https://orcid.org/0000-0002-8380-4420
![Sample our Economics, Finance,Business & Industry journals, sign in here to start your access, latest two full volumes FREE to you for 14 days]({"type":"image" , "src" : "/sda/5931/TOC-Subject-Sample-2021-Economics-Finance-Business-Industry.jpg"})
Abstract
US companies are increasingly responding to demand from investors and other stakeholders for transparent information about companies’ environmental, social, and governance (ESG) performance by issuing ESG reports on a voluntary basis. We examine whether these reports help to resolve the previously documented disagreement among ESG rating agencies about individual companies’ ESG performance. Consistent with this possibility, we find that disagreement among ESG rating agencies is lower for firms that voluntarily issue ESG reports. In particular, disclosures about the environmental and social dimensions help reduce disagreement about the company’s performance on those dimensions. Using textual analysis, we find that longer reports are associated with reduced disagreement among ESG raters while reports with more positive tones or that use a greater number of sticky words are associated with heightened disagreement. The association between ESG disclosure and ESG disagreement is more pronounced when firms obtain third-party attestations on their ESG reports, especially from accounting firms, and when firms adhere to advanced levels of Global Reporting Initiative (GRI) reporting standards. Finally, ESG disagreement is positively associated with disagreement and uncertainty in the capital market, providing strong motivation for firms to voluntarily disclose ESG reports to reduce ESG disagreement.
Acknowledgments
We thank Luc Paugam (the editor), three anonymous reviewers, Valentina Ge (research assistant), Zheng Wang (discussant) at the 2019 AAA Conference, and workshop participants at the University of Maryland-College Park, Rutgers University, and South Dakota State University for their helpful comments and suggestions. We are also grateful for the financial supports from the University of Maryland and Creighton University. All errors and omissions are our own.
Disclosure statement
No potential conflict of interest was reported by the author(s).
Data Availability
Data are available from the public sources cited in the text.
Supplemental Data and Research Materials
Supplemental data for this article can be accessed online at https://doi.org/10.1080/09638180.2022.2088588.
Supplemental Online Appendix for this article can be accessed at https://tinyurl.com/earkwwz, and includes the following parts:
Appendix 1 Specific Linguistic Attributes of ESG Disclosure and the Disagreement about Firms’ ESG Performance
Appendix 2 ESG Disagreement and Commensurability
Appendix 3 ESG Keywords from Bloomberg
Notes
1 While ESG disclosure is mandatory in Europe, it is a type of voluntary corporate non-financial disclosure in the US (e.g., Bagnoli & Watts, Citation2017; Ballou et al., Citation2018; Clarkson et al., Citation2020; Dhaliwal et al., Citation2011). For example, in 2019 only 39% of the companies in the lower half by market cap of Russell 1000 Index issued ESG reports (G&A, Citation2020).
2 The 10th edition of the Concise Oxford English Dictionary defines greenwashing as ‘disinformation disseminated by an organization so as to present an environmentally responsible public image’. See https://www.lexico.com/en/definition/greenwash.
3 A secondary ESG source is an assessment transformed by a process of scoring or a formula from a primary data source by a third party (CFA Institute, Citation2021).
4 We use primary ESG disclosure source and actual ESG reports interchangeably in the paper.
5 The correlation between Bloomberg ESG score and our disclosure indicator is 0.44 (p < 0.01), and the correlation between Bloomberg ESG score and our disclosure quality measure, which is created using textual analysis, is 0.33 (p < 0.01). These relatively low correlations indicate that the two disclosure measures capture different constructs. Since our measure directly comes from the firms’ ESG reports, the differences in the measures are likely due to the Bloomberg ESG score including information beyond the information directly provided by management.
6 About 30% of the observations in D. M. Christensen et al. (Citation2022)’s sample pertain to US firms. By contrast, 100% of the firms in our sample are from the US. Even though we do not have non-US firms that are part of their sample, we replaced our disclosure indicator with the Bloomberg ESG score and attempted to replicate their findings for our sample of US firms. We were unable to document a positive relationship between the Bloomberg ESG score and ESG disagreement for US firms. To shed light on this finding, we used the governance component of the Bloomberg ESG score and examined its association with disagreement among ESG raters about firms’ governance performance. Consistent with their overall results, we find a significantly positive association between the Bloomberg Governance score and Governance disagreement. This result is consistent with the Bloomberg ESG measure likely including mandatory disclosure that results in higher disagreement among rating agencies.
7 The differences between mandatory and voluntary ESG disclosure are important as many countries including the U.S. are still weighing the cost and benefit of mandating ESG disclosure. Mandating ESG disclosure has potential nontrivial costs such as enforcement challenges, compliance challenges, the divergent goals and interests in standard setting, the avoidance strategy of firms using boilerplate language, and the difficulty in trying to harmonize reporting practices of managers with heterogeneous reporting incentives across firms, industries and countries (H. B. Christensen et al., Citation2021). Some of these costs could be part of the sources for ESG disagreements. As the sample consists of close to 70% non-US firms, the setting in D. M. Christensen et al. (Citation2022) includes countries in both mandatory and voluntary regimes. It would be difficult to disentangle whether the results are primarily driven by voluntary or mandatory countries. Our paper uses the US sample only and therefore provides a unique setting where the benefit of an ESG disclosure in resolving ESG disagreement is clearer in the voluntary disclosure regime.
8 Furthermore, our finding of the negative association between the content analysis of environmental and social disclosures and the disagreement on the environmental and social dimensions of ESG performance (but not for the governance dimension) suggests that, in a mandatory setting (i.e., most governance disclosure is mandatory in the US), the benefit of ESG disclosure in resolving ESG disagreement is not detected whereas such benefit is manifested in a voluntary setting (i.e., most environmental and social disclosures are voluntary in the US). The use of actual ESG reports allows us to fully explore this relationship in a voluntary setting.
