Surpluses are treated differently from eliminated candidates in the election

[Brendan Burgess]

In the example shown in post #89 there are:
1,000 votes which express a first preference for A and no second preference (Lets call them Group 1)
1,000 votes which express a first preference for A and a second preference for B (Lets call them Group 2)

I see nothing wrong with allocating Group 1 to A, and transferring all of Group 2 to B, which is the effect that the Irish system has. In fact this is entirely appropriate. Every one of the 2,000 votes is effective. No votes are discarded.

Hi Freelance

I am trying to understand your point here.

"I see nothing wrong with allocating Group 1 (A0) to A, and transferring all of Group 2 (AB) to B,"

By the same token, you should see nothing wrong with allocating all of the ABs to A and then having 1,000 non-transferables.

2,000 people voted for A.
A needed only half of their vote, so what do we do with the other half?
For the ABs we give the 1000 @ .5 to B.

For the A0s , we do nothing with them as they did not specify a preference.

Brendan

 

[Brendan Burgess]

I wonder would it help to put party affiliations on A,B and C


I am a FG voter. I will transfer to FF because I want to keep Sinn Féin out at all costs.
You are a FG voter as well but you cannot bring yourself to vote for anyone else.

Under the Irish system which you agree with, all of your 1,000 votes go to FG and elect him.
But then I get to vote again for FF. That is clearly wrong.

50% of my vote should be used to elect FG and 50% to elect FF.
50% of your vote should go to FG and you expressed no preference with the other 50%.

Brendan

 

[Duke of Marmalade]

I wonder would it help to put party affiliations on A,B and C

View attachment 8957
I am a FG voter. I will transfer to FF because I want to keep Sinn Féin out at all costs.
You are a FG voter as well but you cannot bring yourself to vote for anyone else.

Under the Irish system which you agree with, all of your 1,000 votes go to FG and elect him.
But then I get to vote again for FF. That is clearly wrong.

50% of my vote should be used to elect FG and 50% to elect FF.
50% of your vote should go to FG and you expressed no preference with the other 50%.

Brendan

Or put another way, the Irish system deems that you didn't vote for FG at all as there were enough non-transferables to get A elected. Doesn't make sense.

 

[Brendan Burgess]

Or put another way, the Irish system deems that you didn't vote for FG at all as there were enough non-transferables to get A elected. Doesn't make sense.

That is brilliant!

 

[Freelance]

50% of my vote should be used to elect FG and 50% to elect FF.
50% of your vote should go to FG and you expressed no preference with the other 50%.

What you appear to be advocating is moving towards Cumulative Voting, which is also proportional but a very different flavour. Under our STV system, a vote is whole and indivisible. You cannot give 50% of your vote for one candidate and 50% for a second candidate. Your vote rests with one candidate or the other.

View attachment 8953

The 4,000 shown here is mathematical nonsense. The correct representation of this is:

First Count​	Second Count Transfer​	Second Count Total​
A	2,000​	-1,000​	1,000​
B	100​	1,000​	1,100​
C	900​	0​	900​
Non Transferable	0​	0​	0​
Total Votes	3,000​	0​	3,000​

Hi Freelance

I am trying to understand your point here.

"I see nothing wrong with allocating Group 1 (A0) to A, and transferring all of Group 2 (AB) to B,"

By the same token, you should see nothing wrong with allocating all of the ABs to A and then having 1,000 non-transferables.

The objective of the STV is to achieve proportional representation. One facet is to ensure that a vote does not die if it contains later preferences. It makes no sense to allocate votes with a later preference to a candidate if the effect of this is to cause other votes where there is no later preference to die. So it is perfectly reasonable to allocate the votes with no further preference first, and then those with later preferences. The impact of this in this case is to ensure that all votes in the A's surplus have a later preference and so are transferred.

