RCV is gameable. Here’s how.

Before we talk about Ranked Choice Voting, let’s look at partisan primaries.

Partisan primaries are gameable. A prime example of this is Democratic incumbent Senator Claire McCaskill gaming the 2012 Missouri Senate race by helping the enormously divisive Congressman Todd Akin win the Republican primary before besting him in the general election. Here’s how Senator McCaskill described her plan:

So how could we maneuver Akin into the GOP driver’s seat?

Using the guidance of my campaign staff and consultants, we came up with the idea for a “dog whistle” ad, a message that was pitched in such a way that it would be heard only by a certain group of people. I told my team we needed to put Akin’s uber-conservative bona fides in an ad — and then, using reverse psychology, tell voters not to vote for him. And we needed to run the hell out of that ad.

My consultants put together a $1.7 million plan.

…

As it turned out, we spent more money for Todd Akin in the last two weeks of the primary than he spent on his whole primary campaign.

If we were going to spend that kind of money on ads for Akin, I wanted to get him nominated and start disqualifying him with independent voters at the same time. By that prescription, our ad would have to include Akin’s statement that Obama was a “menace to civilization” and that Akin had said of himself that he was “too conservative” for Missouri. This presentation made it look as though I was trying to disqualify him, though, as we know, when you call someone “too conservative” in a Republican primary, that’s giving him or her a badge of honor. At the end of the ad, my voice was heard saying, “I’m Claire McCaskill, and I approve this message.”

It started to work. Our telephones were ringing off the hook with people saying, “Just because she’s telling me not to vote for him, I’m voting for him. That’s the best ad for Akin I’ve ever seen!” A man wrote a letter to the editor of the Springfield News Leader: “I think it’s time for someone who may be too conservative. Thank you, Senator McCaskill, for running that ad. You have helped me determine that my vote needs to go to Akin.”

Akin won the Republican primary in a close three-way race with 36% of the vote. McCaskill defeated him in the general election in a 55%-39% landslide after Akin ludicrously claimed that “legitimate” rape very rarely leads to pregnancy. Six years later, McCaskill lost to Republican Josh Hawley by a 6-point margin.

The structural reason why McCaskill’s plan could work was that the Republicans could only field a single candidate against her. Given this limitation, it was inevitable that she could meddle in their process for selecting which candidate would face her, and give an edge to the weakest possible challenger. Had the Republicans been able to send multiple candidates against her, such that she’d need to defeat all of them to hold her seat, the fact that one of them was a pushover like Todd Akin wouldn’t have mattered.

From partisan primaries to RCV

Now, let’s turn our attention to Instant Runoff Voting (IRV), aka single-winner Ranked Choice Voting. Under a system like Alaska’s (a Top-4 primary followed by an IRV general election), there could be multiple Republican candidates in the general election. Superficially, this seems like it might preclude McCaskill’s gamesmanship — but the structure of IRV says otherwise.

IRV involves a series of elimination rounds in which the candidate with the fewest votes is eliminated, and their votes are transferred to their supporters’ highest-ranked remaining candidates. When a Republican gets eliminated, their support will be overwhelmingly transferred to other Republicans since voters who rank a Republican first tend to prefer Republicans to Democrats more broadly. Eventually, only two candidates are left, and whichever of them has more votes will be elected.

If the two finalists consist of one Republican and one Democrat, very little is changed from partisan primaries. Republican-leaning voters select one candidate for the final tabulation round, Democratic-leaning voters select the other finalist, and they face off. Had Missouri used Alaska’s system with an IRV general election featuring multiple Republican candidates, McCaskill’s “dog whistle” ad campaign would have still given a huge boost to Akin in determining which Republican McCaskill would face in the final round, and the “too conservative” pseudo-attack would have helped McCaskill against Akin within the final round as well.

(It’s also possible for the final round to feature two Republicans and no Democrats. But for all of the Democrats to be eliminated before the final round, the last remaining Democrat must receive less than 1/3 of the vote. Otherwise, it’s mathematically impossible for her to have fewer votes than two different Republican candidates. This would mean the Democratic candidates were not remotely viable, so contesting the race is hopeless — just another “safe Republican” seat that Democratic strategists would give no mind.)

To be fair, IRV could have still made McCaskill’s strategy less effective. Part of why Akin won the Republican primary was that the more electable Republicans split the vote, and IRV would have addressed this vote-splitting. But the basic structure — Republican voters selecting a single Republican candidate to face off against the Democrat in the final round — would remain, and Democrats could still try to manipulate Republican voters into selecting a less electable candidate.

What systems can’t be gamed like this?

If the Republicans could have tested multiple candidates against McCaskill simultaneously — instead of having to pick a single champion — McCaskill’s manipulations could not have been effective. This is most clearly the case in Condorcet methods, in which every possible head-to-head match between candidates is examined and the candidate who defeats everyone else head-to-head is elected (assuming such a candidate exists, as is nearly always the case). If a Condorcet method had been used, McCaskill’s support for Akin would have done nothing to help her against the other Republican candidates, who she’d still need to beat head-to-head.

STAR Voting is also less vulnerable to McCaskill’s stratagem than IRV despite both these voting methods having a final tabulation round with only two candidates. Under STAR, lower levels of support make a much bigger difference than under IRV, and McCaskill's ads would have made Akin less popular among more liberal voters. This wasn’t an issue with partisan primaries (such voters don’t vote in the Republican primary) and wouldn’t be an issue for Akin under IRV (such voters would rank McCaskill first anyway, so their preferences for other Republicans over Akin would be irrelevant), but would matter a lot under STAR since missing out on a few stars from liberal voters hurts just as much in the scoring phase as missing out on a few stars from conservative voters. IRV makes it easy to reach the final round with relatively narrow support; STAR doesn’t.