Why IRV Produces a Majority Winner

by Alec Slatky // Published July 12, 2010

            The Portland (ME) Charter Commission’s consideration of instant runoff voting (IRV) elicited spirited advocacy from supporters and detractors alike.  Under IRV, as described by FairVote, voters can rank all candidates in order of preference.  If no candidate wins with a first-choice majority, the candidate with the least first-place votes is eliminated and their supporters' second choices are distributed to the remaining candidates.  The process of elimination and redistribution continues until one candidate has more than 50% of the votes.

            Ultimately, the commission sided with supporters, voting 10-2 for a unified recommendation of a directly elected major using IRV (or “ranked choice voting”, as they call it), and the Portland Press Herald reported that “Commissioner Nathan Smith, who is leading the subcommittee [examining concerns about IRV], said Monday that the committee has found no reliable evidence that the new voting system would cause widespread confusion or disenfranchise voters. Ranked-choice voting is the best and most cost-effective way to ensure that the mayor is elected with a majority vote, Smith said.”

            As Nathan Smith noted, one criticism of the voting system was that contrary to claims by pro-IRV activists, IRV does not necessarily produce a “majority winner.”  This criticism is misleading, and does not recognize the true meaning of a “majority winner” in any given election.  This post aims to refute the idea that IRV fails to produce a majority winner.

Compulsory voting

            Compulsory voting is not part of United States policy.  Back in the 1770s, Georgia included it in its state constitution, but there are very few other examples.  The United States regards voting as a right – indeed, one of our most cherished and sacred rights – but not an obligation.  Voting is something people are encouraged to do, as they should be, but is not, nor should it be, an act that citizens are required to perform.

            Compulsory voting restricts the freedom to not vote, which citizens do for a variety of reasons.  Some citizens are not registered, and thus do not have access to the ballot box, a problem that should be remedied.  But plenty of other respectable citizens do not vote for a number of perfectly valid reasons: they may believe the system is illegitimate and fraudulent, and thus want no part of it; they may be religiously compelled to avoid politics in all its forms, like Jehovah’s witnesses; they may believe that since the probability of their vote changing the election is very low, the trip to the polls is not worth their time.  Feel free to disagree with all of these rationales, as most people do, but nothing gives the government the right to force people to act against them.

            The most obvious reason to abstain from voting is the lack of a favorite candidate, or more importantly, the lack of any candidates deemed even acceptable.  The phrase “the lesser of two evils” has become common parlance in American electoral discourse.  Some people do indeed cast their ballots for the lesser of two evils, which is fine.  Others prefer to avoid helping candidates whom they despise, even if one is despised slightly less than the other.  Some might even consider the two candidates equally bad.  The point is, if a voter doesn’t like McCain and doesn’t like Obama, who is to say that they should be forced to make a decision to support one of them?

            In instant runoff voting, the idea behind allowing voters to “bullet vote” – cast a ballot for only one candidate rather than ranking all candidates behind him – is the same.  If a voter has a clear first choice, even a clear second and third choice, but none of those preferred candidates make it to the final runoff, he or she does not have to express a preference for either of the final two candidates.  Voters can decide to do so, like those who choose the lesser of two evils in a plurality election, but they also have the freedom not to.

            Consider a slightly ridiculous example, and bear with me on a rewriting of history.  World War II, before the USA entered it, had 3 parties: the Axis, the Allies, and the U.S.S.R.  The USA is eager to support the Allies, with whom they align on general political views, and they do so.  Unfortunately, the Allies are wiped out, leaving only the Axis powers and the Soviet Union fighting it out.  The following question is soon raised in Congress: should we support the Soviets, the Axis, or neither?  My personal answer would be neither, but someone else might say the Soviets, or even, shockingly, the Axis.  Reasonable people can disagree on this point.  No matter what, the war will be decided – either the Soviets or Axis will emerge victorious – but the USA doesn’t have to take a side with a group it sees as morally bankrupt, and would prefer to see the rest of the world determine the outcome rather than being active participants.  They can take a side, but they are not required to, ethically or politically.  Now consider the same scenario, but where the League of Nations thinks that this war is so important to the fate of the world, that everybody must take a side.  Even with the debate about which faction to side with, if any, legislators in the USA would universally condemn this declaration as an infringement upon the country’s sovereignty.  People might disagree on whether the USA should pick a side, but few would claim that they should be forced to do so.

            Now, I grant that the preceding example was somewhat absurd, and that deciding to fight a war is obviously different than electing a representative, but the principle holds true.  The Allies are akin to a first choice that is eliminated in the runoff, leaving the Axis and the U.S.S.R. as the final two candidates.  Some will express a preference, and some will not, and both decisions must be regarded as equally acceptable.

