Using quantile regression to identify longevity thresholds

Nicola Tedesco, Università degli Studi di Cagliari
Luisa Salaris, Università degli Studi di Cagliari

From the review of longevity studies emerges that there is not an agreement on age-threshold for the identification of long-lived populations and individuals. In general, longevity thresholds can be classified into two large groups, namely “fixed" and "relative" threshold. The first are identified in correspondence of specific ages and its choice may depends on specific research questions and/or may be instrumental to the identification of the population understudy or of specific population subgroups, that accordingly to the selected cutoff age are classified as long-living or not. The applications of “relative” thresholds are the same of the fixed ones, but what changes is the procedure according to which the cutoff age is chosen as the identification of the longevity threshold occurs according to the distribution of deaths and its cumulative percentages. Both longevity thresholds prove to have strengths and weaknesses. However, there are two aspects that deserve special attention: i) survival experience of a population along the entire life cycle can differ from another, despite for example reaching similar level of survival at older ages; ii) when analyzing differential mortality it could be useful to think in terms of population selection, devoting attention not exclusively to the robust component, but also to frail individuals. The questions that arise are numerous: why some people died earlier than others? Which variables are involved in the selection process? Do the estimated effect of variables vary accordingly to the longevity threshold chosen? In the attempt to give an answer, this paper proposes the use of Quantile Regression Models (QRM) as a useful method for the identification of longevity threshold as they allow to examine the evolution of survival and, in the meantime, to check the effect of covariates. The use of QRM is here applied to the study of Villagrande Strisaili (Italy) population.

  See extended abstract

Presented in Session 103: Retirement and ageing