Fundamental characteristics, machine learning, and stock price crash risk

Abstract

We investigate the application of machine learning algorithms for predicting stock price crash risks by employing a set of firm-specific characteristics of the Chinese stock market. The results suggest that machine learning techniques are superior in capturing the nuances of stock price crash risk, particularly through profitability and value versus growth features. These techniques perform well within state-owned enterprises and during periods of low economic policy uncertainty, and predictive insights primarily originate from intra-industry dynamics. In addition, we offer corporate finance- and financial market-based interpretations of machine learning's predictability, as well as a comprehensive understanding of its key determinants.