[EM] New MN court affidavits etc. (correction)

[Abd ul-Rahman Lomax]

At 07:11 AM 11/8/2008, Chris Benham wrote:
>Are you really comfortable supporting and supplying ammunition to a
>group of avowed FPP supporters in their effort to have IRV declared
>unconstitutional?

So many aspects, so little time ....

(1) Brown v. Smallwood outlawed preferential voting, period, in 
Minnesota. Part of the current problem is that FairVote made a highly 
suspect interpretation that tried to assert BvS as being only about 
Bucklin, not about a sequential elimination method like IRV. This was 
based on a comment in BvS that was practically dicta. Other parts of 
BvS made it very clear that casting multiple votes in a single 
election was considered unconstitutional by that court.

(2) Brown v. Smallwood was quite a questionable decision, neither 
following the precedents of other courts, nor, in turn, followed by 
other courts. So its application remained in Minnesota.

(3) As soon as the suit was filed in Minnesota, I recommended to 
activists favoring other methods that resources be put together to 
file friend-of-the-court briefs arguing (1) that BvS applied to all 
forms of preferential voting and, specifically, (2) it did not 
specially apply to methods that fail LnH, and, in particular, applied 
to both IRV and to Bucklin. And that (3) the decision itself should 
be totally overturned.

(4) The danger was that BvS would be confirmed based on LnH failure 
of Bucklin, legitimizing *only* sequential elimination methods. Thus 
Minnesota would get IRV, which, it turns out, is a very questionable 
and, at best, shallow reform; but better and simpler reforms would 
remain impossible.

(5) One exception: what might be the most sweeping reform of them all 
is available with IRV as a base: Asset Voting. It was actually 
invented by Dodgson (Lewis Carroll) as a tweak on STV. All that one 
has to do is to assign any exhausted votes to the first preference 
candidate on the exhausted ballot. However, this does not satisfy 
LNH, technically, even though it clearly gives the voter maximum 
power to influence the outcome. (The vote only transfers to the 
"elector" if the vote would otherwise be wasted. Since it would go to 
the first preference candidate, it cannot harm that candidate. 
However, it fails LnH with respect to lower preference candidates, 
under what would probably be rare contingencies.)

Politically, I'd prefer to see IRV outlawed based on a pure 
confirmation of BvS, than to see the partial overturn based on a 
specification of BvS as applying only to LNH failing methods. But 
better than both these outcomes would be a simple overturn of BvS as 
an abberant decision.

And, yes, that would allow Minneapolis to proceed. I've become *very 
strongly* opposed to IRV, but one of the ways that the general public 
is going to come to this position is to have examples in front of 
them. IRV actually behaves badly, compared to top-two runoff, in 
terms of the opportunities afforded to independent and third-party 
candidates, who have *much* more chance of being elected under 
top-two runoff than they do under IRV.

IRV will elect an "independent or third party candidate" when this 
candidate was a frontrunner, and would have won under FPTP 
(Plurality). Circumstances where other than this happens are limited 
to spoiler-effect scenarios, where there is a strong third party 
candidate, and what I might call distorted vote transfers.

It turns out that vote transfers generally reflect the opinions of a 
subpopulation that, much more than theorists have expected, reflect 
the general opinion. That is, if, say, 60% of all voters prefer A 
over B and C, and C is eliminated, the C voters will generally favor 
A over B by roughly the same ratio, i.e. 60% of the transfers will go 
to A. Yes. This means that the plurality winner of the first round 
will go on, after transfers, to win the election. From the experience 
so far in the U.S., it *always* happens. But nearly all these 
elections, so far, have been nonpartisan elections. In partisan 
elections, it may shift. Or may not, we don't know. Historically, in 
Ann Arber, there was a reversal of the first round result, due to the 
strength of the Human Rights Party and thus vote transfers to the 
Democratic candidate. We might note that HRP strength was ultimately 
demolished after the victory.... but I'm not sure what that means.... 
(The HRP became, eventually, the Minnesota Green Party.)

With top-two runoff, something quite different happens. About 
one-third of the time, the runner-up in the first round goes on to 
win the election. Why?

Well, it really should be studied. There are two possible 
explanations that come to my mind.

(1) A dark horse candidate that wasn't considered a possible winner 
managed to make it to second place; then, once this candidate was 
taken seriously and examined by the public, a decision regarding true 
preference was made.

(2) Voter preferences didn't change, but voter turnout did. This, by 
the way, will shift the result toward what Range voting would 
produce. Reduced voter turnout has generally been considered a bad 
thing. That may be a shallow oversimplification, or, quite simply, 
not true. As long as the difficulty of voting isn't somehow biased 
(i.e., for example, voting is made very easy in affluent 
neighborhoods, and hard in poor ones), differential turnout will 
result in a decision that is more broadly satisfactory (in social 
utility terms). It is a sincere weighting of results according to 
preference intensity.

In reality, I think that both these causes operate. Top-two runoff 
gives minor candidates a much better opportunity to make their case 
to the voters; they need only get to second place, not all the way to 
first place. If the first election is IRV, however, it will be 
resolved based on the votes as cast in a single poll, and, because of 
the effect I've described above, it would be rare that a minor 
candidate would win.

The campaign for IRV in the U.S. found it necessary to turn to an 
economic and convenience argument in order to gain adoptions. In 
selling this, it was implied that IRV was merely a convenient 
short-cut to produce the same results as TTR. In San Francisco, the 
voter information summary, that a majority would still be required to 
win, was blatantly false, for the ballot measure itself removed the 
language from the election code requiring such.

