ENEFOR has made a significant contribution beyond the current state-of-the-art in forecasting oil price volatility, oil prices and economic policy uncertainty.
In particular, we have added to the scarce literature of oil price realized volatility forecasting using the current state-of-the-art Heterogeneous Autoregressive -Realized Volatily (HAR-RV) model, which we mainly extended in two very important ways. First, we investigated for the first time whether the “information channels” by which different asset classes’ volatilities can impact oil price volatility, could also improve oil price volatility forecasts (we named the model HAR-RV-X, where X denoted the exogenous asset classes’ volatilities). Second, we provided for the first time a method that handles exogenous variables in a HAR model in order to proceed with the forecasts. In addition to these major contributions, we also assessed the forecasting accuracy of the HAR-RV-X models based on each individual asset class, their combined forecasts, as well as the forecast-averaging. We further assessed for the first time whether the forecasting accuracy of the HAR-RV-X models can be improved using the time-varying correlations between oil price volatility and the remaining asset classes’ volatilities.
In terms of oil price forecasting, ENEFOR showed for the first time that the Mixed-Data Sampling (MIDAS) models using either daily asset classes’ volatilities or asset classes’ returns, which are constructed from ultra-high frequency data, exhibit significantly higher predictive ability and directional accuracy compared to the current state-of-the-art models for oil price forecasting (e.g. VAR and BVAR models). More specifically, our MIDAS models with the daily realized volatilities of the exogenous assets classes provide predictive gains relatively to the no-change forecast at the level of 75% at the 12-month ahead forecasting horizon. Even more, our MIDAS models also exhibit a very high directional accuracy, especially up to 6-months ahead.
The last working package of ENEFOR we forecasted for the first time the European economic policy uncertainty, showing that even in this case the information extracted from ultra-high frequency of various asset classes provides predictive gains relatively to the no-change forecast.
So far the project has managed to achieve the intended short-run impact. The fellow gained proficiency on the forecasting techniques using ultra high frequency data, which are essential to tackle other energy related forecasting issues. The project also allowed him to develop state-of-the-art programming skills, as well as, consultancy and policy formulation skills, which are necessary to develop other forecasting models, as well as, to obtain additional consultancy experience. The fellow has also strengthened his affiliations with the Greek academic and non-academic sector, which allowed him to expand his research network and obtain consulting experience. He also improved his public outreach profile and effectiveness through the presentations of the project’s output to staff seminars and public talks. Finally, the project provided the fellow with the opportunity to enhance his reputation in the academia and, thus, gain greater visibility, given the prestige of the Marie Sklodowska-Curie fellowship.
The medium- and long-run impacts that are anticipated to be realized are those reported in the project’s proposal.
