Value-at-risk based on generalized error distribution using a quantile approach

This study proposes the use of the quantile method under generalized error distribution (GED) value-at-risk (VaR) forecasts assuming non-normality, fitting the returns data with the GED. This application is more efficient and flexible not only because it can accommodate both normality and non-normality within one model but also because it is as easy to use as the variance-covariance method. Based on the failure rates determined and the results of the Kupiec test, it would appear that the proposed method can considerably enhance the accuracy of estimates of VaR.