[EM] Is Condorcet The Turkey?

[Dave Ketchum]

On Sun, 8 Jun 2003 14:44:19 EDT Dgamble997 at aol.com wrote:

 > Dave you wrote
 >
 > First note: Condorcet considers ALL preferences in pairs simultaneously,
 > though often in two steps:
 > If one candidate is preferred over each and every other candidate,
 > that is the winner.
 >
 > This is exactly what I meant. Condorcet simultaneously compares all
 > pairs of candidates. The voter, has for example, ranked the candidates A
 > 1st, B 2nd, C 3rd, D 4th, etc. In simultaneous paired comparisons the
 > 4th Preference D is rated as having equal value to the 1st preference A.
 > In pairwise comparisons between A and another candidate E this A>B>C>D
 > vote counts one for A, likewise in a similar DE pairwise comparison it
 > will count one for D. Since the voter has ranked A 1st and D 4th s/he
 > clearly supports A more than D.

Agreed that the voter likes each of A, B, and C better than D - and
Condorcet SEES all of this.
 >
 > IRV has the advantage that lower preferences are not considered until
 > higher preferences have been eliminated. This removes the problem that A
 > is preferred to D and yet in a Condorcet count both are counted as one
 > against candidate E. If A, B and C are eliminated we can assume (safely)
 > the voter will wholehearted support D as his/her highest choice
 > remaining in the count.

We have NO assurance that the voter will ever give "wholehearted support"
to D, after saying that A, B, and C are each better.

Suppose that EVERY voter ranked B second, with about 25% ranking each of
A, C, D, and E first:
       IRV will pick one of the 4, though 75% of the voters will have
ranked the winner as third or lower.
       Condorcet will pick B, seeing that while B is no one's first choice
and each of the others has minority support for top rank, none of the
others has wide support.
 >
 > You also wrote
 >
 > Most of the time IRV and Condorcet produce identical results. We only
 > debate around the edges......
 >
 > You statement is probably true regarding the American party system -
 > both IRV and Condorcet will probably give a congress composed entirely
 > or almost entirely of Democrats and Republicans. However for the British
 > party system (the one I'm most familiar with) this is not the case.
 > Since the 1930's the UK has had three parties. A large party on the
 > right ( Conservative), a large party on the left (Labour) and a smaller
 > party ( variously the Liberals, Alliance, Liberal Democrats ) between
 > the two. I attempted an analysis of the results of the 1997 General
 > Election for the 17 seats in the county of Kent under FPTP, IRV and
 > Condorcet.
 >
 > The following assumptions apply:
 >
 > All Conservative voters vote C>LD>LA
 >
 > All Labour voters vote LA>LD>C
 >
 > Liberal Democrat voters vote 50:50 LD>C>LA : LD>LA>C
 >
 > The votes of others have no effect on the election.
 >
 > The votes obtained by the parties were as follows:
 >
 > Conservative (C) 40.5% Labour (LA) 37.2% Liberal Democrat (LD) 17%
 > Others 5.4 %
 >
 > Under FPTP the results were C 9, LA 8, LD 0
 >
 > Under IRV  the results were C 7, LA 8, LD 2
 >
 > Under Condorcet the results were C 1, LA 3, LD 13
 >
 > There is a grossly disproportional result under Condorcet The Liberal
 > Democrats obtain 76.5 % of the seats with 17.0 % of the votes.
 >
 > Were the LD vote to increase by 20% an equally bad result would of
 > course be given by IRV.
 >
 > The moral of this story of course is that there is little proportional
 > about the allocation of a single seat and even less proportional about
 > the sum of the allocation of a series of single seats.
 >
 > Detailed results attached.
 >
 > David Gamble
 >
For the American system either IRV or Condorcet will be more tolerant of
third party votes than Plurality is, so we can expect more of them.  I
would expect Condorcet to do the most good because complete vote count
totals can easily be given to all and give an easily readable strength for
EACH candidate.

I am not sure what to expect of British voters, but the ones in this
example are mostly either pro-C and ANTI-LA or pro-LA and ANTI-C.  This
makes these more tolerant of LD candidates.  57.5% would rather see LD
than LA, 54.7% prefer LD over C, and 5.4% have no preference among these
three parties.  I see Condorcet as doing well for the majority of these
voters.

Let the candidates and voters have some variation, such as the nine or so
current Democratic presidential candidates and their backers are showing
in the US, and the best C, LA, and LD candidates might have a better
chance to win.

BTW - for IRV and Condorcet picking the same winners, I was thinking of
the common case where the same candidate should win by any system, and
even most near ties.  I expect most of the differences between them to be
cases where IRV trips over its spoiler problem (IRV has one - just not as
bad as Plurality's).
-- 
   davek at clarityconnect.com    http://www.clarityconnect.com/webpages3/davek
    Dave Ketchum    108 Halstead Ave, Owego, NY  13827-1708    607-687-5026
               Do to no one what you would not want done to you.
                     If you want peace, work for justice.