Re: [RangeVoting] Re: Range voting, like plurality, causes partisanship?
- At 01:38 AM 12/11/2005, rob brown wrote:
>Let me clarify. I think that given a choice between voting strategically,Yes. That is, some people will vote, quite simply, what they think,
>and voting sincerely, some people will do one and some another. It is hard
>to predict which they will do, but certainly a number will do one and a
>number will do another.
and others will try to anticipate how others will vote and will then
vote in such a way as to obtain the results that they personally
desire. Which is *not* that "the best person wins," but that "a
less-preferred candidate does not win."
There is a subtle difference between "the best person" and "my
favorite." *Of course,* I think my favorite is the best person, but
I'm also aware that this is my opinion, it is not necessarily the
opinion of everyone else.
And, in fact, the person I *actually* want to win the election is the
person who is most widely perceived as appropriate. Am I unique in
this? I don't think so. But the question is not asked of most people,
so they don't every think of the difference.
Voting methods are amalgamation methods. Strategy actually attempts
to defeat the basic purpose of the amalgamation process: to use the
collective intelligence of the society to select officers. It
attempts to substitute the preference of a faction.
Approval, while I support it, will still tend to perpetuate this.
>I think if a system is designed so that voting strategically and votingYes. Stable and predictable is good, if it does not come at the
>sincerely are one and the same, the system will be stable and predictable.
expense of the quality of the outcome. Condorcet methods, I suggest,
tend to harm the quality of the outcome.
So, for me, the question is whether Range will hurt that quality more
than Condorcet methods inherently do.
>Plurality fails miserably on this.It only fails in multicandidate elections. It works pretty well when
there are only two. Which is why plurality and two-party systems are
as if designed for each other. It fails when there are three.
> Approval has the problem that you haveThe major objection to Approval is that it does not allow the
>to be very well informed to vote with optimum strategy, but at least it
>doesn't feel like you are being insincere if you do so (as long as you
>interpret "approve" and "disapprove" to mean "relative to the other options
>with a chance of winning").
expression of preference, so, for example, the Nader voter may indeed
feel insincere in voting as well for Gore. But since implementing
Approval is merely a matter of tossing out the no-overvoting rule, it
should be the default proposal, the point man for election reform.
> Range, however, fails on this. By presentingWe don't agree on this, and it has not been shown. You'll have to be,
>the option of voting in degrees, a voter feels insincere if he doesn't use
>them. But doing so is, almost always, a poor strategic move.
as I said before, much more specific if you want to establish that.
>I think Condorcet does remarkably well on these. I am 100% a strategicThe problem is that "Condorcet" is not a deterministic method; it can
>voter, and very well informed, but I really would not know how to "game" a
>Condorcet ballot other than by listing my choices in the order I actually
happen that there is no Condorcet winner. So Condorcet methods
include a resolution mechanism. And these mechanisms, generally, can
be gamed, at least in theory. Will they actually be gamed?
I think not, just as I think that Range won't be gamed in the way you think.
>I find that unlikely. I can't imagine a lot of Bush voters rating Gore orUnderstand this as a failure of the imagination. Mr. Smith knows,
>Kerry as anything but zero.
from his poll, quite differently.
>But you know, you can't analyze this stuff mathematically, since there is soIndeed.
>much psychology involved.
> That's what happens when you give people a choiceSo you said that you can't analyze this stuff mathematically -- which
>between sincerity and strategy.
means precisely, or at least statistically -- and then you tell us,
quite precisely what will happen. Huh?
> I prefer set things up so there is noExcept that voting systems are methods of deriving decision
>conflict....to vote the best strategy, you specify your sincere
>preferences. Then psychology plays a much smaller role, and the results are
>far more predictable and stable. I'd think you'd like that, being a math
information from the collected opinions of the voters. Psychology
plays a role in this no matter how you slice it. You've elsewhere
acknowledged that Range collects more and better information from the
voters. You also think Approval is a good method;
apparently your understanding of psychology is such that society
functions best, you expect, when the options of citizens as to what
they express are limited. You want to protect them from worrying over choices.
