Six criteria

Saari at Saari at
Fri Jul 3 12:52:16 PDT 1998

In a message dated 98-07-03 13:04:34 EDT, you write:

>Of course, there are only six important criteria: Pareto, Autonomy,
>Neutrality, Non-Dictatorship, Invulnerability by Exaggeration, and
>Independence from Irrelevant Alternatives.
>Unfortunately, it is impossible for an election method to meet
>all six criteria simultaneously. Thus, some criteria have to be

If these are the famous "Arrow" criteria, then one of them (I'm not sure which
name) states that the outcome can depend ONLY on the "preferences" of the
voters.  It thus legislates a ranked "A > B" style of voting.  Thus, for the
case where 60% rate A as excellent, C as very good and the other 40% rate A as
terrible and C as very good, that criteria mandates that any valid election
method MUST choose A over C, even in this case where 100% of the group likes C
whereas only 60% of the group likes A.  THAT criteria is the one that I would
prefer to toss out.

A criteria I would like to include would address whether the method is capable
of finding the center of a simple Gaussian distribution.  At least one popular
method (successive elimination of the candidate receiving the fewest first-
place votes - mentioned in Robert's Rules) fails this test.

Mike Saari

More information about the Election-Methods mailing list