Does the Interaction of National Culture with Governance Structure Influence Quality and Quantity of CSR Disclosure?

Transcription

1 Does the Interaction of National Culture with Governance Structure Influence Quality and Quantity of CSR Disclosure? Shayuti Mohamed Adnan 1, Chris J. van Staden 2 and David Hay 3 1 Phd candidate, Department of Accounting & Finance, The University of Auckland Business School, New Zealand. Corresponding author shayuti_adnan@yahoo.com 2 Professor of Accounting, Department of Accounting and Information Systems, University of Canterbury, Christchurch, New Zealand. 3 Professor of Auditing, Department of Accounting & Finance, The University of Auckland Business School, New Zealand. 1

2 ABSTRACT We provide data on CSR reporting practices of large corporations operating in socially and environmentally sensitive industries in four countries: China, India, Malaysia and the UK. We argue that there are cultural influences in the CSR reporting model. Based on institutional theory, we predict that culture interacts with the governance structure and government ownership in influencing the quality and quantity of CSR disclosure. Observations on 403 annual reports, corporate websites and CSR stand-alone reports of 203 companies are consistent with our expectations. The results show that in China, both the quality and quantity of CSR disclosure increase significantly with the existence of CSR board committees, where the culture is one of collectivism, rather than of individualism. We also found that government-owned companies in Malaysia provide CSR disclosure of a quality higher than non-government owned companies. A similar relationship does not apply to other countries. Our results testify to institutional theory that CSR reporting is influenced by organizational settings and institutional arrangements of a country. For policy makers, it is suggested that CSR reporting policy should be drafted in a way that suits the local culture. 1. INTRODUCTION Findings in the literature are not consistent in describing the motivations for CSR disclosure of companies across countries (Williams, 1999; Adams and Kuasirikun, 2000; Haniffa and Cooke, 2005; Holland and Foo, 2003; Amran and Devi, 2008). For example, previous literature documented that corporate governance influences corporate disclosure practice either positively or negatively, depending on the country of origin (Kamla, 2007; Xiao et al., 2005; Haniffa and Cooke, 2005; Wang et al., 2008; Eng and Mak, 2003). It is also reported that institutional ownership could either encourage or hinder disclosure. However, little explanation has been given for the inconsistencies of these findings. We analyse two factors that are related to the inconsistent findings: issues of content analysis and the existence of interaction variables in the CSR reporting model. Firstly, content analysis involves a degree of subjectivity. Research findings are not comparable because there are differences in unit of analysis, reports to analyse, disclosure scores and indices (see Hooks and Van Staden, 2011). For example, it has been argued that focusing only on annual reports in measuring CSR disclosure provides a distorted picture of CSR practices within companies (Unerman, 2000; Holland and Foo, 2003; Frost et al., 2005). However, research that analyses all means of reporting often has small sample sizes because the data is hand-collected; and thus, the findings lack generalizability. In relation to measurement aspects, researchers analyse disclosure quality and quantity using various methods; and as Hooks and van Staden (2011) point out, little is known about the relationship between the quantity and quality of disclosure. To the best of the authors knowledge, there is a lack of research that employs comprehensive CSR data and has a substantial sample size. Secondly, we argue that there are cultural influences in the CSR reporting model. Cross-country research has documented that CSR reporting practices between countries are different (e.g. see Holland and Foo, 2003; Adams and Kuasirikun, 2000; Buhr and Freedman 2001; Silberhorn and Warren, 2007). Some inferences are also made, showing that culture is important in explaining the variation of CSR reporting practices across countries (e.g. Williams, 1999; Van Der Laan Smith et al. 2005). In addition, there is a claim stating that the current model of CSR has been developed mainly in the West and thus may not be tailored to the needs of societies in the East (Kamla, 2007). If such a claim is true, we would expect to see some variation in CSR reporting practices between countries in Asia and Western countries. More importantly, previous research has shown that a governance mechanism which is effective in one country might not be suitable for another country (e.g. Doidge, 2007). Thus, we aim to minimize the gap in the literature by suggesting that a simultaneous relationship between culture, corporate governance, and ownership structure exists in the CSR reporting model. The major challenge appears to be the need to develop a comprehensive framework for the assessment of the drivers of CSR reporting across countries. 2

3 We put forward three research objectives. The first is to explore the quantity and quality of CSR reporting in China, India, Malaysia and the UK. This is done through a comprehensive content analysis procedure. The second is to investigate the effects of national culture on CSR reporting. Finally, the third objective is to examine whether there is an interaction of cultural variables with factors that influence CSR reporting. Specifically, a total of 203 large corporations operating in eight industries (i.e. energy; oil and gas; materials; manufacturing; transportation; automobiles and components; alcohol, tobacco, casino and gambling; pharmaceutical, biotechnology and drugs, and utilities) were selected from the Compustat Global and Mergent Online database. Sentences in their reports concerned with CSR were identified based on a checklist of 65 items, which had been modified from Global Reporting Initiative (GRI) indicators. CSR disclosure quantity is computed by sentence counting and the quality is determined based on a scale of 0 to 4. Overall, the results show that CSR reporting practice is different across