The Determinants of Limited Household Participation in Risky Financial Markets: Evidence from China Using Explainable Machine Learning

This study takes the limited household participation in risky financial markets as its point of departure. Drawing on microdata from the 2019 China Household Finance Survey (CHFS), we construct a multidimensional analytical framework using machine learning methods. The results indicate that this limitation arises from the interplay of multiple dimensions, with significant nonlinear relationships observed between these factors and household investment behavior. Insufficient development of key driving factors constitutes the main barrier to participation in risky financial markets. Feature interaction analysis reveals a “reversal effect” in how urban–rural disparities, economic attention, income level, and social engagement shape participation behavior. Educational attainment and financial literacy act as “threshold conditions” that enable economic attention to translate into actual investment decisions. The heterogeneity analysis further shows that households at different life-cycle stages as well as across urban–rural settings exhibit distinct participation patterns. These findings provide data-driven insights that can inform policies to promote financial inclusion, enhance investor education, and strengthen household risk management practices.