Forecasting prices of selected metals with Bayesian data-rich models

Abstract The paper presents application of various Bayesian model combination schemes to metals spot prices forecasting. The considered schemes arise from recently gaining attention Dynamic Model Averaging (DMA). Lead, nickel and zinc spot prices are analyzed. Monthly data from 1996 to 2017 are used. The considered schemes seem to be an interesting alternative to some benchmark models. Interestingly, model selection is found more beneficial to tightening forecast accuracy than model averaging.