countries; UK corporations provide the most CSR information, followed by India, Malaysia and China. A considerable amount of support was found indicating that CSR reporting is enhanced by the presence of CSR board committees. In relation to cultural issues, results indicate the interaction of individualism with CSR board committees in the CSR reporting model. In the model without an interaction variable, CSR reporting is prominent in countries in which the society is individualistic. However, the presence of CSR board committees enhances the CSR reporting of companies in which the society is collectivist. Accordingly, the existence of CSR board committees is more effective in China than in other countries. Finally, the results show that in Malaysia and China, the quality of CSR information in the government-linked companies is higher than in their non-government counterparts. A discussion on institutional background and general context of the countries in this study is provided in the explanation of the findings. This study contributes to the literature in its research methodology and theoretical framework. Firstly, we obtained CSR information from various sources of reporting: these include; annual reports, standalone CSR reports, and corporate websites. In addition, the measurement of CSR disclosure is extended to include both the quality and quantity. Hence, we provide a comprehensive database within which CSR reporting is measured. Secondly, the CSR reporting measurement is adapted and modified from GRI, thus enabling this research to be replicated. Thirdly, we contribute to the CSR reporting literature in emerging markets. Finally, we contribute to the institutional theory by analysing the possibility of the theory to describe the heterogeneity, in addition to homogeneity, in accounting practice. The remainder of this paper is organized as follows. Section 2 provides theoretical frameworks, reviews literature and presents the hypotheses. Section 3 describes research design and Section 4 presents results. In the final section the findings are discussed and conclusions are drawn. 2. THEORETICAL FRAMEWORK AND HYPOTHESES DEVELOPMENT Corporate Social Responsibility (CSR) Since the 1970s, issues surrounding CSR have been discussed in several accounting papers (see Mathews, 1997; Gray, 2002; Deegan and Soltys 2007; Parker, 2011); however, researchers have not been able to provide a single definition of what CSR includes (Owen, 2008). The term interchangeably refers to sustainability accounting (Gray, Bebbington and Walter, 1993); social accounting (Cooper et al., 2005); social and environmental accounting (Guthrie et al. 2008); and social, environmental and ethical accounting (Adam and Larrinaga Gonzalez, 2007). Owen (2008) describes the concept of CSR as a moving social construction, that responds to many issues such as ecological and societal change, mood, social inquiry and researchers scope of inquiry (Parker, 2011). Generally, the concept aims to address the limitations of the traditional accounting framework, and it considers the same range of issues, namely the social and environmental impacts of corporate activity (Bebbington, 2001; Guthrie and Boedker, 2006). 3

4 As such, CSR is a broad concept which could be examined from multiple aspects of accounting; these include the definition, measurement, reporting systems and performance. Given the broad aspect, this research only focuses on CSR reporting because generally, a company cannot measure (and report) something that it does not manage. Thus, CSR reporting is viewed from two perspectives (Gray et al., 1995). Firstly, CSR as an addendum to conventional accounting activity which assumes the financial community to be the main users. In this regard, the current accounting system is accepted as the status quo (perhaps with some improvements in relation to voluntary disclosure aspects). Secondly, CSR is placed within the role of information in organization-society dialogue. This viewpoint expects corporations to play a greater role than just the maximization of profit (Adams, 2002; Gray, 2002) Thus, CSR is viewed in the context of an organizational interaction between a society and all its constituent parts including its natural environment, employees, communities, and customers. Institutional theory Institutional theorists assert that institutions are less likely to change than other structures (Zucker, 1977), bringing about stability and inertia or homogenization of organizations (i.e. isomorphism) (DiMaggio and Powell, 1983). The structures are termed as cognitive, normative, and regulative. The cognitive structure describes what seems to be institutionalized in an organization as being culturally supported and conceptually correct (DiMaggio and Powell, 1983). As a result, organizations tend to copy or model procedures used in other organizations, if such practice is perceived as conceptually right (Rahaman et al,. 2004). In the case of CSR reporting, similar firms competing in the same industry are more likely to benchmark and imitate those firms considered to be superior. This process is referred to as mimetic isomorphism. Normative isomorphism deals with pressure to conform to a set of norms and rules developed by a professional and occupational group. This structure maintains the appropriateness of shared values within the organization. For instance, the existence of global standards such as GRI could govern the CSR reporting of multinational corporations, which is generally voluntary. Finally, regulative structure (also termed as coercive isomorphism) refers to the role of regulation in bringing about institutionalization. It refers to the source of pressure to conform to institutionalised procedures (Rahaman et al., 2004). For example in Malaysia, it is argued that the coercive structure, i.e. the government involvement in CSR activities and its investment have influenced CSR reporting of government-linked-companies in Malaysia (Amran and Devi, 2008). Adam and Larrinaga-Gonzalez (2007) suggest that, institutional theory is more potent as an explanation of social and environmental accounting. This argument is based on the premise derived from the process of legitimation. Suchman (1995) and Milne and Patten (2002) assert that legitimation is not only