My recent contributions to public choice

Southern Economic Journal, July, 2008 by William Niskanen

American voters, in their unarticulated collective wisdom, have voted for a divided government for most of the past 50 years. Divided government may not be the stuff of which political legends are made, but the separation of powers is probably a better protection of our liberties when the presidency and at least one house of Congress are controlled by different parties.

5. A Reformulation of Voting Theory

My article "A Reformulation of Voting Theory" was published in 2007 in a conference volume on Public Choice and the Challenges of Democracy (edited by Jose Casas Pardo and Pedro Schwartz) and is the outgrowth of my empirical article on the 2002 congressional elections that was published in 2004.

Our standard theory of voting behavior--the core of public choice--is a mess! The theory of voter behavior is asymmetric with the theory of candidate behavior. More important, the theory does not explain some of the more important recent changes in the outcomes of American elections. My paper summarizes a major problem of the standard theory of voting behavior, develops an alternative theory based on a joint determination of voter behavior and candidate behavior, and presents some evidence from recent elections that is more consistent with the alternative theory.

The standard theory of voter behavior is dramatically asymmetric with the theory of candidate behavior. Voters are assumed to make a joint determination of whether to vote and for whom to vote, based on their understanding of the issue positions of the alternative candidates. Candidates, in contrast, are represented as assuming that voters have made a decision to vote that is invariant to the issue positions of the candidates, but that the decision for whom to vote is still open and is dependent on these issue positions. When the decision whether to vote is invariant to the issue positions of the candidates, a candidate does not risk losing votes from his party base, and there is a strong incentive to choose issue positions close to that of the alternative candidate, gaining two net votes for every swing voter that he attracts. Only in this case (plus the usual assumptions that voter preferences are single peaked on all issues and that there are only two candidates) do both candidates have an incentive to choose issue positions close to the median preferences of those who vote--an encouraging result, perhaps, that suggests that the variance of the issue positions among our elected representatives may be lower than the variance of voter preferences, reducing the bargaining costs of the compromises necessary to govern. The problem is that this result does not seem consistent with the evidence.

My alternative theory assumes that potential voters have not yet made a decision to vote and that candidates face a tradeoff between choosing issue positions that increase the votes from the party base and attracting swing votes. In this case, the vote-maximizing issue positions of both candidates will be somewhere between the median of the party base and the median of all potential voters, depending on the relative effectiveness of increasing votes from the party base and from swing voters. (Karl Rove seems to have been the one to discover that it has been easier to increase two votes from the Republican base than to attract one swing voter.) In this case, the variance of the issue positions of those elected will be higher than that of the electorate. My own contribution to the empirical tests of this hypothesis was an analysis of the votes for Congress in 2002 and 2004, my prior paper on the 2002 election concluding that "U.S. Elections Are Increasingly Biased against Moderates."