Asking local IRV proponents to explain the thwarted majority criticism
Clay Shentrup
Jameson Quinn
Abd ul-Rahman Lomax
>An email I just sent out to about five people, plus some people with the CES.
>
>I keep meaning to ask you about this problem you
>cite with Approval Voting where a majority
>favors X, but Y wins due to a lot of people
>voting for their second favorite. Could you (or
>anyone else reading this) briefly explain that
>(in light of the following points which clarify my perspective)?
>
>1) Is this a "fair" criticism? For instance, IRV
>can elect a candidate who was the favorite of
>only 2 out of 1,000,000+ voters, and who would
>lose head-to-head against all but one other candidate.
>
>But you would probably reply that this scenario
>is extremely unlikely. Or even that this (much
>less contrived) one is highly unlikely:
>
>But in that case, why would you pick on Approval
>Voting for a similarly flukey mathematical
>possibility, instead of focusing on overall average behavior?
The answer is obvious. Because we can make
Approval look bad, to someone who has never
considered the issues. It "sounds bad" that there
can be a candidate "preferred by a majority" who
doesn't win. Most people who have never
considered voting systems and how they work --
and fail -- have never thought about this.
Yes, contrived scenarios are used, and the
reality is this: in Approval voting systems, most
voters do not add additional approvals. They will
only do it if their preference is not strong. So
the "preference" being described is not strong,
in reality. And they are merely asserting this,
since the voters haven't shown the preference on
their ballots. If they did, the preferred candidate would win.
If we don't like this problem with Approval,
there is an obvious fix, so obvious that it was
suggested and widely implemented almost a century
ago, Bucklin voting, which is Ranked Approval.
Bucklin satisfies the majority criterion. Now, if
we allow multiple votes in first rank -- which
I'd highly recommend, why not? -- it can still
"technically" fail the Majority Criterion, if
that criterion is based on what the voters
allegedly think rather than what they put on the ballots.
If you look at this carefully, you will see that,
by this alleged application of the criterion, no
voting system will pass the Majority Criterion,
because voters may mark the ballot differently
based on, perhaps, erroneous voting strategy.
Applying a voting system criterion based on silent preferences?
Look, there are scenarios where the possibility
of the Center Squeeze effect will motivate
*knowledgeable* voters to reverse preference, to
vote in first rank for other than their favorite.
So the method fails the "hidden preference"
interpretation of the Majority Criterion.
When Approval Voting was first (recently, it was
already a method with substantial history)
proposed by Brams, it was called "strategy-free."
That's because the definitions of strategy at
that time only considered preference reversal,
not suppression of preference through equal
preference voting. No problem, said those who
favored IRV, they simply changed the definition.
Was Approval Voting truly "strategy-free" as
claim, in the popular meaning of the word? Of
course not. To vote intelligently and effectively
requires, with all systems, some understanding of what is reasonably possible.
Suppose that there is a small town, and there is
a person whom all persons would favor, if on the
ballot. However, for some reason, she hasn't run.
However, if there were write-ins, she would win,
hands down. She is, legally, a candidate. All
voters prefer her, as to hidden preferences. But,
of course, she doesn't win if the voters don't
write in her name (in first preference if it's a
preferential ballot). This is a reductio ad
absurdem of the intepretation of the Majority
Criterion, as applied to Approval voting. It even
violates the Unanimity Criterion.
No, the voting systems criteria must be applied
based on ballots, not on what the voters allegedly think.
And so Approval passes the Majority criterion, as
originally defined, based on ballots. The voters
have a means of expressing exclusive preference,
it's easy. If they do that, the preferred candidate wins. Simple.
The argument will then be made that voters were
"strategically compelled" to add multiple
approvals. This is only a harm, though, if there
is suppressed *strong preference*. And if a
majority of voters prefer X, but add Y because of
fear that someone *else* will win, this is
massive voter ignorance of the real situation, as
they might discern simply by talking with each other!
