Eric Gorr wrote:
>Ken Kuhlman wrote:
>> The method proposed below, which I call "Correlated Instant
>> Borda Runoff," is basically a tweak of Baldwin to solve the clone problem.
>There seems to be a problem with a case where a high-correlation would
>exist, but there are not actual clones...i.e. your method may unfairly
>eliminate a candidate from the race.
>Why should this not be considered a significant flaw?
Good question. I'll give you two answers.
First, since the eliminated candidate is a Borda loser, they're not going to
win the election anyway, so their elimination can't be unfair.
Second, clones are seen as a special case of correlated pairs. All correlations
devalue the point score of being picked in a certain rank order. An overly strict
definition of what it means to be a clone would result in strange election results.
Consider the election:
50:A>B>C
49:B>C>A
1:C>A>B
Under plain Borda, or a clone-proof version with a strict definition of "clone",
B would win, but 99% of the voters see B & C as (near) indistinguishable. A single
elector that votes otherwise shouldn't undermine the significance of this. "Clone-ness"
thus can't be a binary factor that is either "on" or "off," but a continuous function.
A method that recognizes this should reduce the importance of the BC pair defeating A
in the 49 B>C>A ballots and give the election to A.
----
Election-methods mailing list - see http://electorama.com/em for list info