Lower levels of education are associated with higher mortality. Lower levels of education are also associated with lower income, which is also associated with higher mortality. We investigate the impact of education on mortality and the extent to which this is mediated through income over the life course. We account for both selective educational attainment and selective income over the life course, through inverse propensity weighting (IPW) of the mortality hazard. We decompose the educational gain, that is, the decrease in mortality from more education, in the hazard into an indirect effect of education, running through changes in income and a direct effect of education, running through other factors. We use Swedish conscription data (men only), linked to parental information and individual annual income for the period 1968 till 2012. Our empirical results indicate large educational gains in mortality. We also find that this educational gradient runs through changes in income, especially for the more educated, and does not run through other factors related to education. We conduct several robustness and sensitivity checks that indicate that the results are robust.