Comparative study of techniques estimating non-normal Var

Abstract

Value at Risk (VaR) is a high quartile of the distribution of negative returns, typically the 95th or 99th percentile. It provides an upper bound for a loss that is exceeded only on a small proportion of occasions over a given time horizon. The VaR technique has undergone significant refinement since it originally appeared about a decade ago. We require a dynamic VaR model that is robust during increased volatility and is known to participants before hand. The certainty and transparency of a rule based dynamic margin system would not impinge upon market efficiency while protecting the stock exchange from a default crisis. 2 In the following sections, there is an overview of the appropriate volatility model namely, GARCH. This is followed by a broad outline of the Extreme Value Theory (EVT) used to model points in the tail of a distribution. These two models are combined to present a robust VaR measure. Next section presents the outline for a Power Transformations followed by how this would enable to get an estimate for the VaR. A description for the NM-GARCH model is presented. These are the models where errors have a normal mixture conditional distribution with GARCH variance components. This is followed by the process of determination of alternate VaR using NM-GARCH. The data analysis proposed and prospective results are described for conclusion.