All eyes on you, how far will you go? – The effect of stakeholders’ pressure on the quality of sustainability assurance reports

reports

Supervisor: dr. T.A. (Teye) Marra

MSc A&C, specialization Accountancy & Controlling (EBM869B20 & EBM870B20) University of Groningen

Faculty of Economics and Business – Department of Accounting

June 22, 2020 [Wordcount: 13,180]

"Stakeholders want companies to make a profit, but not at the expense of their staff and the wider community" - Brian Gosschalk – CEO Ipsos Mori

2

TABLE OF CONTENTS

ABSTRACT 3

INTRODUCTION 4

THEORY & HYPOTHESES 8

METHODOLOGY 18

RESULTS 26

ROBUSTNESS TESTS 31

CONCLUSION & DISCUSSION 34

REFERENCES 38

3

All eyes on you, how far will you go? - The effect of stakeholders’

pressure on the quality of sustainability assurance reports

MSc Accountancy & Controlling (EBM869B20 & EBM870B20) June 22, 2020

ABSTRACT

This study empirically examines the influence of stakeholders’ pressure on the quality of sustainability assurance reports, assessed by means of content analysis, from a stakeholder and agency perspective. Academics recently started studying the quality of sustainability assurance, yet, little is known about its determinants (Perego & Kolk, 2012). Using a global panel data sample of 2,033 firm-years over the period 2014 to 2018, I find that environmental protection organizations, consumers and employees exert pressure on firms to provide higher quality sustainability assurance reports. In other words, these stakeholders demand more credible sustainability information to hold firms accountable for their environmental impact, which is in line with stakeholder theory. In contrast to previous findings on stakeholders’ pressure, I find that pressure from institutional investors and creditors prompts firms to deliver lower quality sustainability assurance. Institutional investors and creditors are argued to be able to obtain information from firms via private channels in order to decrease information asymmetry, in accordance with agency theory. My findings indicate that the different types of stakeholders’ pressure are important determinants of sustainability assurance quality. This study’s findings are of great importance to managers since they indicate that stakeholders differently value the quality of external assurance on sustainability information and provide a fruitful starting point for academics to empirically examine the quality of sustainability assurance.

Keywords: Accountability – Managerial capture - Sustainability assurance quality – Sustainability information - Stakeholders’ pressure

Abbreviations

GRI Global Reporting Initiative SA Sustainability Assurance

SAR(s) Sustainability Assurance Report(s) SI Sustainability Information

4

INTRODUCTION

Over the past years, concerns about climate change have risen amongst firms’ stakeholders, governments and the public (European Environment Agency, 2020; Lacy, Cooper, Haywood & Neuberger, 2010; Poortinga et al., 2018). These developments have led to the creation and ratification of the Paris Climate Deal by many countries in 2015. This Deal focuses, amongst other things, on cutting greenhouse gas emissions by firms and people (Mahapatra & Ratha, 2017). Concrete actions to reach the goals of the Paris Climate Deal were recently discussed in Madrid during the United Nations Climate Change Conference, which took place in December 2019 (Green & Spring, 2019). Countries’ representatives could not reach a final agreement on several subjects due to opposing interests. As a response to these environmental concerns, firms increasingly voluntarily disclose environmental, social and other non-financial information (i.e. sustainability information) in sustainability reports (KPMG, 2017; Lacy et al., 2010). These publications of sustainability information (abbreviated: SI) demonstrate firms’ awareness of the social and environmental impact of their actions on their community, country and the world (Vuontisjarvi, 2006). Another attribute of their awareness is trying to deliver their fair share in taking responsibility for the consequences of their actions.

Voluntarily disclosing SI essentially turns private into public information, which is argued to decrease information asymmetry amongst stakeholders (Verrecchia, 2001). Dando & Swift (2003) and Waddock & Goggins (2011), however, claim that the increase in public SI does not directly lead to increased public confidence that firms take sustainability seriously, due to the feeling that firms’ sustainability reports lack in comparability and wholeness (Adams & Evans, 2004; Manetti & Becatti, 2009). According to Adams & Evans (2004), stakeholders feel that firms focus too much attention in sustainability reports on their successful events and leave out events that are harmful to the firm’s reputation, which decreases the credibility of SI to stakeholders. In an attempt to increase the credibility of SI, stakeholders increasingly demand firms to obtain external assurance on their SI (Cheng, Green & Ko, 2015; Zorio, García-Benau & Sierra, 2013). Moroney, Windsor & Aw (2012) found evidence for the credibility-increasing effect of assurance. They found that voluntary environmental information disclosures are of higher quality for firms that have SI assured in comparison to those that do not choose to obtain assurance. The increase in firms choosing to assure SI is noticeable, the percentage of assured sustainability reports from the 250 largest firms in the world (ranked by revenues) rose from 46% in 2011 to 67% in 2017 (KPMG, 2017).

5 Some academics, however, question the quality of assurance on SI (Gürtürk & Hahn, 2016; O’Dwyer & Owen, 2005; Perego & Kolk, 2012). In contrast to audits of financial information, audits on SI are governed by several standards (i.e. ISAE3000, AA1000 and GRI), accountants and non-accountants are allowed to perform audits, different levels of assurance and scope could be applied and a wide variety of procedures and methods are used in the assurance process (Mock, Rao, & Srivastava, 2013). Hence, scholars argue that managers are able to influence the content and process of sustainability assurance (abbreviated: SA) for their own benefits at the expense of transparency and comparability of SI and accountability to stakeholders (Adams & Evans, 2004; Fonseca, 2010; Michelon, Pilonato & Ricceri, 2015; O’Dwyer & Owen, 2005; Smith, Haniffa & Fairbrass, 2011). This phenomenon is called ‘managerial capture’. Fonseca (2010) argues that managerial capture encompasses that managers strategically dominate the process and scope of SA, by merely assuring information that enhances the firm’s image and reputation. Management is for instance able to influence the assurance process by limiting the scope, the level of assurance or by choosing a different assurance provider. This variability in SA poses problems to users of SI to assess whether the information is complete, which items are assured and the quality of assurance.

Freeman (1984) postulates that stakeholders and firms are interdependent since their actions influence each other, and firms maintain relationships with many stakeholders, not only shareholders. This academic developed the stakeholder theory which has led to scholars claiming that firms have to take their key stakeholders into account in sustainability reporting (Huang & Kung, 2010). According to Huang & Kung (2010), stakeholders use SI to hold firms accountable for their environmental impact. In this light, O’Dwyer & Owen (2007) argue that SI should focus on stakeholders’ demands. Some stakeholders, however, view SI disclosures as insufficient and low in credibility, as found by Kuruppu & Milne (2010). In order to increase the credibility of this information, stakeholders are found to demand assurance on SI (Wong & Millington, 2014). Stakeholders are able to stimulate firms to obtain assurance by exerting pressure on them.

Investors’ and creditors’ demand for credible information received specific attention in scientific literature on agency theory. Agency theory posits, amongst others, that there exists information asymmetry between the principal (i.e. investors & creditors) and the agent (i.e. the firm or managers) (Jensen & Meckling, 1976). In an attempt to reduce information asymmetry and monitor agents’ behaviour, principals demand credible financial as well as non-financial information (An, Davey, & Eggleton, 2011). Clinch, Stokes & Zhu (2012) and Fuhrmann, Ott, Looks & Guenther (2017), however, claim and found that only SI assured through a

high-6 quality assurance process increases the credibility of SI and reduces information asymmetry amongst stakeholders.

