One of the most consistent findings in studies of electoral behaviour is that individuals with higher education have a greater propensity to vote. The nature of this relationship is much debated, with US studies generally finding evidence of a causal relationship, while European studies generally reporting no causal effect. To assess whether the US is an exception we rely on an international dataset incorporating 38 countries, the ISSP (International Social Survey Programme) from 1985 to 2010. Both instrumental variable and multi-level modelling approaches reveal that the US is an outlier regarding the relationship between education and voter turnout. Moreover, country-specific institutional and economic factors do not explain the heterogeneity in the relationship of interest. Alternatively, we show that disenfranchisement laws in the U.S. mediate the effect of education on voter turnout, such that the education gradient in voting is greater in U.S. States with the harshest disenfranchisement legislature. As such, the observed relationship between education and voting is partly driven by the effect of education on crime.