The Pareto Criterion is based on votes, not on
sincere preferences. Approval passes Pareto.
Pareto says: "Never elect someone over whom
someone else is voted by everyone."
You claim that Approval violates Independence
from Irrelevant Alternatives, because you
believe that IIAC is about sincere preferences.
I've never seen a precise definition of IIAC
that is about sincere preferences. Can you name
a method that would pass such a criterion? Of
course we can't test such a claim untill we have
a precise definition. Right now all we have is
your claim that Approval fails something that
you haven't defined, and which, if someone defined
it, would surely be failed by all methods.
The only precise definition that I've heard of
for IIAC says: Deleting a loser from the ballots
and then re-counting those ballots should never
change who wins the count.
If you think there's a version of IIAC that
Approval fails, then you should give a precise
definition of it. And then show me a method
that passes it.
You said my worst fear is that in IRV voters will
vote tactically and gain the same outcome that
would happen in Approval. You got that right.
Do you know what the difference is? It's that
with IRV, those voters are forced to dump their
favorite, by voting someone else over their
favorite. In Approval, no one ever has any
incentive to vote someone else over his/her
favorite.
Riker showed that, if the voters have complete
information about eachother's prefences, then
the sincere Condorcet winner will win--no matter
what the voting system is. Voters will use
some sort of strategy to gain that outcome. It's
merely a question of what they have to do.
IRV makes them dump their favorite by voting
someone else over him/her. Approval will never
give anyone reason to do that.
You say Approval doesn't let voters express all
their prefernces. No, but it counts those preferences that they consider the important ones to express. Unlike IRV. When Middle gets
eliminated before your traveling vote reaches
him, then your voted preference for Middle over
Worst never gets counted. IRV capriciously
decides for you what preferences will be counted.
I'd rather decide that for myself, thank you.
As we can do in Approval.
Is IRV better than Plurality? I say IRV is
worse than Plurality. If candidates' support
tapers gradually away from the voter median
position, then transfers coming in from the
extremes will build up nonmedian candidates
enough to eliminate the median candidate.
That's a typically expected situation. What
could be more normal or typical than that
scenario? But in that scenario, IRV does worse
than Plurality. IRV will defeat the Condorcet
winner, the voter median candidate even if he/she
is the Plurality winner--as will often be the
case.
IRV, unlike Plurality or Approval, can act
contrary to how you vote. You can make someone
win by voting him lower, where he'd have otherwise
lost. In fact, Professor Steven Brams has shown
an example in which some voters cause someone
to win by moving him from 1st place to last
place, where he'd have lost had they left him
in 1st place.
Sorry, but that's nonsense. Voters have a right
to expect that the voting system won't act
oppositely when they vote one way instead of
another.
In fact, IRV also violates this criterion:
Adding to the count some identical ballots that
vote X over Y shouldn't change the winner from
X to Y.
No, IRV messes up in every way. IRV is worse
than Plurality.
Aside from all that, by every criterion that
measures for getting rid of the lesser-of-2-evils
problem, IRV is no improvement over Plurality,
whereas Approval and Condorcet bring big
improvement by those criteria.
There are innumerable ways of counting rank
ballots. If you don't really do it right, it's
a mess, and we're better off without it. Condorcet
does it right.
RWE is the same as IRV, but without the eliminations. I'm flexible about its name.
As with IRV, the lowest candidate's ballots give
a vote to their next candidate, but he keeps his
and stays in the election. Of course, even if
he's lowest candidate again, that same ballot
can't again give to the same next candidate.
Count ends when someone has vote total over 1/2
the number of voters. At that time the canddidate
with the most votes wins. Or if no one gets a
majority, the count ends when everyone has given
a vote to every candidate on their ballot. Again,
the one with most votes wins.
Regrettably, RWE is a little wordier to explain
than IRV. But it avoids IRV's worst problem-causer: Eliminations.
RWE would be an excellent compromise with the
IRVists. Seriously, will you please e-mail
Rob Richie and tell him that? You can tell him
that I won't criticize RWE proposals.
Finally, this time I remembered to put in
the paragraph markers. Will they work? Let's find
out.
I still don't know what was the "Arrow" criterion
that Homunculus said that Approval fails.
In my previous posting, I asked Homunc for
the name of the criterion to which he refers,
and a precise definition of it.
That isn't really a lot to ask. Without an
answer on that, Homunc's claim about a criterion
failure for Approval means nothing.
How convenient, to make a claim that you can't
support, and say that you won't read the thread
again. That way you try to avoid responsibility
to explain or justify your claim. No, I'm sorry,
but an unjustified claim remains unjustified
when the person who made the claim runs to evade
questions about his claim.
Humunc, when referring to that "criterion" said
something about a voting system getting the
right answer when everyone agrees. There's one
Arrow criterion about everyone agreeing:
The Pareto Criterion.
Pareto says that a voting system should never
choose an alternative if another altenative
is voted over it by everyone.
That's the neares that any actual Arrow criterion
comes to what Homunc said.
Approval complies with the Pareto Criterion.
Homunc was being incomprehensibly vague when
he alluded to some unspecified criterion that
Approval fails. He didn't say anything.
Homunc said that Approval will cause people
to go with the flow, and vote for the more
publicized candidate (like Gore instead of Nader),
where IRV wouldn't.
First, in Approval, you
never have to vote for anyone instead of anyone.
If you vote for Gore, you can vote for Nader too.
Homunc thinks that, with IRV, you can safely
rank sincerely. I thought I'd clarified that a
long time ago. If in IRV you vote Nader in
1st place, and Gore in 2nd place, Nader might
get enough votes to eliminate Gore. Nader might
then lose to Bush. No that isn't contrived or
far-fetched. In my previous posting I described
two common, ordinary situations where that can
happen and where the voter has incentive to
vote a lesser-evil over his favorite.
Homunc: I tried to write to you by individual
e-mail, but your e-mail address didn' work.
I feel it was reasonable to reply to your
posted claims via a posting.
