I understand how someone can promote IRV when they're not aware of center squeeze, but I cannot understand why they would be so stubbornly persistent in promoting IRV even after understanding its biggest flaws.
The FairVote crowd have really lost the plot, and are just harming their own cause at this point, continually pushing a broken voting system that doesn't actually fix the things they claim to want to fix, and that's getting banned in more places than it gets adopted.
>>...why they would be so stubbornly persistent in promoting IRV even after understanding its biggest flaws.<<
Because they are selling a product that they don't want to admit to customers, particularly the previous customers that had purchased the product, that the product they sell has an avoidable flaw. They want to say that "We've been using IRV for decades and it's been working just fine" and continue on in the same direction hoping that no one (or very few) else will notice the flaw or how the flaw is an avoidable flaw.
My daughter compared it to software revisions. We can call First-Past-The-Post "Democracy v1.0" And IRV (under the moniker of "RCV") "Democracy v2.0" Why can't they just acknowledge the bug, fix the bug, and call it "Democracy v2.1"? They cannot admit to overselling the features of the product in the first place.
We need to remember that, in order to value our votes equally, that Single-Winner, Multi-Winner, and Apportionment of Delegates (Presidential primary) are very different problems.
1. Single-Winner: There is no proportionality to be had with a single winner. It's winner-take-all. So then, the *only* way to value our votes equally is Majority Rule. And Condorcet RCV respects Majority Rule in elections where Hare RCV does not. Condorcet is clearly better at respecting Majority Rule and the equality of our votes.
2. Multi-Winner: Now proportionality comes into the equation and then something like the Weighted-Inclusive Gregory Method (PR STV) might be the method that best reflects apportioning multiple seats in a single legislative district. This, of course, has quota, surplus votes that get transferred, and *then* a Hare-ish elimination of candidates.
3. Apportionment of state delegates to the DNC or RNC: This is a solved mathematical problem and it's used to apportion 435 U.S. House Representatives to the 50 states. Same math, and that method should be used to apportion a known number of delegates to candidates, based on the number of votes that each candidate gets.
Great post. Your point about a candidate who would win a Condorcet-based election to refraining from competing because there is no point doing so because of the way IRV elects winners is one that IRV supporters are blind to, whether willingly or not. Marcus Ogren dubbed it the primordial election here: https://voting-in-the-abstract.medium.com/the-primordial-election-that-is-never-held-e019356faf90 . Since IRV by its own mechanism can throw out a moderate that is preferred by 2/3 of the voters over the other two candidates whilst having almost identical first choice support, candidates who do the math and find they might get "only" 20%-25% of the first place vote will choose to not waste the time and effort to campaign. Because even if they overperform and get to 30%, there's still a good chance they still get eliminated in IRV's penultimate round.
To your point about strategic voting, IRV supporters like talking about the way that other methods can be gamed by voters to change the result in their favor, but they never talk about how voting honestly with IRV causes maximum voter regret. The 2009 Burlington mayoral race and the 2022 Alaska special election showed that voters for Kurt Wright who preferred Andy Montroll to Bob Kiss and voters for Sarah Palin who preferred Nick Begich to Mary Peltola, respectively, didn't get the result they wanted because they voted their honest ranking. Had they known that voting sincerely would result in the election of their least preferred candidate, they would have strategically voted for their second choice as their first choice. In effect, under IRV, voters would be better off strategically voting in order to elect the Condorcet candidate; one might say it would be better to avoid this situation altogether and just use a Condorcet method so they don't have to either experience regret or have to betray their favorite to not elect their least favorite choice.
IRV seems built in a lab to be the anti-Condorcet method, not a Condorcet-like method that Greg Dennis seems to want to prove. Not only does it disincentivize potential Condorcet winners from even competing, it explicitly throws out all Condorcet winners that don't meet the strict IRV elimination thresholds. When Fairvote discussed the cases of Burlington and Alaska, they basically said that "the Condorcet winner didn't deserve to win because they weren't the IRV winner": https://fairvote.org/resources/data-on-rcv/#condorcet-winners They characterized Montroll's 29% as "little core support" and Begich's 28% as "low support" but somehow Kiss's 34% and Palin's 31% were not "low." In reality, Candidate C could come up one vote below Candidate B in the penultimate round and still be eliminated, and the Fairvotes and Greg Dennises of the world would say "well, Candidate B was obviously more worthy than Candidate C to be considered in the final runoff round because they had more 'core support'", as if core support was the only consideration in who was worthy to win. Also, the notion of more "core/enthusiastic/deep" support driving elimination consideration is laughable considering an IRV election may have 7-10 candidates initially where the eventual winner started out with 10%-15% of first place votes, so by the time the penultimate round comes around, the two candidates with higher "core/enthusiastic" votes really have a smaller fraction of their votes being "core/enthusiastic" and a higher fraction of their vote total being the less enthusiastic lower rankings piling up.
>>The 2009 Burlington mayoral race and the 2022 Alaska special election showed that voters for Kurt Wright who preferred Andy Montroll to Bob Kiss and voters for Sarah Palin who preferred Nick Begich to Mary Peltola, respectively, didn't get the result they wanted because they voted their honest ranking. Had they known that voting sincerely would result in the election of their least preferred candidate, they would have strategically voted for their second choice as their first choice.<<
Logan, the real shame in these two examples is that these Wright or Palin voters (who disliked Kiss or Peltola the most) were promised that they could safely vote for their favorite candidate without fear of causing the election of their least-favorite candidate. They were literally promised that if they couldn't get their first choice, then their second-choice vote is counted. That, of course, is never true for voters that vote for the loser in the IRV final round. It's just that in most IRV elections, that loss of one's second-choice vote does not make a difference in the outcome. But in these two elections, it did.
Yes, IRV advocates routinely make the claim that one can "vote for their favorite without fear that their least favorite will be elected" despite it being plainly false both in theory and in practice, most recently for an office as important as a US House seat. And there's no reason to not believe that the same would happen in a Presidential election if IRV was adopted nationwide. 1992 Perot is probably the best example where either Bush or Clinton would've been the Condorcet loser and so Clinton-Perot and Bush-Perot voters would've been better off voting for Perot first. Had IRV been in effect, Perot may have gotten very close to getting in the final round, but still could've been eliminated in favor of the Condorcet loser making the final round. IRV advocates are dumb because their voting method essentially says "yea, the Condorcet loser will by definition always lose an IRV election, but let's put them into the final round for funsies." This isn't like a football playoff where the results of the next round are unknowable so you "have to play the game to see who wins;" the results of the final round are already set in stone, but IRV likes to pretend it's not so they can set up a kabuki final elimination round to fool voters into thinking there was an actual competition between the final two candidates when the loser was already decided.
Greg Dennis's organization Voter Choice Massachussets makes these two claims right on their front page in big bold letters:
1. Ranked Choice Voting Ensures Majority Support by eliminating the “spoiler effect” and guaranteeing the winner earns a majority of the votes in any election.
2. Ranked Choice Voting Expands Voter Choice by freeing you to vote for who you really want, without settling for the “lesser of two evils,” and without fear of “wasting” your vote.
#1 is a combination of straight up lie and lie of omission. As demonstrated earlier, there is a "spoiler effect" in IRV elections, just not the 2000 Ralph Nader type. Montroll and Begich would have won their elections had eventual non-winners Wright and Palin not "spoiled" the elections for them by entering the election, same as non-winner Nader. And the "guaranteeing the winner earns a majority of the votes in any election" point omits that an IRV winner can win despite not earning a majority if the winner was not also a Condorcet winner, because on the actual ballots, by definition they didn't win a majority against the Condorcet winner.
#2 is basically a different way to say it eliminates the "spoiler effect," so obviously is another lie. Voters for Kurt Wright who preferred Andy Montroll to Bob Kiss and voters for Sarah Palin who preferred Nick Begich to Mary Peltola should have settled for the "lesser of two evils" so that their least favorite candidate didn't win, but since they were told by orgs like Fairvote and VCM to not fear ranking their favorite first, they didn't strategize as they might have in a plurality election to ensure they didn't experience maximum regret. Their vote (really their first place ranking) was "wasted" on a candidate that had no chance to win because they were the Condorcet loser, same as Nader voters in 2000. I reiterate that orgs like Fairvote and VCM don't characterize candidates like Wright and Palin as "spoiler candidates" despite that being exactly what they are, so don't know if they're lying because they're committed to IRV ideology or because donors like the fact that IRV has a "center squeeze" problem so want to perpetuate lies that pretend it doesn't exist.
Yes. I didn't go to Greg's site (recently) but I have seen these falsehoods repeated, multiple times, with a variety of different RCV promotional organizations. I have pointed out the falsehood, how it's false, why it's false, and why they should be qualifying the claim enough to make it true. But they don't do that because the simple claim, albeit false, has better marketing appeal.
Another one (from Better Ballot Vermont and several other orgs) is: "To win an RCV election, a candidate must get over 50% of the vote." Straight-out falsehood. Easily proven false by use of counter-examples.
A lie from RCVRC I got in the email in 2022: "Does ranked-choice voting impact how long it takes to know who won the election? ... No! Ranked-choice voting elections can be tabulated as quickly as a few minutes using round-by-round counting software." The disingenuity comes from the omitting the necessity to centralize the ballots or equivalent voting data from all of the polling places. In Maine, it took 4 or 5 days. In Alaska it takes 15 days after the election before they run the IRV software.
I'm trying to be polite to the IRV apologists, but they ain't being honest, and I am not always patient. I am eschewing calling anyone a "liar", but this is hard. Especially with Trumpers. But the IRV apologists just need to stop with the falsehoods, because if they don't, they're gonna be called out. Eventually by me, if no one else does.
"But they don't do that because the simple claim, albeit false, has better marketing appeal."
Yes, IRV advocates are fond of using factually false claims for better marketing appeal. Consumer law would call that fraud, and it's what I'd call it. It's the equivalent of a pharmaceutical company saying that you won't get diarrhea if you take their pill despite the documented cases of diarrhea after taking the pill.
The fact that they know it to be false and keep parroting the lie front and center on their websites and other communications qualifies them as serial liars who shouldn't be taken seriously as anything resembling scholars of election methods.
Yes, "... guaranteeing the winner earns a majority of the votes in any election" is a straight-up falsehood. Greg, why do you continue to prominently post that falsehood? I remember pointing this out to you before.
To put a bow on the last point about IRV rewarding candidates with "more core support," it's obvious that IRV supporters really only considered elections whereby third party spoiler candidates siphoned off less than 5% of the vote. These candidates "spoiled" the election of a candidate who got 46%-48% of the vote to lose to another candidate who didn't have a spoiler who got 47%-49%, and IRV would fix that so that candidates with "high core support" didn't lose because of candidates with "low core support" spoiling the election.
But when confronted with the possibility of an election being 33.5%-33.3%-33.2% whereby the 33.2% candidate is the 2nd choice of the 66.8% of voters who supported the other 2 candidates, IRV supporters just agreed to characterize any candidate that placed third in the penultimate round as a weak spoiler candidate. They did this as a way to deflect criticism about their method not electing the Condorcet candidate based on inflexible elimination criteria that treated a candidate that came up one vote shy of being second the same as a candidate who got only one vote.
Basically, IRV supporters don't think that candidates who make the final round but ultimately lose as "spoilers" for the third place candidate who got eliminated in the penultimate round but would've won in the final round, but that's what they are. Kurt Wright "spoiled" the election for Montroll in the 2009 Burlington mayoral election, because by the stated preferences of the voters, they preferred Montroll over Kiss, but when Wright entered the race, he caused Montroll to lose despite not having a chance to win using an IRV election method. This is functionally the same as Nader causing Bush to win by entering the 2000 election despite not having any chance to win using the plurality election method, but IRV supporters cannot see that.
