Churchill on AV

It has come to my attention that Winston Churchill once described the Alternative Vote (or instant runoff) as:

The child of fraud and the parent of folly.

I do not know the precise context in which he would have said it.

68 thoughts on “Churchill on AV

  1. Whether top-two or majority-plurality, the logic applies only under the assumption of one candidate per party, and parties strong enough to control the use of their labels.

    I am aware of no writing in which Duverger addressed open-entry systems such as the Louisiana system (no longer used there for Congress, as discussed above) or the now-California system.

  2. @Tom Round

    Clay, go phone the Australian Greens and ask them if they want IROV for the mainland lower houses replaced by a system where every [1] Green [2] Labor (or, more often, [1] Green [second-last] Labor) ballot becomes a ballot that says “I like both Labor and Greens equally!”.

    This is one of the most common fallacies in the debate between IRV and AV proponents. IRV advocates seem to think that a more representative ballot equals a more representative election outcome. Thus if they can express that they prefer e.g. Nader to Gore, that’s “better” than saying they support Nader and Gore.

    The problem is that representativeness isn’t just a function of how much data the ballot holds. It’s also affected by

    – the efficiency of the tabulation method, and
    – distortions of the actual data, caused by tactical behavior

    In the first category, IRV massively fails because it discards huge amounts of data, resulting in worse election outcomes. (E.g. it ignored the fact that a large bloc of Burlington voters who preferred the Republican also preferred the Democrat to the Progressive – it did not even consider/user that information.)
    http://scorevoting.net/IgnoreExec.html

    The result is, as I pointed out, bad election outcomes.
    http://scorevoting.net/UniqBest.html

    Australian Greens should want more representative/satisfying election outcomes just as much as any other voters.

    Further, Greens should want a voting method where there is absolutely no reason to fear supporting the Green if you prefer the Green to the frontrunners. With IRV, that’s only the case if the Green already has no chance. If the Green does a spectacular job and shows some hope of winning, it then becomes unsafe to support him, because he’ll more likely be a spoiler than a winner. That’s just a statistically obvious fact.

    Granted, the statement by this Green Party guy implies that voters fear supporting Greens not because they understand the strategy with IRV, but because they are ignorant of IRV. And maybe the same thing would happen with Approval Voting too. But at least we could be assured that math-savvy voters would never have to fear supporting Green because it might waste their vote.

    95% of more “get” the basic idea that you can give a first preference protest vote to a minor party yet still “put the Liberals last”

    Are you saying that these voters honestly think the Liberals are worse than all the other parties, including Family First and One Nation? You don’t think that’s actually exaggeration?

    Minor parties themselves use STV or AV for their own internal elections here, with the exception of the late Australian Democrats, who used Hallett’s variant on Condorcet for their own internal elections. None that I know of use Approval. None that I know of use range voting or other points systems.

    No big surprise since the Bayesian Regret calculations that shows the robust superiority of Score/Approval only came out in 2000, and were written about in popular literature only in 2008. And using a different system than the one used in the actual elections is probably unpalatable in general. I don’t see this as a judgment on the merits of the respective voting methods, as I’m sure most of those responsible for deciding such things don’t have expertise in voting theory. I’m willing to bet most of them couldn’t tell me the definition of monotonicity or Bayesian Regret — so such an ad populum arguments don’t seem convincing to me.

    And Liberal/ National competition in single-member electorates is most certainly not a[n] historical fluke. (The Liberals now hold more rural seats federally than the Nationals do).

    Please tell me how many of the 150 House seats exhibited competition between them in the last election, or the one before that, or the one before that. Examples?

  3. MSS (#51), like Clay, I found your terminology in #39 a bit confusing — until I realized that you were incorporating by reference the separate thread on Lousiana/Washington/California Top Two (or jungle) primaries.

    Clay is right up to a point, that Duverger made a distinction between plurality and two-round majority. But Duverger made an equally important distinction between two-round majority and PR. As Clays quotation makes clear, Duverger linked two-round majority with political systems in which minor parties exist in the shadows of the two major ones, as partners in two broad alliances. Might sound a little like Australia under IRV, actually.

