CSR Disclosure and The Response Of Financial Analysts:
a Meta-Analysis
Joost Jansen
S2574128
Master Accountancy
Supervisor: dr. A. Bellisario
Co-assessor: Prof. dr. D.M. Swagerman
Date: 24-6-2018
2
ABSTRACT
This study investigates the relationship between CSR disclosure and the response of financial analysts. Current studies have investigated this relationship. However, the results are mixed. Several studies found that CSR
disclosure has positive effects on the response of financial analysts, while other studies suggest that CSR disclosure has no value to analysts and is often not used. By performing a meta-analysis on a set of 31 studies, this study aims to clarify the contrasting evidence, and provide a more conclusive answer to this relationship. Findings of the meta-analysis show that CSR disclosure is significant positively related with analyst coverage, and significantly
negatively related with analyst forecast error.
Acknowledgements: I would like to thank Dr. A. Bellisario for all the help and support during the writing process of this
thesis. Also, I want to thank Prof. N. Hussain, who also helped during the writing process and for his knowledge about meta-analyses. Furthermore, I would like to thank my colleague-students.
3
TABLE OF CONTENT
INTRODUCTION ... 4
THEORETICAL BACKGROUND AND HYPOTHESIS DEVELOPMENT ... 7
Legitimacy theory ... 7
Agency theory ... 8
Signaling theory ... 10
Stakeholder theory ... 11
METHOD ... 13
Identification and screening of the records ... 13
Meta-analytical procedure ... 16
RESULTS ... 18
Analyst coverage ... 19
Analyst forecast accuracy ... 20
Analyst forecast error ... 20
Information asymmetry ... 21
Publication bias ... 22
DISCUSSION AND CONCLUSION ... 23
Theoretical implications ... 23
Conclusion and future research ... 25
REFERENCES ... 27
Appendix ... 35
4
INTRODUCTION
Corporate Social Responsibility (CSR) is becoming a “part of the language of Wall Street”, says Gregg Sgambati, head of ESG Solutions at S-Network Global Indexes, in a recent interview with Christopher P. Skroupa. Sgambati is responsible for the firms environmental, social and governance (ESG) products, which create ratings and ESG indexes for the firm. According to Sgambati, ESG is the tool that, because of its quantifiable nature, helps
managers with their investment decisions by integrating sustainability into their decisions. Large investors are also using ESG to assess corporate sustainability, making it a part of the language of Wall Street. Subsequently, companies want to improve their stock price, and if CSR is of interest to analysts, then companies will pay more attention to CSR.
The fact that companies are paying more attention to CSR in recent years can be seen in the 2015 United Nations Sustainable Stock Exchanges (SSE) Model Guidance on Reporting ESG information. According to this report, there has been a remarkable growth in disclosure about ESG issues in recent years. Investors demand more than just financial information, and by reporting on ESG issues, these demands can be addressed (2015 Model Guidance). In the 2016 SSE Report on Progress, the SSE states that there are 58 stock exchanges partnering with the SSE initiative. All these exchanges are making a commitment to improve the sustainability in their markets. Another more recent article of the SSE Initiative states that seven stock exchanges have committed to the campaign of providing guidance on ESG reporting to the issuers (SSE, 2017). The goal of this campaign is to engage and support listed companies to improve their disclosure about ESG issues. These reports show that the disclosure about ESG is becoming of increasing importance.
This increasing importance of disclosure can be particularly seen in the literature on CSR. In the paper of Dhaliwal et al. (2012) a sample of 31 countries was used to examine the relationship between CSR disclosure and analyst forecast accuracy. From this sample, fewer than 100 companies issued a stand-alone CSR report in 1990. In 2007, the number of companies from this sample that issued a stand-alone CSR report was more than 1,000, showing that over the last years, the amount of companies that are disclosing information about CSR issues is growing.
The amount of disclosure a company provides, is a determinant for the number of financial analysts following a firm (Lang and Lundholm, 1996). Financial analysts have the role of providing information such as earnings forecasts or recommendations to investors and brokers, integrating the financial analysts as a part of the capital market (Lang and Lundholm 1996). The authors provided evidence that the number of analysts following a firm is higher for a firm with more informative disclosure policies. However, when firms publicly disclose
information, this can lead to a decline in the demand for services of analysts, because analysts can no longer distribute private information to investors (Healy and Palepu, 2001). Following these findings, and the increasing
5
importance of CSR disclosure, the effect of CSR disclosure on financial analysts’ responses looks like a subject gaining increasing attention within the CSR literature. However, there have been several studies on such relationships, but with outcomes that present contrasting evidence.
According to Gao et al. (2016), firms with higher CSR performance tend to provide better CSR disclosures. More specifically, they found that firms with higher CSR disclosure attract greater analyst coverage. Other studies also found that CSR disclosure increases analyst coverage (Bernardi and Stark, 2016; Cormier and Magnan, 2014). Firms that initiate CSR disclosure have a greater analyst coverage, while these analysts also achieve lower forecast errors (Dhaliwal et al., 2011). Another study by Dhaliwal et al. (2012) found that higher CSR disclosures lead to better analyst forecast accuracy and a higher analyst coverage. Additionally, firms with better CSR report disclosures are associated with higher analyst forecast accuracy (Muslu et al., 2017). Firms with a higher analyst coverage are also more likely to disclose information about CSR, which may be caused by the pressure of the higher analyst coverage on firms to disclose information (Dhaliwal et al., 2014). Cormier and Magnan (2015) found that environmental disclosure contains useful information for financial analysts. These studies show and conclude that the disclosure of CSR information leads to more analyst coverage and also increases the accuracy of the forecasts of analysts.
