At 04:32 PM 2/28/2007, tom smith wrote:
>----- Original Message -----
>From: "Anthony" <cirv@...>
>To: "General Discussion - open to all" <talk@...>
>Sent: Wednesday, February 28, 2007 11:45 AM
>Subject: Re: [Talk] Range Voting Distractions
> > Range Voting, which is not used in any government i know of. There is
> > little practical evidence to support claims made about it. IRV, on the
> > other hand, reduces negative campaigning, probably increases voter
> > turnout, increases sincerity of votes, and ensures a majority winner
> > (when you are picking a single seat, as in President, this is preferred,
> > in my opinion).
What we have here is a statement that there is
"little practical evidence" with regard to Range
Voting, but, then, claims are made about IRV that
are not supported by practical evidence, but
mostly by theoretical considerations or
speculations -- or that are simply not true.
Sauce for the goose is sauce for the gander.
At this point, if we are considering what voting
methods to push for politically, we have to
consider a number of things: how do proposed
reforms behave *in theory*, what practical
experience is there regarding them, and what is
politically feasible. We'll get to feasibility
below, but feasibility should not could our
understanding of the reforms themselves. It's
true that the "ideal reform" might be infeasible,
but that should not change our understanding of
it as ideal, and it may help us to choose from
among options that *are* feasible. That is, if we
understand the ideal, we may be able to choose a
path that will get us in that direction, whereas
if we only look at what is in front of us, as
options, we may choose a blind alley.
Now, the claim that IRV reduces negative
campaigning does not appear to be based on any
evidence whatever, it is speculative. Warren
posted some evidence to the contrary. In the
absence of better substantiation, this one can and should be dismissed.
Increases in voter turnout are also speculative.
I'd *expect* it to happen, but I would also
expect it to happen with Range and Approval, as
well as with a superior ranked method, such as
any Condorcet-compatible method. Why we would
choose IRV if what we want is an accurate ranked
method is beyond me. The only answer I come up
with is political feasibility. But why would we
consider IRV more politically feasible? It's
circular. Because a few reformers, possibly not
very knowledgeable about election systems -- or
for whatever reason -- fixed on IRV early on, we
should be nailed to it until the cows come home?
As to increases in voter sincerity, this could be
false. Yes, if we think of a vote for Gore as
"insincere", under plurality, if you preferred
Nader, then IRV would likely increase sincerity.
However, as with Plurality, IRV is vulnerable to
Favorite Betrayal Incentive. That is, it can harm
you to vote sincerely, under IRV. Does Lorenzo realize this?
What is true that in the situation where your
favorite is a third party candidate with no
chance of winning, you cannot be harmed by voting
sincerely for that candidate in first place.
Essentially, your favorite is moot.
Consider this, from http://wiki.electorama.com/wiki/Instant-Runoff_Voting
>IRV is unusual in that it does not satisfy the
>criterion in some situations, if a voter or
>group of voters decides to rank a preferred
>candidate lower, it can result in that candidate
>winning the election, whereas if they had ranked
>the candidate higher, according to their sincere
>preference, that candidate would not have won.
The web site in question is the wiki of the
Election Methods community. If what is stated
there is in error, anyone can correct it. If you
think it is wrong, you are invited to correct it.
I'd suggest, though, that before you do so, you
make sure you know what you are talking about!
Given that there is another method, very simple,
that is not vulnerable to monotonicity failure,
that never punishes you for voting for your
favorite, that costs nothing in ballot or
counting complexity, that eliminates the spoiler
effect as well or better than IRV, and that only
takes striking a few lines of the election code,
why should we, if we want to promote sincere voting, choose IRV?
Just Count all the Votes!!!
And if we want the ranked ballot of IRV, why not
use a Condorcet method? Once again,
Just Count all the Votes!!!
(With Approval -- which is simply what we have
with the no-overvoting rules stricken from the
Code -- all the votes are counted and the
candidate with the most votes wins. Can't be
simpler, really. It's actually simpler than
no-overvoting FPTP, because the special
procedures for discarding overvotes can be
dropped. With Condorcet methods, all pairwise
votes are considered, instead of only those which
are unmasked by the peculiar candidate-dropping
sequence of IRV. IRV essentially ignores
important votes, and that is why it is not monotonic.)
