Bayesian Population Estimates: Accounting for Uncertainty in Census and Vital Registration Data

Population projections and estimates are important for many purposes, including formulating, implementing and evaluating public policies. There have been rapid advances in implementing probabilistic methods to produce population projections. Population estimates for the past and current years are mostly deterministic. Considering that in most countries demographic data are limited due to coverage and quality problems, it becomes relevant to account for measurement erros for these estimates. This paper proposes a Bayesian framework that uses the cohort-component method to produce annual estimates, accounting for uncertainty in census and vital registration counts. The approach proposed in this paper is applied to population estimates in Brazil from 2010 to 2022, but it is flexible enough to be used in other contexts.