However, other studies present contrasting results. According to Lee et al. (2016), more disclosure of CSR leads to a more transparent information environment, which reduces the role of financial analysts. Because there is more CSR disclosure, the value relevance of the analyst recommendations is reduced, which makes it difficult for analysts to contribute to price discovery. Perego et al. (2016) interviewed several experts in the area of integrated reporting, which combines financial and non-financial information in one report. According to the experts, most external reports that firms disclose lack quality, and firms are not recognizing the potential value of such a report. More specific evidence comes from the study of Campbell and Slack (2011). The authors found that environmental disclosures have no relevance for financial analysts. They came to this conclusion by finding evidence that the environmental information that was disclosed was generally unread by analysts because the information had no use to them. Deegan and Rankin (1997) also found that environmental information does not have material value to financial analysts and their decisions. In an experiment by Milne and Chan (1999), results showed that investment analysts generally do not pay attention to narrative social disclosures. A more recent survey among financial analysts showed that financial analysts rarely use CSR disclosures for their reports (Krasodomska and Cho, 2017). These studies suggest that CSR disclosures have little use for financial analysts. Aerts et al. (2008) found that while environmental disclosure decreases analyst forecast dispersion in continental Europa and North America, the effect of environmental disclosure in environmental sensitive industries in continental Europa is that it actually increases the dispersion between analyst’s forecasts.
6
Overall, as the previous studies show, there are conflicting results about the effect of CSR disclosure on the response of financial analysts. Because of these uneven conclusions, this thesis presents a meta-analytic review of studies addressing this relationship. By performing a meta-analytic review, in particular, I aim to provide a conclusive, integrated picture of current knowledge revolving around this relationship. Being grounded into a statistical procedure that allows to reliably synthesize previous research achievements, in fact, a meta-analytic review of literature affords the drawing of far-reaching conclusions on fragmented evidence of a specific field of study, which would not be available from a single study (Tranfield et al., 2003).
Although there are several meta-analytic reviews in the current literature on CSR, these papers focus on corporate social performance (CSP) or on the effect of CSR on financial performance (e.g., Orlitzky and Benjamin, 2001; Allouche and Laroche, 2005; Wang et al., 2016), and they exclude evidence on the effects of CSR disclosure on financial analysts’ response. Thus, this study will be explicitly targeting this gap by asking the following research question:
How does Corporate Social Responsibility Disclosure affect the response of financial analysts?
Using a random-effects model to perform the meta-analysis, findings show a significant relationship between CSR disclosure and analyst coverage, and analyst forecast error.
The remainder of this paper is structured as followed. In the next section, the theory and hypotheses development will be set out. After that, the research methodology that is used will be further explained. This is followed by the results of the meta-analysis. Lastly, the results will be discussed and a conclusion is formulated, followed by the limitations of the study and suggestions for further research.
7
THEORETICAL BACKGROUND AND HYPOTHESIS
DEVELOPMENT
Epstein and Freedman (1994) have studied that shareholders have interest in reports about their firm’s social activities, including the firm’s environmental activities. More recent studies confirm this, as managers and investors both have an increasing interest in the activities of a firm in the area of corporate social responsibility (Lee et al., 2016). In relation to financial analysts, firms with more disclosure are followed by more analysts (Lang and Lundholm, 1996). The CSR activities of a firm have an effect on the firm value. Because of this, information about the CSR activities can be useful for financial analysts in the forecasting process (Dhaliwal et al., 2012). In this section, I will discuss the disclosure of CSR, and the effect of disclosure on financial analysts, according to different theoretical perspectives and develop hypotheses following these theoretical perspectives. As of right now, there is no accepted theory for social and environmental accounting, but there are several theoretical perspectives used to explain why social or environmental information is disclosed (Deegan, 2002). Based on previous literature (An et al., 2011; Deegan and Rankin, 1997; Su et al., 2016, Healy and Palepu, 2001) I will discuss the relationship between CSR disclosure and financial analysts using four theories that are often used, which are the legitimacy theory, agency theory, signaling theory, and stakeholder theory. An et al. (2011) have used these theories to create a framework to explain voluntary IC disclosure. This framework, and these theories, can also explain the disclosure of CSR information, since the drivers for CSR disclosure are the same as the drivers for voluntary IC disclosure (An et al., 2011). Therefore, these four theories are used to explain the disclosure of CSR information. They will be further discussed in the remainder of this section.
Legitimacy theory
The legitimacy theory suggests that firms disclose information about CSR to legitimize the actions of the firm. Firms have a social contract with society, in which firms act socially desirable and in return get approval for their actions from society, which is needed for survival of the firm (Guthrie and Parker, 1989). Legitimacy is defined as a “generalized perception or assumption that the actions of an entity are desirable, proper, or appropriate within some socially constructed system of norms, values, beliefs, and definitions.” (Suchman, 1995, p. 574). The
disclosure of CSR can therefore be seen as a tool for firms to legitimize their actions. If the actions of a firm do not appear to be legit, or when they are not approved by society, then society can act in such a way that harms the continuity of the firm (Deegan and Rankin, 1997).
The management of an organization will increase its environmental disclosure when the legitimacy of a company is being threatened (Brown and Deegan, 1998). To become legitimate, an organization can attempt to shape the definition of legitimacy towards the organization’s practices, which can be achieved through
8
communication (Dowling and Pfeffer, 1975). However, shaping the definition of legitimacy towards the
organization’s practices is a difficult process. Therefore, it is more likely that a firm will adapt its own practices to what is believed to be legitimate. A way organizations can do this, in order to appear to be legitimate, is to disclose environmental information (O’Donovan, 2002). With the disclosure of information, an organization and its
management can affect the external perceptions on their organization (Deegan, 2002). Organizations that are environmental unfriendly disclose more information about their environmental performance than firms that are environmental friendly (Cormier et al., 2011), in order to deal with the threats to the legitimacy of the organization (Cho and Patten, 2007).
The disclosure of environmental practices affects how financial analysts witness the legitimacy of a firm (Cormier and Magnan, 2015). Information about CSP that is disclosed by companies, is used by analysts for creating their recommendations (Luo et al., 2015). When the disclosed information is inconsistent with the
environmental performance, the credibility of the information decreases for financial analysts, and also for a larger community (Cormier and Magnan, 2015). However, Cormier and Magnan (2015) found that environmental disclosure adds to the quality of financial analysts’ forecasts. Firms that face social pressures will disclose information about their environmental actions, to deal with the threatened legitimacy, thereby also decreasing information asymmetry (Clarkson et al., 2008; Cormier et al. 2011). Managers want to legitimize the environmental activities of the firm, which has an effect on the disclosure. This disclosure reduces forecast dispersion, suggesting a lower deviation in the forecasts of analysts (Aerts et al., 2008). Studies have found results in line with legitimacy theory, that low environmental performance is positively related with environmental disclosures (Cormier and Magnan, 2014; Patten, 2002), suggesting that these firms want to deal with threats of legitimacy (Cho and Patten, 2007). These disclosures also positively affect the number of analysts following the firm and improves their forecasts (Cormier and Magnan, 2014). While firms with low environmental performance do disclose information in an attempt to maintain or obtain their legitimacy, this disclosure is of low quality so that their actual performance is not fully revealed (Hummel and Schlick, 2016; Li et al., 2017).