And it's not Majority Criterion failure, it is a
failure of a different criterion, again
re-interpreted to include Approval, the Favorite
Betrayal Criterion. As originally designed, that
criterion was applied to pure ranked ballots,
where the "Betrayal" was an actual preference
reversal, not mere expression of equal
preference. But if we extend it to include equal
ranking, yes, Approval fails it. It may
strategically require equal ranking in the
presence of a significant preference, under some
circumstances. (Favorite Betrayal is a Criterion
that works with hidden preferences, by
definition. In my view, though, it's cleaner to
not include Approval as an FBC-failing method,
because equal ranking is, in effect, an
expression of a willingness to compromise, and it
means that the real preference is not insuperably
strong. Preference reversal, if strategically
forced, is highly offensive. Preference equality is not.
And, of course, this problem with equal
preference being, sometimes, strategically
forced, is easily solved with ranked approval
methods, like Bucklin, and, as well, with Range,
which drastically reduces the "forcing" involved,
to a level so minor that it can be neglected.
And that "minor" part is what the UnFairVote
people don't want you to see. They want you to
look at "preferred" as if it was black and white,
as if there is no such thing as a minor
preference, one that the voter will easily set aside.
The scenarios they present as examples of bad
behavior are ones where the voters would
*unanimously** approve of the outcome, in the
reasonable scenarios. There could be exceptions,
and the fix for *that* is runoff voting, as well
as increased sophistication in the ballot and
counting. Not IRV, definitely. Bucklin, probably,
with, then, improvements in Bucklin by using a
Range ballot, and by testing for Condorcet failure.
Without going into further detail here, Bucklin
can be made not only Condorcet Criterion
compliant, but also Utility Maximization
compliant, through runoffs when needed. While it
will always be possible to allege that ignorant
voters can screw up an election, there is a
possible voting system that will satisfy *every major criterion* except one.
That's Later No Harm, which is such a highly
questionable criterion that a referee reviewing
the paper that proposed it reported feeling
physcally ill at considering it desirable. IRV is
a system that satisfies it, most don't. The
bottom line: it means that the voter will want to
conceal possible compromises until their
preferred candidate has been taken out back and
shot, is eliminated completely, can't win, so....
maybe *then* they will consider an alternative.
However, even with LNC, it's possible to arrange,
through good ballot design and a runoff system,
that their candidate can still win! Unless the
votes just aren't there. And that's a complex issue.
(If write-ins are possible in a runoff, which, in
my view, is obviously desirable because it gives
voters more power, and power to the voters is
fundamental to democracy, then there is no
elimination, ever. So if the voter, should the
primary be using IRV, perhaps as suggested by
Robert's Rules of Order -- with a majority
requirement -- happens to add a second preference
vote for someone else, who thereby wins the
primary, by the definition of LNH, the voter has
"harmed" the favorite. So not even IRV satisfies
LNH, if used with real runoffs in the case of
majority failure. Voting systems theorists have
almost completely neglected the implications of
runoff elections, confining themselves to
considering top-two runoff in a way completely
divorced from actual usage. Voting systems
theoriest, except for some Approval theorists,
have also completely neglected the most widely
used system in NGOs, repeated ballot until a
candidate gains a majority, with each ballot
being independent (new nominations!). Instead,
they study "runoff" voting, repeated to majority
*with elimination*, and they also think of this
as if it were one election, with no intermediate
process, and with the same voters. The technical
term for this kind of analysis is "sucks.")
>2) Perhaps the fear is, if this
>majority-thwarting happens, voters will repeal
>Approval Voting, then you'll be back to Plurality Voting.
It will never happen, in fact. It's well-known,
with approval, what will happen, instead.
Historically, approval was used in repeated
ballot situations, where voters would, with each
ballot, lower their approval cutoff. They would
start, on the first ballot, with only approving
their favorite. I do not, in fact, advocate
Approval alone, straight, except as a replacement
for IRV or Plurality (which in nonpartisan
elections might as well be the same thing.)
