[EM] Kristofer Musterhjelm

[Kristofer Munsterhjelm]

It seems that we've reached somewhat of an impasse. I point to countries
  where IRV has not broken up the uncontested party duopoly, or have
done so only by leading to an uncontested two-plus-a-part structure
(NatLibs being nominally two parties).

You then state that your particular model of competition enabled by PR,
combined with direct elections by IRV, is the secret that will make it
different this time: because the other countries I point at don't have
this structure, their hardening into two- or two-and-a-half static
duopolies won't generalize to what you're proposing.

Myself, I think that is a great risk to be taking, particularly in
respect to binding the ranked ballot idea to IRV itself. If it
backfires, then electoral reform will be set back for a very long time.
The link between the ranked ballot and IRV will go both ways, and the
ranked ballot, as a concept, will be discredited.

So I think that FairVote is gambling with really high stakes on bad
odds. I say this based on two things. First, theoretically, with regards
to center-squeeze and chaos precisely when IRV results start to differ
from Plurality. Second, on the few countries that have tried IRV, where,
PR or not, the major parties (and the same major parties) stay in power,
and where there are multiple elections, third parties (who should be
local competitors and perhaps even local major parties by your model)
never or almost never get elected. You say that evidence doesn't apply,
and there we are.

Since you're considering something that has not been tested but could
happen, I don't see how I could use further evidence to say that your
conclusion is unlikely and that betting on IRV is a bad bet. For that
matter, I don't know what could show me that your solution would work
unlike the other forms of PR+IRV that have been tried, so we're stuck.

Perhaps going for PR first would resolve the matter. However, you think
that to go for PR first means backing FairVote, and if you back FairVote
on that level, they'll tack IRV onto your support. Then we're right back
where we started: with something that will work if your model works, but
not if not.

Let me do two things, then. First to ask: on what do you base that your
model would make all the difference? You mentioned pseudo-PR in an US
state - did that lead to competitive local elections, even for executive
positions? (I thought this state was Illinois, but apparently not, as my
mail search function can't find any messages of yours mentioning it.)

Second, I'll reply to some of your replies below, but I think we're
converging on what the disagreement is.

David L Wetzell wrote:
> Kristofer Munsterhjelm wrote:

>> 
>> Either the major parties were very good at anticipating the direction 
>> they were to move to the center (and very good also at avoiding any 
>> slipups), or the feedback is weak indeed.
>> [endquote]
> 
> dlw: I agree that with IRV, all the two biggest parties in an area got 
> to do is keep the third party challengers from breaking into the top 
> two.  This retains two-party dominance.  So you're saying the available 
> evidence doesn't favor IRV being improved by use along with PR.  

What I'm saying is, in syllogistic form, this:

1. Non-PR IRV countries don't break out of uncontested two (or
two-and-a-half) party domination.
2. The evidence says PR doesn't make a difference.
3. Therefore, PR can't turn IRV's uncontested duopoly into contested
duopoly.

> This however misses my more important point of prescribing the use of 
> local forms of PR for "more local" elections and IRV for "less local" 
> elections.... And so I think the Aussie mixed approach would work better 
> if the use of PR and IRV were switched between Senate and the IRV 
> chamber. The goal here is to maximize the number of competitive seats so 
> as to keep legislators/parties responsive while retaining a big two to 
> ensure quality control and to trust that major changes will come about 
> thru political cultural changes apart from elections or the circulation 
> of our political elites.  

Okay. I'm just saying that doesn't happen where I have evidence, but as
I said above, none of those use your model.

> dlw:I didn't say it would.  I would never call for IRV alone as a way to 
> secure minority rights or to reconcile a polarized society.  My belief 
> is that what matters most is for us to work out the right balance 
> between single-winner and multi-winner/PR elections and that the sorts 
> of options given voters in single-winner elections are of second order 
> importance.  So I would go a step further and forecast that if IRV had 
> been replaced with some other single-winner election, like Approval 
> Voting, it wouldn't have improved things.

I don't think that's a given.

Approval voting might encounter some difficulties when third parties
grow, because voters have to strategically find out whether they should
approve Gore and Nader or just Nader. There's a critical point where, if
voters vote only Nader, Nader wins, but if they vote both, Gore wins,
but if Nader had slightly less support and they didn't vote both, Bush
would have won. At least Approval voters can try getting around that
by manually taking it into consideration. IRV voters don't have that option.

I think other methods would do much better than IRV and so would support
multiple parties. I know there is little evidence either way because the
methods have been relatively untested, though. I did find descriptions
of a "nursery effect" of Range (where third-party candidates get much
better relative support when people are asked to rate every candidate),
and a similar effect for Majority Judgement based on B&L's paper.

For the sake of the data, I could present "effective number of
candidates" numbers based on the French results at
http://rangevoting.org/OrsayTable.html . This page gives the B&L results
for Majority Judgement and also as inferred by Range using the same ballots.

For Plurality, we get 4.19 effective candidates. For Range, if we
consider the normalized scores as support, we get:

Bayrou: 16.25%, Royal: 14.82%, Sarkozy: 13.24%, Voynet: 9.50%,
Besancenot: 8.74%, Buffet: 7.91%, Bové: 6.59%, Laguiller: 6.49%, Nihous:
3.92%,  Villiers: 5.46%, Schivardi: 3.32%, and Le Pen: 3.76%.

That gives an effective number of candidates of 9.58, significantly
greater than the Plurality count, with a limitation given that the
Plurality count is nationwide and the MJ-as-Range count was only within
Orsay.