That's fine, but it is not democracy.
>On a condorcet method, in a real world election, I think you'd be hardI don't disagree (except for certain problems having to do with
>pressed to find any strategy that is better than "vote your sincere
resolution of Condorcet cycles). However, the problem is precisely
that Condorcet is practically a binary system. In each pairwise
election, each candidate is 0 or 1. The information for intelligent
decisions simply is not there; all that can be done is to amalgamate
*preferences*; you would institutionalize this limitation in the name
I submit this: any election method which takes advantage of the
distributed intelligence of society is not going to be predictable.
And it is precisely this kind of election method that we need.
>Exit polls are not accurate indicators of how people would actually vote.Now, with Range it *seems* that there may be an incentive to vote
>They have no incentive to be insincere in an exit poll.
insincerely. However, I think that is an illusion. My intuition is
that the optimum vote is, as I've mentioned quite a few times now,
the expected utility of the candidate being rated, probably
normalized for the election set (but perhaps including write-ins).
I do think there is a problem with *any* single-stage election
method. In other words, the requirement that election methods be
deterministic, exact ties excepted, is an artificial one which
creates many of the problems we see.
>Also, with range, people's behaviour will change as soon as they learn howYes. And those telling them will be partisans. People are sheep,
>it actually works (as they would if they were used on real elections).
>People would quickly be told "vote the extremes, you are wasting your voting
>power if you don't".
indeed, but they don't necessarily follow the wolves. *Many* people
will ignore that advice. Will they be harming themselves? You haven't
shown that, and I suspect -- I have not proven -- that they will not.
>If everyone actually does follow a strategic course, then you've just gotThat information, as has elsewhere been noted, has a lot of use
>Approval and wasted a lot of money on collecting the rich information.
besides determining the winner. The expense is small, in comparison
to the value of the election. *Of course* if everyone follows a
strategic course, as you asserted is likely, then you wouldn't use
Range. Or at least not a high-granularity Range.
My personal favorite is Approval Plus, where in addition to standard
Approval voting, one may designate a Favorite. This answers the
number one objection to Approval voting, which is that there is no
way to indicate a favorite, thus there is a sense of betrayal in
equating, for some voters, Nader with Gore. Further, there is always
the matter of public election finance. If there is no way to
designate a Favorite, does the finance money get split between the
Democrats and the Green Party?
And then it is also possible to *use* the Favorite information, in
which case the election method is a Condorcet one truncated to three ranks.
> MoreIf this were true, then the objection would be cogent. Is it true? I
>likely, you get a mix of people voting strategically and voting sincerely,
>with the sincere voters being punished for their honesty.
haven't seen a demonstration, just repeated assertions, assuming,
apparently, that a particular outcome is both likely and strongly
undesirable to those who voted sincerely.
>I will agree that the effect is less than plurality. I just think, for allIf you really think this way, then surely you will join the movement
>the effort to reform voting, what a shame to not completely eliminate the
>"strategic vs. sincere voting" issue, and the force that drives people into
to eliminate winner-take-all elections, which is the true cause, and
which will not be addressed by Approval in the least. Consider
http://beyondpolitics.org/wiki for a start.
>Condorcet has a few subtle strategy issues. Very, VERY subtle. I would betHard to tell. But you are probably right.
>my last dollar that they would not have significant effect in a real
>May I ask, what exactly is your problem with Condorcet? I'm just not seeingI think I've expressed a problem; it is inherent in Condorcet and any
method which only allows the expression of preferences without
allowing the expression of preference strengths.
What I'd suggest is that, if there are problems with
strategy-sincerity conflict, that the method be refined. It is almost
certain that range information is likely to result, if properly used,
in a better election outcome. But first we need to define the latter.
If you define it purely in terms of preference, with no regard to
preference strength, then of course Condorcet is going to be your
best method. You *assumed* it.
That assumption includes, I believe, a polarized, partisan environment....
