Risk Measurement Of The Shanghai Composite Index Based On APARCH And POT Models

With financial market volatility and risk increasing, financial crisis events have occurred frequently. This leaves financial regulators and the investors become particularly sensitive to the general collapse of the financial markets. So it’s necessary to study the market influence of crisis events. The most important thing for evaluating the market risk is find an appropriate method. As a a widely accepted risk-measuring tool, VaR is specified as Bank standard risk measurement tool by the Basle Committee in2001.It’s defined as the biggest losses of an asset or portfolio within a specific time in the future to a certain confidence level, or the quantile of portfolio income loss distribution function. Because of characteristics of quantify, synthesis, earthliness, VaR’s regarded as a more accurate measurement for potential loss of different risk resources.VaR based on traditional methods presume yield sequences behave according to normal distribution, it’s conflict with the fact that most financial asset yields have spike thick tail skewed distributions. For this reason, the extreme value theory(EVT) gradually become the mainstream estimation methods for VaR That’s because according to the PBDB theorem:for a certain class of distributions, the losses exceed large enough threshold will obey General Pareto Distribution (GPD).POT method is using GPD fit exceed threshold sequence.Focusing on the daily return data sample of Shanghai Securities Trade Market’s Complex Index1990Dec. to2012March, the dissertation adopted mainly empirical and comparative analysis method to analyze Shanghai stock market risk based on using widely other people’research results for reference.Firstly, the dissertation analyzed statistic features and distribution of returns series in order to choose sound models. The analysis showed that Shanghai stock market returns series have fat tails and excess kurtosis, weak auto-relation and volatility clustering.Secondly, based on above description and analysis, the dissertation firstly adopted the extreme theory method to evaluate VaR based on POT presume the excess threshold series is independent identical distribution, but the fact is that there are correlations between the excess threshold data, as a result VaR based on POT might overestimate the true VaR. To eliminate the affecting, The dissertation apply APARCH model to filter returns series, then adopt POT model to fit its resid sequence, finally get the VaR of the return sequence according to VaR’s additive property.For simplify, The dissertation call the first model POT model and the second APARCH-POT model. By comparing the VaR values of two models, The dissertation conclude that POT model indeed high evaluate the true VaR, which is VaR based on APARCH-POT models is lower than the VaR based on the POT model.Thirdly, by means of the back-testing of Kupiec failure for testing validity of VaR. The dissertation find that VaR model based both models are valid in all confidence levels. But the APARCH-POT is inferiors to the POT model in low confidence.Finally, the dissertation concludes the dynamic VaR of the whole sample based on the APARCH-POT model. With the character of assemble, the dynamic VaR is divide to five intervals. Each of them is given the detail and the reason in the conclusion. From1990to1994, the beginning of the Shanghai index, average VaR is the highest of the fives; From1995to1999, as the regulations estimating, VaR fall to a certain degree, but still remain high; between2000-2006, it’s a smooth period and the VaR is rather low; From2006to2010,VaR in this period rebound quality; From2010to now, VaR runs from high to low.