How Ranked-Choice Voting can elect right-wing candidates in left-wing districts

It looks like Instant-Runoff Voting, a type of ranked-choice voting, is going to be on the ballot in NYC.

Unfortunately, IRV is a poorly-designed plurality-runoff system, which works fine when there are only one or two strong candidates, but can fail whenever there are three or more, causing it to elect unrepresentative candidates.

Here’s an example of IRV electing a Republican in a left-leaning NYC district:

A 3-candidate IRV election in New York City [using the voteline simulator]

Here the Democrat (Blue) and Republican (Red) are equally-spaced away from an absolute “centrist” position (gray dotted line). The electorate (the bell curve) is left of center, and there’s a farther-left Progressive candidate (Green).

Even though the population leans left, and a left-wing candidate should therefore win, that doesn’t happen under IRV: The Progressive splits votes with the Democrat, causing the Democrat to be eliminated first. The Democrat’s votes are then transferred, some to the Progressive, and some to the Republican, and now the Republican has enough to beat the Progressive (since the Republican is a slightly better match for the average voter).

Although the RCV:NYC campaign claims that RCV will “Produce consensus candidates”, we can see that in this scenario, the consensus candidate was eliminated in the first round.

If the Progressive hadn’t run, or if Progressives had voted strategically for the Democrat (like they do under our current system) then the Democrat would have easily won:

A 2-candidate election, with the Progressive dropped out

In other words, the Progressive is acting as a spoiler candidate, just like they would under our current FPTP system. (Many people falsely claim that IRV “eliminates the spoiler effect”, but as you can see here, it does not.)

The problem is that IRV is still a FPTP-based system, because it only pays attention to first-preference votes in each round, so it still suffers from vote-splitting in each round. Adding runoffs hides this problem in some scenarios (when there are only one or two strong candidates), but doesn’t fix the fundamental problem in general.

Under most other voting reform proposals, such as ranked-choice methods that meet Condorcet’s criterion, or score-based methods like STAR Voting or Approval Voting, the Democrat would have won, which is the correct outcome, because the Democrat is the best match for the overall position of the electorate, would beat any other candidate in head-to-head elections (“Condorcet winner”), and has the highest overall approval rating (“utilitarian winner”).

IRV is the most popular voting reform (and the only reform many people have heard of) but unfortunately it’s just “famous for being famous”, not because it’s actually a good choice.

[All images are made by me and CC0 licensed]