Employee treatment and firm performance: evidence from topic modelling in lawsuit announcements
Document Type
Article
Publication Title
Review of Quantitative Finance and Accounting
Publication Date
6-24-2023
Abstract/ Summary
This study uses a machine learning technique to assess whether employee lawsuits are informative in predicting firm performance. Using a Bayesian topic modelling algorithm, we examine a large collection of employee lawsuit announcements between 2000 and 2016. We extract the key topics of lawsuit announcements via a text mining approach of latent Dirichlet allocation. We document that the algorithm generates semantically accurate topics that are predictive of future firm value. We find that announcements of lawsuits with topics of discrimination or injury and death have a considerable effect on firms’ future profitability. In addition, our results suggest that settlement amounts, attorney fees, and other legal expenditures stated in announcements are negatively associated with firm performance. Our results are robust to alternative topic definitions and data specifications.