Smoothing destination-specific migration flows

Accurately estimating age profiles for destination-specific migration is requisite to understanding the determinants of population growth and projecting future change as migration is the primary growth determinant for most regions. In Australia, place-to-place flows based on the age profile of migration derived from census data are commonly used to empirically estimate destination-specific internal migration. However, such flows are heterogeneous and census data is imperfect for accurately generating migration-age profiles. Demographers have addressed this by developing a range of methods for smoothing migration probabilities. These address smoothing on a bi-regional basis, primarily with one destination–origin pairing. We propose a non-parametric method for smoothing destination-specific migration probabilities which can be applied to multi-regional smoothing and is within the generation–distribution framework of Rogers et al. (Environ Plan A 34:341–359, 2002). We demonstrate that, if total age-specific out-migration has already been estimated, smoothing destination-specific migration ratios provides a solution to imperfect input data. Using the example of Australian interstate migration, we show how the method can give an accurate fit to the migration ratio profile across high-curvature ages and a good treatment of sample noise both when the population at risk is low, such as at advanced ages, and when the destination has a low conditional probability of migration. An implementation of the method is available as an Excel add-in.