[Elections-committee] STV vs. Condorcet

Simon Anderson simon at dreamhost.com
Tue Oct 15 23:50:47 UTC 2013


This is a very insightful perspective Thierry. Thanks for sharing it!


Best,
Simon Anderson
CEO, DreamHost


On Tue, Oct 15, 2013 at 1:10 PM, Thierry Carrez <thierry at openstack.org>wrote:

> Hi election committee,
>
> I have a few insights on those various election systems due to past
> experience and Monty suggested that I shared them with the group.
>
> Ranked voting systems are intrinsically superior to first-past-the-post
> or cumulative systems because they let you express a complex preference,
> and adding more candidates doesn't disrupt your ability to express that
> preference. In ranked voting systems, more choice is actually always
> better.
>
> That said, all ranked voting systems are not created equal. You should
> pick one based on the intended results.
>
> STV is designed for proportional representation of factions. It favors
> candidates coming from those factions at the expense of consensus
> candidates.
>
> Condorcet favors consensus candidates and "natural" winners (the
> Condorcet winner), at the expense of giving each faction its representant.
>
> In a very simplified and inaccurate summary, given B>A>C and C>A>B
> ballots, STV tends to pick B&C while Condorcet tends to pick A.
>
> There is not a "bad" system and a "good" system, choosing between them
> depends on the intended results. If the election is all about factions
> and giving them fair representation, then STV is better. If the election
> is about individuals and picking the most consensual candidates, then
> Condorcet is better.
>
> I tend to think that for this specific election (directors representing
> the individual members, not any faction), Condorcet would yield better
> results.
>
> NB: there is an experimental "proportional Condorcet" algorithm at CIVS
> that might make the best of both worlds, but it hasn't been as
> battle-tested as STV and Condorcet. We will be running the TC ballot
> through it to see how it affects the results and report back. I don't
> think such an experimental algorithm should be under consideration for
> the Foundation board.
>
> --
> Thierry Carrez (ttx)
>
> _______________________________________________
> Elections-committee mailing list
> Elections-committee at lists.openstack.org
> http://lists.openstack.org/cgi-bin/mailman/listinfo/elections-committee
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/elections-committee/attachments/20131015/49f322ac/attachment.html>


More information about the Elections-committee mailing list