Poisson regression models of Canadian census division migration flows

Tax-file data are available for moves between census divisions in Canada on a year-by-year basis, and may be the best data source for studying migration flows between censuses. This paper reports the results of fitting a series of models to the 1985 to 1986 flows using a form of regression analysis based on the Poisson distribution that is appropriate for count data. These models confirm results well-known from interprovincial and intermetropolitan migration analysis, but also reveal some more surprising results.