Comparative Analysis of the Quality of Forecasting in Volatility Assessment Models of Financial Assets

The most of previous works on analyzing the prediction quality of realized volatility models consider the security market and just a few researches concern the oil market. With this motivation, the paper aims to compare the prediction ability of the realized volatility models (HAR-RV model) and ARCH family models (GARCH, GJR-GARCH and EGARCH models) on the oil market using Brent oil futures as observed data. We apply Schwartz’s Bayesian Information criterion to choose the best performing conditional variance models and employ the Root Mean Squared Error (RMSE) approach to determine the forecasting errors. After calculating the daily realized volatility, we also use the method of RMSE to determine the forecasting errors in HAR-RV model. Finally, we compare the forecasting errors obtained in each model. The results showed that HAR-RV model generates more adequate forecasts of oil futures volatility than ARCH family models. These results also prove that realized volatility models demonstrate better forecasting performance compared to conditional variance models for the whole energy market as well.