This paper estimates the health returns to education, using data on identical twins. I adopt a twin-differences strategy in order to obtain estimates that are not biased by unobserved family background and genetic traits that may affect both education and health. I further investigate to what extent within-twin-pair differences in schooling correlates with within-twin-pair differences in early life health and parent-child relations. The results suggest a causal effect of education on health. Higher educational levels are found to be positively related to self-reported health but negatively related to the number of chronic conditions. Lifestyle factors, such as smoking and overweight, are found to contribute little to the education/health gradient. I am also able to rule out occupational hazards and health insurance coverage as explanations for the gradient. In addition, I find no evidence of heterogenous effects of education by parental education. Finally, the results suggest that factors that may vary within twin pairs, such as birth weight, early life health, parental treatment and relation with parents, do not predict within-twin pair differences in schooling, lending additional credibility to my estimates and to the general validity of using a twin-differences design to study the returns to education.