>>But when confronted with the possibility of an election being 33.5%-33.3%-33.2% whereby the 33.2% candidate is the 2nd choice of the 66.8%<<
You might have noticed that I posted a form of that extreme example with 99 voters in the comments to Greg's column posted in DemocracySOS.
And BTW, I happen to live in Burlington Vermont (I'm actually in Bernie's ward and work as an election worker and had previously been an elected polling official in this ward). I was here in 2005 when we voted to adopt IRV and was here when the 2009 IRV election occurred and the repeal that happened the year after. The really sad thing is how short the memories of people are. We readopted IRV after 13 years. But I am active in the statehouse where there are some legislators that are beginning to understand this problem.
One thing I would like to resist in this article is the claim Ned makes with respect to the hypothetical Ohio and North Carolina elections he discusses. It is that "Only a Condorcet-consistent electoral system would enable the will of the majority to prevail and elect the rule-of-law Republican instead of the election deniers." In my view, that assertion requires the unsupported assumption that capture of "the general will" must always be a function of favorite picking. If we instead consider the purpose of a single winner election to land on that candidate that most of the electorate thinks would be minimally acceptable (or better than nothing), then no Condorcet-consistent rule is likely to do the trick. And, I take it, using minimal approvability rather than finding the round robin favorite as the goal is just as conducive to "finding somebody in the middle."
What I think we ought to want in single winner elections is to have results that most of the citizenry can live with without raising Cain. That is the goal of Approval Voting. The complaint that it is subject to strategic manipulation through bulleting is significant, but it is not alone in that. At any rate, I am not convinced that modern electorates are incapable of ever coming to understand the benefits of almost always having elected officials who are OK with them even if that means the majority sometimes won't get their fave. I mean, look at the alternatives.
I served on something called The Elections Modernization Committee with Greg for three years in our hometown of Arlington, Mass. and we argued pretty regularly about the relative merits of his beloved RCV and my own preference for Approval Voting. Both FairVote and CES (The main AV support organization) have, over the last few years, been willing to trade a number of what I take to be foundational principles for defensible voting schemes for success in this or that jurisdictional election. It ought to be clear by now that such trades haven't worked particularly well for either group. But, alas.....
I will read this, Walter. My initial pre-reading observation about Approval Voting is that, like all Cardinal methods, AV inherently forces voters to think tactically the minute they walk into the voting booth if there are 3 or more candidates running. The voter has to decide how much to Score or whether to Approve their second-favorite candidate.
If you Approve your 2nd choice, that throws away your vote for your 1st choice *if* the election comes down to a competitive race between your 1st and 2nd choices.
If you don't Approve your 2nd choice, that throws away your vote for your 2nd choice *if* the election comes down to a competitive race between your 2nd favorite and the candidate you hate.
This is why I am for Ordinal, Ranked-Choice Voting. But I want RCV to be done correctly and IRV is not RCV done correctly (unless it's BTR-IRV).
I can't deny the existence of the incentives you describe. I have long wondered, though, whether it's some sort of psychological necessity that most voters must always prioritize their desire to get their favorite candidate elected when that conflicts with any procedure that is more likely to get them--and everyone else--someone they can live with. In my paper on AV linked above, I suggest that if violating the AV rule regarding which candidate(s) one is supposed to vote for under that procedure is and must always be psychologically impossible (i.e., human beings just can't ever learn to follow any rule that puts desire for their favorite to win below any wish to always live in a not-bad jurisdiction), then we're generally screwed. And of course, if only some people CAN learn the value of a cardinal system, they will be the suckers, because if only they are AV-rule-compliant and others continue to be strategic in ways that violate AV instructions, those others are indeed likely to get their most favorite candidates in office. That would mean that what I take to be the best hope for democracy is a dead letter: we are stuck with some sort of ordinal scheme and the unavoidable problems (re cycles, etc.) that they entail. Those problems, of course, are NOT psychological inclinations that might be fixed by learning. The moral is that if the concepts of democracy, "general will," etc. require that majorities at least usually get representatives whom they approve of, it will have to be concluded that human beings are incapable of democracy!
I don't myself make the relevant disposition an eternally inviolable psychological law myself. But, of course, I might be wrong about that. Maybe human beings are simply incapable of changing in the way AV does seem to require. For what very little it's worth, I hope not. You may say I'm a dreamer--but I'm not the only one!
Just wanted to add that an alleged impossibility of overcoming a disposition to do whatever one can to get one's favorite in every situation seems to me at least a smidge analogous to the claim that an "America First" attitude, according to which it must always seem preferable to sensible Americans to ignore everything but what seems to be in their own best (if short-term) interest is utterly inescapable.
That attitude has not always prevailed everywhere, though it is certainly in the ascendancy around the world today.
This is hopefully my last comment, although perhaps it should have been my first, given that it’s responding to something pretty early in this post.
Ned says that I “relied” on the paper by Nick Stephanopoulos about the rate of Condorcet efficiency of IRV but then ignored his acknowledgement. I can forgive him for not knowing this, because it isn’t something I discussed in detail in my essay, but neither of those claims are really accurate.
First, what I really primarily relied on was the full set of data on RCV elections in US history gathered by FairVote, which I have been following for years and well aware of prior to the Stephanopoulos paper. I _cited_ the Stephanopoulos paper, because it gave a subset of that data a peer-reviewed imprimatur, to assuage any doubts about the reliability of the FairVote data given their stance on this issue. I suppose you could say I “relied” on the Stephanopoulos paper for the additional foreign data set it includes, but the US data was of greater interest and relevance to me and my essay.
The post says I ignored Stephanopoulos’ acknowledgement that the data is “necessarily limited to the candidates on the general election ballot,” because it excludes candidates who “either lost in the primary election or didn’t run in the first place.” I didn’t ignore it; I didn’t find it a particularly persuasive qualification for the purposes of the comparison I gave. In a longer essay, I may have explained why, but I’ll do so here in this comment.
Let’s deal with the question of whether potential candidates in the data set could have been excluded in partisan primaries. First, that is just impossible for many elections in the data set. Many were general elections in the data had no primary; some were themselves primaries; and the Alaska elections were preceded with non-partisan primaries, which couldn’t have suffered from the same kind of candidate exclusion problem, at least not anywhere near the same degree.
For the elections that _were_ preceded with partisan primaries, that is of no relevance to a comparison of IRV and Condorcet. Is your argument that using Condorcet in the partisan primary would have helped to depolarize the electorate? It isn't. You acknowledge that these MAGA Republicans would be the Condorcet candidate of their Republican primaries, and instead envisage “law and order” Republicans running as independents instead.
In sum, the potential issue of primary exclusion in the data set has very little relevance to this comparison of IRV and Condorcet. As for the potential issue of centrists candidates simply opting not to run in IRV elections, I’ve already dealt with that in this comment: https://edwardbfoley.substack.com/p/keep-condorcet-in-consideration-as/comment/88782273
In this comment, I’d like to address the Trump/Haley/Harris example. The conclusion that Condorcet would be more effective than IRV in this election hinges on several assumptions of varying plausibility.
Before we get to those assumptions, it’s worth noting that Trump would have likely lost the 2016 Republican primary under IRV [1]. Any argument about the effect of IRV on Trump or the MAGA movement today is at least a bit incomplete without acknowledging that context. Now onto the assumptions ….
First, the example assumes we are operating under a national popular vote. I won’t dwell too much on this assumption, because I don’t think it’s a necessary one for your point. But it’s not a minor adjustment, as NPV has big implications for turnout. By all accounts, Harris had the much stronger ground game [2]. It wasn’t enough to overcome her disadvantage in key swing states in November, but unleashing that ground game across 50 states might have turned her into the Plurality, IRV, and Condorcet winner.
Of course, another assumption is that Haley would choose to run independent at all. Abandoning the party means forgoing critical campaign lists and donor networks, and it means running head first into the immense ballot access barriers to independent candidates that have been constructed nationwide. It may also mean being excluded from the nationally televised debates, regardless of whether they’re run by the CPD or arranged between the campaigns.
Assuming Haley would run, another assumption is that Harris would have encouraged her supporters to rank Haley over Trump under a Condorcet system. I think that’s doubtful. Campaigns are not altruistic enterprises. Leading candidates always attack their leading opponents, particularly in general elections. If polling starts to show Haley as the leading candidate under a Condorcet system, she becomes the target. What was the Biden’s campaign’s response when Haley started to show signs of strength in the Republican primary? As a Biden spokesperson said at the time, “Nikki Haley is no moderate — she’s an anti-abortion MAGA extremist who wants to rip away women’s freedoms just like she did when she was South Carolina governor.” [3] Despite all the concern expressed about Trump being a threat to democracy, there wasn’t even mild recognition from the Harris campaign that Haley winning the primary would remove that threat. That’s not in the nature of campaigns.
To a large degree, political campaigns are ontologically self-centered entities. As such, bullet voting is their natural inclination. It takes a reform like IRV, which puts ranking in the self interest of the campaign, to counteract that inclination. Even in the early days of Australian IRV elections, despite the lack of any real incentive, there was a tendency of the campaigns to encourage bullet voting – to “otherize” all the other candidates and parties – and it took a bit of time for the incentives of IRV to settle in and undo the tendency (and some jurisdictions made ranking mandatory to nip it in the bud.) Similarly, we saw how Palin encouraged bullet voting in August 2022, before succumbing to the incentives of IRV and endorsing a “rank the red” strategy for that November. It’s not a mindset that comes naturally to campaigns.
With a switch from IRV to Condorcet, we would lose the ability to truthfully and unequivocally tell campaigns and their supporters that ranking beyond their first choice won’t hurt that first choice. It could. (Same with the burying strategy, by the way, but that discussion is for another time.) We thereby lose much of the power to counteract the bullet voting inclination, a much larger share of voters will bullet vote, and campaigns will probably continue their “my way or the highway” attitude about their own candidate.
Finally, the example includes the unstated assumption that Haley would be the _only_ additional strong candidate to run independently. Let’s not forget there was a significant schism on the Democratic side on the issue of Israel/Gaza, arguably stronger than any divide on the Republican side. Under IRV, it’s conceivable that a credible Democratic candidate might have launched an independent run to challenge Harris from the left on this issue; or perhaps Jill Stein would have found greater traction and recognition if free of the spoiler label. A poll by IMEU found that “Ending Israel’s violence in Gaza” was the number one issue for Biden 2020 voters that did not vote for Harris [4]. Thus, a robust campaign to the left of Harris on this issue, might have driven leftwing turnout, bringing voters to the polls that were not motivated enough by Harris to cast a ballot, but who would rank her second if they did.
I’ll reiterate my own predictions. If Haley runs under a Condorcet system, and if polling indicates that she would be the Condorcet winner, I do see Trump and Harris mutually attacking her, encouraging bullet voting, and wooing supporters even further to their extremes to boost their odds. That is not to say that Haley wouldn’t have won anyway – she still might. But with the two major coalitions further polarized and neither seeing their candidate elected, I don’t see the system lasting long. Furthermore, I see several reasons why IRV would enable the entry of stronger third-party and independent candidates in a way that boosts turnout and which propels Harris to victory as the Condorcet candidate through second choice transfers.
That said, I freely admit that my conjectures are conjectures. Ultimately, both your and my hypotheticals are loaded with assumption and speculation. That’s why I keep returning to what _has_ happened in practice and what that can tell us about what will happen.
Despite my disagreements on several points, I do agree with the thesis, i.e. that Condorcet should _not_ be “rejected entirely and everywhere.” Where I probably differ is that I don’t think it should be adopted in the US state-wide until we see it in operation at the municipal level, at least somewhere in the world. And before it is adopted at the municipal level, I’d ideally want to see some research into how it has performed in the, albeit limited, places it is and has been used. Something study akin to what Brams and Fishburn did for approval voting [1] would be a good start.