The final outcome achieved is is proportionally representative:

1,000 voted for A (only) and thus he achieved the quota and was elected
1,100 voted for B (100 voted their first preference for B and 1,000 voted for A but, in the event that he was not available to accept their vote, directed their votes to B) and thus B achieved the quota and was elected
900 voted for C who did not achieve the quota and thus was not elected

And the voters wishes are satisfied as follows:
The 1,000 voted for A (only) are satisfied by the election of A
The 100 who voted for B and the 1,000 who voted for B in the event that A was unavailable to accept their votes are satisfied by the election of B
The wishes of those who voted for C are not satisfied because the two other candidates achieved the quota

 

[Duke of Marmalade]

Under our STV system, a vote is whole and indivisible. You cannot give 50% of your vote for one candidate and 50% for a second candidate.

You are being misled by the physicals of our approach which does involve ballot papers being transferred in whole. But a surplus by definition is a fraction of the votes that elected the candidate and it is this fraction which is transferred. Under the Scottish and NI systems this process actually is carried out by carrying forward fractions. In our system we randomly select that fraction (in whole numbers) from the parcels set aside for the next prefs. In our Seanad election a ballot paper is deemed to be 100 votes and the fraction that is carried forward is the number of those votes that represent that fraction. The issue is what is that fraction? See below.

The 4,000 shown here is mathematical nonsense.

This is purely a presentational issue not affecting the argument. The 4,000 threw me at first but Brendan is effectively seeing it as 2 elections. There were 3,000 votes in the first election and 1,000 in the second. You are tracking ballot papers and of course that number stays at 3,000. As I say, irrelevant to the point at discussion but yes misleading.

The objective of the STV is to achieve proportional representation. One facet is to ensure that a vote does not die if it contains later preferences.

Votes with an outstanding preference do not in theory die under any of the systems until all the places have been filled. The Scottish/NI/Seanad elections make this clear as all the transferable votes do survive, but only in fractional form as they have already elected someone. It is less clear in the Irish Local/Dail/European Elections as, whilst in theory, every transferable vote is entitled to survive, albeit at a fraction, instead a fraction of these transferables is selected by an essentially random process to survive in full and the rest are allowed to die even though they were transferable. At the first count this is entirely fair to the second preference candidates but it has a random bearing on the subsequent preferences. Surpluses transferred at later counts are more random as they are chosen from the last sub-parcel that created the surplus.

So I hope I have convinced you that only a fraction of the votes that gave rise to a surplus are transferred. The example is a bit unfortunate in that respect as it gives the impression that there are sufficient whole number of transfers to meet surplus.
So what is that fraction? We have:
Fraction = Surplus/(some) Denominator.
The surplus includes non transferable votes under any system. Under the Scottish system the Denominator is all the votes that elected the candidate. Under the Irish system it is only those which are transferable. It is inconsistent with the numerator and can give rise to big gearing effects, which effectively give the surviving ballot papers the benefit of more than 1 vote.

 

[Brendan Burgess]

1,000 voted for A (only) and thus he achieved the quota and was elected
1,100 voted for B (100 voted their first preference for B and 1,000 voted for A but, in the event that he was not available to accept their vote, directed their votes to B) and thus B achieved the quota and was elected
900 voted for C who did not achieve the quota and thus was not elected

I just can't agree with this.

Why not say that 1,000 voted for AB and their votes were used to elect A.

1,000 Voted for A and did not transfer, so they go nowhere.

You are arbitrarily choosing the 1,000 votes which did not transfer and assigning them to A, thus leaving the 1,000 AB votes available to elect B.

You don't like fractions. OK, so let's take 500 of AB and use them to elect A and then assign the other 500AB to B in the 2nd round.

Use 500 of the A only to A and the other 500 don't transfer.

Whatever way you look at it, if you respect the voters' wishes, you should give only 500 to B.

The Duke had a good way of looking at it. Who "owns" the votes? It's not A. It's not Fine Gael. It's the voters. So respect their wishes.

 

[Freelance]

So respect their wishes.

