The Second-Order Markov Switching Vector Auto-regressions And Its Application On Early Warning Of Chinese Financial Risks

As Chinese economy enters the new normal,the financial sector will also face new development trends.In the process of financial activities,the influence of external or internal factors on financial institutions will lead to financial crisis and even paralysis of financial institutions.At the same time,individual financial institutions may be bankrupted and suffered serious economic losses.Preventing such systemic financial risks has become the focus of Chinese financial work.Therefore,it is urgent to build a financial risk early warning system which is consistent with the Chinese actual situation in.This paper first expounds the traditional financial risk early warning model,the concept and characteristics of financial risk and the theory of financial risk contagion.Combining with theoretical research and practical situation,this paper proposes SOMS-VAR model,which can better reflect the risk accumulation process,and there is no sample information loss problem caused by converting continuous variables into binary variables in the traditional financial risk early warning model.This paper selects 21 quarterly indicators from 2007 to 2016 as the research samples from banking institutions,money market,foreign exchange market,asset bubble market,real economy,government finance and global economy,mainly reflecting the situation since the international financial crisis in 2008.The empirical analysis is based on the four step to build the model:firstly,by comparing the literature on financial pressure index and considering the correlation and sensitivity of indicators,the principal component method is used to synthesize the pressure indices of bank,currency,foreign exchange and asset bubble separately,and these can depict Chinese financial risk situation;secondly,the pressure indices are tested before the model is built and they are stationary,the number of regimes and the order of regression lag were determined,so the SOMS(3)-VAR(4)model is selected;thirdly,the estimation parameters of SOMS(3)-VAR(4)model are obtained by using the maximum likelihood estimation and EM algorithm,and the smoothing probability diagram of regional state transition is compared with the real financial risk situation in China,at the same time,the dynamic correlation among the four subsystems is depicted by the impulse response diagram under different regimes;finally,the SOMS-VAR model and the MSVAR model test results are compared and analyzed,and found that the model in this paper is better.The financial risk situation in the next four years outside the sample is warned lastly.The empirical results show that the SOMS(3)-VAR(4)early warning model well simulates the actual situation,and the time of issuing crisis early warning signal is reasonable and forward-looking.It is found that the financial risk situation is in the medium and high risk regime by predicting the systemic financial risk from 2017 to 2018,which accorded with the actual situation.Judging from the results of the forecast analysis,in the next four years,Chinese financial risks will show a downward trend but it will still be in a medium risk state.