Full details on PR system and how best to use your vote

JackieChan

Guys, any websites that go through this in detail.
In particular I'm interested in how the transfers work.
What happens if one candiate is elected and has a surplus of X, how is this divided. Are all 2nd preferences reviewed or is just a sample taken.

oscarBravo

Wikipedia is your friend.

Gobán Saor

The wiki article is good - I like the idea they have in some countries of keeping the ballots and "re-running" the election with an extra seat instead of having a by-election if a member resigns or dies.

This article http://www.citizensinformation.ie/categories/government-in-ireland/elections-and-referenda/voting/proportional_representation deals with your specific question. In particular, this bit

Where a candidate is elected at the second or at later count, only the votes that brought him/her over the quota are examined in the surplus distribution, i.e., the parcel of votes last transferred to the elected candidate.

tends to be misunderstood. Say, after a number of counts, an FF candidate is 50 votes short of the quota. Suppose he then happens to get 100 vote from the elimination of an FG candidate from his own geographical part of the constituency (a not-unlikely result). Its the subsequent preferences of those 100 votes that are examined to determine the distribution of the 50 vote surplus. Thus you get the odd-looking result of what appears to be an FF surplus going not as you might expect but in accordance with the number 3 and lower preferences of the FG guy. You will invariably find that most of the media pundits constantly fail to pick up on this point on results night coverage.

Gobán Saor

There is a further subtlety to the example I've shown above. The transfer is done by selecting actual ballot papers at random and physically transferring them to the piles of votes for the remaining candidates. This will not matter for the first distribution of these votes, but where any of those votes are further distributed, there is a potential for distortion based on random selection of which number four and lower preference were selected for onward distribution.

This introduction of a random element could be significant especially if, as might well be the case, the selected papers all came from one particular ballot box and hence the lower prefences would tend to have a bias towards candidates from that particular geographic area. This is why there are often rows over the granting of re-checks or re-counts. A re-check doesn't alter the random selection element - it just checks that the bundles were tallied and counted correctly. A full re-count means the random selection will be done again and is another roll of the dice for a very narrowly defeated candidate. Again, this is very badly, if at all, explained by the talking heads on TV and radio.

DadaKopf

Gob&#225 wrote: »

There is a further subtlety to the example I've shown above. The transfer is done by selecting actual ballot papers at random and physically transferring them to the piles of votes for the remaining candidates. This will not matter for the first distribution of these votes, but where any of those votes are further distributed, there is a potential for distortion based on random selection of which number four and lower preference were selected for onward distribution.

This introduction of a random element could be significant especially if, as might well be the case, the selected papers all came from one particular ballot box and hence the lower prefences would tend to have a bias towards candidates from that particular geographic area. This is why there are often rows over the granting of re-checks or re-counts. A re-check doesn't alter the random selection element - it just checks that the bundles were tallied and counted correctly. A full re-count means the random selection will be done again and is another roll of the dice for a very narrowly defeated candidate. Again, this is very badly, if at all, explained by the talking heads on TV and radio.

The transferred ballots are 'mixed', implying randomness (not from the same ballot box), but from the same 'parcels' of transferable votes.

The magnitude of distortion varies, therefore from election to election, but tends to even out over the long-term between 1937 and today.

Using one proportionality index (100 being perfect proportionality, 0 being perfect improportionality), Ireland is 95 for the period 1923-1992. There's definitely room for improvement, but (according to this index), we're OK. However, this does result in a bonus of seats for FF and FG, and underrepresentation for Labour and smaller parties. But, again, these vary due to voter behaviour from election to election.

The mechanics of the transfer makes you wish there was a 100% secure electronic voting system. Then *all* the votes could be recounted to return 100% proportional transfers, probably making Ireland reach 100 on the proportionality index.

royfe1

I was in a discussion earlier on tactical voting and the PR STV system in Ireland

for example, for a National interest strategy, i.e to maximise a FG/Labour government getting in how should I order my preference?

If say my local FG candidate is definitely going to get in anyway, should I give my 1st 2nd etc to say a marginal labour candidate or marginal FG candidate who is borderline on getting in?

