Study On The Computing Methods Of VaR Based On Extreme Value Theory

In recent years, factors like globalization of the economy, liberlization of the finance, competition and unregulation, financial innovation and technological advancement and so on, have rapidly enlarged the financial market scale and have significantly improved its efficiency, and as well as have largely increased its volatility and risk. Especially since the last two years, with the financial crisis sweeping the globe and resulting in a weak world economy, market risk faced by financial assets has increasingly been prominent and complex.As a financial market risk measure method, VaR (Value-at-Risk) appeared at the early part of 1990's. The risk management technique about VaR is a statistical model and method used to estimate and measure finance market risk. The correct estimates of VaR and CVaR are the real challenges to risk managers. The normal distribution is very often inadequate for the description of real financial data with heavy-tail distributions, especially very large quantile that interest to a risk manager. Extreme Value Theory models the tail of the return distribution rather than the whole distribution. It can capture the tail risk that often causes large losses in financing institutions, so it is a good approach for risk measurement in finance field.Extreme value theory is used to analysis the extreme values of random vectors and processes by the statistic methods. EVT mainly studies extreme value and models the tail of distribution financial return, it can effectively forecast and guard against the financial risk on the condition of lacking of sample data. More and more people recognize the great potentials of EVT dealing with the risk of extreme event. Especially EVT can be used in application to value at risk due to modeling the tail of distribution.This paper first introduces the knowledge about financial risk, sums up the present situation about domestic and foreign VaR research, and introduces the concept and computing theorem. Secondly, three kinds of traditional computing methods of VaR, such as History Simulation, Monte-Carlo Simulation and Variance-Covariance Simulation, have been introduced in detail, and then been compared from various aspects. Because the pure VaR method has the shortcoming of dissatisfied uniform risk measure, this paper also presents uniform risk measure method CVaR, and the CVaR method better capture finance data rear part distribution, mading up the shortcoming of VaR. Finally, this paper presents the theory of extreme value and character of tail of distribution and gives the example of VaR with index of Shanghai stock market by EVT, then compares the VaR results of different computation methods and concludes that traditional VaR method is static state model and VaR with EVT is dynamic conservative model and has the ability of forecasting risk out of sample comparing to historical simulation method.This paper aims at improving the applicability and precision of VaR by using the knowledge of Extreme value theory and etc. I believe that this paper is instructive for our financial institutions to control market risk.