Correlatedness of Mortality Assumptions in a Probabilistic Population Projection Model

Christina Bohk, University of Rostock
Roland Rau, University of Rostock

In this paper, we examine an important methodological issue in probabilistic population forecasting: the correlatedness of demographic events, and its impact on the accuracy of projection outcome. Recent studies in demography show that mortality differs among natives and immigrants, i.e. that immigrants' lower mortality approaches natives' higher mortality over descendant generations. We describe an innovative concept to set correlations among mortality assumptions for natives, immigrants, and their descendant generations in the Probabilistic Population Projection Model (PPPM), and we apply this new methodology in two probabilistic projections for Germany up to 2050. In the first projection we do not and in the second projection we do set correlations between mortality assumptions for the different subpopulations. It turns out that setting correlations among mortality assumptions can substantially reduce projection outcomes' variance, and that it can increase projection outcomes' accuracy due to the exclusion of implausible future scenarios in each projection trial.