seppley at alumni.caltech.edu
Wed Nov 6 21:10:54 PST 1996
>If it would make Mr. Eppley happy, then just delete "3. Instant
>Run-Off (a.k.a. MPV)" in the 7-8-96 response since he chooses to
>ignore 1. and 2. in such posting.
Donald is now the only person whose poll response ranks IR ahead of
Condorcet. He's acknowledged having skipped a lot of the detailed
discussion, messages which might have shown him that the arguments
for Condorcet are not "hippopotamus logic." Maybe he now has more
time to look at the details: LOE, generalized majority rule, and
In what sense do you think I ignored your #1 and #2? They were
included in the report, and I didn't suggest that you amend your
response to rank Condorcet or Smith//Condorcet ahead of your #1
or #2. I suggested that you insert several additional methods,
including a combo of Approval & IR, after your #2 and before IR.
Perhaps you think I should have included more about your #1 and #2 in
the report's glossary. I disagree, since no one else included them
in their rankings and you'll be able to describe them to your heart's
content if we ever get the Commentary section done. (I think the
Commentary section is essential if the poll results are to have any
I tried to make it clear that people were permitted to amend their
poll responses at any time, and that people who hadn't responded were
permitted to submit late responses, and that the draft was open to
modification. It would have been more helpful if you had asked to
have the "3. IR" deleted *prior* to the report's posting in ER.
>Such 3. was a very weak 3 (on a 100 to 0 point scale, about a 1
>with simple plurality being zero).
I think I'd rate IR much higher than 1 on a 0-100 scale.
What's your rating of Condorcet? Less than 1?
>I note that any second or later choices (no matter how weak) that
>may be made by anyone at anytime on this list are now apparently
>suspect notwithstanding any qualifying statements or later
Our reports shouldn't be frozen. If people change their views we
should make an effort to keep our public products updated to match
current thinking. It's not to please Mr. Eppley; it's to avoid
misleading the public (which includes ER and CV&D). I invite you
to post a (hopefully brief) message in ER informing them that you've
amended your response.
I think all our ideas should be forever suspect, even first choices.
>I note that any choice in addition to a first choice is
>"inconsistent" with such first choice.
What is inconsistent is to rank a method higher than a more preferred
method. Ranking an addition as "less preferred" than one's first
choice seems perfectly consistent, so I don't understand your point.
Your most preferred method, according to your poll response and your
recent messages, is some combination of Approval and Condorcet.
And you've stated that majority approval is vital, which should lead
you to prefer methods which include approval more than methods which
don't. (Both your #1 and #2 are consistent with this.) Since
Condorcet is part of your most preferred method, then I don't
understand why you listed plain IR instead of plain Condorcet as
your most preferred of the "plain nonapproval" methods. That's
what looks inconsistent, and I don't understand why it's like pulling
teeth now to persuade you to write explicitly that you prefer plain
Condorcet and/or Smith//Condorcet more than plain IR.
>This is all very cute but nonproductive in getting some SW reform
Getting our ducks in a row might be productive. A consensus (i.e.,
zero opposition) in EM would likely be more influential on the course
of single-winner reform than a divided EM.
I understand your dissatisfaction with plain Condorcet and plain IR.
But as I see the big picture, plain IR is the "default" method which
reformers will adopt (because it's the only one they know about and
the only one that respected reformers like Rob Ritchie will mention)
unless we put forth a great deal of effort to explain why other
methods are better than IR. You may have your own "lesser of evils"
dilemma to contemplate: what to do to make sure the "greater evil" IR
isn't the weak reform adopted by an uninformed community of electoral
In other messages, Mike O. and I are trying to point out that your
alarm about nonapproval in plain and Smith// Condorcet is overblown,
so perhaps in a "later development" you may stop seeing Condorcet
as a lesser of evils. Or maybe you'll convince me that adding the
approval option to Condorcet won't risk the passage of the Condorcet
reform--the two things which worry me about the approval option are
that the Anti campaign will frighten the electorate with alarms about
vacuums of power, and that maybe they'd be right to sound those
alarms. Who knows?--as Mike acknowledged, maybe the electorate,
like you, would much rather have vacuums of power. I think our
best reform strategy, though, is to postpone the choice between
Condorcet and Condorcet+Approval (a.k.a. Condorcet+NOTB) until
we're in a position to get massive public input (via polls and
focus groups) on which of these choices they'd prefer. Looking
at it this way, the "+Approval" doesn't make it a wholly new
method to educate people about; it's more like a "one from
column A" (Condorcet, IR, etc.) and "one from column B" (no
expression of disapproval, or one of the ways to tally NOTB
which Mike outlined recently). The properties of columns A
and B are separable.
I have a few questions regarding the Majority Approval standard
(and the similar standard, Majority Disapproval):
Case 1: No candidates are majority-approved.
1.1 Why do you think a good candidate who would be majority-approved
in a re-election can be found? What barriers would keep such a good
"white knight" from competing in the first election but not prevent
him/her from competing in the re-election?
1.2 Are you concerned that the voters would "tactically" or
"capriciously" vote to majority-disapprove all the candidates
even though there's a decent compromise running?
30: A|BC <-- these disapprove the compromise B
30: C|BA <-- these disapprove the compromise B
Enough voters might opt to have no winner indefinitely rather than
compromise, and ensure no one will ever win. Maybe some voters will
prefer anarchy over compromise.
Maybe negative campaigning will persuade voters that decent
candidates ought to be disapproved-- you can expect a lot more
negative ads if the method allows voters to express disapproval,
since opponents of the compromise choice may seek to try again in
a re-election (particularly if they have a big money advantage.)
(The Anti campaign will have a field day alarming people about the
vacuum of power; do you recall Matthew Shugart's emphasis of the
Case 2: At least one candidate is majority-approved (or would
be, except that the "plain" ballots don't allow expression of
2.1 Are you concerned that plain Condorcet might elect a disapproved
candidate when there's at least one approved candidate running (e.g.,
Hitler vs Stalin vs George Washington)? If so, would you provide an
example which illustrates this scenario? If not, would you agree
that Condorcet's counting of pairwise preferences instead of first
choices, and its selection of smallest "largest votes against in a
pair-loss", already does a good job of selecting the majority-
approved candidate, even when the voters can't express approval?
---Steve (Steve Eppley seppley at alumni.caltech.edu)
More information about the Election-Methods