Was Barack Obama a “majority winner” in the 2008 Presidential election?

            The obvious answer to this question is yes, Barack Obama was a majority winner in the 2008 Presidential election.  It would be nearly impossible to find anyone who says otherwise.  Forgetting about the Electoral College, Obama received 69,456,897 votes out of the 131,257,328 cast, equaling 52.92% of the votes.  Since that percentage is greater than 50%, Obama is considered to be a majority winner; put simply, he won a majority of the votes.

            Under any denominator other than votes cast, though, Obama does not emerge as the majority winner.  According to Census estimates, 30.8% of the total voting-age population voted for him; 33.7% of the total citizen voting-age population voted for him; even among just the registered voters, that number only increases to 47.5%, less than a majority.

            But if someone claimed that Obama was not a majority winner and cited these statistics, who would take him seriously?  It is clearly true that Obama did not receive the support – meaning votes – of a majority of the country, but this is not the metric on which we base calling someone a majority winner; if this is the case, then very few Presidents, including the ones who received greater than 50% of the vote, can be called majority winners.  This proposition is foolish.  The nonvoters did exactly that: chose not to vote.  If you choose not to express your preference, you are not counted – in fact, you are choosing not to be counted.  The reason for not voting is irrelevant, because we grant people the right to not vote, with a price: forfeiting your right to be counted.  So if nonvoters voluntarily give up their right to be included in the electoral results, then utilizing them to say that a majority has not been achieved is unfair.

Majorities in IRV

            Joyce McCloy has a new blog post describing the “Instant Runoff Voting Lie” that “IRV…provid[es] a 50% + 1 majority win,” and cites statistics from Cary, North Carolina, which she claims prove that IRV does not always elect a majority winner.  Her statistics are absolutely correct.  The District B City Council race featured three candidates.  In the first round of the instant runoff, Don Frantz received 1151 votes, Vickie Maxwell had 1075, and Nels Roseland finished with 793, in addition to 3 write-in ballots.  Since no candidate achieved a majority, the runoff was conducted using the ranked ballots.  The write-in candidates and Roseland were eliminated, and their 796 votes were distributed to the higher ranked of Frantz and Maxwell, according to the ballots.  After the runoff, Frantz won the election with 1401 votes, which is 50.9% of the runoff vote, but only 46.4% of the initial votes cast.  What happened?

            There were 3022 votes initially cast, and 2754 tabulated in the runoff.  That means that 268 ballots were not included in the runoff, for a very simple reason: they chose to rank neither Frantz nor Maxwell, but instead bullet voted for Roseland (if the write-in voters made up part of those 268, they may have ranked Roseland second but still ranked neither of the other two candidates).  This essentially amounts to not voting in an election between the top two candidates.  In fact, these voters were in the voting booth, so there was no reason for nonvoting like a general opposition to the system.

            There are two conceivable rationales for not ranking Frantz or Maxwell on the ballot.  First, a voter might have been completely indifferent between the two, and in an election would be equally pleased or upset if either of them were victorious.  This voter would not have voted in a plurality election between the top two candidates, because he would not want to express a preference he does not truly have; this applies to IRV as well.  The other possible reason is that both candidates were so abhorrent to the voter that even choosing between the lesser of two evils was unacceptable.  It is not a large assumption to guess that those voters would be nonvoters in a separate election between the top two candidates.

            The argument boils down to this: some number of voters, 268 in Cary, did not care to choose between the top two candidates, even when given the opportunity to do so.  They voluntarily opted out of the election at that point, and thus are not counted, just like nonvoters are not counted, because neither expressed their preference.

            A possible response might be that if those 268 voters knew that the election would come down to Frantz and Maxwell, they would have expressed a preference.  But they knew beforehand that there were three possibilities: Frantz and Roseland are in the runoff, and their votes help Roseland; Maxwell and Roseland are in the runoff, and their votes help Roseland; Frantz and Maxwell are in the runoff, and their votes help nobody.  They knew that ranking either Frantz or Maxwell second could never hurt their first choice candidate, Roseland, and yet they declined to do so.

            IRV voters should not have to rank all candidates, for the same reasons as we permit registered voters to sit out an election.  As long as we do not have compulsory voting as part of our electoral system, voters are allowed to not express a preference.  When that happens, they become part of a pool that is not counted in the election, and thus should not be under consideration when determining a majority winner.

