Does poverty mask an earlier race crossover in mortality in the United States?

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

The aggregate death rates of blacks and whites in the United States “cross” in old age. Recent work on the race crossover emphasizes frailty explanations, which posit that compositional differences can produce a mortality crossover in the aggregate, even if individual-level mortalities do not cross. Conventional wisdom holds that adjusting for compositional differences should increase the age at crossover, possibly beyond the human life span. We object that compositional adjustments are more likely to reduce than to increase the age at crossover: e.g., accounting for the greater prevalence of poverty among blacks should lower black mortality relative to white mortality, thus reducing the age at crossover. We test our hypothesis with covariate-rich longitudinal data on 28 million elderly individuals (U.S. Medicare claims data, 1993-2002). Flexible Poisson models indicate that, net of observed compositional differences, the race crossover occurs several years earlier. Implications for frailty theory are explored.