Modeling and forecasting age at death distributions

Ugofilippo Basellini, Max-Planck Institute for Demographic Research
Carlo G. Camarda, Institut National d'Études Démographiques (INED)

Age at death distributions provide an extremely informative description of mortality, yet they are generally neglected in modeling and forecasting. In this article, we use age at death distributions to model the age-specific pattern of mortality and to inform mortality forecasts. In particular, we introduce a segmented linear transformation model based on the modal age at death and the variability of deaths before and after the mode. This approach allows capturing the compression and shifting dynamics of mortality. We illustrate our methodology by estimating the distribution and life expectancy of two high-longevity countries in the last thirty years. We show that the fitted life expectancies are very close to the observed historical values. Furthermore, we forecast distributions and life expectancies fifteen years ahead by using time series models for the parameters of the segmented linear transformation.