Decomposing social indicators using distributional data

Are the poor less healthy? Does public health spending matter more to them? We decompose aggregate health indicators using a random coefficients model in which the aggregates are regressed on the population distribution by subgroups, taking account of the statistical properties of the error term. This also allows us to test possible determinants of the variation in the underlying subgroup indicators. The approach is implemented with data on health outcomes and poverty measures for 35 developing countries. We find that poor people have appreciably worse health status on average than others, and that differences in public health spending tend to matter more to the poor.