  4. Tom (#37): If “it won’t help minor parties win seats” is your touchstone, then providing ammunition that opponents can use to shoot down an STV-PR initiative (if one is ever put) seems an own [odd?] goal.

    I think this becomes slightly more understandable (although not more sensible) when you link the zealous advocacy for range/score voting to it’s basis in utilitarianism. Suppose that (say) libertarian values are so important to the 10% who vote Libertarian that the “utility” of winning for them outweighs the “utility” to the rest of us of defeating the Libertarian candidate. In other words, those with more passionately held beliefs should have their votes given greater weight than those with less passionate beliefs. This is what range/score voting attempts to accomplish — or, at least, the political value judgment it reflects.

    I think the indifference to PR derives partly from the fact that philosophically it weights votes equally. It doesn’t try to “maximize utility”. The fact that PR is a very hard reform to win gives these folks even more reason to hope against hope that their single winner method can give minor parties a piece of the action.

  5. Bob – “own goal” = accidentally kicking the ball into one’s own goalpost area, which counts as a point for the opposing team anyway.

  6. @Bob Richard

    First, I’d like to ask you once again, do you acknowledge that you were wrong when you said, “no voting rule in single member districts is going to elect legislators from more than two parties — not IRV, not approval, not score voting”? According to Duverger, that was false, but I’ve yet to see you firmly acknowledge that.

    I think this becomes slightly more understandable (although not more sensible) when you link the zealous advocacy for range/score voting to it’s basis in utilitarianism.

    There are several sound mathematical proofs that the social utility of a candidate is just the sum of the individual voters’ utilities, thus utilitarianism is mathematically proven to be the only tenable social utility function.
    http://rangevoting.org/OmoUtil.html
    http://www.rangevoting.org/UtilFoundns.html
    http://www.rangevoting.org/PuzzlePage.html#p36 and #p37

    Put more bluntly, if you suggest any alternative social utility function, I can prove with mathematical certainty that it sometimes picks the wrong outcome, and thus is not the correct preference aggregation algorithm. Your gut might tell you that utilitarianism is wrong or subjective, but the math says your gut is wrong.

    Beyond that altruistic view of utilitarianism, there is the fact (pointed out by the economist Harsanyi) that a utilitarian social utility function gives a random voter the highest expected utility. So if you reject a utilitarian social utility function, you’re not only bad with math, you’re harming yourself by getting election outcomes which give you a lower expected value.

    This is what range/score voting attempts to accomplish — or, at least, the political value judgment it reflects.

    No, Score Voting doesn’t “attempt” to accomplish anything. Bayesian Regret calculations are an attempt to determine how utilitarian a voting method is. Score Voting just coincidentally happens to produce better/lower Bayesian Regret than the other common methods. Warren Smith didn’t know this would happen ahead-of-time; he just ran the calculations and the results favored Score Voting. Moreover, there are other (more complex) methods that are even more utilitarian than Score Voting (but generally thought to be politically infeasible).

  7. I think the indifference to PR derives partly from the fact that philosophically it weights votes equally. It doesn’t try to “maximize utility”.

    This is very wrong on many levels.

    We aren’t indifferent to PR. I have no idea what would make you say such a thing. Warren invented a system called Reweighted Range Voting which is better and simpler than STV, and he invented Asset Voting (which he later discovered had first been proposed by Lewis Carroll).
    http://scorevoting.net/RRV.html
    http://scorevoting.net/Asset.html

    I would argue that you are indifferent to PR, since it is federally illegal in the USA, and that won’t change unless we first repeal that law, which will obviously not happen until we implement a single-winner system which can break free of duopoly, which IRV doesn’t. I.e. if you want PR, you need to start by supporting Score Voting and/or Approval Voting.

    There is no mathematical basis to your claim that PR methods inherently weight votes equally. For instance, all deterministic voting methods can exhibit cases where a voter gets a better result by being tactical — so tactical voters have more “power”. This is yet another case where hard scientific facts refute your unsupported (and apparently un-researched) assertions.

