Bayesian functional models for population forecasting

Abstract

We explore the functional modelling approach to population forecasting within the wider context of Bayesian predictions and model uncertainty. The functional modelling approach can be used to analyse and forecast many different age- and time-specific components for fertility, mortality and migration. For each of these demographic processes, we perform Bayesian model averaging across the outcomes of two functional models to take into account model uncertainty. We illustrate the method with a population forecast for the United Kingdom for 2010�2030. We conclude that regularities in age profiles of demographic processes, where available, provide important information for the forecasts and as such should be included in the forecasting process.