The Maze and the Mirror: Voting Correctly in Direct Democracy

type="main"> This article assesses the conditions under which voters are more likely to vote “correctly” in direct democratic ballots. We look for determinants of correct voting simultaneously at the individual and contextual levels through a multilevel approach. At the individual level we provide special attention to the level of sophistication and the use of cognitive heuristics. At the contextual level we will investigate how the nature and content of political campaigns—more specifically, their intensity and negativity—enhance correct voting. We test our empirical models in the “hard-case” scenario of Swiss direct democracy. We propose an alternative approach for the measure of correct voting, one that makes a full use of survey data but that tries to replicate the original idea of comparison between uninformed and informed voters to establish the “correctness” of a decision. We use two different data sets that cover all direct democratic votes held in Switzerland, at the federal level, between 1999 and 2005; this period covers 23 ballots, during which 75 different projects were voted. To measure correct voting and all individual variables we rely on the VOX survey data, collected on a representative sample of about 1,000 individuals after each direct democratic vote in Switzerland. To measure campaign intensity and negativity we rely on an original data set that codes the content of every campaign ad published in the six major Swiss newspapers during the month prior to each ballot. We show that political sophistication and the use of referential heuristics strongly enhance correct voting. Furthermore, at the contextual level, we show that campaigns with high intensity decrease correct voting, especially when combined with high negativity, even though the direct effect of negativity on correct voting is positive. We contribute to the existing literature in a threefold way. First, our results provide an additional confirmation that correct voting exists under both individual and contextual influences. Second, we assess the presence and determinants of correct voting in direct democracy, a “hard-case” scenario, and find that the share of citizens who vote correctly is in line with the levels found for American citizens during elections, and that the main determinants seem to hold for both elections and direct democracy. Third, our alternative approach to measure correct voting highlights that the original idea of comparison between uninformed and informed voters, usually dealt with through experimental protocols, can be replicated with survey data.