[EM] condorcet loser elimination PR

Alex Small asmall at physics.ucsb.edu
Mon Jul 21 16:44:34 PDT 2003

James Green-Armytage said:
> 	This is going back a little bit, but you suggested a form of Single
> Transferable Vote PR where instead of eliminating the candidate with the
> fewest top-choice votes, you eliminate the Condorcet loser, or some
> approximation of a Condorcet loser based on a Condorcet-completion
> principle.

Not what I proposed.  Here's what I propose:

1)  Elect the Condorcet winner, or the winner of the Condorcet completion
method (from now on I'll use "CW" to denote either the Condorcet winner or
the winner of the Condorcet completion method, just to be succinct).

2)  As in STV or PAV (Proportional Approval Voting) after electing
somebody we need to reduce the weight of some of the ballots.  In STV we
would reduce the weight of all ballots that ranked the candidate first
(first among those remaining, that is).  In Proportional Condorcet (PC) we
would reduce the weight of any ballots that prefer the most recently
elected candidate to the runner-up.

Who is the runner-up?  Whoever the CW would have been had the most
recently elected candidate (i.e. the most recent CW) not been on the
ballot.  This is NOT the Condorcet Loser, at least not as I understand the
term.  I always think of the Condorcet Loser as whoever loses ALL pairwise
contests (if such a candidate exists).  I guess you could say that any
candidate not elected is a Condorcet Loser, but then the Condorcet Loser
is just a loser, and there's no need for a special term.

The proper weight assignment is a tricky matter.  I'll have to think about
that.  But a note on strategy:

(Insert caveats that party affiliation may not be best way to judge
proportionality, but I use party labels for convenience in my example)

Say my preference is (any Libertarian)>(any Democrat)>(any Republican). 
Of course I have a preference within each party, but that's another

If I know from polling data that the first few candidates elected will be
Democrats, and I don't want the weight of my ballot reduced, my best bet
is to rank all or most of the Democrats equal to one another, e.g.

maverick=spineless=stupid=really stupid=socialist=Lieberman

(my sincere preference would be to substitute a > sign in place of each =

That way I'll have maximum impact to help elect Libertarians.  I might
even have an incentive to rank some of the Libertarians equal to one
another, e.g.

techie=drug legalizer=free trader> Rand disciple> druid> ferret legalizer

(my sincere preference would be to substitute a > sign for each = sign,
and I might note that druid and ferret legalizer ran for Gov. and Lt. Gov.
of California in 2002)

However, if I believe that the LAST few candidates elected will be
Republicans, I have no disincentive to rank the Republicans, e.g.

quasi-libertarian> moderate> bad actor> paranoid> militarist> theocrat

OK, the fictitious candidates were a bit of editorializing, but it was fun
;)  The point remains the same:  This Condorcet PR method gives strong
incentives for equal rankings.  This isn't so bad; PR encourages large
fields of candidates, and voters might not be able to study and form an
opinion of every candidate anyway.


More information about the Election-Methods mailing list