Clustering EU countries based on death probabilities

Abstract

Our research is conducted to identify certain grouping of 24 European countries based on their death probabilities. Gathering 2014 data from Human Mortality Database our research objective was twofold. First, we wanted to find homogeneous groups of countries where mortality is similar and for a financial institution they could be grouped as risk communities. Second, we wanted to identify the optimal number of groups as a basis for strategy making. Two different clustering methods were used in our research, k-means and k-median clustering. We applied asymmetric measure (QDEV) in k-median method to handle the differences in country sizes and age groups. Our results are stable but different in k=3 clusters, k-means clustering resulted in a big Western-European cluster and two small-medium Eastern groups; however, k-median clustering gave a homogeneous Eastern group and besides a bigger Western cluster Spain, Italy, and France formed a separated group of countries.