    And we actually have spent some time thinking about how to do Bayesian Regret calculations for multi-winner methods. Again, your point that PR methods don’t try to maximize utility is incoherent. No voting method “tries” to maximize utility. Voting methods just do what they do, and researchers like Warren Smith try to calculate their utility efficiency.
    http://groups.yahoo.com/group/RangeVoting/message/7706

  8. @56, I think Clay’s comments just reinforce my point. Not only does he assert the superiority of utilitarianism, he also asserts it as irrefutable science rather than as a value judgment.

    Give me John Rawls over Jeremy Bentham any day.

    Also, it is silly to assert that Score Voting just coincidentally happens to produce better/lower Bayesian Regret than the other common methods.. No, this is not a coincidence. The “scores” in score voting are self-reported estimates of “utility”. To say that this method maximizes “utility” (at least assuming non-tactical voters and relative to other methods) is entirely circular. If Warren Smith was as surprised as he says he was by his simulation results, then he needed to take another look at the formulas.

  9. @57. Over a period of about 90 years, many countries have switched from winner-take-all to proportional elections. How many of them adopted approval or score voting as an intermediate step? How many of them even found it necessary to adopt two-round runoff as an intermediate step?

    Actually, the leading example of two-round runoff, France, did adopt PR for (I think) two brief periods since World War II and then reverted both times. No one counts France as a victory for PR, whether or not two-round runoff had anything to do with these adoptions (my guess is not).

    Clay, there is valuable research and theorizing on the factors leading to the adoption of PR. Some of the most valuable contributions are by our host, Prof. Shugart (MSS).
    Unfortunately for your point of view, this literature makes zero use of social choice theory, and has almost zero need for it.

    Aside to MSS, tangentially on topic — have you had a chance to look at Alan Renwick’s new book yet? Any comments? Thanks!

  10. @Bob Richard,


    @56, I think Clay’s comments just reinforce my point. Not only does he assert the superiority of utilitarianism, he also asserts it as irrefutable science rather than as a value judgment.

    I cited mathematical proof of that. Your response seems to me to be, “my gut says it’s a value judgment, so I’m not going to read or respond to the fancy math.”

    Nevertheless, I invite you to put your money where your mouth is and actually show me what you believe to be the correct social utility function. If it’s not utilitarian (i.e. additive, anonymous, linear, and non-normalized), I will quickly show you concrete cases where it picks the wrong outcome.

    But you just don’t want to believe me, so go ahead and try it, and I’ll show you.

    Also, it is silly to assert that Score Voting just coincidentally happens to produce better/lower Bayesian Regret than the other common methods.. No, this is not a coincidence. The “scores” in score voting are self-reported estimates of “utility”.

    Wrong. For one thing, tactical behavior distorts the actual utilities severely, causing most voting theory novices to naively assume that Score Voting falls apart under tactical behavior. For all Warren Smith knew, this was correct. It was actually a surprise to him that this wasn’t the case when he empirically tested it.

    Beyond that, the scores are normalized, which an additional distortion away from the actual utilities. So the scores are functions of utility, not exact measures of utility.

    It would be helpful if you would actually study these things before making continuous streams of inaccurate statements about them.

    To say that this method maximizes “utility” (at least assuming non-tactical voters and relative to other methods) is entirely circular.

    You again show that you are uninformed about the process. Even with no tactical behavior, normalization to a standard scale is a distortion of the actual utilities.

    More importantly, we do not assume non-tactical voters. The superiority of Score Voting to alternatives actually increases in proportion to the number of tactical voters. As I have pointed out to you on innumerable occasions, Score Voting behaves as well or better with 100% tactical voters as IRV does with 100% sincere voters.

    If Warren Smith was as surprised as he says he was by his simulation results, then he needed to take another look at the formulas.

    This comes after you have just proved you have no idea how the “formulas” work.

  11. Over a period of about 90 years, many countries have switched from winner-take-all to proportional elections. How many of them adopted approval or score voting as an intermediate step? How many of them even found it necessary to adopt two-round runoff as an intermediate step?

    But in the USA, multi-member House districts are FEDERALLY ILLEGAL (and Senate seats are elected in staggered terms, thus forced to be single-winner).