In view of these arguments, firms want to be legitimate to society. Disclosure of information regarding CSR is one way to appear to be legitimate. These disclosures affect the response of financial analysts. Therefore, I hypothesize that:
Hypothesis 1: CSR disclosure is positively related to the response of financial analysts
Agency theory
Another relevant theory is the agency theory. Jensen and Meckling (1976) define an agency relationship 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” (p. 308).
9
When talking about the principal-agent relationship, in most cases this relationship is between the shareholders (principal) and the management of a firm (the agent) (An et al., 2011). The theory assumes that the actions of the agent do not always act in the best interests for the principal (Jensen and Meckling, 1976). This is because the principal and agent want to maximize their own returns, which could result in different interests for each party, leading to agency problems (An et al. 2011; Jensen and Meckling, 1976). In order to solve agency problems, one possible solution is the disclosure of relevant information (Healy and Palepu, 2001).
These different interests for each party lead to information asymmetry. Information asymmetry is the phenomenon where one party has more information than the other party, leading to an advantage for the party possessing more information (An et al. 2011). This is also applicable to CSR, because managers have more information about the CSR activities of a firm than shareholders and outsiders (Cui et al., 2016). According to Hahn and Kühnen (2013) the performance in the area of sustainability could be a possible cause for asymmetric information between management and other stakeholders. For this reason, a company has an incentive to report on their sustainability performance and activities. For example, it can be difficult for an outsider to obtain information on a firm’s activities regarding sustainability (Hahn and Kühnen, 2013). One way to reduce information asymmetry is transparency (Martínez-Ferrero et al., 2017). In order to be more transparent, organizations disclose information about their CSR actions (Cormier et al. 2011).
Cormier et al. (2009) found that quantitative and indicative social disclosure leads to lower information asymmetry, where indicative social disclosure only slightly decreases information asymmetry. This suggests that only quantitative social disclosure has information value for market participants. Cormier et al. (2011) found that both social and environmental disclosure reduce information asymmetry. The authors found that while both forms of disclosure substitute each other in reducing information asymmetry, the greatest impact on reducing information asymmetry comes from the disclosure of environmental debts, risks and litigations. As mentioned earlier, the sustainability performance of a firm could be seen as asymmetric information. By reporting this information, the information asymmetry can be reduced, suggesting that disclosure of CSR performance leads to a decrease in information asymmetry. (Hahn and Kühnen, 2013). More specifically, Martínez-Ferrero et al. (2016) studied the relationship between voluntary disclosure of CSR information and information asymmetry, finding a negative relationship. They conclude that the voluntary disclosure of CSR information offers more information to the market, which reduces information asymmetry. The disclosure of CSR-related information is higher for firms that have a reputation of having a negative influence on the environment, which are found to report more environmental information (de Villiers and Staden, 2011). One reason for this is that the managers of these firms want to reduce information asymmetry. Other previous studies have also shown that higher social and sustainability reporting quality leads to a decrease in information asymmetry (Cuadrado-Ballesteros et al., 2016; Hummel and Schlick, 2016). The disclosure of social and sustainability information turns private information into public information. Information which is of higher quality is appreciated by investors, who evaluate investment opportunities in the
10
market (Cuadrado-Ballesteros et al. 2016). In a study by Zhong and Gao (2017) the authors found a positive relationship between CSR disclosure and investment efficiency. According to the authors, firms with more CSR disclosure have a higher investment efficiency, which results from a decrease in information asymmetry. This suggests that CSR disclosure has a negative effect on information asymmetry. Michaels and Grüning (2017) also found a negative relationship between CSR disclosure and the bid-ask spread, which the authors used as a proxy for information asymmetry. Their results suggest that CSR disclosure is of value and can also lead to other economic benefits. Analysts also contribute to decreasing the information asymmetry. Because analysts include information about CSP in their recommendations, more information is available for investors, therefore reducing the
information asymmetry (Luo et al. 2015). Supported by such evidence, I hypothesize that:
Hypothesis 2: CSR disclosure is negatively related to information asymmetry
Signaling theory
Signaling theory can also be used to explain the disclosure of CSR information. Originally, Spence (1973) used the signaling theory for the job market. The signaling theory assumes that an employer hires an individual under uncertainty, because the employer is not sure of the productivity of the applicant at the time of hiring. However, the job applicant can send signals, which give information about the job applicant and the contribution of the applicant. Spence (1973) defines signals as “observable characteristics attached to the individual that are subject to
manipulation by him” (p. 357). An example is an applicant acquiring education. The applicant will invest in education if there is an economic benefit to achieve, in this case when the offered wage of the job is higher than the education costs (Spence, 1973; Verrecchia, 1983). While the signaling theory is based on a job market model, it is also applicable to firms disclosing information, where firms can send a signal, in the form of information, to investors (Verecchia, 1983). Information asymmetry leads to problems between parties. Signaling theory is
concerned with handling these problems, which connects the signaling theory to the agency theory (An et al. 2011). While the signaling theory is not specifically implied by the agency theory, a considerable overlap exists between the theories (Morris, 1987).
