[Election-Methods] Range V IRV - crux of "unconstitutionality" worry

Warren Smith wds at math.temple.edu
Thu Dec 27 19:35:17 PST 2007

Actually, upon trying to crystallize exactly what bothers me about IRV
constitution-wise, I realized something:

Suppose you want a voting system (with honest voters)
to be "unmanipulable by sponsoring rival candidates."
I.e. you want that it never can help a candidate's winnign
chances to sponsor a new rival to enter the race.

Obviously, Plurality voting fails this test: it helped 
Bush to sponsor Nader to enter the 2000 race. 
IRV also fails this test.

EVERY non-dictatorial unanimity-respecting single-winner voting
method based on rank order ballots fails this test.  (For example
IRV and Condorcet fail it.) Proof: This is just
a restatement of Arrow's theorem after you note that
any time a new candidate X entering the race swings the winner
from Y to Z, that benefits somebody (namely Z, here).

But honest-utility-voting passes this test (this is an ultra-honest
form of range voting where voters vote their honest utilities... 
not too realistic since no "normalization").

Plain range voting fails the test (with normalized range votes) 
but comes "close" to passing it since it is "close" to utility 
voting.  In fact, it DOES pass it in the time-reversed form 
that "if X drops out of the race, that will not affect the winner."

(This all isn't really a new realization at all, just a new 
wording of something we've known a long time.)

Warren D. Smith
http://RangeVoting.org  <-- add your endorsement (by clicking "endorse" as 1st step)

More information about the Election-Methods mailing list