What does Australia show?

[david.cou...@drake.edu]

Hi! I've been following Election Science for a while now--thanks for your work! I completely agree with Clay's recent post that the gravity of climate change is one example of how critical voting reform is in the US. I worry that we who want to see reform spend a lot of time fighting each other over minor issues when there are a large number of systems that would be a significant improvement over plurality voting. For example, Smith argued in a post that Greens are "suicidally stupid" for supporting IRV, because he thinks IRV is bad for third parties, and AV is clearly better. Australia is always brought up as a case in point, because the lower house (where IRV is used) has very little third party representation (if you accept the claim that the Liberals, Nationals, and Nat-Libs are functionally the same party, which is a bit of a stretch but the subject for another discussion).

Now, there are two reasons why minor parties could be doing so badly in Australia's House. One is the claim Smith makes a lot which is that IRV drives strategic voting and this leads to two-party dominance. The other possibility is that minor parties don't actually make up a majority of many districts, and so no voting system--including AV--would give them more representation in a winner-take-all system. 

It seems easy enough to test whether Australians are voting strategically in their IRV elections: just ask if there are a lot of people who are voting for major parties in House races but minor parties in Senate elections where they have proportional representation (so they have every incentive to vote for their first choice). 

So, I checked the Australian House election of 2013. Labor plus Coalition earned about 78% of the first-place vote. http://en.wikipedia.org/wiki/Australian_House_of_Representatives. Your working theory states that this is the result of a long process of strategic voting where voters have learned that it is a mistake to vote for third party candidates. Sounds pretty convincing, as that's a pretty poor showing for third parties (assuming away the coalition thing). So, in the Senate, where voters aren’t encouraged to vote strategically, how do the major parties fare? 76.4%. 

That’s not a lot of strategic voters.

In my mind, that’s the whole “strategic voting” argument right there. Australians have had 100 years to learn that they should be voting strategically using IRV, but they absolutely don't. Isn't this proof that IRV does *not* lead to strategic voting and hurt third parties?

-- Dave

[Clay Shentrup]

David,

First of all, regardless of why IRV leads to two-party domination, it's clear that it does. Whereas traditional non-instant runoffs (TTR) have led to multipartism.

http://scorevoting.net/TTRvIRVstats.html

Warren discussed potential explanations for this here:

http://scorevoting.net/TTRvIRVreasons.html

Score/approval have even stronger anti-duopoly properties than top-two runoffs. E.g.

http://scorevoting.net/NurseryEffect.html

On Sunday, December 7, 2014 8:09:52 PM UTC-8, david.cou...@drake.edu wrote:

The other possibility is that minor parties don't actually make up a majority of many districts, and so no voting system--including AV--would give them more representation in a winner-take-all system. 

Why does this just happen to be the case in IRV-using countries but not in the TTR-using countries? If it's not due to inherent properties of IRV, then it's an awfully odd coincidence.

It seems easy enough to test whether Australians are voting strategically in their IRV elections: just ask if there are a lot of people who are voting for major parties in House races but minor parties in Senate elections where they have proportional representation (so they have every incentive to vote for their first choice).

That assumes there isn't significant tactical voting in the proportional races.

Isn't this proof that IRV does *not* lead to strategic voting and hurt third parties?

IRV empirically does hurt third parties, regardless of whether tactical voting is the cause. And no, this is not proof.

Consider this claim:

> In fact, there is evidence that fewer than half as many voters vote tactically in Australia (under AV) as in the UK (under FPTP).1

Source: https://www.psa.ac.uk/sites/default/files/TheAlternativeVoteBriefingPaper.pdf

Plurality voting suffers a huge amount of tactical voting, so this is still a lot.

[Clay Shentrup]

Also, it's informative to call the Australian political parties and talk to them, like I did in June of 2010.

https://groups.google.com/d/msg/electionscience/rpD4hOohODw/aESz8ewO_6sJ

[david.cou...@drake.edu]

Clay,

Thanks for the thoughts. Let's go through this.

On Sunday, December 7, 2014 10:43:49 PM UTC-6, Clay Shentrup wrote:

David,

First of all, regardless of why IRV leads to two-party domination, it's clear that it does. Whereas traditional non-instant runoffs (TTR) have led to multipartism.

http://scorevoting.net/TTRvIRVstats.html

I'm not sure what you mean by "it's clear that it does." First, there's a very small sample of countries using IRV--only three, whereas there are a larger number of TTR countries which have mixed results, some multiparty some not. Even if the three countries were characterized correctly, it would be a small enough sample to be hardly conclusive. But, let's go through each one. I don't know much about Fiji so I won't spend too much time on that, but check this out from the 2014 election: http://en.wikipedia.org/wiki/Fijian_general_election,_2014 They are now using a proportional system, and still over 87% of voters chose a candidate from the two major parties. Here again it looks very clear that IRV wasn't suppressing some huge desire for third parties, the fact is that people don't support them even when they have a system designed to give them representation. You could make the argument that IRV built this system, however, so I'll concede that Fiji is inconclusive.

But you're going to have to help me understand the argument for Ireland. The two national parties (from the perspective of sending delegates to the Eurpoean Parliament) are Fine Gael and Sinn Féin. Fine Gael is clearly the major party in Ireland, with 86 seats in either of the two houses of Parliament. Labor has 45, and Fianna Fáil has 34, Sinn Féin 17, and then smaller parties. Now, look at the presidency. The current president is from Labor. The previous president was from Fianna Fáil. The president before that was from a coalition: Labor, Workers, and Independents. Before that Fianna Fáil again. So the argument that Irish presidential races are two-party rests on the extremely dubious contention that the two major parties are Labor and Fianna Fáil, and ignoring the coalition in the 90s. That makes no sense. By any measure, Fine Gael would be one of the major parties in Ireland, and they ran a candidate in 2011 but lost. The second-place candidate for president in 2011 was independent, with Sinn Fein third. In 1997, the second-place candidate was from Fine Gael, with an independent third. In 1990 an independent won from the coalition, with Fianna Fáil second and Fine Gael third. To my eyes, The Irish presidency clearly demonstrates that IRV has *not* led to two party dominance, because you have non-top-two parties not just running, but *winning*. And it's *not* because Fine Gael doesn't participate, and it's *not* because either Labor or Fianna Fáil are just copies of Fine Gael at the presidential level (when Fine Gael loses, their voters don't all choose Fianna Fáil, for example). It's just that voters have not selected the Fine Gael candidate recently. The Irish presidency is actually a counter-example to the claim that IRV stifles minor parties!

And then there's Australia, which also shows quite clearly that first-place votes in IRV parallel almost exactly first-place votes using proportional voting. So let's talk about that.

 

It seems easy enough to test whether Australians are voting strategically in their IRV elections: just ask if there are a lot of people who are voting for major parties in House races but minor parties in Senate elections where they have proportional representation (so they have every incentive to vote for their first choice).

That assumes there isn't significant tactical voting in the proportional races.

Yes, I agree that I am making that assumption. Hopefully you will agree that I am making pretty much *only* that assumption, and so for Australia--the only large country with a long history of IRV *and* a nice experiment with two different voting systems at the same level of government--the question of whether IRV drives tactical voting in the real world comes down to whether proportional voting drives tactical voting in the real world. 

So here I'm going to need your help. I consider it to be on-the-face obvious that there is no better system for third parties than proportional representation, and that there is no voting system that is *less* vulnerable to strategic voting (at least in the case where third-parties pass the hurdle of viability, which they clearly do in Australia). So I'd ask your help here in explaining why tactical voting is likely to be a real problem in Australia.

And here's where I'm coming from on this. I consider myself to be a savvy voter, and I can't personally think of any reason why I wouldn't choose my top-choice party in a representational system if I expected my top choice to meet the minimum threshold. On the other hand, I know for a fact that some voters would vote strategically using AV. How can I be so sure? Because I've done it, more or less. We have plenty of elections in Iowa where we are asked to choose "up to 5" candidates for 5-seat councils, for example. I often engage in "bullet voting" where I choose only my favorite candidate even though I consider others to be acceptable. Is this stupid? Sure, you can argue that it is. *BUT* it is also strongly encouraged by the parties and lots and lots of people do it. So, if you want to demonstrate that IRV is bad for third parties because of strategic voting, then in my case you're going to have to demonstrate that AV is significantly less likely to get people to vote strategically than *representational* elections, and I think that's an impossible task. Because I think the evidence is that people don't actually vote strategically using IRV, even though your analyses suggest they *should.*

But I am interested to hear your reply, because I have certainly been wrong about things in the past and might well be wrong about this.

Warren discussed potential explanations for this here:

http://scorevoting.net/TTRvIRVreasons.html

Score/approval have even stronger anti-duopoly properties than top-two runoffs. E.g.

http://scorevoting.net/NurseryEffect.html

I agree that top-two runoffs don't have strong anti-duopoly properties, largely because turnout in the primaries in the US is so low (I don't know about the other countries). I very much hope that the nursery effect is a real thing, and it may be, but there is no actual real-world data on that, so for now this is a conjecture rather than the fact that you are presenting it as.

 

On Sunday, December 7, 2014 8:09:52 PM UTC-8, david.cou...@drake.edu wrote:

The other possibility is that minor parties don't actually make up a majority of many districts, and so no voting system--including AV--would give them more representation in a winner-take-all system. 

Why does this just happen to be the case in IRV-using countries but not in the TTR-using countries? If it's not due to inherent properties of IRV, then it's an awfully odd coincidence.

Right, but my point is that it *doesn't* actually happen in IRV-using countries, so let's get on the same page on that question before getting into this discussion.  

Isn't this proof that IRV does *not* lead to strategic voting and hurt third parties?

IRV empirically does hurt third parties, regardless of whether tactical voting is the cause. And no, this is not proof.

Again, I am now convinced that there is actually no evidence from elections that IRV is bad for third parties. I think Ireland suggests that it is *good* for third parties. So help get me right on that point first.

 

Consider this claim:

> In fact, there is evidence that fewer than half as many voters vote tactically in Australia (under AV) as in the UK (under FPTP).1

Source: https://www.psa.ac.uk/sites/default/files/TheAlternativeVoteBriefingPaper.pdf

Plurality voting suffers a huge amount of tactical voting, so this is still a lot.

Well, I don't know if it's a lot, because I don't know what they meant by "fewer than half." So, I looked up the citation: Thomas Gschwend, “Comparative Politics of Strategic Voting: A Hierarchy of Electoral Systems”, paper

presented at the annual meeting of the Midwest Political Science Association, Chicago, 20–23 April 2006. Here it is: http://www.sowi.uni-mannheim.de/gschwend/pdf/papers/gschwend-midwest06.pdf . Gschwend finds 3.3% strategic voting in Australia, which seems totally believable to me given the Senate/House results, and the fact that even strategic voters aren't always going to agree on what the best strategy is.

