robert bristow-johnson wrote:

what do you mean: "weight"? rankings are just rankings. if a voter ranks Candidate A above Candidate B (independent of what the absolute rank values are), all that means is that this voter would vote for A if it were a simple two-candidate race with B. and all Condorcet seeks to accomplish is to be consistent with that social choice regardless if Candidate C or Candidate D were in the race or not.

it's pretty simple:

1. if a majority of voters agree that Candidate A is a better choice for office than Candidate B, then Candidate B is not elected. this imposes consistency with the 2-candidate race where we all agree who should be elected and why.

2. the relative merit of Candidate A to Candidate B is not affected by the presence of a third candidate, C. in the converse, this means that removing any loser from the race and the ballot, that this should not change who the winner is. if it does, that loser is a "spoiler". it is precisely the motivation for adopting IRV in the first place.

To my knowledge, Condorcet passes IIA whenever there is a Condorcet winner. If Condorcet winners are frequent, that's a pretty good property.

That is, if candidate A is a Condorcet winner, and you remove some other candidate B, A is still the Condorcet winner. If you add some other candidate C, unless C beats A, A is also still the Condorcet winner.

Some may not like the tradeoffs Condorcet bring (like failing FBC), but it bears keeping in mind, I think. While IIA (general spoiler-independence, as it were) might be too strong to be sensible in the general case, having a method pass it in certain cases is welcome.

Advanced methods can go further, as well: a method that passes independence of Smith-dominated alternatives will not be influenced by candidates outside the Smith set.

(Of course, if there's rarely a CW or if the Smith set is usually large, this doesn't amount to much. Offensive strategy attempts to create cycles in the strategists' favor.)

----
Election-Methods mailing list - see http://electorama.com/em for list info

Reply via email to