Analysis Of The VaR In Time Series Based On MCMC Methods

Risk management is the eternal topic for the financial market participants.Es pecially now,the economic develop rapidly and there are different kinds of financ ial products and financial innovation,so the financial productions' risks is also a problem of which investors concerned about.So there are a lot of qualitative and quantitative modeling to calculate the financial risk.Corporate bonds started latel y in our country and the development process is not smooth,but in January 2015,the new rules for the issuance of corporate bonds promulgated.After,the scale of corporate bond financing sharp increase and thus the risk research has become a n important issue.From the corporate bond market risk measurement,this paper proposes to est ablish a time series model and estimates the risk.Recalling the relevant literature,.first expound the latest development of corporate bond and risk,and then collect ing the SSE Corporate bond yields data from April 1,2014 to February 22,2016,a total of 462 days to study,found that the presence of volatile data gathered,pe ak and fat tail characteristics.Thus consider to establish the GARCH models.Thro ugh the sequence of ADF,PP stationary test,established AR(2)model firstly.Afte r comparing the AIC value about AR(2)-GARCH(1,1)and GARCH(1,1)model,s elect a smaller one.It is AR(2)-GARCH(1,1),which as the basic model in this p aper.Then this paper use ML method MCMC algorithm which based on Gibbs s ampling to estimate the model parameters.The MCMC algorithm could achieve d ynamic simulation.Last,calculating the VaR values in two methods.The conclusio n is in the same level of confidence,the VaR value in MCMC algorithm is big ger than in ML method.At 95% confidence level,if corporate bond market worth1 million,the mean is-53.23 thousands and-52.32 thousands.Under the same method,the VaR average value is bigger than the median.Institutional investors h ave a certain amount of buffer space when estimating corporate bond investment risk,and the MCMC algorithms more suitable than ML.Finally,according to the evidence,presented thinking and summarize in time series modeling process,and make suggestions for investors,financial regulators and issuers,as we can establish a risk measurement model system combined with national conditions,extensive hedging products,enhance company internal and external financial supervision and risk control measures and so on.