9 An important difference between the two studies is their paper focuses on ESG news’ market reactions being moderated by ESG disagreement, while we focus on ESG disagreement's market impact on factors directly related to cost of capital.
10 ASSET4 and KLD use ESG reports as a major information source when rating firms’ ESG performance (Diebecker et al., Citation2019).
11 Prior studies show that ESG raters utilize ESG reports to evaluate and rank firms’ ESG performance and that their judgements and assessments depend on the quality of ESG reports (Scalet & Kelly, Citation2010; Society & Organizations, Citation2018). Specifically, ‘[t]he quality of an ESG report decisively influences a firm’s ESG rating: In their evaluations, ESG rating agencies consider whether these reports are sufficiently comprehensive, balanced, and quantified’ (Society & Organizations, Citation2018, p. 2).
12 Prior studies documented the importance of linguistic characteristics for the information content of various disclosure channels, such as analyst reports (e.g., Huang et al., Citation2018), annual reports (e.g., Li, Citation2008), and conference calls (e.g., Bushee et al., Citation2018).
13 % of words that include at least one 8-word phrase that is identical to a phrase used in the prior year’s report.
14 Many US firms still do not obtain external assurance for their ESG reports. For example, around 23% of US firms in 2013, up from 13% in 2011, received some types of external assurance for ESG disclosure, far below international levels (Bagnoli & Watts, Citation2017, p. 205). In 2019, only 24% of the Russell 1000 ESG reporters ‘seek external assurance for their non-financial ESG disclosures’ (G&A, Citation2020).
15 ASSET4 directly provides an overall score in its database; however, this score includes the economic dimension, which is not a component of the KLD overall score. Therefore, we calculate the ASSET4 overall score as the mean based on the environmental, social, and governance dimensions to maintain comparability with the KLD ESG categories. Nevertheless, the results are robust to using the overall score provided by ASSET4.
16 Social dimension is composed of the five social pillars (MSCI, Citation2017): (1) diversity, (2) employee relations, (3) community, (4) human rights, and (5) product quality. We therefore use the environmental and corporate governance dimensions and construct the social dimension by aggregating the five social pillars from the KLD database.
17 Several studies used Vigeo Eiris as a source of ESG ratings (Ferrell et al., Citation2016; Liang & Renneboog, Citation2017; Meier et al., Citation2021).
18 We only use the overlapped dimensions (i.e., E, S, and G) of ESG rating between the two (KLD and ASSET4) and among the three ESG rating agencies including, Vigeo Eiris. In particular, Vigeo has clear categorical ratings for the Environmental, Social, Governance dimensions, making the combination of the overlapped dimensions much easier.
19 Because linguistic features of ESG reports can only be measured when firms actually issue ESG reports, we can conduct H2 only on the set of firms for which ESG reports are available (i.e., the identified ESG disclosers).
20 Besides the composite disclosure score, TEXTUAL_QUALITY, we also separately examine the underlying textual components (i.e., length, net tone, and stickiness) in our Online Appendix and the results are consistent.
21 Because firms can only obtain assurance on ESG reports that they actually issue, we can conduct H3a only on the set of firms for which ESG reports are available (i.e., the identified ESG disclosers).
22 Assurance-related keywords are listed as follows, ‘third party assurance’, ‘CSR report assurance’, ‘CSR report verification’, ‘assurance statement’, ‘independent assurance’, ‘assurance standard’, ‘assurance scope’, ‘independent assurance’, ‘independent audit’, ‘third-party assurance’, ‘corporate social responsibility report assurance’, ‘corporate social responsibility report verification’, ‘sustainability report verification’, and ‘sustainability report assurance’.
23 KPMG, Ernst & Young, Deloitte & Touche, PricewaterhouseCoopers, Arthur Andersen, Moss Adams, and Grant Thornton.
24 Following prior literature (Dhaliwal et al., Citation2011; Clarkson et al., Citation2020), we focus on standalone ESG reports since they are broader in scope and the linguistic attributes of ESG information tend not be affected by other non-ESG content. Before conducting the textual analysis, we convert ESG reports (PDF files) to text and applied standard pre-processing methods (e.g., eliminate ‘stop words’ and divide the text into single tokens) to reduce measurement error (e.g., Li, Citation2008).
25 The sample starts from 2007 and ends in 2016 because of the availability of CSRHub data and ASSET4 data, respectively.
26 More than half (53%) of the reporting companies in the Russell 1000 Index did not utilize the GRI standard (G&A, Citation2020). Consistent with this view, we only find 12% of firms in our sample that adhere to the GRI standard.
27 Please see Online Appendix for the listed keywords of Environmental, Social, and Corporate Governance from Bloomberg Database.
28 The results still hold using the alternative proxy, ESG_DISPERSION2, as the dependent variable (untabulated).
29 Specifically, Dhaliwal et al. (Citation2011) focus on the fact that voluntary ESG reports are repositories of non-financial measures that can be inputs into valuation models to explain why ESG disclosure is associated with lower cost of capital. Our finding that ESG disagreement is associated with indicators of capital market uncertainty that are linked to cost of capital suggests that reducing ESG disagreement provides a direct channel for voluntary ESG reports to ultimately reduce cost of capital.
30 The results for matched sample analysis still hold using the alternative proxy for disagreements on ESG performance, ESG_DISPERSION2, as the dependent variable (untabulated).
31 We create subsamples with sufficient firm-level variations and further control for the firm fixed effects in tests for all our three hypotheses. While our results are robust to firm fixed effects, we argue that disclosure practices are sticky which make it collinear with firm fixed effects, especially given that we have a relatively short sample period (2007–2016).