methodology and variables. Then, the outcomes of this research are shown. This paper concludes with a discussion section, where I show the main findings and highlight the limitations and possible area’s for future research.

2. LITERATURE REVIEW

The global corporate progression of a shareholder-perspective towards a stakeholder-perspective puts CSR in a greater spotlight. Since CSR helps companies maintaining a good relationship with many of their stakeholders, CSR engagement can be a rewarding exercise. According to Siegel and Vitaliano (2007, p.773), ‘’CSR occurs when firms engage in activities that appear to advance a social agenda beyond which is required by law’’, which implies that CSR heavily relies on voluntary actions by a company. Since these actions are voluntary in nature, management has a lot of influence regarding the engagement in such actions. According to Hawn and Ioannou (2016), the main goal of these actions is to react to the institutional

5

pressures regarding organizational responsibility and transparency. The authors divided CSR actions in two distinct forms: internal and external. Internal actions can be described as actions which are focused on meeting the demands of social actors the company relies on. Examples of internal actions can be CSR-related policies or the implementation of a CSR-committee within an organization. On the other hand, external actions are more aimed at reaching the public outside the organization. These actions have a high rate of visibility, and its main purpose is to increase an organization’s legitimacy. These actions often include symbolic means of communication (Yanow, 1996). Examples of external actions regarding CSR are public claims or reports by a company, which inform the public about their social and environmental commitment/engagement (such as certain targets or initiatives). Both types of action are frequently interconnected, mostly in a way where prior internal action leads to current external action (Hawn & Ioannou, 2016).

The voluntary/unregulated nature of CSR reporting can result in more incomplete, biased or selective information (Garcia-Sánchez, Hussain, Khan & Martinez-Ferrero, 2020). This could result in a gap between CSR performance (internal actions) and CSR disclosure (external actions), which is called CSR decoupling. According to Crilly, Zollo & Hansen (2016), there are two strategies which can explain this decoupling between CSR performance and disclosure. The first strategy is managers making use of the information asymmetry which exists between them and the stakeholders. The stakeholders are not well aware of actual performance, which gives managers the opportunity to follow personal interests. They do so by using CSR information more symbolically, hiding actual performance. This unethical behavior of companies towards their stakeholders is a form of window dressing (Weaver, Trevino & Cochran, 1999). They use symbolic signals to give a satisfactory image regarding certain regulatory requirements and try to enhance stakeholders’ perceptions of overall business legitimacy, while maintaining ‘’business as usual’’. Even though one might think that stakeholders have enough resources at hand to see through these unethical practices, earlier research found that stakeholders are still granting legitimacy to companies who are using symbolic information (Elsbach, 2003; Suchman, 1995; Westphal & Zajac, 2001). Concealing noncompliance or reporting fake CSR actions are examples of this opportunistic symbolic behavior by managers. The second strategy which could lead to CSR decoupling, can be the result of the pressure by firms of adapting to the rapidly changing expectations of stakeholders. The adoption and implementation of new policies is hard to do simultaneously, since the implementation process can be quite a challenge (Meyer & Rowan, 1977). This could also result

6

in a (temporary) difference between internal action (implementing adapted policies) and external action (informing the public about the adaption of certain policies).

There are two types of CSR decoupling. The different strategies which are mentioned above are mostly concerned with so-called CSR ‘talk’. CSR talk, also called the implementation gap, occurs when a company communicates CSR information symbolically, and where the actual implementation of CSR is less than what has been communicated (Wickert, Scherer & Spence, 2016). This type mostly occurs with large sized companies, where CSR reporting is easier and quicker than CSR implementation. Hence, many short-term oriented managers are more likely to engage in CSR reporting, and less in the engagement of actual CSR implementation. The authors also point out that there is an opposite side of CSR talk: CSR ‘walk’. CSR walk, or the communication gap, occurs when a company implements CSR substantively, but does not communicate it to their stakeholders properly. This mostly happens with smaller companies, who can implement CSR relatively easy, but lack the resources to properly inform their stakeholders about it.

Research regarding CSR decoupling is still relatively scarce, but the last few years it has been emerging. In earlier years, much research was done around the notion of symbolic management. Within this line of research, the gap between symbolic/external and substantive/internal actions has been noted often (Meyer & Rowan, 1977; Weaver, 1999; Westphal & Zajac, 2001; Berrone, Gelabert & Fosfuri, 2009). Crilly et al. found two possible underlying responses of companies who have decoupled their internal practices from their external practices. It can be intentional, or also named by the authors as ‘faking it’, which implies that managers create a gap between internal and external actions deliberately. They intend to conceal non-implementation of various CSR policies by using symbolic external actions. Another response by companies can be an emergent one, or ‘muddling through’. This response is the product of competing expectations of stakeholders, where the company tries to find a consensus among managers, who all have different opinions regarding the organizational environment.

Most of the existing research around CSR decoupling has focused on internal and external organizational factors which affect the degree of an organization’s engagement in CSR decoupling. Sauerwald & Su (2018), for example, focused on internal factors which might affect the degree of CSR engagement by companies. They concluded that CEO overconfidence is positively related to CSR decoupling, where the board of directors can mitigate this effect when they have CSR expertise and ownership incentives. External causes of CSR decoupling have been a more popular research area to date, with for example Marquis & Qion (2014) or

7

Schons & Steinmeier (2016), examining the effect of monitoring by external stakeholders. With the focus of this paper on the influence of financial analysts, it adds to the existing literature around external organizational factors which affect the degree of engagement in CSR decoupling.

Besides the fact that research around CSR decoupling is scarce, it is also heterogeneous and divergent. Conducted research, countries and applied theories vary among the articles. This also applies for research regarding the relationship between financial analysts and CSR. There are few studies who have examined this relationship, and since authors used different theoretical frameworks, CSR actions and financial analysts’ outcomes, it remains very fragmented to date (Hinze & Sump, 2019). There is no existing literature regarding the relationship between CSR decoupling and analyst recommendations. However, the impact of CSR performance on financial analysts’ recommendations has been addressed in prior research. Luo et al. (2015) stated that higher levels of CSR performance lead to more positive recommendations by financial analysts. The link between CSR disclosure and analyst recommendation remains understudied to date. However, the relationship between analyst recommendations and CSR disclosure has been examined before, but it showed controversial results (Giannarakis, Zafeiriou, Sariannidis, and Efthalitsidou, 2016). To date, only the study by Garcia-Sánchez et al. (2020) examined the link between analyst coverage and CSR decoupling. A negative relationship was found. The impact of analyst coverage on CSR in general has been examined by Adhikari (2016). The author found that firms with greater analyst coverage receive lower overall CSR scores. However, research with respect to the impact of analyst coverage on CSR disclosure remains understudied.

