At 04:30 PM 3/12/2012, MIKE OSSIPOFF wrote:
Abd:
Sure I do. There are some scenarios that can be asserted that can
lead to a conclusion that if overvoting is allowed in first rank,
Overvoting is equal-ranking, correct?
It's an incautious usage, because "overvoting" that is allowed is not
"overvoting." But the idea is to dump the overvoting rules, i.e., to
"allow overvoting," that is, voting, at the same rank (which could be
the only voting rank, i.e., "Plurality,"), for more candidates than
winners allowed winners.
Bring up something. "One person, one vote" is obviously neglected in
plurality-at-large, "vote for N." It's an artificial interpretation
to make one person, one vote a reason to not allow
"overvoting." What does the number of winners have to do with the
number of votes? The number of winners has to do with analysis of the
ballots, not with the voting, per se. With multiple conflicting
ballot questions, which is logically equivalent to candidate
elections, voters may vote on all the questions, or in none, or on
anything in between.
Further, multiple conflicting ballot questions are completely
equivalent, in standard practice, to approval voting as the "primary"
in repeated election, standard democratic process, which is
compromised into runoff voting in public elections.
That is, a question must gain a majority to pass. If no question gets
that majority, then the "election" is void and, in order to pass a
question on the issue a new process is needed. If one, done. If two
pass, then the one with the most Yes votes prevails.
That is, the principles of approval voting were understood when those
rules were developed.
There is a procedure used under Robert's Rules for "filling in the
blanks." It can be (and sometimes is) used for candidate elections,
but it is used for any question where there is some variable to be
determined. The voice procedure is known to have a problem, because
election can depend on which of multiply acceptable options is
presented for vote first. However, fill in the blanks can have
unlimited "successful" answers, i.e., approved by a majority.
So if the question is, "______ is nominated for the office of Chair."
-- instead of "elected to" -- there is, in effect, a primary, and all
successful candiates -- more than a majority -- are presented in a
runoff, which is the real election. It's a *nomination* procedure.
After nominations, it is standard procedure in Robert's Rules for the
sole candidate to be elected by "acclamation," i.e., without
objection. This is not far from determining the election of a sole
majority-approved candidate elected in a top two runoff election.
This procedure is obviously vulnerable to center squeeze.
However, the sole issue when presenting Count All the Votes, as an
alternative to any existing method, is whether or not it improves
performance overall, and at what cost.
You continued:
voters will use it strategically. Otherwise, from what we know about
Approval Voting, and from the history of Bucklin in certian
elections, I *predict* that most voters won't use them. Mike, do you
have sufficient information to show that this is unlikely to be true?
Certainly, at least as regards Approval: Right now, many or nearly
all progressives,
people who want policies more progressive, humane, or innovative
than those of the Democrats, insist
on "pragmatically" holding their noses and voting for the Democrat
in Plurality. So,
what will they do in Approval? They'll continue voting for the
Democrat, but will also
vote for everyone who is better than the Democrat. They'll vote for
at least two candidates.
Nader and Gore, for instance, if they prefer Nader, but feel a need
to vote for Gore as a
lesser-evil.
This is absolutely correct as to Count All the Votes in approval, and
it is expected. It is not expected with Bucklin, which was the
discussion, where the voter can multiply-approve in ranks. The
questino is whether or not they will multiply-approve in first rank.
Original Bucklin prohibited that, but the issue here is whether or
not the method is improved by allowing it.
Most voters, however, by definition, in most elections, favor a
front-runner. Hence most voters will be rather unlikely to approve a
not-front-runner. Some may (some did with Bucklin, I think, but we
don't have good ballot data), and it's not only harmless, it might be
a good thing.
As for ABucklin (ER-Bucklin), no one can say for sure. It's my
perception that often
one's best strategy will be to only vote for a set of candidates at
first rank position.
When there are completely unacceptable candidates who could win,
one's best strategy is to
top-rank all of the acceptables and not rank anyone else.