- On 12/11/05, warren_d_smith31 <wds@...> wrote:
>Well I have demonstrated in two essays how, even in simple situations with
> Let me reply to several messages at once.
> First of all, I am not sure what the definition of "partisanship" is that
> all have in mind. If you mean "parties exist" then I do not believe that
> you are going to invent a reasonable voting system in which parties will
> not exist.
> Multiwinner PR, single winner, IRV, STV, all have been tried and all led
> to parties
> existing. There is a considerable advantage to candidates in having the
> and experience that parties can provide them. Read books about "how to
> run for office"
> and that will soon become very clear to you.
small numbers of people, there is strong incentive for parties to form under
a system that punishes "similar" candidates.
What I mean by partisanship should be very obvious if you watch anything
having to do with US politics. Two parties that spend a huge amount of
effort trying to defeat each other, rather than reach a consensus.
If by "partisanship" you mean "the problem is politicians taking unnaturally
> extremeI don't argue for proportional representation. I am for reducing the
> then it has been claimed that in fact (quite contrary to Duff) PR
> multiwinner systems lead
> to MORE extremism on issues. There is a book
> Richard S. Katz:
> A theory of parties and electoral systems,
> Johns Hopkins Univ. press 1980.
> that spends its time making that claim and claiming PR is bad, 2-party
> domination is
significance of parties. PR actually requires parties (right?). I want a
system where parties simply become groups that advocate a cause, and may
endorse one or more candidates, but aren't particularly closely tied to any
one candidate. Advocacy groups are fine.
I think we have grown acccustomed to the idea that there should be a "right"
and a "left", and that if I agree with one side on one issue (say, abortion)
I should agree with them on another issue (say, the war in iraq). I think
that is stupid.
He doesn't like either party-list or STV-based PR. (It seems possible to me
> that my own "asset voting" new kind of PR might cure this problem.That's not really what I am saying. I'm all for choice, but I think the
> But that is speculative. Asset voting described in my papers #91 and
> #77 http://math.temple.edu/~wds/homepage/works.html.)<http://math.temple.edu/%7Ewds/homepage/works.html.%29>
> I find Katz's book rather strange, but there is some truth to it. And
> remember, PR brought
> us Adolf Hitler. Anyhow, PR is not going to happen in the USA for
> constitutional reasons,
> so if you want to improve the USA, forget about it.
> If by "partisanship" you mean "the problem is 2-party domination and
> hence lack of
system should gravitate toward centrist candidates who have nuanced views,
rather than ones on either side. And I would be very happy if we get a
full field of candidates with a lot of different mixes of views on different
>Well you are talking about an exit poll being an accurate indicator of how
> >Exit polls are not accurate indicators of how people would actually vote.
> They have no incentive to be insincere in an exit poll.
> --actually, you are definitely wrong about the latter - they do have
> motive to be insincere,
people would vote in a Range election, and that is crazy. People don't even
understand Range now (such as the "vote approval style for most effective
use of your vote"), so using that as some sort of indicator of how they
would vote in a real election defies common sense.
As for motive to be insincere, I see no reason to lie in an exit poll. Why
would I? Do I think I can influence late voters to vote one way or
another? What should I say to bring out those who will vote for my
candidate and make the ones who would vote for the other stay at home?
Trying to figure out how to do so effectively just makes my brain hurt.
Just tell the truth and be done with it.
> and probably in the same signs they would be insincere in reality. And weI've read it, and am simply not convinced. I agree that no voting system
> have EVIDENCE
> voters were insincere to us in several ways. As for "accuracy" that is
> Exit polls are generally regarded as extremely accurate - the most
> accurate polls in
> the world - as indicators of how people actually did vote. But since we
> conducting a poll about how you would have voted if the system were
> different - not
> quite te same thing - a lot of that was lost for us.
> >I will agree that the effect is less than plurality. I just think, for
> >the effort to reform voting, what a shame to not completely eliminate the
> >"strategic vs. sincere voting" issue, and the force that drives people
> >opposing clusters.
