[EM] Condorcet corresponding to some variant of IRV?

[Alex Small]

Dave Ketchum said:
>        Where can you find any happiness by, supposedly, doing IRV, but
> sometimes picking a different winner than IRV would?
>        How can you dependably pick the CW without doing the Condorcet
> method? Why not look closer at Condorcet and start thinking of a
> way to make clear to voters that the CW is the best way to honor their
> desires?

I can think of only one reason to devise a method that's a compromise
between IRV and Condorcet, and it's a purely pragmatic reason:  If IRV
ever becomes wildly popular with the public, we may find that the only way
to pitch Condorcet is to make IRV-style modifications.

Let me be clear:  I am NOT saying that modifying Condorcet to resemble IRV
will give us an election method that is superior to other Condorcet
methods.  I'm saying that, in the face of public opinion, a day may come
when such a compromise is necessary.  However, we are NOWHERE NEAR that
day.

In a conversation with an IRV supporter on another mailing list, I found a
surprisingly simple accomodation.  I had thought that using IRV to resolve
cycles, but electing the Condorcet Winner when one exists, would be the
only compromise he'd contemplate.

To my surprise, he suggested dropping any candidate who is the first
choice of less than 1/(N+1) of the voters, and then using whatever
Condorcet method you like to select among the remaining candidates.  To
me, this is isomorphic to either a primary election or a "ballot access
requirement" (e.g. get a bunch of signatures to appear on the ballot). 
You have some process to determine who will participate in the contest,
and you use Condorcet to pick from among the contestants.

I am NOT saying that dropping people who fail to receive a certain
threshold of votes is the ideal way to do a Condorcet election.  I am
simply saying that, in my opinion, this does little violence to Condorcet
if you view it as a form of primary election.  Or, at the very least, it
does less violence than using IRV to resolve cycles.


For whatever it's worth, the IRV supporter's reason for liking the "drop
anybody below 1/(N+1)" method seemed to come down to fear of situations
like this:

34 B>D>A>C
33 A>D>B>C
33 C>D>A>B

D is the Condorcet winner, even though nobody puts him in first place. 
(OK, extreme example, but you get the spirity of his objection.)  Now, we
can all make our own arguments in favor of electing D anyway, but you have
to admit that this anti-Condorcet objection is a lot better than some of
what we hear.  Condorcet certainly isn't perfect, after all.


Anyway, just my $0.02 worth.




Alex