Score contribution per author:
α: calibrated so average coauthorship-adjusted count equals average raw count
Abstract We test the turnout predictions of the canonical costly voting model through a large-scale, real effort experiment. We recruit 1200 participants through Amazon’s Mechanical Turk and employ a $$2\times 2$$2×2 between subjects design encompassing small ($$N=30$$N=30) and large ($$N=300$$N=300) elections, as well as close and lopsided. As predicted, participants with a higher opportunity cost are less likely to vote; turnout rate decreases as the electorate size increases in lopsided elections and increases the closer the election is in large elections. However, in the large lopsided election the majority turns out to vote at a higher rate than the minority. We rationalize these results as the equilibrium outcome of a model in which voters obtain a small non-monetary utility if they vote and their party wins.