Study On The Early Warning System Of China's Financial Systematic Risk Based On Logit Model

Since the last century,both the developed and developing countries have experienced a number of financial crises,which have caused irretrievable losses to the national economies.As the process of globalization,the destructive power and the infectivity of the financial crisis significantly enhanced,thus no country can be other countries in the financial crises,soothe way to effectively monitor domestic and foreign financial risks and accurate warning became national scholars' Hot spots.Since joining the WTO in 2001,China has played a more important role in the world economy,and its sensitivity to external economic fluctuations has gradually increased.At the same time,China's debt,real estate and other areas accumulate enormous financial risk,with the slowdown in economic growth,these problems may further erupt,to bring the risk of China's financial system.Therefore,it is of great and far-reaching significance to establish an early-warning system of financial risk in line with China's national conditions.First of all,considering the need of the follow-up construction model,this paper chooses the output gap as the proxy variable of financial risk to determine the outbreak time point and the influence length of the financial risk based on the domestic and foreign financial stress index related literature,and take the relevant landmark financial event as supplementary evidence of financial risk timing.Secondly,this paper analyzes the advantages,disadvantages and adaptability of different early warning models based on existing literatures,and combines Logit model with both Chinese characteristics and data characteristics as early warning models.Then,based on domestic and international early warning model construction experience,we choose the credit-to-GDP gap,house price index etc as an explanatory variable.After determining the model and the index,the binary Logit financial risk early-warning model is constructed with quarterly data from 2000 to 2015.At the same time,in order to study the relationship between policy-making and early-warning models,this paper studies the performance of early-warning models under different policy preferences based on the early warning model.Thirdly,this paper explores the using of multivariate Logit model in financial risk early warning,and constructs the multivariate Logit model with single factor and multiple factors.At the same time,the advantages and characteristics of binary and multivariate Logit early warning models are compared.Finally,this paper summarizes the conclusions of the study,generally has the following four points:The multivariate Logit early-warning model with quarterly data is suitable as an early-warning model of financial risk in China.The model can be integrated into multiple levels to adapt to the heterogeneity of different financial risks in different periods.The results show that the accuracy of the early warning model of the Logit early-warning model is between 0.4 and 0.6,and the AUROC is higher than 0.7,which satisfies the accuracy of the early warning model Efficiency experience requirements.In this paper,the relevant indicators of credit,bank fragility and macroeconomic background are selected as early warning indicators according to the theoretical study,but the accuracy and efficiency of the indicators are low.As a warning index,house price has almost no early warning capability,but it can significantly improve the model's early-warning ability.It shows that the mutual promotion between credit expansion and real estate bubble will bring more financial risks.The empirical results show that the multivariate Logit early-warning model can effectively identify the turbulence period after the risk,avoiding confusion with the general period,and can be applied to China's financial risk early warning.But the current statistical data can not fully meet the requirements of the model for the amount of data,the early warning system we now establish is simple and less stable.The discussion of different policy-oriented scenarios shows that the utility of any model is highest when the proportion of risk weight is equal to the proportion of the whole time in the normal period.We found that the higher the TPR,the better the performance of the model in the high risk weight;also the lower the FPR,The better the performance in the low risk weight.