Estimating and explaining differences in income related inequality in health across general practices.

We use data on individual patients in general practices to examine whether income related inequality in self reported health differs across general practices and whether such differences are explained by characteristics of the practices. We allow for the simultaneous determination of health and income by instrumenting income. We also allow for item non response for the income question by a two stage selection model. We find that item non response has little effect on the estimated relationship between income and health but that allowing for simultaneity doubles the estimated effect of income on health. We show that there are significant differences in the effect of income on health across practices and that these differences are related to the number of patients per GP, a measure of practice prescribing quality, and the provision of out of hours services.