I felt that it's important to reply to the
newsgroup, because they heard your comments, and
so I'd like them to also hear my repies. I'll
also reply by individual e-mail.
Two different Australians, who probably don't
know eachother, both told me that it's difficult
for small parties to get their members to
vote sincerely, because those voters feel a need
to vote a big-2 lesser-evil in 1st place, to avoid
wasting their vote.
Your criteria for tactical voting will be
commonly met.
Say there are 3 candidates, Favorite, Middle,
& Worst. You like them in accordance with their
names. The voting system is IRV.
Say, as will often be the case, they all seem
about equal, roughly. That's all we know about
the expected numbers. And Middle is closer to
Worst than to Favorite, and likewise his voters.
That means that Favorite isn't going to win.
He can't win by immediate majority, or by
transfers. Any point in voting Favorite 1st? No.
If you vote Middle 1st, then you just might
save him from elimination, preventing Worst from
winning. Voting Middle 1st can give a better
result than voting sincerely. Voting sincerely
can't give a better result than voting Middle
1st. With respect to vote configurations that
are consistent with the above assumptions,
insincere voting dominates sincere voting in
that common, ordinary situation.
Different situation: Say there are lots of
candidates, and say, as should be typical,
candidates' 1st choice support tapers gradually
from the voter median point. Eliminations will
start from the extremes, and votes will transfer
inward. Soon some nonmedian candidates will
have enough votes to eliminate the median
candidate, even if the median candidate is the
favorite of more people than any other candidate.
When that happens, a majority are losing their
lesser-evil compromise, because they voted
sincerely instead of guessing what compromise they
need and voting him in 1st place.
Both of those common, ordinary situations
have IRV making people vote insincerely in order
to prevent the election of someone whom they
especially dislike.
Anyway, in general, when there are a number of
candidates, how likely is it that your traveling
vote will be on the candidate you need to protect
when you need it to be there? If it is, that's
a nice coincidence.
Look, the whole purpose of rank balloting is so
that we can express all our preferences. But
how much good does it do to express them and not
have them counted? Condorcet reliably counts
all the preferences we vote. Approval admittedly
doesn't let us express all of our preferences,
but it reliably counts every pairwise preference
that we express--every pairwise preference that
we consider important enough to express at the
expense of less important preferences. In
Approval, the voter is the one to choose which
preferences of his will be counted. In IRV,
the IRV count capriciously decides that. I'd
rather decide it for myself, thank you.
You said that your motivation to strategize is
proportional to how well-informed you are.
Not at all! How well informed are all the
progressives who think they need to vote for
Gore instead of Nader in November? One can
be terribly uninformed, and vote some terribly
uninformed strategy as a result.
In Approval, if there's no winnability information
available, no frontrunner probability information,
then your best strategy is to vote for everyone
who is above average in merit.
But in our typical situations, you simply vote
for the same lesser-evil that you think you need
with Plurality--the difference being that you
also can vote for everyone whom you like more,
including your favorite.
You point out that voting is a way to express
support. Exactly! That's why it's better to use
a method that will never force us to dump our
favorite by voting someone else over him. That's
another reason why Approval is better than IRV.
In Approval there's never any incentive for
anyone to vote someone else over their favorite.
That can't be said for IRV.
In what way is Approval Constitutionally
questionable? 1-person-1-vote doesn't mean
the voter only gets one vote. It means that
all voters have the same ways of voting available
to them. Approval, but not IRV, counts the
preferences that everyone expresses.
Your "Instant Approval" is an excellent
mitigation compromise for IRV. The IRVies have
rejected it. We offered it to them years ago.
Someone devised it and called it Runoffs Without
Eliminations (RWE). I strongly urge you to
offer RWE to the IRVies of CVD, and to advocate
for it. I'd never criticize an RWE proposal.
RWE is more than adequaet.
You said Approval fails Arrow's criterion about
getting the right answer when everyone agrees.
Would you please tell me the name of that
criterion, and give me a precise definition of
it? I'm not aware of it. The 1st "criterion"
that I usually find in Arrow articles is
one that says the method must be a rank-balloting
method. Some of us don't consider that a criterion
because it make no requirements about the
outcome. That's why I spoke of outcome-criteria.
Approval meets all the outcome-criteria. IRV
doesn't. Which "criterion" does Approval fail?
The one that says the method must be a rank
method. I'm more interested in outcome-criteria.
What it really amounts to is that the theorem
says: If you want any semblence of majority
rule, and if you don't want the deletion of
a loser from the ballots to change the count's
winner, then don't use rank balloting.
Picking nits? I don't like it when it's common
for people to have reason to regret that they
voted sincerely, or to be afraid to. If that's
a nit, it's a big one.
Mike Ossipoff
Homunculus wrote:
Quick response:
2. I didn't mean to start a flame war between IRV, Condorcet, Approval, et al. To me, any of these systems are so immeasurably better than the US status quo that the differences are immaterial.
I reply:
The merit difference between single-winner
methods are drastic & dramatic. CVD would have
us believe that there aren't important differences. CVD is mistaken. I've posted
several criteria met by Approval but not by IRV,
for instance.
Homunculus continues:
My reading of the political situation in my state is that IRV is the only one of these with a good chance of being adopted in the next few years,
I reply:
CVD always claims that IRV is the only
alternative single-winner method that would have
a chance. They've never supported that claim.
Approval is so much simpler than IRV, and so
much less drastic a change, that it's incomparably
easier to define, propose, enact, & implement.
Approval would have a much better chance than IRV.
But if IRV is the only alternative single-winner
method being promoted in Washington, then
_that_ could cause IRV to be the only method
likely to win. Maybe CVD has no opposition in
Washington.
Homunculus continues:
so I support IRV. If everyone only supports their own pet version we'll get nowhere.
I reply:
So then let's all get in line behind me:-)
Wrong. Let's all support what we believe is
adequate, and may the best method win. I and
some other believe, and can show, that IRV is
entirely inadequate. Why should we support that?
