Forecasting Energy Commodities Prices with Bayesian Model Combination Schemes

Abstract

The biggest advantage of Bayesian approach is that it allows to use data-rich models in a reasonable way. The conventional approach is not suitable when the number of variables exceeds the number of observations. Bayesian formulas remain reasonable in such a case. Researchers usually deal with numerous potentially important variables in forecasting. Newly methods like Dynamic Model Averaging (DMA), Dynamic Model Selection (DMS) and Median Probability Model nicely deal with model uncertainty. The study discusses energy commodity prices development. FRED-MD large macroeconomic database is applied. Forecasting commodities prices is a hard task. Therefore, it seems interesting to check the novel methodology. It allows both explanatory variables and regression coefficients to vary in time. Thus in different periods, different factors can be treated as the important ones.