High Volatility Thick Tails And Using Extreme Value Theory To Measure VaR

T The core of financial risk management is to measure the risk quantitatively. In order to measure the risk accurately, the statistical distribution must be described. In the normal condition, the financial data is credible, the estimated VaR is accurate. But in the abnormal condition, the trustworthy data can not be got and the estimated VaR is not accurate. Financial risk is associated with low-probability events in the tails of asset price distribution. To capture the behavior of these tails, one should therefore rely on models that explicitly focus on the tails. Extreme value theory-based models do exactly it.In This dissertation, the performance of the extreme value theory in Value-at-Risk (VaR) calculations is compared to the performances of other well-known modeling techniques, such as variance-covariance (Var-Cov) method, GARCH and historical simulation in the Chinese stock market. Financial risk management typically deals with low-probability events in the tails of asset price distributions. To capture the behavior of these tails, one should therefore rely on models that explicitly focus on the tails. Extreme value theory based models do that perfectly. We use the Matlab and Eviews software package to estimate the parameters of the extreme value theory model .The result indicate the GPD is a robust tool to produces the accurate forecasts of extreme losses at very high confidence levels.