So you brought up the citation--do you agree that we're talking about something on the order of 3% strategic voting with IRV if we're looking at this empirically? Is that what we're arguing about here?

[Warren D Smith]

>Now, there are two reasons why minor parties could be doing so badly in Australia's House.... The other possibility is that minor parties don't actually make up a majority of many districts, and so no voting system--including AV--would give them more representation in a winner-take-all system.

REPLY: I dispute the implication it is necessary for a candidate to
have a majority to cause him to win. Au contraire, candidates can and
do and should win without a majority, and often no candidate has a
majority.
And furthermore, the "majority" that one or the other of Australia's 2
major parties "have" is often illusory and a side effect of dishonest
voting.

> I checked the Australian House election of 2013. Labor plus Coalition earned about 78% of the first-place vote. http://en.wikipedia.org/wiki/Australian_House_of_Representatives. Your working theory states that this is the result of a long process of strategic voting where voters have learned that it is a mistake to vote for third party candidates. Sounds pretty convincing, as that's a pretty poor showing for third parties (assuming away the coalition thing). So, in the Senate, where voters aren't encouraged to vote strategically, how do the major parties fare? 76.4%.

REPLY:
I dispute the claim "in the Senate, voters aren't encouraged to vote
strategically."
As far as I can tell, the party-advice "how to vote cards" generally,
or always, give the same advice to voters, for both the senate and the
house.
Can you prove the contrary?

But anyhow, in a 78% huge landslide, which is extremely unusual, I rather doubt
anything you could say would convince me of anything, since I'd fully
expect the landslide side to prevail in such circumstances. For
comparison here in the USA,
there has never been anywhere near a 78% landslide for any
presidential election ever.


--
Warren D. Smith
http://RangeVoting.org <-- add your endorsement (by clicking
"endorse" as 1st step)

[Warren D Smith]

>Now, there are two reasons why minor parties could be doing so badly in Australia's House.... The other possibility is that minor parties don't actually make up a majority of many districts, and so no voting system--including AV--would give them more representation in a winner-take-all system.

REPLY: I dispute the implication it is necessary for a candidate to
have a majority to cause him to win. Au contraire, candidates can and
do and should win without a majority, and often no candidate has a
majority.
And furthermore, the "majority" that one or the other of Australia's 2
major parties "have" is often illusory and a side effect of dishonest
voting.

> I checked the Australian House election of 2013. Labor plus Coalition earned about 78% of the first-place vote. http://en.wikipedia.org/wiki/Australian_House_of_Representatives. Your working theory states that this is the result of a long process of strategic voting where voters have learned that it is a mistake to vote for third party candidates. Sounds pretty convincing, as that's a pretty poor showing for third parties (assuming away the coalition thing). So, in the Senate, where voters aren't encouraged to vote strategically, how do the major parties fare? 76.4%.

REPLY:
I dispute the claim "in the Senate, voters aren't encouraged to vote
strategically."
As far as I can tell, the party-advice "how to vote cards" generally,
or always, give the same advice to voters, for both the senate and the
house.
Can you prove the contrary?

AND ANYHOW... in a 78% huge landslide, which is extremely unusual, I

rather doubt
anything you could say would convince me of anything, since I'd fully
expect the landslide side to prevail in such circumstances. For
comparison here in the USA,

there has never been anywhere even close to a 78% landslide for any
presidential election ever. Biggest landslides we ever got were about
61%.

[david.cou...@drake.edu]

On Monday, December 8, 2014 12:24:48 PM UTC-6, Warren D. Smith (CRV cofounder, http://RangeVoting.org) wrote:

>Now, there are two reasons why minor parties could be doing so badly in Australia's House....  The other possibility is that minor parties don't actually make up a majority of many districts, and so no voting system--including AV--would give them more representation in a winner-take-all system.

REPLY: I dispute the implication it is necessary for a candidate to
have a majority to cause him to win.  Au contraire, candidates can and
do and should win without a majority, and often no candidate has a
majority.
And furthermore, the "majority" that one or the other of Australia's 2
major parties "have" is often illusory and a side effect of dishonest
voting.

What I mean here is that in a winner-take-all system, somehow we have to determine who "got the most votes." You are correct that I sloppily referred to that as a "majority" when it could be a plurality. My point was that if people who support minor parties only make up 22% of a district then they will not--and should not--win elections where 40% of voters support one of the major parties. Sometimes major parties win because they have popular positions, and that appears to be what is happening in Australia.

 

> I checked the Australian House election of 2013. Labor plus Coalition earned about 78% of the first-place vote. http://en.wikipedia.org/wiki/Australian_House_of_Representatives. Your working theory states that this is the result of a long process of strategic voting where voters have learned that it is a mistake to vote for third party candidates. Sounds pretty convincing, as that's a pretty poor showing for third parties (assuming away the coalition thing). So, in the Senate, where voters aren't encouraged to vote strategically, how do the major parties fare? 76.4%.

REPLY:
I dispute the claim "in the Senate, voters aren't encouraged to vote
strategically."
As far as I can tell, the party-advice "how to vote cards" generally,
or always, give the same advice to voters, for both the senate and the
house.
  Can you prove the contrary?

Are you saying that the party-advice cards are encouraging their voters to vote for them (to be expected), or are they telling voters something akin to "Here's why you should vote against your first-choice party in the Senate for strategic reasons"? My argument is that the results in a proportional voting election ought to represent voter preferences fairly well. 

 

But anyhow, in a 78% huge landslide, which is extremely unusual, I rather doubt
anything you could say would convince me of anything, since I'd fully
expect the landslide side to prevail in such circumstances.  For
comparison here in the USA,
there has never been anywhere near a 78% landslide for any
presidential election ever.

I think you misunderstood my point. If you look at the Australian Senate elections, what you call the "two" major parties--Labor and the Nat-Lib coalition--combine to occupy over 76% of the seats, because that's about the fraction of the vote they received. In the Australian House elections, 78% of the voters chose either Labor or Nat-Lib first on their IRB ballots, with 22% choosing minor parties. The point is that there is no evidence for IRB driving strategic voting--it seems from the results that about the same number of voters are selecting minor parties first in IRB as are selecting minor parties in proportional representation. Thus, you are mistakenly attributing the lack of minor parties in the House to IRB, when in fact the problem would remain with any winner-take-all system, because minor parties just aren't popular enough to win many districts.

[david.cou...@drake.edu]

To clarify a little more, I can imagine a situation where minor parties should win even if they only have a small amount of preferential support--for example if they are acceptable compromise candidates for a majority. This does not appear to be what's going on in Australia, though. The Greens--the largest minor/non-coalition party in the Senate, are not seen as centrists between Labor and the Nat-Libs.

[Warren D Smith]

> My point was
> that if people who support minor parties only make up 22% of a district
> then they will not--and should not--win elections where 40% of voters
> support one of the major parties. Sometimes major parties win because they
> have popular positions, and that appears to be what is happening in
> Australia.

--maybe a minor should win even in this case. What often happens in
Australia is something like this:
#voters
46% NatLib top, Labour bottom (of 7 ranked in all)
46% Labour top, NatLib bottom (of 7 ranked in all)
8% Somebody else top.

Now what are we to make of this? First of all, this is definitely NOT
honest voting
by the Australians. OBVIOUSLY it is bullshit to claim NatLib is the honestly
favorite party, AND the honestly most hated party, at the same time,
ditto Labour.
What we are seeing here, is a huge distortion. These are not honest rankings.

Second, if, say, 90% of the voters ranked Green 2nd (which I am not
necessarily claiming, but suppose it were so) out of the 7, then in
that circumstance I would say Green probably should have been elected,
and to hell with your mutterings about 22%
being too small.

> Are you saying that the party-advice cards are encouraging their voters to
> vote for them (to be expected), or are they telling voters something akin
> to "Here's why you should vote against your first-choice party in the
> Senate for strategic reasons"? My argument is that the results in a
> proportional voting election ought to represent voter preferences fairly
> well.

--the party advice cards are ranking all parties in an order
recommended by whatever
party prints those cards.

You can see what those cards said, here:
http://www.rangevoting.org/AusAboveTheLine07.html

Not only that, but in Australia you often can, instead of entering
your whole rank-ordering
of all 7, just push one button to "enter the whole ordering
recommended by Labour"
and save yourself time. I.e. the "how to vote cards" are not only
supplied to voters as advice, they in fact are supplied to the polls
as pre-programmed labor-saving voter options.

The Labour-supplied how to vote card always ranks Labour top and
NatLib bottom or nearly; and is by definition strategic.

[Clay Shentrup]

Warren,

He's asking if/why you think e.g. Green voters are tactically ranking Labor top. He's not asking about burial.

[Warren D Smith]

I do not know whether "Green voters are tactically ranking Labor top."
Assuming they are not, i.e. the percentage of true Green-top voters
genuinely is small in Australia -- then so what?
Should that prevent Greens from winning?
I repeat: Greens still may well deserve victory in such a scenario.
I repeat my example:

#voters
46% NatLib top, Labour bottom (of 7 ranked in all)
46% Labour top, NatLib bottom (of 7 ranked in all)
8% Somebody else top.

And... 90% of the voters ranked Green 2nd out of the 7.

I also suggest that you note that the only time a Green won an IRV
seat in the Australian House, it was when one of the 2 Major Parties happened by
fluky circumstances not to run any candidate.
So then the Green effectively became a "replacement major party"
going head to head versus the other (real) major, and sure enough the
Green candidate won.

[Clay Shentrup]

Warren,

While scenarios like that are theoretically possible, this doesn't counter David's argument that, "maybe IRV isn't electing third party candidates because there really isn't enough support for them—not because of anything about IRV." The race you mention, where a Green won because a major party candidate was missing, is only one counter-example.

[Clay Shentrup]

On Monday, December 8, 2014 9:46:30 AM UTC-8, david.cou...@drake.edu wrote:

I'm not sure what you mean by "it's clear that it does." First, there's a very small sample of countries using IRV--only three, whereas there are a larger number of TTR countries which have mixed results, some multiparty some not. Even if the three countries were characterized correctly, it would be a small enough sample to be hardly conclusive.

Actually, Warren calculated the probability that this would have happened purely by chance. It was very tiny.

[Warren D Smith]

Well, "maybe IRV isn't electing third party candidates because there
really isn't enough support for them--not because of anything about
IRV."
might be true.

But I see no evidence it is true. David's so-called evidence did not
impress me.
It could be argued I have not convinced in the other direction either.

Well, sorry. I've merely pointed out that his evidence, wasn't
convincing. I have not gone
further and presented a convincing case he's actually completely wrong.