Stakeholders demand credible SI, but managers are able to capture the SA process, which may decrease the credibility and reliability of SI and the assurance process. Therefore, I study from a stakeholder and agency perspective whether stakeholders’ pressure affects the quality of sustainability assurance reports (abbreviated: SARs). In line with previous studies, the most important stakeholders for SI disclosures and assurance on this type of information are environmental protection organizations, consumers, employees, shareholders, institutional owners and creditors (Fernandez-Feijoo, Romero & Ruiz, 2014; Huang & Kung, 2010; Vitolla, Raimo, Rubino & Garzoni, 2019). These different stakeholder groups are argued and found to be able to exert the most pressure on firms’ SI disclosing behaviour.

Financial audit quality is a widely discussed topic in the field of accounting research (Francis, 2004). Determinants and consequences of sustainability reporting have also been intensely studied in the past (Hahn & Kühnen, 2013). Empirical evidence on the assurance of SI, however, remains scarce (Fuhrmann et al., 2017; Zorio et al., 2013). Various authors argue that little is known about the quality of SARs and its determinants and emphasize the importance of more knowledge creation on this topic (Gürtürk & Hahn, 2016; Hahn & Kühnen, 2013; Kim, Simunic, Stein & Yi, 2011; Perego & Kolk, 2012). Adams & Evans (2004) and Gillet-Monjarret & Rivière-Giordano (2017) stress the importance of more research on the involvement of stakeholders in the SA process. In this light, Perego & Kolk (2012) prompt that the influence of stakeholders’ pressure, as a determinant of SA quality, should be studied to gain insight in quality differences amongst firms’ SARs. Stakeholders’ pressure has been found to be a determinant of sustainability reporting and the decision to obtain assurance on SI (see e.g. Simnett, Vanstraelen, Chua, 2009), this research examines whether this relationship also applies to SA quality. This study aims to answer calls from academics to study determinants of SA quality and fill the gap in the literature on the influence of stakeholders’ pressure, as a determinant, of SARs’ quality (Gürtürk & Hahn, 2016). I aim to study the following research question:

To what extent does stakeholders’ pressure affect the quality of sustainability assurance reports?

The quality of SARs is assessed by means of content analysis (Gürtürk & Hahn, 2016). Using a global panel data sample from 2014 to 2018, I find that increasing pressure from

7 environmental protection organizations, consumers and employees positively affects the quality of SARs. In other words, stakeholders demand more credible, and therefore high-quality assured, SI to hold firms accountable for their environmental impact as their power increases. These results are in line with the stakeholder theory and previous literature on stakeholders’ pressure in relation to SA (see e.g. Wong & Millington, 2014). In contrast to the literature, yet in line with agency theory, I find that increasing pressure from institutional owners and creditors negatively affects the quality of SARs. Institutional owners and creditors are argued to be able to gather information via private channels from firms as their power increases, which may decrease information asymmetry and supersede the demand for high-quality assured SI (Solomon, Solomon, Norton & Joseph, 2011; Wong & Millington, 2014). Information gathered through private channels is argued to be less costly than assured SI and sufficient to fulfil the informational needs (Simnett et al., 2009). In contrast to the other main stakeholders, I do not find evidence that shareholders’ pressure influences the quality of SARs. This may be the result of shareholders thinking that investments in sustainability are too costly and only benefit stakeholders, which decreases the demand for a certain quality assurance (Bénabou & Tirole, 2010). All results, except for creditors’ pressure’s influence on the quality of SARs, are robust to using alternative panel data samples in order to control for coders’ subjectivity and controlling for different assurors and assurance standards applied in the audit. Creditors may not be interested in the quality of SARs and indirectly discourage firms to spend additional resources on SA by maintaining strict debt covenants, as found by Casey & Grenier (2015) and Zorio et al. (2013).

This study contributes to literature on SA in five ways. First, in response to the call from Perego & Kolk (2012), this is the first study to empirically examine and find that stakeholders’ pressure affects the quality of SARs whilst employing a global sample. The different types of stakeholders’ pressure are found to be important determinants of SA quality, which is globally generalizable. The results indicate that stakeholders differently value assurance and demand a predetermined level of quality for which they are willing to exert pressure on firms’ managers. While several stakeholders positively influence the quality of SARs, other stakeholders negatively or not influence this quality. Therefore, it is of importance to individually assess stakeholder groups and their demands for the quality of SARs and not treat them as one group. Second, in contrast to previous studies, I find that institutional investors and creditors negatively affect the quality of SARs. This effect is argued to be the result of institutional investors’ and creditors’ power to obtain SI through private channels that substitutes for assured SI (Solomon et al., 2011). This new perspective may be further explored in future studies. Third, I contribute

8 to the ongoing debate on managerial capture in SA and indicate that this is still present, since powerful stakeholders are found to be able to pressurize managers to provide higher or lower quality SARs (see e.g. Michelon et al. 2015). Fourth, following the call from Gürtürk & Hahn (2016) to employ quantitative methods in SA research, this study is the first to empirically examine the influence of stakeholders’ pressure on the quality of SARs whilst employing quantitative research methods. This creates a new research field for academics to quantitatively study the quality of SA. Fifth, I contribute to the vast body of literature on stakeholders’ pressure in relation to SI and provide evidence that pressure from stakeholders does not only apply to SI, yet also to SA.

The remainder of the paper is structured as follows. At first, the next section contains a literature review of relevant studies and the theoretical framework, which results in the development of the hypotheses. Thereafter, the study’s methodology comprising data collection and analysis are exemplified. The following sections discuss the empirical results of the study and the robustness checks. The final section presents the conclusion and discussion.

THEORY & HYPOTHESES

Quality of sustainability assurance

External assurance on SI is provided by external independent parties in order to verify the information incorporated in sustainability reports (Perego & Kolk, 2012). In this study, a sustainability assurance report (SAR) is defined as the verification statement provided by an independent assuror containing, amongst other things, the reported findings, applied assurance standard, level of assurance and assuror’s opinion (Perego & Kolk, 2012).

Although the increase in firms obtaining assurance on SI is noticeable (KPMG, 2017), the quality of these assurance reports is questioned and criticized by various scholars (Gürtürk & Hahn, 2016; Perego & Kolk, 2012). Academics’ criticism is fuelled by the variability in SARs’ contents and the influence that managers have on the assurance process. In contrast to audits of financial information, many variations are possible in assurance on SI (Mock et al., 2013). While financial audits are governed by the standards of the International Federation of Accountants (i.e. ISAE), audits of SI are most times governed by the ISAE3000 and AA1000 assurance standards. Another difference is that financial audits are performed by financial auditors, whilst assurance on SI can be provided by many different parties, such as, financial auditors, specialist consulting firms, non-governmental organizations and stakeholder panels (Junior, Best, & Cotter, 2014). Firms are allowed to opt for different levels of assurance (i.e.

9 limited or reasonable) and a wide variety of assurance procedures could be applied in the assurance process.

Some academics, as a result of the variations in SARS, argue that managers strategically capture the assurance process for their own benefits and not in order to be accountable to stakeholders and disclose transparent information (Adams & Evans, 2004; Fonseca, 2010; Manurung & Basuki, 2010; Michelon et al., 2015; O’Dwyer & Owen, 2005; Smith et al., 2011). Owen, Swift, Humphrey & Bowerman (2000) claim that managerial capture is present in the assurance process of SI, because managers pursue commercial and managerial interests over the goal of transparency in reporting. More recently, Manurung & Basuki (2010) argue that the materiality of sustainability reporting is impaired because of management’s authority over the scope of the SA engagement, resulting in less relevant SI to stakeholders. Michelon et al. (2015) found that assurance of SI is not related to the quantity or quality of SI disclosures, therefore arguing that managerial capture is present in SA.

