Abd ul-Rahman Lomax wrote below:
In other words, failing the Condorcet Criterion is not a flaw of
Range Voting, it is a strength. However, it is also true that under
some conditions, condorcet failure can indicate a problem with
Range, and so some of my work has been aimed at detecting and fixing
that problem.
...
From there I realized that the ballot could be full Range and that
therefore, from it, a Condorcet winner could be detected, and if
this winner differs from the Bucklin/Approval/Range winner, a runoff
could be triggered. What this amounts to is the rough equivalent of
a ratification combined with a runoff. Range voting easily may fail
to satisfy the basic democratic principle of majority rule, so,
technically, a Range result should be ratified. To be efficient,
though, the ratification may be combined with a consideration of at
least one alternative, and perhaps two or three. With a good voting
system, and with what have become, in the runoff, well-informed
voters, having the results of the first poll -- which could be
printed on the ballot! -- majority ratification of a result is quite
likely.
This is a quite good description of the border area between Condorcet
and Range. This approach may be useful in environments that are not
fully competitive but not fully non-competitive either. I tend to
think that there is no single answer to all the voting needs of the
world. There may be many kind of single-winner elections with
different needs and different voter behaviour. An environment that
hangs somewhere between being majority decision driven and non-
competitive expressed utility driven may not bee very common but
definitely an interesting special case. If the voters are fully
competitive and majority driven (very common in political elections)
then the Range part may be just additional information for polling
purposes, and the majority winner will win the runoff. In that case
basic Condorcet method would be sufficient (one would just lose the
more or less sincere polling info).
The described method could work better if one would arrange the runoff
between the Range winner and the winner of some Condorcet Method (if
they differ) (not only if there is a Condorcet winner) since then the
voters could be confident that they will have their chance of getting
the majority oriented winner, and they would be more free to give more
sincere Range ratings.
One approach would be to allow the voters to declare their strategy.
That would eliminate the need of a separate runoff election. One could
have e.g. three three options for the runoff between the Range winner
and the Condorcet Method winner. 1) Interpret my ballot as ratings. 2)
Interpret my ballot as rankings. 3) Rearrange my ballot so that it
will maximally support the (sincere) Range winner. (Fourth (maybe not
as popular, maybe more rebellious) option could be to maximally
support the Condorcet method winner.) The runoff would be a Range
election between two candidates. Strategy 1 would cast a "weak Range
vote". Strategy 2 would cast a majority/Condorcet oriented vote
(exaggerated "re-normalized" Range vote). Strategy 3 would allow the
voter to change her opinion to support the Range winner (maximally).
(There are some opportunities for strategic voting. The majority
oriented / competitive voters could give insincere ratings to pick
appropriate candidates for the runoff or just to make the poll results
look better (low points to strong competitors, or to candidates that
might rise too high in the next elections, or simply low points to all
others but one's favourites). Strategic ratings could play a role also
in the polls before the actual election.)
This combination of Condorcet methods and Range could be a nice
approach at least in environments where one "tries to teach the voters
to become more non-competitive". This situation might not be very
common in politics, and this balanced state of affairs would might
often not be stable (= a state that would last over several
elections), but certainly this approach is worth a try and a good
discussion point in some environments (maybe first in some clubs and
associations, maybe in some deliberative teams that you often
advocate). Maybe there are also some stable environments that could
use this approach continuously. It could be interesting to study what
would happen if there would be some stable segments of voters falling
into the different strategic categories. But I will not jump into that
for now.
Juho
On Jun 13, 2010, at 6:50 AM, Abd ul-Rahman Lomax wrote:
At 11:53 AM 6/11/2010, Kevin Venzke wrote:
Hi Abd,
The best way to decide how to apply rank ballot criteria to
Approval is to
have a framework for all methods into which you can translate
Approval.
People may not agree with you but at least it's clear what you have
done.
--- En date de : Ven 11.6.10, Abd ul-Rahman Lomax <[email protected]
> a écrit :
> My contention there was that Approval Voting satisfies the
> Majority Criterion if Plurality satisfies the criterion.
But you say this and then quote Woodall's Majority criterion, which
Plurality fails?
Plurality allows voters to place a candidate at the "top of their
preference listings." Does Plurality fail Woodall's Majority
Criterion? That depends on interpretations. The criterion was not
designed and defined well enough to be sure. What is a "preference
listing"?
There is social choice theory and there is election method theory
and they are not actually the same, though they are certainly related.
> The contrary argument, if I can summarize it, is that if a
> majority of voters prefer A to B, but actually vote for both
> A and B, then it is possible that B wins if some other
> voters only vote for B. The wording of the criterion by
> Woodall is
>
> > * Majority. If more than half the voters put the same
> set of
> > candidates (not necessarily in the same order) at the
> top of their
> > preference listings, then at least one of those
> candidates should be elected.
