Demographic Perspectives on Predicting Individual-level Mortality

There are striking disparities in longevity across sociodemographic groups in the United States. Yet, can sociodemographic characteristics meaningfully explain individual-level variation in longevity? Here, we leverage machine-learning algorithms and large-scale administrative data to predict individual-level mortality using an array of social, economic, and demographic predictors measured in early adulthood. We conduct two distinct analyses: a cohort analysis, which predicts the exact age of death for individuals in the same birth cohort, and a period analysis, which predicts whether individuals age 54–95 will die within the next 10 years. We are not able to make accurate predictions in either our cohort analysis (R2= 0.014) or our period analysis (R2= 0.166).Together, these analyses demonstrate that later life longevity is unpredictable using sociodemographic characteristics alone, and underscore the crucial need to account for stochastic processes in demographic theory