In this empirical analysis of voting patterns in five countries on days when one or more national referenda were held, voter turnout appears to decline in the number of concurrent referenda, in contrast to standard theories’ predictions and regardless of method used to hold constant the quality of the referenda. Multiple concurrent referenda imply ‘quantity discounts’ as one may vote on more ballots in one visit to the polling station. They should also draw more voters due to the wider range of interests attracted when more issues are up for vote. Yet, none of this seems to happen in the data. More recent developments, such as rule utilitarian and information-based theories of voting, fare similarly poorly in light of the evidence presented in this article; a social theory of voting does better.