No, what will happen is in the reverse direction,
if there are many candidates. Voters will not add
enough approvals, so there will be majority
failure. This fact was used, with the IEEE, to
dump Approval voting. It was just an excuse,
because *this does no harm* compared with the
alternative that the IEEE favored and went back
to. It's obvious from the history. They
implemented Approval to prevent the spoiler
effect in a particular election with a major
dissident candidate. When the danger was over,
they did not want Approval in place, which would
allow a dissident to participate without causing
the spoiler effect, and thus demonstrate real
support. So they rescinded it, and the argument
they gave was that only a few voters were adding additional approvals.
And that's what will largely happen with
Approval. It doesn't radically change results
from Plurality, except under unusual
circumstances, and the most common of these
represent possible spoiled elections, where a few
percent of voters adding additional approvals can
flip the result to something much fairer.
Multiple majorities in a straight Approval
election will be rare as hen's teeth. If they
occur, it will really mean that the public is
approving more than one candidate, and the
tradition on that is strong, with multiple
conflicting ballot questions: the one with the most approving votes wins.
But Approval, and the technically superior
Bucklin ("Instant Runoff Approval") are better
seen as improved primary methods in a runoff
voting system that requires, at least in the
first ballot, a majority to win. Especially with
Bucklin, we can expect that some voters will
bullet vote, and majorities in the first round of
counting will be almost impossible, again, unless
voters really don't care between two candidates.
There may be majority failure in the whole first
round, because some voters will prefer to wait
for a runoff to make their final choice.
And this all will improve results!
It's odd that this is claimed, if it's being
claimed by the supporters of IRV, because that's
what is actually happening. The oversold IRV,
with the glaring defects of IRV being hidden by
UnFairVote, is possibly damaging the
possibilities of election reform for years, as we
predicted would happen years ago on the Election
Methods mailing list. UnFairVote would not
listen. They are political activists, Richie is a
professional, and he doesn't give a fig about
long-term effects, he might be retired by then.
>Maybe this sounds a little cynical but... how
>would they even know? The vote totals would show
>at least two candidates who are approved by more
>than 50% of the voters, but how would anyone
>know that the second most approved candidate was
>actually favored by a majority of voters? Where
>would the outrage be if this was "invisible"?
Oh, it's possible it could happen. Anything is
possible. "How likely" is the question. It's
*extremely* unlikely, the situation just will not
arise, period. Basically, what is happening is
that quite inventive minds are making up
problems, they are not based on real election results.
Bucklin worked, in many elections. In major
public elections, voters added enough additional
approvals to gain a majority, but generally not
in the first round of counting. Bucklin
encourages additional approvals by giving the
voter's favorite a chance to win on raw
preference. That, by the way, from a social
utility maximization perspective, is a defect,
but it's rather easy to fix. Later on, in primary
elections, majority failure was rather common. I
know of no example of multiple majorities that represented any kind of problem.
Poltical activists do this all the time: make up
arguments based on speculation and what might sound bad.
>In fact, couldn't we argue that the analogous
>problems with IRV (those described above plus
>non-monotonic elections and all that jazz) are
>worse in this regard, because you can actually
>see them (given that the ballot data is made public)?
>
>I think this is a reasonable question, is it not?
It definitely is, not just a reasonable question,
but a fact. Given that it's quite easy to fix the
problem, if we want to go to the complexity of a
ranked ballot -- but without the counting
complexities of IRV -- it really should be moot.
Original Bucklin disallowed overvoting in the
first rank, only one vote was allowed there, so
the method was uncontestably Majority Criterion
compliant. I disapprove of preventing overvoting,
it should be the voter's choice, but voters *who
actually have a significant preference* will not
vote for more than one in first rank. So the
objection becomes totally academic and pendantic.
