Monday, 7 February 2011

Preferences and weighting

It's the world cup, and like always England have been knocked out of the competition.
One England fan decides to support his back up team Brazil.
Of course if England were still in he would prefer them, but he doesn't mind Brazil now England are out.

Brazil is playing Spain in the final and in a pub our England can is sitting with a Spaniard. They are both cheering on their respective teams.

Now who do you think is likely to be most happy if their team wins?

Who cares more would you imagine?
Both teams have one supporter, but who would you imagine has most support?


  1. Er... The one that has the support of not just the England fan, but also 190 million Brazilians?

  2. You lot are very good at avoiding answering the actual questions I ask aren't you.
    I am asking out of the two people, who do you think supports their team more, that is the question. If you cannot answer that question but have to change it to another, then you two have seen the problem with your logic

  3. But your situation bares no resemblance to AV whatsoever. The question is a nonsense.

    Imagining for one moment that the results of football matches are actually decided by the ammount of support the teams have, if these two fans were the only ones that mattered Brazil would have gone out in the first round. The result of the Englsnd/Spain final would be decided randomly, exactly as under FPTP.

  4. Your mixing metaphors james. I am not asking you who will predict results.
    I am taking one aspect of AV and putting it under the spot light.
    I am asking you a very specific question and have been through this entire post.

    Who out of the two supporters is likely to support their teams more?

  5. Sigh... electing an MP is not like supporting a football team.

  6. Sigh...mangos aren't like lions either and both of those would be a problem if i was saying they were.

    @both of you again, Who out of the two supporters is likely to support their teams more? (Didn't think it would be such a difficult question tbh)

  7. SIgh... the Spaniard. Hope that made you happy.

    But most football supporters support one team and actively dislike all the other serious contenders in whichever contest it is, especially whichever team is playing against their team right now. That's very different to an election, where it's quite possible to say "I like Jones and I also like Smith because they are very similar". But you would never hear "I support Chelsea and I support Arsenal too because they are very similar".

    And there are no three-way football matches, which is what you need in order to compare FPTP and AV by using a football analogy. As you know, FPTP and AV are identical in a two-way match. But if there was a three-way FPTP football match (which there isn't), and if my cheering for a team in a pub had a direct effect on how likely a team was to win (which it isn't), and if it was inevitable that England would lose (which it is), then the England fan would be quite likely to cheer for Brazil.

    So ... what was the point you were making?

  8. That it is quite possible to say that a second preference does not have the same support as a first. With that premise destroyed, you can apply that to this decision...

    Last round of AV, shock, its a draw.
    Candidate A has 50% of the votes all first pref, candidate B has 50% of the votes. Theirs consist of 30% 1st, 13% 2nd , 7% third.

    Who is likely to have most support? Or is the support completely even like AV suggests? Course not, 1st preferences on average denote more support, so candidate A should win.

    So if preferences were weighted to reflect their average value we'd have a much more accurate way of electing single winner constituencies.

  9. This comment has been removed by the author.

  10. If it won't affect the result, how is it relevant?

    Our electoral system has never asked voters to quantify a level of support, and nor should it - it's an entirely subjective quantity. We ask people to choose their favourites from the candidates available. That is all. It's impossible to infer anything further whilst maintaining any pretence of democracy.

  11. James, you are doing that thing again. Asking questions to avoid asking very simple questions I ask you.

    "Our electoral system has never asked voters to quantify a level of support, and nor should it - it's an entirely subjective quantity"

    This is the myth that this post is aiming to undo. We do not know how much the Spaniard supports Spain, we do not know how much the English fan supports Brazil.
    However it is safe to say that on average (i.e. done over a bigger pool of input)that the Spaniard will show more support for Spain than the English supporter will show his second preference Brazil.

    With this information we can work out a fairer system than AV's "All levels of preference are worth the same".

    I do not argue that there will not be rogue results, i.e. where someone really likes their 3rd preference so shows the same amount of support as another person shows their 1st preference. However I think it is safe to say that any system that works on the assumption that this is always true is going to give you some pretty weird results.

    Therefore A=D is false.

  12. 'AV's "All levels of preference are worth the same"'

    Where on earth did that come from? My second preference will not even be registered until my first preference candidate has been eliminated. So how can you say my second preference is worth the same as my first preference? If there is a winner in the first round, then all of the second preferences of all of the voters are discarded, ignored, worthless. What makes you think my worthless, ignored, discarded preferences are worth the same as my first preference?