Doing it for MJ is much harder because the grades aren't numeric, but if
you go with the Range generalization and call Excellent 5, Very Good 4,
and so on, and then add or subtract .25 for + or -, you get a relative
support of: Bayrou: 22.41%, Royal: 18.97%, Sarkozy: 18.97%, Voynet:
12.07%, Besancenot: 8.62%, Buffet: 8.62%, Bové: 5.17%, Laguiller: 5.17%,
and rest zero.

So for Range you have an effective number of 9.58 and for Majority
Judgement, 6.37. Both of these are greater than the Plurality count.

Now that I think about it, that might actually be used to see if the
other voting methods would give diversity. Poll people prior to
elections, asking for MJ-style ratings and/or a ranked ballot. Then
check what sort of relative support (and effective number of candidates)
you get from each rule. If you get near-zero for all but the top two,
then that rule probably would keep the uncontested duopoly - the
near-zeroes of the others wouldn't be high enough to make the majority
parties feel challenged. If you get two or three with great support and
the others all at 10-15%, then that could lead to your contested
duopoly. If it's even more fairly distributed, then it could directly
support multiparty rule.

There are a few hidden assumptions that will make the result more fuzzy.
First, the voters might answer strategically in the official
Plurality count but not in the other rules, which would overstate the
other rules' effect. Second, the voters may think that they can't vote
for no-hopes in any of the rules, or may keep themselves from voting so
because "third parties are so small", which would understate the other
rules' effect. Third, dynamics (like IRV's chaos, or cumulative vote's
vote-splitting problem wrt Range) may keep third parties down.

Also, it only works for rules where you can get some measure of a
candidate's relative support (not just "X won, Y didn't" or a social
ordering), and if one of the rules tend to give more evenly distributed
numbers than another, that might distort the relation to actual support.

Perhaps the protocol could be further refined to deal with some of those
problems.

(I got sidetracked again, but it might be useful for other EM members,
so I'll keep it.)

>> David L Wetzell wrote:

>>>   I take as given that the economies of scale in running for
>>>   single-winner offices will make it so that there are two major
>>>   parties.   
> 
> 
>> KM:I'll consider it likely that economies of scale will favor major 
>> parties. It does not, as such, need to favor them so severely that there 
>> are only two viable options for the candidacy, however.
>> [endquote]
> 
> dlw: This is probably why I want a multi-stage presidential election 
> that makes folks pick their 3 favs out of 7.  

That is at least better than unadorned IRV, and it seems that you're
already modifying plain old IRV so that it will be less prone to
behaving strangely. If you are going down that path, there are other
simple modifications you could do to make IRV simpler yet.

For instance, you might say that "when eliminating, do a one-on-one
between the two last placed and eliminate the loser". Instant Condorcet
compliance, and the alteration of the rule is no more complex than
sticking an Approval preround to IRV. So how far can you go? If you
think that other rules aren't going to improve upon IRV, why are you
sticking an Approval preround in there in the first place?

>> In IRV, voting in the way you prefer doesn't hurt you... until the third 
>> parties get large enough that they might affect the elimination order. 
>> When they do, the election can swing the wrong way, so voting the way 
>> you want can hurt you if your choice would have made a difference - i.e. 
>> if minor party candidates have any chance of winning.
>> [endquote]
> 
> dlw: But if the parties and third parties know that when a third party 
> gets stronger it can have a spoiler effect, it should lead to strategic 
> changes to prevent this from happening.  Strategic voting will still 
> happen and will tend to support there being two major parties, but if 
> IRV had been retained in Burlington, those two parties would have been 
> the Progs and the Dems.  Pubs would have started voting strategically 
> for Dems and the center would shift, relative to where it was due to 
> FPTP.  The duopoly would be contested.

Clearly the mere presence of spoilers is not enough to turn an
uncontested duopoly into a contested one, or you would already have a
contested duopoly.

The major parties have occasionally tried to make use of the
spoiler phenomenon by stealthily funding third parties oppositely
aligned to them, with the hope that these minor parties will split the
vote on the other side of the center, thus leading the strategist party
to victory. See http://rangevoting.org/CandCloning.html . Yet even these
strategies - which should lead major parties to be more fearful of minor
ones, as you hope would happen under IRV - have not turned the
uncontested duopoly into a contested one.

<inc?>

That strategy could just as well backfire. The party in danger of being
split could say "hey, these fringes are threatening to tear us apart.
It's more important, now than ever, to stay the course, to remain loyal.
Vote for us, don't let the other guy win". That's what happens in
Plurality, and it might also happen in IRV. Remember that if the minor
parties are small enough, IRV gives the same result as Plurality with
the minor parties removed, so all the major parties have to do is siphon
off enough voters to get the minor parties below that threshold.

> KM: So when we go at this from the logical point of view, the extent to 
> which single-winner methods favor major parties seems to differ. I would 
> point out that top-two runoff works, and that (as I showed in another 
> post of mine) it does give a number of effective candidates 
> significantly greater than two -- but you don't consider top-two runoff 
> a single-winner method. Since I do, I don't think single-winner methods 
> (or should we say, methods that are used to elect individual candidates 
> for single positions) have such a steep barrier that only major parties 
> can participate.
> [endquote]
> 
> dlw: It's a hybrid.  There are more than two effective candidates in the 
> non single-winner stage....

Alright, single-office methods. I still don't see the need to
distinguish the two, but what I really mean is "election methods for
single positions", be they runoff or not.

> And, even if third parties win some seats due to a top-two runoff, it 
> doesn't mean there aren't going to be two major coalitions of parties 
> that will function not unlike two major parties in many ways.  

That's what we've been talking about in the coalitions/transparency part
of the mail. I say that coalitions are unlike parties in that they are
reconfigurable, give voters feedback, and give the voters a more
fine-grained decision.