>3) As one voting geek to another, it's actually
>mathematically proven that the electorate can
>potentially favor Y, even if a majority of its
>members favor X (this was basically the crux of Arrow's Theorem).
><http://www.electology.org/majority-criterion>http://www.electology.org/majority-criterion
Well, that depends on the definition of
"electorate," as distinct from "member." To state
that "the electorate favors Y" requires a social
choice amalgamation method. Arrow disallowed
Approval and Range as "voting systems," from the
start, so his theorem did not apply to them.
>So it's not necessarily wrong to elect someone
>other than the majority-favored candidate
>(albeit, this is total inside baseball that
>99.9% of voters would never care about).
The real problem here is that "majority-favored"
can be a fact, without that preference being
important. Suppose that we have three "voters,"
choosing a pizza. There are three kinds, Pepperoni, Mushroom, Cheese.
Let's state sincere *absolute* ratings for the three voters as this:
Pepperoni Mushroom Cheese
1 90 70 50
2 90 70 50
3 0 80 60
(Voter 3 is Jewish or Muslim, and is practicing,
so the voter really can't eat Pepperoni. The
ratings are not normalized, but represent some
kind of absolute satisfaction index, I just made
it up, but the point will stand. The pepperoni
pizza, we might not, has no nutritional or
pleasure value for voter 3. I suppose the voter
could throw their pizza at the other voters, if
the election is awarded to Pepperoni, so maybe my
0 isn't accurate. Get the picture?)
Majority Criterion requires Pepperoni. So would
the Condorcet Criterion. SU maximization requires Mushroom.
Now, suppose this is a system that requires
majority *ratification*. Suppose the initial
ballot is Range, and the results are that the top
two range winners are presented in a runoff.
Given that the voters will know the results of
the first poll, and may even discuss the reasons
for votes, what will happen in the runoff?
The same happens if there is an approval poll,
and if the approval cutoff is below 70, and then
the result is presented for ratification. But
there is much more data if there is a range poll.
I've seen this, with approval, in a real
organization. The majority had an initial very
strong position, and the majority preference got
maybe 80% approval in an approval poll. But there
was another choice that got something like 98%.
What was done, then, was to present a question,
and the question was, "Shall the 98% approved
result be accepted." The vote was not 98%, it was *unanimous.*
If the voters had been asked, simply, what do you
prefer, a strong majority would have given the 80% answer. Maybe 80%!
Definitely, a majority preference can be other
than what the majority would approve if they have
more information about the preferences of other voters.
It is impossible to *always* tease this out in a
single ballot. To be fully certain takes a series
of ballots in which the voting method can be very
simple, even vote for one, but more sophisticated
methods, like approval, and even better Range,
can make the process far more efficient.
>4) If voters agree that upholding the Majority
>Criterion is more important than maximizing
>their expected utility, then they'll just bullet
>vote, which is equivalent to sincere Plurality
>Voting and hence at least a lot better than
>ordinary tactical Plurality Voting. The nice
>thing about that is that it's the economic
>principle of "revealed preference", which means
>you're getting a more honest (and even more
>thought-out) answer than if you just casually
>ask a lay voter, "do you insist upon a voting
>system which satisfies the Majority Criterion?"
Yes. And that's actually how voters will vote, if
they have a strong preference. If their
preference is weak, they definitely won't care
(or won't care much) if the system doesn't follow it. By definition!
>A great deal of empirical evidence shows that
>most Approval Voting users actually do not
>bullet vote, which would seem to quite clearly
>demonstrate that they think their expected value
>and/or self-expression is more important than
>ensuring they get their way if they happen to be in the majority. E.g.
><http://www.electology.org/pirate-elections-germany>http://www.electology.org/pirate-elections-germany
It depends on the election and context. Bucklin
probably enhances adding approvals, because it
allows the expression of preference as well.
