Compositional changes in educational groups as an explanation for widening differences in health

Cecilia Potente, University of Oxford
Christiaan W. S. Monden, University of Oxford

Recent studies have reported a widening gap in health and mortality between low and high educated individuals in the United States. Several reasons such as differential smoking and obesity trends have been put forward to explain these increasing differences. However, changes in composition of the educational categories over time may also partially explain these differences. Educational expansion gave a higher portion of individuals from diverse background access to education, while the low educated may have become a more (negatively) selected group over time. Using height as a proxy for early life socio-economic status and health, we examine how changes in average height and height dispersion within educational groups over birth cohorts have contributed to the widening gap in health. We use the 1976-2014 National Health Interview Surveys which contain height and health information for samples of individuals in the United States. We limit the analysis to those aged 30-45 to avoid confounding effects due to mortality. Our analytic sample includes the 1928-1984 birth cohorts. First, we test whether the association between height and education has changed across cohorts. Second, through Oaxaca-Blinder decomposition we examine which factors help us explain the widening gap focusing on differences in mean height and variance. This analysis allows to establish whether the lower educated group increasingly consists of negative selected individuals leading to the broadening gap. Preliminary results provide evidence for a notable fall in average height within the less than high school educated group over consecutive birth cohorts. Therefore, they offer some preliminary support to the hypothesis of rising negative selection within the low educated group over cohorts.