Anyone have any thought son how to tactical vote in Ireland, obviously in first past the post systems its easier

Roy

Gobán Saor

DadaKopf wrote:

However, this does result in a bonus of seats for FF and FG, and underrepresentation for Labour and smaller parties.

The larger the number of seats in a constituency, the more accurate the proportionality. There's no reason why we can't have a minimum size of 5 seat constituencies with 6, 7 and 8 or more seats in the cities. Larger constituencies would drastically cut the FF/FG seat bonus. Now guess why we have a lot of three-seaters?:D

Wicklow Boy

Well many people try to do what you are saying and hence the favoured candidate who is a dead cert loses his seat!

Dero

I'd still give the favourite a no. 1, and the marginal FG/Lab 2 & 3, simply because then it's more likely that the top FG candidate would get >quota and your 2nd & 3rd preferences would be as good as a first preference when transferred to the other candidate. It's a lottery as to whether your particular vote would get transferred, but I think you have to vote with the assumption that it will.

Pherekydes

Dero wrote:

It's a lottery as to whether your particular vote would get transferred...

Eh?

For the record:

If a candidate gets more than the quota, his surplus is distributed in the proportion of his second preferences.

Example:

Joe Bloggs (FG) gets 100 votes over the quota.

His 2nd prefs (of all his first prefs) are as follows:
LP 60%
GP 30%
Ind 10%

Then the LP guy gets 60 votes, the GP gal gets 30 votes and the Ind gets 10 votes.

johnlambe

royfe1 wrote:

Anyone have any thought son how to tactical vote in Ireland, obviously in first past the post systems its easier

The problem with PR-STV (as opposed to Condorcet, for example) is that a candidate who may have a lot of 2nd preferences but few 1st preferences could be eliminated before he gets those transfers, and this could lead to a situation where a candidate who is not elected is preferred by most people, to one who is.

Hence, strategic voting can be effective when you give your first preference to a candidate at risk of being eliminated early rather than one who you are confident will not be eliminated early, assuming that you want both candidates to be elected and don't mind which one is elected, if only one is (or if you're very confident that your favourite candidate will get elected despite the strategic voting).

I would expect that the effect of strategic voting is quite small in PR-STV, but I've never seen a statistical analysis of this. (I've been meaning to do this myself, using the data from when the electronic voting machines were used).

The way it works is.
If your no1 goes to a candidate that exceeds the quota then only the surplus is distributed to count again to the no 2 candidates on the surplus batch of votes.

If you vote for the strong candidate and he's elected on the 1st count and his surplus is 2000 votes (eg total 10k and quota=8k) then you have only a one in 5 chance that your next preference is going to be counted.

Thats where tight vote management is crucial.

Theres loads more to this and I havent the time right now to compile a guide so I'll sticky this thread and let people explain.

I'll look in later and tidy the thread and delete/clarify any inaccuracies that I see.

DadaKopf

You're not wrong there. In any case, the Boundaries Commission will have to redraw constituencies before the next election to account for the demographic change. If there was the appetite for change within the next government, it'd be good to increase the seats per constituency. And occasionally, there has been. Except it's been FF holding referenda to replace PR-STV with first-past-the-post!

No one bought it.

grumpytrousers

Gob&#225 wrote: »

There is a further subtlety to the example I've shown above. The transfer is done by selecting actual ballot papers at random and physically transferring them to the piles of votes for the remaining candidates. This will not matter for the first distribution of these votes, but where any of those votes are further distributed, there is a potential for distortion based on random selection of which number four and lower preference were selected for onward distribution.

This introduction of a random element could be significant especially if, as might well be the case, the selected papers all came from one particular ballot box and hence the lower prefences would tend to have a bias towards candidates from that particular geographic area. This is why there are often rows over the granting of re-checks or re-counts. A re-check doesn't alter the random selection element - it just checks that the bundles were tallied and counted correctly. A full re-count means the random selection will be done again and is another roll of the dice for a very narrowly defeated candidate. Again, this is very badly, if at all, explained by the talking heads on TV and radio.

