[EM] IRV's powerful advantage. Method-comparisons by strategy, and in Burlington. Burlington-proofing IRV.

[Michael Ossipoff]

To: Arizona League of Women Voters

From: Michael Ossipoff

Before I say more, I want to emphasize that I really like IRV. It's
one of my favorite methods, maybe my favorite, for the "Green
scenario", a scenario in which Greens have been elected to the
presidency and most of Congress. Conditions then would be different
from current conditions in at least two ways:

1 An electorate that could elect a Green government, in Plurality
elections, is an electorate who must not any longer believe the media
disifnformation that can now so easily induce favorite-burial.
Therefore, IRV's vulnerability that that kind of media manipulation
wouldn't be a problem, and therefore, IRV's big advantages would
recommend it well.

2. The Green (GPUS) platform promises vastly more open, free,
participatory and agenda-free media. Free exchange of information.
Under those media conditions, the current media disinformation problem
wouldn't exist anyway.

Those are the conditions that I refer to as "the Green scenario."  As
you probably know, GPUS offers IRV as the voting system in the
government proposed in the GPUS platform. I fully support and agree
with the Greens' offering of IRV in their platform.

IRV;s big advantages: IRV meets the following criteria:

1. Mutual Majority Criterion (MMC)
2. Later-No-Harm (LNHa)
3. Later-No-Help (LNHe)

The combination of MMC and LNHa is a very powerful combination. It
means that, with IRV, a cohesive majority are free of
strategy-need--free to just rank sincerely.

A few definitions:

A mutual majority (MM) is a set of voters who all prefer a certain
same set of candidates to all of the other candidates.

[end of MM definition]

That set of candidates is that MM's MM-preferred set.

Mutual-Majority Criterion (MMC):

If the members of a MM vote sincerely, then the winner should come
from their MM-preferred set.

[end of MMC definition]

Later-No-Harm (LNHa):

Later-No-Harm says that ranking some candidates below the candidates
of Set A shouldn't take the win from Set A. So you're free to rank as
many candidates as you want to, without fear that doing so will hurt a
higher-ranked candidate.

Here is a more complete and precise (but wordy) wording of LNHa:

Voting for a candidate means voting hir over another candidate.

If a candidate from some set (set A) of candidates would win if you
cast a certain ballot on which you've voted only for them, then voting
for additional candidates shouldn't result in the winner not coming
from Set A, provided that you've not ranked any of those additional
candidates over the Set A candidates, and provided that you've voted
the Set A candidates over the additional candidates if the balloting
system in use allows you do do so when voting for the additional
candidates.

[end of LNHa) definition.]

The reason why MMC and LNHa is such a powerful combination is that,
because of IRV's LNHa compliance, a mutual majority of voters have no
reason not to sincerely rank all of the candidates whom they
like--including all of their MM-preferred set. And that guarantees
that one of their MM-preferred candidates will win.

They have the luxury of choosing between their MM-preferred candidates
via sincere ranking, while ensuring that one of them will win. Very
few voting systems offer that guarantee.

 IRV's LNHa compliance is what empowers its MMC compliance. Together,
they make IRV strategy-free for a mutual majority of voters.

That's why IRV is one of my favorites, maybe my favorite, for the
Green scenario. I fully support and agree with the Greens (GPUS)
offering of IRV in their platform.

Conditions nowadays are a bit different. Our media never mention any
candidate other than those of the Democrat and Republican parties. The
media present the Democrat and the Republican as "The Two Choices".
The implication, the message, from the media is that the winner will
and must always be a Democrat or a Republican. Nearly everyone
believes that. In fact, how many people have even had the opportunity
to hear about the Libertarians or the Greens, for instance?

I refer to that as "current conditions", in contradistinction to the
Green scenario.

Under current conditions, when the voting system is Plurality, IRV, or
the traditional Condorcet methods, if a voter prefers the Green to the
Democrat, and strongly prefers the Democrat to the Republican, then
hir optimal strategy is to (in Plurality) vote for the Democrat, or
(in IRV or traditional Condorcet) rank the Democrat alone in 1st
place.