    In 1996, congresswoman Cynthia McKinney (who later ran as the Green Party candidate for US President) wrote, but failed to pass, bill HR 2545, which would have overridden previous 1967 US law that had outlawed multi-member districts, i.e. had made PR illegal. She re-introduced a similar bill, HR 1189, in 2001. It failed again. Then she tried again with HR 2690 in 2005. It failed yet again.
    http://scorevoting.net/PropRep.html#mckinney

    Regarding France, I don’t know what your point is. If they couldn’t get PR to stick under a top-two-runoff system, I certainly don’t think they’d get it with plurality or IRV, since those systems favor two-party domination, unlike TTR.

    I am well aware that extensive historical information exists about the adoption of PR. I simply think your assessment of that data is severely flawed. You ignore serious political obstacles to PR in the USA. You also routinely make reckless claims about various voting science issues which seem to come from your intuition, and are totally out of line with established facts. For instance, you previously have stated many times that single-winner voting methods cannot break free of duopoly, contrary to Duverger. And you did not understand the simplest facts about Bayesian Regret calculations (namely, that scores are essentially normalized utilities, and that tactical behavior is incorporated), and yet you asserted that my reasoning was circular regarding the utilitarian-ness of Score Voting.

    Prof. Shugart may have made some valuable contributions to the study of PR methods. But I am skeptical of his expertise, given his numerous logical/mathematical fallacies from previous writings.

    For instance, he once complained that Condorcet and Approval, “have the further, and closely related, flaw of possibly allowing a relative unknown, who is the first choice of a small minority, to emerge victorious.” He ignores the fact that IRV can elect a candidate who was the favorite of just 2 voters, and was generally highly disliked compared to other candidates.

    He argues, “I think a flaw that both Condorcet and approval share is the likely favoring of colorless, offend-no-one candidates.” If voters think being “colorless” is bad, then they can prevent colorless candidates from winning, by not voting for them.

    I respond to a slew of his similarly illogical arguments here.
    http://groups.google.com/group/socorg/web/shugart

    Unfortunately for your point of view, this literature makes zero use of social choice theory, and has almost zero need for it.

    This statement is highly irrational in my view. This is like saying that architects have no need to study architecture. They can just make unskilled imprecise drawings, and we shouldn’t be concerned with how well their designs hold up to earthquakes and such.

    When you actually analyze different voting methods scientifically (instead of from Bob Richard’s gut), you find that there are serious differences between them, and they all produce different outcomes that will be more or less representative of the will of voters.

    If you want to support STV because your gut just likes it, with no regard for its actual merits, then you will most certainly end up pushing a poorer system than you would have if you had decided to use science instead of your gut.

  12. > ‘FPTP spoilers hurt major party candidates quite a bit. IRV can protect them from spoilers, without a realistic threat that their duopoly control will be thwarted.

    You are mixing apples and oranges here. Yes, spoilers are a nuisance to large parties under FPTP, but they result in Big Party A losing a seat, not to a minor party, but to Big Party B. Vote-splitting at any given election is a zero-sum tactic. If it helps one major party, it hurts the other.

    Vote-splitting under FPTP does not, in general (Martin Bell and the Canadian aside) help minor parties win a single seat election. Indeed, for the Big Two to tell potential (eg) Green voters “you’re throwing your vote away and helping the Tories to win”, and to have this confirmed by reputable journalists and political scientists who understand how the system works, is a good way to drive down first-preference support for minor parties on election day. Look how many Britons told opinion pollsters they liked Nick Clegg, but then voted tactically for Lab/ Con on election day. Look how many “Nader traders” in the USA have met through vote-swapping websites.

    Moreover, over time, vote-splitting’s advantage among the Big Two tends to cancel out. From 1955 to 1974, when the DLP was still strong in Australia, Labor politicians supported FPTP or at least making preferences optional, while the Libs and Nats defended exhaustive-preferential “number every square”. Now that the Greens are running at 15% in opinion polls and looking to take three Labor House of Reps seats (clearly they haven’t read enough Warren Smith), it’s Labor that’s defending compulsory preferences while a few Liberals (eg, Scott Morrison?) have started to wonder whether “voluntary” is as good when applied to voting systems as when applied to union membership.

