The influence of stakeholder salience on CSR assurance

(1)

Amsterdam Business School

The influence of stakeholder salience on CSR assurance

Name: Bas Booi

Student number: 10694226

Thesis supervisor: Dr. J. J. F. van Raak Date: 20-06-2016

Word count: 12.793

MSc Accountancy & Control, specialization Accountancy Amsterdam Business School

(2)

Statement of Originality

This document is written by student Bas Booi, who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

Abstract

This paper investigates how stakeholder salience influences a firm to engage in corporate social responsibility (CSR) assurance. Prior research suggests that stakeholders with power, legitimacy, and urgency can influence a firm’s CSR agenda, like CSR reporting. Other research indicates that the number of stand-alone CSR reports has grown significantly over the recent years. Also, the call for quality standards and independent assurance on these reports has increased as well. Nonetheless, little research has been done to address the relationship between stakeholder salience and CSR assurance. To test this relationship, a logistic regression has been used to estimate the probability that a firm’s CSR report is externally assured, due to the influence of stakeholder salience. Results for the first hypothesis indicate that there is a positive association between the salience of a shareholder and a firm’s decision to externally assure its CSR report. Additionally, the results for the second hypothesis show that this relation between stakeholder salience and CSR assurance does not count for the stakeholder group directors. Only the shareholder group shows a positive association between its salience and a firm’s CSR assurance. Results for the third hypothesis indicate that a firm’s decision to externally assure its CSR report is not solely based on the power of a stakeholder, but also on the factor urgency.

(4)

1 Introduction

In this thesis I examine how stakeholder salience influences a firm’s decision to engage in corporate social responsibility (CSR) assurance.

In order to establish rules for sustainable development, international organizations, like the World Business Council for Sustainable Development (WBCSD) and the Global Reporting Initiative (GRI) came up with the idea of CSR. In recent years, the number of stand-alone CSR reports has grown significantly (KPMG, 2013). At the same time, the call for quality standards and independent assurance on these reports has increased as well (Simnett et al., 2009).

Holder-Webb et al. (2009) assert that it is not enough for corporations to simply engage in CSR activities. They add that it is also important - and desirable - to make information on these activities available to its stakeholders. The GRI guidelines are currently the world’s most used reporting framework. It has developed detailed CSR reporting guidelines that form the foundation of its Sustainability Reporting Framework. With G4 - released early in 2013 - as their most recent iteration, the guidelines are constantly being refined, (GRI 2013b). While GRI’s mission statement does not specifically address a need for verification (independent assurance), GRI does encourage the development and adoption of principles for verification, as it enhances the quality, usefulness, and credibility of the disclosed CSR information.

GRI recognizes that the verification of CSR reports is at an early stage in its evolution (Cohen & Simnett, 2015; GRI 2013a). Additionally, the call for disclosure of non-financial information has grown in response to the awareness that financial statement omits salient information about a firm (Adams, 2002). However, a firm’s management does not deem all stakeholder groups equally important when conducting their CSR activities. Mitchell et al. (1997) propose a theory of stakeholder identification and salience in response to the many competing definitions of ‘stakeholder’ and the lack of agreement 'who and what really counts' in stakeholder management. The pragmatic reality is that managers simply cannot attend to all actual or potential claims and need to prioritize their managerial attention over these variety of claims. In theory, only the stakeholders who possess the stakeholder salience attributes power, legitimacy and urgency, like shareholders, can influence the responsible stakeholder management team to engage in CSR.

Prior research of Angle et al. (1999) already shows a relationship among the stakeholder attributes, salience, CEO values, CSR, and corporate performance. There is strong support for the attribute-salience relationship and some significant relationships among CEO values, salience, and CSR.

(6)

2

Based on the growing number of CSR reports and the growing demand of independent assurance on these reports, the research question is formulated as follows: How does stakeholder salience influence the engagement of an external assurance provider for corporate social responsibility (CSR) reporting? The purpose of this paper is to answer the research question by determining whether an effect exists between stakeholder salience and CSR assurance, and how this can be described. It addresses how stakeholders motivate management to provide assurance on their CSR activities.

I choose to investigate the relationship between stakeholder salience and CSR assurance for the reason that little research has been done on how the salience of a firm’s stakeholder could affect its policy on CSR information disclosure and verification, even though its influence on other CSR activities, e.g. environmental reporting (Henriques and Sadorsky, 1999), has been broadly explored. Yet, it is also important to understand the relationship between stakeholder salience and CSR assurance, especially with the increased demand from stakeholders for higher quality CSR reports. However, in order to get higher quality CSR information, it is crucial to understand how a stakeholder can become a priority for management to answer his claim.

This thesis makes use of a logistic regression, in which the focus lies on firms listed on the S&P 500 Index. In the first two hypotheses, stakeholder salience is investigated as one variable. In the third hypothesis, the stakeholder salience attributes power, legitimacy, and urgency are analyzed separately. In the first hypothesis investigates if there is a positive association between stakeholder salience and a firm’s decision to externally assure its CSR report by an independent assurance provider. The second hypothesis tests if only the salience of shareholders influences a firm’s decision to externally assure its CSR report. The third hypothesis tests if only the stakeholder salience attribute power has a significant influence on a firm’s decision whether or not its CSR report is externally assured.

Results of the first regression indicate that there is a positive association between the salience of a shareholder and a firm’s decision to externally assure its CSR report. The results of the first regression therefore appear to support the first hypothesis. Additionally, the second regression shows that this relation between stakeholder salience and CSR assurance does not count for the stakeholder group directors. Only the shareholder group shows a positive association between its salience and a firm’s CSR assurance. This is consistent with research of Wood and Jones (1995). The second hypothesis is therefore supported.

The regression results for the third hypothesis indicate that a firm’s decision to externally assure its CSR report is not solely based on the power of a stakeholder. The results of the regressions show that the factor urgency also has a significant influence on the probability that a firm’s CSR report is externally assured.

(7)

3

This study contributes to literature by showing how stakeholders can influence managerial decision-making on CSR disclosure. Based on the stakeholder salience theory of Mitchell et al. (1997), I empirically investigated whether differences in CSR disclosure among companies are systematically related to differences in the level of power, urgency and legitimacy of each stakeholder group with which these companies are confronted. Secondly, this study contributes to the literature by providing a new dimension in understanding the relationship between stakeholder salience and CSR, that of CSR assurance. Prior research of Mitchell et al. (1997) argues that the degree to which managers give priority to competing stakeholder claims may be assessed by a stakeholder’s possession of one (or more) of the three stakeholder salience attributes (power, legitimacy, and urgency). However, the relationship between these attributes, the salience it produces, and the influence it has on a firm’s CSR assurance policy, have not been fully explored yet. From a managerial point of view, to promote the CSR (assurance) of firms that are characterized as defensive and reactive, it is important, when pursuing a CSR agenda, to understand this relationship. Thirdly, the findings of this study contribute to literature, as they highlight that not only the power attribute provides an impact on the salience of the claim to externally assure a CSR report, as asserted by Neill and Stovall (2005). This study shows that also the urgency of the claim has an impact on its salience. Lastly, this study shows that the salience of some stakeholder groups (read: directors) have no impact on the CSR assurance claim.

The paper is organized as follows: Section 2 provides a literature review and develops the hypothesis. Information regarding the sample selection, the methodological approach, the research design, and the handling of validity concerns are presented in section 3. Section 4 illustrates the results and section 5 summarizes and concludes the paper.

