[EM] IRV's "majority winner". What if we let the people choose?

[James Green-Armytage]

Dear James Gilmour, 

	Let's try to follow through with one of these examples until the end.
Let's say that in a presidential election, the ballots cast are

48: Bush > McCain > Gore
3: McCain > Bush > Gore
49: Gore > McCain > Bush

	The Condorcet winner is McCain, and the IRV winner is Bush.
	Now, let's imagine that there is a runoff election between McCain and
Bush. How do you think that people will vote?
	We can probably agree that the Bush > McCain > Gore voters will vote for
Bush. And we can probably agree that the McCain > Bush > Gore voters will
vote for McCain.
	It seems that your issue is with the Gore > McCain > Bush voters. You
seem to be saying that some of them will vote for Bush rather than McCain.
And you're saying that they will do so because, although at first they
thought they liked McCain better, they have somehow decided that he
doesn't deserve to win, because he has so few first choice votes. I'm I
understanding your point correctly?
	At this point, I'm not quite convinced. It seems common-sensical to me
that the Gore > McCain > Bush people voted that way because they really
don't like Bush, and that if they had to get a Republican they'd rather
have the more moderate and respectable McCain. I don't really see a lot of
democrats voting for Bush based on some that he deserves to win because
other people voted for him.
	If anything, I'd expect the opposite, that more democrats would get
behind McCain when it became clear that he was the only alternative to
Bush. Now I know that in the above example it doesn't apply, since McCain
already has 100% second choice support (which I already said isn't so very
realistic), but anyway you get the idea. It just makes sense to me that
the democrats would vote in McCain over Bush, if they knew and accepted
that those were the only two choices available.

********semi-digression*************
	The argument that might give me more pause would be the argument that
some of the initial votes were insincere. For example, let's say that the
sincere preference rankings were something like 
38: Bush > McCain > Gore
10: Bush > Gore > McCain
39: Gore > McCain > Bush
10: Gore > Bush > McCain
3: McCain > Bush > Gore
, but nobody knew before the election who would win the Bush-Gore pairwise
comparison, and so both the Bush > Gore > McCain people and the Gore >
Bush > McCain people engaged in the burial:reversal strategy in order to
gain a possible edge for their first choice candidate.
	I'm not sure how likely that all is, but it constitutes at least some
ground to doubt McCain's legitimacy in the initial example. However, I
still tend to think that a *sincere* Condorcet winner (who wins all
pairwise comparisons) should always be expected to win a runoff election
against any other single candidate.
********end of semi-digression*********

	With all that said, I think that it would be interesting if we could get
some empirical evidence for this, I mean actually have a ranked ballot
followed up by a runoff between the IRV and Condorcet (or completed)
winners. Actually, I'm not even sure that that would be such a bad
election method overall. Thanks, Mike.

my best,
James Green-Armytage