AI-Generated Corporate Environmental Data: An Event Study with Predictive Power

Abstract

Literature has numerous debates about the relation between emerging financial environmental, social, and governance (ESG) factors and financial performance with mixed results. The authors use a unique data set generated by big data analytics (from web-based data mining) for three environmental areas (water, land, and air) to test hypothesis in the extreme events (defined as those that are over/under ±2.58 multiplied by the standard deviation) have a high chance of predicting equity price movements within an window of −3/+10 days, respectively, prior to and after the event. The authors repeat the similar robustness study for a sample of 2018 and the results still holds. The authors interpret these findings to suggest that: (1) studies using continuously AI-generated data for ESG categories can have significant predictive power for extreme events; and (2) that such high correlations can be used to confirm the materiality of some ESG data. The authors conclude with noting limitation of this initial study, and present specific areas for future research.