(8)

4

2 Theory

This chapter first briefly outlines background information on corporate social responsibility (CSR) reporting, CSR assurance, and stakeholder salience. The second paragraph describes the hypothesis development.

2.1 Literature review

This paragraph reviews empirical literature concerning CSR reporting, CSR assurance, and stakeholders salience.

2.1.1 Corporate social responsibility reporting

In order to establish rules for sustainable development, the idea of corporate social responsibility (CSR) was introduced by the World Business Council for Sustainable Development (WBCSD) and the Global Reporting Initiative (GRI). Holder-Webb et al. (2009) assert that it is not enough for corporations to simply engage in CSR activities, but it is also important and desirable to make information about these activities available to stakeholders. Additionally, the call for disclosure of non-financial information has grown in response to the awareness that financial statements miss significant information about a firm (Adams, 2002).

The WBCSD defined CSR as “the commitment of business to contribute to sustainable economic development, working with employees, their families, the local community, and society at large to improve their quality of life”. The CSR reporting and publication rules are based on the standards of the GRI (Hawrysz & Hys, 2012) and can be explained from four different perspectives: legitimacy theory, stakeholder theory, social contract theory, and signaling theory.

Deegan and Unerman (2011) assert that the legitimacy theory relies upon the notion that there is a ‘social contract’ between a firm and society in which it operates. To get the approval from society, firms try to legitimize their actions by engaging in CSR activities and reporting. Getting approval from society ensures their continuing existence. O'Donovan (2002) also argues that for firms to continue operating successfully, it must act within the bounds and norms of what society identifies as socially responsible behavior.

Stakeholder theory embraces the idea that firms must play an active role in the society in which they operate. Wicks et al. (2004) argues that firms should consider the effect of their actions upon their stakeholders. Wearing (2005) states that stakeholder theory stresses the importance of all affected parties by a firm’s actions, either directly or indirectly. It can also be explained by using managerial and ethical branches (Deegan, 2013). The managerial branch posits that firms will respond to those stakeholders who have the necessary economic impact upon the

(9)

5

firms (O'Dwyer, 2003) or to those who are not directly engaged in the firm’s economic activities, but have an interest in the actions of a firm and can influence it (Savage et al., 1991). The ethical branch argues that all stakeholders have a right to know at any time what social and environmental implications arises from the activities of a firm (Deegan, 2013).

The social contract theory views the firm-society relationship from a philosophical thought (Donaldson, 1982). There exists an implicit social contract between business and society and this contract implies some indirect obligations of business towards society. Social contract thinking is explicitly recognized as a form of post conventional moral reasoning (Rest et al., 1999). This theory is further extended by Donaldson and Dunfee (1999). They extend the theory in a way for managers to take decision in an ethical manner. According to the societal approach, firms are responsible to society as a whole. The main idea behind this view is that firms operate by public consent in order to serve constructively the needs of society (Van Marrewijk, 2003). The societal approach is a strategic response to the changing circumstances and the new corporate challenges that previously did not occur, like CSR reporting.

Signaling theory explains why firms have an incentive to voluntarily report information to the capital market. Voluntary disclosure lowers the risk premium for a firm on capital, because the voluntary disclosure of private information (i.e. CSR) ‘signals’ the credibility of a firm and reduces outsider uncertainty (Connelly, Certo, Ireland, & Reutzel, 2011; Mahoney, 2012). Paying a lower risk premium on capital eventually leads to an increased firm value (Omran and El-Galfy, 2014; Thorne, Mahoney, & Manetti, 2014).

Based on the findings of Omran and Ramdhony (2015) there is no universal theory applicable on CSR disclosure for all situations and societies. While legitimacy theory suggests CSR disclosures are part of a process of legitimation, stakeholder theory offers an explanation of accountability to stakeholders. Legitimacy theory seems to be more suitable for firms working in developed countries, while stakeholder theory appears to be most suitable for firm working in developing countries, where a firm can manage its stakeholders and the pressure to comply with existing legislation is less as compared to the developed countries. Social contract theory is appropriate for both developed and emerged economies, as CSR disclosures exist due to an implicit social contract between business and society. This implies some indirect obligations of business towards society. Signaling theory will suit a situation where firms are competing for resources (i.e. Shell and BP). A firm willing to demarcate from other firms will engage in more CSR practices. It is also important that the signal reaches the target audience by reporting on CSR.

(10)

6

2.1.2 Corporate social responsibility assurance

The number of stand-alone CSR reports that has grown significantly over recent years (KPMG, 2013). KPMG’s (2013) survey concludes, ‘‘The high rates of corporate responsibility reporting in all regions suggest it is now standard business practice worldwide’’ (KPMG, 2013). In other words, most firms now engage at least in some form of CSR reporting. In order to improve the relevance and reliability of CSR information, there has been a growing reliance on independent ‘CSR assurance’ (Simnett et al., 2009). The accounting firms account for the majority of the market share for CSR assurance engagements for large international companies, but also other assurance providers maintain a significant market proportion (Cohen & Simnett, 2015; KPMG, 2013).

The most used reporting framework currently is the one of the GRI. The GRI G4 Guidelines organize specific and standard disclosures into three Categories: economic, environmental, and social disclosure. The social category is further divided into four subcategories, which are labor practices and decent work, human rights, society, and product responsibility (GRI 2013b). All of these factors and issues can constitute a comprehensive CSR report and can be reported on independently - such as a greenhouse gas emissions report - as is commonly mandated by regulators. For a comprehensive report, part of the challenge for the assurance provider is to ensure that all significant issues that fall under these broad categories are appropriately reported and that they are reported in accordance with the reporting criteria (Cohen & Simnett, 2015; O’Dwyer 2011).

Assurance can be provided on the comprehensive report, or a specific subject matter area. With regard to assurance standards, the main assurance standard referred to in the assurance reports is the International Auditing and Assurance Standards Board’s ISAE 3000 (2013) - Assurance Engagements Other than Audits and Reviews of Historical Financial Information - or its national equivalent (KPMG, 2011). This standard is used in nearly all instances where the assurer is from the accounting profession. It is also referred to by some assurers from outside the profession. AA1000AS, an assurance standard issued by the non-profit firm AccountAbility, is the standard to most frequently used by independent assurance providers from outside the accounting profession (O’Dwyer & Owen 2005; Simnett 2012). The most common publicly available single issue report is a firm’s report of their greenhouse gas emissions, with the main assurance standard being the IAASB’s ISAE 3410 (2012) - Assurance on a Greenhouse Gas Statement - or its national equivalent (Cohen & Simnett, 2015).

(11)

7

2.1.3 Stakeholders salience

Within the last decades, managers have shifted their philosophy for understanding their business environment from a shareholder approach to a stakeholder approach. This new vision broadens the management’s perception of its roles and responsibilities beyond profit maximization for shareholders. Moreover, the stakeholder approach also includes the interests and claims of non-stockholding groups (Mitchell et al., 1997). Within the stakeholder literature there have been numerous descriptions to define a stakeholder. Based on a chronology of stakeholder definitions made by Mitchell et al. (1997, p. 858), the researchers state that Freeman’s (1984, p. 46) definition of a stakeholder is the most common and used interpretation in explaining a stakeholder. Freeman’s interprets a stakeholder as follows: ‘any group or individual who can affect or is affected or is affected by the achievement of the firm’s objectives’. In other words, an entity must have a legitimate claim or stake in the firm to be considered a stakeholder.