What we mean by a “majority winner”

            A majority winner is based on, essentially, a majority of the willing, a majority of the voters who expressed their preferences.

            If a majority winner means a candidate who won a majority of the electorate, then no, IRV does not always produce a majority winner.  But this has never been the definition used in political discussion.

            If a majority winner means a candidate who won a majority of voters who displayed any preference, including those who did not rank either of the final two candidates on the ballot despite given an opportunity to do so, then no, IRV does not always produce a majority winner.  But this is not the definition when it comes to nonvoters, so it should not be the definition when it comes to the voters who, when given the opportunity to do so, declined to include either of the final two candidates on the ballot.

            If a majority winner means a candidate who won a majority of voters who expressed a preference between the final two candidates, which is the only description logically consistent with the modern connotation, then yes, IRV always produces a majority winner.

Why IRV Produces a Majority Winner, in 4 sentences

            All registered voters can vote; some choose not to.  All voters in IRV can rank all of the candidates; some choose not to.  Since we don’t consider nonvoters when describing the winner of a plurality election as a “majority winner,” there is no reason to consider voters who choose not to rank all of the candidates in IRV when describing the victor as a “majority winner.”  Under all reasonable definitions of the phrase, the mathematics of IRV requires it to produce a majority winner.

Comments currently closed for Why IRV Produces a Majority Winner

  • nice blog
    Posted by cheap vps, 2010-11-02 22:46:13 (3 years ago)

  • Actually IRV does not always elect majority winniers, as FairVote learned when they decided to vote on lunch. Score Voting picked pizza, whereas IRV picked Chinese. 60% of the voters preferred pizza to Chinese - so a majority of voters preferred Score Voting to IRV in this case. And that's in spite of the appearance/claim that voters were tactical with Score Voting but sincere with IRV (there's no way to tell whether the latter claim is true, so we'll take FairVote's word for it).

    The later-no-harm criterion is basically irrelevant, since the vast majority of voters are strategic, and thus their top priority is to maximally support their favorite of the two frontrunners, and maximally oppose the other one. So if you prefer "Third">"Major Party A">"Major Party B", your best tactic (www.electology.org/debate/IrvPlurality) is to insincerely rank "Major Party A" in first place.

    Score Voting and Approval Voting are vastly superior in the context of tactical voting, because even after the voter has tactically supported Major Party A (e.g. A=10, B=0) he has nothing to fear from also giving a maximum score to his sincere favorite, the Third party candidate. Because of this, the Third party candidate can win if enough voters sincerely prefer that candidate -- even if they think he has no chance to win. This demonstrates why it's the favorite betrayal criterion that matters, not the later-no-harm criterion. The later-no-harm criterion is a misleading construct that seems designed specifically to make IRV look good. In fact it is actually harmful for a voting method to pass LNH, since it forces IRV to ignore lots of ballot data.

    Ironically, FairVote's lunch election actually demonstrated IRV's susceptibility to tactical exaggeration. If Pauline or Rob had insincerely/tactically ranked pizza in first place, then Middle Eastern would have been eliminated, and pizza would have won. There was ample opportunity for tactical voting here; only perhaps you couldn't exploit that since you didn't have pre-election polling. But if this had been a real political election, on the other hand, and you had seen a pre-election poll that showed Chinese being preferred to Middle Eastern by 60% of the voters, and pizza being the first choice of only 20% of voters (compared to 40% for the other two "candidates"), then you would have easily been able to deduce that it was "throwing away your vote" to put Middle Eastern in first place. Bloggers and televised pundits would have made note of this, effectively treating Middle Eastern as something of a Ralph Nader (a spoiler that you don't want to waste your vote on). You would have had sound rational justification for strategically insincerely ranking pizza in first place. And even if you didn't actually do it, lots of other voters would do it.

    Like Dale, I was highly skeptical of score voting when I first came across Warren Smith's site back in the summer of 2006. I initially felt that Condorcet was the most compelling method. I naively expected that voters would bullet vote for their favorite candidates, before ever trying to put myself in their shoes and imagine what I would do if e.g. Ralph Nader was my favorite candidate (i.e. quite likely give Gore and Nader a 10).

    Election theory is famously counter-intuitive, and it's quite common for people to naively expect that Score Voting and Approval Voting will fall apart due to tactical voting. It's only after complex mathematical analysis (e.g. Bayesian regret calculations) that the falseness of that argument is completely clear. Prior to Warren Smith's Bayesian regret calculations a decade ago, we just really had no idea how much worse IRV was than approval voting. And it seems FairVote is so locked into a strategy they chose back in the early 90's, that they generally refuse to accept modern election science that turns everything they believed on its head. I think they need to wake up.