3. How did I misquote Arrow?
I shouldn't have saids that you misquoted Arrow.
Every source gives a different version of
"Arrow's" criteria. I've never seen Arrow's
original book or article,and so I of course have
no idea which source is giving Arrow's own
criteria.
In fact, I like your version of Arrow's criteria
better. That's because you list two
outcome-criteria that are failed by IRV, but
passed by Approval. All the other Arrow versions
list outcome-criteria which are all met by
Approval, and one of which is failed by IRV
(Independence from dropped losers, aks
Independence for Irrelevant Alternatives,
abbreviated "IIAC"). With your version, Approval
still passes all the outcome-criteria, but
now IRV fails _two_ of them, not just one.
In additon to IIAC, among your criteria, IRV
fails Participation:
Adding to the count some identical ballots that
vote X over Y should never change the winner from
X to Y.
Homunculus continues:
4. IRV isn't as different from approval as you suggest. Remember, you don't have to rank everyone; if you only "approve" of 2 candidates, only rank those two.
I reply:
But when you need to protect a lesser-evil
from elimination, IRV makes you rank that
lesser-evil in 1st place, over your favorite.
In Approval you never have incentive to vote
anyone over your favorite.
Homunculus continues:
But don't be so factional about this: The problems with approval, IRV, or condorcet arise only in pathological situations, whereas the problems with the current situation are a matter of course.
I reply:
How could you make that claim after I've posted
common, ordinary situations where IRV will
make people vote a less-liked candidate over
their favorite? In fact, with IRV, sincere voting
will be an unstable state of affairs. Sincere
voters will be had by insincere voters, and
sincerity will disappear.
The Australian voters are the people who are
actually using IRV. Do they agree with you
that IRV will never in actual conditions
give incentive to rank a lesser-evil over one's
favorite? No. Australian voters routinely vote
a lesser-evil in 1st place, over their favorite.
So, 1) I've shown that incentive to dump one's
favorite will be common & ordinary in IRV; and
2) Australian experience confirms that prediction.
Homunculus continues:
I like IRV because, of the easily-explicable options, it gives the least emphasis on voting tactics.
I reply:
As I said, IRV will often require insincere
tactics, as borne out in Australia.
Homunculus continues:
I rank nader, bush, gore, as 99, 2, 1 but would "approve" the first two for tactical reasons. Enough people like me and approval voting could easily elect a candidate that the majority dislikes.
I reply:
Approval requires twice as many mistaken compromisers, as compared to IRV & Plurality,
in order to give away an election. That's because
, unlike Plurality & IRV, the compromise doesn't
take the form of voting a lesser-evil over one's
favorite.
And if Nader people vote for Nader & Gore
the 1st time, then Approval will reveal Nader's
true support, and in the next election, people
will know to vote only for Nader, not for the
Democrat. Approval quickly homes in on the
voter median candidate, and stays there.
Approval is the most stable single-winner method.
IRV is one of the least stable, with its
frequent leaps to extremes.
Homunculus continues:
Approval voting also fails in the case of extreme factionalism. If we used approval voting to choose a voting system for the US, and everybody (as you apparently would) voted only for their favorite system because of nitpicking flaws in other systems, we'd effectively be using plurality voting (and we'd probably end up electing plurality voting).
I reply:
If we voted nationally on voting systems, using
Approval for the vote, I'd vote for Approval
and Condorcet, the two adequate methods.
From what you say, you'd vote for IRV, Approval
and Condorcet, and Black's method. What's your
problem with that? Everyone votes for what
they consider adequate, or for what they
consider above-average. The optimal strategy
in Approval, when we don't have winnability
information or frontrunner probabilities, is
to vote for all the above-average candidates or
alternatives.
Hopefully ICANN can take such a vote, after
some discussion.
Mike Ossipoff
Someone called "Drooling Dog" said that, with
the voting system known as Instant Runoff
(aka Preferential Voting, the Alternative Vote,
MPV, Hare, etc.), the winner is at least someone
acceptable to a majority. Not so. Consider this:
40: ABC
25: B
35: CBA
B gets eliminated and A wins, even though a
majority have indicated preference for B over A.
There's only one sense in which IRV (Instant
Runoff) complies with majority rule: After IRV
has eliminated everyone except for the last two
candidates, then whichever of those is preferred
to the other by a majority wins. That, my friend,
is a very poor excuse for majority rule.
Someone replied: "But I feel it's the voters,
not the method, that eliminated B". Yeah?
Did that B>A majority want to eliminate B and
elect someone whom they consider worse than B?
Drooling Dog also said that with IRV you never have to insincerely uprank someone in order to
defeat someone worse. Sure you do. Those CBA
voters could save themselves from an A victory
only by insincerely voting B in 1st place.
I've told in other postings why that will be
a common, ordinary need. In other words, IRV
doesn't get rid of the lesser-of-2-evils problem.
In another posting, I also told why Approval
does much better in that regard. Approval is
the method in which every voter can mark any
1 or more candidates on the ballot, giving
each one marked one whole vote. With Approval
no one ever has incentive to vote anyone else
over their favorite. That can't be said for any
other method. I also defined a weaker
lesser-of-2-evils criterion called WDSC, and
explained that Approval complies with it and
IRV fails it. Condorcet, defined at the websites
that I've referred to in other postings,
meets WDSC and some similar & stronger criteria.
The only single winner reform proposal that's
widely promoted is IRV. But there are other ones.
Some of them are incomparably better than IRV.
Approval & Condorcet are the best.
Mike Ossipoff
Sorry to post so many messages in one day. This
is my 1st day on the bulletin board, and the
only day when I'll post this much. Also, this
is my last message for today.
Metis said that Instant Runoff (IRV) is the
only method that he knows of that doesn't force
voters to vote strategically, to express a
preference that they don't have in order to avoid
having their preference ignored altogether.
Actually, IRV _does_ force voters to strategically
vote preferences that they don't have. Maybe
if you vote Nader in 1st place, he'll eliminate
Gore, and then lose to Bush. Bush wins because
Gore got eliminated while your traveling vote
was on Nader.
