We investigate assortative mating on education using a sample of couples from the Health and Retirement Study. We estimate a reduced-form linear matching function, which links wife’s education to husband’s education and both wife’s and husband’s unobservable characteristics. Using OLS we find that an additional year in husband’s education is associated with an average increase in wife’s education of 0.4 years. To deal with omitted variable bias due to unobservable characteristics, we use a measure of genetic propensity (polygenic score) for husband’s education as an instrumental variable. Assuming that our instrument is valid, our 2SLS estimate suggests that an additional year in husband’s education increases wife’s education by about 0.5 years. Since greater genetic propensity for educational attainment has been linked to a range of personality and cognitive skills, we allow for the possibility that the exclusion restriction is violated using the plausible exogenous approach by Conley et al. (2012). ‘True’ assortativeness on education cannot be ruled out, as long as one standard deviation increase in husband’s genetic propensity for education directly increases wife’s education by less than 0.2 years.