Modelling Human Mortality Trajectories via an Accelerated-Ageing Hazard Function

Elisabetta Barbi, Max Planck Institute for Demographic Research
Francesco Lagona, Università Roma Tre

We suggest an accelerated-ageing hazard function for the analysis of human mortality. Beside the heterogeneity captured by the relative-risks frailty model, a second source of heterogeneity should be addressed, due to possible individual differences in the rate of ageing. Unfortunately, a random term acting multiplicatively on age makes nor the predictive distribution of frailties neither the probabilities of death at the population level available in closed form. To overcome the problem, we suggest and implement a Monte Carlo E-M algorithm. The accelerated-ageing and the relative-risks models have been estimated on a Swedish data set. Although both the models seem reasonable and give significant estimates, the accelerated-ageing model seems to give a better fit compared to that given by the relative-risks model and gives results that might match better with the observation of extreme ages at death that have never been captured by other mortality models.