[Election-Methods] RE : Corrected "strategy in Condorcet" section

[Abd ul-Rahman Lomax]

At 03:24 PM 8/15/2007, Juho wrote:

>Last October I wrote about ranked preferences, i.e. ballots like
>A>>B>C>D>>>E. That allows the voter in a way to define different
>approval like divisions. I think the concept is theoretically pretty
>but I'm not sure if the full set of capabilities is ever needed in
>practice and if the complexity can be justified with the achieved
>benefits.

Ranked ballots that allow equal ranking are vastly superior in 
performance to those that prohibit it, for a number of reasons, one 
of which can be addressed with automated voting equipment, others can't.

And, of course, there are reasons to avoid "automated voting 
equipment," generally.

Three reasons are:

(1) If you prohibit overvoting at a rank, you will see an increased 
number of spoiled ballots, for various reasons. Some of them are 
purely errors, others are deliberate. Voters don't always understand 
the rules, and I've never seen a ballot that said, "Vote for more 
than one and your ballot will be invalidated." (Sometimes the rules, 
as I have seen with IRV rules, only discard votes at the equal ranked 
level or below.) Allowing equal ranking preserves more of voter 
intent than discarding the overvotes.

(Generally. Consider Florida 2000; due to misleading ballot design, 
there were many Gore/Buchanan overvotes. Counting the votes would 
have given some extra votes to Buchanan, which would have been 
harmless. Almost certainly, these were votes intended for Gore. Now, 
it could be argued that they were intended for Buchanan -- though 
that's very unlikely in the case -- and thus that it would be unfair 
to give the votes to Gore. However, by presuming that the voter erred 
and that we cannot therefore discern the voter's intent, we discard 
what the voter has actually expressed. Other votes, for example, may 
have been cast for Buchanan and the voter did not realize that the 
ballot was misleading. We should take the ballots as writ.)

(2) Forced ranking introduces noise.

(3) Allowing equal ranking brings the method closer in S.U. 
maximization to Range. Generally, ranked methods force a single 
preference step; except at the bottom, no other preference 
distinction is permitted. Allowing equal preference turns a ranked 
method into Approval, should a voter decide to only use the first 
rank. Or into Approval at any lower rank. I.e., this is my Favorite, 
but if not this one, then any of these would be acceptable. This 
could fix center squeeze in IRV.