[EM] Approval-based replacement for jungle primary

[Rob Lanphier]

Hi Ted,

Hmm....I have a tweak to the proposal to run by you.  More below...

On Wed, Nov 21, 2018 at 5:31 PM Ted Stern <dodecatheon at gmail.com> wrote:
> I like the MAF proposal. I've posted in favor of PR approval based runoff methods before, almost exactly 2 years ago, in fact.

Thanks!  My proposal comes with a bit of naivety about prior
proposals.  In the past, I've sent an "oh, I have an idea" post it to
this list, and then someone replies "you mean Coombs?" and I ask
"what's Coombs?" and then someone (possibly after emitting a heavy
sigh) composes a very educational email detailing what Coombs is.

I'm assuming the proposal you're talking about is this one:
http://lists.electorama.com/pipermail/election-methods-electorama.com/2016-November/001054.html

I just read it, and I concur.  Applying proportional-style approval
voting to electoral college selection is a neat idea.  The MAF step 2
in my proposal seems to use a very similar mechanism to the one you
describe.  My proposal below might be even more similar...

> My only caveat would be that there is a theoretical possibility that including everyone who gets more than 50% approval could introduce a clone crowding effect.  Do you want any limits on the number?

The clone problem has been in the back of my mind for a while; I'm
glad you brought it up.  A solution that just occurred to me: what if
we generalized the MAF step 2 mechanism.  So, here's the MAF rules in
my blog post yesterday (er...make that Tuesday):
1. The candidate who receives the highest approval rating qualifies
for the general election
2. If less than 75% approve of the leading candidate, then a second
candidate (the “complementary candidate”), who maximizes the approval
of the electorate, also qualifies
3. All candidates who receive over 50% approval also qualify for the
general election

For a dominant party like the Democratic Party in California in 2018,
under this system, their motivation would be to run as many Democrats
as possible, and hope they all make it to the general election.

What if instead, for all candidates who qualify via step 1 or step 3
(getting between 50% and 75% of the vote), an opposition candidate is
chosen?  If the left-wing/right-wing model persists for a while, it
basically means that every Democrat/Green/PeaceAndFreedom candidate
who advances by getting 50% approval, there could be a
Republican/Libertarian/AmericanIndependent who also advances.  If the
Republican and Democratic parties remain dominant, there would
probably be an R for every D, but of course, once people start getting
comfortable approving an army of clones, it could be that clone
coalitions form that bring in these third parties. For reference,
here's the certified political parties in California:
https://www.sos.ca.gov/elections/political-parties/qualified-political-parties/

(and don't hold me to my assessment that
"Democrat/Green/PeaceAndFreedom" == "left" and
"Republican/Libertarian/AmericanIndependent == "right"...I'm a little
uninformed about the nuances of our third parties, and also realize
that, for example, Libertarians would balk at being called "right
wing" in the left/right model)

Advancing an army of clones invites an army of complementary
candidates, so I suspect that would motivate partisan voters to be
stingy about approving an army of clones, and to motivate
parties/coalitions to thin their ranks a little before the primary and
publish focused endorsements, and to motivate candidates to drop out
before the primary if the field is too crowded.

Perhaps a way of formalizing this mechanism would be to change to this
set of rules:
1. Select the candidate who receives the highest approval rating.
This is the "top candidate" and qualifies for the ballot
1a. If the top candidate (and any other candidate) receives greater
than 75% approval, add these candidates to the "highly-approved
candidate pool"
1b. If the top candidate receives less than 75% approval, add the top
candidate to the "majority candidate pool"
2. If the top candidate has been added to the "majority candidate
pool", also add a candidate to the “opposition candidate pool”.  To be
added, this candidate must be the candidate which maximizes the
"majority/opposition ballot satisfaction".  "ballot satisfaction"
generally means voters approve of at least one candidate on a given
ballot.  "majority/opposition ballot satisfaction" is for a ballot
that only contains the "majority candidate pool" and the "opposition
candidate pool".
3. For each candidate who receives over 50% approval, but less than
75% approval:
3a.  Add this candidate to the "majority candidate pool"
3b.  Add a candidate to the "opposition candidate pool" who maximizes
the "majority/opposition ballot satisfaction" of the electorate (as in
step 2)
4.  Eliminate all candidates from the "opposition candidate pool" who
have an overall approval rating under 25%
5.  All candidates remaining in the "highly-approved candidate pool",
the "majority candidate pool" and the "opposition candidate pool"
advance to the general election.

