James Gilmour wrote:
Kristofer Munsterhjelm
Sent: Wednesday, December 24, 2008 9:54 AM
Perhaps real world implementation of Condorcet
systems would have a "first preference" threshold, either on candidates
or on sets: anyone getting less than x% FP is disqualified.
I have not seen any advocate of Condorcet make such a suggestion, but
it has been made for IRV, though not taken up by any serious
IRV advocates, so far as I am aware.
The weak Condorcet winner is, in my view, the political Achilles'
heel of the Condorcet voting system. The corresponding political
defect in IRV is that it can eliminate a Condorcet winner (whether
that is common or not is irrelevant - it is possible). But we
know from experience that real electors and real politicians will
accept that political defect in IRV - evidence: IRV has been
used for public elections for many decades in several countries. In
contrast, despite having been around for about 220 years, the
Condorcet voting system has not been used in any public elections
anywhere, so far as I am aware. That could perhaps change if a
threshold were implemented to exclude the possibility of a weak
Condorcet winner AND if a SIMPLE method were agreed to break
Condorcet cycles.
Technically, Condorcet methods have been used in public elections.
Nanson's method (below-average Borda-elimination) was used in a town in
Michigan. That's one place against IRV's hundreds, though, so I see your
point.
A less "arbitrary" or "hacked upon" manner of fixing your problem might
be to have two elections. The second is between two "winners": the
winner of a Condorcet election, and the winner of a Condorcet election
with a quite high threshold (or the IRV winner, or FPP winner - probably
should be a summable system).
If there's a CW and it's the sincere CW, the second round is pointless.
Otherwise, if people really prefer someone with a certain amount of
first preference votes, not all is lost.
That might be too complex, though, and one of the points of Condorcet is
to not need to have multiple rounds.
As for a simple method, I think Ranked Pairs (or MAM, rather) is quite
simple. Juho thinks Minmax would work, I'm a bit too picky about
criteria; but if it does, that is about as simple as you get.
Schulze is complex but has "precedence" (history) in organizations:
mainly technical/computer-related organizations, but also Wikimedia and MTV.
----
Election-Methods mailing list - see http://electorama.com/em for list info