Tuesday 22 March 2011

No2WS

From Mark Reckons via Neil Harding.

31 comments:

Bayard said...

Of course, there were no nuclear bombs before women got the vote.

Deniro said...

The counting system used in AV does not work. It incorrectly gives special significance to the second choice votes of voters who vote for lower ranking first vote candidates. The second choice votes of voters who vote for higher ranking first choice candidates are just as significant in assertaining which candidate has a majority endorsement. Therefore it declares a winner incorrectly. The AV count system simply does not work.
An alternative use of first and lower preference votes could be to produce a total score for each candidate by summating their respective votes , weighted by preference, and maybe giving extra weight to first preference votes.

Deniro said...

More than one or even all of the candidates can receive majority endorsment if all the second preference votes are counted. AV incorrectly picks one by incorrectly not counting some of the second preference votes.

Scott Wright said...

"The counting system used in AV does not work. It incorrectly gives special significance to the second choice votes of voters who vote for lower ranking first vote candidates. The second choice votes of voters who vote for higher ranking first choice candidates are just as significant in ascertaining which candidate has a majority endorsement. Therefore it declares a winner incorrectly. The AV count system simply does not work.
An alternative use of first and lower preference votes could be to produce a total score for each candidate by summating their respective votes , weighted by preference, and maybe giving extra weight to first preference votes.

More than one or even all of the candidates can receive majority endorsement if all the second preference votes are counted. AV incorrectly picks one by incorrectly not counting some of the second preference votes."


Well although I fully agree with you, I am inclined towards a yes vote for AV just because if the system is changed & the country continues to run largely unscathed. The non-argument that a different voting system is a bad thing due to "perpetual coalition" will be quashed and people can continue to campaign for an actual proportional system. The constituency link is clearly nonsense when we have people like Nick Clegg the nancy southerner in a Sheffield seat.

Mark Wadsworth said...

B, indeed. Or Islamic extremism. Funny that.

Den: "The counting system used in AV does not work. It incorrectly gives special significance to the second choice votes of voters who vote for lower ranking first vote candidates."

It does not give special significance to second pref! That is a myth!

Imagine somebody's popping to the corner shop and they ask you if you want anything. You tell him "I'd like a bottle of coke, but if they're sold out, bring me an orange juice".

He comes back with an orange juice because coke was sold out. Had your second prefence been 'special significance'?

Or to paraphrase what you said:

"The counting system used in FPTP does not work. It incorrectly gives special significance to the second choices of voters who did not dare vote for their preferred candidate."

Deniro said...

The crucial point is that more than one or even all of the candidates can receive majority endorsment if all the second preference votes are counted. AV incorrectly picks one by incorrectly not counting some of the second preference votes.

Mark Wadsworth said...

Den, FPTP supporters come up with some contorted arguments, but that is beyong sanity. The whole point of AV is that your second vote is only counted if your first is discarded. Only one of your votes counts!

To use the shop example again, do you expect your friend to return from the shop with a coke AND an orange juice? Nope, you expect one or the other.

Deniro said...

Eg

a 45 second c
b 35 second a
c 20 second b

b wins but a has more first votes than b and more second votes than b

Mark Wadsworth said...

Den, you are using highly contrived figures.

Under FPTP, the chances are that the bulk of people who would give C first vote and B second vote under AV, would have given their ONLY vote to B and he would have won anyway. So why is it OK for him to win under FPTP but not under AV?

Deniro said...

You have just defended a vote system on the basis that you know the outcome beforehand so it doesn't matter what system is used anyway. That doesn't make sense. You say my figures are contrived? But my figures are objective and not contrived . Thats the point. Contrived figures have to be used to make AV work, as you have just done.

another example a similar set of votes showing the wrong mid placed candidate winning.

a 28 second pref b
b 20 second pref b
c 19 second pref b
d 17 second pref c
e 16 second pref c

c wins but b has more first and second prefs

AV can never work (accept by chance) because it is based on the addition of lower pref votes but doesn't count all the votes cast.

An alternative use of first and lower preference votes could be to produce a total score for each candidate by summating their respective votes , weighted by preference, and maybe giving extra weight to first preference votes.

Mark Wadsworth said...

Den: "My figures are objective and not contrived . That's the point. Contrived figures have to be used to make AV work, as you have just done."

??? Which figures have I contrived? It appears to me that you would only be happy with numbers where the result under AV would be the same as the result under FPTP (which will be the case in nine out of ten constituencies anyway).

You've now given an example where 20 people vote "B first, B second" so you don't even appear to understand AV at all!

Deniro said...

Corrected

a 28 second pref b
b 20 second pref d
c 19 second pref b
d 17 second pref c
e 16 second pref c

This is an example where the wrong mid placed candidate wins so I am not contriving figures to make the FPTP candidate the winner. Under FTPB A would win. Under the stated aim of AV to "select a winner who has majority endorsement" b should win. But because AV does not work C wins by chance.
AV does not work on its own terms. I'm not saying I don't like it because the FPTP candidate doesn't win , I'm saying it mathematically does not acheive its stated aim.

Mark Wadsworth said...

Den, that's still contrived, and we end up with C winning with 19% of first votes, instead of B winning with 20% of first votes.

Big deal.

You seem to think that B 'should' have won this one, the FPTPers will say that A 'should' have won and the AVers will say that he did win.

The FPTPers can shut up as well, because 28% of first votes and none of the second votes is hardly a resounding endorsement, is it?

Deniro said...

what do you mean by

"the FPTPers will say that A 'should' have won and the AVers will say that he did win."

Mark Wadsworth said...

DEN, my bad. The AVers will say that C did win. Which he did.

Deniro said...

but b has more first votes and more second votes than c

Mark Wadsworth said...

Den: "But b has more first votes and more second votes than c."