>
> The way this criterion is worded, Approval satisfies the
> criterion based on actual votes, but not necessarily based
> on internal preferences, for that majority might, for
> example, all prefer A to B but actually vote for A and B.
But within Woodall's framework, all methods have to be interpreted as
rank. Woodall uses an interpretation of Approval so that it will
fit in
his framework. You don't have to use that interpretation. But arguing
about whether a method satisfies a Woodall criterion without any
attempt
to stay within Woodall's framework seems futile.
Basically, the term "Majority Criterion" went on to be used by
others in various ways, with various definitions and interpretations.
I think you may know that my position on "voting systems criteria"
in general is that they are a failed attempt to create objective
standards for comparing voting systems. They do have uses, but they
become ammunition in political wars over voting systems, and the
actual practical implications of the criteria are often obscure. In
particular, some criteria *seem* to be obviously desirable, but turn
out, on deeper examination, to be contradictory to what people, the
"electorate," would be likely to decide if all the cards were laid
on the table, all the preferences were thoroughly known, and then
the group of people made the optimal decision for their collective
welfare.
Great example is the Condorcet criterion. The Condorcet winner can
be, under conditions that may be revealed if voters cast sincere and
accurate Range ballot, contradictory to the optimal decision, *as
the voters will agree when the true preferences and preference
strengths are revealed.*
In other words, failing the Condorcet Criterion is not a flaw of
Range Voting, it is a strength. However, it is also true that under
some conditions, condorcet failure can indicate a problem with
Range, and so some of my work has been aimed at detecting and fixing
that problem. What I came across, early on in this study, was the
fact that standard deliberative process, with its procedure and
requirement of a majority for any decision, was far more
sophisticated than anything I've seen said about it from voting
systems theorists. Repeated ballot is a very powerful method! But
with the vote-for-one limitation, it can be inefficient; in actual
practice, candidate withdrawals and voting shifts handle what could
be handled more directly with Approval. Range could make the process
even more efficient, and then I came to the understanding that
Bucklin was an Approval system, simulating a series of repeated
ballot Approval elections, driven by a Range ballot.
From there I realized that the ballot could be full Range and that
therefore, from it, a Condorcet winner could be detected, and if
this winner differs from the Bucklin/Approval/Range winner, a runoff
could be triggered. What this amounts to is the rough equivalent of
a ratification combined with a runoff. Range voting easily may fail
to satisfy the basic democratic principle of majority rule, so,
technically, a Range result should be ratified. To be efficient,
though, the ratification may be combined with a consideration of at
least one alternative, and perhaps two or three. With a good voting
system, and with what have become, in the runoff, well-informed
voters, having the results of the first poll -- which could be
printed on the ballot! -- majority ratification of a result is quite
likely.
Determining elections with a single ballot is a basic violation of
democratic principles, when a true majority has not been found that
has explicitly approved the result. Certainly it may be argued that
efficiency requires it, but the claim that multiple ballots are
impractical has been shown to be false again and again. Sure, there
is a cost, but ... apparently it has been considered worthwhile,
often. But don't expect the top two political parties, in general,
to support it. They would rather risk an occasional spoiler. Hence
top two runoff has *mostly* been confined to nonpartisan elections.
Fine. That's the place to start. Don't dump the best voting system
in common use to replace it with a system with the same problems,
but without the benefits -- other than fixing the spoiler effect
which TTR also fixes. Instead, replace the primary with a system
that can develop far more information from the electorate, that will
more commonly find a majority, because it counts all the votes, and
that is much more likely to find a condorcet winner, but which can
also detect an approval and range winner. Bucklin, driven by a Range
ballot with adequate resolution, used as a primary in a runoff
system, could be, with certain runoff rules, not only Majority and
Condorcet criterion compliant, but, as well, social utility
maximizing, reserving to the majority the right to reject a "range
winner" if they choose. Usually, I contend, they won't, unless it is
so close that the difference is academic with little damage being
done to social utility.
This is new, I believe, the concept of maximization of social
utility with ratification when necessary, thus preserving majority
rule (which is very important). Many theorists in the field have
simply assumed that a runoff would ratify the Condorcet winner, a
result of not considering how turnout affects results, and how
turnout is a basic component of democratic process. People who don't
care don't bother to vote, even if they have *some preference*. Thus
standard vote-for-one, particularly with repeated ballot, was more
sophisticated than we thought.
----
Election-Methods mailing list - see http://electorama.com/em for
list info
----
Election-Methods mailing list - see http://electorama.com/em for list info