If you wanted to make the case third party guys ought to be winning more often
(or not), then one approach would be to do score-style polls inside
Australia, see if a lot of Greens would have won the elections if
they'd used score-voting elections.
I do not know what the result of that exercise would be. Perhaps such polls
have already been done -- if so all you'd need to do is to find the poll data to
answer that.

[Jim Mueller]

All of the voy
Ting systems that have been discussed have the major flaw that they intentionally deny all percentages of the voters the representation of their choice.
I urge you to support Direct Representation.
Let each candidate represent the people who voted for him or her.
The candidates/representatives then have weighted votes equal to the number of people who voted for them.
Every voter gets the representation of their choice. All candidates get rewarded proportional to their popularity. No gaming or strategies are necessary.
Independents and minor party candidates become viable giving the electorate the wide range of choices that they are looking for.
Legislation will more closely reflect the will of the majority.

Sent from my iPad

> --
> You received this message because you are subscribed to the Google Groups "The Center for Election Science" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to electionscien...@googlegroups.com.
> For more options, visit https://groups.google.com/d/optout.

[Jim Mueller]

Giving all of the political power to the one person who gets the most votes is so primitive.
Adopt Direct Representation and let all of the candidates/representatives serve the people who voted for them.

Sent from my iPad

[Warren D Smith]

Re trying to find some Australian score-style polls...

The problem is, if I were an Australian pollster, I would avoid
score-style polling
because I'd want to best-emulate future (rank-order elections).
So they probably are rarely done there,
even though quite commonly done nowadays in USA.
I tried some google searching and failed to find any,

I did find this mildly related thing which is interesting:

http://www.visionandvalues.org/2010/09/could-an-atheist-be-elected-president-a-look-down-under/
Former PM Julia Gillard is a declared atheist. Also a woman. Also single.
Supposedly all traits that would have made it impossible for her to be
elected US president.

This blog post by a pollster (unusually) did some approval-style
polling in Australia:
http://jwsresearch.com/news_files/chisholm-and-melbourne-ports-safe-but-isaacs-set-to-fall.pdf

[Warren D Smith]

> http://www.visionandvalues.org/2010/09/could-an-atheist-be-elected-president-a-look-down-under/
> Former PM Julia Gillard is a declared atheist. Also a woman. Also single.
> Supposedly all traits that would have made it impossible for her to be
> elected US president.

--she also dyes her hair, a trait which definitely does NOT prevent
you from being US president...

[Clay Shentrup]

The question isn't whether Greens should be winning, but whether there should be more than two successful parties.

A related question: why doesn't e.g. the ALP break up into two smaller parties representing its two major factions? Perhaps because of having PR in the senate.

[Warren D Smith]

http://blogs.abc.net.au/antonygreen/2011/09/do-australians-follow-how-to-votes.html

claims to provide info about how many Australians just follow "how to
vote" advice.
(This info is not necessarily to be relied on, it contradicts other
info saying over 90%.
May depend on which election.)

But anyhow, clearly a substantial fraction of the Australian
electorate uses how-to-vote
orderings completely verbatim, say somewhere between 40% and 100%.

[david.cou...@drake.edu]

I also suggest that you note that the only time a Green won an IRV
seat in the Australian House, it was when one of the 2 Major Parties happened by
fluky circumstances not to run any candidate.
So then the Green effectively became a "replacement major party"
going head to head versus the other (real) major, and sure enough the
Green candidate won.

That's just not true. Adam Bandt won the seat in 2010. In the 2010 election, he ran against Cath Bowtell of Labor (who got 38% of the primary vote) and Simon Olson of the Liberal party (who got 21% of the primary vote). http://www.abc.net.au/elections/federal/2010/guide/melb.htm

In 2013, he won again against Cath Bowtell again and Sean Armistead (Lib), despite the fact that the Libs shifted on their voter cards from having voters rank Labor last to having them rank Greens last.

http://www.abc.net.au/news/federal-election-2013/guide/melb/ 

[david.cou...@drake.edu]

It would be tiny if you assumed random distributions of party preferences. By definition the two "major parties" are going to be the largest parties in any country, so the question is what do you assume about the distribution of the remaining voters. I'd be interested to see how the calculations were done, because you would have to make some pretty significant assumptions. Anyway, as I said before, I think Ireland is an obvious counter-example because you have minor parties winning the Presidency.

[Warren D Smith]

On 12/9/14, david.cou...@drake.edu <david.cou...@drake.edu> wrote:
>
>>
>>
>> I also suggest that you note that the only time a Green won an IRV
>> seat in the Australian House, it was when one of the 2 Major Parties
>> happened by
>> fluky circumstances not to run any candidate.
>> So then the Green effectively became a "replacement major party"
>> going head to head versus the other (real) major, and sure enough the
>> Green candidate won.
>>
>
> That's just not true. Adam Bandt won the seat in 2010. In the 2010
> election, he ran against Cath Bowtell of Labor (who got 38% of the primary
> vote) and Simon Olson of the Liberal party (who got 21% of the primary
> vote). http://www.abc.net.au/elections/federal/2010/guide/melb.htm

--I stand corrected. Sorry, I seem to have been confused and/or out
of date -- I think
I was thinking about Michael Organ, the first Green to win an
Australian House seat?
http://en.wikipedia.org/wiki/Michael_Organ

Anyhow, the fact this Green, Bandt, managed to win versus both majors
only helps the argument I was trying to make, I guess.

[david.cou...@drake.edu]

I just want to clarify what I am arguing here, because I don't think it's a matter of opinion at all. The hypothesis that you put forward is that IRV will lead to two-party domination *because* people will not vote for their preferred candidate in the first round if that person is not from one of the two major parties. They will worry that their candidate does not have a chance, and so their candidate will not have a chance. The hypothesis holds if there are a significant number of people who prefer minor parties but don't rank them first. The real-world test of the hypothesis is to use a population of people who simultaneously indicate their preference in an election, but also rank candidates using IRV. The hypothesis fails the test in real voting in Australia, because you have about the same number of people voting for minor parties in the IRV election as in the proportional election, indicating that first choice votes, at the very least, are not being made "strategically," but instead are identifying real preferences.

This is not a matter of you have your opinion and I have my opinion. This is a matter of a real-world test of a hypothesis, and the hypothesis failing. In science, at that point we either reject or revise the hypothesis.

-- Dave

[Warren D Smith]

On 12/9/14, david.cou...@drake.edu <david.cou...@drake.edu> wrote:

> It would be tiny if you assumed random distributions of party preferences.
> By definition the two "major parties" are going to be the largest parties
> in any country, so the question is what do you assume about the
> distribution of the remaining voters. I'd be interested to see how the
> calculations were done, because you would have to make some pretty
> significant assumptions. Anyway, as I said before, I think Ireland is an
> obvious counter-example because you have minor parties winning the
> Presidency.

--In Ireland the "presidency" is an unimportant largely symbolic
position, the PM has the real power.
Second, in Ireland only 2 parties have ever won the presidency...
It always was the major FF party, except that in 1990 and 2011,
the "third" party, Labour, won.

The 1990 election was somewhat fluky and pathological:
http://rangevoting.org/Ireland1990.html

In the 2011 election, FF did not run any candidate, so Labour was
effectively a major party, and duly won.

[Warren D Smith]

On 12/9/14, david.cou...@drake.edu <david.cou...@drake.edu> wrote:

> I just want to clarify what I am arguing here, because I don't think it's a
>
> matter of opinion at all. The hypothesis that you put forward is that IRV
> will lead to two-party domination *because* people will not vote for their
> preferred candidate in the first round if that person is not from one of
> the two major parties. They will worry that their candidate does not have a
>
> chance, and so their candidate will not have a chance.

--I would rephrase that. Since it is my hypothesis, maybe I should
word it, not you.
The hypothesis I have, is IRV yields 2-party domination.
I am not necessarily knowing everything about why, I am just claiming it is so.
It is an experimentally confirmed fact. In Australia, world's most IRV country,
3 election cycles for House (IRV) in a row, 2001-2004-2007
the total number of races was 150*3=450
and the total number of of 3rd party winners was zero, despite
typically 7 candidates
for each seat. That is clear, massive, unquestionable, 2-party domination.

In Ireland, presidency is the only IRV seat. So there is not a lot of data.
The whole history of IReland provides less data than just 1 Australia
House cycle.
But anyhow: only 2 parties have ever won the presidency.

In Fiji, they had IRV for a while, and 2-party domination looked to be
setting in, but then their government was overthrown by a coup.

In Papua New Guinea, they recently introduced (2007 was the first time
it was used)
an IRV-like system they call "limited preferential vote." It seems to
be IRV3, you
rank order your top 3 choices only. Also used in San Francisco and
not a good voting system, especially with PNG now having over 20
parties in office.
I do not think it is ok to draw conclusions from PNG because 2-party
domination has not had time to set it yet. In the USA it took about
70 years for it to set in. And it might be
IRV3 and IRV differ about 2-party domination.

In short, my hypothesis is supported by the experimental evidence. Period.
As you say,

"This is not a matter of you have your opinion and I have my opinion. This
is a matter of a real-world test of a hypothesis, and the hypothesis
failing. In science, at that point we either reject or revise the
hypothesis."

>The hypothesis holds
> if there are a significant number of people who prefer minor parties but
> don't rank them first.

--NO, I did NOT make that claim. That number of people might
be large in Australia, or it might not, I do not know. You also do not know.
I did not claim to know. I also did not claim this question had any
relevance whatever.
But there is some limited evidence we discussed that the number of Green-liking
voters in Australia is significant, certainly it is far larger than
their fraction of House seats.

I do, however, make several other claims, which I actually do make:
1. that Greens should win even in some circumstances where NOBODY
ranks them top, 2. Australia in House is 2-party dominated, and that
is the fault of IRV. I am not the only person claiming that, it is
widely claimed all over Australia.
3. The reason IRV leads to 2-party domination in Australia, is,
that a large fraction of voters vote in the style
A>essentially all others>B
or
B>essentially all others>A
where A and B are the two major parties. These behaviors also are
recommended by the two major parties over and over again. It is a
theorem that if at least 75% behave
this way, then a third-party can never win. Ever. This 75% threshold
is easily exceeded in Australian reality. That is why third parties
essentially never win in Australian IRV elections.
4. WHY the voters behave that way, is something I do not know. I suspect it
has something to do with the how-to-vote cards, or naive strategic
notions, or herd
jungle instincts.
All I know is, they do it. Exactly why they do it, I cannot say.
5. I also claim this massive behavior by the Australian voters, is
OBVIOUSLY massively dishonest.

