Fair point. As this piece is largely a response to the Condorcet proponents, the primary audience was those already reading and comprehending that work.
To briefly answer the question, both RCV and all Condorcet methods use a ranked ballot. You can read about how RCV tallies the votes at the FairVote website [1]. There is no single way to tally the votes under Condorcet voting, because a Condorcet method, as described in the piece, is a property of the voting method, not an individual method itself. There are several methods of tallying the votes that qualify as Condorcet methods [2].
As a fellow wonk, this may be a little too wonky. I support RCV, multimember districts, reapportionment of the Senate, popular vote for President, and a shorter election season among other things...but I've yet to see a proposal to count votes in such a way where every candidate is judged head to head against every other candidate to determine the winner. That's way more complicated than RCV and seems nearly impossible to explain to an electorate whose buy-in is required to change the process.
And it's easy to explain to anyone who knows what a Round Robin tournament is: "If more voters mark their ballots preferring Candidate A to Candidate B, then Candidate B is not elected."
Is that too hard to explain to understand?
Somehow, you Hare advocates have to persuade us why B should be elected even if more voters prefer A.
I'm not advocating anything in particular, but you haven't persuaded me that it makes sense to count your second choice as or more heavily than your first, or use a system that allows candidates with very little support as a first choice to vault over multiple other candidates with more support because they're less offensive to a polarized base.
The data you present in your paper on RCV is hard to follow, and I wouldn't base my voting system on a single anomoly that does not neatly break down along ideological lines, but tactical voting goes both ways: if I'm understanding your system correctly, and choosing between a progressive, moderate, and conservative, and I'm confident the progressive will make it to the final round in a progressive-leaning district, I'm now incentivized to vote P>C>M even though it's not really my preference because I want the progressive to face the conservative rather than the moderate. Similarly, a strategic conservative in a conservative district may take the calculated risk of voting C>P>M to help their candidate's chances in the final round.
No system is ever going to be perfect, and what you're proposing seems better than first-past-the-post, but I don't think it makes more sense to let the voters who prefer the 1st place candidate decide which of the 2nd or 3rd candidates they'd rather face in a final round than it does to let voters who prefer the 3rd place candidate choose which of the top 2 they prefer.
I sorta promoted that in my paper, but have later come to understand that "Straight-Ahead Condorcet" is better for legislation. The statehouse legislative counsel, the sponsoring rep, and I all agreed that "The Law should say what it means and mean what it says."
Comparing candidates head to head is necessary if you want a democratic election. You can't just throw some people's preferences away because "it was easier to count that way".
The root of RCV's flaws is that it does exactly that: Gives some voters more power than others to affect the outcome of the election, counts some voters' preferences while discarding others. That's why it can elect a candidate who was not the preference of the majority of voters.
No it isn't. Democracy just means everyone gets a vote. It doesn't mean everyone's 5th, 6th, and 7th preferences need to be factored ino that vote.
And any system can and will elect a candidates who was not the preference of a majority of voters. If I end up with my 3rd choice out of 5 because my preference over my 4th choice was registered, I'm not sure how relevant that is or should be, but either way I'm not getting my preference. I mean, at this point why not have voters assign a point value on a scale of 0-7 for each candidate and rank candidates based on the total points they recieve? Surely, this would be a more accurate way of registering voter preferences than simply ranking them.
Frank, you're wrong. This is from my paper, so I'll just quote it.
_____________
(Principle) 1. “One person, one vote”. Every enfranchised voter has an equal influence on government in elections because of our inherent equality as citizens and this is independent of any utilitarian notion of personal investment in the outcome. If I enthusiastically prefer Candidate A and you prefer Candidate B only tepidly, your vote for Candidate B counts no less (nor more) than my vote for A. The effectiveness of one’s vote – how much their vote counts – is not proportional to their degree of preference but is determined only by their franchise. A citizen with franchise has a vote that counts equally as much as any other citizen with franchise. For any ranked ballot, this means that if Candidate A is ranked higher than Candidate B then that is a vote for A, if only candidates A and B are contending (such as in the RCV final round). It doesn’t matter how many levels A is ranked higher than B, it counts as exactly one vote for A.
_____________
You're advocating, essentially, for Score or Borda or something like that when you "assign a point value on a scale of 0-7 for each candidate". But that is wrong.
We voters are not Olympic judges scoring a performance of figure skating. We voters must not be burdened (unnecessarily) with the burden of tactical voting. ***Any*** cardinal method inherently requires the voter to consider tactics whenever there are 3 or more candidates in the race. Essentially they have to decide how high to score their second-favorite candidate (or whether to Approve their second-favorite if it's Approval Voting). The voter is faced with that the minute they step into the voting booth.
Please read my paper and read the North Dakota Supreme Court opinion at the very beginning of the paper. "One-person-one-vote" means our votes count equally in spite of our differences in enthusiasm of support for our candidates. Anything else is a "false or fictitious method of marking the ballots". A "scheme".
Also a quote (translated from French) of Borda in response to a critique from Condorcet is" "My system is only intended for honest men." And that's a problem. We need a system that is resilient to the efforts of dishonest and manipulative people trying to game the system. Borda or Score (or even Approval, since it is also Cardinal) is not resilient in that. It requires voters to consider that how they rank or score Candidate A might affect the effect of their vote regarding Candidates B vs. C. We must not burden the voters with that worry.
I'm not sure anyone's votes are counted equality no matter which system you use, Robert.
What you seem to be suggesting is that voters who prefer the top performer in a 3-way race become instrumental in deciding which of the 2nd and 3rd place candidates they want to go against in the final round. This does not seem fairer or more equitable than giving voters who prefer the 3rd place candidate an opportunity to vote for one of the top 2. If anything, the opposite.
Not that adding complexity will help sell it to voters, but perhaps adding a proportional or differential threshold would give us the best of both systems. In a situation where the first round distribution is 41/39/20, the only thing that makes intuitive sense is to take the 2nd choice of voters for the last place candidate into account. Better to give ~40% their top choice and 10%+ their second than give 20% their top choice and 30%+ their second, especially when that second choice could be a very distant second.
In a situation where the first round distribution is more like 35/34/31, however, there's a much stronger case for not eliminating any candidate, factoring in everyone's second choice, and trying to determine the winner by tallying the race as 3 separate head-to-head matchups. (Of course, you need a backup plan if everyone goes 1-1.)
Also, as an aside, 1827 voters in that Burlington election did get their 2nd choice, so Howard Dean was clearly not mistaken. He said voters "could" not "would" get their 2nd choice. Given that each of the 6 possible choice combinations were made by some number of voters, a minimum of 865 and a maximum of 3375 voters were guaranteed to get their last choice depending on who won. Not that elections should be determined by the fewest number of people getting their last choice, but that part of your argument comes across as bad faith, and clearly flawed.
> I'm not sure anyone's votes are counted equality no matter which system you use, Robert.
If there is a Condorcet winner and if that Condorcet winner is elected, you have satisfied majority rule and have counted every vote equally in all possible election scenarios. That is clear.
Of course if the CW is not elected, or if there is no CW to elect, then we know that there will be a simple majority of voters that can accurately complain that their votes were not counted as much as the fewer voters supporting the non-majority candidate who was elected. In the case that there is no CW to elect, that is unavoidable. Arrow prevails. But in the case that there *is* a CW and that candidate was not elected, you failed to honor the vote of the majority, the voters supporting that majority candidate over the candidate who *was* elected can accurately complain that their votes did not count as much as the votes from the fewer number of voters supporting the candidate who was elected.
There is so much stuff to respond to, and I am trying to write a reasonably complete response to Greg's opinion.
What seems clear to you is not at all clear to me.
In BTR-STV, voters for the TOP candidate in the semi-final round enjoy having their vote counted twice--they picked the top candidate, and also have their 2nd choice counted, (potentially) becoming the deciding factor between the 2nd and 3rd place candidates. Only the top choice is counted for everyone else in that round. Maybe there are good reasons to do it that way, but you're not counting every vote equally. For that, you'd need to take the sum of the 1st and 2nd place votes for all 3 candidates to determine the rankings of the top 3.
You are trying to make the case for a system that would elect a candidate who is 3rd or 4th in terms of voters' top choice when they end up in 1st if you factor in people's 2nd and 3rd choices. I can envision scenarios where this makes sense, and could or would be considered the "right," "best," or "most widely accepted" choice.
However, I can also envision a scenario like the one presented in this post, where a candidate who is in a distant third and only the top choice for say 15% of voters ends up beating the other two candidates in a head-to-head.
If we were to collect more data on the relative strength of support for each candidate, we would find this kind of "consistent majority" is not at all the same as one where the 3rd place candidate is the first choice for roughly 30% of the electorate. If some two-thirds of your "support" is coming from people who are only supporting you because they'd vote for a turd sandwich over your opponent, that doesn't seem like a better option than someone who is the first choice for 40-45% of the electorate and is getting begruding support from another 6-15%. Even if the former ends up being a larger majority, it's not a stronger one.
