According to recent studies educational inequality in health is widening in the United States, but reasons are still not clear. This study examines to what extent widening educational differences in health in the United States can be attributed to changes in marital status composition of the low and high educated groups over cohorts. Therefore, using repeated cross-sectional data from the National Health Interview Survey 1976-2014 we consider how the rising educational inequalities in health can be explained by the changing selection into different marital categories based on education. First, we examine how the changing heterogeneity in educational levels within various marital categories influences the observed relationship marriage-health over cohorts and how he changing heterogeneity in marital status within different educational groups influences the observed relationship education-health over cohorts. Then, we decompose the health differences between less-than-high-school and  college educated individuals using Oaxaca-Blinder method, and we calculate the proportion attributable to marital status in different cohorts. The results show a sharp increase over cohorts in the proportion of educational differences in health explained by marriage. We believe this analysis highlights the need for a greater attention to the changing demographic characteristics in educational categories when looking at trends over time.