[EM] output ranking

Anthony Simmons asimmons at krl.org
Tue Apr 10 13:37:28 PDT 2001

>> Subject: [EM] output ranking

Certainly, expecting a ranking that carries the
same credibility as the winner would be quite
different from just choosing one winner.  On the
other hand, suppose we wanted to ask who came in
second in the recent Presidential election.
There may be no official criterion, but if the
results are known publicly, we can apply various
criteria anyway.

So it seems like it would be okay to have several
different, not necessarily consistent, results.
Would be instructive.

>> >One other possibility, if it's a possible possibility:  In
>> >addition to arriving at a winner, how about trying to get a
>> >ranking out of the poll.  That seems to be appropriate in a
>> >vote on election systems.

>> Sure. The easiest way would just be to rank
>> the alternatives according to their final
>> Approval scores. Better would be to delete the
>> winner from the ballots, and then start the count
>> all over. But on the Approval portion of the
>> ballot, there could be a question of whether
>> people would have voted differently among the
>> remaining alternatives if the deleted winner
>> hadn't been there initially. Maybe someone voted
>> for a certain other alternative only because he
>> perceived it to be the thing that could keep the
>> deleted winner from winning, and wouldn't have
>> voted for it if that winner hadn't been on the
>> ballot. Or maybe someone refrained from voting
>> for something because it could be a rival to the
>> deleted winner, and would have had no reason not
>> to vote for it if that winner hadn't been on the
>> ballot.

>> So making an output ranking by deleting
>> winners doesn't seem to work for these polls.
>> I suppose it could be done just by ordering
>> them by their final Approval scores, but I
>> wouldn't want to use that for any material
>> purpose, like saying that the topics will be
>> voted on in the order that they scored in the
>> final Approval count of the poll-topics poll.
>> For that, I'd rather have new poll-topic
>> polls. Because if we could really know how
>> people would vote among the remaining
>> alternatives, _that_ would be the best way to
>> pick the next winner. And the only way to know
>> that would be to hold another poll-topics
>> poll.

>> Ordering the alterantives by their final
>> Approval scores seems ok if it isn't going to
>> decide an action, like the order of subsequent
>> poll topics. But if the repeated poll-topics
>> polls were considered too much trouble &
>> delay, then of course the ordering by final
>> Approval scores might well be chosen even when
>> the result makes a material difference, as
>> when determining the order in which polls will
>> be done.

>> But of course people only voted on one poll,
>> and it wouldn't be democratic to use that
>> result to write an order for subsequent polls.
>> That's something that would have to be
>> specified before the vote on poll topics. At
>> that time, of course, a way of determining
>> that polling order would have to be specified.
>> If it's going to be determined from one poll,
>> then I guess it would have to be based on the
>> order of the final Approval scores.

>> Mike Ossipoff

More information about the Election-Methods mailing list