Re: Results for Project membership procedures
Lucas Nussbaum <lucas@lucas-nussbaum.net> writes:
> On 15/12/08 at 15:28 -0800, Russ Allbery wrote:
>> I suspect this is because the obvious "please, dear deity, stop talking
>> about things constantly and just do them" vote ranks 3 above 2 above 1,
>> so I doubt many votes transferred from 3 to 1 when 3 was eliminated.
> There's no such transfer in plurality voting (where you only vote for
> one option). That transfer happens in instant-runoff voting, but I
> didn't compare our condorcet results with IRV results. That's not easy
> to do accurately because you can't rank several options at the same
> level with IRV, so you would have to ignore a lot of ballots.
I suspect I confused things by not being sufficiently clear. I meant that
the difference between the Condorcet outcome and the plurality outcome was
probably due to the 3>2>1 voting pattern leading to a lot of transfers
from 3 to 2 in Condorcet, so although 1 wins plurality, 2 ends up with
more preference votes once 3 is eliminated.
(And by choice 2, I mean the one listed as choice 2 on the voting page,
which is shown first before choice 1 and is also known as amendment A.)
It's the standard case for Condorcet producing a different outcome than
plurality: the compromise is less popular by itself, but is much more
popular with a minority of the voters who would prefer some other choice.
It happens quite a bit in Hugo voting (IRV, if I recall correctly), where
the leader in first place votes often doesn't win.
--
Russ Allbery (rra@debian.org) <http://www.eyrie.org/~eagle/>
Reply to: