The Application Of Bayesian AGARCH Model In Chinese Commercial Bank Interest Ra Risk Measurement

The paper is to research the risk of interest rate in China’s commercial banks. Taken the representative overnight interest rate which is belong to Shanghai interbank interest rate as research objects, the paper construct bayesian AGARCH model to analyze the volatility of returns and VaR calculation of interest rate. In this paper, the main content is as follows:The first chapter, I elaborate the background of selected topic and research significance, carrying on the literature review.The second chapter, I introduce the management system of interest rate risk in our country’s commercial bank which is included definition, recognition, measurement.The third chapter, the empirical interval is from January 4th,2007 to May 29th, 2013.1 elect SHIBOR overnight interest rate data as the research object, then analyze the characteristics of SHIBOR yield sequence. It is observed that the returns has a series of characteristics which is inclued wave agglomeration, rush and fat-tailed features, asymmetry, obvious characteristics of the ARCH effect, so I construct AGARCH(1,1,1) model and estimate the model parameters by maximum likelihood method. The prediction interval is from May 30th,2013 to August 15th,2014. Then predict the next issue of conditional mean and conditional variance in order to measure interest rate risk of our country commercial bank in chapter 5.When construct AGARCH model, it is difficult to estimate the model parameters by maximum likelihood method because some of the model parameters need nonnegative and in order to improve the precision of the model in chapter 3. So bayesian AGARCH model is established in chapter 4. Setting the posterior distribution of the model through the prior distribution of parameters and likelihood function of sample, use the MH sampling algorithm to estimate the parameters. Through empirical I find out the value of α1+λ1+β1 is 0.83153, less than 1, the volatility has not the character of continuity.λ1 is equal to 0.02001, greater than 0. Returns sequence has the "leverage effect". Furtherly sensitivity analysis is carried on the model, we can see from the results that in spite of the initial value which we set is different, but the parameters estimation result of bayesian AGARCH(1,1,1) model is basically consistent. Eventually I compares the bayesian AGARCH model to AGARCH model, it is found out that using bayesian AGARCH model to estimate parameters, the error is significantly less than the maximum likelihood method. I predict the next issue of conditional mean and conditional variance in order to measure interest rate risk of our country commercial bank in chapter 5.The fifth chapter, calculate the value of VaR of interest rate in China’s commercial banks. We accord AGARCH model and bayesian AGARCH model to predict the next issue of conditional mean and conditional variance. The result show that under the confidence level of 95%, long positions, for example, the average daily value of VaR is 0.0878 based on the bayesian AGARCH model, it is small than 0.1010 based on the AGARCH model in the prediction period. Then testing the validity of the VaR model by failure test, the failure rates of bayesian AGARCH model is 4.92% which is more close to 5% of the expected failure rate. So using bayesian AGARCH model to measure our China’s commercial interest rate risk which is more accurate.The sixth chapter is the summary of the artical and the ending.