I heard at one point (and this might turn out to be wrong) that the city of Marquette, MI, keeps a file on its historic use of Nanson’s method. To my knowledge, no scholar has ever written about this use in any detail beyond citing that it happened. I still find it perplexing that there is this leap to advocate for broad adoption for Condorcet methods before any investigation of its past or current use.
Now, I wonder why, Greg, you think that it's always someone else who has to be first? What if that was the argument in the early 19th century regarding Thomas Hare and the Single Transferrable Vote? If a proposed reform is going to take root, someone's gotta be first.
Yea, that's just another dumb talking point by IRV advocates because a) as you said, it's got to be tried somewhere first, and IRV/Hare could've been argued against using that exact same logic and b) IRV orgs seem fond of not wanting non-IRV single winner methods to be tried at the local level because they like IRV better. They're not an "everything bagel" single winner voting method reform group; they are specifically out there to ensure only RCV gets adopted. Greg perfectly encapsulates this because he claims that he just wants to see a study of the efficacy of Condorcet methods, but then he gets into the weeds of hypothetical strategic voting scenarios in order to discredit Condorcet methods so that he can point that out to elected officials and say "so this is why you shouldn't even try a Condorcet method."
As a condition for STV being adopted for the Tasmanian Parliament, it was first tried in two local municipalities for candidate elections. That makes complete sense to me. I would apply the same standard to any other voting method.
IRV orgs spill loads of ink trying to discredit Condorcet with hypothetical doom scenarios like the one you painted in your article so as to scare elected officials away from it. You and your IRV orgs are not interested in advocating for a scientific approach of testing out numerous election methods in many different cities; you are an IRV-only promoter and other methods are vigorously opposed, not supported, in your websites.
Yes, I'm familiar with these uses of the Schulze method for referendum and online decision-making. Importantly, none of those are candidate elections. But sticking with those for just the moment, what have those experiences been like? Have the campaigns resembled candidate campaigns? Has there been any local interest or movement to translate those uses into actual candidate elections? Where is the research studying these experiences to help us understand if there are any lessons for its applicability more broadly? The complete lack of interest in these questions is perplexing.
Robert, you must be unfamiliar with the history and adoption of STV. Before being adopted by the Tasmanian Parliament, it was introduced first to the cities of Hobart and Launceston. In other words, if my conditions were imposed upon STV, it would be exactly where it is today.
Greg, the ballot is the same as for IRV and the meaning of the ballot is the same: If the voter ranks A above B, that means that in an election involving only A and B, this voter is voting for A. It's not Borda nor Score. No points. So, just like IRV, all the ranked ballot is, is the contingency vote information for the particular voter: How would this voter vote in the different contingencies involving different subsets of candidates?
So, every one of those 800 elections you might say are tested as successful, all of them but 4, those IRV elections are "successful" *solely* because they elected the Condorcet winner. In those IRV elections, no voter suffered voter regret for voting sincerely and ranking the candidates sincerely. And the reason they don't regret their vote is because the Condorcet winner was elected. Whether they like the Condorcet winner or not, there is nothing else they could have done to their ballot to get a better outcome from their political POV. No tactical voting would have helped them.
Condorcet RCV is well tested with all of these 800 IRV elections except 4. If those elections performed well, it's because they elected the Condorcet winner. IRV works pretty good when it elects the Condorcet winner, and when it doesn't, it works like shit. IRV is only as good as it can elect the Condorcet winner.
If, instead, exactly the same elections occurred, except they were tallied according to Condorcet rules, the ballots would be identical, the candidates are the same, the voters would mark the ballots exactly the same and the same outcome would result. Except for those 4 elections.
Now 2 of those 4 elections had no Condorcet winner. A Condorcet-TTR method would elect exactly the same candidate.
The 2 elections that had a Condorcet winner that was not elected with IRV are Burlington 2009 and Alaska 2022 (August). Those are the only two elections with a different outcome. Now what might have happened had those two elections elected the center candidate (Montroll in Burlington, Begich in Alaska)? What might have occurred regarding the repeal effort?
At least in Alaska, people would understand that this was the only way Begich would have been elected, otherwise it's Peltola. Do you think the GOP would have been so opposed tp RCV with a GOP winning instead of a Democrat?
In Burlington, the only difference in topology is that it was a *"come-from-behind"* victory. The supporters of Kurt Wright would still be clinging to their misconception that their candidate had the most voter support, whether Bob or Andy were elected. But they had the additional slap in the face when they understood that Kurt ended up as the spoiler and that, despite the promise, their second-choice vote for Andy was never counted. No one told them in advance that, if their candidate loses in the final round, their second-choice vote will not count. Then why should they bother to rank a second-choice?
About history, I only know what's available for the pedestrians. i.e. https://en.wikipedia.org/wiki/History_and_use_of_instant-runoff_voting . Hill, Hare, Andrae. I think it's Hare who is credited with coining the term: *"Single Transferable Vote"*. Condorcet considered IRV before any of them and rejected it for the very reasons pointed out here.
The other dumb thing that IRV people aren't getting is the loss of process transparency that comes with the loss of local tabulation and Precinct Summability. https://drive.google.com/file/d/1YtejO54DSOFRkHBGryS9pbKcBM7u1jTS/view . Consider the recent presidential election in Venezuela. What exposed that election as stolen? This is part of my Installment #4 to your essay in DemocracySOS. Still working on Installment #3, which is responding to your "theoretical analysis". I'm trying to be fair.
The IRV elections are test cases for Condorcet only under the assumption that voters vote and candidates campaign in exact the same way under both voting methods. You seem confident they will be the same, or perhaps that they will only differ in ways that are better under Condorcet. I am highly skeptical of that assumption, for reasons I've given here and elsewhere. It is that precise assumption that I would like to see investigated before further advocacy takes place. I understand your faith is so strong in the assumption that you don't need an actual scientific investigation into it. I think that's the crux of the disagreement.
The ballots are exactly the same (with one exception that in nearly any Condorcet method, not BTR-IRV, voters may equally rank multiple candidates). The meaning of the ballots are exactly the same (rank A above B is exactly one vote for A).
Then there are these promises like RCV will elect the candidate having a simple majority of voter support, that the spoiler effect is avoided, that voters are free to vote their hopes instead of their fears (that they can vote for whoever they really want without risk of helping elect the candidate they hate), that whenever they cannot get their first-choice candidate, their second-choice vote is counted instead.
So we engineers like to put things into black boxes, in order to test hypotheses. There's an input to the black box and there is an output. The input are the actual cast vote records for all of these "800" elections you refer to. In nearly all of these elections, whether it's Hare or Condorcet in the box, the output is the same and the promises above are kept.
But in two elections, we have the same input, but different outputs. One of the systems in a block box actually materially satisfy the conditions promised and the other system in the other black box measurably fails to satisfy the conditions promised. They are not equal. One of them works to do the job and the other one does not.
And this *still* hasn't even touched the issue of the loss of process transparency due to the loss of decentralized local tabulation of the vote that we get right now with FPTP. This component of process transparency, called "Summability" or "Precinct Summability", is the ability to know how an election will turn out, just from adding up the tallies published at each polling place. While Summability did not save the Venezuelan presidential election from being stolen (dictators will dictate), it *did* expose that election as stolen. If Venezuela had an independent judicial system that had teeth, Maduro would not be the president right now. Precinct Summability (and the rule of law and a strong and uncorrupt judicial system) keeps elections honest.
Why would you want to sacrifice the security of process transparency (which might allow a corrupt election official in Georgia to "just ,... find, uh, 11780 votes") just so you can keep doing things the same way and occasionally failing at delivering exactly what it is we want RCV to do? When this sacrifice is unnecessary?
Ned, first, thank you so much for responding to the points in my essay. I read and respect a lot of your work, but do not have many disagreements with you beyond this issue. There’s a lot to respond to here, and to keep things somewhat organized, I think I’ll devote one comment to each of the responses.
At multiple points, this essay suggests that Condorcet candidates may bow out of IRV races, knowing they will be eliminated early. To my knowledge, no one has ever presented any evidence that this is a significant feature of real IRV races to date. To the contrary, there were 13 candidates in the 2021 NYC Democratic mayoral primary, and 17 in the 2021 Minneapolis mayoral general election. If anything, one might wish for a few to bow out :)
Furthermore, it’s not fair to compare the incentives to bow out under plurality to those under IRV. Candidates bow out under plurality in part because they realize they have no shot, yes, but perhaps more often because _others_ believe they can’t win, causing the donations to dry up to the point where no campaign can actually be run. They also bow out because they are shamed and derided as “spoilers,” and they don’t want that label attached to them for the remainder of their political career. Under plurality, the “I won’t vote for (or donate to) you, because you can’t win” quickly becomes a self-fulfilling prophecy.
Under IRV, that underdog candidate has pretty strong incentives to run and stay in the race. First, those people who would otherwise say “I don’t want to throw my vote away” can now take a chance on you as their first choice. Although you may start in the back of the pack, you could build your support through a "virtuous cycle" of voters taking a chance, an inversion of the viscious cycle that plurality offers. Second, even if you doubt your ability to do that, running offers you a way to elevate your political profile and key issues without the spoiler label being affixed. Third, in the election itself, you can play kingmaker. To earn your second choice endorsement, the major candidates may offer you policy concessions or influence in their administration.
I know it wasn’t a major point of your post above, so apologies for dwelling on it, but whatever the reasons for the rarity of Condorcet failure in IRV elections, low candidate entry does not seem to be one of them.
"To my knowledge, no one has ever presented any evidence that this is a significant feature of real IRV races to date." How does one present evidence that someone didn't run due to IRV? Maybe both sides of the "Condorcet winner who chooses not to run" debate have no way to measure it. But it stands more to reason that since IRV exhibits the spoiler effect due to center squeeze (while Condorcet does not), IRV elections will feature the same tendency for candidates that are strong but maybe not seen as top-2 frontrunners to stay out of the race.
Your examples of the 13 and 17 candidate races proving that IRV doesn't cause potential Condorcet winners to bow out is a red herring because you are treating the existence of many fringe candidates in the race as being evidence that a strong centrist candidate who would've been the Condorcet winner also not bowing out. Perennial and fringe candidates have existed forever under plurality voting, so it's not some unique feature of IRV that candidates who normally wouldn't win will decide to run because they aren't going to be "derided" as a spoiler.
Your 4th paragraph is just standard IRV agitprop because nothing you say about IRV in that paragraph couldn't also be said of a Condorcet method. You are pretending that it's a unique feature of IRV in order to distract from the fact that IRV exhibits the very pathologies that organizations such as yours claim don't exist. I promise you, you can just admit your organization was wrong about its core claims and choose to advocate for a Condorcet method that won't exhibit the pathologies of IRV.
I know that you probably know that no single winner method is 100% free of pathologies (Arrow, Gibbard, etc.), but your organization straight up lies on its website that IRV doesn't exhibit the very pathologies that it exhibits as a matter of both theory and in practice. It's always the same thing with IRV orgs: "other methods can elect the wrong winner, but not ours, nosiree." Condorcet advocates never say that Condorcet methods can't ever be free of pathologies that can elect the "wrong" winner, but IRV advocates routinely say IRV is free of any pathologies that can elect the "wrong" winner, and that's just intellectually dishonest at best and fraud in advertising at worst.
"How does one present evidence that someone didn't run due to IRV?"
