This method would provide guaranteed Droop proportionality while still have range/score like effects. It is basically PR-STV for electing the candidates but RRV for deciding who to eliminate.
So, 1) Voter casts range/score vote 2) Ranked ballot can be inferred 3) Run PR-STV as normal until an elimination step is required 4) Reweight all the ballots based on current winning set and eliminate the lowest rated candidate 5) goto 3 This is slightly different to RRV as it eliminates the lowest candidate rather than elects the most popular candidate. Has that been considered as a variant for RRV (i.e. for each candidate? Assume all the other candidates are the winning set and work out the candidates rating and then eliminate the lowest candidate) Voters might insist on more rating resolution so that they can distinguish between ranks. Another option would be to require that the ranks are consistant with the ratings. For example, the rule could be that the ranking is inferred from the ratings with any ranking being used as a tie breaker. So a person might vote A: 100 (1) B: 100 (2) C: 100 (4) D: 50 (3) E: 0 (5) The rating would then be A>B>C>D>E as if there is a conflict between the rating and the ranking, the ratings win. This reduces the risk of people accidently spoiling their vote. Alternatively, people could be given 2 separate ballots and provide completely independent ratings and rankings. It might reduce strategy for either. ---- Election-Methods mailing list - see http://electorama.com/em for list info