Various academics studied the quality of external assurance on SI, in a response to questions about the quality of the assurance process (see e.g. Gürtürk & Hahn, 2016). These studies are mainly descriptive in nature and used content analysis to examine the quality of SARs. O’Dwyer & Owen (2005) critically examined the quality of SARs from EU-companies, by employing content analysis, that were short-listed for the European Sustainability Reporting Awards Scheme. These SARs are deemed to be from the ‘best’ organizations and assurance is provided by the ‘best’ providers. They studied whether assurance practices foster more transparent reporting and accountability to stakeholders by comparing assurance practices and reports to international guidelines provided by the Global Reporting Initiative1 _{(abbreviated:}

GRI), AccountAbility2_{, the IFAC}3 _{and the FEE}4_{. These academics found that management’s}

control over the SA process is severe, stakeholders’ involvement in the assurance process is limited and the independence of the assuror is questionable.

More recently, by analysing the contents and quality of SARs from a set of firms from the Fortune Global 250 over a period of 10 years, Perego & Kolk (2012) found great variability in the quality of SARs because of different assurance providers and audit standards applied in

1 _{The Global Reporting Initiative (GRI) is an independent global organization that aims to support businesses and} governments in sustainability reporting and assurance by keeping stakeholders’ interests in mind (GRI, 2020). 2 _{AccountAbility is an international consulting and standards firm that developed the AA1000 assurance standard} (AccountAbility, 2020).

3 _{The International Federation of Accountants (IFAC) is an influential organization for the accounting profession} worldwide (IFAC, 2020). They developed the ISAE3000 standard which governs audits other than audits and reviews of financial information (like sustainability information) (IFAC, 2013).

4 _{The Fédération des Experts Comptables Européens (FEE) (or Accountancy Europe) is an organization that} represents the accounting profession in Europe (FEE, 2006).

10 sustainability audits. The authors developed a coding scheme based on the O’Dwyer & Owen (2005) framework to assess the quality of SARs. Perego & Kolk (2012) analysed and coded the contents of SARs based on their coding book and assigned a quality score ranging from 0 (i.e. lowest quality) to 27 (i.e. highest quality) to each SAR. The quality of SARs was found to differ considerably between countries, assurance providers and industries (Perego & Kolk, 2012). Although the quality increased over time, the quality was still argued to be quite low. These academics found that, on average, accountants tend to provide higher quality assurance than non-accountants. EU-firms and firms operating in environmentally sensitive industries (e.g. mining industries) are also associated with higher quality SARs than non-EU-firms and firms from other industries.

Gürtürk & Hahn (2016) expanded the Perego & Kolk (2012) coding framework and employed it to examine the quality of SARs. The Gürtürk & Hahn (2016) framework is also applied in this study to assess the quality of SARs and is described in more detail in the Methodology section and Appendix A. In line with Perego & Kolk (2012), Gürtürk & Hahn (2016) found differences in the quality of SARs provided by auditors and non-auditors. Accountants’ assurance statements are deemed less extensive and the set of procedures used less diverse, resulting in lower quality SARS than those from non-accountants. Gürtürk & Hahn (2016) devote these differences to the primary use of the ISAE3000 standard by auditors. Non-auditors used the more stakeholder-oriented AA1000 standard more often (79%).

Furthermore, Zorio et al. (2013) found that higher quality SARs are associated with larger firms and audits performed by accountants. The academics employed a sample of Spanish firms and a content analysis, comparable to the O’Dwyer & Owen (2005) framework, to assess the quality of SARs.

Stakeholders’ pressure

A company’s success is determined by a variety of stakeholders (e.g. consumers, employees, NGOs, creditors and suppliers) that are in pursuit of different interests (Buchholz & Rosenthal, 2005; Freeman, 1984; Laplume, Sonpar & Litz, 2008). The influence of stakeholders’ pressure, as a determinant of different characteristics of environmental and other non-financial information, has been intensely studied in the past (see e.g. Hahn & Kühnen, 2013). For instance, Huang & Kung (2010) found that external (i.e. governments and consumers) and internal stakeholders (i.e. employees and shareholders) exert pressure on firms to disclose more environmental information by using a sample of Taiwanese firms. In light of SI quality, Fernandez-Feijoo et al. (2014) found that pressure from consumers, employees, shareholders

11 and environmental protection organizations enhances the transparency of sustainability reports whilst employing a global sample. More important, Wong & Millington (2014) surveyed investors, procurers and environmental protection organizations in order to examine the influence of stakeholders’ pressure on firms’ decision to assure their SI. They found that stakeholders that value SI are more likely to demand assurance, which is in line with stakeholder theory.

Based on the literature review presented, managerial capture over the assurance process could still be present due to management’s influence over the assurance process (Michelon et al., 2015). Differences in SARs’ quality are mostly explained from the point of different assurance standards and assurors. Since management is hypothesized and found to be able to influence the assurance process and reports, quality differences in assurance reports cannot solely be the result of different assurance providers or assurance standards. Stakeholders demand credible SI, for which they may pressurize firms to disclose high-quality assurance statements through their influential position. In this light, various academics stress the importance of more research on determinants of SARs’ quality and the role of stakeholders in the assurance process on SI (Gürtürk & Hahn, 2016; Hahn & Kühnen, 2013; Kim et al., 2011; Perego & Kolk, 2012). To the best of my knowledge, there is not a study that examines the influence of stakeholders’ pressure on the quality of SARs. The following paragraph contains a description of the applied theories in this study and the rationales for their use.

Theoretical background

This study takes a stakeholder and agency theory perspective to examine the influence of stakeholders’ pressure on the quality of SARs. Stakeholder theory is argued to be the most suitable theory to study the determinants of sustainability reporting (Snider, Hill, & Martin, 2003). This claim is corroborated by the vast body of literature on the influence of stakeholders’ pressure on the quality and quantity of sustainability disclosures and the application of assurance on SI (see e.g. Darnall, Seol & Sarkis, 2009; Fernandez-Feijoo et al., 2014; Moroney et al., 2012; Smith et al., 2011; Vitolla et al., 2019). Stakeholder theory concentrates on the holistic relationship between stakeholder groups in the society and the firm, and views the firm as part of the broader societal system (An et al., 2011). In addition to stakeholder theory, agency theory is applied, which is mostly used to study assurance on SI (Wong & Millington, 2014). Agency theory focuses on the relationship between capital providers (i.e. shareholders and creditors) and the firm. This theory operationalizes information asymmetry, which provides additional explaining power for capital providers’ informational needs (An et al., 2011).