With this study I extend the CSR decoupling and financial analysts literature, and by using a research method which has been used more often, this study tries to reduce the degree of heterogenicity within this area of research. In table 1, I summarized earlier research around CSR decoupling, financial analysts and complementary fields of research. The table is sorted by year of publication, from earliest to latest research.

8

Table 1. Prior research in the field of CSR Decoupling and Financial Analysts

Authors Contribution Data Source Theories Country

Prior literature regarding (the notion of) CSR decoupling Berrone et

al. (2009)

Impact of environmental symbolic policies and substantive actions on

environmental legitimacy

Multiple databases Institutional theory and stakeholder theory U.S. MacLean & Benham (2010)

Illustrating how decoupling creates a legitimacy façade

Case study Neo-institutional theory

U.S. Crilly et al.

(2012)

Two ways of decoupling policies from practice: intentional or

emergent

Rating agencies, interviews, archival data

Institutional theory Global Kim &

Lyon (2012)

The costs and benefits of symbolic management: the gap

between symbolic and substantive actions

Data from internet and specialized energy industry data companies

Institutional theory U.S.

Walker & Wan (2012)

Effects of substantive and symbolic environmental actions

on firm performance

Financial Post’s ranking of top 500 Canadian

firms

Institutional theory, signaling theory and

prospect theory

Canada Christensen

et al. (2013)

The positive effect of CSR talk on actual social change within

organizations

Literature review No particular Global Marquis &

Qian, (2014)

The strategic response of firms to government signals regarding

appropriate CSR

CSMAR, National Bureau of Statistics

China, RKS

Institutional theory China Cho et al.

(2015)

Sustainability disclosure literature: organizational facades

and organized hypocrisy

Literature review Signaling theory and legitimacy theory Global Jamali et al. (2015) Selective decoupling of CSR information by enterprises in developing countries

Empirical fieldwork Institutional theory India Hawn &

Ioannou (2016)

Effect of the interplay between external and internal CSR actions

on market value

Multiple databases Neo-institutional theory 33 countries Schons & Steinmeier (2016)

The link between CSR actions and financial performance

Thomson Reuters’ Asset4 database Neo-institutional theory and stakeholder theory Cross-country Wickert et al. (2016)

Conceptual model regarding different patterns of CSR

engagement

Literature review Many, complements institutional theory Global Luo et al. (2017) CSR reporting as organizational response to government pressures CSR reports, rating agency Resource dependence theory China Sauerwald & Su (2018)

Relationship between CEO overconfidence and CSR

decoupling

S&P 500 firms Agency theory U.S. Graafland

& Smid (2019)

Analyzing the

incidence/conditions that affect decoupling among CSR policies

and programs

Ratings from sustainability rating

agency

Institutional theory and slack resource

theory 24 countries Tashman et al. (2019) CSR decoupling in emerging market multinational enterprises

Multiple databases Neo-institutional theory

9

Prior literature regarding financial analysts coverage, recommendations & CSR Jo (2003) Companies who are socially

responsible attract more analysts

Multiple databases Social norm compliance

U.S. Harjoto &

Jo (2013)

CSR related activities do not attract analysts

KLD stats Social pressure theory

U.S. Chang et al.

(2014)

Good CSR performance leads to higher hold recommendations by

analysts I/B/E/S Information economics Taiwan Jo & Harjoto (2014)

Lagged CSR performance has no impact on analyst coverage

Multiple databases Social pressure theory & agency

theory U.S. Ioannou & Serafeim (2015) Less pessimistic recommendations for firms with

high CSR ratings in recent decades

KLD, I/B/E/S, Compustat

Agency theory & stakeholder theory

U.S.

Luo et al. (2015)

Increase in CSR performance positively affects analyst

recommendations

Bloomberg Stakeholder theory U.S.

Adhikari (2016)

Greater analyst coverage leads to less social responsibility by firms

Difference-in-difference technique

No particular Global Gao et al.

(2016)

The quality of CSR disclosure is positively related to analyst

coverage

Multiple databases Information economics

Netherlands Giannarakis

et al. (2016)

Relationship between CSR disclosure and analyst

recommendations is controversial

Bloomberg Voluntary disclosure theory

Global

Lee et al. (2018)

CSR reports lead to better information environment, which

decrease informativeness recommendations

MSCI ESG, I/B/E/S, Compustat

No particular U.S.

Garcia et al. (2020)

Relationship between CSR decoupling and financial market

outcomes

10

3. HYPOTHESIS DEVELOPMENT

3.1 Financial analysts

Of late, the participants in the capital market attach more value to corporate investments in CSR. This change of investors’ perception regarding CSR activities motivated managers to increase investments in CSR and report these increases to their investors (Lee et al., 2018). According to Jegadeesh, Kim, Zain & Sawani, (2004), since financial analysts forecast future performance, rate stock investment potential and create research reports for companies, they function as important intermediaries regarding information for investors, and for other stakeholders as well.

There are two types of financial analysts, which are referred to as ‘’sell-side’’ and ‘’buy-side’’ analysts (Fogarty & Rogers, 2005). Sell-side analysts are working for brokerage firms, where their main goal is to make research reports for anyone who is willing to invest in securities which are traded publicly. They try to gather as much useable information as possible, to assess the future performance of a company, which is a very important proxy for security holders. In the end, the analysts will come up with a recommendation report. Buy-side analysts are mostly employed by institutional investors, who own their own set of portfolios of securities. The main difference between buy-side and sell-side analysts, is that they do not make their reports openly available for the public. These reports are solely meant for their employers. Their way of gathering information differs as well, since they use a lot of sell-side analyst reports for their own reports (Schipper, 1991). In this study, I focus on the recommendations by sell-side analysts. I further refer to them as financial analysts.

