Anomalies in agricultural trade: A Bayesian classifier approach

Abstract

This study examines the uncertainty-agricultural trade nexus. Uncertainty effects on macroeconomic indicators such as consumption and investment have been well studied. However, less is known about the relationship between uncertainty and international trade, particularly the heterogeneity of that linkage across sectors. Application of a novel data-driven methodology—anomaly detection and classification via a Naïve Bayesian Classifier—to monthly data at the HS-4 level finds that agricultural imports are reduced when economic policy uncertainty is high. The effects of policy-related uncertainty are more persistent than that of supply-side fluctuations. Anticipatory stock-piling occurred when uncertainty is specific to trade policy.