The TLDR version: Hare RCV could be improved to ensure the election of the "most widely accepted candidate" in multi-way races where 3 or more candidates are all running neck-and-neck. BTR, however, does not seem to be the solution, as it is also capable of producing outcomes that may be widely considered undesirable.
Frank, BTR-IRV ends up electing the same candidate as Condorcet-Plurality (as a "two-method" Condorcet). Now, maybe instead of Condorcet-Plurality, it should be Condorcet-TTR (top-two runoff). That would elect the same as the IRV winner in the case of a cycle. Perhaps that would be better.
Again, a cycle happens even more rarely than just an IRV failure to elect the CW. What we need is a law that has a result that voters can accept as fair in this case of an *extremely* close election.
If we need a law that produces a result that voters can accept as fair in this case of an election that is not just *extremely close* but has 3 or more candidates who are *all* extremely close to one another, we also need a law that doesn't run the risk of catapulting a candidate from a distant third in terms of top preference to winning as BTR and Condorcet-Plurality are capable of doing.
The way to solve this would be to have some sort of formula that prescribes when to use TTR, and when BTR or Condorcet-Plurality would go into effect. This is why I suggested that in the case of 3 candidates, the 3rd place candidate on top preference alone would either need to have a minimum of say 28% (or perhaps 30%), or be closer to the 2nd place candidate than the 2nd place candidate is to the 1st place candidate. If either of those conditions are met, BTR/Condorcet Plurality seems most likely to produce the most widely accepted result. If neither of those conditions are met, TTR is more likely to produce the most widely accepted result.
"Even if one believes that center-squeeze scenarios weaken RCV’s propensity for moderating our politics, it’s implausible that they could be consequential at such a low frequency. Those who insist otherwise owe us either a careful argument as to why such rare events could have a substantial impact, or an argument as to why they might be more common in the future."
Moving beyond the frequency of center squeezes, the central argument of this post is poorly reasoned.
"If the Alaska race used Condorcet, Peltola and Palin would not improve their odds by earning more second choice rankings of Begich supporters."
This can only be true if Begich is a strong enough frontrunner that the probability of Palin and Peltola tying for first is negligible. If (say) Palin and Begich are competing for first, then under Condorcet they each have a strong incentive to appeal to to Peltola's supporters - an incentive that does not exist under RCV unless Peltola is liable to be eliminated in the first round. These incentives for Republican candidates to appeal to Democratic voters (and vice versa) are incentives against political extremism.
Alternatively, suppose Begich isn't a frontrunner (maybe he's a lot less popular in general in the alternative reality), and Palin and Peltola are the leading candidates. In that case, Condorcet provides every bit as strong of an incentive for Palin and Peltola to appeal to Begich's voters. In order to be the Condorcet winner, Peltola needs to have more voters rank her ahead of Palin than there are voters ranking Palin ahead of Peltola, and it doesn't matter whether these voter's first choice is Peltola, Begich, or a write-in. The claim the second-choice rankings are valuable under RCV, but not under Condorcet, is completely false.
"As Left and Right would soon discover, in a center-squeeze configuration they have an incentive to encourage their respective supporters to “bullet vote,” meaning rank them first and leave the rest of the rankings blank."
If Palin tells her supporters to bullet vote, that helps Peltola win; it does not help Palin. Palin getting her supporters to bullet does not cause her to be ranked higher than her opponents on any ballots, so it cannot help her become the Condorcet winner. Maybe you're claiming that Palin and Peltola would make a "deal with the devil" to tell each of their supporters to bullet vote, but individual voters are best off not heeding such suggestions.
"While bullet-voting offers no advantages to candidates under RCV, under Condorcet it could take the Center candidate out of the running, effectively setting up a plurality-style contest between the Left and Right candidates where they might stand a better shot at winning."
Even if there is a deal between the more extreme candidates to promote bullet voting and this successfully takes the centrist out of contention, it doesn't mean there would be a Plurality-style contest between Left and Right. Instead, it would be an RCV-style contest, with both the Left and Right candidates competing to be ranked higher than the other by the Center candidate's supporters.
I do agree with you and Ned Foley on one point: RCV provides better incentives to reduce extremism than Plurality. My research that found that Condorcet methods offer much stronger incentives for depolarization than RCV also found that RCV has better incentives than Plurality (https://www.sciencedirect.com/science/article/abs/pii/S026137942400057X).
My piece is primarily a response to the arguments and anecdotes put forth by Foley, et al, which focuses on the “classic” center-squeeze scenario, which I think I faithfully characterized above. The only two instances in US history in which RCV did not elect the Condorcet winner are examples of this scenario. The August 2022 Peltola race in particular has seen the lion’s share of attention from the most prominent Condorcet proponents, and for all the flaws I see in their arguments, at least they stem from something that actually happened. Thus, my focus on it.
I say this because much of your comment is devoted to exploring variations of the real-world anecdote. Once you resort to statements like "suppose the facts of the anecdote were different than the are” (e.g. “If Palin and Begich are competing for first…”), it ceases to be an anecdote at all. There are hundreds and hundreds of real RCV elections to draw on. If the problem is at all a practical one, let’s look at those. I repeat my plea for real-world empiricism.
I see your linked substack post does give a real-world example: the 2022 race in Alaska State Senate District E. I’m happy about that. But what’s most interesting about that example is that it completely undercuts your theory. As you admit in the post, Giessel ran a campaign that reached out to all voters in a consensus-building fashion even though your model says she wouldn’t have. As I state in my piece, “when a theory doesn’t explain reality, it’s time to find a new theory.”
With respect to bullet voting, I think you have a fair point that it may be unlikely to help the candidates. But it might: it could create a Condorcet cycle which could theoretically lead either candidate to win. That bullet voting might help that candidate might be reason enough for campaigns to disseminate that strategy.
The reason? Well, one aspect of campaigns that I think goes underappreciated is to the degree to which they are ontologically self-centered entities. To a large extent, the campaign’s candidate is the only one that matters. As a result, they have a natural inclination to promote bullet voting, regardless of the voting method. They routinely do in plurality bloc elections today. No one else matters. Case in point: early Australian elections, where campaigns and parties advocated bullet voting even though they had no such electoral incentive. It took a bit of time for the incentives of RCV to be fully absorbed (and some jurisdictions began requiring ranking anyway). Or consider Palin’s race in August 2022, where she advocated bullet voting, before eventually succumbing to the RCV incentives and endorsing “rank the red” that November.
So I don’t think it would take a "deal" or any agreement of sorts. It just takes following the natural instincts and mentality of a campaign, with perhaps with a dash of tacit collusion. But no formal or informal deal. I think it takes a lot to dislodge a campaign of the “we’re all that matters mentality.” Because bullet voting provides only disadvantages under RCV, I think RCV has the necessary incentives. With Condorcet, bullet voting might help your candidate, so one cannot truthfully, unequivocally say it will not. And in a center squeeze scenario, it may be the only shot at defeating the Center candidate.
If bullet voting strategies were to take hold, you are right that it could set up more of an RCV-style contest than a plurality-style one. But I wouldn't count on it. When you have both major coalitions disseminating a bullet voting strategy, that would be the dominant message of "how to vote" that all voters hear, and so most voters will adopt.
I'm happy to focus on the actual voter preferences expressed in the Alaska special election; I talked about the "what ifs" because you brought up hypotheticals that were strongly divorced from the Alaska election. You wrote, "While bullet-voting offers no advantages to candidates under RCV, under Condorcet it could take the Center candidate out of the running, effectively setting up a plurality-style contest between the Left and Right candidates where they might stand a better shot at winning." In the Alaska special election, Palin did encourage her supporters to bullet vote, and her supporters' second-choice support for Begich nevertheless caused him to be preferred over Peltola. Such speculation seems to presuppose a weaker centrist than Begich who would be defeated by appeals for bullet voting.
As for bullet voting leading to a Condorcet cycle, this can be mathematically possible in a very close election. That said, it's weird to focus on bullet voting here since ranking candidates in the direction of the cycle you want to create is more effective. And it often isn't mathematically possible for a coalition of strategic voters to create a Condorcet cycle; I haven't looked into it for Alaska, but the numbers don't work out for strategic voters to have changed the outcome of the 2009 Burlington center squeeze election under Minimax (https://medium.com/@voting-in-the-abstract/the-center-squeeze-a-deep-dive-under-several-voting-methods-911495ba2ed9). The biggest problem with this argument is that it takes modest claim (it is mathematically possible for telling one's supporter to bullet vote to gain a candidate an advantage by causing a Condorcet cycle) and assumes that it implies a far stronger claim ("To justify the bullet-voting strategy, candidates Left and Right would likely spike their rhetoric with dismissal and demonization of the Center candidate and centrism generally") without considering the feasibility of getting enough voters to bullet vote, countervailing strategic incentives, etc.
