Bayesian cohort component population projections: an exploration of different model specifications
Peter W.F. Smith, University of Southampton
James Raymer, University of Southampton
In this paper, we explore the use of Bayesian methods for cohort component population projections. The rationale for considering a Bayesian approach is that it offers a more natural framework than traditional probabilistic methods to project future populations with uncertainty measures. We focus on only a small part of the picture, that is, to explore the consequences of choosing different specifications of age-specific fertility and mortality in a closed cohort projection model in terms of its forecasted populations and measures of uncertainty. For illustration, we use a historical time series of fertility and mortality from England and Wales. The data consist of age-specific rates for single years of age going back in time to 1841 for mortality and 1974 for fertility. Finally, we present and compare the projected populations, with associated measures of uncertainty, and evaluate the merits of the various specifications for including fertility and mortality.
Presented in Session 110: Issues in stochastic forecasting