[EM] Condorcet 2 - The Sequel ( the same people say the same things)

[Alex Small]

Forest Simmons said:
> Alex said he was willing to agree that IRV might have some advantage
> over Condorcet in the one dimensional case, because the "centrist"
> candidate might be positioned there opportunistically.

It's not about opportunism per se.  It's about the possibility that the
centrist might grow fat and inefficient like any other monopoly.  However,
I also believe that Condorcet is more likely to break us out of the 1D
mold and enable true competition.

Or, at the very least, from what I know of IRV's track record there's very
little evidence of IRV breaking out of the 1D mold.

> Note that an order preserving transformation like x->x^3 could easily
> change which candidate is closest to the mean or midrange, but it will
> never change the Condorcet winner if the Condorcet winner is positioned
> precisely with the median voter.
>
> This means that our analysis doesn't depend on what units the one
> dimensional issue spectrum is measured with, but only on the relative
> order.  Any distortion of the issue space by astute politicians would
> have to alter the order of the points before it could fool Condorcet.

If one were to try and model competition in a 2D issue space, there are
two plausible approaches:

1)  A relative scale from zero to 1, where the position on the scale
denotes the cumulative percentage of voters to the left of that position. 
Left refers to some specified issue, not necessarily the conventional
"left" or "right" in US politics.  If our issue were foreign policy,
somebody positioned at zero might be more pacificist than 100% of the
population.  A candidate positioned at 0.25 would be more pacificist than
75% of the population, etc. until we reach 1, where a candidate is less
pacifist than any voter out there.

This has the nice property of ignoring any possible distortions of issue
space.  It focuses on all that really matters:  Voters.

2)  An issue-based index might still be desirable.  The above approach has
the nice aspect of telling where a candidate stands relative to the
electorate on a particular issue, but it fails to give any info on how
voters might rank candidates.  If a candidate is more pacifist than me, or
more socially authoritarian than me, he might still be my favorite.  It
depends on where the other candidates stand.  And I prefer A on one issue,
and B on another, how do I determine whether or not A is better than B?

So the abstraction of "absolute" stances on issues might be nice.  Yes,
the notion of "center" is subjective, but at least for theoretical
purposes it might be useful for figuring out where candidates should
position themselves when there are tradeoffs among issues.

For instance, we might give all of the candidates and voters a
questionaire on economics, and another quiz on social issues.  Assign each
candidate and voter a score from those questionaires.  A candidate's
distance from a given voter could be defined as

|candidate economic score - voter economic score| + |candidate social
score-voter social score|  (where |x| = absolute value of x).

This might be a useful tool for modeling how people make trade-offs.



Alex