Just ask anyone who votes Democrat in our Plurality elections. They'll
tell you that they must hold their nose and vote for the Democrat even
though they don't like hir. They're thereby voting their optimal
strategy, given their assumptions and beliefs, as described in my
definition of "current conditions".

Those same "current-conditions" beliefs and assumptions likewise make
favorite-burial the optimal strategy in IRV.

Favorite-burial makes a joke of the election result. It's a drastic
falsification of voter preference. A strategic need for favorite
burial has no place in a democracy's voting system.

That's why IRV can't be recommended for current conditions.

There's a criterion called the Favorite-Betrayal Criterion. It says
that no one should ever have strategic incentive to vote someone over
hir favorite.

In wordier, more complete and precise form, here is FBC:

A ballot top-votes a candidate, or votes hir at top, if that ballot
doesn't vote anyone over that candidate, and votes that candidate over
at least one other candidate.

If a ballot votes a set of candidates, Set A, at top, and a candidate
in Set A wins, then moving an additional candidate to top, shouldn't
cause the winner to be a candidate who is not then voted at top on
that ballot.

(In other words, the winner should be either that newly top-voted
candidate, or someone in Set A)

[end of FBC definition]

Plurality and IRV fail FBC.

In the seriously pathological and abnormal "current conditions", it's
essential to use a voting system that doesn't fail FBC, in order to
avoid the gross distortion of voter preferences caused by a strategic
need for favorite-burial.   ...and the obvious adverse societal
consequences of that.

What methods meet FBC?

Approval meets FBC.  So does Score.

There are a few rank methods that meet FBC, but, as a practical
matter, Approval and Score are better, more winnable, proposals than
the new rank methods that meet FBC.

Score is a generalization of Approval, in which the voter can assign
any number of points (within a pre-specified range, such as 0-10), to
each candidate. Score might work better, but Approval would work fine,
and is probably more winnable.

Arguments that we've heard against Approval:

"Approval violates 1-person-1-vote, and gives more power to voters who
approve more candidates.":

Actually, each voter has the same voting power. 1-person-1-vote (1p1v)
referred to the greater voting power of voters who were in smaller
legislative districts.

Suppose that I approve all of the candidates. How much voting power
does that give me? Zero. I have no influence on the election outcome.

Suppose that there are 10 candidates. I approve 9 of them. You approve
only the one that I don't approve.

Your ballot completely cancels out my ballot.

I approve 9 candidates. You approved 1 candidate. Your ballot canceled
out my ballot.

The unequal power of voters in Approval is a fallacy and a myth.

"Approval has tactical incentives":

It's been proven that every voting system has tactical incentives.
But, because Approval meets FBC, Approval does _not_ have IRV's
favorite-burial tactical incentive. In fact, "tactical incentive" is
an understatement. Under current conditions, IRV has a
_strategic-need_ for favorite-burial.

FairVote &/or Richie has said that, in Approval, voters will use the
tactic of "bullet-voting", approving only their favorite. What?? If
voters were inclined to do that, then they'd be doing it now too, in
Plurality. Instead, we see millions of people voting for a compromise,
instead of for their favorite. ...People "holding their nose" and
voting for the Democrat. In Approval, it's obvious that those same
people would approve the Democrat--but they'd also approve everyone
whom they like better, including their genuine favorite.

In Approval, unlike ever before, everyone would feel free to fully
support their favorite.

Yes, in Approval there could be need to also approve a compromise. But
remember that IRV gives strategic need to vote that compromise _over_
one's favorite. Favorite-burial.

In Approval polls, I just tell people to approve the candidates or
alternatives that they like. When people do that, the winner is the
candidate liked by the most people. The most liked candidate.

IRV and Approval in Burlington:

In Burlington, IRV was thrown out. When the Democrat was eliminated,
and most of hir votes transferred to the Progressive, the progressive
won, though a majority preferred the Democrat to the Progressive.
We've been warning, for years, that IRV will often do that. When IRV
did that in Burlngton, when IRV elected the Progressive, IRV violated
the wishes of the majority who preferred the Democrat to the
Progressive.

