Nonparametric estimation of the gravity model of migration: application of partitioning Poisson estimates to Japanese inter-municipal migration by age–sex group

In the analysis of migration flows, the gravity model is used as a standard framework. However, the analysis of interregional migration within a country by age and gender group using the gravity model is relatively new. This study nonparametrically estimated the relations between the number of inter-municipal migrations and origin–destination distance by age–sex groups in Japan. We assumed the number of inter-municipal migrations is proportional to the origin and destination populations. Poisson local constant estimates were used. The resulting estimator was the weighted average of the migration preference index, i.e., the ratio of the number of interregional moves to the product of the origin and destination populations, so that the interpretation of estimation results was easy and estimation computationally light. Partitioning estimate was used to reduce downward bias caused by a negative relation between the number of inter-municipal migrations and the origin–destination distance. The relation between the number of inter-municipal migrations and the origin–destination distance was fairly smooth, and the elasticity of the number of inter-municipal migrations with respect to distance somewhat larger in absolute value in shorter distances. In addition, age group 90-, especially men, were found to move frequently when compared to age groups 15–19 and 20–24 years.