That IRV "ensures a majority winner" is, simply,
nonsense. It only does such a thing by a
technical trick. If you didn't vote for one of
the top two remaining after all eliminations, it
discards your vote. Thus the winner has a
"majority" of all votes remaining. Now, if you
have IRV with compulsory full ranking, yes,
again, technically, a "majority" winner will
remain. But only because they held a gun to your head, so to speak.
The fact is, though, that there can be a
candidate, *easily*, under IRV, who would lose to
another candidate in a pairwise election, as
shown by the votes. If you can actually see the
votes. In other words, there can be not only a
majority in favor of a candidate, but a *strong*
majority, over all other candidates, and that
candidate loses in IRV if they don't have
sufficient first-place support. This is "Center
Squeeze." It would be better if IRV advocates
actually undersood what they are talking about.
There is no single-poll election method that can
ensure a true majority winner. The closest,
though, would be Range with an Approval cutoff.
If a majority of voters approve a candidate, we
can consider this a majority winner. Approval
does this, but even Approval (or this Range
method) can't ensure a majority winner, if no
majority votes to approve the candidate with the
highest vote count. The best system if having a
majority winner is considered important would be
the hybrid Range/Condorcet method that would
consist of a Range ballot that was analyzed by
sum-of-votes -- range -- and by pairwise
comparisons -- condorcet -- and, where the
condorcet and range winners differed, there is an actual runoff between them.
*This* gets as close as possible to having a true
majority winner. Actually, however, *any*
election method can generate a true majority
winner if it is *ratified* by a vote after the
election, and if the candidate selected fails to
be ratified by a majority, the candidate becomes
ineligible, and a new election is held.
Now, for practical reasons, we may not want to do
this, but the fact is that it *could* be done
with Asset Voting. And if you don't know what
Asset Voting is, you should. Under Asset, votes
are considered "assets," which can be allocated
by candidates who receive them. Asset Voting can
be considered to create an electoral college with
the candidates as members. As such, it can hold
*many* votes. So it could use *any* election
method to pick a nominee and then vote on the
nominee. If a majority reject the nominee, that
nominee is out. This is actually standard
deliberative process, and this is how we'd elect
candidates if we used such process. We don't do
this because we think it impractical to
deliberate on a large scale, but reducing the
scale by selecting proxies (which is what
candidates under Asset Voting amount to) could make it possible.
This may seem impractical at present, until we
realize that any political party could adopt such
a process internally, without requiring changes
in law.... So the question I have is "Do we
really want democracy," and, if the answer is
yes, why don't we start to actually have it? We could. It's possible. Now.
> > I have reviewed the data on their site. I don't agree with them that
> > Single Transferrable Voting is inferior to Range Voting.
First of all, there is a lot of information on
that site. STV is quite a good method,
multi-winner. I expect that it would approach
Range Voting in performance. Range Voting is
superior because Range Voting actually uses a
measure of result quality to determine the winner.
Let's put it this way: how do we determine what
is a good election outcome, assuming that we know
how all the voters rate the candidates? Range is
simply a means of determining the degree of voter
satisfaction with an outcome. The Range winner is
simply the one who maximizes satisfaction. So, if
we assume honest votes, the Range winner *has* to
be the best winner. *By definition!*
(There are still issues, such as how to
amalgamate individual "utilities" to a "social
utility," I'm not pretending there are no
difficult problems or controversies, but what
I've written here is clearly true in round
outline. Range is the only method that uses a
true measure of election quality to select the
winner. Other methods use indirect measures, for
example, Plurality assumes that the best winner
is the one with the most first-place votes.
Except that is a reference to the voting system
itself. It's an *assumption* that such a winner
is best, though, as I think we agree, that
assumption is clearly false in reality.)
When it is claimed that Range is not in common
use, this simply is not true. Range is in
*common* use in opinion polling. If it were not
considered a reasonable way to amalgamate
preferences and opinions, it would not be so
commonly used. But, of course, polling differs
from elections in that polls don't determine
outcomes. However, the most common objection
raised to Range is that people won't vote
sincerely, and that this will allegedly harm the
outcome. This is actually a compliment to Range!