Firms disclose CSR information to send a signal to their stakeholders (Lee et al., 2016). According to Lee et al. (2016), by signaling this, the firm provides more information to the stakeholders, in order to serve the need of their information. By sending a signal, in the form of disclosure, firms decrease information asymmetry because they disclose information that investors do not have (Watson et al., 2002). An important aspect of the signaling theory is that the signal that is communicated by a firm usually contains positive information. By doing so, the firm shows its positive qualities to outsiders (Connelly et al., 2011). For this reason, managers want to send a signal because they want their firm to be distinguished as a firm of high quality or performance. In order to do this successfully, the signal must be credible (Watson et al., 2002; Xiao et al., 2004). Managers can distinguish the firm
11
as a firm of high quality, because there is information asymmetry and investors are therefore not fully aware of the quality of the firm (An et al. 2011). The study of Mahoney et al. (2013) has shown that firms that voluntarily disclose CSR information, in the form of standalone CSR reports, generally have high CSR performance scores. Similar results were found in a study by Clarkson et al. (2008), which shows that firms with superior environmental performance, voluntarily disclose more information about their environmental performance and consequences. These results suggest that firms primarily disclose CSR information to signal their quality to stakeholders.
Following the perspective of the signaling theory, according to which the information that firms disclose usually contains positive information, Lang and Lundholm (1996) provide evidence that more analysts will follow a firm with informative disclosure policies. When a firm discloses more information, the demand for analyst reports increases as well, leading to an increase in the number of analysts that follow a firm with more disclosure. Firms with more voluntary disclosures are also followed by more analysts (Healy et al., 1999). Moreover, financial analysts are interested in the voluntary disclosure of strategic, non-financial and governance information, which could suggest an interest in information about CSR (Hamrouni et al., 2017). This relationship was found in previous studies, which found a positive relationship between the CSR disclosures of a firm, and the number of analysts following the firm (Qiu et al., 2016; Dhaliwal et al. 2012; Baldini et al., 2016; García-Sánchez and Noguere-Gámez, 2017). In line with the signaling theory, where the communicated signal is usually positive, a firm’s past profitability has a positive effect on the disclosure of information about their social activities (Connelly et al. 2011; Qiu et al. 2016). Firms with superior social responsibility performance and who voluntarily disclose information about CSR, attract more analysts (Dhaliwal et al., 2011). In addition, firms with higher CSP also provide greater quality disclosures, leading to a higher analyst coverage, because the greater quality of the disclosure helps the analysts to gather information (Gao et al., 2016).
In view of these arguments, I hypothesize that:
Hypothesis 3: CSR disclosure is positively related to analyst coverage
Stakeholder theory
Stakeholder theory suggests that organizations should focus on the relationships it has with its stakeholders in society (An et al., 2011). A stakeholder is “any group or individual who can affect or is affected by the
achievement of the organization’s objectives” (Freeman, 1984, p. 46). There are stakeholders who demand that managers engage in CSR (McWilliams and Siegel, 2001). According to the stakeholder theory, an organization and its management should act in a way that is important to their stakeholders. More specifically, when an organization engages in CSR activities, conflicts between managers and stakeholders can be resolved, resulting in higher firm
12
value (Jo and Harjoto, 2012). Besides undertaking activities that is in the interest of stakeholders, the organization should also disclose information to their stakeholders about these activities (Guthrie, 2006).
The concept of stakeholder theory is that an organization should be accountable for its actions, which can be achieved with the disclosure of information (An et al., 2011). Gelb and Strawser (2001) argue that firms have incentives to engage in activities that are socially responsible, one such an activity is disclosing information. The increasing demand of CSR disclosure puts pressure on managers to disclose information about CSR activities (Lee et al., 2016). Additionally, managers have an incentive to disclose information to show stakeholders that they are conforming to their expectations (Deegan, 2002). Firms can voluntarily disclose CSR information to serve the need of stakeholders. This disclosure about CSR activities contributes to the information environment, which may make it more transparent. This can align the interests of stakeholders and which can also affect the demand for financial analysts (Lee et al. 2016).
According to Hope (2003), there is a positive relation between firm-level disclosure and forecast accuracy. The information disclosed by firms provides analysts with more information, and can also provide information on future earnings, which can increase the accuracy of analysts’ forecasts (Lee et al. 2016; Lang and Lundholm, 1996). The disclosure of forward looking non-financial information also increases analyst forecast accuracy, and reduces dispersion between the forecasts of analysts as well (Vanstraelen et al. 2003). Aerts et al. (2008) found similar results in their study on environmental disclosure. They found that the analyst forecast dispersion is lower when a firm discloses information about its environmental activities, which suggests that the forecast accuracy increases. More specifically, studies have found a positive relationship between CSR disclosure and analyst forecast accuracy, which suggests that the issuance of a CSR report increases the forecast accuracy because of the increased transparency for analysts (Dhaliwal et al. 2011; Dhaliwal et al. 2012). The findings further show that this relationship is stronger in countries that are stakeholder-oriented, which have a business culture where stakeholders are seen to have a legitimate interest in the activities of companies (Dhaliwal et al. 2012). Cormier and Magnan (2015) found that both financial, and non-financial stakeholders would benefit from clear information about activities regarding the environment. The issuance of a stand-alone CSR report, which are addressed to all stakeholders, are useful for analysts, leading to lower forecast errors, thus improving the forecast accuracy (Garrido-Miralles et al., 2016). In a study by Muslu et al. (2017), disclosure scores were assigned to CSR reports. They found that CSR reports with higher scores lead to a higher analyst forecast accuracy, therefore emphasizing the importance of the CSR reporting quality.
Thus, according to the previous arguments, firms will disclose more CSR information as a result of the pressure coming from stakeholders. The disclosure leads to a more transparent information environment, which may increase the accuracy of analyst forecasts. In view of these arguments, I hypothesize that:
13
METHOD
This study performs a systematic review of the literature in the form of a meta-analysis. Meta-analysis offers a statistical procedure to synthesize the results of different studies, to obtain an overall reliable answer which would not be available from a single study (Tranfield et al. 2003). As mentioned earlier, the results from previous studies on the relationship between CSR disclosure and the response of financial analysts, come to different conclusions. Therefore, there is a need to clarify the fragmentation from the current literature. A meta-analysis is a useful statistical tool to do this, since a meta-analysis is effective in synthesizing inconsistent result across studies, by comparing studies with similar characteristics, to provide a conclusive answer or a summary effect. (Pomeroy and Thornton, 2008; Song et al., 2008; Borenstein et al., 2009). For the collection of papers to be included in the meta-analysis, the process of Moher et al. (2010) was followed, which includes identification, screening, determining the eligibility of papers and finally the decision whether to include papers in the meta-analysis or not. These steps are explained in the following paragraphs.