Indeed; and it was this randomness that made me think that Electronic Voting was never going to work as long as we had PR-STV. From my (admittedly terribly limited ) knowledge of computers, it's rather difficult to *actually* get a truly random stream of numbers etc from one, and even if you could, there's no 'accountability' per se for where they came from.

It's why, I guess, in this day and age, the Lottery balls are done with, well, balls and not a computer spewing out 6 numbers from a darkened room.

Pherekydes

Tristrame wrote:

If you vote for the strong candidate and he's elected on the 1st count and his surplus is 2000 votes (eg total 10k and quota=8k) then you have only a one in 5 chance that your next preference is going to be counted.

The surplus is transferred to the remaining candidates in proportion based on all of the 2nd preferences.

Slow coach

Slow coach wrote:

The surplus is transferred to the remaining candidates in proportion based on all of the 2nd preferences.

yes in regard to the first count only my bad.In subsequent counts the one in 5 example comes into play.

Dero

Slow coach wrote:

Eh?

For the record:

If a candidate gets more than the quota, his surplus is distributed in the proportion of his second preferences.

Example:

Joe Bloggs (FG) gets 100 votes over the quota.

His 2nd prefs (of all his first prefs) are as follows:
LP 60%
GP 30%
Ind 10%

Then the LP guy gets 60 votes, the GP gal gets 30 votes and the Ind gets 10 votes.

OK. I understand that much. However, what I am confused about is the following:
Joe Blogs above got 100 votes over the the quota. Assume the quota was 1000, so JB got 1100 first preferences. If 100 are re-distributed among the remaining candidates;

a) which 100 are they?
b) what happens to the other 1000 ballot papers?

If I am completely misunderstanding this, please enlighten me. :-)

Slow coach

Slow coach wrote:

If a candidate gets more than the quota, his surplus is distributed in the proportion of his second preferences..

Yes but by that stage it is a lottery as to whether your vote is counted as part of the candidates surplus or not.
If say the second candidate is elected after the 3rd count for example then only the last batch of votes counted equal to the surplus is distributed and it's the next preference on those votes only that is looked at and determines the distribution of the surplus.
If your vote is earlier in the pile it dies.

Therein lies the lottery.

Dero

Dero wrote:

a) which 100 are they?
b) what happens to the other 1000 ballot papers?

If I am completely misunderstanding this, please enlighten me. :-)

They are the last 100 to be counted.
The other votes stay in the previously elected candidates pile never to be looked at again unless theres a total recount.
In the case of a total recount, they are not recounted as such, they are re examined to find any ballots mis counted eg spoiled votes.
Other than that in the example all bar the 100 in the last batch counted are now dead votes their lower preferences will never be looked at.

I might add that those 100 votes are redistributed in the case of the first count in direct percentage wise proportion to where the total of all the 2nd preferences went but again it's only the "100" that are distributed and again you have a lottery here as to which "100" that ends up being.
It will be the last 100 counted so the lottery is whether thats your vote or not when they examine the next preference.

Dero

Thank you Tristrame. That's what I thought, and that's exactly what I meant when I said lottery in my first post - it's a lottery as to whether your particular ballot paper will be used for transfers.

Tristrame

Tristrame wrote:

I might add that those 100 votes are redistributed in the case of the first count in direct percentage wise proportion to where the total of all the 2nd preferences went but again it's only the "100" that are distributed and again you have a lottery here as to which "100" that ends up being.
It will be the last 100 counted so the lottery is whether thats your vote or not when they examine the next preference.

To clarify further.

Say 80 of that surplus goes to a local candidate who is then eliminated on the next count.
Those 80 out of the original 100 will be in the special position of being counted again and in that case 80 out of the "special" lottery winning votes from the first count "100 vote" surplus distribution that have their 2nd preferences move on to whoever the 2nd prefernce is for.
In other words a random 80 voters will still be in play in deciding who gets elected where as the other 20 will stay in the uneliminated candidates total.

Meanwhile the rest of the first elected candidates votes from whom the random "100" were taken from in the first place will have no more say at all.

