Complex Voting Question - Potential flaw in PR System

lempsipmax · 25-02-2011 12:52pm #1

Take the following scenario:

Snap, Crackle and Pop are the only 3 candidates running for election in a 2 seat constituency.
3000 votes were cast making the quota 1001 votes.
After the first count Snap is elected with 1002 First Preferences.
Crackle gets 999 first preferences
Pop gets 999 first preferences
Snaps surplus of one vote is going to be transferred.

Now here is my question:
How is the vote to be transferred selected? As some of the Snap votes have given Crackle their second preference while others have given second preferences to Pop, the selection of the vote to transfer actually determines the outcome.
Is it selected at random?
And how is it randomised?

There is a 3rd possibility and that is that the surplus vote does not contain a second preference, I assume this would result in a recount?

johnnyskeleton · 25-02-2011 12:55pm

Usually what will happen is that the electoral counters are old party hacks from Pop's party and when they are "randomly" selecting the transfer, it will just happen to be the one with Pop as second preference.

Seriously though, such a situation is highly unlikely to arise, and the detriment of remedying the system to avoid that peril would probably destroy the entire PR system. Therefore, I think some minor unfairness in unusual circumstances can be accepted. It is not at all like the Florida fiasco where GW Bush got elected, apparently by magic.

Inquitus · 25-02-2011 1:00pm

The solution to this is a reliable and secure system of e-voting which can quickly review all 2nd preference votes and pro-rate the excess of one exactly.

Riskymove · 25-02-2011 1:01pm

lempsipmax wrote: »

Is it selected at random?
And how is it randomised?

the 1002 votes Snap got should be examined for 2nd preferences

whichever of Pop and crackle got more 2nd preferecnes would get the vote

Cookie_Monster · 25-02-2011 1:02pm

Inquitus wrote: »

The solution to this is a reliable and secure system of e-voting which can quickly review all 2nd preference votes and pro-rate the excess of one exactly.

+1
electronic is the way to go.

Also cuts out the huge expenses of the following days (paying for counters etc) and means we don't have to wait forever to figure it all out.

Godge · 25-02-2011 1:04pm

Riskymove wrote: »

the 1002 votes Snap got should be examined for 2nd preferences

whichever of Pop and crackle got more 2nd preferecnes would get the vote

That is exactly what happens.

rjt · 25-02-2011 1:04pm

There are all sorts of similar problems. In fact it's possible for someone to be elected in a constituency of 3 seats, but if another seat is added to the constituency that person would not have been elected (with the exact same voters)!

Have a look through: http://en.wikipedia.org/wiki/Issues_affecting_the_Single_Transferable_Vote.

lempsipmax · 25-02-2011 1:04pm

Seriously though, such a situation is highly unlikely to arise

I disagree. If you were to multiply out the numbers in my scenario. Say there were hundreds or thousands of votes being transferred, the selection of the votes to be transferred could easily swing the election of the next candidate.

In fact a large proportion of the votes which elect candidates are transferred are they not (maybe someone could hazard a guess as to how many)?

Some might argue that randomness in the situation could 'even it's self out', however I find it unacceptable. If someone was determining which bunch of first preferences to use to elect a candidate and which to assign as a surplus, it would be possible that a knowledge of likely transfer patterns in a given geographic area could allow some degree of control over which way the surplus votes were likely to transfer.

lempsipmax · 25-02-2011 1:07pm

rjt wrote: »

There are all sorts of similar problems. In fact it's possible for someone to be elected in a constituency of 3 seats, but if another seat is added to the constituency that person would not have been elected (with the exact same voters)!

Have a look through: http://en.wikipedia.org/wiki/Issues_affecting_the_Single_Transferable_Vote.

Fascinating link. I will have a read of it in a while, looks like it will answer my questions.