>> KM: If you disqualify the runoff method, then pretty much all you're 
>> left with is Plurality and IRV. In both Plurality and IRV nations, you 
>> see major parties winning almost all the time, even when counterbalanced 
>> with PR (e.g. Australia). So then you could of course say that from our 
>> {Plurality, IRV} based sample, all single-winner methods elect 
>> overwhelmingly from major parties, and you could say that it's because 
>> of economics of scale.
> 
>> To me, it seems these economics of scale are of the sort that are needed 
>> to compensate for the methods' failure modes. You need to unite to win 
>> in Plurality, and that kind of unification, almost by definition, favors 
>> the major parties greatly. You also need to unite to win in IRV because 
>> IRV has problems once the third parties get large enough.
> 
> dlw: Basically, We're saying the same thing.  

Then I'll follow up with "... but you don't necessarily need to unite to
win in [insert well-behaved method here]: it scales well beyond two and
a half candidates whereas IRV does not". And so we arrive at the same
point as before. You think the marketability of IRV is worth taking the
circuitous route that /may/ pay off with your particular form of PR
backing it.

>> KM:If I'm right, then the extent of major-party dominance isn't given by 
>> the single-winner approach, but by the tool used. What we'd really need 
>> would be examples of non-{IRV, Plurality} methods being used. How did 
>> the Bucklin method fare in Minnesota?
> 
> (And I know about Borda and Nauru Borda. The latter is pretty much 
> Plurality, and the former has severe cloning problems that rewards those 
> who can field the most candidates.)
> 
> dlw: The only counter example you give me from single-winners leading to 
> two major party rule is an "impure" single-winner method.  This does not 
> suggest that the use of other pure single-winner methods would end 
> two-party rule.  

You're the one who wants to exclude top-two runoff as being somehow
impure. What matters here are elections to single positions, like
governor or president. These can be done using what you call "pure"
methods or by "impure" ones, and top-two runoffs have been used in the
US before, so it's not like they're a completely alien concept to the
voters there.

You're right, I don't have any other examples than TTR. That's because I
don't have any other widely used election methods from which to get my data.

What data I do have tends to support Duverger's law: Plurality
(parliamentary or presidential) leads to national two-party domination
except where the country is large enough (India) where it leads to
regional two-party domination instead. Parliamentary PR leads to
multiparty rule. IRV leads to two-party domination even when countered
by PR (but not by your particular form of PR or fixed IRV method, of
course, because I have no data of that).

Finally, if it turns out all pure methods lead to two-party rule, then
you've got little to lose. Think of it this way: if superior method X
can break free of two-party rule while IRV can't, you should support X.
If no method can, then you'll get the same outcome (two-party rule)
whether you support X, IRV, or status quo, so go for X.

> dlw: It seems the key difference between us is how to make the two party 
> dominance more contested.  

Myself, I prefer going to multiparty rule right off the bat. I'm
discussing these things in terms of your contested two-party model to
try to show that even if your model would make two parties capable of
covering multidimensional space (both economic and
extent-of-centralization axes), IRV wouldn't give you what you want.

> I think that so long as some of the PR elections enable third party
> candidates to get elected then they can play the two major parties
> off of each other to make the overall system more fair and enable the
> rise of a minor party if the two major parties fail to adapt.  You
> think we need another single-winner alternative to get the job done,
> because IRV+PR in its limited usage hasn't been adequate at
> faciliting the circulation of the elites.

Yes. If I'm a Green voter, IRV is the Democrat party: a well-known
lesser evil, but not lesser enough to make a difference.

>> I also think that if you have to have an IRV3 second round, the first
>> round should be proportional. Otherwise, the cloning problems I have 
>> mentioned earlier may happen and thus make IRV3 itself redundant.

> dlw: But with IRV3/AV3 then if there was a narrow victory between the 
> third and fourth place candidates in the first stage you could do the 
> tabulations for two different sets of three finalists and only have to 
> do a recount if their final outcome in the second stage differed between 
> the two.
> As for PR in the first round, cloning problems can be resolved by 
> increasing intra-party discipline and developing ways to fight 
> skullduggery with exposure of campaign finances.  What AV3 has in favor 
> as a first stage is simplicity.  It's damn easy to tabulate the number 
> of rankings each candidate receives, just as the number of votes get 
> tabulated with a FPTP election.

It's also damn easy to just use Approval. "Count all the votes", then
pick the person who got most.

The whole clone-resistance thing is somewhat off topic, so I'll cut it.
My point was really just that if you use a majoritarian method to pick
multiple eligibles, and there are clones or near-clones of the winner,
they will all win. (See also Walabio's remark regarding the Wikimedia
foundation.)

> As for the Dems crowding teh 3rd stage, if they're the more popular 
> party, it doesn't really matter that they are the only finalists.  
> What does matter is that the Democrat who wins will be more appealing to 
> Rs on average than would be the case if it was FPTP.  

Sure, but that'll make the IRV stage even less important, so cut it out
and just use the AV results. We already know, from the AV round, which
Democrat is most appealing to the Republicans since he got the
Democratic approvals *and* the most other approvals.

> dlw: I believe the top two runoff in France has retained the existence 
> of two major parties, albeit with some significant minor parties and 
> lots of me-threes and so I don't see it's use as upsetting my point. 
>  They use hybrids between single and multi-elections.  This is why 
> there's scope for more parties and some who play a significant role.

If you think TTR has retained the existence of two major parties with
some significant minor parties, then would not that be a better fit for
your contested duopoly model than IRV? The data I've shown says that
French parties are more evenly matched than Australian ones, as more
third party candidates get elected by TTR than by IRV. Is really IRV
that more marketable, or your particular form of PR so good that it'll
make (your PR + IRV) better than TTR?

