Brexit

[KarenRei]

"Flextension" granted until 31 January. May be terminated earlier if the UK comes to a deal.

ED: Apparently it's not been formally "granted", only decided; there's still a written procedure, which is to be completed within 24 hours.

Odds of a 9 Dec. election increasing, now that no deal is being taken off the table. Lib Dems and SNP supporting it. Proposing to do it using legislation to bypass the Fixed Terms Parliament Act so that they only need a simple majority, and thus Labour can't veto it. Downside for the government is that Parliament will get a chance to amend any such bill.

 

[Cogarch]

Shame the flextension can't work in the other direction as well..

 

[EVNow]

Interesting line of attack by Labor.

Jeremy Corbyn on Twitter

This confirms what we know would happen with a Johnson-Trump trade deal. The Tories would put our NHS at the mercy of US pharmaceutical, tech and private healthcare companies. We will stop a No Deal Brexit and Johnson's sell out Brexit deal.

​

 

[KarenRei]

So, Labour and Corbyn are now backing an early election. It's going to happen. I don't think the date (9th or 12th of December?) has been locked in yet.

 

[KarenRei]

So, looks like you guys will be going to the polls on the 12th. Congrats

 

[Webeevdrivers]

So the question remains. If there was another Brexit vote, would it pass?

 

[KarenRei]

So the question remains. If there was another Brexit vote, would it pass?

Really depends on the structure. There's four options "approved" by the EU:

• No Deal
• Boris's Deal
• May's Deal
• No Brexit

They might be able to add some, like a customs union, with EU approval. But it's definitely not a binary choice anymore. So how do you fairly do a multi-choice vote? If one wanted to rig it against deals, for example, they could just go with first-past-the-post, so votes split among the deal options. Not a fair way to do it. You could also arrange it into a series of multiple binary choices - the way you group them could help ensure a particular outcome that the person setting it up intends on achieving (same effect as gerrymandering an electoral district). Also not a fair way to do it.

IMHO, I think the best way to do it might be Baldwin's Method - that is, condorcet voting with instant-runoff tiebreaking.

• Condorcet voting means you rank your choices, and then every pair of possibilities is compared against each other, with all of the other options removed, in a bunch of one-on-one races. If any possibility would win against all other possibilities in a one-on-one race, it's the condorcet winner. Just seems fair, right?
  
  
• The downside to condorcet voting is that there is not a guaranteed winner. You can get a situation where A is more popular than B, which is more popular than C, which is more popular than A. In that case, you need a resolution mechanism. IMHO, the fairest method is Schulze method - but it's a really complicated algorithm and hard to explain to people, so I think most people would see it as basically some black magic behind the scenes and mistrust it. Dodgson's method (invented by Lewis Carroll) is a bit easier to explain and still IMHO very fair, but perhaps still to complicated for the general electorate (maybe I underestimate people, however...) Instant runoff, however, is really easy to explain to people - after each round, you just remove the option that got the fewest top-ranking votes. Eventually you end up with a condorcet winner.

You can just skip the concept of a condorcet winner altoghether and go with instant runoff voting, but in practice, instant runoff is terrible as a sole mechanism - arguably worse than first-past-the-post. Let's say that there were three options:

A) Hard Left
B) Reasonable Middle Ground That Everyone Rather Likes
C) Hard Right

Let's say that your electorate can be divided into three groups:

Leftists (34% of the electorate): Thinks (C) is terrible (0/10), (A) is awesome (10/10), and B is rather nice (9/10)
Centrists (31% of the electorate): Think (A) and (C) are terrible (a mix of 0 or 1/10), and (B) is awesome (10/10)
Rightists (35% of the electorate): Thinks (A) is terrible (0/10), (C) is awesome (10/10), and B is rather nice (9/10)

So roughly 2/3rds of the electorate despise (A) and (C), while almost everyone likes (B). But instant runoff immediately kicks (B) out of the race. Only the extremists remain, and (C) slightly edges out (A) in the final round to win. This is a big problem with instant runoff: it tends to kick out broadly popular options ("everyone's second-favourite choice"), while leaving behind options that specific voting groups think are #1, even if said options are broadly unpopular among the broader electorate.