In the late nineties, information regarding sustainability was not a main concern in the assessments of financial analysts (Deegan & Rankin, 1997). Since the importance of CSR information has grown over the past few decades, many financial analysts started marking this type of information as a key criterion (Garcia-Sánchez et al., 2020). Prior research showed that there is a significant relationship between CSR initiatives and corporate financial performance (Orlitzky et al., 2003). This led to the perspective that information regarding CSR can constitute useful input for company valuation and analysis processes (Dhaliwal et al., 2012). Luo et al. (2015) pointed out that the majority of financial analysts do not endorse stocks that involve CSR risks, even when financial performance seems to be promising. Since CSR performance often contains uncertain and ambiguous information, general investors value the opinion of

11

financial analysts. They are able to connect CSR performance with firm stock returns (Luo et al., 2015).

Besides analyzing financial information, financial analysts can play an important role as monitoring agents as well. This interpretation originates from the agency theory. The agency theory is based on the assumption that people are self-interested, and use opportunism to get what they want (Dees, 1992). In this perspective, the self-interested person is the agent, which operates on behalf of the principal. There may be an information asymmetry between the principal and the agent, which implies that the agent has more information at hand than the principal. The agent may use this information for his/her own gain, at the expense of the interests of the principal. For example, managers (the agents) could use an entrenchment strategy by committing themselves to certain CSR initiatives, to gain more stakeholder support. This increase in support decreases the chance for managers of being replaced in endorsed takeovers (Cespa & Cestone, 2007). While increasing CSR initiatives does not sound dissatisfactory at first, this entrenchment strategy leads to more inefficiency by management, which in turn negatively affects firm performance. So while these CSR initiatives seem to be for the greater good, it actually fulfills a manager’s agenda. To keep an eye on their agent, a principal may use monitoring mechanisms, to see whether the actions of the agent are in line with the principal’s interests (O’Donnel, 2000). This monitoring function reduces the information asymmetry between the principal and the agent. In this setting, financial analysts can act as a monitoring mechanism, since they follow and analyze a company closely. They assess company strategies, ask questions to management and analyze information regarding firm performance (Adhikari, 2016).

Financial analysts acting as monitoring mechanisms can pose pressure on companies who are engaging in CSR decoupling. Since CSR information is in general very voluntary, selective and unregulated in nature (Garcia-Sánchez et al., 2020), this type of information is subject to a lot of management discretion. Within the agency theory, this discretion regarding CSR information can move managers to act in their own best interest, instead of the best interest of financial and non-financial stakeholders. The pressure put on companies by financial analysts could enable managers to alter their engagement in CSR: from self-serving interest to serving the best interest of their stakeholders. And since financial analysts are important intermediaries for investors and other stakeholders (Jegadeesh et al., 2004), they are able to punish companies when management is serving their own interests by implementing it into their recommendation reports. In the next two sections, the interplay between a firm’s engagement in CSR decoupling and the role of financial analysts as a monitoring mechanism will be explained in more detail.

12

3.2 CSR decoupling and Financial analysts’ recommendations

Financial analysts follow a company closely to collect information, which they interpret for their recommendation reports. They use these reports to give advice to their clients, by informing whether they should buy, sell or hold their shares of a certain firm (Ioannou & Serafeim, 2015). Buying and selling can be seen as positive and negative recommendations, respectively. Information like CSR activities could help financial analysts, since they operate in an environment based on information. However, according to Lee et al. (2016), the presence of alternative information can be a double-edged sword for analysts. On the one hand, an increase of information can be beneficial for analysts since it increases their potential of using and analyzing this information for their recommendations. On the other hand, when the supply of information has increased, it can contribute to a more efficient pricing environment which reduces the value of using analysts’ recommendations.

Organizations tend to disclose their CSR information for the participants of the financial markets, including financial analysts, to reduce the information asymmetry between them (Bushman, Piotrosky & Smith, 2004). However, for legitimizing purposes, they also need to disclose CSR information to non-financial stakeholders as well (Orlitzky, Schmidt & Rynes, 2003). Ultimately, the disclosed CSR actions by a firm are validating or undermining its legitimacy efforts and the economic impact of these actions. Due to these external pressures, organizations might diverge their CSR disclosure from their CSR performance, which leads to CSR decoupling. Financial analysts use the reported CSR information by firms for their recommendation reports, in aid of drawing their expectations of these firms. When analysts solely use disclosed CSR information which is not in line with the firm’s actual performance, financial analysts use misleading information which could lead to wrong financial forecasts. This could lead to credibility or reputational risks (Cornier & Magnan, 2015). To mitigate these risks, analysts use private information as well, which is acquired through their industry expertise. General investors are less able to gain access to this private information because they have less industry expertise, and lack time and resources (Luo et al., 2015).

To examine the link between CSR and financial analysts, the agency theory can be used. Financial analysts can function as information moderators between the principal and the agent, where shareholders and other stakeholders are the principal and management is the agent. When firms are engaging in CSR decoupling, and analysts find out that the reported information is deceiving, i.e. not in line with actual performance, they can punish companies on behalf of the principal by negatively reflecting on this in their recommendation reports. So when companies

13

engage in CSR decoupling, it negatively affects financial analysts’ recommendations. In other words, it leads to more sell recommendations by financial analysts.

Hypothesis 1: CSR decoupling is positively linked to sell recommendations by financial

analysts.

3.3 Analyst coverage and CSR decoupling

To draw their expectations or recommendations of a company, financial analysts follow a company closely, in order to gather enough information. This closely following companies by analysts can be seen as a form of external monitoring. Jo and Harjoto (2014) argue that when analysts are monitoring companies, they basically act as an external corporate governance mechanism. When a company is monitored by analysts, they might feel more social pressure to make their business more legitimate in the eyes of their stakeholders. For example, As a reaction to such social pressure, companies are more likely to engage in charitable giving (Zhang, Tong, Su & Cui, 2015). Also, according to Zhang et al. (2015), financial analysts following a company may lead to an increase of attention from the public. This could consequently lead to CSR initiatives being more value-enhancing, which gives incentives to management to engage in CSR. So financial analysts following a company could ultimately lead to an increase of CSR initiatives by that company.

