Re: Election results

On Tue, Aug 07, 2007 at 02:51:24PM +0100, MJ Ray wrote:
> Antti-Juhani Kaijanaho <antti-juhani(at)kaijanaho(dot)fi> wrote:
> > As I'm sure you know - as it was discussed during the soc ctte debates
> > at Debian - the election method used here fails proportionality.
>
> Why be sure? Assumption is the mother of all ... To be clear, I
> didn't know Condorcet-SPI was already discussed

I apologise for the assumption :)

As far as I can tell, the SPI method and the Debian method are similar
enough that the issue is common to both. Of course, I can't find
anywhere a precise definition of the SPI method for *multi-winner*
elections, but the result page makes me believe it's what I'd expect.

I'll revisit if you'll point me to the definition of the multi-winner
SPI method :)

> Where's that discussion, please?

The messages you mentioned were the ones I was referring to.

> Anyway, should/how could we try to fix this proportionality failure?

My current favourite is the Schulze STV method that generalizes the
classic Schulze ("cloneproof Schwarz sequential dropping") method to
multiple-winner elections. It is claimed to satisfy proportionality,
though I haven't bothered checking the argument myself.

--
Antti-Juhani Kaijanaho, Jyväskylä
http://antti-juhani.kaijanaho.fi/newblog/
http://www.flickr.com/photos/antti-juhani/