Multidimensional Mortality Selection and the Black-White Mortality Crossover

Elizabeth Wrigley-Field, Columbia University
Felix Elwert, University of Wisconsin-Madison

There is a gap between recent empirical work on the black-white mortality crossover and theoretical work on mortality selection. In the theoretical models, the crossover results from greater mortality selection among blacks against a single dimension of heterogeneous mortality risk. But recent empirical work implicitly assume that the heterogeneity that generates the crossover is multidimensional. Yet there is no explicit model of the crossover in the context of multidimensional mortality selection. We provide such a model and explore its properties. We show that even a simple two-dimensional mortality selection model produces unpredictable crossover timing, casting doubt on the interpretation of recent studies. The instability in crossover timing stems from the fact that, in a multidimensional setting, the counter-intuitive mortality selection dynamics that are well-known in the unidimensional setting (i.e., compositional differences created by mortality; mortality crossovers) occur at many levels simultaneously and interact in complex ways.