When examining the relationship between analyst coverage and CSR decoupling, the agency theory can be used as well. Within an organizational setting, financial analysts can act as an external monitoring function for the principal (stakeholders), where management is the agent. Analyst coverage is an important factor when it comes to the effectiveness of external monitoring. Analyst coverage can be defined as the amount of financial analysts who follow and make forecasts and recommendations regarding a company. When analyst coverage increases, the external monitoring function of financial analysts becomes more effective, since a greater amount of analysts follow a particular company. Consequently, the information asymmetry between stakeholders and management decreases. Companies may feel more pressure to make sure that their performance and reporting are more aligned, since stakeholders can see through symbolic gestures with the aid of this external monitoring by financial analysts. I predict that when the analyst coverage of a company increases, the gap between CSR reporting and CSR performance decreases. So the relationship between analyst coverage and CSR decoupling is negative. To draw on this perspective, I present the following hypothesis.

14

Hypothesis 2: There is a negative association between analyst coverage and CSR

decoupling.

4. METHODOLOGY

4.1 Sample

In this study, I use data from public companies with a time period ranging from 2006 to 2016. The sample consists of 4,947 firm-year observations for analyst recommendations and 23,990 firm-year observations for CSR decoupling. The observations include 4,433 companies operating in 49 industries in 58 countries. Since there are many countries included in this research, there is a high degree of external validity. Furthermore, by using companies from 49 industries based on the Fama & French industry classification, the results are generalizable for multiple business sectors. In this research, panel data is used. This type of data includes the effect of measurement over time, by identifying each company and each year. In other words, the company observations in this sample are being examined through multiple points in time. I use regressions based on random effects, with the inclusion of three kind of dummies to control for firm year, country and industry effects.

4.2 Dependent variables

For both hypotheses, I use seperate dependent variables. In the first hypothesis, analyst recommendations is the dependent variable. I conduct two different regressions regarding recommendations. One which uses the percentage of analysts who recommend buying shares, and one which uses the percentage of analysts who recommend selling shares. These

variables are obtained by using the database Thomson Reuters’ Eikon.

In the second hypothesis, I use CSR decoupling as dependent variable. In the literature review, I discussed two types of CSR decoupling, the implementation gap (CSR talk) and the communication gap (CSR walk). In this study, the main focus regarding the CSR decoupling types lays on the implementation gap for two reasons. First, the data I use is mostly from companies with a relatively large firm size, who are more prone to the implementation gap (Wickert et al., 2016). Second, the implementation gap can be a product of managers’ self-serving behavior and the external monitoring by analysts can mitigate this type of behavior.

15

So it is more likely that financial analysts reduce the implementation gap, instead of the communication gap. The degree of engagement in CSR decoupling is measured based on the method of Tashman et al. (2019) and Hawn & Ioannou (2016). CSR decoupling is the difference between what a company says about their CSR performance, and their actual CSR performance. This misalignment can be measured by comparing values of CSR performance with measures of CSR disclosure. The data regarding CSR performance is gathered from the Thomson Reuters Asset4 database. I use one year lagged performance scores, assuming that it takes one year for performance data to be legitimized. For CSR disclosure rates, I use the Bloomberg database to collect CSR disclosure scores. Since both databases use ESG-metrics for their performance and disclosure scores, they are well suited to be compared to each other. Furthermore, the data of both databases have already been standardized, which improves comparability as well. I dropped the data of companies who did not receive ESG scores in any of the two databases. Also, I use absolute CSR decoupling scores, which implies that there are no observations smaller than zero. The measurement of CSR decoupling is defined as follows, where f is firm and t is year. The disclosure and performance measures are based on the scores by Bloomberg and ASSET4 respectively.

CSRDecouplingft = CSRDisclosureft – CSRPerformanceft-1

4.3 Independent variables

In the first hypothesis, the independent variable is CSR decoupling. How this variable is measured has been explained in the dependent variable section. This measurement procedure is also used with CSR decoupling as an independent variable.

In the second hypothesis, the independent variable is analyst coverage. Analyst coverage is the amount of analysts following a single company. In this study, this variable is the mean of analysts following a company throughout a given year. This data is obtained by using the Thomson Reuters Asset4 database.

4.4 Control variables

I also implement some control variables, to mitigate the risk of the possibility of alternative explanations, and to mitigate the risk of biased results. For financial analysts, I use firm size as control variable. Larger firms might lead to more trading commissions for analysts, and can be beneficial for investment banking business (Ioannou & Serafeim, 2013). Since analysts are financially dependent on both commissions and banking business, they can be inclined to be

16

more optimistic regarding recommendations. As a proxy for firm size, I use the log of total assets. The second control variable is R&D intensity. Prior research has shown that

recommendations by analysts and the degree of the use of R&D are positively related

(Palmon & Yezegel, 2012). I use leverage and return on assets (ROA) as control variables for analyst recommendations as well. Fama and French (1992) showed that these variables are associated with market returns. At last, I implement some corporate governance control variables. According to Byard, Li and Weintrop (2006), when companies improve their corporate governance, it benefits the users of financial information provided by the company, including financial analysts. Consequently, this improves financial analysts’ forecast

accuracy, which could ultimately lead to an improvement in the quality of their

recommendations. So corporate governance quality can affect recommendation reports by financial analysts. I include board size, board independence, CEO duality and board gender diversity as corporate governance control variables.