When identifying (defining) a stakeholder, three fundamental concepts are important: ‘power’, ‘legitimacy’, and ‘urgency’. The first concept portrays power of a stakeholder to influence a firm (Etzioni, 1964). The second one outlines the legitimacy of a stakeholder’s claim on a firm (Suchman, 1995). A stakeholder claim could be due to a risk or a legal, contractual, moral, or financial claim. Urgency expresses the degree to which stakeholder claims call for immediate attention of management. The degree of urgency does not only depend on time sensitivity, but also on how critical the relationship (or the importance of the claim) is with the stakeholder (Mitchell et al., 1997).

When a manager has identified its stakeholders, he needs to prioritize the competing stakeholder claims according to their power, legitimacy, and urgency. This is due to the fact that a manager cannot attend to all actual or potential claims. In the stakeholder literature this degree of stakeholder importance is called ‘stakeholder salience’. The theory behind stakeholder salience accepts the reality that managers can’t (and don’t) consider all possible stakeholders to be evenly important. The higher a stakeholder scores on all three attributes, the higher his salience will be. High salience stakeholders (those with the most power, legitimacy, and urgency) will have the most influence on a firm and a manager will give these stakeholders the greatest priority (Mitchell et al., 1997).

The theory of stakeholder salience describes three ranges of salience: latent stakeholders (low salience), expectant (moderate salience), and definitive stakeholders (high salience). The low and moderate salience ranges each have three stakeholder subclasses. The high salience range consists of only one subclass (Mitchell et al., 1997).

(12)

8

The first subclass within the low range consists of dormant stakeholders. They score high on power, but lack legitimacy or urgency, and therefore have little to no interaction with the firm. The second class consists of the discretionary stakeholder. They score high on legitimacy, but lack the power and urgency to influence a firm. The last low range subclass stakeholders are the demanding stakeholders. Their claims score high on urgency, but have no power or legitimacy. For example people a unjustifiable grudge to a company.

The first subclass of expectant stakeholders consists of the dominant stakeholders. These stakeholders are powerful and legitimate, but don’t score high on urgency. For example significant creditors. They are important to the firm, but don’t have urgent claims. The second class are the dependent stakeholders. This group depends on other groups to carry out their will, because they lack the power to enforce their legitimate and urgent claim. The last group of expectant stakeholders are the dangerous stakeholders. These stakeholders have an urgent claim in the direction of a firm and have the power, but not the legitimacy, to enforce it. For example, unlawful tactics used by activists against - for example – Shell (Mitchell et al., 1997).

Finally, the stakeholders within the high salience range are called the definitive stakeholders. They poses all three attributes.

A key presumption of the stakeholder salience theory is that it is dynamic. The attributes of a stakeholder can change over time and therefore, a stakeholder can change from class. For example dependent stakeholders can become definitive if their cause is picked up by a dominant stakeholder (Mitchell, et al., 1997).

2.2 Hypothesis development

In the previous paragraph I discussed the empirical literature of the three main themes – CSR, assurance, and stakeholder salience – that arise from my research question. Based on the growing number of CSR reports and the growing demand of stakeholders on independent CSR assurance, I want to examine the relationship between initiatives of stakeholders and the ability and motivation of managers to provide assurance on their CSR activities. As discussed in the previous paragraph, empirical evidence shows that stakeholders can influence the CSR activities of a firm. Therefore, supportive stakeholder behavior can influence CSR activities like engaging in CSR assurance by an independent assurance provider.

(13)

9

While studying the relation between stakeholders and CSR, Wood and Jones (1995) argue that the majority of studies imply that while a variety of stakeholder groups deem CSR to be important – because of its impact on them – management mostly (or only) consider shareholders as an appropriate stakeholder group. The implication of this assessment is that even if the relationship between CSR and shareholder evaluations may represent a judgment that managers are dealing effectively with stakeholders, such an overall relationship cannot be asserted until the strands that make it up – the links among stakeholder expectations, effects, evaluations, and behaviors with respect to a firm – have been delineated (Peloza & Papania, 2008). Another implication of Wood and Jones (2005) their statement is that managers are not consistent in identifying and assessing their stakeholder groups across all firms. For example, external stakeholders such as governments and regulatory bodies are highly salient for firms operating in highly regulated environments (Miles, 1987), while customers may be a more important stakeholder group for firms operating in consumer goods category. Even within a same industry there can be highly differentiated stakeholder priorities across firms.

Henriques and Sadorsky (1999) demonstrate that the environmental position of the firm (defensive, reactive, accommodative or proactive) influences the priorities that the firm assigns to different stakeholder groups. Firms that are characterized as being reactive in their approach to environmental practices are, for example, more concerned about their relationship with the media, instead of their relationship with their employees or customers. Regardless of industry or firm characteristics, the stakeholder characteristics of power, legitimacy, and urgency - outlined by Mitchell et al. (1997) - do appear to determine how managers deal with (both external and internal) stakeholders and their demands (Agle et al., 1999).

𝐇_𝟐

The salience of shareholders influences a firm’s decision to externally assure its CSR report by an independent assurance provider.

The identification and subsequent support among stakeholders can have significant positive benefits on the CSR assurance. However, this relationship is contingent upon the stakeholder’s ability to possess the stakeholder salience attributes power, urgency, and legitimacy. For example, a consumer who does not purchase products tied by the firm’s business environment is not in a position to reward the firm with increased legitimacy; nor is a NGO with limited political influence likely to be able to encourage consumers to support a firm (Peloza & Papania, 2008).

Neill and Stovall (2005) found that the attributes of stakeholder salience can shift over a firm’s life cycle. Therefore, the levels of CSR activities can change as well. As long as power is the

(14)

10

primary attribute driving CSR within these firms. In order for a firm to engage in purposeful CSR activities, the stakeholder pursuing a CSR agenda must possess the power to impose such an agenda on management. Neill and Stovall (2005) conclude that without power, legitimacy and urgency may not provide sufficient stakeholder salience to promote CSR activities. This is not surprising since stakeholders without power, combined with legitimacy and urgency, are less likely to affect a firm’s performance than those that do possess these characteristics (Peloza & Papania, 2008).

Where stakeholders do exhibit power are in a position to punish the firm if they feel that their interests have been violated. McGuire et al. (1988) found that firms engaging in socially irresponsible behavior risk having their implicit contracts transformed into explicit contracts, which are often more costly. For example, if a firm fails to meet promises to government officials in regard to actions that affect the environment (like dumping), government agencies may find it necessary to pass more stringent regulations, constituting explicit contracts, to force the firm to act in a socially responsible manner.

𝐇_𝟑

The stakeholder salience attribute power has a significant influence on a firm’s decision whether or not its CSR report is externally assured.

The next chapter discusses the used research method to answer the research question. The first paragraph describes the selected sample and sample period. The next paragraph outlines its research design.

(15)

11

3 Methodology

This chapter outlines the used research method to answer the research question. The first paragraph describes the selected sample and sample period. The next paragraph describes its research design.

3.1 Sample selection

To investigate the effect of stakeholder salience on the probability that a firm’s CSR report is externally assured, the analysis focuses on the U.S. firms listed on Standard & Poor’s 500 Index (short: S&P 500). The number of stand-alone CSR reports with these firms has grown significantly over recent years. Based on a survey by KPMG (2013), 86 of the largest 100 firms within the U.S. now engage in some form of CSR reporting.

The sample period for the analysis extends from 2010 through 2014. Data on whether a firm’s CSR report is externally assured or not is collected from Datastream’s ESG Asset 4 database. The code used for CSR assurance is CGVSDP030. The Datastream code LS&PCOMP provides data on the S&P 500 firms.