That is, in my opinion, we should really
understand what election methods are best with
*sincere* votes. We can then look at the effect
of insincere voting. Who does it harm, and how?
From my analysis, the person most harmed by
insincere voting, under Range, is the insincere
voter. Insincere voting, with Range, is based on
a misunderstanding of how the system works. Now,
this is currently being debated by many; however,
Warren's work with simulations indicates that,
even where voters are insincere (they vote
"strategically" which means that under Range they
vote extremes where in fact their opinions are
more moderate), Range performs better than other methods.
It makes sense. My own approach to this problem
has been to note that Range never provides an
incentive to vote insincere reversal, i.e., to
rate a less-favored candidate higher than one
favored. The most that it arguable favors is to
vote equal rating where in fact the voter has a
preference. Now, in Range, it can happen that if
you rate, say, your second preference lower than
your first, you could allow a third candidate
whom you prefer even less to win (by not having
given the second preference a higher rating).
However, this is actually quite unlikely to
happen in high-resolution Range, if it is Range
0-99, the difference between a notch down for the
second preference and a full rating for the
second preference is essentially 1/100 of a vote.
Further, this behavior is an individual issue.
The winner in the election mentioned, the alleged
problem election, is the one who was, overall,
rated more highly by the entire electorate. It's
not a monotonicity failure because you did not
harm a candidate by rating that candidate more
highly. Rather, you failed to rate a candidate,
not your favorite, highly enough to create a
victory for him or her. And if you thought
similarly to many other voters, we might also
note that you and your cohort must also have
rated that third candidate reasonably high -- or you are a small faction.
Applying election criteria that were designed for
use with ranked systems to Range can produce some
misleading results. Range ballots collect far
more information about voter preferences than
ranked ballots, so it would not be surprising
that Range can generate more accurately maximized
outcomes. In particular, what is generally
missing from ranked ballots is preference
strength, which can lead to preposterous outcomes
in the extreme, where a candidate who might cause
a civil war will be chosen over a candidate who will please *everyone*.
This neglect of preference strength is the
Achilles heel of ranked methods, and even fully
Condorcet-compliant methods are vulnerable. IRV is even worse.
Sure, these are theoretical considerations. But
they are confirmed by simulations, which would be
the next level of knowledge. As to actual
experience, there is little data comparing
election methods in actual practice. We can see
that, probably, IRV leads to or reinforces
two-party domination, whereas actual top-two
runoff apparently does not. But to really *know*
how election methods behave in the real world, we
need actual elections, which can be hard to get.
> I don't agree
> > with them that the election method is more important than multi-member,
> > non-majority districts whatsoever. Truth is, if you do multi-member
> > districts, it doesn't really matter which voting method you use, third
> > parties will probably do quite well. Choice Voting, which is both IRV
> > and Single Transferrable Voting (STV), is what we are promoting, as is
> > Fairvote.org.
It is deceptive to link STV and IRV. STV is an
excellent method, used multiwinner, because the
vulnerabilities of IRV mostly don't apply.
"Choice Voting" is also a name that could apply
to Range, which gives *more* choice to the voters.
It's been argued that Range ballots don't allow
equal rating while showing a preference. It would
be simple to fix that, but my question is "Why
bother?" Essentially, this would encourage
insincere *ratings*. If you really consider two
candidates equally qualified, and the system
allows you sufficient resolution (0-9 is probably
enough, but 0-99 certainly is), why should
*society* care which of them you "prefer"? It's noise, essentially.
But preferring A to B because you will try to
leave the country if B is elected, and preferring
A to B so slightly that it reverses on odd days,
are exactly the same to a ranked method. (Because
of the presence of many candidates, this is less
true. Indeed, there is a paper which used ranked
analysis to show a defect in the Condorcet
method... the example took many candidates to
make the point, and depended upon an assumption
that -- most likely -- the preference expressed
between A and G in A>B>C>D>E>F>G was stronger
than that expressed in A>G>B>C>D>E>F. Range, of
course, deals with this directly.
