It’s almost election day in the US, and most states remain single-vote systems. Maine and Alaska, however, have authorized ranked-choice voting (RCV), where voters indicate ordinal preferences (rank in order: 1, 2, 3, …) for the slate of candidates. Cities across the US, like San Francisco, New York, and Minneapolis, are also experimenting with this new way to vote.
As with most things nowadays, the recent changes come with controversy. I’ll cover the pros and cons of ranked-choice voting, as well as point out an overall truth: no electoral system is perfect.
Let’s start off with unpacking “no electoral system is perfect”. This is actually a mathematically-derived result, with Arrow’s impossibility theorem covering ordinal preferences, the Gibbard-Satterthwaite theorem covering cardinal preferences (assigning a value to each candidate), and Gibbard’s theorem covering more general systems (including single-vote, approval voting, and majority judgement).
Arrow’s Impossibility Theorem states that, given at least 2 members in a society and at least 3 options to choose from, no rank-order system can satisfy all three “fairness” criteria:
- If every voter prefers one choice over another (e.g. X and Y), then the group also prefers X over Y
- If every voter’s preference between two choices (e.g. X and Y) remains unchanged, then the group’s preference between X and Y will also remain unchanged (even if voters’ preferences between other pairs change, like X and Z, or Z and W).
- There is no dictator who can always determine the group’s preference
Ranked-choice voting has gained popularity because it offers the ability to express preference for non-mainstream candidates without “wasting” a vote in the overall contest. In current single-vote systems, votes for third-party candidates in sufficient number can cannibalize enough vote share from another top preferred candidate to cost them a victory, even though the majority of voters may have preferred that top preferred candidate to the actual winner.
With ranked-choice, if no candidate has a majority of votes, the candidate with the lowest number of votes is removed from consideration, and their votes are distributed according to voters’ next preferred choice. By continuing to do this until you have a winner, you help ensure that a wider array of voters’ preferences are taken into consideration.
Those are the positives. As we stated before, no system is perfect, and the biggest flaw with RCV is that it can lead to some “wonky” outcomes. Here’s a relatively simple example to illustrate:
- Consider Candidates A, B, C, and D under ranked-choice voting
- In Round 1: A gets 20% of the vote, B gets 40% of the vote, C gets 30% of the vote, and D gets 10% of the vote
- No candidate has a majority, and D has the lowest share of votes. For sake of simplicity, assume that half of these voters have preferences D->A, and half have D->C
- Round 2: A gets 20% + 5% = 25% of the vote, B still has 40%, C has 35%
- No candidate has a majority, so votes get redistributed again. Again, suppose that half of A’s voter pool from the second round have preferences A->B (or D->A->B), and half have A->C (or D->A->C).
- Round 3: B now has 52.5% of the vote, and while C ends up with 47.5%, B has reached majority and wins.
With subsequent narrowing each round, you eventually get to one candidate with a majority of votes. However, as Arrow’s theorem points out, this solution isn’t guaranteed to be the preference of the group. Here’s another example:
- Consider Candidates A, B, C, and D under ranked-choice voting
- In Round 1: A gets 39.7% of the vote, B gets 30.9%, C gets 27.8%, and write-ins make up the rest.
- C has the lowest share of votes and is dropped. 50% of C’s voters chose B as their #2, 30% of C’s voters chose A as their #2, and 20% of C’s voters either didn’t write in a #2 or wrote in a non-candidate.
- In Round 2: A now has 51.48% of the vote, and B has 48.52%. A wins.
- Now suppose we review B voters to see what their preferences are, and 75% would have chosen C as their #2. If C were up and B was the lowest-ranked choice, candidate B would have won.
The second example isn’t a hypothetical, it’s the 2022 Special Election in Alaska where Mary Peltola became the first Democrat to represent the state in 50 years. We don’t know exactly how Palin voters (Candidate B) would have voted for Begich (Candidate C), but both are Republicans and Palin is more extreme (so we’d assume a higher % would choose Begich as their second choice, or at least a significantly lower % would have ranked Peltola as #2). If Peltola (Candidate A) had only been facing 1 Republican candidate, it’s likely that the pull of partisanship and only 2 choices would have put Palin or Begich (whoever won the primary) over the edge instead.
On one hand, you can argue that Begich was a spoiler candidate (Palin had a higher % vote share than Begich in the free-for-all primary as well) compared to a top-2 final, since 58% of people chose a Republican as their 1st choice. On the other hand, this is RCV in action: when voters have multiple candidates and can actually say what order you prefer them, they can more accurately express preferences instead of having to stick to a duality. At the end of the day, a sizable chunk of people wanted Begich to win, but also preferred being represented by Peltola instead of Palin if it came down to the two, and RCV made these preferences a reality. Peltola may be a member of the Democrats, but she’s carved out a uniquely Alaskan position of “Fish, Family, and Freedom.”
Ranked-choice voting isn’t a perfect system, but it’s a pretty intriguing premise overall, and may be the key to fostering bipartisanship: appealing to having voters rank you as #2 in addition to the #1 choice requires appealing to a broader constituency. At the same time, there’s already a ballot measure to repeal RCV in Alaska, and past attempts at RCV have led to similar repeals.
I’m curious to see the results this year, and whether this system, or many other potential systems, expand across the US.
Andrew Jabara 