A non-parametric approach to decompose the young adult mortality hump by causes of death

Adrien Remund, Université de Genève and Institut National d'Études Démographiques (INED)
Tim Riffe, Max-Planck Institute for Demographic Research
Carlo G. Camarda, Institut National d'Études Démographiques (INED)

We propose a new non-parametric approach to decompose the young adult mortality hump by cause of death. This novel method isolates the cause-specific contributions to the temporary deviation of the age-specific mortality rates in early adulthood, irrespective of their share in the absolute number of deaths. We test the robustness of our method using a toy example with only three causes of death. Preliminary results on the US male data from 1960 to 2010 show that accidents were initially generating a narrow and sharp hump, but from 1990 suicides and poisoning made the hump flatter and wider. The different age-specific contributions of each cause of death suggests that they may be generated by different underlying processes, which opens up new substantive lines of explanation for this yet under-studied phenomenon.