Forecasting crude oil market volatility: can the Regime Switching GARCH model beat the single-regime GARCH models?

TL;DR

This paper compares single-regime GARCH models with the two-regime Markov Regime Switching GARCH model for crude oil volatility forecasting, finding the two-regime model generally outperforms others in in-sample estimation and daily data forecasts.

Contribution

It provides a comprehensive evaluation of the forecast performance of single-regime versus two-regime GARCH models across different data frequencies and horizons, highlighting the advantages of the Markov Regime Switching GARCH model.

Findings

Two-regime MRS-GARCH outperforms single-regime models in in-sample estimation.

MRS-GARCH provides more accurate daily volatility forecasts.

Linear GARCH models excel in VaR forecasting compared to nonlinear models.

Abstract

In order to obtain a reasonable and reliable forecast method for crude oil price volatility, this paper evaluates the forecast performance of single-regime GARCH models (including the standard linear GARCH model and the nonlinear GJR-GARCH and EGARCH models) and the two-regime Markov Regime Switching GARCH (MRS-GARCH) model for crude oil price volatility at different data frequencies and time horizons. The results indicate that, first, the two-regime MRS-GARCH model beats other three single-regime GARCH type models in in-sample data estimation under most evaluation criteria, although it appears inferior under a few of other evaluation criteria. Second, the two-regime MRS-GARCH model overall provides more accurate volatility forecast for daily data but this superiority dies way for weekly and monthly data. Third, among the three single-regime GARCH type models, the volatility forecast of the nonlinear GARCH models exhibit greater accuracy than the linear GARCH model for daily data at longer time horizons. Finally, the linear single-regime GARCH model overall performs better than other three nonlinear GARCH type models in Value-at-Risk (VaR) forecast.