So we know: in Australia, there is massive dishonest/strategic voting
going on. We know it causes 2-party domination under IRV. We do not
know, or at least I do not, exactly why the Australian voters do what
they do.

[david.cou...@drake.edu]

3. The reason IRV leads to 2-party domination in Australia, is, 
that a large fraction of voters vote in the style 
A>essentially all others>B 
or 
B>essentially all others>A 
where A and B are the two major parties. These behaviors also are 
recommended by the two major parties over and over again.  It is a 
theorem that if at least 75% behave 
this way, then a third-party can never win.  Ever.  This 75% threshold 
is easily exceeded in Australian reality.   That is why third parties 
essentially never win in Australian IRV elections. 

Ah! I think we have reached the crux of the issue. Let me generalize the theorem: If at least 75% of the electorate select either A or B first, then it doesn't matter *how* they vote down ballot one bit, either A or B will win, always. 

The structure

 A > all others > B

B > all others > A

provides third parties with the best chance to win, not the worst. Please correct me if I'm wrong about this last part, because I may be missing one of the peculiarities of IRV elimination, but I don't think I'm wrong about the 75% thing.

If 75% of voters *prefer* either of the major parties, then you have two party popularity, not two party dominance caused by a distorted voting system. Again I claim that we know what voters prefer because they have little incentive to vote strategically in the proportional election. My point is that there is nothing *wrong* with this if that's what voters want, and moreover there's nothing *correctible* about this if you are stuck with winner-take-all elections, but there's where we disagree.

Now, I think I can see where you're coming from with AV. In a situation where two major parties hate each other, it might be possible for a minor party to win because some voters would use the "anything but" method of voting, and select nearly everyone on the ballot except the hated other major party. In this case, both A voters and B voters would choose C, and C would win. So, yes, I can imagine a world in which AV led to the selection of an unpopular third party (or, at least, a third party that was the first choice of a small group of people). 

So, now comes the part where IRV is at a disadvantage vis-a-vis AV, because we know what the results are with IRV. But here's what I'm absolutely certain would happen with AV, and I have that level of certainty because it happens in AV-like elections with multiple seats open. The major parties would encourage their voters to use bullet voting, not real approval voting. Now, voters may ignore their advice, but it wouldn't take much bullet voting before a minor party candidate has no chance to win. Of course, you can argue all you want that it would be *irrational* for voters to follow party advice on this topic, but I'd argue that it's *irrational* for voters to follow party advice on candidate order in an IRV election. And yet, they often do, as you point out. 

-- Dave

[Warren D Smith]

> Ah! I think we have reached the crux of the issue. Let me generalize the
> theorem: If at least 75% of the electorate select either A or B first, then
>
> it doesn't matter *how* they vote down ballot one bit, either A or B will
> win, always.

--yes. With IRV.

> The structure
> A > all others > B
> B > all others > A
> provides third parties with the best chance to win, not the worst.

--huh? Zero chance does not seem to be best. It seems worst.

> If 75% of voters *prefer* either of the major parties, then you have two
> party popularity, not two party dominance caused by a distorted voting
> system.

--Well. OK, if we had the new structure (David's)
A>all others
B>all others
and we knew those top rankings were honest, then that'd be one thing.
But we actually have the different structure (Warren's)
A > all others > B
B > all others > A.
Now I agree as far as IRV is concerned these two structures yield
equivalent results. But as far as many other voting methods
are concerned, they are not equivalent.
Now "Warren's stricture" which happens in Australia to good
approximation, plainly
involves massively dishonest voting, which hence is plainly very
prevalent in Australia.
If, however, we partly covered our eyes and only noticed David's (sub)structure,
then it would no longer be obvious Australia has massively dishonest voting --
those could plausibly have been their honest feelings! But when we
uncover our eyes, and see the whole picture, ("Warren's structure") it
IS obvious
Australia's voting really is massively dishonest.

Where is that dishonesty? Is it located near the bottom or the top
of the rankings? I do not know. All I know is, the whole thing is
massively dishonest.
It is like, if somebody claims 2+7=53. You do not know which number
they got wrong,
but you definitely know they were wrong.

> Again I claim that we know what voters prefer because they have
> little incentive to vote strategically in the proportional election.

Excuse me, but:
1. what the hell gives you that idea? And can you give me one jot of
evidence voters are less strategic in PR elections?
2. The fact is, in the Australian PR elections, they also use
"Warren's structure" a huge fraction of the time --
which we already agree is plainly massively dishonest. So you're
just wrong. Australia's voters are massively strategic=dishonest,
regardless of whether they use IRV or use PR. By "massive" I mean,
"the percentage of
voters who give some ordering different than their honest order, is
large, of the same order as the total number of voters."

> My point is that there is nothing *wrong* with this if that's what voters want,

--you do not know what voters want. And in fact, as we just went
over, you are intentionally partly covering your eyes, which seems a
strange way to try to deduce what they want.

> and moreover there's nothing *correctible* about this if you are
> stuck with winner-take-all elections, but there's where we disagree.

--Wrong. Score voting is a winner-take-all system, i.e. there is one single
election winner. But in this score voting election
#voters their rating vote
45 R=9, G=8, D=0
45 D=9, G=8, R=0
10 G=9, D=4, R=0
observe G is the winner. (G still would win here even with 50-50-0
not 45-45-10.)
Notice that by theorem G could never have won with IRV since 90%>75%.
But G does win with score. Thus this problem IS correctable, even if
we are "stuck with winner take all elections."

And as you've said before, this is not merely a matter of subjective
"disagreement" between us. It is simply an objective fact.

> Now, I think I can see where you're coming from with AV. In a situation
> where two major parties hate each other, it might be possible for a minor
> party to win because some voters would use the "anything but" method of
> voting, and select nearly everyone on the ballot except the hated other
> major party. In this case, both A voters and B voters would choose C, and C
> would win. So, yes, I can imagine a world in which AV led to the selection
> of an unpopular third party (or, at least, a third party that was the first
> choice of a small group of people).

--I do not know what you are talking about here. In any event, I was
not trying to
make such an argument about that fantasy world. I was trying to make
arguments about the real world.

> But here's what I'm absolutely
> certain would happen with AV, and I have that level of certainty because it
> happens in AV-like elections with multiple seats open. The major parties
> would encourage their voters to use bullet voting, not real approval
> voting.

--Well, numerous Approval style polls have been conducted, and some as
exit polls, in real elections, where real voters were encouraged to
vote as they would have if the AV study were the real election. In
those studies, multi-Approval ballots have been common. For example,
see
http://rangevoting.org/French2007studies.html
where it is mentioned that in the France 2007 presidential election
the average voter approved 2.33 candidates, and in France 2002 she
approved 3.15.
Both a lot more than 1. (See "Distribution" on that page.)

> Now, voters may ignore their advice, but it wouldn't take much
> bullet voting before a minor party candidate has no chance to win.

--that argument by you has some validity and strikes me as a good argument for
"score voting" not approval voting. In fact, I observed in a poll I
did, and also in other
polls by others, an effect I dubbed the "Nursery effect"...
http://rangevoting.org/NurserySumm.html
which is: third parties get a lot more votes (relative to the major
two's vote counts)
if we use score voting rather than approval voting. By "a lot," I
mean, factor of over 10
often happens. For this reason, I fear that approval voting might
still fall into 2-party domination even while score voting could
escape that trap. (E.g. see above example
with R versus D versus G.)

> course, you can argue all you want that it would be *irrational* for voters
> to follow party advice on this topic, but I'd argue that it's *irrational*
> for voters to follow party advice on candidate order in an IRV election.
> And yet, they often do, as you point out.
> -- Dave

--Indeed, I do not know what voters, or parties, will do. The
studies so far with AV exit polls, of course have been conducted
without the parties particularly trying to influence voters on how to
act about the poll study.

There also has been long real world use of score voting for real governments.
Sparta was run with score voting & Venice with 3-level score voting, both
for centuries. Unfortunately the records of what happened in the Sparta
elections seem lost, and in Venice they are not necessarily lost, but
I do not have them.
As far as I can tell, though,
1. in neither place was 2-party domination experienced.
2. in both places, their governments were extremely successful
in the sense of lasting a long time despite what would seem to be very
adverse circumstances. They both lasted longer than the USA or
Australia so far, and in circumstances much more adverse. Indeed, I
think no democracy that has yet existed
based on a non-score voting method, can compete with those two for
longetivity + overcoming adversity, except perhaps England.

[Clay Shentrup]

On Tuesday, December 9, 2014 8:28:01 AM UTC-8, david.cou...@drake.edu wrote:

The hypothesis that you put forward is that IRV will lead to two-party domination *because* people will not vote for their preferred candidate in the first round if that person is not from one of the two major parties.

I repeat what I said in my first reply: "regardless of why IRV leads to two-party domination, it's clear that it does."

[Clay Shentrup]

On Tuesday, December 9, 2014 2:44:51 PM UTC-8, david.cou...@drake.edu wrote:

But here's what I'm absolutely certain would happen with AV, and I have that level of certainty because it happens in AV-like elections with multiple seats open. The major parties would encourage their voters to use bullet voting, not real approval voting. Now, voters may ignore their advice, but it wouldn't take much bullet voting before a minor party candidate has no chance to win.

That's only if they sincerely favor one of the two major parties. If they prefer e.g. Green, then they vote Green, or maybe Green+Democrat.

In our recent Maine gubernatorial exit poll, the independent won with approval voting, despite having only 8.4% with plurality voting.

http://scorevoting.net/Maine2014Exit.html

Bullet voting, in general, is a nonsense argument.

www.electology.org/bullet-voting

[Clay Shentrup]

Regarding "bullet voting", let me go even further. What that term means is "sincere plurality-style voting". Note that word: sincere.

The whole reason plurality voting ordinarily leads to duopoly is tactical voting. Thus this "criticism" is saying that approval voting would actually solve the biggest part of the problem that leads to duopoly.

[david.cou...@drake.edu]

On Tuesday, December 9, 2014 10:20:36 PM UTC-6, Clay Shentrup wrote:

Regarding "bullet voting", let me go even further. What that term means is "sincere plurality-style voting". Note that word: sincere.

The whole reason plurality voting ordinarily leads to duopoly is tactical voting. Thus this "criticism" is saying that approval voting would actually solve the biggest part of the problem that leads to duopoly.

This is a really good point. I would definitely like to have a system that leads to more sincere first-choice voting, and I agree 100% that the first-choice candidate in AV will get the vote, and the question is just what happens next. Let me make it clear that I'm not trying to argue against AV or score voting here. I totally believe that we should try both in public elections and I would vote for any initiative that called for doing so. My goal in this discussion is 1) to learn where my thinking may be sloppy, and 2) to see if I can convince you guys to step down the anti-IRV vitriol, because I think IRV is probably the best chance we have at voting reform and I think a lot of the criticisms you lob at it are unfair. 

