EM vote on SW methods: summary of votes to date
dfb at bbs.cruzio.com
Tue Jul 30 02:42:02 PDT 1996
Steve Eppley writes:
> Mike O wrote:
> >As for how to count equal rankings in an IRO count, I claim that
> >both equally-ranked 1st choices should receive a _full_ vote.
> That doesn't violate one-person one-vote? Nor create incentives to
Sure, it does violate 1-person-1-vote, as exemplified in
1-Vote Plurality & MPV. That's why it's better. Because that
traditional 1 person 1 vote is the problem. The reason why
MPV does such a good job of keeping Plurality's problem is
that, like Plurality, it only lets a person vote for 1
alternative at a time, in 1 big overall-count, as opposed
MPV advocates have objected to that version of MPV, the one
that permits someone to give whole votes to several alternatives,
on the grounds that it violates 1-person-1-vote. I don't expect
to convince MPV advocates about that, but I suggest the whole
votes interpretation of MPV equal-ranking in our election,
because I'm certain that Kevin didn't intend to give
those 2 1st choices mere half-votes, in the MPV count.
Not that most of us are very interested in the MPV count
anyway, of course.
It could be argued that when someone can give whole votes
to more than 1 alternative, that that person is getting more
voting power than the other voters. Well, if that's so, then
what if someone voted _all_ the alternatives in 1st place. Then
that person would be having exactly zero effect on the election
result, and wouldn't be exerting any power on the election.
With any sw method, there's no reason to not let someone give
a whole vote to more than 1 alternative. In Condorcet, if I
list several 1st choices, then I'm counted for each of them against
everything else. Same in other pairwise methods. So why not
in MPV, or Plurality. Of course when that's allowed in Plurality,
Kevin's equal 1st choice ranking of those 2 alternatives doesn't
cast more than 1 vote for either of them against the others, even
if we give each a whole vote. It doesn't give the 2 Condorcet
versions any more help against the other methods. All it does
is avoid the MPV problem of which one to give the initial
vote to. It's like Approval compared to Plurality. The equal
1st choice ranking could be done because it isn't certain which
one has the best chance to win, and should therefore be protected
from early elimination, avoiding the usual MPV dilemma in that
situation. Or it could be done merely because someone considers
both alternatives good, and doesn't wish to vote either over
In Approval, if you vote for A & B, but not for C, D, or E,
then what you're doing is casting a vote that says that
A & B are better than C, D, or E. But you're still only
castng _one _ vote to that effect, and aren't being given
more voting power to over-rule someone who doesn't believe
as you do about that.
And suppose that your 2nd choice, B, wins because you did that.
Do the C voters have anything to complain about? Because you
were counted among the majority who voted for B against C?
Most people voted for B against C, and you were counted
_once_ as doing that. C voters have nothing to complain about.
So, in any sw method, there's no reason, when someone lists
several 1st choices, not to count that ballot as giving each
of those alternatives the full status of 1st place.
> rank a second choice equal to one's first choice?
Yes, it does give incentive to do that, but it removes any
incentive to rank a 2nd choice _over_ a 1st choice. I consider
that a good trade.
Actually, the incentive to not rank one's 1st choice alone in
1st place is something that must be blamed on _MPV_. It's just
a question of whether we want to create incentive to rank
the 2nd choice equal to the 1st chocie, or to rank the 2nd
choice over the 2nd choice.
No way would I ever claim that whole votes for equally-ranked
1st choices gets rid of MPV's problem. But I claim that it
alleviates the problem, by eliminating the need for defensive
MPV advocates have never accepted that modification. Just as
well, since there are much better methods than any form of MPV.
Since EM has so few MPV advocates, and I'm not one of them,
I have no real complaint if only half votes are given to
the equal 1st choices. Chances are that wouldn't bother Kevin
either, since he isn't an MPV advocate and, probably isn't
concerned about how MPV is counted, or what the MPV result
is. Besides, counting half-votes or whole-votes for the
equally-ranked MPV 1st choices on that ballot aren't going to
change the election result anyway, which is an additional
reason why none of us would object either way.
So I suggest that the person doing the MPV count (and that would
be Steve, since he did all the counts for the report) should
do the MPV count whichever way he prefers, with the assurance
that no one's going to object, and it won't change the election
Though I doubt very much that this would be necessary, as a
last resort, Kevin could say that, for the purposes of the
MPV count, he ranks one of those 1st choices in over the
other, and no alternatives in the same rank position--
in keeping with the traditional MPV rules. But that seems
unnecesary, since the solution in the previous paragraph
seems perfectly adequate.
> Well, at least it keeps the Instant Runoff definition as simple as
> possible. Its only alleged advantage, according to its proponents
> in CV&D. :-)
Yes, and a simpler definition doesn't outweigh MPV's lack of
merit. What if you could get a useless product real cheap,
or get a usable one for a slightly higher price?
I agree about not buying the "foot in the door" justification
for MPV: Proposing MPV to introduce people to rank-balloting
would be going to a lot of campaign work & expense, for
no real reform. And once MPV was adopted, people wouldn't,
any time soon, be interested in another reform sw method
when they'd just adopted a "reform" sw method. Maybe
a few generations later they would--but maybe not even then.
Because, consider how long people have put up with Plurality's
inadequacy. Maybe the "(tiny) step in the right direction" would
turn out to be all she wrote. Let's do it right the 1st time,
instead of saying "Guess what: The method that we just sold
you, & asked for your campaign work & contributions for, isn't
really very good, and now we'd like you to do it all over again.
This one's actually good; trust us."
> ---Steve (Steve Eppley seppley at alumni.caltech.edu)
More information about the Election-Methods