I agree that nobody can say for sure. It's not been tested. "Best
strategy" depends on preference strength. When preference strength is
low, voters may well elect to top-rank more than one. Bucklin, in
truly contested elections, frequently collapsed all the ranks into
pure Approval voting, and voters should be aware of that. And even
then majority failure sometimes took place. (Bucklin was oversold,
just like IRV more recently. But Bucklin, under similar conditions,
probably is more efficient at finding majorities in real elections,
because it counts more of the votes, and it often did this by
creating a "comeback election.") The claim that the best strategy is
voting all acceptable in top rank seems to neglect preference
strength among all "acceptable" candidates.
Note that this argument is the opposite of the UnFairVote argument
that voters will bullet vote, out of fear of "harming" their
favorite. That, of course, assumes high preference strength, in which
case the bullet vote is fully sincere.
Runoff voting with Bucklin ballots and analysis, however, may address
majority failure, and once that is in place or being considered, we
might consider far more sophistication in analysis, by using a Range
ballot, combined with Condorcet analysis. The goal would be to place,
in a runoff, when one is necessary to find majority approval, all
reasonable possible winners. Range ballots provide possible input
information for all voting methods. (And it's even possible to use a
Range ballot, which permits equal ranking, as imput for Borda, by
dithering. I.e., ballots with "overvotes" would be dithered into N
counts, following all the possible permutations of order. But why one
would want to do this, I don't know. Borda-ER is Range, properly
implemented (i.e., when there are "overvotes" there are corresponding blanks.)
(And this exposes that the corruption of claiming that Borda is
superior to Range is based on believing that the method is improved
by *forcing* voters to rank, even though they have no preference.
That's the same as requiring mandatory voting and full ranking of all
candidates in IRV ballots, as is done in Australia. The idea that
voting system performance is improved by forcing voters to make
choices has not been uncommon in voting system history. It's
anti-democratic, in fact. One of the reasons plurality works as well
as it does is that voters who don't care don't vote! -- mostly.)
When the fear about failing to elect one of the better candidates
isn't so great, one might
want to distinguish between some of them by ranking them at several
rank-levels, though that
increases the risk that someone else will win.
Right. But by definition, then, the voter isn't greatly concerned
about that possibility. Nader voters, in Bucklin, will almost
entirely rank Nader first, and may rank, in Bucklin, Gore at second
or third rank. Whether this is truly sincere or is strategic depends
on their fear that Bush will win by a majority in the second round.
No voting system can avoid all forms of "strategic voting," where
voters vote other than how they would vote with zero-knowledge. This
is the implication of Dhillon and Mertens' work with what they called
Rational Utilitarianism, i.e, Range voting combined with the
adjustment of "fully sincere normalized Range votes" by the
probability of relevance of the vote. In that concept, preference
strength in an "irrelevant" pairwise election, were the irrelevance
relatively certain, is assigned no voting power by the voter.
(We can think of Range as a series of pairwise elections, where the
voter may cast fractional votes in each election, and the sum of
these votes is one full vote, a system constraint equivalent to
normalization. In order to vote with maximum effectiveness, then, per
game theory, the "incremental" preferences, stepping up the list of
candidates in order of preference, are valued according to expected
probabilities. It all sounds complicated, but we do this
instinctively with choices all the time. We put little or no effort
or money (ie., power) into choices we consider impossible of realization.)
But sometimes
maybe not. Imagine an Approval election in which the ABucklin option
is allowed. If someone you don't
like has an early majority, maybe largely from Approval ballots,
then you're out of luck.
Yup. On the other hand, if the ballot can be interpreted as a Range
ballot, and if that shows what I'll call Range Reversal, it is
possible that this early-majority candidate won't be declared
elected. There could then be a runoff between a Range winner and,
probably, this candidate. Range Reversal takes place when raw
preference order of a majority is reversed by the preference strength
of a minority.
Many voting systems theorists are disconcerted by the idea of
presenting that choice to the voters in a runoff, because they assume
that the voters will have the same preferences. But in real life, the
majority often gives up its first personal preference in favor of
wider acceptability of result. The outcome cannot be predicted from
the primary votes. The outcome depends on *absolute preference
strengths.* Those are not easily collected on a ballot, unless there
is a cost to voting, like a Clarke tax, which is not likely to be on
the table soon, I'd say.