So, a couple of comments about bullet voting. I do not believe that it is in any way a silly argument. I have been lobbied to use it in multi-seat races, and the political pressure to do so in single-seat races using AV would be even greater. You point to the French poll to indicate that I am being unreasonable in my fear of bullet voting. Here's how I see that poll. The researchers sent letters to all of the participants explaining to them the principles of approval voting and such. Then, after the official voting, voters went to take the poll on whom they would have chosen had approval voting been the method that was being used. Psychology says that people taking surveys will 1) feel obliged to try out the tool that the survey is asking about, and 2) want to please the survey takers. Thus, the fact that in a survey where participants were *told* the researchers were investigating this new voting method, and the sole difference between the old voting method and the new voting method is that you get to choose more than one candidate if you want, I am not at all surprised that respondents chose more than one candidate! Two other things make me skeptical that these results would hold up in the real world. First, people make different choices on surveys than they do in the voting booth, and more importantly, people are likely to respond very differently after a big push by the major parties to get their voters to bullet vote. Will the parties make such a push? Absolutely! (See Warren's post about how AV can hopefully lead to third party victories!)

Now, this is speculation on my part because we don't have mass elections using AV or Score. But it is not fair to say that worries about bullet voting are silly. This discussion is all about whether AV is likely to break the duopoly in countries where a majority of people vote for one of the two major parties. If I had money to bet, I would say no. Look, take Warren's example with score voting in a previous post. in that case, if one third of the voters engaged in bullet voting (giving their side a 10 an everyone else a 0), then the Democrat would win. And that's making the crazy assumption that both Democrats and Republicans would rank the Green an 8 when they weren't bullet voting. If Dems and Reps rank the Green a 5 then the Dem wins, even without bullet voting. If Dems rank the Green an 8 and Reps rank the Green a 2, then the Dem wins, without bullet voting. And so on. So, yeah, I think bullet voting is going to lead to results that are little different for minor parties than those they see with IRV. That's especially true since the minor parties are often seen as less centrist than the major parties.

Again, we should test AV and Score in the real world. But I think you're doing harm by calling IRV supporters "suicidally stupid."

[david.cou...@drake.edu]

I continue to contend that claim is based upon a biased reading of the results. The fact that smaller parties have recently held the presidency in Ireland, and that the biggest party has not *even though they run* suggests to me that this claim is far, far from clear. And we've had a lot to say about Australia. I'm guessing that if I look into Fiji I'll find a more nuanced picture also.

Why do I think this? Because every time I look at an absolute claim on the election science site, I get more suspicious. I *did* think it was very weird that Instant Runoff gave rise to very different results than delayed runoff. I decided to check what warranted a "Q" in your list of delayed runoff countries. The first on the list was Argentina. Holy cow, the Argentine Presidency has been held ONLY by candidates from the Judicialist Party (Menem, Saa, Duhalde, Kirchner, and Kirchner) or the Radical Civic Union (Alfonsin, de la Rua) SINCE THE END OF MILITARY RULE 30 YEARS AGO. How in the heck does that warrant a Q? If I check the other Qs on Wikipedia will I find the same thing??? I'm throwing up my hands here.

[Warren D Smith]

> This is a really good point. I would definitely like to have a system that
> leads to more sincere first-choice voting, and I agree 100% that the
> first-choice candidate in AV will get the vote, and the question is just
> what happens next.

--well, I don't know about 100%, but it appears almost 100% of
the time, I think about 98% in our Maine poll, real voters appear to
approve their favorite. I mention this to note that when you do these kinds of
polls, you soon realize that voters do not
necessarily do or think what you think, and you don't know why they do it.
It's a good cure for the overconfidence some have, that they
understand what voters think and will do.

> to see if I can convince you guys to step down the
> anti-IRV vitriol, because I think IRV is probably the best chance we have
> at voting reform and I think a lot of the criticisms you lob at it are
> unfair.

--I doubt you'll succeed. In fact, if anything, the massive number of
lies and inaccuracies I've seen over many years from IRV propagandists
has had the effect on me of making me more not less, anti-IRV. In
the USA during my life, I think almost all (every?) IRV
enaction has been a change away NOT from plurality, but from plur+top2runoff
(i.e. with a second round, genuine runoff). And this change, I
think, is actually the opposite of an improvement -- for example
plur+top2 empirically escapes 2-party dominatuon while IRV does not.
Whereas if they had instead concentrated on changing plain plurality
to IRV, then that would in my view have clearly been an improvement.
(Not enough of an improvement, but improved.) So in short, the entire
effect of Rob Richie's life work, has probably been to WORSEN the
voting systems of the USA, and this so-called "reform" was
accomplished via large scale lying and self-deception.
Think you are going to make me happy about that?

> So, a couple of comments about bullet voting. I do not believe that it is
> in any way a silly argument. I have been lobbied to use it in multi-seat
> races, and the political pressure to do so in single-seat races using AV
> would be even greater. You point to the French poll to indicate that I am
> being unreasonable in my fear of bullet voting. Here's how I see that poll.
>
> The researchers sent letters to all of the participants explaining to them
> the principles of approval voting and such.

--did they? I don't think there were such letters, but I could be wrong.
Certainly all the exit polls I've been involved in, no letters, and
we did not even know who would participate so we could not have sent
letters even if we wanted to & had the money.

> Then, after the official
> voting, voters went to take the poll on whom they would have chosen had
> approval voting been the method that was being used. Psychology says that
> people taking surveys will 1) feel obliged to try out the tool that the
> survey is asking about, and 2) want to please the survey takers.

--there also have been many many approval-style and core-style polls
conducted pre-election in the USA and other places by polling agencies
usually by phone.
(EXIT polls are pretty rare to use score & approval, but it is pretty
common pre-election polls.) None of those do letters. In none of
those do the voters have any idea how to "please the pollster" nor
does the pollster tell them about the ideas behind approval voting or
nor does he even tell then it IS "approval voting."

Typical score-style polls:
http://www.rangevoting.org/LAmay2001polls.txt
http://www.rangevoting.org/UIowaPolls.txt

Review of USA elections 1968-now:
http://www.rangevoting.org/RangePolls.html

In nearly every USA election with a goodly
number of well known candidates the sum of approval percentages exceeds 100%.
A random example might be
1992 USA presidential election, where the favorability numbers were
Perot 36.4
Bush 39.6
Clinton 40.8
SUM 116.8%
using 3-choice scale "favorable," unfavorable," or "no opinion".
If we discard the no-opinion pollees then the sum would be higher, in
this case probably it would rise to about 140%.

> Thus, the
> fact that in a survey where participants were *told* the researchers were
> investigating this new voting method, and the sole difference between the
> old voting method and the new voting method is that you get to choose more
> than one candidate if you want, I am not at all surprised that respondents
> chose more than one candidate!

--also I remark in our Maine poll we just did, the voting methods used
were plurality, IRV, and approval, not just approval. As Clay
remarked, our and other polls in Maine using
various voting methods all found CUTLER won, except under plurality where
he came way last.

> Two other things make me skeptical that
> these results would hold up in the real world. First, people make different
> choices on surveys than they do in the voting booth,

--oh. Well, if you are going to take the stance that polls all are
garbage and nothing whatever can be told from them, we will have to
disagree. And you will be standing on
extremely thin ice and look exceedingly desperate.

> and more importantly,
> people are likely to respond very differently after a big push by the major
> parties to get their voters to bullet vote.

--they might indeed. However, the very fact that the parties in your
scenario felt they had to make such a "big push" at all, would be
quite a difference versus now...

> Now, this is speculation on my part because we don't have mass elections
> using AV or Score.

--yes. I do not know how the parties will behave. But as I said, centuries
of experience using score voting in Sparta & Venice did not, as far as
anybody can tell, lead to 2-party domination.

> But it is not fair to say that worries about bullet
> voting are silly.

--I agree the bullet worry is not totally silly, but I also think some
of those making arguments about bullet voting have done so in silly
ways. Indeed, the most prominent
people making those arguments have been "FairVote" (IRV propagandists)
who scream
"Approval? Horrible due to bullet bugaboo!" but neglect to
inform us that the very IRV elections they themselves laud as great
successes, suffered higher bullet voting rates than approval polls &
elections studies generally do.

It is to laugh.

> This discussion is all about whether AV is likely to
> break the duopoly in countries where a majority of people vote for one of
> the two major parties. If I had money to bet, I would say no.

--I would also worry about that; as I said the "Nursery effect"
suggests to me it is a lot more likely score voting will break 2-party
domination, than approval.

> Again, we should test AV and Score in the real world. But I think you're
> doing harm by calling IRV supporters "suicidally stupid."

--well, IRV supporters who also are third-party bigwigs, ARE
suicidally stupid, since IRV yields 2-party domination. Third
parties should support a voting system that is not known to lead to
2-party domination.

[Warren D Smith]

> Why do I think this? Because every time I look at an absolute claim on the
> election science site, I get more suspicious. I *did* think it was very
> weird that Instant Runoff gave rise to very different results than delayed
> runoff. I decided to check what warranted a "Q" in your list of delayed
> runoff countries. The first on the list was Argentina. Holy cow, the
> Argentine Presidency has been held ONLY by candidates from the Judicialist
> Party (Menem, Saa, Duhalde, Kirchner, and Kirchner) or the Radical Civic
> Union (Alfonsin, de la Rua) SINCE THE END OF MILITARY RULE 30 YEARS AGO.
> How in the heck does that warrant a Q? If I check the other Qs on Wikipedia
>
> will I find the same thing??? I'm throwing up my hands here.

--well, I just looked at http://en.wikipedia.org/wiki/Politics_of_Argentina
and it has a table giving party composition of legislature. Actually it only
gives the parties winning seats in the latest election (many seats
were not being contested and are not tabulated). Anyhow, far as that
table goes, I count
A 19
B 14
C 6
D 8
E 47+8
F 28
G 4
H 3
I=6
J=41+14
K=20
L=4
M=2
N=6
P=13
seats won by parties I have represented by single letters.

This seems like multi-party rule to me, certainly far more so than
Australia house or USA legislature, hence warranting a "Q".

[Clay Shentrup]

On Wednesday, December 10, 2014 8:18:54 AM UTC-8, david.cou...@drake.edu wrote:

So, a couple of comments about bullet voting. I do not believe that it is in any way a silly argument. I have been lobbied to use it in multi-seat races, and the political pressure to do so in single-seat races using AV would be even greater.

Ralph Nader lobbied his supporters to vote for him. Yet an ANES exit poll found that 90% of those supporters voted for someone else anyway. Why? Because they didn't think he could win. So lots of people will choose not to bullet vote.

