[EM] British Colombia considering change to STV

[Kristofer Munsterhjelm]

Raph Frank wrote:
> On Sat, May 2, 2009 at 11:33 PM, Juho Laatu <juho4880 at yahoo.co.uk> wrote:
>> It practice that seems to set the limits
>> to max 4 and min 2 parties/groupings per
>> constituency represented in the Dail.
>> <http://en.wikipedia.org/wiki/Members_of_the_30th_D%C3%A1il>\
> 
> The small constituency sizes do hurt the smaller parties.  This is one
> of the big issues I would have.

I think Schulze's MMP idea would work well here. Use STV (or some other 
neutral method) for district seats, then "top up" by nationwide MMP. His 
concept includes ways of fixing the decoy list problem (basically, 
downweighting votes for a party if it got more top-up seats than it 
deserves).

Because the base method is STV, one doesn't need to have very many 
top-up seats, so the complaint that the system elects representatives 
one didn't want ("I voted for party X but only got the party elite!") is 
weakened.

It's kind of a hack, and it only ensures party proportionality, but 
people probably aren't going to want to rank 10+ candidates required to 
get proportionality the direct way.

So, a proposal might be: have an average of >5 seats, with a minimum of 
2 or 3, (and perhaps also a maximum of 10 or 11, so people won't get 
exhausted trying to vote in the megadistricts). If needed, have a 
commission redistrict, or do it algorithmically; or just use the natural 
first-level borders ("counties"). Then, for the next level up from the 
districts (may be national, or may be an intermediate level like 
raions), add a small fraction of top-up seats, perhaps a tenth of the 
number of seats encompassed by that particular intermediate-level district.

>>> Party list systems aren't neutral at all.
>> Yes. Or one could say that they may
>> allow votes to individuals but they do
>> not allow voters to define any arbitrary
>> inheritance order of the vote (unlike in
>> STV).
> 
> Well, the assume that the party is primary.  The inheritance ordering
> is decided by the party rather than the voter.
> 
> The tree system is a possible compromise as it allows the candidates
> some say in what branch of the tree they are in.  At least that the
> candidate will be closer to the voter than the party leadership.
> 
> I think a candidate list system is better though as it allows more
> general inheritance ordering.  Ofc, it is always going to be a
> tradeoff between precision and complexity (both for the count and for
> the voter).
> 
> Closed party list
> Open party list
> Tree based lists
> Candidate list
> PR-STV
> 
> All, except PR-STV could be handled at the national level.

How about this? A voter picks a party, and also, if he wants to, ranks 
their candidates as in a single-winner system. For each district, the 
party runs the method (probably some sort of Condorcet method), and the 
winners get the seats.

For instance, consider a county that gave three seats to the yellow 
party. The voters voted so that the social ordering within the party is:

Y1 > Y2 > Y3 > Y4 > Y5.

Then Y1, Y2, and Y3 are the party representatives for that district.

There are two problems with this idea. First, strategic voters may vote 
for a party they don't support, just so that their ranked vote can move 
a disliked candidate first (same problem as in an open primary, but 
weaker). Perhaps there would be a way of making a method that ensures 
that if some voter votes for a party, he gives them more power than he 
can take away by the ranking.
Second, it's not proportional within the party. If the Y party gets ten 
seats, those will all consist of ten Y-centrists (assuming there are 
that many centrists). One could use STV, but at that point, why not skip 
the middleman and use STV-MMP? One could also argue that parties are 
coherent units anyway (and that ordinary open list and closed list is 
even worse), but however you put it, it would reduce proportionality.

The system above could be implemented fairly easily: a voter picks a 
party's ballot paper, marks the candidates if he wants, and then puts 
the paper in an envelope (to conceal it) before dropping the envelope in 
the ballot box. This, by its nature, makes it impossible to rank the 
candidates of any other party than the one the voter's voting for.

(Also, the within-district ranked votes are summable if one uses a 
system like Condorcet.)

>> And of
>> course to keep the "irrelevant" candidates
>> out
> <snip>
>> Also money has been used somewhere.
> 
> In Ireland, you have to pay a deposit to be on the ballot, if you
> don't get at least 25% (I think) of a quota, you lose your deposit.

That seems reasonable. The lower the percentage, the more people are 
encouraged to run, so set it just at the level where having more 
candidates would clutter the ballot unduly.

>> In Finnish open
>> list elections I'm used to have some 150
>> candidates.
> 
> OTOH, you only have to pick one of them, rather than provide a ranking
> for say 10 of them.

The system above would let voters bullet vote, rank, truncate, whatnot 
as they see fit.

> This also applies with threshold based party list systems.  If I vote
> for a party which gets 4.9% of the vote in a system with a 5%
> threshold, then I am throwing my vote away.
> 
> This creates a disincentive to vote for parties that are close to the
> threshold.

No, it creates a disicentive to vote for parties that are below the 
threshold, not just close to it. It's like the spoiler effect, but 
lessened: "Do I vote for X, who is above the threshold, and thus my vote 
may count, or do I vote for Y, who is not, thus perhaps making X's 
opposition stronger?".

If you want to reduce this further, you could say that there's a 
disincentive to vote for X when the number of votes for X is just above 
the amount required to have an integer number of seats (including 0). 
When the party has a large surplus, your vote may give it another seat, 
but when it doesn't, the chances are less. However, if the barrier is 
sufficiently low, people vote anyway (as they do in party list PR).

> In the extreme case, you could have 19 parties with 4.9% of the vote
> and 1 party with 6.9% of the vote, and the 6.9% party gets all the
> seats in the legislature.

Yes, and the Wikipedia article about election thresholds mention this. 
It states that in Russia in 1995, the 5% threshold made 45% of the 
population unrepresented (meaning there were many small parties indeed).

> To fix this, at minimum, wasted votes should be reassigned.  I would
> also allow surplus transfers for independents (maybe including a rule
> that such transfers have to go to parties).

Another possible way would be to use a multiwinner method. People rank 
parties. The voter's ballot is altered so that each party is replaced by 
a clone set of size equal to the number of candidates that party fields. 
  After counting, one runs a multiwinner system with number of winners 
equal to the size of the assembly. Each party then gets as many 
candidates as there are members of its clone set in the outcome.

If you're using a divisor method, my "Webster-flavored multiwinner 
method" may be suitable (unless it is flawed; I am not certain if it is).

>> It is also possible to develop systems
>> that mix both styles. That could mean
>> e.g. default inheritance order (tree or
>> even candidate specific) for short
>> (exhausted) votes but allowing voters to
>> define their own order / deviate from
>> the default order if they so wish.
> 
> I agree with that.
> 
> In Australia, they have the 'above-the-line' system.  It is PR-STV,
> but you must either vote for all the candidates (and there can be
> 100+) or just pick a party's list.

Above the line, along with the requirement to rank everybody, changes it 
from true PR to party list. It would be better if they permitted truncation.

> There would be a national list PR election to fill half of the seats
> and then the other half would be returned from 5+ seater local
> constituencies.
> 
> I am not a big fan of MMP like methods that can be abused (as seen in
> Italy) or require that a vote for a candidate is assumed to be a vote
> for his party (the standard fix for the abuse).

Schulze's STV-MMP tries to fix this issue without linking candidate 
votes to party votes.