[EM] cyclic preferences

[James Green-Armytage]

replying to this message:
http://lists.electorama.com/pipermail/election-methods-electorama.com/2004-July/013453.html


Dear Jobst,

	There are two separate issues here. 
	The first deals with whether a voting system should make allowances for
irrational preferences.
	The second deals with whether cyclic preferences are irrational.

	Let me address the second question first.
	In your example, how does the voter infer that there will be a
considerable opposition to issue X if B and C both have a majority over A,
but B and C are tied? B favors issue A just as much as A does, and B is
tied for the lead, so why is issue X any more dead than it is in the
scenario where A was tied with C for the lead?? 
	Keep in mind that the majorities that the tied candidates have over the
losing candidates could be of any possible margin, anywhere from 1 vote to
nearly the whole electorate... so it's hard to infer how well an issue is
doing based on what the scenario is.
	Anyway, if you are successful in demonstrating a situation where cyclic
pairwise voting might be rational due to inferences about other things
that would be going on in the scenario where it makes a difference, that's
interesting...
	...BUT...
	...that doesn't mean that a person's PREFERENCES are cyclical. That is, a
person still has a transitive ordering of which candidates are better or
worse or tied or unknown. Maybe there are strategic reasons to vote a
worse candidate over a better candidate, pairwise, but I think that you
are no longer talking about your actual preferences, you are talking about
how you vote strategically. 
	So, I think that to have a cyclical PREFERENCE ORDERING is irrational,
but to cast pairwise votes in a cyclical manner may conceivably be
rational, and strategically viable.

	Now, let me return to the role of rationality assumptions in voting
methods design. I'm not saying that voting methods designers should impose
an idea of rationality on voters just for its own sake. However, I suggest
that separate pairwise votes in place of ranked orderings would lead to
two things. One, of course, is nuisance. But more importantly, I think
that it will lead to strategic manipulation. 
	By having a method that allows cyclical pairwise voting, I think that
what you are enabling most of all is not expression of sincere cyclical
preference (if it does exist, which I doubt), but rather new and weird
forms of strategic manipulation.

my best,
James