The thematic content of ESG reports: A topic modeling approach

This paper aims to show the impact of the thematic content of ESG reports on information asymmetry between managers and investors. Using a sentence-based topic modeling algorithm, we determine and empirically quantify 30 topics in a large collection of 1,667 ESG reports between 2007 and 2020. We find that the algorithm produces a semantically meaningful set of topics and that there exist clear patterns in the topics discussed in ESG reports, along with significant temporal shifts in managers’ focus on environmental, social or governance issues. We also show that the content of such reports helps explain investors’ behavior and reduces information asymmetry around and after the release date of the report. Furthermore, we evidence that there are substantial differences in the value relevance of the topics discussed in the reports and that investors react more to ESG reports that discuss environmental matters, while they tend to find the social and governance pillars less informative. Finally, there exists a concave relationship between the number of topics discussed in ESG reports and their informativeness. Overall, this paper theoretically contributes to prior literature by showing that investors are not only sensitive to whether ESG reports are disclosed, but also to what information is provided.