[EM] Raph: Sainte-Lague. Transfers in party-list PR.

Michael Ossipoff email9648742 at gmail.com
Wed Jul 18 05:58:46 PDT 2012

>> And the parties needn't publish whole rankings. Each party need only
>> publish one other party to which it wants to transfer to if it
>> transfers.
> I think full rankings would be better.  If excess is being
> transferred, then even large parties will transfer some of their vote.

Yes, but if a large party suplus-transfers to another party,and it,
too,as a result, acquires a quota and must transfer,then the
destination for that next transfer can just be that 2nd party's
transfer-choice. All of the transfers, surplus and elimination, could
be accommodated by each party publishing a single transfer-choice.

But either would do. Published by each party, it could be a single
transfer-choice, or a complete ranking.
> If the rule is that any party which gets less than 2 seats is
> eliminated...

I'd only eliminate parties that failed to get a seat at all. With 1.4
or 2 as the 1st Sainte-Lague denominator.

>> But,again, just as with the transfers between parties, it seems to me
>> that each candidate need only designate one other candidate to whom
>> s/he wants to transfer, if s/he transfers.
>> ..instead of publishing whole rankings.
> Well, the idea is that it would be a national level PR-STV election,
> but with a compromise to keep the ballot counting manageable.

Sure, and the ballot-rankings could be replaced, as you suggested, by
rankings published by candidates. That would be the closest approach
to STV. And maybe that would be preferable.I'm just saying that it
would still work if each candidate published only a single
transfer-choice. I'm not saying that would be better--only that it
would be one possible way of doing it.

>> There are many quite workable solutions. Parties (or other
>> organizations) could agree among their members, that people with names
>> beginning with certain letters would vote for certain candidates. Or
>> that voters in certain geographical regions would vote for certain
>> candidates. Or, as you suggested, a party or organisation could
>> directly advise each voter which candidate to vote for, in the form of
>> direct communications with voters, or via a published list that pairs
>> voters with candidates.
> The issue I have with this is that it means that voters are less free
> to make intra-party choices, but that is inherent in the system.

Yes, it sacrifices some expression of voter-choice, in return for a
simpler count. But the candidate-ranking STV, or
candidate-single-transfer-instruction STV, would still only require
one _big_ count of all the ballots, and then the remaining counting
would be the kind that anyone could do at home. So the need for SNTV's
simplification isn't great. So of course I'd prefer the
candidate's-choice versions of STV, the ones without ballot rankings,
in which voters merely vote for a candidate.  ...because, with those,
the voter hirself chooses to which candidate hir vote will go to

Though SNTV eliminates the need for any further counting, after the
initial count of ballots, the work that SNTV eliminates is actually
very little work.

> This happens somewhat with "vote management" under PR-STV.  A
> candidate who is popular in his own right is discouraged from seeking
> first choices from party supporters.  This means that he gets elected
> on "personal" vote and doesn't waste party supporters.

I suppose that could save some other members of his party from
1st-round elimination.
>> But the voter could also be allowed to vote for several candidates of
>> a party, and the candidates could be seated in order of their numbers
>> of approvals.
> There are lots of possibilities.  Approval isn't very proportional.

No, but I was just suggesting, in an open-list election, Approval for
determining the order in which a party's list-candidates would take
the seats won by the party in the Sainte-Lague allocation. But I was
just mentioning it as a possibility. I'm sure that the way currently
used by the countries that allow several votes for each voter in their
open list elections is perfectly good as-is. It has been working well
for a long time.

> I think Asset voting might be interesting.

I haven't seen a definition of it,.

>> Isn't that how it already is, in those countries that let voters vote
>> for a _set_ of candidates on a party list (or even several party
>> lists) if they want to?
> I am not sure, I thought open list was mostly vote for 1.

Sometimes. But some countries give each voter a specific number of
votes, greater than one, which s/he can divide among several
candidates of one party, or (at least in some of those countries)
among candidates of several parties. Each party gets a party vote for
each vote received by one of its candidates. It's an interesting
system, with a lot of freedom for a voter. The voter isn't limited to
voting for one party's candidates, but can "mix and match". Some of
the PR systems in use are very interesting, beautifully creative,and
allow great voter-freedom.

 >I think the
> Netherlands has a system where it is party list, but if you get more
> than 25% of a quota, you skip the list (I guess they have a tie break
> rule if to many get more than 25%).

What I like about what I read about the Netherlands system was that
it's purely at-large PR. Not by district.

>> There's been inexplicably little innovation in single-winner methods.
>> A few impressively progressive countries have been using a different
>> single-winner method (IRV, also called Preferential Voting or the
>> Alternative Vote). They deserve credit for progressiveness, and iRV
>> would even be ok, if people would avoid over-compromise,and would be
>> very particular what they regard as acceptable.
> It seems from Australia that IRV leads to 2 party domination.

That confirms my insistence that FBC is essential, given the way
people are inclined to strategize.

> However, I am not sure if there are some subtle underlying effects of
> IRV, maybe 3rd parties have more influence, even if they don't win.

IRV has its good points, but I feel that FBC is necessary in
single-winner elections.

> I was thinking that an interesting tweak for IRV would be to modify
> the rule for eliminations.
> It assumes equal ranks are allowed.  Each candidate would have 2
> totals.  For ballots where there are 2 or more candidates at the top
> rank, there would be different rule for the candidates election total
> and elimination total.
> For election, the vote would be split equally between top ranked
> candidates and for elimination it would operate at full strength for
> all candidates.

It wouldn't pass FBC.

In fact, someone on EM demonstrated that, even if all of the
equal-ranked candidates received a full vote, as in Approval, the
resulting IRV would still fail FBC. I had proposed a full-votes
equal-rankings IRV, but it was shown to violate FBC.
> In the PR-STV context, it means that intra-party elections are
> determined by approval voting, rather than IRV.

In STV, you wouldn't want equally ranked candidates getting whole
votes, because it would violate proportionality.In STV there probably
isn't any need for ranking several candidates equally.

Mike Ossipoff

More information about the Election-Methods mailing list