Range Voting

[Wildy]

So, there are two things that IRV proponents claim: one is that it "lets you vote for your true favorite" (which I see you've caveated; good) and the other that "it helps third parties."

The problem I have with IRV is that, as the second claim begins to approach being true, the first claim becomes more and more false.

IRV "works" fine when there are two big parties and a lot of little ones; people can rank the little ones highly, but they get eliminated, and your votes get moved back to the "proper" major party that they belong too. Okay, fine.

While you're more or less right when it comes to IRV I would disagree with the bold bit. Otherwise what you're saying is that the "favourite" party of a person is the major one simply because the preferences flow to those parties.

But when the little parties grow, that breaks down.

So consider, step one; A and B are major parties, C is a (very well-performing! 25%!) 3rd party.

45%: A > B > C
30%: B > A > C
25%: C > B > A

IRV handles this well. C is eliminated, and B wins by a narrow 55% to 45%.

But then the next election comes, and C is able to get their message out, and grows in popularity among those B-voters:

45%: A > B > C
25%: B > A > C
30%: C > B > A

C is in 2nd place now! So now the winner is... A? By a better than 2:1 margin? Well that doesn't seem right. No one's opinion of A has improved (and in fact, some people moved them from 2nd to 3rd). What happened here is that C spoiled the election, which is easy to see; if you ignore C, there's still a 45%/55% split among voters, favoring B in the A vs. B contest.

Except that according to your example those 25% of people who still vote for B prefer A over C. So while C has managed to improve their opinion among the electorate and garner more first preference votes, they've failed to garner enough support among A and B voters to get them to preference C above the other major party. You can say that there's a 45/55% split between A and B, but you would also have to admit that there is a 70/30% split between C and "not C".

By the way, considering that C managed to take 5% of B's voters why is it that B voters aren't preferencing C before A?

And then what happens in the NEXT election, is voters punish C.

Do you have any real world evidence of this happening?

They go back to voting for B, because they'll be damned if they let A win again!

Or they'd just preference C before A, which is a valid option as well. If B voters preference B > C > A, and assuming that all the percentages are the same, you would end up with C winning over A.

Because the truth is you CAN'T vote for your favorite under IRV, not when it actually matters, which means IRV WON'T help third parties either.

Only for strange definitions of "favourite". I would say that the first preference would be the favourite simply because it's the first preference. Otherwise you're telling me that in the last election I voted in my "favourite" was the ALP because they were my fifth preference simply because they're a major party.

 

[Xulld]

I don't think there would be primaries anymore with this system. The parties wouldn't risk having the wrong person win .

I think there would be if only to narrow down the total number of candidates. Range voting sure sounds great, but having an unlimited number of folks on a ballet does not sound like a desirable outcome.

IMHO, a party politics needs to go the way of the dodo.

 

[bonzombiekitty]

But we don't have 3 candidates in general election usually.

Yeah, we usually don't. As someone else mentioned that is a problem caused by FPTP voting. In elections with three candidates, that third candidate can be a spoiler. To take the most recent example - in 2000 the people who voted for Nader would have, by a landslide, rather have had Gore than Bush elected, but enough of them voted for Nader that Bush won.* Going back even further, Perot helped out Clinton a ton. I am not a fan of the idea that the presence of a more preferable candidate helps out a less preferable candidate.

IRV helps to eliminate that, but range voting does an even better job.

*Note: I don't blame Nader, Gore should have been able to win even with Nader.

 

[bonzombiekitty]

I think there would be if only to narrow down the total number of candidates. Range voting sure sounds great, but having an unlimited number of folks on a ballet does not sound like a desirable outcome.

But it doesn't have to have an unlimited number. All you need to do is do something that we already do - require a certain number of signatures before being placed on the ballot.

 

[Xulld]

But it doesn't have to have an unlimited number. All you need to do is do something that we already do - require a certain number of signatures before being placed on the ballot.

Good point, what would be the advantages to different systems to limit the candidates?

 

[Almo]

Step one: get rid of the electoral college. That will help at least the presidential races.

 

[psionl0]

But then the next election comes, and C is able to get their message out, and grows in popularity among those B-voters:

45%: A > B > C
25%: B > A > C
30%: C > B > A

C is in 2nd place now! So now the winner is... A? By a better than 2:1 margin? Well that doesn't seem right. No one's opinion of A has improved (and in fact, some people moved them from 2nd to 3rd). What happened here is that C spoiled the election, which is easy to see; if you ignore C, there's still a 45%/55% split among voters, favoring B in the A vs. B contest.

This reminds me of the election result in Alfred Cove in 2005.

Alfred Cove has always been a blue ribbon liberal seat. However, in 2005, the endorsed liberal candidate (Graham Kierath) was running against a popular high profile independent (Janet Woollard) as well as the endorsed labor candidate (Michael Kane).