In the original MAF proposal, the "opposition candidate pool" (i.e.
the "complementary candidate") is no more than one candidate.  In this
proposal, the opposition candidate pool could grow to two candidates
if two candidates are added to the majority candidate pool..

This doesn't seem like a radical departure from the rules I described
in my Tuesday blog post.  The added complexity bothers me, but this
seems to solve a problem with the original proposal.  In the old
proposal, the single complementary candidate is chosen as an
alternative to the leading candidate, even if a third candidate is
also added.  In this proposal, it would seem rare to advance just
three candidates; either there's one candidate in each pool ("highly
approved", "majority", "opposition") or there as a candidate pruned
from the "opposition candidate pool" for being under 25% approval.

My hunch: it would take at least 2-3 election cycles before more than
two candidates advance to the general.  I suspect bullet voting would
be common in early elections, and it would take a while before
sophisticated campaign strategies emerge (e.g. like candidates
endorsing each other, holding joint events, and advertising for one
another).  Most elections would result in a single "majority"
candidate, and an "opposition" candidate.

I came up with the set of rules above as I was composing this email,
because I wanted to make the rules fit the examples below.  My first
draft had rules ensuring that if the majority candidate pool had N
candidates, the opposition candidate pool would only have N-1
candidates. In this version, it's possible for the opposition
candidate pool to have just as many candidates as the majority
candidate pool.  But it didn't match my examples below, and I liked my
examples better than I liked my draft rules, so I rewrote the rules.
Now that I have rules I like, I've tweaked my example scenarios to fit
the rules:

Test scenario #1: Let's say that seven candidates qualify to advance,
and the top candidate only receives 55% approval.  It seems that the
order that the primary candidates should enter their respective pools
should be like this:
#1 - Top candidate - first in majority candidate pool
#2 - first in opposition candidate pool (complementing the top
candidate in the majority pool)
#3 - second in majority candidate pool (with 54% approval)
#4 - second in opposition candidate pool (complementing the candidates
above in the majority pool)
#5 - third in majority candidate pool (with 53% approval)
#6 - third in opposition candidate pool (complementing the candidates
above in the majority pool)
#7 - fourth in majority candidate pool (with 52% approval)
#8 - fourth in opposition candidate pool (complementing the candidates
above in the majority pool)

In my original draft, the candidate with the lowest overall approval
score would be eliminated from the opposition pool so that the
majority pool had four candidates, and the opposition pool only had
three, and thus only seven candidates advanced.  In my current rules,
it's possible for eight candidates to qualify, but my new rule #4
above ("Eliminate all candidates from the opposition candidate pool
who have an overall approval rating under 25%") could knock it down to
seven.  Or six,  Or even four.

Test scenario #2: Let's say the top candidate gets greater than 75%
approval. That's a pretty strong indication that the top candidate is
the median candidate.  But if three other candidates also get greater
than 50% approval, it only seems fair to give them a hearing in the
general election.  Thus, when the top candidate gets greater than 75%
approval, it seems the order should go like this:
1 - Top candidate - first in highly-approved candidate pool
2 - first in majority candidate pool
3 - first in opposition candidate pool (complementing the first
candidate in the majority candidate pool)
4 - second in majority candidate pool
5 - second in opposition candidate pool (complementing the candidates
above in the majority candidate pool)
6 - third in majority candidate pool
7 - third in opposition candidate pool (complementing the candidates
above in the majority candidate pool)

I originally wrote "I'm pretty sure it'd be possible to write a set of
rules to achieve this.  I'm just not going to do it tonight".  I
*think* I pulled it off.  That's why I didn't send this mail a couple
hours ago.  Now I really should send this email.  :-)

Rob
p.s for those of you who prefer reading blog stuff on Medium (or feel
like clicking on the applause link), here's the Medium version of this
proposal:
https://medium.com/@robla/replacing-the-jungle-primary-c1e844a5333b