Or, to be more concise: you say TTR is a hybrid. If you accept hybrid
methods, okay. If not, why are hybrids unqualified but pure ones acceptable?

>>    dlw: I'd rather use 3-5 seat forms of PR in "more local" elections
>>    to ensure plurality.  I'm okay with using IRV for "less local"
>>    elections, because I value both plurality and hierarchy at the same
>>    time.  

But you just said that TTR gives both plurality (some significant minor
parties) and hierarchy ("retained the existence of two major parties").
So, by what you're saying, TTR too should be valuing both plurality and
hierarchy.

> dlw: I believe Germany's MMP system uses Plurality plus PR.  
> IRV ties the two major parties more so to the center.  It allows thrid 
> party candidates to raise important issues in "less local" elections and 
> get treated with respect by the major party candidates.  It's a clear 
> improvement over FPTP, it just doesn't end effective two party domination.  

Germany uses Plurality plus compensating PR. That is, if party X gets
40% of the FPTP support but only 20% support overall, X doesn't get 20%
of the PR vote - it gets enough to make its total share of seats 20%.

With compensating PR, you can use any method you'd like. You could even
use Borda, because even if the other method errs egregiously, the
compensation will make up for it. Compensating PR with a bad
single-member method does hurt geographical representation but it
doesn't hurt ideological (at least per-party representation).

So I suppose I could accept IRV with compensating PR. I don't trust
FairVote to leave the "compensating" part alone, though, because they
think IRV is the end-all be-all, or at least market as if it were.

> KM:Is IRV3/AV3 single-winner or just single-office?
> 
> dlw: I'd say it's single-winner by your criterion.  The first stage 
> tabulates the votes differently than the second stage.  

Even stranger. It's a multistage method but it's single-winner, yet TTR
isn't single-winner. Let's just cut this distinction altogether and work
with only single-office methods, no matter what goes on inside the box
that is the election method. :-)

If you want to make a distinction, let's make it this: "proper"
single-winner/single-office methods involve only the voters. Other
methods involve the candidates after election (parliamentarism, for
instance). Though even for this, some other EM members might object,
pointing at Asset and SODA.

> dlw: It's not strictly summable, because there's the back and forth to 
> determine the three finalists so that the rankings information can be 
> sorted into the ten relevant categories.  But the 10 categories can be 
> summarized at the precinct level.  

Indeed. IRV advocates have said that summability doesn't matter, either,
because IRV is "almost-summable" in a similar manner: the center can
gather first place counts, then tell the districts which candidates to
eliminate, and rinse and repeat.

>>>  dlw: It's also possible that one can bring too much info into the
>>>  open and thereby gain an degree of obscurity.  Lots of inevitable
>>>  compromises in politics are easier to work out when there's less
>>>  transparency.   

>> KM:I can only speak for my own country, but that has never been a problem
>> here. I haven't heard of such being a problem except for places where
>> they have other, more serious, problems too (like the participants of
>> the political system trying to game said system by any means available 
>> and so actively muddy the situation as much as they can).
>> Perhaps Juho can tell us whether it's a problem in Finland, but I don't 
>> think it is there, either.
>> [endquote]

> dlw: It's easier to do that with smaller country with a relatively 
> homogenous culture. It's also hard to move directly from a two-party
> dominated system to a system such as yours. This is part of why I
> focus on alternatives that retain two-party domination but prevent
> single-party domination. It is the destructive cut-throat competition
> from the tendency to single-party domination in the US that has 
> caused a lot of the problems that people pin on us having a two-party
> system.

Okay, so you support a contested duopoly in favor of PR on pragmatic
grounds. That's fine, and I've said as much - even argued that IRV won't
give what you want even if you want "only" a contested duopoly. In this
section, however, we were talking about the advantages of PR in general,
not as a pragmatic next step. So in the long view, I think the
transparency of PR is more good than bad.

(As a side note: Norwegian society has become less homogenous - though
probably still more homogenous than continental Europe - the last
decades, due to immigration. This has, predictably, led a right-populist
party to grow as has been the case in continental Europe. However, they
have never really held governmental power, and have been limited to
being part of the opposition in the legislature. Here, PR works in the
moderating way it is supposed to: they have some influence as
opposition, but they don't have enough support to govern, and they're
not moderate enough to form a coalition.)

> dlw: Even if I agreed about the desirability of such a system, it's not 
> in the cards for the US.  

Not yet, at least. Perhaps it would realign by itself if you got PR with
a sufficiently good single-winner rule not to block it.

We shouldn't impose presidentialism or parliamentalism upon the people.
We should give them tools that are good enough that the voting dynamic
gets them where they want to be (and then changes when the people's
opinion changes). From the data I can get at, IRV isn't it.

> BTW, did you watch the documentary film "too much Norway"??  I liked it. 
>  I'm mainly Swedish-American.

I haven't. Maybe I should take a look when I have the time!

> The US is on the verge of a significant major realignment of its two
> major parties, IMO. We'll still have two major parties, but there'll
> be two different parties. All you need is a catalyst issue that
> splits both the existing parties. Methinks that the use of PR in
> "more local" elections to change our country's political dynamics and
> circumvent $peech by increasing the number of competitive elections
> could be the issue that brings about a change in regime.

Then there has to be something about your own type of PR that increases
the chance of such a realignment happening, to such a degree that it
overpowers the IRV-component's tendency to limit that realigment - a
tendency that isn't overpowered by other forms of PR.

