[EM] Approval-based replacement for jungle primary
robla at robla.net
Sat Dec 1 19:25:53 PST 2018
Thanks for pointing back to the conversation Chris Benham started. I
need to read and fully absorb that conversation, but I'm going to try
replying in advance of that. More inline...
On Thu, Nov 29, 2018 at 1:27 PM Ted Stern <dodecatheon at gmail.com> wrote:
> Regarding your modification suggestion, I'm not sure that you fixed
> anything. Including everyone above 50% approval potentially creates an
> over-crowded ballot.
If a candidate requires 50% approval to advance to the general
election ballot (to get into the "majority candidate pool" as I call
it in my 2018-11-22 email), I have a hard time believing that an
"overcrowded ballot" will really be a problem. If people become
familiar enough with approval voting that a majority of voters are
approving a large number of candidates, that seems like a really good
problem to have. In some of the offline conversations I've had, I've
received pushback that 50% approval is way too high. It seems that if
we get an overcrowded ballot because we allow every candidate that has
50% approval to advance, then by definition, a majority of the
electorate *wants* an overcrowded ballot. Plus, it seems like there
are policy countermeasures that could be introduced as disincentives
for a unified block of candidates to operate this way (e.g. having a
fixed pool of money for public campaign financing divided evenly among
the candidates in the majority candidate pool)
As you point out, my "Majority Approval Filter" [MAF] draft proposals
also call for allowing an "opposition candidate pool" to advance, such
that it's possible for candidates with lower than 50% approval to
advance. My initial draft (from my 2018-11-20 email) allowed for
exactly one candidate to advance via the opposition candidate pool.
My second draft (from my 2018-11-22 email) allowed for an opposition
candidate pool that could be potentially as large as the majority
I'm probably going to use electowiki to create a third draft of MAF
when I get around to it. More about that below...
> However, as Chris Benham states in the included
> post above, Approval Winner plus complementary (or, as I like to call it,
> excluded) approval winner could lead to Condorcet violation.
I did some reading of the prior discussions on this topic. In
particular, I looked at Chris's email:
[EM] Top-two Approval Pairwise Runoff (TTAPR)
...which in turn refers to earlier conversations (e.g. "earlier
suggestion of mine for a 2 trips to the polling station method using
simple Approval ballots."). In my brief scan of those conversations, I
found this as a possible entry point:
"A better 2-round method that uses approval ballots"
I'm not sure I get how a Condorcet winner violation can occur with
MAF, and I don't know which of Chris Benham's earlier suggestions
you're pointing me to. I'm willing to believe there's something
obvious I'm missing (because that happens a lot), but there's nothing
obvious to me yet about that reference.
One possible interpretation: my "Maximum approval top-two" [MATT]
proposal from 2018-11-20 is subject to a Condorcet winner violation.
I'll acknowledge that that is true, since not every candidate in the
majority candidate pool advances in MATT.
At this point, though, I'm planning to refine and advocate for MAF.
MATT is a proposal-of-last-resort if (for whatever reason) it becomes
impossible to advance an initiative that replaces the general election
first-past-the-post voting method with a system that works well with
three or more candidates. MAF is great since (if my hunch is correct)
it will *usually* only advance two candidates (at least at first), and
thus won't seem so weird to voters that are used to California's
current system that exclusively advances two candidates per office.
> Let's imagine a Condorcet-like election with a score ballot with more
> than two ratings, so it's not just approval. Say 0 to 3, with any score
> above zero indicating approval.
> Infer rankings from the ratings (higher score ranks above lower score),
> and tabulate the pairwise array.
> Then advance the following candidates to the general election: the
> Smith set, plus the Approval winner, plus the AW-complement. This group
> is guaranteed to include the Condorcet winner, plus at least one other
> candidate. If desired, according to your criterion above, you could
> also include the AW-runner-up.
This is an interesting solution to the problem that isn't obvious to
me yet. As of right now, I'm not yet convinced that we need to make a
more complicated ballot than an approval ballot for the primary.
Moreover, I think there's tremendous benefit to allowing both the
primary election and the general election to use the same
approval-style ballot instructions.
Here's the end goal for my requirements for a third draft of MAF,
using similar vocabulary as my 2018-11-22 proposal: 1) advance three
pools of candidates ("highly-approved candidate pool", "majority
candidate pool", "opposition candidate pool") to the general election.
2) provide electoral motivation for candidates to maximize their
overall approval score, and not obsess about their base voters 3)
ensure the resulting general election ballot achieves a very high
"ballot satisfaction" score.
My first draft had a rather arbitrary and small opposition candidate
pool (one candidate). My second draft probably has one that is too
big. I've got ideas for my third draft that I think passes the
Goldilocks test :-)
More information about the Election-Methods