Wonderfull democracy isn't it ?
It's like animal farm some votes are more equal than others

royfe1

Thanks for the discussion, My head hurts now *groan*

Pal

can somebody please explain how they keep track of transfers in a recount. or as the above says, they dont.

a recount can dramatically differ if this is true.

Pherekydes

royfe1 wrote:

My head hurts now *groan*

So does mine! You learn something new every day!

simu

It's annoying. Although, would eliminating this random aspect change anything results-wise?

nesf

Tristrame wrote:

Wonderfull democracy isn't it ?
It's like animal farm some votes are more equal than others

Doesn't electronic voting removing the lottery part?

nesf

simu wrote:

It's annoying. Although, would eliminating this random aspect change anything results-wise?

Of course it would... But the argument goes that because it's random it's fair.

simu

nesf wrote:

Of course it would... But the argument goes that because it's random it's fair.

Well, fairness is probable! It's a bit HHG2TG! Why was this randomness accepted by the designers of the system? Were our super-duper e-voting machines to eliminate this or would that require some sort of legislative amendment?

Dero

I remember hearing that our e-voting system was to have some form of random selection to simulate the current system because people were/are comfortable with it. I'd personally prefer if all the votes were counted all the way down - something which could be done much more easily with an electronic system, but that's a discussion for another day...;)

nesf

Dero wrote:

I remember hearing that our e-voting system was to have some form of random selection to simulate the current system because people were/are comfortable with it. I'd personally prefer if all the votes were counted all the way down - something which could be done much more easily with an electronic system, but that's a discussion for another day...;)

Yeah but the (easy) option of removing the lottery element would be there if we had electronic voting versus the present where counting down every vote would lengthen counts and recounts by quite a bit which is a good argument for having a lottery in the first place.

It would require a change in legislation though wouldn't it?

Voipjunkie

This explains exactly how the system works

http://www.environ.ie/en/LocalGovernment/Voting/PublicationsDocuments/FileDownLoad,1895,en.pdf

daveirl

This post has been deleted.

oscarBravo

Electronic voting could eliminate the random aspect of surplus distribution, but the broken system we purchased didn't, because yes: it would require a legislative change.

JustSomeone

Still not entirely clear on this - specifically on the "parcel" mentioned above.

Simple example - quota is 16,000 and Joe Bloggs gets 17,000 first prefs, 1,000 of his votes are to be "transferred". What I don't understand is which 1,000.

Are 1,000 ballots selected randomly from the 17,000?

Are they selected from a subset of the 17,000, if so, how is the subset defined?

Is there scope for a candidate/agent to challenge the subset selected on the basis that it is unrepresentative?

If 300 of the 1,000 selected ballots have no next preference, are further ballots selected until 1,000 transferable votes are selected?

In the event of a recount, are the same ballots selected?

Thanks.

JustSomeone

grumpytrousers wrote:

It's why, I guess, in this day and age, the Lottery balls are done with, well, balls and not a computer spewing out 6 numbers from a darkened room.

Wouldn't that be cool... vote transfers being done on that basis? Balls with pictures of candidates' heads bouncing around in a lotto yolkie.

grumpytrousers

JustSomeone wrote:

In the event of a recount, are the same ballots selected?

In the event of a recount, I believe the same ballots are selected; the re-count is ONLY a check that the numbers are 'right'...

would love to hear answers to the other questions you raise tho'...

Zombrex

grumpytrousers wrote:

Indeed; and it was this randomness that made me think that Electronic Voting was never going to work as long as we had PR-STV. From my (admittedly terribly limited ) knowledge of computers, it's rather difficult to *actually* get a truly random stream of numbers etc from one, and even if you could, there's no 'accountability' per se for where they came from.

<Off topic>
That is not quite true. A computer itself cannot generate a truly random number itself, but it can generate a number based on a random input source.

So for example if you are setting up some kind of private key for encrypting email you might be asked to bang keys for a bit on a keyboard in a randomly pattern. This produces a random number for the computer to generate larger random numbers from.

In the case of something like a polling booth a device that feeds in the current light levels, or wind speed and direction on top of the polling booth, or any other random data, could be used to generate a random number.