    Posted by Clay Shentrup, 2010-07-23 14:34:42 (3 years ago)
  • @Alec

    Voters might act strategically in approval, and might hurt themselves or their candidates. Yes. Voters might act strategically in instant runoff, and might hurt themselves or their candidates. Yes.

    Yes, but the important questions are HOW OFTEN and HOW MUCH?

    All electoral methods inject some evil, which is why we need some sort of holistic framework which can collect the negative effects of ALL possible failures and find out what method creates the least of all evils.

    That's what Dr. Warren Smith's simulation was for. Here is what he found: score voting. Second place: approval voting.

    I'm not saying it's perfect. I'm not saying these simulations are without flaws (although many concerns have been raised against them, and then incorporated into the simulation; the results didn't change significantly). These results are at least COMPELLING and SUGGESTIVE that approval and score are at least worth TRYING; much more-so than IRV is. And yet, FairVote continues to press for a sub-optimal process; I find it all so frustrating. Don't you want the best possible voting system? Don't you see that all your arguments are qualitative and based on intuition? Don't you see that we have arguments that are quantitative and based on analysis?

    I wasn't always an approval-advocate; I sat down with Smith's work and tried to prove to myself that he was WRONG. I couldn't. I, like people here, started with some intuitive argument; then I sat down and started running the math to prove its correctness. But the math always worked out in favor of approval. So take a look; a good, hard, long, QUANTITATIVE look. That's all I ask.

    Posted by Dale Sheldon-Hess, 2010-07-16 16:25:16 (3 years ago)

  • @ Mike

    Thanks for commenting. And you raise an important point about voter education. Only 5% of voters in Cary said that it was not easy to understand the IRV ballot; even if we assume that all of those people bullet voted for Roseland, it accounts for barely half of the exhausted ballots. Under the more realistic assumption that voters who did not understand the IRV ballots were proportionally distributed, only 15% of the exhausted ballots can be explained by this phenomenon.

    You’re right that voter education is extremely important to any new voting system, and should be emphasized. There are still people under the current system, though, who don’t know how to register to vote. Though this problem needs to be fixed, we don’t count them in describing a winner of a plurality election as a “majority winner.” Similarly, while voter education efforts are not 100% successful, IRV can still be described as producing a “majority winner” in the context I specified.

    I don’t want to de-emphasize your point that voter education is crucial to the success of any election, though. I couldn’t agree more.

    @ Dale

    Under approval voting, strong third-party candidates certainly can be spoilers and flip the election to the least preferred candidate; they do not necessarily cause the second choice to win.

    Let’s use your example: candidate A (GOP) has 4 supporters, candidate B (Dem) has 2 supporters, and candidate C (Green) has 3 supporters. We can go through the mindset of the voters for each in order.

    Candidate A has a legitimate shot at winning the election. All 4 supporters prefer B to C, but choose not to approve of either on the ballot. They realize that approving of B will hurt C’s chances to win, but also know that it hurts A’s chances to win. If A were a fringe candidate, then supporting B would be a no-brainer, but since their preferred candidate has a great shot at winning the election, A’s supporters choose not to vote for anyone else.

    Candidate C, a third party candidate, also has a legitimate shot at winning the election, and the thinking of his supporters is the same as for A’s. All 3 prefer B to A, but choose not to approve of either on the ballot. The first thought might have been to vote for B as well, because that’s normally what third parties would do. They would approve of both their candidate and one who had a real shot at winning the election. But now the Green Party is no longer a patsy, and their members think, “Well, we used to approve of two candidates. But that was because we had no shot at getting our guy elected. Now, we’re polling equal to or better than the Democrat and there’s a real chance that we can win this thing, so why let that chance slip away by hedging our bets and voting for B?” And so C supporters choose to only approve of C.

    For Candidate B, the thought process can go either way, but it doesn’t really matter. If the supporters think like those of the other two, and believe that B has a legitimate shot, they won’t want to screw that up and vote for either of the other two more radical candidates. In this way, all voters approve of only one candidate, and A wins, making C a spoiler. If B’s supporters are bearish about B’s chances, they will approve of their second choice, to make sure their last choice doesn’t emerge victorious. Under your example, that means 1 more vote for A, and 1 more vote for C, again leading A to victory and making C a spoiler.

    I think the point of the preceding example can be summed up as follows: if third parties continued to follow traditional third-party strategy – voting for themselves and the less hated of the major two parties – even when the third party was strong, then yes, they would elect their second choice candidate. But if, in accordance with their rise to a level of popularity approximately equal to that of the major parties, they adopt a major-party strategy – that is, vote only for your most preferred candidate – then a third party can act as a spoiler and help elect its least preferred candidate. The second scenario seems just as likely, if not more likely, than the first.

    When voters vote honestly, later-no-harm can come into play, and lead to the election of a second choice candidate. But there are two problems with this assumption. First, it’s hard to describe an honest vote in approval voting (even more difficult in score) because there is no universal cutoff point for approval. Some voters may honestly approve of 5 candidates, some of 1, and it’s impossible to demonstrate that either was being strategic; therefore, it’s also impossible to demonstrate that either was being honest. Second, when voters are thinking strategically (though we may not know it), as in my example above, they may take the later-no-harm criterion into account, and vote in such a way as to render it irrelevant. By ensuring that their vote won’t help a second-place candidate defeat a first-place candidate, they are taking a risk between their most and least preferred candidates, incurring the same description you assigned to IRV.

    Posted by Alec Slatky, 2010-07-16 13:23:20 (3 years ago)
  • (Oops; thanks for the heads up on the link!)

    "Since approval voting (and score voting) violate the later-no-harm criterion, voters will be aware that approving of a second choice can actually defeat the first choice."

    This is true: which is precisely why approval (and score) is such a GREAT system.

    Because in every other system--including IRV--when you get to that tipping point, where a third party has a chance of winning, the risk the voter takes is that their LAST choice wins instead of their first. That approval delivers you your SECOND choice is a huge improvement! Both for you, for that third party (who gets out of it without being a spoiler and with measurable support from the voters), and for everyone who had a stake in the horse race between the top-two candidates (because the more-preferred one of that pair still wins the election).

    Later-no-harm is a trap. Because it concerns itself with the damage that the CANDIDATE takes (that is, they are not elected), where we should concern ourselves with the damage that the VOTER or the ELECTORATE takes. The trap comes in that it has been shown that any voting system which satisfies later-no-harm* is prone to these "first-or-last" decision points, meaning that the voter's damage is "least-favorite-elected"; the worst possible outcome, and the fear of that outcome will lead to electing the lesser of two evils. Approval, by forsaking LNH, is able to limit the voter's damage to "second-choice-elected", which on average leads to a much better outcome for everyone involved and a healthier political climate for future elections.

    *Technically, later no harm AND independence of clones; but both IRV and approval clear IoC, so let's put it aside for now.

    Posted by Dale Sheldon-Hess, 2010-07-16 11:23:40 (3 years ago)
  • This article puts forth sensible, logical arguments, until it gets to "They voluntarily opted out of the election at that point, and thus are not counted, just like nonvoters are not counted, because neither expressed their preference." Call it "voluntary", but really 268 is probably closer to the number of people who didn't understand that, unlike non-IRV elections, to vote against the one they liked the least they had to rank all others. It's an education issue for a more complicated voting system, bound to leave some people out more than others.
    Posted by Mike LaBonte, 2010-07-16 11:14:05 (3 years ago)

  • @ All commenters:

    I’ll try to clarify again what I mean by majority rule. As I said earlier to Markus, “[Joyce McCloy] thinks that IRV candidates who receive fewer votes in the runoff than 50% of all votes cast are not majority winners, implying that candidates who receive greater than 50% of the total votes in the runoff are majority winners.” She said that Don Frantz won with a plurality of the votes because he had less than 50% of the original votes, which is the position that I was trying to refute, because bullet voters are analogous to nonvoters.

    By majority winner I don’t mean a Condorcet winner, which seems to be a main criticism. It is of course correct that IRV fails the Condorcet criterion, as do all other non-Condorcet methods. (Incidentally, I should have another blog post coming out soon – maybe the beginning of next week – about the Condorcet criterion. Hope to continue the discussion there.)

    @ Dale

    The logic of your example is extremely clear, and indisputably correct. However, I disagree with the presumption that approval voting would lead to a different outcome. It is possible that when three candidates are almost equally strong, the two candidates who are most aligned will have voters approve of both, ensuring a defeat for the other side. But it also seems quite likely that once a third party becomes formidable, the approval voting strategy would no longer be to approve of a preferred third party and an acceptable major party. Since approval voting (and score voting) violate the later-no-harm criterion, voters will be aware that approving of a second choice can actually defeat the first choice.

    In your example, for instance, candidates A, B, and C are relatively close in votes, and the media would probably indicate as such. The C voters might think, why should I vote for B when C has a very good shot at victory? And so the C voters only vote for C. Same goes for A. Then, B’s votes can go either way. If they think B has a legitimate shot at winning, they might vote for only one. If they think B has no shot at victory, they might each vote for two choices. In both cases, A ends up winning, and C is a spoiler.

