Combination forecasts of China's oil futures returns based on multiple uncertainties and their connectedness with oil

The first objective of this paper is to investigate the overall predictive power of multiple uncertainties as a whole on China's crude oil futures returns. For this purpose, we combine the forecasts of individual uncertainties through various combination methods, namely mean, median, trimmed mean and discount mean square prediction error (DMSPE). The second objective is to explore the predictive power of the connectedness indices between crude oil and uncertainties on oil returns. To address this objective, we generate the one-way connectedness and the net pairwise directional connectedness through a dynamic connectedness construction method under the TVP-VAR framework and evaluate the overall predictive power of these two types of connectedness indices on oil returns by the same steps as we do to address the first objective. The empirical results show that the combination forecasts of the multiple uncertainties and the connectedness indices covered in this paper generally outperform the benchmark at forecast horizons of >1 month. And the connectedness indices show significant superiority to the uncertainties themselves. This paper not only provides a new perspective on the use of uncertainties to forecast oil returns but also extends the application of connectedness indices from economic inference to forecasting purposes, providing new and stable predictors for forecasting oil returns.