Quantifying Cross-Border Risk Shocks to China’s New Energy Industry: Evidence from Frequency-Domain Networks and Interpretable Machine Learning

Abstract

This paper constructs a transnational industry-level risk network linking China’s new energy value chain with major developed-market industries to quantify cross-border risk shocks and examine their associated macroeconomic conditions. The results show that developed-market technology- and resource-related industries, including semiconductors, software, hardware, raw materials, and energy, are associated with substantial positive risk shocks to China’s new energy sector, whereas the automotive industry acts as a net risk receiver. Frequency-domain decomposition indicates that high-frequency spillovers dominate overall risk connectedness. Using interpretable machine learning, we find that international macroeconomic variables consistently possess stronger predictive power than domestic variables, accounting for approximately 75–85% of total feature importance. Furthermore, accumulated local effects analysis uncovers pronounced nonlinear threshold effects: (1) U.S. bond yields exhibit a U-shaped association with risk spillovers, turning sharply positive above approximately 2.8%; (2) VIX effects remain dormant below 30, after which predicted spillovers rise steadily; (3) global supply chain pressure activates an immediate positive marginal effect beyond zero; (4) trade policy uncertainty exerts a step-function impact even at low values, while economic policy uncertainty follows an S-shaped accumulation pattern. These findings offer key insights for strengthening China’s new energy supply chain resilience and risk management.