[EM] CR style ballots for Ranked Preferences

[Dave Ketchum]

I wander in looking for something better than plurality for uses up to
and including electing a governor in NY - where we may have a dozen
candidates next year:
     Rules for voters MUST be simple and understandable.
     Rules for deciding on a winner must be reasonably simple, and NOT
declare as winners obvious losers.

So we need preference voting.
Must be able to combine votes from thousands of precincts.

IRV clearly fails, due to easily declaring wrong winners - also has
trouble due to vote patterns being important (Condorcet only counts
pairs, with results that can be easily combined into sums for use in
deciding on a winner).

So far I LIKE Condorcet, though picking the best way to pick a winner
from a set of counts needs thought.

I like that without the >>> that I see below.

REAL question:  Could Condorcet both be improved AND remain explainable
to my voters with >>> introduced?  My suspicion is that the answer is no.

Dave Ketchum

On Mon, 10 Sep 2001 11:41:37 -0700 Roy One wrote:
> 
> Forest Simmons wrote:
> > In general, for N candidates a scale of zero to N(N-1)/2 is barely
> > sufficient to encode the relative preference strengths in this way.
> 
> Just for clarification, this implies that there must be a > for there
> to be a >>, and a >> to be a >>>, etc. Hence the name "ranked
> preferences": the relative strengths of the preferences are not
> quantified, but it is known which jumps are bigger than others.
> 
> The Condorcet Criterion should be modified in some way to take
> advantage of the extra information; however, it should not try to
> quantify the >s (for example, weight >> twice as much as >).
> 
> One (non-Condorcet) tabulation method I like is what I call Run-Off
> Approval. It (or something similar) may have been describe
> previously. I am pretty sure it is not summable. For each ballot,
> count a vote for each candidate to the left of the Big Jump.
> Eliminate all candidates who do not have a vote from a majority of
> voters. If more than one candidate is still in contention, repeat the
> process, using the biggest remaining jump on each ballot as the
> dividing line. If all candidates are eliminated, choose the plurality
> winner from the last round.
> 
> A Condorcet version of this might be to construct a pairwise matrix
> marking only the victories across the Big Jump, and requiring a
> quorum of votes to pick a winner. If no winner is found in the first
> round, consider the top two biggest jumps on each ballot. And so on.
> A liability here is that strategy might have voters choosing
> preferences all of the same strength, so that every contest is
> counted every round. One solution to that would be to count the
> leftmost of any equal-sized jumps as the biggest (making A >> B >> C
> > D the same as A >>> B >> C > D).
> 
> > Suppose that we have the following ballot with fully ranked
> > preferences:
> >
> > A >> B >>> C > D
> >
> ...
> > But how about the other four preferences (the ones that straddle
> the
> > triple ">>>" ) ?
> >
> > Should they all be considered as equal in strength, as in Dyadic
> > Approval, or should some of them be considered stronger than B >>>
> C ?
> 
> It seems logically consistent to count A >>>>>> D, as long as that is
> a qualitative assessment, and not a quantitative one. I think that a
> pairwise tabulation method will automatically take that into account, though.
> 
> =====
> ------- Roy One --- royone at yahoo.com --- Speaking for myself -------
-- 
 davek at clarityconnect.com    http://www.clarityconnect.com/webpages3/davek
   Dave Ketchum     108 Halstead Ave, Owego, NY  13827-1708    607-687-5026
             Do to no one what you would not want done to you