Now I discuss the control variables for CSR decoupling. Following research by Tashman et al. (2019), I use ROA as a control variable, because it can provide some extra financial resources to invest in CSR (Bansal, 2005). Research & Development intensity is a suitable control variable as well, since it is positively related with the degree of

innovativeness of a company. A very innovative company may be better able to improve their CSR performance (McWilliams & Siegel, 2000). I also use firm size as control variable for CSR decoupling. Firms of smaller size are not always capable of implementing CSR properly, since they have less experience (Lepoutre and Heene, 2006). Following research by Hussain, Rigoni and Orij (2018), I use sales growth as a control variable, since this gives a firm opportunities to invest in CSR. The last set of control variables for CSR decoupling are variables regarding corporate governance. For determining which corporate governance to use, I followed research by Sauerwald & Su (2019). I use board size as control variable, since it can affect the effectiveness of the monitoring function of the board. A more effective monitoring board may decrease the possibility of management to engage in CSR decoupling. Furthermore, I use board independence, since board independence is positively associated with the effectiveness of board monitoring. The third corporate governance variable is CEO duality. CEO duality allows a CEO to pose more power and influence on the organization, which can affect reporting quality. The last corporate governance control variable is board gender diversity. Bolino & Turnley (2003) found that female directors use less symbolic management than male directors, which should negatively affect the degree of CSR decoupling. At last, I also use dummies. Year dummies, country dummies and industry

17

dummies have been included in my analyses. I use the Fama & French industry classification for the industry dummies, where company sic codes are matched to 49 different industries. I summarized all variables and corresponding proxies in table 2.

4.5 Data analysis

For the first hypothesis, the following formulae have been created, for buy and sell recommendations separately. β is the correlation coefficient and ε is the error term.

BUYft = β0 + β1 CSRDft + β2 ROAft + β3 LEVft + β4 RD_INTft + β5 SIZEft + β6 B_SIZEft +

β7 INDEPft + β8 DUALft + β9 GENDERft + β10 YD ft + β11 CD ft + β12 IDft + ε ft

SELLft = β0 + β1 CSRDft + β2 ROAft + β3 LEVft + β4 RD_INTft + β5 SIZEft + β6 B_SIZEft +

β7 INDEPft + β8 DUALft + β9 GENDERft + β10 YD ft + β11 CD ft + β12 IDft + ε ft

For the second hypothesis, I use the following formula:

CSRDft = β0 + β1 COVft + β2 ROAft + β3 S_GRft + β4 RD_INTft + β5 SIZEft +β6 B_SIZEft +

18

Table 2. Overview of the variables

Variable Type Variable Name Proxy Measurement

Dependent Variables

CSR Decoupling CSRD The absolute difference between CSR disclosure score and lagged

CSR performance score Buy Recommendations BUY The percentage of analysts who

recommend buying shares Sell Recommendations SELL The percentage of analysts who

recommend selling shares Independent

Variables

Analyst Coverage COV The average number of analysts following a company in a year CSR Decoupling CSRD The absolute difference between

CSR disclosure score and lagged CSR performance score Control variables Return on Assets ROA The ratio of net income and total

assets

Leverage LEV The ratio of debt compared to equity

R&D Intensity RD_INT The degree of R&D costs compared to total costs Firm Size SIZE Calculated as log of total assets in

USD

Sales Growth S_GR Ratio of change in net sales compared to previous year Board Size B_SIZE Amount of members on a

company’s board Board Independence INDEP The ratio of outside directors

within a company’s board CEO Duality DUAL Whether the chairman of the board

is also the CEO of the company Board Gender Diversity GENDER The ratio of female board members

within a company’s board

Dummies Firm Year Dummy YD A dummy for each firm year

Firm Country Dummy CD A dummy for each country included in the sample Firm Industry Dummy ID A dummy for each industry, based

on the classification by Fama & French

19

5. RESULTS

5.1 Descriptive statistics

Before showing the regressions for the hypotheses and determine whether these hypotheses can be statistically supported, I first show the descriptive statistics of all variables in table 3. This table shows the mean, standard deviation (SD), minimum and maximum values for every variable. Table 4 shows the results of the Pearson correlation matrix. This matrix tests for multicollinearity. There is one coefficient in this matrix which can be seen as very high. That is the relationship between buy and sell recommendations, with a coefficient of -0.60.

However, since these two are strongly related and are used in separate regressions, it does not pose a real threat of multicollinearity. All variables have been winsorized, to reduce the effect of outliers.

Table 3. Descriptive statistics

Variable Mean SD Min Max

CSRD 28.466 18.147 0.000 86.393 COV 2.496 0.806 0.000 3.761 BUY 50.135 26.861 0.000 100.000 SELL 14.082 17.340 0.000 100.000 ROA 5.600 7.915 -45.080 35.490 LEV 0.249 0.141 0.000 0.801 RD_INT 0.014 0.048 0.000 0.394 SIZE 15.793 1.635 11.529 21.000 S_GR 0.071 0.223 -0.622 1.329 B_SIZE 10.521 3.484 4.000 25.000 INDEP 72.676 24.407 0.000 100.000 DUAL 0.410 0.492 0.000 1.000 GENDER 12.120 11.450 0.000 50.000

20

Table 4. Pearson correlation matrix

CSRD COV BUY SELL ROA LEV RD_INT SIZE S_GR B_SIZE INDEP DUAL GENDER

CSRD 1.000 COV 0.267 1.000 BUY -0.061 0.072 1.000 SELL 0.095 0.036 -0.606 1.000 ROA 0.071 0.136 0.060 -0.065 1.000 LEV 0.001 -0.008 -0.012 0.042 -0.108 1.000 RD_INT -0.004 -0.029 0.014 -0.024 -0.002 -0.005 1.000 SIZE 0.239 0.345 -0.013 0.035 -0.144 0.080 -0.142 1.000 S_GR -0.074 0.010 0.169 -0.119 0.189 -0.022 0.011 -0.045 1.000 B_SIZE 0.157 0.184 -0.022 0.045 -0.047 0.050 -0.055 0.049 -0.043 1.000 INDEP 0.200 0.181 -0.012 0.127 0.054 0.058 0.068 0.070 0.017 0.027 1.000 DUAL 0.025 0.074 -0.000 -0.023 0.013 0.013 -0.051 0.090 -0.029 0.056 -0.041 1.000 GENDER 0.252 0.149 -0.069 0.096 0.037 0.037 0.031 0.096 -0.056 0.078 0.383 -0.0114 1.000

21 5.2 Main regressions first hypothesis

For the first hypothesis, I use two separate regressions to test whether CSR decoupling affects financial analysts’ recommendations of buying or selling shares. To follow my hypothesis, CSR decoupling should decrease the amount of buying recommendations, and increase the amount of selling recommendations. I made multiple models to test the hypothesis.

