The Dynamic of Mortality Explained with a Reduced Number of Key Ages

Abstract

Fitting and forecasting the age pattern of mortality has been of interest to researchers and practitioners for a century and a half. In fact, many researchers have focused on the decomposition of the life length into homogeneous age groups for which a similar pattern of mortality can be discerned, and their secular dynamics. In this paper, we propose to use a mortality improvement rate model where the changes in mortality rates are linearly related to changes in three target rates corresponding to three “key ages”. These ages represent age groups where mortality evolves according to a specific pattern. Thus, we decide to fit and forecast mortality for the entire age range using these three key ages, chosen based on their demographic meaning. The key age model is fitted to seven populations and genders, and the in-sample and out-of-sample outcomes are compared to two mortality models used as benchmark. Despite its small number of parameters, the model shows good in-sample and out-of-sample accuracy.

Keywords: Mortality Improvements, Key Age, Mortality Curve, Mortality Patterns, Cross-Validation.