With a condorcet method, of course, the results are obvious. (B) utterly crushes both (A) and (C) in any one-on-one race, and so wins handily. As it should.

BUT.... this is of course totally my take, and has nothing to do with what anyone in the UK will decide!

 

[KarenRei]

I don't know, do you think people would find Dodgson's method too complicated? Basically, you first have to explain the concept of condorcet voting. Then you explain the resolution mechanism, which is: if there's a conflict where A is more popular than B, B is more popular than C, and C is more popular than A, then you look at each option and figure out: what's the fewest number of changes to people's vote ordering that you'd have to make in order to make this option the unambiguous condorcet winner? The option for which you have to change the fewest votes for it to come out on top is the winner.

Too complicated? Maybe the fact that Lewis Carroll came up with it would give it some cachet?

 

[Webeevdrivers]

Really depends on the structure. There's four options "approved" by the EU:

• No Deal
• Boris's Deal
• May's Deal
• No Brexit

They might be able to add some, like a customs union, with EU approval. But it's definitely not a binary choice anymore. So how do you fairly do a multi-choice vote? If one wanted to rig it against deals, for example, they could just go with first-past-the-post, so votes split among the deal options. Not a fair way to do it. You could also arrange it into a series of multiple binary choices - the way you group them could help ensure a particular outcome that the person setting it up intends on achieving (same effect as gerrymandering an electoral district). Also not a fair way to do it.

IMHO, I think the best way to do it might be Baldwin's Method - that is, condorcet voting with instant-runoff tiebreaking.

• Condorcet voting means you rank your choices, and then every pair of possibilities is compared against each other, with all of the other options removed, in a bunch of one-on-one races. If any possibility would win against all other possibilities in a one-on-one race, it's the condorcet winner. Just seems fair, right?
  
  
• The downside to condorcet voting is that there is not a guaranteed winner. You can get a situation where A is more popular than B, which is more popular than C, which is more popular than A. In that case, you need a resolution mechanism. IMHO, the fairest method is Schulze method - but it's a really complicated algorithm and hard to explain to people, so I think most people would see it as basically some black magic behind the scenes and mistrust it. Dodgson's method (invented by Lewis Carroll) is a bit easier to explain and still IMHO very fair, but perhaps still to complicated for the general electorate (maybe I underestimate people, however...) Instant runoff, however, is really easy to explain to people - after each round, you just remove the option that got the fewest top-ranking votes. Eventually you end up with a condorcet winner.

You can just skip the concept of a condorcet winner altoghether and go with instant runoff voting, but in practice, instant runoff is terrible as a sole mechanism - arguably worse than first-past-the-post. Let's say that there were three options:

A) Hard Left
B) Reasonable Middle Ground That Everyone Rather Likes
C) Hard Right

Let's say that your electorate can be divided into three groups:

Leftists (34% of the electorate): Thinks (C) is terrible (0/10), (A) is awesome (10/10), and B is rather nice (9/10)
Centrists (31% of the electorate): Think (A) and (C) are terrible (a mix of 0 or 1/10), and (B) is awesome (10/10)
Rightists (35% of the electorate): Thinks (A) is terrible (0/10), (C) is awesome (10/10), and B is rather nice (9/10)

So roughly 2/3rds of the electorate despise (A) and (C), while almost everyone likes (B). But instant runoff immediately kicks (B) out of the race. Only the extremists remain, and (C) slightly edges out (A) in the final round to win. This is a big problem with instant runoff: it tends to kick out broadly popular options ("everyone's second-favourite choice"), while leaving behind options that specific voting groups think are #1, even if said options are broadly unpopular among the broader electorate.

With a condorcet method, of course, the results are obvious. (B) utterly crushes both (A) and (C) in any one-on-one race, and so wins handily. As it should.

BUT.... this is of course totally my take, and has nothing to do with what anyone in the UK will decide!

Very informative. Thank you.

 

[Fred42]

I don't know, do you think people would find Dodgson's method too complicated? Basically, you first have to explain the concept of condorcet voting. Then you explain the resolution mechanism, which is: if there's a conflict where A is more popular than B, B is more popular than C, and C is more popular than A, then you look at each option and figure out: what's the fewest number of changes to people's vote ordering that you'd have to make in order to make this option the unambiguous condorcet winner? The option for which you have to change the fewest votes for it to come out on top is the winner.