Advanced Bucklin can allow the expression of full
Range. It all works more reliably in a runoff
voting system, which treats majority failure as
"the electorate was not ready to decide."
>Another way of thinking about this is to imagine
>you have a room full of people, and you can
>either give 51% of them €100, and the rest
>nothing — or you can give eververyone €99. A
>majority of the people would prefer the formerr,
>if they knew that they were in the 51%. But if
>they don't know, then 100% of them would prefer
>the latter (to the extent that they understand
>basic economics). This highlights Harsanyi's
>point that the option which maximizes the
>welfare of the group is also the one which any
>randomly selected member of the group would
>prefer, given he were unsure of his identity.
Yeah, but most people will look at that
explanation and be suspicious that the wool is
being pulled over their eyes, even if it's the
absolute truth. It *is* true, I think, without looking up Harsanyi.
But one thing I'll disagree with, the proposition
that most people would prefer the selfish option.
They will only prefer it if they don't see
significant damage to others. Most people, in
fact, care about community welfare, and will
readily give up a small benefit in exchange for a
larger benefit for others. Fortunately. The world
would be quite an ugly place if this were not so.
Driving down the street, I see a car wanting to
pull out of a driveway, but there is a lot of
traffic, so he's stuck. I can slow myself down *a
little* to let him in, which is a major benefit
for him by comparison. So what do I do? It's a
no-brainer for me, I slow down and let the person in.
Some people will think that most people won't let
the driver in. Especially in Boston or New York,
eh? However, something else is operating there,
which is that most people *won't notice* the
driver waiting. They are not deliberately
choosing to be so selfish, they are simply
unaware, because the social environment causes
them to be living in a bubble, unaware of other *people* out there.
Very few groups of three would consciously choose
Pepperoni in the situation described. If the
group did so decide, voter 3 would leave the
group, and would be gaining value, almost
certainly, as a result. That overall social
choices are made that resemble the Pepperoni
choice, sometimes, is due to ignorance, not vicious selfishness.
Hence the value of decision-making systems that
enhance our knowledge of the preferences and
sensibilities of others. I've proposed that using
a Range ballot even if the method is -- anything!
it could be IRV! -- would have a great value in
ultimately maximizing social utility.
That would be opposed by people who benefit from
ignorance, or think they do. In fact, they don't.
That is, itself, ignorance. Winning an election
can result in total annihilation for the winner,
and the winner's family, or even the winner's
entire social class, if the election result was
truly unfair. We don't see that here in the
United States, usually, because we have a broadly
sustained social consensus, or, at least, it's not so blatant.
>Conclusion
>
>What I'm trying to really distinguish here is to
>what extent this majority-thwarting argument is
>about doing the "game-theoretical right thing"
>vs. "doing the politically practical thing". But
>it seems like in both cases, it's not actually a
>problem. Could you at least indulge us before
>you continue to make this argument in the future?
The majority is obviously not thwarted by
Approval Voting, in any significant way. Even
less by ranked approval methods or Range. The
majority, in the examples that they cite, got
what it explicitly approved. You want an example
of "majority thwarted," look at Burlington's
Center Squeeze election, and that had the real
effect that people are fearing for Approval.
There, a majority of voters voted *against* the
winner, by voting only for other candidates, and
that's actually normal for IRV in many elections.
There, the method is so poor, there are voters
who would likely have gotten a better result by
staying home and not voting. And they can tell that from the votes.
In any case, the objection is purely pedantic,
highly unlikely to actually happen. And if it
does happen, it is highly likely to be considered
fair. And you can make that totally explicity by
requiring a runoff with multiple majorities, as
one might require with majority failure.
The question, though, will be, is it worth that special election?
One fix is to elect through a primary that is off
the general election, preceding it by a few
weeks. Runoffs, then, are requesting decision by
a generally wider electorate. It works, but I'm not sure I'm in favor of it.
We won't know of the possible dimensions of the
"multiple majority effect" until there are real
elections where it's a possibility. It is highly
unlikely, from Bucklin history, to be a problem.
