Measuring systemic risk in China: A new hybrid approach incorporating ensemble learning and risk spillover networks

To address the limitations of traditional systemic risk indices in measuring nonlinearity and network interdependence, we introduce ESRISK, a novel systemic risk measure that incorporates ensemble learning and risk spillover networks. Our approach can effectively analyze the complex nonlinearity in high-dimensional data, enabling more accurate quantification of systemic risk in China's financial system. Comprehensive evaluations reveal that ESRISK outperforms prevailing systemic risk measures, particularly in predictability, accuracy in measuring systemic risk, and effectiveness in early warning detection of systemic events. Moreover, ESRISK demonstrates superior predictive power for macroeconomic downturns. Our findings highlight the importance of applying machine learning methods and considering inter-institutional spillovers when measuring systemic risk in China's financial ecosystem.