Commodity network and predictable returns

Abstract We investigate the lead–lag relation in the cross‐section of commodity returns. We estimate dynamic and directional networks for 32 commodities and then construct a new predictor termed commodity network momentum, exploring cross‐commodity information spillover. Network momentum positively and significantly predicts future commodity returns, controlling for existing commodity characteristics. Unlike previous lead–lag studies, the predictive relation is consistent with overreaction rather than underreaction. The relation is stronger for attention‐grabbing commodities and commodities with lottery‐like properties and with higher limits to arbitrage. Extrapolation from connected commodities contributes to this predictive relation. Overall, our paper highlights the role of information spillover in commodity return predictability.