Voting Error across Multiple Ballot Types: Results from Super Tuesday (2020) Experiments in Four American States

Abstract

This paper reports findings on voting error from online experiments involving real presidential candidates on hypothetical ballots. Over 6,000 respondents in four American states in February and March of 2020 were given the opportunity to vote in a simulated Democratic Party presidential primary and a simulated blanket (all-party) presidential primary. Experimental subjects were randomly assigned one of three different voting methods for each contest: a traditional single-mark (a.k.a. “exclusive” or “categorical”) ballot, a ranking (a.k.a “ordinal”) ballot, and a grading (a.k.a. “evaluative” or “range”) ballot. Results show that, on the most basic and widespread measure of voting error, the traditional ballot produced more void (a.k.a. “residual” or “rejected”) votes than the reform alternatives. At the same time, the rates of mismarked votes, or those with at least one violation of the ballot’s instructions, were higher for the reform alternatives than for the traditional ballot. Though the opportunity to mark each and every candidate on ranking and grading ballots generated more violations of the instructions, it also enabled more voters to express a clear, countable judgment of at least one candidate. Other results suggest that group-based inequalities in voting error – measured by the discrepancies in void rates across age, gender, and race cohorts – are smaller with reform ballots. These findings cast doubt on the conventional wisdom that single-mark voting is inherently “natural” or “intuitive” for all voters, irrespective of personal traits, group identities, or socialization. They also defy the presumed theoretical trade-off, or zero-sum game, between more expressive ballot types and more accessible (easy-to-use) ones.