Well, you could start with at least one anecdote. We have plenty of examples of candidates bowing out of races today under plurality, both before and during the election, because of the pathologies of plurality. Many people, for example, asked why Bernie Sanders, after a life of running as an independent candidate, chose not to run as a third-party or independent candidate in the general election in 2016. He answered "What I did not want to do is run as a third party candidate, take votes away from the Democratic candidate and help elect some right-wing Republican. I did not want responsibility for that." That's pretty decent evidence that the voting system denied an additional strong candidate in that general election.
In 20 years and 800 elections under IRV, we have yet to have even a single anecdote of a credible centrist saying they'd stay out of the race due to anything attributable to a pathology of IRV.
Finally, I have never stated that "IRV is free of any pathologies," as you say. In fact, my essay cites the center squeeze pathology openly.
Do you consider yourself a Condorcet advocate? If so, what is an example election where you consider Condorcet to elect the "wrong" winner?
The organization you are affiliated with (and every other IRV advocacy group) makes two statements that unequivocally state that IRV doesn't suffer from pathologies that it does suffer. From https://voterchoicema.org/ :
1. Ranked Choice Voting Ensures Majority Support by eliminating the “spoiler effect” and guaranteeing the winner earns a majority of the votes in any election. FALSE
2. Ranked Choice Voting Expands Voter Choice by freeing you to vote for who you really want, without settling for the “lesser of two evils,” and without fear of “wasting” your vote. FALSE
So long as you are affiliated with any IRV advocacy group that advertises IRV falsely, you are effectively stating that IRV is free of pathologies on par with being a magic supplement that cures everything with no side effects.
Also, anecdotes aren't data, so not sure why you responded with my question with how to present evidence to actually do a study with a request for an anecdote. I actually conceded there was no actual way to know the full population of people who didn't run because of IRV, but you are assuming because there isn't an anecdote where someone actually said the words "I don't think I can win an IRV election, therefore I'm not mounting a campaign" that it's not a calculation that is affecting decisions to run for office. The thing is that you actually have no idea how many times a "credible centrist" looked at internal polling that showed them at 15%-20% in a three way race and they quietly decided to not run. It's like saying that no crime is happening because it's not being reported; a lack of observation doesn't invalidate the hypothesis.
What's funny is that you are doing the thing of inventing hypothetical scenarios where voters try to game the system while also criticizing Condorcet for having no real world examples of being tried. How do YOU know that voters will vote strategically that way if there are no real world examples? What we do know is that in both the Burlington and Alaska examples, Wright and Palin should have told their supporters to abandon them in favor of Montroll and Begich, respectively, yet didn't do so despite it being strategically better for their state to not elect the most "far left" candidate in the race. It seems that candidates are not likely to "stand down" in the face of certain defeat and instruct their voters to vote strategically, so why are you assuming they will do that in a hypothetical Condorcet election when they didn't do it in actual IRV elections?
As for being a Condorcet advocate, it's less about advocating for Condorcet and more about truth in advertising, and your organization currently has false claims on its website and other IRV organizations have been putting out those same lies about IRV for decades. That these obvious lies seem to barely have gotten IRV nowhere near widespread adoption makes me wonder why you sacrifice intellectual integrity for such meager results. It seems that IRV orgs can't even sell a free lunch, but maybe because people realize it's a toadstool and sauerkraut sandwich covered in arsenic sauce. I'm actually more interested in PR than single winner, but can't stand the intellectual bankruptcy of the IRV people.
I said you could start with anecdotes, but you don't have to end with one. One could conduct a statistical study of "candidate exit" of the kind available in past polisci literature (example [1]). I put the challenge of identifying such an anecdote as a low bar, but if you have a candidate exit study, or you plan to conduct one, I'd be happy to review it.
Yawn. Still not addressing the core fact that your organization pushes easily disproven lies, lies which have real consequences for how people choose to vote under IRV. I bet Wright-Montroll and Palin-Begich voters would love to know why your organization said that voting for their first choice wasn't wasting their vote on them when it actually did. I mean, I'd never lie to people like that, but you do you I guess.
I really don't care too much about your question about "candidate exit" because you can't even defend IRV's main defect and your organization's role in lying about how it doesn't exist. It probably does happen from time to time that a candidate drops out due to how IRV works, but you really can't measure it.
The study you linked to examines money raised, which is an objective thing that a political science researcher can pull from a website. What if someone chooses not to run before raising a single penny? If they choose to drop out after raising money and then seeing initial internals that say they are coming in third by "too much" and decide to drop out, how is that measured in a dataset with any way of knowing if someone didn't put out an official statement? Maybe they had family/medical issues, or felt another candidate entered the race that was better than them and agreed with the same basic issues... it literally could be anything, whether you could measure the money or not. The paper, from what I could tell, had no dataset that told us WHY a candidate dropped out. It is only telling us that candidate exit is highly influenced by money raised, controlling for other factors, but in no way are the authors able to figure out if money raised was due to "IRV viability," or a scandal, or changing political tides. It has the not-surprising conclusion that "candidate who doesn't get money drops out at higher rates than those who do get money," so don't know why it's relevant to our discussion.
I did admit it's pure conjecture, but also it's conjecture that should be taken seriously because we know that candidates drop out due to viability concerns all the time in plurality, so why not in IRV? Candidates could easily also drop out due to viability in a Condorcet election, but we're specifically talking about whether a Condorcet candidate would drop out of an IRV election, which they would obviously not do if it were a Condorcet election.
I think that's where you're having trouble dealing with this conceptually (amongst other things). Foley specifically is talking about Condercet candidates dropping out of elections that happen using methods that suffer from failure of the Condorcet criterion, which both plurality and IRV fail. Does a Condorcet candidate run if they think they can't win an IRV election, the same that a Condorcet candidate may not run in a plurality election?
You used Bernie as an example, but read what he said again: "What I did not want to do is run as a third party candidate, take votes away from the Democratic candidate and help elect some right-wing Republican. I did not want responsibility for that." He's basically saying here that he's not the Condorcet candidate and that Hilary would be the Condorcet candidate for whom he'd be spoiling the election. Whether he'd run in an IRV contest as a non-Condorcet candidate is not the issue because he isn't expecting to be the Condorcet candidate. The issue is a) whether it would affect the decision to run for a candidate if they were the Condorcet candidate but probably couldn't get to the final 2 round in an IRV race and b) is it desirable for the Condorcet candidate to win.
With respect to a), again, I don't know because politicians run for office based on a host of factors, but viability to win PER THE RULES OF THE ELECTION METHOD has always been a factor, so we shouldn't not think it would matter here. And with b), IRV orgs are obstinate in thinking that no matter how close the 3rd place person is to the 2nd place person, that somehow disqualifies them from being considered as a credible candidate (whether the margin is 5%, 3%, 1%, or one vote). You seem fine with vote splitting when it affects the penultimate round (since you support IRV), which is an odd stance to take for someone who claims to not like multiple candidates splitting the vote so as to not determine the "true majority winner."
Just one more anecdote I forgot to mention. Independent Tiffany Bond, who ran as a "centrist" alternative to Golden and Poliquin, explicitly said she would not have run in the 2018 Maine House race had IRV not been in place. Many examples of these candidates running and staying in the race under IRV.
This is a tangent, but note that Bond wasn't the Condorcet candidate in this race, despite being the "centrist." There is something else to be said here about the assumption that the candidate objectively in the "center" on policy is necessarily the Condorcet candidate -- an assumption that pervades much of this commentary and in which I indulged, for the sake of argument, in my essay. I'll try to say more of that in another comment, if I have time.
Again, this is a red herring. Foley is claiming that under IRV, a centrist candidate CAN be squeezed out by the IRV elimination algorithm, not that the centrist candidate is always the Condorcet candidate. Your twisting of his example into something he didn't say is indicative of the deceptive form of argumentation that IRV advocates are known for, so not surprised.
Another assumption you are making is that because Bond branded herself as the centrist alternative to Golden and Poliquin, that automatically makes her the centrist in the race. But if Bond wasn't the Condorcet candidate, then by the determination of the voters themselves, she wasn't likely the centrist candidate in reality. Golden, being the Condorcet candidate, has more claim to being the centrist in the race than does Bond, but you treat candidate advertising as truth rather than data. This is another example of IRV advocates writing an anecdote in a way to promote IRV and pull down Condorcet that is just factually incorrect.
Foley has offered several examples, including Rob Portman, Nikki Haley, and others, and assumes they are the Condorcet candidate because he judges them to be politically centrist. He is one operating on that assumption in this piece and others.
What's your point? He's assuming they are and pointing out that using hypothetical ballots that under IRV that they wouldn't get elected despite being preferred to the other candidates in head to head matchups.
You were the one assuming that Bond was the centrist candidate despite actual ballot data that points to the opposite conclusion as she wasn't the Condorcet candidate. You are the one making editorial decisions that contradict actual facts, but that's par for the course for IRV advocates.
Using ACTUAL ballot data, it's clear that in the 2009 Burlington mayoral race, the Democrat Montroll was the centrist candidate; using ACTUAL ballot data, it's clear that in the 2022 Alaska special election race, the non-MAGA Republican Begich was the centrist candidate. They did FEEL like the centrist candidate in both races, but until the ballots were cast, we didn't know who was closest to the center of the electorate.
You seem to be like every other IRV advocate in that you are intellectually dishonest. You know that Foley is using hypothetical examples to prove a point about how IRV doesn't elect the centrist candidate but say that he's making unfounded assumptions about who is a centrist (despite it being hypothetical). But you instead choose to take real ballot data that shows a candidate as not being the Condorcet candidate and claiming Condorcet doesn't always elect a centrist despite having no objective way to determine whether Bond is actually the centrist. Other than she has no D or R next to her name or based on her self-serving campaign branding, how can you possibly know she's the centrist candidate if not for the actual ballot data?
>>You seem to be like every other IRV advocate in that you are intellectually dishonest. You know that Foley is using hypothetical examples to prove a point about how IRV doesn't elect the centrist candidate but say that he's making unfounded assumptions about who is a centrist (despite it being hypothetical).<<
Ya know, the dumb thing is, Logan, is *of course* we know who the Center candidate is. This is the candidate who, in the semifinal round, gets far more 2nd-choice votes from both of the other candidates than these candidates receive from the other. Right cannot expect a lot of 2nd-choice votes from voters for Left. Nor can Left expect a lot of 2nd-choice votes from voters for Right. It's because *both* sets of "extremist" voters are marking Center as their 2nd-preference instead of the "extremist" candidate on the other side.
>>You seem to be like every other IRV advocate in that you are intellectually dishonest.<<
And I agree. There is something to this *persistent* intellectual dishonesty. I *want* to have good and productive conversations with other voting systems reformers and intellectuals, but these IRV/FairVote apologists seem to think that civil conversation means that they can literally change the facts. The numbers. And the meanings of the numbers. That insults the intelligence of the rest of us and that becomes wearisome.
I understand how someone can promote IRV when they're not aware of center squeeze, but I cannot understand why they would be so stubbornly persistent in promoting IRV even after understanding its biggest flaws.
The FairVote crowd have really lost the plot, and are just harming their own cause at this point, continually pushing a broken voting system that doesn't actually fix the things they claim to want to fix, and that's getting banned in more places than it gets adopted.
>>...why they would be so stubbornly persistent in promoting IRV even after understanding its biggest flaws.<<
Because they are selling a product that they don't want to admit to customers, particularly the previous customers that had purchased the product, that the product they sell has an avoidable flaw. They want to say that "We've been using IRV for decades and it's been working just fine" and continue on in the same direction hoping that no one (or very few) else will notice the flaw or how the flaw is an avoidable flaw.
My daughter compared it to software revisions. We can call First-Past-The-Post "Democracy v1.0" And IRV (under the moniker of "RCV") "Democracy v2.0" Why can't they just acknowledge the bug, fix the bug, and call it "Democracy v2.1"? They cannot admit to overselling the features of the product in the first place.