You can make Bucklin Majority Criterion compliant
by forbidding overvoting in the first rank.
But ... that is a *theoretical improvement* while
actually harming voters! -- i.e., by discarding
their ballots. With Bucklin, because of the
ranking, voters can express and distinguish
between true equal approval and multiple
approvals in the presence of a significant difference.
I'm sure you know all this, Clay, I'm just rehearsing the playlist.
The primary issues I see with Approval, plain old
Count All the Votes, are this:
1. Approval is an obvious improvement over Plurality.
2. Approval allows voters for no-hope candidates
to still support their favorite, *and*
participate in the general election. These voters
will see Approval as a clear improvement.
3. In partisan elections, in most places, most
people will bullet vote, because they *actually
prefer* a front-runner. They will not care about
Approval, as to their own vote, unless they
decide to make some expression with a write-in,
probably, or some of them might vote for a minor
candidate if they want to encourage that
candidate. If enough do this, of course, that
candidate could win. Surprise! But that's a
choice the voters make. I don't recommend voting
that way, i.e., you prefer a major candidate, but
vote for a minor one to *encourage some
direction*. Instead, support Bucklin or Range,
which would, more safely, allow you to do that.
4. Count All the Votes is just plain fair. There
is no sound reason to discard overvotes, and
there has been obvious, and famous, harm.
5. Approval and the ranked version, Bucklin, are
even better as a primary method in a real runoff
system, and where such a system exists, that's
the reform to push, not IRV, which demolishes the
benefit from top two runoff while pretending to
emulate it. A good primary method will avoid most
runoffs, thus reducing average election costs,
and it's possible to tweak the rules to maximize
this. For example, it should be possible if the
original ballot collects enough information, to
predict with reasonable accuracy the result of a
runoff, even if the explicit approvals don't show
a majority. Some top two runoff rules already do
this, only requiring a runoff if the *plurality*
is not above a certain level. That can be done
with much more sophistication and accuracy, so
that runoffs are only required when the certainty
is low. In order to intelligently do this, one
must have substantial experience with real runoffs!
6. Most election experts, I'll claim, consider
Approval a better single-ballot method than IRV.
But it's a false dichotomy. Straight-on
comparison should be between Bucklin and IRV
(same ballot, essentially!). Straight approval
should be compared with Plurality, where it's an
obvious improvement, or does no harm. And top-two
runoff, which seeks a true majority, should then
be compared with methods that leave the runoff in
place as a possibility, but that seek to
determine if it is really necessary, or if a
majority choice can be reasonably inferred from the ballot. Or known.
7. Thus improving top two runoff is a matter of
using more expressive ballots in the primary (and
maybe in the runoff as well), which can avoid
unnecessary runoffs. Robert's Rules of Order
considers repeated elections to have benefits of
their own, by allowing the voters a closer look,
when they were not ready to make a collective
decision in the first ballot, and then they can
make choices in the next ballot based on what
they know from the first, and additional campaign arguments.
8. Top Two Runoff is the most widely implemented
election reform, it was hailed as such for years
by political scientists. It should not be
discarded, it should be improved. IRV is not an
improvement, it's a serious step back from TTR,
as can be shown, easily, from a comparison of
results. The only down side of TTR is cost, and
almost all that cost can be ameliorated. With
IRV, the cost of runoffs is replaced by serious
canvassing costs, the cost savings, if real, are
not nearly as great as alleged. The real cost
savings of Approval and Bucklin as TTR primary
methods, are an obvious improvement over raw TTR
with vote-for-one in the primary and simple top-two elimination for a runoff.
