Volatility Forecasting With Range-Based EGARCH Models

We provide a simple, yet highly effective framework for forecasting return volatility by combining exponential generalized autoregressive conditional heteroscedasticity models with data on the range. Using Standard and Poor's 500 index data for 1983–2004, we demonstrate the importance of a long-memory specification, based on either a two-factor structure or fractional integration, that allows for some asymmetry between market returns and volatility innovations. Out-of-sample forecasts reinforce the value of both this specification and the use of range data in the estimation. We find substantial forecastability of volatility as far as 1 year from the end of the estimation period, contradicting the return-based conclusions of West and Cho and of Christoffersen and Diebold that predicting volatility is possible only for short horizons.