[EM] IRV vs Condorcet vs Range/Score

[Raph Frank]

On Wed, Oct 15, 2008 at 5:49 PM, Kristofer Munsterhjelm
<km-elmet at broadpark.no> wrote:
> Since Condorcet is a majoritarian method (as it needs to be in order to be a
> good single-winner method), the resulting parliament would also be
> majoritarian. This means that a party whose candidates consistently got low
> support in all districts would find none of those elected.

Condorcet should elect lots of centerists.  Plurality elects a group
of off-center candidates (centerists in their own party).

With opinions along a single axis from 0 to 100, condorcet should
elect candidates clustered around 50, but plurality will elect
candidates clustered around 33 and 67 (or maybe 40 and 60), but few
centerists.

PR elects a legislature spread reasonably evenly (though extremists,
i.e 0-5 and 95-100 might be under represented due to thresholds and
such).

A PR legislature has the advantage that it will react to unexpected
occurances in a similar way to the general population.

> But Condorcet is also a better single-winner method than Plurality, so the
> candidates would be better representatives of the majority. They would
> probably be good compromise candidates, so the parliament would be less
> party-based than one elected by PR methods.

It really depends on the effects of parties and condorcet.  Would a
large number of voters effectively bullet vote.

> This might not be all that good for traditional parliaments; I don't know if
> the majoritarian nature would lessen competition (except in dominant-party
> states, where using Condorcet would be an improvement on Plurality in that
> respect), but it would reduce the voice of the minority.

Condorcet allows a faction to replace its candidate without losing power.

Under plurality, you must vote for your party's candidate or your
party will lose power due to spoiler.  You can't replace the candidate
but keep the party.

You would be able to do that under condorcet.

> It might be a good choice for an upper house decided to be a moderating
> counterweight to a populist lower house.

I think that is reasonable.  An upper house selected by a good single
seat system and a lower house selected by PR.

The viewpoint of the upper house would tend to move alot less.  It
would operate like a median, so be less effected by noise.

Ofc, it depends on how the voters vote.  If they vote by party, then
it may still be top-2 based where each of the top 2 comes from one of
the main parties.

> It's possible that IRV is a bad single-winner method precisely because STV
> is a good PR method (with many seats). In STV, one doesn't need to find a
> candidate that covers a large area of opinion space, since other candidates
> can be used to cover those areas, but in IRV there are no other positions to
> be used in such a manner. I'm not sure if this is the reason, but it would
> fit well with my simulation results that "majoritarian IRV" (eliminate until
> only k remains for council size k) is a surprisingly good PR method, at
> least in absence of strategy.

I think that may be because without strategy nobody is systematically
discriminated agains.