strategic, but also institutional in nature (Suchman, 1995; Milne and Patten, 2002). Thus, proponents of institutionalism depict legitimacy as a result of congruency between the organization and its cultural environment, with a greater focus on the cognitive rather than the evaluative aspects (Amran and Devi, 2008). This point shows that institutional theory is useful in describing CSR reporting. Ball and Craig (2010) assessed the capacity of institutional theory in providing an understanding of organizational change in response to social and environmental issues. Specifically, they questioned firstly if sufficient attention has been paid to the discussion of CSR reporting in a broader institutional environment; and secondly whether proponents of normative models of social and environmental accounting have paid sufficient attention to the choices about accounting which are embedded in cultural and historical frameworks. (page 284). They comment that institutional theory has not been regarded as a theory or organisational change, but usually as an explanation of the similarity (isomorphism) and stability of organizational arrangements in a given population (page 284). Thus, they proposed an extension of the institutional theory by incorporating Lounsbury (1997) s framework in the model. They argue that (the) variation in political outlook can help us to understand how the timing, organization, and acceptance of environment accounting initiatives are likely to differ according to national context (Lounsbury, 1997). 4

5 Similarly, Harrison and McKinnon (1986) provided a framework, which suggests that accounting changes are driven by three main factors: responsive events, intrusive events and cultural environment.they argue that these events influence accounting systems, and that there is a cultural influence within and between the various systems in an organization. This point shows that culture is an important aspect that should not be omitted when CSR reporting is being described. Accordingly, we argue that cultural aspects and institutional factors should simultaneously influence CSR reporting practice across countries. For example, globalization and the development of voluntary standards in relation to CSR reporting such as GRI are considered as the external events that shape the strategy of CSR reporting in an organization. However, choices on accounting policy still exist at the organizational level. For example, the options to disclose CSR information in an corporate annual report, or to have CSR board committees in place, or to seek professional assurance for CSR information are all instances of managerial choice. In other words, it is an institutional lens that shapes these choices. Nonetheless, cultural attributes, including norms and values, also determine the option that an organization takes. A model that attempts to explain this scenario is presented in Figure 1. The model describes that the institutional environment exists in two layers: international and organizational. At each level, isomorphism (i.e. coercive, normative and mimetic isomorphism) describes the CSR reporting practice. However, cultural attributes tie the knot within and between isomorphism. Culture and CSR reporting << Insert Figure 1>> Withrop (1991) defines culture as the arrangement of beliefs and customs through which social relations are expressed. It can also be interpreted as a set of standards for behaviour considered authoritative within a society (Withrop, 1991). Hofstede (1980, p. 5) refers to culture as the collective programming of the minds which distinguishes the members of one group from another. Such mental programming could exist in several layers at various levels including national, regional, ethnic, religious affiliation, gender and social class levels in addition to organizational and corporate levels. Given this definition, based on a survey of 117,000 IBM employees, Hofstede (1980) expresses culture in five dimensions, namely: power distance, individualism, uncertainty avoidance and masculinity. These five cultural taxonomies explain the difference in the behaviour of individuals from country to country. Power distance refers to the extent to which the unequal distribution of power is tolerated within a society. Individualism is the extent to which the individual acts independently as opposed to collectivism, where people prefer to be in a group. Uncertainty avoidance refers to the situation where people feel threatened by unknown situations. Masculinity represents stress on achievement, heroism, assertiveness and material success; whereas, feminine society emphasizes relationships, modesty, caring for the weak and the quality of life 4 (Hofstede, 1980; Hofstede, 2001). In the context of Hofstede s national culture, Williams (1999) found that uncertainty avoidance and masculinity are statistically related to voluntary environmental and social disclosure (VESAD). The study hypothesizes that the levels of uncertainty avoidance negatively influence the extent of VESAD. To illustrate, companies that operate in a society which has high levels of uncertainty avoidance have a preference for secrecy because of the necessity to restrict information in order to avoid possible conflict and uncertainty of competition and to ensure the preservation of security in the society (Williams, 1999). Whereas, firms in more masculine-biased societies disclose less social and environmental information because they encounter lower social expectations and demands for information related to environmental and social matters. 4 Later work by Hofstede (2001) added another variable, long versus short term orientation but this has not been studied as widely. 5