Majority election requirements are a major safeguard in a democracy, 
and they've generally been swept away, when push came to shove. 
Further, access to runoff ballots by write-in candidates has been 
terminated. It was explicitly outlawed by San Francisco, and that was 
tested in the California Supreme Court, and sustained.

Voting systems theorists have generally ignored these details, 
preferring to stick to studying the simper single-round systems, and 
analyzing TTR only with an assumption that voter preferences are 
fixed things, and totally ignoring preferential turnout.

But the devil is in the details.

Because of differential turnout, Plurality is probably a better 
method than general election theory, neglecting the effect of 
preference strength on turnout, would predict. However, there is some 
suppression of this effect, I'm sure, in general elections, where 
many races are decided together. In a runoff election, there is more 
focus, voters turn out if they care about the result of that 
particular election.

By using Plurality with a majority requirement, and thus a runoff 
election is a possibility, the spoiler effect can be detected. 
However, even better methods for a first round are possible. It could 
be Bucklin, for example, or even IRV; however, IRV's "failure to find 
a compromise candidate" -- Robert's Rules strongest objection to 
sequential elimination preferential voting -- would leave this method 
unable to find a condorcet winner. Unless.

Unless write-in candidates are allowed in the runoff rounds, as they 
were in Long Beach, California, and which is the norm in California 
unless specific laws prohibit it. There, the mayor was prevented by 
term limits from being on the ballot. She ran anyway, and got a 
first-round plurality. Still not allowed to be on the ballot in the 
runoff, she won with a plurality in the second round (there was 
another write-in candidacy that prevented her from gaining a second 
round majority).

This really should be understood. Majority vote, FPTP with a majority 
requirement or the balloting is repeated, is Condorcet-compliant. All 
the supporters of the Condorcet winner have to do is vote 
consistently with their preference, which is true with any method. 
And, yes, the number of ballots is theoretically unlimited, except 
that in real practice they don't go on forever. However, Asset Voting 
addresses this brilliantly, allowing *negotiation* to determine the 
winner, additional polls would probably never be necessary, but the 
salutary effects of top-two runoff would remain (because the winner 
is utterly unconstricted, presumably, could even be a person who 
wasn't on the original ballot.)

At present, I'm promoting one simple idea: the best method (most in 
current use in the U.S. is Top Two Runoff, because it allows voters 
to truly reconsider their votes, making it possible for a dark horse 
to win, and actually doing this in maybe one election out of ten or 
so. To replace this with IRV on the argument that it saves money is, 
besides being quite speculative and quite possibly wrong, penny-wise 
and pound-foolish, if we place value on democratic and majority-rule 
values. IRV is a plurality method, it cheerfully (and usually in the 
U.S., whenever a majority wasn't obtained in the first round, i.e., 
when votes are transferred) elects with only a plurality of the votes.

>Will you have any complaint when in future they are trying to do the same
>thing to some Condorcet method you like and IRV supporters help
>them on grounds like it fails Later-no-Harm, Later-no-Help, and
>probably  mono-add-top?

Unlikely.

But I think I addressed this. BvS should be overturned. IRV should be 
allowed in Minneapolis, though possibly some details should be 
addressed. I particularly dislike that IRV is sold as somehow 
respecting majority rule, because the practical result of IRV is 
quite the opposite. Lewis Carroll addressed this over a century ago, 
inventing what we now call Asset Voting as a method of enfranchising 
the common voter, who easily doesn't have the knowledge to rank 
multiple candidates, but knows the one candidate most trusted.

The problem of exhausted ballots gets swept under the carpet by IRV 
promoters, who have attempted, generally, to minimize or distract 
from it. IRV is sold as if all ballots will be fully expressed; but 
in Australia, often cited as an example of how IRV works, that is 
accomplished by considering ballots as spoiled that aren't fully 
ranked, a provision that is unlikely to fly in the U.S. In areas in 
Australia that use Optional Preferential Voting, ballot exhaustion is 
common, and, of course, election by plurality is, likewise, and the 
phenomenon of "comeback elections" turns into practically universal 
election of the first round winner (i.e., the method almost exactly 
duplicates Plurality.)

FairVote got stuck on IRV long ago, because of their long-term goals, 
and generally disregarded the election methods community and all the 
built-up expertise on the topic, in favor of political strategy and 
promotion, decided centrally by the organization (and possibly by, in 
effect, a single individual). As a result, they may indeed be showing 
some success, but with a method vastly inferior to what is possible. 
Bucklin was used, worked, and should do at least as good a job of 
dealing with the spoiler effect as IRV. It's more likely to find a 
majority (possibly one out of three majority-failure IRV elections 
would find a majority, because Bucklin is more efficient at finding 
majorities). It's cheaper to implement, because it is simply a 
sum-of-votes method, precinct summable. Combine Bucklin with a true 
majority requirement, you'd have quite a sophisticated method. And 
there are many other possibilities.

Range, of course, is ideal, properly implemented, because it uses the 
best method for evaluating election outcomes as the method itself. 
And it can be made Majority Criterion compliant (though runoffs when needed.)

(Chris wouldn't agree about MC compliance, but by that, I mean that a 
majority have approved the winner. They might have preferred someone 
else -- but that is only theoretically a problem, it's quite unlikely 
in actual practice unless voters have no knowledge of who is likely to win.)