Investigating commodity price interdependence with Granger causality networks

This paper investigates the interdependence among commodity prices. Although the literature on this topic is extensive, it often struggles with the issue of dimensionality. To address this challenge, the paper proposes a solution based on network theory and a sparse estimation approach. The analysis relies on a large dataset of about 50 monthly commodity prices (1980–2024), grouped into energy, metals, agriculture, food, and other raw materials. A Commodity Price Network is constructed via Granger causality tests, using both pairwise and sparse VAR models, applied to price levels and first differences to account for potential non-stationarity. The results show that network topology is sensitive to the methodological approach. Nonetheless, consistent patterns emerge: metals and energy commodities maintain a central role, alongside certain agricultural products. In the sparser network configuration, energy commodities exhibit an average outgoing modularity 2.2 times higher than that of metals. Conversely, metals display an average ingoing modularity 5 % greater than energy commodities. Notably, iron shows 2.5 times more outgoing links than crude oil. • Analyzing price interdependence across many commodities and natural resources is highly challenging. • We propose an approach using Granger-causality price networks from alternative VAR models. • The analysis covers almost 50 commodity prices observed at a monthly frequency from 1980 to 2024. • Prices are grouped into energy, metals, agriculture, food, raw materials to assess interdependence within and between groups. • Evidence of widespread interdependence confirms the centrality of key metals and some agricultural commodities.