In the first model, I test the dependent variable of the buying recommendations with the control variables. The coefficients of the control variables ROA, financial leverage, R&D intensity, firm size, board size, board independence, CEO duality and board gender diversity are 0.402, -3.108, -0.380, 0.776, -0.094, -0.039, -1.011 and -0.096 respectively. ROA shows a significant positive relationship with buy recommendations, at the p<0.01 level. Furthermore, board gender diversity shows a significant negative relationship with buy recommendations, at the p<0.05 level. The other control variables are not significantly related to buy

recommendations.

In the second model, I test the dependent variable of selling recommendations with the control variables. Here, the coefficients of the control variables ROA, leverage, R&D intensity, firm size, board size, board independence, CEO duality and board gender diversity are -0.262, 3.466, -16.351, -0.615, -0.040, 0.040, 0.991 and 0.091 respectively. The control variables show much more significance, compared to buy recommendations. Especially ROA and board gender diversity, which are significant at the p<0.01 level, where ROA is

negatively associated with sell recommendations and board gender diversity is positively associated with sell recommendations. Furthermore, R&D intensity shows a significant negative relationship with sell recommendations, at the p<0.05 level. Three control variables are significant at the p<0.10 level, which are financial leverage, firm size and board

independence. Where financial leverage and board independence show a positive relationship with sell recommendations, firm size is negatively related. The other control variables, board size and CEO duality, do not show any significance related to sell recommendations.

In the third model, I test buying recommendations with the independent and control variables. This model shows a positive coefficient of 0.028 between CSR decoupling and buy recommendations. This indicates that CSR decoupling leads to more buy recommendations, which is contrary to my hypothesis. However, since the relationship does not show any significance, it can not be statistically supported.

In the fourth model, I test sell recommendations with the independent and control variables. This model shows a positive coefficient of 0.017 between CSR decoupling and sell recommendations. Since this positive coefficient would indicate that CSR decoupling leads to

22

more sell recommendations, it is in line with my hypothesis. However, since the relationship does not show any significance, as is the case with buy recommendations, it can not be statistically supported. Since both regressions based on the relationship between CSR decoupling and buy/sell recommendations can not be statistically supported, my first hypothesis must be rejected.

Table 6. Regressions with analyst recommendations as dependent variable

Variable (1) (2) (3) (4) CSRD 0.028 0.017 ROA 0.402 *** -0.262 *** 0.421 *** -0.259 *** LEV -3.108 3.466 * -2.020 3.301 RD_INT -0.380 -16.351 ** -3.201 -13.502 SIZE 0.776 -0.615 * 1.243 ** -0.976 ** B_SIZE -0.094 -0.040 -0.102 -0.065 INDEP -0.039 0.040 * -0.587 * 0.052 ** DUAL -1.011 0.991 -1.192 1.254 * GENDER -0.096 ** 0.091 *** -0.119 ** 0.103 *** CONSTANT 50.692 *** 7.493 34.152 * 13.338

Year Dummies Included Included Included Included

Country Dummies Included Included Included Included

Industry Dummies Included Included Included Included

R-Squared 0.151 0.132 0.157 0.139

n 4,947 4,947 4,299 4,299

*p<0.10, ** p<0.05, ***p<0.01 (two-tailed).

5.3 Main regression second hypothesis

For the second hypothesis, I test whether it can be statistically supported that analyst coverage has a negative relationship with CSR decoupling. Here I use multiple models as well, to test this hypothesis.

In the first model, I test the dependent variable of CSR decoupling with the various control variables. The coefficients of the control variables ROA, financial leverage, R&D intensity, firm size, sales growth, board size, board independence, CEO duality and board gender diversity are 0.061, -2.304, 3.820, -2.764, 0.292, 0.051, 0.324 and 0.107 respectively. Many of the control variables are significantly related with CSR decoupling, at the p<0.01

23

level. Most of them are positively related to CSR decoupling, including ROA, firm size, board size, board independence and board gender diversity. Only sales growth seems to be

negatively associated with CSR decoupling. R&D intensity and CEO duality do not show any significance.

In model 2 I test all the variables, including the independent variable analyst coverage, in relation to CSR decoupling. This model shows a positive coefficient of 1.221 between analyst coverage and CSR decoupling. It is the opposite of what was expected, since a negative relationship was hypothesized. It is also noticeable that the relationship, while being positive instead of negative, is significant at the p<0.01 level. The hypothesis must be rejected since the relationship between analyst coverage and CSR decoupling is not in line with the hypothesized expectation, with statistical support.

Table 7. Regressions with CSR decoupling as dependent variable

Variable (1) (2) COV 1.221 *** ROA 0.061 *** 0.054 *** RD_INT -2.304 -2.412 SIZE 3.820 *** 3.552 *** S_GR -2.764 *** -2.765 *** B_SIZE 0.292 *** 0.287 *** INDEP 0.051 *** 0.049 *** DUAL 0.324 0.302 GENDER 0.107 *** 0.103 *** CONSTANT -53.602 *** -51.537 ***

Year Dummies Included Included

Country Dummies Included Included

Industry Dummies Included Included

R-squared 0.257 0.260

n 23,990 23,990

24 5.4 Additional regressions: robustness test

In this research, a robustness test is used to analyze whether different CSR decoupling data leads to the same statistical results. For this robustness test, following research by Garchia-Sánchez et al. (2020), I use the data collected by Hawn & Ioannou (2016). They calculated the gap between external and internal actions as well, which is referred to here as ‘abs_EI_gap’. This CSR decoupling score is based on absolute numbers and performance is lagged by one year. This way, it closely resembles the CSR decoupling scores used in this research. In table 8, multiple models are made.

In the first and second model, I test the first hypothesis with buy and sell recommendations as dependent variable, respectively. Regarding the abs_EI_gap, the

relationship between CSR decoupling and analyst recommendations has changed compared to the main regressions. Where both buy and sell recommendations show a positive association with CSR decoupling in my main regressions, a negative association has been found in the robustness test. However, Since the relationships are not statistically supported in both the main regressions and the robustness test, the change in this relationship does not bare any statistical implications.

