[EM] The Coming California Single Seat Election

[Forest Simmons]

On Tue, 19 Aug 2003, Adam Tarr wrote:

> Forest Simmons wrote:
>
> >Suppose that the (only, as far as you are concerned) issues, i1, i2, and
> >i3, are equally important to you, and that (in your opinion) candidate A
> >beats candidate B on two out of three, candidate B beats candidate C on
> >two out of three, and candidate C beats candidate A on two out of three:
> >
> >i1: A>B>C
> >i2: B>C>A
> >i3: C>A>B
> >
> >Why would you be insane to say that you prefer A to B to C to A?
>
> Well, if that were the case, then your preferences in candidates would make
> sense, but your stances on the issues appear to be schizophrenic.  All
> you've done here is abstracted the argument from candidates to issues.
>
> I guess the question becomes, can you imagine three stances on an issue,
> such that you prefer stance A to stance B, stance B to stance C, and stance
> C to stance A?  I can't see any rational reason for that.
>

Actually it's three separate issues:

On the first issue you like A's stance, B is in the middle, and you detest
C's stance.

On the second issue you like B's stance, C is in the middle, and you hate
A's stance.

On the third issue you like C's stance, A is in the middle, and you hate
B's stance.

So far nothing unusual or insane, right?

If it came to a choice between A and B, you could say, "I like A's stance
better than B's on the first and third issues; that's two out of three, so
rather than try to compare apples to oranges, I'll go with A."

If it came to a choice between B and C, you might say, "I like B's stance
better than C's on the first and second issues; that's two out of three,
so I'll take B."

If it came to a choice between C and A, you might say, "I like C's stance
better than A's on the second and third issues;  I'll go with the two out
of three winner, C."

I could counter (in a Borda vein) by saying (for example) that although B
does better than C on two out of three issues, C does much better than B
on that third issue, so they are really tied because they have the same
average rank.

But in order to admit that you have to be willing to compare apples to
oranges, (for example) by swallowing Borda, or (for another example) by
assigning utilities to each candidate's stance on each issue, a total of
nine utilities.

If we require that of a voter, then we might as well require them to fill
out utility ballots.

For the sake of those just tuning in to this thread, I'm not saying that
it is desirable to allow the voter to express a cyclic preference, I'm
just saying that such a preference pattern is not necessarily a
manifestation of irrationality, and that some voters might rationally
prefer the option of expressing the cycle over the hard work of assigning
utilities (or whatever it takes) to come up with a linear ranking of the
candidates.

Forest