Clustering mortality data over time and across countries

Pedro Menezes de Araujo , University College Dublin
Isobel Claire Gormley, University College Dublin
Thomas Brendan Murphy, University College Dublin
Ugofilippo Basellini, Max Planck Institute for Demographic Research

Cluster analysis applied to mortality data is useful for various tasks, such as improving mortality forecast accuracy, constructing model life tables, and uncovering patterns in mortality profiles across countries, including inequality and divergence phases. Most existing approaches either cluster country–year observations or cluster countries alone. In this work, we propose a framework to cluster mortality patterns over time as well as country groups. We first cluster the mortality curves for all countries and periods (which we call mortality levels), and then cluster the countries based on their mortality level dynamics over time, linking both analyses. This allows the identification of countries that have followed similar mortality developments, helping to describe mortality stages, patterns of convergence and divergence, and inequality trends. To illustrate our method, we use period life table data from the World Population Prospects, covering 71 countries from 1960 to 2019. Our method identifies diverse mortality levels and country clusters, with inequality increasing until the 1990s and decreasing more recently. Overall, we provide an integrated approach to describe mortality evolution across time and countries, contributing to a better understanding of global mortality dynamics and their changing inequalities.

See extended abstract

 Presented in Session 78. Assessing and Improving the Quality of Mortality Data