[EM] full rankings, voter desire for

[Abd ul-Rahman Lomax]

At 11:54 PM 10/15/2005, Chris Benham wrote:
>Abd,
>You wrote:
>
>>Note that if the method allows equal ranking, adding clones does 
>>not require additional ranks.
>How on earth do you work that out?  "Require" for what purpose?

If a method does not allow equal ranking, and if full ranking is 
desired, adding clones adds additional ranks without improving the 
expected outcome for the voter. I was using "clone" to mean an 
additional candidate who matches an existing candidate in rank, such 
that the voter is equally happy (or unhappy) with the outcome if 
either of them wins.

If full ranking is not provided and overvoting is not allowed, clones 
consume ranking space with no immprovement for the voter.

This is a very strong argument for allowing overvoting, it improves 
ballot efficiency. (Most Condorcet proposals seem to allow 
overvoting, i.e., ranking candidates identically, equivalent to 
Approval voting). It's not important if full ranking is provided, but 
providing full ranking, if the candidate set becomes large is 
impractical. I've seen it argued here that elections are rare that 
have *many* candidates on the ballot, but the fact that it can happen 
means that public election methods must be able to deal with the situation.

Practically speaking, there appears to be substantial resistance to 
election reforms that require *many* ranks. It is one of the 
obstacles to implementing IRV; so San Francisco only implemented a 
few ranks. I don't know if they allowed overvoting, but the failed 
IRV initiative in Washington specifically prohibited it (as I recall, 
the ballot was considered truncated at the overvoted rank.)

>You seem to be assuming that it doesn't matter which member of a set 
>of clones wins,

Yes, for anyone who considers them clones.

>  which is odd
>because it is perfectly possible that the two rival front-runners 
>are members of  the same set of clones.

This is different usage of clones, unless I misunderstand: clones in 
this meaning are those candidates such that every voter ranks them 
the same relative to every other candidate. So if every voter ranks 
A>B>D and A>C>D, and there are no other candidates, then A and B are 
clones. This does not negate voters having preferences within the set 
B,C. But this is not what I meant by clone.