The Theoretical Method Of Threshold Factor Model And Its Application In The Measurement Of Systemic Risk

At present,the world’s major economies are still in the wave of global financial market integration.While integration has deepened economic exchanges between countries and enhanced the vitality of international financial markets,the high degree of interconnectedness has also added additional risks to economies’ own financial systems.The subprime crisis that spread from America to the rest of the world between 2007 and 2009 is a classic example.However,the stock market,as an "economic barometer",will be directly affected by the economic crisis,and then have a "domino effect" on the whole financial sector,which is also known as a "contagious" risk of systemic risk.Therefore,in order to further implement the "systemic financial risk does not occur to hold the bottom line"(zhou,2017),effectively reduce the risk of shock on the market at home and abroad,it is necessary to explore the scientific and effective method to realize the accuracy of Chinese stock market systemic risk measure,and put forward effective measures to actively avoiding and management.However,the increasing complexity of financial markets and the diversification of financial products lead to more sources and types of risks,which makes it difficult to accurately measure systemic risks.As an investment product with higher returns,stock will inevitably expose investors to various risks and bear the economic losses caused by risks.Therefore,it is of great theoretical and practical significance to measure the systematic risk of China’s stock market scientifically.In this context,this paper calculated systematic risk based on the stock market represented by 28 first-level industries of shenwan based on the threshold factor model proposed by Massucci(2017)and Liu and Chen(2018).The threshold factor model breaks through the existing "curse of dimensionality" problem,and can not only study the regional transformation behavior of high-dimensional time series,but also consider the time-variability of factor loading matrix,which has good interpretability and predictive ability.In view of the threshold effect of the factor loading matrix,the Lagrange multiplier,likelihood ratio and Wald test methods are also presented,and the asymptotic distribution under the null hypothesis and the alternative hypothesis is analyzed.Then,monte carlo stochastic simulation experiment is conducted to verify that the above test statistics have good large-sample properties and finite sample performance.In order to find the time-varying and dynamic correlation between each industry and the overall systemic risk level,the dynamic correlation multivariate generalized autoregressive conditional heteroscedasticity model(DCC-GARCH)was used to estimate the dynamic correlation.By decomposing conditional covariance matrix into conditional covariance and conditional correlation coefficient matrix,this model can reduce the number of parameters to be estimated and better describe the transfer process of correlation between different time series.In this paper,1,334 dynamic correlation coefficients of each industry from March 5,2014 to August 27,2019 were obtained and the average value was taken as the contribution rate of the industry to the level of systemic risk.The results showed that the contribution rate of non-bank finance to systemic risk was 0.9588,followed by real estate,banking,light industrial manufacturing and mechanical equipment,which were 0.9569,0.9504,0.9500 and 0.9426,respectively.The lowest contribution rate was 0.7413 for electrical equipment,0.7139 for food and beverage,and 0.6996 for utilities.