This will be common. Say that Favorite, Middle,
& Worst are known to be roughly equal in
1st choice support. So no one has a majority.
Say Middle is closer to Worst, and so Middle
will transfer to Worst if Middle is eliminated.
Now, as a Favorite voter, do you have any reason
to rank sincerely? No. Favorite can't win.
He can't have a majority, and noting will transfer
to him from Middle. By voting Middle in 1st
place, you can protect Middle from immediate
elimination, and help him win, keep Worst from
winning. Voting Middle 1st can get a better result
than voting Favorite 1st, but voting Favorite
1st can't get a better result than voting
Middle 1st. Voting Middle in 1st place, therefore,
dominates voting Favorite in 1st place.
Strategically, voting Middle in 1st place, over
Favorite, is all that makes sense, if you're
a Favorite voter whose 2nd choice is Middle.
So, Metis, IRV forces you to vote an unfelt
preference. It forces you to reverse a preference.
With Approval, no one ever has any incentive to
vote someone else over their favorite.
Metis spoke of preferences ignored. IRV ignores
your preference for Gore over Bush when it
eliminates Gore while your vote is on Nader.
That's why I say that IRV is unconstitutional,
because, when it violates the one person
one (counted) vote principle, it violates
the equal protection clause, when it capriciously
chooses whose preferences to count & whose
to ignore.
Mike Ossipoff
I already replied to this message by Homunculus,
but I've been told that my messages don't have
much chance of getting read when they're
"anonymous coward" letters, as my initial
reply was, because I hadn't gotten registered
yet. So I'd like to post my reply one more time.
Sorry about the duplicate posting--this will
be my last duplicate.
Homunculus needs to read the guidelines for this
bulleting board. We're asked not to use abusive
language. Since IRV is rejected and disliked by
nearly all academics who discuss it, does that
mean that the academics are all trolls?
Homunculus says that IRV is generally considered
better than Plurality. Generally by whom?
The marching homunculi that CVD recruits?
And isn't it a pitiful testimonial when the
best that can be said for IRV is that some people
think it's better than Plurality?
Homunculus says that the Supreme court endorsed
IRV. Wrong. The article that he links to doesn't
discuss IRV. IRV does elimination. The article
that Homunculus linked to doesn't mention
elimination. What they're talking about is
Bucklin's method. Bucklin is pretty good. It's
slightly better than Approval, though not as
good as Condorcet.
Let me list 2 websites:
http://www.electionmethods.org
and
http://www.barnsdle.demon.co.uk/vote/sing.html
[end of website listings]
Those websites define Approval & Condorcet,
and define many criteria for evaluating &
comparing voting systems.
By the way, Homunculus, why are you afraid to
use your actual name?
Homunculus misquotes Arrow. A brief Arrow summary:
If you want even the most minimal majority rule
compliance, and if you don't want the deletion
of a loser from the ballots to change the count's
winner, then don't use rank balloting.
Ironically, the Irvies mention Arrow, trying to
argue that it means that we can't improve on IRV
and must give up & use the abysmally inadequate
IRV--when actually, Arrow's results-criteria say
that Approval is better than IRV. (Approval
passes the independence from dropped losers
criterion, but IRV fails it).
Mike Ossipoff
An anonymous coward referring to himself as
Shithead said that Nader really does have
no chance. If that's true, it's only because
of all the people preferring Nader to Gore
won't vote for Nader because they believe that
Nader has no chance.
Do you see the magnificant irony there? (Not
you, annonymous coward; I don't suppose you
would). I have yet to find one person who
is voting for Gore instead of Nader because he/she
likes Gore better than Nader. They always say
they like Nader better, but are voting for Gore
because Nader can't win. Wrong. If everyone
who preferred Nader to Bush voted honestly,
Nader just might win.
In general, we hear everyone express disgust
and dismay about the lousy two choices that we
have. Hey--doesn't it occur to anyone that if
none of us like those "two choices", then maybe
they aren't really the only two choices??
Mike
T. Rat said correctly that if we could specify
all of our preferences, and these preferences had
some impact on the election, then we'd be free
of the old lesser-of-2-evils problem.
True. But the words "..and these preferences had
some impact on the election" are crucial.
In IRV, your voted preference for Gore over Bush
won't ever be counted if Gore gets eliminated
while your traveling vote is still on Nader.
That's why I say that Instant Runoff is
unconstitutional, due to its capricious choice
of whose preferences to count and whose to not
count. It violates the equal protection clause
of the Constitution, because it violates the
principle of one person, one (counted) vote.
Homunculus earlier said that the Supreme Court
vindicated IRV. But when you check his link,
it turns out that it wasn't IRV that they
defined. IRV has eliminations. The method
described in Homunculus's link to something from
the Supreme Court made no mention of elimination.
What they described was Bucklin's method. Bucklin
is a good method, somewhat better than Approval,
though not as good as Condorcet.
Mike Ossipoff
Matt Leese said that it wouldn't do any good to
change our voting system, because we have a 2
party system, with only 2 candidates. Yes?
Ask anyone why they're voting for Gore instead
of Nader. They'll say that of course Nader is
better, but he can't win. So they must vote for
Gore. That's not unusual either, when a
progressive is runnign against the Democrat.
That should show you that the only reason why
we have a 2-party system is because we have a
lousy voting system. Plurality and Instant
Runoffa are the 2 worst voting systems.
Mr. Leese is correct that there's no perfect
voting system, and that some voting systems are
much better than others. For instance, Approval
and Condorcet are much better than Instant
Runoff (IRV). Condorcet's method is defined &
described at the websites:
http://www.electionmethods.org
http://www.barnsdle.demon.co.uk/vote/sing.html
Approval is merely Plurality done right:
Each voter may vote for as many candidates as
he wants to, marking them on their ballot.
Each candidate for whom you vote gets one whole
vote from you. As now, the candidate with the
most votes wins.