>> KM:And if you want to optimize a certain metric, it's best to directly 
>> optimize it, rather than something that is corelated with it. If you 
>> want transparency, I think it's better to make the political system 
>> transparent than to try to find some change (IRV) that leads to a change 
>> (handicapping) that leads to transparency. If any link of that chain 
>> fails, the whole thing fails, so the fewer links, the better.

> dlw:The use of 3-5seat forms of PR in "More local" elections isn't a 
> long chain.  That's what I'm betting on far more so than IRV to improve 
> the state of the US's democracy.  

So let's reach a consensus. Drop IRV. Focus on PR. The proportional
representation organizations of the 20th century didn't need IRV to get
local PR, nor to keep that PR until the major parties used time-specific
dirty tricks (red-baiting).

>> KM:The voters don't directly determine what the coalitions are like. The 
>> voters don't veer off from one extreme to another, either, so they have 
>> some idea of which coalitions are realistic.
> 
> dlw: that'd be the kicker.  

PR is much more accurate than Plurality in reporting support, I think.
If 20% of the vote means 20% of the seats, then people vote for other
parties than the big-two. A side effect is that it's possible to see how
popular the parties actually are, rather than how popular they are
filtered through the lens of strategic "lesser-evil" Plurality.

So PR will help in giving the voters some idea of what coalitions are
viable.

>> Tea-partyists or Occupiers have little chance to alter the power dynamic 
>> directly by voting in the current system, since they would vote 
>> Republican or Democratic respectively anyway - the only way they *can* 
>> change things is either internally (where the tea party has had quite a 
>> bit of success) or by threatening not to vote at all.

> dlw: I'm guessing that's what a good deal of the Occupiers are going to 
> do: only vote strategically, based on strategically chosen electoral reforms.
> 
> I'm trying to get some of them to rally around the use of American forms 
> of PR.  If they wanted to get revenge on city hall for treating them so 
> badly, they could focus on really local electoral reform, with PR for 
> city council elections, and a two stage mayoral election where the 
> second stage would be by the quasi-PR-elected city council members)

If you can get local PR again, all the better! Just don't chain it to IRV.

>> Another pattern that might emerge is that one party appeals to one
>> group, the other to the other. Then the first party says "these guys
>> are all corrupt, as you've seen, so vote for us instead". They win
>> and resume their corrupt practices with the second party in 
>> opposition. Then the first and second party switches places and the
>> same thing happens with signs reversed. Each party has a solid base
>> to which go the spoils, and an undecided middle who vote for "the
>> other party" (whoever was in opposition the last time around) because
>> they want things to change. No party wants to permanently claim the
>> middle group by becoming more honest, because they gain more by
>> perpetuating the system than by cleaning themselves up.

> dlw: But when there are many such bases of single-seat elections, like 
> say 50 states, then it becomes harder for the two parties to keep up the 
> shell game.  Cuz, the system in fact doesn't require that the two 
> dominant parties are the same in every state.  As such, there's scope 
> for outsiders to come in and threaten the duopoly.

The system would be weakened if the two dominant parties aren't the same
in every state, yes. But again, the Australian state elections tend to
have the same dominant parties.

(In all fairness, the Australian parties don't run that kind of shell
game, but they haven't been displaced either.)

> dlw: undoubtedly, but this has always held true in part, not in full, 
> and it has not stymied MLKjr-like movements that have succeeded in 
> moving the center in ways that forced both parties to stop whistling 
> dixie...

Sure. You can only deny reality for so long before it shifts under your
feet. It's better for the system (and, I would argue, the people) if the
system forces the parties to keep themselves attached to the reality of
the people's wishes instead of periodically slipping, though.

>> KM:This might be fixed by making the field competitive enough that the 
>> model doesn't pay, but I can't find evidence that IRV gives honest 
>> parties a chance to unseat the old major parties and become new major 
>> parties. At least I know that a multiparty system will give competition, 
>> if for no other reason than that it's much harder to buy off ten parties 
>> than two.

> dlw: I do not believe electoral reform can end the fact that "modern 
> democracy" is an unstable mix of "popular democracy" and 
> plutocracy/kleptocracy.  But I believe it can bolster "popular 
> democracy" if it is accepted that plutocracy/kleptocracy is going to 
> continue.  I stand by the notion that "PR" for "more local" 
> elections(that is set up so third parties can play a constructive role, 
> able to win some seats and play the two major parties off of each other) 
> and IRV3/AV3 for "less local" and multi-stage for the "least local" 
> election would be a potent mix.  This is not the Australian model.  

No, but it is IRV + PR. Your type of PR has to be significantly more
different than the other IRV+PR combinations for "this time, it's
different" to apply. The other IRV+PR combinations do have diversity in
the PR chambers, but even that fails to make the IRV positions diverse.

>>     Perhaps this is akin to the extent to which one would expect
>>     strategic behavior to appear in single-winner elections.

> Yes it is.  It's also decent marketing to simplify the choice so the 
> reformers don't get divided and conquered by the Stalwarts.   

I don't see the relation. "This" in my quoted mail was about how people
may strategize more in PR when the process is seen as something to be
gamed and won at all costs, and that this might apply to single-winner
elections too. I'm not talking about different single-winner methods,
but about how people approach them -- and I have argued that IRV does
badly with even honest votes (because of center squeeze and chaos). You
might argue that IRV is better at repelling strategy, but you can get
Condorcet compliance (which helps with a lot of the ills of IRV) and
still retain strategy resistance, as Green-Armytage's papers show.

>> dlw: I guess that's why I'm psyched up about pushing long and hard for 
>> American forms of PR more than any particular single-stage rule.  I 
>> believe that it will refocus folks to taking care of their own gardens 
>> in ways that will trickle up to make important changes possible at the 
>> top.

