Tuesday, October 22, 2013

October Op-Eds Weave RCV Spell over Minneapolis

Minneapolis is conducting their local elections by ranked choice voting (AKA instant runoff voting) for the second time this November. I've already written about this a number of times. But with the election drawing closer, RCV proponents are getting op-eds published in an attempt to hype the systems purported advantages.

Please indulge me while I reiterate my refutations of these common, and false, RCV talking points.

  • RCV does not "ensure majority outcomes": With the large number of candidates running for mayor, and the inability of the Democratic-Farmer-Labor party to settle on an endorsement, there is an extremely high likelihood that the winner of that election will not have a majority of the votes cast, since a large number of votes will be exhausted. But be on the lookout, because reported results will likely conceal the total number of ballots cast, as they did in the 2010 Oakland mayoral election, giving only the number of "continuing ballots" in the final round.
  • RCV does not "encourage more respectful campaigns": We know this because of experience in California. It does seems that the first couple RCV elections are more respectful, but this is solely because of RCV advocates' claiming this to be true, and uncertain politicians being afraid that they might be right. But after having some experience with the new system, trying some negative ads, and seeing the effect on their polling, the politicians realize they've been tricked, and go back to their old ways.
  • RCV does not eliminate "fear of a wasted vote": This is the most persistent RCV myth, because it appears true at first glance. And it is true... when there are only two significant candidates. Since most elections have only two significant candidates, it can take quite some time for the truth to come out. But we only need to look at Burlington, VT to remind ourselves. It's unclear if we'll see it in Minneapolis this year though. I give it about 1-in-5 odds of occurring.

I do agree with the op-ed authors that this is perhaps the most important test of RCV yet. And I've already made one testable prediction about the outcome of these races: That the mayoral race winner will not win a majority of the votes cast. And now, I'd like to add a second testable prediction, one for which I have even higher confidence: There will be a higher percentage of spoiled ballots for this election than there were in the 2005 election (the last one before the change to RCV). This op-ed didn't repeat the claim, but previous ones had implied (falsely) there was only one spoiled ballot in the 2009 election. The actual rate was over 4.1%, while the 2005 rate was less than 1.1%. I expect similar performance this year.

Those are my claims. If any RCV proponent would like to make a friendly wager of it, you can step up in the comments.


  1. BTW, I'm thinking something like: If neither of my predictions come true, I will donate to FairVote, if both of them come true, whoever is willing to take my bet will donate to the Center for Election Science. (And if one and not the other, it's a wash.) Amounts negotiable.

  2. I think that the Burlington debacle should forever be held up as the example of what's wrong with IRV/RCV. As you pointed out in your link about the Burlington election, it exhibited favorite betrayal, center squeeze, non-monotonicity, and participation failure all in the same election!

    Much is made about the fact that RCV only elects those with "core" support, but nothing is said about whether or not those who place second in an RCV election are better off. Wright, the 2nd place finisher, was going to lose the election whether it was RCV, some Condorcet method or if he had faced off against Kiss or Montroll one-on-one. But it was only RCV that disallowed consideration of his voters' preferences, which were running 3-1 in favor of Montroll. This means that the candidate with the most "core" support was another ignored candidate along with Montroll. No consideration was to be given to their full ranking because the RCV process had to be honored, regardless of the wishes of Wright's voters.

  3. Indeed. "Core support" is a cover for "refuse to compromise."

  4. Well, the non-machine-supported candidate got elected and the election brought a lot of attention to the acheivement gap for minorities in Mpls and folks are happy with it. They'd like the vote-counting done more quickly, but that cd easily be fixed by treating the up to 3 rankings as approval votes in a first round that determined 3 finalists.

    I think that a lot of the BR models that predict a major diff in BR-results depend heavily on the a priori odds of the candidates. It may be true that the average number of candidates is 7, but it isn't true that all 7 of those candidates have voter-utilities drawn from the same distribution. Typically, there's a mixed distribution with some candidates being "competitive" and others "non-competitive".

    IRV lets the number of "non-competitive" candidates proliferate without spoiling the competition between the "competitive" candidates. It's possible that there cd be a feedback from the election rule to the expected number of competitive candidates, but in important political campaigns economies of scale of running a widescale campaign mitigate any feedback.

    As such, my prediction is that IRV improves on FPP closely comparably with AV or a Condorcet-like method if the poisson process that determines the number of competitive candidates(-1) has a mean of 1.3 or so. (let's say the mean currently is .5) You'd need a feedback big enough to get to like 1.7 before other election rules wd start to significantly improve on IRV in terms of BR. And if you complicated the model y relaxing the assumption of cardinal utility then Approval (or score) Voting wdn't pull away from the pack.

    Hope you are well.

    I think it's possible that a GOP civil war cd get the GOP establishment to support the use of IRV(modified as I advocate) for congressional elections since it'd take away the out-sized influence of the tea-party on the GOP and help Dems and moderate GOPers elect more moderate GOPers so we'd never have to worry about a gov't shutdown or default like we had or almost had happening again. And that cd get the Dems in the market for American forms of PR, since the GOP wd likely keep control of the HOuse due to the undue concentration of Dem-leaning voters in urban areas.

  5. IRV doesn't end the incentive for negative voting by the top candidate against their closest competitor, but it does mitigate it, more so when there's uncertainty as to who is the closest competitor.

    It's here to stay, maybe mutating some into the second round rule for a "top four primary" that replaces "top two primary" in CA/WA/LA, and you'd better get used to it...

  6. seems likely that GOP establishment might be pushing top 2 primary as a way to subvert tea-party. Methinks, FairVote's top 4 primary, using irv for the 4 candidates leftover in the 2nd round, that actually acheives what the proponents of top 2 primary claim it acheives would be the path of least resistance rejoinder.


  7. So no mea culpa about IRV not being so bad???

    1. I never said it was worse than plurality, but not going to apologize for saying that it's the worst thing other than plurality.