After the first preferences were counted, Graham Kierath had 37.9% of the vote easily outscoring Woolard at 24% and Kane at 22.8% Under FPTP Kierath would have been declared the winner even though nearly 2 out of 3 voters voted against him. However, the preferences of Kane (and a few low scoring independents) ensured that Woollard won instead.

Naturally, Kierath ran around screaming, "I was robbed!" and a number of media commentators shared that opinion. However, I believe that this is a case where democracy prevailed because the votes for the "spoiler" candidate (Kane) still counted even though he wasn't elected.

How would Alfred Cove have fared under range voting? It is difficult to tell. Graham Kierath was an unpopular minister in the former liberal government (he had lost his seat in 2001) and it is likely that those not voting for him would have given him a zero rating. We don't know how voters would have rated the unknown candidates and it is always possible that an unknown candidate with a lot of middle rankings might beat a more popular candidate with some high and some low rankings.

 

[psionl0]

This partially depends on how many scoring levels you allow. The example uses 100, but you could do it with 10, or 3, or just 2 (which is approval voting). Obviously, with 100, it'll be a little tricky to translate that between a paper format and a machine-readable format. But smaller scales would be quite easy.

This is what will make range voting difficult to sell. It requires voters to fill in a ballot paper in a radically different manner to what they do now.

You can achieve a similar effect to range voting without changing the way people fill in ballot papers by adopting some form of Borda Count where the number a voter gives to a candidate represents the relative level of the voter's disapproval of that candidate.

 

[Tsukasa Buddha]

This is what will make range voting difficult to sell. It requires voters to fill in a ballot paper in a radically different manner to what they do now.

You can achieve a similar effect to range voting without changing the way people fill in ballot papers by adopting some form of Borda Count where the number a voter gives to a candidate represents the relative level of the voter's disapproval of that candidate.

But the Borda count is crazy susceptible to strategic voting, which is why I don't like it.

 

[psionl0]

But the Borda count is crazy susceptible to strategic voting, which is why I don't like it.

I suspect this is also true of range voting.

In single member electorates, borda counting/range voting is likely to give a more accurate representation of voter intentions than AV or FPTP.

For multi-member electorates I haven't seen anything that beats STV proportional representation yet.

 

[mudlock]

This is what will make range voting difficult to sell. It requires voters to fill in a ballot paper in a radically different manner to what they do now.

You can achieve a similar effect to range voting without changing the way people fill in ballot papers by adopting some form of Borda Count where the number a voter gives to a candidate represents the relative level of the voter's disapproval of that candidate.

I don't see how range ("pick a number for each candidate") is radically different from Borda ("pick a number for each candidate"). (You could think of range as "Borda... only you can put more than one candidate in each rank, and/or skip ranks.)

Don't get hung up on "100 different possible values!" Range can be very effective with much less; 10, or 5, or just two. And with two, it's more-similar to our current voting method than any other possibility! You fill in the bubble next to the candidate you like; you're just allowed to fill in the bubble for more than one.

 

[mudlock]

I suspect this is also true of range voting.

"Susceptibility to strategic voting" is a very wishy-washy thing to define. At some point, you have to argue from some numbers.

The graphic on this page:

rangevoting.org/BayRegsFig.html

(Again, excuse the false-link; still too new to be allowed to make real ones.)

Shows the expected levels of total voter dissatisfaction using various voting methods (including Borda, range, approval ("range 2"), instant runoff, plurality, and Condorcet), using both "honest" votes and "strategic" votes, as calculated in computer simulation.

You'll note that range with honest voters has the best overall results, and that range and approval share the best strategic results; honest Borda is quite good, and is actually better than approval or than strategic range, but strategic Borda is quite a bit worse.

I suspect that voters tend towards acting strategically, which is why I like approval: really great results from a really simple system.

For multi-member electorates I haven't seen anything that beats STV proportional representation yet.

Any sort of proportional representation is great. I don't think STV is especially better than other PR methods. For consistency sake, since I favor approval, I think a PR method with ballots similar to approval ballots would be good (re-weighted approval voting, for instance; based on re-weighted range voting rangevoting.org/RRV.html )

 

[psionl0]

"Susceptibility to strategic voting" is a very wishy-washy thing to define. At some point, you have to argue from some numbers.

I did run some numbers which suggested that if two parties (only) fielded candidates in a 6 member electorate, a borda count (even without strategic voting) would not differentiate between a 60/40 vote and a 50/50 vote (ie 3 members from each party would be returned) while PR would correctly return 4 candidates from the party that got 60% of the vote.

Obviously, spending a few minutes mucking around with a spreadsheet isn't going to provide any "gospel" truths but now we have evidence that approval/borda counting can produce some pretty skewed results.

