An Empirical Test of Duncan Black's Median Voter Theorem.

A unique Condorcet winning alternative exists when voting is over a single dimension and the voter's preferences are single-peaked. The generality of voting through binary agendas can be tested directly to determine the frequency a Condorcet winning alternative is chosen when voting is by pure majority rule. In this setting, voters were given single peaked preferences and a binary agenda to compare a sequence of proposals. The sample consists of 1358 votes on a public good dimension, modeling whenever the electorate has a quasi-transitive ordering over the alternatives. The findings reveal the voter's understate their support for preferred alternatives, while voting consistently over these alternatives. The results simulate public opinion data examining the effects on support for candidates and political parties on the division of votes, partisan bloc voting, and the distribution of political views in Presidential elections from 1968 to 2004. The findings strongly indicate the robustness of the median voter's theorem for explaining what happens in national general elections. Because the findings are explaining divisions of votes and predict bloc voting, the results estimated for voter preference revelation are consistent with polling data on self-reported ideology more so than partisanship. However, the evenness of the divisions of the votes and replication of both partisanship and partisan competition demonstrate the implications of a model of pure majority rule with fixed voter preferences, varying small numbers of alternatives, and voting through a binary agenda. ..PAT.-Unpublished Manuscript [ABSTRACT FROM AUTHOR]