Dependency, centrality and dynamic networks for international commodity futures prices

Abstract This paper adopts a network approach to measure dependency among a set of international commodity futures prices. We first use partial correlations to construct a static dependency network for a vector of variables, and then illustrate within-system connections in a minimum spanning tree (MST) to evaluate centrality of the variables. Rolling-window estimation is then applied to address time variations in both dependency and centrality networks. We show that crude oil price plays a pivotal role in connecting together components in the networks and there is clear evidence of time-varying within-system dependency. Our method demonstrates a new and easy-to-apply way to investigate dependency. The empirical results provide new evidence to the recent intensive discussions on financialisation in energy and commodity markets.