> It indeed is a shame not to completely eliminate this issue. However,
> there just is
> no voting system that eliminates it completely. Period. That is a
> theorem by Gibbard.
> So since it cannot be eliminated, the best one can do is to reduce it.
> Now range voting is provably better than EVERY system that employs
> ranked-ballot type votes,
> in several ways related to honesty, see
eliminates strategy completely. But Range doesn't even come close, nor does
I would love to see you, in a real election, try to vote on a condorcet
ballot in a "strategic" way that is in any way different from the "sincere"
way. I don't think you could do it.
[Non-text portions of this message have been removed]
- At 11:52 AM 12/11/2005, warren_d_smith31 wrote:
>And remember, PR broughtWorth remembering well. Not necessarily an argument against PR, but
>us Adolf Hitler.
definitely an argument against considering PR a panacea, at least
party-list PR. Asset Voting could be quite a different animal, since
it allows people to run for office without waste. That is, if you
don't win, your efforts are not wasted if you receive *any* votes. So
you don't *have* to campaign, you merely need to be trusted by a few people.
(I assume that if Asset were implemented, write-ins would be allowed,
and that qualifications for candidates -- such as petition numbers --
would be easy to satisfy.)
> Anyhow, PR is not going to happen in the USA for constitutional reasons,PR has happened in the U.S. It's not going to happen any time soon
>so if you want to improve the USA, forget about it.
for the federal government, except possibly with the electoral
college, which could rather easily become a PR institution. In fact,
I have a mechanism which should theoretically work, to get from here
to there. As with most of these ideas, it takes more than one person
with the idea to have any meaning practically.
But states could elect assemblies by PR; yes, it would take state
constitutional changes, but those are a dime a dozen, practically.
State constitutions are *much* easier to amend than the federal one.
>Also, Condorcet methods often are complicated to describe and cannotThis may assume complete ranking. If multiple candidates per rank are
>be run on today's
allowed, then a truncated ballot can be used, and it could be argued
that more information than that is not really useful. It should be
noted that plurality is a Condorcet method with a maximally truncated ballot.
(There is also a problem in that individual ballot data is important
for multirank Condorcet systems. So totalizing machines would not cut
the mustard, you cannot derive the necessary preference information
from them. However, I've always thought the dependence on such
machines silly. It is quite easy to manually count ballots, and there
are volunteers aplenty who will help, and systems to prevent such
volunteers from biasing the counts. Australian elections use
handwritten rank numbers. Talk about hard to count!)
>As for motive to be insincere, I see no reason to lie in an exit poll. Why--because you want the poll results to say something. That's why.
>Just tell the truth and be done with it.--I wish...
We got a lot more alleged Nader voters (that is, who told us they voted
Nader in the real election) than the official results said existed.
This was too big to be a statistical fluke.
So either the official results lie or our pollees lied to us.
I think the latter. Everybody wants to vote strategically (Bush or Kerry)
to avoid wasting their vote, but many want to pretend Nader has
a lot of support (because they think he is the best) so that he will
in fact, get more support. So they lie to the pollster.
In any event, I advise not arguing
with experimental facts. You seem to have a penchant for that.
("Clearly, voters will behave thusly, otherwise they are crazy."
Another thing, we got a lot of Bush and Kerry zero-scores. This
obviously was due to strategy not complete honesty. So our
voters clearly were strategizing to some degree.
- --- warren_d_smith31 <wds@...> wrote:
> Let me reply to several messages at once.By "partisanship", I was assuming that we were talking about the
> First of all, I am not sure what the definition of "partisanship"
> is that you
> all have in mind. If you mean "parties exist" then I do not
familiar two party dominated system, with each party diametrically
opposed to the other.
> If by "partisanship" you mean "the problem is politicians takingThat was not at all what I meant. PR certainly leads to more
> unnaturally extreme
> then it has been claimed that in fact (quite contrary to Duff) PR
> multiwinner systems lead
> to MORE extremism on issues.
extremism in elected candidates, elected by a small minority. There
are plenty of real examples.