Of course I've no idea if any of these systems are used in electronic voting. From what I read a lot of these electronic voting machines seem to be put together by people with limited experience in securing systems. They don't seem very well designed at all.

These guys seem to have got it right
http://www.wired.com/techbiz/media/news/2003/11/61045
<back on topic>

grumpytrousers

<staying off topic>Yup, that's more or less what my mate who works in computers was saying down the pub...but my point remains, that there isn't, in electronic voting, the random variable that would allow a level of randomness be introduced. And the kicker is, of course, that even if there was, it'd be completely invisible <as you were>

daveirl

daveirl wrote:

This post has been deleted.

That was only for the trial runs though wasn't it?
Ie they couldnt have the trial e voting not using randomness and the pencil voting being random.

daveirl

This post has been deleted.

JustSomeone

Oops, looks like my questions were actually answered in the citizens info link above. Not that it hasn't generated more

If a candidate receives more than the quota on any count, the surplus votes are transferred to the remaining candidates in proportion to the next available preferences indicated by voters (i.e., the next preference on each vote for a candidate who has not been elected or eliminated). For example, if candidate A receives 900 votes more than the quota on the first count and on examining all of his or her votes, it is found that 30% of these have next available preferences for candidate B, then candidate B does not get 30% of all candidate A's votes, candidate B gets 30% of his/her surplus, i.e., 270 votes (30% of 900).

So, in my example above, all 17,000 votes would be examined to identify how many of the 17,000 votes have selected a second preference, and what percentage of these votes go to each of the other candidates. So, if 12,000 of Deputy Elect Blogg's 17,000 first preferences have indicated a second preference, and of those 12,000, 9,000 listed Mary Moggs as 2nd pref and 3,000 listed Ted Toggs, then Mary would get 3/4 of the surplus (ie 750) and Ted would get 1/4 250.

This seems reasonable enough. No randomness, no subselections and repeating the process should produce the same results.

Interestingly, it only applies on the surplus from the 1st count.

After that:

Where a candidate is elected at the second or at later count, only the votes that brought him/her over the quota are examined in the surplus distribution, i.e., the parcel of votes last transferred to the elected candidate.

In this case the "parcel" is made up of all of the votes most recently transferred to the candidate.

So, to answer my own questions:


Are 1,000 ballots selected randomly from the 17,000?
Nope. They are selected according to very definite rules.

Are they selected from a subset of the 17,000, if so, how is the subset defined?
On the first count, all are taken into account, thereafter the most recent set of transfers are considered.

Is there scope for a candidate/agent to challenge the subset selected on the basis that it is unrepresentative?
Nope.

If 300 of the 1,000 selected ballots have no next preference, are further ballots selected until 1,000 transferable votes are selected?
The full surplus is transferred, according to the ratio calculated by examining the appropriate "parcel" of ballots.

In the event of a recount, are the same ballots selected?
Yes. By definition. There is no room for discretion or randomness.


And the absence of a random element, Wicknight and Grumpy Trousers, means that an electronic system would not need to consult wind, rain, or particle physics to generate a random number, and so those charged with delivering any electronic voting system can continue to concentrate on how to make the system robust, verifiable and tamper-proof.


Cheers!

silvine

Very enlightening.

However, I am still not sure what is the best way to tactically vote?

Should I put my favourite candidate - who is sure to get in - at number one?

Or should I place the shoe in further down my preference list in the hope that:

a) Giving my number one preference to candidate who has a battle on his hands to get elected will do greater good
b) My vote will win the selection lotto for transfers and the shoe in will get my second preference

Voipjunkie

silvine wrote:

Very enlightening.

However, I am still not sure what is the best way to tactically vote?

Should I put my favourite candidate - who is sure to get in - at number one?

Or should I place the shoe in further down my preference list in the hope that:

a) Giving my number one preference to candidate who has a battle on his hands to get elected will do greater good
b) My vote will win the selection lotto for transfers and the shoe in will get my second preference

If the candidates are from the same party then the party should have tried to manage the vote so that people are asked in one area to vote 1 ,2 and in another 2,1
If they are separate parties then keeping your second favourite candidate in the race is more important than plumping for the shoo in ( of course if too many people do that they are no longer a shoo in)
It is one of the things with STV in that a candidate who could have far more lower preferences is eliminated early for lack of first preferences.

