Evaluation of Volatility Forecasts in a VaR Framework

Abstract

Many methods can be considered to select which volatility model has a better forecast accuracy. In this work a loss function approach in a Value at Risk (VaR) framework is chosen. By using high-frequency data it is possible to achieve a consistent estimate of the VaR bootstrapping the intraday increments of an asset. The VaR estimate is used to find a threshold discriminating low from high loss function values. The analysis concerns the high-frequency data of a stock listed on the New York Stock Exchange.