Definition of Condorcet's method
seppley at alumni.caltech.edu
Tue Apr 23 01:21:22 PDT 1996
Bruce Anderson wrote:
>Since Steve and Demorep seem to have thought that Condorcet(M) was
>Condorcet(1/2), I think that I am adding to clarity here.
I never thought Condorcet is Condorcet(1/2). But I'm interested in
whether Condorcet(1/2) is a worthy method.
For a time, I thought that Condorcet used *margin* of defeat as its
measure of size of defeat, since that meaning appealed to my
intuition so strongly that I read what I wanted into others'
writing. This error was unrelated to Condorcet(1/2). I give credit
to Bruce for clarifying my mistake.
A few months ago in EM, we briefly discussed whether it makes a
difference if the Condorcet algorithm tallies the equally ranked
candidates on each ballot (what Bruce calls the "xq(i,j)" term) as
half votes or ignores them. At the time, since I thought size of
defeat was measured by margin of defeat, it seemed to make no
difference how the equal rankings are tallied since the margin is
the same either way.
I appreciate some of what Bruce has written here, but I don't
understand why he's so concerned with the definition of "pairwise".
Like good accountants, we'll probably have two or three sets of
"books"--definitions of our terms--depending on who the audience will
be. If Bruce wants to contribute his time to rigorously defining
terms for our academic audience, that's certainly his choice, but not
how I plan to spend my time. I'd rather use the heuristic that we'll
save time overall if we deal with ambiguities when they're a problem
instead of dealing with potential ambiguities preemptively.
In the case of "pairwise", I haven't seen any compelling argument by
Bruce why we need to be able to rigorously categorize methods as
pairwise or nonpairwise. Categorization is not our mission.
More information about the Election-Methods