The original dataset collected from Datastream consisted of 2.525 firm-year observations (505 unique firms) for the sample period from 2010 through 2014. The years 2015 and 2016 are excluded, as most of these firms have not provided sufficient data yet whether they engage in CSR assurance or not. Also. from the 2.525 firm-year observations, 1.687 observations (226 unique firms) did not provide any data whether or not they engage in CSR assurance in the

1_{Data output is either ERROR, INVALID, N/A (Not available), numerical data (instead of Y/N), or there is no}

output data.

TABLE 1 Sample selection

Description of the data Sample size

firms observations Datastream’s dataset of firms listed on the S&P 500 Index in the

fiscal years 2010 to 2014 505 2.525

Missing/insufficient data on firm’s CSR assurance (226) (1.687) Datastream’s dataset of S&P 500 firms that participate in CSR

activities in the fiscal years 2010 to 2014 279 838

Missing/insufficient data on variables1 ₍₅₃₎ ₍₂₁₀₎

(16)

12

sample period from 2010 through 2014. Therefore, these 1.687 observations are deleted from the sample. Additionally, of the remaining 838 observations (279 unique firms), 210 observations (53 unique firms) didn’t provide sufficient data on the variables used in the analysis. After deleting these firm-year observations, the final sample consists 628 firm-year observations (226 unique firms), as shown in table 1.

3.2 Research design: logistic regression

Considering that the outcome variable - CSR assurance – has a binary output, a logistic regression is the best fitting model to measure the relationship between stakeholder salience and CSR assurance. It generates the coefficients, standard deviations, and significant levels to predict the probability of the presence of CSR assurance within a firm (Hosmer et al., 2013).

Determining (accounting) proxies based on stakeholder salience can be quite difficult. However, Andrikopoulos and Webber (2014) provide a wide range of accounting proxies to measure the stakeholder salience attributes and their dynamic behavior. These proxies provided by Andrikopoulos and Webber (2014) are also the base of this study.

Andrikopoulos and Webber (2014) outline eleven stakeholder groups in their paper, all with their own proxies to measure the stakeholder factors power, legitimacy, and urgency. The two stakeholder groups used in this study are shareholders and directors2_{. The salience of both}

groups are analyzed separately (due to the fact that their interests within a firm could differ). The data to measure stakeholder salience is collected from Datastream’s ESG Asset 4 database. Measuring shareholders’ power is done on the basis of a shareholders’ right to vote on an executive compensation pay (Datastream code CGSRDP038). With a dummy variable, the output data is either 0 (= no) or 1 (= yes). Directors’ power is based on its limitation to be removed by a shareholder (CGSRDP071). The probability that a firm’s CSR report is externally assured is negatively related to the limitation to remove a board member. When it is easier to remove a board member, he is more likely to voluntary engage in CSR assurance. With a dummy, the output data is either 0 (= yes) or 1 (= no).

Indices on shareholders’ concentration provide a basis to measure shareholders’ legitimacy. In this case, the percentage ownership of the single biggest owner (by voting power) of a company (CGSRDP045) is being used for measuring the shareholders’ legitimacy. It is

2_{Initially, the analysis consisted of four stakeholder groups: (1) shareholders, (2) government, (3) society, and (4)}

environment. However, both the Datastream database and the Wharton Research Data Services (WRDS) database didn’t provide sufficient data (firm-year observations) to analyze the relationship between CSR assurance and the stakeholder groups government, society, and environment. Eventually, besides the shareholder group, the stakeholder group directors provides sufficient data (and the same amount of firm-year observations) to test my hypotheses.

(17)

13

important to note that the shared voting right and dispositive power are also taken into account to determine the percentage ownership of the single biggest owner. The data output of the proxy is a percentage between 0-100. The higher the shareholder’s concentration, the higher the probability that a firm’s CSR report is externally assured (due to the fact that a more powerful shareholder is in a better position to impose his CSR agenda). Director’s legitimacy is based on a directors’ terms of office, like whether or not a CEO’s compensation is linked to the total shareholder’s return (CGCPDP041). When his compensation is linked to the shareholder’s return, the probability that the firm’s CSR report is externally assured is lower. This is due to the fact that a director will put greater priority on the firm’s profit (and shareholder’s return), than on the firm’s CSR activities. When the CEO compensation is not linked to the shareholder’s return, the director is more likely to voluntary give up some profit to use for CSR activities (and assurance). With a dummy, the output data is either 0 (= yes) or 1 (= no).

Shareholders’ urgency can be tested with evidence of urgent communication between a firm and its shareholders. Datastream’s ESG Asset 4 database provides data if a company has been in the media due to controversies linked to shareholders rights (CGSRO07V). Firms that are in the spotlight due to shareholder rights controversies show their failure in corporate social responsibility. The variable is measured with a dummy, with a 0 (= N) when a firm didn’t had any shareholder rights issues and with 1 (= Y) when a firm has been in the media due to shareholder rights controversies (which leads to a higher urgency for shareholders to claim more assurance). Analyzing directors’ urgency is according to the number of board meetings during the year (CGBFDP024). A higher frequency of board meetings is positively related to a firm’s CSR assurance.

The remaining independent variables in the equation are control variables. Prior research shows that the factors firm size, leverage, return on assets (ROA) could affect a firm’s CSR assurance (e.g., Simnett et al., 2009; Peters & Romi, 2005; Dhaliwal et al., 2011). The three3

variables are therefore included in the regression as control variables. Datastream’s databytes and Worldscope are used to attain the data.

Firm size is measured by a firm’s total assets in U.S. dollars (DWTA). It represents the sum of total current assets, long term receivables, investment in unconsolidated subsidiaries, other investments, net property plant and equipment and other assets. Prior research shows that the size of a firm is positively associated with CSR assurance, due to the fact that a large-sized

3_{Another control variable that was considered was firm age. However, the Datastream databases did not provide}

information on the sample firm’s age. It provided only scant data to calculate firm age. Therefore, I decided not to include this variable in my regression. Even though prior research shows that it is a factor that could affect a firm’s choice of CSR assurance (e.g., Simnett et al., 2009; Peters & Romi, 2005; Dhaliwal et al., 2011)

(18)

14

firms’ scale of operations, resource usage, and public visibility have a greater impact on a larger group of stakeholders (Udayasankar, 2008).

Leverage (WC08236) is measured by the ratio of total debt to total assets. It shows how much capital comes in the form of debt, like loans, or assesses the ability of a firm to meet its financial liability. Mishra and Modi (2013) show that there is a relationship between CSR and a firm’s specific risk. Considering that more information disclosure leads to a lower risk premium means that more CSR assurance leads to a lower debt, due to the lower risk premium. Leverage is therefore negatively associated with CSR assurance.

The last control variable is the return on assets (ROA; WC08326). ROA is measured by the ratio of net income to total assets. Dehning and Richardson (2002) describe ROA as accounting measures on how well a firm uses all of its assets. It measures how profitable a company is relative to its total assets. ROA gives an idea as how efficient management is at using its assets to generate earnings. Research of Dewi et al. (2014) shows that ROA is positively associated with a firms’ his CSR performance. Thus, ROA is positively associated with CSR assurance.

