[EM] Issues, Condorcet, and IRV (was: IRV vs. plurality)

[Alex Small]

Somebody said:
>>At the risk of repeating myself, my argument against Condorcet can be
>> summarised as follows:
>>
>>I do not support Condorcet because I believe that in practice,
>> regardless  of the theoretical and conceptual advantages it may
>> possess, it would be  too favourable towards parties who succeed in
>> positioning themselves in  the centre and correspondingly
>> discriminatory against wing parties.

Suppose that we have 3 candidates, and voters and candidates mostly fall
along a 1D left-right spectrum.  Let's call the candidates Left, Right,
and Middle.  Let's also assume (I'll justify this momentarily) that the 3
candidates will always maintain their relative positions.  Assume (for the
purposes of this model) that if somebody suddenly went to the other side,
he'd lose credibility.  People would compare his track record with his new
stances and he'd be regarded as a liar.

Given the assumption of no role-reversal, Left and Right can guarantee
that Middle won't win an IRV race moving toward the middle.  They don't
have to land right on top of Middle, they just have to get close enough to
him so they can steal first-place votes.  If I'm slightly left of center,
and Left is also slightly left of center, even if he's to the left of me I
might prefer him to Middle as long as he isn't TOO far from Middle.

This analysis suggests that, as long as most voters/issues/candidates fall
along a 1D spectrum, IRV will maintain a 2-party duopoly with a weak
center trying but failing to win elections.  The "weak center" will have
to take a very different position and make the issue space 2D.  Maybe go
to the left on social issues and to the right on economic issues.  Or
maybe turn "moderation" into its own issue:  "We're the reasonable people
who try to do what's right, the others are just stuck in an ideological
vendetta."  In any case, going 2D is the only hope for the center if we
use IRV.  Only then can the "third party" hope to pick up more first-place
votes.

By contrast, with Condorcet the middle will ALWAYS win if the issues are
1D, unless either the left or right happens to have an outright first
place majority.  There will always be 4 types of voters in a 3-way race:

Middle>Left>Right
Middle>Right>Left
Left>Middle>Right
Right>Middle>Left

Unless Left or Right has an outright majority, the 2 Middle factions can
always join with one of the "wings" to defeat the other "wing" pairwise. 
Left and Right desperately need to go 2D so they can persuade some people
to vote Left>Right>Middle or Right>Left>Middle (at that point the labels
become less meaningful, but I think everybody knows what I mean).

So Condorcet gives 2 of the 3 parties an incentive to "go 2D" in this
example, while IRV only gives 1 party an incentive to "go 2D" in this
model.

What interests me is that, based on what I hear of politics in other
countries (which is admittedly not a lot), it seems like most countries
have some sort of left-right division.  The factions may not have the same
respective stances as the American left and right, but that sort of
division still seems to stand.  Even PR doesn't seem to do much to cure
this problem.  Of course, I am quite willing (and EAGER!) to be corrected
on this, to be told that my gloomy assessment is false.

If in fact humans get easily trapped in 1D political molds, then maybe IRV
is almost more desirable than Condorcet (commence flaming now).  As long
as we're 1D, there will be 2 competitive parties in IRV but Condorcet
might yield single-party monopoly.

So, in summary:

If it's possible for cultures to break out of the 1D political stalemate
then I'm enthusiastically pro-Condorcet.  If it isn't possible for the
most part, if fiscally conservative/socially liberal mold-breaking is
impossible, then I may actually be more sympathetic to IRV than I was
previously.  (Commence flaming now.)



Alex