The Forecasting Ability of GARCH Models for the 2003–07 Crisis: Evidence from S&P500 Index Volatility

Abstract

This article studies the ability of the GARCH family of models to accurately forecast the volatility of S&P500 stock index returns across the financial crisis that affected markets in 2003–07. We find the GJR-GARCH (1,1) model to be superior in its ability to forecast the volatility of the initial crisis period (2003– 06) compared to its realized volatility, which acts as a proxy for the actual. This model is then extended to make forecasts for the crisis period. We conclude that the model’s ability to forecast volatility across the crisis is not substantially affected, thus supporting the use of the GARCH family of models in forecasting volatility.