The work is devoted to the estimation of the average treatment effect of higher education on health, assuming the existence of structural differences in the dependencies between the studied factors by gender and the birth cohort of individuals. The effects are estimated separately for each cohort of men and women using parametric (multivariate recursive probit model) and nonparametric (matching) methods. The use of fundamentally different approaches makes it possible to verify the robustness of the results to violating the assumptions of the applied methods. During the study the statistical evidence in favor of the existence of significant negative average treatment effects of higher education on the probability of obesity and the probability of hypertension and significant positive effects on the likelihood of high self-esteem of health was revealed for women. For men, the statistical evidence in favor of the presence of positive average treatment effects on the probability of having heart diseases, hypertension, allergy, and high self-esteem of health was found. A significant differentiation of the average treatment effects depending on the age cohort was revealed. For certain types of diseases, the cohort diminishing hypothesis for education is confirmed. The conclusion is made about the necessity of studying the impact of education on individuals’ health separately for men and women born in different time periods, as well as the inexpediency of using aggregated health indicators when analyzing the impact of education on health. The results of the study allow determining the nature of cohort changes in the impact of education on health of the Russian population and can be taken into account when forming state policy on education as a source of positive external effects on health.