Riskymove · 25-02-2011 1:08pm

lempsipmax wrote: »

If someone was determining which bunch of first preferences to use to elect a candidate and which to assign as a surplus, it would be possible that a knowledge of likely transfer patterns in a given geographic area could allow some degree of control over which way the surplus votes were likely to transfer.

all the votes are emptied into one pile and mixed before counting so the geographical spread is reduced as a factor

as I mentioend above the proportion of 2nd preferences determines the distribution of the surplus

Zubeneschamali · 25-02-2011 1:08pm

When Snap's votes are examined, only two have a 2nd preference marked: one for Crackle and for Pop.

It's a draw!

Riskymove · 25-02-2011 1:10pm

Zubeneschamali wrote: »

When Snap's votes are examined, only two have a 2nd preference marked: one for Crackle and for Pop.

It's a draw!

actually a Judge (appointed by Pop's father- Pop Pop) decides that a number of Crackle's votes are invalid and declares Pop the winner

johno2 · 25-02-2011 1:19pm

Cookie_Monster wrote: »

+1
electronic is the way to go.

Only if it's an open source system with multiple levels of auditing allowed. There is no place in a free democracy for a secret proprietary voting system. It's too easy for the guys behind the scenes to manipulate the results. In fact I think that's the real reason that Bertie was so keen on implementing such a system.

Cookie_Monster wrote: »

Also cuts out the huge expenses of the following days (paying for counters etc) and means we don't have to wait forever to figure it all out.

Not really. The voting system we used in 2002 cost €52 million in secure storage costs up to 2007. That about €16 per vote that was cast.

johno

johnnyskeleton · 25-02-2011 1:24pm

lempsipmax wrote: »

I disagree. If you were to multiply out the numbers in my scenario. Say there were hundreds or thousands of votes being transferred, the selection of the votes to be transferred could easily swing the election of the next candidate.

That's my point. If you do multiply out the numbers, the seeming unfairness of chosing one random transfer is defeated by chosing 1,000 transfers which will have a more representative sample. Of course, they still won't be perfect, but the system cannot have perfect transfers.

lempsipmax wrote: »

In fact a large proportion of the votes which elect candidates are transferred are they not (maybe someone could hazard a guess as to how many)?

Invariably there will be people elected on transfers. But I can't think of a single constituency that has only 3 candidates, 3000 voters and an almost perfect split of preference. Typically, there will be perhaps 10 candidates, tens of thousands of voters and a wide variety of vote. Typically, the national number of seats will more or less tally with the polls and national voter preference, whereas in the UK system there can be massive swings.

Some might argue that randomness in the situation could 'even it's self out', however I find it unacceptable. If someone was determining which bunch of first preferences to use to elect a candidate and which to assign as a surplus, it would be possible that a knowledge of likely transfer patterns in a given geographic area could allow some degree of control over which way the surplus votes were likely to transfer.

Well look if you want to point out academically a flaw that could arise in specific but unlikely circumstances, that's fine. However, in the real world this is unlikely to arise because greater numbers means a greater representative sample.

In reality the numbers transferred from a successful candidate will usually be in the thousands and will be selected fairly randomly. Even where the numbers to be transferred are less than a hundred, they will still be picked randomly. But even if they were not, in actual elections such a small number of votes doesn't make all that much difference. And if it does, then they can have a recount.

The alternative is to throw out the whole system. So what do we replace it with? I'm not saying PR is perfect, but it is less imperfect than other voting systems such as first past the post and the whacky electoral colleges system in the US.

lempsipmax · 25-02-2011 1:39pm

johnnyskeleton wrote: »

The alternative is to throw out the whole system. So what do we replace it with? I'm not saying PR is perfect, but it is less imperfect than other voting systems such as first past the post and the whacky electoral colleges system in the US.

Good points. I dont have an alternative suggestion. Its just something that occurred to me during a discussion of the PR system.

Ciaran · 25-02-2011 4:58pm

lempsipmax wrote: »

I disagree. If you were to multiply out the numbers in my scenario. Say there were hundreds or thousands of votes being transferred, the selection of the votes to be transferred could easily swing the election of the next candidate.

