GARCH model and predictive performance of volatility forecasting: evidence from oil market

Abstract

This study investigates the predictive performance of the GARCH (1, 1) model in forecasting the return volatility of the three major items on the oil market, WTI crude, Brent crude and heating oil, for horizons of one, five, twenty and thirty days ahead. The empirical results indicate that the predictive performance of the GARCH model is best for short horizons, from one day ahead for WTI, one to five days ahead for Brent and one to twenty days ahead for heating oil, at the 5% significant level. For horizons from twenty days ahead, no robust forecasting model can be observed for any of these items. Thus, although the results show that the GARCH (1, 1) model has high predictive performance in forecasting the volatility for the short-term horizons, it cannot produce reliable results when the forecast horizon increases. The findings also show how volatility clustering on returns emerges on the oil market, which creates periods of high volatility followed by small movements.