. As voters and candidates better figure out IRV, I would expect it to "reduce" mostly back to Choose-one Voting with similar candidate and voter strateg

I will caution you that this expectation is contradicted by the available evidence, which indicates that RCV does lower the barriers to running, and may even encourage more, and more diverse, candidate entry, as theory suggests. I'm not sure how long-term you're expecting this effect to require, but the research studies this article cites span a period of decades of use in Bay Area cities.

I don't think IRV inherently elects more polarizing candidates than Choose-one Voting because of its mechanics. What I've seen is that in the US right now, because of the way IRV is sold, after it's implemented somewhere, candidate (exit) strategy and voter strategy decreases (at least initially). That would cause either method to elect more polarizing candidates, and I'm pointing in particular to (rare) IRV elections like Burlington, VT 2009 and this ongoing Alaska Special General Election with 3 distinct front-runners. As voters and candidates better figure out IRV, I would expect it to "reduce" mostly back to Choose-one Voting with similar candidate and voter strategy. The transparency and familiarity of Choose-one Voting helps it to "stabilize".

So Sass predicts that the result will be that a more divisive candidate wins under RCV IRV Hare than would under Choose One.

This prediction is not supported by real-world data. In fact, the opposite seems to be true.

Research shows that the claim that IRV alters incentives and campaign dynamics in ways that appear to decrease negative campaigning and increase campaign civility is mostly supported at the local level, and furthermore, early evidence is promising that IRV makes primaries work better in avoiding polarizing candidates.

Thus far, primaries that use RCV have generally produced “consensus” candidates, affirming expectations that RCV can have a moderating effect on primaries, or at least have the effect of blocking the path of more polarizing candidates who might have enough base support to win under plurality rules.Another apparent advantage to using RCV in the primary process is that because RCV encourages sincere (as opposed to strategic) voting, ranked ballots in party primaries can more clearly demonstrate factions’ relative priorities and voting power than traditional single-mark ballots, thereby giving party leaders, who will need to unify those factions around the party nominee (and in the long term keep their big tents together and safe from a hostile takeover), a better sense of how and with whom to bargain for future cooperation.
source

So Sass predicts that the result will be that a more divisive candidate wins under RCV IRV Hare than would under Choose One.

I find that one particularly unlikely, although it is not clear if this means "if the whole election was under choose-one from the beginning" vs. looking at the RCV ballots after the fact, and counting only first choices choose-one style.

Still, I think it's a stretch to say RCV would elect a more divisive candidate than choose-one. Overall I think Sass's bias against RCV is out of proportion with its actual real-world flaws.

Factions are "competitive" if there is enough support for them that they could conceivably win. When there are factions within a party, Sass says, there are usually just two, the moderate faction and the more extreme faction. He says that we rarely see elections in the US with more than two factions, because people are used to that count, which results from use of Choose-one Voting.

Now in Alska, you have Peltola, Palin, and Begich. Predicts Palin will win, and thinks she is not the Condorcet winner, per polling. With choose-one, the Republicans would strategically vote for Begich instead of Palin. So Sass predicts that the result will be that a more divisive candidate wins under RCV IRV Hare than would under Choose One.

When voters only have one mark they can make on their ballot, they look for a bandwagon to jump on so they don’t waste it. This increases the relative importance of money and fame in elections because those are the most obvious signs of a bandwagon. The scarcity of power that voters have causes them to tactically make the safest choice they can to be the most effective with that limited power.

Peltola, Begich, and Palin
Yup’ik Eskimo
(f)

Sass points out that the CES polling shows that voters want to support multiple candidates.

https://electionscience.org/ces-updates/voting-system-ces-conducts-the-republican-liberty-caucus-straw-poll/

Sass says that a precinct in Texas or rural Wyoming would not have the grasp of tech to publish the cast-ballot data over the Internet.

You should be able to rank all the candidates if you want to.

Raynaud system.

Sass thinks in the case of 4 or more candidates, equal-ranking systems do not escape Arrow.

Ranked Robin does not pass Independence of Irrelevant Alternatives. But Sass thinks this is a matter of degree, and it might not be that important. Clone candidates help each other.

Condorcet systems are "usually" monotonic, but all fail participation.

"Tournament-style Borda" is equivalent to the tiebreaker in Ranked Robin.

There's no possibility that single-choice plurality voting in Burlington 2009 would have this 3-way race as it was. All the media and voters and everyone (candidates included) would have come to some pre-election idea of Montroll or Kiss as the primary non-Republican and the other as dangerous spoiler. They would have had that argument. And the result would be likely Montroll winning but maybe Kiss or even the Republican and then all the spoiler-blame fall-out. There's NO chance it would have avoided all those well-known dynamics.

and that is why Montroll would have won

Isn't stating that with such certainty doing pretty much the same thing as FairVote stating the opposite with such certainty?

We don't know what would have happened under choose-one plurality. We don't even know if Montroll would have run, maybe the Democrats would have run a candidate that was less centrist.

I agree that FairVote being disingenuous is a problem, but I do wish people wouldn't use that alone to bash (or otherwise lessen their enthusiasm for) positive change.

(My opinion on IRV is increasingly "it's doing the right thing, it just isn't doing it enough." Basically, it is pulling its punches.)

Given the preference data available, I think it is a very reasonable expectation that Montroll would not have won with either choose-one or top-two-runoff. Granted, I wouldn't go so far as to call it "certain," but it's surely reasonable.

