I'm not obsessed with it, honest. It's just that people keep talking about it, and I keep running across these bad arguments. Brian Barder has one. He starts out by dismissing one fallacious piece of reasoning against AV, but goes straight on to substitute another. Here is what Brian says:
There are cogent arguments against AV, but the main one being used by the luminaries of the No campaign, Lord [formerly John] Reid, William Hague, Margaret Beckett and Sir Malcolm Rifkind, among others, isn't one of them. Their claim that under AV some voters will have more votes than others is nonsense. All the valid votes are counted again at every recount. Those giving their first preferences to the two candidates who come first and second, and who are therefore never eliminated from the next recount, don't get their second and lower preferences redistributed and counted, but that's not a disadvantage: their first preferences continue to count right to the last round.
The No to AV campaign needs to focus on the real objection to AV, i.e. the fallacy in the only serious claim for it, that AV, unlike First Past the Post, ensures that all MPs have the support of the majority of their voters. But this is simply not so. An MP whose majority depends on votes transferred from other candidates eliminated in early counts no more has the support of a majority of voters than an MP elected on a minority vote under FPTP: in both cases, a majority of those voters preferred and voted for someone else.
Brian can only arrive at this conclusion by treating the first round as the real, or more real, contest and the other rounds as somehow less real or, at any rate, of lesser validity. But that isn't so. As I explained in this earlier post, AV functions by conducting all the possible rounds in one go. When you rank A, B and C in that order, you are voting for A in the three-way contest, for A as between either A and B or A and C, and for B as between B and C. (The same sort of thing can be constructed for constituencies with four, five, six etc candidates, but I've left it at three for the sake of simplicity and economy.) Voters don't know, to begin with, which of those four contests will be the decisive one; but, if they understand the method, then they do know that any of the four contests may be the decisive one and also how their vote will count in each case. Let's suppose that candidate A is eliminated on the first round, and B wins on the second, with the redistributed second preferences from A's ballots. Then B has a majority of the voters behind her - in that contest. To say that she didn't have a majority in the round involving A, B and C is neither here nor there. That contest is not privileged with respect to the other three, A vs B, A vs C, and B vs C. If we're using AV, that's what we're doing. We're not then bound by the ghost of FPTP.
Imagine that A, rather than being eliminated after the first count, had choked on a chicken bone and died two days before the election. Imagine that he'd just failed to get his nomination papers (or whatever) in in time. Would B not have a majority? She would. With some other array of candidates she maybe wouldn't. But no particular array is normatively more real and to be used as a benchmark against which to assess the others. If the Troony Party had put up a candidate everything again might have been different. Having a majority is a contest-relative achievement and when there's more than one contest taking place simultaneously, all of them are equally valid. Perhaps this is a use for the multiverse.
