[EM] IRV and PR-STV Re: examination of plaintiff's memorandum re IRV in Minneapolis

[Abd ul-Rahman Lomax]

At 10:45 AM 11/10/2008, Raph Frank wrote:
>On Mon, Nov 10, 2008 at 3:32 AM, Abd ul-Rahman Lomax
><abd at lomaxdesign.com> wrote:
> > Is this complicated? Yes. Is it fair? Well, up to the election of the last
> > candidate, yes, it is clearly fair. With the last candidate, the election
> > effectively becomes the same as an instant runoff voting election, with the
> > problems associated with that.
>
>I don't see why the last seat being filled is that much different from
>the others.  The negative effects are always there, though because
>factions aren't quite solid coalitions in practice, PR-STV doesn't
>display the effects quite so much (maybe).

Okay, sometimes it isn't the last seat. Rather than saying "last 
seat," I should say, "up until the point where non-election 
eliminations begin." Scenarios can certainly be constructed where 
there is a problem before the last seat. Normally, though, the first 
eliminations are obviously reasonable. In IRV scenarios, *usually*, 
the problem elimination (the one which drops the Condorcet winner) is 
the last one. Hence my comment about "last seat."

Raph's approach, by the way, shows the weakness of study through 
constructed examples. As he notes, "because factions aren't quite 
solid coalitions in practice," which is like saying "because people 
aren't quite like stereotypical caricatures of the party of the 
candidate they choose in a single election to vote for," what is 
usually missing from examples is any consideration of how realistic 
the example is.

It's the old approach to considering election methods, part of the 
"criterion" approach. Approval Voting -- by some definitions -- 
doesn't pass the majority criterion. Therefore, toss it in the trash. 
Even if the Approval result is clearly the more just one, the 
unifying one, resulting in significantly better overall voter 
satisfaction. Even if the violated majority preference is quite weak 
(which it must be for the majority criterion to be violated under Approval).

There is only one way to study election method performance, with any 
hope of objectivity,, and that is utility analysis, the approach of 
Warren Smith. The same kind of mistake could be made there: an 
artificial utility profile might warp the results, but some of the 
simulations he's done do attempt to generate, at the outset, 
reasonable utility profiles, perhaps using issue distance in n-space.

>For example, assume that 2 seats are being filled
>
>A faction
>
>16: A1>A2>A3
>5: A2>A1>A3
>14: A3>A2>A1
>
>B faction
>
>14: B1>B2>B3
>5: B2>B1>B3
>16: B3>B2>B1
>
>C faction
>
>30: C
>
>A2 and B2 (and C) are the condorcet winners in their own factions.

This is truly a bizarre example, by the way. Yes, Raph noted the 
problem. Rigid factions, no crossover, no independent voters.

With the above scenario, though, the first election is A1.  The 
factions have voted entirely within their own faction. (In the real 
world, such a tight faction would internally agree, in separate 
process, on a single candidate, probably A1, *though we cannot tell 
without understanding the preference strengths*.)

So the faction totals are A:35, B:35, C:30. This would be a razor's 
edge election if only one seat were being chosen.

Because of the rigid isolation within the factions, we have, 
effectively, three IRV elections, one for each party. Hence the first 
seat can pass over the faction's Condorcet winner, due to the low 
first preference support for A2.

Eliminations have occurred prior to *any* seat being chosen.

And this shows why I've long proposed that the place to start with 
election reform is outside the actual election process, but with 
pre-election process. Choose and support Condorcet winners *before* 
the election, using better election methods or, even better, 
deliberative methods (such as FA/DP).

Theoretically, a political party (even more a "faction") should be 
able to choose its own election process. If not, democracy has been 
lost, only a surface sham remains.


>PR-STV will run as follows
>
>Quote = 35 (approx)

The Droop Quota is 34.

