Testing for treatment effect heterogeneity: Educational reform, genetic propensities, and family background

We study the heterogeneous effects of a Swedish educational reform that increased compulsory schooling from seven to nine years. Following a pre-registered Analysis Plan, we examine how the reform differentially affected 20 outcomes in various domains (education, the labor market, fertility, substance use, and health) for individuals with different genetic propensities towards these outcomes. To estimate the effects of the reform, we leverage the gradual roll-out of the reform with a differences-in-differences estimator that compares municipalities and birth cohorts before and after the reform. Among 105 pre-registered tests of heterogeneous effects as a function of genetic propensities (across outcomes, sexes, and family SES background), we find two significant interactions after correction for multiple hypothesis testing. One of these implies that, among females from high-SES families, the reform had a relatively larger effect on earnings for those with a lower genetic propensity for educational attainment. Our results suggest that sizable heterogeneous effects of educational reforms may not be ubiquitous but do exist, and highlight how careful study designs with preregistration can help uncover them.