Empirical Studies Of The VaR And ES Models On Fattailed Distribution

In recent years the international finance crisis repeatedly occurred, Caused the broad scholars and the financial supervising and managing department to the risk survey high value and the research, Simultaneously comprehensively lets loose along with our country money market, affects the money market undulation factor day by day increases, the financial risk measure question will change more and more importantly. At present, the mainstream method financial risk measure for is value at risk (VaR) which emerges only after 93, It is refers the market in under the normal undulation situation, some one portfolio most greatly loses under specified probability, But VaR also has the very many exploitation conditions, these conditions are not often consistent with the market condition. Also under consistency risk measurement standards, because VaR in general turns out to be not even a convex measure and in particular not subadditive even when the two random variables are independent, and VaR concern is the frequency of loss, rather than the size of the loss, we introduce the concept of expected shortfall, which refers the even value of the loss for surpassed VaR, it cares about is the loss size, but is not the loss frequency, more importantly it is the consistency risk measurement. At the same time we compared risk Riskindicators 's of VaR, ES, Variance, discussed between the relations of them, given two kind of computations expressions for expected shortfall under the normal distribution, But regarding under other distributions, the ES computation is very difficult, We produced estimator of ES model based on the extreme value theory. Finally we did the exhaustive experiment design, we did empirical study based on SHANG HAI Composite index, After and did corresponding backtesting. The test results indicate it is feasible to divide financial data by R/S statistical and then use the corresponding model, the traditional VaR models improved clearly by use extreme value theory, Settled the problem of underestimating the tail risk value of VaR in thick tail distribution, which also showed the importance of extreme value theory in risk management, On the other hand the use of the ES based on the extreme value theory, can better address the problem that the VaR model is not consistency risk measurement in the non-normal distribution...