We need to remember that, in order to value our votes equally, that Single-Winner, Multi-Winner, and Apportionment of Delegates (Presidential primary) are very different problems.
1. Single-Winner: There is no proportionality to be had with a single winner. It's winner-take-all. So then, the *only* way to value our votes equally is Majority Rule. And Condorcet RCV respects Majority Rule in elections where Hare RCV does not. Condorcet is clearly better at respecting Majority Rule and the equality of our votes.
2. Multi-Winner: Now proportionality comes into the equation and then something like the Weighted-Inclusive Gregory Method (PR STV) might be the method that best reflects apportioning multiple seats in a single legislative district. This, of course, has quota, surplus votes that get transferred, and *then* a Hare-ish elimination of candidates.
3. Apportionment of state delegates to the DNC or RNC: This is a solved mathematical problem and it's used to apportion 435 U.S. House Representatives to the 50 states. Same math, and that method should be used to apportion a known number of delegates to candidates, based on the number of votes that each candidate gets.
Great post. Your point about a candidate who would win a Condorcet-based election to refraining from competing because there is no point doing so because of the way IRV elects winners is one that IRV supporters are blind to, whether willingly or not. Marcus Ogren dubbed it the primordial election here: https://voting-in-the-abstract.medium.com/the-primordial-election-that-is-never-held-e019356faf90 . Since IRV by its own mechanism can throw out a moderate that is preferred by 2/3 of the voters over the other two candidates whilst having almost identical first choice support, candidates who do the math and find they might get "only" 20%-25% of the first place vote will choose to not waste the time and effort to campaign. Because even if they overperform and get to 30%, there's still a good chance they still get eliminated in IRV's penultimate round.
To your point about strategic voting, IRV supporters like talking about the way that other methods can be gamed by voters to change the result in their favor, but they never talk about how voting honestly with IRV causes maximum voter regret. The 2009 Burlington mayoral race and the 2022 Alaska special election showed that voters for Kurt Wright who preferred Andy Montroll to Bob Kiss and voters for Sarah Palin who preferred Nick Begich to Mary Peltola, respectively, didn't get the result they wanted because they voted their honest ranking. Had they known that voting sincerely would result in the election of their least preferred candidate, they would have strategically voted for their second choice as their first choice. In effect, under IRV, voters would be better off strategically voting in order to elect the Condorcet candidate; one might say it would be better to avoid this situation altogether and just use a Condorcet method so they don't have to either experience regret or have to betray their favorite to not elect their least favorite choice.
IRV seems built in a lab to be the anti-Condorcet method, not a Condorcet-like method that Greg Dennis seems to want to prove. Not only does it disincentivize potential Condorcet winners from even competing, it explicitly throws out all Condorcet winners that don't meet the strict IRV elimination thresholds. When Fairvote discussed the cases of Burlington and Alaska, they basically said that "the Condorcet winner didn't deserve to win because they weren't the IRV winner": https://fairvote.org/resources/data-on-rcv/#condorcet-winners They characterized Montroll's 29% as "little core support" and Begich's 28% as "low support" but somehow Kiss's 34% and Palin's 31% were not "low." In reality, Candidate C could come up one vote below Candidate B in the penultimate round and still be eliminated, and the Fairvotes and Greg Dennises of the world would say "well, Candidate B was obviously more worthy than Candidate C to be considered in the final runoff round because they had more 'core support'", as if core support was the only consideration in who was worthy to win. Also, the notion of more "core/enthusiastic/deep" support driving elimination consideration is laughable considering an IRV election may have 7-10 candidates initially where the eventual winner started out with 10%-15% of first place votes, so by the time the penultimate round comes around, the two candidates with higher "core/enthusiastic" votes really have a smaller fraction of their votes being "core/enthusiastic" and a higher fraction of their vote total being the less enthusiastic lower rankings piling up.
>>The 2009 Burlington mayoral race and the 2022 Alaska special election showed that voters for Kurt Wright who preferred Andy Montroll to Bob Kiss and voters for Sarah Palin who preferred Nick Begich to Mary Peltola, respectively, didn't get the result they wanted because they voted their honest ranking. Had they known that voting sincerely would result in the election of their least preferred candidate, they would have strategically voted for their second choice as their first choice.<<
Logan, the real shame in these two examples is that these Wright or Palin voters (who disliked Kiss or Peltola the most) were promised that they could safely vote for their favorite candidate without fear of causing the election of their least-favorite candidate. They were literally promised that if they couldn't get their first choice, then their second-choice vote is counted. That, of course, is never true for voters that vote for the loser in the IRV final round. It's just that in most IRV elections, that loss of one's second-choice vote does not make a difference in the outcome. But in these two elections, it did.
Yes, IRV advocates routinely make the claim that one can "vote for their favorite without fear that their least favorite will be elected" despite it being plainly false both in theory and in practice, most recently for an office as important as a US House seat. And there's no reason to not believe that the same would happen in a Presidential election if IRV was adopted nationwide. 1992 Perot is probably the best example where either Bush or Clinton would've been the Condorcet loser and so Clinton-Perot and Bush-Perot voters would've been better off voting for Perot first. Had IRV been in effect, Perot may have gotten very close to getting in the final round, but still could've been eliminated in favor of the Condorcet loser making the final round. IRV advocates are dumb because their voting method essentially says "yea, the Condorcet loser will by definition always lose an IRV election, but let's put them into the final round for funsies." This isn't like a football playoff where the results of the next round are unknowable so you "have to play the game to see who wins;" the results of the final round are already set in stone, but IRV likes to pretend it's not so they can set up a kabuki final elimination round to fool voters into thinking there was an actual competition between the final two candidates when the loser was already decided.
Greg Dennis's organization Voter Choice Massachussets makes these two claims right on their front page in big bold letters:
1. Ranked Choice Voting Ensures Majority Support by eliminating the “spoiler effect” and guaranteeing the winner earns a majority of the votes in any election.
2. Ranked Choice Voting Expands Voter Choice by freeing you to vote for who you really want, without settling for the “lesser of two evils,” and without fear of “wasting” your vote.
#1 is a combination of straight up lie and lie of omission. As demonstrated earlier, there is a "spoiler effect" in IRV elections, just not the 2000 Ralph Nader type. Montroll and Begich would have won their elections had eventual non-winners Wright and Palin not "spoiled" the elections for them by entering the election, same as non-winner Nader. And the "guaranteeing the winner earns a majority of the votes in any election" point omits that an IRV winner can win despite not earning a majority if the winner was not also a Condorcet winner, because on the actual ballots, by definition they didn't win a majority against the Condorcet winner.
#2 is basically a different way to say it eliminates the "spoiler effect," so obviously is another lie. Voters for Kurt Wright who preferred Andy Montroll to Bob Kiss and voters for Sarah Palin who preferred Nick Begich to Mary Peltola should have settled for the "lesser of two evils" so that their least favorite candidate didn't win, but since they were told by orgs like Fairvote and VCM to not fear ranking their favorite first, they didn't strategize as they might have in a plurality election to ensure they didn't experience maximum regret. Their vote (really their first place ranking) was "wasted" on a candidate that had no chance to win because they were the Condorcet loser, same as Nader voters in 2000. I reiterate that orgs like Fairvote and VCM don't characterize candidates like Wright and Palin as "spoiler candidates" despite that being exactly what they are, so don't know if they're lying because they're committed to IRV ideology or because donors like the fact that IRV has a "center squeeze" problem so want to perpetuate lies that pretend it doesn't exist.
Yes. I didn't go to Greg's site (recently) but I have seen these falsehoods repeated, multiple times, with a variety of different RCV promotional organizations. I have pointed out the falsehood, how it's false, why it's false, and why they should be qualifying the claim enough to make it true. But they don't do that because the simple claim, albeit false, has better marketing appeal.
Another one (from Better Ballot Vermont and several other orgs) is: "To win an RCV election, a candidate must get over 50% of the vote." Straight-out falsehood. Easily proven false by use of counter-examples.
A lie from RCVRC I got in the email in 2022: "Does ranked-choice voting impact how long it takes to know who won the election? ... No! Ranked-choice voting elections can be tabulated as quickly as a few minutes using round-by-round counting software." The disingenuity comes from the omitting the necessity to centralize the ballots or equivalent voting data from all of the polling places. In Maine, it took 4 or 5 days. In Alaska it takes 15 days after the election before they run the IRV software.
I'm trying to be polite to the IRV apologists, but they ain't being honest, and I am not always patient. I am eschewing calling anyone a "liar", but this is hard. Especially with Trumpers. But the IRV apologists just need to stop with the falsehoods, because if they don't, they're gonna be called out. Eventually by me, if no one else does.
"But they don't do that because the simple claim, albeit false, has better marketing appeal."
Yes, IRV advocates are fond of using factually false claims for better marketing appeal. Consumer law would call that fraud, and it's what I'd call it. It's the equivalent of a pharmaceutical company saying that you won't get diarrhea if you take their pill despite the documented cases of diarrhea after taking the pill.
The fact that they know it to be false and keep parroting the lie front and center on their websites and other communications qualifies them as serial liars who shouldn't be taken seriously as anything resembling scholars of election methods.
If I could do an "angry emoji" I would, Logan.
Yes, "... guaranteeing the winner earns a majority of the votes in any election" is a straight-up falsehood. Greg, why do you continue to prominently post that falsehood? I remember pointing this out to you before.
To put a bow on the last point about IRV rewarding candidates with "more core support," it's obvious that IRV supporters really only considered elections whereby third party spoiler candidates siphoned off less than 5% of the vote. These candidates "spoiled" the election of a candidate who got 46%-48% of the vote to lose to another candidate who didn't have a spoiler who got 47%-49%, and IRV would fix that so that candidates with "high core support" didn't lose because of candidates with "low core support" spoiling the election.
But when confronted with the possibility of an election being 33.5%-33.3%-33.2% whereby the 33.2% candidate is the 2nd choice of the 66.8% of voters who supported the other 2 candidates, IRV supporters just agreed to characterize any candidate that placed third in the penultimate round as a weak spoiler candidate. They did this as a way to deflect criticism about their method not electing the Condorcet candidate based on inflexible elimination criteria that treated a candidate that came up one vote shy of being second the same as a candidate who got only one vote.
Basically, IRV supporters don't think that candidates who make the final round but ultimately lose as "spoilers" for the third place candidate who got eliminated in the penultimate round but would've won in the final round, but that's what they are. Kurt Wright "spoiled" the election for Montroll in the 2009 Burlington mayoral election, because by the stated preferences of the voters, they preferred Montroll over Kiss, but when Wright entered the race, he caused Montroll to lose despite not having a chance to win using an IRV election method. This is functionally the same as Nader causing Bush to win by entering the 2000 election despite not having any chance to win using the plurality election method, but IRV supporters cannot see that.
>>But when confronted with the possibility of an election being 33.5%-33.3%-33.2% whereby the 33.2% candidate is the 2nd choice of the 66.8%<<
You might have noticed that I posted a form of that extreme example with 99 voters in the comments to Greg's column posted in DemocracySOS.
And BTW, I happen to live in Burlington Vermont (I'm actually in Bernie's ward and work as an election worker and had previously been an elected polling official in this ward). I was here in 2005 when we voted to adopt IRV and was here when the 2009 IRV election occurred and the repeal that happened the year after. The really sad thing is how short the memories of people are. We readopted IRV after 13 years. But I am active in the statehouse where there are some legislators that are beginning to understand this problem.
