Environmental and social (E&S) considerations have become integral to investment decisions in the past two decades, and more public firms are making E&S disclosures. Meanwhile, dozens of ESG reporting frameworks have emerged, and many jurisdictions have issued or are in the process of issuing mandatory E&S disclosure regulations. Regulators and standard setters have also started to cooperate and consolidate their efforts around E&S disclosures. In a new study, we provide the first large-sample evidence on the evolution of E&S disclosure.

Using over 210,000 annual reports between 2001 and 2020 from public firms in 30 countries, we apply a natural-language processing technique called word embedding to create an E&S dictionary. This machine learning technique helps overcome the challenges of extracting and quantifying E&S disclosure from rich annual report data. We first select seed words that fall into seven Environmental (E) and Social (S) subtopics from major voluntary ESG reporting frameworks. These subtopics include Climate Change, Natural Resources, Pollution & Waste, Ecosystem, Human Capital, Products & Customers, and Other Stakeholders. We then apply the word-embedding technique to identify words and phrases that have meanings similar to those of the seed words and expand them into a comprehensive E&S dictionary.

After generating and validating the E&S dictionary, we next develop empirical measures for both the quantity and quality of E&S disclosures. First, we measure the quantity of E&S disclosure in an annual report, using the word count of sentences containing E&S keywords. Second, we capture E&S disclosure quality using 1) the occurrences of boilerplate language in E&S disclosure, calculated as the percentage of E&S-related sentences containing generic boilerplate phrases; 2) the specificity of E&S disclosure, defined as how often E&S-related text refers to specific people, places, organizations, times, or numbers; 3) the stickiness of E&S disclosure for the same firm over time, measured by the degree of similarity in a firm’s E&S disclosure year-over-year; and 4) the occurrence of visual cues in E&S disclosure, calculated as the number of infographics within an E&S disclosure narrative.

We use these measures to test our main research question on the global evolution of E&S disclosure. After controlling for various firm- and country-level disclosure determinants, we document significant upward trends in the quantity of E&S disclosure and related infographics in annual reports. However, we also find that firms in recent years tend to provide E&S disclosures that contain more generic language and less specific information, which echoes practitioners’ concerns about the usefulness of E&S disclosure. We further notice that these trends do not appear to be industry specific. Overall, our findings imply that E&S disclosure quality is not improving as E&S reporting practices mature. In addition, we explore how the content of E&S disclosure in annual reports has evolved during our sample period and find that the amount of disclosure related to climate change and human capital has significantly increased relative to the other subtopics. Specifically, issues pertaining to energy efficiency, greenhouse gases, discrimination, and carbon emissions are more frequently discussed, while topics related to product quality and customer service are mentioned less often.

In our second research question, we explore whether voluntary reporting frameworks and mandatory disclosure regulations have helped change E&S disclosure. Six major ESG reporting frameworks, including the Global Reporting Initiative and the Climate Disclosure Standards Board, emerged during our sample period. We first document that the percentage of firms in our sample that have voluntarily adopted ESG reporting frameworks has soared from zero to more than 20 percent over the last two decades. We next investigate whether firms’ use of voluntary reporting frameworks is associated with any disclosure changes. We find that, after adopting voluntary reporting frameworks, firms provide more E&S disclosures in annual reports, and these disclosures use less generic language, contain more infographics, and exhibit higher similarity year-over-year.

As for mandatory disclosure, 23 of our 30 sample countries issued E&S disclosure mandates during our sample period, and 20 of them issued multiple E&S disclosure mandates. For each country, we identify the mandate with the earliest effective date (hereafter, the first E&S disclosure mandate), as well as the one with the most comprehensive E&S disclosure requirements (hereafter, the most comprehensive E&S disclosure mandate) in our sample period. We document that firms respond to the most comprehensive mandates by providing more E&S disclosure but do not respond similarly to the first mandates. Moreover, we find that E&S disclosure became more generic after both types of mandates took effect, and less specific after the most comprehensive mandates were in place. Together, our findings suggest that overall effects of E&S disclosure mandates seem to concentrate on the quantity rather than the quality of E&S reporting.

Our study makes several important contributions. First, to our knowledge, this is the first study to systematically examine the textual attributes of annual report E&S disclosures in a large sample of international firms over a period of 20 years. In addition to documenting significant trends in E&S disclosure quantity and quality, we also find that E&S disclosure quality improves after the adoption of voluntary ESG reporting frameworks but deteriorates after disclosure mandates, implying the potential value for regulators to collaborate with these established voluntary ESG standard setters. Second, our study provides academic researchers with a comprehensive dictionary and readily available measures to quantify E&S disclosure. Finally, our empirical analyses support evidence-informed policymaking related to ESG reporting. For example, many regulators across jurisdictions are still in the process of establishing (e.g., the SEC) or expanding (e.g., EU) E&S disclosure requirements. Our findings of increasing boilerplate language and decreasing specificity in E&S disclosure reaffirm the urgency of strengthening disclosure quality.

This post comes to us from professors Yan Lin at the University of Macau, Rui Shen at Chinese University of Hong Kong (Shenzhen), and Jasmine Wang and Julia Yu at the University of Virginia. It is based on their recent paper, “Global Evolution of Environmental and Social Disclosure in Annual Reports,” available here.