Study on backtesting techniques of VaR for China stock markets

Abstract

VaR is a popular type of technique that focuses on risk measurement, whereas, the precision of VaR model is relied on the backtesting of VaR. Therefore, the approaches of backtesting are initially significant for us to measure the model of VaR. In order to obtain a relatively correct VaR model, the author builds a GARCH model and eliminates the heteroscedasticity effect of this model. Since the importance of risk measurement, the backtesting methods of VaR model should not be ignored. Insofar as we can see, there are two methods of backtesting which are unconditional coverage backtesting and conditional coverage back testing. The results of this study reveal that a back testing VaR model should not just depend on the test of unconditional coverage, but rely on the conditional coverage backtesting. The reason to consider the independence of exception and unconditional coverage backtesting is that the joint test is a more sound and powerful test for us to measure the VaR model. In this study, the author has critically assessed the accuracy of a VaR model to China’s stock market and compared the methods of backtesting. The daily VaRs are calculated by using the Shanghai composite index for the period 2002-2011.