Constructing Bayesian Stochastic Volatility Model To Study The Volatility Of Internet Financial Product Yu'e Bao Returns And Its Dynamic VaR Measure

In recent years,the study on the volatility of the returns of Internet financial products and its risk measurement have become more and more important to investors and regulators.This paper takes the yield rate of Yu'e bao as the research object,after the analysis of its Internet financial characteristics,the bayesian random fluctuation model was established to study its fluctuation characteristics.On this basis,the bayesian POT model was established to measure its dynamic VaR.The research content of this paper is mainly divided into three parts:The first part is to explore the characteristics of the Internet finance of Yu'e bao data after the overview of its operation mode,laying a foundation for the establishment and test of the model.This paper explores the characteristics of Yu'e bao's Internet finance from the perspectives of volatility,tail and peak.For the volatility,we divided a small sample and compared its standard deviation with the overall standard deviation,founding that the fluctuation of the last period had a significant impact on the current fluctuation.When studying the tail and peak parts,the yield data of Yu'e bao were standardized with the standard normal distribution and the Student's t-distribution as the comparison criteria.In determining tail length characteristics,choosing a point on the end of the sample distribution a numerical value as the left endpoint of a certain interval,fixing in the area between the probability value,to determine the area between the right endpoint,so as to calculate the area between the length,by comparing three kinds of distribution of the interval length,finding the Yu'e bao sample data corresponding to the interval length of the longest,and its tail is longer;In determining tail thickness characteristics,for the three distributions,exactly the same tail interval was selected to calculate the probability of sample points falling within the interval respectively,founding that the probability of Yu'e bao yield rate was the smallest,so its tail was thinner.In determining the characteristics of the peak,probability nonzero and probability differentiability rules are established around its mode for each distribution,a sufficiently small neighborhood radius is determined,and the probability of sample data falling into this neighborhood is calculated.By comparison,it is found that Yu'e bao yield distribution has a more pointed peak.The second part is to establish the Bayes-SV-T model on the basis of the first part,to study the volatility of Yu'e bao yield,and to lay a foundation for the construction of Bayes-POT model in the third part.Firstly,based on the characteristics of Internet finance and the specific advantages of SV model compared with GARCH model in model form,parameter number,parameter estimation and model test,the model is set as SV-T model.Second,determining the likelihood function model and seting each parameter of the prior distribution on the basis of its nuclear,deriving the joint posterior distribution and full conditional posterior distribution of each parameter,Using Gibbs sampling to estimate the parameters,and get the Bayes-SV-T model.Then,the model was used to analyze the volatility,founding that the volatility had a strong persistence,with the highest volatility level of 0.6528 and the lowest volatility level of 0.039.Moreover,after July 16,2015,the average volatility level value was taken as the horizontal axis,and the fluctuation curve presented a sinusoidal feature with a period of 890 days and an amplitude of 0.2.Finally,the random simulation method is used to test the model,and the fractional value of the simulated sample is compared with the real sample,and the error is found to be about 0.1%,which verifies the reliability of the model and predicts rate of return.The third part is to further build the Bayes-POT model of Yu'e bao yield on the basis of the Bayes-SV-T model above,and conduct dynamic VaR measure for its tail risk.Firstly,the category of POT model and the threshold of the disturbance term of Bayes-SV-T model are determined.Secondly,the Bayes-POT model was established and solved to calculate the dynamic VaR value of the rate of return.Then,Kupiec likelihood ratio and bayesian method are used to test and compare the VaR value,and the results show that the risk measure model constructed in this paper is reliable and the bayesian method is better.Finally,the model is used to extrapolate the dynamic VaR value.