Corporate social responsibility disclosure: a topic-based approach

Abstract

In this study, we investigate the potential differences in topic-specific corporate social responsibility (CSR) disclosure between companies located in liberal market economies (LMEs) and coordinated market economies (CMEs). We also examine the potential convergence of the reporting practices that characterise these two economies over time. We analyse a sample of 5,939 CSR reports issued by European and U.S. firms over 2008–2019. We use textual analysis to examine how explicitly such reports address specific CSR topics. Following Matten and Moon (2008), we focus on three thematic areas: ‘human resources’, ‘environmental protection’, and ‘society at large’. Each area comprises three distinct topics. Our results show that companies operating in LMEs report more explicitly on these thematic areas, with one exception: those operating in CMEs report more explicitly on parental policies. Additionally, the reporting practices of companies operating in these two types of economies converge for most of the topics under study. For the disclosure of parental leave policies, biodiversity, and waste, no distinct trend is observable.

Keywords:

• CSR disclosure
• explicit CSR
• textual analysis
• topic-based analysis

Acknowledgements

We appreciate the thoughtful and relevant comments from Daniel Hsiao (discussant at the 2017 American Accounting Association Annual Meeting). We also wish to thank the participants of the 2017 American Accounting Association Annual Meeting in San Diego, the 40th European Accounting Association Annual Congress in Valencia, the 4th French Conference on Social and Environmental Accounting Research in Toulouse and research workshops and seminars at the University of Bath, the University of Padova, the University of Zurich, the University of Regensburg and WU Vienna. We also thank Kevin Dix, Martin Saurer, Dominik Scherrer and Sandra Taubländer for their excellent research assistance. Charles Cho acknowledges the financial support provided by the Erivan K. Haub Chair in Business & Sustainability at the Schulich School of Business and the Global Research Network program through the Ministry of Education of the Republic of Korea and the National Research Foundation of Korea (NRF-2016S1A2A2912421).

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 Masculinity is based on Hofstede’s culture dimensions and it denotes a ‘preference in society for achievement, heroism, assertiveness and material rewards for success’ (Hofstede, 2016).

2 As indicated earlier, there is generally a disconnect between CSR performance and CSR disclosure, particularly in the case of U.S. (LME) firms (Neu et al., 1998; Patten, 2002; Cho et al., 2010, 2015), as well as significant lobbying activities from these firms to avoid social and environmental responsibilities by greenwashing their performance through selective reporting (Cho et al., 2018).

3 In 2000, the GRI launched its first version of the reporting guidelines. The current version, the GRI Standards, was launched in 2016.

4 Only a small proportion of the companies listed in the STOXX North America 600 Index are headquartered in Canada.

5 We collect the reports either manually or from www.corporateregister.com.

6 We exclude integrated reports because they differ significantly from standalone reports. In addition, and in any case, while CSR reports, sustainability reports, environmental reports, and social reports have seemingly different names, they are similar in nature and content. If a firm provides more than one separate CSR report, we include the most comprehensive report.

7 We base our categorisation on Matten and Moon (2008). In our robustness section we provide an alternative categorisation.

8 Our sample selection focuses on firms that provide standalone CSR reports, whereas the CSR reporting variable in Stolowy and Paugam (2018) includes both separate CSR reports and CSR sections in annual reports.

9 Regarding employees and parental policies, there are substantial differences in employee protection rights as well as the duration and amount of maternity leave benefits between LMEs and CMEs. The most striking example is the United States, which has almost no legislation to grant paid maternity leave and protect working mothers after giving birth. With respect to diversity, a number of European member states have defined long-term gender quotas regarding their management and supervisory bodies (e.g. Norway, France, Italy, Finland, the Netherlands, and Germany).

10 Regarding climate change, LMEs tend to delegate responsibility for greenhouse gas emission reductions to private firms and rely on flexible trading schemes, whereas CMEs tend to set specific targets. For instance, in March 2020 the European Commission published a proposal for a European Climate Law that mandates member states to achieve net-zero greenhouse gas emissions by 2050. Furthermore, climate change mitigation and adaptation as well as the protection and restoration of biodiversity are among the six environmental objectives outlined in the EU’s Taxonomy Regulation. Concerning waste, the EU has a long tradition of strong regulation lasting from early regulations on recycling (e.g., REACH, WEEE, and RoHS) to recent regulation such as the Waste Framework Directive (2015/851). The transition towards a circular economy and thus the prevention of waste is covered by the environmental objectives of the EU’s Taxonomy Regulation.

11 Regarding education, in LMEs there is a strong distinction between public and private schools and universities; the latter depend on private funding such as scholarships and donations. Public funding is less important for these schools. Such an educational environment creates leeway for private (firm) engagement in terms of both direct donations of resources and student scholarships. CMEs, on the contrary, lack the tradition of private schools and university, as education is typically viewed as a public good offered by the state at moderate prices. Therefore, (private) firm engagement in education is less common in CMEs. Regarding philanthropy and community, owing to the relatively high corporate tax rates in CMEs and general reservations towards private engagement, European companies are traditionally less engaged in philanthropic activities and therefore less explicit on this topic in their voluntary CSR disclosure.

12 A convergence of CSR reporting can occur, even if there is an explicitisation in CSR reporting among firms located in both economics. In this case, a negative β₃ indicates that the explicitisation of CSR disclosure is less strong for firms located in LMEs than for firms located in CMEs.

13 The classification of industries into environmentally sensitive categories is described in greater detail in Appendix I.

14 The words of the phrases ‘parental leave’, ‘maternity leave’, ‘paternity leave’, ‘charitable giving’, ‘climate change’, ‘global warming’, ‘Kyoto Protocol’, ‘carbon emission’, and ‘greenhouse gas’ are required to exist exactly side by side (separated only by stop words, if at all).

15 ‘Loading’ means that the respective keyword appears in a report at least once.

16 In the robustness section of the paper, we obtain the same results for the subsample of GRI reporters (gri = 1) and GRI non-reporters (gri = 0).

17 Our sample includes 14 reports with a total word count of only one word. These words are PDF reports that comprise image files for which the text is only partly extractable.

18 We present the results for the ordinary least squares regressions with t-statistics based on Huber–White robust standard errors. The results remain substantially unchanged if we use standard errors clustered at the firm level or industry level.

19 The results remain unchanged with respect to the direction and significance of our main findings when we also winsorise our dependent variables at the top and bottom 1%.

20 When splitting the esi variable into indicator variables for the industries considered, the results indicate that this overall pattern is found for all the esi industries except for the industry ‘metals’, which is negatively related to all the disclosure variables.

21 We also apply alternative measures for gri in our baseline regression models, namely, GRI_BB and GRI_Textual. GRI_BB equals one if the firm reports in accordance with the GRI Sustainability Reporting Guidelines/Standards and zero otherwise and is obtained from the Bloomberg database. We also apply textual analysis and scan all the sample reports for the occurrence of ‘GRI’ or ‘Global Reporting Initiative’; based on these results, we derive GRI_Textual that equals one if a firm’s report mentions the terms ‘GRI’ or ‘Global Reporting Initiative’ and zero otherwise. These three variables (gri, GRI_BB, and GRI_Textual) are highly correlated with one another (Pearson correlation coefficients equal 0.8 or higher). When rerunning our regression analyses with these alternative proxies for gri, our main findings remain unchanged (untabulated).

22 There is only one exception: post is positive and significant for discl_parental when we move the cut-off year to 2012.

23 These measures are based on the entire CSR disclosure rather than on specific topics.