Based on the above data, the following equations, where, for fiscal year 𝜏 and firm 𝑖, will be used to test H1 and H2. The first equation analyzes the shareholder group:

log(𝑝_𝑐𝑠𝑟_𝑎𝑠𝑠𝑢𝑟𝑎𝑛𝑐𝑒) = 𝐿𝑛 ( 𝑝_𝑐𝑠𝑟_𝑎𝑠𝑠𝑢𝑟𝑎𝑛𝑐𝑒

1−𝑝_𝑐𝑠𝑟_𝑎𝑠𝑠𝑢𝑟𝑎𝑛𝑐𝑒) = 𝛽0+ 𝛽1𝑆ℎ𝑎𝑟𝑒ℎ𝑜𝑙𝑑𝑒𝑟_𝑆𝑎𝑙𝑖𝑒𝑛𝑐𝑒𝑖𝜏+

𝛽2𝑆𝑖𝑧𝑒𝑖𝜏+ 𝛽3𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒𝑖𝜏+ 𝛽4𝑅𝑂𝐴𝑖𝜏

The second equation tests H₁ and H₂ with the directors group:

1−𝑝_𝑐𝑠𝑟_𝑎𝑠𝑠𝑢𝑟𝑎𝑛𝑐𝑒) = 𝛽0+ 𝛽1𝐷𝑖𝑟𝑒𝑐𝑡𝑜𝑟_𝑆𝑎𝑙𝑖𝑒𝑛𝑐𝑒𝑖𝜏+

𝛽₂𝑆𝑖𝑧𝑒_𝑖𝜏+ 𝛽₃𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒_𝑖𝜏+ 𝛽₄𝑅𝑂𝐴_𝑖𝜏

𝑝_𝑐𝑠𝑟_𝑎𝑠𝑠𝑢𝑟𝑎𝑛𝑐𝑒 is the probability that a firm’s CSR report is externally audited, The output is 1 when a CSR report is assured and 0 when a firm’s CSR report is not assured. 𝛽1𝑆ℎ𝑎𝑟𝑒ℎ𝑜𝑙𝑑𝑒𝑟_𝑆𝑎𝑙𝑖𝑒𝑛𝑐𝑒𝑖𝜏 is the computed coefficient of the three shareholder salience

variables for firm 𝑖 at year 𝜏. Its output is between 0 and 34_{. 𝛽}

1𝐷𝑖𝑟𝑒𝑐𝑡𝑜𝑟_𝑆𝑎𝑙𝑖𝑒𝑛𝑐𝑒𝑖𝜏 is the

4_{To calculate the coefficient 𝛽}

1𝑆ℎ𝑎𝑟𝑒ℎ𝑜𝑙𝑑𝑒𝑟_𝑆𝑎𝑙𝑖𝑒𝑛𝑐𝑒𝑖𝜏, a dummy with a binary output is created for the proxy

shareholder’s legitimacy. The output of the dummy is 1 when the value of the observation is higher than the median of the proxy and 0 when the observed value is lower.

(19)

15

computed coefficient of the three director salience variables for firm 𝑖 at year 𝜏. Its output is also between 0 and 35_{. 𝛽}

2𝑆𝑖𝑧𝑒𝑖𝜏 is the coefficient for firm size. It shows the total assets of firm 𝑖 (in

dollars) at year 𝜏. 𝛽₃𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒_𝑖𝜏 is the coefficient for leverage. It shows the ratio of total debt to total assets of firm 𝑖 at year 𝜏. 𝛽4𝑅𝑂𝐴𝑖𝜏 is the ratio of net income to total assets of firm 𝑖 at year

𝜏. When only 𝑆ℎ𝑎𝑟𝑒ℎ𝑜𝑙𝑑𝑒𝑟_𝑆𝑎𝑙𝑖𝑒𝑛𝑐𝑒 has a p-value smaller than 0.05, it will support Wood and Jones (1995) argument that a firm only deem shareholders as an appropriate stakeholder group.

To test H3, the following equation is used:

1−𝑝_𝑐𝑠𝑟_𝑎𝑠𝑠𝑢𝑟𝑎𝑛𝑐𝑒) = 𝛽0+ 𝛽1𝑃𝑜𝑤𝑒𝑟𝑖𝜏+ 𝛽2𝐿𝑒𝑔𝑖𝑡𝑖𝑚𝑎𝑐𝑦𝑖𝜏+

𝛽₃𝑈𝑟𝑔𝑒𝑛𝑐𝑦_𝑖𝜏+ 𝛽₄𝑆𝑖𝑧𝑒_𝑖𝜏+ 𝛽₅𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒_𝑖𝜏+ 𝛽₆𝑅𝑂𝐴_𝑖𝜏

𝛽₁𝑃𝑜𝑤𝑒𝑟_𝑖𝜏 is the coefficient of the power proxy from either the shareholder group or the director’s group of firm 𝑖 at year 𝜏. 𝛽2𝐿𝑒𝑔𝑖𝑡𝑖𝑚𝑎𝑐𝑦𝑖𝜏 is the coefficient of the described legitimacy

proxy from either the shareholder group or the director’s group of firm 𝑖 at year 𝜏. 𝛽₃𝑈𝑟𝑔𝑒𝑛𝑐𝑦_𝑖𝜏 is the coefficient of the described urgency proxy from either the shareholder group or the director’s group of firm 𝑖 at year 𝜏. The control variables are measured the same way as in the first two equations. When only the independent variable 𝑃𝑜𝑤𝑒𝑟 has a p-value smaller than 0.05, it will support Neill and Stovall’s (2005) conclusion that power is the only stakeholder factor that really significantly influences a firm’s CSR activities.

The adequacy of the model will also be tested by its overall goodness-of-fit and robustness (Hosmer & Lemeshow, 2000). The models’ robustness is tested with the Wald test. The models’ goodness-of-fit is tested on the basis of the Chi-square goodness-of-fit test and the Hosmer and Lemeshow test (Hosmer et al., 2013).

Although numerous pseudo-R2_{values have been developed, I will not use this value to}

interpret the overall goodness-of-fit of the models. There are different ways to compute the pseudo-R2_{, all with their own computational issues which cause them to be artificially high or}

low. Its interpretation is highly subjective.

The next chapter discusses the results of the regressions and whether 𝐻₁ and 𝐻₂ are supported or not.

5_{To calculate the coefficient 𝛽}

1𝐷𝑖𝑟𝑒𝑐𝑡𝑜𝑟_𝑆𝑎𝑙𝑖𝑒𝑛𝑐𝑒𝑖𝜏, a dummy with a binary output is created for the proxy

director’s urgency. The output of the dummy is 1 when the value of the observation is higher than the median of the proxy and 0 when the observed value is lower.

(20)

16

4 Results

This chapter outlines the empirical results of the analysis. The first paragraph give a description of the statistics and explains the main differences. Paragraph 2 displays and explains the multivariate results of the regressions and argues whether the hypotheses are true or not.

4.1 Descriptive statistics

Before performing the regression analysis, it is important to limit extreme observation values – outliers - in the statistical data. The distribution of the statistics of the (continuous) control variables can be profoundly influenced by outliers. By setting the observations in the smallest percentile equal to the smallest value of the second percentile, and the largest percentile equal to the largest value of the 99th_{percentile, the influence by outliers will be limited. This technique of}

reducing the influence of extreme values is called winsorizing. Winsorized estimators are usually more robust than their more standard forms.

Panel A in table 2 (Appendix A) and 3 (Appendix B) present the descriptive statistics for the sample. Table 2 outlines the descriptive statistics of the shareholder group; table 3 presents the descriptive statistics of the directors. It displays the median, mean, standard deviation, 25th

(Q1) and 75th_{(Q3) percentile of the independent variables, and the lowest and highest observed}

value in the dataset.

