Thursday, March 31, 2011

Radio 5's AV attempt

Via Guido Faukes just to show once again that without access to the raw data you can say whatever you like - AV is too complicated and it didn't produce a majority. Or the BBC screwed up.

We don't know because the raw data is not yet available to us. To illustrate the problem let's take the result from his blog regarding the Yay! FPTP system. With an electorate of 200 the Conservative Party polled 22.7% of the vote. Numerically that means they received 45.4 votes. So we have 0.4 of a person out there? Well no obviously this means that either some of the ballots were spoilt or not all of the electorate voted.


Without that information we can't point at either system and say that one is better than the other. Without it we get the following mess:

I'm using the full percentages which means that of the 2.6% of the BNP vote. 2% second placed UKIP and 0.6% only voted for one party; rather than saying of the 2.6% of the BNP vote 77% second placed UKIP and 23% only used one vote.

Of the combined 5.3% UKIP and BNP vote. 2.6% went to the Conservatives, 1.9% to the Greens, and 0.7% to Labour. With 0.1% not using their second or third preference vote (the 0.7% is a running total on the graph).

[Out of interest is anyone else surprised at the high preference vote for the Green party? This is a side to BNP UKIP we wouldn't see except for AV]

Of the combined 15.2% UKIP, BNP, and LIbDem vote. 3.3% went to the Conservatives, 6% to the Greens, and 5.4% to Labour, with 0.6% no votes.

Finally of the combined 27.8% UKIP, BNP, LibDem and Conservative vote 11.9% went to the Greens and 7.9% to Labour, with 8% no votes.

What's the second, third and fourth preference splits between those combined values? No idea.

As an exercise to show AV in practice this isn't great; as an attempt to compare the two systems it's worse. Not the BBC's finest hour (or two).

[Update - As per Andy's comment they've made a right pig's ear of this. Assuming they've actually assigned the figures correctly, and again without the raw figures we can't tell, here's how the results should have been tallied:


PartyAV1AV2AV3AV4AV5
Labour35.10%35.31%36.05%41.67%54.02%
Green21.90%22.03%23.97%30.18%45.98%
Conservative21.90%22.03%24.67%28.15%
Liberal Democrat15.20%15.29%15.31%
UKIP3.30%5.33%


BNP2.60%




Note how the figures now add to 100% for each column and produce a greater than 50% result]

2 comments:

Andy said...

They did the count completely wrong. Everyone votes for their first choice. The candidate with the lowest votes is eliminated and the second choice of these voters is given to the rest of the candidates. The candidate with the lowest amount of votes in this round is then eliminated. The second choice of the voters who voted this candidate as their first choice is given to other voters. The voters whos votes were given to this candidate from the previous round are then recounted for their third choice voter. However they clearly took all voters whos candidate was eliminated in the first round and then gave their third choice votes to the other candidates and effectively removed the votes from another candidate(s). This is evident in that the percentage of votes for parties not eliminated reduced in some rounds. Which means their votes were removed. A candidate percentage of the vote can never decrease unless they are eliminated.

FlipC said...

The AV vote result starts about 1:08 on the broadcast.

Having listened to it they've definitely messed up the percentages at the very least.

In an FPTP count any spoilt votes are removed so the percentage given is that of the votes and not the ballot. This should hold true for AV that means the total in each round of the candidates votes should equal 100%.

I've added an adjusted graph.