But all the talk of candidates being strongly incentivized to promote bullet voting is premised on claims like "If the Alaska race used Condorcet, Peltola and Palin would not improve their odds by earning more second choice rankings of Begich supporters. In a center squeeze, second choice rankings from the Center do not bring the Left and Right candidates any closer to victory, and therefore centrist appeals and moderating rhetoric are of little value." Here's why this is wrong:
First, this quote fails to emphasize that Begich would have won, so focusing on the incentives of Peltola and Palin means focusing on the incentives of the losers. These are less important than the incentives of the winner since the losers don't get to govern. And to win, Begich needs to convince Democrats that he's better than Palin and Republicans that he's better than Peltola. These are incentives for staying in the center.
Second, for the purpose of defeating Begich, Peltola and Palin would have strong incentives to reach out to one another's supporters to pick up second-choice rankings. Consider which would yield better incentives for reducing political polarization: Palin being incentivized to appeal to Begich's supporters (largely moderate Republicans) under RCV to defeat Peltola, or Palin being incentivized to appeal to Peltola's supporters (mostly Democrats) under Condorcet to defeat Begich? Personally, I'm going with the latter. Under RCV, Palin could write off all of Peltola's supporters as a lost cause and alienate them to her heart's content.
Third, Palin wouldn't know who her strongest opponent would be with certainty. (I'm not sure I can say the same for Peltola and Begich; Palin did abysmally against Begich head-to-head.) In the face of this uncertainty, Palin would have incentives both to reach out to Peltola's supporters to gain an edge on Begich and to Begich's supporters to gain an edge on Peltola.
"I see your linked substack post does give a real-world example: the 2022 race in Alaska State Senate District E. I’m happy about that. But what’s most interesting about that example is that it completely undercuts your theory. As you admit in the post, Giessel ran a campaign that reached out to all voters in a consensus-building fashion even though your model says she wouldn’t have."
This misinterprets my theory. I claim that (a) candidates will mostly follow their electoral incentives, though they will sometimes deviate from them, and (b) RCV incentivizes candidates to primarily focus on winning first-choice support, even though being ranked #2 often has nonzero value. I believe we agree on (a); as you've noted, Palin failed to heed her electoral incentives. (I'm uncertain of your stance on (b).) I think State Senate District E was either a case in which Giessel didn't properly follow her electoral incentives or a case in which the electoral incentives that led to her running such an inclusive campaign weren't determined by the voting method. (For example, maybe reaching out to Democrats was effective for gaining first-choice support, which would have been equally effective under Plurality.)
The article is a bit long, I agree, but I needed to give enough background for those who weren't familiar with Condorcet voting or with any of the relevant theory. As Frank's comment suggests, for some the article wasn't long enough :)
I am not an employee of FairVote; this article is not a FairVote product; nor is any FairVote affiliation cited. It's a bit funny to criticize them for "misrepresentation," while misrepresenting the origin of the piece.
This is installment #1 in the rebuttal to Greg's opinion column here.
RCV elections in the U.S. failed to elect a Condorcet winner in at least 4 elections. I heard that an RCV election in Moab Utah also failed and is a 5th example but I don't have specific information. But two of these RCV elections (Minneapolis 2021, Oakland 2022) had no Condorcet winner, they were in a Rock-Paper-Scissors cycle. Arrow ultimately prevails. The other two RCV elections that had a Condorcet winner who was not elected both resulted in an elected candidate lacking in perceived legitimacy and a concerted repeal effort that succeeded (for 13 years) in Burlington Vermont in 2010 and almost succeeded (0.2% margin, after $15 million was spent to defeat repeal) in Alaska in 2024.
The case for Condorcet RCV is the case for RCV. The purpose of RCV is, in single-winner elections having 3 or more candidates:
1. ... that the candidate with majority support is elected. Plurality isn't good enough. We don't want a 40% candidate elected when the other 60% of voters would have preferred a different *specific* candidate over the 40% plurality candidate. But we cannot find out *who* that different specific candidate is without using the ranked ballot. We RCV advocates all agree on that.
2. Then whenever a plurality candidate is elected *and* voters believe that a different *specific* candidate would have beaten the plurality candidate in a head-to-head race, then the 3rd candidate (neither the plurality candidate nor the one people think would have won head-to-head) is viewed as the spoiler, a loser whose presence in the race materially changes who the winner is. We want to prevent that from happening. All RCV advocates agree on that.
3. Then voters voting for the spoiler suffer voter regret and in future elections are more likely to vote tactically (compromise) and vote for the major party candidate that they dislike the least, but they think is best situated to beat the other major party candidate that they dislike the most and fear will get elected. RCV is meant to free up those voters so that they can vote for the candidate they really like without fear of helping the candidate they loathe. All RCV advocates agree with that.
4. The way RCV is supposed to help those voters is that if their favorite candidate is defeated, then their second-choice vote is counted. So voters feel free to vote their hopes rather than voting their fears. Then 3rd-party and independent candidates get a more level playing field with the major-party candidates and diversity of choice in candidates is promoted. It's to help unlock us from a 2-party system where 3rd-party and independent candidates are disadvantaged.
Now the case for Condorcet, instead or Hare (IRV), is that Condorcet succeeds at those purposes when Hare fails. And when Hare fails unnecessarily (that is, we cannot blame Arrow for the failure), the results are *never* good. And IRV advocates just cannot come to terms with that fact.
In Burlington Vermont 2009 [and also more recently in the Alaska 2022 (August special election)], RCV (in the form of IRV) failed in every one of those core purposes for adopting RCV. And it's an unnecessary failure because the ballot data contained sufficient information to satisfy all four purposes, but the tabulation method screwed it up.
In 2000, 48.4% of American voters marked their ballots that Al Gore was preferred over George W. Bush while 47.9% marked their ballots to the contrary. Yet George W. Bush was elected to office.
In 2016, 48.2% of American voters marked their ballots that Hillary Clinton was preferred over Donald Trump while 46.1% marked their ballots to the contrary. Yet Donald Trump was elected to office.
In 2009, 45.2% of Burlington voters marked their ballots that Andy Montroll was preferred over Bob Kiss while 38.7% marked their ballots to the contrary. Yet Bob Kiss was elected to office.
[And more recently in August 2022, 46.3% of Alaskan voters marked their ballots that Nick Begich was preferred over Mary Peltola while 42.0% marked their ballots to the contrary. Yet Mary Peltola was elected to office.]
That's not electing the majority-supported candidate. Andy would have defeated Bob in the final round by a margin of 6.5% had Andy met Bob in the final round. The 3476 voters that preferred Bob had votes with more effect than the 4064 voters that preferred Andy. Each of the 3476 voters for Bob had a vote that counted more than the vote from each of the 4064 voters for Andy.
[Or in Alaska, each of the 79000 voters that preferred Democrat Mary Peltola over moderate Republican Nick Begich had a vote that effectively counted more than a vote from each of the 87000 voters preferring Begich over Peltola. Those are not equally-valued votes, not "One person, one vote".]
Then, because Kurt Wright displaced Andy from the final round, that makes Kurt the spoiler, a loser in the race whose presence in the race materially changes who the winner is. When this failure happens, it's always the loser in the IRV final round who becomes the spoiler.
[Similarly in Alaska, Sarah Palin displaced Nick Begich from the final round, which makes Palin the spoiler, a loser in the race whose presence in the race materially changes who the winner is.]
First, none of this actually addresses my specific arguments about the centrifugal effects or the likelihood of bullet voting under Condorcet in center-squeeze situations.
Second, there have been "legitimate and concerted" RCV repeal efforts, including successful ones, in jurisdictions when RCV _did_ elect only Condorcet winners, too. The only real lesson here: any reform that changes who wins office will face backlash from some part of the establishment, because some part of the establishment will be negatively affected. In fact, if you make an electoral reform change and the establishment doesn't push back in some way, you should ask yourself whether you've made a substantial change at all.
Why was Condorcet repealed in Marquette, MI? Do you know know? Has any Condorcet proponent done a study or written about the experience in any detail? Have you or anyone else studied the use and effects of Condorcet in the organizaitons that have used it? I haven't seen it. No, there's statements of certainties about its effects without any level of interest into whether those predictions align with what the effects have been in practice. They aren't even acknowledged as predictions.
It's Installment #1. I am methodically going from beginning to end. It was a really long column.
>> The only real lesson here: any reform that changes who wins office will face backlash from some part of the establishment, because some part of the establishment will be negatively affected. <<
That's a falsehood. It's not the only real lesson. Just because some folks don't want to learn the entire lesson does not mean the rest of us should not.