Identification and screening of the records
In order to collect a comprehensive dataset with respect to CSR disclosure and financial analysts, a literature search was performed in different ways. To perform an effective search, a set of keywords was created for the search. These keywords were based on literature in the area of CSR (Tranfield et al., 2003), by looking at the different dimensions of CSR. The set of keywords should be large enough in order to find a sufficient quantity of papers, but also precise so that irrelevant papers are eliminated (Duff, 1996). This led to the following keywords being used in the identification process (table 1):
Table 1: Keywords used
Category Keywords
CSR Disclosure Corporate social disclosure; Social disclosure; Environmental disclosure;
Environmental reporting; Sustainability disclosure; Philanthropy disclosure; ESG reporting; ESG disclosure; Corporate philanthropy
Financial analysts Analyst coverage; Analyst following; Analyst forecast; Analyst forecast error; Analyst recommendations; Information asymmetry
14
The search terms were used in a combination of a disclosure search term and an analyst search term. To make sure that both search terms are incorporated in the search, a Boolean operator (AND) was added in the search field. Searches for papers were made primarily by using the SmartCat search engine from the University of Groningen. SmartCat is the search engine of the University of Groningen, which includes all the printed and electronic collections of the university, thus providing a large coverage for papers. In addition to SmartCat, searches were made on the SSRN website and Google Scholar. For the meta-analysis, papers were searched that are written in English. In order to get a sufficient quantity of papers, there is no limit to the years that the papers are published.
A possible bias for meta-analysis is the publication bias, or the file-drawer problem (Rosenthal, 1979). The publication bias occurs because it is more likely for significant results to be published than non-significant results (Field and Gillet, 2010). According to Pomeroy and Thornton (2008), by using both published and unpublished papers, the effect of publication bias can be reduced. However, the quality of unpublished papers may be inconsistent, since they are not fully peer-reviewed (Pomeroy and Thornton, 2008). In order to mitigate the publication bias, the papers included in the meta-analysis also include non-published papers (i.e. working papers). These were searched through SSRN or Google Scholar. The searches made in SmartCat, SSRN and Google Scholar led to the identification of 1155 papers.
Next to the database searches, additional searches were made through other sources. One other source was by going through previous meta-analysis studies in the area of CSR. Additionally, papers that were found through the database searches and that could be of interest for the meta-analysis were also scanned. While there are no previous meta-analytical reviews studying the relationship between CSR disclosure and the response of financial analysts, there are previous meta-analytical studies studying other relationships in the area of CSR (Orlitzky and Benjamin, 2001; Orlitzky et al., 2003). The reference lists of these studies were scanned, in order to identify additional studies that could be of interest and use for the meta-analysis (Duff, 1996; Greenhalgh and Peacock, 2005). Lastly, papers recommended by personal contacts were examined. My supervisor and colleague-students contributed with references which could be of interest for the meta-analysis. These additional searches through other sources led to the identification of 185 records. After removing duplicates that were found in the searches, a total of 615 papers were identified by searching in databases and additional sources.
Following the identification of the papers, papers that were clearly not about the relationship between CSR disclosure and financial analysts, or which are not written in English, were excluded. This led to the exclusion of 443 papers. From the remaining 172 papers, the titles and abstracts were screened for relevance. This was done in order to check if the papers were about the relationship between CSR disclosure and financial analysts. By screening the articles, 36 papers were excluded from the sample.
15
Eligibility and inclusion
After the screening process, the remaining 133 papers that remained were checked for eligibility. This was done by looking at the title, abstracts, method, results and conclusions of the study. For a paper to be included in the meta-analysis there are several requirements. First, as this meta-meta-analysis is about the relationship between CSR disclosure and financial analysts, papers that were about CSR in general or CSR performance and did not mention disclosure, were excluded. Second, besides having a CSR-disclosure related variable, the paper should also include a financial analysts-related variable and vice versa. There are multiple variables for each aspect, but the papers should at least include a variable in each aspect. For example, a paper can have a variable for the issuance of a CSR report (CSR disclosure-related) and a variable measuring the number of analysts following the firm (financial analysts-related). The following financial analysts-related variables are used for the meta-analysis: “analyst coverage”, “analyst forecast accuracy”, “analyst forecast error”, “analyst recommendations” and “information asymmetry”. The papers were checked for these, or similar, variables and their definitions to see if the variables are suitable for the analysis. Some papers do not include the specific variables mentioned. However, it could be possible that another variable is used as a proxy. For example, Clarkson et al. (2008) use the stock price volatility as a proxy for information asymmetry.
Another requirement is that there should be a statistical method performed to measure the effect of these variables on each other. This meta-analysis will be performed using correlation coefficients. There were some papers found that did not include a correlation matrix, or that did not report a correlation coefficient of the relationship between the CSR-disclosure variable and the financial analyst-related variable. The authors of these papers were contacted to ask if they could provide the correlation matrix or value. However, after receiving answers from these authors, it was not possible to receive these data.
From the 133 papers that were checked for eligibility, 104 were excluded because they did not meet the requirements mentioned above to be included in the meta-analysis. This resulted in a final sample of 29 papers to be included in the meta-analysis, with papers from 2001 till 2017. A list of studies included in the meta-analysis can be found in Appendix B. Information and characteristics of these studies were collected in a spreadsheet. This includes general information about the studies, but also information about the correlation values of the relationships between the CSR Disclosure-related variable and the financial analyst-related variable. Some studies that were found reported multiple correlations on different aspects of CSR disclosure (i.e. environmental disclosure and social disclosure). If this occurred, the correlations were averaged to one correlation value. Figure 1 shows a summary of the search process that was performed for the meta-analysis.
16
Meta-analytical procedure
To perform the meta-analysis, the HOMA meta-analytic procedure was followed (Hedges and Olkin, 1985). There are two statistical models which are used for a meta-analysis, the fixed-effects and random-effects model (Hedges and Vevea, 1998). Within the fixed-effects model, it is assumed that there is one true effect size for all studies, while the random-effects model assumes that this varies between studies (Borenstein et al., 2009). Because the studies included in the sample are performed independently, it is not likely that the studies are functionally equivalent and have one true effect size. Therefore, in this meta-analysis, the effects model is used. In addition, the random-effects model is also more appropriate to generalize to other studies outside of the sample (Hedges and Vevea, 1998), which allows researchers to draw conclusions beyond studies included in the sample.
