Search This Blog

Tuesday, August 29, 2023

A Caution About Instant Runoff Voting

Beyond the Spoiler Effect: Can Ranked Choice Voting Solve the Problem of Political Polarization?
University of Illinois Law Review, Forthcoming

Nathan Atkinson University of Wisconsin - Madison
Edward B. Foley Ohio State University (OSU) - Michael E. Moritz College of Law
Scott Ganz Georgetown University - McDonough School of Business

Ranked-Choice Voting (RCV) is growing in popularity among election reformers, who have coalesced in particular around Instant Runoff Voting (IRV), a specific form of RCV that has recently been adopted in Maine and Alaska and will likely be proposed in many more states as ballot initiatives in the coming years. While reformers hope that IRV can ameliorate extremism and political polarization, this paper presents empirical evidence that undercuts these hopes. For instance, Alaska’s very first election following the state’s adoption of IRV signaled that the method may fail to elect the candidate most preferred by a majority of the state’s voters. Extrapolating from Alaska’s experience, and using a nationally representative sample of over 50,000 voters, we analyze the prospective effects of adopting IRV in every state. This analysis shows that IRV tends to produce winning candidates who are more divergent ideologically from their state’s median voter than do other forms of RCV. And the effect is most pronounced in the most polarized states—precisely the electorates for which IRV is being promoted as an antidote to existing divisiveness. We conclude by highlighting other formulations of RCV that result in more representative outcomes and are thereby better positioned to combat extremism and political polarization.

They explain:

Instant runoff voting does result in the majority rule its advocates tout, but only in the sense that the winning candidate is definitionally preferred to the runner-up candidate by a majority of the electorate. When no more than two candidates are electorally viable, as is frequently the case under two-partyrule, IRV ensures that the candidate with the widest support wins the election. However, a third viable candidate undermines the guarantee that the winner under IRV—or even the runner-up—will be the candidate with the “broadest support from all voters.” In fact, one can easily construct examples where a candidate supported by “a true majority of the voters” is not included in the final two-candidate matchup.

Consider an election with five voters and three candidates: a left-leaning candidate, Linda; a right-leaning candidate, Rachel; and a centrist candidate, Carl. Two voters are liberals and prefer Linda to Carl to Rachel. Two voters are conservatives and prefer Rachel to Carl to Linda. And the lone centrist voter prefers Carl to Linda to Rachel. In an election conducted under IRV, Carl receives only one first-choice vote, and so is eliminated after the first round. The one vote for Carl then transfers to Linda, who wins the runoff election against Rachel by a count of 3-2. Linda does in fact have the “broadest support from all voters” in the contest between Linda and Rachel. However, if the election had been between Carl and Linda, then Carl would have had the “broadest support from all voters,” defeating Linda by a vote of 3-2. Likewise, Carl would have beaten Rachel by the same vote. Yet, because of IRV’s focus on first-place choices in determining the order in which candidates are eliminated from contention, Carl is unable to survive the first ballot. Recall that under Condorcet’s method, Carl would be the most deserving candidate: a majority of voters prefer him to either alternative. But centrist Carl does not win the IRV election when both the liberal and the conservative alternatives have larger bases.