However, the cost of voting in a runoff election is a kind of Clarke
tax. We know that when voters are highly motivated, they turn out in
droves for runoffs (which has happened when Center squeeze in top-two
runoff presented Poor Choice vs. Truly Awful Choice, not to mention
Good Choice vs. the latter, where the strength would be even
greater.) The reverse effect is likely the case when both runoff
candidates are considered acceptable by most. This is the
implication: voters with greater preference strength are more likely
to turn out in a runoff election, so if the weak preference of a
majority is being outweighed, for determining a Range winner, by the
strong preference of a minority, then there is an increased
possibility of that Range candidate to win the runoff.
There are *lots* of reasons why runoff elections are sometimes
desirable. It's not just a question of static preferences, because
preferences may shift in the runoff campaign, as voters are
presumably now looking at a reduced set of candidates and learning
more about them specifically.
Candidates
in your ranking who haven't yet received your Abucklin votes lose
because you've missed your chance
to help them (as you could have if you'd top-ranked them).
Yes. But it's actually unlikely. Bucklin is apparently much more
likely to still allow majority failure, than to find multiple
majorities in the first and second ranks. Voters *do* bullet vote,
whether it's advantageous to them or not.
Looking at ABucklin as an option in an Approval
election, that vote-management option doesn't look like necessarily
always a good idea. That suggests
that, in an ABucklin elecion, voting Approval-style might often be
the best way to vote. But that's just
my subjective impression.
I think it's inaccurate. Voters who have a weak preference, where it
is difficult for them to determine preference order, *will* multiply
approve in the relevant rank. Voters with a strong preference will not.
Remember, if they strongly prefer A to B, but because they fear C
will win, they vote in first rank for A and B, they risk regret if B
wins in first rank, because of their votes. Nevertheless, if A truly
has no chance of winning, period, this isn't a loss for them. From
the Dhillon-Mertens analysis, they will rationally not put voting
strength into the A/B pairwise election.
However, given the historical performance of Bucklin, they have
little expected cost from ranking their clear favorite at top rank,
and then the compromise candidate below that. Remember, if this
election is such that C could win by a majority in first rank, there
is probably no way for the voter to avoid that win, realistically.
The hope is to create multiple majorities in first rank for C and B.
The cost is Failure to Prefer the Favorite (this is not FBC). That is
a cost that voters will consider, it is real for them, even when it
makes no sense from game theory.
I don't see this possible strategic thinking to have much real impact
on voters.
Many voters only care about voting for their favorite, no matter what
system you give them, unless you *force* them to add additional
preferences.
But not many progressives, regrettably. Nearly all progressives
refuse to vote for
their favorite, voting instead for a "lesser-evil".
"Progressives" may or may not do this, it depends. However, advanced
voting systems, generally, allow them a more palatable choice.
There's no reason to believe that all those people will stop voting
for a lesser-evil
in the 1st Approval election. But they'll be able to also vote for
everyone who is better.
Yes. If we get Count All the Votes, it will not escape notice.
Bucklin allows them even more expressivity. Bucklin/Range would allow
them maximal expressivity (at least on the ballot!)
But yes, after the 1st Approval election, when the reported vote
totals show that a progressive candidate can beat a Republican,
hopefully many or most of those voters will stop voting for the
Democrat, and will only vote
for their genuine favorite(s).
Not at first. That will happen when the previously no-hope candidate
comes within reach of winning. There is a transitional period where
there are three major candidates, and this is what tests advanced
voting systems. It's where IRV can fall flat on its face.
The bullet voting from progressives will quite likely arise if it is
full Approval, as the perceived probability of a progressive win
increases. But if voting according to this causes the Republican to
win? (As it might)? Fingers will be burnt. No, the retreat from
mutiple approvals, Democrat and Progressive will more sensibly
continue until the real contest is between the Democrat and the
Progressive, with the Republican being no-hope.
