Jobst, I don't know that this criterion has been discussed in general. The CR example I gave came up a few years ago when were comparing cumulative votes and ordinary CR, whose ballot restrictions correspond to L(1) and L(infinity) norms, repectively. It was noted that for these norms the optimal strategies occur at corners of the corresponding unit balls. The only norm that doesn't give rise to corners or bulges in its unit ball is the L(2) norm, which corresponds to a perfectly round unit sphere. Forest
<<winmail.dat>>
---- Election-methods mailing list - see http://electorama.com/em for list info