    If G*d grants me enough years of life I will address your other points. In a nutshell, any electoral system that can allow a candidate who is the first preference of more than 50.1% of the voters to be defeated by her own supporters’ second preferences is defective.

  13. What Toms aid, moreover the strength of your theoretical case might improve if your empirical evidence about Australia showed some ground in reality. You cannot prove your case by making empirical claims that are simply untrue (we are still looking for the famous NatLibs) and then proclaiming that your mathematics is indisputable. That’s counting angels on pinheads.

  14. @Tom Round

    any electoral system that can allow a candidate who is the first preference of more than 50.1% of the voters to be defeated by her own supporters’ second preferences is defective.

    This is the opposite of the truth. Any voting method which always elects majority winners is defective. This is mathematically proven.
    http://www.electology.org/criteria/majority

    (This was the major reason why Arrow’s Theorem was so disconcerting.)

    As for spoilers, they may hurt both parties equally overall. But when it comes to incumbents (the people with the power, who make the laws and executive orders), they are generally very safe. I’ve heard that (in the USA at least) they are more likely to be unseated by death or resignation than by electoral defeat. That may be an exaggeration, but even then, probably not much of one.

    A major factor there seems to be that if they won, it was usually because their party was already dominant in their district, such that the “real” contest is their party’s nomination. Yet voters are unlikely to nominate someone other than the incumbent, because they fear losing to the other party, and the incumbent has already proved that he can win.

    Spoilers may hurt the non-incumbent party too, but in these cases it doesn’t matter, since they were likely to lose anyway.

    This is just a theory. I don’t know of any serious research into the subject. But I think it’s plausible, and it may be specious to conclude that spoiler-mitigating voting methods are inherently unfavorable to major parties.

    Now that the Greens are running at 15% in opinion polls and looking to take three Labor House of Reps seats

    AU has had IRV in the House since (I think) 1918, right? If this were to happen, it could be a great thing. But why do you think the House would remain two-party dominated all that time, and then finally break free? Does IRV just take 90 years or so to break out of duopoly? Is it related to the increased political awareness and distributed campaign donations brought about by the rise of the Internet? Surely you can understand why I’m skeptical.

  15. @Alan

    the strength of your theoretical case might improve if your empirical evidence about Australia showed some ground in reality.

    First of all, mathematical proofs aren’t any less correct when a person who discovers them or discusses them makes other claims that you disagree with. The Poincaré conjecture is proven, and doesn’t cease to be true even if Grigori Perelman blogs that the moon is made of blue cheese.

    we are still looking for the famous NatLibs

    I cited several examples where that term was used. You said one was from the UK. Okay, my mistake, but that doesn’t discount all the others.

    And besides, the term “NatLib” is irrelevant to the real issue, of whether the coalition is effectively one party. My study (including a phone call to the AU Green Party recently) has found that they are effectively one party, despite a few historical flukes.

    Again, if you want to dispute that, show some EVIDENCE. Look at the 2007 House election, and show me some of the 150 seats which exhibited competition between the coalition parties. That would be a nice start. You know, actual evidence instead of allegories about angels on pinheads.

    Regarding the proof that the social utility function is “utilitarian”, you could attempt to refute that by, say, showing an alternative one which cannot be refuted via reductio ad absurdum. I’m quite confident you won’t be able to, and thus you’ll have to stick with cute but academically meaningless allegories.

    But I could be wrong. No way of knowing since you haven’t even attempted a rebuttal.

  16. No, Clay, you cited several examples where a different phrase was used. Unlike me, you did not give a count of those terms or any other testable evidence. Continually making new claims, and altering old claims to evade intellectual responsibility, does not contribute to rational intellectual inquiry.

    Nor does trying to define the argument in your own terms. The data about the house of representatives would be interesting, but essentially meaningless without data from the state parliaments.

  17. @Tom Round,

    Do you have any evidence that failing the Majority Criterion is a flaw rather than a virtue? The mathematical case to the contrary seems cut and dried to me.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s