6 Willams s (1999) model has provided evidence to explain the variation in CSR disclosure by companies in Asia-Pacific. However, the model was tested based only on quantity of disclosure. Aspects of quality of information are thus identified as a research gap that is examined in this study. Van Der Laan Smith et al. (2005) performed an analysis on CSR disclosure which is more extensive than Williams (1999) study because they analyse both the extent and quality of disclosures. They conducted content analysis on 32 Norwegian/Danish companies and 26 US companies in the electric power generation industry. Based on Hofstede s (2001) masculinity-femininity concept, they contend that a masculine society is more concerned about power and economic status, whilst a feminine society puts more emphasis on social goals such as relationships, helping others and the physical environment. As a result, CSR reporting is expected to be better (in terms of quality and extent) in a feminine society, as opposed to in a masculine society. Given this, Van Der Laan Smith et al. (2005) hypothesize that the amount of CSR reporting in Norwegian/Danish companies should be greater than in US companies. The findings provide significant support for the variations in quality and extent of CSR reporting in these countries; culture is inferred as important in explaining the variation. On the basis of Hofstede s framework, Gray (1988) developed four additional cultural accounting values that specifically relate to corporate reporting practices. These include; professionalism versus statutory control, uniformity versus flexibility, conservatism versus optimism and secrecy versus transparency. Gray (1988) suggests that disclosure relates to secrecy. Secrecy increases with uncertainty avoidance and power distance, and decreases with individualism and masculinity. If this assumption is correct, it would suggest that culture plays a role in explaining the difference in the corporate disclosure of countries throughout the world. Gray (1988) also suggests that Asian managers are more secretive, collective, and have a high tendency to avoid uncertainty. Accordingly, financial reports prepared by Asian corporations are expected to contain less voluntary information than that of other societies. For example, in the context of China, Chinese society is characterized as having high levels of collectivism and power distance and strong uncertainty avoidance. Chinese society tends to adhere to rules and regulations and companies tend to disclose little voluntary information in their annual reports. Therefore, it is argued that Chinese culture does not promote voluntary disclosure (Huafang and Jianguo, 2007). Chau and Gray (2001) confirm the secretive nature of Chinese reporting practices in their comparative studies of companies in Hong Kong and Singapore and those in the US and the UK. However, in a more recent study, Qu and Leung (2006) found that regardless of the notable secrecy level in Chinese society, Chinese listed companies are now more willing than in previous decades to provide voluntary information in their corporate annual reports. In the study, Qu and Leung (2006) investigated whether voluntary disclosure with regard to corporate governance can be found in Chinese listed companies annual reports as a result of the changed cultural environment. Content analyses of 120 companies showed that 85 per cent of the sample made disclosures. In essence, the results demonstrated that disclosure in Chinese society has improved despite the argument that the society is generally secretive (Qu and Leung, 2006). Several recent studies also reported similar results, finding that Chinese corporate reports are not generally secretive; factors such as listing status, globalizations and ownership motivate the voluntary disclosure in China (Wang et al., 2008; Cheung et al., 2010). Accordingly, the relationship between national culture and CSR reporting might not be straightforward. For example, it can be argued here that an individualistic society pays less attention to CSR reporting as it values the maximization of shareholders wealth more than societal or environmental needs. This viewpoint would expect a negative relationship between individualism and CSR reporting. However, the relationship could be positive if the shareholders, in making their investment decisions, also require social and environmental information in addition to the monetary data. 6

7 Thus, we first assess whether CSR reporting practices vary across countries and if culture influences CSR reporting. Later, we will establish the relevant arguments on the interaction of culture with factors that influence CSR reporting. The following hypotheses are stated: H 1 : The quality and quantity of CSR information are different across China, India, Malaysia and the UK H 1a : National culture influences the quantity and quality of CSR information across China, India, Malaysia and the UK CSR reporting and Corporate Governance The existence of a CSR board committee is considered important in explaining CSR reporting practices across countries. For example, drawing upon legitimacy theory, Wahyuni et al. (2009) argue that companies with an environmental committee are more likely to voluntarily disclose information on greenhouse gas emissions than companies without such a committee. The presence of a committee shows a company s concern in regard to legitimizing their environmental reputation (Neu et al. 1998). Likewise, engaging the board of directors as members of an environmental (or social) committee demonstrates the company s endeavours to improve its environmental (or social) performance. As a result, we would expect the existence of CSR board committees to enhance CSR reporting. As the existence of a CSR board committee is part of the corporate governance subject, the relevant literature on this issue is examined. Researchers in the voluntary disclosure stream reported that corporate governance influences accounting disclosure (e.g. Chen and Jaggi, 2000; Eng and Mak, 2003). However, the literature provides inconsistent evidence on the way in which corporate governance influences corporate reporting. For example, Haniffa and Cooke (2002; 2005) found that companies produce less voluntary information when the number of non-executive directors on the board is high. This finding is opposite to their prediction and the predictions of agency theory. Supposedly, a high number of outside directors represents good monitoring of activities by the board and limits managerial opportunism (Fama and Jensen 1983). Therefore, board composition (measured by the proportion of outside directors) should be positively associated with voluntary disclosure. Ho and Wong (2001) assert that inconsistent findings are attributed to the roles of governance mechanisms in corporate disclosure policy, which can be