One thing I would like to resist in this article is the claim Ned makes with respect to the hypothetical Ohio and North Carolina elections he discusses. It is that "Only a Condorcet-consistent electoral system would enable the will of the majority to prevail and elect the rule-of-law Republican instead of the election deniers." In my view, that assertion requires the unsupported assumption that capture of "the general will" must always be a function of favorite picking. If we instead consider the purpose of a single winner election to land on that candidate that most of the electorate thinks would be minimally acceptable (or better than nothing), then no Condorcet-consistent rule is likely to do the trick. And, I take it, using minimal approvability rather than finding the round robin favorite as the goal is just as conducive to "finding somebody in the middle."
What I think we ought to want in single winner elections is to have results that most of the citizenry can live with without raising Cain. That is the goal of Approval Voting. The complaint that it is subject to strategic manipulation through bulleting is significant, but it is not alone in that. At any rate, I am not convinced that modern electorates are incapable of ever coming to understand the benefits of almost always having elected officials who are OK with them even if that means the majority sometimes won't get their fave. I mean, look at the alternatives.
I served on something called The Elections Modernization Committee with Greg for three years in our hometown of Arlington, Mass. and we argued pretty regularly about the relative merits of his beloved RCV and my own preference for Approval Voting. Both FairVote and CES (The main AV support organization) have, over the last few years, been willing to trade a number of what I take to be foundational principles for defensible voting schemes for success in this or that jurisdictional election. It ought to be clear by now that such trades haven't worked particularly well for either group. But, alas.....
In my view, there isn't any particular method that will be perfect for every sort of election, and the religious attachment to one or another procedure for every conceivable sort of group decision is unwise. Furthermore, I don't think that any single-winner method, even AV, should be used exclusively in any election that, to satisfy other important principles, ought to have more than one winner. But I continue to believe that there are certain advantages to AV that are ignored in a lot of the lit I have seen over the last few years. For anybody interested, I suggest a look at chapters 7 and 8 of my book on democratic theory, this paper on a few neglected virtues of AV: https://www.qeios.com/read/ZETKEQ, and a paper that I hope to be forthcoming in the journal Prolegomena. I have also discussed a couple of Ned's (and Drutman's and Pildes's) recent writings on Condorcet, and PR here: https://luckorcunning.blogspot.com/2025/01/ned-foley-richard-pildes-and-larry.html and here: https://www.3-16am.co.uk/articles/larry-diamond-edward-b-foley-and-richard-h-pildes-eds-electoral-reform-in-the-united-states-proposals-for-combating-polarization-and-extremism?c=a-hornbook-of-democracy-book-reviews Comments are most welcome.
I will read this, Walter. My initial pre-reading observation about Approval Voting is that, like all Cardinal methods, AV inherently forces voters to think tactically the minute they walk into the voting booth if there are 3 or more candidates running. The voter has to decide how much to Score or whether to Approve their second-favorite candidate.
If you Approve your 2nd choice, that throws away your vote for your 1st choice *if* the election comes down to a competitive race between your 1st and 2nd choices.
If you don't Approve your 2nd choice, that throws away your vote for your 2nd choice *if* the election comes down to a competitive race between your 2nd favorite and the candidate you hate.
This is why I am for Ordinal, Ranked-Choice Voting. But I want RCV to be done correctly and IRV is not RCV done correctly (unless it's BTR-IRV).
I can't deny the existence of the incentives you describe. I have long wondered, though, whether it's some sort of psychological necessity that most voters must always prioritize their desire to get their favorite candidate elected when that conflicts with any procedure that is more likely to get them--and everyone else--someone they can live with. In my paper on AV linked above, I suggest that if violating the AV rule regarding which candidate(s) one is supposed to vote for under that procedure is and must always be psychologically impossible (i.e., human beings just can't ever learn to follow any rule that puts desire for their favorite to win below any wish to always live in a not-bad jurisdiction), then we're generally screwed. And of course, if only some people CAN learn the value of a cardinal system, they will be the suckers, because if only they are AV-rule-compliant and others continue to be strategic in ways that violate AV instructions, those others are indeed likely to get their most favorite candidates in office. That would mean that what I take to be the best hope for democracy is a dead letter: we are stuck with some sort of ordinal scheme and the unavoidable problems (re cycles, etc.) that they entail. Those problems, of course, are NOT psychological inclinations that might be fixed by learning. The moral is that if the concepts of democracy, "general will," etc. require that majorities at least usually get representatives whom they approve of, it will have to be concluded that human beings are incapable of democracy!
I don't myself make the relevant disposition an eternally inviolable psychological law myself. But, of course, I might be wrong about that. Maybe human beings are simply incapable of changing in the way AV does seem to require. For what very little it's worth, I hope not. You may say I'm a dreamer--but I'm not the only one!
Just wanted to add that an alleged impossibility of overcoming a disposition to do whatever one can to get one's favorite in every situation seems to me at least a smidge analogous to the claim that an "America First" attitude, according to which it must always seem preferable to sensible Americans to ignore everything but what seems to be in their own best (if short-term) interest is utterly inescapable.
That attitude has not always prevailed everywhere, though it is certainly in the ascendancy around the world today.
This is hopefully my last comment, although perhaps it should have been my first, given that it’s responding to something pretty early in this post.
Ned says that I “relied” on the paper by Nick Stephanopoulos about the rate of Condorcet efficiency of IRV but then ignored his acknowledgement. I can forgive him for not knowing this, because it isn’t something I discussed in detail in my essay, but neither of those claims are really accurate.
First, what I really primarily relied on was the full set of data on RCV elections in US history gathered by FairVote, which I have been following for years and well aware of prior to the Stephanopoulos paper. I _cited_ the Stephanopoulos paper, because it gave a subset of that data a peer-reviewed imprimatur, to assuage any doubts about the reliability of the FairVote data given their stance on this issue. I suppose you could say I “relied” on the Stephanopoulos paper for the additional foreign data set it includes, but the US data was of greater interest and relevance to me and my essay.
The post says I ignored Stephanopoulos’ acknowledgement that the data is “necessarily limited to the candidates on the general election ballot,” because it excludes candidates who “either lost in the primary election or didn’t run in the first place.” I didn’t ignore it; I didn’t find it a particularly persuasive qualification for the purposes of the comparison I gave. In a longer essay, I may have explained why, but I’ll do so here in this comment.
Let’s deal with the question of whether potential candidates in the data set could have been excluded in partisan primaries. First, that is just impossible for many elections in the data set. Many were general elections in the data had no primary; some were themselves primaries; and the Alaska elections were preceded with non-partisan primaries, which couldn’t have suffered from the same kind of candidate exclusion problem, at least not anywhere near the same degree.
For the elections that _were_ preceded with partisan primaries, that is of no relevance to a comparison of IRV and Condorcet. Is your argument that using Condorcet in the partisan primary would have helped to depolarize the electorate? It isn't. You acknowledge that these MAGA Republicans would be the Condorcet candidate of their Republican primaries, and instead envisage “law and order” Republicans running as independents instead.
In sum, the potential issue of primary exclusion in the data set has very little relevance to this comparison of IRV and Condorcet. As for the potential issue of centrists candidates simply opting not to run in IRV elections, I’ve already dealt with that in this comment: https://edwardbfoley.substack.com/p/keep-condorcet-in-consideration-as/comment/88782273
In this comment, I’d like to address the Trump/Haley/Harris example. The conclusion that Condorcet would be more effective than IRV in this election hinges on several assumptions of varying plausibility.
Before we get to those assumptions, it’s worth noting that Trump would have likely lost the 2016 Republican primary under IRV [1]. Any argument about the effect of IRV on Trump or the MAGA movement today is at least a bit incomplete without acknowledging that context. Now onto the assumptions ….
First, the example assumes we are operating under a national popular vote. I won’t dwell too much on this assumption, because I don’t think it’s a necessary one for your point. But it’s not a minor adjustment, as NPV has big implications for turnout. By all accounts, Harris had the much stronger ground game [2]. It wasn’t enough to overcome her disadvantage in key swing states in November, but unleashing that ground game across 50 states might have turned her into the Plurality, IRV, and Condorcet winner.
Of course, another assumption is that Haley would choose to run independent at all. Abandoning the party means forgoing critical campaign lists and donor networks, and it means running head first into the immense ballot access barriers to independent candidates that have been constructed nationwide. It may also mean being excluded from the nationally televised debates, regardless of whether they’re run by the CPD or arranged between the campaigns.
Assuming Haley would run, another assumption is that Harris would have encouraged her supporters to rank Haley over Trump under a Condorcet system. I think that’s doubtful. Campaigns are not altruistic enterprises. Leading candidates always attack their leading opponents, particularly in general elections. If polling starts to show Haley as the leading candidate under a Condorcet system, she becomes the target. What was the Biden’s campaign’s response when Haley started to show signs of strength in the Republican primary? As a Biden spokesperson said at the time, “Nikki Haley is no moderate — she’s an anti-abortion MAGA extremist who wants to rip away women’s freedoms just like she did when she was South Carolina governor.” [3] Despite all the concern expressed about Trump being a threat to democracy, there wasn’t even mild recognition from the Harris campaign that Haley winning the primary would remove that threat. That’s not in the nature of campaigns.
To a large degree, political campaigns are ontologically self-centered entities. As such, bullet voting is their natural inclination. It takes a reform like IRV, which puts ranking in the self interest of the campaign, to counteract that inclination. Even in the early days of Australian IRV elections, despite the lack of any real incentive, there was a tendency of the campaigns to encourage bullet voting – to “otherize” all the other candidates and parties – and it took a bit of time for the incentives of IRV to settle in and undo the tendency (and some jurisdictions made ranking mandatory to nip it in the bud.) Similarly, we saw how Palin encouraged bullet voting in August 2022, before succumbing to the incentives of IRV and endorsing a “rank the red” strategy for that November. It’s not a mindset that comes naturally to campaigns.
With a switch from IRV to Condorcet, we would lose the ability to truthfully and unequivocally tell campaigns and their supporters that ranking beyond their first choice won’t hurt that first choice. It could. (Same with the burying strategy, by the way, but that discussion is for another time.) We thereby lose much of the power to counteract the bullet voting inclination, a much larger share of voters will bullet vote, and campaigns will probably continue their “my way or the highway” attitude about their own candidate.
Finally, the example includes the unstated assumption that Haley would be the _only_ additional strong candidate to run independently. Let’s not forget there was a significant schism on the Democratic side on the issue of Israel/Gaza, arguably stronger than any divide on the Republican side. Under IRV, it’s conceivable that a credible Democratic candidate might have launched an independent run to challenge Harris from the left on this issue; or perhaps Jill Stein would have found greater traction and recognition if free of the spoiler label. A poll by IMEU found that “Ending Israel’s violence in Gaza” was the number one issue for Biden 2020 voters that did not vote for Harris [4]. Thus, a robust campaign to the left of Harris on this issue, might have driven leftwing turnout, bringing voters to the polls that were not motivated enough by Harris to cast a ballot, but who would rank her second if they did.
I’ll reiterate my own predictions. If Haley runs under a Condorcet system, and if polling indicates that she would be the Condorcet winner, I do see Trump and Harris mutually attacking her, encouraging bullet voting, and wooing supporters even further to their extremes to boost their odds. That is not to say that Haley wouldn’t have won anyway – she still might. But with the two major coalitions further polarized and neither seeing their candidate elected, I don’t see the system lasting long. Furthermore, I see several reasons why IRV would enable the entry of stronger third-party and independent candidates in a way that boosts turnout and which propels Harris to victory as the Condorcet candidate through second choice transfers.
That said, I freely admit that my conjectures are conjectures. Ultimately, both your and my hypotheticals are loaded with assumption and speculation. That’s why I keep returning to what _has_ happened in practice and what that can tell us about what will happen.