In fact a large proportion of the votes which elect candidates are transferred are they not (maybe someone could hazard a guess as to how many)?

The vast majority of transferred votes are from eliminated candidates, where all votes are transferred. Surpluses are usually not that large. Ties between two candidates can happen in any voting system and there's not much that can be done to avoid them.

Don Din · 25-02-2011 6:40pm

lempsipmax wrote: »

Take the following scenario:
Snap, Crackle and Pop are the only 3 candidates running for election in a 2 seat constituency.

3000 votes were cast making the quota 1001 votes.

After the first count Snap is elected with 1002 First Preferences.

Crackle gets 999 first preferences

Pop gets 999 first preferences

Snaps surplus of one vote is going to be transferred.

Now here is my question:
How is the vote to be transferred selected? As some of the Snap votes have given Crackle their second preference while others have given second preferences to Pop, the selection of the vote to transfer actually determines the outcome.
Is it selected at random?
And how is it randomised?

There is a 3rd possibility and that is that the surplus vote does not contain a second preference, I assume this would result in a recount?

Beacause Snap is elected in the first count, all of Snaps votes are checked to see what the second preference is. Lets assume that Cackle gets 60%, Pop gets 30% and 10% don't transfer then Cackle will be the surplus vote in preference to Pop and will be elected without reaching the quota! That's what happens.

dvpower · 25-02-2011 6:44pm

Inquitus wrote: »

The solution to this is a reliable and secure system of e-voting which can quickly review all 2nd preference votes and pro-rate the excess of one exactly.

You wouldn't even need e-voting, just e-counting.
The existing ballot papers could be easily scanned and ocr'd

Don Din · 25-02-2011 6:45pm

Riskymove wrote: »

the 1002 votes Snap got should be examined for 2nd preferences

whichever of Pop and crackle got more 2nd preferecnes would get the vote

They are. People are confused about what happens with a surplus in the 1st count and a surplus arising from a transfer from another candidate. When the surplus arises in the 1st count then all the votes of the elected candidate are checked to see what proportion of the surplus should be transfered to the other candidates. A random bundle of the votes from the surplus are then physically transfered.

Capt'n Midnight · 25-02-2011 7:57pm

Inquitus wrote: »

The solution to this is a reliable and secure system of e-voting which can quickly review all 2nd preference votes and pro-rate the excess of one exactly.

if voters can verify the way they voted they could sell their votes which is not a good thing.

If voters can't verify the way they voted then the ability to slip a few votes in a key marginal means it you can't trust people not to be tempted, whether the operators, the programmers or third parties like NSA/Mossad/Mi6/Betting syndicates

The paper system is cheaper , more fun (watch them sweat) and harder to fake results from (ballots are numbered and punched and generally have different hand writing) - problems like personation would be the same in either event.

Capt'n Midnight · 25-02-2011 8:03pm

Riskymove wrote: »

actually a Judge (appointed by Pop's father- Pop Pop) decides that a number of Crackle's votes are invalid and declares Pop the winner

RTE radio had a returning office and he said 2% of votes were potentially spoiled. 1% of votes had more than a single 1 on them and so were totally wasted

The other 1% were up for grabs , in some cases if the intention is clear then they would be allowed. ( though if you put down a pair of 3's then your vote won't transfer after 1 and 2 )

So Mr Judge in this example has about 1% of lee way

In practice when there is a recount the result may change by a few votes either way, and then the other party could also ask for a recount , and after a while someone conceeds defeat or the whole thing goes to court.

Ppat · 25-02-2011 9:05pm

Why not have a Snap election?!!!

Fergus · 26-02-2011 1:22pm

dvpower wrote: »

You wouldn't even need e-voting, just e-counting.
The existing ballot papers could be easily scanned and ocr'd

As soon as the full record of each ballot has been compiled, the database should be posted online so everyone is free to process the counts and transfers with their own software. That would be interesting cos you could easily compare outcomes under random transfer SVT, fractional transfer, etc.