[end of Approval definition]
Because Approval makes only that small change
in Plurality, it's the easiest reform to propose,
easiest to explain, and easiest to implement.
What has to be changed to implement Approval?:
Where it now says on the ballot "Vote for 1", it
would instead say "Vote for 1 or more". That's it.
Approval is the only method that can make the
following guarantee: No one ever has incentive
or need to vote another candidate over his/her
favorite, when Approval is used.
That certainly isn't true with Instant Runoff,
where you'll often need to vote a lower choice
compromise in 1st place (dumping your real
favorite) in order to keep someone even worse
from winning. That's the same old lesser-of-2-
evils problem that voters have now.
Condorcet's method, and many other criteria,
are defined at the websites that I listed above.
Mike Ossipoff
Slashdot Mirror
Pareto says: "Never elect someone over whom someone else is voted by everyone."
You claim that Approval violates Independence from Irrelevant Alternatives, because you believe that IIAC is about sincere preferences.
I've never seen a precise definition of IIAC that is about sincere preferences. Can you name a method that would pass such a criterion? Of course we can't test such a claim untill we have a precise definition. Right now all we have is your claim that Approval fails something that you haven't defined, and which, if someone defined it, would surely be failed by all methods.
The only precise definition that I've heard of for IIAC says: Deleting a loser from the ballots and then re-counting those ballots should never change who wins the count.
If you think there's a version of IIAC that Approval fails, then you should give a precise definition of it. And then show me a method that passes it.
You said my worst fear is that in IRV voters will vote tactically and gain the same outcome that would happen in Approval. You got that right.
Do you know what the difference is? It's that with IRV, those voters are forced to dump their favorite, by voting someone else over their favorite. In Approval, no one ever has any incentive to vote someone else over his/her favorite.
Riker showed that, if the voters have complete information about eachother's prefences, then the sincere Condorcet winner will win--no matter what the voting system is. Voters will use some sort of strategy to gain that outcome. It's merely a question of what they have to do. IRV makes them dump their favorite by voting someone else over him/her. Approval will never give anyone reason to do that.
You say Approval doesn't let voters express all their prefernces. No, but it counts those preferences that they consider the important ones to express. Unlike IRV. When Middle gets eliminated before your traveling vote reaches him, then your voted preference for Middle over Worst never gets counted. IRV capriciously decides for you what preferences will be counted. I'd rather decide that for myself, thank you. As we can do in Approval.
Is IRV better than Plurality? I say IRV is worse than Plurality. If candidates' support tapers gradually away from the voter median position, then transfers coming in from the extremes will build up nonmedian candidates enough to eliminate the median candidate. That's a typically expected situation. What could be more normal or typical than that scenario? But in that scenario, IRV does worse than Plurality. IRV will defeat the Condorcet winner, the voter median candidate even if he/she is the Plurality winner--as will often be the case.
IRV, unlike Plurality or Approval, can act contrary to how you vote. You can make someone win by voting him lower, where he'd have otherwise lost. In fact, Professor Steven Brams has shown an example in which some voters cause someone to win by moving him from 1st place to last place, where he'd have lost had they left him in 1st place.
Sorry, but that's nonsense. Voters have a right to expect that the voting system won't act oppositely when they vote one way instead of another.
In fact, IRV also violates this criterion:
Adding to the count some identical ballots that vote X over Y shouldn't change the winner from X to Y.
No, IRV messes up in every way. IRV is worse than Plurality.
Aside from all that, by every criterion that measures for getting rid of the lesser-of-2-evils problem, IRV is no improvement over Plurality, whereas Approval and Condorcet bring big improvement by those criteria.
There are innumerable ways of counting rank ballots. If you don't really do it right, it's a mess, and we're better off without it. Condorcet does it right.
RWE is the same as IRV, but without the eliminations. I'm flexible about its name. As with IRV, the lowest candidate's ballots give a vote to their next candidate, but he keeps his and stays in the election. Of course, even if he's lowest candidate again, that same ballot can't again give to the same next candidate.
Count ends when someone has vote total over 1/2 the number of voters. At that time the canddidate with the most votes wins. Or if no one gets a majority, the count ends when everyone has given a vote to every candidate on their ballot. Again, the one with most votes wins.
Regrettably, RWE is a little wordier to explain than IRV. But it avoids IRV's worst problem-causer: Eliminations.
RWE would be an excellent compromise with the IRVists. Seriously, will you please e-mail Rob Richie and tell him that? You can tell him that I won't criticize RWE proposals.
Mike Ossipoff
I still don't know what was the "Arrow" criterion that Homunculus said that Approval fails.
In my previous posting, I asked Homunc for the name of the criterion to which he refers, and a precise definition of it.
That isn't really a lot to ask. Without an answer on that, Homunc's claim about a criterion failure for Approval means nothing.
How convenient, to make a claim that you can't support, and say that you won't read the thread again. That way you try to avoid responsibility to explain or justify your claim. No, I'm sorry, but an unjustified claim remains unjustified when the person who made the claim runs to evade questions about his claim.
Humunc, when referring to that "criterion" said something about a voting system getting the right answer when everyone agrees. There's one Arrow criterion about everyone agreeing: The Pareto Criterion.
Pareto says that a voting system should never choose an alternative if another altenative is voted over it by everyone.
That's the neares that any actual Arrow criterion comes to what Homunc said.
Approval complies with the Pareto Criterion.
Homunc was being incomprehensibly vague when he alluded to some unspecified criterion that Approval fails. He didn't say anything.
Homunc said that Approval will cause people to go with the flow, and vote for the more publicized candidate (like Gore instead of Nader), where IRV wouldn't.
First, in Approval, you never have to vote for anyone instead of anyone. If you vote for Gore, you can vote for Nader too.
