The Unexpected Consensus Among Voting Methods

Historically, the theoretical social choice literature on voting procedures in economics and political science routinely highlights worst case scenarios, emphasizing the inexistence of a universally ‘best’ voting method. Indeed, the Impossibility Theorem of Nobel Laureate Kenneth Arrow proved that no voting method can ever satisfy all of his requirements simultaneously.

But as the U.S. Presidential election of 2000 reminds us, voting methods continue to court controversy, and there are many efforts under way to reform the electoral system at all levels of government. Unfortunately, the popular debate and the scientific debate about voting methods have long been preoccupied with theoretical claims that are often supported only with simple hypothetical thought experiments. Likewise, the theoretical social choice literature on voting procedures in economics and political science primarily highlights worst-case scenarios and the mathematical impossibility of a single universally ‘best’ election method. New research published in the July issue of Psychological Science, a journal of the Association for Psychological Science, tested whether methods of voting, including instant runoff, achieved more similar results than previously thought.

Michel Regenwetter, a professor of psychology and political science at the University of Illinois, along with his colleagues analyzed four presidential elections of the American Psychological Association (APA) using state-of-the art decision modeling and statistical methods. APA elections are particularly useful for such analyses because, in contrast to a common two-person runoff election, the APA ballots provide individual voter preference rankings and the APA elections involve multiple candidates (five).

Using these data, the researchers compared instant runoff voting to three other classic methods: Condorcet, in which all candidates are placed head to head against each rival, the winner being the one who wins the most contests; The Borda count in which points are given to each candidate based on ranking by the voter; and the more familiar Plurality system, in which each voter gives one vote to one single candidate.

Although past research has routinely depicted these procedures as irreconcilable, Regenwetter and his colleagues found strong support for consensus among these procedures in these elections. Their findings contrast two centuries of pessimistic thought experiments and computer simulations and demonstrate the need for more systematic empirical research on voting than exists to date.