Posting the actual scans of the ballots would allow the selling of votes cos you could mark it in some obscure non-spoiling way and prove you had voted as agreed. If there is any means to prove your particular voting was included, then that's open to selling. You should be at least able to prove your particular ballot paper was included in the count though.

blubloblu · 26-02-2011 1:34pm

Something like this? http://www.ted.com/talks/david_bismark_e_voting_without_fraud.html

baalthor · 26-02-2011 2:01pm

johnnyskeleton wrote: »

The alternative is to throw out the whole system. So what do we replace it with? I'm not saying PR is perfect, but it is less imperfect than other voting systems such as first past the post and the whacky electoral colleges system in the US.

A non STV PR system such as voting for party lists, just like the vast majority of European countries?

telekon · 26-02-2011 2:27pm

johnnyskeleton wrote: »

Seriously though, such a situation is highly unlikely to arise, and the detriment of remedying the system to avoid that peril would probably destroy the entire PR system...

Why? To be fair, there's very little to choose between them...they are three very similar candidates all standing on a breakfast related platform.

johnnyskeleton · 26-02-2011 2:30pm

baalthor wrote: »

A non STV PR system such as voting for party lists, just like the vast majority of European countries?

List systems are PR too. The OP's objection is to PR. He doesn't accept the argument that on a larger scale the flaw is less obvious, so he probably wouldn't accept the benefits of a list or partial list system too.

Capt'n Midnight · 26-02-2011 2:42pm

Fergus wrote: »

As soon as the full record of each ballot has been compiled, the database should be posted online so everyone is free to process the counts and transfers with their own software. That would be interesting cos you could easily compare outcomes under random transfer SVT, fractional transfer, etc.

Posting the actual scans of the ballots would allow the selling of votes cos you could mark it in some obscure non-spoiling way and prove you had voted as agreed. If there is any means to prove your particular voting was included, then that's open to selling. You should be at least able to prove your particular ballot paper was included in the count though.

you have the very big problem that if there are 8 or more candidates proving which way you voted is trivial.

Most constituencies have a lot of candidates so it's very easy to create a unique pattern once you avoid likely voting patterns. Bottom 8 candidates gives 80,640 combinations, so almost certainly unique, especially when you consider that most voters will follow patterns. (Factorial 8 x 2)*

Ignoring that 1% of votes will be up to the Returning office deciding if they have been spoilt (and thus not available on line till all the dust has settled and even then they may query if a close thing) it would take as long to upload the votes as to count them. OCR of random handwriting would have too high a failure rate especially when the sample of handwriting is only a few characters. - Maybe one of those tick the box things - but more likely for the person to make a mistake since you could easily have hundreds of boxes to choose from.

Lets imagine someone uses A,B,C,B instead of 1,2,3,4 - they can easily write a word on the paper (they will have half the alphabet to choose from) and then fill in the rest of the ballot - how can a machine detect if coincidence or not even with a spell checker and parsing and Bayesian filtering.

*You can double the number of combinations by not filling in the last choice
For 8 choices the bottom 7 give 10,080 combinations. But you can get even more combinations by mixing your effective first preference with candidates who are certain to be eliminated early on including possibly a running-mate of the targeted candidate. Once these are placed higher than potential rivals it shouldn't affect the outcome.
cba doing out the full maths but 2 no-hopers amongst 8 should then be enough to prove which way you voted in this election.

http://www.irishtimes.com/newspaper/ireland/2011/0226/1224290927713.html

Tipperary North, Tipperary South and Kildare South have the fewest candidates at eight

Fergus · 26-02-2011 2:58pm

Good point about using unique patterns. I wasn't suggesting OCR for reading the ballots though.

limklad · 26-02-2011 4:14pm

Cookie_Monster wrote: »

+1
Electronic voting is the way to go

Also cuts out the huge expenses of the following days (paying for counters etc) and means we don't have to wait forever to figure it all out.

Electronic is the way to go but have paper as backup to minimise voting fraud.

