Friday, 29 April 2011

AV Oddity Part 2

Let's say that we have three candidates
10,000 people turn out for the vote

A has 34% of the Vote
B has 33%+1 of the vote
C has 33-1 of the vote..

All C second preference A.

A wins with 6699 votes 3400 first and 3299 second preferences. =67%
A Majority of 1699 votes

Let's re-run that election but this time two people change their mind from B to C

A has 34% of the Vote
B has 33%-1 of the vote
C has 33+1 of the vote..

All of B second preference C

C Wins on 6600 votes 3301 first and 3299 second preferences. =66%
A Majority of 1100 votes.

So we go from a majority for A to a majority for C from just two votes...thats a 2799 vote swing!

AV is not logical, is "disturbingly unpredictable" and "unacceptably unfair"

Due to it's elimination process, it turns the competition from a 'winner takes all' to a 'loser takes all' where the will of whoever loses being the deciding factor

NB. this example is not likely to happen on this scale, but is only a magnification of the unfairness and inequality that exists in every AV election

Part two ...

A has 34% of the Vote
B has 33%-1 of the vote
C has 33+1 of the vote..

Perhaps, lets say in the second example, after the two voters have changed their mind, that only 1%of the total vote (100 votes) belonging to B transfers to C and the rest are exhausted..

We now have A on 3400 votes and C on 3401

C wins the election on 34%+1 of the total vote (or 50%+1 of final round votes)

In this example just two votes had a swing of 3299 votes..

and because of those two votes instead of getting a winner that had 66% of the total vote...who happens to be the FPTP winner (don't ever let someone tell you AV is only better than FPTP) we have an AV winner that is on 34% of the total vote...about half!

Vote No to AV on May 5th


  1. D,

    All you are showing is that you can't talk about a "swing" under AV in the same was as you talk about a "swing" under FPTP. You have not shown that the outcome of the election is unfair. Nor have you shown that AV is unpredictable: your example is extremely marginal, which means that just one or two changes in people's votes could change the outcome.

    And it's very contrived to have all of the C voters putting A as second preference and all of the B voters putting C as second preference.

  2. Yes...that is my point entirely! In the first one it is marginal...yet is counted as a huge 67/33% win..though it isn't a huge win.

    The second one it is classed as a huge win, when really by approval voting and other more comprehensive voting systems, it would have been a huge loss..

    Changing two votes shouldn't change one majority to another!

    As for the likelihood I did concede it was an extreme example, as this type of thing normally goes on at micro level inside the AV election system, I have blown it up and reduced the number of candidates to show it working at a macro level so everyone can see.

    The principles however are exactly the same. Overall Support or approval isn't what is used to decided AV elections...just the specific order that support is looked at

  3. There are only three parties in this example - which isn't realistic.

    Another thing that isn't realistic is that the 2nd preferences would ALL go to one party. The example has been set up to give almost equal 1st pref votes (unrealistic), but then highly unequal 2nd pref votes. This is perhaps the definition of contrived.

    If you run it again and the 2nd pref votes are split equally among the remaining parties, you'll see that A wins both times with a majority of 49-50 votes - which is fair.

  4. Pitch Twit, I have already said that it was a blown up example, but like I have said, the problems exist under every election with av to an extent