Regulating risk culture in the insurance industry using machine learning

Abstract

This study examines the relationship between risk culture and regulation in the insurance industry using textual analysis and machine learning. By analyzing 10-K disclosures, we classify firms into distinct risk culture clusters and find that the risk culture of insurance firms is significantly shaped by their uncertain risk strategies, constraints in defining, implementing, and reporting risks, as well as litigious decisions and risk management practices. A temporal prediction analysis indicates that large insurers maintaining a poor risk culture trend are less likely to reverse it compared to those improving. Moreover, insurance firms show enhanced risk culture post-Dodd–Frank Act. Our findings underscore the potential benefits of regulations aimed at monitoring and overseeing insurers' risk practices.