Voipjunkie

daveirl wrote:

This post has been deleted.

That is correct and was one of the arguments against the e-voting machines in that it did not eliminate the random element but tried to replicate the random element.

JustSomeone

Giving your first preference to the struggling candidate ensures that s/he has a better chance of getting in.

If s/he is eliminated, then the vote transfers to your lower preferences. No lotto. your vote continues down your list until it elects someone, ends up in the last eliminated candidate's pile, or runs out of preferences.

Voting for the shoe-in first will probably stop your vote there when the shoe-in is elected, as only surpluses (a fraction of the total vote for the shoe-in) will be transferred.

I wouldn't dream of advising you how to vote, what I propose to do is vote for people in the following order.

The people I want in, and who are unlikely to get in
The people I want in, and who are likely to get in
The people I am neutral on
The people I want to keep out, and are unlikely to get in

The last point above is probably academic, but I like it as a way of annoying the people I want to keep out, and who were seen as likely to get in but are still waiting after umpteen counts.

JustSomeone

Voipjunkie wrote:

That is correct and was one of the arguments against the e-voting machines in that it did not eliminate the random element but tried to replicate the random element.

What random element?
The Citizens Info page (http://www.citizensinformation.ie/categories/government-in-ireland/elections-and-referenda/voting/proportional_representation) doesn't mention any random element.

JustSomeone

JustSomeone wrote:

What random element?
The Citizens Info page (http://www.citizensinformation.ie/categories/government-in-ireland/elections-and-referenda/voting/proportional_representation) doesn't mention any random element.

read page one and two of this thread....

JustSomeone

Tristrame wrote:

read page one and two of this thread....

I've read the full thread, and can see that a few people have said that there is a random element.

This doesn't tally (!) with what the Citizen's Info site says, or with what the Dept of the Environment says: http://www.environ.ie/en/Publications/LocalGovernment/Voting/FileDownLoad,1895,en.pdf

I'm not denying the knowledge of the people at the start of the thread, just that the official docs don't tell the same story.

Does anyone know of any official documentation which admits the random element?

daveirl

This post has been deleted.

grumpytrousers

it's not an official document, but if you read stephen collins in todays irish times at around page 9, top left of the page, he explains how PR-STV works. towards the end of the article, the element (and it's only an element) of randomness is introduced. It tends to occur when a surplus is being distributed, and *while* that happens, somebody hits a quota, the surplus of the surplus (if you will) is distributed in a 'random' fashion...

somebody with an ireland.com login might be able to throw more light...

*edit* don't need login

http://www.ireland.com/focus/election2007/voting/voting20.html

SeanW

The finer details of PR-STV can be confusing.
Ok, here's my problem: I want to vote as tactically as possible, to make it a fine-tuned FF bashing instrument. But I also want good representation for Longford (consitutency Longford/Westmeath).

The bookies http://www.paddypower.com/bet?action=go_type&category=SPECIALS&disp_cat_id=31&ev_class_id=33&ev_type_id=8065&ev_oc_grp_ids=45148&bir_index=
give the Mullingar based Labour candidate the best odds at 1-50. So he's a dead cert. No low-pref from me.
My intended number one, James Bannon FG, is in a tight, potentially losing fight with Peter Kelly FF. So the first 2 or 3 preferences should be going to FG candidates.
On the other hand, I think this is the first time the Greens are running a candidate here and I'd like not to have theirs lose out in the first count. But the bookies think she's a dead loser anyway so I don't want to waste a first pref.

I am also assuming that either the Greens or Labour would do business with FF in the right/wrong circumstances.

That gives me a choice of 3 schemes to control my first 5 preferences.
1:
Local FG
Green
Other FGs
Labour
Others

2:
Local FG
Other FGs
Green
Labour
Others

3:
Green
Local FG
Other FGs
Labour
Others