Several studies have found a positive association between education and health. Confounding factors that a ect both education choices and health, such as (ob- served) parental background and (unobserved) intelligence, may play an important role in shaping this association. In this paper we estimate the impact of education on diseases in old age, accounting for this endogeneity. Our estimates are based on administrative data on men born in 1944–1947, who were examined for military service in the Netherlands between 1961–1965, linked to national death and medication use records. We assume medication use identifies diseases. We estimate a structural model, consisting of (i) an ordered probit model for the educational attainment, (ii) a Gompertz mortality model for survival up to old age, (iii) a probit model for medication use in old age and, (iv) a measurement system using IQ tests to identify latent intelligence. Educational choices, surviving up to old age and medication use all depend on observed individual factors and on latent intelligence. Based on the estimation results, we derive the impact of education on diseases in old age. Our empirical results reveal a strong effect of education on physical diseases, but low or no effect of education on depression and anxiety.