However, a PR elected chamber does not lead to extremist resolutions.
Apart from making sensationalist noise, extremists elected by PR
have little power. A PR chamber is more likely to get stuck in
inaction, being unable to form a majority consensus.
> If by "partisanship" you mean "the problem is 2-party dominationOn this subject, I suspect that two party domination is here to stay,
> and hence lack of
> lack of democracy" then IRV and plurality are bad since lead to
> 2-party domination.
that it wouldn't be broken by approval, range or condorcet. I think
it is more important that the two parties are encouraged to evolve,
as opposed to being stabilised by the system. I believe that
plurality and IRV both stabilise an existing big party because both
(for different reasons) are heavily biased against candidates running
on a platform largely similar to that of a big party.
Warren's arguments along the lines of the nursery effect are
persuasive. Under range, I see that there is no systematic bias
against a candidate running with similar but distinct policies to a
major party. If this candidate better represents the people, then
range will reward that candidate. This would either lead to this
candidate's popularity growing in time, though I think it more likely
that it will lead to the major party being motivated to shift its
position. Both are good outcomes.
Under plurality, the above candidate damages his position due to vote
Under IRV, he can be ignored, with his true popularity never being
considered (preferences from the major party voters are ignored).
It is not clear to me exactly what would happen under approval or
Send instant messages to your online friends http://au.messenger.yahoo.com
- On 12/11/05, warren_d_smith31 <wds@...> wrote:
>Well, I am not saying everyone will be sincere in the exit polls. I am
> >As for motive to be insincere, I see no reason to lie in an exit poll.
> >would I?
> --because you want the poll results to say something. That's why.
> >Just tell the truth and be done with it.
> --I wish...
> We got a lot more alleged Nader voters (that is, who told us they voted
> Nader in the real election) than the official results said existed.
> This was too big to be a statistical fluke.
> So either the official results lie or our pollees lied to us.
> I think the latter. Everybody wants to vote strategically (Bush or Kerry)
> to avoid wasting their vote, but many want to pretend Nader has
> a lot of support (because they think he is the best) so that he will
> in fact, get more support. So they lie to the pollster.
> In any event, I advise not arguing
> with experimental facts. You seem to have a penchant for that.
> ("Clearly, voters will behave thusly, otherwise they are crazy."
> But... no.)
saying that you have no established facts to back up your implication that
the exit polls give any idea of how people would really vote in a Range
election, given that strategy in voting is highly different than the
strategy of answering exit polls. Even if there is strong strategy to
answering exit polls, as your example shows, it is clearly different from
I have common sense that says it is unlikely to be a good indicator. But
you have nothing to convince me that it would be a good indicator, other
than speculation, which seems based on weak logic at best.
[Non-text portions of this message have been removed]
- At 02:05 PM 12/11/2005, warren_d_smith31 wrote:
>Can you provide some URLs where we can find out about qualcommTry this: http://www.qualcomm.com/ir/proxy.html
I read the 2004 Proxy solicitation.
>Floyd Norris recently wrote an interesting NY Times piece aboutI've written about this many times. The problem is the way that the
>corporate democracy and the lack of it.
>Do Companies Need a Little Democracy?
>December 9, 2005 - front page of business section.
>He basically argued that corporate democracy was a myth and mostly
>it is a 1-party system. He wondered if things could be better.
proxy system has been implemented. It gives existing management far
too much advantage, so that investor oversight becomes difficult or
impossible. Essentially, existing management may solicit proxies,
which many clueless investors readily provide without realizing the
implications. Others simply don't want to bother.
Large institutional investors sometimes hire companies which do
nothing but serve as proxies.
Smaller investors could do the same, essentially, without any change
in the system. This is typical of FA/DP plans for changing society:
they use the existing structures in a new way.
Basically, if small investors in a company simply join a shareholder
interest group with is an FA/DP organization, and they participate
minimally, just enough to decide whom in the organization to
provisionally trust, and they choose such a person as proxy -- in the
FA/DP organization, not necessarily as the proxy to the corporation
-- they have created a body capable of intelligent decision, in which
every shareholder is represented by someone they trust, or by someone
chosen by someone they trusted, directly or indirectly.
