| Abstract: | "A full multiregional projection model requires migration data that are simultaneously classified by age and gender and region of origin and region of destination. Except for a very small number of regions, these data requirements are so high that aggregation of the data (which is equivalent to simplification of the model) is called for. This paper investigates the extent to which the full internal migration matrix can be simplified without seriously affecting the performance of the resulting multiregional model. Using IPF (iterative proportional fitting) methods, a log-linear analysis of alternative model specifications is made, using data for Italy, the Netherlands and the UK....A reasonable balance between goodness-of-fit and parsimony is found for the model in which time interacts with the main effects only (i.e. with age/sex, with origin and with destination)." |
