Subnational Fertility Projections in Brazil – a Bayesian Probabilistic Approach Application

Gabriel Borges, Instituto Brasileiro de Geografia e Estatística (IBGE) and University of California, Berkeley

The main objective of this paper is to project fertility rates for the subnational level in Brazil using the UN’s Bayesian probabilistic approach in order to compare with the most recent fertility information and with the deterministic projection released by IBGE. Results show that out-of sample projections systematically overestimate the TFR, since there was an acceleration of fertility decline in the 2000s. Social advances observed in this decade might have influenced such rapid fertility decline. For projections with starting point in 2010, IBGE’s estimations still present lower figures than those projected by the Bayesian model, in addition to a different convergence pattern in 2030. Bayesian approach offers a promising alternative to subnational level projections in Brazil. However, some adjustments may be added, like the incorporation of age-specific fertility rates and some covariates related to fertility, in addition to use regions that have closer characteristics to the Brazilian context.