Precise modeling of aging populations

Abstract

Aggregate continuous time formulations used in system dynamics models result in the implicit mixing of individual constituents of levels. Normally, this is acceptable as the heterogeneous nature of model variables implies that some individuals counted in a level's value will pass others and exit earlier. In models with a focus on chronological aging and age-related characteristics, this phenomenon, which we call cohort blending, can result in large distortions. Using aging chains can reduce these distortions, but cohort blending persists even with one-year groupings. As an alternative, we introduce an approach we call "continuous cohorting" in which populations are tracked with cohorts sized the same as the computational interval of the model. This approach eliminates the blending problem with minimal notational and moderate computational burden. The resulting models display quantitative, though limited qualitative, differences from their traditional counterparts and are more easily defensible with demographers and others who study population dynamics. © 2013 System Dynamics Society.