Systematic Risk Measurement Based On CoVaR And Network Approach And Its Application

The long-term destructive impact of the 2008 financial crisis on the global economy has made people realize that the micro-prudential supervision focusing on the risk of a single institution can no longer prevent the occurrence of crises,and can no longer meet the requirements of supervision.It is necessary to resort to macro-prudential supervision and strengthen the management of systemic risks.After the crisis,many countries have taken effective measures to strengthen the supervision of systemic risks.Meanwhile,the relevant financial regulatory authorities in our country have also repeatedly emphasized we must prevent and defuse major risks,and firmly hold that systemic risk will not occur.At present,the development environment at home and abroad is facing many uncertain factors,so strengthening the research on systemic risk is of great significance to maintaining the stable development of our country’s economy.The salient feature of the modern financial system is the close connections between institutions,which not only diversify risks and improve financial stability,but also become channels for risk contagion between institutions,and may even lead to systemic risks.As a tool to describe complex systems,the network can effectively reflect the connections between institutions and the process of risk contagion.CoVaR,a risk metric,can effectively measures the tail dependence of institutions under extreme events.This paper proposes to combine the network method and CoVaR to study systemic risk.On the one hand,the network method can be used to consider the risk contagion between institutions in extreme events,and on the other hand,the network between institutions and the structural characteristics of the network can be introduced into CoVaR to improve the accuracy of the risk level of an institution.The main research contents of this paper are as follows:The first part extends the definitions of binary CoVaR and ΔCoVaR to the multivariate case,and proposes to establish a tail risk network by ΔCoVaR.ΔCoVaR not only reflects tail dependence between institutions,but also measures the incremental change of risk from normal state to extreme state.Thence building a network usingΔCoVaR can reflects the process of risk contagion between institutions.The explicit expression of ΔCoVaR is derived under the multivariate t distribution,and the timevarying ΔCoVaR is estimated under the DCC-GARCH model.A time-varying tail risk network is established based on ΔCoVaR,which avoids estimation error caused by window rolling method.The empirical results show that the time-varying tail risk network established by ΔCoVaR can effectively reflect the evolution process of the risks of financial system,industry and institution over time,which verifies the validity of the model.The second part introduces the network between institutions into the CoVaR estimation equation,and proposes a CoVaR model considering community structure under stochastic block model.The model can take into account the difference of risk contagion caused by different groups of institutions,and can significantly improve the estimation of the risk level of institutions.This paper presents the parameter estimation method of the model,and proves the consistency of the estimated parameters.The large-sample nature of estimated parameter is verified by numerical simulations under different network structures.The model is applied to the listed banks of our country,and the results show that the CoVaR model considering the community structure can effectively measure the risk level of institutions,and our model is better than other models.The third part introduces the influence of multiplex networks into the CoVaR,and proposes a risk measure suitable for multiplex networks.The CoVaR model considering multiplex network structure can evaluate risk contagion from various channels and improve the accuracy of estimate of institutional risk level.To avoid variable redundancy caused by tight connections,which will lead to inaccurate CoVaR estimation,this paper proposes to identify the institutions that have an important impact,and then introduce these institutions into the CoVaR model.The consistency of estimated parameter is proved.Applying the model to listed banks of our country,the results show that compared with the monoplex network,the multiplex network can more effectively reflect the risk level of the institution.At the same time,compared with other CoVaR models,the CoVaR model considering the influence of the multiplex network can obtain more accurate risks level of institution.