Recent Advances in Spatial Interaction Modeling Based on Poisson Gravity Models

Abstract

THIS PAPER DEMONSTRATES THE POSSIBILITY OF IMPROVEMENT ON SPATIAL INTERACTION MODELS USING THE POISSON GRAVITY MODELS. A LARGE DATA SET CONTAINING MIGRATION FLOWS BETWEEN 46 PREFECTURES IN JAPAN IN 1960 AND 1985 IS USED. THE RESULTS OF DOUBLY CONSTRAINED POISSON GRAVITY MODELS AND LOG-LINEAR MODELS ARE COMPARED WITH RESPECT TO THE RELATIONSHIPS BETWEEN THE ORIGIN AND THE DESTINATION. THE ADDITION OF OTHER EXPLANATORY VARIABLES REFERRING TO BOTH THE ORIGIN AND THE DESTINATION TO THE MODEL IS ALSO EXPLORED.