We discuss some issues associated with the empirical analysis of the relationship between socioeconomic status and health. We point out that, in addition to elaborate empirical modeling and good data, a conceptual framework is helpful both for making sense of one's own results and for the purpose of reconciling results across studies. We find that when we align the empirical specification with the Grossman model, a negative effect of income on health emerges. Even though this unexpected finding can be rationalized, we think that some caution regarding standard dynamic panel data techniques is warranted in this context.