At 01:54 AM 6/14/2008, Jonathan Lundell wrote:
I don't think that the point is all that important wrt ranked
voting, though. The relevant question is whether there are clones
(which needn't be perfect clones) in the election. And the answer is
that the presence of clones is not dependent on the race being de
jure (or even de facto) partisan.
I think that something has been overlooked. There is an interesting
phenomenon that I've seen, and it seems to be consistent in
nonpartisan elections. Suppose we have three candidates, A, B, and C.
C is eliminated in the first round. Where do C's votes go? It turns
out that a pretty good prediction can be made from the existing votes
for A and B. If vA is the vote for A and similarly with vB, whatever
vote transfers are made from C come in the ratio of vA/(vA+vB) for A
and vB/(vA+vB) for B. In other words, it is as if the voters who
prefer C are a relatively unbiased sample of the voting population.
So vote transfers don't alter expressed preference order.
That was not at all a result that I expected to find. But, once
found, it seems to make sense in nonpartisan environments. Cloning is
a different matter, somewhat. If equal ranking were possible, the
clone issue would disappear.
A case in point is the just-concluded Democratic primary for
California State Senate District 13. A primary is, for our purposes,
non-partisan. The election had three candidates with a chance of
election, a progressive incumbent, a challenger to the left, and a
challenger to the right who ordinarily would have no chance at
election in this district, but saw a shot as a consequence of a
split between the first two. (You'll agree that we can treat a party
primary as a nonpartisan election, yes?)
Perhaps. Ideological position tends to matter more with party
primaries. I'd not expect a Kucinich supporter to fall back to
Clinton, say. (By this I mean that most Kucinich supporters would
prefer a different candidate than Clinton, who was relatively
conservative. The lower the level of the office though, the closer
the office to the people, the less this would be true and the larger
would loom the personality and name recognition of the candidate.
The very fact that you can describe this election in terms of
ideology indicates to me that it is, in this sense, a partisan
election. So I thought I would agree with you, but, in fact, it's a
partisan election with the "parties" being wings of the party.
"Our purposes" are not served by a narrow definition of "partisan"
that refers to formally organized parties, but rather by an
understanding of whether ideological categories are strongly
meaningful to voters, as distinct from personalities and general
consideration of trustworthiness and the like. Where there is this
kind of ideological association in the minds of voters, then vote
transfers from a candidate may go almost entirely to another
candidate, whereas what is happening in the elections I've studied is
that transfers are far more diffuse, and tend to not change the
relative position of the remaining candidates.
The spoiler effect is a result of ideological association, in the
examples I've seen. Nader didn't take votes away from Bush, for the
most part, and Buchanan or Badnarik didn't take votes away from Gore.
Where there are spoilers, the vote transfers can reverse the social
ordering and we see a "comeback election."
In the event, the left challenger pulled out a plurality win, but
it's easy to see how it might have been otherwise. Under IRV (or
approval, for that matter), the right challenger's supporters might
well have elected the incumbent; or the right challenger might have
beaten the incumbent and left challenger if they had more evenly
split the left vote.
It's also easy to see how this real-world election could have
resulted in heavily strategic voting in an approval election, and
that either IRV or a Condorcet-compliant method would have likely
resulted in a better result, for some reasonable definition of "better".
"Strategic voting" is really a misnomer with Approval. No preference
reversals are involved, as was always the case when strategic voting
developed its bad name. If the election has a majority requirement
(as is sometimes the case with party primaries), that "strategic
voting" is simply voters experimenting with their approval cutoff.
There is no insincere approval cutoff; the idea that there is comes
from an incorrect and confused association with the idea that
"approval" is some fixed thing, that it is independent of context,
when, in fact, what we "approve" depends on what we think attainable.
Bucklin voting in primaries seems to have had, it's been claimed,
about 10% usage of the lower choices. There really should be a lot
more research done, there were many elections, and we simply don't
know the results. 10%, though, is enough for minor candidates not to
spoil the election. Much higher vote percentages were seen in the
Duluth election that was reversed -- very unfairly -- in Brown v.
Smallwood. Smallwood had won the election despite being second place
in the primary round. That was a nonpartisan election, so comebacks
can occur. But this was Bucklin, and what we see are not vote
tranfers, but additional approvals.
All that aside, my point is more narrow. The distinction between
(nominally) partisan and nonpartisan isn't particularly relevant to
the question of election method.
Explain the difference between elections where there is a comeback
and elections where thre is not. Don't consider whether the election
is partisan or not. IRV isn't showing comeback elections, and while
the sample is still small, it's significant. Top two runoff does show
that effect. The primary given as an example doesn't seem relevant at
all. It was what might be called "semi-partisan," because of
contending wings of the party, and it wasn't either IRV or top-two runoff.
I mention nonpartisan because of mechanism I see that explains why,
contrary to expectations (including my own), IRV is reproducing the
results of plurality: voters aren't associating the candidates
together in some consistent spectrum of preference, as would happen
more strongly in partisan elections.
----
Election-Methods mailing list - see http://electorama.com/em for list info