That's not true either, is it?

B has 1% more first votes than C (that much is true), but he has 28% of the second votes as against 33% of second votes for C.

Why you think it reasonable or fair for the 19% of people who preferred C to B to count as second votes for B as well as first votes for C is a mystery to me. We're trying to stick with 'one man one vote'.

In the event of a tie, clearly these 19% of the votes belong to C and not to B.

Deniro said...

this is the result

a first votes 28 second 0
b first votes 20 second 47
c first votes 19 second 33
d first votes 17 second 20
e first votes 16 second 0

evidently b has the majority approval.

"In the event of a tie, clearly these 19% of the votes belong to C and not to B."


There was no "tie"

Redistributing the 33 from c to b is no different from redistributing the e and d to c

Mark Wadsworth said...

Den, sorry, nope. It's 'one man one vote', remember? You can't count the 19% first votes for C as second votes for B as long as C is still in the ring.

D and E are eliminated and their votes redistributed. So after two elimination rounds, the table actually looks like this:

a first votes 28
b first votes 20
c first votes 19 + second votes 33 = 52

If you like, we can then eliminate Mr B and redistribute his votes to D, who is already out of the running, so we end up with this:

a = 28 votes
c = 52 votes

Deniro said...

Thats the problem , the counting stops after 2 rounds and a's and c's second votes never get counted. The outcome depends on the order in which the candidates are knocked out not the actual number of votes. Which is incorrect.

Mark Wadsworth said...

D, using your contrived figures, in which order would you eliminate candidates? You seem determined to prove that C 'shouldn't' win, so having eliminated D and E, whom would you eliminate next, and why?

A, B or C?

Remember: AV is much the same as the system used to elect the French President, the leader and deputy leader of Labour party and so on.

So any logical approach says that B would be eliminated next, his second votes go to D, who is already OUT!

If you like, you can eliminate A next and allocate his second votes to B, this still leads to:

C: 52 votes.
B: 48 votes.

Deniro said...

"D, using your contrived figures, in which order would you eliminate candidates? "

I am pointing out that such an "elimination" approach does not work.

(In the French Presidential election that you mention there is a two round system So all the second votes are counted.)

Mark Wadsworth said...

D, why doesn't an elimiation process work? (apart from being very long winded).

As many candidates can enter as they like, it's one-man-one-vote and in each subsequent round the lowest scoring candidate is eliminated. There's no concept of 'second votes' - in each round every voter gets one vote.

AV is just a quicker and easier way of doing it.

Mark Wadsworth said...

Den, in any event, it seems highly unlikely to me that Mr A would get such a high share of first votes but not a single second vote. That is beyond the bounds of statistical plausibility.

Deniro said...

"it seems highly unlikely to me that Mr A would get such a high share of first votes but not a single second vote. That is beyond the bounds of statistical plausibility."

well the second votes for d could have been cast for a . Still the same problem.


"in each round every voter gets one vote."

Not true . Voters who's second vote is for a knocked out candidate don't get a vote, unlike the French system.

See with a set of votes yourself how the workings different from the French system

The biggest (huge) error of AV is that it is not mathematically coherent.
As you can see from 100 voters there is more than 100 votes but AV stops counting a 51 mistaking it for 51%.

By the way you critisised my three candidate example by saying that in FPTP the c vote would have gone to the b candidate. actually small party supporters of AV beleive it will swell their first votes so you might just as well say that the b votes would have gone to the candidate occupiying the position where the c candidate is.

Anyway good blog

The Daily Mash seems on form again
http://www.thedailymash.co.uk/politics/politics-headlines/osborne-hands-average-family-%C2%A345-to-give-to-british-gas-201103233652/

Mark Wadsworth said...

Den: "Not true. [Under AV] voters whose second vote is for a knocked out candidate don't get a vote, unlike the French system."

Not true.

Under French system, you have as many opportunities to vote as there are candidates.

Under AV, you have as many opportunities to vote as there are candidates.

It may well be that under AV, if your first and second choice candidates are eliminated in the early rounds, that neither your first nor second votes count - THIS IS WHY YOU GET A THIRD AND FOURTH VOTE ETC.

If you don't want to use them, that is absolutely fine, in the same as some people might e.g. have voted for Le Pen in the first round but not bothered voting for Sarko or Sego in the second round.

If Le Pen supporters were to claim that they had no opportunity to vote in the second round, then they are clearly liars, their candidate got booted out, end of discussion.

Mark Wadsworth said...

Den: "Well the second votes for d could have been cast for a. Still the same problem."

In that case, A would have won. If that's how the votes were cast then he's the winner, fine by me.

Deniro said...

No c would still win.


Explain how a would win (to check you understand the av process)

Mark Wadsworth said...

Den

We count first votes
a 28
b 20
c 19
d 17
e 16

E is eliminated in first round and 16 second votes go to C.

a 28
b 20
c 35
d 17

D is eliminated in second round and his second votes go to A (using your amended figures from earlier today)

a 45
b 20
c 35

B is eliminated, his second votes were for D and don't need to be counted.

Final result:

a 45
b 35

A wins.

Deniro said...

Oh I see .

Actually the ammendment I was defining was changing the "second votes for d" to "second votes to a" on the b ballots.

Mark Wadsworth said...

Den, in that case C wins, as he ends up with 52 votes. You could allocate 20 second votes to A but that still only gets him 48.

52 > 48, end of discussion.

Look, of course there will be the odd funny result under AV, just as there is under FPTP, MMC, PR or anything else, but they tend to even out.

For example, in the 2009 Euro's, UKIP missed the last MEP place in the North East by a few hundred votes, but we scraped the fourth MEP place in Wales by a few hundred votes, so fair enough.

Across hundreds of constituencies (or a smaller number of MMC's) this will all even out.