[EM] Incorporating slates into STV (was: PRfavoringracialminorities)

[Raph Frank]

On Tue, Aug 19, 2008 at 12:16 AM, Aaron Armitage
<eutychus_slept at yahoo.com> wrote:
> I'm also stuck with plurality, and I agree entirely that plain-vanilla STV
> s a vast improvement

Right.  Whatever PR-STV's flaws, they are minor compared
to plurality.

> If we wants slates that run as teams and are voted on, and also the
> ability to single out a single candidate on a slate, the easiest way I can
> think of would be to abandon the transferable vote altogether and use a
> different mechanism. We would generate every possible assembly and compare
> them one-on-one, Condorcet style. The comparison would ignore those
> candidates who receive seats in both outcomes, and compare only the seats
> filled by different people. We give each seat a "weight" of 1, and both
> outcomes divide the weight of each seat evenly among the voters who prefer
> that outcome's winner over the alternative. The potential assembly that
> spreads its "weight" most evenly wins the comparison. It wouldn't be that
> simple, of course: if there are fifty differences between two outcomes,
> which ones are we actually comparing? But I just thought of this today.

I think you are aiming for CPO-STV.  This is a condorcet based
STV method and IIRC it meets Droop proportionality.

The way it works is that it has a metric for comparing any
2 possible outcomes to say which outcome wins and by how
much.  This can then be plugged into any standard condorcet
method.

Only candidates who hit the
quota *and* are in both outcomes have their surplus transferred.
This may cause some other candidate to hit the quota and
have their surplus transferred too.

The theory is that there is no point in transferring
surplus from a candidate who is in only one outcome
as that voter would want their full weight voting for that
outcome and it doesn't matter which candidate in the
outcome has the vote.

No candidates are eliminated and the score
for each outcome is equal to the sum of the votes
held by each candidate in that outcome.

E.g. if there was 2 outcomes being considered
A,B,C and C,D,E and after the first count the results were

Quota = 100
A: 50
B: 75
C: 125 (in both outcomes)
D: 25
E: 25
F: ignored
G: ignored

C's 25 surplus would be transferred giving, say

Quota = 100
A: 50
B: 75
C: 100
D: 50
E: 25
F: ignored
G: ignored

A,B,C would score: 225
C,D,E would score: 175

Thus A,B,C would beat C,D,E.

The outcome which is the condorcet winner of all possible
outcomes is then declared the winner.

The big issue with this method is that it is very
computationally complex as for even small constituencies
(not to mind an entire assembly), there can be a large
number of outcomes and each one much be compared
pairwise to all others.

>
> By ranking a slate, you choose any member of that slate over anyone ranked
> below it, thus making it more even spread and therefore more likely to
> win, but you can still vote for any particular candidate. Unlike STV with
> slates, there would probably be no independent candidates, because for the
> candidate there is no strategic downside to running on a slate. In fact,
> there's no reason there couldn't be overlapping slates.
>

As long as they are aids to the voter and don't restrict them, then
they are a positive thing.  Ofc, that assumes that voters aren't de
facto forced to use them like in Australia with the above the line
system.

> In an American context, I don't picture STV with slates as being a semi-
> party list system. They would work better if the slates are simply
> candidates who choose to campaign as a team, and who are subject to ballot
> access rules similar (or identical) to what independent candidates.
> Probably they would state party affiliations in some way, though. With the
> second system, the political parties probably would be slates you could
> vote for, encompassing or overlapping with other, smaller slates and
> individual candidates' campaigns.

It really depends on district size.  If the districts are less than 10
seats each,
I think slates are just making things overly complex.

The candidates could just send out literature asking their supporters
to vote give the other candidates their 2nd and 3rd choice etc.

There would likely be less than 20 candidates running.