Either direction might be good, but indeed, I have no objections to
trying to get PR by itself.

>> KM: Let's have proportional control, not bang-bang control. It's gentler 
>> on the system and reaches the (changing) target more quickly.
> 
> dlw: Let's move from where the US is right now, not aim for what we 
> think is the sun.  

Okay then, let's move from where the US is right now into a domain of
PR. By your reasoning, we can leave single-winner reform until PR is done.

After it is done, let's investigate which single-winner method to use,
now that third party support makes it obvious that Plurality doesn't
work. That investigation might be done by appointed groups, by
organizations, whatever. The important thing is not to lock yourself
into IRV early. If you do, the others may say "why have an investigation
into single-winner reform? We already have IRV!", or they may have said
"this IRV upon which you want to base your PR... it gives the wrong
results, we're going to repeal it".

Ideally, we'd move in parallel: a good single-winner method and a good
PR method. But it's okay. I can hold on the single-winner method. Just
don't advocate a good PR method and a bad single-winner method, or you
may get what you wish for.

>>> dlw : I agree.  This is why I'm pushing for the electoral reform
>>> that is most likely to be adopted the soonest.

>> KM:If it isn't good enough, you'll simply have traded off gradualism in 
>> the political process in general for gradualism in adopting the election 
>> method. Since there are many more decisions not involving elections than 
>> there are decisions involving them, the benefits to the smooth 
>> transition of the election method is outweighed by the disadvantages in 
>> having to wait for another extra-political movement to correct it 
>> properly the next time (or for extra-political movements to fix what 
>> would otherwise be fixed through the political process).

> dlw: If a multitude of LTPs are fostered then they'd be a ferment of 
> ideas.  We wouldn't have to wait so damn long. 

I think you'd have to wait longer with IRV. Either it backfires
(Burlington - even if you think it's due to moneyed interests, IRV gave
them enough targets to fire at), or it simply acts as Plurality with the
minor parties removed (with a dash of random oscillation if the minor
parties get too large). In the former case, IRV's setback is electoral
reform's setback. In the latter, you've just made it harder to get a
single-winner reform that amounts to something, because people will
associate single-winner reform with IRV, which you already got.

>> KM:Further, if the method isn't good enough, the political movements may 
>> end up moving away from it. "We tried that, it didn't work", as I've 
>> said before. FairVote's attempts to link the ranked ballot and IRV (such 
>> as by using the name "Ranked Choice Voting") would make such a reversal 
>> more likely. "We tried the ranked ballot, and IRV didn't work/just gave 
>> us weird results, so forget about it" - something like that.
> 
> dlw: That isn't a common outcome and we can't assume the ground has been 
> poisoned by a few bad apples.  

If there were only a few bad apples - and perhaps you could then say
that IRV's strange behavior only happens some of the time, and it still
gives third parties enough of a chance to be worth the change. But IRV
has been replaced with Plurality or runoffs in more than just a few
places - let's look at some of them.

You have Burlington. Little more needs be said: reverted after the
second election, where IRV didn't pick the CW.

Aspen, Colorado, started to use IRV in 2007. The voters then, in 2009,
rejected keeping it, and in 2010, reverted to runoff in a referendum.
See http://www.aspentimes.com/article/20101103/NEWS/101109959 for that.
Note in particular the mention of non-monotonicity, and of one of the
critics saying traditional runoff lets people know the candidates better
(i.e. change their opinion between rounds). IRV was sold as a
replacement to TTR and didn't suit the task.

In 2007, Cary, North Carolina also used IRV. They elected not to
participate in the IRV pilot after this (in 2009). Although the post may
not be neutral, I didn't find any others and the CaryNews page is down,
so see
https://votingmatters.wordpress.com/2009/05/17/cary-nc-tries-irv-then-says-no-more/ 

 Again, IRV was marketed as a replacement for TTR and failed in that
task - in particular, IRV's "majority" winner was no majority winner at all.

In Pierce County, WA, the voters passed IRV by 53%. It was challenged
three times and upheld until 2009, when the voters repealed it. Earlier
in 2009, the county's auditor said that repealing IRV might save money
(see http://tinyurl.com/cap9btx), and the county council decided 6-1 to
repeal (though they had to put the question to the voters afterwards,
which they did. See http://tinyurl.com/brlemo9 )

Even the first US uses of IRV (in 1912 party primaries, though it's
closer to contingent vote) were reverted: none of the elections picked
anyone but the Plurality winner, so the parties didn't see the point.

So I'll readjust a little. It seems that IRV doesn't manage to show that
it's "worth it". It elects Plurality winners often (which we'd expect if
it was essentially "Plurality with too-minor-to-win parties removed"),
and when it doesn't, it can easily get it wrong (as when it elected Kiss).

>> KM:Well, that would get around most of our disagreement. If you use a 
>> proportional method like STV, then IRV never enters the picture, while 
>> you also get the election form you want.

> dlw: I meant STV in the first round to pick the 7 finalists.  IRV could 
> still be used in the final round...

How about using STV, then just use Approval after it? You still get your
STV and thus your PR, but you don't have to put your faith in IRV.

> It's 3 stages.  STV in the first, everyone picks their 3 favs of 7 
> candidates in the 2nd(along with picking three electors for their 
> congressional district), and then voting by proxies until there is a 
> majority of electors in favor of one candidate in the 3rd.  

Is it STV, then Approval, then IRV, then electoral college/proxies -- or
just STV then IRV then proxies?

>> Instead, one should try to show better systems to the friendly
>> politicians. Being politicians, they don't have time to go into
>> detail and find out which of the many ranked methods are good, but
>> that's where the declaration comes into play.