As to the stuff about election method being more
important than multimember, I have no idea what
he is talking about. There is a lot of material
on that site, but certainly this idea, if it is
there somewhere, i.e., somebody wrote it -- there
are a lot of writers represented on the site --
is not central to what we discuss. The Range
Voting list deals with all kinds of
election-related issues, including proporational
representation, apportionment, voting machines, etc. And it's all important.
As to it not mattering what election method is
used for multimember districts, this is simple
naivete. It can matter very much. Some methods
will give total domination to a majority party.
> > I have talked to the folks from Rangevoting, and they seem more
> > interested in discrediting IRV, which has become a successful and
> > progressing movement, rather than actually working to implement Range
> > Voting somewhere.
This simply isn't true. He might have that
impression, I can understand that, but there is
an effort under way in Colorado, for example. And
if we include Approval as a Range method -- and
most of us do -- there is actually a fairly broad
consensus that this is a very good first step,
given how terminally simple it is. There is a
somewhat hopeful effort to get it done by 2008 in
Colorado, at least in some trials.
The IRV movement likes to describe itself as
"successful," but that is, in my view, premature.
There have a been a *few* successes, that is,
jurisdictions that are now using IRV. Time will
tell if that is stable. There are probably more
people voting, from time to time, in "Approval"
elections than in IRV elections, in the U.S.
This is because, of course, of the rules
regarding conflicting referenda, which are
essentially decided by Approval -- if two
conflicting referenda pass, the one with the most votes wins. That's Approval.
The real bottom line is that there has been more
money poured into IRV, by CVD, plus some third
parties have been essentially hoodwinked into
supporting it. Absolutely, IRV is an improvement,
but it is neither the simplest nor the most effective improvement.
This is a democracy, I assume. We should consider
all the options, certainly at this level, the
level of discussion. And we should *discuss*
before we *commit*, don't you think?
It has been the refusal of IRV forces to
seriously discuss the issues that creates this
impression of our being "interested in
discrediting IRV." The fact is that we want
*knowledge* of IRV to be more accurate and
widespread. There is a lot of nonsense
promulgated about IRV, and we aren't the ones
promulgating it. Some of it was in this post,
especially if we include conditionally true
statements that are reasonably false when stated without the conditions.
(IRV encourages sincere voting. unstated: as long
as third parties are moot. unstated: compared to
plurality, not to other available methods
including simply allowing "overvotes.")
> > I can honestly say that i would support Range Voting over our current
> > system. I am not willing to organize a Range Voting campaign though
> > (unless someone wants to pay me to, as i can be bought). I would
> > support Libertarians trying it somewhere and would help in whatever way
> > i could so we can see.
Good. We can agree on that. Most of us already
would, in theory, support IRV "over our current
system." I've named a caveat: the tactics of CVD
lead me to be reluctant to help them "win."
They've really made some outrageous statements,
dismissing the knowledge and efforts of those who
have spent years studying election methods and solutions.
But the method we should *really* focus on is
Approval. Simple, cost-free, probably actually
reduces costs. Few of us think it is ideal
(though, actually, a few experts do, because it
forces consideration of compromises), but it
solves the first-order spoiler effect at far
lower cost than IRV, and it does not negate the
possibility of further reform. We think it may
whet the appetite. "Hey, great, now I can vote
for my favorite *and* cast a vote that counts.
But why can't I indicate which of these is my favorite. I want that!"
And then we could give it to them, in one of two
major ways, we could argue about it *then*. We
give it to them with a Range method or we give it
to them with a ranked method. I think Range is
better, all things considered. But it can be done
gradually, and it can be done in different ways
in different places. The simplest reform would be
to add a third possibility (which means two spots
to mark by a candidate) that allows the
specification of a favorite or first rank. Then
the question is how to count that. And we have
written a great deal about this, I won't repeat it here.
> > These people don't even know what Duverger's Law is. Neither did I,
> > until i met a political science professor online who spent a few hours
> > lecturing to me on the phone about why we have a two party system.
> > Rangevoting being used to pick single-members is still going to produce
> > a two-party system.
The fact is that we shouldn't be voting for
"members" at all. STV does pretty well, but I'm
not really represented until I have maximum
choice in *picking* my representative. I agree,
by the way, with the judgement that Range,
single-member, isn't going to end two-party
domination. Indeed, it will not alter most
election outcomes. But Warren, I think, may have
a different opinion. What Range *will* do is to
allow third parties to gain strength, measurably,
without sabotaging elections. Thus, in the long
run, we can see party reversal, where a third
party becomes one of two majors, the other declining.
I've come to have a bit less respect for certain
famous political science professors when I
realized what narrow boxes they sometimes think
in. We really don't know what effect Range will
have. But I'm not pushing for Range *right now*.
I'm pushing for Approval. Simple, cheap,
effective, allows third parties to register vote
totals that show their real support. Obviously,
that, by itself, isn't going to make them rise to parity.
I'm more concerned, however, about how the
*major* parties make decisions. The process is
erratic and full of hazards. Range would be
excellent for *advising* parties about how to
proceed. Warren thinks it would be good for
primaries, and he originally started the Center
in an effort to convince parties to start using
it as soon as possible. I think that Range isn't
ready for prime time, though his simulations are
starting to convince me. But I have no doubt that
Range is proper for polling, and that parties
could use it for that purpose. I'd rather see
political parties choose candidates by
deliberative process, through conventions where
representation is full. Parties could, indeed, use Asset Voting for that.
And, of course, so could independent citizen's
groups dedicated to reforming the system. What if
election reformers formed such an organization?
One that is *not* nailed to an agenda, but that
simply functions to facilitate communication, coordination, and cooperation.
> No one has suggested single-member district
> > elections using IRV is going to grow third parties; it only makes them
> > viable in larger races (President, Governor, etc.) and ensures majority
> > winners, eliminating the spoiler effect used to cow the sheople into
> > voting for one of the loser majors.
It is questionable that IRV makes third parties
"viable" in larger races. It allows them to
function, but without chance of winning. As soon
as they get within striking distance, IRV becomes
quite problematic. That is the point that we keep
making over and over. Going to IRV is setting up
a possible problem down the road.
> > I would love to come speak to any group about some of these issues. I
> > am promoting IRV and STV, and we are working on a city now to see if
> > they will adopt STV/IRV (STV for city council, 7 seats done in one
> > election; IRV for mayor).
Why not STV for council and Approval for Mayor?
Better, in my opinion, would be Asset Voting for
both.... At least I think that the option should
be understood. Frankly, I prefer parliamentary
systems, where the presiding officer or manager
is elected to serve at the pleasure of the
parliament. If the parliament is truly
representative, this is far more democratic.
The argument is, however, that it is unstable. It
is an essentially anti-democratic argument, it
assumes that the people are incompetent to value
stability. (And, in this case, it would not be
the people, but representatives of the people,
presumably more knowledgeable and in a proper system, generally wiser.)
> > I simply think Choice Voting is superior. If you want to know why, ask
> > www.fairvote.org.
We've tried. They don't respond with anything
that makes sense. My opinion is that they are
well aware of the shortcomings of IRV, but they
have decided to keep them under wraps as much as
possible because they made the decision that IRV
was the most politically feasible reform.
However, the people running CVD seem to me to be
so politically incompetent that I suspect that
their judgement on this might not be sound. (They
seem to have gone out of their way to be
offensive, which is not what skilled politicians do.)
(How could they be politically incompetent and
still have had some success? The likely answer:
Money, and serendipity. Election reform is a
coming topic, and they were there at the right time.
Lorenzo here does not really address Range
Voting. The only argument he gives against it is
that it is allegedly untried. He does not address
the known problems with IRV, he merely claims
that it is superior, but the reasons that he
gives for it being superior don't lead to that conclusion.
It is standard among political advocates who have
made a conclusion about what to support. They continue, regardless.
However, perhaps this will come to Lorenzo's
attention and he will at least investigate
Approval, which should be an option on the table
everywhere, it is so terminally simple. He should
also take a look at Warren's simulation results
-- which are being improved every day.
Contrary to what might be assumed, the
simulations are not an effort to prove Range
Voting the best. They are an honest attempt to
start to measure election method performance
under different situations, to start to do a
little more than guess about it. They are not
perfect, and the assumptions on which they are
based are not perfect, but Warren is quite aware
of this and is open about it. The work deserves
support, and we might note that, however
imperfect it is, it is better than anything else we have.