12 Stakeholder theory

The main idea behind stakeholder theory is that managers should not only focus on the interests of shareholders but on those of all stakeholders in their business operations (Freeman, 1984). A firm has many different stakeholders but the most prominent and powerful stakeholders’ demands are decisive for a firm’s behaviour (Dong, Burritt, & Qian, 2014). In this light, Guthrie, Petty & Ricceri (2006, p. 256) state: “According to stakeholder theory, an organization’s management is expected to undertake activities deemed important by their stakeholders and to report on those activities back to the stakeholders... stakeholder theory highlights organizational accountability beyond simple economic or financial performance”. This theory has led to several implications, such as, a company’s survival being dependent on the ability to manage relationships with the firm’s various stakeholder groups (Elijido-Ten, Kloot & Clarkson, 2010) and the perceived pressure from stakeholders increasing the likelihood that firms obtain assurance on SI (Darnall et al., 2009). The degree of importance of the stakeholder (group) for a firm’s operations determines the efforts that firms exert in managing the particular stakeholder relationship (Gray, Owen, & Adams, 1996). Another important aspect of stakeholder theory is accountability, which in accounting terms entails providing an objective and complete overview of a firm’s performance in social and environmental fields in order to be accountable to stakeholders (O'Dwyer & Owen, 2007). Environmental protection organizations, consumers, employees, shareholders, institutional owners and creditors are argued to be the most important stakeholders for firms’ SI (see e.g. Fernandez-Feijoo et al., 2014; Vitolla et al., 2019). The influence of pressure from these stakeholder groups on the quality of SARs is examined in this study.

Agency theory

Two particular groups of stakeholders received specific attention in finance, accounting, auditing and economic literature, namely shareholders and creditors. The relationship between shareholders and the firm can be characterized as a principal-agent relationship (or agency relationship), which is the result of the separation of ownership and management (Jensen & Meckling, 1976). The shareholder is the principal (i.e. owner) and the firm’s management the agent in this relationship. A similar agency relationship applies between creditors and management (An et al., 2011). Agency theory assumes that managers make decisions, whilst investors and creditors bear the risks related to these decisions. Principals, as a result, aim to monitor agents’ decisions to prevent managers from taking decisions that are not in their best interests. A key assumption of agency theory is that there exists ‘information asymmetry’

13 between the two parties (Jensen & Meckling, 1976). Information asymmetry indicates that managers are assumed to have an informational advantage over principals since they are involved in the daily operations (An et al., 2011). Principals aim to reduce information asymmetry through, amongst others, monitoring activities. Non-financial information is also employed in these monitoring activities and is argued to decrease information asymmetry (Verrecchia, 2001).

In this light, Blackwell, Noland & Winters (1998) argue that firms choose to assure SI in order to reduce information asymmetry with creditors. Solely auditing information, however, need not be sufficient to reduce information asymmetry, as found by Clinch et al. (2012). The quality of the audit process too affects the credibility of the reported information and, thus, may help to decrease information asymmetry. Fuhrmann et al. (2017) corroborate this finding and argue that solely a high-quality SA process decreases information asymmetry.

The hypotheses that are tested in this study are developed in the following paragraph. The pressure from environmental protection organizations, consumers and employees on the quality of SARs is examined from a stakeholder theory perspective. Pressure from shareholders, institutional owners and creditors on the quality of SARs is studied from a stakeholder and agency perspective.

Hypotheses development

Environmental protection organizations

As a response to past, current and future environmental issues, environmental protection organizations were founded to voluntarily monitor firms’ environmental behaviour and compliance to laws (Vitolla et al., 2019). Firms with a history of detrimental environmental behaviour are increasingly pressurized by these organizations to behave more sustainable, produce more environmentally friendly products and report on the improvement (Huang & Kung, 2010). This stakeholder group is argued to demand credible information about sustainability issues to monitor and hold firms accountable for the environmental consequences of their production processes and other polluting activities. According to Fernandez-Feijoo et al. (2014) and Liu & Anbumozhi (2009), environmental protection organizations are able to exert pressure on firms by influencing and mobilising the media and public opinion in support or against the firm. Protection organizations, thereby, monitor firms’ pollution levels and other environmentally unfriendly activities and lobby and report to (local) governments to implement regulations or punish firms through fines (Vitolla et al., 2019). Academics found that pressure from environmental protection organizations positively affects the environmental disclosure

14 level (Huang & Kung, 2010) and the transparency of disclosed SI (Fernandez-Feijoo et al., 2014). Perego & Kolk (2012) found descriptive evidence that firms in more polluting industries are more likely to deliver higher quality SARs. In order to empirically assess this claim, the following hypothesis is proposed:

H1: Environmental protection organizations’ pressure positively affects the quality of sustainability assurance reports.

Consumers

A firm’s reputation and image are essential elements that consumers take into account before (re)purchasing a product or service (Huang & Kung, 2010). Important factors of corporate reputation are perceived corporate social responsibility and corporate sustainable management, which are both found to increase consumers’ repurchase intentions (Becker-Olsen, Cudmore & Hill, 2006; Shin, Van Thai, Grewal & Kim, 2017). Consumers are especially interested in SI about pollution from production processes, consumers’ health and safety and environmentally friendly behaviour to form an opinion about a firm’s reputation (Vitolla et al., 2019). In this respect, Park (1999) found that firms disclose more extensively on investments in sustainability initiatives and provide higher quality disclosures when they have better relationships with consumers. Fernandez-Feijoo et al. (2014) argue and found that firms are more likely to disclose more transparent SI when pressure from consumers is higher. Huang & Kung (2010) argue that consumers exert pressure on firms through purchasing and repurchasing behaviour. Since managers are able to capture the assurance process on SI in order to increase brand image and reputation, consumers could pressurize firms to provide higher quality assurance in order to obtain more credible and reliable information (Fonseca, 2010). Consumers’ demand for high-quality assured SI results in the following hypothesis:

H2: Consumers’ pressure positively affects the quality of sustainability assurance reports.

Employees

Comparable to environmental protection organizations, employees become more environmentally aware of firms’ actions as well (Huang & Kung, 2010). Employees realize that bad environmental performance may lead to fines and a diminished reputation, which could ultimately result in decreasing economic prospects that may hurt employees (Vitolla et al., 2019). In order to monitor firms’ environmental behaviour, employees use SI. They demand this information to be disclosed and of a predetermined level of quality. In light of this, Vitolla

15 et al. (2019) found that increasing pressure from employees positively affects the quality of sustainability reporting. Employees are, amongst others, able to exert pressure on firms through external trade unions or organizations (e.g. environmental protection organizations) to make sure their demands reach the firm’s management (Huang & Kung, 2010). Their power to exert pressure increases with the number of employees, since they usually become more organized and united (Vitolla et al., 2019). In order to monitor and hold firms accountable for their environmental behaviour, employees demand credible high-quality SI, leading to the following hypothesis:

H3: Employees’ pressure positively affects the quality of sustainability assurance reports.

Shareholders

Shareholders consider firms’ sustainable performance to be an important facet and increasingly use environmental, social and governance data in valuing firms (Grewal, Hauptman, & Serafeim, 2017). Amel-Zadeh & Serafeim (2018) argue that shareholders use SI, amongst others, to gain more insight in financial risks related to environmental behaviour (e.g. waste management & carbon emission) and to form an opinion about the corporate reputation. In this light, SI disclosures are found to be associated with lower costs of capital (Dhaliwal, Li, Tsang, & Yang, 2011) and more accurate analysts’ forecasts (Dhaliwal, Radhakrishnan, Tsang, & Yang, 2012). Fuhrmann et al. (2017), however, found evidence that some shareholders do not consider all SI to be credible and refrain from using not credible SI for valuation purposes. Managers may increase the credibility of SI by providing high-quality assurance on this information, which is found to decrease information asymmetry amongst investors (Fuhrmann et al., 2017). According to Liu & Anbumozhi (2009), shareholders exert pressure on firms through investment choices to sell or refrain from buying shares. These decisions may influence the attractiveness of firms’ shares and the cost of capital.