either, complementary, or substitutive (Kelton and Yang, 2008). Governance is complementary when it can restrain managers from withholding information for their own benefit. As a result, good governance strengthens the quality and comprehensiveness of corporate reporting. On the other hand, governance mechanism is substitutive when it reduces information asymmetry and opportunistic behaviours in the firm. Thus, corporate governance and voluntary disclosure are negatively associated in a substitutive-governance environment (Kelton and Yang, 2008). Additionally, it can also be argued that country differences could influence the governance mechanism and thus the disclosure. For instance, Eng and Mak (2003) reported that the relationship between the number of outside directors and voluntary disclosure is moderated by country variables. They found that in Hong Kong, disclosure increases when a company has a large number of outside directors. In contrast, in Singapore, the amount of disclosure decreases with an increase in the number of outside directors. Hence, we argue that national culture plays a role in explaining the association between corporate governance and disclosure. For example, using Hofstede s masculine-feminine concepts, Van Der Laan Smith et al. (2005) view corporate governance structures from two perspectives: contractarianism and communitarianism (Bradley et al., 1999). They argue that governance structures in US companies are contractarian (shareholder) oriented, whereas Norwegian/Danish companies are communitarian (stakeholder) oriented (see also Simnett et al., 2009). Corporate governance structures in contractarian countries (i.e. the USA) revolve around shareholder relationships and promoting 7

8 shareholder value whereas the structures in communitarian countries (i.e. Norway and Denmark) deal with social responsibilities, which go beyond the achievement of economic efficiency. Cultural differences and institutional factors contribute to systematic differences in situational factors and management characteristics among countries. This in turn, will be reflected through the level and type of CSR disclosure, as the management respond to their relevant stakeholders (Van Der Laan Smith et al., 2010). Thus, CSR reporting in Norwegian/Danish companies is expected to be different from that in companies in the USA. This leads to the following hypotheses: H 2 : Companies governance structure influences the quality and quantity of CSR disclosure. H 2a : Companies governance structure interacts with culture in influencing the quality and quantity of CSR disclosure. CSR and Ownership Structure Mohd Ghazali (2007) examined the ownership structures of Malaysian firms to see whether these structures influence the CSR disclosure. She assessed whether ownership concentration, director ownership or government ownership, in addition to corporate characteristics of 87 non-financial companies, bear any relationship to a company s CSR disclosure. The regression results show that two ownership variables, director ownership and governmental ownership, significantly influenced the CSR disclosures of Malaysian firms. She found that owner-managed companies (i.e. companies in which the directors hold a higher proportion of equity shares) disclosed significantly less CSR information, while companies in which the government is a substantial shareholder disclosed significantly more CSR information in their annual reports. A similar finding was reported by Amran and Devi (2008), demonstrating that the government plays an important role in increasing the quantity of CSR information disclosed in the annual reports. Likewise, Eng and Mak (2003) examined the impact of ownership structure on voluntary disclosure. They argue that companies having a government link disclosed more voluntary information to mitigate the higher agency cost and weaker governance of these firms. An analysis carried out on 158 firms in Singapore showed a positive association between the disclosure and a government-linked company. Accordingly, CSR disclosure of government-owned companies should be high because the appointed directors of these companies need to align their decisions with the aspirations of the government and society (e.g. Amran and Devi, 2008; Naser et al., 2006; Donelly and Mulcahy, 2008). Furthermore, government-owned companies face more pressure from the society than do nongovernment owned companies. Thus, the government-owned companies might use the disclosure as part of their legitimization strategy. This viewpoint expects a positive relationship between government ownership and CSR reporting. However, it is too simplistic to assume that disclosure is used as a legitimization strategy in government institutions because the accounting system in itself is influenced by institutional factors. For instance, research evidence from China demonstrates that government or state-owned organizations face less pressure to voluntarily provide information in their annual reports since the information should be easily obtainable and accessible from other reporting means. Therefore, this viewpoint expects the relationship between government ownership and corporate social reporting to be negative (Huafang and Jiangou, 2007; Xiao et al., 2004). Thus, we argue that there is a simultaneous relationship between culture and ownership structure in the CSR reporting model. For example, in a culture where power and political parties play important roles in business settings, the government influence on CSR is pertinent (Smith et al., 2007). However, the same situation may not apply to other countries in which business settings are isolated from the political institutions. This point attests to the importance of cultural and institutional context in the CSR reporting practice. 8

