We go beyond estimating the effect of college attainment on longevity by uncovering the mechanisms behind this effect while controlling for latent skills and unobserved heterogeneity. We decompose the effect with respect to a large set of potential mechanisms, including health behaviors, lifestyles, earnings, work conditions, and health at the start of the risk period (1993–2017). Our estimates are based on the Wisconsin Longitudinal Study and show that the effect of education on longevity is well explained by observed mechanisms. Furthermore, we find that for women, the positive effect of education on longevity has been historically masked by the negative effect of education on marriage. An adjustment for the relationship between education and marriage based on data for more recent cohorts increases the explained effect of education on longevity for women. We discuss the implications for policies aimed at improving health and longevity and reducing health inequality.