Homunc thinks that, with IRV, you can safely rank sincerely. I thought I'd clarified that a long time ago. If in IRV you vote Nader in 1st place, and Gore in 2nd place, Nader might get enough votes to eliminate Gore. Nader might then lose to Bush. No that isn't contrived or far-fetched. In my previous posting I described two common, ordinary situations where that can happen and where the voter has incentive to vote a lesser-evil over his favorite.
Homunc: I tried to write to you by individual e-mail, but your e-mail address didn' work. I feel it was reasonable to reply to your posted claims via a posting.
Mike Ossipoff
I felt that it's important to reply to the newsgroup, because they heard your comments, and so I'd like them to also hear my repies. I'll also reply by individual e-mail. Two different Australians, who probably don't know eachother, both told me that it's difficult for small parties to get their members to vote sincerely, because those voters feel a need to vote a big-2 lesser-evil in 1st place, to avoid wasting their vote. Your criteria for tactical voting will be commonly met. Say there are 3 candidates, Favorite, Middle, & Worst. You like them in accordance with their names. The voting system is IRV. Say, as will often be the case, they all seem about equal, roughly. That's all we know about the expected numbers. And Middle is closer to Worst than to Favorite, and likewise his voters. That means that Favorite isn't going to win. He can't win by immediate majority, or by transfers. Any point in voting Favorite 1st? No. If you vote Middle 1st, then you just might save him from elimination, preventing Worst from winning. Voting Middle 1st can give a better result than voting sincerely. Voting sincerely can't give a better result than voting Middle 1st. With respect to vote configurations that are consistent with the above assumptions, insincere voting dominates sincere voting in that common, ordinary situation. Different situation: Say there are lots of candidates, and say, as should be typical, candidates' 1st choice support tapers gradually from the voter median point. Eliminations will start from the extremes, and votes will transfer inward. Soon some nonmedian candidates will have enough votes to eliminate the median candidate, even if the median candidate is the favorite of more people than any other candidate. When that happens, a majority are losing their lesser-evil compromise, because they voted sincerely instead of guessing what compromise they need and voting him in 1st place. Both of those common, ordinary situations have IRV making people vote insincerely in order to prevent the election of someone whom they especially dislike. Anyway, in general, when there are a number of candidates, how likely is it that your traveling vote will be on the candidate you need to protect when you need it to be there? If it is, that's a nice coincidence. Look, the whole purpose of rank balloting is so that we can express all our preferences. But how much good does it do to express them and not have them counted? Condorcet reliably counts all the preferences we vote. Approval admittedly doesn't let us express all of our preferences, but it reliably counts every pairwise preference that we express--every pairwise preference that we consider important enough to express at the expense of less important preferences. In Approval, the voter is the one to choose which preferences of his will be counted. In IRV, the IRV count capriciously decides that. I'd rather decide it for myself, thank you. You said that your motivation to strategize is proportional to how well-informed you are. Not at all! How well informed are all the progressives who think they need to vote for Gore instead of Nader in November? One can be terribly uninformed, and vote some terribly uninformed strategy as a result. In Approval, if there's no winnability information available, no frontrunner probability information, then your best strategy is to vote for everyone who is above average in merit. But in our typical situations, you simply vote for the same lesser-evil that you think you need with Plurality--the difference being that you also can vote for everyone whom you like more, including your favorite. You point out that voting is a way to express support. Exactly! That's why it's better to use a method that will never force us to dump our favorite by voting someone else over him. That's another reason why Approval is better than IRV. In Approval there's never any incentive for anyone to vote someone else over their favorite. That can't be said for IRV. In what way is Approval Constitutionally questionable? 1-person-1-vote doesn't mean the voter only gets one vote. It means that all voters have the same ways of voting available to them. Approval, but not IRV, counts the preferences that everyone expresses. Your "Instant Approval" is an excellent mitigation compromise for IRV. The IRVies have rejected it. We offered it to them years ago. Someone devised it and called it Runoffs Without Eliminations (RWE). I strongly urge you to offer RWE to the IRVies of CVD, and to advocate for it. I'd never criticize an RWE proposal. RWE is more than adequaet. You said Approval fails Arrow's criterion about getting the right answer when everyone agrees. Would you please tell me the name of that criterion, and give me a precise definition of it? I'm not aware of it. The 1st "criterion" that I usually find in Arrow articles is one that says the method must be a rank-balloting method. Some of us don't consider that a criterion because it make no requirements about the outcome. That's why I spoke of outcome-criteria. Approval meets all the outcome-criteria. IRV doesn't. Which "criterion" does Approval fail? The one that says the method must be a rank method. I'm more interested in outcome-criteria. What it really amounts to is that the theorem says: If you want any semblence of majority rule, and if you don't want the deletion of a loser from the ballots to change the count's winner, then don't use rank balloting. Picking nits? I don't like it when it's common for people to have reason to regret that they voted sincerely, or to be afraid to. If that's a nit, it's a big one. Mike Ossipoff
The merit difference between single-winner methods are drastic & dramatic. CVD would have us believe that there aren't important differences. CVD is mistaken. I've posted several criteria met by Approval but not by IRV, for instance. Homunculus continues: My reading of the political situation in my state is that IRV is the only one of these with a good chance of being adopted in the next few years, I reply: CVD always claims that IRV is the only alternative single-winner method that would have a chance. They've never supported that claim. Approval is so much simpler than IRV, and so much less drastic a change, that it's incomparably easier to define, propose, enact, & implement. Approval would have a much better chance than IRV. But if IRV is the only alternative single-winner method being promoted in Washington, then _that_ could cause IRV to be the only method likely to win. Maybe CVD has no opposition in Washington. Homunculus continues: so I support IRV. If everyone only supports their own pet version we'll get nowhere. I reply: So then let's all get in line behind me :-)
Wrong. Let's all support what we believe is
adequate, and may the best method win. I and
some other believe, and can show, that IRV is
entirely inadequate. Why should we support that?
3. How did I misquote Arrow?
I shouldn't have saids that you misquoted Arrow.
Every source gives a different version of
"Arrow's" criteria. I've never seen Arrow's
original book or article,and so I of course have
no idea which source is giving Arrow's own
criteria.