All this means that there are very good reasons to go to Bucklin or
some other more advanced system, which allows voters to "plump" for
their favorite in the first round, and then add additional approvals
if the contest is still live. Real Bucklin voters sometimes
suppressed the second rank, compromising only in the third rank,
which makes perfect strategic sense. And if there is a runoff
possible, there will be even further suppression of additional
approvals. That's testing the sincere preference strengths, with an
associated cost: the risk and inconvenience and expense of a runoff election.
Voters will balance these things, in a much more sophisticated way,
I'm sensing, than we often expect.
Maybe for their one most favorite candidate. Maybe for several
best candidates who are all significantly better than the others.
All of the Approval strategies
that we've discussed here amount to voting for every candidate who
is better than your expectation
for the election.
Yup. Pretty simple way of putting it. You will discriminate between
the candidates according to your real preference strengths and your
assessment of relevance of that discrimination.
From conversations with Democrat-voters, it's my opinion that,
among those who have actually looked at or listened to candidates' and parties'
policy proposals, no one considers the Democrat their favorite. I
don't think that the Democrats
have any serious favorite-voters. They're only lesser-evils. Their
genuine support doesn't exist. With
the enactment of Approval, those fictitious hollow-men known as
Democrats will cease to appear to exist.
Well, that's a political opinion. I'm not going there. With better
primary systems, "Democratic candidates" will improve, perhaps.
If we go to Asset, the whole party system becomes a fish bicycle.
Indeed, that was the thinking behind Carroll's invention
of Asset Voting.
>Those were only municipal elections, of course. You can't use
them to predict
>voting in national or state elections. In important elections,
>people would soon
>learn what voting strategy is in their best interest.
Let's start with small scale elections, eh? First of all, there are
as yet no national elections in the U.S.
I just mean elections for national office. So I include elections
for presidents and congress-members
(including senators) as national elections. Of course I should say
"national-office elections" instead
of "national elections".
While state elections create certain offices that function on a
national level, all are actually state elections, electing an office
that is only elected for that state, with no participation from
voters in other states. Presidential elections might seem to be an
exception, but, of course, they are not. Those elections are electing
electors, per the state rules (not national rules, which only
determine the number of electors assigned to each state).
Innovation is easier to achieve locally, but it takes a while for
local innovation to filter up to
the national-office level. It would be nice if, somehow, Approval
could be enacted for state or national
offices without going through the long local-offices-first process.
It's not only unlikely, it may not even be advisable. We may have all
the bright ideas in the world, but they are not necessarily true
improvements, and they can even be disastrous. How about, before
betting the farm, we *test* the ideas?
I like ABucklin, and it seems to me that the best route to it, the
natural route, is as an option
in Approval elections. As I often say, it's difficult to oppose or
criticize an option. After all, how
someone uses their Approval vote will be understood to be their
business. So how could anyone object to
a vote-management option such as ABucklin, in Approval elections?
Agreed. It allows the voters more freedom of expression.
Bucklin with a Range ballot obviously does that even more. And it's
possible to interpret that Range ballot in many different ways. It
can be interpreted purely with the Bucklin method. The stepping down
of approval cutoff can be extended below whatever rating is
considered "approval," it could go all the way to the bottom. I
prefer, however, from democratic principles, to not elect a candidate
without explicit majority approval of the result, as a minimum.
That's where runoffs come in.
But I feel that the real improvement on Approval is getting rid of
the co-operation/defection problem. That's
accomplished by AOC, GMAT and MMT. Options such as those, for an
Approval election, (they aren't mutually compatible as options in the same
election) are therefore the ones that I'd suggest first. Later I'd
suggest ACBucklin or AOCBucklin (wherein a voter could optionally make
any non-top listing of a candidate conditional).
Traditional Bucklin, with Count All the Votes, would seem to do the
trick, and the ballot is simple and easy to understand. Ranked
Approval is another way to put it. Instant Runoff Approval. If there
is a majority requirement, it's real Runoff Approval, with instant stages.
The chicken dilemma is often presented as a reason why this will
fail. But to play that game requires fairly high preference strength.
Most voters don't have that. In the classic Burr Dilemma, we had
electors, at a time when these where highly involved politicians, not
ordinary voters, and they were voting publicly, I think. It's quite a
different situation.
