Research On Systemic Risk Spillovers And Early Warning Of Chinese Sectors

Since the 2008 global financial crisis,crises have occurred frequently in various countries,and China has also experienced turbulent times such as the domestic stock market turmoil,the trade friction between China and US,and the Covid-19.Systemic risks will not only undermine the stability of the financial system,but will also seriously hinder the normal development of the national economy and cause a massive loss of social wealth.In order to better prevent and respond to systemic risks,countries around the world have attempted to build a macroprudential supervision policy framework.China has also raised the prevention of systemic risks to the level of national security.As the primary task of the three critical battles,the practical significance of "Forestall and Defuse major risks" is self-evident.Therefore,precisely quantifying the magnituse of systemic risk spillovers,understanding the contagion mechanisms of systemic risks,identifing the main sources of systemic risks and effective forewarning are undoubtedly crucial for forestalling and defusing systemic risks.This paper expands the connotation of systemic risk from the financial system to the economic and financial system,and focuses on the systemic risk spillover effect and early warning among China’s real economy industries.Based on literature and theoretical analysis,taking the Wind primary industry index as the research object,this paper uses the tail event-driven method to construct inter-industrial risk spillover network in China in the years of 2002—2021,measures systemic risk spillover,and identifies the systemically important sector and risk transmission mechanism.Next,this paper uses the Gated Graph Neural Network(GGNN)and Graph Isomorphism Network(GIN)early warning model for risk warning,the input of the early warning model is the risk spillover network at each time point,and the output is the height representation vector of the risk spillover network at each time point.The empirical results suggest that firstly,the level of inter-industrial systemic risk spillover in China has the characteristics of conforming to the evolution trend of crisis,which will increase significantly when the market is overheated or a crisis occurs.The risk out-flow level in various industries is heterogeneous,and the risk in-flow level keeps stable state;Secondly,the risk transmission structure is found to be time-varying,the sectors,such as information technology,industry,materials,discretionary consumption and real estate,are systemically important and the core nodes of the network;Thirdly,in recent years,the level of risk spillover between the real industry and the financial industry has increased,and the financial industry has mostly played the role of systemic risk receiver.There has always been a strong two-way risk spillover in the financial and real estate industries;Fourth,it is feasible to carry out systemic risk early warning based on the risk spillover network.After the risk spillover indicator is added to the early warning index system,the prediction accuracy of logistic regression and support vector machine has been significantly improved.After a comprehensive comparison of the prediction performance of the domestic stock market turmoil data in 2015,it is found that GGNN and GIN early warning model have better forewarning performance,the prediction accuracy is over 95%,and it can effectively forewarn before the crisis occurs.