Too complicated? Maybe the fact that Lewis Carroll came up with it would give it some cachet?

It doesn't have to be complicated. Everyone gets a first and second choice. if someone among the first choice votes has a majority, they win. Else drop everyone who didn't get 5% of the first choice votes and apply their second choice votes for those candidates who are left. If someone now has a majority, they win. Else drop the candidate with the fewest votes and utilize their second choice votes for those candidates remaining, etc. until you have someone with a majority. This is fair and would tend to lead to centrist victories. Also it allows room for small parties to be in the fray without gumming up the will of the majority. People could choose to vote for their favorite left wing or right wing or loony party and then give their second vote to a larger party that could win.

 

[KarenRei]

It doesn't have to be complicated. Everyone gets a first and second choice. if someone among the first choice votes has a majority, they win. Else drop everyone who didn't get 5% of the first choice votes and apply their second choice votes for those candidates who are left. If someone now has a majority, they win. Else drop the candidate with the fewest votes and utilize their second choice votes for those candidates remaining, etc. until you have someone with a majority.

You're describing a (very slightly tweaked) version of Instant Runoff Voting (IRV). Scroll up to see the serious problems IRV has in practice. It tends to choose options that are strongly liked by small groups but hated by large groups, rather than options that are widely liked among broad groups. It's also very vulnerable to strategic voting.

Then there's the general issue of basic fairness. Concorcet winners *should* win. If we have the following ranking of votes for options A, B and C in some poll, where 1 is the best-liked and 3 is the worst-liked

A: 3 1 3 2 1 3 3 1 3 2 1 3 3 1 1 3 3
B: 1 2 2 1 2 2 1 2 2 1 2 2 1 2 2 2 2
C: 2 3 1 3 3 1 2 3 1 3 3 1 2 3 3 1 1

Who should win? I mean, just look at it: OBVIOUSLY B should win. It doesn't even need explanation; DUH, B should obviously win. And why should it win? Because if you made the election between either A vs B:

A: 2 1 2 2 1 2 2 1 2 2 1 2 2 1 1 2 2 = 6 #1s
B: 1 2 1 1 2 1 1 2 1 1 2 1 1 2 2 1 1 = 11 #1s

.... or B vs. C:

B: 1 1 2 1 1 2 1 1 2 1 1 2 1 1 1 2 2 = 11 #1s
C: 2 2 1 2 2 1 2 2 1 2 2 1 2 2 2 1 1 = 6 #1s

... B crushes it's competition. It'd win in a 1-on-1 with against either option, so why shouldn't it win? There's no moral reason why it shouldn't win. Yet what does instant runoff choose? The competition looks like:

Round 1:
A: 3 1 3 2 1 3 3 1 3 2 1 3 3 1 1 3 2 = 6 #1s
B: 1 2 2 1 2 2 1 2 2 1 2 2 1 2 2 2 2 = 5 #1s
C: 2 3 1 3 3 1 2 3 1 3 3 1 2 3 3 1 1 = 6 #1s

B is eliminated.

Round 2:
A: 2 1 2 1 1 2 2 1 2 1 1 2 2 1 1 2 2 = 8 #1s
C: 1 2 1 2 2 1 1 2 1 2 2 1 1 2 2 1 1 = 9 #1s

Option C wins. Despite, as a reminder, in a 1 on 1 race, B would have beaten C 10 to 6 (e.g. would have won with 65% of the vote). Is that right? Is that fair?

Instant Runoff is "simple", but it leads to bad outcomes. It's many people's "introduction to voting systems other than first-past-the-post", but it's just not a good system. A condorcet criterion is a basic element of any fair multi-choice voting system.