Generally speaking, voters behave the way that candidates and parties tell them to. If all three candidates had stayed in the race, then they are signalling to voters "hey, you should still vote for me," and I don't think there's any reason to believe that voters' would change their first-preference behavior given how competitive the race was.

More likely, under choose-one, one of Montroll or Kiss would have dropped out of the race before election day; strategy will occur, but it typically manifests as candidate exit rather than voter behavior.

On that same page where Fairvote talks about Condorcet and such:

It may be disputed whether it would have been better for Montroll to win the election despite attracting so little core support. However, it is certain that Montroll would have also lost under a two-round runoff election or a single-choice plurality election.

This is plainly wrong. Under a plurality election, we know with certainty that voters will be strategic, and that is why Montroll would have won. Their wording is missing the key point, and can be true only with this change:

it is certain that Montroll would have also lost under a two-round runoff election or a single-choice plurality election — if the voters were all completely honest, which we know plurality voting not to be.

Fairvote in this argument is trying to assert that IRV cannot be worse than single-choice plurality. They assert that by selectively ignoring the entire issue of strategic voting, even though they focus elsewhere on strategic voting. Fairvote is a model for motivated reasoning over fair reasoning.

If 90% of those elections had only one candidate, for instance, it wouldn't be particularly meaningful that RCV selected them.

It is true that the vast majority of elections have an "obvious" winner (aka true majority of first-preference votes) and many only have two or maybe even one real candidate(s).

Although I will say I can corroborate the claim that no IRV election (besides obviously Burlington) seems to fail Condorcet at least for the elections in San Francisco and Minneapolis, which were the ones I could find ballot data for.

While FairVote's editorials can be a little over-exaggerated and their terminology purposefully vague (never defining 'majority'), sometimes to the point of untruthfulness---they are a political advocacy group after all---I have never seen reason to question the basic facts they present.

If 90% of those elections had only one candidate, for instance, it wouldn't be particularly meaningful that RCV selected them.

My understanding is that for all the Bay Area RCV elections there is full ballot data published. https://www.fairvote.org/every_rcv_election_in_the_bay_area_so_far_has_produced_condorcet_winners

Is there a reason you have to suspect that they aren't telling the truth, or that there might be a significant number of elections where ballot data isn't available but Condorcet winner didn't get elected? Just curious. It's not something I personally would be worried about.

@psephomancy Are you looking for more than the statement from FairVote?

Here is where they say it: https://www.fairvote.org/research_rcvwinners

As for the ballot data for those 440 elections, I don't have it, but someone may.

Yeah, I've heard this claim a bunch of times but never with actual data showing that there was a Condorcet winner and that they won.

Of the 440 single-winner RCV elections in the United States since 2004 in which we have sufficient ballot data to assess whether the Condorcet winner won the election, all but one — the 2009 mayoral election in Burlington, Vermont — were won by the Condorcet winner.

Here is where they say it: https://www.fairvote.org/research_rcvwinners

As for the ballot data for those 440 elections, I don't have it, but someone may. @Andy-Dienes , did you say you had parsed a ton of data? Personally I think we should convert as much as we can to a standardized format and make them available on this site. (with javascript versions of them, since that would allow easy linking them into CodePens etc.... see https://codepen.io/sergelerner/pen/dXdqPe )

Yes this is true. 440 elections, all had a Condorcet winner. One of them didn't pick that Condorcet winner.

Where is the data for this?

I particularly like this chart showing how the IRV election for London Breed played out:

[]

It is interesting in that it makes clear certain things that aren't so clear in other ways of expressing results, including Condorcet matrices.... for instance it clearly shows how most of Jane Kim's votes went to Mark Leno. It is clear that Mark Leno would have done a lot worse in a plurality race, as Jane Kim would have been a spoiler.

You can do the same Sankey diagrams for Condorcet methods that eliminate candidates in rounds, like Baldwin's method or BTR-IRV.

I have a question or two arising from what you and Rob B. have been saying back and forth. You distinguish the historical votes in the US where we have better ballot data than just what Choose-one Voting gives, as to whether they involved competition among more than two factions. Are those factions among the candidates, or factions among the voters? And how do you discern whether such factions existed for a given one of those elections?

They say "IRV & bottom-2 runoff will usually have the same result but different campaign incentives. IRV rewards those w/ strong 1st choice support. Bottom-2 rewards those who avoid polarizing stances. We like IRV since it encourages real stances, not just campaigning to avoid the bottom"

It's really odd for them to claim that people would campaign differently when both methods "usually have the same result." If they have the same result in 339 out of 440 elections (which they proudly proclaim in multiple articles).... why would campaign incentives be so different?

Also odd that what they are arguing here seems to be the exact opposite of what they argue elsewhere. They are basically saying "Polarization is bad. Our voting system is better than others, because ours doesn't polarize people like those other systems do. Except, the reason we're better than Condorcet compliant systems, is that those systems have the exact same effect of reducing polarization.... but they do it even more than we do, and that's probably just too much".

I don't see how that makes sense. They really are talking out of two sides of their mouths.

two competitive factions,

Are you talking about factions among the candidates, or factions among the voters?

not a race to 50%+1.

That's an absolute majority. Is there any seriously-proposed system that won't elect a candidate who has majority support (as their first choice), provided that the majority knows they are the majority and chooses their strategy in a way that takes that into account and is aimed at winning?