Additive outliers, GARCH and forecasting volatility

The Generalized Autoregressive Conditional Heteroskedasticity [GARCH] model is often used for forecasting stock market volatility. It is frequently found, however, that estimated residuals from GARCH models have excess kurtosis, even when one allows for conditional t-distributed errors. In this paper we examine if this feature can be due to neglected additive outliers [AOs], where we focus on the out-of-sample forecasting properties of GARCH models for AO-corrected returns. We find that models for AO-corrected data yield substantial improvement over GARCH and GARCH-t models for the original returns, and that this improvement holds for various samples, two forecast evaluation criteria and four stock markets.