 

[EVNow]

A) Hard Left
B) Reasonable Middle Ground That Everyone Rather Likes
C) Hard Right

Let's say that your electorate can be divided into three groups:

Leftists (34% of the electorate): Thinks (C) is terrible (0/10), (A) is awesome (10/10), and B is rather nice (9/10)
Centrists (31% of the electorate): Think (A) and (C) are terrible (a mix of 0 or 1/10), and (B) is awesome (10/10)
Rightists (35% of the electorate): Thinks (A) is terrible (0/10), (C) is awesome (10/10), and B is rather nice (9/10)

In reality, B is loathed by most. In US context, for eg.,
A) Left (say Dem nominee) - 45%
B) Corporate Sponsored (Howard Shultz) - 5%
C) Right (say Rep nominee) - 45%

In reality, if there is a person / thing who both the left & right support 9/10, they (it) will win hands down.

 

[KarenRei]

In reality, B is loathed by most. In US context, for eg.,
A) Left (say Dem nominee) - 45%
B) Corporate Sponsored (Howard Shultz) - 5%
C) Right (say Rep nominee) - 45%

In reality, if there is a person / thing who both the left & right support 9/10, they (it) will win hands down.

In my example (not yours), B is loathed by none. "3" is least favourite - B has none of those; literally nobody considers B their least favourite candidate, while many consider them . A and C are predominantly those. "2" is middle favourite. And I could easily do the same comparison where B is all 1s and 2s out of 10 candidates, or even 100 candidates; I kept it to 3 candidates for simplicity.

Instant Runoff voting is dangerous, and it only gets more dangerous the more options there are in the picture. It tends to pick extremists rather than people who are broadly liked. You may luck into it being an extremist who happens to be on your side, but you're just as likely to luck into it being an extremist on the opposite side.

And again: the condorcet criterion is just basic fairness. If any candidate / option would win in a one-on-one against every other candidate, then obviously they should win.

If you want to apply condorcet voting to your example: Do you think Schulz would win against the Dem nominee? The polling suggests strongly "No". And vs. the Republican nominee? Again, the polling suggests strongly "No". So he's not the condorcet winner. But, if in a one-on-one he'd win against both of them? Then yes he should be elected. Because the general public wants him more than either other option, and that's how democracy is supposed to work.

 

[KarenRei]

... didn't finish an edit before the edit button disappeared:

".. literally nobody considers B their least favourite candidate, while many consider them their favourite. A and C are predominantly 3, aka those 'least favourites'. "2" is middle favourite. And ..."

 

[bhtooefr]

There is also approval voting, but it's vulnerable to strategic voting as well (don't vote for a candidate you approve of because it harms the candidate you prefer... but then isn't that the case in a Condorcet system too?)

However, in your proposed matchup, in a straight 1 vote per voter per candidate counting for approval voting, and assuming 1% equals one voter (to make the math easier), and assuming one centrist each approves of A and C, and 90% of leftists and rightists vote for B:

A: receives ~35 votes
B: receives ~93 votes
C: receives ~36 votes

B wins.

(Note that you could also have a system where you rate (not rank) candidates. In that system, preference could be allocated more accurately - in your scenario (10 votes per candidate per voter, with your ratings used as-is (a mix of 50% 0/10 for both A and B, 25% 1/10 for A and B each), A receives ~348 votes, B receives ~931 votes, and C receives ~358 votes. Again, B wins.)

There's also last-past-the-post, which I've half-jokingly suggested for the US.

In first-past-the-post, in your scenario, the result is utterly dependent on strategic voting.

In last-past-the-post, ~7 voters are opposed to B, versus let's say ~50 voters being opposed A, and ~49 voters being opposed to C. However this balances out - it must add up to 100 votes, and realistically, it'll be votes against A and C mostly, most likely it'd be 1 vote against B in practice, with 50 against A, and 49 aganst C - B wins.

 

[KarenRei]

Yeah, Approval Voting is definitely preferable to Instant Runoff. But lacks the condorcet fairness criteria and, as you note, is very vulnerable to strategic voting. Still, it's decent. One of its main advantages is how simple it is to explain to people.

I studied rating (not ranking) systems, and eventually came to the conclusion that the only think that gives you over ranking is better ability to do strategic voting. I.e., it's not a good thing.