The real lesson is that when RCV *fails* to respect Majority Rule and the Equality of our votes, in a tangible manner, the repeal effort actually can point to a legitimate reason for repeal. I don't wanna give folks like Phil Izon anything to hang their hat on. As RCV is used more and more, the frequency of these failures will increase and this failure will become less anomalous. As the other commenter pointed out, RCV is *not* growing all that fast and it is *not* prevalent in the U.S. It's early in the voyage that we need to make the course corrections. Problem is the resistance and denial of RCV organizations that there need be any course correction. Then when we arrive at a different place (which is democracy that respects majority rule and the equality of our vote and preventing spoiled elections and allowing voters' second-choice votes to actually count and freeing voters to "vote their hopes not their fears") then you're going to rely on people's short memories and say that's never where we wanted to go in the first place, to justify not making the course correction when we could have been least costly and most effective.
There are other real lessons to learn here that you have not addressed in your column, but I intend to bring it up.
I asked Steven Hill if DemocracySOS would consider publishing a rebuttal from me. That would be a far more careful reasons and concise rebuttal than we are going to get in comments. I have heard no response after two emails.
So, if that is the case, the rebuttal is coming in bits and pieces.
Now, one thing that is disingenuous, is to represent "The Case for Condorcet Voting" with an example that is cherry-picked to make the Condorcet results appear nonsensical. Now I presented the case for Condorcet RCV *solely* as the case for RCV. The very reasons we want Condorcet RCV are the stated reasons that RCV advocates make for RCV, yet those RCV advocates are usually plugging Hare or IRV.
Now the reason to use a Condorcet-consistent method to tally the vote, instead of Hare, is to prevent a failure of IRV to respect majority rule and the equality of our votes. This failure does not happen often, but when it *does* happen, the consequences are *never* good. So I will cherry pick (or make up) another example to show this. But I will not disingenuously call this "The Case for Hare (Instant Runoff) Voting". This is presented to point out the essential flaw of IRV instead. And that case is that of a close 3-way race, so my example is gonna make it *really* close. All three candidates are plausible winners.
99 voters in total
34 Right voters
---30 R>C>L
--- 1 R>L>C
--- 3 R only
32 Center voters
---13 C>L>R
---11 C>R>L
--- 8 C only
33 Left voters
---30 L>C>R
--- 1 L>R>C
--- 2 L only
Now that's a pretty close 3-way race. Who should win? Who truly has the most voter support? Who will Plurality elect? Who will IRV elect?
Now the FPTP people will say that Right should win in this close race because more voters prefer Right as their first choice than any other candidate.
The IRV people will say that Left should win because Left is preferred over Right by 46 to 45 votes (after Center is eliminated). Even though Right gets more first-choice votes (barely), Left is preferred over Right by 1 vote (again, barely).
But Center is widely preferred over either Left (62 to 34) or Right (62 to 35). You can call Center "milquetoast" if you want, but the electorate would *greatly* prefer Center over either Left or Right in this extreme example. Neither FPTP nor IRV will see that.
That is the cherry-picked example that really presents the Case for Condorcet.
> Ranked Choice Voting (RCV) is one of the fastest-growing voting reforms in the United States.
You can't be serious. RCV has been failing spectacularly, and is now banned in more states than it has been adopted in.
I genuinely don't understand why you people are so committed to RCV. You finally understand the Center Squeeze effect, and you recognize that it prevents RCV from fixing polarization, and that it perpetuates the same problems as our current system, but instead of allowing evidence to change your mind, you just double down and keep pushing an unpopular reform with bogus claims that its flaws don't actually matter in practice, even after it demonstrably fails.
I don't get it. "Getting RCV adopted" is obviously your goal, but shouldn't "improving democracy" be your goal?
Yes, improving democracy is my goal. RCV is one of several reforms which I believe are important vehicles to doing that.
I haven't "finally" understood the center-squeeze effect. I've known about it for more than 20 years. If you read the piece, you will see that I do _not_ recognize "that it prevents RCV from fixing polarization."
Any reform that changes who wins office will face backlash from some part of the establishment, because some part of the establishment will be negatively affected. (In fact, if you make an electoral reform change and the establishment universally likes it, you should ask yourself whether you've made a substantial change at all.)
Despite that, RCV continues to make net gains every year in terms of adoption. There hasn't been a successful repeal of an implementation in many years -- strictly new adoptions. That isn't to say that a repeal here or there won't happen -- progress is normally a bumpy road and establishment figures often wield real power.
If you read the piece, you will find plenty of evidence and citations that contradict your "failing spectacularly" claim.
Installment #2. Regarding Alaska Special Election August 2022:
>>Did this election demonstrate the propensity for RCV to elect an “extremist?” Far from it. Peltola was a clear moderate ... <<
Okay, this is not what the failure of IRV in Alaska is really about. There *is* the Center Squeeze, which this election demonstrates. When there is a majority failure with IRV (a failure to elect the Condorcet winner when there is one), it is *always* the candidate in the Center who is robbed. One of the other candidates is the spoiler and the other non-Center candidate is the beneficiary of the spoiled election. Now that doesn't mean that IRV leans Left or Right. It just leans away from the Center. The mechanism for this is that in the semifinal round (when there are 3 candidates) IRV is opaque to the second-choice rankings of the ballots and cares only about the first-choice rankings (or votes promoted to that level) and counted as votes.
Now, that is much different than the IRV final round (which does it right). In the final round, it doesn't matter how low the two remaining candidates are ranked on the ballots. They could be ranked fourth and fifth, and if Candidate A is ranked higher than Candidate B, then that is a full vote for A. When choosing between two candidates in the final round, IRV is working. But what is at issue is *which* two candidates and also if it worked when there were more than two candidates and the method was opaque to lower rankings?
The failure of IRV in Alaska in August 2022 is that if failed majority rule, and as a consequence, did not value voters' votes equally. If our votes aren't to be valued equally, then I want my vote to count more than yours. (If that is unacceptable to you, then let's make certain we're counting our votes equally.)
In that August 2022 election, more Alaskans, 87899 to 79461 (an 8438 voter margin), wanted Nick Begich instead of Mary Peltola and marked their ballots saying so. But Mary Peltola was sent to Washington to represent the Alaskan people in Congress.
This last November, again, more Alaskan voters marked their ballots that Begich is preferred to Peltola by nearly the same margin, 7876 (164861 to 156985). But this time Begich is sent to Washington.
Both elections about 8000 more Alaskans said they wanted Begich rather than Peltola. And, both times, marked their ballots saying so. Both times Instant-Runoff Voting was used.
What was different?
Sarah Palin was in the race in 2022 and not in the race in 2024. And the outcomes were different winners.
Sarah Palin was techically the spoiler, a loser whose presence in the race alters who the winner is. More Alaskans preferred Begich and they voted saying that. Why should Palin's running for the same seat change how Alaskans feel about Begich vs. Peltola?
Mary Peltola in August 2022, unfortunately shares a distinction with George W. Bush in 2000 and with Donald Trump in 2016. All three candidates were elected to office when the public record indicates that more voters marked their ballots preferring a different specific candidate for that office.
Mary Peltola is a wonderful person, I would most certainly have voted for Peltola had I been living in Alaska. She's clearly moderate, but in the context of Peltola vs. Begich vs. Palin, Peltola is not the centrist candidate among voters in Alaska. IRV falsely promised to fairly deal with the GOP split vote problem with two Republicans running in the general election, but it didn't. IRV propped up the *weaker* of the two Republicans against the Democrat. Sarah Palin could not beat Mary Peltola head-to-head in the final round, but Begich *could*. Yet IRV did not allow that side-by-side comparison of voter support and that is the flaw.
There should be an IRV final round occurring between *every* pairing of candidates. If more voters marked their ballots that Candidate A is a better choice for office than Candidate B, we should not elect Candidate B, if that can, at all, be avoided. If we elect Candidate B, we are guaranteed a spoiled election, one of the candidates on the Left or Right is the spoiler, the Center candidate will be the candidate who is robbed, and the other candidate on the Right or Left is the lucky beneficiary of a spoiled election.
It would be helpful if the article laid out the basics of what Cordorcet voting is...and RCV for that matter too.
Fair point. As this piece is largely a response to the Condorcet proponents, the primary audience was those already reading and comprehending that work.
To briefly answer the question, both RCV and all Condorcet methods use a ranked ballot. You can read about how RCV tallies the votes at the FairVote website [1]. There is no single way to tally the votes under Condorcet voting, because a Condorcet method, as described in the piece, is a property of the voting method, not an individual method itself. There are several methods of tallying the votes that qualify as Condorcet methods [2].
[1] https://fairvote.org/our-reforms/ranked-choice-voting-information/#how-ranked-choice-voting-works
[2] https://en.wikipedia.org/wiki/Condorcet_method
As a fellow wonk, this may be a little too wonky. I support RCV, multimember districts, reapportionment of the Senate, popular vote for President, and a shorter election season among other things...but I've yet to see a proposal to count votes in such a way where every candidate is judged head to head against every other candidate to determine the winner. That's way more complicated than RCV and seems nearly impossible to explain to an electorate whose buy-in is required to change the process.
