Population Association of America Conference

Atlanta, Georgia

April 8, 2022

Why Your Microsimulation Misses the Mark: Identifying and Solving Sources of Divergence Between Microsimulation and Cohort Component Models of Population Change

Rachel Bacon, Center for Mind and Culture
George Hodulik, Center for Mind and Culture
David Voas, University College London
Ivan Puga-Gonzalez, Center for Modeling Social Systems
Wesley J. Wildman, Center for Mind and Culture

Abstract: As computer capabilities expand, simulations can model larger populations of artificial agents that cover both past and future periods. It is intuitive to use widely available demographic rates for artificial agents’ risk of childbirth, death, or migration when designing a simulation. Ignoring the underlying assumptions of the cohort component method (CCM) model associated with those demographic rates can result in a divergence in expected births, deaths, and total population. We demonstrate an alternative simulation design that reduces divergence by splitting the fertility process into two parts, while relying on the same demographic data that the United Nations provides for their CCM estimates and projections. Our results show that selecting the order of demographic events for a simulation is sensitive to assumptions associated with demographic rates and sources. Simulation designers who aim to track standard population estimates will benefit from an understanding of CCM projection methods.