9 Given the above, the relationship between government ownership and CSR reporting is investigated in this study. The following states the hypotheses: H 3 : Government ownership influences the quantity and quality of CSR reporting. H 3a : Culture interacts with government ownership to influence the quantity and quality of CSR reporting. 3. RESEARCH DESIGN Country We examined CSR reporting in China, India, Malaysia and the UK. In order to minimize the complexities involved in the institutional environments of many countries, these four countries were chosen for this study. We believe that CSR issues and their relationship to governance structures and culture are prominent in these countries for four reasons: firstly, China and India are amongst the largest emerging economies. However, in China, drastic economic growth has caused problems relating to air quality, land use, water and ecological conservation. This in turn, affects the neighbouring countries as well as the rest of the world (see Zhang and Wen, 2008; Wang et al., 2008; Liu et al., 2010). Secondly, China, India and Malaysia have experienced significant reforms in corporate governance structure. Thirdly, the UK is at a relatively advanced stage in the promulgations for relevant CSR standards and practices, comparing to the other three countries. Finally, these four countries represent diverse cultural settings. For example, high power distance exists in Malaysia; whereas Chinese society is the most collectivist in nature. Appendix 1 presents the national culture score as presented by Hofstede (2001) for the four countries. Sample The sample was obtained from Compustat Global and Mergent Online Database. A total of 203 large corporations operating in socially and environmentally sensitive industries were selected from the Compustat Global and Mergent Online database. For several reasons, we controlled for the size and industry effects in the sample selection process. Firstly, with regard to size, previous literature generally provides a consensus that large companies are generally good reporters (see Patten, 1992; Hackston and Milne, 1996; Kolk 2003; KPMG 2005; Owen, 2007). Secondly, large companies face greater political and public pressures than do small companies due to the resources and profits they generate. Thirdly, research provide empirical findings that, in general, industry affiliation motivates social and environmental disclosure; companies operating in environmentally and socially sensitive industries tend to provide more CSR disclosure than their counterparts reporting (see Halme and Huse, 1997; Milne and Patten, 2002; Yongvanich and Guthrie, 2005; Deegan and Blomquist, 2006; Guthrie et al. 2008; Belal, 2008). Thus, drawing upon legitimacy theory, the CSR issue should be more prominent in large companies operating in socially and environmentally sensitive industries. Specifically, the sample was chosen from eight industries: energy, oil and gas; materials; manufacturing; transportation; automobiles and components; alcohol, tobacco, casino and gambling; pharmaceutical, biotechnology and drugs, and utilities. Using data from Compustat Global and Mergent Online, we ranked all the companies in each country based on their market capitalization. Companies from the UK, India and Malaysia were selected from the top 100 lists. Chinese companies were chosen from the top 200 lists because some reports were not in English and could not be used. We obtained 203 companies from this sampling frame 5. Table 1 (Panel A) tabulates the sample companies. Overall, the major part of the sample comes from two industries: materials and manufacturing (47 per cent). Thirteen companies (6.4%) are operating in the socially sensitive 5 We compare the sample with data from GLOBAL 2000 list (the list ranks the world s top 2000 companies based on market capitalization, assets, profit and sales value). All seventy companies which were in GLOBAL 2000 were in our sample. This procedure confirms that we have captured all large corporations operating in the selected industries. 9

10 industries (alcohol, tobacco, casino and gambling); while the rest (93.6%) are in environmentally sensitive industries. << Insert Table 1>> << Insert Table 2>> CSR Disclosure Information on CSR reporting was obtained through content analysis of each of the 2008/2009 company annual reports, CSR stand-alone reports and corporate websites (see Table 2). Annual reports were obtained from the Mergent Online database and from corporate websites. The relevant sections of annual reports were printed and all related sections on corporate websites (including the CSR stand-alone report) were converted to pdf. files and saved 6. Content analyses were conducted on both soft copy and hard copy files. CSR information in annual reports, CSR stand-alone reports and corporate websites were assessed separately. A total of 403 observations were made including the annual reports, CSR sections on websites and stand-alone CSR reports. Table 1 (Panel B) shows that all companies provide CSR disclosure in annual reports. From 203 companies, sixty five per cent (132 companies) have CSR sections in their corporate websites. For every 10 companies that have corporate websites, 5 provide stand-alone CSR reports 7. However, the ratio could have been influenced by the number of CSR reports produced by UK companies, which is relatively high. By country, the UK is ranked first as all the companies have a CSR section in their corporate websites; 41 of them (82 per cent) have stand-alone CSR reports produced annually. India is ranked second, followed by Malaysia and China. China and Malaysia respectively produced 22 and 28 CSR sections in their corporate websites; and they produced the same number of CSR stand-alone reports for the year 2008/2009. CSR sentences were identified based on a checklist of 65 items, which had been modified from Global Reporting Initiatives (GRI) indicators. The GRI index was used in this study for several reasons. Firstly, the GRI provides an internationally recognized framework for CSR reporting (Frost et al., 2005), which is relevant in a study that examines CSR reporting practices at an international level. Secondly, using an internationally recognized framework to measure CSR disclosure can enable the replication of this study. Thirdly, the GRI is comprehensive; it covers all reporting aspects such as the social, environmental and economic performances. Fourthly, GRI is also considered as the latest and most innovative measure for CSR reporting 8. Finally, previous studies examining CSR issues such as environmental reporting (e.g. Hasseldine et al., 2005; Van Staden and Hooks, 2007), ethical and social reporting (e.g. Adams and Kuasirikun, 2000) and sustainability reporting (e.g. O Dwyer and Owen, 2005; Frost et al., 2005) utilized GRI as a framework to develop their disclosure indices. Specifically, the index has seven main categories: environment, society, labour practice, human rights, product relation, economics and profiles. CSR disclosure was measured using scores for both quantity and quality. CSR Quantity To compute CSR quantity, sentences that met the decision criteria were identified and added according to the categories. Table 3 presents the number of CSR sentences per annual report, 6 All websites sections were saved in July and August, The same CSR disclosure was not assessed twice. Thus, we discarded seven corporate websites that had been duplicated from the corporate annual reports. 8 The first version of GRI was issued in June 2000; and revised two years later (Frost et al., 2005). The latest version is available on line as G3 ( ) 10