There's legislative language available.
https://drive.google.com/file/d/1DGvs2F_YoKcbl2SXzCcfm3nEMkO0zCbR/view
And it's easy to explain to anyone who knows what a Round Robin tournament is: "If more voters mark their ballots preferring Candidate A to Candidate B, then Candidate B is not elected."
Is that too hard to explain to understand?
Somehow, you Hare advocates have to persuade us why B should be elected even if more voters prefer A.
I'm not advocating anything in particular, but you haven't persuaded me that it makes sense to count your second choice as or more heavily than your first, or use a system that allows candidates with very little support as a first choice to vault over multiple other candidates with more support because they're less offensive to a polarized base.
The data you present in your paper on RCV is hard to follow, and I wouldn't base my voting system on a single anomoly that does not neatly break down along ideological lines, but tactical voting goes both ways: if I'm understanding your system correctly, and choosing between a progressive, moderate, and conservative, and I'm confident the progressive will make it to the final round in a progressive-leaning district, I'm now incentivized to vote P>C>M even though it's not really my preference because I want the progressive to face the conservative rather than the moderate. Similarly, a strategic conservative in a conservative district may take the calculated risk of voting C>P>M to help their candidate's chances in the final round.
No system is ever going to be perfect, and what you're proposing seems better than first-past-the-post, but I don't think it makes more sense to let the voters who prefer the 1st place candidate decide which of the 2nd or 3rd candidates they'd rather face in a final round than it does to let voters who prefer the 3rd place candidate choose which of the top 2 they prefer.
You don't even need a round robin tournament. A single elimination bracket is Condorcet compliant, too.
Do you mean Bottom-Two Runoff (BTR-IRV)?
I sorta promoted that in my paper, but have later come to understand that "Straight-Ahead Condorcet" is better for legislation. The statehouse legislative counsel, the sponsoring rep, and I all agreed that "The Law should say what it means and mean what it says."
That's what we did for H.424 https://legislature.vermont.gov/bill/status/2024/H.424
Comparing candidates head to head is necessary if you want a democratic election. You can't just throw some people's preferences away because "it was easier to count that way".
The root of RCV's flaws is that it does exactly that: Gives some voters more power than others to affect the outcome of the election, counts some voters' preferences while discarding others. That's why it can elect a candidate who was not the preference of the majority of voters.
No it isn't. Democracy just means everyone gets a vote. It doesn't mean everyone's 5th, 6th, and 7th preferences need to be factored ino that vote.
And any system can and will elect a candidates who was not the preference of a majority of voters. If I end up with my 3rd choice out of 5 because my preference over my 4th choice was registered, I'm not sure how relevant that is or should be, but either way I'm not getting my preference. I mean, at this point why not have voters assign a point value on a scale of 0-7 for each candidate and rank candidates based on the total points they recieve? Surely, this would be a more accurate way of registering voter preferences than simply ranking them.
Frank, you're wrong. This is from my paper, so I'll just quote it.
_____________
(Principle) 1. “One person, one vote”. Every enfranchised voter has an equal influence on government in elections because of our inherent equality as citizens and this is independent of any utilitarian notion of personal investment in the outcome. If I enthusiastically prefer Candidate A and you prefer Candidate B only tepidly, your vote for Candidate B counts no less (nor more) than my vote for A. The effectiveness of one’s vote – how much their vote counts – is not proportional to their degree of preference but is determined only by their franchise. A citizen with franchise has a vote that counts equally as much as any other citizen with franchise. For any ranked ballot, this means that if Candidate A is ranked higher than Candidate B then that is a vote for A, if only candidates A and B are contending (such as in the RCV final round). It doesn’t matter how many levels A is ranked higher than B, it counts as exactly one vote for A.
_____________
You're advocating, essentially, for Score or Borda or something like that when you "assign a point value on a scale of 0-7 for each candidate". But that is wrong.
We voters are not Olympic judges scoring a performance of figure skating. We voters must not be burdened (unnecessarily) with the burden of tactical voting. ***Any*** cardinal method inherently requires the voter to consider tactics whenever there are 3 or more candidates in the race. Essentially they have to decide how high to score their second-favorite candidate (or whether to Approve their second-favorite if it's Approval Voting). The voter is faced with that the minute they step into the voting booth.
Please read my paper and read the North Dakota Supreme Court opinion at the very beginning of the paper. "One-person-one-vote" means our votes count equally in spite of our differences in enthusiasm of support for our candidates. Anything else is a "false or fictitious method of marking the ballots". A "scheme".
https://drive.google.com/file/d/1jIhFQfEoxSdyRz5SqEjZotbVDx4xshwM/view
Also a quote (translated from French) of Borda in response to a critique from Condorcet is" "My system is only intended for honest men." And that's a problem. We need a system that is resilient to the efforts of dishonest and manipulative people trying to game the system. Borda or Score (or even Approval, since it is also Cardinal) is not resilient in that. It requires voters to consider that how they rank or score Candidate A might affect the effect of their vote regarding Candidates B vs. C. We must not burden the voters with that worry.
Your value judgements are noted, but aren't relevant here.
My criticism of BTR is it violates your first principle even more than you claim TTR does.
My criticism of a multi-way head to head method is that treating varying levels of preference equally does not create functional governing coaltions.
My criticism of both is it incentivizes gaming the system in a polarizing manner.
Frank, Condorcet RCV is simply committed to majority rule and the equality of our votes:
If more voters mark their ballots that Candidate A is preferred to Candidate B, then Candidate B is not elected.
That's it. That's the Condorcet RCV ethic in a nutshell.
https://drive.google.com/file/d/1jIhFQfEoxSdyRz5SqEjZotbVDx4xshwM/view
https://link.springer.com/journal/10602/volumes-and-issues/34-3
I'm not sure anyone's votes are counted equality no matter which system you use, Robert.
What you seem to be suggesting is that voters who prefer the top performer in a 3-way race become instrumental in deciding which of the 2nd and 3rd place candidates they want to go against in the final round. This does not seem fairer or more equitable than giving voters who prefer the 3rd place candidate an opportunity to vote for one of the top 2. If anything, the opposite.
Not that adding complexity will help sell it to voters, but perhaps adding a proportional or differential threshold would give us the best of both systems. In a situation where the first round distribution is 41/39/20, the only thing that makes intuitive sense is to take the 2nd choice of voters for the last place candidate into account. Better to give ~40% their top choice and 10%+ their second than give 20% their top choice and 30%+ their second, especially when that second choice could be a very distant second.
In a situation where the first round distribution is more like 35/34/31, however, there's a much stronger case for not eliminating any candidate, factoring in everyone's second choice, and trying to determine the winner by tallying the race as 3 separate head-to-head matchups. (Of course, you need a backup plan if everyone goes 1-1.)
Also, as an aside, 1827 voters in that Burlington election did get their 2nd choice, so Howard Dean was clearly not mistaken. He said voters "could" not "would" get their 2nd choice. Given that each of the 6 possible choice combinations were made by some number of voters, a minimum of 865 and a maximum of 3375 voters were guaranteed to get their last choice depending on who won. Not that elections should be determined by the fewest number of people getting their last choice, but that part of your argument comes across as bad faith, and clearly flawed.
> I'm not sure anyone's votes are counted equality no matter which system you use, Robert.
If there is a Condorcet winner and if that Condorcet winner is elected, you have satisfied majority rule and have counted every vote equally in all possible election scenarios. That is clear.
Of course if the CW is not elected, or if there is no CW to elect, then we know that there will be a simple majority of voters that can accurately complain that their votes were not counted as much as the fewer voters supporting the non-majority candidate who was elected. In the case that there is no CW to elect, that is unavoidable. Arrow prevails. But in the case that there *is* a CW and that candidate was not elected, you failed to honor the vote of the majority, the voters supporting that majority candidate over the candidate who *was* elected can accurately complain that their votes did not count as much as the votes from the fewer number of voters supporting the candidate who was elected.
There is so much stuff to respond to, and I am trying to write a reasonably complete response to Greg's opinion.
What seems clear to you is not at all clear to me.
In BTR-STV, voters for the TOP candidate in the semi-final round enjoy having their vote counted twice--they picked the top candidate, and also have their 2nd choice counted, (potentially) becoming the deciding factor between the 2nd and 3rd place candidates. Only the top choice is counted for everyone else in that round. Maybe there are good reasons to do it that way, but you're not counting every vote equally. For that, you'd need to take the sum of the 1st and 2nd place votes for all 3 candidates to determine the rankings of the top 3.
You are trying to make the case for a system that would elect a candidate who is 3rd or 4th in terms of voters' top choice when they end up in 1st if you factor in people's 2nd and 3rd choices. I can envision scenarios where this makes sense, and could or would be considered the "right," "best," or "most widely accepted" choice.
However, I can also envision a scenario like the one presented in this post, where a candidate who is in a distant third and only the top choice for say 15% of voters ends up beating the other two candidates in a head-to-head.