Here's a simple hypothetical example to show why approval voting is so far superior to IRV for breaking out of duopoly. Let's say we have the Center party.

Real preferences

35% Rep > Center > Dem

33% Dem > Center > Rep

17% Center > Dem > Rep

15% Center > Rep > Dem

Now let's say we just have a tiny bit of tactical voting—10%. First-round totals look like this:

Rep 37%

Dem 35%

Center 28%

Center gets eliminated, and all voters "see" is this first-round total (and most emphasis will be on the percentages between the final two candidates). This is how the media reports IRV totals here in the Bay Area.

With approval voting, some of those Dem and Rep voters will sincerely approve Center. And some will do it tactically, because they believe e.g. Rep=10, Center=7, Dem=0, and are hedging their bets. So your results will look more like:

Rep 48%

Dem 48%

Center 41%

This is a much more aggressive move away from duopoly. This is very much the kind of effect you see in our recent Maine gubernatorial exit poll:

http://scorevoting.net/Maine2014Exit.html

[david.cou...@drake.edu]

Yeah, but the Argentine legislature is not elected using delayed runoff. The Chamber of Deputies is proportional representation (d'Hondt), and the Senate is elected by a three-seat representational sort of thing (winning party/coalition gets two seats, second place gets one seat, in 24 districts, so there are 72 Senators).

The race in Argentina that is delayed runoff is president, and that seems clearly two-party dominated.

[Jameson Quinn]

I think it's important to be clear about the difference between our personal opinions about different voting systems and our stance as the CES and as activists. This thread, so far, has been a debate between David on the one side, and Warren and Clay on the other. In this debate, the CES does not have an official position; and scorevoting.net is Warren's personal website, not an official CES site. 

I think it's healthy to have this debate. For instance, it seems clear to me that David is right about Argentina being mislabeled in the table; though I'm sure it was an honest mistake.

In the end, CES is on the side of science, which means seeking truth through evidence and debate. I am interested in what both sides of this debate are saying, and I look forward to seeing if we can reach some common ground. For now, I'm just stepping in to make sure that we don't give the impression of closed-mindedness.

--

[Warren D Smith]

On 12/11/14, david.cou...@drake.edu

<david.cou...@drake.edu> wrote:
> Yeah, but the Argentine legislature is not elected using delayed runoff.
> The Chamber of Deputies is proportional representation (d'Hondt), and the
> Senate is elected by a three-seat representational sort of thing (winning
> party/coalition gets two seats, second place gets one seat, in 24
> districts, so there are 72 Senators).
>
> The race in Argentina that is delayed runoff is president, and that seems
> clearly two-party dominated.

--aha. Well, in that case, I will now remove Argentina from the dataset
entirely, since it is pretty irrelevant with only 1 seat elected by
the methods in question.
(I don't really buy the notion of "2-party domination" if there is
only 1 seat affected and it has had only about 7 elections in all of
history.)

If you find more countries in that data you do not like, or have other data,
let me know and we can continue to modify that page with better data.

[Warren D Smith]

> --aha. Well, in that case, I will now remove Argentina from the dataset
> entirely, since it is pretty irrelevant with only 1 seat elected by
> the methods in question.
> (I don't really buy the notion of "2-party domination" if there is
> only 1 seat affected and it has had only about 7 elections in all of
> history.)
>
> If you find more countries in that data you do not like, or have other
> data,
> let me know and we can continue to modify that page with better data.
>

--although actually, I point out the page had said
"Delayed runoff is used in presidential elections in (where "Q" =
qualifies as multiparty country at time of writing) Argentina ..."

which in fact was a true statement about Argentina: (1) it did use
runoff for president
(2) it was multiparty country.
At the time I wrote the page it was not as easy as now to figure out
how all the non-presidential seats were elected... there may be more
countries affected by an analogous criticism.

--By the way, another thing you said was, in Ireland, you thought the
fact FF had not run a candidate in the latest presidential election,
was somehow indicative of something about IRV.

I think it is more indicative of the fact the presidency of Ireland is
a nearly powerless symbolic position that the FF party may not
actually give much of a fig about owning. The closest analogy in the
USA seems the "surgeon general." The president of Ireland has less
power than the lowliest MP, far as I can see. We have an Irish
member who has sometimes posted, who perhaps may comment on this (?).

Also, Higgins, the Labour candidate and winner, had formerly been in
the FF, which may also have affected their thinking.
Also, FF had recently suffered a huge drop in popularity, and they may
not have wanted to exacerbate that by publicly losing the presidential
campaign, considering that
more damaging than the small chance of winning this unimportant, but
well publicized, position was worth.

[david.cou...@drake.edu]

OK, so I looked through the list, and checked whether there were multiple parties represented in the elections where delayed runoff is used. Here is what I came up with. Where a date is given, that's because that was when military rule ended and elections began (or something similar). I then call things "Q" for "multiparty in relevant election," "2" for one- or two-party in relevant election, or "N" for not enough info/inconclusive/not delayed runoff.

Argentina: Two parties since election. 2

Brazil: Multiparty (1985) Q

Central African Republic: Minor parties only have 2 seats out of 105, but independents have 26. I don't really know what that means, but clearly no strong minor parties. 2

Chile: Multiparty in presidential elections since 1990. Q

Colombia: This is sort of a special situation. 

Colombia—two party dominated 1958-2002 (Liberal/Conservative). 2002 Uribe was elected after the FARC had taken control of 100 municipalities, and people freaked out. He won in the first round with 53% of the vote. Then in 2010, the liberal and conservative parties had all but disappeared, and Santos was elected from what became of Uribe’s party. Parliament is proportional. I'd call this 2.

Congo: Not sure how this one got on the list. There's only been one party in power since Mobutu Sese Seko was deposed. Current prez took over when his father was assassinated. Parliament is proportional rep. 2

Greek Cyprus: Legislature proportional. Prez multiparty. Q

Finland: Only two in presidency since runoffs began in 1994. Legislature is proportional. 2

France: Q

Gabon: 2

Ghana: 2

Iran: Q

Liberia: Very difficult to say. Single party rule until 1980, since then each party’s rule has ended in civil war. Interestingly, legislature is done with fptp elections, and yet it is highly multiparty (12 parties represented, with top two controlling less than half of the house total).    N

Macedonia:: legislature proportional. Presidency was held by nonpartisan Gligorov after breakup of Yugoslavia. Since 1999, two parties. 2

Mali: Q

Mauritania: 2

Moldova: Uses proportional representation. Probably shouldn't be on list. N

Monoco: "Too small to have parties." N

Nicaragua: Proportional representation, but only two parties. Presidency appears to be plurality. N

Niger: Presidency hard to classify due to frequent coups and military rule (1974-1989, 1999, 2010-2011). Parliament multi-seat constituencies. N

Peru: Parliament representational, president multiparty. Q

Poland: One house proportional, the other fptp. President multiparty. Q

Portugal: legislature proportional, prez two party (Socialist or Social Dem) since 1986. 2

Romania: legislature proportional. President: interim presidents and suspensions make this difficult to tell. N

Senegal: Parliament, multiseat constituencies, president 2-party (current pres is from new party--APR--but he's from former president's party PDS in a personal dispute). 2

Slovakia: Parliament proportional, presidency is generally won by independents. Not sure what this means, but we'll call it Q.

Uruguay: parliament proportional, prez 3-party. Q

So, by my count there are 10 countries where delayed runoff leads to multiparty winners, and 11 with one- or two-party domination.

About this I would say a couple of things. First, Warren should probably update the page, because it makes it look like IRV is terrible for minor parties, but delayed runoff is somehow functionally very different because he has 20 listed as multiparty and only 4 as 2-party. There's a big difference between 1/5 and 1/2. 

Second, if we think about the countries that are listed as IRV, I would have this to say:

Ireland: minor parties win presidency: Q

Australia: 2 (6 minor party seats out of 150)

Malta: Malta has fewer than 400,000 residents. Is that really a country? But, ok, 2.

Fiji: Fiji had, as far as I can tell, 2 elections using IRV. The first one had numerous minor parties, the second did not, and then there was a coup. Perhaps it was moving toward 2-party, perhaps not, who can tell. It's also a country of 900,000 people, so make of that what you will. I'd call this N.

So, by my count, we have three countries where we can say something about IRV, and of those two are two-party, one is multiparty. That is indistinguishable from the delayed-runoff results of about half of the countries with multiparty. Even if you dispute whether we should count Ireland because it is not a powerful position, and just go with Australia and Malta, that's not very convincing that there's a difference between IRV and delayed runoff (if the chance of having multiparty with DR is about 0.5, then there would be a 25% probability that any two randomly chosen countries to run IRV would both wind up 2-party dominated). 

What does this show? To me it suggests that there really is no way to say that there is an observed difference between IRV and DR in terms of whether they drive 2-party domination. This makes logical sense, since the two ought to suffer from the same propensity for strategic voting and so on. 

[Warren D Smith]

--1.thanks, that update was helpful and I will hopefully soon update
that web page.

> Second, if we think about the countries that are listed as IRV, I would
> have this to say:
>
> Ireland: minor parties win presidency: Q

--I'm pretty unimpressed. By minor "Parties" you actually meant
"party", and only 2 parties have ever won presidency -- I repeat, only
2 parties have ever won presidency -- and presidency is unimportant
and nobody cares about it.

Seems to me, a good sign of 2-party domination is "only 2 parties have
ever won."
If you disagree, it is you who are going on on a limb, not me.
I'm calling it 2-party dominated, for what it is worth (which is: not
much, since just 1 Australian house round is way more IRV elections
than Ireland has had, all time).

Plus the only 2 times that so-called minor party won in Ireland, both
were flukes: once FF did not run, and other time there were some
scandals and pathologies.

> Australia: 2 (6 minor party seats out of 150)

--huh?
http://en.wikipedia.org/wiki/Australian_House_of_Representatives
I count 3 out of 150, which by the way is pretty amazingly large.
In 3 consecutive election cycles 2001-4-7 they had elected zero.
That zero was what I was basing my old writing on. But now, voila, there are 3.
Which is a good deal bigger than zero, but not 6.
(Realize the Libs, LibNats, CLP, and Nats are all the same party for
practical purposes.)
Anyhow, this is still 2PD.

> Malta: Malta has fewer than 400,000 residents. Is that really a country?

--Malta has 423282 says 2013 census.
Malta actually does not use IRV, they use PR-STV, but they may at one
point have used IRV, I haven't been able to figure that out. In any
case they've had a long history of massive 2PD.

