On 2 Jun 1996 Bruce Anderson wrote: -snip- >B. Some Procedures for Creating New (Compound) Voting Methods > from Previously Defined Voting Methods: > >If M1 and M2 are two (not necessarily distinct) ranked-ballot >voting methods, then define M1//M2 to be the ranked-ballot voting >method that applies M2 to the set of winners according to M1 (using >the preference orders as expressed on the voters ballots restricted >to that set of M1-winners). -snip- That parenthetical clause at the end is one option, but what about the alternative? Suppose the voters' preference orders include all the candidates when tallying the M2 scores, not just the M1-winners? In the case of Smith//Condorcet, the result is the same either way: Condorcet looks only at voter opposition in pair-losses, and the Smith set will include every candidate who pair-defeated any candidate in Smith. But in general the M1//M2 result depends on whether the voters' preference orders are stripped by //. Bruce, why did you propose only the one alternative? ---Steve (Steve Eppley [EMAIL PROTECTED])
