The predictive content of co-movement in non-energy commodity price changes

Resumen

The predictive content of the co-movement either of a large range of commodities, or the co-movement within a specific category of raw material prices is evaluated. This paper reports success in using small scale factor models in forecasting the nominal price of non-energy commodity changes on a monthly basis. Therefore, communalities of commodities in the same category, estimated by the Kalman filter, can be useful for forecasting purposes. Notably, category communalities in oils and protein meals, as well as metals seem to substantially improve the forecasting performance of the random walk model. In contrast, co-movement in extensive data of commodity prices, estimated through Principal Components, has poor predictive power over non-energy commodity prices, compared to the small-scale factors.