IRV run on the election is really three elections, collapsing a 
primary into the final. Each faction elects its winner, the original 
eliminations clearly do this first. Thus this particular election is, 
in effect, a set of three factional elections, each choosing one 
winner. And that is how the eliminations proceed. So, in this 
constructed scenario, because IRV can pass over a Condorcet winner, 
one of its flaws, the factional condorcet winners lose the initial 
election. But the factions each get one rep except for the smallest one.

Thus the general result is fair, though I consider it unsatisfactory. 
The lack of satisfaction is due to there being only two reps for 
three roughly equal factions. Elect three reps, and see what you get.

Droop quota is 26.

A1, B1, and C are elected, no problem. Except, of course, for the 
choice within each faction. This example shows why, if I were a party 
leader, I'd strongly oppose eliminating party primaries to "save the 
expense and hassle." (Okay, party primaries *at public expense* could 
be eliminated, but, then, parties should remain free to choose their 
own candidates. Not to choose who runs, but who carries the party 
banner, who can be identified on the ballot as a candidate of a party 
(rather than merely self-identified. If parties receive any 
recognition on the ballot, though, the public has a right to demand 
some kind of openness to the process, some way in which it can be 
known how a candidate has received the party stamp of approval. 
Parties could *choose* to run multiple candidates, of course, as they 
would when there are enough seats up for grabs.)

STV's problem is the same as that of IRV, it is in the candidate 
eliminations. Any candidate elected before eliminations begin is 
clearly legitimate. (This is *certain* if the Hare quota is being 
used, and quite reasonable under the Droop quota.)

(I.e., Hare quota winner has won 1/N of the votes in first 
preference, and is thus clearly legitimate to, in an assembly, 
represent 1/N of the assembly voting power. With the Droop quota, the 
candidate has 1/N of the voting power but has received 1/(N+1) of the 
votes, thus is exercising, possibly, some unearned edge of power.)

There are Asset Voting schemes which fix this, or certainly 
ameliorate it. The method can be STV. With Asset Voting, the A2 
voters, understanding their candidate's position, would vote only for 
A2, perhaps. No A faction candidate would reach the quota, then, 
without the approval of A2. Given that A2 is a unifying candidate 
(assuming that the A1>A2 and A3>A2 preference strengths are weak, 
what is likely is that either A2 accepts A1 as the winner, 
considering this a reasonable result, by transferring those 5% of 
votes, or A2 gains the votes of the A3 voters through elimination of 
A3 and then A1.

If the A1 and A3 voters "retaliate" by truncating, themselves, not a 
problem. A1, A2, and A3, all members of the same faction, negotiate, 
on behalf of their supporters, who is to win the faction's votes and 
thus a seat, and, benefit that comes with it, they get to use up 
their last remaining unassigned votes. In Asset systems I'd propose, 
they might choose A4!, someone eligible but not on the ballot.

Having spent 34 votes, the A faction still has a vote left, which can 
be spent by the asset holder(s). In this case, there is nothing to 
spend it on, but in some elections, it could make a difference in a 
later election choice.

Asset voting, folks, was invented as a tweak to STV, back in the 
1880s. It was a brilliant idea, from one brilliant polymath. Warren 
Smith reinvented it and gave it the name Asset Voting, early this 
century. It was previously mentioned by Mike Ossipoff on this mailing 
list in the 1990s.

It is a variation, really, on Delegable Proxy, and Delegable Proxy, 
when it comes to pass that vote assets are being held by many 
candidates -- I expect that Asset Voting will lead to *massive* 
numbers of "candidates," so many that a booklet will be published 
with those eligible, through registration, to receive votes -- and 
the registration fee will be, basically, the shared publication cost 
to put a name in the booklet, i.e., *cheap*. Like a few dollars; 
indeed, most of the cost would be, I think, in the cost of collecting 
the fee and processing and confirming the registrations. Votes would, 
then, be for candidate *number*, the number taken from the booklet 
available with the ballot.