On the other hand, some academics argue that shareholders think that investments in sustainability are too costly and only benefit stakeholders (Bénabou & Tirole, 2010). This statement is corroborated by Di Giuli & Kostovetsky (2014) who argued and found evidence that sustainability investments are associated with negative future stock returns and a diminishing ROA. According to Masulis & Reza (2015), shareholders punish managers for maintaining a private agenda in sustainability investments and placing stakeholders’ over shareholders’ values. Managers, for example, use sustainability investments to improve their personal reputation or social networks through the use of company funds. Shareholders, as a

16 response, are argued to be less eager to invest in firms when the costs for gaining and maintaining relationships with non-shareholding stakeholders are high (Schwarzmüller, Brosi, Stelkens, Spörrle & Welpe, 2017). Shareholders’ reluctance towards stakeholders’ interests (e.g. SI disclosures and assurance or sustainability investments) and sustainability may decrease shareholders’ demand for high-quality disclosure and assurance of SI. Thus, academics’ rationales for shareholders’ demand for the quality of SARs are twofold, leading to the following hypothesis:

H4a: Shareholders’ pressure affects the quality of sustainability assurance reports.

Institutional owners

Institutional owners (i.e. pension funds, investment banks, mutual funds & venture capital funds) are a type of shareholder that usually hold a large block of firms’ shares (Christofi, Christofi, & Sisaye, 2012). Institutional ownership increased over the years and this type of investor is especially interested in sustainability, for instance, climate change (Cotter & Najah, 2012). Cotter & Najah (2012) and Solomon et al. (2011) argue that institutional owners are, amongst other reasons, interested in SI to monitor firms’ exposure to social, reputational and financial risks related to climate change. In this light, Dyck , Lins, Roth & Wagner (2019) argue that institutional owners demand credible information for monitoring activities. Once institutional investors own a large part of firms’ shares, they are able to exert pressure on firms via their voting rights in order to stimulate environmentally friendly behaviour (Prado-Lorenzo, Gallego-Alvarez, & Garcia-Sanchez, 2009). Dyck et al. (2019) argue that institutional investors also exert pressure on firms through investment choices to sell or refrain from buying shares. Selling shares or refusing to invest in firms is argued to decrease the attractiveness of the firm to other shareholders, which may result in a higher cost of capital. Liesen, Hoepner, Patten & Figge (2015), for example, argued and found that firms that face greater pressure from institutional investors are more likely to disclose information on greenhouse gas emissions. These investors monitor firms’ behaviour and therefore demand high-quality assured SI in order to reduce information asymmetry, monitor firms and hold organizations accountable for their actions.

On the other hand, Funaoka & Nishimura (2019) and García-Meca & Consuelo Pucheta-Martínez (2017) argue that institutional owners are able to obtain SI via private channels from firms through, amongst others, their voting rights and possible close ties with the firm’s management. In this light, Solomon et al. (2011) found that institutional investors are able to

17 obtain SI via private channels to assist firms in environmental risk management. This ability to obtain private information may impair institutional owners’ demand for high-quality assurance since they are able to obtain information via private channels. Therefore, I postulate:

H4b: Institutional owners’ pressure affects the quality of sustainability assurance reports.

Creditors

Firms’ environmental and social performance may influence their current and future profitability (Yang, He, Zhy, & Li, 2018). A negative (environmental) reputation or possible litigation costs related to firms’ environmental misdemeanours could negatively influence firms’ profitability. Creditors, as a response, use financial as well as environmental and social information to estimate firms’ current and future profitability and ability to pay back loans’ principal and interest (Elsakit & Worthington, 2013). Creditors also use SI in loan negotiations, risk assessments and monitoring activities of firms’ environmental behaviour (Barako & Brown, 2008; Huang & Kung, 2010). Yang et al. (2018) corroborate this claim and find that SI disclosures reduce information asymmetry amongst firms’ creditors. In light of this, Huang & Kung (2010) found that highly leveraged firms disclose more SI than less leveraged firms. According to Dawkins & Fraas (2010), creditors exert pressure on firms by demanding a risk premium in lending decisions from firms that are deemed poor environmental performers. Roberts (1992) argues that creditors may also pressurize firms to change policies by recalling loans or declining to extend further credit. Creditors may want to reduce information asymmetry through high-quality assured SI (Fuhrmann et al., 2017).

On the contrary, Casey & Grenier (2015) found that highly leveraged US-firms are less likely to obtain assurance on SI, since banks are indirectly discouraging firms to spend resources on assurance by maintaining debt covenants. In this light, there is evidence that banks do not pressurize clients to disclose high-quality SARs. Another argument for creditors not demanding high-quality assurance on SI is provided by Tan, Tsang, Wang & Zhang (2020) who argue that banks are also able to retrieve information at their clients via private channels which is argued to be less costly and of high-quality (Solomon et al., 2011; Thompson & Cowton, 2004). Creditors’ hypothesized ability to obtain private information may supersede the demand for high-quality SARs, as with institutional owners. Evidence and arguments on creditors’ demand for SA quality are mixed, which leads to the following hypothesis:

18 The hypotheses are illustrated in the conceptual model presented in Figure 1.

Figure 1. Conceptual model

METHODOLOGY

The sampling procedure started with collecting (stand-alone) sustainability reports and integrated reports accompanied by external assurance statements from publicly listed firms uploaded on the public sustainability disclosure database of the GRI, in line with Fernandez-Feijoo et al. (2014). These reports are published by firms operating in many different countries, which is explicated in Table 2. Currently, approximately 60,000 sustainability reports are uploaded on the GRI’s database (GRI, 2020). BSc students Business Administration of the University of Groningen collected 2,008 sustainability reports accompanied by external assurance statements covering fiscal years 2014 to 2018 from the GRI database for their bachelor’s thesis in order to code on content. Thereafter, MSc Accountancy & Controlling students, including myself, from the same university collected 373 sustainability reports accompanied by external assurance statements covering fiscal years 2016 to 2018 for their master’s thesis. The MSc students did not only use the GRI database, but also directly retrieved

19 external assurance statements from organizations’ websites when these were not available on the GRI database. Two corrections were made to the initial sample collected by the BSc students. First, a correction for duplicate observations (64). Second, a correction for assurance statements that did not clearly state the covered fiscal year (44).

Data on ownership structures, the number of employees and the year of incorporation were collected from the Orbis database and were matched against the data on SARs. Since firms’ ownership structures at specific historical moments in time are often not publicly available, 106 observations had to be dropped due to missing key-data. Financial and organizational data were retrieved from the Thomson Reuters Worldscope database. After matching the data from the Thomson Reuters Database against data on SARs, another 134 observations were omitted due to missing observations, resulting in a final sample of 2,033 firm-years. Table 1 presents the structure of the final sample. A more in-depth analysis of the final sample per fiscal year and continent of origin is presented in Table 2. A global sample is chosen in order to prevent bias resulting from national or continental laws and culture, to enhance the generalizability of the results and to holistically capture the global quality of SA and its determinants. Firms from Africa and Oceania are underrepresented in the sample since Africa is less developed and Oceania is by far the smallest continent. In line with Perego & Kolk (2012), firms located in the American continent are less represented than European and Asian firms, since the litigation risks are higher in the US and US-firms are more inclined to provide assurance when this is obligated by law, which it currently is not.

