At 04:30 AM 10/3/2005, Rob Lanphier wrote:
90 voters: A=7, B=6
10 voters: A=0, B=10
It has been proposed that Range votes be normalized, otherwise voters
who honestly recognize that no candidate is perfect and thus does not
rank at least one as a 10 will suffer vote dilution. It is possible
that ballot instructions could warn the voter that their vote will be
diluted if they don't rank at least one at the top rank. Or, with
full normalization [top and bottom], the above would become
90 voters: A=10, B=0
10 voters: A=0, B=10.
Obviously, A wins by a landslide.
A:630
B:640
B wins, even though 90% of voters prefer A to B.
By a slight preference. Still if only top normalization is used, the
vote becomes
90: A=10, B=8.52
10: A=0, B=10
still a strong victory for A.
But what if there were the same votes present in the system for A and
B, but there were other candidates, and it was clear from the A votes
that the preference for A over B was indeed very weak, as the numbers
would indicate. (There might be other candidates ranked as 10s, for
example, but scattered so that none of those candidates will actually
win. And others as zeros.)
I could easily argue that B *should* win that Range election.
But normalization makes sense, for what it does is to compensate for
the fact that people express different intensities of feeling with
what might be the same levels of difference between candidates. I see
no reason to reward extreme ranking; therefore normalization. 90% of
the people prefer A and 10% rate him as zero? It is patently obvious
that those 10% are exaggerating, or they are using a corrupt standard.
There is no possible way Range will ever get serious support, given that
weakness.
Normalization completely answers this particular objection.
Normalization, by the way, would probably be to 1, no matter what the
numbers on the ballot. The reason: each voter gets 1 vote. Range is
an Approval method. A vote of up to 1 may be cast for each candidate.
Range is essentially Approval with fractional voting possible.
If it manages to pass constitutional muster, it goes against
what I suspect is the instinct of most voters out there, including
myself. I cannot be brought to recommend a system that suffers from
such a glaring defect.
First of all, I think the defect has been misunderstood. Even if
Range is not normalized, most voters will know that voting less than
100% is a weak vote. And then there is an easy fix. Whether or not
votes in Range should be normalized is controversial within the Range
community; the side other than the one I've argued says that people
should be free to express such weak preferences as result in the
election shown above. Personally, I think that Mr. Lanphier's
objection will be a common one, if Range is not normalized, and I
also think that many will stick with that objection. Whether it is
valid or not.
----
Election-methods mailing list - see http://electorama.com/em for list info
