Generalized Cross Entropy estimators and the estimation of interregional trade flows

This paper addresses the challenge of estimating interregional trade flows, a crucial concern for subnational input–output and Computable General Equilibrium models. A commonly used solution is the gravity approach; however, entropy maximization can serve the same purpose. We examine two distinct entropy specifications: the data-constrained Generalized Cross Entropy (GCE) estimator and the moment-constrained GCE estimator. The performance of each estimator is evaluated through numerical simulations. Our results show that entropy-based estimators yield more accurate results than the gravity model, once the possibility of potential errors in row or column totals is taken into account. The larger the potential error and the higher the number of regions, the greater the difference in accuracy between the gravity model and GCE estimators. Additionally, we find that the moment-constrained GCE estimator appears to be more accurate than the data-constrained GCE estimator under specific conditions.