Research has shown that education is positively correlated with political party membership at the individual level. It is thus puzzling that increased education at the aggregated level in most Western countries has not resulted in an aggregate increase in levels of party membership. One explanation for this paradox is provided by the sorting model of education, according to which there is no direct effect of education on political participation; education affects individuals’ social network positions, which in turn affects political participation. Prior research on the sorting model has focused on the observable predictions derived from the model. The hypothesized causal mechanism, i.e. social network position, has not been sufficiently tested. This article employs Swedish data with comprehensive measures of social relations and utilizes structural equation modelling to test the hypothesized causal relationship. The results confirm that social network position mediates the effect of education on active political party membership.