[EM] STV seat count, and start small and locally

Kristofer Munsterhjelm km_elmet at lavabit.com
Fri Mar 9 11:12:56 PST 2012

On 03/03/2012 07:59 PM, Richard Fobes wrote:
> (My comments are interspersed because there are multiple topics here.)
> On 2/29/2012 2:02 PM, Kristofer Munsterhjelm wrote:

>> I'm not entirely sure what you're saying. If you're saying that you
>> can't have more than two seats per district and still have
>> proportionality unless you use party list PR, that's obviously wrong.
>> But if you say that you can use an MMP compensatory mechanism to get
>> proportionality beyond the effective threshold, then I get what you're
>> saying. So I'll assume that :-)
> I am very much against party lists.
> I am saying that there is a limited degree of proportionality that can
> be achieved without asking voters to indicate their favorite political
> party.
> Yes, an MMP-like compensatory mechanism (that makes use of
> ballot-expressed party preferences) improves proportionality -- in ways
> that STV cannot (because STV does not look at party preferences).

I see that; the reason is that the voters would not accept having to 
rank enormous numbers of candidates that say, a 20-candidate STV 
election would give. One may deal with this by letting the voters vote 
on some sort of high level choice, and then extrapolate proportionality 
from there. That is what MMP does (where the high level choice is by 
party), and also what delegation methods do (where the high level choice 
is given by the candidate ordering or opinions).

I'm not sure if the disproportionality is so bad at 5-7 seats. It seems 
to work in nations where they use it. I also look differently at "STV 
does not look at party preferences", as I have said before. Since STV 
does not look at party preferences, it can give proportionality by what 
the voters want rather than just proportionality by party, and 
therefore, it seems reasonable to me to have STV as broad as one can 
have while not burdening the voters too much.

>>> Using STV to fill more than two seats would lead to very unfair results
>>> in some situations. Those situations don't exist now, but they can (and
>>> I believe would) arise if the voting system changes (such as adopting
>>> STV).
>> I think you said minor parties could get undue power in three-seat
>> district STV with the two parties + minor situation that you have today,
>> but I also guess that's not what you're referring to (since you say
>> "those situations don't exist now").
>> So what kind of unfairness are you envisioning? STV with five-seat
>> districts seems to work where it's been used, in the sense that it does
>> produce multipartyism and the voters don't complain about vote
>> splitting. At least if they do, I don't know it.


> Note that two-seat STV can be used in conjunction with MMP "compensatory
> mechanisms." These results would be reasonable in the United States
> because of the two-party dominance.
> What would not work well would be 3-seat, 4-seat, or 5-seat STV (either
> with or without MMP compensation).
> Another way to express this concept is to refer to it as roundoff error.
> Specifically, one single winner cannot represent all the voters in a
> district. In a similar way, three winners would be very disproportional
> in a district that has a balance (of voters) between just two political
> parties.

Yes, but you said the problem with large-seat STV would not come into 
effect until there were more than two parties. The rounding effect you 
speak of would be a problem with two parties but even out to a greater 
extent when you have multiple parties, thus be worst under two-party rule.

Also note that the rounding error disproportionality would become lower 
with many seats. If you have two seats, then in the worst case (not 
considering Banzhaf-type voting power but raw support) is that 33% are 
not represented (if they were >33%, they would get the second seat by 
the Droop proportionality criterion). For one seat, the corresponding 
number is 50% (majority criterion). For five, 16.7%.

But perhaps you argue from a voting power point of view. That is, you'd 
say something like: "if I have an odd-seat district and R has a majority 
minus two votes, and D has the rest, then D gets a majority of the seats 
and thus absolute power, which is not very proportional". That is true. 
Yet increasing the seat size will shrink the margin in which you'd get 
that kind of outcome, and if it's still a problem, you could use an even 
number of seats or a maximally composite number (e.g. a factorial like 
6) to make ties among the voters translate to ties in the council. As 
the margin shrinks, the likelihood that swings don't cancel out in 
different districts will also decrease. Nations that use plain 
multiwinner systems don't seem to have gerrymandering, so the swings 
would be more random.

>>> In other words, I agree that reform must start at the "local" level, but
>>> I think that some state-level changes would fit your idea of "local". (I
>>> don't know if there are cities that are ripe for proportional
>>> improvements.)
>> I know too little about US politics to comment, but if you're right,
>> that's good, and I hope your strategy can work :-) Do you have any
>> specific plans on how to advocate substantial electoral reform in Rhode
>> Island?
> One strategy is to educate as many people as possible about the
> existence of the Declaration of Election-Method Reform Advocates.
> I'm also pursuing other strategies that I'll reveal as appropriate.
> I'm not focusing any direct attention on the Rhode Island situation.
> I'm aware of it because last spring there was a demonstration going on
> at the Rhode Island state capitol building when I was across the street
> at the train station waiting to be picked up.

Okay. I agree that educating people about the existence of the 
Declaration is a good thing.

>>> I'll add that here in the state of Oregon there was a ballot measure
>>> about adopting an "open primary", so there are opportunities to adopt
>>> election-method change at the state level if it's the right change. (It
>>> failed; I opposed that change for what I hope are obvious reasons.)
>> Was that a partisan or nonpartisan open primary?
> Here "open primary" refers to having all the primary-election candidates
> in the same race and then choosing the two candidates who have the "most
> votes" to compete in a runoff election.
> So few people understand what "most votes" really means, so lots of
> voters thought this would be a good idea. Afterwards I talked to one of
> the backers of that measure, and he had no clue about how voting really
> works.

As far as I understood, you still have to vote for the same party for 
all positions in an open primary (i.e. partisan), but if it's a blanket 
primary (nonpartisan), then you can vote for different parties' 
candidates for different positions. I may have been mistaken, though.

If it is truly a "vote for anyone you want, then vote between the two 
winners", then that sounds a lot like top-two runoff, and top-two runoff 
doesn't have the centralizing (two-party-genic) power of plain 
Plurality. Is an open primary different from top-two runoff? If not, why 
is it a bad thing compared to ordinary Plurality with closed primaries?

More information about the Election-Methods mailing list