    Also, on an unrelated note, your website keeps liking as leastevil.blogpsot.com, rather than leastevil.blogspot.com. I fixed it for you, but just wanted to give you a heads up – if you visit the first site, you'll see why I was pretty shocked that someone from there would be interested in electoral reform!

    @ Clay

    Thanks for commenting. Again, refer to my statement at the top of this comment that a “majority winner” is not a Condorcet winner. I mean that in the final runoff, the winner gets a majority of the votes, and that including exhausted ballots in describing someone as a majority winner is counterintuitive.

    Your main criticism, as I see it, is that IRV does not always elect a Condorcet winner – C in your example. The same criticism can be applied to score and approval voting. If A and C are portrayed as the main front-runners, and D voters are generally known to prefer B to the other two, then voters for A, B, and C all have a strong incentive to vote only for their top candidate, providing a scenario under which approval and score voting produce the same result.

    As far as the idea that IRV degrades to Plurality Voting, I must take issue with the “ALWAYS exaggerate” portion of your link. The scenario in which IRV fails, as the link puts it, is one where there are more than two front-runners, so by definition there is no “favorite frontrunner” but rather three approximately equal candidates, and so saying that ranking the “favorite frontrunner” higher – presumably one of the more moderate major party candidates, as the scenario discusses – is more likely to lead to defeat of the least preferred candidate is inaccurate.

    Of course, in some cases it is rational to be strategic, but these cases are fewer in number than under score or approval voting, where strategy is inherent in every decision (score – what precise ranking should I give; approval – where should the cutoff be). And score and approval fail the later-no-harm criterion, which is one of their major flaws, so saying that IRV is worse in every way is also unfair.

    @ John

    Thanks for commenting. As I was trying to demonstrate, a plurality winner is one who wins an election, but less than 50% of the votes. But under IRV, everyone has a chance to rank all the candidates, so if a candidate wins will less than 50% of the votes, that means that some people did not choose between the final two candidates; they shouldn’t be factored into the description of “plurality” or “majority” when nonvoters aren’t included either.

    And yes, I think typing html directly is the way to go – I don’t know too much about computers but I believe that’s correct.

    @ AllAboutVoting

    You are right – IRV shows a majority between the two final candidates, not the whole field of candidates. This is why it doesn’t pass the Condorcet criterion. As far as top-two runoff goes, the increased attention on the final two candidates tends to lead towards significant negative campaigning and not intelligent debate. It’s certainly conceivable that a few voters who declined to rank either of the final two candidates under IRV would vote in a runoff, but they would almost certainly be offset by the usual decline that accompanies a runoff.

    I still fail to see why “using all ballots cast” makes any more sense than including all registered voters in determining a majority.

    Posted by Alec Slatky, 2010-07-16 08:37:48 (3 years ago)
  • Alec: My point from the earlier post:
    • The claim that IRV produces a majority winner is as absurd as claiming that it produces a winner with unanimous support
    And some other points:
    • IRV shows a majority between two candidates only and not the whole field of candidates. (Clay gives a good example of this where an eliminated candidate was preferred by the electorate to the IRV winner. This is quite common with IRV.)
    • TTR is not a great system but at least with that system you have an electorate/media that can become more familiar with the remaining candidates which will inform both their participation and choice. IRV's election of Ed Jew in a San Francisco election is a frequent example of this where TTR would have presumably brought in more media attention and revealed some of the flaws of the winner that were no well known to the electorate during the IRV race.
    • When a legitimate majority can be found in a single election (using all ballots cast - not just continuing majorities as in IRV) then I favor avoiding a runoff for the usual cost and participation reasons. IRV achieves it's majority in a way that I consider to be capricious and illegitimate. Direct counts via, eg, approval voting can achieve what I consider to be a legitimate majority (eg. candidate with the highest vote total that received at least #voters/2 votes.)
    Posted by AllAboutVoting, 2010-07-14 09:08:06 (3 years ago)
  • Who funds the Center for Voting and Democracy and fairvote.org? SourceWatch has no information on your source of funds.
    Posted by john francis lee, 2010-07-14 05:39:09 (3 years ago)
  • Having read the analysis below, I take issue with giving equal weight to the choices in all four tiers of vote. 35% A > C > D > B 17% B > C > D > A 32% C > D > B > A 16% D > B > C > A The first choice is not even clearly preferred , let alone the fourth tier! I imagine the fourth tier "preference" as being along the lines of "Oh my god I hope B (for example) never does win but she's better than E! So you are cheating in your calculations when you weight all four tiers equally. If you can't win the argument... baffle 'em with bullshit... is what it looks like to me :) After tier 1, and not even then really, I see now way to quantify the degree of aversion an individual voter has for the candidate in that tier,,, except arbitrarily. And 100% support don't cut it.
    Posted by john francis lee, 2010-07-14 05:36:46 (3 years ago)
  • "This proposition is foolish. " The proposition alluded to is that a plurality is not a majority. It is not foolish. What is foolish is allowing the sloppy, misleading use of language that allows the designation of one elected by a plurality as a "majority" winner. This is an important distinction to make. I am a proponent of a national Recall, Referendum, and Initiative amendment to our Constitution. The effect of reclaiming our sovereignty will put back TOTAL power into our hands to amend the Constitution, to overturn laws passed by the Congress and signed by the President, and to initiate and pass laws ourselves,,, laws that cannot be undone by the Congress or the President, but only by ourselves. The only way to safeguard such direct, awesome power is to require a majority, not a plurality, in its exercise. I agree with all you've said above. Your choice of words is wrong, of course, and unfortunate. Please see http://www.28amen.org, and http://www.uspvp.org if you are interested (and you should be :) in a national Referendum, Recall, and Initiative amendment, and a Campaign Finance Amendment as well. And never call a plurality winner a majority winner again! Do you here? :)
    Posted by john francis lee, 2010-07-14 05:26:29 (3 years ago)