In fact, I like your version of Arrow's criteria
better. That's because you list two
outcome-criteria that are failed by IRV, but
passed by Approval. All the other Arrow versions
list outcome-criteria which are all met by
Approval, and one of which is failed by IRV
(Independence from dropped losers, aks
Independence for Irrelevant Alternatives,
abbreviated "IIAC"). With your version, Approval
still passes all the outcome-criteria, but
now IRV fails _two_ of them, not just one.
In additon to IIAC, among your criteria, IRV
fails Participation:
Adding to the count some identical ballots that
vote X over Y should never change the winner from
X to Y.
Homunculus continues:
4. IRV isn't as different from approval as you suggest. Remember, you don't have to rank everyone; if you only "approve" of 2 candidates, only rank those two.
I reply:
But when you need to protect a lesser-evil
from elimination, IRV makes you rank that
lesser-evil in 1st place, over your favorite.
In Approval you never have incentive to vote
anyone over your favorite.
Homunculus continues:
But don't be so factional about this: The problems with approval, IRV, or condorcet arise only in pathological situations, whereas the problems with the current situation are a matter of course.
I reply:
How could you make that claim after I've posted
common, ordinary situations where IRV will
make people vote a less-liked candidate over
their favorite? In fact, with IRV, sincere voting
will be an unstable state of affairs. Sincere
voters will be had by insincere voters, and
sincerity will disappear.
The Australian voters are the people who are
actually using IRV. Do they agree with you
that IRV will never in actual conditions
give incentive to rank a lesser-evil over one's
favorite? No. Australian voters routinely vote
a lesser-evil in 1st place, over their favorite.
So, 1) I've shown that incentive to dump one's
favorite will be common & ordinary in IRV; and
2) Australian experience confirms that prediction.
Homunculus continues:
I like IRV because, of the easily-explicable options, it gives the least emphasis on voting tactics.
I reply:
As I said, IRV will often require insincere
tactics, as borne out in Australia.
Homunculus continues:
I rank nader, bush, gore, as 99, 2, 1 but would "approve" the first two for tactical reasons. Enough people like me and approval voting could easily elect a candidate that the majority dislikes.
I reply:
Approval requires twice as many mistaken compromisers, as compared to IRV & Plurality,
in order to give away an election. That's because
, unlike Plurality & IRV, the compromise doesn't
take the form of voting a lesser-evil over one's
favorite.
And if Nader people vote for Nader & Gore
the 1st time, then Approval will reveal Nader's
true support, and in the next election, people
will know to vote only for Nader, not for the
Democrat. Approval quickly homes in on the
voter median candidate, and stays there.
Approval is the most stable single-winner method.
IRV is one of the least stable, with its
frequent leaps to extremes.
Homunculus continues:
Approval voting also fails in the case of extreme factionalism. If we used approval voting to choose a voting system for the US, and everybody (as you apparently would) voted only for their favorite system because of nitpicking flaws in other systems, we'd effectively be using plurality voting (and we'd probably end up electing plurality voting).
I reply:
If we voted nationally on voting systems, using
Approval for the vote, I'd vote for Approval
and Condorcet, the two adequate methods.
From what you say, you'd vote for IRV, Approval
and Condorcet, and Black's method. What's your
problem with that? Everyone votes for what
they consider adequate, or for what they
consider above-average. The optimal strategy
in Approval, when we don't have winnability
information or frontrunner probabilities, is
to vote for all the above-average candidates or
alternatives.
Hopefully ICANN can take such a vote, after
some discussion.
Mike Ossipoff
Someone called "Drooling Dog" said that, with the voting system known as Instant Runoff (aka Preferential Voting, the Alternative Vote, MPV, Hare, etc.), the winner is at least someone acceptable to a majority. Not so. Consider this: 40: ABC 25: B 35: CBA B gets eliminated and A wins, even though a majority have indicated preference for B over A. There's only one sense in which IRV (Instant Runoff) complies with majority rule: After IRV has eliminated everyone except for the last two candidates, then whichever of those is preferred to the other by a majority wins. That, my friend, is a very poor excuse for majority rule. Someone replied: "But I feel it's the voters, not the method, that eliminated B". Yeah? Did that B>A majority want to eliminate B and elect someone whom they consider worse than B? Drooling Dog also said that with IRV you never have to insincerely uprank someone in order to defeat someone worse. Sure you do. Those CBA voters could save themselves from an A victory only by insincerely voting B in 1st place. I've told in other postings why that will be a common, ordinary need. In other words, IRV doesn't get rid of the lesser-of-2-evils problem. In another posting, I also told why Approval does much better in that regard. Approval is the method in which every voter can mark any 1 or more candidates on the ballot, giving each one marked one whole vote. With Approval no one ever has incentive to vote anyone else over their favorite. That can't be said for any other method. I also defined a weaker lesser-of-2-evils criterion called WDSC, and explained that Approval complies with it and IRV fails it. Condorcet, defined at the websites that I've referred to in other postings, meets WDSC and some similar & stronger criteria. The only single winner reform proposal that's widely promoted is IRV. But there are other ones. Some of them are incomparably better than IRV. Approval & Condorcet are the best. Mike Ossipoff
Sorry to post so many messages in one day. This is my 1st day on the bulletin board, and the only day when I'll post this much. Also, this is my last message for today. Metis said that Instant Runoff (IRV) is the only method that he knows of that doesn't force voters to vote strategically, to express a preference that they don't have in order to avoid having their preference ignored altogether. Actually, IRV _does_ force voters to strategically vote preferences that they don't have. Maybe if you vote Nader in 1st place, he'll eliminate Gore, and then lose to Bush. Bush wins because Gore got eliminated while your traveling vote was on Nader. This will be common. Say that Favorite, Middle, & Worst are known to be roughly equal in 1st choice support. So no one has a majority. Say Middle is closer to Worst, and so Middle will transfer to Worst if Middle is eliminated. Now, as a Favorite voter, do you have any reason to rank sincerely? No. Favorite can't win. He can't have a majority, and noting will transfer to him from Middle. By voting Middle in 1st place, you can protect Middle from immediate elimination, and help him win, keep Worst from winning. Voting Middle 1st can get a better result than voting Favorite 1st, but voting Favorite 1st can't get a better result than voting Middle 1st. Voting Middle in 1st place, therefore, dominates voting Favorite in 1st place. Strategically, voting Middle in 1st place, over Favorite, is all that makes sense, if you're a Favorite voter whose 2nd choice is Middle. So, Metis, IRV forces you to vote an unfelt preference. It forces you to reverse a preference. With Approval, no one ever has any incentive to vote someone else over their favorite. Metis spoke of preferences ignored. IRV ignores your preference for Gore over Bush when it eliminates Gore while your vote is on Nader. That's why I say that IRV is unconstitutional, because, when it violates the one person one (counted) vote principle, it violates the equal protection clause, when it capriciously chooses whose preferences to count & whose to ignore. Mike Ossipoff