    Anyway, here's an FPTP election in which candidate A has 50% of the votes, all from people who genuinely think candidate A is the best candidate, and candidate B has 50% of the votes, consisting of 30% who think B is the best, 13% who would have voted for C if C had any chance of winning, and 7% who would have voted for D or C if either D or C had any chance of winning. Now which candidate has the most support? Isn't FPTP wonderful and transparent?

  13. This comment has been removed by the author.

  14. And I'm sorry, but it really isn't possible to say my second preference has less support than your first preference. That information just isn't available. Sorry, it just isn't! In the second and subsequent rounds, there will be some second and lower preferences counted. There's no way of knowing whether those lower preferences represent less support than the remaining first preferences. True, on average the support represented by first preferences will be higher than that represented by second preferences, but that's irrelevant, because only some second preferences are counted. (Second preferences are worth less than first preferences, as you hopefully know by now!)

  15. Anthony, how many votes is a first preference worth? One . How many votes is a second preference worth. One .

    You argue that it is impossible for me to know that your second pref has less support than my first. In individual cases this is true. (though even you conceed that on average second preferences indicate less support than first).

    However where is your evidence that your second preference IS worth the same? That also doesn't exists (and isn't as likely as you have admitted) so why are you still supporting a system that is more often than not, wrong about the level of support for its preferences?

    In your example, A has the most support i'd say and under FPTP, unless your suggesting that people never vote for more than two parties (?!) if 50% of the electorate believe one candidate to be the best they will win, because although 50% don't agree with candidate A, they don't agree with eachother either.
    Howver under AV this can be a draw.

  16. "X" doesn't exactly communicate much about the level of support as far as I can tell.

    You seem to be advocating some kind of score/range voting method. Is that right?

  17. Yep, that's correct.
    X doesn't show how much support they get. But we can assume that they are the candidate they want their vote to go to out of all the people up for election. Which is the same as everyone else. Once you start taking candidates away, this is no longer true and the support that each candidate has begins to vary more and more as the rounds go on.

  18. So in neither FPTP nor AV is the number of votes equivalent to the level of support. But at least AV allows you to express your opinions in a bit more detail.

  19. Grr... For some reason the comment posting widget here seems immensly unreliable. Tried to write this yesterday, but it wouldn't let me log in:

    You seem to fundamentally misunderstand the point of democratic elections. They do not measure a level of support, and nor should they. I would have liked to have voted Pirate in 2010, but they weren't standing in Bradford West. Should my vote only have counted for 75% because the party I voted for was my second choice?

    You seem to be saying that the team who's supporters shout the loudest should win. That's not the way the game works. Elections aim to establish which candidate from a list is preferred by the largest share of the voters. AV does that better than FPTP.

  20. Antony, almost, both Fptp and AV are as bad as each other in the FIRST round. However, when you take away a candidate n redistribute, this makes it less representative...A-side worse and worse as the rounds go on.

    AV allows SOME to show their full level of support only.

  21. I disagree, I believe quality of support is just important as quantity of supporters ( especially if these ' supporters' only support you because all their choices they support more than you are already eliminated). I think in an ideal world, you vote should reflect how much you support the candidate otherwise donkey voting decides the winner. I haven't decided on the best weighting method, though after some research I'm sure we could work something out.

  22. In that case, I suggest you go and get a research grant from somewhere, do some serious investigation, get some evidence, present your findings to people who care about electoral reform, win them to your side and start a campaign to have your "improvements" implemented.

    In the meantime, I suggest you vote yes to making the modest improvement we are actually being offered.

  23. I am carrying out small scale research already, AV on the other hand gives preferences of different level same value so cannot be a step forward

  24. OK, so explain why you think it was fine for the second preferences of Ken Clarke's voters in the second round of the Tory leadership election to count the same as in the first?

  25. I don't think they should...but it is more complicated than that.

    As there is in essence a fresh campaign in between each round of normal run off voting, each candidate can re-interate their case to voters that perhaps were zoning out when they first listened to them because they knew who they were voting for already.

    If they realised they were wrong and they really supported the new candidate then I guess they should count the same...though i doubt this is the case and there would just be a compromise.

    Which is why i don't like rounds. I do however like preferential voting, Maybe a Borda count would have been better, or a cordorcet maybe.

    Either way I am glad to say that the eventual winner would have been picked if they had used a multi-winner fptp system (or plurality at large as its called) to select the candidates to put to the postal vote anyway. Go DC!