A Machine Learning Based Anatomy of Firm-Level Climate Risk Exposure

This applied research paper was written by Kai Li (Peking University HSBC Business School) and Tingyu Yu (The Hong Kong University of Science and Technology, Southern University of Science and Technology). We construct various measures of firm-level climate risk exposure by utilising two natural language processing techniques (LDA and word2vec) on firms’ quarterly earnings conference call transcripts. The unsupervised learning method automatically generates five topics, all aligned with popular concerns about climate change. We then conduct an empirical analysis on the topic that puts high weight on words about natural disasters. This disaster exposure measure has a significant negative association with firms’ sales growth and profitability. Moreover, firms with higher disaster exposure tend to earn higher expected stock returns than those counterparts with lower exposure, suggesting that firms’ disaster risk exposure significantly affects the cost of equity and market valuations. A long-short portfolio based on this exposure measure generates a positive return of 5% per annum, which cannot be explained by common risk factors and other firm characteristics