It isn't good to have an angry majority. That's why IRV was thrown out
in Burlington.

What would Approval have done in Burlington?

Well, suppose that, maybe based on polls, the public perception is
that there are more Progressives than Republicans. Then, the
Republicans obviously should approve the Democrat. The Democrat will
have approvals from the {Democrat, Republican} majority, and will win.

Or, if it's publicly perceived that there are more Republicans than
Progressives, then the Progressives obviously should approve the
Democrat, and the Democrat will have approvals from the {Progressive,
Democrat} majority, and will win.

In other words, majority rule is not violated. There is no angry
majority wanting to throw-out the voting system.

So: Not only does Approval pass FBC, while IRV fails FBC, but Approval
would avoid the majority-rule violation that got IRF thrown out in
Burlington.

Burlington-Proofing IRV:

But IRV can be fixed. I believe that Condorcet-IRV, which I call
Unbeaten//IRV, was proposed in a 1926 book by George Hallett, who was
Rob Richie's great uncle.

Here are some preliminary definitions:

X beats Y of more ballots rank X over Y, than rank Y over X.

If X isn't beaten by anyone, then X is unbeaten.

Unbeaten//IRV:

If there is an unbeaten candidate, elect hir.

Otherwise, do IRV.

[end of Unbeaten IRV definition]

That method wouldn't do the majority-rule violation that got iRV
thrown out in Burlington. Unbeaten//IRV is a good method that would be
more successful than iRV.

Hallett recommended Unbeaten//IRV as his best recommendation for a
single-winner voting system. He advocated ordinary IRV as a 1st
proposal, because of its greater simplicity.

By the way, speaking of simplicity, here is a simpler and briefer way
to define iRV:

Repeatedly, cross off from the rankings the candidate who currently
tops the fewest rankings.

[end of brief IRV definition]

It's obvious that that wording implies that that process continues
till there's only one un-eliminated candidate, and that s/he wins.

Sure, IRV calls for electing someone when s/he acquires a majority.
But that isn't necessary--S/he'd win anyway.

Unbeaten//IRV is very much good enough, but it can be improved on. If
there's no unbeaten candidate, and so IRV is used, then why not stop
as soon as, during the IRV eliminations, and unbeaten candidate.

Here's a method that does that:

Benham's Method:

Do IRV until there's an unbeaten candidate. Elect hir.

[end of Benham's method definition]

Obviously, if there were initially an unbeaten candidate (the one that
Unbeaten//IRV would elect), Benham would elect hir too.

Here's another alternative to IRV:

Approval-IRV (AIRV):

Same as IRV, except that voters may rank as many candidates as they
want to any rank position.

If you rank several 1st choices, then they each get an initial whole
vote from you.

When all of your rank N candidates are eliminated, then each of your
rank N+1 candidates gets a whole vote from you.

[end of AIRV definition]

Here's something similar to Unbeaten//IRV. It's more complicated and
wordy to define, and just very slightly better:

Woodall's method:

The Smith set is the smallest set of candidates such that every
candidate in the set beats every candidate outside the set.

Do IRV until only one member of the initial Smith set remains
un-eliminated. Elect hir.

[end of Woodall definition]

Conclusion:

1. IRV would be an excellent method, my favorite or one of my
favorites, for the Green scenario.

2. Due to its FBC failure, IRV is not adequate for current conditions.

3. Approval and Score meet FBC, and wouldn't commit the majority-rule
violation that got IRV thrown out in Burlington.

4. I recommend 3 methods that avoid IRV's Burlington problem:
......a) Unbeaten//IRV
......b) Benham's method
......c) Woodall's method

5. I recommend a simple method that avoids most of IRV's
favorite-burial need, and would have given the Burlington Republicans
a non-drastic strategy to elect the Democrat:
......AIRV.

Michael Ossipoff