> dlw: No, that's where the strong push for American forms of PR comes 
> into play.  It's simpler than pushing 4 other alternative 
> single-winner/single-stage elections.

Signing/supporting the declaration and working for PR isn't mutually
exclusive. The declaration says, among other things "PR is nice but it
isn't the focus of this document".

> And now is a great time to push for election rules that'll handicap
> the rivalry between the two major parties and enable third party
> candidates to get elected. This can be done without ending two-party
> domination, which is why that's the path of least resistance.

If IRV handicaps the rivalry enough. If.

> dlw: I don't dispute that, I was proffering why we shouldn't focus on 
> getting the right single-winner/single-stage election rule.  Methinks 
> LTPs will expedite the adaptation of voters but not make them change 
> their opinions really quickly, since they'd be able to take advantage of 
> their small size to proceed more often by consensus, and make them 
> harder for the major parties to control.  

Okay, so go for PR. Leave IRV alone, and don't swallow IRV just to get PR.

>> KM: Furthermore, if people change their opinions quickly, their standard 
>> deviation might be greater - and a greater sigma amplifies the oddities 
>> of IRV. To see this, consider the opposite situation, where the standard 
>> deviation is so small everybody agrees on the complete ordering. Then 
>> every nondeterministic method that passes the unanimity criterion would 
>> elect the same winner.
>> [endquote]
> 
> I don't anticipate wild fluctuations in people's political preferences.

Then it's all the more important that the rule gets it right - that the
center corresponds to the voters' center, not to a shard of strange
behavior, or an island where some other candidate wins.

> dlw: The politics of electoral reform does not support that happening 
> any time soon.  They'd be smarter to push for switching the use of PR in 
> between the houses...

What do you mean by "in between the houses"? Local elections? Note that
I was talking about Australia. I don't think you'd get PR in both the
Senate and House, right off, in the US. The model most fitting with what
bicameralism was intended as, I think, would be to have the House of
Reps by PR and the Senate either picked by the state legislature (that
way you avoid IRV vs Condorcet, etc., but you'd also have a
constitutional amendment to make) or a good singlewinner method (by
which I don't count IRV).

> dlw: I think we agree a lot on this.  I know I wrote about this before 
> above, but 1. STV, 2. 3 of 7, And for the third stage, we could let the 
> Electoral College keep on voting (not unlike the bishops selecting the 
> next pope) until one of the three finalists gets a majority.  That 
> should do the trick.

Replace 2. with Approval (or any of the better-than-Approval rules like
MJ or advanced Condorcet) and it's good. I might quibble that you
wouldn't need STV in that case, but it's minor in comparison - keep STV
if you want, it's no big deal compared to IRV.

> dlw: I said there'd tend to be up to 4 strong candidates, due to the 
> cost and benefits of running for a single-winner office.  I want more 
> uncertainty about who the top 2-3 will be at the onset of an election, 
> not wholly unlike as is happening right now for the Republican party 
> nomination of a presidential candidate.  My point is that the purported 
> diffs among single stage, single-winner elections get reduced 
> considerably when this is the case.  I do not foresee it changing in the 
> near future in the US and so I'd rather stick with IRV as the 
> front-runner alternative to FPTP. 

Three-candidate Yee diagrams for IRV (as pictures of the dynamic frozen
at the time of election) have their share of strange shapes, too. So I
disagree.

> dlw: It's a matter of degree of emph.  I myself think progressives 
> should have rallied hard to keep IRV, since the system undoubtedly would 
> have corrected itself....as stated before my interp was that the 
> election of the Prog was the birth pains of the formation of there being 
> two different dominant major parties in Burlington.  

It would have corrected itself if the voters noticed that they shouldn't
be voting honestly - and in particular, that the Republican voters
shouldn't vote R first because that makes More-Liberal win. I'd rather
prefer that the voters didn't have to twist themselves or their
expressions to fit the voting method.

(Alternately, it would have corrected itself if the Republicans scaled
down their efforts - if the party, rather than the voters, reasoned "if
we keep on doing this, More-Liberal will win, and we rather want Liberal
to win, so we better not interfere.")

>> KM: IRV's artifacts didn't stop there, and so gave the campaigners more 
>> ammunition. I'd say IRV's failure to elect the Democrat was a good 
>> example of its center squeeze, as well - its tendency to pick the wing 
>> with the greatest support rather than the center with even greater 
>> support because the center's votes are obscured by first preferences for 
>> the wings. The Republican candidate was, in a sense, a spoiler because 
>> his presence hid the Democrat that would otherwise have won.

> dlw: I'm sorry this is where things get ugly among electoral reform 
> advocates.  My first thought was to respond that there was strong 
> monetary support for folks to bring up such artifacts, with less than 
> altruistic motivations.

Sure, but you can't bring up what isn't there, and you can't do it if
enough people are satisfied enough with the outcome. The anti-PR
arguments in New York didn't attack the composition of the method, but
rather tried to convince people that "more diversity" was precisely what
they didn't want - in the more extreme cases, that STV was a Russian
poison that would fracture US society. That's not an argument against
the workings of the method. That's an argument against the result.

>> (And as a correction: Burlington doesn't have FPTP, it has top-two 
>> runoff with an odd 40% threshold instead of the more sensible 50%.)
>> [endquote]
> 
> dlw: Well that's better than FPTP.  
Agreed. A Burlington group tried to get it to 50% rather than 40%, but
the proposal failed. A right-wing page tried to associate the 50%
organization with IRV (by the logic that 50% leads to runoffs every time
which would lead to renewed call for IRV), but I don't know if that
reasoning was widespread enough to have any effect. I don't live in
Burlington, after all.