Table 1. Final sample

Observations

Initial sample from BSc students 2,008

Correction for duplicate values and unclear fiscal years (108)

Initial sample from MSc students 373

Missing data on ownership structures (106) Missing financial and organizational data (134)

20 Table 2. Sample per fiscal year and continent of origin

Observations per fiscal year Observations per continent Fiscal year Observations Percent Continent Observations Percent

2014 427 21.0% Africa 66 3.2% 2015 502 24.7% America 353 17.4% 2016 532 26.2% Asia 787 38.7% 2017 461 22.7% Europe 750 36.9% 2018 111 5.4% Oceania 77 3.8% Total 2,033 100.0% Total 2,033 100.0%

Dependent variable: Quality of sustainability assurance reports

In line with previous studies, the quality of SARs is measured by performing a deductive content analysis based on the O’Dwyer & Owen (2005) framework. This framework finds its roots in guidelines and recommendations provided by AccountAbility (2003a, b), the FEE (2002, 2006) and the GRI (2002). These organisations aim to improve the comparison between sustainability reports, the credibility of reports and stakeholder inclusiveness in sustainability reporting. The guidelines specify the minimum contents of SARs. The contents of assurance statements are coded by applying the coding rules of Gürtürk & Hahn (2016), who expanded and modified the rules developed by Perego & Kolk (2012). The latter developed 19 generally accepted and employed criteria in order to code the content of SARs. Gürtürk & Hahn (2016) added four criteria based on more recent guidelines, developments in sustainability reporting and observations of SARs. The coding book and coding rules applied in the content analysis are disclosed in Table 6 in Appendix A. Based on the coding book, SARs were assessed on 23 criteria in which a maximum score varying from 1 to 3 could be obtained per criteria. In total, SARs could obtain a minimum score of 0 and a maximum score of 33, in which 0 corresponds to the lowest and 33 the highest degree of quality. SARs were mainly assessed on referencing, mentioning or (clearly) describing certain facets (e.g. the title of the underlying sustainability report, the name and type of assuror and the report date) in the coding procedure.

In order to increase the reliability of the content analysis, multiple coders were involved in the coding procedure (Gürtürk & Hahn, 2016). According to Potter & Levine-Donnerstein (1999), shared judgments are fostered by using this method. Biases that might result from individual estimations in the coding procedure are decreased in this way. Intercoder reliability, which equals the amount of agreement amongst multiple coders concerning a measured variable, is increased by pilot and final testing the coding scheme (Neuendorf, 2016).

21 SARs were coded in two phases. The first phase concerns SARs coded by BSc students for their bachelor’s thesis. Upfront, thesis supervisors allocated SARs to individual students to analyse on content. In order to increase intercoder reliability, a general meeting was organized before the coding procedure in which thesis supervisors exemplified the coding book and coding rules to the BSc students. Multiple students coded identical SARs which were analysed afterwards on variances in obtained scores. These variances remained within acceptable limits to use for statistical testing.

The second phase of the coding procedure comprised the SARs from the fiscal years 2016 to 2018, which was conducted by four MSc students Accountancy & Controlling in order to collect data for their master’s thesis. Upfront, a pilot test to increase intercoder reliability was conducted in which eight SARs were randomly drawn from the panel data sample and coded separately on content by each coder. In line with Gürtürk & Hahn (2016), differences in obtained scores and rationales for choices were discussed and agreement on the interpretation of the coding book and rules was accomplished, which increases the reliability of the measurement. Every coder independently coded an equal part of the sample and uncertainties or disagreements concerning individual coding rules were discussed afterwards, resulting every time in agreement. These discussions translate into decent training for coders and increasing robustness of the framework. Through documentation of agreements on the interpretation of coding rules and following discussions, the reliability, replicability, consistency and internal validity of the study are increased. The coding framework’s construct validity is increased through the use of the framework in previous studies from O’Dwyer & Owen (2005), Perego (2009), Perego & Kolk (2012) and Gürtürk & Hahn (2016) (Smith, 2017).

Independent variables: Stakeholders’ pressure

Studies that examine the influence of stakeholders’ pressure on the quality and quantity of sustainability reporting and the decision of firms to obtain external assurance on SI were used to operationalize the independent variables (see e.g. Fernandez-Feijoo et al., 2014; Vitolla et al., 2019). In line with Liesen et al. (2015) and Dyck et al. (2019), all continuous independent and control variables are lagged at t-1 in relation to the dependent variable. This lag is integrated in the regression in order to better state a causal relationship between the independent and the dependent variable instead of a relationship of association (Bellemare, Masaki, & Pepinsky, 2017). A description of the measurement and data collection for each variable is presented in Table 7 in Appendix B.

22 Environmental protection organizations’ pressure (ENVPRE)

Firms that perform activities with significant environmental impact are argued to experience more pressure from environmental protection organizations than firms that perform other activities, in line with Fernandez-Feijoo et al. (2014), Simnett et al. (2009) and Vitolla et al. (2019). ENVPRE is defined as a dummy variable that takes a value of ‘1’ if the firm operates in an industry that could have a significant impact on the environment and a ‘0’ for firms that do not operate in these industries. The following industries are classified as having a great impact on the environment: agriculture, automotive, aviation, chemical, construction, construction materials, energy, energy utilities, forest and paper products, logistics, metal products, mining, transport, waste management and water utilities (Branco & Rodrigues, 2008; Gamerschlag, Möller & Verbeeten, 2011; Tagesson, Blank, Broberg & Collin, 2009). The industry classification is based on each firm’s NAICS 2017 (North American Industry Classification System) code, retrieved from the Orbis database. The classification of industries that have a great impact on the environment and industries that have not, was manually performed by the researcher, in line with Fernandez-Feijoo et al. (2014) and Vitolla et al. (2019). Most NAICS 2017 industry classification descriptions are comparable to the classifications from the literature described above, therefore decreasing subjectivity.

Consumers’ pressure (CUSPRE)

Consumers’ pressure’s operationalization is derived from Fernandez-Feijoo et al. (2014) and Vitolla et al. (2019) who argue that firms operating in industries that are well-known by the public, because their goods or services are bought by many, are more affected by consumers’ pressure than other firms. Consumers’ pressure is measured by a dummy variable that adopts the value ‘1’ for firms that operate in industries that are more visible to consumers and ‘0’ for firms operating in other industries. The core industry classification is derived from Branco & Rodrigues (2008) and Sweeney & Coughlan (2008). Fernandez-Feijoo et al. (2014) added several industries that experience more consumers’ pressure, which are also incorporated in this study. The following industries are deemed to experience more severe consumers’ pressure: beverage and tobacco products, commercial services, consumer durables, financial services, food, healthcare, household and personal products, media, textiles and apparel, tourism and leisure, toys, universities, waste management and water utilities. The industry classification is identical to the process described for the variable environmental protection organizations (ENVPRE).

23 Employees’ pressure (EMPPRE)

Pressure from employees is measured by firm size and defined as the natural logarithm of the total amount of employees, in line with Fernandez-Feijoo et al. (2014), Huang & Kung (2010) and Vitolla et al. (2019). Fernandez-Feijoo et al., (2014) argue that the operationalization of employees’ pressure is related to firm size (Aldama, Amar & Trostianki, 2009). These scholars argue that employees working in larger firms are better organized and are therefore able to pressurize managers to hear and adhere to their demands (Huang & Kung, 2010). Data on the number of employees was retrieved from the Thomson Reuters Worldscope database and merged with data from the Orbis database. The correlation between data from the two databases was for each fiscal year over 99%, therefore the databases could be merged. Missing data was hand-collected by the researcher, which comprised approximately thirty percent of the observations.

