Probabilistic Population Projection for Brazil Subnational Areas

Felipe Souza, UFRN
Flávio Freire, UFRN - Universidade Federal do Rio Grande do Norte
Everton Lima, Unicamp
Marcos Gonzaga , UFRN

This study presents probabilistic population projections for two Brazilian states—São Paulo and Rio Grande do Norte—based on hierarchical Bayesian models following the methodology adopted by the United Nations. The objective is to evaluate their applicability to the Brazilian subnational context by estimating 80% prediction intervals for population size by sex and age group from 2025 to 2070. Projections were produced using the cohort-component method, with a hierarchical Bayesian approach applied to fertility, mortality, and migration through the bayesTFR, bayesLife, and bayesMig packages in R, respectively, and integrated via the bayesPop framework. Data from 1980–2023 were obtained from the Brazilian Institute of Geography and Statistics (IBGE) and the Ministry of Health’s DATASUS systems (SIM and SINASC). Adjustments were made for birth and death underregistration, and net migration rates were derived from census data and interpolated for intermediate five-year periods. Results indicate strong convergence between the Bayesian median projections and IBGE’s deterministic projections, with the added advantage of quantifying forecast uncertainty. In São Paulo, population decline is expected between 2035 and 2040, with accelerated aging leading to about 315 elderly per 100 young people by 2070. In Rio Grande do Norte, demographic change is even more intense, with population decline beginning in the same period and an aging index projected at 386 elderly per 100 young people by 2070. These results underscore Brazil’s rapid demographic transition and the usefulness of Bayesian methods for generating realistic and policy-relevant uncertainty intervals in subnational population projections.

See paper

 Presented in Session P8. Demographic Trends, History, Data and Methods