The Method I suggest is : You should received the Paper vote as Normal. Attend the Voting Machine, Insert Paper Vote to a printer, vote and verify voting on the Machine and wait for printing to be finished and verify vote and then place the paper vote in the traditional voting box.

Have the Electronic voting machine should have a touch screen, with simple display and interface like most games and ticket machines to match current paper Forms with Candidates pictures/Names/Parties with a box beside their name.You move the Number by touch in your preference. You be allowed to to vote using only 1 or 1,2 or 1,2,3 etc like current written form.

You make your selection and press Vote and a Prompt to make sure you want to proceed and go back if you change your mind.

The Voting Machine will print your preference on voting paper and prompt you to verify your printed vote. Once done you click Vote on the Machine and place your printed vote into the Traditional Voting Box.

Finish of Voting: Each Presiding officer will verify each paper vote issued with number of Electronic Voting. If there a discrepancy then paper votes issued and electronic votes Presiding officer must issued discrepancy for Returning officer to decide to invalidate on that electronic polling box. Therefore Paper votes are only counted as backup using Traditional Methods.

Also three Lights should be present on top of the voting Booth.
Green (Voting have been done and ready for new voter)
Orange (Voting in progress),
Red (Need Help from presiding officer)

Count Centre:
Valid checking of electronic voting needs to be done for a sample of boxes using Traditional statistical methods.
A sample of boxes are count and the various combination of transfers are counted and verified with electronic voting machine to verify everything is fine. Any Electronic problems with votes then that polling station then Paper votes are counted for that polling station. The Returning officer have full decision to invalided electronic voting of certain polling stations is there is any discrepancy only.

The Election Counts. The Electronic voting Machines Votes will be uploaded to a secure central computer (non-Internet networked) which will show all the votes and order of transfers for each polling station and printed for public viewing just like information of tallies are viewed but only after the Presiding officer validate the Electronic polling station counts. Returning Officers Upload each count to a Internet connected secure database for General Public Viewing.

The Traditional order of Counts for each round of Count and number of votes will be announce by the Returning officer just like Traditional Paper voting.
The delay of announcement over three hours to allow each constituency to publicly stagger announcements at different times to allow public view by the General Public and also should be allowed to allow the candidates to assimilate any defeat and compose themselves before facing the Public.

All candidates before announcement should enter a Private Candidates Meeting Room room outside public viewing. Any Mobile phone or any recording device should be banned and criminal prosecutions if anybody breach the privacy of that room. Returning officer is exempt from having a secure communication device with his/her staff other than current traditional rules applied to the Returning Officer and their staff. Returning Officer and their staff must abide by the privacy rules while in the Candidates Meeting Room to maintain independence and neutrality.

Then Candidates who failed miserably with no chance of been near 20% of the quota after transfers should be told privately before the count be publicly announcement with all candidates held in a room with the presiding officer to avoid public humiliation and allow them to recover their composure before the public announcement and before any of them giving interviews for the media. The Returning Officer or the staff must not reveal the completed results before the Public Returning Officer announcement of the Counts and Election of Candidates to the Public as it is Traditionally done.

baalthor · 26-02-2011 5:12pm

johnnyskeleton wrote: »

List systems are PR too. The OP's objection is to PR. He doesn't accept the argument that on a larger scale the flaw is less obvious, so he probably wouldn't accept the benefits of a list or partial list system too.

The issue described only happens with Single Transferable Vote PR.
For example, in a pure list PR system, voters mark X beside the party they want to vote for. Then the number of votes for each party are counted and divided by the total valid poll to give a % for each party. Multiply that by the number of seats in parliament and you that party's proportional representation.
No multiple counts, surpluses, transfers, eliminations and two-day vote counts. These are not an inherent part of PR, just of STV-PR.
Of course there are many other versions of PR, each with advantages and disadvantages but only STV has the multiple counts and transfers which is why our election counting takes so long.

Complex Voting Question - Potential flaw in PR System

Comments