9. Approval (and Bucklin) add very little cost,
they are both Count All the Votes methods. The
increased costs of counting *a few more votes*
(Approval) or ranked votes (Bucklin) -- where
maybe a third of voters will add additional
approvals, it depends -- are small. It can also
be conditional, i.e., with Bucklin, lower ranked
votes were, historically, only counted when a
majority had not been found in the higher rank,
so the election really started out with
vote-for-one. Approval methods, including Bucklin
and Range, are Count All the Votes methods, with
each precinct being able to report the full
ballot sums without any dependence on what other
precincts have counted. IRV makes it all be
linked, so that an early error propagates through
all precincts, and, when found, all other
precincts may have to recount. That's why IRV canvassing can be such a mess.
10. Plain Approval has an obvious defect which
will cause a lowering of additional approvals by
voters, the inability to multiple approve while
still expressing a preference. Bucklin fixes
this, easily. The likely effect of this problem
with Approval, though, is simply to increase the
likelihood of majority failure. It does not harm
results over Plurality, it simply moves them
toward Plurality. This defect will not be a
problem for most voters in 2-party partisan
elections, because these voters, by definition,
generally prefer one of the major candidates (if
the 2-party system is working); voters will only
add additional approvals if they actually prefer
another candidate. And if they actually prefer
another candidate, they will clearly see Approval
as an improvement over Plurality, since it allows
them more (*much more!*) freedom of expression.
And they'll be deleriously happy with Bucklin,
which lets them have their cake and participate in the major election as well.
Remember, Bucklin was *very* popular with voters,
as was shown when the Minnesota Supreme Court
shot it down. Read the request for
reconsideration! I don't recall any situation
where Bucklin, for public elections, was
eliminated by the public, whereas IRV has been
eliminated that way, quite a few times. Not
always fairly, by the way, that's a separate
issue. Bucklin, ranked choice approval, was easy
to vote, easy to understand, and highly effective. What's not to like?
Plenty. Bucklin, we must remember, in the first
Bucklin election, elected a SOCIALIST!!!! Grand
Junction, Colorado. That this was a nonpartisan
election, and that the Socialist did not
implement a Socialist agenda, apparently, but
just served the town, didn't seem to matter to
the people who argued against it on that basis,
this is in the historical record. They lost, when
this was presented to the public, probably
because the public wasn't quite so disturbed, but
it may well have counted in the behind-the-scenes
activity that led to San Francisco dropping
Bucklin, with the California Supreme Court not
caring that the will of the voters was being
thwarted, by a blatant technical maneuver, the
buying of a single voting machine, with no
intention to actually use such machines widely.
Same thing as in Minnesota, the Supreme Court
there actually reversed an election, awarding it
to the candidate with a plurality in the first
round, breaking tons of precedents. They did not
care that the will of the voters was plain from
the ballots, that the Bucklin result was
obviously fairer, and they also did not care that
if the election had been plurality, without the
additional ranking possible, voters might well
have voted differently, there might have been a
different candidate set, etc. (Advanced voting
systems encourage increased numbers of
candiates.) From legal principles, if they were
going to invalidate the method because of a true
failure, they'd have still accepted the result as
the most just of alternatives, or, alternatively,
they'd have required a new election. Instead,
they obviously wanted to drive a stake through
the reform, and not just for Duluth, they wanted
to have a wider impact. Don't try to use an
advanced voting system, we'll get you if you do.
What I find fascinating is that the public did
accept this, i.e., did not organize to defeat or
bypass those courts. Perhaps the impending Great Depression had an effect.
Clay Shentrup
Dale Sheldon-Hess
> Not one of the recipients has responded. Well, except for Richard Winger. He
> very politely complimented me, saying it was one of my best pieces yet or
> something like that. But no one has responded to the actual substance of the
> email.
Winger is slowly coming around. His blog (Ballot Access News,
ballot-access.org) had been consistently pro-irv and dismissive of
approval since I began following it two years ago, but in the last few
months, it seems essentially neutral toward both.
Minds are changing. Slowly, but changing.
--
Dale Sheldon-Hess
Clay Shentrup
Winger is slowly coming around.