11 corporate websites and CSR stand-alone report and their combinations. Overall, the disclosure was substantially about the environment (27%) and labour relations (24 %). Disclosure on society (17.9%) and CSR profiles of companies (17%) was considerably similar. The least amount of disclosure was on human rights issues (1.7%). The disclosure pattern in annual reports is significantly different from that on websites. In the annual reports, disclosure on CSR profiles, the environment, labour and social issues were relatively the same. In contrast, sentences on websites could be ranked, with the first relating to environmental matters, followed by labour relations and societal issues. Companies in China, Malaysia, and India disclosed a very small amount of information concerning human rights issues. << Insert Table 3>> << Insert Table 4>> Table 4 summarises the descriptive statistics of CSR disclosure quantity and quality by countries for annual report, the CSR section on corporate websites and CSR stand-alone reports and their combinations. In relation to CSR quantity, UK corporations provided the highest rate of CSR disclosure for all modes of reporting (mean of ), followed by India, Malaysia and China. This ranking is consistent with the rate of disclosure on corporate websites. Mean of QuantityWeb (217.46) is higher than mean of QuantityAR (85.07), indicating that there are more CSR sentences in corporate websites than in annual reports. The disclosure range in websites is also higher than that in annual reports, suggesting that the data is widely dispersed, there is a huge difference between a company that provided the least CSR disclosure and that with the most CSR disclosure. Interestingly, across the emerging market countries, Malaysia provided the highest amount of CSR in annual reports; whereas, China consistently provided the least CSR disclosure both in annual reports and corporate websites. Overall, the results suggest that the setting of CSR reporting is unique in each country. CSR Quality The content analysis was extended by scoring each CSR sentence in the report to determine quality. Each sentence was scored from 0 to 4, with 0 for no disclosure; 1 for general rhetorical statement or policy stated; 2 for specific endeavour, descriptive information of implementation and monitoring; 3 for quantitative statement and 4 for the use of targets in addition to publication of quantified results. Table 5 shows the examples of sentences and their scores. << Insert Table 5>> The score was computed separately for each mode of reporting. Thus, a company could have multiple scores in separate reports. In the final analysis, we computed a combined score, which took the highest score in each index. For example, if a company provides disclosure on policy in the annual report (say on recycling issue), the quality score for the annual report is 2. In its CSR report, the issue is extended and elaborated to include quantitative information and benchmarking against the past year s performance. Thus, the score for the CSR report would be 4. Eventually, the combined score for the index on the recycling issue is also 4 (the highest score for either of the annual report or the CSR report). Table 4 also presents the descriptive statistics for the 65-item CSR disclosure index 9. Overall, the quality scores for CSR disclosure are generally low (total mean of 26.54). The total mean for each country also reveals a significant difference in CSR quality across countries. In addition, the mean quality score for CSR stand-alone reports and websites (27.69) is higher than for annual reports (15.02). The table also presents mean quality using dichotomous scores (QualityAllDic), computed 9 The maximum score for each item in the index is 4. However, some items have less than four scales (see Hooks and van Staden (2011) and van Staden and Hooks (2007)). Thus, maximum score can be given for a company is 236. Results were converted to a percentage. 11

12 based on the binary score of '0' and '1'(e.g. Williams 1999; Hanifa and Cooke, 2005) 10. Although the mean of the dichotomous scores is higher than the discrete scores, the pattern of disclosure quality across country is consistent; the UK provided the highest quality of disclosure, followed by India, Malaysia and China. << Insert Table 6>> To enhance the richness of our data, we also present CSR quality using a matrix scale, by counting the number of times in which each category of information appeared in the analysis (see Beck et al., 2010). For example, if 13 out of 65 items of CSR disclosure have a score of 1, the matrix score for category 1 will be 20%. Table 6 presents the CSR disclosure using matrix scales by annual report, corporate websites and their combinations. Panel A shows that, on average, each company scored 64.82% for category 0, indicating that the company did not disclose 42 out of 65 items in the disclosure index. The highest score for the disclosed items in Annual Reports and Websites was 13.05%, at category 4 (i.e. disclosure with quantitative information in comparison to the last year s performance, benchmarks or standards). This means that overall, the companies either disclose CSR information at a very high quality, or provide no disclosure at all. However, in the annual reports, scores for category 1 was the highest; indicating that most companies disclosed CSR information in the annual report, but at the barest