A large number of studies in labor economics estimate the returns to schooling using data on monozygotic twins, under the assumption that educational attainment is random within twin pairs. This exogeneity assumption has been commonly questioned, however, but there is to date little evidence on the topic. Using a large dataset of twins, including comprehensive information on their health status at the age of 18 and later educational attainment, we investigate whether educational attainment is related to early health status within monozygotic twin pairs. In general, we obtain no indication of this being so. As a result, we find little evidence that early health differences between twins would bias the estimates of the returns to schooling available in the literature.