  • IRV does not pick a "majority winner". Here's an example (oversimplified for clarity, but you can make it arbitrarily more complex/realistic to your liking).

    #voters their vote
    35% A > C > D > B
    17% B > C > D > A
    32% C > D > B > A
    16% D > B > C > A

    A huge 67% majority of voters would rather have candidate C than candidate B. And candidate C received nearly twice as many first-place votes as candidate B, 32% to 17%. And an even larger 83% super-majority of voters would rather have candidate D than B.

    So to be precise, IRV ensures that the winning candidates is preferred by a majority to at least one other candidate, and nothing more.

    Additionally, IRV degrades to Plurality Voting when voters are strategic. Score Voting and Approval Voting are better in essentially every way, particularly in that they pass the Favorite Betrayal Criterion (so a voter need never feer supporting his sincere favorite candidate).

    Posted by Clay Shentrup, 2010-07-14 02:25:43 (3 years ago)
  • Thanks for the warm greeting. I think we mostly agree that the goal of reform should be to allow new, popular ideas--new parties--to grow and, eventually, win elections. What worries me about IRV is that, unless that growth is astoundingly fast in the inter-electoral time (pushing them from under-25% of first-ranks, all the way to a high enough percentage to win), there is still the possibility of spoilers at some point along that growth curve. And becoming a spoiler is what will kills a third party. And the way that would go down under IRV is that the third party candidates moves up to 2nd in the first-rank tally by taking a handful more votes from the Condorcet winner (who up to this point had been a consistent winner), so the CW is now eliminated for being in third, and if only a few of his supporters break for the other guy, then the third-party candidate just spoiled the election in the eyes of all his supporters; his party is dead. (I realize that is a bit confusing in text... I have a graphical example: http://leastevil.blogspot.com/2009/06/presented.html
    Posted by Dale Sheldon-Hess, 2010-07-13 12:56:05 (3 years ago)

  • @ Dale

    First, thanks for commenting. I think your inference is quite a logical leap, though. Winning a majority of the votes between the "two strongest candidates" says absolutely nothing about what political party they are. Are the Democrats and Republicans the two strongest parties now? Of course. Would instituting IRV change that immediately? Of course not. But the "two strongest candidates" could be a strong third party, or an independent. When I say "two strongest candidates" I mean the two final candidates, and I apologize if this was unclear in context. But this assumes absolutely nothing about the political positions of the candidates.

    As a general aside (and this is simply my own personal opinion), I dislike the two-party system strongly. I'm a registered member of a third party, so I'm quite familiar with how plurality disadvantages third parties. Nearly all alternative voting systems will improve the situation for third parties. But the point of electoral reform shouldn't be merely to break down the two-party system. It should make it easier for a third party to become successful, but an immediate dismantling of the two-party system should not be the goal. I say this not because of stability arguments - which I think are nonsense - but because if America supports two parties, then the representatives should reflect that. Breaking down the barriers put up by the two-party system will take not only electoral reform, but a change in the mindset of the American public, and successful campaigns for support by third parties.