In ordinary non-conditional ABucklin, of course the C/D problem
could be dealt with in the various ways we've
discussed for ordinary Approval and RV. So ordinary ABucklin isn't
without merit. It's just that I personally
feel that, as Approval-election vote-management options, AOC, GMAT
or MMT offer a more important kind of improvement over ordinary
Approval.
I'm much more concerned with getting *any kind* of approval, as well
as in preserving the advantages of seeking a real majority and of
runoffs when needed. I am not attached to the specific rules of
primary and runoff, but I would like to see reform follow a
reasonable upgrade path. I see that path moving toward including
range polling in the method.
My sense is that a quite sophisticated method can be built using
range polling, including explicit approval cutoff, that is easy to
vote with reasonable strategy, and that incorporates a seeking of
majority approval for a winner.
We must notice, however, that any voting system that handles the
spoiler effect will encourage additional candidacies, thus stressing
the system. To my my mind, the only way to *ultimately* address this
is with Asset methods, which allow, in effect, unlimited runoff
polling. Single-winner elections should probably be handled with
ordinarily deliberative process in a fully or highly representative Assembly.
, the largest jurisdiction to
hold an election is a state. What we think of as presidential
elections are actually local elections of pledged electors.
[endquote]
An uncontroversial tacit agreement in these discussion is that the
president should be elected
by a direct national election, dispensing with the electoral college.
I don't agree with that, in fact. Absent a constitutional amendment,
I'd prefer approaches that fix the electoral college, returning it to
its original function. That's quite possibly not attainable with the
very strong two-party system. Asset at lower levels will very likely
eliminate the necessity of party endorsement for election. Once that
is in place, reforming the national level will come within reach.
There are methods to force this if a substantial number of states desire it.
Direct national election is *intrinsically* vulnerable to
manipulation of major media. The electoral college was designed to
allow a national election without requiring national camplaigns. This
was corrupted by the influence of national parties, leading to a
situation where a majority party in each state would be shooting the
national party in the foot by moving to fair representation. Hence
any reform must be designed to sidestep this, to not require a party
to act against its own interests.
The real electoral college problem is the all-or-nothing system that
became almost universal. That is what causes the electoral college to
sometimes be drastically unrepresentative. In Florida, the obviously
fair result of that election would have been a division of electors.
Such a division would have awarded the election to Gore. It would not
have been close. It was only close if every factor fell on the Bush
side, which is what happened, ending with the unprecedented
interference of the U.S. Supreme Court.
Otherwise the electoral college system was *brilliant.* The founders
simply failed to agree upon a fair way of choosing electors, so they
decided, to complete the constitutional convention, to leave it to
the states, thus setting up conditions where the majority in each
state would act for its own benefit, leading to the all-or-nothing system.
Of course it could be reasonably argued that parliamentary
government would be better, and I have no quarrel with that.
But proposing a better way to elect the president is much more
modest than proposing the drastic change from presidential
to parliamentary system.
We get way ahead of ourselves by considering national election
reform. We can't even get good voting systems in dogcatcher
elections. When we can do that, maybe we can look at wider issues.
Good voting systems are of general application, both in and out of government.
Consider this: some national NGOs hold elections by mail ballot.
Asset for such organizations could vastly improve the reliability.
Short of Asset, but following the spirit of Robert's rules,
preferential ballot can be used to seek a majority in one round, with
repeated rounds where there is majority failure. Robert's Rules
*never* recommends electing by a plurality, but compromising after
two rounds is better than compromising within one, where voters
really don't have enough information, often, to make
utility-maximizing choices.
You continued:
There is a lot of crap out there on what strategy is in the voter's
best interest
[endquote]
...like the crap that says we should vote for a lesser-evil in Plurality.
That's simple game theory, in circumstances where a vote different
from that is irrelevant. It's not crap. There is a lost performative
here, Mike. "crap" is a judgment. (Like "lesser evil.") If it's an
isolated judgment, if the voter knows that, then not voting for a
"lesser evil" is simply not taking responsibility for using the power
one has to lessen evil.
If it's truly *evil,* sure. Maybe it's better to flee.
