Effects of the Affordable Care Act on black-white health disparities in the United States: A nonparametric approach

This paper examines the effects of the Affordable Care Act (ACA) on health disparities between blacks and whites in the United States using a nonparametric decomposition method. The Blinder-Oaxaca (B-O) decomposition method (1973) has been widely used to explain the difference in means of a dependent variable between two groups by decomposing the gap which can or cannot be explained by the difference in characteristics of each group. The unexplained gap may be partly caused by discrimination. However, there are two major concerns about the B-O decomposition method. One concern is the so-called “index number problem”, which means that the analysis results are sensitive to the choice of the reference group in decomposition. The other concern is that the results may vary with the choice of covariates used in the decomposition. Since estimation of the underlying equation is required to conduct the decomposition analysis, an omitted variable bias cannot be avoided as well. The B-O method implicitly assumes that the estimators are valid for all the individual characteristics in case two groups may not be comparable, resulting in overestimation of the component of the gap attributable to the difference in the rewards. Alternative methods have been developed recently using either nonparametric or semiparametric decomposition methods in analyzing the gender wage differences. (DiNardo et al (1996), Nopo (2004), Anspal (2015)). In this study, we adopted a nonparametric decomposition method, originally proposed by Nopo (2004). Using the National Health Interview Survey (NHIS) 2010-2016 pooled cross-sectional data, we examine the impact of healthcare enrollment on health disparities by comparing the pre-ACA period (2011-2013) and the ACA-period (2014-2016). NHIS from the Centers for Disease Control and Prevention is the nation's largest in-person household health survey, collected by US Census since 1957. We find large unexplained gaps in health status between blacks and whites, and we demonstrate that while one important factor – healthcare – appears to explain some portion of the gaps, there remains a huge unexplained portion. Decomposition results using the nonparametric estimation are consistent with the parametric estimates. However, the nonparametric analysis results via matching should be substantially more accurate. Advantageous whites who more likely to have high income, high education, an the like, seemed to be in the out-of-support range and were dropped after matching. The results also show the importance of comparing the individuals on support. Otherwise, one might overestimate the explained gap by assuming all individuals are comparable.