Research On Measurement Of Interest Rate Risk Of Chinese Commercial Banks Based On VaR Model

With the market-opening and reformation in China financial area, theamplitude and frequency of wave motion of the interest rate are furtherincreasing. After2010, fluctuation of China interest rate market is increasing tostrengthen, and commercial banks would be confronted with more and morechallenges of interest rate risk. Compared with foreign study of measure ofinterest rate risk, there are difference between domestic research level on thisarea and oversea. In China, most of commercial banks practice to control andmanage interest rate risk with traditional static method such as SensitivityAnalysis and Duration, and these methods come with some shortcomings inaccuracy and timeliness. There are some advantages over VaR method inaccuracy, terseness and dynamics. Nowadays, measure method of interest raterisk based on VaR is so immature that VaR is estimated and tested in accordancewith intra-sample data, and the research on VaR is lagging based on statisticalnature of interest rate sample. Utilizing the advanced method and experience offoreign interest rate risk measure, it is so important to build quantized VaR modelaccording to Chinese interest rate features in order to enhance the abilities ofinterest rate risk management and to improve the level of domestic riskmanagement in financial area.According to summary of research on interest rate risk measure, we arguedthat foreign research on this area is advanced, abundant and cutting-edge, anddomestic research on this area is late beginning and rapid expansion in recentdays. We make analysis to Sensitivity Analysis, Duration and VaR, and draw aconclusion that VaR is better than other methods and more effectual. ChoosingShibor data(2007.4-2014.4) as experimental data, We compare and evaluate36model based on GARCH-family using intra-sample data and chooseAR(1)-GARCH(1,1)-G, AR(1)-TGARCH(1,2)-N, AR(1)-EGARCH(1,2)-G,AR(1)-EGARCH(2,1)-N to forecast, estimate and back-test VaR for out-sampledata. In conclusion, VaR estimated by GARCH-family is more conservative andsample-sensitivity which lead to deviation. On the basis of analysis of stagefeatures of Shibor, We build piecewise GARCH model to estimate VaR, and backtest it. The test result show that piecewise GARCH is better than traditionalGARCH to fit stage features of Shibor, and is more available and precise.