In the third model, I test the second hypothesis with the abs_EI_gap as dependent variable. This model shows a significant positive association of analyst coverage with CSR decoupling. This is in line with the findings of the main regressions regarding the second hypothesis.

25

Table 8. Robustness check based on Hawn & Ioannou (2016) CSR decoupling variable

Variable (1) (2) (3) ABS_EI_GAP -50.730 -31.936 COV 0.000 *** ROA 0.401 *** -0.263 *** -0.000 *** LEV -3.062 3.498 * RD_INT 0.0135 -16.102 ** 0.001 SIZE 0.701 -0.664 * -0.002 *** S_GR 0.000 B_SIZE -0.097 -0.041 -0.000 *** INDEP -0.039 0.040 * 0.000 *** DUAL -1.023 0.984 -0.000 GENDER -0.096 ** 0.091 *** 0.000 *** CONSTANT 52.766 *** 8.823 0.040 ***

Year Dummies Included Included Included

Country Dummies Included Included Included

Industry Dummies Included Included Included

R-Squared 0.151 0.133 0.570

n 4,947 4,947 27,809

*p<0.10, ** p<0.05, ***p<0.01 (two-tailed).

5.5 Additional regressions: regressions based on the separate ESG scores

Since the CSR decoupling scores are based on the difference between ESG-scores for disclosure and performance, it can be interesting to analyze the ESG pillars separately. This can improve the overall understanding of CSR decoupling. For the three pillars,

environmental, social, and corporate governance, I use the following formulae:

Edecouplingft = Edisclosureft – Eperformanceft-1

Sdecouplingft = Sdisclosureft – Sperformanceft-1

Gdecouplingft = Gdisclosureft – Gperformanceft-1

I included the results of these regressions in Appendix A. The first panel, panel A, is based on environmental decoupling. The environmental decoupling scores show some interesting

26

results. First of all, environmental decoupling is positively related and significant at the p<0.10 level in relation to sell recommendations. This is in line with my hypothesis, stating that CSR decoupling leads to more sell recommendations. Furthermore, the regression regarding buy recommendations show a negative coefficient of environmental decoupling, which is also in line with my hypothesis. However, this coefficient does not show any significance. The relationship between analyst coverage and environmental decoupling is in line with the second hypothesis, since a significant positive relationship has been found.

The regressions based on social decoupling, in panel B, give some insightful results as well. Compared to environmental scores, the coefficient of social decoupling in relation to buy and sell recommendations show opposite results. Social decoupling shows a negative and positive relationship with buy and sell recommendations respectively. However, no

significance has been found. Analyst coverage is positively related to social decoupling with a high level of significance, which is in line with my second hypothesis.

At last, in panel C, corporate governance decoupling is used as dependent/independent variable. Here, the coefficient of corporate governance decoupling regarding buy and sell recommendations is in line with my first hypothesis. The coefficient for buy

recommendations is negative, and for sell recommendations positive. However, since they are not statistically significant, these numbers can not be supported. Furthermore, a negative relationship between analyst coverage and corporate governance decoupling has been found. This is in line with the second hypothesis. However, this coefficient does not show any significance.

6. DISCUSSION AND CONCLUSION

6.1 Main findings

This research contributes to the existing literature around CSR decoupling and financial analysts, by exploring the relationship between them. CSR decoupling is defined as the difference between CSR disclosure and CSR performance. First, this article explores the effect of CSR decoupling on financial analysts’ recommendations. Second, it examines the effect of analyst coverage on CSR decoupling. The findings in this research are based on 23,990 and 4,947 observations for CSR decoupling and analyst recommendations

27

respectively. The observations include 4,433 companies operating in 49 industries in 58 countries, from 2006 to 2016.

Regarding financial analysts’ recommendations, the results of this research do not show any significance between buy and sell recommendations and CSR decoupling. So the first hypothesis can not be supported. There are a few possible explanations why there is no significant relationship between CSR decoupling and buy/sell recommendations. First, I expected that the majority of the companies in this research were prone to the implementation gap. However, this seemed not to be the case. Of the 28.518 total observations for CSR decoupling, 18.202 of them were understating their CSR performance. Compared to 10.316 companies who overstated their CSR performance, the communication gap is relatively large. Given the fact that the CSR decoupling type is different than I expected, the theoretical explanations which focused on the implementation gap are not completely in line with the CSR decoupling outcomes of this research. Furthermore, since the CSR decoupling score is only based on two separate databases, it might not capture all the relevant data to compute a company’s overall CSR decoupling (Margolis et al., 2009). Another possible explanation of not finding a significant relationship, is that there was relatively few analyst recommendation data, which led to only 4,947 observations. This small size negatively affects the overall regression quality, since it is harder to draw conclusions from a smaller sample size. Interestingly, the additional regression based on environmental decoupling scores partly supports the first hypothesis. It shows a significant positive relationship between

environmental decoupling and sell recommendations.

The second hypothesis states that analyst coverage negatively affects the engagement in CSR decoupling. This hypothesis is derived from the agency theory, where financial analysts can be seen as monitoring mechanisms. When the analyst coverage increases, a company may feel more pressure to reduce their CSR decoupling. The findings of my main regression suggest otherwise: a positive relationship was found. A possible explanation for this positive relationship is the fact that many companies included in the sample understated their CSR performance. The theory regarding the second hypothesis focuses on overstated CSR performance. Consequently, the theory loses some of its relevance in relation to the results. Since the majority of companies in this research engage in CSR decoupling by understating their CSR performance, it can be theorized that analyst coverage is positively associated with CSR performance, but not specifically with CSR disclosure. Since analysts look beyond reported information, companies might feel the pressure to increase their CSR performance but not necessarily their CSR disclosure. Another interesting point to address is

28

that analyst coverage is not significantly related to the decoupling of the ESG pillar corporate governance. This would indicate that analyst coverage does not affect the gap between corporate governance disclosure and corporate governance performance.

