Multiregional Probabilistic Population Projections for Open Populations, Incorporating an Embedded Spatial Interaction Model of Interregional Migration

Jacques Poot, University of Waikato
Michael Cameron, University of Waikato

Most multiregional population projections are undertaken either top-down, (where a national projection model is run initially and then sub-national models are run and subsequently moderated to ensure that the regional projections sum to the national projection) or bottom-up (where regional projections are run without recourse to a national projection and the national projection subsequently obtained by simple aggregation). While in a closed system (with zero aggregate net migration), the two approaches may give similar results, in the case of an open system (with large spatio-temporal variations in external migration) the latter is preferred on theoretical and performance grounds. Current practice is to either assume a given net migration level or to assume age-sex specific in-migration and out-migration rates. Instead, we treat in-migration and out-migration simultaneously by modelling gross inter-regional migration flows by means of an Alonso spatial interaction model and by modelling regional immigration and emigration separately. We illustrate our methodology with the example of projecting the populations of the 66 Territorial Authorities in New Zealand, which range in population from a few thousand to 1.5 million. We explicitly incorporate parameter uncertainty by means of Monte Carlo simulations to yield probabilistic projections.

Keywords: Population projections, forecasts, and estimations, Spatial Demography

See extended abstract.