A new high-frequency volatility forecasting model for crude oil market-evidence from RGARCH-CARR-MIDAS model

Predicting of high-frequency volatility in crude oil has emerged as a prominent research area in recent years. To address the issue of excessive parameters in the Realized GARCH model used for high-frequency volatility prediction, the RGARCH-CARR model was proposed. However, this model fails to capture the long memory and leptokurtosis. To overcome these limitations, we introduce MIDAS regression to develop the RGARCH-CARR-MIDAS model. Furthermore, we examine the effectiveness of this model through empirical studies and robustness tests, which demonstrate its superiority regarding in-sample fitting and out-of-sample volatility forecasting.