Simultaneous Modeling of Heterogeneous Subpopulations within one Framework

Demographic forecasting models simulate the likely future development of a population with assumptions in mortality, fertility, and migration.
If demographic behavior is heterogeneous in a population,
projection accuracy can be affected substantially.
To overcome this problem,
we propose a new general framework that can be used to disaggregate heterogeneous populations in as many homogeneous subpopulations as needed that can then be projected with separate assumptions in mortality and fertility.
The advantage of our general framework is that subpopulations cannot only be constructed with typical characteristics like age and sex,
but also with other characteristics like country of origin or level of education.
We apply this general framework in projections for Germany up to 2050 in order to show the impact of expected demographic heterogeneity on projection outcome.
In the first projection, we use the variables age and sex to construct subpopulations,
and in the second and third projection, we add the variables migration status and reason to migrate.
As expected, our preliminary results indicate that projection error accumulates over time,
and that it is higher for the first than for the second projection.
We also discuss how this general framework can be easily applied to conduct multiregional projections.