>> KM:As for the weird results, those can happen when the election is hard 
>> to call (close to the border of the cell in the Yee diagram given by 
>> candidate positions at the time of election) or voter opinions move 
>> slowly (for their sigma) so that the center-squeeze and nonmonotonicity 
>> phenomena show up. While not ubiquitous, the latter puts strain on IRV 
>> precisely when it has to meet the challenge - when it exits the 
>> Plurality domain and minor candidates start to matter.

> dlw: But a lot of them are not consequential if they're essentially 
> "sour-grapes" phenomena, like with non-monotonicity.  If a large number 
> of Republicans had a serious change of heart from supporting their 
> candidate to supporting the Prog candidate at the other end of the 
> spectrum then they might have gotten the Democrat candidate elected. 
>  Yes, that could happen theoretically, but it is not by any means a 
> common occurrence.

Non-monotonicity is as much, if not more a "dual election" error as a
strategic weakness. If you improve X's condition yet X loses, then the
method got it wrong either in the first place (when X didn't lose yet
should) or in the latter (when X did lose yet shouldn't by consequence
of the first), and you can't know which is the wrong result, so it might
be the real election.

As for frequency, http://rangevoting.org/Monotone.html says monotonicity
failure occurs under impartial culture once about every 7 elections
(14.5% of all elections) or 10% in a 1D left-right situation. These
figures are all for 3-candidate elections. Now you could argue that
these all happen in the unrealistic area of voting space (i.e. for
ballot combinations far from real elections) and so overestimate, but
then I could argue that they all happen in the realistic area of voting
space and so underestimate. Neither of us have any evidence.

>>>    dlw:The big thing with FairVote is that they get people out of the
>>>    FPTP stupor, which is a major step in the US!!! 

>> KM:FairVote gives by showing people that an alternative to FPTP is 
>> possible. They then take away by telling people that IRV is both a good 
>> alternative and the only alternative. I don't agree about the former, 
>> and the latter is somewhat of a self-fulfilling prophecy. FairVote 
>> markets IRV so heavily that their actions serve to turn IRV into "the 
>> only alternative".

> dlw: IMO, their marketing campaign is pragmatically based.  There are an 
> infinite number of single-winner electoral rules.  Y'all on this list 
> have managed to get it down to 4 other alternatives to FPTP.  That's a 
> lot for most folks to take in.  They've framed the debate the best for 
> the US's FPTP system.  Outsiders like you all are muddying the waters 
> and not providing a single clear-cut alternative.  Later on, there'll be 
> more scope for using or considering a variety of election rules for 
> different elections.  

They do seem to be pragmatic about the facts, though.

For instance, they marketed IRV as "the" ranked-ballot method, calling
IRV "Ranked Choice Voting" as if there were no other method of
processing a ranked ballot. Later, they've found it necessary to mention
some other rules on their pages, but they still keep trying to link the
ranked ballot with IRV.

Also, they claim that IRV will always give a "majority winner". This
evidently means "a winner where a majority prefers this candidate over a
specific other candidate", in other words, a Condorcet non-loser. That
is quite different from a true majority winner - a candidate who has a
majority (the same majority) preferring him over all other candidates.

Furthermore, in certain cases IRV doesn't even give a majority winner by
FairVote's definition. This is in particular true when IRV is run on
"three ranks only" machines. Since there are many truncated ballots, the
winner is only a "majority winner" by the ballots that rank both the
candidate in question and whoever remained in the last round. Australia
gets around this by mandating full ballots, but that's not a very good
solution.

They also play a bit loose when they say, regarding nonmonotonicity,
that "these mathematical paradoxes that while in theory are interesting
for mathematicians to doodle around on their sketch pads, in fact have
no basis in reality... it's also possible that a meteorite will strike
Earth and wipe out life... but probably not for a few more million
years"; or when they say that Burlington went off "without a hitch" and
there were no spoilers.

Later, they've then said that of course there were no spoilers, because
they define spoilers as minor party candidates who are so weak that they
have no chance of winning, but still affect the outcome. If you make
"minor" minor enough, and define minor by first preference votes, IRV
will eliminate all the minor candidates first. By that definition, IRV
will never have problems with "spoilers", but that's kind of
tautological in their altered definition of a spoiler.

All of this might make sense if they've set their sights on IRV and then
want to get it, come hell or high water; but stretching the facts like
this doesn't make me more convinced of their cause. There's such a thing
as going too far.

>> dlw:I think a key part of politics is about recognizing the need for 
>> hierarchy.  FairVote is the leader of electoral reform in the US.  I'd 
>> rather encourage them to push for IRV3/AV3(at least in bigger elections) 
>> and support their push for PR strongly than to ask them to back off from 
>> what's been a significant work-horse for them.  

Then I'll say, at least for IRV: "It's better to vote for what you want 
and not get it than to vote for what you don't want and get it". I don't 
think IRV will make a difference.

(Well, I'm not American, so I can't "vote" in this sense. But if I were...)

> dlw: Why not say,  "the use of PR in "more local" elections will create 
> a greater ability for third parties to spoil single-winner elections, 
> thereby increasing the demand for single-winner election reform.  Right 
> now, the plurality of support among electoral reform activists is for 
> the use of a form of IRV to replace FPTP.  We think that will change 
> later down the road, since there are other options, but we'd rather just 
> stay united in pushing hard for American forms of PR than cause dissent 
> over an issue that is secondary in importance.  

Because I think they can get locked into IRV, and because I think IRV
won't break up the hierarchy enough even to make a contested duopoly,
and because I don't think your particular form of PR will be so
different that it can succeed in giving diversity to IRV elections where
Ireland's and Australia's failed.