Abstract

This paper provides an overview of the commodities market, considering three relevant attributes: predictability, similarity, and efficiency. We examine the monthly spot and futures prices time series for 22 commodities from January 1984 until January 2022 with 457 observations. We estimate the permutation entropy (í µí°¸íµí°¸í µí±) and Fisher information measure í µí°¹ í µí± using the Bandt & Pompe method (BPM). We employ the value of these two complexity measures to construct the Shannon-Fisher Causality Plane (SFCP), which allows us to evaluate the disorder and assess the randomness present in the monthly spot and futures prices time series for these commodities. Moreover, we apply í µí°¸íµí°¸í µí± and í µí°¹ í µí± to classify the commodities using complexity hierarchy. We find that the commodities that are located farther from the random ideal position (í µí°¸íµí°¸í µí± = 1, í µí°¹ í µí± = 0) in the SFCP, such as Natural gas, Europe; Iron ore, cfr spot, and Potassium chloride are marked by lower entropy, higher predictability and lower efficiency. In contrast, the commodities that are located near the random ideal position (í µí°¸íµí°¸í µí± = 1, í µí°¹ í µí± = 0) in the SFCP, such as Crude oil-Brent; Crude oil-average, and Silver are characterized by higher entropy, lower predictability and higher efficiency. The K-means algorithm and the hierarchical cluster grouped commodities into only three distinct groups, which is a strong indication that commodity prices have very similar behaviour.