minimum requirement. Panel B shows some variation in the categories of information for each country. Companies in China and Malaysia consistently disclosed brief statements on CSR issues (i.e. category 2 ) in the annual reports and on corporate websites. Interestingly, companies in India disclosed CSR information of the best quality (i.e. category 4 ) in their corporate websites, but not in their annual reports. UK corporations consistently provided sentences, which were quantitative and comparable with some benchmarks or the past years performance in both annual reports and corporate websites. Reliability and Validity of the CSR Reporting Measurement We undertook three procedures to minimize the subjectivity involved in the measurement process. Firstly, to ensure consistency, content analysis was conducted based on clearly specified decision rules. In addition, only one coder (i.e. the first author) conducted the content analysis throughout the whole research period. Secondly, reliability tests were performed by having two additional coders, who repeated all the processes in the content analysis. Each coder analysed 20 companies, with correlation results of 0.8 and 0.9 respectively. Thirdly, to ensure accuracy, our disclosure scores were compared with the GRI grades awarded to the company being studied 11. The correlation tests show a significant positive relationship between our quality scores and the GRI grades (p=0.000), indicating that our instruments are perfectly matched with those given by the professional organizations. Finally, as the dichotomous scores are arguably less subjective than the weighted scores, correlation test between CSRQualityAll and CSRQualityDic was computed. A correlation of (p=0.000) was found, indicating a high level of accuracy. Independent Variables Independent variables in this study are corporate governance, government ownership and national culture. Although we acknowledge that corporate governance can be measured using various measurements (see Larcker et al., 2007), we only take two constructs: board composition and the existence of a CSR committee on the board. Government ownership is measured based on a dichotomous scale of 1 for companies with government shares of more than 50% and 0 for privately owned companies. These variables were obtained from annual corporate reports. Finally, national culture variables are measured based on individualism, power distance, uncertainty avoidance 10 Maximum score can be given for a company is 65. Results were presented in percentage. 11 From 203 companies, 33 companies (16.3%) used GRI as a framework to prepare CSR reports. 12

13 and masculinity index (Hofstede, 1991; Hofstede, 2001; see current studies that use these indices: Jansen et al. 2009; Waldman et al. 2006; Smith, 2006; Kim and Gray, 2009; Tang and Koveos, 2008). Control Variables This research incorporates five control variables in the regression analysis:csr assurance statement, Big-4 auditor, listing status, proportion overseas subsidiaries, and market capitalization (proxy for size). These variables are motivated from previous research (see Wang et al., 2008; Webb et al., 2008; Simnett et al., 2009; Cheung et al., 2010). All the variables, their measurements and source of information have been presented in Table 2. Table 7 presents descriptive results for all the independent and control variables. Of particular interest is the mean for Big 4 Auditor (0.744). This value shows that 74.4% of the annual reports in the sample were audited by the auditors from Big4 firms. In particular, all 50 companies in the UK have Big4 firms as their auditors. Variance for market capitalization is high, probably resulting from the Chinese firms which have been selected from the top 200 lists. Finally, the mean of government shares (0.167) includes the UK corporations, which are all privately owned companies. << Insert Table 7>> << Insert Table 8>> Table 8 presents correlations among the independent variables 12. Overall, there is no serious multicollinearity problem among the independent variables (no correlation of greater than 0.7, except for the national culture variables and government ownership). Initial tests reveal multicollinearity amongst the four national culture variables. As a result, only two cultural variables (i.e. individualism and masculinity) were included in the multivariate test. GovtShares and GovtDic were also correlated at Thus, one of the proxies was excluded in the multivariate analysis. In particular, the existence of CSR board committees, CSR assurance, board composition and proportion of overseas subsidiaries are associated significantly with the country and its national culture variables. These relationships indicate that choices of governance structure differ from one country to another. In other words, country differences play a role in governance structure choice. Likewise, variables such as; Big4 auditors, proportion of overseas subsidiaries and government ownership also differ between countries. The implications of this result are examined in the hypothesis testing section. 4. RESULTS Regression Results All the independent variables are regressed separately against CSR quality and quantity. There are six models in the regressions tests: CSR quality obtained from annual reports (QualityAR), CSR quality obtained from corporate websites and CSR stand-alone reports (QualityWeb), total CSR quality (QualityAll), CSR quantity obtained from annual reports (QuanAR), CSR quantity obtained from CSR stand-alone reports and corporate websites (QuantityWeb) and total CSR Quantity (QuantityAll). The results are summarized in Table 9. << Insert Table 9>> Regarding the quality of CSR disclosure, Table 9 (Panel A) demonstrates that overall, the model for CSR quality has a greater explanatory power than that of CSR quantity (adjusted R 2 of 67.8% and 57.0% respectively). The existence of CSR board committee, CSR assurance statement and listing status enhance both the quality and quantity of CSR disclosures in all modes of reporting. Individualism influences CSR reporting negatively; i.e. the more collectivist the society, the less the 12 Of interest, Table 8 also presents two variables that have been controlled for in the sample selection process (i.e. industry and country) 13