Shareholders’ pressure (SHAPRE)

Shareholders’ pressure is determined by the concentration of ownership, based on the operationalization of Liu & Anbumozhi (2009), Huang & Kung (2010) and Vitolla et al. (2019). In line with Huang & Kung (2010), ownership concentration is measured as the total percentage of shares held by the largest shareholders. The top 10 shareholders are argued to be the largest shareholders (Liu & Anbumozhi, 2009). Data on ownership structures was retrieved from the Orbis database per fiscal year. Ownership structures were calculated and the data was checked and cleaned in Microsoft Office Excel.

Institutional owners’ pressure (INSTPRE)

Institutional owners’ pressure’s operationalization is derived from Liesen et al. (2015), who argue that firms with increasing institutional ownership experience more pressure to disclose SI. In line with and An & Zhang (2013) and Dyck et al. (2019), institutional owners’ pressure is measured by the stake (i.e. total percentage of shares owned) institutional investors have in the firm. Institutional owners are (investment)banks, financial companies, hedge funds, venture capitalists, insurance firms, private equity firms and mutual funds & trusts (Funaoka & Nishimura, 2019). The data collection process is identical to the process described for the variable shareholders’ pressure (SHAPRE).

Creditors’ pressure (CRPRE)

Creditors’ pressure is measured by the degree to which firms rely on external debt financing (Huang & Kung, 2010; Liesen et al., 2015; Liu & Anbumozhi, 2009; Prado-Lorenzo et al.,

24 2009). Increasing leverage is argued to result in increasing pressure from creditors since firms become more dependent on their creditors. In line with Liesen et al. (2015), creditors’ pressure is measured by dividing total debt by common equity. Data was retrieved from the Thomson Reuters Worldscope database.

Control variables

In order to reduce biases in the results, several control variables are included in the study that might affect the quality of SARs. Since research on determinants of SARs’ quality is scarce, few control variables are incorporated in the regression (Zorio et al., 2013). Firstly, this study controls for the age (AGE) of the firm, since older firms are argued to be more stable than younger firms (Vitolla et al., 2019). In this light, Liu & Anbumozhi (2009) found that older firms disclose higher quality sustainability reports than younger firms. Since higher quality sustainability reports are found to be associated with higher quality SARs, this variable is controlled for (Moroney et al., 2012). In line with Vitolla et al. (2019), the age of the firm is measured as the firm’s year of incorporation until the end of 2018. The data is retrieved from the Orbis database and approximately twenty percent of the missing values were hand-collected.

Secondly, this study controls for firm performance, since external assurance on SI is a costly decision (Simnett et al., 2009). Firms with lower performance may decide to withdraw from higher quality SARs, in order to save costs. Hahn & Kühnen (2013) argue that firms with higher performance may experience more flexibility to choose for assurance. In line with Casey & Grenier (2015), profitability (PRFT) is measured as the total return on assets, that is total income before extraordinary items over total assets.

Thirdly, Perego & Kolk (2012) argued and found descriptive evidence for differences in SARs’ quality due to country-specific characteristics. They found that EU-firms are associated with higher quality SARs than non-EU-firms. A dummy variable (EU) that takes a value of ‘1’ for EU-firms and a value of ‘0’ for all other firms, is integrated in the regression to control for this effect.

Statistical analysis model and techniques

The sample comprises panel data, which allows for panel data analysis techniques (Vitolla et al., 2019). Panel data techniques possess more explanatory power than techniques used for cross-sectional data (Martínez-Ferrero & García-Sánchez, 2017). Each observation in the panel is characterized by an individual dimension (firm) ‘i’ and a time dimension (year) ‘t’

25 (Cuadrado-Ballesteros, Martínez-Ferrero, & García-Sánchez, 2017). All continuous variables are winsorized at the first and 99th _{percentiles (Dyck et al., 2019). In order to test whether a}

random effects model or a pooled OLS model could be employed, a Breusch-Pagan Lagrange Multiplier test is performed (Vitolla et al., 2019). The test indicates that a random effects model is more appropriate than the pooled OLS model since the panel data structure influences the results. Thereafter, a Hausman specification test is performed, which tests whether random individual firm effects are correlated with the explanatory variables and therefore indicate either a random or fixed effects model is preferred (Smith, 2017). The test’s outcome is insignificant, which indicates that the random effects model is the most appropriate model to employ in this study (Martínez-Ferrero & García-Sánchez, 2017). The random effects model is more efficient than the fixed effects model considering that it allows for the inclusion of time-invariant variables in the regression models such as ENVPRE, CUSPRE, AGE and EU, which would be omitted by the fixed effects model and decrease the model’s explaining power.

In line with Martínez-Ferrero & García-Sánchez (2017), I test whether year-fixed effects bias my result. Based on a F-test, I conclude that year-fixed effects influence the results and therefore year dummies for each fiscal year are incorporated in the regression models to control for this effect.

In order to prevent biases in the random effects model, I test whether heteroskedasticity or autocorrelation (i.e. serial correlation) are present in the panel data sample (Martínez-Ferrero & García-Sánchez, 2017). In line with Cuadrado-Ballesteros et al. (2017), heteroskedasticity and autocorrelation are respectively tested through the Breusch-Pagan/Cook-Weisberg and Wooldridge test. Heteroskedasticity indicates that the error terms of the regression model do not have equal variances across time and are therefore influenced by the test variables. Based on the outcomes of the test, I conclude that heteroskedasticity is present in the model. In order to mitigate the effects of heteroskedasticity, I run the random effects model with White’s heteroskedasticity robust standard errors, which clusters individual firms and therefore controls for correlation across residuals per unique firm (White, 1980). I test whether autocorrelation is present in the panel by employing the Wooldridge test (Wooldridge, 2011). Based on the outcomes of the test, I conclude that autocorrelation is not present in the panel.

The Statistical model presented below is used to test whether different forms of stakeholders’ pressure influence the quality of SARs. As mentioned above, i represents individual firms in the panel data sample and t takes a value of (fiscal year) 2014 to 2018. ENVPRE, CUSPRE, EMPPRE, SHAPRE, INSTPRE and CRPRE are the explanatory variables

26 in the regression. AGE, PRFT and EU are the control variables. 𝜇 symbolises the unobserved heterogeneity amongst firms and 𝑣 the error term.

𝑆𝐶𝑂𝑅𝐸 = 𝛽 + 𝛽 𝐸𝑁𝑉𝑃𝑅𝐸 + 𝛽 𝐶𝑈𝑆𝑃𝑅𝐸 + 𝛽 𝐸𝑀𝑃𝑃𝑅𝐸 + 𝛽 𝑆𝐻𝐴𝑃𝑅𝐸 + 𝛽 𝐼𝑁𝑆𝑇𝑃𝑅𝐸 + 𝛽 𝐶𝑅𝑃𝑅𝐸 + 𝛽 𝐴𝐺𝐸 + 𝛽 𝑃𝑅𝐹𝑇 + 𝛽 𝐸𝑈 + ∑𝛽 𝑌𝐸𝐴𝑅 + 𝜇 + 𝑣 [Statistical model]