    Okay, enough of my personal tangent. Basically, if you are interpreting my statements as an admission that IRV entrenches the two-party system, you are mistaken. I mention political parties nowhere in this article, nor in the comments, so trying to discern any hidden message along those lines is a fruitless search.

    Again, as I said to Markus, I'm refuting the definition of "majority winner" as given by Joyce McCloy. Just because I apply the plurality connotation of the phrase to what it means under IRV, does not mean I am equating the effects of the two systems.

    Posted by Alec Slatky, 2010-07-13 12:12:43 (3 years ago)

  • @ Markus Schulze

    You're quite right, but I was trying to touch on a different point, because "majority winner" is not really a voting criterion. The closest analogue would be the Condorcet criterion, but as I pointed out to Stephen, this post is not about the idea of a "majority winner" in that sense (which I don't think is the general connotation at all), but what Joyce McCloy was referring to. She thinks that IRV candidates who receive fewer votes in the runoff than 50% of all votes cast are not majority winners, implying that candidates who receive greater than 50% of the total votes in the runoff are majority winners. I was simply trying to point out the incompatibility of these two sentiments.

    Posted by Alec Slatky, 2010-07-13 11:46:55 (3 years ago)
  • "What I’m referring to, essentially, is the definition of “majority winner” we have under plurality or two-round runoffs [...] The meaning for IRV is essentially the same: it elects the winner with a majority of the votes from the two strongest candidate" So you're saying that IRV will keep us trapped in a two-party system? Glad to finally hear that admission.
    Posted by Dale Sheldon-Hess, 2010-07-13 10:40:24 (3 years ago)
  • Well, it would be better to say that IRV satisfies independence of clones, majority for solid coalitions, and the Condorcet loser criterion. These terms are well defined.
    Posted by Markus Schulze, 2010-07-13 10:18:24 (3 years ago)

  • @Both commenters: thanks for commenting. One clarification I could make is that by majority winner, I do not mean a majority of first place (or even second place) support, because the nature of voting makes that impossible. I was specifically responding to Joyce McCloy’s post, which asserted that because a candidate received less than 50% of the original vote, he shouldn’t be considered a majority winner.

    @AllAboutVoting

    I’m not quite sure what you’re getting at. Yes, IRV proceeds through a series of eliminations, but there is a world of difference between a choice between two candidates and one candidate. It’s possible to express a preference between two, but not one. Any notion of majority or unanimity in a choice of one is silly; comparing two candidates – the two final candidates based on their breadth and depth of support – lends itself well to the idea of a majority winner. I know your post was sarcastic, but I’m trying to discern what your point is.

    @Stephen Unger

    The example you’re referring to is a demonstration of how IRV fails the Condorcet criterion, and you are absolutely correct in your explanation. However, plurality, approval, range, and Borda count voting all also fail this criterion, and could potentially fail to elect a Condorcet winner, or candidate who defeats all others in a pairwise election. Can we never call any candidate that wins under these systems a majority winner? Condorcet methods have their merits, certainly, but it seems unfair to consider these the only methods that can produce a majority winner.

    What I’m referring to, essentially, is the definition of “majority winner” we have under plurality or two-round runoffs, the two most prominent systems we have in America today. The meaning for IRV is essentially the same: it elects the winner with a majority of the votes from the two strongest candidates – strongest, and not broadest. By your definition, someone who gets 65% of the votes in the runoff cannot be classified as a majority winner, which seems to conflict with the general usage of the phrase.

    Posted by Alec Slatky, 2010-07-13 09:22:50 (3 years ago)
  • Consider the very simple example below of a 3-candidate IRV election. 2 B>A>C 2 C>A>B 1 A>B>C 60% of the voters (3 out of 5) cast votes indicating a preference for A over B, and 60% of the voters ranked A over C. By any reasonable definition of majority preference, we would have to say that A is preferred by a majority. But the IRV procedure declares B to be the winner, defeating C in the second round, after A is knocked out in round-1. This example illustrates the point that the IRV process guarantees only that the winner is preferred by a majority over ONE other candidate (the one encountered in the final round.) In an IRV election a candidate preferred by majorities over EVERY other candidate might lose.
    Posted by Stephen Unger, 2010-07-13 07:55:40 (3 years ago)
  • sarcasm on Absolutely true. In fact I recommend that you make a minor twist to the IRV algorithm so that you continue one elimination further. Then every election will indicate unanimous support for the winner. sarcasm off
    Posted by AllAboutVoting, 2010-07-12 17:00:29 (3 years ago)