Bayesian Dynamic Estimation of Mortality Schedules in Small Areas

TL;DR

This paper introduces Bayesian dynamic models to estimate and smooth mortality schedules in small populations, addressing data sparsity and instability issues, with promising results on simulated and real Brazilian data.

Contribution

It presents a novel application of relational Bayesian dynamic models for mortality estimation in small areas, improving over existing methods.

Findings

Bayesian models effectively smooth mortality estimates in small populations.

The approach outperforms recent methodologies in accuracy.

Preliminary results show promising application to real data.

Abstract

The determination of the shapes of mortality curves, the estimation and projection of mortality patterns over time, and the investigation of differences in mortality patterns across different small underdeveloped populations have received special attention in recent years. The challenges involved in this type of problems are the common sparsity and the unstable behavior of observed death counts in small areas (populations). These features impose many dificulties in the estimation of reasonable mortality schedules. In this chapter, we present a discussion about this problem and we introduce the use of relational Bayesian dynamic models for estimating and smoothing mortality schedules by age and sex. Preliminary results are presented, including a comparison with a methodology recently proposed in the literature. The analyzes are based on simulated data as well as mortality data observed in some Brazilian municipalities.