Modeling and Forecasting Evaluation of Different Models of Short-Term Memory, Long-Term Memory, Markov Switching and Hyperbolic GARCH in Forecasting OPEC Crude Oil Price Volatility

Predictability in financial markets is very complex, and the reasons for this complexity can be summarized as non-standard data, nonlinear data flow, and large variations in data. Determining the proper pattern for forecasting volatility can play a significant role in decision making. In the old econometric models it is assumed that the component of error constant during the sample period. But in many financial time series it is observed that during periods of volatility is very sever. Under these conditions, the assumption of the exictence of the equivalence of variance is no longer reasonable. In the present paper, the GARCH, IGARCH, EGARCH, GJR-GARCH, FIEGARCH, HYGARCH, and MRS-GARCH two-regime models were evaluated in prediction of OPEC crude oil price volatility during 2010-2016 based on their RMSE error criterion. The results of this evaluation show the superiority of the Markov Switching GARCH Model on the 5 and 22-day horizons. Also, the long-term FIEGARCH memory model in predicting horizons of 1 and 10 days has better performance in predicting oil price volatilities than other competing models.