Research on invalid voting has expanded over the past few years. Scholars largely agree on the most important determinants of invalid voting, foremost compulsory voting, quality of democracy, and ballot structure. However, disagreement prevails concerning several factors; more exactly, whether the effect is positive or negative. We examine determinants of invalid voting in the four most recent local elections in a single country, Sweden, all municipalities included. Applying a subnational research design allows us to control for several determinants that have been reported to be of significance in explaining varying levels of invalid voting, and, consequently, to examine factors with theoretically contradictory directions of effects. Two theoretical models of invalid voting are tested, an institutional and a societal model, each consisting of four variables. The latter explains considerably more of the variation in the dependent variable than the former. The most important determinant is education: higher levels of education among the population decrease the share of invalid votes. Concerning the institutional model, only district magnitude significantly affects the dependent variable, indicating that larger district magnitude depresses invalid voting. However, the effect disappears when societal variables are included, due to higher levels of education among the population in municipalities with high district magnitudes. Proportion of the electorate born abroad also significantly affects levels of invalid voting; higher proportions decrease invalid voting. We suggest that this is a consequence of political sophistication, interest, and being politically informed combined with who decides to vote and who decides to abstain amongst immigrants.