Ahem, that sentence did go on, didn't it. The end of it is: -- 
Delegable Proxy can be used by the asset holders to negotiate their 
final vote transfers. That would be entirely voluntary, but if 
intermediate transfers are recorded (and reversible are changeable, 
up to the point where a binding decision is made), a functioning 
Electoral College will have been created that can serve other 
purposes as well. It would make Direct/Representative democracy 
possible. It could unseat winners, creating an easy impeachment process.

(I've discussed the details many times; I've suggested that, when a 
holder of a seat loses the confidence of his or her supporters, there 
would be latency, where the seat holder has reduced voting power, 
always having only the voting power of maintained supporters, but 
would retain deliberative rights, i.e., to speak to the assembly and 
enter motions. Thus a minor loss of vote support would only be that: 
a minor loss of voting power. Not a drastic removal from office.)

Truly, the devil is in the details. For many, what happens with 
exhausted ballots is an unimportant detail. The problem, on the face, 
doesn't exist in STV as usually practiced in Australia, because full 
ranking is mandatory. But that, by depriving voters of the freedom to 
truncate (to decide that they are not willing to support any other 
candidates than those they ranked), a basic democratic right has been 
eliminated: the right to deprive a candidate of what they call in 
Australia an absolute majority, resulting in election failure and 
need for further process. Asset Voting provides an efficient way for 
the further process to proceed without an expensive public election, 
but only, effectively, a "runoff election" where the Asset holders 
vote, in public. Given the nature of this body, reduced in size, 
operating publicly, that "election" could really be a continuous process.

It would be exciting to watch, I'd think, to see as assigned votes 
approach a quota. Today, this could be done on a web site; given that 
votes are public, the security issues are minor. (Someone hacking and 
presenting false votes would be quickly discovered, by the actual 
asset holder! -- and, of course, before validating final results, 
officials would make sure that the electors confirm what is recorded 
for them.) The FA/DP process, if used by electors, would be advisory 
only, the electors would directly vote whatever they choose, but 
presumably as advised by their own chosen proxies in the FA process.

Let me emphasize this: this is a tweak on STV, remedying a problem 
with STV that was first noticed by Lewis Carroll, and making it 
unnecessary for voters to have sufficient knowledge to fully rank. 
That, indeed, was his concern, that the "common voter" would be quite 
likely to only know one thing well: their favorite. I don't know that 
he realized that it could have a major impact beyond that, that it 
could pretty much make the *need* for political parties obsolete. 
They would still exist, I'm sure, but much more as voluntary 
coordinating bodies than as necessities of the process. (STV in 
Australian, say, would be pretty chaotic without the political 
parties and their recommended preference lists, there would be lots 
more "donkey voting.")

So I am hoping, in fact, that the defendants in the Minnesota case 
prevail, even though it means that IRV will proceed in Minneapolis. 
My opposition to IRV, which is strong, is not an absolute; if 
Minneapolis wants to waste money and effort on a method that, in 
practice, is little or no improvement over Plurality in nonpartisan 
elections, and that has pretty much only one benefit in partisan 
elections, the reduction of the spoiler effect, when there are other 
reforms that are cheaper, simply, fairer, and which more completely 
address the spoiler effect, that's their privilege. The sky will not 
fall, Kathy. The country will not therefore shift to IRV in a 
landslide. These are *city* offices being elected, not state ones, if 
I'm correct.

As to top-two runoff, consider what would happen if top-two runoff 
were tweaked with the Asset fix. Most runoffs would be avoided, there 
would be a runoff only if the asset holders could not put together a 
majority within a set period of time. There would be no eliminations 
in that runoff, I'd assume, though it might be a plurality election. 
Presto! Condorcet winners would be quite likely to prevail in the 
runoff, or before. *Range winners* could beat Condorcet winners, when 
they differ (which is rare). Think about it. Differential turnout 
tests preference strength.

(The runoff would also be an Asset election, but with a tight time 
limit for vote transfers. Enough to negotiate. The groundwork for the 
negotiation would have been laid in the primary.)

The voters would see the same ballot as now. The only difference is 
that no votes are wasted, except those cast foolishly for a candidate 
who will waste them. TAANSTAAFL.