Adam Tarr said: >>Independence of Irrelevant Alternatives, > > Well, it's a ranked method that's pareto efficient and non-dictatorial, > so it fails.
Well, technically it's only a ranked method if there are sufficiently many ranks so that no voter is FORCED to rank 2 candidates equal (voters are of course ALLOWED to). So the "Generalized Bucklin" would be a ranked method, but MCA wouldn't be unless there are only 3 candidates. As an aside, whether or not a method is ranked is even a little more complicated than whether there's enough levels to allow maximum expressivity. Imagine CR with 3 candidates and 5 levels, 0 to 4. If it were a purely ranked method then somebody who gives the ratings A=4, B=2, C=0 would have the exact same impact on the election as A=4, B=1, C=0. Why? Because in a ranked method the only thing that matters is that you prefer A to B, not _how much_ you prefer A to B. But that isn't true in CR, so CR isn't a ranked method. Alex ---- Election-methods mailing list - see http://electorama.com/em for list info
