Dear Richard,
suppose d[X,Y] is the number of voters who strictly prefer candidate X to candidate Y.
First, let's clarify that "prefer" here refers to preferences as expressed on ballots (otherwise this criterion wouldn't be testable).
I suggest the following criterion ("Independence from Nonsupporting Pareto-Dominated Alternatives"; INPDA):
Adding candidate Z with d[A,Z] > 0 and d[Z,A] = 0 must not change candidate A from a winner to a loser.
Tideman's ranked pairs method violates this criterion while my beatpath method satisfies this criterion. Hehehe.
Hmm, so does IRV. Not a very discriminating criterion, is it?
-- Richard
---- Election-methods mailing list - see http://electorama.com/em for list info
