A Head-to-Head Comparison of Alternative Voting Rules

The plurality voting rule used in most U.S. elections has been the subject of debate, with some noting that it may misrepresent voters' preferences and allow for candidates who lack broad support to be elected. To better understand how alternative voting rules may perform in real-world elections, Mathematica built a rigorous agent-based model that is grounded in real-world survey data from the American National Election Studies. We simulated more than half a million Senate elections to quantify and compare the performance of eight voting rules on the extent to which they resisted spoiler effects and elected winners who were centrist, representative, and majority preferred.

We assessed the rules' performance under several scenarios that captured the influences of political tribalism, ballot truncation, polarization, primary election type, and the number of candidates competing. We found a stark, consistent performance gap between Condorcet- and non-Condorcet voting rules but minimal differences within the class of Condorcet voting rules. The most widely debated voting rules in electoral reform discourse-plurality and instant run-off voting-showed lower performance on all measures analyzed. We also noted that Copeland was the only voting rule not vulnerable to spoiler effects.

The results suggest that alternatives to the status quo (plurality) that go beyond familiar options like instant run-off voting may warrant consideration. Extensions to further the realism of our model include incorporating ballot errors, candidate and voter strategy, and systems of influence (such as political governing bodies and endorsements), which can affect candidates' viability and familiarity to voters.