Following the HOMA meta-analytic procedures, the effect sizes of each studies were combined to a weighted mean effect size. These procedures were performed using Meta-Essentials workbook by Suurmond et al. (2017). The data of the papers in the sample was collected in a spreadsheet. From this spreadsheet, the following data was taken
1155 of records identified
through database searching
185 of additional records
identified through other searches
615 records after duplicates
removed
172 records screened
133 full text articles assessed
for eligibility
29 studies included in
quantitative synthesis
39 records excluded
104 records
excluded
443 records excluded
17
to enter into the software to run the meta-analysis: study name, correlation value, and the number of observations. This was done for the relationship between (1) CSR disclosure and analyst coverage, (2) CSR disclosure and analyst forecast accuracy, (3) analyst forecast error, and (4) information asymmetry. There is no meta-analysis performed on the relationship between CSR disclosure and analyst recommendations. There were however 3 studies found that reported a correlation on this relationship. However, since there are only three studies which reported data, no reliable analysis can be performed.
The software ran the analysis to calculate the weighted mean of correlations. However, this is not done using the correlation values. First, according to the HOMA-method, the correlations values are converted to Fisher’s Z values using the following equation (Field, 2005):
!"#= 1 2 ()*
1 + , -1 − ,
-The analysis is then performed using the Fisher’s Z values. -The weighted average of these values is calculated, after which the values are converted back to a correlation value, using the equation (Field, 2005):
,- =
/(12#)− 1 / 12# + 1
18
RESULTS
In this section, the results of the meta-analysis are presented. The main results are presented in table 2. It shows the results for the relationship between CSR disclosure and the analyst-related variables. Consequently, it reports the number of studies included in the meta-analysis, the total sample size, the mean correlation, the 95% confidence interval (CI), the Q-statistic, the I² index, and the failsafe-N. The Q-statistic and I² index are to ways to assess the heterogeneity in the meta-analysis. While the Q-statistic can be used to assess heterogeneity, it does not assess to the extent of true heterogeneity. Therefore, the I² index is more suitable. The I² index measures the total variability in the meta-analysis that is due to true heterogeneity (Huedo-Medina et al., 2006). For example, with an I² index of 75%, if all studies would have the same sample size, the dispersion between the studies would be 75% of the dispersion with different sample sizes. According to Huedo-Medina et al. (2006), percentages of 25%, 50%, 75% respectively represent low, medium and high heterogeneity.
Table 2: Results Relationship between CSR disclosure and: K N Mean (r) 95% C.I. min 95% C.I. max Q(p) I² Fail-safe N Analyst coverage 18 179.874 0.227 0.16 0.30 737,196 (0) 97,694 25.283 Analyst forecast accuracy 5 1.119 0.060 -0.073 0.192 6,470 (0,167) 38,174 0
Analyst forecast error 8 125.208 -0.045 -0.087 -0.002 26,209 (0) 73,292 154 Information asymmetry 8 14.267 -0.071 -0.119 0.017 28,383 (0) 75,337 75
K = number of effect sizes; N = Total sample size; Mean = mean of population correlation; 95% C.I. = 95% confidence interval of p; Q(p) = Cochran’s homogeneity test statistic (probability of Q); I² = scale free index of heterogeneity, Fail-safe N: the number of studies that need to be added to make the results insignificant.
19
Analyst coverage
The analysis on the relationship between CSR disclosure and analyst coverage consisted of 18 studies with a total amount of 179.874 observations. The mean correlation of the relationship between CSR disclosure and analyst coverage is 0.227, with a 95% CI between 0.16/0.30. This value shows that CSR disclosure is significantly positively related with analyst coverage (p < 0.001), which means that more CSR disclosure leads to a higher analyst coverage. Hypothesis 3, which suggests that CSR disclosure is positively related to analyst coverage, can therefore be accepted. The Q-statistic of analyst coverage is 737,196 and the I² index is 97.7%, indicating that a very high degree of heterogeneity exists between the studies included in the analysis. Figure 2 shows the forest plot of the meta-analysis with the analyst coverage variable. The dots represent each study. The lines that are connected to the dots represent the confidence interval. At the bottom, the mean correlation is presented, which is also the value that is presented in table 2. Studies with a larger sample size have a smaller confidence interval.
Figure 2: Forest plot analyst coverage. ● = Effect size of study, the lines connected to ● show the 95% confidence interval. The last ● is the mean effect size.
20
Analyst forecast accuracy
For the relationship between CSR disclosure and analyst forecast accuracy, 4 studies were included in the meta-analysis, with a total amount of 1.119 observations. However, one study did have two different samples, which were both analyzed. The mean correlation is 0.060, with a 95% CI between -0.073/0.192. The value shows that CSR disclosure is positively, but not significantly (p > 0.05) related with analyst forecast accuracy. Based on this outcome no support was found for hypothesis 4, which suggests that CSR disclosure is positively related with analyst forecast accuracy. The Q-statistic (Q(p) = 6,47) and I² index (38,17%) show some signs of heterogeneity, although there is not as much heterogeneity as the other analyses with other variables. The I² index (38,17%) means that 38,17% of the total variability among the effect sizes is a result of true heterogeneity, and thus not caused by sampling error. Figure 3 shows the forest plot of the meta-analysis on analyst forecast accuracy.
Figure 3: Forest plot analyst forecast accuracy. ● = Effect size of study, the lines connected to ● show the 95% confidence interval. The last ● is the mean effect size.
Analyst forecast error
The mean correlation for CSR disclosure and analyst forecast error is -0.045, with a 95% CI between -0.087/-0.002. There were 8 studies included in this analysis, with 125.208 observations. The results indicate a significant negative relationship (p < 0.05) between CSR disclosure and analyst forecast error. Many studies used in this analysis use analyst forecast error as an inverse measurement of analyst forecast accuracy. Therefore, based on this analysis, hypothesis 4 can be accepted. Because the analysis between CSR disclosure and analyst forecast accuracy did not show a significant relationship, hypothesis 3 can be moderately accepted. As with analyst coverage, the meta-analysis between CSR disclosure and analyst forecast error shows signs of high heterogeneity (Q(p) = 26,209, I² = 73,29%). In this analysis, 73,29% of the total variability among the effect sizes is due to true heterogeneity.
