What Does the Evidence Show about the Shape of Mortality at the Oldest Ages?

Dennis Feehan, Princeton University

Many important theoretical and policy questions require an understanding of the trajectory of death rates at advanced ages. Unfortunately, studying mortality at advanced ages can be very difficult: since absolute numbers of deaths are typically very small, we must be careful to distinguish empirical regularities from stochastic noise. In this abstract, I present an analysis of the best available data on mortality above age 80 with the aim of understanding what we can conclude about the shape of the hazard function over that range. I also review and evaluate the options available for assessing the quality of a hazard model's fit to data. This problem has been discussed at length in the literature, and is of importance to researchers and policymakers in a host of fields. I expand on previous analyses by adding new data, considering more functional forms, and improving the strategy for choosing which model provides the best fit.