Estimating Level and Structure of Mortality in Small Areas: Counterfactual Approaches

Everton Lima, Cedeplar, UFMG
Flávio Henrique M. de A. Freire, Universidade Federal do Rio Grande do Norte
Diana Sawyer, Cedeplar, UFMG
Josivan Justino, Universidade Federal do Rio Grande do Norte

In this paper we aim to study the quality of mortality data for small areas in Brazil and estimate the level and structure of mortality using alternative and comparable approaches. The main idea is to compare different approaches of mortality estimation, in order to get a range of estimates of mortality schedules for small areas. We applied standard demographic methods (Death Distribution Methods) and empirical Bayesian (EM) statistics and Expectation Maximization (EM) algorithm to produce estimates of levels and structure of mortality in two states in Brazil. We then compare the estimates produced using the two methods to investigate the performance of both for less populated areas. Our results indicate large the spatial heterogeneity with respect the coverage of death counts across small areas. Furthermore, it is important to correct for under-registration of death counts before any estimate of mortality is applied.