[1] https://fairvote.org/simulating_instant_runoff_flips_most_donald_trump_primary_victories/
[2] https://www.cbsnews.com/news/harris-trump-2024-election-ground-game/
[3] https://www.washingtonpost.com/elections/2023/11/17/haley-abortion-six-week-south-carolina
[4] https://www.imeupolicyproject.org/postelection-polling
Despite my disagreements on several points, I do agree with the thesis, i.e. that Condorcet should _not_ be “rejected entirely and everywhere.” Where I probably differ is that I don’t think it should be adopted in the US state-wide until we see it in operation at the municipal level, at least somewhere in the world. And before it is adopted at the municipal level, I’d ideally want to see some research into how it has performed in the, albeit limited, places it is and has been used. Something study akin to what Brams and Fishburn did for approval voting [1] would be a good start.
I heard at one point (and this might turn out to be wrong) that the city of Marquette, MI, keeps a file on its historic use of Nanson’s method. To my knowledge, no scholar has ever written about this use in any detail beyond citing that it happened. I still find it perplexing that there is this leap to advocate for broad adoption for Condorcet methods before any investigation of its past or current use.
[1] https://link.springer.com/chapter/10.1007/978-3-642-02839-7_3
>>I don’t think it should be adopted in the US state-wide until we see it in operation at the municipal level, at least somewhere in the world.<<
Take a look at this, Greg: https://en.wikipedia.org/wiki/Schulze_method#Government
Silla, Spain
Turin, Italy
San Donà di Piave, Italy
Borough of Southwark, London , UK
Now, I wonder why, Greg, you think that it's always someone else who has to be first? What if that was the argument in the early 19th century regarding Thomas Hare and the Single Transferrable Vote? If a proposed reform is going to take root, someone's gotta be first.
Yea, that's just another dumb talking point by IRV advocates because a) as you said, it's got to be tried somewhere first, and IRV/Hare could've been argued against using that exact same logic and b) IRV orgs seem fond of not wanting non-IRV single winner methods to be tried at the local level because they like IRV better. They're not an "everything bagel" single winner voting method reform group; they are specifically out there to ensure only RCV gets adopted. Greg perfectly encapsulates this because he claims that he just wants to see a study of the efficacy of Condorcet methods, but then he gets into the weeds of hypothetical strategic voting scenarios in order to discredit Condorcet methods so that he can point that out to elected officials and say "so this is why you shouldn't even try a Condorcet method."
As a condition for STV being adopted for the Tasmanian Parliament, it was first tried in two local municipalities for candidate elections. That makes complete sense to me. I would apply the same standard to any other voting method.
IRV orgs spill loads of ink trying to discredit Condorcet with hypothetical doom scenarios like the one you painted in your article so as to scare elected officials away from it. You and your IRV orgs are not interested in advocating for a scientific approach of testing out numerous election methods in many different cities; you are an IRV-only promoter and other methods are vigorously opposed, not supported, in your websites.
Yes. They have certainlyh been particularly hostile to Approval Voting.
Yes, I'm familiar with these uses of the Schulze method for referendum and online decision-making. Importantly, none of those are candidate elections. But sticking with those for just the moment, what have those experiences been like? Have the campaigns resembled candidate campaigns? Has there been any local interest or movement to translate those uses into actual candidate elections? Where is the research studying these experiences to help us understand if there are any lessons for its applicability more broadly? The complete lack of interest in these questions is perplexing.
Robert, you must be unfamiliar with the history and adoption of STV. Before being adopted by the Tasmanian Parliament, it was introduced first to the cities of Hobart and Launceston. In other words, if my conditions were imposed upon STV, it would be exactly where it is today.
Greg, the ballot is the same as for IRV and the meaning of the ballot is the same: If the voter ranks A above B, that means that in an election involving only A and B, this voter is voting for A. It's not Borda nor Score. No points. So, just like IRV, all the ranked ballot is, is the contingency vote information for the particular voter: How would this voter vote in the different contingencies involving different subsets of candidates?
So, every one of those 800 elections you might say are tested as successful, all of them but 4, those IRV elections are "successful" *solely* because they elected the Condorcet winner. In those IRV elections, no voter suffered voter regret for voting sincerely and ranking the candidates sincerely. And the reason they don't regret their vote is because the Condorcet winner was elected. Whether they like the Condorcet winner or not, there is nothing else they could have done to their ballot to get a better outcome from their political POV. No tactical voting would have helped them.
Condorcet RCV is well tested with all of these 800 IRV elections except 4. If those elections performed well, it's because they elected the Condorcet winner. IRV works pretty good when it elects the Condorcet winner, and when it doesn't, it works like shit. IRV is only as good as it can elect the Condorcet winner.
If, instead, exactly the same elections occurred, except they were tallied according to Condorcet rules, the ballots would be identical, the candidates are the same, the voters would mark the ballots exactly the same and the same outcome would result. Except for those 4 elections.
Now 2 of those 4 elections had no Condorcet winner. A Condorcet-TTR method would elect exactly the same candidate.
The 2 elections that had a Condorcet winner that was not elected with IRV are Burlington 2009 and Alaska 2022 (August). Those are the only two elections with a different outcome. Now what might have happened had those two elections elected the center candidate (Montroll in Burlington, Begich in Alaska)? What might have occurred regarding the repeal effort?
At least in Alaska, people would understand that this was the only way Begich would have been elected, otherwise it's Peltola. Do you think the GOP would have been so opposed tp RCV with a GOP winning instead of a Democrat?
In Burlington, the only difference in topology is that it was a *"come-from-behind"* victory. The supporters of Kurt Wright would still be clinging to their misconception that their candidate had the most voter support, whether Bob or Andy were elected. But they had the additional slap in the face when they understood that Kurt ended up as the spoiler and that, despite the promise, their second-choice vote for Andy was never counted. No one told them in advance that, if their candidate loses in the final round, their second-choice vote will not count. Then why should they bother to rank a second-choice?
About history, I only know what's available for the pedestrians. i.e. https://en.wikipedia.org/wiki/History_and_use_of_instant-runoff_voting . Hill, Hare, Andrae. I think it's Hare who is credited with coining the term: *"Single Transferable Vote"*. Condorcet considered IRV before any of them and rejected it for the very reasons pointed out here.
The other dumb thing that IRV people aren't getting is the loss of process transparency that comes with the loss of local tabulation and Precinct Summability. https://drive.google.com/file/d/1YtejO54DSOFRkHBGryS9pbKcBM7u1jTS/view . Consider the recent presidential election in Venezuela. What exposed that election as stolen? This is part of my Installment #4 to your essay in DemocracySOS. Still working on Installment #3, which is responding to your "theoretical analysis". I'm trying to be fair.
The IRV elections are test cases for Condorcet only under the assumption that voters vote and candidates campaign in exact the same way under both voting methods. You seem confident they will be the same, or perhaps that they will only differ in ways that are better under Condorcet. I am highly skeptical of that assumption, for reasons I've given here and elsewhere. It is that precise assumption that I would like to see investigated before further advocacy takes place. I understand your faith is so strong in the assumption that you don't need an actual scientific investigation into it. I think that's the crux of the disagreement.
The ballots are exactly the same (with one exception that in nearly any Condorcet method, not BTR-IRV, voters may equally rank multiple candidates). The meaning of the ballots are exactly the same (rank A above B is exactly one vote for A).
Then there are these promises like RCV will elect the candidate having a simple majority of voter support, that the spoiler effect is avoided, that voters are free to vote their hopes instead of their fears (that they can vote for whoever they really want without risk of helping elect the candidate they hate), that whenever they cannot get their first-choice candidate, their second-choice vote is counted instead.
So we engineers like to put things into black boxes, in order to test hypotheses. There's an input to the black box and there is an output. The input are the actual cast vote records for all of these "800" elections you refer to. In nearly all of these elections, whether it's Hare or Condorcet in the box, the output is the same and the promises above are kept.
But in two elections, we have the same input, but different outputs. One of the systems in a block box actually materially satisfy the conditions promised and the other system in the other black box measurably fails to satisfy the conditions promised. They are not equal. One of them works to do the job and the other one does not.
And this *still* hasn't even touched the issue of the loss of process transparency due to the loss of decentralized local tabulation of the vote that we get right now with FPTP. This component of process transparency, called "Summability" or "Precinct Summability", is the ability to know how an election will turn out, just from adding up the tallies published at each polling place. While Summability did not save the Venezuelan presidential election from being stolen (dictators will dictate), it *did* expose that election as stolen. If Venezuela had an independent judicial system that had teeth, Maduro would not be the president right now. Precinct Summability (and the rule of law and a strong and uncorrupt judicial system) keeps elections honest.
Why would you want to sacrifice the security of process transparency (which might allow a corrupt election official in Georgia to "just ,... find, uh, 11780 votes") just so you can keep doing things the same way and occasionally failing at delivering exactly what it is we want RCV to do? When this sacrifice is unnecessary?
Ned, first, thank you so much for responding to the points in my essay. I read and respect a lot of your work, but do not have many disagreements with you beyond this issue. There’s a lot to respond to here, and to keep things somewhat organized, I think I’ll devote one comment to each of the responses.
At multiple points, this essay suggests that Condorcet candidates may bow out of IRV races, knowing they will be eliminated early. To my knowledge, no one has ever presented any evidence that this is a significant feature of real IRV races to date. To the contrary, there were 13 candidates in the 2021 NYC Democratic mayoral primary, and 17 in the 2021 Minneapolis mayoral general election. If anything, one might wish for a few to bow out :)
Furthermore, it’s not fair to compare the incentives to bow out under plurality to those under IRV. Candidates bow out under plurality in part because they realize they have no shot, yes, but perhaps more often because _others_ believe they can’t win, causing the donations to dry up to the point where no campaign can actually be run. They also bow out because they are shamed and derided as “spoilers,” and they don’t want that label attached to them for the remainder of their political career. Under plurality, the “I won’t vote for (or donate to) you, because you can’t win” quickly becomes a self-fulfilling prophecy.
Under IRV, that underdog candidate has pretty strong incentives to run and stay in the race. First, those people who would otherwise say “I don’t want to throw my vote away” can now take a chance on you as their first choice. Although you may start in the back of the pack, you could build your support through a "virtuous cycle" of voters taking a chance, an inversion of the viscious cycle that plurality offers. Second, even if you doubt your ability to do that, running offers you a way to elevate your political profile and key issues without the spoiler label being affixed. Third, in the election itself, you can play kingmaker. To earn your second choice endorsement, the major candidates may offer you policy concessions or influence in their administration.
I know it wasn’t a major point of your post above, so apologies for dwelling on it, but whatever the reasons for the rarity of Condorcet failure in IRV elections, low candidate entry does not seem to be one of them.
"To my knowledge, no one has ever presented any evidence that this is a significant feature of real IRV races to date." How does one present evidence that someone didn't run due to IRV? Maybe both sides of the "Condorcet winner who chooses not to run" debate have no way to measure it. But it stands more to reason that since IRV exhibits the spoiler effect due to center squeeze (while Condorcet does not), IRV elections will feature the same tendency for candidates that are strong but maybe not seen as top-2 frontrunners to stay out of the race.
Your examples of the 13 and 17 candidate races proving that IRV doesn't cause potential Condorcet winners to bow out is a red herring because you are treating the existence of many fringe candidates in the race as being evidence that a strong centrist candidate who would've been the Condorcet winner also not bowing out. Perennial and fringe candidates have existed forever under plurality voting, so it's not some unique feature of IRV that candidates who normally wouldn't win will decide to run because they aren't going to be "derided" as a spoiler.
Your 4th paragraph is just standard IRV agitprop because nothing you say about IRV in that paragraph couldn't also be said of a Condorcet method. You are pretending that it's a unique feature of IRV in order to distract from the fact that IRV exhibits the very pathologies that organizations such as yours claim don't exist. I promise you, you can just admit your organization was wrong about its core claims and choose to advocate for a Condorcet method that won't exhibit the pathologies of IRV.
