Having been on the fence, either side of the fence, and back on the fence again, I’m convinced that most people are making the AV decision using reasoning that’s irrelevant and dangerously nonsensical.  All that matters, in any voting system, is what the effect is on who wins.  Simple as that.

So the main arguments from the YES campaign are all terrible:

1. MPs would need to get 50%: Not true.  They need 50% of the ACTIVE votes, which, by the time you’ve eliminated all but one candidate, means all the votes that were for them.  If all the candidates you put a number against were eliminated, your vote is no longer active.  It’s not therefore very hard to get 50% of all the votes that were for you.
2. The end of tactical voting: Partly true.  AV does seem to present a partial solution to vote splits.  More on this in a second.
3. It’s tried and tested: Laughable.  Nothing’s more tried and tested than the system we already have.
4. Shuts out extremist parties: Not true, and the reality is probably the opposite, assuming you replace ‘extremist’ with ‘minor’ which is the same thing without the bias.

The main arguments of the NO campaign are equally poor:

1. It works, don’t fix it: Right, moving on.
2. It will cost a lot to implement: These arguments aren’t getting any better are they.
3. It’s complex: True, but misses the point.  It’s simple for the end user.  THe complexity is in understanding the impact of the way the vote is distributed, which I agree that only a minority of people will understand.
4. It will produce more hung parliaments: True but irrelevant.  If the will of the electorate is a hung parliament, then so be it.  And research has shown that the Labour landslide in 1997 would have been even bigger under AV.

So the arguments are mostly nonsense.  What matters is how the numbers stack up when the votes are counted.  Here are a few scenarios:

1. Split vote: A majority of the electorate support an ideology shared by several candidates who each offer it slightly different flavours.  The competing ideology fields only one candidate.  Under FPTP the minority ideology would probably win because the rest of the vote would be split between the similar candidates.  Under AV, this is fixed by counting second and third preferences.  AV wins.
2. Tight multi-way marginal: All candidates polling roughly equal amounts of the vote.  Under FPTP, there’s a simple winner, though they win with a minority of the vote.  Under AV, the winner depends hugely upon the ORDER in which the candidates are ranked after the first preferences are counted.  Say you have A, B, C and D, on 27, 26, 24 and 23 respectively.  D is eliminated and all his supporters back A as their second pref.  A wins.  Now imagine the votes for A and B were exactly the same, but C and D were reversed – C got 23 and D got 24. C is now the first to be eliminated and all his supporters back B as their second preference.  A remains on 27, B is now on 49, and C on 24.  Not an outright win, but B needs only one out of C’s 24% to go over the line.  So despite the result of the most popular candidates remaining the same, small changes in the position of the minor candidates, and therefore the order in which they are eliminated, causes the overall result to change.  This seems dangerously unpredictable and far too nuanced to be properly understood by the average voter.  It becomes even worse when you consider that it’s possible, though unlikely, that a candidate that wins based on the AV formula could have been in line to *lose* if *more* people had voted for them.  This incomprehensible concept is called non-monotonicity and in AV is caused by the changing order of the eliminations.  Frankly if this doesn’t send you running scared back to FPTP I don’t know what will.
3. Clear majority: One candidate enjoys a significant amount more support than any of the others.  In this case, any voting system that doesn’t elect that candidate is simply not democratic.  So frankly it matters not one iota whether you use FPTP or AV, it won’t change the result.

In the end, we can only choose between these two systems, not any other system that might be considerably fairer, and based on the above reasoning, AV seems to me a bit like keeping a pet lion – it might be impeccably trained and curl up and purr on cue, but none of that alters the fact that if it wanted to, it could eat you.