Executive Summary: "Understanding Racial Disparities in Health: The Income-Wealth Paradox" by John Mullahy, Andra T. Wenzlow, and Barbara L. Wolfe

There are large and well-documented differences between the health of non-Hispanic whites and the health of other racial and ethnic groups. In the United States, African Americans have particularly poor health outcomes compared to whites. A significant portion of these differences, however, may be due to differences in income and wealth. On average, nonwhites have lower incomes and less wealth than non-Hispanic whites and these differences appear to be growing. We examine the extent to which income and wealth “explain” racial differences in health at different points over the income-wealth distribution, comparing the self-reported health status of non-Hispanic whites with those of individuals of other races and ethnicities and with those who classify themselves as African American among individuals 25-54.

We use data from the Survey of Consumer Finances covering the years 1989, 1992, 1995, and 1998. These SCF data are used to estimate a probit model for the binary outcome “poor or fair health,” controlling for individual characteristics, transformed measures of family income and wealth, and interaction terms. Income and wealth are transformed using the inverse hyperbolic sine function (IHS) to accommodate the highly nonlinear relationship between the economic measures and health. The use of IHS permits the inclusion of nonpostive values which are found in reports of net wealth. Using our probit estimates, we compute the total difference in health between a white and nonwhite prototype, then determine what portion of this difference is explained by differences in income and wealth of whites and nonwhites at various percentiles of the income-wealth distribution. For example, a non-Hispanic white individual with 10th percentile income and 10th percentile wealth is predicted to have a 0.306 probability of reporting poor or fair health. In comparison, holding other characteristics fixed, a nonwhite individual with nonwhite 10th percentile income and wealth has a higher predicted probability of reporting poor or fair health: 0.497, or a difference of 0.191. If a nonwhite prototype is instead given the income and wealth of a white person at the10th percentile, we obtain a prediction of 0.357, accounting for 73% of the difference in reporting poor or fair health between the two types. While 73% of the race gap between poor nonwhites and poor whites is explained by income and wealth differences, only 20% is explained among wealthier individuals of each race.

            We find that the largest unadjusted racial differences in health are between poor whites and poor nonwhites, but somewhat surprisingly after adjusting for income, wealth, and other demographic characteristics, health differences between nonwhites and whites are not significant among those in the lower half of the income-wealth distribution. In contrast after adjusting for these income, wealth, and other observed characteristics, most of the smaller measured differences in health between whites and nonwhites in the upper half of the income-wealth distribution remain. These results then suggest that differences in income and wealth account for most of the health differences between nonwhites and whites in the lower half of the income-wealth distribution and that unexplained racial differences in reported health status increase with socioeconomic status among individuals aged 25–54.