Recently, the paper "Fundamental Characteristics, Machine Learning, and Stock Price Crash Risk", co-authored by Prof. Fuwei Jiang, Associate Prof. Feifei Zhu, Associate Prof. Ma Tian (School of Economics, Central University of Nationalities, and PhD graduates of our college), has been accepted by the Journal of Financial Markets, a leading international journal in the field of finance. Journal of Financial Markets is an international authoritative journal in the field of finance, and is a 3-star journal in the ABS list and an A* journal in the ABDC list (Leading Journal in Field).
With the economic slowdown in recent years coupled with the impact of the global epidemic, the share prices of listed companies are facing greater downside risks. Based on the share price data of listed companies in China's A-share market, this paper utilizes machine learning algorithms to predict and analyze the risk of share price collapse by combining the fundamental characteristics of listed companies. The empirical results show that the model has a good risk prediction ability on the full sample by capturing the changes in the profitability class and value growth class indicators, with better performance in state-owned enterprises and in periods of low economic policy uncertainty. In addition, this study provides an economic explanation of the predictive nature of machine learning based on corporate finance and financial market perspectives.
The main contributions of this study are as follows. First, this paper pioneers a high-dimensional perspective-based examination of the predictability of stock price crash risk and its economic mechanisms. Second, this paper further refines the theory of crash risk formation mechanism based on corporate finance and market mechanism. While traditional theories suggest that the causes of crash risk are mainly focused on the role of mechanisms such as management agency issues and market structure, this project explores the applicability of the two types of theories in the Chinese market while using machine learning to sum up and analyze the features representing the different theories, and tries to merge the two types of theoretical models into a discussion. Finally, this paper complements related studies on machine learning interpretability, such as the disentanglement analysis for industry systematic risk and individual stock-specific risk, and the predictive analysis of models based on crash risk causation theories, a process that helps to ensure the robustness of the models in subsequent complex environments.
Written by Feifei Zhu
Reviewed by Yu Chao Peng