I think a PR method with ballots similar to approval ballots would be good (re-weighted approval voting, for instance; based on re-weighted range voting rangevoting.org/RRV.html )

Although some voters might give approval to some opposition candidates, approval voting for the most part works in a similar manner to FPTP and, like FPTP, it could deliver a party a "clean sweep" of a multi party electorate even though it got little more than 50% of the vote.

 

[mudlock]

Although some voters might give approval to some opposition candidates, approval voting for the most part works in a similar manner to FPTP and, like FPTP, it could deliver a party a "clean sweep" of a multi party electorate even though it got little more than 50% of the vote.

Right. Because approval isn't a PR method. Re-weighted approval is though. They're different methods, and the PR one is (of course) a bit more involved.

It's a bit of a tangent, but I guess I could take the time to explain how re-weighted approval actually works

Like in most PR methods, multiple candidates will win in each district.

Each voter approves of as many candidates as they like, just like in normal approval voting. The first candidate elected is the one with the most approvals (the most "points") just like in normal approval voting.

But then, every voter's ballot who approved the winning candidate, has the weight of their ballot cut in half; all of their remaining approvals are now worth only half a point. The second candidate elected is whoever has the most points now, after the re-weighting. Each time, every voter's ballot is re-weighted based on how many times they've "gotten their way"; if one of your approved candidates has won, you're counted as half weight; if two, a third; three, a quarter; etc.

You're welcome to do the math, but I assure you, if there's just two parties, 100% voter loyalty, and that 60/40 split you describe, you'll always have a congress that's as close to that 60/40 split as fractions will allow.

(With one winner, the 60s get the only seat. With 2, it's 50/50; 3 is 66/33; 4 is a point tie, so either 50/50 or 75/25; 5 is 60/40; 6 is 66/33; etc.)

STV isn't bad. Iit does just as good a job in the scenario we've described, although the counting procedure is even more complex. But whereas the single-winner version of approval voting is a very good voting method in its own right, the single-winner version of STV--instant runoff voting--is pretty poor, as I've discussed above.

 

[psionl0]

You're welcome to do the math, but I assure you, if there's just two parties, 100% voter loyalty, and that 60/40 split you describe, you'll always have a congress that's as close to that 60/40 split as fractions will allow.

I must be doing something wrong. I still get 3 seats each.
ETA when I use 40/3 = 13 instead of 40/3 = 16, I get the correct answer. I will still leave the maths here as it seems to be illustrative.

Round 1			Round2			Round 3	
							
A1=40	B1=60		A1=40				
A2=40	B2=60		A2=40	B2=30		A2=20	B2=30
A3=40	B3=60		A3=40	B3=30		A3=20	B3=30
A4=40	B4=60		A4=40	B4=30		A4=20	B4=30
							
B1 elected		A1 elected		B2 elected	
							
Round 4			Round 5			Round 6	
							
A2=20			A3=13	B3=20		A3=13	
A3=20	B3=20		A4=13	B4=20		A4=13	B4=15
A4=20	B4=20						
							
A2 elected?		B3 elected		B4 elected.	
							
							
Round 4			Round 5			Round 6	
							
A2=20			A2=20			A3=13	
A3=20	B3=20		A3=20			A4=13	B4=15
A4=20	B4=20		A4=20	B4 = 15			
							
B3 elected?		A2 elected.		B4 elected.

 

[mudlock]

I must be doing something wrong. I still get 3 seats each.
ETA when I use 40/3 = 13 instead of 40/3 = 16, I get the correct answer. I will still leave the maths here as it seems to be illustrative.

Round 1			Round2			Round 3	
							
A1=40	B1=60		A1=40				
A2=40	B2=60		A2=40	B2=30		A2=20	B2=30
A3=40	B3=60		A3=40	B3=30		A3=20	B3=30
A4=40	B4=60		A4=40	B4=30		A4=20	B4=30
							
B1 elected		A1 elected		B2 elected	
							
Round 4			Round 5			Round 6	
							
A2=20			A3=13	B3=20		A3=13	
A3=20	B3=20		A4=13	B4=20		A4=13	B4=15
A4=20	B4=20						
							
A2 elected?		B3 elected		B4 elected.	
							
							
Round 4			Round 5			Round 6	
							
A2=20			A2=20			A3=13	
A3=20	B3=20		A3=20			A4=13	B4=15
A4=20	B4=20		A4=20	B4 = 15			
							
B3 elected?		A2 elected.		B4 elected.

?? It looks like both your alternatives say to elect 4 Bs and 2 As when there are 6 winners; not 3 and 3 like you say in your text. I think you got it right

 

[psionl0]

?? It looks like both your alternatives say to elect 4 Bs and 2 As when there are 6 winners; not 3 and 3 like you say in your text.

Yes, I initially made a mistake (hence the ETA). It is 4 and 2 as you stated.

I think you got it right

Thanks. You provide some interesting discussion points. I still don't see this as an improvement on STV-PR but for multi-member electorates that have been using FPTP, this would definitely be a way to go.