[EM] Fwd: The Top-Set. Method-Merit.

[Jameson Quinn]

>
> ...instead of including some dismal pessimistic compromise like Hillary.
>
This is, of course, moving away from abstract voting theory into concrete
politics. I'm going to follow that move. I don't want this to turn into a
thread on the list, though; so, if you want to answer this, feel free to
respond to individually to me and/or Mike (at the "Janet Robinson"
address), without including the list.

So, you consider Hillary to be a "dismal pessimistic compromise". It seems
likely you're saying this from the left, as otherwise she wouldn't be any
kind of compromise, but simply the greater evil.

In an absolute sense, you may well be right. But I think the relevant
comparison set is that of historical presidents, and perhaps major party
nominees as well. In that comparison set, I'd argue that Hillary is
actually among the more progressive, more honest, and more competent as an
administrator; and that she's not significantly worse than average in terms
of her inherent charisma and skills as a campaigner.

Note that I'm NOT saying that she is actually progressive, honest, or
competent in an absolute sense. Far from it. I'm just saying that she is
above average for a successful politician at the national level.

If all that's true, why is her approval level among the lowest in history
for a nominee (aside from Trump)? I'd say that she's faced an unprecedented
level of media bias. Her actions regarding Whitewater, Benghazi, the email
server, and the Clinton Foundation have all faced extreme scrutiny, and
while she clearly acted wrongly in some ways, at the end of the day there
are more "troubling questions" than troubling answers. I think that the
average politician has done much worse.

How is any of this relevant to voting theory? I think that Mike is
suggesting that there should be some kind of absolute cutoff for "honest
approval". And I see the allure of that attitude. But I think that it's
clear that there are plausible sets of candidates among which a progressive
voter would be strategically well-advised to approve Hillary. (And I can
easily imagine corresponding scenarios where a conservative or libertarian
voter should strategically approve of someone they find unpalatable.)

I have more to say on this, but I'll continue quoting Mike for a bit first:

> Your "bottom-set" consists of everyone who isn't in your top-set.
>
> Approval's 2-valued ratings perfectly match the candidates' 2 important
> categories (top-set & bottom-set).
>
> Approving your top-set maximizes the probability of electing from your
> top-set.
>
> Approval is the method that asks the relevant question.
>
> Based in the above, it _is_ possible to objectively say what the best
> method is. It's Approval.
>
> Yes, people want to vote not only _for_ their top-set--they want to also
> vote _among_  their top-set.
>
> In other words, they want it all.
>
Yes! Of course they want it all! And I think that a system like U/P can
give it all! Your best top-set is "preferred", your broader operational
top-set is "acceptable", and others are "unacceptable".

> That's too much to ask of a voting-system. You'll increase the probability
> of electing from your bottom-set.
>
I have no idea what the logic is here. There is no sense in which U/P is
more likely to elect a voter's bottom-set than approval.

....

> I suggest that the best rank methods consist of 3 Bucklin versions and
> MAM. Which to choose depends on what people insist on, and what kind
> offensive and defensive strategy is expected.
>

I agree that MAM is good.

> 3 Bucklin versions:
>
> 1. Plain Bucklin:
>
> No equal-ranking or skipping.
>
> 2. B or A:
>
> You can vote Plain Bucklin or Approval
>
Why would you consider either of these better than something like U/P or MJ

> 3. ER Bucklin:
>
> Equal ranking allowed. No skipping.  ...except that if you equal-rank n
> candidates in a round, then the count rule automatically skips n-1 rounds
> before giving votes at your next rank.
>
I don't entirely understand what you mean by "automatically skips n-1
rounds", except that it seems that this makes the system more ranked than
rated.

I'd say that any graded Bucklin method that uses strictly-above-median
tallies as a tiebreaker — such as, in the simplest-to-describe case, U/P —
is better than any of the three above.

Not that the above aren't better than many methods. But I wouldn't call
them best.

...

> MJ advocates obfuscate it & its strategy, for themselves, by their
> non-Bucklin wording of it.
>
I have sympathy for this criticism, which is why I'm now pushing U/P over
MJ.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.electorama.com/pipermail/election-methods-electorama.com/attachments/20160907/96bfe56c/attachment-0001.htm>