The FA/DP organization then determines consensus courses of action,
where it can find them. Even where it cannot, recommendations go back
to the individual shareholder through the Delegable Proxy network, so
recommendations are coming from someone the shareholder has decided
to trust. The whole process is open so individual shareholders can
see for themselves how their representative came to his or her
conclusions. And then the shareholder chooses an actual proxy to the
company. There is no need for the FA/DP organization to come to an
agreement on a single proxy; it is enough that the organization is a
way for shareholders to cooperate with others of like mind to share
proxies and thus be actually represented at the shareholder meeting.
I see utterly no reason, as a shareholder, that I would not join such
an organization. There is no down side. There are no long meetings to
participate in, unless I want to do that. There is no cost, unless I
voluntarily decide, for example, to share in a proxy's expenses.
Collectively, such expenses would be peanuts, compared to the value
of the stock, if the corporation is worth anything.....
Except there is one reason not to join: inertia. But that won't stop
it, it only needs a few people to get going.... It will simply take time.
- At 02:59 PM 12/11/2005, rob brown wrote:
>I don't argue for proportional representation. I am for reducing theAs normally implemented, yes. But it is not intrinsic to PR. Asset
>significance of parties. PR actually requires parties (right?).
Voting essentially allows interest groups to form ad-hoc for the
purpose of creating representation.
If you don't know what Asset Voting is, you should read Warren's
description. His implementation is fairly complex, I came up with a
simplified version called FAAV, Fractional Approval Asset Voting.
The common element in all Asset Voting variations is that candidates
receiving votes may either use these votes to elect themselves, or
they may reassign them to help elect someone else.
Suppose you decide to have a 50-member assembly. You hold an election
under Asset Voting. Candidates which receive 2% of the vote are
elected. Then the fun begins. Those who were elected immediately will
normally have extra votes to allocate. And those who were not elected
have *all* their votes to allocate.
The voter may divide his or her vote. Essentially, they are voting
for an elector, or for a committee of electors with fractional votes
each. In original Asset, voters voted a decimal fraction, such as
0.345, with all fractions adding up to 1.000. Only a math professor....
Fractional Approval is much simpler. You vote for one, that candidate
gets one vote (same as with Warren's scheme). Vote for more than one,
i.e., vote for N, each candidate gets 1/N vote.
The system would work quite well with Fractional Approval not
allowed. But I see no reason whatever to prohibit it (just as the
prohibition of overvoting in standard election practice in the U.S.
is unjust and unnecessary, resulting in the loss of many ballots --
this particular trick turned Florida in 2000).
Sometimes people say that they could not trust anyone with their
vote. As if they were not already trusting elected representatives
with their votes!
The skill involved in good government is largely the skill of
deciding whom to trust, to whom to delegate authority and power. So a
good governmental officer is probably also a good person to *choose*
governmental officers. So you can simply vote, with Asset, for the
candidate you trust the most, and see where it leads. I expect it
would generally lead to better outcomes.
Dividing Approval votes into fractions would be a horrible idea if
not for the Asset reassignment, which causes none of those votes to be wasted.
(It's an interesting analysis to note that the only vote that isn't
moot is a vote cast for a winner. This is the simplest argument for
why Approval Voting does not violate the one-person, one-vote
principle. Only one vote *at most* of multiple votes cast isn't moot.
Take out all the others, the election result does not change. The
voter has not cast more than one effective vote. Another way to look
at it is that the voter is voting separately in each pairwise
election; in some pairwise elections, the voter effectively abstains,
either by voting for neither or by voting for both. Never is more
than one vote from one voter given to an individual candidate. It is
ironic that the same is true for IRV, but one objection I've seen
raised by IRV supporters to Approval is that the latter supposedly
violates the one-person, one-vote principle. If Approval does, so
does IRV. Both involve casting, at one time, multiple votes.)