Their wishes were respected in that 2,100 voters got a representative outcome that appealed to them

1,000 voted for A (only) and thus he achieved the quota and was elected
1,100 voted for B (100 voted their first preference for B and 1,000 voted for A but, in the event that he was not available to accept their vote, directed their votes to B) and thus B achieved the quota and was elected

900 didn't get what they wished for

Your interference with the system would reduce this number to 1,900 votes achieving the voters desired outcome. A with 1,000 votes and C with 900 votes.

 

[Brendan Burgess]

1,100 voted for B (100 voted their first preference for B and 1,000 voted for A but, in the event that he was not available to accept their vote, directed their votes to B) and thus B achieved the quota and was elected

But do you not see that the opposite is just as true?

1000 voted for A with a second preference for B.
Therefore the 1000 who voted for A with no second preference went undistributed?

Again, you should not choose arbitrarily where the votes go. That should be left to the voters.

If 2000 vote for A, then half of each vote, or group of you don't like fractions, should go to A. The rest should be allocated according to their wishes.

Half should go to B and half should remain unallocated.

Brendan

 

[Duke of Marmalade]

2,000 voted for A. Only half their votes are needed.
Only half your vote was needed, you still have half a vote in hand.
Where do you want your surplus half to go?
I don’t want it to go anywhere.
Ok we’ll give your half vote to somebody who does want your half vote to go somewhere.
Did you not hear me? I don’t want it to go anywhere.
Doesn’t make sense.

Someone is eliminated.
Look your preference has been eliminated but you still have your vote, where do want it to go?
Nowhere.
Ok we’ll give it to someone who does want to use it.

 

[Brendan Burgess]

OK, here is another illustration of why the Irish system is wrong and not just different.

Here is the situation where half of those who vote for A, give their second preference to B. This is the result:



Now consider a situation where everyone who votes for A, gives their second preference to B.



We get the same result! B gets 1,000 transfers under both circumstances.

The Irish system does not differentiate between the two scenarios.

But the Scottish system does. If B gets 50% of A's 2nd preferences, they get 50% of the surplus. If B gets 100% of A's 2nd preferences, they get 100% of the surplus.

Brendan

 

[Duke of Marmalade]

OK, here is another illustration of why the Irish system is wrong and not just different.

Here is the situation where half of those who vote for A, give their second preference to B. This is the result:

View attachment 8969

Now consider a situation where everyone who votes for A, gives their second preference to B.

View attachment 8970

We get the same result! B gets 1,000 transfers under both circumstances.

The Irish system does not differentiate between the two scenarios.

But the Scottish system does. If B gets 50% of A's 2nd preferences, they get 50% of the surplus. If B gets 100% of A's 2nd preferences, they get 100% of the surplus.

Brendan

Powerful argument Boss. Seems to have convinced the doubters.

 

[losttheplot]

OK, here is another illustration of why the Irish system is wrong and not just different.

Here is the situation where half of those who vote for A, give their second preference to B. This is the result:

View attachment 8969

Now consider a situation where everyone who votes for A, gives their second preference to B.

View attachment 8970

We get the same result! B gets 1,000 transfers under both circumstances.

The Irish system does not differentiate between the two scenarios.

But the Scottish system does. If B gets 50% of A's 2nd preferences, they get 50% of the surplus. If B gets 100% of A's 2nd preferences, they get 100% of the surplus.

Brendan

That looks like the votes for Cyprian Brady.
"At the 2007 general election, held under Ireland's single transferable vote system, Brady polled just 939 first preference votes (2.7%) in the first count. However he was elected on the fourth count, due in large part to a pre-arranged electoral strategy seeing large transfers from his running mate, the Taoiseach Bertie Ahern,[8] and became only the second ever TD (after Brian O'Higgins) to be elected with less than 1,000 first-preference votes." - Wikipedia

Someone told me to always continue the preference all the way down the ballot, even if you wouldn't normally vote for the other candidates. I never understood why, as I assumed a ballot with no preference wouldn't be transferred. Now I see why it's important to continue the prefernce to the end.

