[EM] Example's dual has 2 of 3 candidates over -1/3 quota

Craig Carey research at ijs.co.nz
Wed Sep 13 18:27:58 PDT 2000


At 19:02 00.09.13 -0400 Wednesday, DEMOREP1 at aol.com wrote:
 >Mr. Carey wrote-
 >
 >Mike Ossipoff has been writing about reducing/minimizing the
 >"need for insincerity".
...
 >D- In the single winner case, the sincere/ insincere situation happens when
 >there is (guess what)- a divided majority.
 >
 >Polls before the election show *roughly* a *sincere* possible vote of
 >
 >26 ABC
 >25 BAC
 >49 C[A=B]
 >
 >Some of the C voters may want to be insincere and rank A > B or B > A.
 >
 >Some of the first choice A and B voters may then want to be insincere. Not 
so
 >amazing.
 >
 >I say so what. Majority rule is majority rule.



This is more a comment on a method finding the wrong winners, I suppose,
than on sincerity-like properties.

The majority rule is a rule is not desirable since it says there should
be 1 winner in all elections where it is specified that there be 0
winners. It is discredited. There is no respectability to the majority
rule because it is a rule that fails in 1 candidate elections. Other
rules manage to survive tests up to 2 or 3 candidate elections. So it
is not a bit bad.

To fix the rule, it can be said that the candidate with more than
               1
           --------
            1 + NW
of the total votes, must win. NW = the number of winners.

If it is a good method, it may satisfy the duality rule. FPTP does, and
so does my IFPP:

The winner of the above election is the loser of this:

-26 ABC
-25 BAC
-49 C[A=B]

Using the fixed majority rule, any candidate with over 1/3 of the total
of the votes wins. Hence both A and B win since they have 1st preference
votes -26 and -25, which is greater than -(33+1/3).

Hence C lose the election example that Demorep1 gave.

If something else wins Demorep's example, then the method could be a
bad method.

Demorep wrote "Some of the C voters may want to be insincere and rank
A > B or B > A.". If they do that, it seems from the above that it
ought not make any difference.

I got another example privately that has 2 of the 3 canidates over the
1/3 quota for losers. I have yet to look at that.

 >
 >My standard mantra- an election method works on the votes cast (not added or
 >removed votes -- unless some major felonies are being committed).
 >

The standard mantra is wrong.

 >To get ONLY *sincere* votes would require something like lie detectors
 >connected to the (now no longer secret) ballots.  No thanks.

((Demorep has a definition of his own. Can the data test methods (i.e.
formulae returning a set (of winners), and receiving a vector representing
ballot paper counts, ?.))



More information about the Election-Methods mailing list