As described in paragraph 3.2, the values of the variables 𝑆ℎ𝑎𝑟𝑒ℎ𝑜𝑙𝑑𝑒𝑟_𝑠𝑎𝑙𝑖𝑒𝑛𝑐𝑒𝑖𝜏

(table 2) and 𝐷𝑖𝑟𝑒𝑐𝑡𝑜𝑟_𝑠𝑎𝑙𝑖𝑒𝑛𝑐𝑒𝑖𝜏 (table 3) are a computation of the three stakeholder salience

factors, with an output between zero and three. The median of both variables is one, which indicates that in the higher half of the observations, in both variables, at least one (of the three) stakeholder salience factor(s) has an positive output. The maximum output of 𝑆ℎ𝑎𝑟𝑒ℎ𝑜𝑙𝑑𝑒𝑟_𝑠𝑎𝑙𝑖𝑒𝑛𝑐𝑒𝑖𝜏 and 𝐷𝑖𝑟𝑒𝑐𝑡𝑜𝑟_𝑠𝑎𝑙𝑖𝑒𝑛𝑐𝑒𝑖𝜏 is three, when all three factors in the

variable are positive.

The median of 𝑃𝑜𝑤𝑒𝑟𝑖𝜏 in table 2 is one, which indicates that in at least 50% of the

observations the shareholders have the right to vote on an executive compensation pay. The median of 𝑃𝑜𝑤𝑒𝑟𝑖𝜏 in table 3 is zero, which indicates that in at least 50% of the observations a

board member cannot be removed by a shareholder without limitations.

The values of the variable 𝐿𝑒𝑔𝑖𝑡𝑖𝑚𝑎𝑐𝑦𝑖𝜏 in table 2 represent the percentage of

ownership of the single biggest owner. The highest percentage of ownership observed in the sample is 99.8% (and the lowest 1.06%). In the higher half of the observations, the single biggest owners have at least 8.08% share in the firm. The variable 𝐿𝑒𝑔𝑖𝑡𝑖𝑚𝑎𝑐𝑦𝑖𝜏 in table 3 outlines the

(21)

17

descriptive statistics whether a CEO compensation is linked to the shareholders return. In at least 50% of the observations the CEO compensation is linked to the shareholders return.

The median of the variable 𝑈𝑟𝑔𝑒𝑛𝑐𝑦_𝑖𝜏 is zero, which means that most firms were not in the media due to shareholder controversies. The statistics of the variable 𝑈𝑟𝑔𝑒𝑛𝑐𝑦𝑖𝜏 in table 3

indicate that the higher half of the firm-year observations the directors had at least eight board meetings per year (median = 8), with a maximum of 25 board meetings per year.

The results of the control variables are continuous. The results of 𝑆𝑖𝑧𝑒𝑖𝜏 are scaled to $

1.000.000,-. Also, the observed values in the smallest percentile and the largest percentile of the control variables - 𝑆𝑖𝑧𝑒𝑖𝜏, 𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒𝑖𝜏, and 𝑅𝑂𝐴𝑖𝜏 - are set equal to the 1st and 99th percentile.

This is due to extreme observation values within the smallest and largest percentiles of all three control variables.

The results of control variable 𝑆𝑖𝑧𝑒𝑖𝜏 (table 2 or 3) indicate that there is a large diversity

in the data. The median of the observed values is 20.1842, which indicates that a firm with a firm value (in total assets) of $ 20.1 million separates the lower 50% with higher 50%, while the highest observed firm value in the sample is $ 841 million.

The lowest value of 𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒𝑖𝜏 (table 2 or 3) is zero, which indicates that the lowest

observed total debt to total assets ratio is zero, while the highest observed total debt to total assets ratio is 0.531. The lowest return on assets (𝑅𝑂𝐴𝑖𝜏; table 2 or 3) observed is – 1.19%, while

the highest observed value is 22.57%.

Panel B of both tables shows the covariance matrix (displayed as correlation matrix). This matrix outlines the correlations between the independent variables. A high correlation means that the effect of one independent variable also influences (read: disturbs) the results of the other variables. For the logistic regression it is desirable that the correlation between the independent variables are low. A variable correlated with itself is always 1. As panel B in both tables show, there is no sign of high correlation between the independent variables, which is consistent with my predictions.

4.2 Multivariate results

The next subparagraphs present the results of the logistic regression and whether the hypotheses are true or not. Additionally, all models are tested on their goodness-of-fit and robustness. The used goodness-of-fit tests are the Chi-square goodness-of-fit test and the Hosmer and Lemeshow test. The Wald test of robustness is used to test if the significance of the independent variables.

(22)

18

4.2.1 Regression results 𝐇𝟏 and 𝐇𝟐

Table 4 and 5 provide the results of the first two equations from paragraph 3.2. Table 4 represents the regression results of the shareholder group and table 5 of the directors. Both regressions are based on the same sample, which consists of 628 firm-year observations.

The Chi-square goodness-of-fit test (Prob > chi2_{) for table 4 shows a p-value smaller}

than the critical value of 0.05. This indicates that the overall model is statistically significant. The Hosmer-Lemeshow (HL)6_{statistic presents a p-value of 0.085. A p-value greater than the critical}

value of 0.05 indicates that the model is correctly specified and that the observed and expected observations are close to each other. Based on both results, the overall goodness-of-fit of the first model is correctly specified. And last, the Wald test7_{for table 4 shows that the results are also}

significant when the 628 firm-year observations are clustered into the 226 unique firms.

TABLE 4

Results from logistic regression of shareholder salience’s influence on CSR assurance

Sample consists of 628 firm-year observations from 2010-2014

Log likelihood = -414.66062

Number of obs. LR chi2₍₄₎

Prob > chi2

Pseudo R2

Hosmer & Lemeshow Prob > chi2

= = = = = 628 41.27 0.0000 0.0474 0.0850 Coef. Std. error z-score

𝑆ℎ𝑎𝑟𝑒ℎ𝑜𝑙𝑑𝑒𝑟_𝑆𝑎𝑙𝑖𝑒𝑛𝑐𝑒_𝑖𝜏 0.4206*** 0.1330*** 3.16***

𝑆𝑖𝑧𝑒_𝑖𝜏 0.0057*** 0.0015*** 3.91***

𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒_𝑖𝜏 -0.0033 0.0173 -0.60

𝑅𝑂𝐴_𝑖𝜏 0.0065 0.3051 0.38

**

*** = Significant on a 1% level = Significant on a 5% level

As predicted, the coefficients 𝑆ℎ𝑎𝑟𝑒ℎ𝑜𝑙𝑑𝑒𝑟_𝑆𝑎𝑙𝑖𝑒𝑛𝑐𝑒𝑖𝜏, 𝑆𝑖𝑧𝑒𝑖𝜏, and 𝑅𝑂𝐴𝑖𝜏 are positive and

𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒𝑖𝜏 is negative. Furthermore, 4 shows that the variables 𝑆ℎ𝑎𝑟𝑒ℎ𝑜𝑙𝑑𝑒𝑟_𝑆𝑎𝑙𝑖𝑒𝑛𝑐𝑒𝑖𝜏 and

𝑆𝑖𝑧𝑒_𝑖𝜏 are significant on a 1% level. The result of 𝑆ℎ𝑎𝑟𝑒ℎ𝑜𝑙𝑑𝑒𝑟_𝑆𝑎𝑙𝑖𝑒𝑛𝑐𝑒_𝑖𝜏 indicates that the probability that a firm’s decision to externally assure its CSR report or not is significantly

6_{The results of the Hosmer and Lemeshow goodness-of-fit test for table 4 are presented in Appendix C (table 8).} 7_{The results of the Wald test of robustness for table 4 are presented in Appendix G (table 12).}

(23)

19

influenced by the salience of a shareholder. The results of 𝑆𝑖𝑧𝑒𝑖𝜏 are consistent with research of

Udayasankar (2008), who already showed that the size of a firm significantly influences a firm’s CSR assurance policy, as large-sized firms have a higher public visibility and greater impact on a larger group of stakeholders.