By the time an election is being held, the point where a voter's
judgment that both major candidates are "evil" could lead to choices
other than supporting the lesser-evil, has past. The present system
requires action *before* the election, a whole lot of it; that's
where the choices to be presented on the ballot are determined.
As part of this pre-election activity, we should reform the voting
system so that we can sincerely express our preferences, with little
or no harm to the excercise of voting power to mitigate evil (and
promote good).
Maybe sometimes vote
for a compromise, the most winnable acceptable candidate, or a
candidate agreed-upon by a similar-believing
large set of voters. But never vote for an evil, even if a lesser one.
That depends on the definition of evil, doesn't it? Set the bar too
high for that, define people who do represent a major chunk of
society, not to mention a majority, as evil, you have cut yourself
off from humanity. That's actually an evil in itself.
After all, if a candidate is evil, and might win, and spread his
evil, then killing him could become morally justified, to prevent
harm. This is the province of obsession and insanity, it's a form of
paranoia, that readily identifies the other as evil.
You continued:
, and there is a large class of voters who vote for what
is in the society's best interest (in their opinion, of course, but
these voters will value *consensus* and will recognize that getting
their own preference is not necessarily best for the society).
[endquote]
Maybe some will vote that way in RV. I don't know. But surely most people feel
that their favorite candidate would be best for society, and that
has a lot to do with
why he's their favorite.
It's a tautology, as you have set it up. However, that opinion may
not be strong at all, and the person may be quite willing to accept
that the choice of someone else is reasonable, and some healthy
people actually know that their own preferences are not necessarily
real, i.e., accurate representations of true benefit to society, or
even to themselves.
I might even think that Mr. Do-Good is absolutely the best candidate,
except for one problem. Too many people feel otherwise, and he would
not be able to effectively govern. Perhaps it would cause a
revolution, the worst kind, where the forces are more or less evenly
balanced, where no side can quickly win, avoiding the massive harm of
long term civil war.
Might some know that their candidate is bad for society, but good for
their own private special interest?
Few think that way, my opinion. It might actually be the case, and
this might be influencing their decisions, but most people could not
sustain this opinion, only sociopaths could handle that. It runs
contrary to normal human social instincts. In order to sustain a
level of activity harmful to society, the harm of it must be
rationalized away; if the harm becomes truly obvious, most will
abandon it, or, if attachment to belief is too strong, it will fade with time.
Maybe, but probably most people have convinced themselves
that he's best for society in some meaningful sense.
Yes, though I don't know about "most people." Most people, in the
U.S., don't vote. Give them Asset, my sense is that far more would
vote, because they would then have a measurable and visible effect on
the election. (as long as a full-representation system is used,
meaning that an adequate number of seats in an assembly or electoral
college are being elected, and that there are fair ways for the
"dregs" to be represented. If that is done, *no vote is wasted.*)
However, also, "how much the best" is always a question. I know that
many times I've voted in elections with no truly clear idea of who
was best, in a particular race. I made a choice, because that is what
the system required. Sometimes I abstain, but if I can figure out
some preference, I may vote that way. This is weak preference
strength, while I might at the same time strongly prefer that another
candidate *not* be elected. Count All the Votes allows me to express
this, meaningfully.
Range allows me to more accurately express real preference strength.
Bucklin can use a Range ballot, the Bucklin method easily adapts to
it. And then it becomes possible to test for Range winners, even
before that's the actual method.
Step by step implementation, each step considered an improvement over
the prior steps, with little cost.
Whether the existing system is Plurality or Top Two Runoff, Count all
the Votes improves it. Then a more detailed ballot improves it, i.e,
Bucklin. Same with Plurality and Top Two Runoff. Bucklin has always
been considered a ranked method, but because A>B is treated
differently than A>.>B, it's really more of a Range ballot. (And the
allowance of overvoting at a rank is also range-like, and, Mike, this
*was* part of traditional Bucklin (as to lower ranks than first).
Some versions of Bucklin used fractional votes for the lower ranks as
well. Range, again.
----
Election-Methods mailing list - see http://electorama.com/em for list info
