This paper analyzes the relationship between education and health outcomes using a natural experiment in Turkey. The compulsory schooling increased from 5 to 8 years in 1997. This increase was accompanied by a massive construction of classrooms and recruitment of teachers in a differential rate across regions. As in previous studies, we confirm that the 1997 reform substantially increased education in Turkey. Using the number of new middle school class openings per 1000 children as an intensity measure for the 1997 reform, we find that, on average, one additional middle school class increases the probability of completion of 8 years or more of schooling by about 7.1 percentage points. We use this exogenous increase in the educational attainment to investigate the impact of education on body mass index, obesity, smoking behavior, and self-rated health, as well as the effect of maternal education on the infant's well-being. Using ordinary least squares, we find that there is a statistically significant favorable effect of education on health outcomes and behavior. However, this relationship becomes insignificant when we account for the endogeneity of education and health by instrumenting education with exogenous variations generated by the 1997 reform and the accompanying middle school class openings. The insignificance of the health effect may be due to lack of statistical power in our data, or to the fact that this policy affects only relatively low levels of schooling and the health effects of education need to be examined at higher levels of schooling.