[EM] Cyclic Ambiguity in Condorcet

Dave Ketchum davek at clarityconnect.com
Tue Mar 12 13:37:22 PST 2002

On Mon, 11 Mar 2002 12:08:18 -0500 Narins, Josh wrote:

 > A> It's a bloody unlikely result.
 > B> The same current system for handling a perfect tie in 
winner-take-all can
 > be used (I would imagine this usually would involve throwing it into the
 > Legislature)
 > What's the big problem, here?
 > Three way ties are no more confusing than two way ties, if you ask me.
 > I'm new to all this, so, please be gentle.
Perhaps you were trying to apply commonsense where you were reading 
backing claimed for some theory.

Candidates CAN be equally liked by voters, vote counting BETTER be able to
recognize this, and the response BETTER be something more attractive than
"Joe went off in a corner to flip a coin and SAID the response was that
his candidate won".  All the systems I know of can get there.

I will throw in some more thought, since I have not seen much mention of
Condorcet recently.

There are two major voter interests, them caring about one or both:
       Indicate their preference, even is they see no chance of winning.
       Indicate their choice, among those with chance of winning.

So what do systems offer:
       Plurality - only able to specify one candidate, they cannot do both.
       Plurality with rerun - the rerun only gives a choice between
leaders, which they MAY care about.
       Approval - this sounds simple, until as a voter one tries to decide
how much approval must be felt to deserve approving one more candidate.
It is weak because, after investing as much effort as with other chances
to list multiple candidates, voter can not indicate which is best liked.
       IRV - this works fine for a couple strong candidates and any number
of weak ones.  Voter lists candidates in order of preference, knowing that
an indicated preference for a weak candidate will not interfere with also
indicating a preference between the strong candidates.
            IRV gets in trouble when voters get serious about LIKING third
party candidates.  IF, at any time while crossing off losers, one of the
strong candidates has less first place votes than the other strong
candidate or any of the third party candidates, that strong candidate has
LOST - its visible first place votes get crossed off and it never will
look any better as other losers get crossed off and make some of that
loser's votes visible to be instantly crossed off.
            Because of the above problem, strategy enters in to try to
avoid it happening.
       Condorcet - which I LIKE, for it avoids IRV's problem and thus has
no need for strategy to try to avoid them.  Thus the ballot and voting is
IDENTICAL with what IRV claims to offer.
            Condorcet can be bothered with cyclic ambiguity - A>B, B>C, and
C>A.  If it was exact, IRV could see the same tie.  If not exact, formulas
can be specified in advance.  Note that the formulas get sticky only near
ties - a place where who wins is not significant except as fuel for arguments.
            Condorcet has a couple other advantages over IRV for major
elections such as governor:
                 Even for a dozen candidates, the exact numbers to be
reported by a precinct, town, county, etc., can be printed on a sheet of
paper for human review and understanding, and adding in a few absentee
ballots cannot change winner unless the vote was that close to a tie.
                 With IRV each vote pattern has to be reported for central
analysis to be used in deciding order of deleting losers, changing order
can have major effects in totals, and thus can change winners even if no
tie had seemed near.
            As to approval - what I said above is important - I CAN want to
say who I like BEST.
   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.

More information about the Election-Methods mailing list