6.2 Theoretical and managerial implications

The findings of this research are partially comparable with research by Hawn & Ioannou (2016), who find that companies undertake more internal than external actions regarding CSR. They mentioned a few possible explanations of why internal actions can be more predominant over external actions. First, a company might not disclose due to secrecy policies. Second, companies might experiment with CSR performance before actually disclosing it. Third, they want to mitigate the risks of increased pressure by stakeholders or social movements. The last possible explanation is that companies have a more risk-adverse orientation in terms of analyst perceptions regarding CSR disclosures.

The findings regarding CSR decoupling and buy/sell recommendations implicate that there is no significant relationship between them. Only CSR performance has been examined in relation to analyst recommendations. Luo et al. (2015) found that higher CSR performance leads to more positive recommendations by analysts. The relationship between CSR

disclosure and analyst recommendations has not been examined to date. To get a better understanding of why CSR decoupling is not related to analyst recommendations, more knowledge regarding the relationship between CSR disclosure and analyst recommendations might give us better insights. There are studies which address the opposite relationship. Giannarakis et al. (2016) explored the effect of analyst recommendations on CSR disclosure, but found controversial results.

The results of the regressions for the second hypothesis show that the hypothesis can not be accepted, since a positive significant relationship between analyst coverage and CSR decoupling has been found. This is not in line with research by Garcia-Sánchez et al. (2020), who found that an increase in analyst coverage is negatively associated with CSR decoupling. However, since their coefficient was excessively low, it did not show very high statistical significance. Even though the occurrence of the communication gap may suggest that the theory around the hypothesis is less relevant, Qian (2019) suggested that when analyst

coverage is absent, aggressive disclosure strategies may be adopted by firms. In line with this strategy, a decrease of analyst coverage may result in an increase of the communication gap. This is in line with the second hypothesis in this research, which suggests that analyst

29

hard to conclude which role analysts play when it comes to the occurrence of the communication gap.

6.3 Limitations and future research

This research has some limitations which I point out in this section. Furthermore, I provide some information regarding opportunities for future research.

The first limitation of this study is that the CSR decoupling score is based on the ratings of two different database. First, it led to a relatively broad decoupling score, which can be narrowed down by more qualitative research in the field. It is also possible that the

calculated CSR decoupling scores are not capturing all the relevant information. For example, for disclosure scores, companies have multiple ways of disclosing information regarding their CSR performance. Also, since I used data from other databases, my research is subject to the interpretations and judgment of those who create the databases. This could lead to some reliability issues. According to Margolis et al. (2009), CSR scores based on data by data providers could lead to unreliable findings and spurious relationships. The second limitation is that there was relatively few recommendation data, which decreases the overall reliability on the regressions. For both types of recommendations, there were only 4.380 observations. Third, this study solely focused on public listed firms, since financial analysts are more concerned with this type of company. Consequently, given the scope of this research, CSR decoupling developments at smaller firms are ignored.

This study also offers some opportunities for future research. First, since the results showed that firms were more prone to the communication gap instead of the implementation gap, it would be interesting to increase some focus on this particular type of gap. Prior research regarding CSR decoupling has mostly been focusing on the implementation gap (Hawn & Ioannou, 2016). This study showed that the communication gap is an interesting side of CSR decoupling as well. It can lead to a change in perspective regarding firms

engaging in CSR decoupling. Second, since the measurement of CSR decoupling in this study is relatively broad, it would be interesting to enhance our understanding regarding this

phenomenon by examining it more closely. For example, by conducting qualitative research or examining the ESG metrics more closely. A third possible area for future research is looking closer at the political, social and cultural effects on CSR decoupling. By using a global setting, these factors are, even though controlled for, somewhat ignored. At last, the relationship between CSR disclosure and financial analysts’ recommendations is an

30

area of research. This creates a good opportunity for future research, since it could give insights whether CSR disclosure reduces information asymmetries or is somewhat ignored since the users of the disclosed information are aware of the impression management behind it.

31

7. REFERENCES

Adhikari, B.K. (2016). Causal effect of analyst following on corporate social responsibility.

Journal of Corporate Finance, 41, 201-216.

Bansal, P. (2005). Evolving sustainably: A longitudinal study of corporate sustainable development. Strategic Management Journal, 26(3): 197–218.

Berrone, P., Gelabert, L., & Fosfuri, A. (2009). The impact of symbolic and substantive actions on environmental legitimacy.

Byard, D., Li, Y., & Weintrop, J. (2006). Corporate governance and the quality of financial analysts’ information. Journal of Accounting and Public policy, 25(5), 609-625. Bushman M., Piotrosky J.D., Smith A.J. (2004). What determines corporate tranparency?

Journal of Accounting Research, 42(2): 207–252.

Bolino, M. C., & Turnley, W. H. (2003). More than one way to make an impression: Exploring profiles of impression management. Journal of Management, 29, 141–160 Cespa, G. & Cestone, G. (2007), “Corporate social responsibility and managerial

entrenchment”, Journal of Economics and Management Strategy, Vol. 16 No. 3, pp. 741-771.

Chang, Y., Chen, T. H., Chou, H. H., & Shen, Y. F. (2014). Corporate Social Responsibility and Analyst's Recommendation. International Review of Accounting, Banking & Finance, 6(2).

Cormier, D., & Magnan, M. (2015). The economic relevance of environmental disclosure and its impact on corporate legitimacy: An empirical investigation. Business Strategy and the Environment, 24(6), 431-450.

Cottle, S., Murray, R., & Block, F. (1989). Graham and Dodd’s security analysis (5th ed.). New York: McGraw-Hill.

Crilly, D., Zollo, M., & Hansen, M. T. (2012). Faking it or muddling through? Understanding decoupling in response to stakeholder pressures. Academy of Management Journal, 55(6), 1429-1448.

Daily, C. M. (1995). The relationship between board composition and leadership structure and bankruptcy reorganization outcomes. Journal of Management, 21: 1041-1056.

Deegan, C. (2002). The legitimizing effect of social and environmental disclosures—A theoretical foundation. Accounting, Auditing &Accountability Journal,15, 282–311. Deegan, C., & Rankin, M. (1997). The materiality of environmental information to