[EM] STV - the transferrable part is OK (fair), the sequential round elimination is not

[Juho]

On Nov 2, 2009, at 11:30 PM, Raph Frank wrote:

> On Mon, Nov 2, 2009 at 8:38 PM, Juho <juho4880 at yahoo.co.uk> wrote:
>> On Nov 2, 2009, at 1:53 PM, Raph Frank wrote:
>>> Districts with 7+ seats seem reasonable, and give reasonable
>>> proportionality.
>>
>> I guess there is some practical limit to how may candidates the  
>> voters are
>> willing to evaluate and rank. Districts of 7+ already offer  
>> reasonable
>> proportionality (approximate quite well the x% of votes => x% of  
>> seats
>> principle). Also the number of candidates should be small enough in  
>> this
>> case so that the voters need not rank too many candidates (e.g. 10
>> candidates from each party).
>
> Well, even with a larger number of seats, a voter would waste very
> little of their vote, even if they only voted for 2-3 candidates.
>
> Assuming you only ranked 3 candidates and they all get elected with
> double the quota, only 12.5% of your vote would be exhausted.

Number of candidates is maybe the actual problem and the number of  
seats mainly influences the number of candidates. If the number of  
candidates is large then it may be necessary to rank numerous  
candidates to be sure that at least one of them will be elected (and  
the vote is not wasted).

There is also the risk that voters will vote for the strongest  
candidates and not their (possibly weaker) favourites because of this  
problem.

I'm thinking e.g. the Finnish elections where currently there can be  
some 150+ candidates to rank. It might be necessary to rank quite many  
candidates if I don't want to support the incumbents.

>
> In practice it is rare that candidates get much more than 10% above
> the quota (except the candidates who are elected on the first count).
>
> A reasonable rule would be to keep ranking until you hit a candidate
> who has a reasonable chance of being elected, but isn't so popular
> that he will gain much more than a quota in the first round.

Yes. But maybe I should rank until I'm quite sure that at least one of  
the ranked candidates will be elected. In elections where there are  
numerous candidates (ref. Finland) it is also important to rank those  
good candidates that may not be elected this time but whom I want to  
promote so that they will be elected in the next elections (i.e. the  
next time potential winners will be picked by the voters in these  
elections and not by the party officials (that may offer just a  
limited set) just before the next elections).

>
>> Also the number of districts has an impact here. If there are e.g. 10
>> districts of size 7 there could be a party with 10% support and no  
>> seats
>> although from a nation wide perspective 10% of the votes would  
>> justify 7
>> seats.
>
> True, however that assumes that the party has very constituent  
> support.

Yes, districts tend to favour local groupings over evenly spread ones.

>
> If it varies a little from region to region, then maybe they would win
> a few seats at least.
>
> The could also decide to focus their resources from the whole country
> on the 7 regions that they are most likely to win a seat in.  (though
> that might get a backlash due to using "outsiders").
>
>> Yes, districts with independent elections set similar limitations  
>> in all
>> systems. In list based systems it is just somewhat easier to extend  
>> them
>> e.g. so that proportionality will be counted at country level.  
>> Candidate
>> lists could still be regional if one so wants (the summed up votes  
>> would
>> determine proportions at the country level, and seats could then be
>> propagated back down (as in the Finnish proposal)).
>
> You could also pretend that there is just 1 national constituency and
> voters just happened to only vote for local candidates.

Yes. Usually the number of seats in each district is based on  
population. It would be an interesting trial if the number of seats in  
each district would be based on the number of valid votes in that  
district. That might improve the turnout :-).

>
> Also, you could list local candidates on the ballot, but give a
> write-in slot.  The write in could allow voters to vote for a
> candidate from other regions.

One could also allow anyone to vote any candidate from any region but  
still allocate a fixed number of seats to each district. The voter  
could then vote for her favourite (and thereby guarantee that she will  
be elected) even if that favourite would be from another district.

Juho


>
> This reduces the complexity of the ballot for locals, but also allows
> voters to vote for a write in candidate if they wish.
>
>> Yes, this is one way to extend STV to offer better proportionality  
>> at the
>> country level. This method seems to combine some list type features  
>> with STV
>> voting.
>>
>> (Btw, did you consider the possibility of parties running their  
>> most popular
>> candidates (that will be elected in any case) outside the party  
>> list. Is
>> that a valid strategy in this method?)
>
> It depends on what you mean here.
>
> It doesn't suffer from the same problem as MMP, where you can gain
> extra votes by using a decoy list.  Only votes which would otherwise
> be exhausted are transferred to the national level.
>
> A voter who votes for an independent doesn't also get to cast a party
> vote, so you can't have your supporters support a fake independent
> locally while still voting for the party with their party vote.
>
> However, the method would still have the standard issues with vote
> management.  This is pretty much inherent to PR-STV.  If party
> supporters vote for the weaker party candidates instead of a very
> popular candidate, then when the popular candidate is elected, fewer
> of the party supporters' votes are used up.