 

[Duke of Marmalade]

That looks like the votes for Cyprian Brady.
"At the 2007 general election, held under Ireland's single transferable vote system, Brady polled just 939 first preference votes (2.7%) in the first count. However he was elected on the fourth count, due in large part to a pre-arranged electoral strategy seeing large transfers from his running mate, the Taoiseach Bertie Ahern,[8] and became only the second ever TD (after Brian O'Higgins) to be elected with less than 1,000 first-preference votes." - Wikipedia

Someone told me to always continue the preference all the way down the ballot, even if you wouldn't normally vote for the other candidates. I never understood why, as I assumed a ballot with no preference wouldn't be transferred. Now I see why it's important to continue the prefernce to the end.

Yes, in Australia your ballot paper is not valid unless completely filled. So no bias towards multi-party candidates down under. The Scottish system does allow you to stop giving preferences but they eliminate the bias by transferring the surplus in the correct fraction.

 

[Brendan Burgess]

That looks like the votes for Cyprian Brady.

That might be a good example to illustrate the point.

I wonder if I could get the all the counts and the papers showing how the surpluses were distributed.

Brendan

 

[jdwex]

Dublin Central (Dáil constituency) - Wikipedia

en.wikipedia.org

Has the 2007 election counts. He had a margin of 1800 or so over Pascal at the last count

 

[JohnCondon]

The numerical examples trying to show that the Irish system results in individual vote counts being greater than 1 have a subtle double-counting error and are incorrect.

However, we all know that there is effectively a piggy-back effect happening in our system: the transferable vote can be transferred and sometimes it relies on the strength of non-transferrable votes for the surplus number to be achieved and for it to be transferred. Brendan I think this is what you're getting at in your vote inflation point, but I don't think it helps to frame this as inflated vote values, or as vote values being greater than 1.

I think it would help your argument (and there is a good argument for what you're saying) if you avoid asserting that vote values can be greater than 1. For example, if you assert that a vote can have a count value greater than 1, then I think a reasonable test - that anyone would ask for - would be to check that the valid poll and total number of votes per count are not matching in the count results. We all know this does not happen, and that would naturally cause a backtrack on the claim that some vote values are greater than 1. If you did not backtrack on it, I believe the claim would end up relying on a very weak, and just incorrect, definition of what a vote value is, and what it means for a vote to be counted.

The whole argument is hingeing on who should get a say on where the surplus goes. I think our system is not unreasonable (and it is certainly not plainly wrong) where only those that express an opinion on the destination of the surplus can influence where it goes.

 

[Brendan Burgess]

For example, if you assert that a vote can have a count value greater than 1, then I think a reasonable test - that anyone would ask for - would be to check that the valid poll and total number of votes per count are not matching in the count results. We all know this does not happen, and that would naturally cause a backtrack on the claim that some vote values are greater than 1.

Hi John

It does happen!

And I gave an example, but it confused people.


A has 2,000 first preferences. 1,000 are used to get elected. B gets 1,000 of his surplus and by the same token 1,000 are non-transferable.

If you do it correctly, the second column would have 500 for B and 500 for non-transferable and the total poll would remain at 3,000.

Brendan

 

[Brendan Burgess]

The numerical examples trying to show that the Irish system results in individual vote counts being greater than 1 have a subtle double-counting error and are incorrect.

Where is the error?

The only double-counting I see is the double counting of A's 2nd preferences which went to B.

Brendan

 

[Brendan Burgess]

John

Here is a quiz for you which might help illustrate the point.

A gets 2,000 first preferences
The quota is 1,000
the surplus is 1,000

How many transfers will B get in the following scenarios?


A’s 2nd preferences e.g. in Scenario X, 2,000 people voted A no 1 and B no 2.