If we were to collect more data on the relative strength of support for each candidate, we would find this kind of "consistent majority" is not at all the same as one where the 3rd place candidate is the first choice for roughly 30% of the electorate. If some two-thirds of your "support" is coming from people who are only supporting you because they'd vote for a turd sandwich over your opponent, that doesn't seem like a better option than someone who is the first choice for 40-45% of the electorate and is getting begruding support from another 6-15%. Even if the former ends up being a larger majority, it's not a stronger one.
The TLDR version: Hare RCV could be improved to ensure the election of the "most widely accepted candidate" in multi-way races where 3 or more candidates are all running neck-and-neck. BTR, however, does not seem to be the solution, as it is also capable of producing outcomes that may be widely considered undesirable.
Frank, BTR-IRV ends up electing the same candidate as Condorcet-Plurality (as a "two-method" Condorcet). Now, maybe instead of Condorcet-Plurality, it should be Condorcet-TTR (top-two runoff). That would elect the same as the IRV winner in the case of a cycle. Perhaps that would be better.
Again, a cycle happens even more rarely than just an IRV failure to elect the CW. What we need is a law that has a result that voters can accept as fair in this case of an *extremely* close election.
Isn't TTR already the standard RCV system?
If we need a law that produces a result that voters can accept as fair in this case of an election that is not just *extremely close* but has 3 or more candidates who are *all* extremely close to one another, we also need a law that doesn't run the risk of catapulting a candidate from a distant third in terms of top preference to winning as BTR and Condorcet-Plurality are capable of doing.
The way to solve this would be to have some sort of formula that prescribes when to use TTR, and when BTR or Condorcet-Plurality would go into effect. This is why I suggested that in the case of 3 candidates, the 3rd place candidate on top preference alone would either need to have a minimum of say 28% (or perhaps 30%), or be closer to the 2nd place candidate than the 2nd place candidate is to the 1st place candidate. If either of those conditions are met, BTR/Condorcet Plurality seems most likely to produce the most widely accepted result. If neither of those conditions are met, TTR is more likely to produce the most widely accepted result.
"Even if one believes that center-squeeze scenarios weaken RCV’s propensity for moderating our politics, it’s implausible that they could be consequential at such a low frequency. Those who insist otherwise owe us either a careful argument as to why such rare events could have a substantial impact, or an argument as to why they might be more common in the future."
I agree. Here's my argument for why center squeezes have a substantial impact despite occurring infrequently: https://medium.com/@voting-in-the-abstract/rarely-occurring-pathologies-can-frequently-be-relevant-9b9dc8e9fe22
Moving beyond the frequency of center squeezes, the central argument of this post is poorly reasoned.
"If the Alaska race used Condorcet, Peltola and Palin would not improve their odds by earning more second choice rankings of Begich supporters."
This can only be true if Begich is a strong enough frontrunner that the probability of Palin and Peltola tying for first is negligible. If (say) Palin and Begich are competing for first, then under Condorcet they each have a strong incentive to appeal to to Peltola's supporters - an incentive that does not exist under RCV unless Peltola is liable to be eliminated in the first round. These incentives for Republican candidates to appeal to Democratic voters (and vice versa) are incentives against political extremism.
Alternatively, suppose Begich isn't a frontrunner (maybe he's a lot less popular in general in the alternative reality), and Palin and Peltola are the leading candidates. In that case, Condorcet provides every bit as strong of an incentive for Palin and Peltola to appeal to Begich's voters. In order to be the Condorcet winner, Peltola needs to have more voters rank her ahead of Palin than there are voters ranking Palin ahead of Peltola, and it doesn't matter whether these voter's first choice is Peltola, Begich, or a write-in. The claim the second-choice rankings are valuable under RCV, but not under Condorcet, is completely false.
"As Left and Right would soon discover, in a center-squeeze configuration they have an incentive to encourage their respective supporters to “bullet vote,” meaning rank them first and leave the rest of the rankings blank."
If Palin tells her supporters to bullet vote, that helps Peltola win; it does not help Palin. Palin getting her supporters to bullet does not cause her to be ranked higher than her opponents on any ballots, so it cannot help her become the Condorcet winner. Maybe you're claiming that Palin and Peltola would make a "deal with the devil" to tell each of their supporters to bullet vote, but individual voters are best off not heeding such suggestions.
"While bullet-voting offers no advantages to candidates under RCV, under Condorcet it could take the Center candidate out of the running, effectively setting up a plurality-style contest between the Left and Right candidates where they might stand a better shot at winning."
Even if there is a deal between the more extreme candidates to promote bullet voting and this successfully takes the centrist out of contention, it doesn't mean there would be a Plurality-style contest between Left and Right. Instead, it would be an RCV-style contest, with both the Left and Right candidates competing to be ranked higher than the other by the Center candidate's supporters.
I do agree with you and Ned Foley on one point: RCV provides better incentives to reduce extremism than Plurality. My research that found that Condorcet methods offer much stronger incentives for depolarization than RCV also found that RCV has better incentives than Plurality (https://www.sciencedirect.com/science/article/abs/pii/S026137942400057X).
Thanks for your thoughtful comments, Marcus.
My piece is primarily a response to the arguments and anecdotes put forth by Foley, et al, which focuses on the “classic” center-squeeze scenario, which I think I faithfully characterized above. The only two instances in US history in which RCV did not elect the Condorcet winner are examples of this scenario. The August 2022 Peltola race in particular has seen the lion’s share of attention from the most prominent Condorcet proponents, and for all the flaws I see in their arguments, at least they stem from something that actually happened. Thus, my focus on it.
I say this because much of your comment is devoted to exploring variations of the real-world anecdote. Once you resort to statements like "suppose the facts of the anecdote were different than the are” (e.g. “If Palin and Begich are competing for first…”), it ceases to be an anecdote at all. There are hundreds and hundreds of real RCV elections to draw on. If the problem is at all a practical one, let’s look at those. I repeat my plea for real-world empiricism.
I see your linked substack post does give a real-world example: the 2022 race in Alaska State Senate District E. I’m happy about that. But what’s most interesting about that example is that it completely undercuts your theory. As you admit in the post, Giessel ran a campaign that reached out to all voters in a consensus-building fashion even though your model says she wouldn’t have. As I state in my piece, “when a theory doesn’t explain reality, it’s time to find a new theory.”
With respect to bullet voting, I think you have a fair point that it may be unlikely to help the candidates. But it might: it could create a Condorcet cycle which could theoretically lead either candidate to win. That bullet voting might help that candidate might be reason enough for campaigns to disseminate that strategy.
The reason? Well, one aspect of campaigns that I think goes underappreciated is to the degree to which they are ontologically self-centered entities. To a large extent, the campaign’s candidate is the only one that matters. As a result, they have a natural inclination to promote bullet voting, regardless of the voting method. They routinely do in plurality bloc elections today. No one else matters. Case in point: early Australian elections, where campaigns and parties advocated bullet voting even though they had no such electoral incentive. It took a bit of time for the incentives of RCV to be fully absorbed (and some jurisdictions began requiring ranking anyway). Or consider Palin’s race in August 2022, where she advocated bullet voting, before eventually succumbing to the RCV incentives and endorsing “rank the red” that November.
So I don’t think it would take a "deal" or any agreement of sorts. It just takes following the natural instincts and mentality of a campaign, with perhaps with a dash of tacit collusion. But no formal or informal deal. I think it takes a lot to dislodge a campaign of the “we’re all that matters mentality.” Because bullet voting provides only disadvantages under RCV, I think RCV has the necessary incentives. With Condorcet, bullet voting might help your candidate, so one cannot truthfully, unequivocally say it will not. And in a center squeeze scenario, it may be the only shot at defeating the Center candidate.
If bullet voting strategies were to take hold, you are right that it could set up more of an RCV-style contest than a plurality-style one. But I wouldn't count on it. When you have both major coalitions disseminating a bullet voting strategy, that would be the dominant message of "how to vote" that all voters hear, and so most voters will adopt.
I'm happy to focus on the actual voter preferences expressed in the Alaska special election; I talked about the "what ifs" because you brought up hypotheticals that were strongly divorced from the Alaska election. You wrote, "While bullet-voting offers no advantages to candidates under RCV, under Condorcet it could take the Center candidate out of the running, effectively setting up a plurality-style contest between the Left and Right candidates where they might stand a better shot at winning." In the Alaska special election, Palin did encourage her supporters to bullet vote, and her supporters' second-choice support for Begich nevertheless caused him to be preferred over Peltola. Such speculation seems to presuppose a weaker centrist than Begich who would be defeated by appeals for bullet voting.