Contrast to prior literature (Mishra and Modi, 2013; Dewi et al., 2014), both 𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒𝑖𝜏

and 𝑅𝑂𝐴𝑖𝜏 are not significant on a 1% or 5% level. This means that, based on this study, CSR

assurance is not significantly influenced by a firm’s leverage or return of assets.

Looking at the results of the stakeholder group directors (table 5), only the variable of 𝑆𝑖𝑧𝑒_𝑖𝜏 is significant. The variable of interest - 𝐷𝑖𝑟𝑒𝑐𝑡𝑜𝑟_𝑆𝑎𝑙𝑖𝑒𝑛𝑐𝑒𝑖𝜏 – is not significant. This

indicates that a firm’s CSR assurance is not significantly influenced by the salience of a director. The results in table 5 confirm Wood and Jones (1995) assertion that CSR is mostly (or only) influenced by shareholders.

TABLE 5

Results from logistic regression of director salience’s influence on CSR assurance

Number of obs. LR chi2₍₄₎

Prob > chi2

Pseudo R2

𝐷𝑖𝑟𝑒𝑐𝑡𝑜𝑟_𝑆𝑎𝑙𝑖𝑒𝑛𝑐𝑒_𝑖𝜏 0.0526 0.1034 0.51 𝑆𝑖𝑧𝑒𝑖𝜏 0.0053*** 0.0014*** 3.78*** 𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒_𝑖𝜏 -0.0019 0.0055 -0.35 𝑅𝑂𝐴_𝑖𝜏 0.0061 0.0171 0.36 ** *** = Significant on a 1% level = Significant on a 5% level

Moreover, the Chi-square goodness-of-fit test (Prob > chi2_{) for table 5 shows a p-value smaller}

than the critical value of 0.05, which indicates that the overall model used for table 5 is statistically significant. The HL8_{statistic of table 5 presents a p-value of 0.2913, which is greater}

(24)

20

than the critical value of 0.05. The model is correctly specified. The Wald test9_{for table 5 shows}

that the results are not significant enough when the 628 firm-year observations are clustered into the 226 unique firms. This is due to the fact that only 𝑆𝑖𝑧𝑒_𝑖𝜏 is statistically significant within the model.

The results of table 4 and 5 indicate that there is a positive association between stakeholder (shareholder) salience and a firm’s decision to externally assure its CSR report. Therefore, we can reject the null hypothesis. H1 is supported. Additionally, consistent with Wood

and Jones’ (1995) assertion, this relation between stakeholder salience and CSR assurance is only observed within the shareholder group. Based on the regression, the salience of the directors did not influence the firm’s CSR policy on external assurance. Therefore, H2 is also supported.

4.2.2 Regression results 𝐇𝟑

Table 6 and 7 present the results for the third hypothesis, in which the shareholder’s and director’s salience attributes – power, legitimacy, and urgency – are analyzed separately.

The Chi-square goodness-of-fit test (Prob > chi2_{) for table 6 shows a p-value smaller}

than the critical value of 0.05, which indicates that the regression model used for table 6 is statistically significant. The Hosmer-Lemeshow (HL)10_{statistic presents a p-value of 0.9673. The}

model is correctly specified. The Wald test11_{also shows that the results of table 6 are still}

significant when the 628 firm-year observations are clustered into the 226 unique firms they represent. Therefore, the results in table 6 are robust enough to use for the analysis.

The results of table 6 indicate that two of the three stakeholder salience variables are statistically significant: 𝑃𝑜𝑤𝑒𝑟𝑖𝜏 and 𝑈𝑟𝑔𝑒𝑛𝑐𝑦𝑖𝜏. The variable 𝑃𝑜𝑤𝑒𝑟𝑖𝜏 is significant on a 1% level and

𝑈𝑟𝑔𝑒𝑛𝑐𝑦𝑖𝜏 is significant on a 5% level. The high significance of the variable 𝑃𝑜𝑤𝑒𝑟𝑖𝜏 is

consistent with the findings from Neill and Stovall (2005), who determined that the primary attribute driving CSR is power. In order for a firm to engage in purposeful CSR, the stakeholder must possess the power to impose his agenda on management. This could also explain why 𝐿𝑒𝑔𝑖𝑡𝑖𝑚𝑎𝑐𝑦_𝑖𝜏 is not significant within the regression. A stakeholder could have a legitimate claim on a firm, but without the power to impose his claim, the firm is less likely to answer it. This assertion should also be true for the factor urgency. Even though a stakeholder claim is urgent, it cannot be imposed successfully on management when power is absent. Nonetheless, based on the results of table 6, not only 𝑃𝑜𝑤𝑒𝑟_𝑖𝜏, but also 𝑈𝑟𝑔𝑒𝑛𝑐𝑦_𝑖𝜏 has a significant influence

9_{The results of the Wald test of robustness for table 5 are presented in Appendix H (table 13).}

10_{The results of the Hosmer and Lemeshow goodness-of-fit test for table 6 are presented in Appendix E (table 10).} 11_{The results of the Wald test of robustness for table 4 are presented in Appendix I (table 14).}

(25)

21

on a firm’s CSR assurance. This could due to the fact that table 6 is based on the shareholder group. Shareholders’ urgency will be much more important to management than any other stakeholder group, considering that shareholders are the owners of the firm. The results of table 6 are therefore not directly applicable on other stakeholder groups. Furthermore, just as in the previous regressions, the coefficients of the control variables are consistent with my predictions and only 𝑆𝑖𝑧𝑒𝑖𝜏 has a significant influence on 𝑝_𝑐𝑠𝑟_𝑎𝑠𝑠𝑢𝑟𝑎𝑛𝑐𝑒.

TABLE 6

Results from logistic regression of shareholder salience’s influence on CSR assurance

Number of obs. LR chi2₍₆₎

Prob > chi2

Pseudo R2

𝑃𝑜𝑤𝑒𝑟_𝑖𝜏 1.3344*** 0.2460*** 5.42*** 𝐿𝑒𝑔𝑖𝑡𝑖𝑚𝑎𝑐𝑦_𝑖𝜏 -0.0167 0.0123 -1.36 𝑈𝑟𝑔𝑒𝑛𝑐𝑦𝑖𝜏 -1.4881** 0.6620** -2.25** 𝑆𝑖𝑧𝑒_𝑖𝜏 0.0052*** 0.0014*** 3.79*** 𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒𝑖𝜏 -0.0046 0.0058 -0.078 𝑅𝑂𝐴_𝑖𝜏 0.0118 0.0180 0.66 **

Now it is time to analyze the director’s group, which is presented in table 7. The Chi-square goodness-of-fit test (Prob > chi2_{) for table 7 shows a p-value smaller than the critical value of}

0.05, which indicates that the overall model used in table 7 is statistically significant. The HL12

statistic of table 7 is 0.9676, which indicates that the model is correctly specified. The Wald test13

for table 7 shows that the results of the independent variables in table 7 are not significant enough when the 628 firm-year observations are clustered on the basis of their unique firm.

