Automatic linear causal relationship identification for financial factor modeling

摘要

Given a comprehensive set of financial factors, we use linear non-Gaussian SEM to automatically identify the causal relationships buried in the factor set. The causal structure is allowed to have cyclic edges, explicitly accommodating ‘mutual causality’ which is well acknowledged but rarely modeled in standard economic theory. The method takes advantage of both artificial intelligence and economic related techniques, and identifies one stable model from several distribution-equivalent equilibrium models for each dataset. Empirical studies on 15 financial factors reveal some interesting findings, especially for the risk-return relationship modeling and capital structure determinants discovery.