As for bullet voting leading to a Condorcet cycle, this can be mathematically possible in a very close election. That said, it's weird to focus on bullet voting here since ranking candidates in the direction of the cycle you want to create is more effective. And it often isn't mathematically possible for a coalition of strategic voters to create a Condorcet cycle; I haven't looked into it for Alaska, but the numbers don't work out for strategic voters to have changed the outcome of the 2009 Burlington center squeeze election under Minimax (https://medium.com/@voting-in-the-abstract/the-center-squeeze-a-deep-dive-under-several-voting-methods-911495ba2ed9). The biggest problem with this argument is that it takes modest claim (it is mathematically possible for telling one's supporter to bullet vote to gain a candidate an advantage by causing a Condorcet cycle) and assumes that it implies a far stronger claim ("To justify the bullet-voting strategy, candidates Left and Right would likely spike their rhetoric with dismissal and demonization of the Center candidate and centrism generally") without considering the feasibility of getting enough voters to bullet vote, countervailing strategic incentives, etc.
But all the talk of candidates being strongly incentivized to promote bullet voting is premised on claims like "If the Alaska race used Condorcet, Peltola and Palin would not improve their odds by earning more second choice rankings of Begich supporters. In a center squeeze, second choice rankings from the Center do not bring the Left and Right candidates any closer to victory, and therefore centrist appeals and moderating rhetoric are of little value." Here's why this is wrong:
First, this quote fails to emphasize that Begich would have won, so focusing on the incentives of Peltola and Palin means focusing on the incentives of the losers. These are less important than the incentives of the winner since the losers don't get to govern. And to win, Begich needs to convince Democrats that he's better than Palin and Republicans that he's better than Peltola. These are incentives for staying in the center.
Second, for the purpose of defeating Begich, Peltola and Palin would have strong incentives to reach out to one another's supporters to pick up second-choice rankings. Consider which would yield better incentives for reducing political polarization: Palin being incentivized to appeal to Begich's supporters (largely moderate Republicans) under RCV to defeat Peltola, or Palin being incentivized to appeal to Peltola's supporters (mostly Democrats) under Condorcet to defeat Begich? Personally, I'm going with the latter. Under RCV, Palin could write off all of Peltola's supporters as a lost cause and alienate them to her heart's content.
Third, Palin wouldn't know who her strongest opponent would be with certainty. (I'm not sure I can say the same for Peltola and Begich; Palin did abysmally against Begich head-to-head.) In the face of this uncertainty, Palin would have incentives both to reach out to Peltola's supporters to gain an edge on Begich and to Begich's supporters to gain an edge on Peltola.
"I see your linked substack post does give a real-world example: the 2022 race in Alaska State Senate District E. I’m happy about that. But what’s most interesting about that example is that it completely undercuts your theory. As you admit in the post, Giessel ran a campaign that reached out to all voters in a consensus-building fashion even though your model says she wouldn’t have."
This misinterprets my theory. I claim that (a) candidates will mostly follow their electoral incentives, though they will sometimes deviate from them, and (b) RCV incentivizes candidates to primarily focus on winning first-choice support, even though being ranked #2 often has nonzero value. I believe we agree on (a); as you've noted, Palin failed to heed her electoral incentives. (I'm uncertain of your stance on (b).) I think State Senate District E was either a case in which Giessel didn't properly follow her electoral incentives or a case in which the electoral incentives that led to her running such an inclusive campaign weren't determined by the voting method. (For example, maybe reaching out to Democrats was effective for gaining first-choice support, which would have been equally effective under Plurality.)
Long long article with much less meritorius content than verbosity. It's *really* awful long.
And its wrong. And, like FairVote often does, it misrepresents facts.
I'll re-read it and comment again with some concise content.
The article is a bit long, I agree, but I needed to give enough background for those who weren't familiar with Condorcet voting or with any of the relevant theory. As Frank's comment suggests, for some the article wasn't long enough :)
I am not an employee of FairVote; this article is not a FairVote product; nor is any FairVote affiliation cited. It's a bit funny to criticize them for "misrepresentation," while misrepresenting the origin of the piece.
Greg, I said "like FairVote". I am comparing your position and your argument to those of FairVote. Let me itemize the ways. That takes time.
Hay DemocracySOS, would you be willing to publish a response from me in a separate column?
This is installment #1 in the rebuttal to Greg's opinion column here.
RCV elections in the U.S. failed to elect a Condorcet winner in at least 4 elections. I heard that an RCV election in Moab Utah also failed and is a 5th example but I don't have specific information. But two of these RCV elections (Minneapolis 2021, Oakland 2022) had no Condorcet winner, they were in a Rock-Paper-Scissors cycle. Arrow ultimately prevails. The other two RCV elections that had a Condorcet winner who was not elected both resulted in an elected candidate lacking in perceived legitimacy and a concerted repeal effort that succeeded (for 13 years) in Burlington Vermont in 2010 and almost succeeded (0.2% margin, after $15 million was spent to defeat repeal) in Alaska in 2024.
The case for Condorcet RCV is the case for RCV. The purpose of RCV is, in single-winner elections having 3 or more candidates:
1. ... that the candidate with majority support is elected. Plurality isn't good enough. We don't want a 40% candidate elected when the other 60% of voters would have preferred a different *specific* candidate over the 40% plurality candidate. But we cannot find out *who* that different specific candidate is without using the ranked ballot. We RCV advocates all agree on that.
2. Then whenever a plurality candidate is elected *and* voters believe that a different *specific* candidate would have beaten the plurality candidate in a head-to-head race, then the 3rd candidate (neither the plurality candidate nor the one people think would have won head-to-head) is viewed as the spoiler, a loser whose presence in the race materially changes who the winner is. We want to prevent that from happening. All RCV advocates agree on that.
3. Then voters voting for the spoiler suffer voter regret and in future elections are more likely to vote tactically (compromise) and vote for the major party candidate that they dislike the least, but they think is best situated to beat the other major party candidate that they dislike the most and fear will get elected. RCV is meant to free up those voters so that they can vote for the candidate they really like without fear of helping the candidate they loathe. All RCV advocates agree with that.
4. The way RCV is supposed to help those voters is that if their favorite candidate is defeated, then their second-choice vote is counted. So voters feel free to vote their hopes rather than voting their fears. Then 3rd-party and independent candidates get a more level playing field with the major-party candidates and diversity of choice in candidates is promoted. It's to help unlock us from a 2-party system where 3rd-party and independent candidates are disadvantaged.
Now the case for Condorcet, instead or Hare (IRV), is that Condorcet succeeds at those purposes when Hare fails. And when Hare fails unnecessarily (that is, we cannot blame Arrow for the failure), the results are *never* good. And IRV advocates just cannot come to terms with that fact.
___________________________________________________________________
In Burlington Vermont 2009 [and also more recently in the Alaska 2022 (August special election)], RCV (in the form of IRV) failed in every one of those core purposes for adopting RCV. And it's an unnecessary failure because the ballot data contained sufficient information to satisfy all four purposes, but the tabulation method screwed it up.
In 2000, 48.4% of American voters marked their ballots that Al Gore was preferred over George W. Bush while 47.9% marked their ballots to the contrary. Yet George W. Bush was elected to office.
In 2016, 48.2% of American voters marked their ballots that Hillary Clinton was preferred over Donald Trump while 46.1% marked their ballots to the contrary. Yet Donald Trump was elected to office.
In 2009, 45.2% of Burlington voters marked their ballots that Andy Montroll was preferred over Bob Kiss while 38.7% marked their ballots to the contrary. Yet Bob Kiss was elected to office.
[And more recently in August 2022, 46.3% of Alaskan voters marked their ballots that Nick Begich was preferred over Mary Peltola while 42.0% marked their ballots to the contrary. Yet Mary Peltola was elected to office.]
That's not electing the majority-supported candidate. Andy would have defeated Bob in the final round by a margin of 6.5% had Andy met Bob in the final round. The 3476 voters that preferred Bob had votes with more effect than the 4064 voters that preferred Andy. Each of the 3476 voters for Bob had a vote that counted more than the vote from each of the 4064 voters for Andy.
[Or in Alaska, each of the 79000 voters that preferred Democrat Mary Peltola over moderate Republican Nick Begich had a vote that effectively counted more than a vote from each of the 87000 voters preferring Begich over Peltola. Those are not equally-valued votes, not "One person, one vote".]
Then, because Kurt Wright displaced Andy from the final round, that makes Kurt the spoiler, a loser in the race whose presence in the race materially changes who the winner is. When this failure happens, it's always the loser in the IRV final round who becomes the spoiler.
[Similarly in Alaska, Sarah Palin displaced Nick Begich from the final round, which makes Palin the spoiler, a loser in the race whose presence in the race materially changes who the winner is.]
First, none of this actually addresses my specific arguments about the centrifugal effects or the likelihood of bullet voting under Condorcet in center-squeeze situations.
Second, there have been "legitimate and concerted" RCV repeal efforts, including successful ones, in jurisdictions when RCV _did_ elect only Condorcet winners, too. The only real lesson here: any reform that changes who wins office will face backlash from some part of the establishment, because some part of the establishment will be negatively affected. In fact, if you make an electoral reform change and the establishment doesn't push back in some way, you should ask yourself whether you've made a substantial change at all.
