Do Citizens Vote Strategically (if they vote at all)? Evidence from U.S. National Elections

Two prominent facts emerge from data on U.S. presidential and congressional elections. First, often people vote a split-ticket, that is, they vote for different parties' candidates for President and for Congress. Second, many citizens do not go to vote and, after going to vote, some decide to vote in one election but not in the other (typically more people vote for President than for Congress). This paper addresses two main questions: (1) To what extent is split-ticket voting the natural result of individuals who vote in each election according to their immediate policy preferences? In particular, what is the proportion of citizens who vote ''sincerely'' versus ''strategically''? (2) Can we simultaneously account for the patterns of abstention and voting observed in the data? To answer these questions, we propose and estimate, using individual-level data on voting choices in presidential and congressional elections from 1972 to 2000, a unified model of turnout and voting with asymmetric information. Our main findings are as follows. While a majority of citizens behave ''sincerely'', a significant proportion of citizens behave ''strategically''. Split-ticket voting is not only generated by strategic behavior but also by sincere behavior. In contrast to the implications of the ''balancing theories'', we find that a big portion of split-ticket behavior comes from extreme voters and that those middle-of-the-road voters that split their ticket do so as a result of voting according to their policy preferences in each election separately without any need for balancing purposes. We use the estimated model to conduct counterfactual experiments to assess the effect of sincere behavior, information, and abstention on electoral outcomes and the insurgency of divided governments.