21
Figure 4: Forest plot analyst forecast error. ● = Effect size of study, the lines connected to ● show the 95% confidence interval. The last ● is the mean effect size.
Information asymmetry
Lastly, the mean correlation for CSR disclosure and information asymmetry is 0.071, with a 95% CI between -0.159/0.017, indicating a negative, and non-significant relationship (p > 0.05) between CSR disclosure and information asymmetry. Based on these results, hypothesis 2, which suggests that CSR disclosure is negatively related to information asymmetry, is rejected. This analysis included 8 studies with a total amount of 14.267 observations. The results also show signs of high heterogeneity (Q(p) = 28,383, I² = 75,34%), suggesting high variety between the studies, not due to sampling error but due to true heterogeneity.
Figure 5: Forest plot information asymmetry. ● = Effect size of study, the lines connected to ● show the 95% confidence interval. The last ● is the mean effect size.
22
Based on the meta-analyses that were performed in this study, hypothesis 1 can be moderately accepted. Hypothesis 1 states that CSR disclosure is related to the response of financial analysts. Two out of four meta-analyses showed significant results for the response of financial analysts to CSR disclosure, thus hypothesis 1 can be moderately accepted.
Publication bias
In order to test for publication bias, failsafe N tests were performed to the failsafe number. One way of calculating this is by the method of Rosenthal (1979). This method is used to calculate the number of studies that must be in the sample to make the results of the meta-analysis insignificant. This is done in order to address the “file-drawer” problem. According to Rosenthal (1979), the extreme view of the file-drawer problem is that 95% of the studies that found non-significant result are within file-drawers in the lab, meaning they are not published. The tolerance level for the publication bias is set at 5K + 10 (K = number of studies). If the failsafe number is higher than this amount, then publication bias does not apply. The calculation of the fail-safe N is by the following equation (Rosenthal, 1979), where X is the number of studies required to make the result of the meta-analysis non-significant:
4 = 5
2,706 [5 !; 1− 2,706]
Using this method, the failsafe numbers were calculated for the meta-analysis by using the Meta-Essentials software by Suurmond et al. (2017). The fail-safe numbers were calculated for analyst coverage (25.283), analyst forecast accuracy (0), analyst forecast error (154) and information asymmetry (75). In this case, the only case where publication bias could have an effect, is the meta-analysis with CSR disclosure and analyst forecast accuracy.
Summarizing the results from the meta-analyses performed, there is a positive significant relationship between CSR disclosure and analyst coverage (, = 0.227, > < 0.001) , a positive non-significant relationship between CSR disclosure and analyst forecast accuracy (, = 0.060, > > 0.05), while a negative significant relationship is found for CSR disclosure and analyst forecast error (, = −0.045, > < 0.05), and a negative non-significant relationship between CSR disclosure and information asymmetry (, = −0.071, > > 0.05). As to heterogeneity, all meta-analyses, with the exception of analyst forecast accuracy, showed high degrees of heterogeneity. High
heterogeneity is caused because of the between study variance. These findings will be further addressed in the discussion section.
23
DISCUSSION AND CONCLUSION
Theoretical implications
The current literature has shown that there is contrasting evidence on the relationship between CSR disclosure and the response of financial analysts. While some studies show that CSR disclosure increases analyst forecast accuracy and analyst coverage, and decrease information asymmetry (Dhaliwal et al. 2012; Michaels and Grüning, 2017), other studies found results that CSR disclosure is not of value of analysts (Campbell and Slack, 2011; Krasodomska and Cho, 2017). The aim of this meta-analysis was to provide more insight into this relationship and to find a more conclusive answer to the research question: how does Corporate Social Responsibility Disclosure affect the response of financial analysts? To find an answer to this research question, the relationship has been looked at from different theoretical perspectives, in order to explain the relationship and to develop hypotheses. The findings of the meta-analysis contribute to the current literature, clarifying the contrasting evidence that previously has been found. Consequently, the findings have several theoretical implications, and also create opportunities for future research.
First, based on the legitimacy theory, results of the meta-analysis found support for the first hypothesis. The findings (see table 2) show that when firms disclose CSR information, an effect can be seen in the response of financial analysts. This can be derived from the analyses on CSR disclosure and analyst coverage, and analyst forecast error, which both show significant results that are in line with the expectations based on previous studies. Based on the legitimacy theory, firms disclose information in order to appear legitimate (O’Donovan, 2002). These disclosures have to be consistent with the actual performance of the firm, otherwise it will harm the legitimacy and reduce the credibility of the information for financial analysts (Cormier & Magnan, 2015). In fact, O’Dwyer (2001) found that corporate social disclosure does not always achieve legitimacy. However, the results of the meta-analysis suggest that the information is perceived as legitimate by financial analysts, as is shown by the increased analyst coverage and analyst forecast accuracy. This suggests that firms do disclose the information in order to be legitimate, and not just to appear to be legitimate. In fact, this is also seen in the study of Christensen (2016), who found that firms that disclose information about CSR are less likely to engage in misconduct. This shows that CSR disclosure improves the ability of firms to manage their operations (Christensen, 2016), but it also supports the findings of this meta-analysis that CSR disclosure can be used to be legitimate.