Why was Condorcet repealed in Marquette, MI? Do you know know? Has any Condorcet proponent done a study or written about the experience in any detail? Have you or anyone else studied the use and effects of Condorcet in the organizaitons that have used it? I haven't seen it. No, there's statements of certainties about its effects without any level of interest into whether those predictions align with what the effects have been in practice. They aren't even acknowledged as predictions.
It's Installment #1. I am methodically going from beginning to end. It was a really long column.
>> The only real lesson here: any reform that changes who wins office will face backlash from some part of the establishment, because some part of the establishment will be negatively affected. <<
That's a falsehood. It's not the only real lesson. Just because some folks don't want to learn the entire lesson does not mean the rest of us should not.
The real lesson is that when RCV *fails* to respect Majority Rule and the Equality of our votes, in a tangible manner, the repeal effort actually can point to a legitimate reason for repeal. I don't wanna give folks like Phil Izon anything to hang their hat on. As RCV is used more and more, the frequency of these failures will increase and this failure will become less anomalous. As the other commenter pointed out, RCV is *not* growing all that fast and it is *not* prevalent in the U.S. It's early in the voyage that we need to make the course corrections. Problem is the resistance and denial of RCV organizations that there need be any course correction. Then when we arrive at a different place (which is democracy that respects majority rule and the equality of our vote and preventing spoiled elections and allowing voters' second-choice votes to actually count and freeing voters to "vote their hopes not their fears") then you're going to rely on people's short memories and say that's never where we wanted to go in the first place, to justify not making the course correction when we could have been least costly and most effective.
There are other real lessons to learn here that you have not addressed in your column, but I intend to bring it up.
I asked Steven Hill if DemocracySOS would consider publishing a rebuttal from me. That would be a far more careful reasons and concise rebuttal than we are going to get in comments. I have heard no response after two emails.
So, if that is the case, the rebuttal is coming in bits and pieces.
Now, one thing that is disingenuous, is to represent "The Case for Condorcet Voting" with an example that is cherry-picked to make the Condorcet results appear nonsensical. Now I presented the case for Condorcet RCV *solely* as the case for RCV. The very reasons we want Condorcet RCV are the stated reasons that RCV advocates make for RCV, yet those RCV advocates are usually plugging Hare or IRV.
Now the reason to use a Condorcet-consistent method to tally the vote, instead of Hare, is to prevent a failure of IRV to respect majority rule and the equality of our votes. This failure does not happen often, but when it *does* happen, the consequences are *never* good. So I will cherry pick (or make up) another example to show this. But I will not disingenuously call this "The Case for Hare (Instant Runoff) Voting". This is presented to point out the essential flaw of IRV instead. And that case is that of a close 3-way race, so my example is gonna make it *really* close. All three candidates are plausible winners.
99 voters in total
34 Right voters
---30 R>C>L
--- 1 R>L>C
--- 3 R only
32 Center voters
---13 C>L>R
---11 C>R>L
--- 8 C only
33 Left voters
---30 L>C>R
--- 1 L>R>C
--- 2 L only
Now that's a pretty close 3-way race. Who should win? Who truly has the most voter support? Who will Plurality elect? Who will IRV elect?
Now the FPTP people will say that Right should win in this close race because more voters prefer Right as their first choice than any other candidate.
The IRV people will say that Left should win because Left is preferred over Right by 46 to 45 votes (after Center is eliminated). Even though Right gets more first-choice votes (barely), Left is preferred over Right by 1 vote (again, barely).
But Center is widely preferred over either Left (62 to 34) or Right (62 to 35). You can call Center "milquetoast" if you want, but the electorate would *greatly* prefer Center over either Left or Right in this extreme example. Neither FPTP nor IRV will see that.
That is the cherry-picked example that really presents the Case for Condorcet.
> Ranked Choice Voting (RCV) is one of the fastest-growing voting reforms in the United States.
You can't be serious. RCV has been failing spectacularly, and is now banned in more states than it has been adopted in.
I genuinely don't understand why you people are so committed to RCV. You finally understand the Center Squeeze effect, and you recognize that it prevents RCV from fixing polarization, and that it perpetuates the same problems as our current system, but instead of allowing evidence to change your mind, you just double down and keep pushing an unpopular reform with bogus claims that its flaws don't actually matter in practice, even after it demonstrably fails.
I don't get it. "Getting RCV adopted" is obviously your goal, but shouldn't "improving democracy" be your goal?
Brief corrections and responses:
Yes, improving democracy is my goal. RCV is one of several reforms which I believe are important vehicles to doing that.
I haven't "finally" understood the center-squeeze effect. I've known about it for more than 20 years. If you read the piece, you will see that I do _not_ recognize "that it prevents RCV from fixing polarization."
Any reform that changes who wins office will face backlash from some part of the establishment, because some part of the establishment will be negatively affected. (In fact, if you make an electoral reform change and the establishment universally likes it, you should ask yourself whether you've made a substantial change at all.)
Despite that, RCV continues to make net gains every year in terms of adoption. There hasn't been a successful repeal of an implementation in many years -- strictly new adoptions. That isn't to say that a repeal here or there won't happen -- progress is normally a bumpy road and establishment figures often wield real power.
If you read the piece, you will find plenty of evidence and citations that contradict your "failing spectacularly" claim.
Installment #2. Regarding Alaska Special Election August 2022:
>>Did this election demonstrate the propensity for RCV to elect an “extremist?” Far from it. Peltola was a clear moderate ... <<
Okay, this is not what the failure of IRV in Alaska is really about. There *is* the Center Squeeze, which this election demonstrates. When there is a majority failure with IRV (a failure to elect the Condorcet winner when there is one), it is *always* the candidate in the Center who is robbed. One of the other candidates is the spoiler and the other non-Center candidate is the beneficiary of the spoiled election. Now that doesn't mean that IRV leans Left or Right. It just leans away from the Center. The mechanism for this is that in the semifinal round (when there are 3 candidates) IRV is opaque to the second-choice rankings of the ballots and cares only about the first-choice rankings (or votes promoted to that level) and counted as votes.
Now, that is much different than the IRV final round (which does it right). In the final round, it doesn't matter how low the two remaining candidates are ranked on the ballots. They could be ranked fourth and fifth, and if Candidate A is ranked higher than Candidate B, then that is a full vote for A. When choosing between two candidates in the final round, IRV is working. But what is at issue is *which* two candidates and also if it worked when there were more than two candidates and the method was opaque to lower rankings?
The failure of IRV in Alaska in August 2022 is that if failed majority rule, and as a consequence, did not value voters' votes equally. If our votes aren't to be valued equally, then I want my vote to count more than yours. (If that is unacceptable to you, then let's make certain we're counting our votes equally.)
In that August 2022 election, more Alaskans, 87899 to 79461 (an 8438 voter margin), wanted Nick Begich instead of Mary Peltola and marked their ballots saying so. But Mary Peltola was sent to Washington to represent the Alaskan people in Congress.
This last November, again, more Alaskan voters marked their ballots that Begich is preferred to Peltola by nearly the same margin, 7876 (164861 to 156985). But this time Begich is sent to Washington.
Both elections about 8000 more Alaskans said they wanted Begich rather than Peltola. And, both times, marked their ballots saying so. Both times Instant-Runoff Voting was used.
What was different?
Sarah Palin was in the race in 2022 and not in the race in 2024. And the outcomes were different winners.
Sarah Palin was techically the spoiler, a loser whose presence in the race alters who the winner is. More Alaskans preferred Begich and they voted saying that. Why should Palin's running for the same seat change how Alaskans feel about Begich vs. Peltola?
Mary Peltola in August 2022, unfortunately shares a distinction with George W. Bush in 2000 and with Donald Trump in 2016. All three candidates were elected to office when the public record indicates that more voters marked their ballots preferring a different specific candidate for that office.
Mary Peltola is a wonderful person, I would most certainly have voted for Peltola had I been living in Alaska. She's clearly moderate, but in the context of Peltola vs. Begich vs. Palin, Peltola is not the centrist candidate among voters in Alaska. IRV falsely promised to fairly deal with the GOP split vote problem with two Republicans running in the general election, but it didn't. IRV propped up the *weaker* of the two Republicans against the Democrat. Sarah Palin could not beat Mary Peltola head-to-head in the final round, but Begich *could*. Yet IRV did not allow that side-by-side comparison of voter support and that is the flaw.
There should be an IRV final round occurring between *every* pairing of candidates. If more voters marked their ballots that Candidate A is a better choice for office than Candidate B, we should not elect Candidate B, if that can, at all, be avoided. If we elect Candidate B, we are guaranteed a spoiled election, one of the candidates on the Left or Right is the spoiler, the Center candidate will be the candidate who is robbed, and the other candidate on the Right or Left is the lucky beneficiary of a spoiled election.