Contd: Re: [EM] Condorcet, Same Ranks, Ties

[MIKE OSSIPOFF]

>Right now IRV is increasingly supported as the only alternative to
>plurality. IRV advocates are very loud and very unified, giving them
>a cascading unfair advantage in the more progressive press
>organizations, especially through their mouth pieces in Center for
>Voting and Democracy, and in the Green Party. (I.e. Nader talks about
>IRV, but no other possibility).

Yes, and _The Progressive_ magazine has had articles advocating
IRV, and prints IRVie letters, but hasn't printed any of my
rebuttals.

>If those who believe that IRV is one
>of the worst alternatives, but also believe Plurality is worse than
>IRV, wish to make any headway, we need to unify behind some
>alternative that is easily implemented and understood. It seems to me
>that Ranked Pairs and Plain Condorcet/Schultz would be at the top of
>the lists for a unifying method.

Yes, Condorcet's method, in its various versions, is the rank-count
that can be defended, and which turns out to be the one that does
best by popular standards.

Which version? PC has the briefest definition. When answering people
who say that Condorcet is too complicated, I tend to offer them PC.

Smith//PC does better than PC by some criteria that the academics like.
I personally like it better, but not enough to outweigh PC's simplicity.

BeatpathWinner and SSD are first-rate for public elections.
(Cloneproof SSD is only needed for replacing SSD in small committees).

I've found that people who are really interested in single-winner reform
prefer the best to the simplest, and would prefer SSD or BeatpathWinner.

I defined SSD for someone with absolutely no familiarity with voting
systems or rank-counts, and she immediately understood it & liked it.
She wasn't just being polite, because, previously, when I defined
SD (A simpler method than SSD, which mentions the word "cycle" in
its definition), she didn't like it because of the strange notion
of cycles.



>Ranked Pairs does seem, to me,
>easier to explain than PC/S.

Ranked Pairs is easier to demonstrate in an example than SSD is,
and it's also true that Ranked Pairs has briefer definitions, such
as the ones that Blake & I posted. (But I question whether Blake's
definition is complete).

But, for the person who asks questions, Ranked Pairs requires
additional explanation, beyond the definition, to justify it,
whereas SSD's rules are their own obvious justification.

If Blake's definition is valid, then Ranked Pairs does have a good,
plausibly-justified definition.

>Also all single rank methods (IRV, RP,
>PC/S) suffer from the problem of making bad ballots more likely, and
>RP and PC/S suffer from the problem of an overuse of randomness to
>resolve ties.

But remember that in public elections, ties are vanishingly unlikely
anyway.

As for bad ballots, equal ranking is agreeable to Condorcetists.
The only bad ballot would be if someone ranks the same candidate at
2 rank positions. We've argued what to do in that instance, but I don't
suppose there's a satisfactory answer. I'd suggest to average the
2 rank numbers, and give him the rank equal to that average. If the
average is between 2 ranks, then randomly give him one or the other
of those 2 ranks. Then there's no way to spoil a ballot, short of
writing one's name on it.

>
>If I am to reverse the advocation of IRV in the Green Party without
>appearing to support plurality, I need to advocate an alterative.

It's difficult to say which Condorcet version is the best. Most
Greens haven't heard of the Clone Independence Criterion, and so
maybe it would be best to include an alternative count rule as a
tiebreaker. Maybe Borda, then Plurality. After they've adopted
Condorcet, then someone could bring up ICC and suggest Random Ballot
as the tiebreaker.

Would the Greens reject SSD by saying it's too complicated? I don't
know. I think BeatpathWinner is at least as complicated to define as
SSD, and doesn't have SSD's obvious & natural motivation & justification. PC 
is simplest, but it's criticizable. The criticisms
can be satisfactorily answered, but maybe something uncriticizable would be 
better, like SSD. Maybe Smith//PC is a good compromise for
a proposal to the Greens. It has the advantage that one need only
define PC, and then later mention the Smith set. And the advantage
of meeting most of the academic criteria that PC doens't meet.
Again, the Greens haven't heard of the Clone Criterion, so that won't
be a problem.

But I suggest testing these methods on some Greens. First try
SSD. Use a diagram. Make some dots on a paper, and draw an ellipse
around some of them, saying "These are unbeaten from outside this
area. The winner should come from among these. So if no one is
unbeaten, then we drop pair-defeats from in here, to find if that
makes someone unbeaten."  Then draw a smaller ellipse inside the
1st one, and point out that what you said about the bigger ellipse
is true of that one too, and so the winner should come from that
smaller ellipse, the "innermost unbeaten set". After that
explanation, the usual definition of SSD will be accepted.

Try SSD first. If some Greens accept it, and they very well might,
then that should be the proposal. If not, then try Smith//PC.
If they consider that too complicated, then try PC. They won't say
that's too complicated.

For elections in committees & meetings, suggest Cloneproof SSD, which
differs from SSD only in its stopping rule--it stops when the
current Schwartz set contains no cycle, rather than when someone is
unbeaten.

Norm or I can send an algorithm for implementing Cloneproof SSD.

Mike Ossipoff

_________________________________________________________________
Get your FREE download of MSN Explorer at http://explorer.msn.com