Testing for food market integration revisited

Abstract

This article considers the statistical performance of four commonly used econometric tests for market integration: the Law of One Price, the Ravallion Model, cointegration and Granger causality. A spatial price equilibrium (SPE) model, that is subject to both production shocks and general price inflation, and mimics many of the key characteristics of integrated food markets, is constructed. The model is used to generate food price time series of lengths that are typical of the short sample sizes available in most developing countries, for both instantaneously integrated and independent markets. A series of Monte Carlo experiments on these artificial food price time series are performed, which show that all four of the conventional tests for market integration are statistically flawed.