Abstract: Demographic analysis of the age-specific components of life tables is almost entirely endogenous. If exogenous variables such as smoking or wealth are considered, it is usually in relation to a single indicator such as period life expectancy at birth or infant mortality. A classic example is the model of life expectancy at birth as a function of time and GDP p.c. by Preston (1975). To explore the effect of covariates across all ages, this paper uses a method from Compositional Data Analysis (CoDA) that allows one to treat life table densities of deaths by age as dependent entities in a regression model. The regression framework offers access to a very rich array of techniques, including Difference-in-Difference analysis. The method is illustrated by an analysis of potential explanatory variables associated with the age-specific changes in US State life tables across the years before (1980-89) and during (2010-18) the opioid crisis leading to Deaths of Despair. Potential applications to non-human life tables will be mentioned.
The meeting will be held as a physical seminar in the Institute of Biology seminar room, but it will also be possible to participate online via Zoom.