Connections between Commodity Prices and Global Macroeconomy : Evidence from Machine Learning and GVAR

We investigate the connections between some specific commodity prices and global macroeconomic performance in a two-stage approach. At the first stage we employ machine learning techniques to identify from a large set of globally traded commodities the ones with the strongest predictive power on global macroeconomic activity. After this, we use a global vector autoregressive (GVAR) model to assess the global economic reactions. Our results indicate that out of the focused 59 commodities, only 6 are revealed to be the best predictors regarding the development of global macroeconomic activity indicators. The results also indicate that the effects are not unanimous across commodities or macro variables. We also show that the commodity market exposure is much stronger among the advanced countries (like the Euro Area), other developed economies, and China, compared to the emerging economies such as in Africa, Asia and Latin America, both at individual country and regional levels