Hybrid Format

Evaluation of Small-area Estimation Methods for Mortality Schedules

Esther Denecke
Laboratory of Digital and Computational Demography, October 05, 2022

Esther Denecke from the Laboratory of Digital and Computational Demography aims to provide guidance to choose the best method to estimate mortality by reviewing demographic methods and evaluating some of them on real and simulated data.

Abstract

There is a growing demand for estimating mortality in small populations. In small populations, however, death counts are erratic. To address the problem of unstable mortality estimates, demographers have developed Bayesian and frequentist methods. While some of these methods have been compared to others, it remains unclear which method to choose for practical implementation and under which circumstances. She aims to provide such guidance by reviewing demographic methods and evaluating some of them on real and simulated data. For the practical evaluation, she focuses on a Bayesian hierarchical model based on principal components, frequentist TOPALS regression and Demographic-Splines. She is specifically interested in usability, data requirements, and the performance in different situations and under different assumptions. Preliminary results show that the two frequentist methods which estimate mortality schedules independently for each area indicate larger gaps in Austrian life expectancy than the Bayesian model which smooths over space and time.

About

Esther Denecke is currently a PhD student in the Lab of Digital and Computational Demography here at the institute. She is also affiliated with the University of Rostock. Her main research interests include the estimation of mortality indicators, especially in small areas/small populations. Methodologically, she is especially interested in exploring the circumstances under which methods work well or fail to produce reliable results.