Some Comments on Bruce's article

[Mike Ossipoff]

First, it isn't that there's an "official" version of Condorcet's
method here; It's that there's a _proposed_ one here. The
votes-against version of Condorcet's method, where the the count
doesn't invent un-voted & mutually contradictory prefrences on a ballot,
is the Condorcet's method version that has been proposed on this
list. The only methods relevant to this discussion are those that
are proposed here as the best public proposal.

Also, if the translation of Condorcet's words, in Duncan Black's
book, says what I believe it says, then Bruce is incorrect when
he claims the Condorcet version that he defines is consistent
with Condorcet's proposal:

For 1 thing, I believe that Condorcert proposed considering only
an alternative's _defeats_. Bruce's definition ignores the
distinction between a pairwise defeat & a pairwise victory, and
is therefore NOT consistent with Condorcet's proposal. Bruce
goes by an alternative's smallest vote-for. But this could occur
in a "low-turnout" victory--a pairwise victory where few people
have voted a preference.

Also, Condorcet, as I understand it, proposed a count based on 
preferences expressed by voters. I believe that he didn't say anything
about making up preference votes that the voter didn't express. 
And so any method that does so seems inconsistent with Condorcet's
proposal for that reason, since its result isn't completely determined
by expressed preferences.

I don't know if Bruce's brief recent Condorcet definition includeds
Bruce's proposal to invent un-expressed preference votes, but it
does ignore defeats vs victories, and so the paragraph before
last applies to it.

I've already pointed out that my 1st Condorcet vs Copeland
letter to this list discussed the academic criteria named
by Bruce, and talked about how they relate to Condorcet's
method. Bruce's tally of what meets those criteria doesn't
have meaning, without a discussion of how important it is
to fail those criteria. I discussed why I don't consider
it important that plain Condorcet fails some of those criteria.
Of course Smith//Condorcet meets them all.

***

Bruce, unless I'm mistaken, says that Approval fails the majority
criterion. Actually, if we define that criterion psychologically,
saying that the favorite of a majority should always win, then
Copeland & Regular Champion fail the Majority Winner Criterion too.
That's because in Copeland, or Regular Champion, as in Approval,
voters will sometimes be forced to vote a compromise equal to
something they like more (or even over it, unlike in Approval).
In that way, in Approval, and in Regular Champion & Copleand,
an alternative that's favorite to a majority can lose. 

Bruce is using a bit of a double-standard here, applying stricter
standards to Approval than to Copeland & Regular Champion (& all
the Copeland versions).

If, however, we define MW (Majority Winner Criterion) according
to actual votes, instead of psychologically, then Approval
meets it:

Non-psychological statement of MW:

If a majority of all the voters indicate an alternative as 
sole 1st choice on their ballots then that alternative should
win.

By that vote-based definition of MW, Approval meets that criterion.

Again, we've got a little double standard going here.

***

I just wanted to mention these things, but I don't want the
fact that I've answered these topics to be taken as endorsement
of Bruce's timing in posting his articles & definitions. Right
now a vote has been proposed, and we're voting on a method
recommendation to ER.

If you want to discuss Bruce's articles & definitions, I suggest
voting now, and then discussing Bruce's articles & definitions
later.

***

Mike





--