As for the relationship between CSR disclosure and analyst coverage, the meta-analysis produced significant results, showing a positive relation. Therefore, hypothesis 3, based on the signaling theory, can be accepted. According to the signaling theory, firms want to be distinguished as a firm of high quality. For this reason, managers want to signal their quality through disclosures (Watson et al. 2002). The results of the meta-analysis show that a positive relation is found for CSR disclosure and analyst coverage, suggesting that CSR disclosure is a signal of high
24
quality and which shows the quality of the firm. This is in line with Gao et al. (2016), who suggest that greater quality disclosures lead to a higher analyst coverage. As the results show that a high-quality signal leads to a higher analyst coverage, it suggests that the signal is of value for analysts. Moreover, it suggests that analysts will follow a firm with CSR disclosures that present good news, as firms want to show their positive qualities to outsiders (Connelly et al., 2011). In addition, Zhou et al., (2017) mention that an integrated report signals that sustainability is integrated in the company’s business. This supports the view that firms want to signal their quality. Similar results are found in the study of Dhaliwal et al. (2011), who found that firms with superior CSR performance and who disclose CSR information achieve a higher analyst coverage. Thus, from a signaling perspective, these results suggest that firms that provide CSR disclosures have a higher CSR performance, providing disclosures that are more informative. This can explain the increase in analyst coverage, since firms with more informative disclosure policies attract more analysts (Lang & Lundholm, 1996).
Based on the stakeholder theory, hypothesis 4 stated that there is a positive relationship between CSR disclosure and analyst forecast accuracy. Results from the meta-analysis show a positive, although not significant, relationship. This non-significant relationship may be caused by the small sample (K = 5) included in the analysis (Borenstein et al. 2009). However, other studies use analyst forecast error as an inverse measure of analyst forecast accuracy. The meta-analysis performed on CSR disclosure and analyst forecast error did show a significant negative relationship. These results are in line with the arguments based on the stakeholder theory. Stakeholders put pressure on managers to disclose CSR information (Lee et al., 2016), which leads to more transparency, causing analysts to create better forecasts and increase the forecast accuracy (Dhaliwal et al., 2011; Dhaliwal et al. 2012). The results from the meta-analysis and the arguments from the stakeholder theory suggest that the pressure from stakeholders appears to make firms disclose more CSR information in order to avoid conflicts with stakeholders, thereby also creating a more transparent information environment. As the results show, higher CSR disclosures reduce the analyst forecast error, thus increasing the accuracy of analysts’ forecasts.
The last meta-analysis performed between CSR disclosure and information asymmetry showed a negative relationship between CSR disclosure and information asymmetry. However, these results are not significant. Therefore, the second hypothesis cannot be accepted. On the other hand, the negative relationship found in the meta-analysis does suggest that CSR disclosure can decrease information asymmetry. This is in line with the arguments based on the agency theory. Martinez-Ferrero et al. (2017) mention that information asymmetry can be reduced by being transparent, to which organizations respond by disclosing CSR information (Cormier et al., 2011). Because the results of the meta-analysis are not significant, this could indicate that current studies on the relationship between CSR disclosure and information asymmetry still present contrasting evidence. A possible reason for this could be the quality of the information. Hummel and Schlick (2016) studied the effect of both high-quality CSR disclosure and
25
low-quality CSR disclosure on Tobin’s Q1. While high-quality CSR-disclosure lead to a decrease in information
asymmetry, low-quality CSR-disclosure lead to an increase in information asymmetry. Similar results were found by Cuadrado-Ballesteros et al. (2016), who found that a higher social reporting quality leads to a decrease in information asymmetry. These results, together with the results from the meta-analysis, suggest that the quality of CSR disclosure can be an important part of the usefulness of the information, where low-quality disclosures may possibly create more information asymmetry, thereby increasing agency problems.
Conclusion and future research
Overall, the aim of this meta-analysis was to perform a systematic review of the contrasting evidence on the relationship between CSR disclosure and the response of financial analysts. The disclosure of CSR information has been described from several theoretical perspectives, in order to develop hypotheses to answer the research question of how CSR disclosure affects the response of financial analysts. Following the HOMA-procedure to perform the meta-analysis, results show that CSR disclosure generally has a positive effect on the response of financial analysts, resulting in a reduced analyst forecast error and a higher analyst coverage. Other results show a non-significant negative relationship between CSR disclosure and information asymmetry.
These results provide clarity on the contrasting evidence in the current literature. While some previous studies have found positive effects of CSR disclosure on the response of financial analyst, other studies found that CSR disclosure is not of value to financial analysts (Campbell and Slack, 2011; Deegan and Rankin, 1997; Krasodomska and Cho, 2017). The results of this study show that more analysts will follow a firm if it discloses CSR information, thereby also reducing the error of analysts’ forecasts. Therefore, the results suggest that CSR disclosure is of value to financial analysts, leading to more and better forecasts. Furthermore, the results also suggest that CSR disclosure is a signal of high quality. Other findings of the study show that CSR disclosures decreases information asymmetry, although this is a non-significant relationship.
While the results of this study contribute to the current literature, clarifying the fragmentation of current studies, it is not without limitations. Despite an extensive literature search process in order to collect a sufficient amount of studies, it is possible that there are studies that were not found. A larger amount of studies would increase the reliability of the meta-analysis. In addition, while tests were performed to check for publication bias, there is still a chance that this can play a role. However, evidence was found that publication bias does not threaten the validity of conclusions of meta-analyses (Dalton et al. 2012).
26
The outcome of the meta-analysis provides opportunities for future research. First, the results of the meta-analyses found high heterogeneity. As mentioned, the analysis on CSR disclosure and information asymmetry showed a non-significant negative relationship. Future research can study this, to provide more clarity on why there is still contrasting evidence on this relationship. For instance, there could be moderators that affect this relationship, like legal environments (Cuadrado-Ballesteros et al. 2016).
Another opportunity for future research is to study how analysts use the information for their forecasts. Since the results have shown a generally positive effect of CSR disclosure on the response of analysts, it would be of interest to study how this information is used, and what is of most importance to financial analysts. This is also recommended by Dhaliwal et al. (2012).
As the findings of the meta-analysis between CSR disclosure and analyst coverage show a significant positive relationship, based on the signaling theory, it suggests that the disclosures are showing positive information. Future research can study what the effect is of disclosing positive CSR information and negative CSR information, and if analysts will respond differently to this information.
Finally, the studies used in the meta-analysis use different methods of measuring CSR disclosure, leading to high heterogeneity in the results. This suggests that there is variance in the studies, which is not caused by sampling error. Future research could perform methods like a subgroup-analysis or meta-regression to find out what the reason behind this variance is (Borenstein et al., 2009). This could provide further evidence which affects the relationship between CSR disclosure and the response of financial analysts.