RESULTS

Descriptive statistics concerning the final sample are presented in Table 3. The dependent variable’s (i.e. SCORE) descriptive statistics are in line with Gürtürk & Hahn (2016), who examined the quality of SARs and employed the same framework as this study does. This study presents an overall mean of 17.77 for SCORE, which is slightly higher than the Gürtürk & Hahn (2016) study that presented a mean score of 16.86. Their study examined SARs from fiscal year 2013, which may signal that the quality of SARs increased over time. This study found a standard deviation of 4.31 for SCORE, whilst the Gürtürk & Hahn (2016) study presented a standard deviation of 5.41, providing descriptive evidence that SARs’ quality may have converged over time. The lowest quality score reported is 7 and the highest 28 after winsorizing the data, in line with Gürtürk & Hahn (2016). The proxies for environmental protection organizations’ (ENVPRE) and consumers’ pressure (CUSPRE) respectively present a mean of 0.475 and 0.315, indicating that more firms experience pressure from environmental protection organizations than from consumers in my sample, in line with Vitolla et al. (2019). The natural logarithm of the amount of employees in this sample is 9.71 on average, whilst Vitolla et al. (2019) presented an average natural logarithm of the amount of employees of 9.63. This indicates that this study examines firms of a comparable size. 30.2 percent of the shares are held on average by the top 10 shareholders (SHAPRE), in contrast to Huang & Kung (2010) that displayed less concentrated ownership, namely 17.4 percent of the shares. Institutional owners (INSTPRE) hold on average 27.5 percent of the shares in this sample, whilst Dyck et al. (2019) reported 21.5 percent institutional ownership on average. Pressure from creditors (CRPRE) results in an average leverage ratio of 1.274. Firms in the sample display on average a return on assets ratio of 0.039 and on average are incorporated 58 years ago. Lastly, 36.9 percent of the firms are domiciled in Europe.

27 The Pearson Correlation Matrix is presented in Table 4, which presents all pairwise correlation combinations and significance levels for the variables included in the model. The highest pairwise correlation states -0.622 between ENVPRE and CUSPRE, yet it remains below the 0.7 threshold for multicollinearity issues. In order to verify whether multicollinearity does not bias the results and interpretation of this study, a variance inflator factor (VIF) analysis is performed on all variables included in the model. The highest displayed VIF values are for CUSPRE and ENVPRE respectively 1.78 and 1.72 and remain within acceptable boundaries. Therefore, I conclude that multicollinearity does not pose a problem for this study.

Test of hypotheses

Table 5 presents the results of the panel data’s statistical analysis whilst employing the random effects model in order to test the six hypotheses. The hypotheses focus on whether stakeholders’ pressure affects the quality of SARs (SCORE). The coefficient’s sign and level of significance are of interest to the interpretation of the results. The control model displays the results of the regression in which all control variables are regressed on the dependent variable (SCORE). In Models Ⅰ to Ⅵ, the individual effects of the different types of stakeholders’ pressure on SCORE are tested in univariate analyses. The explanatory variables’ levels of significance and the signs of the coefficients in the individual models are mostly in line with the results of the multivariate regression presented in model Ⅶ. This model demonstrates the impact of all explanatory and control variables on the quality of SARs.

As for the first hypothesis, higher environmental protection organization’s pressure (ENVPRE) is significantly positively associated with higher quality SARs (p<.01). This result is in line with Perego & Kolk (2012), who found descriptive evidence that firms operating in more polluting industries tend to obtain higher quality SARs. My findings also corroborate with Huang & Kung (2010) and indicate that environmental protection organizations monitor firms operating in more polluting industries through SI and pressurize these firms to provide higher quality SA in order to be more credible. Therefore, firms operating in these industries tend to deliver higher quality SARs on average.

28 Table 3. Descriptive statistics

N = 2,033

Variable Mean Std. dev. Median Q1 Q3 Min Max

SCORE 17.774 4.312 17.000 15.000 21.000 7.000 28.000 ENVPRE .475 .499 0.000 0.000 1.000 0.000 1.000 CUSPRE .315 .465 0.000 0.000 1.000 0.000 1.000 EMPPRE 9.711 1.504 9.782 8.777 10.736 5.389 12.967 SHAPRE .302 .201 .252 .156 .401 .026 .949 INSTPRE .275 .207 .220 .120 .370 .009 .872 CRPRE 1.274 1.575 .750 .373 1.464 .001 9.244 PRFT .039 .053 .032 .009 .060 -.125 .238 AGE 58.484 43.966 44.000 24.000 84.000 5.000 201.000 EU .369 .483 0.000 0.000 1.000 0.000 1.000

Table 4. Pearson correlation matrix

SCORE ENVPRE CUSPRE EMPPRE SHAPRE INSTPRE CRPRE PRFT AGE EU

SCORE -1.000 ENVPRE -0.102*** -1.000 CUSPRE -0.039* -0.622*** -1.000 EMPPRE -0.045** -0.114*** -0.019*** -1.000 SHAPRE -0.040* -0.062*** -0.063*** -0.106*** -1.000 INSTPRE -0.127*** -0.132*** -0.020 -0.016 -0.422*** -1.000 CRPRE -0.044** -0.116*** -0.216*** -0.097*** -0.063*** -0.031 -1.000 PRFT -0.006 -0.084*** -0.024 -0.044**- -0.067*** -0.069*** -0.264*** -1.000 AGE -0.100*** -0.024 -0.074*** -0.170*** -0.042* -0.016 -0.093*** -0.020 -1.000 EU -0.070*** -0.008 -0.036 -0.128*** -0.033 -0.069*** -0.115*** -0.078*** -0.252*** -1.000

Notes: Table 4 presents the pairwise correlations between all variables included in the model for the final sample (i.e. 2,033 observations). ***, ** and * respectively represent (two-tailed) significance at the .01, .05 and .10 level.

29 Table 5. Random effects regression results

Independent variable Predicted sign Control model

Model Ⅰ Model Ⅱ Model Ⅲ Model Ⅳ Model Ⅴ Model Ⅵ Model Ⅶ

ENVPRE + .893 (3.41) *** 1.903 (5.45)*** CUSPRE + .627 (2.33) ** 1.875 (5.12)*** EMPPRE + .175 (2.09)** .148 (1.78)* SHAPRE +/- -.688 (-1.33) .453 (0.79) INSTPRE +/- -2.297 (-4.07)*** -2.068 (-3.26)*** CRPRE +/- -.099 (-1.34) -.146 (-1.96)** PRFT + .275 (0.14) .687 (0.35) .407 (0.21) .384 (0.20) .354 (0.18) .561 (0.29) -.297 (-0.15) 1.018 (0.50) AGE + -.008 (-2.83) *** -.008 (-2.81)*** -.009 (-2.99)*** -.009 (-3.10)*** -.008 (-2.85)*** -.008 (-2.86)*** -.008 (-2.78)*** -.010 (-3.50)*** EU + -.269 (-0.98) -.262 (-0.97) -.237 (-0.87) -.315 (-1.16) -.273 (-1.00) -.195 (-0.72) -.240 (-0.88) -.089 (-0.33) YEAR-FIXED EFFECTS

Yes Yes Yes Yes Yes Yes Yes Yes

N 2,033 2,033 2,033 2,033 2,033 2,033 2,033 2,033

R² 0.013 0.023 0.015 0.018 0.015 0.029 0.015 0.056

RHO 0.520 0.516 0.521 0.521 0.518 0.512 0.519 0.499

Notes: Table 5 presents the results from the panel data regression with random effects of the different types of stakeholders’ pressure on the quality of SARs. The value in parenthesis presents the t-statistic with White’s heteroskedasticity robust standard errors. ***, ** and * respectively indicate significance at the .01, .05 and .10 level (two-tailed).