Dynamic Modeling And Empirical Study Of Risk Network From Quantile Perspective

With the development of globalization and the progress of science and technology,economies have formed an increasingly close network through frequent trading activities and international capital flows.The outbreak of the subprime mortgage crisis in the United States in 2008 not only caused a huge shock in the world financial market but also subsequently induced the depression of the real economy.This has made researchers and policymakers realize that the subsequent spread of systemic risks may lead to more serious consequences than itself.No economy can remain unscathed,and the outbreak of any crisis may spread to every corner of the world.How to prevent such systemic risks has also received a lot of attention and discussion.The Basel Ⅲ Accord put forward the macro-prudential supervision framework which takes the financial system as a whole and sets higher capital adequacy requirements for the important nodes of the financial network,the global systemically important banks(G-SIBs).In this context,to better identify and test systemic risks and quantify their impact on the real economy,this paper proposes a quantile-based time-varying risk measure-time-varying quantile correlation(TV-QCOR)and a network analysis framework.This paper proposes a new kind of time-varying risk measure,the time-varying quantile correlation coefficient.Its local polynomial estimation method is given.The new measure can be used to gauge the dynamic dependence between two financial series at a specific quantile level.Based on this measure,we propose a method to test the contagion effect of a financial crisis.Then we give the relationship between TV-QCOR and Copula function under the assumption of the standard normal marginal distribution.The simulation results show that TV-QCOR can effectively describe a variety of dependent structures.In the empirical study,we find that TV-QCOR detects different degrees of risk spillovers between the United States and other countries during the subprime crisis.Furthermore,the paper proposes the time-varying quantile partial correlation coefficient(TV-QPCOR).Since TV-QCOR describes the dependence between a pair of variables,and nodes in the network often cross-interact with each other,this is the socalled "pseudo correlation" problem.To avoid this,we use the concept of partial correlation coefficient to make TV-QPCOR describe the dependence of two nodes while controlling the interference of other nodes.On this basis,we introduce the risk network connected by TV-QPCOR.Through the study of the topological properties of quantile risk networks,we find that the structure of the network is closely related to the quantile level.Based on the observation of quantile network,we introduce the network structure into the quantile regression model and study the influence of the heterogeneity of network and exogenous variables on different quantiles.We focus on two typical scenarios.The first one is the network structural change of listed companies and their profitability in the context of the COVID-19 pandemic.We found that the network structure of listed companies changed a lot after the pandemic.The second example studies the risk spillover effect of China’s listed banks in the credit default swap spread market.We found the risk spillover effect between Chinese banks at different quantile levels and the empirical evidence of the "too big to fail" phenomenon in Chinese financial institutions,which deepened our understanding of China’s inter-bank risk structure and credit default swap market.