Changes in age-at-death distribution in low-mortality countries: a nonparametric approach

Nadine Ouellette, Université de Montréal
Robert R. Bourbeau, Université de Montréal

The mortality decrease of the last century has led to substantial modifications in the shape of the survival curve. Recent studies have suggested a number of indicators measuring these changes, which are easy to implement and offer intuitive interpretation. However, these indicators are developed in a parametric setting, which implies fairly rigid theoretical assumptions and modeling structure. In this paper, we propose a flexible nonparametric approach based on regression splines, specifically B-splines with penalties known as P-splines, to improve our monitoring of transformations in the age-at-death distribution over time. The adult modal age at death, which is strictly influenced by adult mortality, and its associated quantiles are used to measure and explain changes in adult and old age mortality. Among the various low-mortality countries studied, some are showing convincing evidence that they may have come through their phase of compression of mortality.