The concentration index and decomposition analysis are commonly used in economics to measure and explain socioeconomic inequalities in health. Such analysis builds on the strong assumption that a health production function can be estimated without substantial bias implying that health is caused by socioeconomic outcomes, which is hard to prove.
This article contributes to the decomposition literature by applying a twin design to standard decomposition analysis of socioeconomic health inequalities in Sweden. The twin-based decomposition estimates, which control for unobserved endowments at the twin-pair level, are much lower in magnitude than estimates obtained via typical OLS on the same sample. This demonstrates that OLS-based decompositions are severely upward biased due to underlying confounders, exaggerating the contribution of income and education to health inequality, which in turn limits the usefulness of such decompositions for policy purposes.
Scandinavian Journal of Economics 2015; 118(1)M: 25-48