You can invert any "best" voting mechanism to a "least worst" situation - and for a number of voting mechanisms, do both at the same time - but IMHO every time I've gamed a number of them out I've never been satisfied with the results. It never fixes any of the weaknesses of a given system (e.g. strategic voting-weak systems remain weak, non-condorcet methods remain non-condorcet, reversed IRV remains vulnerable to extremism, etc), and can sometimes make them worse.

Of course, we're drifting into Strong Arrow's Theorem territory here...

 

[KarenRei]

Good table of comparing voting methods and which criteria they meet here:

Schulze method - Wikipedia

Note that Condorcet methods usually fail LIIA, Participation Consistency, Later-No-Harm, and Later-No-Help, because they're premised on lower-ranked options not carrying meaningful information.

Except, of course, they do. For example, if you believe that:

A > B > C > D

Sure, you want A to win. But the fact that you think that B > C > D also has - and should have - an impact on the outcome.

* LIIA automatically fails in any case where the voting paradox (such as A > B, B > C, C > A) exists. The whole IIA concept is shown to be flawed by the "red bus / blue bus" example. Say that you have choice between a car and a red bus, and 50% would choose the car, and 50% the red bus. Now introduce a new option, a blue bus - and presume that people don't really care about bus colour. Then you now expect that people would choose the car 50% and each of the bus colours 25% each. But IIA suggests that the ratio in choice between the car and the red bus should remain equal by the introduction of the blue bus. The problem with any "irrelevant alternatives" criteria is that in reality there is no such thing a an "irrelevant alternative.
* Participation Consistency presumes that you not voting should never hurt A - but in reality, you're not simply casting votes for A, you're also casting votes for B over C and C over D. So for example, if there was ambiguity over whether B or D was best, and that ambiguity allowed A to just barely edge past, and your vote makes it clear that B is significantly better than D, then you can resolve the ambiguity in a way that lets B edge past A.
* Later-No-Harm claims that upping the ranking on a less preferred candidate should never cause a more preferred candidate to lose. But of course it should be able to; you're boosting the overall rank of said less preferred candidate. It's been described as a property that "objectively ... is not even a desirable property with honest voters"
* Later-No-Help is the inversion of the above, that upping the ranking on a less preferred candidate should never help a more-preferred candidate to win. But upping the ranking of one less preferred candidate is akin to lowering the ranking of another - one that might be a threat to your preferred candidate. You're adding more information to the system that said candidate is problematic.

In short, Condorcet methods use all information, because all candidates are being compared to all others in one-on-one races. Any criterion which presumes that certain information that you provide (or don't provide) will be irrelevant should automatically fail.

 

[KarenRei]

Oh, and here's a great example of IRV weirdness. Take the following votes:

• 5 voters prefer A then B then C
• 4 voters prefer B then C then A
• 2 voters prefer C then A then B

C is eliminated. The race then becomes

• 5 voters prefer A then B
• 4 voters prefer B then A
• 2 voters prefer A then B

A wins.

Now, go back to our original ballot and invert everyone's opinions. Their worst options become the best, and vice versa:

• 5 voters prefer C then B then A
• 4 voters prefer A then C then B
• 2 voters prefer B then A then C

Now we run the voting. B is eliminated, leaving us with:

• 5 voters prefer C then A
• 4 voters prefer A then C
• 2 voters prefer A then C

A wins again, even though people voted in exactly the opposite manner!

IRV just plain sucks.

 

[Jetex Jim]

And now Brexit has morphed into a General Election.

Boris Johnson lies (tee hee) in a ditch and President Trump has called into Nigel Farage's phone in show and suggested a Tory/Brexit party alliance. (Honestly, President Trump took time out from being impeached to call a London radio phone in.) Thousands of 50p coins that were minted to celebrate Brexit on October 31st must now be destroyed.

When is Brexit over? We don't want it to be over. Like Game of Thrones or Star Wars we just never want it to end. It can't end its too much fun.

Admittedly, these latest plot twists strain ones credulity but the adventure continues,

 

[KarenRei]

When is Brexit over? We don't want it to be over. Like Game of Thrones or Star Wars we just never want it to end. It can't end its too much fun. Admittedly, these latest plot twists strain ones credulity but the adventure continues,

It's a shame that they're not bringing Bercow back for Brexit season 4. He was my favourite character in season 3.