> But, ok, 2.
> Fiji: Fiji had, as far as I can tell, 2 elections using IRV. The first one
> had numerous minor parties, the second did not, and then there was a coup.
> Perhaps it was moving toward 2-party, perhaps not, who can tell. It's also
> a country of 900,000 people, so make of that what you will. I'd call this
> N.

--Fiji:
http://www.rangevoting.org/FijiPol.html
seemed to have pretty clearly developed 2PD

> So, by my count, we have three countries where we can say something about
> IRV, and of those two are two-party, one is multiparty.

--nope, I'd still count all three and a half as 2-party dominated.

> That is
> indistinguishable from the delayed-runoff results of about half of the
> countries with multiparty. Even if you dispute whether we should count
> Ireland because it is not a powerful position, and just go with Australia
> and Malta, that's not very convincing that there's a difference between IRV
> and delayed runoff (if the chance of having multiparty with DR is about
> 0.5, then there would be a 25% probability that any two randomly chosen
> countries to run IRV would both wind up 2-party dominated).
>
> What does this show? To me it suggests that there really is no way to say
> that there is an observed difference between IRV and DR in terms of whether
> they drive 2-party domination.

--Well, first of all, with only "3 and a half" IRV countries, you are
right this is not a statistically significant conclusion, and I have
no way to avoid that, because
you are never going to get a statistically significant conclusion
based only 3.5 datapoints.

Second, nevertheless, with all 3.5 being 2-party dominated, if I were
a betting man, I'd bet IRV will yield 2-party domination. Especially
since the USA has a lot more 2-party-domination-genic factors than
those places. So if USA switched to IRV I'd bet money it
will stay 2PD.

Third, it looks to me like your updates re delayed runoff mean there
are still a goodly number of places where it led to escaping 2PD.
Optimally I think that page should be rewritten to discuss both
presidency & legislature. Also, Duverger claimed delayed runoff
led to "multiparty-ism moderated by alliances" based on his historical study,
which is why this is called "Duverger's law" in political science:
http://www.rangevoting.org/DuvTrans.html

> This makes logical sense, since the two
> ought to suffer from the same propensity for strategic voting and so on.

--absolutely not. With delayed runoff, every voter, no matter how
strategic they are, will vote honestly in round 2. (Their dishonesty
will be confined to round 1, and there is
evidence that it is large there. For example in French presidential
election 2007,
1-5% of voters said they'd voted against their 1st round choice in the
2nd round,
a particularly blatant form of dishonesty; in their 2012 election 20%
said to TNS-Sofres they'd voted dishonestly in round #1.)
With IRV, many voters will vote dishonestly in each round. There is
no wholy-honest round. A considerable difference, and one which seems
to lead to substantially different results:
http://www.rangevoting.org/TTRvsIRVrevdata.html

[david.cou...@drake.edu]

So, to get clear on the argument for why delayed runoff is better than IRV in your mind: In delayed runoff people will vote honestly in the last round. It's a plurality thing before that, so there are all kinds of opportunities for spoilers and so on, but at least the last round is honest. Fine, I'll give you that nobody has an incentive to vote strategically in the runoff round (though lots of people apparently don't have an incentive to vote at all, if US runoff elections are an indication, so perhaps you're trading disenfranchisement for honesty?). 

What would it take for this to be functionally important? My understanding of your arguments for how people vote strategically via IRV is that they use polls to figure out who is likely to be the front runner, and then they wind up voting early on for the likely frontrunner that they prefer over the likely frontrunner that they don't want, and put the person that they don't like at the bottom of the list below the communists and the nazis and all. If polls are any good, then to cast a "strategic" ballot where you're voting against your preferred candidate in the final round, you would have had to rank your less preferred candidate higher than your preferred candidate. And you would have had to do that *knowing* that the less preferred candidate had a really good shot at being in the final round. I agree that strategic voting can be a problem in IRV (as bullet voting can be a problem in approval or range), but this level of strategic error strikes me as very unlikely. Perhaps you have counter-evidence?

The time I can see it happening is if there was all of this strategic voting going on and so the polls didn't really predict what would happen, and so the morning after the election everybody woke up and was surprised by how things turned out so differently from what they expected. The top two candidates were not who they expected and *holy cow* they actually voted *for* the person they hated instead of the person they wanted to win. That would be terrible, I agree, but if that kind of thing happened more than occasionally then I feel like people would realize that trying to predict polls and vote strategically was much more difficult than they thought, and, heck, they might as well vote honestly if strategy was going to bite them in the rear like that. So, long-term win for honest voting. :)

[Warren D Smith]

On 12/22/14, david.cou...@drake.edu

<david.cou...@drake.edu> wrote:
> So, to get clear on the argument for why delayed runoff is better than IRV
> in your mind: In delayed runoff people will vote honestly in the last
> round.

--that indeed is an obvious difference.

> (though lots of people apparently don't
> have an incentive to vote at all, if US runoff elections are an indication,

--also true. However, among those who do vote, I cannot see any
reason any of them would want to be dishonest in an A versus B runoff.
Hence I would expect 99% honesty, and maybe 1% deranged voters, in
said runoff.

In the USA, turnouts are usually small, say 50% in presidential
elections, 33% in nonpresidential elections, and in primaries and
obscure elections often much smaller
than that, like 10%. Runoffs can have very high and very low
turnouts, it depends on the runoff. (This contrasts with many other
countries, including Australia, where not voting is
actually illegal and hence the turnouts exceed 90% in every election.)

That 10% is a very bad thing because in many elections due to the
USA's enormous gerrymandering, the primary is the real election. With
say 10% of the voters from only 1 party deciding what is effectively
the real election, and those 10% are a very biased self-selected
sample highly enriched in the most extreme members of that 1 party, we
get
the election of maniacs who have virtually nothing to do with what
that place actually wanted. [E.g. just last election Michael Grimm
(R, Staten Island NY) was re-elected despite on camera threatening to
throw a reporter off the balcony for enquiring about
his campaign financial shenanigans. Grimm also was under federal
indictment for 20 alleged crimes and is expected to plead guilty.
http://www.nydailynews.com/news/politics/new-york-rep-michael-grimm-threatens-reporter-asked-fundraising-allegations-article-1.1594857
How did somebody like that get elected? Simple, New York state is
gerrymandered,
and he's in a sure-to-elect-republicans district, which probably would
still have elected him even if he'd gone ahead and thrown that
reporter off the balcony in front of all the TV cameras. In fact, he
won the general by a greater margin than his party primary
every time anybody ran against him in the primary. Next door in
Harlem NY, Charlie Rangel (D), a congressman apparently comparably
corrupt to Grimm, is continually re-elected, and is now one of the
USA's longest serving congressmen.]
This has caused today's congress to be among the most polarized (only
rivaled by the congress just before the US civil war), and also the
least productive, ever.

> What would it take for this to be functionally important? My understanding
> of your arguments for how people vote strategically via IRV is that they
> use polls to figure out who is likely to be the front runner, and then they
> wind up voting early on for the likely frontrunner that they prefer over
> the likely frontrunner that they don't want, and put the person that they
> don't like at the bottom of the list below the communists and the nazis and
> all. If polls are any good, then to cast a "strategic" ballot where you're
> voting against your preferred candidate in the final round, you would have
> had to rank your less preferred candidate higher than your preferred
> candidate. And you would have had to do that *knowing* that the less
> preferred candidate had a really good shot at being in the final round. I
> agree that strategic voting can be a problem in IRV (as bullet voting can
> be a problem in approval or range), but this level of strategic error
> strikes me as very unlikely.

--your fantasies are very long and very detailed. Apparently you have
the idea that if
you construct a very long and detailed fantasy, then ascribe that
fantasy to me, then because it is so long and detailed, it must be
unlikely, and therefore you ascribe to me, unlikely--to-be-true
thoughts. Thus proving I'm a fool.

The truth about my thoughts is simpler. I do not know why voters do
what they do,
especially when it involves complex thinking. If anybody constructs
complicated theories about why they do whatever, and proclaims those
theories must be true due to introspection and their alleged vast
understanding of human nature, I do not believe them. In fact, voters
are weird and plenty are not like you and me (or anybody else); and
often exhibit enormous ignorance. In a recent USA poll, 2/3 of them
were unable to name the three branches of government. In San
Francisco 2005, where IRV had recently been introduced, 13% of voters
polled said they "Did not understand IRV entirely" or "Did not
understand it at all." Probably the true fraction exceeds 13% because
other polls have shown voters pretend greater understanding that they
really have, e.g. large fractions are willing to give their opinions
about non-existent things, if asked.
So do you think they understand how voting systems like IRV and
(especially, which is
quite a lot more complicated still) PR-STV, work? Especially such phenomena as
non-monotonicity? I have doubts. Given such a large level of
ignorance, any claims due to the Vast Power Of Your Comprehension Of
Human Nature, of understanding of what voters will do and why, is
dubious to me.

What we can do, is instead of SAYING what voters think (rather like
Sigmund Freud, who famously claimed understanding of how human minds
work, without need for experimental evidence -- in case you're
wondering I detest Freud and his followers), we can ENQUIRE about what
they want and how they behave.

Behave: In Australia, rank-order ballots, we find that about 90% do
the exaggerate-the-two-major-parties-to-extreme-ends behavior.
Why do they do that? What is going through their minds? I DO NOT KNOW.
And if you clam to know, based purely on the power of your inner
understanding, then you are a fool. I simply know that they do it, and
they've done it for decades, and it has large mathematical and
historical consequences,and it is obviously massively dishonest.
I do not know why they do it.

Want: Also in Australia, there have been polls asking voters if they'd like
to get rid of IRV and replace it by plain plurality. Every such poll I know of
(2010, 1984, 1974) "plurality" won:
http://rangevoting.org/WhatVotersWant.html
by an aggregate margin of 59-41 (ignoring "don't know"s) which margin
is comparable to the largest "landslides" in USA presidential election history.

>Perhaps you have counter-evidence?

--I simply don't need to counter fantasies.
It is you who needs to counter reality.

> The time I can see it happening is if there was all of this strategic
> voting going on and so the polls didn't really predict what would happen,
> and so the morning after the election everybody woke up and was surprised
> by how things turned out so differently from what they expected. The top
> two candidates were not who they expected and *holy cow* they actually
> voted *for* the person they hated instead of the person they wanted to win.

--you just cannot stop theoretical fantasizing, can you?
The fantasies grow ever more and more detailed and complicated in your mind for
paragraph after paragraph of detailed fantasies.
with no experimental evidence being provided to decide any aspect of them.
Apparently in your mind THE VAST POWER OF YOUR UNDERSTANDING OF HUMAN
NATURE trumps any need for experiment, you're so hugely vastly
convinced you understand the innermost details of people's thoughts,
99.99% of whom, you never met.
Why? What's the matter with simply accepting experimental reality?

