The effect of data aggregation on a Poisson regression model of Canadian migration

This study examines whether the national migration system of Canada described by a Poisson regression model exhibits any aggregation effects. The results discussed below are preliminary findings from the second stage of larger study. The original data set consists of a 266 by 266 matrix of 1986 migration flows for Canada derived from income tax returns. The 266 regions represent the census divisions. The study uses 260 of these regions, excluding the divisions from the territories due to lack of comparable socio-economic data. The Poisson model used is essentially an unconstrained gravity model in which migration flows between origin region and destination region are predicted with an exponential function of the origin population, the destination population, and the separating distance. -from Authors