I already replied to this message by Homunculus, but I've been told that my messages don't have much chance of getting read when they're "anonymous coward" letters, as my initial reply was, because I hadn't gotten registered yet. So I'd like to post my reply one more time. Sorry about the duplicate posting--this will be my last duplicate. Homunculus needs to read the guidelines for this bulleting board. We're asked not to use abusive language. Since IRV is rejected and disliked by nearly all academics who discuss it, does that mean that the academics are all trolls? Homunculus says that IRV is generally considered better than Plurality. Generally by whom? The marching homunculi that CVD recruits? And isn't it a pitiful testimonial when the best that can be said for IRV is that some people think it's better than Plurality? Homunculus says that the Supreme court endorsed IRV. Wrong. The article that he links to doesn't discuss IRV. IRV does elimination. The article that Homunculus linked to doesn't mention elimination. What they're talking about is Bucklin's method. Bucklin is pretty good. It's slightly better than Approval, though not as good as Condorcet. Let me list 2 websites: http://www.electionmethods.org and http://www.barnsdle.demon.co.uk/vote/sing.html [end of website listings] Those websites define Approval & Condorcet, and define many criteria for evaluating & comparing voting systems. By the way, Homunculus, why are you afraid to use your actual name? Homunculus misquotes Arrow. A brief Arrow summary: If you want even the most minimal majority rule compliance, and if you don't want the deletion of a loser from the ballots to change the count's winner, then don't use rank balloting. Ironically, the Irvies mention Arrow, trying to argue that it means that we can't improve on IRV and must give up & use the abysmally inadequate IRV--when actually, Arrow's results-criteria say that Approval is better than IRV. (Approval passes the independence from dropped losers criterion, but IRV fails it). Mike Ossipoff
An anonymous coward referring to himself as Shithead said that Nader really does have no chance. If that's true, it's only because of all the people preferring Nader to Gore won't vote for Nader because they believe that Nader has no chance. Do you see the magnificant irony there? (Not you, annonymous coward; I don't suppose you would). I have yet to find one person who is voting for Gore instead of Nader because he/she likes Gore better than Nader. They always say they like Nader better, but are voting for Gore because Nader can't win. Wrong. If everyone who preferred Nader to Bush voted honestly, Nader just might win. In general, we hear everyone express disgust and dismay about the lousy two choices that we have. Hey--doesn't it occur to anyone that if none of us like those "two choices", then maybe they aren't really the only two choices?? Mike
T. Rat said correctly that if we could specify all of our preferences, and these preferences had some impact on the election, then we'd be free of the old lesser-of-2-evils problem. True. But the words "..and these preferences had some impact on the election" are crucial. In IRV, your voted preference for Gore over Bush won't ever be counted if Gore gets eliminated while your traveling vote is still on Nader. That's why I say that Instant Runoff is unconstitutional, due to its capricious choice of whose preferences to count and whose to not count. It violates the equal protection clause of the Constitution, because it violates the principle of one person, one (counted) vote. Homunculus earlier said that the Supreme Court vindicated IRV. But when you check his link, it turns out that it wasn't IRV that they defined. IRV has eliminations. The method described in Homunculus's link to something from the Supreme Court made no mention of elimination. What they described was Bucklin's method. Bucklin is a good method, somewhat better than Approval, though not as good as Condorcet. Mike Ossipoff
Matt Leese said that it wouldn't do any good to change our voting system, because we have a 2 party system, with only 2 candidates. Yes? Ask anyone why they're voting for Gore instead of Nader. They'll say that of course Nader is better, but he can't win. So they must vote for Gore. That's not unusual either, when a progressive is runnign against the Democrat. That should show you that the only reason why we have a 2-party system is because we have a lousy voting system. Plurality and Instant Runoffa are the 2 worst voting systems. Mr. Leese is correct that there's no perfect voting system, and that some voting systems are much better than others. For instance, Approval and Condorcet are much better than Instant Runoff (IRV). Condorcet's method is defined & described at the websites: http://www.electionmethods.org http://www.barnsdle.demon.co.uk/vote/sing.html Approval is merely Plurality done right: Each voter may vote for as many candidates as he wants to, marking them on their ballot. Each candidate for whom you vote gets one whole vote from you. As now, the candidate with the most votes wins. [end of Approval definition] Because Approval makes only that small change in Plurality, it's the easiest reform to propose, easiest to explain, and easiest to implement. What has to be changed to implement Approval?: Where it now says on the ballot "Vote for 1", it would instead say "Vote for 1 or more". That's it. Approval is the only method that can make the following guarantee: No one ever has incentive or need to vote another candidate over his/her favorite, when Approval is used. That certainly isn't true with Instant Runoff, where you'll often need to vote a lower choice compromise in 1st place (dumping your real favorite) in order to keep someone even worse from winning. That's the same old lesser-of-2- evils problem that voters have now. Condorcet's method, and many other criteria, are defined at the websites that I listed above. Mike Ossipoff