12_{The results of the Hosmer and Lemeshow goodness-of-fit test for table 7 are presented in Appendix F (table 11).} 13_{The results of the Wald test of robustness for table 7 are presented in Appendix J (table 15).}

(26)

22

Additionally, none of the director salience variables significantly influences the probability that a firm’s CSR report is externally assured. This is consistent with the result of table 5.

At last, just as in the previous regressions, the coefficients of the control variables are consistent with my predictions and only 𝑆𝑖𝑧𝑒𝑖𝜏 has a significant influence on 𝑝_𝑐𝑠𝑟_𝑎𝑠𝑠𝑢𝑟𝑎𝑛𝑐𝑒.

TABLE 7

Results from logistic regression of director salience’s influence on CSR assurance

Number of obs. LR chi2₍₆₎

Prob > chi2

Pseudo R2

𝑃𝑜𝑤𝑒𝑟_𝑖𝜏 0.2584 0.1666 1.55 𝐿𝑒𝑔𝑖𝑡𝑖𝑚𝑎𝑐𝑦𝑖𝜏 -0.2254 0.1996 -1.13 𝑈𝑟𝑔𝑒𝑛𝑐𝑦_𝑖𝜏 0.0464 0.0291 1.59 𝑆𝑖𝑧𝑒_𝑖𝜏 0.0047*** 0.0014*** 3.39*** 𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒_𝑖𝜏 -0.0025 0.0055 -0.45 𝑅𝑂𝐴_𝑖𝜏 0.0068 0.0173 0.40 **

In conclusion, based on the results of table 6 and 7, the probability that a firm’s CSR report is externally assured is not solely influenced by the power of a stakeholder. On a 5% significance level, the stakeholder factor urgency also plays a role whether or not a firm’s CSR report is externally assured. Therefore, we can reject H3. However, keep in mind that these results are

based on the shareholder group. The director group did not provide the same results as the stakeholder group. The results of this research are therefore not necessarily applicable on other stakeholder groups and needs to be further investigated.

The last chapter outlines the conclusion of this thesis. It also discusses its theoretical and managerial contributions, the limitations of the study and its research implications.

(27)

23

5 Conclusion

The call for more disclosure of non-financial information is a consequence to the awareness that financial statements miss significant information concerning the firm’s corporate social responsibility and its sustainable development (Adams, 2002; Holder-Webb et al., 2009). Nowadays it is not enough for firms to simply engage in CSR. It is also important and desirable to report these activities and disclose its information for verification. Especially now the number of issued CSR reports significantly increase each year (KPMG,2013). When firms report and disclose information on its CSR activities, the manager assesses which stakeholder groups he deems most important to report to. The pragmatic reality is that management simply cannot attend to all actual or potential claims. They need to prioritize their managerial attention over these variety of claims. The theory of identifying and selecting ‘those who really counts’ is called stakeholder salience (Mitchell et al., 1997). In literature (e.g. Angle et al., 1999) only the stakeholders who possess the most stakeholder salience - stakeholder power, legitimacy and urgency - can influence a firm’s responsible stakeholder management team to engage in their CSR agenda.

In this thesis I investigate how stakeholder salience influences a firm’s decision to engage in CSR assurance. It addresses how stakeholders motivate management to provide assurance on their CSR activities, and reports. For the first hypothesis I analyzed if there is a positive association between stakeholder salience and a firm’s decision to externally assure its CSR report by an independent assurance provider. Based on research of Wood and Jones (1995), I tested in my second hypothesis if only the salience of shareholders influences a firm’s decision to externally assure its CSR report by an independent assurance provider. In the third and last hypothesis I tested Neill and Stovall’s (2005) assertion that only the attribute power has a significant influence on a firm’s decision whether or not its CSR report is externally assured. This thesis makes use of a logistic regression, in which the focus lies on firms listed on the S&P 500 Index. In the first two hypotheses, stakeholder salience is investigated as one variable. In the third hypothesis, the stakeholder salience attributes power, legitimacy, and urgency are analyzed separately. The variables firm size, leverage, and return on assets served as control variables.

The results of the shareholder group indicate that there is a positive association between stakeholder salience and a firm’s decision whether its CSR report is externally assured or not. The results of the first regression therefore appear to support the first hypothesis. Additionally, the second regression indicates that this relation between stakeholder salience and CSR assurance does not count for the stakeholder group directors. Only the shareholder group shows a positive association between its salience and a firm’s CSR assurance. This is consistent with research of

(28)

24

Wood and Jones (1995) and the second hypothesis. Therefore, the second hypothesis is supported.

The regression results for the third hypothesis indicate that a firm’s decision to externally assure its CSR report is not solely influenced by the power of a stakeholder. The results of the regressions indicate that the factor urgency also has a significant influence on a firm’s decision to externally assure its CSR report.

This study contributes to literature by showing how stakeholders can influence managerial decision-making on CSR disclosure. Secondly, this study contributes to the literature by providing a new dimension in understanding the relationship between stakeholder salience and CSR, that of CSR assurance. The relationship between the stakeholder salience attributes, the salience it produces, and the influence it has on a firm’s CSR assurance policy, had not been fully explored yet. Thirdly, the findings of this study contribute to literature, as they highlight that not only the power attribute provides an impact on the salience of the claim to externally assure a CSR report, as asserted by Neill and Stovall (2005). This thesis shows that also the urgency of the claim has an impact on its salience. Also, this study shows that the salience of some stakeholder groups (read: directors) have no impact on the CSR assurance claim. It shows that, while one stakeholder group’s claim (read: shareholders) can impact CSR assurance significantly, without that claim even being legitimate, while at the same time the claim of another stakeholder group (read: directors) does not even impact CSR assurance. The consequences of the findings could suggest that legitimate stakeholder parties like society, the environment (read: environmental organizations), and local communities could, when they do not poses power or urgency to impose their claim, get excluded in from the development process of CSR assurance. This could be a big issue to the credibility of CSR assurance, due to the fact that stakeholder groups as these are put into second place, after less legitimate stakeholders like shareholders. The results of this thesis also have important practical implications for regulators. For management the results provide insights into the most important stakeholders, on which they may develop strategies for proactively disclosing CSR information. However, where management does not proactively discloses important CSR information, the government could issue and enforce laws and regulations to enforce the disclosure of important CSR information, which in turn could affect the practice of the auditing community.

A limitation in this study is that the analysis only included firms listed on the S&P 500 Index. The S&P 500 Index only consists of large-sized firms in the United States. Therefore, the results of this research are not necessarily applicable to smaller firms or non-US firms.

(29)

25

An implication of the results is that they are only based on two stakeholder groups, while Andrikopoulos & Webber (2014) summarize a total of eleven stakeholder groups. Initially, the analysis consisted of four stakeholder groups: shareholders, government, society, and environment. However, both the Datastream database and the Wharton Research Data Services (WRDS) database didn’t provide sufficient data to analyze the relationship between CSR assurance and the stakeholder groups government, society, and environment. Therefore, three of the four initial stakeholder groups were removed and replaced by only the stakeholder group directors. Still, only the shareholder group provided satisfying results, which indicates that the results of one stakeholder group are not necessarily applicable to other stakeholder groups. There are still nine stakeholder groups untested and they need to be investigated further.

The influence of stakeholder salience on CSR assurance

Amsterdam Business School