> That would be terrible, I agree, but if that kind of thing happened more
> than occasionally then I feel like people would realize that trying to
> predict polls and vote strategically was much more difficult than they
> thought, and, heck, they might as well vote honestly if strategy was going
> to bite them in the rear like that. So, long-term win for honest voting. :)

--yawn. Look, I do not believe in your utterly unsupported internal
fantasies, so you might as well just write "blah blah blah" for all
the serious attention they are likely to get from me or anybody else.
If some random guy spouts his thoughts abut Human Nature
in a bar, who pays attention?

If we have three polls saying landslide margins of Australian wish
they did not have IRV, I'd recommend paying attention to that. Over
and above the bar ranter guy.

[Clay Shentrup]

Sometimes I truly believe there is a gene or something—or some kind of defect in neural circuitry affecting logical comprehension—that drives the thinking of IRV advocates. I just rekindled this discussion with David, after he posted this article in the Des Moines Register.

http://www.desmoinesregister.com/story/opinion/columnists/iowa-view/2016/06/16/how-we-can-change-our-failed-two-party-system/85978282/

Notice this quote:

IRV voting won’t suddenly make third parties viable — elections will still be won overwhelmingly by Democrats and Republicans because these parties are well organized and have a lot of members. But it would mean that people could vote their preferences, not their fears, and strong independents would not be tarred as spoilers.

Astonishing. It's as if everything we said to him just vanished from his mind. After I mentioned the Bayesian Regret figures, he shot back this final email.

---

The model [Warren's BR sims] assumes people will not engage in bullet voting, but will, instead, give accurate scores to all candidates. It makes assumptions, however, about IRV strategy, because Warren happened to think those were warranted. I consider the lack of bullet voting to be an entirely unreasonable assumption refuted by actual evidence from real elections, but instead based upon what one guy thinks seems like a “rational” action. With significant amounts of bullet voting, Score and Approval voting degrade into the current system, and you’re back to square one. The fact that the model doesn’t have AV sometimes doing as badly as the current system demonstrates my point about assumptions in the model. I imagine the truth lies somewhere in-between, but even after we spoke at length on this topic last year you guys "constantly repeat false and misleading claims about [AV] and other systems. As you just did in your [e-mail], even though the facts have been explained to you at great length. This is infuriating.” So, there you go.

I do not believe that my statement [that I quoted at the top of this post, from the Register] is misleading. The spoiler effect that concerns me is when people feel like they can’t vote for their favorite candidate because it will lead to a worse outcome. I don’t believe that will happen. [Oh, and readers were just supposed to read his mind and know that he was only talking about a specific kind of spoiler scenario—the one that concerns him.] We live in extremely polarized times. The number of people who will vote for a liberal first but a conservative second against another liberal is quite small. There’s all kinds of analysis on this out on the web. Yet you assume that those people are common enough to sway the outcome, but of course NONE of the voters for the “spoiler” liberal would have chosen the conservative second. Because if the second choice of the “spoiler” is, for a chunk of people smaller than for the “spoiled” candidate, the conservative, then no spoiling happened. So, you need very particular circumstances for a third person to be a spoiler. Anyway, if my worse sin is using the word “avoids” when I should have said “avoids, under real-world scenarios,” then mea culpa. I’m not the one posting as mathematical fact a model that makes very particular and unwarranted assumptions about human nature in order to make his system look better.

Anyway, thanks for the link. It reminded me why I was less concerned when you guys cut me off (or whatever). I had forgotten that he had also put together an entirely incorrect table of IRV resulting in 2-party domination, and when called on it did some hand-waving and pretended not to have made up the result. Then argued that AV is better because it probably worked in Athens and Sparta, though he wasn’t really sure on that. 

So, look, this is not a productive use of my time. I feel like I engaged in the conversation to learn, and CES engaged in the conversation to persuade. I feel like you do so with extremely faulty models that Warren puts into intentionally opaque language so that nobody bothers to check them (as long as he doesn’t submit it for peer review, which of course he has decided is beneath his dignity, probably because people kept rejecting the papers for being wrong). I just don’t see any point in continuing to converse under those terms.

So, go ahead and have the last word. I will read it, as I have read all of your stuff. Perhaps I’ll even see something valuable that will help me to widen my perspective, as I did last time we conversed. But I am not going to respond. I consider you and Warren to be intellectually entrenched and unwilling to question your assumptions. For all I know you think the same of me. In any event the conversation takes a lot of my time and it is not worth it. 

Again, I wish you all the best in your endeavors, as I’d love to be proven wrong about AV in practice.

— Dave

[Clay Shentrup]

Reminder that David is an associate professor of environmental science and policy at Drake University. Let your mind stew on that one.

[Toby Pereira]

I've probably said this before, but on the Bayesian Regret thing, while Warren wouldn't deliberately fix the results, every simulation like that contains assumptions, so I wouldn't use it as a final proof of anything, even though it does look much better for score and approval than for IRV. It should serve as encouragement to others to do similar work of their own.

[Warren D Smith]

> be a significant improvement over plurality voting. For example, Smith
> argued in a post that Greens are "suicidally stupid" for supporting IRV,
> because he thinks IRV is bad for third parties, and AV is clearly better.
> Australia is always brought up as a case in point, because the lower house
> (where IRV is used) has very little third party representation (if you
> accept the claim that the Liberals, Nationals, and Nat-Libs are
> functionally the same party, which is a bit of a stretch but the subject
> for another discussion).
>
> Now, there are two reasons why minor parties could be doing so badly in
> Australia's House. One is the claim Smith makes a lot which is that IRV
> drives strategic voting and this leads to two-party dominance. The other
> possibility is that minor parties don't actually make up a majority of many
>
> districts, and so no voting system--including AV--would give them more
> representation in a winner-take-all system.
>
> It seems easy enough to test whether Australians are voting strategically
> in their IRV elections: just ask if there are a lot of people who are
> voting for major parties in House races but minor parties in Senate
> elections where they have proportional representation (so they have every
> incentive to vote for their first choice).

--
1. You simply observe that the vast majority, about 85%,
of Australians rank one other majors top, other bottom or nearly, on
their ballot
with 7 candidates on average running in all.
This obviously is dishonest strategic voting: How can the most popular
2 parties in Australia be rated "worst"? It is exaggerated. It is
dishonest.
2. You observe the mathematical theorem about IRV, that if at least
T percent of the voters behave this way, then it is impossible for a
third-party candidate to win. Ever. What is the value of T? 75%
works.
3. Plainly 75% is exceeded by actual Australian behavior.
4. Therefore it is suicidally stupid for third parties to want IRV.
And indeed, the Australian third parties do not want it, and the
Australian public also does not want it (say polls).




>
> So, I checked the Australian House election of 2013. Labor plus Coalition
> earned about 78% of the first-place vote.
> http://en.wikipedia.org/wiki/Australian_House_of_Representatives. Your
> working theory states that this is the result of a long process of
> strategic voting where voters have learned that it is a mistake to vote for
> third party candidates.

> Sounds pretty convincing, as that's a pretty poor
> showing for third parties (assuming away the coalition thing). So, in the
> Senate, where voters aren’t encouraged to vote strategically, how do the
> major parties fare? 76.4%.
>
> That’s not a lot of strategic voters.

--that is a lot of strategic voters. If a huge supermajority of
voters are strategic,
what the hell could possibly be larger? What you are confused about
here (I think)
is you are attributing great intelligence to the voters, and arguing
the distinction between the IRV (House) and STV-PR (Senate) election
makes a big difference in their thinking, then finding a contradiction
with the evidence.

The truth is the voters are naive and in fact the vast majority of
Australians cannot even explain their STV-PR system to you.
Nevertheless they do have the idea/feeling that exaggerating their
vote will increase its impact. I call this "naive exaggeration
strategy." As I said, the vast majority, about 85%\, of Australians
employ this dishonest voting strategy and it utterly destroys IRV.



> In my mind, that’s the whole “strategic voting” argument right there.
> Australians have had 100 years to learn that they should be voting
> strategically using IRV, but they absolutely don't.

--they have had 100 years, and the result is, they absolutely do, by massive
sup[ermajority, with the absolutely TOTALLY HUGELY OBVIOUS ELEPHANT IN
ROOM RESULT being that third parties can virtually never win
Australian IRV house seats.

> Isn't this proof that
> IRV does *not* lead to strategic voting and hurt third parties?

--you appear to have ignored the huge evidence sitting in front of your face,
and somehow concluded it shows the opposite.

[Clay Shentrup]

David just independently came up with the Burr Dilemma. I showed him this link:

http://scorevoting.net/BurrSummary.html

Some of our subsequent discussion follows:

On Tue, Jun 28, 2016 at 5:45 AM, David Courard-Hauri <david.cou...@drake.edu> wrote:

Ha ha. That’s not well-studied, that’s a hand-waving argument based, as I keep saying, in what your dear leader considers to be rational voting behavior.

The Burr Dilemma has indeed been voluminously discussed since 2007. E.g. this post from 2008.

https://groups.yahoo.com/neo/groups/rangevoting/conversations/messages/7728

 

He makes the claim that maybe one side will bullet vote, but the other won’t, because that seems logical to him (point 6).

You're making the assumption that one side will be tactical but the other side won't. Warren is making the assumption that both sides will have a similar composition of strategic/honest voters. Warren's view is in line with decades of empirical data.

The latter point really drives home the fallacy he's making. When Warren says that the C-voters will tactically vote for A or B as well, David sees that as an argument for asymmetrical behavior: A/B voters will bullet vote but C-voters won't. He doesn't understand that this is symmetrical behavior, because both parties are voting tactically.

[Toby Pereira]

It's not necessarily that the A/B voters and the C voters have a different composition of strategic/honest voters, but they're in very different situations from each other so it's reasonable to suppose that their behaviour is likely to be different, whether or not this is objectively logical.

"Because our Bayesian regret computer simulations employed thus-logical strategic voters (in those sims involving strategic voters) the BR measurements were unable to see this whole problem (or only saw a small effect from it)."

On Tuesday, 28 June 2016 21:33:24 UTC+1, Clay Shentrup wrote:

David just independently came up with the Burr Dilemma. I showed him this link:

http://scorevoting.net/BurrSummary.html

Some of our subsequent discussion follows:

[Clay Shentrup]

On Tuesday, June 28, 2016 at 3:51:01 PM UTC-7, Toby Pereira wrote:

It's not necessarily that the A/B voters and the C voters have a different composition of strategic/honest voters, but they're in very different situations from each other so it's reasonable to suppose that their behaviour is likely to be different, whether or not this is objectively logical.

They are in opposite situations. Their tactical votes should be the precise opposite of each other.