I know that you probably know that no single winner method is 100% free of pathologies (Arrow, Gibbard, etc.), but your organization straight up lies on its website that IRV doesn't exhibit the very pathologies that it exhibits as a matter of both theory and in practice. It's always the same thing with IRV orgs: "other methods can elect the wrong winner, but not ours, nosiree." Condorcet advocates never say that Condorcet methods can't ever be free of pathologies that can elect the "wrong" winner, but IRV advocates routinely say IRV is free of any pathologies that can elect the "wrong" winner, and that's just intellectually dishonest at best and fraud in advertising at worst.
"How does one present evidence that someone didn't run due to IRV?"
Well, you could start with at least one anecdote. We have plenty of examples of candidates bowing out of races today under plurality, both before and during the election, because of the pathologies of plurality. Many people, for example, asked why Bernie Sanders, after a life of running as an independent candidate, chose not to run as a third-party or independent candidate in the general election in 2016. He answered "What I did not want to do is run as a third party candidate, take votes away from the Democratic candidate and help elect some right-wing Republican. I did not want responsibility for that." That's pretty decent evidence that the voting system denied an additional strong candidate in that general election.
In 20 years and 800 elections under IRV, we have yet to have even a single anecdote of a credible centrist saying they'd stay out of the race due to anything attributable to a pathology of IRV.
Finally, I have never stated that "IRV is free of any pathologies," as you say. In fact, my essay cites the center squeeze pathology openly.
Do you consider yourself a Condorcet advocate? If so, what is an example election where you consider Condorcet to elect the "wrong" winner?
The organization you are affiliated with (and every other IRV advocacy group) makes two statements that unequivocally state that IRV doesn't suffer from pathologies that it does suffer. From https://voterchoicema.org/ :
1. Ranked Choice Voting Ensures Majority Support by eliminating the “spoiler effect” and guaranteeing the winner earns a majority of the votes in any election. FALSE
2. Ranked Choice Voting Expands Voter Choice by freeing you to vote for who you really want, without settling for the “lesser of two evils,” and without fear of “wasting” your vote. FALSE
So long as you are affiliated with any IRV advocacy group that advertises IRV falsely, you are effectively stating that IRV is free of pathologies on par with being a magic supplement that cures everything with no side effects.
Also, anecdotes aren't data, so not sure why you responded with my question with how to present evidence to actually do a study with a request for an anecdote. I actually conceded there was no actual way to know the full population of people who didn't run because of IRV, but you are assuming because there isn't an anecdote where someone actually said the words "I don't think I can win an IRV election, therefore I'm not mounting a campaign" that it's not a calculation that is affecting decisions to run for office. The thing is that you actually have no idea how many times a "credible centrist" looked at internal polling that showed them at 15%-20% in a three way race and they quietly decided to not run. It's like saying that no crime is happening because it's not being reported; a lack of observation doesn't invalidate the hypothesis.
What's funny is that you are doing the thing of inventing hypothetical scenarios where voters try to game the system while also criticizing Condorcet for having no real world examples of being tried. How do YOU know that voters will vote strategically that way if there are no real world examples? What we do know is that in both the Burlington and Alaska examples, Wright and Palin should have told their supporters to abandon them in favor of Montroll and Begich, respectively, yet didn't do so despite it being strategically better for their state to not elect the most "far left" candidate in the race. It seems that candidates are not likely to "stand down" in the face of certain defeat and instruct their voters to vote strategically, so why are you assuming they will do that in a hypothetical Condorcet election when they didn't do it in actual IRV elections?
As for being a Condorcet advocate, it's less about advocating for Condorcet and more about truth in advertising, and your organization currently has false claims on its website and other IRV organizations have been putting out those same lies about IRV for decades. That these obvious lies seem to barely have gotten IRV nowhere near widespread adoption makes me wonder why you sacrifice intellectual integrity for such meager results. It seems that IRV orgs can't even sell a free lunch, but maybe because people realize it's a toadstool and sauerkraut sandwich covered in arsenic sauce. I'm actually more interested in PR than single winner, but can't stand the intellectual bankruptcy of the IRV people.
Right on.
I said you could start with anecdotes, but you don't have to end with one. One could conduct a statistical study of "candidate exit" of the kind available in past polisci literature (example [1]). I put the challenge of identifying such an anecdote as a low bar, but if you have a candidate exit study, or you plan to conduct one, I'd be happy to review it.
[1] https://csap.yale.edu/sites/default/files/files/apppw-dmt-11-30-22.pdf
Yawn. Still not addressing the core fact that your organization pushes easily disproven lies, lies which have real consequences for how people choose to vote under IRV. I bet Wright-Montroll and Palin-Begich voters would love to know why your organization said that voting for their first choice wasn't wasting their vote on them when it actually did. I mean, I'd never lie to people like that, but you do you I guess.
I really don't care too much about your question about "candidate exit" because you can't even defend IRV's main defect and your organization's role in lying about how it doesn't exist. It probably does happen from time to time that a candidate drops out due to how IRV works, but you really can't measure it.
The study you linked to examines money raised, which is an objective thing that a political science researcher can pull from a website. What if someone chooses not to run before raising a single penny? If they choose to drop out after raising money and then seeing initial internals that say they are coming in third by "too much" and decide to drop out, how is that measured in a dataset with any way of knowing if someone didn't put out an official statement? Maybe they had family/medical issues, or felt another candidate entered the race that was better than them and agreed with the same basic issues... it literally could be anything, whether you could measure the money or not. The paper, from what I could tell, had no dataset that told us WHY a candidate dropped out. It is only telling us that candidate exit is highly influenced by money raised, controlling for other factors, but in no way are the authors able to figure out if money raised was due to "IRV viability," or a scandal, or changing political tides. It has the not-surprising conclusion that "candidate who doesn't get money drops out at higher rates than those who do get money," so don't know why it's relevant to our discussion.
I did admit it's pure conjecture, but also it's conjecture that should be taken seriously because we know that candidates drop out due to viability concerns all the time in plurality, so why not in IRV? Candidates could easily also drop out due to viability in a Condorcet election, but we're specifically talking about whether a Condorcet candidate would drop out of an IRV election, which they would obviously not do if it were a Condorcet election.
I think that's where you're having trouble dealing with this conceptually (amongst other things). Foley specifically is talking about Condercet candidates dropping out of elections that happen using methods that suffer from failure of the Condorcet criterion, which both plurality and IRV fail. Does a Condorcet candidate run if they think they can't win an IRV election, the same that a Condorcet candidate may not run in a plurality election?
You used Bernie as an example, but read what he said again: "What I did not want to do is run as a third party candidate, take votes away from the Democratic candidate and help elect some right-wing Republican. I did not want responsibility for that." He's basically saying here that he's not the Condorcet candidate and that Hilary would be the Condorcet candidate for whom he'd be spoiling the election. Whether he'd run in an IRV contest as a non-Condorcet candidate is not the issue because he isn't expecting to be the Condorcet candidate. The issue is a) whether it would affect the decision to run for a candidate if they were the Condorcet candidate but probably couldn't get to the final 2 round in an IRV race and b) is it desirable for the Condorcet candidate to win.
With respect to a), again, I don't know because politicians run for office based on a host of factors, but viability to win PER THE RULES OF THE ELECTION METHOD has always been a factor, so we shouldn't not think it would matter here. And with b), IRV orgs are obstinate in thinking that no matter how close the 3rd place person is to the 2nd place person, that somehow disqualifies them from being considered as a credible candidate (whether the margin is 5%, 3%, 1%, or one vote). You seem fine with vote splitting when it affects the penultimate round (since you support IRV), which is an odd stance to take for someone who claims to not like multiple candidates splitting the vote so as to not determine the "true majority winner."
Just one more anecdote I forgot to mention. Independent Tiffany Bond, who ran as a "centrist" alternative to Golden and Poliquin, explicitly said she would not have run in the 2018 Maine House race had IRV not been in place. Many examples of these candidates running and staying in the race under IRV.
This is a tangent, but note that Bond wasn't the Condorcet candidate in this race, despite being the "centrist." There is something else to be said here about the assumption that the candidate objectively in the "center" on policy is necessarily the Condorcet candidate -- an assumption that pervades much of this commentary and in which I indulged, for the sake of argument, in my essay. I'll try to say more of that in another comment, if I have time.
Again, this is a red herring. Foley is claiming that under IRV, a centrist candidate CAN be squeezed out by the IRV elimination algorithm, not that the centrist candidate is always the Condorcet candidate. Your twisting of his example into something he didn't say is indicative of the deceptive form of argumentation that IRV advocates are known for, so not surprised.
Another assumption you are making is that because Bond branded herself as the centrist alternative to Golden and Poliquin, that automatically makes her the centrist in the race. But if Bond wasn't the Condorcet candidate, then by the determination of the voters themselves, she wasn't likely the centrist candidate in reality. Golden, being the Condorcet candidate, has more claim to being the centrist in the race than does Bond, but you treat candidate advertising as truth rather than data. This is another example of IRV advocates writing an anecdote in a way to promote IRV and pull down Condorcet that is just factually incorrect.
Foley has offered several examples, including Rob Portman, Nikki Haley, and others, and assumes they are the Condorcet candidate because he judges them to be politically centrist. He is one operating on that assumption in this piece and others.
What's your point? He's assuming they are and pointing out that using hypothetical ballots that under IRV that they wouldn't get elected despite being preferred to the other candidates in head to head matchups.
You were the one assuming that Bond was the centrist candidate despite actual ballot data that points to the opposite conclusion as she wasn't the Condorcet candidate. You are the one making editorial decisions that contradict actual facts, but that's par for the course for IRV advocates.
Using ACTUAL ballot data, it's clear that in the 2009 Burlington mayoral race, the Democrat Montroll was the centrist candidate; using ACTUAL ballot data, it's clear that in the 2022 Alaska special election race, the non-MAGA Republican Begich was the centrist candidate. They did FEEL like the centrist candidate in both races, but until the ballots were cast, we didn't know who was closest to the center of the electorate.
You seem to be like every other IRV advocate in that you are intellectually dishonest. You know that Foley is using hypothetical examples to prove a point about how IRV doesn't elect the centrist candidate but say that he's making unfounded assumptions about who is a centrist (despite it being hypothetical). But you instead choose to take real ballot data that shows a candidate as not being the Condorcet candidate and claiming Condorcet doesn't always elect a centrist despite having no objective way to determine whether Bond is actually the centrist. Other than she has no D or R next to her name or based on her self-serving campaign branding, how can you possibly know she's the centrist candidate if not for the actual ballot data?
>>You seem to be like every other IRV advocate in that you are intellectually dishonest. You know that Foley is using hypothetical examples to prove a point about how IRV doesn't elect the centrist candidate but say that he's making unfounded assumptions about who is a centrist (despite it being hypothetical).<<
Ya know, the dumb thing is, Logan, is *of course* we know who the Center candidate is. This is the candidate who, in the semifinal round, gets far more 2nd-choice votes from both of the other candidates than these candidates receive from the other. Right cannot expect a lot of 2nd-choice votes from voters for Left. Nor can Left expect a lot of 2nd-choice votes from voters for Right. It's because *both* sets of "extremist" voters are marking Center as their 2nd-preference instead of the "extremist" candidate on the other side.
>>You seem to be like every other IRV advocate in that you are intellectually dishonest.<<
And I agree. There is something to this *persistent* intellectual dishonesty. I *want* to have good and productive conversations with other voting systems reformers and intellectuals, but these IRV/FairVote apologists seem to think that civil conversation means that they can literally change the facts. The numbers. And the meanings of the numbers. That insults the intelligence of the rest of us and that becomes wearisome.